Commit Graph

137 Commits

Author SHA1 Message Date
Georgi Gerganov
4d98d9a656
sync : ggml (SD ops, tests, kernels) (#4444)
* sync : ggml (SD ops, tests, kernels)

ggml-ci

* cuda : restore im2col

ggml-ci

* metal : fix accuracy of dequantization kernels

ggml-ci

* cuda : restore correct im2col

ggml-ci

* metal : try to fix moe test by reducing expert size

ggml-ci

* cuda : fix bin bcast when src1 and dst have different types

ggml-ci

---------

Co-authored-by: slaren <slarengh@gmail.com>
2023-12-13 21:54:54 +02:00
slaren
799a1cb13b
llama : add Mixtral support (#4406)
* convert : support Mixtral as LLAMA arch

* convert : fix n_ff typo

* llama : model loading

* ggml : sync latest ggml_mul_mat_id

* llama : update graph to support MoE

* llama : fix cur -> cur_expert

* llama : first working version

* llama : fix expert weighting in the FFN

* ggml : ggml_get_rows support 2D indexing [n_tokens, n_experts] (cpu only)

* ggml : add n_as argument to ggml_mul_mat_id

* ggml : fix ggml_get_rows to take into account ne02 / ne11

* metal : add more general support for ggml_get_rows + tests

* llama : add basic support for offloading moe with CUDA

* metal : add/mul/div use general kernel when src1 not cont

* metal : reduce the kernel launches for ggml_mul_mat_id

* ggml : get_rows : support non-contiguos tensors with gaps, generalize up to 3D

* ggml : update get_rows f16 and q

* cuda : support non-contiguous src1 in get_rows

* llama : offload missing ffn_moe_silu

* metal : fix ggml_get_rows to work with non-cont src1

* metal : add indirect mat-vec kernels for all quantization types

* llama : do not quantize expert gating tensors

* llama : add n_expert and n_expert_used to hparams + change quants

* test-backend-ops : add moe test

* cuda : fix get_rows when ncols is odd

* convert : determine n_ctx correctly

* metal : fix ggml_mul_mat_id for F32

* test-backend-ops : make experts more evenly probable (test_moe)

* test-backend-ops : cleanup, add moe test for batches

* test-backend-ops : add cpy from f32 -> all types test

* test-backend-ops : fix dequantize block offset

* llama : fix hard-coded number of experts

* test-backend-ops : simplify and disable slow tests to avoid CI timeout

* test-backend-ops : disable MOE test with thread sanitizer

* cuda : fix mul_mat_id with multi gpu

* convert : use 1e6 rope_freq_base for mixtral

* convert : fix style

* convert : support safetensors format

* gguf-py : bump version

* metal : add cpy f16 -> f32 kernel

* metal : fix binary ops for ne10 % 4 != 0

* test-backend-ops : add one more sum_rows test

* ggml : do not use BLAS with ggml_mul_mat_id

* convert-hf : support for mixtral-instruct (#4428)

* convert : typo fix, add additional hyperparameters, use LLaMA arch for Mixtral-instruct

* convert : use sentencepiece tokenizer for Mixtral-instruct

* convert : make flake8 happy

* metal : fix soft_max kernels

ref: 1914017863

* metal : limit kernels to not use more than the allowed threads

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
Co-authored-by: Radek Pilar <github@mrkva.eu>
2023-12-13 14:04:25 +02:00
Georgi Gerganov
fe680e3d10
sync : ggml (new ops, tests, backend, etc.) (#4359)
* sync : ggml (part 1)

* sync : ggml (part 2, CUDA)

* sync : ggml (part 3, Metal)

* ggml : build fixes

ggml-ci

* cuda : restore lost changes

* cuda : restore lost changes (StableLM rope)

* cmake : enable separable compilation for CUDA

ggml-ci

* ggml-cuda : remove device side dequantize

* Revert "cmake : enable separable compilation for CUDA"

This reverts commit 09e35d04b1.

