Commit Graph

Select branches

Hide Pull Requests

main

patches

sm86

Mono Color

• 0ec8422171 [Bugfix] Fix incorrect channel order for idefics3 in edge case (#29881) Isotr0pyandgemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> 2025-12-03 00:03:52 +08:00
• 2eb4fe9129 [examples] Resettle pooling examples. (#29365) wang.yuqiandHarry Mellor 2025-12-02 23:54:28 +08:00
• 51c57b51dd [Bugfix] Fix DeepSeek R1 MTP weight loading (#29545) Matthew BonanniandBenjamin Chislett 2025-12-02 10:52:18 -05:00
• 60c3d413af [Multimodal][Core] Optimize multimodal preprocessing cache by hashing image bytes instead of pixel values (#29621) ImaGoodFellaandCyrus Leung 2025-12-02 14:49:02 +01:00
• 68ffbca7e4 [Chore] Use tokenizer.encode and tokenizer.decode directly (#29851) Cyrus Leung 2025-12-02 20:30:40 +08:00
• 951445a52d Remove default values from InitVars so that they're not stored (#29859) Harry Mellor 2025-12-02 12:16:37 +00:00
• d8c6210eea Add Mistral Large 3 and Ministral 3 (#29757) 3 people 2025-12-02 11:29:00 +01:00
• 8bbcf8b6e7 [vLLM Benchmark Suite] Add default parameters section and update CPU benchmark cases (#29381) 3 people 2025-12-02 01:00:23 -08:00
• 70fb77b4dc [BugFix] add max-num-batched-token to scheduler hash (#29829) Boyuan Feng 2025-12-02 00:55:02 -08:00
• 48d15a32aa [CI] Fix Bad_words test for tokenizer encode/decode asymmetry (#28193) 杰兮andzhyajie 2025-12-02 16:02:12 +08:00
• 3b221cb661 [BugFix] respect VLLM_LOGGING_LEVEL in logger (#29761) Boyuan Feng 2025-12-01 23:49:16 -08:00
• 0037b5746a [Core] Eliminate redundant is_encoder_decoder lookups (20-40us/step) (#29800) Wushi Dong 2025-12-01 23:08:07 -08:00
• f5b0846ba0 Fix some Transformers nightly tests (#29802) Harry Mellor 2025-12-02 07:05:27 +00:00
• 13ea39bc09 [CPU]Parallelize over tokens in int4 moe (#29600) Zhang Xiangze 2025-12-02 14:21:39 +08:00
• 4b612664fd [CI] Renovation of nightly wheel build & generation (take 2) (#29838) Shengqi Chen 2025-12-02 14:17:10 +08:00
• 653591d5e7 [Chore] Move tokenizer initialization methods (#29793) Cyrus Leung 2025-12-02 13:33:37 +08:00
• e2fbfc955e [CI][AMD] spec_decode:eagle skip FLASH_ATTN for deepseek on ROCm (#29827) Divakar Verma 2025-12-01 23:27:46 -06:00
• a690fb5bd6 [CI][ROCm] Fix test_correctness_sliding_window (#29243) Divakar VermaandCyrus Leung 2025-12-01 22:53:27 -06:00
• 81fe3f82af [BugFix] Fix index error in ngram_proposer (#29779) usberkeley 2025-12-02 12:48:11 +08:00
• 53bf71b0f0 [Misc] Update conftest for entrypoints/sagemaker test folder (#29799) Zuyi Zhao 2025-12-01 19:56:39 -08:00
• f441d36cee Add missing return in _check_vllm_model_embed_input_ids (#29834) Johnny Yang 2025-12-01 19:22:50 -08:00
• 22274b2184 [Misc] Add ReplicaId to Ray metrics (#24267) Seiji Eicherandrongfu.leng 2025-12-01 19:21:44 -08:00
• fc95521ba5 [Misc] Throw error on unintended access to scheduler_config.max_model_len (#29771) Wei Wei 2025-12-01 18:58:44 -08:00
• d0cd728907 [Core] Support reseting all running requests' KV while calling reset_prefix_cache (#28827) Zhuohan LiandNick Hill 2025-12-01 18:25:05 -08:00
