Commit Graph

Select branches

Hide Pull Requests

main

patches

sm86

Mono Color

• 804e5377aa [Bugfix][HiSparse][PD] Align region-mapped pulls across logical block sizes (#57077) Nicolò Lucchesi 2026-09-16 18:39:07 +02:00
• 42919b49c5 [P/D] Report prefill worker cache hits in prompt_tokens_details (#54222) Robert ShawandClaude 2026-09-16 11:23:31 -04:00
• dff1bde84d [Model] Qwen4Exp: fp8_e4m3 main KV cache on the QSA path (#55557) 4 people 2026-09-16 17:07:52 +02:00
• e97ff80613 [BugFix][PCP] Handle missing DP metadata in one-sided EP (#54016) 3 people 2026-09-16 16:24:59 +02:00
• ca542ee154 [Model][LoRA] Enable LoRA support for ModernBertModel (#57148) linitra24 2026-09-16 22:16:09 +08:00
• c81eb3e69c [LoRA] Support modules_to_save for sequence classification (#53555) linitra24andJee Jee Li 2026-09-16 22:15:42 +08:00
• ec4a3a5370 [CI] Bump Transformers version to 5.17.0 (#56108) 3 people 2026-09-16 14:34:16 +01:00
• f8b5c11468 [Bugfix][Kimi-K3] Do not classify a stateless first chunk as a decode (#51483) sashko-zakharchukandzjy0516 2026-09-16 15:24:48 +02:00
• 975dca5bb5 [Perf][GDN] Scatter mixed speculative outputs into the caller buffer (#57140) 5 people 2026-09-16 14:03:57 +01:00
• 8b1d188046 [Bugfix][Mooncake] Report request-level KV load failures under HMA (#56855) Nicolò Lucchesi 2026-09-16 14:55:40 +02:00
• 903285fbcc [Bugfix] Let an optional Literal flag accept the None it advertises (#52370) 3 people 2026-09-16 05:14:54 -07:00
• 6a2fdf9ac6 [Model] Optimize Sarvam MLA routing and preserve FP32 router logits (#56034) Harshit KediaandClaude Opus 5 2026-09-16 17:32:47 +05:30
• e582b9f4a2 Merge remote-tracking branch 'origin/main' into sync/lesj/qwen4exp-bugfixes lesj0610 2026-09-16 20:33:18 +09:00
• 3bb7826214 [Agent] Add agent instructions for parser directories (#56883) Flora Feng 2026-09-16 06:55:26 -04:00
• 118e17f5ff [Platform] Move env check function to platform interface (#48599) wangxiyuanandmergify[bot] <37929162+mergify[bot]@users.noreply.github.com> 2026-09-16 18:08:36 +08:00
• d6a1677d55 [Model][DSv4.1] FlashMLA mega attention and the NVFP4 compressed KV cache (#56935) 5 people 2026-09-16 03:07:50 -07:00
• ceb87de065 [Misc] Remove no-op self-assignments across vLLM (#55988) Adababyandshaolila 2026-09-16 17:17:07 +08:00
• 5f203baedb [Docs] Split slash-combined docstring parameters (#57141) Harry MellorandClaude Opus 5 2026-09-16 10:16:32 +01:00
• 82731931ae [Bugfix][CPU] Fall back to CpuPlatform when zentorch fails to import (#54923) Ganesh RandCursor 2026-09-16 14:45:12 +05:30
• 03f67b3ad1 [Perf][DSV4.1] Pad shared experts for native MegaMoE fusion (#56568) Canlin GuoandCodex 2026-09-16 17:08:42 +08:00
• 0384e72693 [UX] Add thinking support to vllm chat (#57045) Misha Goin 2026-09-16 04:54:10 -04:00
• 35f6047c1a [XPU][Bugfix] Fix Qwen2-Audio ValueError on audio clips longer than 30s (#56912) 3 people 2026-09-16 10:54:02 +02:00
• f37c550bf6 [Feature][Frontend] Expose effective attention block size for DCP (#56538) 3 people 2026-09-16 01:29:43 -07:00
• 0d8173d153 [Bugfix][Frontend] Lazy-import model_hosting_container_standards to prevent log suppression (#51366) wenjinhust 2026-09-16 16:25:49 +08:00
• 8be5205abb [Parser] Fix: correct parser frontend handling of reasoning end and boundary tokens (#56635) dev 2026-09-16 13:47:22 +05:30
