Michael Goin
2a052011ca
[Kernel] Support MoE Fp8 Checkpoints for Mixtral (Static Weights with Dynamic/Static Activations) ( #4527 )
...
Follow on to #4332 to enable FP8 checkpoint loading for Mixtral and supersedes #4436 .
This PR enables the following checkpoint loading features for Mixtral:
Supports loading fp8 checkpoints for Mixtral, such as this "nm-testing/Mixtral-8x7B-Instruct-v0.1-FP8" test model
Supports static or dynamic activation quantization with static weight quantization (all per tensor)
Supports different scales for each expert weight
Supports Fp8 in QKV layer
Notes:
The Expert Gate/Router always runs at half / full precision for now.
If there are different weight scales between QKV layer (for separate QKV weights), they are re-quantized using layer.weight_scale.max() so we can have a single gemm for performance.
2024-05-04 11:45:16 -07:00
Cody Yu
bc8ad68455
[Misc][Refactor] Introduce ExecuteModelData ( #4540 )
2024-05-03 17:47:07 -07:00
Cade Daniel
ab50275111
[Speculative decoding] Support target-model logprobs ( #4378 )
2024-05-03 15:52:01 -07:00
Lily Liu
43c413ec57
[Kernel] Use flashinfer for decoding ( #4353 )
...
Co-authored-by: LiuXiaoxuanPKU <llilyliupku@gmail.com>
2024-05-03 15:51:27 -07:00
Sebastian Schoennenbeck
f8e7adda21
Fix/async chat serving ( #2727 )
2024-05-03 11:04:14 -07:00
SangBin Cho
3521ba4f25
[Core][Model runner refactoring 1/N] Refactor attn metadata term ( #4518 )
2024-05-03 10:20:12 -07:00
youkaichao
344a5d0c33
[Core][Distributed] enable allreduce for multiple tp groups ( #4566 )
2024-05-02 17:32:33 -07:00
SangBin Cho
0f8a91401c
[Core] Ignore infeasible swap requests. ( #4557 )
2024-05-02 14:31:20 -07:00
Michał Moskal
32881f3f31
[kernel] fix sliding window in prefix prefill Triton kernel ( #4405 )
...
Co-authored-by: SangBin Cho <rkooo567@gmail.com>
2024-05-02 11:23:37 -07:00
alexm-nm
7038e8b803
[Kernel] Support running GPTQ 8-bit models in Marlin ( #4533 )
2024-05-02 12:56:22 -04:00
youkaichao
2a85f93007
[Core][Distributed] enable multiple tp group ( #4512 )
...
Co-authored-by: Zhuohan Li <zhuohan123@gmail.com>
2024-05-02 04:28:21 +00:00
Ronen Schaffer
5e401bce17
[CI]Add regression tests to ensure the async engine generates metrics ( #4524 )
2024-05-01 19:57:12 -07:00
SangBin Cho
0d62fe58db
[Bug fix][Core] assert num_new_tokens == 1 fails when SamplingParams.n is not 1 and max_tokens is large & Add tests for preemption ( #4451 )
2024-05-01 19:24:13 -07:00
Danny Guinther
b8afa8b95a
[MISC] Rework logger to enable pythonic custom logging configuration to be provided ( #4273 )
2024-05-01 17:34:40 -07:00
sasha0552
c47ba4aaa9
[Bugfix] Add validation for seed ( #4529 )
2024-05-01 19:31:22 +00:00
Nick Hill
a657bfc48a
[Core] Add multiproc_worker_utils for multiprocessing-based workers ( #4357 )
2024-05-01 18:41:59 +00:00
leiwen83
24750f4cad
[Core] Enable prefix caching with block manager v2 enabled ( #4142 )
...
Co-authored-by: Lei Wen <wenlei03@qiyi.com>
Co-authored-by: Sage Moore <sagemoore@utexas.edu>
2024-05-01 11:20:32 -07:00
leiwen83
b38e42fbca
[Speculative decoding] Add ngram prompt lookup decoding ( #4237 )
...
Co-authored-by: Lei Wen <wenlei03@qiyi.com>
2024-05-01 11:13:03 -07:00
SangBin Cho
6f1df80436
[Test] Add ignore_eos test ( #4519 )
2024-05-01 08:45:42 -04:00
Jee Li
d6f4bd7cdd
[Misc]Add customized information for models ( #4132 )
2024-04-30 21:18:14 -07:00
Robert Caulk
c3845d82dc
Allow user to define whitespace pattern for outlines ( #4305 )
2024-04-30 20:48:39 -07:00
Florian Greinacher
a494140433
[Frontend] Support complex message content for chat completions endpoint ( #3467 )
...
Co-authored-by: Lily Liu <lilyliupku@gmail.com>
Co-authored-by: Cyrus Leung <tlleungac@connect.ust.hk>
2024-04-30 16:28:46 -07:00
Robert Shaw
111815d482
[Kernel] Support Fp8 Checkpoints (Dynamic + Static) ( #4332 )
...
Co-authored-by: Philipp Moritz <pcmoritz@gmail.com>
Co-authored-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
Co-authored-by: mgoin <michael@neuralmagic.com>
Co-authored-by: Tyler Michael Smith <tyler@neuralmagic.com>
Co-authored-by: Cody Yu <hao.yu.cody@gmail.com>
2024-04-30 21:46:12 +00:00
leiwen83
4bb53e2dde
[BugFix] fix num_lookahead_slots missing in async executor ( #4165 )
...
Co-authored-by: Lei Wen <wenlei03@qiyi.com>
2024-04-30 10:12:59 -07:00
youkaichao
f4f921b7f1
[Core][Distributed] use cpu group to broadcast metadata in cpu ( #4444 )
2024-04-29 13:52:22 -07:00
Robert Shaw
73c8d677e5
[Kernel] Marlin Expansion: Support AutoGPTQ Models with Marlin ( #3922 )
...
