Commit Graph

Select branches

Hide Pull Requests

main

patches

sm86

Mono Color

• 6aa6ad8992 [BugFix] Fix implicit and incorrect assumption on ECConnector is_producer (#34783) Qi Wang 2026-03-04 06:01:30 -08:00
• c8c3935b70 [Bugfix][Model] Fix FP8 k_scale/v_scale not loaded for Qwen3-MoE (#35656) Raghavan 2026-03-04 18:45:38 +05:30
• bb6888b8b1 [Bugfix][CPUOffloadingManager] Prevent eviction of already-stored blocks in LRU/ARC prepare_store() (#35846) Ronen Schaffer 2026-03-04 14:25:33 +02:00
• 1aaec59d79 [MISC] fixed tool_parser mypy errors (#35640) Taneem IbrahimandHarry Mellor 2026-03-04 06:23:12 -06:00
• 1659b2e058 [Feature] Add basic metrics for /realtime endpoint (#35500) 3 people 2026-03-04 03:56:32 -08:00
• d6e04f4c43 [Bugfix] Cap FULL decode cudagraph sizes for Mamba/hybrid models (#34094) (#34571) haosdentandzjy0516 2026-03-04 18:56:22 +08:00
• a8f66cbde8 [XPU] bump vllm-xpu-kernels to v0.1.3 (#35984) Kunshang Ji 2026-03-04 18:23:31 +08:00
• 16d2ad1d38 [Hardware] Replace torch.cuda.empty_cache with torch.accelerator.empty_cache (#30681) Kunshang JiandHarry Mellor 2026-03-04 17:49:47 +08:00
• 5dc3538736 [ROCm][Bugfix] Fall back from CK MXFP4 MoE when GEMM dimensions are unsupported (#35893) Chuan (Richard) Li 2026-03-04 00:30:54 -08:00
• 36bf213181 [Bugfix] Add missing dynamic_arg_dims for Qwen3-ASR torch.compile (#35869) Nathan PriceandIsotr0py 2026-03-04 02:29:01 -06:00
• 6f0dd93801 [Core] Remove busy loop from idle buffer readers (#28053) 3 people 2026-03-04 00:44:20 -07:00
• 5d199ac8f2 Support Audio Extraction from MP4 Video for Nemotron Nano VL (#35539) +212 people 2026-03-04 08:20:33 +01:00
• 9e0f44bec4 [cohere][fix][spec-decode]: fix crash when allowed_token_ids is set without penalties (#35654) Komal Kumar Teru 2026-03-04 12:50:15 +05:30
• 097eb544e9 [Bugfix] Improve engine ready timeout error message (#35616) lailoo 2026-03-04 13:54:32 +08:00
• 7cdba98edf [BugFix] Support tool_choice=none in the Anthropic API (#35835) ShiJie ZhongandChauncey 2026-03-04 13:24:46 +08:00
• 3c85cd9d74 [Rocm][CI] Fix ROCm LM Eval Large Models (8 Card) (#35913) Charlie Fu 2026-03-03 22:50:13 -06:00
• edba15045a [Bugfix] Guard mm_token_type_ids kwarg in get_mrope_input_positions (#35711) Andreas Karatzas 2026-03-03 22:12:51 -06:00
• e379396167 [Refactor] Clean up processor kwargs extraction (#35872) Cyrus Leung 2026-03-04 11:53:53 +08:00
• 6e9f21e8a2 [Chore] Remove debug code in model implementation (#35883) Isotr0py 2026-03-04 11:50:58 +08:00
• c1d963403c [model] support FireRedASR2 (#35727) 3 people 2026-03-04 11:41:30 +08:00
• 77e6dcbbfa [PluggableLayer][MM] Add PluggableLayer for RelPosAttention (#33753) Shanshan Shen 2026-03-04 11:41:27 +08:00
• 70c73df69e [Bugfix] Fix EVS implementation for Qwen3 VL (#33607) William Zhang 2026-03-03 18:18:11 -08:00
• 9a9d442464 Enable bnb for multiple indices weight (#35838) xjxandIsotr0py 2026-03-04 09:46:47 +08:00
• f7da9cdffc [ROCm][CI] Support async weight transfer example with platform-aware determinism (#35710) Andreas Karatzas 2026-03-03 19:44:14 -06:00
