* Speedup tokenization
On current master it takes ~3.2 seconds to tokenize
Wikitext. With this change it becomes ~525 ms.
* Fixit: it was missing the piece after the last found occurence
---------
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
* Make ggml-cuda.cu build with QK_K = 64
Using LLAMA_CUDA_FORCE_DMMV = ON and -nommq it runs and produces
a meaningful result.
* k_quants tuning for Falcon-7b
---------
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
* tests : write a Python tokenizer test (wip)
* llama : prefix input text for tokenization with whitespace
* llama : distinguish pieces from decoded text + fix detokenization
* common : add comments
* examples : no longer manually add leading space when tokenizing
* tests : use Python to generate tokenizer tests for C++
* tests : add option to tokenize text files
ggml-ci
* tests : add test-tokenizer-1.py
* llama.cpp : fix LF token
* hellaswag : move the concat space for clarity
* tests : add falcon tests (py + cpp, currently do not pass Unicode)
ggml-ci
* common : temporary separate llama_detokenize calls for SPM and BPE
---------
Co-authored-by: klosax <131523366+klosax@users.noreply.github.com>
* ci : add lora test
ggml-ci
* move lora summary to the top, add lora logs
ggml-ci
* ci : decrease CPU ppl runs to 2 to avoide 20 min timeout
ggml-ci
* add 7b lora test
use 1 thread for CUDA generation tests
ggml-ci
* add test with q8_0 (cpu only)
ggml-ci
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Allow convert.py to convert to q8_0
Fix issue with bounded_parallel_map and greedy consuming iterator
Display elapsed time during conversion
* Add --concurrency option
Minor improvements to help text
Clean up bounded_parallel_map function a bit
* Massive speed improvement thanks to Cebtenzzre
* Refactor types
The use of special characters within source files can break compiling on some computers with different region and language settings. Using Unicode escape sequences should allow for the code to be compiled on all setups without needing to change your computers settings or switch regions.
* Fix bug in main.cpp where penalize_nl=false has no effect. It modifies the underlying logits array, but at this point we are already working on the candidates copy.
* Suppress redefinition warning for NOMINMAX on mingw. In my installation, this macro is already defined by /usr/lib/gcc/x86_64-w64-mingw32/11/include/c++/x86_64-w64-mingw32/bits/os_defines.h:45.
* main : fix indentation
* main : pass ctx to llama_token_nl()
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
Problem
-------
`nix build` fails with missing `Accelerate.h`.
Changes
-------
- Fix build of the llama.cpp with nix for Intel: add the same SDK frameworks as
for ARM
- Add `quantize` app to the output of nix flake
- Extend nix devShell with llama-python so we can use convertScript
Testing
-------
Testing the steps with nix:
1. `nix build`
Get the model and then
2. `nix develop` and then `python convert.py models/llama-2-7b.ggmlv3.q4_0.bin`
3. `nix run llama.cpp#quantize -- open_llama_7b/ggml-model-f16.gguf ./models/ggml-model-q4_0.bin 2`
4. `nix run llama.cpp#llama -- -m models/ggml-model-q4_0.bin -p "What is nix?" -n 400 --temp 0.8 -e -t 8`
Co-authored-by: Volodymyr Vitvitskyi <volodymyrvitvitskyi@SamsungPro.local>
* llama.cpp : fix spm whitespace escaping + clean up
* main.cpp : spm - add whitespace in front of prompt
* test-tokenizer-0.cpp : spm - add whitespace in front of prompt
* Add llama_beam_search().
* Add '// Beam search' heading to llama.{h,cpp} after llama_grammar_accept_token().
* Add space around * pointers and & references.
* Add spaces around comparison and assignment operators.
* Prefer west const.
* Use llama_ prefix for structs in global namespace.
* Delete obsolete comment from an earlier revision.
* Change eos to eob in llama_beam and llama_beam_view structs.
* server : add n_probs param in chat UI
* server : keep message data array & show in probabilites component
* server : add simple popover component
* server : fix completion_probabilities undefined if not set n_probs
* server : implement Probabilites
* server : handle bytes
* server : make n_probs max to 10 for easy scroll
* server : adjust for dark/light mode
* server : Fix regenerated prompt
* server : update index.html.hpp
* server : convert prob to percentage + show original value as div title
* server : fix Probabilites not used if included empty str
* server : skip byte pair in display probabilites
* server : remove array check of completion_probabilities in messages
* skip empty array or byte pair (> 1) in Probabilites
* generate index.html.hpp
* fix incorrect prob convert if the str is already a known token
* use final response to show probabilities on stop
* revert unnecessary change
* correct probabilites usage
* remove unused function
* always send partial response for get correct probs of last to_send
* fix typo
* fix content of format_final_response
* refactor probs render & make pColor transparent if not found
* send empty string when got stop_pos in partial
* avoid unnecessary empty data event & send rest of partial tokens on stop
* use <br /> for new line
* skip -1 tok in loop to avoid send '' on end
* trim last new lines on stop
* revert unnecessary change
* use hipblas based on cublas
* Update Makefile for the Cuda kernels
* Expand arch list and make it overrideable
* Fix multi GPU on multiple amd architectures with rocblas_initialize() (#5)
* add hipBLAS to README
* new build arg LLAMA_CUDA_MMQ_Y
* fix half2 decomposition
* Add intrinsics polyfills for AMD
* AMD assembly optimized __dp4a
* Allow overriding CC_TURING
* use "ROCm" instead of "CUDA"
* ignore all build dirs
* Add Dockerfiles
* fix llama-bench
* fix -nommq help for non CUDA/HIP
---------
Co-authored-by: YellowRoseCx <80486540+YellowRoseCx@users.noreply.github.com>
Co-authored-by: ardfork <134447697+ardfork@users.noreply.github.com>
Co-authored-by: funnbot <22226942+funnbot@users.noreply.github.com>
Co-authored-by: Engininja2 <139037756+Engininja2@users.noreply.github.com>
Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
Co-authored-by: jammm <2500920+jammm@users.noreply.github.com>
Co-authored-by: jdecourval <7315817+jdecourval@users.noreply.github.com>
* metal: better memory alloc w/ concurrency dispatch
The ggml-alloc should only free tensors at memory barriers.
* ggml-alloc: avoid return silently
In certain cases, the allocate_node() function may silently return
without performing any memory allocation.
