Commit Graph

Select branches

Hide Pull Requests

main

patches

sm86

Mono Color

• c5d7fb9ddc [Doc] fix third-party model example (#9771) Russell Bryant 2024-10-28 22:39:21 -04:00
• 76ed5340f0 [torch.compile] add deepseek v2 compile (#9775) youkaichao 2024-10-28 14:35:17 -07:00
• 97b61bfae6 [misc] avoid circular import (#9765) youkaichao 2024-10-28 13:51:23 -07:00
• aa0addb397 Adding "torch compile" annotations to moe models (#9758) Yongzao 2024-10-29 04:49:56 +08:00
• 5f8d8075f9 [Model][VLM] Add multi-video support for LLaVA-Onevision (#8905) 3 people 2024-10-29 02:04:10 +08:00
• 8b0e4f2ad7 [CI/Build] Adopt Mergify for auto-labeling PRs (#9259) Russell Bryant 2024-10-28 12:38:09 -04:00
• 2adb4409e0 [Bugfix] Fix ray instance detect issue (#9439) Yan Ma 2024-10-28 15:13:03 +08:00
• feb92fbe4a Fix beam search eos (#9627) Robert Shaw 2024-10-28 02:59:37 -04:00
• 32176fee73 [torch.compile] support moe models (#9632) youkaichao 2024-10-27 21:58:04 -07:00
• 4e2d95e372 [Hardware][ROCM] using current_platform.is_rocm (#9642) wangshuai09 2024-10-28 12:07:00 +08:00
• 34a9941620 [Bugfix] Fix load config when using bools (#9533) madt2709 2024-10-27 10:46:41 -07:00
• e130c40e4e Fix cache management in "Close inactive issues and PRs" actions workflow (#9734) Harry Mellor 2024-10-27 17:30:03 +00:00
• 3cb07a36a2 [Misc] Upgrade to pytorch 2.5 (#9588) bnellnmandyoukaichao 2024-10-27 05:44:24 -04:00
• 8549c82660 [core] cudagraph output with tensor weak reference (#9724) youkaichao 2024-10-27 00:19:28 -07:00
• 67a6882da4 [Misc] SpecDecodeWorker supports profiling (#9719) 科英 2024-10-27 12:18:03 +08:00
• 6650e6a930 [Model] Add classification Task with Qwen2ForSequenceClassification (#9704) kakao-kevin-usandKevin-Yang 2024-10-27 02:53:35 +09:00
• 07e981fdf4 [Frontend] Bad words sampling parameter (#9717) Vasiliy Alekseev 2024-10-26 19:29:38 +03:00
• 55137e8ee3 Fix: MI100 Support By Bypassing Custom Paged Attention (#9560) ErkinSagiroglu 2024-10-26 13:12:57 +01:00
• 5cbdccd151 [Hardware][openvino] is_openvino --> current_platform.is_openvino (#9716) Mengqing Cao 2024-10-26 18:59:06 +08:00
• 067e77f9a8 [Bugfix] Steaming continuous_usage_stats default to False (#9709) Sam Stoelinga 2024-10-25 22:05:47 -07:00
• 6567e13724 [Bugfix] Fix crash with llama 3.2 vision models and guided decoding (#9631) 3 people 2024-10-25 16:42:56 -06:00
• 228cfbd03f [Doc] Improve quickstart documentation (#9256) Rafael Vasquez 2024-10-25 17:32:10 -04:00
• ca0d92227e [Bugfix] Fix compressed_tensors_moe bad config.strategy (#9677) Michael Goin 2024-10-25 15:40:33 -04:00
• 9645b9f646 [V1] Support sliding window attention (#9679) Woosuk Kwon 2024-10-24 22:20:37 -07:00
• a6f3721861 [Model] add a lora module for granite 3.0 MoE models (#9673) Will Johnson 2024-10-25 01:00:17 -04:00
• 9f7b4ba865 [ci/Build] Skip Chameleon for transformers 4.46.0 on broadcast test #9675 (#9676) Kevin H. Luu 2024-10-24 17:59:00 -10:00
