Commit Graph

Select branches

Hide Pull Requests

main

patches

sm86

Mono Color

• 1cf6555214 [Bugfix] Pin EPLB and MLA host-to-device transfer buffers (#56138) Kevin H. LuuandOpenAI Codex 2026-09-11 01:57:47 -07:00
• d0dfe587d5 [Bugfix] Fall back to full decode graphs for noncompiled models (#55095) 3 people 2026-09-11 03:56:05 -05:00
• 9dcab80215 [Bugfix][Bench] Fix bench mm-processor crash in shared request sampling (#56300) Nicolò Lucchesi 2026-09-11 10:53:04 +02:00
• 69db1c26b4 [CI/Build][Rust Frontend] Publish vllm-proto on crates.io (#56365) 4 people 2026-09-11 01:28:52 -07:00
• e3f755b732 [CPU] Speedup LM Head on Arm CPUs (#55352) Fadi ArafehandLi, Jiang <jiang1.li@intel.com> 2026-09-11 09:21:14 +01:00
• 6376c601ed [XPU][Tests] Enable test_per_token_group_quant_int8 on XPU (#55681) pmanczak 2026-09-11 09:59:13 +02:00
• fadfe1c7d4 [Bugfix][Core] Retire Mamba states across null gaps (#55450) 4 people 2026-09-11 09:50:47 +02:00
• a9271c750f [ROCm][CI] Accept base-v2-preview images during content hash lookup for ROCm base images (#56356) Micah Williamson 2026-09-11 02:30:39 -05:00
• 5c642796e8 [Docs] Add Nebius Serverless AI deployment guide (#56414) Alexander 2026-09-11 07:58:49 +01:00
• 4b839c3378 [Attention] Extend XQA decode support on SM90 (#50439) 3 people 2026-09-11 08:19:55 +02:00
• fbf51c7026 [Bugfix] Fix FlashInfer KV sharing with omitted K/V (#55864) 3 people 2026-09-11 14:07:52 +08:00
• fe4d4109c2 [Frontend][last/N] Move all non-OpenAI content out of the OpenAI folder. (#56369) wang.yuqi 2026-09-11 13:33:26 +08:00
• 5ff50f3996 [Bugfix][MM] Fix swapped H/W in dummy video profiling inputs (#56385) Hung Hoang (黄兴）andCursor Agent 2026-09-11 12:25:51 +07:00
• ae48466cf3 [Perf][ROCm] Add AITER custom AG/RS (DP only) (#48247) Simon Danielsson 2026-09-11 07:19:57 +02:00
• 980c16c8e4 [PCP][Spec Decode] Adds PCP support for single-module MTP and replicated DSpark. (#56107) 4 people 2026-09-11 01:07:53 -04:00
• e7edf17cea [EC] Automatically enable embedding inputs on EC/KV consumers (#56379) Tianyu Guo 2026-09-11 12:53:37 +08:00
• b5d4186300 [Bugfix][LoRA] Use stored rsLoRA scaling factor in MoE expert packing (#55548) Kushal 2026-09-11 12:52:33 +09:00
• ce08bb5b34 [Bugfix][Mooncake] Fix heterogeneous PP transfer completion (#56033) 3 people 2026-09-11 10:57:19 +08:00
• 07b7553465 [Feature][SimpleCPU] Load fine-grained hybrid prefix hits (#54736) YukioZzz 2026-09-11 09:36:37 +08:00
• 828f4f19b4 [ROCm][Bugfix] Route GLM-5.3-Flash MTP through ragged sparse MLA (#55239) 3 people 2026-09-10 21:13:14 -04:00
• 84030bbe3d [Bugfix][Core] Stop zero-progress preemption cascades for deferred KV frees (#49675) 焦恺文 2026-09-11 08:40:10 +08:00
• 5b6cf93e8e [XPU] Add forward_xpu to Mixer2RMSNormGated and FusedRMSNormGated (#54968) Marceli FylcekandClaude Opus 5 2026-09-11 01:58:28 +02:00
• a89de950a4 [CI/Build] Pin HyperCLOVAX V2 test model revision (#56335) Tyler Michael SmithandOpenAI Codex 2026-09-10 19:26:11 -04:00
• 360f33f61c [Bugfix] Set stop_sequence explicitly in streaming message_delta event (#55325) Colin McNamara 2026-09-10 17:21:21 -05:00
• 9163190dda [ROCm][Bugfix][Perf] Tune multi-stream shared experts use; wvSplitKrc fixes (#56098) MattandClaude Opus 5 2026-09-10 15:14:22 -07:00
