Commits

Branch selector

main

User selector

All users

All time

Commit history

Commits on May 5, 2026

• Merge branch 'vllm-project:main' into main

  inisis

  authored
  0fc4abb

  Copy full SHA for 0fc4abb

• [Model] Use AutoWeightsLoader for CohereMoe (#41690)

  Show description for 0c620d2
  Crystora

  authored
  0c620d2

  Copy full SHA for 0c620d2

• [Model] Fix Gemma4 MoE activation mismatch (#41574)

  Show description for 6bb924b
  lucianommartins

  authored
  6bb924b

  Copy full SHA for 6bb924b

• [BugFix] Preserve max_seq_len in ubatch metadata during CUDA graph capture (#40961)

  Show description for eaec7be
  czhu-cohere

  authored
  eaec7be

  Copy full SHA for eaec7be

• [Ray] Enable RayExecutorV2 by default (#41421)

  Show description for f04fd16
  jeffreywang88

  authored
  f04fd16

  Copy full SHA for f04fd16

• [Hardware][Power]Add Power VSX Attention Backend and fix l2 Cache Crash (#40451)

  Show description for 420b0a5
  

  4 peopleauthored
  420b0a5

  Copy full SHA for 420b0a5

• [ROCm][Quantization][2/N] Refactor quark_moe w4a8 w/ oracle (#39136)

  Show description for 1e95004
  BowenBao

  authored
  1e95004

  Copy full SHA for 1e95004

• [Perf][2/n] Eliminate GPU<->CPU syncs in pooling code (#41433)

  Show description for 416f9cd
  njhill

  authored
  416f9cd

  Copy full SHA for 416f9cd

• [XPU] enable is_act_and_mul for xpu (#37481)

  Show description for 685bf81
  

  xuechendi

  and

  jikunshang

  authored
  685bf81

  Copy full SHA for 685bf81

• [Model Runner V2] Rebuild attn metadata between draft decode steps (#41162)

  Show description for e1e4646
  TheEpicDolphin

  authored
  e1e4646

  Copy full SHA for e1e4646

• [Feature] TurboQuant: support hybrid models and uniform quantization (#39931)

  Show description for 4f2af1a
  

  4 peopleauthored
  4f2af1a

  Copy full SHA for 4f2af1a

Commits on May 4, 2026

• [Bug] Fix status update address for non-MOE model within external dp mode (#40839)

  Show description for 577b962
  yewentao256

  authored
  577b962

  Copy full SHA for 577b962

• [ROCm][CI] Fix MLA prefill scale for DeepSeek GSM8K (#41569)

  Show description for 1cb0838
  AndreasKaratzas

  authored
  1cb0838

  Copy full SHA for 1cb0838

• [Docs] Add non-causal support to attention backend docs (#41643)

  Show description for be5983b
  MatthewBonanni

  authored
  be5983b

  Copy full SHA for be5983b

• [NVFP4][fix] Fix `layer.weight` -> `w13` typo in NVFP4 MOE emulation kernel preparation (#41630)

  Show description for 9c07342
  fxmarty-amd

  authored
  9c07342

  Copy full SHA for 9c07342

• feat: update xgrammar==0.2.0 to use structural tags for strict tool calling + reasoning for more models (#40894)

  Show description for 844df54
  

  4 peopleauthored
  844df54

  Copy full SHA for 844df54

• [bugfix] Fix prompt logprobs on request eviction during chunked prefill (#41411)

  Show description for 422dd02
  

  joa-stdn

  and

  claude

  authored
  422dd02

  Copy full SHA for 422dd02

• Fix Nano Nemotron text-only weight loading (#41205)

  Show description for 8c78094
  

  4 peopleauthored
  8c78094

  Copy full SHA for 8c78094

• [ROCm] ROCm7.2.2 + profiler fix + AITER 0.1.12.post2 (#41386)

  Show description for e724b0e
  

  gshtras

  and

  Rohan138

  authored
  e724b0e

  Copy full SHA for e724b0e

• [Bugfix] KimiK2ReasoningParser: guard against buffered end-token in streaming (#41068)

  Show description for 712ad02
  

  4 peopleauthored
  712ad02

  Copy full SHA for 712ad02

• Limit gpu utils and lower max BS on test_transcription_api_correctness.py (#41649)

  Show description for 321fa2d
  ekagra-ranjan

  authored
  321fa2d

  Copy full SHA for 321fa2d

• [Bug] Fix `tests/compile/test_config.py` AttributeError: 'NoneType' object has no attribute 'dtype' (#41288)

  Show description for 3e1ad44
  yewentao256

  authored
  3e1ad44

  Copy full SHA for 3e1ad44

• Test nemotron nano-v2 and nemotron nano-v3 separately, disable super-omni redundant tests (#41616)

  Show description for 8decbfa
  netanel-haber

  authored
  8decbfa

  Copy full SHA for 8decbfa

• Revert "[Doc] Fix RTD build: pytorch.org/docs/stable/objects.inv returns 404" (#41618)

  Show description for 62ba751
  stecasta

  authored
  62ba751

  Copy full SHA for 62ba751

• [Bugfix] Apply ruff-format to hyperclovax.py (#41620)

  Show description for 6f53753
  stecasta

  authored
  6f53753

  Copy full SHA for 6f53753

• [CI] Stabilize cpu offload compressed tensors test (#41102)

  Show description for 6ec9bbe
  AndreasKaratzas

  authored
  6ec9bbe

  Copy full SHA for 6ec9bbe

• [ROCm][CI] Align spec decode logprob test prefill settings (#41335)

  Show description for 01d4d1a
  AndreasKaratzas

  authored
  01d4d1a

  Copy full SHA for 01d4d1a

• [Transformers v5] Vendor HCXVisionConfig for compatibility (#38447)

  Show description for c103c02
  

  HanFa

  and

  claude

  authored
  c103c02

  Copy full SHA for c103c02

• [CI] Clean up remote servers on pytest parent exit (#41570)

  Show description for 67058ca
  AndreasKaratzas

  authored
  67058ca

  Copy full SHA for 67058ca

• [Bench] Forward --seed to CustomDataset and CustomMMDataset shuffle (#40788)

  Show description for 894a025
  Aktsvigun

  authored
  894a025

  Copy full SHA for 894a025

Commits on May 3, 2026

• [Bugfix] Fix degenerate KV cache stride causing TMA cudaErrorIllegalInstruction (#40737)

  Show description for 66dfee7
  

  4 peopleauthored
  66dfee7

  Copy full SHA for 66dfee7

• [Bugfix] Fix FP8 Bias Loading (#41424)

  Show description for db9a84e
  alex-jw-brooks

  authored
  db9a84e

  Copy full SHA for db9a84e

• [Bugfix][Ray] Fix RayExecutorV2 actor name collision with DP > 1 (#40398)

  Show description for cb03fee
  tomeras91

  authored
  cb03fee

  Copy full SHA for cb03fee

• Disable flashinfer autotune temporarily due to correctness issues (#41524)

  Show description for c51df43
  wzhao18

  authored
  c51df43

  Copy full SHA for c51df43

• [Doc] Add Qwen3-30B-A3B-Thinking-2507-FP8 to batch invariance verified models (#41513)

  Show description for 54dc64d
  taneem-ibrahim

  authored
  54dc64d

  Copy full SHA for 54dc64d