* cuda : remove assert for rope

* tests : add test-backend-ops

* ggml : fix bug in ggml_concat

* ggml : restore `ggml_get_n_tasks()` logic in `ggml_graph_plan()`

* ci : try to fix macOS

* ggml-backend : remove backend self-registration

* ci : disable Metal for macOS cmake build

ggml-ci

* metal : fix "supports family" call

* metal : fix assert

* metal : print resource path

ggml-ci

---------

Co-authored-by: slaren <slarengh@gmail.com>
2023-12-07 22:26:54 +02:00
Georgi Gerganov
bcc0eb4591
llama : per-layer KV cache + quantum K cache (#4309)
* per-layer KV

* remove unnecessary copies

* less code duplication, offload k and v separately

* llama : offload KV cache per-layer

* llama : offload K shift tensors

* llama : offload for rest of the model arches

* llama : enable offload debug temporarily

* llama : keep the KV related layers on the device

* llama : remove mirrors, perform Device -> Host when partial offload

* common : add command-line arg to disable KV cache offloading

* llama : update session save/load

* llama : support quantum K cache (#4312)

* llama : support quantum K cache (wip)

* metal : add F32 -> Q8_0 copy kernel

* cuda : add F32 -> Q8_0 copy kernel

ggml-ci

* cuda : use mmv kernel for quantum cache ops

* llama : pass KV cache type through API

* llama : fix build

ggml-ci

* metal : add F32 -> Q4_0 copy kernel

* metal : add F32 -> Q4_1 copy kernel

* cuda : wip

* cuda : add F32 -> Q4_0 and F32 -> Q4_1 copy kernels

* llama-bench : support type_k/type_v

* metal : use mm kernel only for quantum KV cache

* cuda : add comment

* llama : remove memory_f16 and kv_f16 flags

---------

Co-authored-by: slaren <slarengh@gmail.com>

* readme : add API change notice

---------

Co-authored-by: slaren <slarengh@gmail.com>
2023-12-07 13:03:17 +02:00
Georgi Gerganov
ef47ec18da
ggml : add ggml_soft_max_ext (#4256)
* metal : implement soft_max_ext

* cuda : implement soft_max_ext

* ggml : implement soft_max_ext (CPU)

* batched-bench : print threads

ggml-ci

* metal : simplify soft_max encoding

ggml-ci

* cuda : use 512 threads for soft_max instead of 32

* ggml : update soft max cpu

* cuda : do warp-based block reduce

* cuda : increase max block size to 1024

* cuda : fix warp reduction initialization of shared mem

* metal : warp-based reduction for soft max kernel

* metal : warp-based reduce for rms_norm

* metal : simplify soft max kernel

ggml-ci

* alloc : fix build with debug
2023-12-01 10:51:24 +02:00
slaren
8a052c131e
ggml-cuda : support stablelm rope (#4156)
* ggml-cuda : support stablelm rope

* remove unused freq_base kernel parameter

* add n_dims parameter to llm_build_k_shift, default to n_rot via overload

* llama : fix llm_build_k_shift args

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-11-24 18:04:31 +01:00
Haohui Mai
55978ce09b
Fix incorrect format strings and uninitialized variables. (#4133)
* Fix incorrect format strings and uninitialized variables.

* Address comments

* Add the missing include statement
2023-11-23 22:56:53 +01:00
Kerfuffle
2923f17f6f
Clean up ggml-cuda.cu warnings when compiling with clang (for ROCM) (#4124)
* ggml-cuda.cu: Clean up warnings when compiling with clang

* ggml-cuda.cu: Move static items into anonymous namespace

* ggml-cuda.cu: Fix use of namespace start macro

* Revert "ggml-cuda.cu: Fix use of namespace start macro"

This reverts commit 26c1149026.

* Revert "ggml-cuda.cu: Move static items into anonymous namespace"

This reverts commit e29757e0f7.
2023-11-18 08:11:18 -07:00
Andrew Godfrey
b83e149ec6
cuda : get_row_rounding F32 (#4095)
* Fix #4017

* Update ggml-cuda.cu

Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>

* Update ggml-cuda.cu

Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>

---------

Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2023-11-17 10:01:15 +02:00
Georgi Gerganov
4f447a4833
llama : fix data units (#4101)
* llama : fix data units

ggml-ci

* Revert "llama : fix data units"

This reverts commit f5feac831f.