• ae71862c51 [Bugfix][Frontend] Check EC requirements for each metadata item (#56070) shaohuaxiandCodex 2026-09-10 12:23:34 -07:00
• a36dfc93cb [Bugfix][Multimodal] Restore cached audio inputs with UUIDs (#56310) Jiatai WangandOpenAI Codex 2026-09-10 12:20:03 -07:00
• 7de70fa7ae [Bugfix][ROCm] Create linear layer biases with requires_grad=False (#56161) Micah Williamson 2026-09-10 13:54:52 -05:00
• 48cb12c184 [ROCm][Bugfix] Fix profiler in TheRock image (#56190) Matt 2026-09-10 11:39:33 -07:00
• 7cdd9304ae [KV Connector] Support symmetric DCP disagg for hybrid mamba models (#55531) 4 people 2026-09-10 14:12:15 -04:00
• 983b7e28c9 [Bugfix] Avoid MistralCommonBackend for HF tokenizers (#54192) Julien Denize 2026-09-10 19:57:44 +02:00
• 8359e15aae [ROCm][Feature] Support KV connectors with ROCM_AITER_UNIFIED_ATTN (#53695) 3 people 2026-09-10 19:48:42 +02:00
• e6cb56337b [Core] MRV2 support for fast-prefill (#56145) Nick Hill 2026-09-10 10:36:04 -07:00
• 7470082f57 [ROCm][Perf] Add bpreshuffled blockscaled fp8 GEMM (#51692) Simon Danielsson 2026-09-10 18:18:37 +02:00
• 3078e7cbe1 [Docs]: quote variable-bearing wheel URLs (#56286) Josiah Davis 2026-09-10 11:13:09 -05:00
• 6ee5bb0a0b [DCP][Kernel][Perf] Fuse the empty-shard LSE mask into the A2A pack kernel (#54889) Rita Brugarolasandmergify[bot] <37929162+mergify[bot]@users.noreply.github.com> 2026-09-10 08:43:00 -07:00
• bdea2777ea [Bugfix][MoE] Fix batched CUTLASS workspace overallocation (#55579) Itay AlroyandTyler Michael Smith 2026-09-10 18:33:56 +03:00
• db723d245a [Bugfix] Honor explicit empty and zero CLI arguments (#55710) Cenab Batu BoraandCodex 2026-09-10 09:07:42 -06:00
• 46bae7e98c [CI/Build][CPU] Fix flaky rust downloads, broken prune flag, and triton-cpu cache coupling (#56247) Li, Jiang 2026-09-10 22:27:26 +08:00
• 2e0ee66cab [Perf] Use UVA-backed contents for MRV2 apply_write (#55819) bjf-frzandNick Hill 2026-09-10 22:27:16 +08:00
• 7d8d71e989 [Bugfix] Qwen3-VL(-MoE): pass architectures to with_hf_config for pipeline parallelism (#43272) Alexander CrettenandandClaude 2026-09-10 08:49:35 -05:00
• 364679feab [CI] Enable ruff INP to require __init__.py under vllm/ (#56264) Harry MellorandClaude Opus 5 2026-09-10 14:49:17 +01:00
• f9083cb83a [Mypy] Fix typing for R/S models (#54157) Taneem Ibrahim 2026-09-10 08:43:08 -05:00
• 73fb19151f [Fast Start] Support fp4 (#55465) siyuandIsotr0py 2026-09-10 21:22:17 +08:00
• 40e6042ec8 [Feature][Spec Decode] MTP with separate (possibly quantized) lm head for nemotron (#54574) YoavMironandBenjamin Chislett 2026-09-10 15:26:31 +03:00
• 9b959b8657 [Model] DeepSeek-V4.1-Flash Model Definitions (#56228) 4 people 2026-09-10 05:16:26 -07:00
• e83d57d8ae [Doc] Sync KV event medium terminology after #48123 (#51646) 3 people 2026-09-10 19:42:31 +08:00
• 93911fcf68 [NIXL][PCP] Report replicated-PCP ranks > 0 as done sending instead of hiding them (#53903) Lucas WilkinsonandNicolò Lucchesi 2026-09-10 06:58:05 -04:00
• 08426d51ef [Docs][Security] Clarify reporter credit and CVE publication timing (#55476) Juan Pérez de AlgabaandCursor 2026-09-10 12:19:39 +02:00
• 86aca66191 [Kimi K3 Perf] Avoid KDA mixed-batch gather/scatter, 5.2%~7.7% E2E Throughput Improvement (#56159) Wentao Ye 2026-09-10 05:57:01 -04:00
