Commit Graph

Select branches

Hide Pull Requests

main

patches

sm86

Mono Color

• 91dab0eb7f [ROCm][Perf] Allocate the pinned PLE prefetch buffer lazily (#58797) Mathew Oddenandopencode+glm-5.3+vllm 2026-09-30 04:43:16 -05:00
• 50319916a3 [ROCm][CI] Move Basic Models (Other) to MI355 (#59400) djramicandAndreas Karatzas 2026-09-30 11:42:25 +02:00
• 16d4ac4ba3 [ROCm][MoE] Support MiMo-V2.6 MXFP4 on gfx942 (#58262) yhcheong (vllmellm) 2026-09-30 17:23:00 +08:00
• 749192ff41 [CI] Skip the IPC weight-checker test on non-CUDA platforms (#59398) aoshen02andClaude Opus 5.5 2026-09-30 17:20:32 +08:00
• 713ec07655 [Frontend][RL] Track HTTP weight operation outcomes and concurrency (#55781) 4 people 2026-09-30 17:19:43 +08:00
• ae66f9a97c [Hardware][PowerPC] Prioritize bfloat16 for auto dtype on PowerPC (#58528) 4 people 2026-09-30 14:44:17 +05:30
• fb91712b38 [Model] Make the BERT/RoBERTa embedding class a class attribute (#59348) Avishek GoswamiandAvishek Goswami 2026-09-30 14:27:21 +05:30
• 3627a6a124 [ROCm][Perf][GLM-5.3-Flash] Stride-aware decode KDA (#57979) Simon Danielsson 2026-09-30 10:16:29 +02:00
• b81984879a [K2 Horizon] fold partial-RoPE permutation into q/k (and norm) weights (#55335) Alexandra Sidorova 2026-09-30 12:03:37 +04:00
• 0a30bc3f9a [Model] Extend device-side mm normalization to GLM4V/GLM5Next (#55389) cjackal 2026-09-30 16:44:46 +09:00
• 8ee4069033 [ROCm][DSv4.1][Perf] Use the shared prefill chunk plan in the ROCm sparse prefill (#58539) Zheng Gong 2026-09-30 15:31:51 +08:00
• 87228cff01 [Rust Frontend] Replay roundtrip output grammars through XGrammar (#59143) Bugen ZhaoandClaude Opus 5.5 2026-09-30 15:29:01 +08:00
• fcd317cf33 [ROCm][DSv4][Perf] Use the shared prefill chunk plan in the ROCm sparse prefill (#58405) Zheng GongandCursor 2026-09-30 15:26:57 +08:00
• 583637f126 [CPU][Perf] Add vectorized Sampler Kernel (#53913) Rehan KhanandLi, Jiang <jiang1.li@intel.com> 2026-09-30 12:46:28 +05:30
• 77e8b02c8d [ROCm][Perf] Add opt-in a4w4 (FP4 activation) MoE for DeepSeek V4.1 on AITER (#58819) Fangzhou AiandCursor Agent 2026-09-30 03:07:44 -04:00
• e5f3d08d72 [Frontend][RL] Align sleep-mode API responses and operation metrics (#52864) 3 people 2026-09-30 14:16:38 +08:00
• 29773c722b [Observability] log sub process killing in process manager force kill (#52314) Ning Xie 2026-09-30 14:13:24 +08:00
• 866fa130fa [Core] Bound UniProc EngineCore startup threads to available CPUs (#58946) Andreas Karatzas 2026-09-30 01:02:55 -05:00
• 5faf81a429 [CPU][Whisper] Support W4A16 quantized Whisper on the CPU WNA16 kernel (#58268) 3 people 2026-09-30 11:22:53 +05:30
• fcf4750466 [Bugfix][Frontend] Constrain forced named tool choice with empty parameters to a JSON object (#45290) Ting SUNandFlora Feng 2026-09-30 13:44:35 +08:00
• 6945d142e5 [XPU] Use encoder-only model runner for EC producer instances (#59320) liuzhenwei 2026-09-30 13:43:21 +08:00
• 1117140edb [Bugfix][Model] Fix DiffusionGemma silently freezing attention mask under CUDA graph replay (#51994) Fabian Joswigandmergify[bot] <37929162+mergify[bot]@users.noreply.github.com> 2026-09-30 07:34:26 +02:00
• 329e6eee3f [Bugfix][Frontend] Report named Anthropic tool calls as tool_use (#47598) Ting SUNandFlora Feng 2026-09-30 13:32:38 +08:00
• ac68c30872 [Bugfix][Frontend] Ignore reused prompt token ids for media in /v1/responses (#59173) Shijie Lyu 2026-09-29 22:21:09 -07:00