* llama : disambiguate data units

ggml-ci
2023-11-17 10:00:15 +02:00
slaren
1cf2850d52
ggml-cuda : increase max graph size (#4084) 2023-11-15 14:58:13 +02:00
Georgi Gerganov
3d68f364f1
ggml : sync (im2col, GPU conv, 32-bit arm compat) (#4060)
ggml-ci
2023-11-13 16:55:52 +02:00
Georgi Gerganov
4760e7cc0b
sync : ggml (backend v2) (#3912)
* sync : ggml (backend v2) (wip)

* sync : migrate examples and llama.cpp to dynamic graphs (wip)

* sync : update tests + fix max op params to 64

ggml-ci

* sync : ggml-cuda

ggml-ci

* llama : fix save/load state context size

ggml-ci

* sync : try to fix build on tvOS

* sync : pass custom graph sizes in training examples

* sync : update graph copies to new ggml API

* sync : update sync-ggml.sh with new files

* scripts : fix header in sync script

* train : fix context size calculations

* llama : increase inference graph size up to 4096 nodes

* train : allocate grads for backward graphs

* train : allocate grads for gb_tmp
2023-11-13 14:16:23 +02:00
Kerfuffle
bb50a792ec
Add ReLU and SQR CUDA ops to (partially) fix Persimmon offloading (#4041)
* Add ReLU and SQR CUDA ops to fix Persimmon offloading

* Persimmon loader: More helpful error on CUDA/ROCM when offloading too many layers
2023-11-13 01:58:15 -07:00
Meng Zhang
46876d2a2c
cuda : supports running on CPU for GGML_USE_CUBLAS=ON build (#3946)
* protyping the idea that supports running on CPU for a GGML_USE_CUBLAS=on build

* doc: add comments to ggml_cublas_loaded()

* fix defined(...)
2023-11-07 08:49:08 +02:00
slaren
2833a6f63c
ggml-cuda : fix f16 mul mat (#3961)
* ggml-cuda : fix f16 mul mat

ggml-ci

* silence common.cpp warning (bonus)
2023-11-05 18:45:16 +01:00
Jared Van Bortel
132d25b8a6
cuda : fix disabling device with --tensor-split 1,0 (#3951)
Co-authored-by: slaren <slarengh@gmail.com>
2023-11-05 10:08:57 -05:00
slaren
48ade94538
cuda : revert CUDA pool stuff (#3944)
* Revert "cuda : add ROCM aliases for CUDA pool stuff (#3918)"

This reverts commit 629f917cd6.

* Revert "cuda : use CUDA memory pool with async memory allocation/deallocation when available (#3903)"

This reverts commit d6069051de.

ggml-ci
2023-11-05 09:12:13 +02:00
slaren
abb77e7319
ggml-cuda : move row numbers to x grid dim in mmv kernels (#3921) 2023-11-03 12:13:09 +01:00
Kerfuffle
629f917cd6
cuda : add ROCM aliases for CUDA pool stuff (#3918) 2023-11-02 21:58:22 +02:00
Georgi Gerganov
c7743fe1c1
cuda : fix const ptrs warning causing ROCm build issues (#3913) 2023-11-02 20:32:11 +02:00
Oleksii Maryshchenko
d6069051de
cuda : use CUDA memory pool with async memory allocation/deallocation when available (#3903)
* Using cuda memory pools for async alloc/dealloc.

* If cuda device doesnt support memory pool than use old implementation.

