Commits

Branch selector

 main

User selector

All users

All time

Commit History

Commits on Mar 26, 2026

• [Refactor] Remove unused utils (#38153)

  Show description for bf5eec6
  yewentao256

  authored
  bf5eec6

  Copy full SHA for bf5eec6

• DOC: Documentation pages fixes (#38125)

  Show description for b1cb1d3
  mtsokol

  authored
  b1cb1d3

  Copy full SHA for b1cb1d3

• [XPU] Disable xpu graph by default (#38193)

  Show description for 6ae8bbd
  jikunshang

  authored
  6ae8bbd

  Copy full SHA for 6ae8bbd

• [Doc] Fix outdated reference to CUDAGraphManager (#38209)

  Show description for a9213c0
  DarkLight1337

  authored
  a9213c0

  Copy full SHA for a9213c0

• [Model] Use helper function to run MM processors with token inputs (where applicable) (#38018)

  Show description for 502c41a
  DarkLight1337

  authored
  502c41a

  Copy full SHA for 502c41a

• [Bugfix] Fix DeepGemm E8M0 accuracy degradation for Qwen3.5 FP8 on Blackwell (#38083)

  Show description for 5206901
  vadiklyutiy

  authored
  5206901

  Copy full SHA for 5206901

• [cpu][ci] remove soft-fail for Arm CI and add quant model tests (#37691)

  Show description for 71161e8
  fadara01

  authored
  71161e8

  Copy full SHA for 71161e8

• [Model] Add torch.compile support for InternVL vision encoder (#38049)

  Show description for 38de822
  tianrengao

  authored
  38de822

  Copy full SHA for 38de822

• [Bugfix] Fix benchmark_fused_collective.py (#38082)

  Show description for 2bfbdca
  jeejeelee

  authored
  2bfbdca

  Copy full SHA for 2bfbdca

• Add `/v1/chat/completions/batch` endpoint for batched chat completions (#38011)

  Show description for 2908094
  MatejRojec

  authored
  2908094

  Copy full SHA for 2908094

• [Bugfix][CI] Fix Marlin FP8 Linear Kernel for Compressed Tensors Format (#38092)

  Show description for e6bf9f1
  

  BadrBasowid

  and

  gemini-code-assist[bot]

  authored
  e6bf9f1

  Copy full SHA for e6bf9f1

• Relocate Encoder CUDA graph manager (#38116)

  Show description for 144030c
  

  WoosukKwon

  and

  njhill

  authored
  144030c

  Copy full SHA for 144030c

• [Tool Parser][1/3] Pass tools to ToolParser constructor (#38029)

  Show description for e2db2b4
  sfeng33

  authored
  e2db2b4

  Copy full SHA for e2db2b4

• [Revert] Remove DeepGEMM availability check in DeepseekV32IndexerMetadataBuilder (#38076)

  Show description for 87f05d6
  chaunceyjiang

  authored
  87f05d6

  Copy full SHA for 87f05d6

• [Misc] Optimized check to encapsulate both CUDA and ROCm platforms (#34549)

  Show description for 36f6aed
  AndreasKaratzas

  authored
  36f6aed

  Copy full SHA for 36f6aed

• Disable dual stream execution of input projection for Qwen3 (#38152)

  Show description for 9704a5c
  xyang16

  authored
  9704a5c

  Copy full SHA for 9704a5c

• Fix minimax m2.5 nvfp4 kv scales weight loading (#37214)

  Show description for 7405603
  wzhao18

  authored
  7405603

  Copy full SHA for 7405603

• [Bugfix] Fix Qwen3.5-FP8 Weight Loading Error on TPU (#37348)

  Show description for d7d51a7
  jrplatin

  authored
  d7d51a7

  Copy full SHA for d7d51a7

• Various Transformers v5 fixes (#38127)

  Show description for 3c3c084
  hmellor

  authored
  3c3c084

  Copy full SHA for 3c3c084

Commits on Mar 25, 2026

• [Cohere] Enable Cohere-Transcribe (#38120)

  Show description for 7b54f60
  ekagra-ranjan

  authored
  7b54f60

  Copy full SHA for 7b54f60

• [ROCm]: Update rope+kvcache fusion conditions and disable custom op by default (#36716)

  Show description for a0e8c74
  Rohan138

  authored
  a0e8c74

  Copy full SHA for a0e8c74

• [MultiModal] add support for numpy array embeddings (#38119)

  Show description for 70a2152
  

  3 peopleauthored
  70a2152

  Copy full SHA for 70a2152

• [ROCm] Utilize persistent MLA kernel from AITER (#36574)

  Show description for 978fc18
  SKPsanjeevi

  authored
  978fc18

  Copy full SHA for 978fc18

• [ROCm] Fix MoE kernel test failures on gfx950 (#37833)

  Show description for 7d6917b
  

  3 peopleauthored
  7d6917b

  Copy full SHA for 7d6917b

• [Core][KV Connector] Remove use of num_cached_tokens in error handling (#38096)

  Show description for e38817f
  markmc

  authored
  e38817f

  Copy full SHA for e38817f

• [Frontend] Move APIServerProcessManager target server fn (#38115)

  Show description for 72cad44
  njhill

  authored
  72cad44

  Copy full SHA for 72cad44

• [Misc] Reorganize inputs (#35182)

  Show description for ba2f0ac
  DarkLight1337

  authored
  ba2f0ac

  Copy full SHA for ba2f0ac

• [MoE Kernel] Flashinfer nvfp4 cutedsl moe kernel integration (#38050)

  zyongye

  authored
  678b3c9

  Copy full SHA for 678b3c9

• [2/n] Migrate per_token_group_quant to torch stable ABI (#36058)

  Show description for bf4cc9e
  mikaylagawarecki

  authored
  bf4cc9e

  Copy full SHA for bf4cc9e

• [CI/Docs] Improve aarch64/DGX Spark support for dev setup (#38057)

  Show description for 1ac2ef2
  

  3 peopleauthored
  1ac2ef2

  Copy full SHA for 1ac2ef2

• [compile] Add some more startup tests for top models (#38046)

  Show description for 6e37c46
  zou3519

  authored
  6e37c46

  Copy full SHA for 6e37c46

• [Refactor] Rename `WAITING_FOR_FSM` to `WAITING_FOR_STRUCTURED_OUTPUT_GRAMMAR` (#38048)

  Show description for 1bf2ddd
  yewentao256

  authored
  1bf2ddd

  Copy full SHA for 1bf2ddd

• [Kernel] Optimize SM120 CUTLASS blockwise FP8 GEMM (#37970)

  Show description for e722118
  

  Nekofish-L

  and

  mgoin

  authored
  e722118

  Copy full SHA for e722118

• [Bugfix] Preserve CUDA arch suffix (a/f) for SM12x — fixes NVFP4 NaN on desktop Blackwell (#37725)

  Show description for 4a76ad1
  

  RobTand

  and

  LucasWilkinson

  authored
  4a76ad1

  Copy full SHA for 4a76ad1

• [Feature] EPLB Support for GPU Model Runner v2 (#37488)

  Show description for d7e93e1
  

  yewentao256

  and

  WoosukKwon

  authored
  d7e93e1

  Copy full SHA for d7e93e1