74c47921fa
[Fix] Llama3 Load/Omit CheckpointIO Temporarily ( #5717 )
...
* Fix Llama3 Load error
* Omit Checkpoint IO Temporarily
2024-05-14 20:17:43 +08:00
7806842f2d
add paged-attetionv2: support seq length split across thread block ( #5707 )
2024-05-14 12:46:54 +08:00
8754abae24
[Fix] Fix & Update Inference Tests (compatibility w/ main)
2024-05-05 16:28:56 +00:00
f79963199c
[inference]Add alibi to flash attn function ( #5678 )
...
* add alibi to flash attn function
* rm redundant modifications
2024-04-30 19:35:05 +08:00
5cd75ce4c7
[Inference/Kernel] refactor kvcache manager and rotary_embedding and kvcache_memcpy oper… ( #5663 )
...
* refactor kvcache manager and rotary_embedding and kvcache_memcpy operator
* refactor decode_kv_cache_memcpy
* enable alibi in pagedattention
2024-04-30 15:52:23 +08:00
5f00002e43
[Inference] Adapt Baichuan2-13B TP ( #5659 )
...
* adapt to baichuan2 13B
* add baichuan2 13B TP
* update baichuan tp logic
* rm unused code
* Fix TP logic
* fix alibi slopes tp logic
* rm nn.Module
* Polished the code.
* change BAICHUAN_MODEL_NAME_OR_PATH
* Modified the logic for loading Baichuan weights.
* fix typos
2024-04-30 15:47:07 +08:00
3c91e3f176
[Inference]Adapt to baichuan2 13B ( #5614 )
...
* adapt to baichuan2 13B
* adapt to baichuan2 13B
* change BAICHUAN_MODEL_NAME_OR_PATH
* fix test_decoding_attn.py
* Modifications based on review comments.
* change BAICHUAN_MODEL_NAME_OR_PATH
* mv attn mask processes to test flash decoding
* mv get_alibi_slopes baichuan modeling
* fix bugs in test_baichuan.py
2024-04-25 23:11:30 +08:00
a8fd3b0342
[Inference/Kernel] Optimize paged attention: Refactor key cache layout ( #5643 )
...
* optimize flashdecodingattention: refactor code with different key cache layout(from [num_blocks, num_kv_heads, block_size, head_size] to [num_blocks, num_kv_heads, head_size/x, block_size, x])
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
2024-04-25 14:24:02 +08:00
5d4c1fe8f5
[Fix/Inference] Fix GQA Triton and Support Llama3 ( #5624 )
...
* [fix] GQA calling of flash decoding triton
* fix kv cache alloc shape
* fix rotary triton - GQA
* fix sequence max length assigning
* Sequence max length logic
* fix scheduling and spec-dec
* skip without import error
* fix pytest - skip without ImportError
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
2024-04-23 13:09:55 +08:00
e37ee2fb65
[Feat]Tensor Model Parallel Support For Inference ( #5563 )
...
* tensor parallel support naive source
* [fix]precision, model load and refactor the framework
* add tp unit test
* docstring
* fix do_sample
2024-04-18 16:56:46 +08:00
be396ad6cc
[Inference/Kernel] Add Paged Decoding kernel, sequence split within the same thread block ( #5531 )
...
* feat flash decoding for paged attention
* refactor flashdecodingattention
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
2024-04-18 16:45:07 +08:00
56b222eff8
[inference/model]Adapted to the baichuan2-7B model ( #5591 )
...
* Adapted to the baichuan2-7B model
* modified according to the review comments.
* Modified the method of obtaining random weights.
* modified according to the review comments.
* change mlp layewr 'NOTE'
2024-04-15 16:53:02 +08:00
f8598e3ec5
[Fix] Llama Modeling Control with Spec-Dec ( #5580 )
...
- fix ref before asgmt
- fall back to use triton kernels when using spec-dec
2024-04-10 18:19:44 +08:00
e60d430cf5
[Fix] resolve conflicts of rebasing feat/speculative-decoding ( #5557 )
...
- resolve conflicts of rebasing feat/speculative-decoding
2024-04-10 18:13:49 +08:00
d85d91435a
[Inference/SpecDec] Support GLIDE Drafter Model ( #5455 )
...
* add glide-llama policy and modeling
* update glide modeling, compitable with transformers 4.36.2
* revise glide llama modeling/usage
* fix issues of glimpsing large kv
* revise the way re-loading params for glide drafter
* fix drafter and engine tests
* enable convert to glide strict=False
* revise glide llama modeling
* revise vicuna prompt template
* revise drafter and tests
* apply usage of glide model in engine
2024-04-10 11:07:52 +08:00
a37f82629d
[Inference/SpecDec] Add Speculative Decoding Implementation ( #5423 )
...
* fix flash decoding mask during verification
* add spec-dec
* add test for spec-dec
* revise drafter init
* remove drafter sampling
* retire past kv in drafter
* (trivial) rename attrs
* (trivial) rename arg
* revise how we enable/disable spec-dec
2024-04-10 11:07:52 +08:00
04aca9e55b
[Inference/Kernel]Add get_cos_and_sin Kernel ( #5528 )
...