• c91ed47c43 [Bugfix] Remove xformers requirement for Pixtral (#9597) Michael Goin 2024-10-24 18:38:05 -04:00
• 59449095ab [Performance][Kernel] Fused_moe Performance Improvement (#9384) Charlie Fu 2024-10-24 17:37:52 -05:00
• e26d37a185 [Log][Bugfix] Fix default value check for image_url.detail (#9663) Michael Goin 2024-10-24 13:44:38 -04:00
• 722d46edb9 [Model] Compute Llava Next Max Tokens / Dummy Data From Gridpoints (#9650) Alex Brooks 2024-10-24 11:42:24 -06:00
• c866e0079d [CI/Build] Fix VLM test failures when using transformers v4.46 (#9666) Cyrus Leung 2024-10-25 01:40:40 +08:00
• d27cfbf791 [torch.compile] Adding torch compile annotations to some models (#9641) Yongzaoandyoukaichao 2024-10-25 00:31:42 +08:00
• de662d32b5 Increase operation per run limit for "Close inactive issues and PRs" workflow (#9661) Harry Mellor 2024-10-24 17:17:45 +01:00
• f58454968f [Bugfix]Disable the post_norm layer of the vision encoder for LLaVA models (#9653) litianjian 2024-10-24 22:52:07 +08:00
• b979143d5b [Doc] Move additional tips/notes to the top (#9647) Cyrus Leung 2024-10-24 17:43:59 +08:00
• ad6f78053e [torch.compile] expanding support and fix allgather compilation (#9637) Yongzaoandyoukaichao 2024-10-24 16:32:15 +08:00
• 295a061fb3 [Kernel] add kernel for FATReLU (#9610) Jee Jee Li 2024-10-24 16:18:27 +08:00
• 8a02cd045a [torch.compile] Adding torch compile annotations to some models (#9639) Yongzaoandyoukaichao 2024-10-24 15:54:57 +08:00
• 4fdc581f9e [core] simplify seq group code (#9569) youkaichaoandZhuohan Li 2024-10-24 00:16:44 -07:00
• 3770071eb4 [V1][Bugfix] Clean up requests when aborted (#9629) Woosuk Kwon 2024-10-23 23:33:22 -07:00
• 836e8ef6ee [Bugfix] Fix PP for ChatGLM and Molmo (#9422) Cyrus Leung 2024-10-24 14:12:05 +08:00
• 056a68c7db [XPU] avoid triton import for xpu (#9440) 3 people 2024-10-24 13:14:00 +08:00
• 33bab41060 [Bugfix]: Make chat content text allow type content (#9358) Vinay R Damodaran 2024-10-24 01:05:49 -04:00
• b7df53cd42 [Bugfix] Use "vision_model" prefix for MllamaVisionModel (#9628) Michael Goin 2024-10-23 22:07:44 -04:00
• bb01f2915e [Bugfix][Model] Fix Mllama SDPA illegal memory access for batched multi-image (#9626) Michael Goin 2024-10-23 22:03:44 -04:00
• b548d7a5f4 [CI/Build] Add bot to close stale issues and PRs (#9436) Russell Bryant 2024-10-23 18:45:26 -04:00
• fc6c274626 [Model] Add Qwen2-Audio model support (#9248) Yunfei ChuandDarkLight1337 2024-10-24 01:54:22 +08:00
• 150b779081 [Frontend] Enable Online Multi-image Support for MLlama (#9393) Alex BrooksandCyrus Leung 2024-10-23 11:28:57 -06:00
• 9013e24f7b [torch.compile] Adding torch compile annotations to some models (#9614) Yongzao 2024-10-24 01:07:48 +08:00
• fd0e2cfdb2 [Misc] Separate total and output tokens in benchmark_throughput.py (#8914) Michael Goin 2024-10-23 12:47:20 -04:00
