Pull Requests vllm-project/vllm

[CI][XPU] Skip CUDA-IPC weight sync metrics test on Intel intel-gpu ci/build nvidia

#59556 opened 2026-10-01 06:42 by zhenwei-intel

[Bugfix][Frontend] Check reused prompt token ids against the vocab before streaming bug ready

#59555 opened 2026-10-01 06:24 by shijie-lyu

[Spec Decode][Watermarking] Support DFlash with dual_key_gumbel documentation speculative-decoding needs-rebase mrv2 dflash

#59554 opened 2026-10-01 06:13 by shernshiou

[ROCm][Perf] Fuse the MXFP4 activation quant into the GDN gated RMSNorm rocm torch.compile quantization

#59553 opened 2026-10-01 06:13 by mjkvaak-amd

k3 chunk kda aiter enablment kimi k3

#59552 opened 2026-10-01 06:12 by omuhamma

[ROCm][Bugfix] Fix ROCM_ATTN sliding-window boundary bug rocm

#59550 opened 2026-10-01 05:42 by tangzzycc

[Rust Frontend] Support logprob_token_ids on OpenAI endpoints rust

#59549 opened 2026-10-01 05:14 by zupengwang

[KV Offload] Give each rank its own CPU region outside the replicated layout

#59547 opened 2026-10-01 04:35 by Etelis

[KV Offload] Split the CPU offload region across shm files

#59546 opened 2026-10-01 04:35 by Etelis

[KV Offload] Split replicated-layout pre-faulting across workers

#59545 opened 2026-10-01 04:35 by Etelis

[KV Offload] Pin only each worker's own offload slots

#59544 opened 2026-10-01 04:35 by Etelis

[Core] Opt-in per-request return of the last hidden states in generation (MRV2) documentation mrv2 scheduler

#59543 opened 2026-10-01 04:29 by aminry

[Bugfix][Spec Decode] Keep heterogeneous-vocab draft models on Model Runner V1 bug

#59541 opened 2026-10-01 04:27 by nishantrevur

[Core] Take the top-k/top-p Triton scratch from WorkspaceManager mrv2

#59540 opened 2026-10-01 04:20 by aoshen02

[Core] Reuse one CUDA graph capture stream per device nvidia

#59537 opened 2026-10-01 04:15 by aoshen02

[Bugfix][MRV2] Keep GDN prefill checkpoint metadata local to each cache group bug ready mrv2 verified

#59536 opened 2026-10-01 04:14 by ai-jz

[Core] Move FlashInfer MLA decode workspaces into WorkspaceManager deepseek nvidia DSv4 DSv4.1

#59535 opened 2026-10-01 04:09 by aoshen02

[Perf][Qwen4Exp] Merge QSA QKVG and indexer QK projections qwen

#59533 opened 2026-10-01 03:19 by gau-nernst

[Perf][DSv4.1] Run decoder replay layers in CUDA graphs and trim in PIECEWISE graphs ready torch.compile deepseek nvidia mrv2 DSv4.1

#59532 opened 2026-10-01 03:09 by Juntian777

[ROCm] Add gfx1151 W8A8 wvSplitK blockscale skinny path performance rocm ci/build deepseek DSv4

#59531 opened 2026-10-01 03:02 by keneoneth

[Bugfix][Bench] Clean up synthetic video files and writers bug performance

#59529 opened 2026-10-01 02:18 by Prudhvivuda

[Bugfix] Keep the GLM-5.3 kpool tail and Qwen4 QSA ring out of the null block bug needs-rebase qwen glm

#59528 opened 2026-10-01 01:55 by ivanium

[Core] Release the checkpoint page cache after weight loading

#59527 opened 2026-10-01 01:42 by gracehonv

[CI] Run pip-compile pre-commit hooks only when requirements change ci/build

#59526 opened 2026-10-01 01:20 by tahsintunan

[CI] Use vllm_runner in fusions_e2e conftest for reliable GPU cleanup ready torch.compile

#59525 opened 2026-10-01 01:06 by divakar-amd

[Bugfix][NIXL] Roll back incomplete remote-agent handshakes bug kv-connector

#59524 opened 2026-10-01 01:00 by xijiaat

[ROCm] Enable the cuMem CUDA-graph pool offload on ROCm rocm torch.compile nvidia mrv2

#59523 opened 2026-10-01 00:48 by aoshen02

[Feature][KV Connector] Add selective Mooncake recurrent checkpoint leases documentation performance kv-connector mooncake

#59522 opened 2026-10-01 00:47 by 0z5a

[CI][Test] Fix flaky weight sync metrics multi-API-server test needs-rebase

#59521 opened 2026-10-01 00:23 by aarushjain29

[ROCm][CI] Run distributed vision tests in the test_areas Distributed… rocm ci/build

#59519 opened 2026-10-01 00:10 by aarushjain29