• 9ca6dbba71 fix(multimodal): tolerate malformed EXIF metadata during hashing (#56527) (#56576) Ama Senevirathneandamasen02 2026-09-16 13:26:50 +05:30
• b68408043d [Bugfix][CPU] Add per-tensor FP8 W8A16 kernel to fix Ministral crash on CPU (#56985) Li, Jiang 2026-09-16 15:49:17 +08:00
• f12fe10c5c [Bugfix][KV Offload] Track cache recency once per request (#51787) MINJUN GILandOr Ozeri 2026-09-16 16:47:26 +09:00
• ab35354c21 [ROCm][AITER] Skip AITER norm kernels when flattening to 2D would copy (#55991) 3 people 2026-09-16 15:26:56 +08:00
• f30a195bbb [Bugfix] Fix incorrect Mamba block allocation estimate that prevents request admission (#57050) Wei Zhao 2026-09-16 03:23:20 -04:00
• 18553c5c9f [Kernel] Use Murmur3 RNG for Gumbel sampling (#51367) 4 people 2026-09-16 14:52:26 +08:00
• 0f8fa53acf [Bugfix][KV Offload] Check cgroup memory before SHM allocation (#54014) 3 people 2026-09-16 14:17:50 +08:00
• e0301afb2a [Bugfix][Responses] Clean up MCP tool sessions once before closing them (#56988) shaohuaxiandCodex 2026-09-15 23:07:06 -07:00
• b3b13c1292 [ROCm][CI] Fix MLA RoPE fused-kernel tests for TheRock image (#57112) MattandAndreas Karatzas 2026-09-15 22:57:40 -07:00
• 323a29d8f2 [Bugfix] Unreadable prompt_embeds payload should be a 400, not a 500 (#55451) 3 people 2026-09-16 00:53:36 -05:00
• 4fe9e6f6e5 [Refactor][GLM-5.3-Flash] Move sparse_attn_indexer_kpool into the model folder and split AMD/NVIDIA (#55358) Jiangyun ZhuandTJian 2026-09-16 13:50:35 +08:00
• 961c5b6b7e [Bugfix][KV Offload] Restore MTP-retained sliding-window history (#56709) 3 people 2026-09-16 13:14:09 +08:00
• 711768fc17 [CI][ROCm] Add opt-in TheRock builds for AMD CI (#56351) Andreas KaratzasandCodex 2026-09-16 00:09:45 -05:00
• 21d2dd56d8 Merge remote-tracking branch 'origin/main' into sync/lesj/qwen4exp-bugfixes lesj0610 2026-09-16 13:41:12 +09:00
• af1c01499b [Bugfix] Make GPU sync checks safe under torch.compile (#56904) 3 people 2026-09-15 21:37:08 -07:00
• 6ecd97f1b7 [Perf][DSpark] Add KV-only context insertion across V4.1 cache formats (#56441) 5 people 2026-09-16 12:34:55 +08:00
• a31ec3a68b [LoRA][Nemotron] Add LoRA support for Nemotron VL models (for the language model only) (#56231) danisereb 2026-09-16 07:20:28 +03:00
• 75dc588269 [KV Offload] Add per-request max_load_tokens control (#55885) alberto 2026-09-16 04:57:38 +01:00
• 622934e61f [XPU] Honor device_ids for worker device placement (#56015) Chaojun ZhangandKunshang Ji 2026-09-16 11:53:23 +08:00
• c16bb6068f [Bugfix][Rust Frontend] Tolerate NaN-corrupted logprobs in engine-core (#51026) IslamandBugen Zhao 2026-09-16 07:10:58 +04:00
• ccbaed5b91 [XPU] Route to fused_qk_rmsnorm_rope_gate triton kernel (#56096) Marceli FylcekandKunshang Ji 2026-09-16 05:05:54 +02:00
• 4ace5cd9d8 [XPU] fix_test_worker_memory_snapshot_for_xpu (#50097) Qiming ZhangandKunshang Ji 2026-09-15 20:05:43 -07:00
• fdfcbbcac7 [CI/Build][Rust Frontend] Retire Buf schema publishing (#57046) 3 people 2026-09-15 20:02:25 -07:00
• 22bb158c21 [Bugfix][ROCm] Fix MiniMax-M3 fused MXFP8 block scale (#53674) maithilijoshi20 2026-09-16 11:59:43 +09:00
• bfd713bf8b [Rust Frontend] Support HF config overrides --hf-overrides (#56931) Bugen Zhao 2026-09-16 10:52:55 +08:00
