Pull Requests vllm-project/vllm

[Metrics] Clamp prefill KV computed token metric v1

#42256 opened 2026-05-11 00:17 by sladyn98

Add tool parser for Nvidia Nemotron family models documentation tool-calling nvidia

#42255 opened 2026-05-11 00:14 by sniper35

The Best of Times v1 nvidia

#42254 opened 2026-05-10 23:59 by vamsiDT

docs: add Agent Friendly badge documentation

#42253 opened 2026-05-10 23:47 by AbhiOnGithub

[MoE Refactor] Mk construct needs-rebase cpu gpt-oss nvidia

#42252 opened 2026-05-10 23:27 by bnellnm

[Perf] Auto-compile trivial CustomOp fallbacks to complete GemmaRMSNorm fusion under enforce_eager

#42251 opened 2026-05-10 22:38 by Ray-RP

[Bugfix][Model] Gemma4 MoE routing closure captures per_expert_scale, breaking functional_call substitution bug

#42250 opened 2026-05-10 21:47 by NoeliaBentancor

Add recompute partial rollout sleep mode documentation frontend v1 cpu

#42249 opened 2026-05-10 21:26 by LYMDLUT

[ROCm] Avoid full KV cache dequant in MLA decode fallback rocm v1

#42248 opened 2026-05-10 19:40 by Bortlesboat

[ROCm] Normalize fp8 scales through float32 rocm

#42247 opened 2026-05-10 19:03 by Bortlesboat

Fix EXAONE-4.5 to align with Transformers update

#42246 opened 2026-05-10 18:00 by lkm2835

[Bugfix] Fix prompt_logprobs non-determinism with prefix caching (issue #42019) bug v1

#42245 opened 2026-05-10 17:35 by factnn

Avoid silent weights corruption when loading Nemotron Nano VL with reusable-buffer loaders like runai distributed streaming multi-modality verified

#42244 opened 2026-05-10 17:34 by noa-neria

[Performance] Build logprobs context IDs incrementally v1

#42243 opened 2026-05-10 17:19 by lokashrinav

[LoRA] Support 2D and 3D MoE LoRA adapter at the same documentation frontend qwen

#42242 opened 2026-05-10 17:11 by jeejeelee

[Perf] [Qwen3.5] Reenable torch compile for `rearrange_mixed_qkv` in GDN linear attention qwen

#42241 opened 2026-05-10 17:08 by tjtanaa

[Bugfix][ROCm] Force splitK=0 in AiterFp8BlockScaledMMKernel for determinism bug rocm

#42240 opened 2026-05-10 16:36 by maeehart

[CI][Bugfix] De-flake Fusion E2E TP2 test bug v1

#42238 opened 2026-05-10 15:36 by haosdent

[Bugfix] Rewrite Gemma4 streaming tool parser bug tool-calling

#42237 opened 2026-05-10 15:32 by whytem

[DSv4] Improved dequant gather K cache kernel v1 DSv4

#42236 opened 2026-05-10 15:29 by gau-nernst

[Kernel][Performance] Add FlashInfer cutedsl NVFP4 GEMM backend nvidia

#42235 opened 2026-05-10 15:27 by mmangkad

[Bugfix] Fix scipy audio resampling ratio bug multi-modality

#42233 opened 2026-05-10 15:17 by BWAAEEEK

[Bugfix][GGUF] Fix deepseek2 architecture not supported when loading without config.json bug deepseek

#42232 opened 2026-05-10 15:10 by AjAyrAo43

[Distributed] SymmMem fused allreduce+RMSNorm: portable Triton kernel performance

#42230 opened 2026-05-10 14:09 by LeoYangXY

[Model] Add Mistral3 MM LoRA token helper methods multi-modality mistral

#42228 opened 2026-05-10 13:04 by coldpark

[Bugfix] Fix queue cleanup deadlock in multi-turn benchmark bug performance

#42227 opened 2026-05-10 12:09 by idan-friedman

[CPU] Fix rotary embedding for CPU without flash-attn ops

#42225 opened 2026-05-10 12:01 by jmamou

[MM][CG] Enable encoder Cudagraph for Step3VL documentation v1 multi-modality nvidia

#42224 opened 2026-05-10 11:12 by JisoLya

[CI] Narrow basic_correctness.yaml source dependencies ci/build

#42223 opened 2026-05-10 10:31 by khluu

[CI] Narrow models_language.yaml source dependencies ci/build

#42222 opened 2026-05-10 10:30 by khluu