* Removed redundant cublasSetStream

---------

Co-authored-by: Oleksii Maryshchenko <omaryshchenko@dtis.com>
2023-11-02 19:10:39 +02:00
Georgi Gerganov
4d719a6d4e
cuda : check if this fixes Pascal card regression (#3882) 2023-11-02 08:35:10 +02:00
cebtenzzre
2fffa0d61f
cuda : fix RoPE after #2268 (#3897) 2023-11-02 07:49:44 +02:00
slaren
d02e98cde0
ggml-cuda : compute ptrs for cublasGemmBatchedEx in a kernel (#3891)
* ggml-cuda : compute ptrs for cublasGemmBatchedEx in a kernel

* fix warnings
2023-11-01 23:10:09 +01:00
cebtenzzre
898aeca90a
llama : implement YaRN RoPE scaling (#2268)
Co-authored-by: cebtenzzre <cebtenzzre@gmail.com>
Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 18:04:33 -04:00
Andrew Godfrey
73bdcb395e
finetune : add -ngl parameter (#3762)
* Add '-ngl' support to finetune.cpp

* Add fprintf in ggml_cuda_op_add

When I tried CUDA offloading during finetuning following the readme, I got an assert here.
This probably isn't an important case because inference later gives a warning saying you should use f16 or f32 instead when using lora

* Add 'finetune.sh', which currently fails when using GPU

"error: operator (): Finetuning on tensors with type 'f16' is not yet supported"

* tweak finetune.sh

* Suppress some warnings in ggml.c

* Add f16 implementation to ggml_compute_forward_add_f16_f32

* Add an f16 case to ggml_add_cast_impl and llama_build_lora_finetune_graphs

* finetune.sh: Edit comments

* Add "add_f16_f32_f32_cuda"

* Tweak an error message

* finetune.sh: Add an optional LLAMA_MODEL_DIR variable

* finetune.sh: Add an optional LLAMA_TRAINING_DIR variable

* train : minor

* tabs to spaces

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
Co-authored-by: cebtenzzre <cebtenzzre@gmail.com>
2023-11-01 13:49:04 +02:00
Georgi Gerganov
2f9ec7e271
cuda : improve text-generation and batched decoding performance (#3776)
* cuda : prints wip

* cuda : new cublas gemm branch for multi-batch quantized src0

* cuda : add F32 sgemm branch

* cuda : fine-tune >= VOLTA params + use MMQ only for small batches

* cuda : remove duplicated cuBLAS GEMM code

* cuda : add CUDA_USE_TENSOR_CORES and GGML_CUDA_FORCE_MMQ macros

* build : add compile option to force use of MMQ kernels
2023-10-27 17:01:23 +03:00
Georgi Gerganov
6961c4bd0b
batched-bench : print params at start 2023-10-25 10:26:27 +03:00
Georgi Gerganov
b2f7e04bd3
sync : ggml (conv ops + cuda MSVC fixes) (#3765)
ggml-ci
2023-10-24 21:51:20 +03:00
Georgi Gerganov
2b4ea35e56
cuda : add batched cuBLAS GEMM for faster attention (#3749)
* cmake : add helper for faster CUDA builds

* batched : add NGL arg

* ggml : skip nops in compute_forward

* cuda : minor indentation

* cuda : batched cuBLAS GEMMs for src0 F16 and src1 F32 (attention ops)

* Apply suggestions from code review

These changes plus:

```c++
#define cublasGemmBatchedEx hipblasGemmBatchedEx
```

are needed to compile with ROCM. I haven't done performance testing, but it seems to work.

I couldn't figure out how to propose a change for lines outside what the pull changed, also this is the first time trying to create a multi-part review so please forgive me if I mess something up.

* cuda : add ROCm / hipBLAS cublasGemmBatchedEx define

* cuda : add cublasGemmStridedBatchedEx for non-broadcasted cases

* cuda : reduce mallocs in cublasGemmBatchedEx branch

* cuda : add TODO for calling cublas from kernel + using mem pool

---------

Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-10-24 16:48:37 +03:00
Jan Ploski
f5f9121de1
llm : add MPT support (#3417)
* CUDA: added support for ggml_clamp (see also: https://github.com/ggerganov/ggml/issues/545)

* mpt : added an implementation based (mostly) on falcon integration, modified with deltas from ggml/examples/mpt

* mpt : protect against "clip_qkv": null in mpt-7b

* mpt : quick fix to avoid "Strange model" warning when quantizing MPT models

* mpt : addendum to changeset:84e30e8 - leave parameter clamp_kqv out from metadata rather than use 0.0 to indicate "no clamping" (more compliant with the current GGUF spec?)

