Commit Graph

Select branches

Hide Pull Requests

main

patches

sm86

Mono Color

• 2bea22002d [BUG] Fix KV offload shared-memory slot selection using worker rank (#58497) Tony Lin 2026-10-10 20:25:11 +08:00
• c41b2639e2 [Bugfix][Responses API] Complete cache-write token accounting (#60973) Kay YanandCodex 2026-10-10 17:08:36 +08:00
• a98247ab4d [Model] Drop the 4x M-RoPE cache headroom for Qwen3-VL, Qwen3.5 and Qwen3.8 (#59853) SeongJun Leeandlesj0610 2026-10-10 15:51:52 +09:00
• c1692c85f1 [Misc] mv chat_utils.py to renderers. (#60522) wang.yuqi 2026-10-10 14:50:34 +08:00
• 699c892dd5 [Perf] Skip the M-RoPE warmup when every holder of the RoPE fuses it (#60246) SeongJun LeeandIsotr0py 2026-10-10 15:47:39 +09:00
• ff440c7805 [BreakableCG] Add eager break for Mamba-family ops (#57881) liuzhenweiandmergify[bot] <37929162+mergify[bot]@users.noreply.github.com> 2026-10-10 14:06:57 +08:00
• a205fb3177 [Bugfix][Model] Restore AXK1 config defaults dropped with the vendored config (#60954) Andreas KaratzasandClaude 2026-10-10 00:54:52 -05:00
• 46fb84c2d4 [Model] Support EAGLE3/DFlash aux hidden states for Granite (#60855) devpramodandClaude Opus 5.5 2026-10-10 01:34:10 -04:00
• fa5cfd48d5 [Refactor] Deduplicate trace header handling in BaseServing (#60828) Chauncey 2026-10-10 13:27:58 +08:00
• 7d0b4e57aa [ROCm] Remove MRV1 model exceptions (#56360) Nick Hill 2026-10-09 22:09:19 -07:00
• 8ea646b827 [Rust Frontend][gRPC] Remove GenerateStream producer task so disconnects abort requests (#60922) 4 people 2026-10-09 21:59:35 -07:00
• 364bc81a9e [Bugfix][LoRA] Fix GPT-OSS gate/up interleaving with mixed MoE LoRA formats (#60812) linitra24 2026-10-10 12:52:51 +08:00
• 5ca26880ad Optimize ppc64le vLLM runtime image size (#48074) Puneet Sharma 2026-10-10 10:09:58 +05:30
• acec1b1d84 [Bugfix][XPU] Remove stale W4A8 weight after repacking (#57511) Guancheng FuandKunshang Ji 2026-10-10 12:27:35 +08:00
• d5f356dae0 [HiSparse] Own GPU residency instead of editing core block-table rows (#60647) Matthew BonanniandClaude Opus 5.5 2026-10-10 00:25:44 -04:00
• 9bd074a996 [Attention] Decode an NVFP4 KV cache with XQA on SM12x (#60452) SeongJun LeeandMisha Goin 2026-10-10 13:24:00 +09:00
• 29e4d2d42e [Frontend][Build] Add vllm download-kernels for FlashInfer precompiled kernels (#58765) Misha GoinandClaude Opus 5.5 2026-10-10 00:22:35 -04:00
• be0823357f [TEST][XPU][CI] Fix ci uva kernel test instability 2 (#60404) Miłosz GrunwaldandKunshang Ji 2026-10-10 06:09:13 +02:00
• f1a3c60558 [Feature][Attention] Cover a whole spec-verify in one DiffKV split-KV program (#59085) Giulio De Pasquale 2026-10-10 04:55:45 +01:00
• 2d69083c4a [ROCm][MXFP4] Align the AITER W4A4 MoE intermediate size to 128 (#55104) 3 people 2026-10-10 11:14:01 +08:00
• 10cc2f6ae2 [Bugfix] Bound GDN speculative decode query widths (#55504) DimensionSTPandOpenAI Codex 2026-10-10 12:00:16 +09:00
• 02aa10539f [Bugfix][ModelLoader] Make enable_weights_track check quantized layers (#57108) 3 people 2026-10-10 10:44:06 +08:00
• 1dfe9fbdae [Core] Release FlashInfer allreduce workspace with --enable-nccl-comm-suspend (#59360) aoshen02andClaude Opus 5.5 2026-10-10 10:26:40 +08:00
• 193922d6ff [ROCm][Perf][DSv4.1] Mono decode layer for DeepSeek-V4.1-Flash on MI355X (#60397) Fangzhou AiandClaude Opus 5.5 2026-10-09 21:53:18 -04:00
• 3c802af3cf [Feat][Frontend] structured decisions: DiffusionGemma read strategy (#60466) 4 people 2026-10-09 19:33:22 -06:00