• f22ff2958c [Bugfix] Fix coord_socket assertion in DPEngineCoreProc for offline DP mode (#35916) Jaewon 2026-03-03 16:10:11 -08:00
• d15c3b90fc [Core] Move save_tensorized_model logic to Worker (#35825) Nick Hill 2026-03-03 15:31:59 -08:00
• 97286a20ed [Model Runner V2] support dp & ep for spec decoding (#35294) zhrrrandGiancarlo Delfin 2026-03-04 07:19:45 +08:00
• 12b38c0f45 [CI/Build] Allow mounting AWS credentials for sccache S3 auth (#35912) Amr Mahdi 2026-03-03 14:30:47 -08:00
• 467886a0c4 [Model Runner V2] Fix inputs_embeds=None bug for MM models (#35917) Woosuk Kwon 2026-03-03 13:47:45 -08:00
• a9b8b13e5c [Bugfix] Fix misnamed parameter in compressed_tensors_moe.py (#35813) bnellnmandRobert Shaw 2026-03-03 16:29:57 -05:00
• e7213003cb [ROCm][CI] Fix TP size issue for test_gpt_oss (#35887) Micah Williamson 2026-03-03 14:57:34 -06:00
• 3a8eef5869 [ROCm][Bugfix]: Disable AITER Triton ROPE by default (#35601) Rohan Potdar 2026-03-03 13:43:56 -06:00
• 97995f6376 [MoE Refactor] Create MK for TRTLLM Kernels (#32564) 3 people 2026-03-03 13:39:50 -05:00
• 881a6b011b [CI] Temporarily Disable Llama4 MoE Refactor Test (#35870) Robert Shaw 2026-03-03 13:36:15 -05:00
• 8e1fd5baf0 [CI] Bump num_speculative_tokens to 3 in nightly DeepSeek tests (#35882) Matthew Bonanni 2026-03-03 12:26:44 -05:00
• ae88468bcc fix: Ensure invalid audio files return 400 error (#34715) JasonCohereandNicolò Lucchesi 2026-03-03 16:47:39 +00:00
• e05cb3b93e TRTLLM gen-full attn Test Coverage (#34986) ojhaanshikaandAnshika Ojha 2026-03-03 08:35:34 -08:00
• 28ef9ba399 [BugFix] Add support for MTP num_speculative_tokens > 1 with sparse MLA (#34552) Lucas WilkinsonandMatthew Bonanni 2026-03-03 10:21:57 -05:00
• fb7fdc49c4 [ROCm] [CI] Add new fusion test cases that are relevant to vLLM IR Ops (#34307) TJianandvllmellm 2026-03-03 22:24:21 +08:00
• ea463978bb [Frontend][1/n] Improve pooling entrypoints | classify. (#35604) wang.yuqiandCyrus Leung 2026-03-03 22:05:36 +08:00
• 440f0e7dc6 [Bugfix] Avoid src/dst as None in irecv/isend_tensor_dict (#35754) Li, Jiang 2026-03-03 21:56:08 +08:00
• fd4a90f337 [CI] And PPL test for Qwen3.5. (#35853) wang.yuqiandgemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> 2026-03-03 21:15:51 +08:00
• ad9d09e2b8 [Perf] [Hybrid] Copy num_accepted_tokens in non-blocking way when not using prefix caching (#35442) Thomas Parnell 2026-03-03 13:15:43 +01:00
• 4beebfd146 [CI/Build][Intel] Add new performance benchmarks for Intel Gaudi 3 (#31025) Szymon ReginisandKunshang Ji 2026-03-03 12:48:24 +01:00
• b8401cde0e add regression test (#35834) hallerite 2026-03-02 23:32:15 -08:00
• 5dfc5abe94 [ROCm] [Release] Change the package from aiter to amd-aiter (#35198) TJian 2026-03-03 15:13:39 +08:00
• 8fa68a8ce4 Fix TYPE_CHECKING stub defaults in envs.py to match actual runtime defaults (#35645) lin-shh 2026-03-03 00:59:43 -05:00
• 35a6f0bfe2 [Misc] Fix typos in comments: explict→explicit, paramaters→parameters (#35648) lin-shh 2026-03-03 00:59:14 -05:00
• 3a6cbf16e2 [MISC] Removed unused function find_all_indices() from tool_parsers/utils.py (#35683) Taneem Ibrahim 2026-03-02 23:58:42 -06:00
