Commit Graph

Select branches

Hide Pull Requests

main

patches

sm86

Mono Color

• 9cd956c7e6 [CI/Test] Add expert parallelism coverage to external LB tests (#53497) Tony LinandKevin H. Luu 2026-09-04 15:56:59 +08:00
• 605ca45b91 [XPU] Fix device assignment for DP external LB (#53037) Tony Lin 2026-09-04 15:51:25 +08:00
• a8693df504 [SpecDecode]Fix spec decode warmup device selection (#55245) Kunshang JiandGitHub Copilot 2026-09-04 00:48:45 -07:00
• a5c9179e73 [Qwen3.8-Flash-Next] Compact indexer logits workspace to improve prefill efficiency (#54915) Thien Tran 2026-09-04 15:43:20 +08:00
• 8f816a3f66 [MRV2][Metrics] Support CUDAGraphStat in MRV2 (#52358) YizhouandOpenAI Codex 2026-09-04 15:19:50 +08:00
• 29af8bd672 [XPU][UT] skip GLM-5.3-Flash test on XPU (#55266) Qiming Zhang 2026-09-03 23:45:06 -07:00
• e862c2f45b [Docs] Add example for Renderer.render_cmpl() usage (#54265) Shaurya Pratap Srivastava 2026-09-04 12:05:37 +05:30
• ae2d1ca96f [Rust Frontend] Record Mooncake/NIXL KV-connector metrics (#52755) Ilya MarkovandBugen Zhao 2026-09-04 08:21:54 +02:00
• 8a728663c1 [Bugfix] Reject tokenizer-less Qwen VL processor init (#54886) Luyi Xiaoandmergify[bot] <37929162+mergify[bot]@users.noreply.github.com> 2026-09-03 21:25:35 -07:00
• da55494a12 docs: note that enforce_eager also disables torch.compile (#55271) Anshul Desai 2026-09-03 21:25:30 -07:00
• 78300cdabf [Bugfix][Docs] Package glm5next nvidia subtree and fix its docstrings (#55214) Harry MellorandClaude Opus 5 2026-09-04 05:25:26 +01:00
• e35298628f [Bugfix][KV Connector] Fix DecodeBenchConnector prefix block selection (#54878) Jeff (Junze) MaandOpenAI Codex 2026-09-04 00:05:58 -04:00
• 156050598e [ROCm][CI] Bump ROCk base to ROCm 10.0 (#55246) 3 people 2026-09-03 22:38:07 -05:00
• 8a0a7ee40a [EC Connector] P2P NIXL + CPU EC Connector (#47941) omerpaz95andClaude Opus 5 2026-09-04 06:23:56 +03:00
• 69cf055936 [Performance][DSv4] Size dequant gather launch grid by rows (#55061) aoshen02andaoshen02 2026-09-04 11:06:34 +08:00
• a26b71d868 [Bugfix][PD] Pad resumed speculative decode requests (#55126) Ziming HuangandOpenAI Codex 2026-09-04 11:05:33 +08:00
• 25268f0c9c [KVConnector] Add retention interval to OffloadingConnector (#51886) bnellnmandTyler Michael Smith 2026-09-03 22:42:20 -04:00
• 7dc30f5a66 [ROCm][CI] Build and publish TheRock nightly docker images (#55014) 4 people 2026-09-03 21:34:22 -05:00
• d9e2b5238d [Core][KV Events] Echo session_id on GPU BlockStored events (#51381) 4 people 2026-09-04 10:22:50 +08:00
• 19c018ec05 [Bugfix][DCP] Materialize prefill keys on non-owner ranks (#54908) Yunxiao NingandOpenAI Codex 2026-09-04 10:12:07 +08:00
• 560ef78bfe [Perf][Model Runner V2] Compact sampling masks on GPU instead of unpacking the full-vocab bitmask on CPU (#54901) 3 people 2026-09-04 10:03:36 +08:00
• d6bce42983 [Rust Frontend] Report reasoning tokens in chat completion usage (#54883) Bugen Zhao 2026-09-04 09:47:24 +10:00
• 579aef4e8d [ROCm] Bump AITER to 0.1.21.post1 (#52826) 4 people 2026-09-03 18:28:40 -05:00
• bc2ee48073 [Perf] Prefetch the weight before the PDL wait in fused_q_kv_rmsnorm (#55020) Yongye ZhuandClaude Fable 5 2026-09-03 15:40:16 -07:00