* Add get_cos_and_sin kernel
* fix code comments
* fix code typos
* merge common codes of get_cos_and_sin kernel.
* Fixed a typo
* Changed 'asset allclose' to 'assert equal'.
2024-04-01 13:47:14 +08:00
68e9396bc0
[fix] merge conflicts
2024-03-25 14:48:28 +08:00
87079cffe8
[Inference]Support FP16/BF16 Flash Attention 2 And Add high_precision Flag To Rotary Embedding ( #5461 )
...
* Support FP16/BF16 Flash Attention 2
* fix bugs in test_kv_cache_memcpy.py
* add context_kv_cache_memcpy_kernel.cu
* rm typename MT
* add tail process
* add high_precision
* add high_precision to config.py
* rm unused code
* change the comment for the high_precision parameter
* update test_rotary_embdding_unpad.py
* fix vector_copy_utils.h
* add comment for self.high_precision when using float32
2024-03-25 13:40:34 +08:00
4eafe0c814
[fix] unused option
2024-03-21 11:28:42 +08:00
aabc9fb6aa
[feat] add use_cuda_kernel option
2024-03-19 13:24:25 +08:00
6e30248683
[fix] tmp for test
2024-03-14 16:13:00 +08:00
d02e257abd
Merge branch 'feature/colossal-infer' into colossal-infer-cuda-graph
2024-03-14 10:37:05 +08:00
f366a5ea1f
[Inference/kernel]Add Fused Rotary Embedding and KVCache Memcopy CUDA Kernel ( #5418 )
...
* add rotary embedding kernel
* add rotary_embedding_kernel
* add fused rotary_emb and kvcache memcopy
* add fused_rotary_emb_and_cache_kernel.cu
* add fused_rotary_emb_and_memcopy
* fix bugs in fused_rotary_emb_and_cache_kernel.cu
* fix ci bugs
* use vec memcopy and opt the gloabl memory access
* fix code style
* fix test_rotary_embdding_unpad.py
* codes revised based on the review comments
* fix bugs about include path
* rm inline
2024-03-13 17:20:03 +08:00
f7aecc0c6b
feat rmsnorm cuda kernel and add unittest, benchmark script ( #5417 )
2024-03-08 16:21:12 +08:00
cefaeb5fdd
[feat] cuda graph support and refactor non-functional api
2024-03-08 14:19:35 +08:00
600881a8ea
[Inference]Add CUDA KVCache Kernel ( #5406 )
...
* add cuda KVCache kernel
* annotation benchmark_kvcache_copy
* add use cuda
* fix import path
* move benchmark scripts to example/
* rm benchmark codes in test_kv_cache_memcpy.py
* rm redundancy codes
* rm redundancy codes
* pr was modified according to the review
2024-02-28 14:36:50 +08:00
2a718c8be8
Optimized the execution interval time between cuda kernels caused by view and memcopy ( #5390 )
...
* opt_view_and_memcopy
* fix bugs in ci
* fix ci bugs
* update benchmark scripts
* fix ci bugs
2024-02-21 13:23:57 +08:00
730103819d
[Inference]Fused kv copy into rotary calculation ( #5383 )
...
* revise rotary embedding
* remove useless print
* adapt
* fix
* add
* fix
* modeling
* fix
* fix
* fix
* fused kv copy
* fused copy
* colossalai/kernel/triton/no_pad_rotary_embedding.py
* del padding llama
* del
2024-02-21 11:31:48 +08:00
b21aac5bae
[Inference] Optimize and Refactor Inference Batching/Scheduling ( #5367 )
...
* add kvcache manager funcs for batching
* add batch bucket for batching
* revise RunningList struct in handler
* add kvcache/batch funcs for compatibility
* use new batching methods
* fix indexing bugs
* revise abort logic
* use cpu seq lengths/block tables
* rm unused attr in Sequence
* fix type conversion/default arg
* add and revise pytests
* revise pytests, rm unused tests
* rm unused statements
* fix pop finished indexing issue
* fix: use index in batch when retrieving inputs/update seqs
* use dict instead of odict in batch struct
* arg type hinting
* fix make compress
* refine comments
* fix: pop_n_seqs to pop the first n seqs
* add check in request handler
* remove redundant conversion
* fix test for request handler
* fix pop method in batch bucket
* fix prefill adding
2024-02-19 17:18:20 +08:00
1f8c7e7046
[Inference] User Experience: update the logic of default tokenizer and generation config. ( #5337 )
...
* add
* fix
* fix
* pause
* fix
* fix pytest
* align
* fix
* license
* fix
* fix
* fix readme
* fix some bugs
* remove tokenizer config
2024-02-07 17:55:48 +08:00
6fb4bcbb24
[Inference/opt] Fused KVCahce Memcopy ( #5374 )
...
