Skip to content

Pull requests: vllm-project/vllm

New pull request New
5,609 Open 32,577 Closed
5,609 Open 32,577 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

[Bugfix][Frontend] Expand the assistant tokens mask through multimodal placeholder expansion bug Something isn't working
#57254 opened Sep 16, 2026 by he-yufeng Contributor Loading…
[Bugfix][Core] Keep cache-registered Mamba states out of retirement bug Something isn't working kv-cache-manager
#57253 opened Sep 16, 2026 by sungsooha Contributor Loading…
[Bugfix][ROCm] Add record_logical_topk_ready to ROCMAiterMLASparseImpl (GLM-5.3-Flash boot crash) bug Something isn't working glm rocm Related to AMD ROCm
#57252 opened Sep 16, 2026 by mustafayildirim Loading…
2
[Metrics][KV Offload] Add Prometheus metrics for SimpleCPUOffloadConnector documentation Improvements or additions to documentation kv-connector
#57251 opened Sep 16, 2026 by mevince Contributor • Draft
1
[Core] structured generation mode for DiffusionGemma model (Jev-like) documentation Improvements or additions to documentation
#57250 opened Sep 16, 2026 by mmastrac Loading…
3 of 4 tasks
2
[ROCm][CI] Fix Nixl+Offloading PD edge cases on TheRock image ci/build kv-connector rocm Related to AMD ROCm
#57249 opened Sep 16, 2026 by mawong-amd Contributor Loading…
4 tasks
[Bugfix][Parser] Preserve Gemma4 content when reasoning parsing is disabled bug Something isn't working tool-calling
#57247 opened Sep 16, 2026 by Prudhvivuda Contributor Loading…
2
[Bugfix][ROCm][MoE] Fix AITER MoE routing fallback to eager torch.topk for scoring_func=softmax bug Something isn't working rocm Related to AMD ROCm
#57246 opened Sep 16, 2026 by shantipriya-amd Contributor • Draft
3 of 4 tasks
[LoRA] Add DeepSeek V4.1 text and vision adapter support deepseek Related to DeepSeek models DSv4 DSv4.1 Related to DeepSeek-V4.1 models
#57245 opened Sep 16, 2026 by HollowMan6 Contributor Loading…
3 of 4 tasks
2
[Bugfix][Frontend] Handle malformed Anthropic tool arguments bug Something isn't working frontend
#57244 opened Sep 16, 2026 by subhashpolisetti Contributor Loading…
[Bugfix][Frontend] Support scalar priority and wire priority into LLM.chat() bug Something isn't working frontend
#57243 opened Sep 16, 2026 by subhashpolisetti Contributor Loading…
[Bugfix] Default missing detail for Responses API input images bug Something isn't working frontend tool-calling
#57241 opened Sep 16, 2026 by yzong-rh Contributor Loading…
4 tasks done
[WIP][Model][ROCm][DCP] DeepSeek-V4 decode context parallelism deepseek Related to DeepSeek models DSv4 needs-rebase rocm Related to AMD ROCm
#57239 opened Sep 16, 2026 by edwinlim0919 Contributor • Draft
3 of 9 tasks
1
[CI] Shard attention by whole units with six primary and two mirror replicas ci/build cpu Related to CPU backends deepseek Related to DeepSeek models dflash glm nvidia rocm Related to AMD ROCm
#57238 opened Sep 16, 2026 by Thangnguyenvn98 Contributor Loading…
3
[CI] Folder-shard Spec Decode Speculators + MTP 1→4 ci/build deepseek Related to DeepSeek models qwen Related to Qwen models speculative-decoding
#57237 opened Sep 16, 2026 by Thangnguyenvn98 Contributor Loading…
[A2A] Allow EPLB + SharedExpert Overlap + DeepEPv2 ready ONLY add when PR is ready to merge/full CI is needed
#57236 opened Sep 16, 2026 by robertgshaw2-redhat Collaborator Loading…
6
[Perf][TP] Skip full-vocabulary gather for exact greedy sampling qwen Related to Qwen models
#57235 opened Sep 16, 2026 by WildWestOat Loading…
1
[Bugfix][Multimodal] Tolerate malformed EXIF in Molmo 2 image preprocessing bug Something isn't working multi-modality Related to multi-modality (#4194)
#57234 opened Sep 16, 2026 by Hotragn Contributor Loading…
[Rust Frontend] Bump vllm-proto to 0.3.0 rust
#57233 opened Sep 16, 2026 by alec-flowers Contributor Loading…
[ROCm][Bugfix] Reduce CUDA graph divergences bug Something isn't working nvidia rocm Related to AMD ROCm
#57229 opened Sep 16, 2026 by mawong-amd Contributor Loading…
4 tasks
[Core] Don't issue a blocking collective RPC during engine handshake
#57226 opened Sep 16, 2026 by okorzh-amd Contributor Loading…
4 tasks
7
[LoRA] