Commit Graph

Select branches

Hide Pull Requests

main

patches

sm86

Mono Color

• a33b3bac5d [Bugfix][Qwen2.5-VL] Honor video fps for temporal M-RoPE (#47736) Ting SUNandmergify[bot] <37929162+mergify[bot]@users.noreply.github.com> 2026-09-22 13:56:12 +08:00
• 5abf85d27a [Bugfix][Frontend] Validate mixed prompt embedding mask lengths (#57006) shaohuaxiandCodex 2026-09-21 22:50:31 -07:00
• 4800024b98 [Security] Reject min_tokens that exceeds the filled max_tokens default (#57731) Juan Pérez de Algaba 2026-09-22 07:50:23 +02:00
• d50723df04 [Rust Frontend] Introduce parser-owned output grammar interfaces (#55269) Bugen ZhaoandClaude Fable 5.1 2026-09-22 13:35:07 +08:00
• f84325c48c [Frontend][Rust] Add mm-processor benchmark for Rust frontend (#51922) 4 people 2026-09-22 13:08:35 +08:00
• 0961bbae28 [Spec Decode] Enable async scheduling for DFlash (#58065) Jeff (Junze) Ma 2026-09-22 01:05:44 -04:00
• 42a85a4976 [ROCm][Bugfix] Explicitly reject FSE=1 with DPA+ETP deployment for DeepSeek-V4 (#57919) Shanshan Shen 2026-09-22 12:45:53 +08:00
• 639461eda4 [RL][Sleep] Retain frozen weights across level-2 sleep (#57891) 3 people 2026-09-22 12:36:55 +08:00
• c4d424c2e2 [Bugfix][Spec Decode] Cap DFlash/DSpark profiling query batch (#56448) Yicong Wang 2026-09-22 12:36:32 +08:00
• ff3c9cba3f [Rust Frontend] Add MiMo V2.5 parser support (#57933) Bugen ZhaoandCodex 2026-09-22 12:27:09 +08:00
• c723a831a8 [Fast Start] Cache the MTP draft model in a separate daemon group (#57312) siyuandIsotr0py 2026-09-22 11:13:38 +08:00
• f07e227ee9 [Docker] Expose bundled vllm-rs on PATH (#57606) 3 people 2026-09-21 20:03:03 -07:00
• 1fee69838e [XPU] Fix Nemotron FP8 LM-eval config: drop CUDA-only moe_backend and wire to new Buildkite job (#49685) Chaojun ZhangandCopilot 2026-09-22 10:05:26 +08:00
• e9f169d16b [CI] Add pre-commit check that new tests are tethered to Buildkite jobs (#54867) 4 people 2026-09-21 19:36:45 -06:00
• 79468c20ef [Bugfix][GDN] Fix stateless first-chunk classification (#51565) cherry77-cloudandzjy0516 2026-09-22 09:11:49 +08:00
• 0bce411a07 [Bugfix] Annotate MTP draft KV cache groups positionally on the hybrid grouping path (#55390) Navjot SinghandCodex 2026-09-21 20:31:26 -04:00
• a369a7becc [Perf][Attention] Reduce GLM sparse MLA preparation overhead (#57458) 4 people 2026-09-21 17:24:25 -07:00
• c64b15cde5 [Scheduler] Soften Long Prefill Tokens Threshhold (#57951) Robert ShawandRobert Shaw 2026-09-21 20:19:42 -04:00
• 78d7836525 [CI] Split (H200) LM Eval Large Models into per-model jobs (#57965) 3 people 2026-09-21 20:19:17 -04:00
• 7d06dd1ffa [Pooling] MRV2 pooling shutdown model ref (#57737) Taneem Ibrahim 2026-09-21 18:54:57 -05:00
• 13058faa49 [Frontend] Add reusable TP1 initialized-engine snapshots (#51360) 5 people 2026-09-21 16:18:49 -07:00
• 8bf051cd51 [Bugfix][Kernel] Skip the fused silu-mul block-quant fast path when a swiglu clamp is set (#57984) Garrett GoonandClaude Opus 5 2026-09-21 18:38:38 -04:00
• 8644d2af2f [CI][Build] Harden triton-cpu sleef submodule fetch in CPU image build (#57871) vllm-agentandkimi-no-na-wa 2026-09-21 15:30:14 -07:00
• 382970ee6c [Kimi-K3][AMD] Return KDA and MLA projection outputs directly (#50592) 3 people 2026-09-22 04:44:41 +08:00
• 9fe1cbd463 [Docker] Use zstd for CI images and offer a Docker Hub variant (#55608) Nils Mattesonandmergify[bot] <37929162+mergify[bot]@users.noreply.github.com> 2026-09-21 13:40:15 -07:00
• 12972ae40b [Bugfix][ROCm] Reject unsupported EP for monolithic AITER MXFP4 MoE (#57866) Andreas KaratzasandOpenAI Codex 2026-09-21 15:30:21 -05:00
• 9179bcd416 [Bugfix][MoE] Reject hash routing for unsupported monolithic backends (#57867) Andreas KaratzasandOpenAI Codex 2026-09-21 15:28:35 -05:00
• 54020c3c3e [Engram] Drop redundant VLLM_PLE_CPU_OFFLOAD env var (#57937) Nicolò Lucchesi 2026-09-21 20:16:02 +02:00
• 17322375a2 [ROCm][Perf] Route the fused shared-expert gate GEMM through the platform dispatcher (#54185) Mikko Tukiainenandmergify[bot] <37929162+mergify[bot]@users.noreply.github.com> 2026-09-21 20:17:55 +03:00
• fba33e39cb [Docs] Add an Engram feature page explaining Engram usage in vLLM (#57910) Nicolò Lucchesi 2026-09-21 19:15:28 +02:00
• 986e217158 [Feature] Add first-class KV hints request envelope for programmatic KV management (#53423) Karen Chung 2026-09-21 09:56:11 -07:00
• 4f14516790 [ROCm][CI] Use ROCm backend for DeepSeek V4.1 ViT test (#57931) djramicandAndreas Karatzas 2026-09-21 18:39:31 +02:00
• 7853700d4c [MM] Move get_dummy_processor_inputs into MM processor (#57967) Cyrus Leung 2026-09-22 00:39:20 +08:00
• 4e61ad8ee1 [Bugfix] unskip InternViT test for transformers v5 compatibility (#55767) Sahil PatelandCyrus Leung 2026-09-21 17:38:25 +01:00
• 874a6b5a06 [ROCm] Refactor tuned gemms (#55001) 4 people 2026-09-21 11:30:16 -05:00
• 8c0825090d [Bugfix][NIXL] Fix DCP pulls across MLA cache regions (#57389) Lucas WilkinsonandOpenAI Codex 2026-09-21 11:57:49 -04:00
• 6b858751f6 [docs] Fix legacy hf CLI references (vllm) (#57958) LucainandClaude Opus 5 2026-09-21 17:32:27 +02:00
• dc479a2d83 [Bugfix] Fix external LB DP rank handling when replicas share nodes (#53743) Tony Lin 2026-09-21 23:19:14 +08:00
• e34685dfc0 [ROCm][DSv4] Enable DSpark adaptive verification (#52362) Tuukka Sarviandlarryli2-amd 2026-09-21 18:03:41 +03:00
• 21aa17c128 [Bugfix][Frontend] Accept diarized transcription responses in run-batch (#57948) sfigs 2026-09-21 11:55:38 -03:00
• 887cf91e30 [Pooling] Fix normalization of chunked long-text embeddings (#57498) Taneem Ibrahim 2026-09-21 09:46:39 -05:00
• c8824a1e42 [MM] Move supports_multimodal_inputs and cache out of registry (#57913) Cyrus Leung 2026-09-21 22:43:07 +08:00
• 7488b5780f [ROCm][Tests] Reduce host memory when downcasting FP32 HF references (#57869) Andreas KaratzasandOpenAI Codex 2026-09-21 09:42:47 -05:00
• 8b98b7d0b4 [Frontend] Add streaming parity tests and docs for derender (#57922) Martin Hickey 2026-09-21 15:13:52 +01:00
• 606d124b37 [Bugfix][Multimodal] Tolerate malformed EXIF in Molmo 2 image preprocessing (#57234) Hotragn PettuganiandClaude Opus 5 2026-09-21 09:27:40 -04:00
• db7f1f6714 [CI][XPU] Deselect Ray UT in XPU V1 test Job (#57889) Chaojun ZhangandKunshang Ji 2026-09-21 20:54:58 +08:00
• 7268f6e38e [Security] Prefer fresh multimodal payloads over a stale receiver cache (#57833) Juan Pérez de AlgabaandCursor 2026-09-21 14:50:48 +02:00
• 82daf9f575 [CI][ROCm] Add ten AMD parity groups (#57876) Andreas KaratzasandOpenAI Codex 2026-09-21 06:08:06 -05:00
• 15859bb3a1 [Test] Organize structured output utility tests (#57545) CaiJohnandjohn 2026-09-21 18:56:28 +08:00
• 04c1f4a407 [Bugfix][ROCm][DSv4.1] Disable SWA bounded replay on ROCm (#57906) Fangzhou AiandCursor 2026-09-21 06:29:14 -04:00
• 0b7f11a1ee [AuxOutput] Add block-keyed storage for routed-expert outputs (#45635) 6 people 2026-09-21 18:25:05 +08:00
• 0aee727ff6 [CI][ROCm] Increase timeout for MI300 Distributed DP Extended (#57903) djramic 2026-09-21 11:18:59 +02:00
• 8902dbb177 [CI][ROCm] Mirror remaining portable suites without duplicate coverage (#57877) Andreas KaratzasandOpenAI Codex 2026-09-21 04:16:26 -05:00
• b8cf275382 [AMD][Minimax-M3][perf] Enable packed LBHNC AITER QK-norm fusion for MiniMax-M3 on ROCm (#54535) 3 people 2026-09-21 01:38:30 -07:00
• f05b88751e [CI][ROCm] Make Python-only installation failures blocking (#57870) Andreas KaratzasandOpenAI Codex 2026-09-21 03:29:02 -05:00
• 76ffb0d374 [Model][Engram] Share host tables across co-located DP replicas by default (#57651) 3 people 2026-09-21 01:26:56 -07:00
• 0ff0477ff9 [Perf][ROCm] Add a ROCm path for Hy4 and compile the backbone (#57526) akii96 2026-09-21 11:20:43 +03:00
• 67513c8b67 [Bugfix][DSV4.1] Restrict mHC overlap to full CUDA graphs (#57874) Woosuk KwonandOpenAI Codex 2026-09-21 01:16:59 -07:00
• 97dc6b19d2 [Perf][Attention] Avoid redundant sparse attention metadata operations (#57885) Woosuk KwonandOpenAI Codex 2026-09-21 00:08:43 -07:00
• 3c84ad13ed [Bugfix] Fix xgrammar feature gate bypass when JSON Schema type is a list (#48416) 3 people 2026-09-21 14:59:20 +08:00
• 9598487bc1 [Bugfix][Model] Mistral3: align placeholder grid with public processor (#53758) Oliver Holworthyandmergify[bot] <37929162+mergify[bot]@users.noreply.github.com> 2026-09-21 07:36:59 +01:00
• 59f1527280 [Bugfix] Fix Whisper engine crash on audio clips longer than 30s (#57769) Tianyao Wu 2026-09-21 13:44:40 +08:00
• 8ec00ec2bc [Tests] Release HF embedding weights after prompt extraction (#57868) Andreas KaratzasandOpenAI Codex 2026-09-21 00:27:29 -05:00
• 6654bcbfca [MRV2][XPU] use xpu sample kernel in mrv2 sampler (#57277) liuzhenweiandKunshang Ji 2026-09-21 13:24:41 +08:00
• 76ee20c1ec [ROCm][Tests] Avoid pooling cleanup waits against live shared engines (#57864) Andreas KaratzasandOpenAI Codex 2026-09-21 00:21:35 -05:00
• a15dbbd63d [XPU][CI] fallback to 7-args of moe_align_block_size (#57855) liuzhenweiandKunshang Ji 2026-09-21 13:15:29 +08:00
• 3df4ae153e [Core][KDA] Generalize Mamba prefill checkpoint builder and exporter (#57783) Chaunceyandzjy0516 2026-09-21 12:46:11 +08:00
• ff0c885494 [Bugfix] DiffusionGemma: fix multimodal support (#57589) 4 people 2026-09-20 22:05:24 -06:00
• 9a70c233cd [Perf][DSV4.1] Fuse TP all-reduce with mHC input preparation (#57643) Woosuk KwonandCodex 2026-09-20 20:28:06 -07:00
• eb87980585 [Build] Fix DeepGEMM CUDA 12.9 release builds (#57554) 3 people 2026-09-20 20:14:00 -07:00
• 86ce4d10e2 [BUGFIX][HY4] Record indexer completion event for full CUDA graph capture (#57811) Jee Jee Li 2026-09-21 10:59:11 +08:00
• 3918f3c5a3 [Profiler] Unify platform-aware torch profiling (#57460) czhu-cohere 2026-09-20 19:17:23 -07:00
• 6dfd87f596 [ROCm][DSv4.1] Keep the Engram tables in host memory on ROCm (#57491) JohnQinAMD 2026-09-20 19:17:11 -07:00
• 3e67044972 [Doc] Refresh Kthena integration guide (#57314) Zhou ZihangandJee Jee Li 2026-09-21 09:42:55 +08:00
• 1185254851 [Bugfix][KV Connector] Propagate cache reset failure during sleep (#54581) 3 people 2026-09-21 09:25:02 +08:00
• 7ddd461a28 [Docs] Add ECMooncakeConnector usage example for EPD (#57742) jiangkuaixue123andCodex 2026-09-21 09:15:02 +08:00
• d2983f2f16 [Bugfix][Spec Decode] Stop dummy draft decode steps from writing KV through stale block-table rows (#56734) Yifan QiaoandClaude Fable 5.1 2026-09-20 18:06:25 -07:00
• 9b49f92344 [Perf][DSV4.1] Overlap mHC coefficients for small TP batches (#57603) Woosuk KwonandCodex 2026-09-20 17:33:17 -07:00
• 17e50b9b76 [XPU][UT]Bugfix when the process can't see all the world_size meet accuracy issue. (#57779) YiSheng5andKunshang Ji 2026-09-21 06:50:25 +08:00
• 9679173788 [KVConnector][MoRIIO] Transfer hybrid mamba/KDA recurrent state in READ mode (#51052) YukioZzz 2026-09-21 05:30:34 +08:00
• 01f1f58f10 [ROCm][CI] Query HIP device memory for test GPU teardown waits. (#57450) 4 people 2026-09-20 16:20:39 -04:00
• 4868312128 [Refactor] Remove dead kernel code (#57621) Wentao Ye 2026-09-20 13:39:23 -04:00
• 49ee12d742 [Misc] Rename --enable-mamba-fine-grained-prefix-cache (#53945 follow-up) (#57382) roikoren755andmergify[bot] <37929162+mergify[bot]@users.noreply.github.com> 2026-09-20 19:50:34 +03:00
• 7d99c2c4fd [Feature][Humming] Humming feature integration (#56685) 3 people 2026-09-20 23:11:01 +08:00
• 51d5b3ac8a Merge remote-tracking branch 'origin/main' into sync/lesj/qwen4exp-bugfixes lesj0610 2026-09-20 22:55:38 +09:00
• 27757dde02 [DSV4.1] Add encoder cuda graph support for deepseek-v4.1-flash (#56625) Isotr0py 2026-09-20 21:51:13 +08:00
• 9b2f34cad4 [Feature] support bf16 MoE router and mxfp4 MoE for MiMo V2 (#57784) ZyannandAbatom 2026-09-20 21:25:56 +08:00
• 10e6a7f210 [Frontend] Expose per-request spec decode metrics in generate API (#43310) 3 people 2026-09-20 20:21:33 +08:00
• bf01fc4a31 [GLM-5.3-Flash] Route kpool indexer top-k through the shared SparseIndexerTopk dispatcher (#57546) Chauncey 2026-09-20 19:17:59 +08:00
• f648eed23d [Bugfix][KV Offload] Skip non-prefix-cacheable groups in SimpleCPUOffload (GLM-5.3-Flash kpool tail and QSA) (#56810) 3 people 2026-09-20 18:58:55 +08:00
• 1b9fa3eaa8 [Perf][GLM] Fuse the kpool tail slot mapping into one Triton kernel (#57534) 3 people 2026-09-20 18:20:55 +08:00
• db1bfdd4fb [Bugfix][GLM-5.3-Flash] Address kpool tail blocks by the padded indexer stride in the NVIDIA prefill seed kernel (#57477) 3 people 2026-09-20 18:15:04 +08:00
• 0748d3bd57 [Model][LoRA] Enable LoRA support for VoyageQwen3BidirectionalEmbedModel (#57708) mahird3 2026-09-20 11:50:00 +02:00
• e5fce7b56b [Core][Kernel] Share persistent workspaces for Marlin and Humming (#57421) Misha GoinandErinYin 2026-09-20 05:28:33 -04:00
• 0e110f696d [ROCm][DSV4][Perf] Use FP8 WO_A output projection (#54894) yinfengLiuandShanshan Shen 2026-09-20 17:13:55 +08:00
• 27b7757f63 [Bugfix][Frontend] Bound the prompt after multimodal expansion (#57076) Dilber 2026-09-20 14:22:27 +05:30
• e378275a8f [Model] Pass intermediate_tensors to the model when capturing CUDA gr… (#57745) Mi Jiazhiandjiazhimi 2026-09-20 14:23:29 +08:00
• d05da62e9c [XPU] fix incorrect gdn kernel log for XPU path (#54871) Yan Ma 2026-09-20 14:00:31 +08:00
• 601e1c7c51 [Perf] Give a prefill-only batch the model state's number of logit rows (#57416) Matt MastracciandClaude Fable 5.1 2026-09-19 22:58:30 -06:00
• 0c90804c0b [Bugfix] DiffusionGemma: cast the self-conditioning soft embed to the buffer dtype (#57462) 3 people 2026-09-19 22:58:02 -06:00