Co-authored-by: alexm <alexm@neuralmagic.com>
Co-authored-by: mgoin <michael@neuralmagic.com>
2024-04-29 09:35:34 -07:00
Prashant Gupta
d6e520e170
[Core] Support offline use of local cache for models ( #4374 )
...
Signed-off-by: Prashant Gupta <prashantgupta@us.ibm.com>
Co-authored-by: Travis Johnson <tjohnson31415@gmail.com>
2024-04-27 09:59:55 -07:00
Nick Hill
81661da7b2
[BugFix] Fix min_tokens when eos_token_id is None ( #4389 )
...
Co-authored-by: DefTruth <31974251+deftruth@users.noreply.github.com>
2024-04-27 09:52:46 -07:00
Ruoyu Qin
dfea173148
[Bugfix] Abort requests when the connection to /v1/completions is interrupted ( #4363 )
2024-04-27 09:48:37 -07:00
Roy
7134303cbb
[Bugfix][Core] Fix get decoding config from ray ( #4335 )
2024-04-27 11:30:08 +00:00
Austin Veselka
eefeb16464
[Kernel] Full Tensor Parallelism for LoRA Layers ( #3524 )
...
Co-authored-by: Antoni Baum <antoni.baum@protonmail.com>
2024-04-27 00:03:48 -07:00
Cyrus Leung
8947bc3c15
[Frontend][Bugfix] Disallow extra fields in OpenAI API ( #4355 )
2024-04-27 05:08:24 +00:00
Cody Yu
a62aaf1df5
[Misc][Refactor] Generalize linear_method to be quant_method ( #4373 )
2024-04-26 16:41:14 -04:00
SangBin Cho
603ad84815
[Core] Refactoring sampler and support prompt logprob for chunked prefill ( #4309 )
2024-04-26 13:02:02 +00:00
Cyrus Leung
a74dee9b62
[Bugfix] Fix parameter name in get_tokenizer ( #4107 )
2024-04-25 19:10:48 -07:00
Woosuk Kwon
468d761b32
[Misc] Reduce supported Punica dtypes ( #4304 )
2024-04-23 18:54:33 -07:00
youkaichao
91f50a6fe2
[Core][Distributed] use cpu/gloo to initialize pynccl ( #4248 )
2024-04-23 18:32:19 -07:00
Cyrus Leung
1e8f4252aa
[Bugfix][Frontend] Raise exception when file-like chat template fails to be opened ( #4292 )
2024-04-23 18:19:03 +00:00
James Fleming
2b7949c1c2
AQLM CUDA support ( #3287 )
...
Co-authored-by: mgoin <michael@neuralmagic.com>
2024-04-23 13:59:33 -04:00
Cade Daniel
62b8aebc6f
[Speculative decoding 7/9] Speculative decoding end-to-end correctness tests. ( #3951 )
2024-04-23 08:02:36 +00:00
SangBin Cho
050f285ff6
[Core] Scheduling optimization 2 ( #4280 )
2024-04-23 08:02:11 +00:00
SangBin Cho
ad8d696a99
[Core] Scheduler perf fix ( #4270 )
2024-04-22 21:11:06 +00:00
GeauxEric
a37d815b83
Make initialization of tokenizer and detokenizer optional ( #3748 )
...
Co-authored-by: Yun Ding <yunding@nvidia.com>
Co-authored-by: Roger Wang <ywang@roblox.com>
2024-04-21 22:06:46 +00:00
nunjunj
91528575ec
[Frontend] multiple sampling params support ( #3570 )
2024-04-20 00:11:57 -07:00
Cody Yu
a22cdea371
[Kernel][FP8] Initial support with dynamic per-tensor scaling ( #4118 )
...
Provide an initial support to FP8 computation. This PR is inspired by HuggingFace TGI: huggingface/text-generation-inference#1726
This feature can be enabled with --quantization fp8 or -q fp8 when launching an engine.
Algorithm:
We still load a model checkpoint in FP16/BF16. After the weights are loaded, Fp8LinearMethod calculates the per-tensor scaling factor of weights and quantizes the weights accordingly. The scaling factor will then be stored for future use. Meanwhile, the per-tensor scaling factor for activations is calculated in every forward pass.
Initial Results:
Currently tested Mistral-7B on 1xH100. With prompt length ~5 and decoding length 128:
BF16: 1.47s
FP8: 1.66s
I'll try to use larger models and try to find more performance bottleneck. Meanwhile, you're welcome to try this code.
2024-04-20 04:28:57 +00:00
Ayush Rautwar
138485a82d
[Bugfix] Add fix for JSON whitespace ( #4189 )
...
Co-authored-by: Ubuntu <ubuntu@ip-172-31-13-147.ec2.internal>
2024-04-19 20:49:22 -07:00
Jee Li
d17c8477f1
[Bugfix] Fix LoRA loading check ( #4138 )
...
Co-authored-by: simon-mo <simon.mo@hey.com>
2024-04-19 00:59:54 -07:00
youkaichao
8a7a3e4436
[Core] add an option to log every function call to for debugging hang/crash in distributed inference ( #4079 )
...
Co-authored-by: Simon Mo <simon.mo@hey.com>
2024-04-18 16:15:12 -07:00
James Whedbee
e1bb2fd52d
[Bugfix] Support logprobs when using guided_json and other constrained decoding fields ( #4149 )
2024-04-18 21:12:55 +00:00
Michał Moskal
e8cc7967ff
[Bugfix][Kernel] allow non-power-of-two head sizes in prefix prefill ( #4128 )
2024-04-18 00:51:28 -07:00