* mpt : standardized all tensor names to follow GGUF spec

* mpt : addendum to changeset:1be89c40 - use "req" parameter of GGUF_GET_KEY macro instead of duplicate code

* mpt : fixed comment s/gptneox/mpt/

* mpt : remove tabs, trailing whitespace

* mpt : removed ne01 + n_past == ne00 assertion from alibi (cuda/f32) and rope_shift from build_mpt

* mpt : updated convert-mpt-hf-to-gguf.py to reflect changes made to convert-gptneox-hf-to-gguf.py in pr:3252

* comment out n_past instead of marking it unused

* mpt : removed hardcoded +178 from convert script in favor of utilizing hparams["vocab_size"]

* mpt : remove unused tokenizer_json in convert script

* ggml : remove obsolete n_past assert in ggml_alibi

* llama : print clam_kqv and max_alibi_bias hparams

---------

Co-authored-by: Cebtenzzre <cebtenzzre@gmail.com>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 10:50:23 +03:00
Georgi Gerganov
db3abcc114
sync : ggml (ggml-backend) (#3548)
* sync : ggml (ggml-backend)

ggml-ci

* zig : add ggml-backend to the build
2023-10-08 20:19:14 +03:00
slaren
f5ef5cfb18
ggml-cuda : perform cublas mat mul of quantized types as f16 (#3412)
* ggml-cuda : perform cublas matrix multiplication of quantized types as fp16

* rename CC_TURING to CC_VOLTA

* disable fp16 mat mul completely with multi GPU
2023-09-30 18:12:57 +02:00
slaren
16bc66d947
llama.cpp : split llama_context_params into model and context params (#3301)
* llama.cpp : split llama_context_params into model and context params

ggml-ci

* fix metal build

* fix freq_base/scale default to model value

* llama-bench : keep the same model between tests when possible

* move n_threads to llama_context_params, add n_threads_batch

* fix mpi build

* remove kv_size(), cuda scratch fixes

* remove low-vram option

* add n_threads_batch to system info, refactor to get_system_info()

* add documentation about --threads-batch to the READMEs

* llama-bench fix

* main : fix rope freq/scale warning

* llama.cpp : add llama_get_model
common : add llama_tokenize from model

* remove duplicated ctx/model functions

ggml-ci

* cuda : print total VRAM used
2023-09-28 22:42:38 +03:00
Georgi Gerganov
ec893798b7
llama : custom attention mask + parallel decoding + no context swaps (#3228)
* tests : verify that RoPE is "additive"

* llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask)

* ggml : ggml_rope now takes a vector with positions instead of n_past

* metal : add rope_f16 kernel + optimize cpy kernels

* llama : unified KV cache + batch inference API

* llama : add new llama_decode() API that works with llama_batch

* llama : add cell_max heuristic for more efficient kv_cache

* llama : extend llama_kv_cache API

* llama : more robust cell_max heuristic + wip shift

* metal : disable concurrency optimization

* llama : add llama_kv_cache_shift_seq + no more context swaps

* llama : apply K-cache roping for Falcon and Baichuan

* speculative : fix KV cache management

* parallel : example for serving multiple users in parallel

* parallel : disable hot-plug to avoid cache fragmentation

* fixes : speculative KV cache + llama worst-case graph

* llama : extend batch API to select which logits to output

* llama : fix worst case graph build

* ggml-cuda : update rope implementation for parallel decoding (#3254)

* ggml-cuda : update rope implementation for parallel decoding

* better solution for p0 computation

* fix rope

* simpler rope implementation

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

* make : add parallel to build + fix static functions in llama.cpp

* simple : fix token counting

* parallel : various improvements

* llama : fix cell_max logic + rename functions

* parallel : try smaller batches when the KV cache is fragmented

* parallel : fix sequence termination criteria

* llama : silence errors KV cache errors

* parallel : remove new line from prompt

* parallel : process system prompt once + configurable paramters + llama API

* parallel : remove question with short answers

* parallel : count cache misses

* parallel : print misses on each request

* parallel : minor

* llama : fix n_kv to never become 0

* parallel : rename hot-plug to continuous-batching

* llama : improve llama_batch API + simplify parallel example

* simple : add parallel decoding support

* simple : improve comments + free batch

* ggml-cuda : add rope f16, restore performance with parallel decoding (#3272)