• fa8804ad9c [responsesAPI][4] fix responseOutputItem Kimi K2 thinking bug (#29555) Andrew XiaandAndrew Xia 2025-12-01 18:11:35 -08:00
• 4b40924998 [ROCm] Fallback pytorch GELU with tanh approximation to GELU() (#29244) Divakar Vermaandgemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> 2025-12-01 20:02:22 -06:00
• c0dfc89485 SM120 / NVFP4: add device guard and runtime SM dispatch to cutlass_scaled_fp4_mm (#29711) Hendrik Holtmannandmgoin 2025-12-02 02:24:18 +01:00
• 44822d7ff2 [BugFix] Preserve spec decoding uniform decode when scheduling (#29759) Nick Hill 2025-12-01 17:15:52 -08:00
• 342c4f1472 Updated CI mirror 2025-11-25 (#29434) Alexei-V-Ivanov-AMDandKevin H. Luu 2025-12-01 17:44:33 -06:00
• 1336a1ea24 Revert #29787 and #29690 (#29815) Kevin H. Luu 2025-12-01 13:42:03 -08:00
• eaf81485ed [Ascend]: Fixed the issue where OOT Platform vllm-ascend could not enable SP in Eager mode (#28935) Nengjun Ma 2025-12-02 04:02:18 +08:00
• 38caf7fa1a Update FAQ on interleaving sliding windows support (#29796) Finbarr Timbers 2025-12-01 12:15:19 -07:00
• cabc77cc86 [Core][Observability] Add KV cache residency metrics (#27793) shivampr 2025-12-01 10:27:53 -08:00
• ec7035c9d4 [ci] Make distributed 8 gpus test optional (#29801) Kevin H. Luu 2025-12-01 10:22:05 -08:00
• fc6acc88ca [Bugfix] Missing cached item in the MultiModalReceiverCache (#28525) 4 people 2025-12-02 02:18:07 +08:00
• d0985c5feb [Hardware][AMD] Remove ROCm skip conditions for transformers backend tests (#29782) BADAOUI Abdennacer 2025-12-01 19:03:13 +01:00
• 092bb73b8a [Frontend] add 'verbose_json' and 'timestamp' feature on Whisper Transcription/Translation (#24209) sangbumlikeagod 2025-12-02 02:19:17 +09:00
• 5d43f7372e [Doc] Update description disable_any_whitespace (#29784) FredericOdermatt 2025-12-01 17:48:33 +01:00
• 37593deb02 [CI] fix url-encoding behavior in nightly metadata generation (#29787) Shengqi Chen 2025-12-01 23:17:20 +08:00
• f5516039c5 [Doc] fix heading levels (#29783) Liu Jinyi 2025-12-01 22:49:22 +08:00
• 36db0a35e4 [CI] Renovation of nightly wheel build & generation (#29690) Shengqi Chen 2025-12-01 21:25:39 +08:00
• 5cfa967efa [Bugfix] TypeError: 'NoneType' object is not callable (#29414) Marcin Ostrowski 2025-12-01 14:16:44 +01:00
• b95db244ee [v1] Add real sliding window calculation to FlexAttention direct BlockMask building (#26015) Isotr0pyandbaonudesifeizhai 2025-12-01 21:12:51 +08:00
• ad9d656bfa [multimodal][test] Reduce memory utilization for test_siglip to avoid OOM (#29504) Zhengxu ChenandCyrus Leung 2025-12-01 07:41:48 -05:00
• f37e8938d2 [XPU] Fix AWQ skipped layer detection in IPEX quantization (#29774) Fanli Lin 2025-12-01 20:00:52 +08:00
• f0a28bf661 [Misc] Unify tokenizer registration (#29767) Cyrus Leung 2025-12-01 19:34:58 +08:00
• 86e178f7c4 [crashfix] Eagle + multimodal can crash on mm cache miss (#29750) Mickaël SeznecandRoger Wang 2025-12-01 10:29:33 +01:00
• 014ece97c7 [Frontend] Add tool filtering support to ToolServer (#29224) daniel-salibandChauncey 2025-12-01 03:03:57 -05:00