• 9509fc8ae6 [Perf][Kimi-K3] Cut MLA decode concat/cache epilogue latency (#54896) Yongye ZhuandClaude Fable 5 2026-09-03 15:39:58 -07:00
• c7e6e36fa9 [Rust Frontend] Add support for TLS in render server (#54999) 3 people 2026-09-03 23:56:44 +02:00
• 6fdee17f84 [CI] Zen5 image build (#50314) 4 people 2026-09-03 17:34:35 -04:00
• e41011129b [CI] Avoid logging test server environment values (#54379) Taneem Ibrahim 2026-09-03 16:00:29 -05:00
• cee0f92c02 [ROCm][Perf] Optimize MiniMax-M3 decode indexer and top-k (#54682) 3 people 2026-09-03 16:20:18 -04:00
• 2a336d8239 [warmup] overlap renderer warmup and engine core initialization (#54557) Ning XieandCyrus Leung 2026-09-04 03:44:21 +08:00
• fc8f10792c Fix DeepSeek V4 FlashMLA auto KV cache dtype (#45091) 3 people 2026-09-04 03:29:14 +08:00
• 2db1c4dc31 [Rust Frontend] Support --lora-modules for static adapter loading (#54837) Will EatonandBugen Zhao 2026-09-03 15:12:45 -04:00
• 21a2211040 [Rust Frontend] Use token-attributed text in reasoning and unified parsers (#54884) Bugen Zhaoandmergify[bot] <37929162+mergify[bot]@users.noreply.github.com> 2026-09-04 04:05:05 +10:00
• 8bf39632b8 [ROCm][Perf] Add low-M FP32 router GEMM for gfx950 (#54845) Fangzhou Ai 2026-09-03 13:42:25 -04:00
• b7624069ff [Fusion] Manual ActivationQuantFusionPass initial application (#51415) 3 people 2026-09-03 13:37:31 -04:00
• d4d703caf9 [Bugfix][Model] Fix FP8 PLE loading in mixed ModelOpt checkpoints (#54882) sychen52andVadim Gimpelson 2026-09-03 10:09:39 -07:00
• d410fc12f3 [Kernel] Enable Kimi-K3 SiTU on the CuteDSL MoE backend and the SM107 low-latency GEMM plan (#54606) Bolin SunandClaude Fable 5 2026-09-03 12:49:31 -04:00
• 98ed0856f3 [Model] add GLM-5.3-Flash support (#53906) Jiangyun Zhu 2026-09-04 00:40:35 +08:00
• da8ec28268 [Bugfix][KV Offload] Do not let a recurrent group's unhashed block truncate the load boundary (#52807) Yifan Jiang 2026-09-03 09:08:46 -07:00
• 4ae6228284 [Bugfix][KV Connector] Populate SimpleCPUOffload BlockStored metadata (#54325) mevince 2026-09-03 08:58:07 -07:00
• e55b93f296 [Core] Deprecate "all" mamba cache mode (#55041) Nick Hill 2026-09-03 08:56:59 -07:00
• 31e9c13685 [Bugfix][KV Connector] Safely fill circular buffers in DecodeBench (#54879) Jeff (Junze) MaandOpenAI Codex 2026-09-03 11:16:22 -04:00
• edc0fb7e03 Optimize PLE MTP metadata transfers (#55054) bhsueh_NV 2026-09-03 23:02:59 +08:00
• bf95f58d10 [Core] Triton kernel for small-batch top-p only masking (#54651) Nick Hillandmergify[bot] <37929162+mergify[bot]@users.noreply.github.com> 2026-09-03 07:52:47 -07:00
• bb363db9a5 feat: Add support for reasoning_token_count to reasoning parser (#54982) JasonCohereandCursor 2026-09-03 13:02:49 +01:00
• 848ab131bc [Perf] Accumulate Conformer attention scores with baddbmm (#55062) SunandCodex 2026-09-03 19:17:52 +08:00
• 758c79e1c5 [Bugfix] Retain vocab embeddings during replacement (#55083) Taneem Ibrahim 2026-09-03 06:17:38 -05:00
• ee0a4c46ae [Bugfix] Account for PCP in multi-node world size validation (#55111) 3 people 2026-09-03 18:52:42 +08:00
• facd9a74a1 [Model Runner V2][Spec Decode] Skip DP sync for all speculator uniform decodes (#54856) Giancarlo Delfin 2026-09-03 03:32:33 -07:00