* fused kv memcopy
* add TODO in test_kvcache_copy.py
2024-02-07 17:15:42 +08:00
8106ede07f
Revert "[Inference] Adapt to Fused rotary ( #5348 )" ( #5373 )
...
This reverts commit 9f4ab2eb92
2024-02-07 14:27:04 +08:00
9f4ab2eb92
[Inference] Adapt to Fused rotary ( #5348 )
...
* revise rotary embedding
* remove useless print
* adapt
* fix
* add
* fix
* modeling
* fix
* fix
* fix
2024-02-07 11:36:04 +08:00
35382a7fbf
[Inference]Fused the gate and up proj in mlp，and optimized the autograd process. ( #5365 )
...
* fused the gate and up proj in mlp
* fix code styles
* opt auto_grad
* rollback test_inference_engine.py
* modifications based on the review feedback.
* fix bugs in flash attn
* Change reshape to view
* fix test_rmsnorm_triton.py
2024-02-06 19:38:25 +08:00
1dedb57747
[Fix/Infer] Remove unused deps and revise requirements ( #5341 )
...
* remove flash-attn dep
* rm padding llama
* revise infer requirements
* move requirements out of module
2024-02-06 17:27:45 +08:00
21ad4a27f9
[Inference/opt]Optimize the mid tensor of RMS Norm ( #5350 )
...
* opt rms_norm
* fix bugs in rms_layernorm
2024-02-02 15:06:01 +08:00
249644c23b
[Inference]Repalce Attention layer and MLP layer by shardformer to optimize the weight transpose operation，add fused_qkv and fused linear_add ( #5340 )
...
* add fused qkv
* replace attn and mlp by shardformer
* fix bugs in mlp
* add docstrings
* fix test_inference_engine.py
* add optimize unbind
* add fused_addmm
* rm squeeze(1)
* refactor codes
* fix ci bugs
* rename ShardFormerLlamaMLP and ShardFormerLlamaAttention
* Removed the dependency on LlamaFlashAttention2
* rollback test_inference_engine.py
2024-02-01 15:49:39 +08:00
5f98a9d68a
[Infer] Optimize Blocked KVCache And Kernels Using It ( #5325 )
...
* revise shape of kvcache (context attn kernel)
* revise shape of kvcache (flash decoding kernel)
* revise shape of kvcache (kvcache copy) and attn func
* init of kvcache in kvcache manager
* revise llama modeling
* revise block size retrieval
* use torch for rms_norm benchmarking
* revise block size retrieval
2024-01-30 16:06:09 +08:00
e8f0642f28
[Inference]Add Nopadding Llama Modeling ( #5327 )
...
* add nopadding llama modeling
* add nopadding_llama.py
* rm unused codes
* fix bugs in test_xine_copy.py
* fix code style
2024-01-30 10:31:46 +08:00
4f28cb43c0
[inference]Optimize the usage of the mid tensors space in flash attn ( #5304 )
...
* opt flash attn
* opt tmp tensor
* fix benchmark_llama
* fix code style
* fix None logic for output tensor
* fix adapted to get_xine_cache
* add comment
* fix ci bugs
* fix some codes
* rm duplicated codes
* rm duplicated codes
* fix code style
* add _get_dtype in config.py
2024-01-26 14:00:10 +08:00
bfff9254ac
[inference] Adapted to Rotary Embedding and RMS Norm ( #5283 )
...
* adapted to rotary_embedding
* adapted to nopad rms norm
* fix bugs in benchmark
* fix flash_decoding.py
2024-01-22 10:55:34 +08:00
6e487e7d3c
[kernel/fix] Performance Optimization for Decoding Kernel and Benchmarking ( #5274 )
...
* prevent re-creating intermediate tensors
* add singleton class holding intermediate values
* fix triton kernel api
* add benchmark in pytest
* fix kernel api and add benchmark
* revise flash decoding triton kernel in/out shapes
* fix calling of triton kernel in modeling
* fix pytest: extract to util functions
2024-01-19 15:47:16 +08:00
9e2342bde2
[Hotfix] Fix bugs in testing continuous batching ( #5270 )
...
* fix bug
* fix bugs
* fix bugs
* fix bugs and add padding
* add funcs and fix bugs
* fix typos
* fix bugs
* add func
2024-01-18 16:31:14 +08:00
86b63f720c
[Inference]Adapted to the triton attn kernels ( #5264 )
...
* adapted to the triton attn kernels
* fix pad input
* adapted to copy_kv_to_blocked_cache
* fix ci test
* update kv memcpy
* remove print
2024-01-17 16:03:10 +08:00
2a73e828eb
fix bugs related to processing padding mask
2024-01-11 13:46:14 +00:00
fa4fbdbffb
adapted to pad_context_forward
2024-01-11 13:44:06 +00:00
47e53eaa1c
fix bugs in attention.py and request_handler.py
2024-01-11 13:44:06 +00:00
3ad1f3b78b
fix beam_width
2024-01-11 13:39:56 +00:00
b2eb9cd186
Fixed a typo
2024-01-11 13:39:56 +00:00
Runyu Lu