Commits vllm-project/vllm

Merge branch 'main' into fused-embedding-kernel

jeejeelee committed 5 hours ago

Verified d9351d33

[CI][Ray] Fix flaky multi-node assignment test after placement-group teardown (#53621)

yzong-rh committed 6 hours ago

Verified 680e2177

[CI][Test] Deflake test_mem.py sleep-mode asserts via allocator bookeeping (#54312)

okorzh-amd committed 11 hours ago

Verified 6cddad41

[Bugfix][Model Runner V2][Spec Decode] Decouple the draft's gumbel noise stream from the target's (#54282)

TheEpicDolphin committed 13 hours ago

Verified fe755c88

[Bugfix][Frontend] Only echo the assistant turn in batched chat completions (#52529)

Kaif10 committed 13 hours ago

Verified 7fbfca26

[Bugfix][Frontend] Validate stop_token_ids against vocab size (#54196)

QwertyJack committed 13 hours ago

Verified 5b0e5b69

[Perf] Reuse topk SparseMatrix routing metadata in GPT-OSS MoE forward (#45457)

ShengleiFu committed 13 hours ago

Verified 129087dd

[Nixl][PD] DCP support for MLA models (#50611)

NickLucche committed 14 hours ago

Verified 7f4793ea

[Multimodal] Deprecate PyAV video decoder backend (#54231)

Isotr0py committed 14 hours ago

Verified 4fc943b8

[Mypy] Fix typing for J models (#54130)

taneem-ibrahim committed 17 hours ago

Verified fd5d3aea

[ROCm][CI] Stage E gating (#50920)

AndreasKaratzas committed 23 hours ago

Verified cacc429f

[Bugfix][Spec Decode] Capture the widest uniform decode batch by default (#50488)

rchalamala committed 1 day ago

Verified d3d79ffc

[Perf][KV Connector] Pin token_indices before non_blocking H2D in hf3fs helper (#54293)

JasonKeyiL committed 1 day ago

Verified a758a9f6

[Test] Assert co-located RayExecutorV2 stores publish distinct ports (#54310)

aoshen02 committed 1 day ago

Verified f4f3bcd4

[Bugfix][MRV2] Release model and KV cache on in-process engine shutdown (#54162)

okorzh-amd committed 1 day ago

Verified fb680251

[Bugfix][V1] Keep an encoder cache entry until its last occurrence is freed (#54284)

Hotragn committed 1 day ago

Verified 3958a420

[Attention][DCP] Enable FlashInfer MLA for DSpark drafting (#54277)

GirasoleY committed 1 day ago

Verified 6d4562c5

[Hy4] support Hy4-preview model (#54160)

thisjiang committed 1 day ago

Verified b2f68583

[Perf] Avoid h2d copies from non-pinned CPU tensors (#54299)

njhill committed 1 day ago

Verified 6c18a546

[CI/Build] Use file rendezvous for UniProc loader fixtures (#52367)

yu-xin-c committed 1 day ago

Verified 5182f270

[BugFix] Disable TP for Qwen3-Omni audio encoder when heads % TP != 0 (#50858)

CalvinXKY committed 1 day ago

Verified 738bc881

[Perf] Pin CPU tensors before non_blocking H2D in three MM paths (#54292)

JasonKeyiL committed 1 day ago

Verified 46a83642

[Bugfix][AMD] Annotate draft KV cache groups on the hybrid grouping path (#52047)

okorzh-amd committed 1 day ago

Verified 93ab92be

[Bugfix][Mooncake] Save exact Mamba boundary states (#51358)

ivanium committed 1 day ago

Verified 6b110bad

[Bugfix][Models] Register sleep-managed runtime buffers (#53507)

Ronald1995 committed 1 day ago

Verified 16d6c376

[KV Connector] Support MooncakeStore with hybrid DCP prefix caching (#53324)

wzhao18 committed 1 day ago

Verified 7fd9cc03

[Perf][MLA Sparse] Pin req_id_per_token before non_blocking H2D on XPU and ROCm (#54295)

JasonKeyiL committed 1 day ago

Verified 43196f24

[Rust Frontend] Take the raw buffer in mm tensor lowering when possible (#53528)

roachsinai committed 1 day ago

Verified 4c6c9d56

[Bugfix][Distributed] Gate cross-node MNNVL custom all-reduce by group capability (#53253)

JulianZJN committed 1 day ago

Verified 99013d77

[Bugfix] Fix ncclCommQueryProperties heap overflow with NCCL >= 2.31 (#53008)

Xuan-1998 committed 1 day ago

Verified 026d5af7