Commits

Branch selector

main

User selector

All users

All time

Commit history

Commits on Apr 15, 2026

• [FlashAttention] Don't overwrite `flash_attn_interface.py` when installing precompiled (#39932)

  Show description for c77e596
  MatthewBonanni

  authored
  c77e596

  Copy full SHA for c77e596

• [Bugfix][Perf] Indexer upcast WK to BF16 for fusion (#38928)

  Show description for ac3dac5
  benchislett

  authored
  ac3dac5

  Copy full SHA for ac3dac5

• [Bug] Fix batch invariant test issue, bs=1 with `max_seq_num = 1` (#39320)

  Show description for 39ac640
  yewentao256

  authored
  39ac640

  Copy full SHA for 39ac640

• [Speculative Decoding] Add DFlash speculators config parsing (#38300)

  Show description for 0b790a2
  ZhanqiuHu

  authored
  0b790a2

  Copy full SHA for 0b790a2

• [Bugfix][NIXL] Fix `_logical_to_kernel_block_ids` conversion for non-mamba models (#39724)

  Show description for 41488f2
  ZhanqiuHu

  authored
  41488f2

  Copy full SHA for 41488f2

• [CI] Only build release Docker images when NIGHTLY=1 (#39882)

  Show description for 102d51c
  

  khluu

  and

  claude

  authored
  102d51c

  Copy full SHA for 102d51c

• [Mooncake] Fix mixed MLA+Eagle block-size validation (#39596)

  Show description for 55e1a8e
  

  4 peopleauthored
  55e1a8e

  Copy full SHA for 55e1a8e

• Bug/test eagle dp v2 (#39838)

  Show description for 21e5a9f
  Monishver11

  authored
  21e5a9f

  Copy full SHA for 21e5a9f

• [Test] Fix @create_new_process_for_each_test("fork") in interactive shell pipeline (#29130)

  Show description for 8ad6ff0
  markmc

  authored
  8ad6ff0

  Copy full SHA for 8ad6ff0

• [BugFix] KeyError on scope["method"] for realtime api websocket in AuthenticationMiddleware (#36934)

  Show description for f2145ef
  

  daniebrill

  and

  gemini-code-assist[bot]

  authored
  f2145ef

  Copy full SHA for f2145ef

• [KVConnector][LMCache] Propagate cache_salt through MP connector for per-user cache isolation (#39837)

  Show description for ed33310
  royyhuang

  authored
  ed33310

  Copy full SHA for ed33310

• [SpecDecode][Benchmark] Add SPEED-bench support to benchmarking CLI (#36029)

  Show description for 3cc328a
  

  talorabr

  and

  benchislett

  authored
  3cc328a

  Copy full SHA for 3cc328a

• [XPU] properly handle q_descale on XPU as quant query input not supported (#39676)

  Show description for 3beb57a
  

  yma11

  and

  jikunshang

  authored
  3beb57a

  Copy full SHA for 3beb57a

• FIX: support language_model.backbone naming in NemotronH Nano VL quantization config (#39901)

  Show description for 8b55319
  

  danielafrimi

  and

  root

  authored
  8b55319

  Copy full SHA for 8b55319

• [BugFix][Graph] fix: handle empty sym_shape_indices in PiecewiseBackend. (#39395)

  Show description for db8d4a4
  chaunceyjiang

  authored
  db8d4a4

  Copy full SHA for db8d4a4

• [XPU][MXFP4] add mxfp4 quant op for XPU (#39857)

  Show description for fc701c8
  zufangzhu

  authored
  fc701c8

  Copy full SHA for fc701c8

• [Metrics] Add request_id to FinishedRequestStats to enable correlation between metrics and requests (#39710)

  Show description for 68be0f8
  Csrayz

  authored
  68be0f8

  Copy full SHA for 68be0f8

• [Quantization][Autoround][CPU] Add W4A16 Support (#38192)

  Show description for 60995c0
  Zhenzhong1

  authored
  60995c0

  Copy full SHA for 60995c0

• fix online fp8 for MiniCPM models (#39862)

  Show description for 29e5d10
  yma11

  authored
  29e5d10

  Copy full SHA for 29e5d10

• [kv_offload+HMA][3/N]: Remove block_size from KVEvents (#36644)

  Show description for 235e1f9
  orozery

  authored
  235e1f9

  Copy full SHA for 235e1f9

• [Bugfix] Fix tool_calls Iterable consumed when debug logging is enabled (#34844)

  Show description for 431cea3
  

  95 peopleauthored
  431cea3

  Copy full SHA for 431cea3

• [CI][NIXL] Fix PD CI breakage: pin nixl-cu{12,13} versions (#39851)

  Show description for 799973a
  ZhanqiuHu

  authored
  799973a

  Copy full SHA for 799973a

• [Bugfix] Respect VLLM_WEIGHT_OFFLOADING_DISABLE_PIN_MEMORY in prefetch offloader (#37699)

  Show description for bcc2306
  

  he-yufeng

  and

  Isotr0py

  authored
  bcc2306

  Copy full SHA for bcc2306

• [Test] Refactor hard coded device string in test files under compile/quantization/models/model_executor folders (#38901)

  Show description for 3abf858
  wincent8

  authored
  3abf858

  Copy full SHA for 3abf858

• [Attention Backend] TurboQuant: 2-bit KV cache compression with 4x capacity (#38479)

  Show description for f4b42df
  

  3 peopleauthored
  f4b42df

  Copy full SHA for f4b42df

• [Model Runner V2] Disable piecewise cudagraph mode fallback for eagle draft decodes (#39773)

  Show description for 3bfe55a
  TheEpicDolphin

  authored
  3bfe55a

  Copy full SHA for 3bfe55a

• [CI] Add PyTorch nightly build and test pipeline (#37226)

  Show description for b569620
  atalman

  authored
  b569620

  Copy full SHA for b569620

Commits on Apr 14, 2026

• [Bugfix] Disable FlashInfer CUTLASS MoE on SM121 (DGX Spark) (#39825)

  Show description for 65b9808
  

  mgoin

  and

  claude

  authored
  65b9808

  Copy full SHA for 65b9808

• [Hybrid] Simplify accepted token counting in spec decode for hybrid models (#38372)

  fuscof-ibm

  authored
  507df79

  Copy full SHA for 507df79

• [Bugfix][Mooncake] Fix thread-local CUDA context for NVLink transfers in _send_blocks (#39548)

  Show description for 1696c86
  

  zhewenl

  and

  Zhewen Li

  authored
  1696c86

  Copy full SHA for 1696c86

• [Bug] Fix batch invariance nvfp4 support (#39820)

  Show description for 2ad1029
  yewentao256

  authored
  2ad1029

  Copy full SHA for 2ad1029

• fix(lmcache): correct store for cached requests while enable prefix cache (#39719)

  Show description for b2f749d
  maobaolong

  authored
  b2f749d

  Copy full SHA for b2f749d

• [Reasoning][Frontend] Add model config to adjust_request in reasoning parser (#37848)

  Show description for 70ed015
  

  3 peopleauthored
  70ed015

  Copy full SHA for 70ed015

• [MoE Refactor] Refactor ZeroExpertFusedMoE into new framework (#35549)

  Show description for 19ec9a0
  bnellnm

  authored
  19ec9a0

  Copy full SHA for 19ec9a0

• [MoE] Move GPT OSS Triton kernel experts into fused_moe/experts/ (#39007)

  Show description for 1a9353b
  

  Jackmin801

  and

  robertgshaw2-redhat

  authored
  1a9353b

  Copy full SHA for 1a9353b