• 0d3ede3e3b [Bugfix][Model] Enable torch.compile for StableLM (#54969) djramic 2026-09-03 12:21:54 +02:00
• c21751c90b [Kernel] Warm up Qwen GDN gated RMSNorm (#54251) 3 people 2026-09-03 18:19:48 +08:00
• 4cc0cb6f76 [CI][AMD] Avoid expandable segments in LoRA TP tests (#55094) Andreas Karatzas 2026-09-03 04:55:10 -05:00
• e6eb9074e9 [CI] Bump Transformers version to 5.16.1 (#53905) Harry Mellor 2026-09-03 10:12:43 +01:00
• 096d8e8ce6 [ROCm][CI] Add MiniMax reduce RMS kernel coverage (#55057) Andreas Karatzas 2026-09-03 04:07:59 -05:00
• 1f76efaa21 [Model] Add K2-Horizon model support (#55063) Bowen Tan 2026-09-03 01:17:08 -07:00
• 092334c828 [Bugfix] Wait for offload keys before storing chunks (#52923) 3 people 2026-09-03 16:07:35 +08:00
• 859dd39512 [Bugfix][Core] Wait for the previous PP tensor sends before the next forward pass (#54962) Derek WeitzelandClaude Fable 5.1 2026-09-03 03:01:36 -05:00
• ee17d0d869 Use server-generated keys for late-interaction query caches (#51445) Clinton ThomasandLucas Bourtoule 2026-09-03 03:15:58 -04:00
• 3f1af35e01 [Bugfix][Multimodal] Handle prefix-covered items in SHM worker cache (#54994) waizuichougouandNick Hill 2026-09-03 14:39:34 +08:00
• 27a94d1ce4 [CI] Fix DeepSeek-V4 registry platform guard (#55042) Taneem Ibrahim 2026-09-02 22:27:01 -05:00
• 5d09eb2cf3 [Bugfix][MoE] Preserve unquantized weight storage on ROCm (#46009) 6 people 2026-09-03 11:06:09 +08:00
• e47356c63e [ROCm][Installation] Add mooncake package to image using public wheels (#55002) Giuseppe GrossiandAndreas Karatzas 2026-09-02 22:48:36 -04:00
• 5e4e927b5e [CI] Force HTTP/1.1 for runtime Git installs (#55044) Kevin H. Luu 2026-09-02 19:14:22 -07:00
• 0e14198a63 [Skills] Add kernel benchmark sanity references (#54995) Woosuk Kwon 2026-09-02 17:58:55 -07:00
• ad127d9a0f [Perf][Rust Frontend] Coalesce decoded chunks per engine update (#55012) Bugen Zhao 2026-09-03 08:44:23 +10:00
• 443febe723 [Agents] Expose Triton kernel-writing skill to Claude (#55028) Woosuk KwonandOpenAI Codex 2026-09-02 15:36:12 -07:00
• cf3263d571 [Agents] Add Triton kernel-writing skill (#55019) Woosuk KwonandOpenAI Codex 2026-09-02 15:29:16 -07:00
• 6c6376a09e [CI] Remove deleted nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16 and its arch aliases (#55026) Kevin H. LuuandKimi Code 2026-09-02 15:18:46 -07:00
• a0d3e5c16e [XPU] Add fused GemmaRMSNorm path for eager execution (#53678) 3 people 2026-09-03 06:11:20 +08:00
• 963054ed58 [CI] Exclude nightly-dev tags from nightly DockerHub cleanup (#55023) Kevin H. Luu 2026-09-02 15:09:18 -07:00
• e3e1241003 [ROCm][CI] Extend Multimodal Processor Shard timeout on AMD CI (#55011) Micah Williamson 2026-09-02 16:11:46 -05:00
• 60857baa53 [Bugfix][Rust Frontend][Renderer] Align DeepSeek V4 historical developer message handling (#54854) Reid 2026-09-03 04:43:57 +08:00
• a56654d6de [K3 Perf] Enable DSV3 GEMM for inner-contiguous and row-strided tensors, 12%~81% kernel performance improvement (#54565) Wentao Ye 2026-09-02 15:22:40 -04:00
• 0e3ac4907d [ROCm][CI] Fix false multi-node detection on native CI (#54989) Sheral KumarandAndreas Karatzas 2026-09-02 14:53:07 -04:00