• a3af78384a [Agents] Add API compatibility checking skill (#58003) Flora FengandClaude Fable 5.1 2026-09-30 00:59:48 -04:00
• 834ad456d3 [Refactor] Remove dead tests code (#58916) Wentao Yeandmergify[bot] <37929162+mergify[bot]@users.noreply.github.com> 2026-09-30 00:39:03 -04:00
• 90e13fc757 [Bugfix][HiSparse] Never allocate GPU pages without host backing (#59036) 3 people 2026-09-30 00:13:29 -04:00
• e006d761a5 [Bugfix][CPU][DiffusionGemma] Support narrower canvas w/ sync scheduling (#59107) Mia Ojeda 2026-09-29 21:28:50 -06:00
• 1b77cc3d67 [Benchmark] Add sweep warmup and failure recovery (#57305) Louie Tsai 2026-09-29 20:16:41 -07:00
• 42d9f8ccdb support rl feature : weight checker (#51350) 7 people 2026-09-30 11:13:43 +08:00
• a8e069f81b [Bugfix][MiniMax M3] Share target embeddings with MTP under PP (#58648) 4 people 2026-09-30 11:04:10 +08:00
• 2798f66860 [Bugfix][HiSparse] Resolve MTP verification rows with a union residency kernel (#59235) Matthew BonanniandClaude Opus 5.5 2026-09-29 23:01:45 -04:00
• bec710fccf [CI] Fix mock type narrowing in LoRA serving test (#59344) Kevin H. LuuandCodex 2026-09-29 19:55:04 -07:00
• 678baf5372 [Dependency] Upgrade FlashInfer to 0.7.0.post1 (#59323) Wei Zhao 2026-09-29 22:34:36 -04:00
• 8039d3cbfd [Rust Frontend] Use the MiMo structural-tag builder (#59148) Bugen ZhaoandClaude Sonnet 5.5 2026-09-30 09:34:46 +08:00
• 105a4e097b [Bugfix][Frontend] Reject LoRA adapters named after a served model (#59286) Nick HillandClaude Opus 5.5 2026-09-29 18:29:38 -07:00
• 765872e7ed [Core][BugFix] Tag prefix-cache extra keys by source (#51899) 5 people 2026-09-29 21:24:59 -04:00
• 09c47db1ca [XPU][CI]Skip test_abort_timeout_on_prefiller in nightly (#58307) xiangdongandKunshang Ji 2026-09-30 08:29:59 +08:00
• 7ca31ef599 [Bugfix][Responses API] Build streamed final response from streamed items (#59307) Flora Feng 2026-09-29 20:16:05 -04:00
• 4e0a414c44 [Kernel][Perf] Add TP=2/4/8 per-rank shapes to the sm_120 batch-invariant matmul table (#58495) Lio Einaudiandmergify[bot] <37929162+mergify[bot]@users.noreply.github.com> 2026-09-30 07:51:50 +08:00
• be255076d0 [Bugfix] Reject a prefix_match_unit that a single KV cache group cannot honor (#58021) Harshilandmergify[bot] <37929162+mergify[bot]@users.noreply.github.com> 2026-09-29 16:12:15 -07:00
• eec2a86b1e [Security] Bump nltk, aiohttp, pillow, and datamodel-code-generator (#59249) Juan Pérez de AlgabaandCursor 2026-09-30 01:00:36 +02:00
• 0755e69e75 [Bugfix][Responses API] Preserve built-in tool output call IDs (#55596) RongJie GandFlora Feng 2026-09-30 06:45:14 +08:00
• c37f86e572 [Bugfix] GLM-5.3-Flash: fp8 plan dtype on SM90 sparse MLA, and right-size the indexer prefill workspace (#55222) 3 people 2026-09-30 01:36:39 +03:00
• 50239dfc4d [WideEP] Change DeepEPv2 to auto select hybrid mode by default (#57991) Tyler Michael SmithandCodex 2026-09-29 18:36:30 -04:00
• 797505747f [Bugfix][Frontend] Honor parallel_tool_calls=false in the Responses API (#59298) Flora Feng 2026-09-29 18:36:11 -04:00
• a7a9ac2745 [Core] Combine per-engine utility results like the Rust client (#59240) 3 people 2026-09-30 06:20:51 +08:00
• 5e887a078b [Bugfix] Route Step3p5 forced tool choices through XML parser (#51810) cherry77-cloudandFlora Feng 2026-09-30 05:58:46 +08:00
• 4e7af3b0de [Bugfix][Frontend] Enforce parallel_tool_calls=false in the required-tool grammar (#50502) Ostring24andFlora Feng 2026-09-30 05:33:28 +08:00