* ggml-cuda : add rope f16, restore performance

* offload KQ_mask with all models

* fix rope shift

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

* llama : disable MPI for now

ggml-ci

* train : make KQ_pos memory buffer permanent via dummy scale op

* ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275)

ggml-ci

* parallel : fix bug (extra BOS) + smaller token_prev array

* parallel : fix cases where the input prompts can overflow the batch

* parallel : add disabled experimental batch chunking in powers of two

* llama : llama.h formatting + comments

* simple : add README.md

* llama : fix kv cache heuristic when context is less than 32

* parallel : fix crash when `-n -1`

* llama : simplify returns if/else branches

* metal : use mm kernels for batch size > 2

* examples : utilize new llama_get_logits_ith()

* examples : add example for batched decoding

* examples : do not eval prompt 2 times (close #3348)

* server : clear the KV cache beyond n_past before llama_decode

* server : avoid context swaps by shifting the KV cache

---------

Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 19:04:36 +03:00
slaren
da0400344b
ggml-cuda : perform cublas fp16 matrix multiplication as fp16 (#3370)
* ggml-cuda : perform cublas fp16 matrix multiplication as fp16

* try to fix rocm build

* restrict fp16 mat mul to volta and up
2023-09-28 13:08:28 +03:00
Johannes Gäßler
ee66942d7e
CUDA: fix peer access logic (#3231) 2023-09-17 23:35:20 +02:00
Johannes Gäßler
111163e246
CUDA: enable peer access between devices (#2470) 2023-09-17 16:37:53 +02:00
Johannes Gäßler
578d8c8f5c
CUDA: fix scratch malloced on non-main device (#3220) 2023-09-17 14:16:22 +02:00
Vlad
5dbc2b3213
Enable build with CUDA 11.0 (make) (#3132)
* CUDA 11.0 fixes

* Cleaner CUDA/host flags separation

Also renamed GGML_ASSUME into GGML_CUDA_ASSUME
2023-09-16 16:55:43 +02:00
Johannes Gäßler
0a5eebb45d
CUDA: mul_mat_q RDNA2 tunings (#2910)
* CUDA: mul_mat_q RDNA2 tunings

* Update ggml-cuda.cu

Co-authored-by: Henri Vasserman <henv@hot.ee>

---------

Co-authored-by: Henri Vasserman <henv@hot.ee>
2023-09-13 11:20:24 +02:00
Johannes Gäßler
4f7cd6ba9c
CUDA: fix LoRAs (#3130) 2023-09-13 00:15:33 +02:00
Johannes Gäßler
89e89599fd
CUDA: fix mul_mat_q not used for output tensor (#3127) 2023-09-11 22:58:41 +02:00
Johannes Gäßler
d54a4027a6
CUDA: lower GPU latency + fix Windows performance (#3110) 2023-09-11 19:55:51 +02:00
Johannes Gäßler
8a4ca9af56
CUDA: add device number to error messages (#3112) 2023-09-11 13:00:24 +02:00
Georgi Gerganov
b3e9852e47
sync : ggml (CUDA GLM RoPE + POSIX) (#3082)
ggml-ci
2023-09-08 17:58:07 +03:00
Jiahao Li
35195689cd
2x faster (rms) norm cuda kernels (3.7% e2e improvement) (#2985)
* 2x faster (rms) norm cuda kernels

* Fix code style
2023-09-04 08:53:30 +02:00
Engininja2
f04d002844
cuda : vsubss4 for older versions of ROCm/clang (#2942) 2023-09-01 23:33:19 +02:00
Johannes Gäßler
92b1bbd2ec
CUDA: fix RoPE asserts, block sizes (#2833) 2023-08-28 14:23:55 +03:00