• 7d1d8660d6 [MRV2] Miscellaneous code structure updates (#60754) Nick Hill 2026-10-09 17:58:09 -07:00
• 9f1c783886 [Test] Allow one fp8 step on K bytes in the kpool decode fuzz test (#60725) 3 people 2026-10-09 19:26:53 -05:00
• f8e6e365ae [Bugfix][Mooncake] Save the EAGLE attention proof with hybrid Mamba partial tails (#60695) 4 people 2026-10-09 17:08:08 -07:00
• 03a7d3776f [Perf] Support minimum KV splits configuration for TokenSpeed MLA decode (#60345) Wei ZhaoandCodex 2026-10-09 19:33:47 -04:00
• 04b02fc8a5 [Core] shm tensor arena for efficient cpu->gpu worker broadcast (#51207) 3 people 2026-10-09 19:13:24 -04:00
• fa94ad4143 [Test] Allow one bf16 ulp in the fused mHC post/pre residual check (#60724) Rohan PotdarandClaude Opus 5.5 2026-10-09 17:54:06 -05:00
• 4c43254d52 [HiSparse] Report host-pool block residency as HiSparse metrics (#59485) Matthew BonanniandClaude Opus 5.5 2026-10-09 18:45:34 -04:00
• 783ac12071 [CI] Accept whitespace around /ci commands and reply instead of going silent (#60925) Kevin H. LuuandClaude Opus 5.5 2026-10-09 15:43:37 -07:00
• ed7ae705a1 [Bugfix][MRV2] Use int64 index mappings to fix row-offset overflow (#60736) 3 people 2026-10-09 14:38:11 -07:00
• 51144dd83e [Bugfix][Spec Decode] Trim token_ids/logprobs left past a stop string under speculative decoding (#47616) 3 people 2026-10-10 05:22:31 +08:00
• 35e920d671 [Rust Frontend] Support --stream-interval and per-request stream_interval (#60589) Bugen ZhaoandClaude Opus 5.5 2026-10-10 05:10:42 +08:00
• ffdde24f54 [Docs] Document adaptive verification without a confidence head (#60913) Giancarlo Delfin 2026-10-09 13:54:40 -07:00
• e0cb3efce0 [Bugfix][Structured Output] Decode choice specs as JSON in outlines and LMFE (#60748) Kai HoangandFlora Feng 2026-10-10 03:52:08 +07:00
• 02e55d2af3 [KV Cache] GLM-5.3-Flash: use the generic packed KV layout; CircularBufferSpec tail with speculative ring slots (#57169) 7 people 2026-10-10 03:48:37 +08:00
• 3766a67a32 [WideEP] Explain how to fix a too-old NCCL for DeepEP v2 (#60449) Misha GoinandClaude 2026-10-09 15:34:58 -04:00
• 04462732a4 [Bugfix] Reject max_num_scheduled_tokens=0 at config time instead of hanging (#60714) 3 people 2026-10-09 23:15:10 +04:00
• ad8a6065f4 [Bugfix] Validate guidance tokenizer compatibility before EngineCore (#52298) Willow LopezandArtem Perevedentsev 2026-10-10 02:59:02 +08:00
• 1dc60dd367 [CI] Pin pydantic in the mypy pre-commit hook (#60894) Misha GoinandClaude 2026-10-09 14:42:41 -04:00
• a0f28d73a8 [sharded state loader] support pp in sharded state loader (#52110) Ning XieandSimon Mo 2026-10-10 02:08:31 +08:00
• c811943137 [Bugfix][Frontend] Reject beam search with echo and logprobs in /v1/completions (#60876) premsurawutandClaude Opus 5.5 2026-10-10 01:06:26 +07:00
• 5dc9bcd8cd [Rust Frontend] Pass reasoning controls through render -> generate (#60546) Bugen ZhaoandClaude Opus 5.5 2026-10-10 02:04:49 +08:00
• ce4603cad6 [Bugfix][Rust Frontend] Pass through raw characters in byte-level decode (#59861) Bugen ZhaoandClaude Opus 5.5 2026-10-10 02:04:00 +08:00
• 9a585c62aa [Rust Frontend] Box rarely-set fields of per-token output types (#60588) Bugen ZhaoandClaude Opus 5.5 2026-10-10 02:01:16 +08:00
• ac305c44cb Merge branch 'main' into patches MOVZX 2026-10-09 17:41:13 +00:00