• 0673de9800 [Rust Frontend][Metrics] Add per-request preemption histogram (#57033) ReidandBugen Zhao 2026-09-16 10:51:17 +08:00
• 7d4bba5498 [Doc] Show how to get Responses prompt token IDs (#55912) Francisco Javier ArceoandOpenAI Codex 2026-09-15 22:48:18 -04:00
• b0898a4937 [Misc] Allow install empty package from pip (#41074) wangxiyuan 2026-09-16 10:02:37 +08:00
• a4d2d9d95e [ROCm][Kimi-K3] Enable a4w4 flydsl kernels for KimiK3 (#53940) 5 people 2026-09-15 18:38:02 -07:00
• c8d1cf077a [ROCm] [Bugfix] Enable Load and Inference of GLM-5.3-Flash Quark MXFP4 Checkpoint (#56176) Colin ZandAndreas Karatzas 2026-09-15 18:25:29 -07:00
• 1fd119def5 [Pooling] Cap max-length padding for chunked embeddings (#56505) Taneem Ibrahimandmergify[bot] <37929162+mergify[bot]@users.noreply.github.com> 2026-09-15 19:31:51 -05:00
• 85c1f58d50 [BugFix][Model Runner V2][Spec Decode] Fix decode instance's multi-layer MTP kv caches during P/D (#55055) Giancarlo Delfin 2026-09-15 17:31:25 -07:00
• 031f5810c1 [Build][NVIDIA] Update public Rubin dependencies and MSA compatibility (#56545) Shang WangandCodex 2026-09-15 16:29:13 -07:00
• 0fefffc934 [KVConnector][NIXL] Support attention-HMA layouts in pipeline-parallel push prefill (#50494) 4 people 2026-09-15 15:43:56 -07:00
• c6fa1f05d1 [Perf][Kernel] Add sampled filtering for persistent top-k (#56346) Misha GoinandOpenAI Codex 2026-09-15 17:15:59 -04:00
• 1257512305 [DSA] Wire DeepGEMM sparse MQA logits into the DeepSeek V4.1 indexer (#56254) Jared WenandClaude Fable 5.1 2026-09-16 05:07:14 +08:00
• f2aad6aa70 [MRV2] Buffer util simplifications (#56888) Nick Hill 2026-09-15 13:36:51 -07:00
• b6ce714354 [Bugfix] Format kernel-import errors eagerly so warning_once does not retain them (#54098) 3 people 2026-09-15 13:35:54 -07:00
• 38ca7a899c [Bugfix][HiSparse] Preserve per-layer offsets in KV cache bindings (#57027) Lucas WilkinsonandOpenAI Codex 2026-09-15 16:27:11 -04:00
• 18f8aa0465 [Refactor] Remove dead kernel code (#56845) Wentao Yeandmergify[bot] <37929162+mergify[bot]@users.noreply.github.com> 2026-09-15 16:17:53 -04:00
• df42d112ee [Config] Infer HiSparse attention config from HiSparseConnector (#57041) Nicolò Lucchesi 2026-09-15 21:44:26 +02:00
• 2fcc524788 [Bugfix] Avoid nested score cancellation handlers for /v1/score alias (#57024) Taneem Ibrahim 2026-09-15 14:41:54 -05:00
• dffbb714e4 [ROCm][Perf] Optimize DSV4.1 K=512 decode top-k on gfx950 (#56743) Fangzhou Ai 2026-09-15 14:34:24 -04:00
• 2f46a1d9c6 [CI] Keep Qwen3 Omni DSpark config fixture complete (#57044) Kevin H. Luu 2026-09-15 11:19:14 -07:00
• 35dc273072 [ROCm][Spec Decode] Add Aiter MLA decode support non-causal draft block (#55966) 5 people 2026-09-15 10:18:09 -07:00
• 241e9391ed [Bugfix][Attention] Stabilize sparse-MLA DCP for GLM PCP evals (#55879) 4 people 2026-09-15 09:58:30 -07:00
• 836bb3839f [Bugfix] Prevent out-of-bounds access in FlashInfer SM90 sparse MLA mixed batches (#56969) Chauncey 2026-09-15 23:03:58 +08:00
• 9446ea1680 [Bugfix][Spec Decode] Only create draft_id_to_target_id when draft vocab differs (#53458) 4 people 2026-09-15 22:59:02 +08:00
• 1e47ec00d2 [Bugfix][Gemma 4] Don't read fft_length when profiling unified audio (#56721) Abhinav VermaandClaude Opus 5 2026-09-15 20:27:32 +05:30