• f44d1ddc8c [BugFix] Fix cmake based incremental install (wrong vllm install dir) (#35773) Lucas WilkinsonandWentao Ye 2026-03-03 00:58:16 -05:00
• 48a54c1e0d [CI/Build] Trigger processor tests on registry update (#35824) Cyrus Leung 2026-03-03 13:55:57 +08:00
• 8b9e8b7454 [ROCm][CI] Fix Assertion Logic For test_gpt_oss (#35806) Micah Williamson 2026-03-02 23:08:04 -06:00
• c21d0039ec [Refactor] Fix maxsim cuda platform and add cli to control it (#35427) Wentao YeandCyrus Leung 2026-03-02 23:48:31 -05:00
• 7d8bbe6f42 [CI/Build] Automatically patch video metadata for multimodal processor test (#35822) Isotr0py 2026-03-03 12:27:45 +08:00
• 25e02647c2 [Core] Add optional flags to check for repetitive token patterns in engine output (#35451) aykoppol 2026-03-02 20:23:25 -08:00
• a0a5178ab4 [Model Runner V2] Use ModelState.prepare_attn() for cuda graph capture [5/N] (#35774) Woosuk Kwon 2026-03-02 20:06:27 -08:00
• 8ea8ba275e [V0 deprecation] Remove Swin model (#35821) Isotr0py 2026-03-03 12:03:41 +08:00
• 4f85bae9d6 [Docs][Model Runner V2] Add Design Docs (#35819) Woosuk Kwon 2026-03-02 19:58:14 -08:00
• 0a7165fd71 [ModelRunnerV2] Rename sampler functions and variables for clarity (#35459) Andy Lo 2026-03-03 04:48:56 +01:00
• 6521ccf286 [CI] Temporarily Disable Nightly Failures (#35770) Robert ShawandRobert Shaw 2026-03-02 20:49:13 -05:00
• 8ebd872f50 [Tool Parser] Fix Qwen3Coder streaming parameter loss with speculative decode (#35615) Martin VitandClaude Opus 4.6 2026-03-03 02:40:37 +01:00
• 168ee03e1c [Model Runner V2][Perf] align dummy_run tokens to uniform decode for dp cudagraph (#35376) zhrrr 2026-03-03 09:10:47 +08:00
• 9dd656f0ea [XPU][NIXL] Add GPUDirect RDMA support for XPU (#35270) liuzhenweiandKunshang Ji 2026-03-03 08:42:49 +08:00
• c8b678e53e [Model] Add support for nvidia/llama-nemotron-rerank-vl-1b-v2 (#35735) Jakub Zakrzewski 2026-03-03 01:32:14 +01:00
• 18c29c746b [ROCm][CI] Fix backslash-continuation in pytest marker re-quoting and treat exit code 5 as success (#35798) Andreas Karatzas 2026-03-02 18:07:51 -06:00
• 96fc09503a [All Reduce] Change default backend of Flashinfer All Reduce to trtllm (#35793) Hanjie Qiu 2026-03-02 18:57:38 -05:00
• 1b82b433fc [Bugfix] Fix MM processor test for Qwen3.5 (#35797) Roger Wang 2026-03-02 15:05:08 -08:00
• 9319044ee9 [MoE][Perf] Wrap DSV3 QKVAProj GEMM in custom op for torch.compile (#35751) Robert ShawandRobert Shaw 2026-03-02 18:03:49 -05:00
• c42dc402c1 clean unused cudagraph_batch_sizes (#35552) Boyuan Feng 2026-03-02 14:00:16 -08:00
• fa6a6be519 [Bugfix] Fix missing sequence_lengths in qwen3_omni_moe_thinker (#35741) Ye (Charlotte) Qi 2026-03-02 13:11:56 -08:00
• cad21918e3 [BUG] Fix rlhf_async example (#35788) Aaron Hao 2026-03-02 12:36:40 -08:00
• 53700bf49b [ci] Add Ray compatibility check informational CI job (#34672) Jeffrey Wang 2026-03-02 12:06:16 -08:00
• a13d8c03c9 [KVConnector] Auto-downgrade to PIECEWISE cudagraph mode for layerwise async ops (#31057) Yashwant Bezawada 2026-03-02 14:04:47 -06:00
• 9433acb8df [Spec Decode] Add hidden states extraction system (#33736) Fynn Schmitt-Ulms 2026-03-02 14:29:09 -05:00