• 1a3a7a8f01 [Bugfix] Treat max_tokens as output upper bound, not input-space reservation (#57035) 4 people 2026-10-09 22:45:22 +05:30
• 89ac8d971b [CI][Test] Allow 10% ngram acceptance drop under preemption on CUDA in test_async_scheduling_accuracy (#60715) Andrey TalmanandAndrey Talman 2026-10-09 12:59:17 -04:00
• b74233860d [CI][ROCm] Wait for VRAM to settle before building LLM in test_full_graph (#60872) stefankoncarevic 2026-10-09 18:48:43 +02:00
• 75e3b17c05 [Frontend] Add cache salt support to generative scoring (#60841) Taewoong KimandCyrus Leung 2026-10-10 01:07:50 +09:00
• 90ba2f34b5 [Build] Find compatible published wheels for precompiled installs (#58871) 3 people 2026-10-09 08:05:54 -07:00
• 970dc63478 [Docs] Add Giga-Embeddings bidirectional Qwen3 usage to embed docs (#60854) Abhijit Roy 2026-10-09 20:11:38 +05:30
• a6b39f0262 [Bugfix] Clean up async client sockets after event loop closure (#57757) Taneem Ibrahim 2026-10-09 09:32:47 -05:00
• fc39317392 [Bugfix][Kernel] Remove invalid NC from Kimi-K3 KDA autotune keys (#60446) Warren Dengandmergify[bot] <37929162+mergify[bot]@users.noreply.github.com> 2026-10-09 07:25:41 -07:00
• 3d52578e13 [Security] Do not retain Phi-4 MM audio attention matrices (#60614) Juan Pérez de Algabaandmergify[bot] <37929162+mergify[bot]@users.noreply.github.com> 2026-10-09 16:23:35 +02:00
• 81c993658e [ROCm][Perf] Fuse main QK-norm/RoPE/gate and KV-cache write into the AMD QSA prepare launch (#59732) Mikko TukiainenandCursor 2026-10-09 16:14:49 +03:00
• 9adba10f3f [Bugfix] Fix pre-commit check (#60840) Li, Jiang 2026-10-09 20:48:32 +08:00
• 1ee69dde9c [Bugfix][LoRA] Reject PEFT adapter features vLLM does not implement (#60225) Suraj Sharanandlinitra24 2026-10-09 16:39:58 +04:00
• c17fadcfdc [XPU] fix fused_input_norm for MM (#59865) Yan Ma 2026-10-09 20:16:35 +08:00
• 0cc0460870 [XPU]Fix the accuracy issue when meet topk_ids=-1 on DP+EP scenarios (#57787) YiSheng5andKunshang Ji 2026-10-09 20:01:54 +08:00
• 196483e25b [Tools] Recipes sweep workflow (#57307) Louie Tsai 2026-10-09 04:44:09 -07:00
• 1c47ebfc57 [Docs] Fix VLLM_CPU_KVCACHE_SPACE default and units (#60053) Praneeth Reddy AandLi, Jiang <jiang1.li@intel.com> 2026-10-09 17:13:55 +05:30
• d095e61c0b [Bugfix][Spec Decode] Fix GLM-5.3-Flash DFlash2/EAGLE3 aux hidden-state capture with the upstream config (#60763) Chen ChengandIsotr0py 2026-10-09 07:35:48 -04:00
• 7368fadebd [CI Failure] Move MTEB tests to h200_35gb (#60817) wang.yuqi 2026-10-09 19:34:38 +08:00
• 83ed0c7c09 [Bugfix][NIXL] Fix receive post-process for different P/D block sizes (#58860) Luca MotzandClaude Opus 5.5 2026-10-09 13:30:11 +02:00
• 28d76f3bad [Bugfix][Structured Output] Reject unsupported regex inside groups (#60359) Shreyansh JainandArtem Perevedentsev 2026-10-09 16:46:13 +05:30
• b027ac8312 [XPU] Support register KV offload mmap region as pinned host memory on XPU (#51956) Chaojun Zhang 2026-10-09 18:58:07 +08:00
• 5cadabfafd [Docs] Add Qwen2.5-Coder-7B-Instruct to batch invariance tested models (#46368) 3 people 2026-10-09 13:41:29 +03:00
• 4e318ac563 [XPU] fix online fp8_per_channel quantization (#56027) Yan Ma 2026-10-09 18:38:01 +08:00
• efd141047d [CI] Enroll a third batch of five steps in automatic sharding (#60757) Thang NguyenandClaude Opus 5.5 2026-10-09 06:26:47 -04:00
• b8bd335754 [CI] Enroll five more steps in automatic sharding (#60662) Thang NguyenandClaude Opus 5.5 2026-10-09 06:26:27 -04:00
