Commits

Breadcrumbs

History for

onmain

User selector

All users

All time

Commit history

Commits on May 28, 2026

• [Bugfix][Kernel] TRTLLM NVFP4 MoE chunking (#43599)

  Show description for 381edde
  amitz-nv

  authored
  381eddeView commit details

  Copy full SHA for 381edde

  View code at this pointBrowse repository at this point

Commits on May 27, 2026

• Fix Qwen3-VL and Qwen3-omni-thinker accuracy degradation from deepstack inputs under torch.compile (#43617)

  Show description for 5963c19
  andakai

  authored
  5963c19View commit details

  Copy full SHA for 5963c19

  View code at this pointBrowse repository at this point

• [Bugfix][DFlash]allocate the proper number of lookahead slots (#43733)

  Show description for 7fb9c01
  

  benchislett

  and

  NickLucche

  authored
  7fb9c01View commit details

  Copy full SHA for 7fb9c01

  View code at this pointBrowse repository at this point

• Validate against some config fields being set to 0 (#43794)

  Show description for 2c2c966
  hmellor

  authored
  2c2c966View commit details

  Copy full SHA for 2c2c966

  View code at this pointBrowse repository at this point

• [Quantization] Fix Humming RoutedExperts import (#43540)

  Show description for 206b72c
  

  fallintoplace

  and

  yewentao256

  authored
  206b72cView commit details

  Copy full SHA for 206b72c

  View code at this pointBrowse repository at this point

• [ROCm] mori: add InterNodeV1LL inter-node kernel selection via VLLM_MORI_INTERNODE_KERNEL (#41751)

  Show description for 05c50c7
  jatseng-ai

  authored
  05c50c7View commit details

  Copy full SHA for 05c50c7

  View code at this pointBrowse repository at this point

• Fix early CUDA init (#43791)

  Show description for 41688e2
  hmellor

  authored
  41688e2View commit details

  Copy full SHA for 41688e2

  View code at this pointBrowse repository at this point

• [BugFix] HFValidationError with cloud storage URIs when HF_HUB_OFFLINE=1 (#39155)

  Show description for 1654609
  injaeryou

  authored
  1654609View commit details

  Copy full SHA for 1654609

  View code at this pointBrowse repository at this point

• [Bugfix] Map reasoning_effort to enable_thinking in chat template kwargs (#43401)

  Show description for 52a31cc
  

  ashwing

  and

  chaunceyjiang

  authored
  52a31ccView commit details

  Copy full SHA for 52a31cc

  View code at this pointBrowse repository at this point

• [Kernel] Enable TritonW4A16LinearKernel as CUDA fallback for non-Marlin-aligned W4A16 shapes (#43731)

  Show description for 2272062
  lucianommartins

  authored
  2272062View commit details

  Copy full SHA for 2272062

  View code at this pointBrowse repository at this point

• [ROCm][GPT-OSS] Avoid repeated compile-time `cos_sin_cache.to(bf16)` casts in rotary path (#42833)

  Show description for de12f5c
  akii96

  authored
  de12f5cView commit details

  Copy full SHA for de12f5c

  View code at this pointBrowse repository at this point

• [Frontend] Add MiniCPM5 XML tool call parser (#43175)

  Show description for 683033d
  

  3 peopleauthored
  683033dView commit details

  Copy full SHA for 683033d

  View code at this pointBrowse repository at this point

• [MRV2][BugFix] Fix KV connector handling in spec decode case (#43719)

  Show description for 8c94938
  

  njhill

  and

  yewentao256

  authored
  8c94938View commit details

  Copy full SHA for 8c94938

  View code at this pointBrowse repository at this point

• [ROCm][Perf] Expose AITER MoE sorting dispatch policy via env var (#39177)

  Show description for 7b54690
  nholmber

  authored
  7b54690View commit details

  Copy full SHA for 7b54690

  View code at this pointBrowse repository at this point

• [KVConnector][Mooncake] Wire reset_cache cascade end-to-end (#42694)

  Show description for 1fc2cee
  

  3 peopleauthored
  1fc2ceeView commit details

  Copy full SHA for 1fc2cee

  View code at this pointBrowse repository at this point

• [DSv4] Refactor compressor & Fix ROCm compatibility (#43710)

  Show description for adaa5e4
  WoosukKwon

  authored
  adaa5e4View commit details

  Copy full SHA for adaa5e4

  View code at this pointBrowse repository at this point

• [Deprecation] Deprecate functions as scheduled for v0.21.0 (#43358)

  Show description for c02c758
  

  yewentao256

  and

  mergify[bot]

  authored
  c02c758View commit details

  Copy full SHA for c02c758

  View code at this pointBrowse repository at this point

• [MLA][Attention] Add OOT MLA prefill backend registration mechanism (#43325)

