Commits

Breadcrumbs

History for

onmain

User selector

All users

All time

Commit History

Commits on Apr 24, 2026

• [MoE] Move cutlass moe to fused_moe/experts/ (#40574)

  Show description for 079a4cf
  

  Jackmin801

  and

  claude

  authored
  079a4cf

  Copy full SHA for 079a4cf

• [Deprecate] Deprecate LLM.reward offline api, use LLM.encode instead. (#40688)

  Show description for 9744b69
  

  3 peopleauthored
  9744b69

  Copy full SHA for 9744b69

• [Bugfix] Avoid mutating `chat_template_kwargs` in `HYV3ReasoningParser` initialization (#40713)

  Show description for c662b43
  

  BugenZhao

  and

  gemini-code-assist[bot]

  authored
  c662b43

  Copy full SHA for c662b43

• [BUG]: fix HF tokenizer concurrent borrow in tool parsers (#40059)

  Show description for c9bf77d
  

  3 peopleauthored
  c9bf77d

  Copy full SHA for c9bf77d

• [Bugfix] Treat <tool_call> as implicit reasoning end in Qwen3 parser (#35687)

  Show description for 92762ed
  qmx

  authored
  92762ed

  Copy full SHA for 92762ed

• [Feat] Unified Synthetic Acceptance Rate for V1 and V2 (#40662)

  Show description for 626daa2
  

  benchislett

  and

  gemini-code-assist[bot]

  authored
  626daa2

  Copy full SHA for 626daa2

• [Core] Avoid seq_lens_cpu GPU->CPU sync (#40654)

  Show description for fe85a92
  njhill

  authored
  fe85a92

  Copy full SHA for fe85a92

• [EPLB] Remove asyncio infrastructure from Async EPLB (#40730)

  Show description for 62b1bbe
  SageMoore

  authored
  62b1bbe

  Copy full SHA for 62b1bbe

• [MoE] Move remaining PrepareAndFinalize to prepare finalize folder (#39009)

  Show description for 447c372
  

  Jackmin801

  and

  robertgshaw2-redhat

  authored
  447c372

  Copy full SHA for 447c372

Commits on Apr 23, 2026

• [Spec Decode] Move `SpecDecodeBaseProposer` out of `eagle.py` (#40732)

  Show description for cde8d24
  MatthewBonanni

  authored
  cde8d24

  Copy full SHA for cde8d24

• [CI] Split disaggregated tests into own test-area (#40623)

  Show description for 8824f50
  NickLucche

  authored
  8824f50

  Copy full SHA for 8824f50

• [Model] Support Hy3 preview (#40681)

  Show description for d0009dd
  

  stevenkuang-tencent

  and

  jeejeelee

  authored
  d0009dd

  Copy full SHA for d0009dd

• [Bugfix] Include inductor and functorch configs in compilation cache key (#40627)

  Show description for 424033f
  zou3519

  authored
  424033f

  Copy full SHA for 424033f

• [Compilation] Refactor SiluMul activation+quant Fusion Pass (#39684)

  Show description for 2196bac
  BadrBasowid

  authored
  2196bac

  Copy full SHA for 2196bac

• [UT][Hardware] let torchrun example tests use the default backend (#39879)

  Show description for 4a79262
  zhenwei-intel

  authored
  4a79262

  Copy full SHA for 4a79262

• [Misc] Support Human-readable (k/K/m/M..) json cli arg (#40473)

  Show description for 9c2492e
  NickLucche

  authored
  9c2492e

  Copy full SHA for 9c2492e

• [Responses] Add tool_choice/tools validation to match OpenAI behavior (#40399)

  Show description for 8317ced
  

  sfeng33

  and

  mergify[bot]

  authored
  8317ced

  Copy full SHA for 8317ced

• [TurboQuant] enable FA3/FA4 for prefill paths (#40092)