• c9355e25e8 [XPU][Bugfix] Add forward_xpu to Ernie4_5_VLRotaryEmbedding (#55942) Jakub ByczkowskiandCopilot 2026-09-10 11:48:00 +02:00
• 6fd08c45e6 [XPU][Bugfix] Fix FalconH1 pipeline-parallel execution (#55913) 3 people 2026-09-10 11:47:29 +02:00
• fe26c705f7 [Model] Add Bailing V3 VL support (#55921) 3 people 2026-09-10 17:46:56 +08:00
• 6ff479e1f7 [ROCm] Add better kv dtype error discoverability (#55236) Giuseppe Grossi 2026-09-10 05:17:41 -04:00
• b28c3e1568 [BugFix] Retain both replay boundaries so an EAGLE resend of a block-aligned prompt still hits (#54713) Toby Mao 2026-09-10 01:39:24 -07:00
• 1768273c13 [Perf][GLM-5.3-Flash] Decode hot-path cleanups: strided KDA recurrent inputs, NoPE MQA query without concat, no duplicate router GEMM (#55736) Jared WenandClaude Fable 5.1 2026-09-10 16:28:33 +08:00
• b47b01cf33 [Model][Frontend] Support DeepSeek-V4.1-Flash in Rust and Python frontends (#56208) Bugen ZhaoandIsotr0py 2026-09-10 18:15:06 +10:00
• 9521c60bdc [ROCm][Perf] Kimi-K3 Fused kernels for KDA prefill reland (#54038) 3 people 2026-09-10 15:37:54 +08:00
• be1cb9834b [Kernel] Optional Q-norm in fused DSv4 MLA epilogue; group_size=32 for packed FP8 quant (#56215) Yongye ZhuandClaude Opus 5 2026-09-10 00:35:59 -07:00
• 9e25706560 [Bugfix] Tolerate misspelled DSML tool_calls wrapper (#56141) Flora Feng 2026-09-10 03:04:29 -04:00
• 83990f5bcc [Bugfix][MRV2] Initialize DCP metadata after batch partitioning (#55212) 5 people 2026-09-09 23:50:09 -07:00
• 912d2e79fe [multimodal][feat] add torchaudio backend to AudioResampler (#52598) 3 people 2026-09-10 14:42:38 +08:00
• 7e91760650 [ROCm][Perf] Route large DSV4 sparse prefill to AITER OPUS (#54855) 4 people 2026-09-09 23:39:12 -07:00
• 442d36031c [MM][CG] Enable encoder CUDA Graph for MiniCPM-V (#42785) Yunzhu LuandIsotr0py 2026-09-10 14:28:48 +08:00
• c3ccc0e957 [Model] Add DeepSeek-V4 CPU backend (#55355) Li, Jiang 2026-09-10 14:14:48 +08:00
• 49632fb3ac [CI][XPU] Reject CUDA-IPC weight cache on non-CUDA/ROCm platforms (#56010) Qiming Zhang 2026-09-09 23:05:07 -07:00
• 588a813a60 [ROCm] [BugFix] Fix AITER MXFP4 ASM-GEMM crash on unfused shared experts (#55213) Colin Z 2026-09-09 22:23:36 -07:00
• 479611c0c6 [Frontend][EPD] Use JSON arrays for multimodal metadata (#56090) Tianyu Guo 2026-09-10 13:04:14 +08:00
• 2a02f6efe3 [CI][ROCm][Disagg] Add GLM-5.2-FP8 to MoRIIO model catalog (#53885) 3 people 2026-09-10 09:42:54 +05:30
• ea40bb9e90 [feature] Watermarked generation and detection (Gumbel-max algorithm) (#54053) 3 people 2026-09-10 06:00:15 +02:00
• 8c34a37232 [Refactor] Extract maybe_run_omni() and defer CLI imports past omni early-return (#56103) zvier 2026-09-10 11:21:58 +08:00
• 6e1d051fc9 [CI][XPU] skip test_models_text on XPU (#56038) Qiming Zhang 2026-09-09 20:03:09 -07:00
• 5c0aeb5954 [CI][XPU] Disable model runner V2 for XPU quantization test for some partially pre-quantized models (#56179) Chaojun Zhang 2026-09-10 11:02:57 +08:00
• 3735c2d5f5 [Security][Rust Frontend] Normalize HTTP method labels in metrics (#56058) 3 people 2026-09-10 04:54:40 +02:00
• 41cffa9af0 [Rust Frontend] Strongly type wire dtypes and multimodal modalities (#56174) 3 people 2026-09-10 12:11:22 +10:00