  Show description for aa61381
  MatthewBonanni

  authored
  aa61381View commit details

  Copy full SHA for aa61381

  View code at this pointBrowse repository at this point

• [Attention] Make FlexAttention and FlashAttention use num-blocks first layouts (#42095)

  Show description for 7e33081
  

  3 peopleauthored
  7e33081View commit details

  Copy full SHA for 7e33081

  View code at this pointBrowse repository at this point

• [Perf] Optimize Fp8BlockScaledMMLinearKernel input_scale tensor using new_empty() (#43677)

  Show description for d8eebe6
  xyang16

  authored
  d8eebe6View commit details

  Copy full SHA for d8eebe6

  View code at this pointBrowse repository at this point

• [ROCm][CI] Fix ROCm multimodal Qwen2.5-VL activation compile and Phi4MM ragged image mask handling (#43647)

  Show description for 5bdb181
  AndreasKaratzas

  authored
  5bdb181View commit details

  Copy full SHA for 5bdb181

  View code at this pointBrowse repository at this point

• [Bugfix] Split attention groups by num_heads_q for spec-decode drafts (#43543)

  Show description for dede691
  lucianommartins

  authored
  dede691View commit details

  Copy full SHA for dede691

  View code at this pointBrowse repository at this point

Commits on May 26, 2026

• [Bugfix][V1] Fix TOCTOU race causing intermittent `EADDRINUSE` on multi-API-server DP startup (#42585)

  Show description for 812e7e7
  

  vadiklyutiy

  and

  gemini-code-assist[bot]

  authored
  812e7e7View commit details

  Copy full SHA for 812e7e7

  View code at this pointBrowse repository at this point

• [KV Connector] MooncakeStore: drop dead discard_partial_chunks parameter (#43627)

  Show description for d98cbf4
  

  3 peopleauthored
  d98cbf4View commit details

  Copy full SHA for d98cbf4

  View code at this pointBrowse repository at this point

• [Kernel] Porting fuse_minimax_qk_norm to manual fusion (#43410)

  Show description for 6e50386
  jeejeelee

  authored
  6e50386View commit details

  Copy full SHA for 6e50386

  View code at this pointBrowse repository at this point

• [DSv4] Drop _get_compressed_kv_buffer in DeepseekCompressor (#43690)

  Show description for 193ce88
  WoosukKwon

  authored
  193ce88View commit details

  Copy full SHA for 193ce88

  View code at this pointBrowse repository at this point

• Add LM head quantization support for ModelOpt (#42124)

  Show description for 6f5b533
  meenchen

  authored
  6f5b533View commit details

  Copy full SHA for 6f5b533

  View code at this pointBrowse repository at this point

• [ROCm] Remove MegaMoE integration in deepseek v4 (#43629)

  Show description for c8414a8
  WoosukKwon

  authored
  c8414a8View commit details

  Copy full SHA for c8414a8

  View code at this pointBrowse repository at this point

• [MoE Refactor] W4a8 int8 oracle (#42789)

  Show description for f51bbc6
  

  bnellnm

  and

  robertgshaw2-redhat

  authored
  f51bbc6View commit details

  Copy full SHA for f51bbc6

  View code at this pointBrowse repository at this point

• [MoE Refactor] Migrate ModelOptMxFp8FusedMoE to oracle (#42768)

  Show description for b226dda
  

  bnellnm

  and

  robertgshaw2-redhat

  authored
  b226ddaView commit details

  Copy full SHA for b226dda

  View code at this pointBrowse repository at this point

• [Feat][DSV4] Fuse q pad into deepseek v4 fused kernel (#43162)

  zyongye

  authored
  6ab6ffbView commit details

  Copy full SHA for 6ab6ffb

  View code at this pointBrowse repository at this point

• Fix CuPy runtime deps and restore humming (#43530)

  Show description for a970fb5
  mmangkad

  authored
  a970fb5View commit details

  Copy full SHA for a970fb5

  View code at this pointBrowse repository at this point

• [XPU] Fix fused MoE LoRA kernel crash on XPU by using platform-agnos num_compute_units (#43646)

  Show description for 861b977
  chaojun-zhang

  authored
  861b977View commit details

  Copy full SHA for 861b977

  View code at this pointBrowse repository at this point

• [Model] Use AutoWeightsLoader for InternLM2 (#38278)

  Show description for ebd0692
  

  javierdejesusda

  and

  DarkLight1337

  authored
  ebd0692View commit details

  Copy full SHA for ebd0692

  View code at this pointBrowse repository at this point

• [Reasoning] [Bugfix] Reject invalid thinking_token_budget values (#43402)

  Show description for 739af5c
  

  linzm1007

  and

  mergify[bot]

  authored
  739af5cView commit details

  Copy full SHA for 739af5c

  View code at this pointBrowse repository at this point