Pull Requests vllm-project/vllm

AMD CI Test - unskip moe_sum test rocm

#32039 opened 2026-01-09 17:32 by hongxiayang

[CPU][BugFix] Disable AOT Compile for CPU

#32037 opened 2026-01-09 17:19 by fadara01

[responsesAPI] add unit test for optional function tool call id

#32036 opened 2026-01-09 17:06 by qandrew

[Docs] Add docs about OOT Quantization Plugins documentation

#32035 opened 2026-01-09 16:48 by mgoin

[Refactor] Remove numpy split in async scheduling ready v1

#32034 opened 2026-01-09 16:39 by yewentao256

[CI/Build] Publish CPU images for each release ci/build

#32032 opened 2026-01-09 15:24 by nathan-weinberg

[NIXL][Bugfix] Failure logging overhaul + early metadata free on failure v1 kv-connector

#32031 opened 2026-01-09 15:12 by NickLucche

Add: acceptance length tests speculative-decoding v1

#32030 opened 2026-01-09 14:44 by rahul-tuli

[Refactor] Separate sequence and token pooling types documentation new-model frontend ready qwen

#32026 opened 2026-01-09 11:57 by DarkLight1337

[Frontend] Try multiple prompts in benchmark initial test run performance

#32024 opened 2026-01-09 10:31 by MatteoFari

[Bugfix] fix memory inconsistency in cross-process shared memory

#32022 opened 2026-01-09 10:18 by slippersss

[WIP] Optimize greedy sample. tpu v1

#32021 opened 2026-01-09 10:09 by whx-sjtu

[LoRA][Perf] Improve FusedMoE LoRA performance for small rank

#32019 opened 2026-01-09 09:52 by xyang16

[Frontend] `finish_reason` must be `tool_call` whenever a tool is called frontend

#32018 opened 2026-01-09 08:55 by sanghoon-yn

[MLA] Support DCP + FP8 v1

#32014 opened 2026-01-09 05:19 by LucasWilkinson

[Fix] Qwen3-VL-MoE bitsandbytes 4 bit quant qwen

#32013 opened 2026-01-09 05:09 by Datta0

[MISC] Add strict contiguity check for FlashInfer attention tensors ready v1 nvidia

#32008 opened 2026-01-09 02:21 by vadiklyutiy

Reduce the kernel overhead when num of active loras is smaller than max loras. Multiple cuda graphs are captured for each num of active-loras. v1 nvidia

#32005 opened 2026-01-09 01:41 by yugong333

fused_moe_kernel - cast accumulator after applying router weights

#32002 opened 2026-01-09 00:54 by gnovack

[fix] add cutedsl to global sf ready nvidia

#32001 opened 2026-01-09 00:39 by jiahanc

[Misc] Enable async scheduling by default with spec decoding ready

#31998 opened 2026-01-08 23:31 by njhill

[CI/Build][Hardware][AMD] Fix test_forward_error rocm v1

#31997 opened 2026-01-08 23:03 by rjrock

[MoE Refactor] Move `select_experts` from `FusedMoEQuantMethod` -> `FusedMoE` needs-rebase

#31996 opened 2026-01-08 22:10 by bnellnm

[ROCM] Add ROCm image build to release pipeline rocm ci/build

#31995 opened 2026-01-08 21:56 by dllehr-amd

[Misc][PD] Fix `get_attn_backend` usage in transfer connectors ready v1 kv-connector

#31988 opened 2026-01-08 18:20 by NickLucche

[Feature][#29390]: Add timeout support to MultiprocExecutor.collective_rpc and FutureWrapper v1

#31986 opened 2026-01-08 17:31 by SandishKumarHN

[Kernel] Optimize Sliding Window Attention in 3D Triton Kernel

#31984 opened 2026-01-08 17:18 by jvlunteren

[Bugfix] Fix Fp8 Triton for non-gated MoE (Nemotron)

#31983 opened 2026-01-08 17:09 by danisereb

[Misc] Clean up world_size > avail_gpu warning for ray v1

#31981 opened 2026-01-08 17:05 by ruisearch42

Add mergify label job for "bug" match ci/build

#31980 opened 2026-01-08 16:59 by mgoin