8000 Commits · xeb/vllm · GitHub 8000

Repository navigation

Commits

Commits on Sep 23, 2026
- [Refactor][Quantization] Make FP8 and MLA weight transforms reusable pure functions (#57732)
  
  aoshen02
  and
  zyongye
  authored
  973a3be
Commits on Sep 22, 2026
- [Bugfix] Narrow AuxOutput KV restrictions to known PD connectors (#58150)
  aoshen02
  authored
  feb87b1
- [RL][Sleep] Retain frozen weights across level-2 sleep (#57891)
  
  aoshen02
  and
  codex
  authored
  639461e
Commits on Sep 17, 2026
Commits on Sep 16, 2026
- [Frontend] Return prompt metadata from /inference/v1/generate (#53187)
  aoshen02
  authored
  0983aef
Commits on Sep 15, 2026
- [Bugfix] Fix stale HPC QK-norm weights after weight refit (#56706)
  aoshen02
  authored
  eb61707
- [Bugfix] Validate routed-expert prompt offsets before engine submission (#56844)
  
  authored
  84e5f3e
Commits on Sep 8, 2026
- [Bugfix][ROCm] Mask paged attention V cache padding (#53856)
  aoshen02
  authored
  7555414
Commits on Sep 6, 2026
- [Governance] Add aoshen02 as code owner for RL components (#55529)
  
  aoshen02
  and
  codex
  authored
  808f8cd
- [Core] Sync DP state on the first step of a wave (#52957)
  aoshen02
  authored
  039ea82
Commits on Sep 5, 2026
- [Bugfix][DSv4] Seed the -1 sentinel in the prefill sparse index workspace (#55299)
  aoshen02
  authored
  7fbd44c
Commits on Sep 4, 2026
- [Bugfix][V2] Warm up kernels before capturing CUDA graphs (#55341)
  
  aoshen02
  and
  njhill
  authored
  685074c
- [Bugfix][MoE] Allow TRTLLM FP8 block-scale MoE with SwiGLU clamp (#55069)
  
  authored
  f19431e
- [Performance][DSv4] Size dequant gather launch grid by rows (#55061)
  
  aoshen02
  and
  aoshen02
  authored
  69cf055
- [Perf][Model Runner V2] Compact sampling masks on GPU instead of unpacking the full-vocab bitmask on CPU (#54901)
  
  authored
  560ef78
Commits on Sep 1, 2026
- [Core] Release NCCL communicator memory in sleep mode (#51485)
  
  aoshen02
  and
  claude
  authored
  4ac452a
Commits on Aug 29, 2026
- [Test] Assert co-located RayExecutorV2 stores publish distinct ports (#54310)
  
  aoshen02
  and
  claude
  authored
  f4f3bcd
Commits on Aug 28, 2026
- [BugFix] Bind RayExecutorV2 TCPStore before publishing its port (#50969)
  
  authored
  67e86d1
Commits on Aug 26, 2026
- [Bugfix][DP] Synchronize the device on pause completion (#52914)
  aoshen02
  authored
  17da485
- [RL] Add rank-local IPC weight updates (#52497)
  
  authored
  ee5ae2a
Commits on Aug 18, 2026
- [Bugfix][DP] Don't assume the engines started when forwarding a wake (#51481)
  
  aoshen02
  and
  claude
  authored
  7ddb507
Commits on Aug 13, 2026
- [Frontend] Add routed-experts prompt offset (#51906)
  
  authored
  7553aac
Commits on Aug 12, 2026
- [Model] Support R3 capture with DeepGEMM MegaMoE (#51831)
  aoshen02
  authored
  4eef91c
- [Bugfix] Avoid repeated layerwise reload warning scans (#49505)
  
  aoshen02
  and
  codex
  authored
  f2906b4
Commits on Aug 11, 2026
- [Bugfix][Model Loader] Defer post-load attention weight processing (#49519)
  
  aoshen02
  and
  ywang96
  authored
  99e62b8
- [Frontend] Add content_parts to /inference/v1/generate for raw multim… (#51478)
  
  authored
  3358490
Commits on Aug 10, 2026
- [BugFix] Use file:// rendezvous for single-node executors to eliminate startup port races (#50999)
  
  authored
  11ba93f
Commits on Aug 8, 2026
- [BugFix] Reject invalid data-parallel RPC ports (#51469)
  
  aoshen02
  and
  codex
  authored
  22a1759
- [Bugfix] Fix ZMQ port TOCTOU race in shm_broadcast MessageQueue (#50960)
  
  authored
  f7ef489
Commits on Aug 7, 2026
- [Bugfix] Fix get_open_port() livelock on DP-reserved ports and cover get_open_ports_list (#50965)
  
  authored
  448344c
Commits on Aug 6, 2026
- [BugFix] Dense multinode DP rescope with regression test (#49212)
  
  aoshen02
  and
  andylolu2
  authored
  8cfa01c
- [Bugfix] Size and iterate w13 by shard count for non-gated MoE (#51125)
  
  authored
  7c77868
Commits on Aug 5, 2026
- [R3] Unify routed expert shape configuration (#50940)
  
  aoshen02
  and
  nvyihengz
  authored
  6153dbe

0