• 1945a94575 [Bugfix][Tests] Stabilize B12X linear kernel checks (#54996) Luke Alonso 2026-09-02 11:46:08 -07:00
• 9b38e3ad53 [CI][MoE] Moe kernels test cleanup (#54954) stefankoncarevic 2026-09-02 20:31:36 +02:00
• 3e9d364ff7 [CI/Build][ROCm] Guard the two CUDA-only tests in test_bf16_skinny_gemm (#54984) stefankoncarevic 2026-09-02 20:21:18 +02:00
• 2691c6cc53 [Bugfix][CI] Set cudagraph_mode=FULL for the Ernie4.5-VL ViT cudagraph test (#54957) stefankoncarevic 2026-09-02 20:18:56 +02:00
• bf7a14d307 [CI][ROCm] Add DSpark evals (#54852) 3 people 2026-09-02 13:14:32 -05:00
• 1356635d83 [New model][Multimodal] Add DeepSeek-V4-Flash-Vision-Exp support (#54566) Isotr0py 2026-09-03 02:09:09 +08:00
• 488e6fd53c [CI] Revert flaky test_quark_int8_w8a8_moe (#54991) fxmarty-amd 2026-09-02 19:58:14 +02:00
• 605c3ddcba [BUILD] Bump cutlass to v4.7.1 (#54190) Shengqi Chen 2026-09-03 01:17:41 +08:00
• d539de1c5d [Docs] Add missing return annotations flagged by griffe (#54980) Harry MellorandClaude Opus 5 2026-09-02 17:56:07 +01:00
• 3b45d053b4 [Bugfix][Model] Fix CohereASR streaming audio-token estimate (unit + subsampling) (#53829) Hung Hoang (黄兴）andCyrus Leung 2026-09-02 22:55:16 +07:00
• 41848caa63 [NIXL] Use int32 array for indices to avoid intermediate conversion (#51952) Ilia Yastrebov 2026-09-02 16:30:16 +02:00
• c6bca6e585 [Bugfix][Multimodal] Scope cache hash kwargs by modality (#54918) waizuichougou 2026-09-02 22:20:26 +08:00
• 872084fb77 [CI] Add Kimi-K3-pruned75-DSpark-TP4 gsm8k eval (#54817) Misha GoinandOpenAI Codex 2026-09-02 10:18:38 -04:00
• ba6c60e98f [Bugfix][KV Offload] Scale UniformTypeKVCacheSpecs groups by DCP (#50883) drakoshaandClaude 2026-09-02 17:11:48 +03:00
• 35faf957d6 [Bugfix][KV Offload] Ignore stale async lookup results (#54872) AlexHuangandOpenAI Codex 2026-09-02 22:10:26 +08:00
• ffe3bb3c72 [Hardware][XPU] Register matmul and linear batch-invariant kernels for XPU (#49209) 4 people 2026-09-02 15:31:38 +02:00
• 7894394b07 [Online quantization] Add targeted online quantization configuration based on user patterns (#51285) fxmarty-amd 2026-09-02 14:55:15 +02:00
• 2a4e3cc3db [Bugfix][KV Offload] Ensure tracker progress for oversized offers (#54759) Cheng RuiandCodex 2026-09-02 04:09:29 -07:00
• f81eb41934 [Bugfix] adjust_dcp_kv_cache_interleave_size for NixlConnector only (#54803) Nicolò Lucchesiandcjackal 2026-09-02 12:31:14 +02:00
• 3140531773 Include chat template fallbacks in package_data (#53762) Tarun Kumar 2026-09-02 15:14:06 +05:30
• 76ba32160a [CI] Shard CPU jobs above the 24h P90 threshold (#54751) Kevin H. Luu 2026-09-02 02:16:05 -07:00
• b205750fe0 [Bugfix][ROCm][Build] fix profiler hang due to queue interposition bug (#54171) 3 people 2026-09-02 10:59:23 +02:00
• 1c26e57d3c [Bugfix] Raise for unavailable piecewise CUDA graphs (#54782) Isotr0py 2026-09-02 16:52:24 +08:00
• 584e8f0dda [Model] Fix GLM-OCR MTP weight loading (#49869) jackLeiandOpenAI Codex 2026-09-02 01:38:36 -07:00
• 87deddc7a4 [CI][Spec Decode] Add MTP placeholder-token regression coverage (#54893) Andreas KaratzasandOpenAI Codex 2026-09-02 03:29:38 -05:00
• c6005000d3 [CI][ROCm] Prefetch safetensors weights in AMD CI (#54898) Andreas KaratzasandOpenAI Codex 2026-09-02 03:28:39 -05:00