• e5ac6a4199 [Bugfix] Fix divide by zero when serving Mamba models (#9617) Tyler Michael Smith 2024-10-23 12:40:43 -04:00
• dbdd3b5e5a [misc] comment to avoid future confusion about baichuan (#9620) youkaichao 2024-10-23 09:14:44 -07:00
• e7116c017c [Bugfix] Fix _init_vision_model in NVLM_D model (#9611) Cyrus LeungandIsotr0py 2024-10-23 22:09:04 +08:00
• 31a08f5bd2 [Model] Add min_pixels / max_pixels to Qwen2VL as mm_processor_kwargs (#9612) Alex Brooks 2024-10-23 08:05:18 -06:00
• c18e1a3418 [VLM] Enable overriding whether post layernorm is used in vision encoder + fix quant args (#9217) Cyrus LeungandIsotr0py 2024-10-23 19:27:37 +08:00
• 3ff57ebfca [Model] Initialize Florence-2 language backbone support (#9555) Isotr0py 2024-10-23 18:42:47 +08:00
• 2394962d70 [Hardware][XPU] using current_platform.is_xpu (#9605) Mengqing Cao 2024-10-23 16:28:21 +08:00
• 51c24c9736 [Build] Fix FetchContent multiple build issue (#9596) Luka Govedič 2024-10-23 00:43:07 -04:00
• 831540cf04 [Model] Support E5-V (#9576) Cyrus Leung 2024-10-23 11:35:29 +08:00
• 29061ed9df [Misc] Add an env var VLLM_LOGGING_PREFIX, if set, it will be prepend to all logging messages (#9590) Flex Wang 2024-10-22 20:17:28 -07:00
• 65050a40e6 [Bugfix] Generate exactly input_len tokens in benchmark_throughput (#9592) Chen Zhang 2024-10-22 17:45:35 -07:00
• 208cb34c81 [Doc]: Update tensorizer docs to include vllm[tensorizer] (#7889) Seth KimmelandKaunil Dhruv 2024-10-22 15:43:25 -07:00
• b17046e298 [BugFix] Fix metrics error for --num-scheduler-steps > 1 (#8234) yulei 2024-10-23 06:43:03 +08:00
• d1e8240875 [Bugfix] Fix spurious "No compiled cutlass_scaled_mm ..." for W8A8 on Turing (#9487) Lucas Wilkinson 2024-10-22 18:41:13 -04:00
• cb6fdaa0a0 [Misc] Make benchmarks use EngineArgs (#9529) Jeremy Arnold 2024-10-22 17:40:38 -05:00
• 23b899a8e6 [Bugfix] fix detokenizer shallow copy (#5919) Aurick Qiao 2024-10-22 18:38:12 -04:00
• 17c79f3c36 [torch.compile] auto infer dynamic_arg_dims from type annotation (#9589) youkaichao 2024-10-22 13:43:37 -07:00
• cd5601ac37 [BugFix] Prevent exporting duplicate OpenTelemetry spans (#9017) Ronen Schaffer 2024-10-22 21:11:53 +03:00
• 434984e665 [Frontend] Support custom request_id from request (#9550) Yuhong GuoandYuhong Guo 2024-10-23 02:07:30 +08:00
• 32a1ee74a0 [Hardware][Intel CPU][DOC] Update docs for CPU backend (#6212) 4 people 2024-10-22 10:38:04 -07:00
• 08075c3448 [Bugfix] Eagle: change config name for fc bias (#9580) gopalsarda 2024-10-22 21:44:22 +05:30
• bb392ea2d2 [Model][VLM] Initialize support for Mono-InternVL model (#9528) Isotr0py 2024-10-23 00:01:46 +08:00
• 9dbcce84a7 [Neuron] [Bugfix] Fix neuron startup (#9374) xendoandJerzy Zagorski 2024-10-22 14:51:41 +02:00
• a48e3ec052 [CI/Build][LoRA] Temporarily fix long context failure issue (#9579) Jee Jee Li 2024-10-22 19:32:51 +08:00