  Show description for fe9c3d6
  

  3 peopleauthored
  fe9c3d6

  Copy full SHA for fe9c3d6

Commits on Apr 22, 2026

• [Bugfix] Fix RMS norm + quant fusion on DeepGEMM UE8M0 path for B200 (#40552)

  Show description for b8401a9
  Lucaskabela

  authored
  b8401a9

  Copy full SHA for b8401a9

• [gRPC] Add standard gRPC health checking (grpc.health.v1) for Kubernetes native probes (#38016)

  Show description for 9c271f9
  V2arK

  authored
  9c271f9

  Copy full SHA for 9c271f9

• [Bugfix][Parser] Fix Mistral pre-v11 tool parser failing on trailing model output (#40531)

  Show description for cfa4921
  

  3 peopleauthored
  cfa4921

  Copy full SHA for cfa4921

• FlexAttention non-causal support (#40394)

  Show description for 29f64c5
  

  fynnsu

  and

  mergify[bot]

  authored
  29f64c5

  Copy full SHA for 29f64c5

• Fix test_startup.py for torch 2.12 (#40636)

  Show description for eb6661d
  angelayi

  authored
  eb6661d

  Copy full SHA for eb6661d

• [NVFP4] NVFP4 MOE emulation fallback for H100/MI300/MI350, standardize `TritonExperts` usage for OCP MX emulation (#35737)

  Show description for d622e27
  

  fxmarty-amd

  and

  kylesayrs

  authored
  d622e27

  Copy full SHA for d622e27

• [Bugfix][CI] Fix `v1/kv_connector/unit/test_nixl_connector_hma.py::test_fewer_blocks_with_hma` (#40597)

  Show description for 33ef194
  NickLucche

  authored
  33ef194

  Copy full SHA for 33ef194

• [Bugfix] LoRA for DeepSeek V3.2 (#35077)

  Show description for a250f1b
  

  3 peopleauthored
  a250f1b

  Copy full SHA for a250f1b

• [Model] Add block-local attention and YaRN for local layers to Gemma3 (#39823)

  Show description for 1236748
  philip-essential

  authored
  1236748

  Copy full SHA for 1236748

• [fix] flaky test_mla_attn_quant_fusion.py (#40530)

  Show description for 4254aeb
  carlyou

  authored
  4254aeb

  Copy full SHA for 4254aeb

• [kv_offload+HMA][8/N]: Support multi-group worker transfer (#38453)

  Show description for aad88f8
  

  orozery

  and

  mergify[bot]

  authored
  aad88f8

  Copy full SHA for aad88f8

• [compile] mla + group fp8 fusion (#38877)

  Show description for 4506319
  

  carlyou

  and

  mergify[bot]

  authored
  4506319

  Copy full SHA for 4506319

• [Multimodal] Add PyAV video backend for concurrent video decoding (#39986)

  Show description for 6f2c71b
  

  jaseelmohd2

  and

  Isotr0py

  authored
  6f2c71b

  Copy full SHA for 6f2c71b

• [AMD][CI][BugFix] Override normalize_e4m3fn_to_e4m3fnuz for fnuz machines in test_moe_layer_no_parallel (#40550)

  Show description for 2463f00
  rasmith

  authored
  2463f00

  Copy full SHA for 2463f00

• test: add nan/inf clamp regression test for fused_topk_bias (#40553)

  Show description for 4679495
  jhaotingc

  authored
  4679495

  Copy full SHA for 4679495

• [Bugfix] avoid warmup if text only expectation in multi_modal run (#40409)

  Show description for 6ff8dea
  khushali9

  authored
  6ff8dea

  Copy full SHA for 6ff8dea

Commits on Apr 21, 2026

• [Startup][UX] Enable CUDAGraph memory profiling by default (#38284)

  Show description for 96a85c5
  

  MatthewBonanni

  and

  tlrmchlsmth

  authored
  96a85c5

  Copy full SHA for 96a85c5