• 285cbce6bd [Rust Frontend] Resolve unified and split parser selections consistently (#56018) Bugen Zhao 2026-09-10 11:19:30 +10:00
• 978c22f54b [XPU] update triton-xpu 3.8.0 shim layer (#56014) Yan Ma 2026-09-10 08:44:11 +08:00
• 65f3fca568 [Bugfix][Parser] Seed-OSS turn-boundary tokens + boundary-fallback tests (#54264) Mikhail PodvitskiiandFlora Feng 2026-09-10 01:29:30 +02:00
• 4ebf61ebda [Cohere] Bound remaining request priorities to the MessagePack int64 range (#56146) Taneem Ibrahim 2026-09-09 17:57:01 -05:00
• 83252ea899 [Performance][ROCm] Integrate aiter indexer scoring and top-k kernels into MiniMax-M3 sparse attention path (#52664) ykamiset 2026-09-09 17:40:33 -05:00
• 6983a0883d [ROCm][CI] Split MI300 Distributed Compile by graph partition mode (#53602) Aarushi Jain 2026-09-09 17:07:51 -05:00
• dcd544486b [ROCm][Bugfix] Support shared KV prefill in AITER attention (#55887) Andreas Karatzas 2026-09-09 16:51:34 -05:00
• 26fec6d183 [Bugfix] Fix Step-3.5 reasoning parser for structured outputs (#53174) yzong-rh 2026-09-09 17:10:23 -04:00
• 8c87c333b8 [Perf] Fix TRTLLM ragged prefill perf regression (#55499) Wei ZhaoandOpenAI Codex 2026-09-09 16:38:20 -04:00
• 56d001faf0 [Bugfix] Detect OpenAI content format when message.content is passed through macro parameters (#53824) 4 people 2026-09-10 04:10:28 +08:00
• cce50657b7 [ROCm][CI] Use a platform-independent GEMM in the merged-column fuser test (#56130) djramic 2026-09-09 22:08:39 +02:00
• e8064a96d0 [Bugfix][V2] record_stream idx_mapping in the PP draft broadcast (#55745) Chris EastwoodandNick Hill 2026-09-09 15:55:56 -04:00
• c69d5d72a6 [CI][ROCm] Increase timeout for AMD MI355 Language Models (Standard) (#56114) Aarushi Jain 2026-09-09 13:10:26 -05:00
• 3fb676bfad [ROCm][CI] Fix moe layer tests for fp8 dtype compatibility (#56106) Divakar VermaandAndreas Karatzas 2026-09-09 13:48:14 -04:00
• a207d7ce1d [Bugfix][LoRA] Log when an adapter applies no weights (#55310) Guancheng Fu 2026-09-10 01:44:46 +08:00
• 22d95d1adc [Bugfix] Make mm_device_do_normalize encoder-cudagraph safe (#55370) 3 people 2026-09-10 02:43:22 +09:00
• 83fe99399e [Spec Decode] Add NVFP4 DSpark gathered top-k projection (#55713) 3 people 2026-09-09 19:33:46 +02:00
• af51d0ef03 [Pooling] Report LoRA adapter names in responses (#56004) Taneem Ibrahim 2026-09-09 12:31:13 -05:00
• 1454b71727 Fix ROCm AITER FP8 KV test tolerances. (#53590) Aarushi JainandAarushi Jain 2026-09-09 12:22:25 -05:00
• 42d76ee35a [Docs] Fix griffe docstring indentation warning in SupportsMRoPE (#56112) Harry MellorandClaude Opus 5 2026-09-09 18:11:40 +01:00
• 719284fe15 [Core][Model] Unify XD-RoPE into M-RoPE and derive the channel count (#56078) Harry MellorandClaude Opus 5 2026-09-09 17:23:30 +01:00
• c55e15a44e [Structured Output] Keep invalid structured-output requests from stopping the engine (#51450) 3 people 2026-09-09 11:32:01 -04:00
• 7e95735eb9 [Bugfix] Avoid FlexAttention recompiles when request counts change (#55888) Andreas Karatzas 2026-09-09 10:10:25 -05:00
• d8d53f17c2 [Rocm][Kimi-k3] Add pipeline_parallel support for the kimik3 model (#53664) 3 people 2026-09-09 22:50:24 +08:00
• 4e990dfcec [ROCm] Bump AITER to v0.1.21.post2 (#55968) Micah WilliamsonandFangzhou Ai 2026-09-09 09:49:59 -05:00