• 6c5af09b39 [V1] Implement vLLM V1 [1/N] (#9289) Woosuk Kwon 2024-10-22 01:24:07 -07:00
• 3ddbe25502 [Hardware][CPU] using current_platform.is_cpu (#9536) wangshuai09 2024-10-22 15:50:43 +08:00
• 0d02747f2e support TP in qwen2 bnb (#9574) chenqianfzh 2024-10-22 00:13:23 -07:00
• f7db5f0fa9 [Doc] Use shell code-blocks and fix section headers (#9508) Rafael Vasquez 2024-10-22 02:43:24 -04:00
• ca30c3c84b [Core] Remove evictor_v1 (#9572) Kuntai Du 2024-10-21 23:55:49 -05:00
• c0292211ce [CI/Build] Replaced some models on tests for smaller ones (#9570) Wallas Henrique 2024-10-22 01:52:14 -03:00
• 74692421f7 [Bugfix]: phi.py get rope_theta from config file (#9503) Falko1andIsotr0py 2024-10-22 04:53:36 +02:00
• 29acd2c34c [Bugfix][OpenVINO] fix_dockerfile_openvino (#9552) ngrozae 2024-10-22 04:47:52 +02:00
• f085995a7b [CI/Build] Remove unnecessary fork_new_process (#9484) Cyrus Leung 2024-10-22 10:47:29 +08:00
• b729901139 [Bugfix]: serialize config by value for --trust-remote-code (#6751) Travis JohnsonandCyrus Leung 2024-10-21 20:46:24 -06:00
• 76a5e13270 [core] move parallel sampling out from vllm core (#9302) youkaichao 2024-10-21 17:31:44 -07:00
• ef7faad1b8 🐛 Fixup more test failures from memory profiling (#9563) Joe Runde 2024-10-21 19:10:56 -05:00
• 575dcebe9a [CI] Make format checker error message more user-friendly by using emoji (#9564) Kuntai Du 2024-10-21 18:45:15 -05:00
• 711f3a7806 [Frontend] Don't log duplicate error stacktrace for every request in the batch (#9023) Wallas Henrique 2024-10-21 18:49:41 -03:00
• 15713e3b75 [BugFix] Update draft model TP size check to allow matching target TP size (#9394) Nick HillandBaoyuan Qi 2024-10-21 22:14:29 +01:00
• d621c43df7 [doc] fix format (#9562) youkaichao 2024-10-21 13:54:57 -07:00
• 9d9186be97 [Frontend] Reduce frequency of client cancellation checking (#7959) Nick Hill 2024-10-21 21:28:10 +01:00
• 5241aa1494 [Model][Bugfix] Fix batching with multi-image in PixtralHF (#9518) Michael Goin 2024-10-21 14:20:07 -04:00
• ec6bd6c4c6 [BugFix] Use correct python3 binary in Docker.ppc64le entrypoint (#9492) Varad Ahirwadkar 2024-10-21 23:13:02 +05:30
• 8ca8954841 [Bugfix][Misc]: fix graph capture for decoder (#9549) yudian0504 2024-10-22 01:33:30 +08:00
• f6b97293aa [Model] FalconMamba Support (#9325) Dhia Eddine Rhaiem 2024-10-21 20:50:16 +04:00
• 496e991da8 [Doc] Consistent naming of attention backends (#9498) Thomas Parnell 2024-10-21 16:29:57 +02:00
• 696b01af8f [CI/Build] Split up decoder-only LM tests (#9488) Cyrus LeungandNick Hill 2024-10-21 12:27:50 +08:00
• 855e0e6f97 [Frontend][Misc] Goodput metric support (#9338) Andy Dai 2024-10-20 11:39:32 -07:00
• 4fa3e33349 [Kernel] Support sliding window in flash attention backend (#9403) Chen Zhang 2024-10-20 10:57:52 -07:00
• 962d2c6349 [Model][Pixtral] Use memory_efficient_attention for PixtralHFVision (#9520) Michael Goin 2024-10-20 01:29:14 -04:00