• f0fbc98c48 [Frontend] Share max_num_queued_reqs across API server processes (#54746) Chauncey 2026-09-15 22:50:26 +08:00
• 0136df94b0 [Bugfix][Spec Decode][MoE] Avoid uninitialized EPLB state in DeepSeek V4.1 DSpark drafter (#56387) zack 2026-09-15 22:35:47 +08:00
• 4bac767695 [Perf][Nemotron] Skip redundant latent-MoE all-reduce at TP>1 (~13% decode win) (#52301) 4 people 2026-09-15 22:33:51 +08:00
• a2e8f1ea94 [XPU][CI]Skip test_chat_completion_with_tools in Intel GPU CI (#56934) xiangdongandKunshang Ji 2026-09-15 22:04:09 +08:00
• e6960af33b [Refactor] StructuredOutputManager x Speculative Decoding Refactor (#48200) yzong-rh 2026-09-15 09:48:49 -04:00
• 64856080b0 [Bugfix] Measure complete pooling responses (#56760) Taneem Ibrahim 2026-09-15 08:21:05 -05:00
• 78eaaad2c7 [Rust Frontend] Add iteration token histogram (#56990) Bugen ZhaoandCodex 2026-09-15 20:12:15 +08:00
• bb55077411 [Perf][Kernel] Integrate Mega-mHC from DeepGEMM for DeepSeek V4.1 (reopen of #56255) (#56962) 4 people 2026-09-15 04:01:06 -07:00
• 5be911ee87 [Bugfix] Load reasoning parser plugins before headless engine config (#53124) ALEX LILUZandJiangyun Zhu 2026-09-15 18:31:25 +08:00
• 24bfbd5d4d [Refactor] Derive is_reasoning_end from the engine grammar (#56200) Flora Feng 2026-09-15 06:07:58 -04:00
• 142020c6c2 [Bugfix][Rust Frontend] Honor add_generation_prompt in DeepSeek V3.2/V4/V4.1 renderers (#56593) Yeonwoo SungandBugen Zhao 2026-09-15 19:03:26 +09:00
• 073f883b56 [Perf][DSpark] Collapse DeepSeek-V4.1 draft states before SP all-gather (#56903) 3 people 2026-09-15 02:58:57 -07:00
• abb4cb9fc5 [CI] Shard V1 KV Connectors 1→4 (#56324) 3 people 2026-09-15 05:58:42 -04:00
• 2983fc6286 [CI] Fix NIXL push worker test stub after failure deferral (#56955) Kevin H. Luu 2026-09-15 02:00:58 -07:00
• 2c2cbd84fd [Frontend] Use XGrammar schema constraints for DeepSeek V4.1 (#56408) Yixin DongandFlora Feng 2026-09-15 04:36:16 -04:00
• 676650397f [Feature][PCP] Support decode-only FULL CUDA graphs (#53867) 4 people 2026-09-15 16:27:46 +08:00
• ca67438c08 [Bugfix][Frontend] Handle aborted requests in beam search (#56249) shaohuaxiandCodex 2026-09-15 01:17:30 -07:00
• 2e2fdaa99a [CI] Extend Arm CPU kernel shard timeout (#56601) Kevin H. LuuandSherlock 2026-09-15 01:00:40 -07:00
• 8263ea12bd [Build] Move DeepGEMM pin to the vLLM fork and update the Mega MoE call convention (#56876) Yongye ZhuandClaude Opus 5 2026-09-15 00:57:18 -07:00
• 000c7df9ff [KV Offload] Add KVCR secondary-tier adapter (#53624) 3 people 2026-09-14 23:55:04 -07:00
• 51918e252a [Bugfix][SparseMLA] Fix piecewise cudagraph capture crash in index group (#56825) Jared WenandKimi Code 2026-09-15 14:51:35 +08:00
• 79e205e8cd [Doc] Clarify ITL vs TPOT Prometheus metrics (#55283) +174184 people 2026-09-15 12:05:13 +05:30
• f547c23ec9 [Bugfix][Kimi-K3] Keep transient checkpoints out of prefix-cache eviction (#56794) 3 people 2026-09-14 23:18:24 -07:00
• ec6b14c044 [CI] Select the supported DCP backend for PCP eval (#56677) Kevin H. Luu 2026-09-14 23:05:59 -07:00
• cd10ed6f9f [feature] [xgrammar] support patternProperties/propertyNames/unevaluatedProperties kw for object types (#42904) 3 people 2026-09-15 13:58:52 +09:00
• 7b7e117c29 [CI] Extend DSv4 engine readiness timeout on main (#56735) Kevin H. LuuandOpenAI Codex 2026-09-14 21:54:57 -07:00