• 1ee7f78e80 [Perf] Reduce redundant Triton sampler warmup specializations (#58605) Misha Goinandmergify[bot] <37929162+mergify[bot]@users.noreply.github.com> 2026-09-29 17:23:27 -04:00
• d882bddbea [Bugfix][Mamba] Keep the prompt-end prefill checkpoint under sparse retention (#59146) 3 people 2026-09-30 05:00:16 +08:00
• 1d331f8341 [Bugfix] Suppress HarmonyError Unexpected token while expecting start token 200006 (#59254) yzong-rh 2026-09-29 16:40:15 -04:00
• faacc13565 [Perf][KV Connector][Mooncake] Pack hybrid/MLA KV into coalesced transfer regions (#57952) 3 people 2026-09-29 22:29:04 +02:00
• 5dd693122c [MyPy][2/N] Fix mypy errors in small tests/ dirs (#51043) 3 people 2026-09-29 21:27:45 +01:00
• a96ee5935d [MRV2] Add DRY as a custom logits processor example (#50584) sashko-zakharchuk 2026-09-29 21:46:00 +02:00
• 19a4d66a0b [Bugfix] Avoid JSON constraints for native tool parsers (#47512) hubuntandFlora Feng 2026-09-30 03:37:12 +08:00
• 7aa8372ef7 [ROCm][CI] Run Basic Correctness Sleep Mode on MI300 for now (#59262) Aarushi Jain 2026-09-29 14:25:01 -05:00
• 30ae8248b6 [KVConnector][MoRIIO] Support K3 DSpark hybrid READ (#57700) YukioZzzandmergify[bot] <37929162+mergify[bot]@users.noreply.github.com> 2026-09-30 03:14:24 +08:00
• 35cd03ebff [CI] Mint the CRCR report's OIDC token after the wait, not before (#59259) Andrey Talman 2026-09-29 14:50:12 -04:00
• 4dc57d44c4 [Frontend] Attach resolved logprobs to streaming derender chunks (#55029) Shimi BandielandClaude 2026-09-29 11:43:44 -07:00
• df4dbe46e7 [CI][ROCm] Increase timeouts for AMD MI300 jobs near their limits (#59227) djramicandAndreas Karatzas 2026-09-29 20:25:49 +02:00
• 91de237333 [Bugfix][DP] add_dp_placement_groups does not require ray[default] (#57648) Chanbin Lim 2026-09-29 11:15:13 -07:00
• 9a190f142c [Mypy] Fix mypy typing for Zamba2 models (#58255) 3 people 2026-09-29 14:13:52 -04:00
• 467d81d9a0 [ROCm] Upgrade MoRI version on rocm dockers required for WideEP DP16 DI CI enablement and fixes for combine API (#56073) 3 people 2026-09-29 23:37:31 +05:30
• afac509a33 [Bugfix][Rust Frontend] Skip engine-derived metrics under --disable-log-stats (#59205) Bugen ZhaoandClaude Opus 5.5 2026-09-30 02:01:54 +08:00
• f5ed3d9c03 [Fast Start] Charge daemon-held weights against gpu_memory_utilization (#57298) 3 people 2026-09-30 01:58:35 +08:00
• cb4f016c5d [Bugfix] Fix standalone torch.compile cache loading after relocation (#52142) Jung JiyuandMisha Goin 2026-09-30 01:53:20 +09:00
• cd94b21264 [Bugfix][Frontend] Return 400 for malformed RL dev route bodies (#59236) 3 people 2026-09-30 00:46:56 +08:00
• 31876d8ea1 [Bugfix] Fix TorchCodec audio IO correctness (#58364) Isotr0py 2026-09-30 00:34:48 +08:00
• c3a4a36c20 [CI][ROCm] Fix stale basic_correctness path in Model Runner V2 Distributed (#59201) stefankoncarevicandAndreas Karatzas 2026-09-29 18:32:04 +02:00
• 2454b5a4f0 [Mypy] Fix mypy typing for Whisper models (#58254) Ashraf BhuiyanandWentao Ye 2026-09-29 12:10:28 -04:00
• f4917dadc8 [Bugfix][Quantization] Stop sleep(level=2) from zeroing compressed-tensors KV scales (#57163) 3 people 2026-09-29 23:34:15 +08:00
• 8aaeef343a [Bugfix][MiMo] Declare embedding_fields so an EPD pair can serve images (#58938) Giulio De Pasquale 2026-09-29 16:33:01 +01:00
• ac7f3e11ea [Bugfix] Profile maximum DeepSeek V4.1 vision features (#57898) 3 people 2026-09-30 00:21:27 +09:00