• 62de4f4257 [Frontend] Resettle pooling entrypoints (#29634) wang.yuqi 2025-12-01 15:30:43 +08:00
• 83805a6078 [CI] Skip paddleocr_vl for transformer 4.57.3 (#29758) Huamin Li 2025-11-30 20:38:06 -08:00
• 1ab8fc8197 Make PyTorch profiler gzip and CUDA time dump configurable (#29568) Yifei Zhang 2025-12-01 12:30:46 +08:00
• f72a817bdf [MoE] CuteDSL MoE with Nvfp4 DeepEP dispatch (#27141) 3 people 2025-11-30 18:05:32 -06:00
• ec38a7368d [Model Runner V2] Use packed mask for prompt bin counts (#29756) Woosuk Kwon 2025-11-30 14:15:42 -08:00
• 21c2627934 [Misc]Remove redundant hidden_size property in ModelConfig (#29749) Xingyu LiuandHarry Mellor 2025-12-01 01:14:23 +08:00
• 39d28108f4 [Feat] Support non-gated activations in NVFP4 modelopt path (#29004) Omer Ullman Argov 2025-11-30 18:02:40 +02:00
• cd719de5cb Fix RoPE failures in Transformers nightly (#29700) Harry Mellor 2025-11-30 14:29:32 +00:00
• 8c363ed666 [ROCm][Attention] Sliding window support for AiterFlashAttentionBackend (#29234) Pleaplusone 2025-11-30 19:31:50 +08:00
• 64bc09ba27 [Core] Enable inputs_embeds_size separate from hidden_size (#29741) Cyrus Leung 2025-11-30 17:31:12 +08:00
• 47539cfd3e [Bugfix] Fix mismatched nvfp4 gemm output shape (#29742) Isotr0py 2025-11-30 17:15:01 +08:00
• 2afcec4dec [Misc] Update TokenizerLike interface and move get_cached_tokenizer (#29730) Cyrus Leung 2025-11-30 14:59:47 +08:00
• 9381b5cde0 [Doc]: Fix typo in fused_moe layer (#29731) 朝 2025-11-30 14:29:13 +08:00
• 66b5840287 [Bugfix][sleepmode][fp8 kv cache]: Fix FP8 KV cache + sleep(level=2) gibberish output (#28783) VensenandTJian 2025-11-30 14:24:25 +08:00
• 82c795d6f2 Fix AttributeError about _use_fi_prefill (#29734) Huamin Li 2025-11-29 22:04:55 -08:00
• e1464c3a08 [Quantization] Enable compressed-tensors AWQ for Turing GPU (#29732) Isotr0py 2025-11-30 14:04:28 +08:00
• a491b0911b [LoRA] Support FusedMoE LoRA Triton kernel for mxfp4 (#29708) Xin YangandJee Jee Li 2025-11-29 18:37:25 -08:00
• b9d0504a36 [Bugfix] Revert test_tokenization.py (#29729) Jee Jee Li 2025-11-30 00:35:15 +08:00
• 1656ad3704 [Kernel][Quantization] add w4a8 support for marlin kernel (#24722) 3 people 2025-11-29 23:19:33 +08:00
• fa59fe417f [Chore] Move detokenizer_utils to vllm/tokenizers (#29727) Cyrus Leung 2025-11-29 22:25:17 +08:00
• fe3398fab2 [Chore] Enable passing tokenizer=None into MM processor (#29724) Cyrus Leung 2025-11-29 22:25:10 +08:00
• ad7f714d62 hfrunner.classify should return list[list[float]] not list[str] (#29671) Chukwuma Nwaugha 2025-11-29 13:57:00 +00:00
• f4341f45d3 [Doc]: fix code block rendering (#29728) dublc 2025-11-29 21:46:48 +08:00
• 34a984274e [Misc] Refactor tokenizer interface (#29693) Cyrus Leung 2025-11-29 20:02:21 +08:00
• f223ed4181 [Model Runner V2] Fuse penalties and temperature into single kernel (#29720) Woosuk Kwon 2025-11-29 02:29:16 -08:00
• 04a797cd0e [Doc]: fixing typos in various files. (#29717) Didier Durand 2025-11-29 10:15:39 +01:00
• 6afc0ffaf6 [Model Runner V2] Add sample/ directory and reorganize files (#29719) Woosuk Kwon 2025-11-29 00:41:01 -08:00