• 59ffd73e85 [Bugfix][Kernel] Restore FP fusion for indexer Q quantization (#60716) Warren Deng 2026-10-09 02:46:41 -07:00
• 2186b89a9a [Bugfix][Model] Voxtral realtime: fix boot OOM and engine crash at max_model_len (#45022) Damien LaineandCyrus Leung 2026-10-09 11:40:29 +02:00
• 2c99ee9333 [Chore] Remove vendored code made redundant by Transformers 5.16.1 (#60623) Harry MellorandClaude Opus 5.5 2026-10-09 10:29:03 +01:00
• bc713cf8cc [Bugfix]MooncakeConnect add query timeout and retry (#55923) Chao Lei 2026-10-09 17:19:33 +08:00
• 18c19eaf23 [Frontend] Add cache salt and tracing to structured decisions (#60786) Taewoong KimandChauncey 2026-10-09 17:53:50 +09:00
• a6f43e57d6 [ROCm] Enable the cuMem CUDA-graph pool offload on ROCm (#59523) 3 people 2026-10-09 16:29:55 +08:00
• bacbbe1878 [Bugfix][NIXL] Guard record_transfer against None telemetry (#57734) The Anh Nguyen 2026-10-09 15:22:18 +07:00
• 560fae64cd [ROCm][Bugfix] Pass per-sequence context lengths to the AITER paged MQA-logits kernels (#59420) Sriram Kumar 2026-10-09 11:16:20 +03:00
• 8a03ccf7b3 [Frontend] Add return_mm_kwargs to render requests (#60728) Maroon AyoubandClaude 2026-10-09 11:14:13 +03:00
• f43e4c39e3 [Structured Output] Allow propertyNames with additionalProperties on xgrammar (#57594) 3 people 2026-10-09 09:49:47 +02:00
• efaca9547d [Perf][DSv4.1] Exchange packed FP8 Engram rows across DP and feed wkv MXFP8 (#59923) 3 people 2026-10-09 00:30:18 -07:00
• 82e89e9944 [Perf][Qwen4Exp] Retune H200 M=4 merged QSA LL-GEMM plan (#60776) 3 people 2026-10-09 15:18:50 +08:00
• 19c0bed584 [Fusion] Support 3D activation quantization fusion and enable it for Pixtral (#59612) 3 people 2026-10-09 09:12:03 +02:00
• dd30786fef [CI] Enable Zen CPU image build on AMD CI (#60016) 3 people 2026-10-09 12:10:21 +05:30
• f0a5f111f2 [XPU] Add tuned Mamba SSU configs for Intel Arc Pro B60 (#56765) Paweł MańczakandKunshang Ji 2026-10-09 08:29:29 +02:00
• ad73a4740d [Perf][DSv4.1] Trim the last KV-source layer in decoder replay (#59894) 3 people 2026-10-09 14:07:49 +08:00
• 8cbd5d0300 [Frontend] structured decisions: noul and score question types (#60696) 4 people 2026-10-08 22:48:16 -06:00
• 88a419649e [ROCm][Kimi-K3] Use vLLM's FP8 max in the fused AttnRes quantization (#60717) Rohan PotdarandClaude Opus 5.5 2026-10-08 23:45:13 -05:00
• fa312090cb [Frontend] Accept each connection on the least loaded API worker (#60723) 3 people 2026-10-08 21:40:13 -07:00
• 81d9c45e4b [Docs] [Multimodal] Note that per-request media_io_kwargs needs --trust-request-mm… (#60598) Rajeev Warrier 2026-10-09 09:53:07 +05:30
• e31b823bf5 [Bugfix][Tool Parser] Use LongCat tags in the LongCat structural tag (#60747) 3 people 2026-10-09 11:20:54 +07:00
• 9824159d94 [Bugfix] Fix InternLM2 tool parser dropping characters from streamed arguments (#60709) 3 people 2026-10-09 08:20:49 +04:00
• 112a282915 [Bugfix][Frontend] Keep streamed completion text_offset in step with tokens (#60668) Vineeth Sai VarikuntlaandFlora Feng 2026-10-08 21:20:43 -07:00
• 51ef239768 [Rust Frontend] Carry reasoning controls over gRPC (#59836) 3 people 2026-10-08 20:11:19 -07:00
• a1cd0ec650 [Model] Add zerank-2 reranker (#60415) Rémi Delhaye 2026-10-09 05:01:51 +02:00
• 702ce313f4 [Perf][Multimodal] Sort multimodal encoder inputs by modality (#60196) zzaebokandmergify[bot] <37929162+mergify[bot]@users.noreply.github.com> 2026-10-09 10:55:36 +08:00