• 741edeebee [Bugfix][CI] Assert the logger call the Anthropic merge warning makes (#59202) stefankoncarevic 2026-09-29 15:27:34 +02:00
• cd947f5c62 [MM] Fix compiled ViT attention output layouts (#58182) 3 people 2026-09-29 15:27:03 +02:00
• d95d1dcfb9 [Bugfix][Frontend] Force reasoning mode for GLM-5.3 chat templates in the GLM MoE parser (#56994) Shijin ZhangandChauncey 2026-09-29 21:26:34 +08:00
• dfc8e0f3e2 [Frontend] Support strict MiMo-V2.6 tool calling (#58019) 5 people 2026-09-29 08:47:47 -04:00
• 79a0307d74 [Config] Remove Engram CUDA-alike device restrictions (#59171) jackandQwertyJack 2026-09-29 20:42:02 +08:00
• 998490cd9f [Docs] Speed up docs build ~5x (#58705) Harry MellorandClaude Opus 5.5 2026-09-29 12:16:33 +01:00
• 05d89636fc [ROCm][CI] Mirror generic GEMM-RS/AR on MI355 (#58433) 3 people 2026-09-29 05:28:15 -05:00
• 3f3fbe286f [Bugfix][Rust Frontend] Account for new requests in DP routing (#58956) 3 people 2026-09-29 18:24:28 +08:00
• 208bb38c80 [Bugfix][Rust Frontend] Fix startup with config-only model (#58643) Louie Tsai 2026-09-29 03:11:53 -07:00
• af5b4857e1 [Fast Start] Support PP (#55477) 3 people 2026-09-29 17:37:43 +08:00
• 77e52645e9 [Bugfix][Model] MiMo: keep fused fp8 qkv_proj pairing state across weight-loading calls (#58142) yhcheong (vllmellm) 2026-09-29 17:37:20 +08:00
• 4861833ae2 [Bugfix][Core] Allow AuxOutput reset after abort without running requests (#59060) aoshen02andOpenAI Codex 2026-09-29 17:23:32 +08:00
• 3e2a7e74a5 [Feat][Model] Enable KDA prefill checkpoints for GLM-5.3-Flash (#56960) Chauncey 2026-09-29 17:14:30 +08:00
• 3acddf01e7 [Bugfix][ROCm] Keep Mooncake bootstrap ports bound during startup (#58967) Andreas KaratzasandOpenAI Codex 2026-09-29 04:14:26 -05:00
• 6ebb5bd64f [ROCm][CI] Expand MI355 mirrors and route MIG-sized jobs to DPX (#59137) Andreas Karatzas 2026-09-29 03:40:07 -05:00
• 70dc122f81 [MM] Enable device normalization for Llama Nemotron VL Embed/Rerank (#57928) Jakub ZakrzewskiandOpenAI Codex 2026-09-29 10:33:52 +02:00
• 61349e342d [Misc] Avoid repeated warnings for merged Anthropic system messages (#59172) Chauncey 2026-09-29 16:31:22 +08:00
• 4b2e1cfa7b [Model] Decoder-side SWA bounded replay for DeepSeek-V4.1 (#58132) Yifan QiaoandClaude Fable 5.1 2026-09-29 01:25:46 -07:00
• 9af952c555 [Bugfix][Rust Frontend] Support raise_exception in chat templates (#59020) 3 people 2026-09-29 13:31:52 +05:30
• 77d9cda688 [ROCm][Model][Bugfix] Fix GLM-5.2 shared-expert fusion and MTP on the ROCm DSA path (#58904) 4 people 2026-09-29 10:57:23 +03:00
• 30f5c01d4e [DSv4.1] Avoid runtime recompiles of _ring_slot_mapping_kernel (#59119) Juntian LiuandClaude Opus 5.5 2026-09-29 00:31:30 -07:00
• 6cbbea8721 [Bugfix] Hoist $defs/definitions in Cohere parser tool schema composition (#49602) Václav ČadekandCursor Agent 2026-09-29 09:11:56 +02:00
• 78fc1e6274 [ROCm][CI] Add MI355 TP2 AR-RMS and B200 fusion mirrors (#58284) Andreas KaratzasandOpenAI Codex 2026-09-29 02:10:26 -05:00
• 69d5c81fcd [ROCm][CI] Mirror split kernel groups on MI355 and fix exposed tests (#58654) Andreas KaratzasandOpenAI Codex 2026-09-29 02:08:31 -05:00
• 0ae480ff1a [CI][ROCm] Mirror large-model GSM8K evaluations on MI355 (#58683) Andreas KaratzasandCodex 2026-09-29 02:06:44 -05:00
• d86cf91240 [MM][Mistral3] Image preprocessing optimization (#57531) 4 people 2026-09-29 08:53:38 +02:00