• 39e63dec7c [LoRA] Cleanup LoRA unused code (#29611) Jee Jee LiandCyrus Leung 2025-11-29 14:52:58 +08:00
• 4a80ad0a25 [Model Runner V2] Don't use UVA buffer for prefill_len (#29713) Woosuk Kwon 2025-11-28 20:27:16 -08:00
• 4b17ce6815 Add gpu memory wait before test_async_tp (#28893) Angela YiandLuka Govedič 2025-11-28 20:19:05 -08:00
• e23f665d83 [BugFix] Fix DBO failing with TypeError: 'NoneType' object is not iterable (#29698) Lucas Wilkinson 2025-11-28 23:19:01 -05:00
• ca1b1e7296 [Model Runner V2] Refactor prefill token preparation (#29712) Woosuk Kwon 2025-11-28 19:49:17 -08:00
• 762a4a6ca9 [Frontend] Perform offline path replacement to tokenizer (#29706) Tsukasa OI 2025-11-29 11:32:08 +09:00
• b2c50eda50 [Bugfix] Fix wrong mock attribute (#29704) Cyrus Leung 2025-11-29 10:30:41 +08:00
• 1dcafb3dea [Model Runner V2] Support penalties using bin counts (#29703) Woosuk Kwon 2025-11-28 17:53:17 -08:00
• ea3370b428 [ROCm][Bugfix] Patch for the Multi-Modal Processor Test group (#29702) Andreas Karatzas 2025-11-28 19:31:44 -06:00
• c625d7b1c6 [Bugfix] Fix O(n²) multimodal string prompt processing (#29667) Mert UnsalandRoger Wang 2025-11-28 16:10:39 -08:00
• 6173682b6e [compile] Include enable_sleep_mode into caching factors. (#29696) Zhengxu Chen 2025-11-28 18:58:38 -05:00
• 9726e64530 bugfix: correct attn output with base 2 or e (#28840) Augusto Yao 2025-11-29 07:52:12 +08:00
• 3fd1fb0b60 Revert "[LoRA] Support FusedMoE LoRA Triton kernel for mxfp4 (#28971)" (#29697) Huamin Li 2025-11-28 15:26:52 -08:00
• a51f4186f2 [Bugfix] fix dots.llm1.inst (#29687) Jiangyun ZhuandRoger Wang 2025-11-29 07:25:26 +08:00
• 7675ba30de [Misc] Remove redundant ClassRegistry (#29681) Cyrus Leungandgemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> 2025-11-29 07:24:47 +08:00
• 7c1ed45848 [CI/Build]: make it possible to build with a free-threaded interpreter (#29241) Ralf GommersandCyrus Leung 2025-11-29 00:21:46 +01:00
• 1986de1375 [Perf] Optimize EAGLE prepare_inputs_padded with triton kernels (#28597) Benjamin Chislett 2025-11-28 17:25:05 -05:00
• 3461e7efd8 [Frontend] Remap -O to -cc commandline flag (#29557) Yanan CaoandClaude 2025-11-28 13:51:12 -08:00
• fecae12cd7 Remove all_special_tokens_extended from tokenizer code (#29686) Harry Mellor 2025-11-28 20:26:51 +00:00
• 8d9338fae4 [Chore] Rename Processor to InputProcessor (#29682) Cyrus Leung 2025-11-29 01:35:41 +08:00
• d40c854009 [CI/Build] Rework CPU multimodal processor test (#29684) Isotr0py 2025-11-29 01:10:29 +08:00
• 4332955602 [Docs] Add CLI reference doc for vllm bench sweep plot_pareto (#29689) Harry Mellor 2025-11-28 17:10:08 +00:00
• f946a8d743 [Chore]: Reorganize model repo operating functions in transformers_utils (#29680) Isotr0py 2025-11-29 00:46:51 +08:00
• 6f9d81d03b [V0 deprecation] Clean up legacy paged attention helper functions (#28043) Isotr0py 2025-11-29 00:44:33 +08:00
• fae6943068 [Doc]: fixing typos in multiple files. (#29685) Didier Durand 2025-11-28 17:41:41 +01:00