• d1a6e96d9e [torch.compile] Improve cold and warm start compile tests (#35709) Richard Zou 2026-03-02 14:27:06 -05:00
• 2a9e3347e9 [BugFix][Model]Fix the garbled code in Ernie4.5-VL caused by fast_moe_cold_start (#35587) CSWYF3634076 2026-03-03 02:56:33 +08:00
• cc0d565f40 [CI/Build] Enable Qwen3.5 tests on CI (#35763) Isotr0py 2026-03-03 01:43:53 +08:00
• 358e4d5ba7 [CI][HPU] Pin vllm commit compatible with vllm-gaudi - HPU tests (#35307) Patryk Wolsza 2026-03-02 18:02:26 +01:00
• 792a74b973 [Doc] Improve UX of --enable-log-requests (#35723) Cyrus Leung 2026-03-03 00:24:09 +08:00
• 4034c3d32e [Core] Move test utility to test file (#35672) Turner Jabbour 2026-03-02 08:56:03 -07:00
• 7560d674c9 [CI] Fix mypy for vllm/device allocator (#35518) Martin HickeyandHarry Mellor 2026-03-02 15:53:18 +00:00
• d9c7730877 [Performance] Extract kv update ops from MLA attention backends (#34627) 3 people 2026-03-02 16:43:19 +01:00
• ada4f4fadd [Fix Bug]num_active_loras always equals to zero (#34119) Runkai TaoandJee Jee Li 2026-03-02 10:17:46 -05:00
• 7e9149d9a9 [Docs] Add breadcrumbs for better UX (#35749) Harry Mellor 2026-03-02 14:31:54 +00:00
• 87c98b0236 [MyPy][BugFix] Check profiler is assigned before calling start() on it (#35505) Martin HickeyandHarry Mellor 2026-03-02 13:23:42 +00:00
• de7dd634b9 Fix unresolved-import errors when using Astral's ty by removing src.root (#35681) Tyler Michael Smith 2026-03-02 05:26:47 -05:00
• 9a87b0578f [Feat] Supports Anthropic Messages count_tokens API (#35588) Chauncey 2026-03-02 17:48:54 +08:00
• 510bc9e1df [Misc] Cleanup useless current_platform import (#35715) wangxiyuan 2026-03-02 17:36:54 +08:00
• cbd361fd46 [CPU][Distributed] Fix Enable _CPUSHMDistributed only when TP/PP ranks share the same SHM group name (#34169) Charles Ashby 2026-03-02 04:34:35 -05:00
• c212202d93 [Misc] Bound NIXL upper bound version (#35495) Nicolò Lucchesi 2026-03-02 09:57:07 +01:00
• ec27b36b4b [CI] Defining extended V1 e2e + engine tests (#35580) Andreas Karatzas 2026-03-02 02:10:54 -06:00
• 3fd1d4ec2c [Rocm][CI] Fix LM Eval Large Models (H100) test group (#34750) Charlie Fu 2026-03-02 01:43:38 -06:00
• cb21972a97 [Kernel] Integrate SM100 MXFP8 blockscaled grouped MM and quant kernels (#34448) EdalatiAliandMichael Goin 2026-03-02 02:31:19 -05:00
• c34963f138 [ROCm][CI] Disable skinny GEMMs in language model standard tests to fix non-determinism (#35152) Andreas Karatzas 2026-03-02 01:04:18 -06:00
• f26650d649 [ROCm] add amd-quark package in requirements for rocm to use quantized models (#35658) Hongxia YangandHongxia Yang 2026-03-02 01:02:43 -05:00
• 92f5d0f070 [XPU] fix mxfp4 activation type (#35691) Kunshang Ji 2026-03-02 11:48:39 +08:00
• a60985b07e Fix deprecated v1 config tests (#35327) Jesse Cai 2026-03-01 17:32:03 -08:00
• 8b5014d3dd [Attention] FA4 integration (#32974) 3 people 2026-03-01 18:44:57 -05:00
• 57a96e26c9 Revert "[Bugfix] Disable TRTLLM attention with KV transfer enabled (#33192)" (#34832) zhanqiuhu 2026-03-01 17:32:37 -05:00
• e82fbeec7b [torch.compile] Undo the fast_moe_cold_start hack in torch>=2.11 (#35475) Richard Zou 2026-03-01 16:44:22 -05:00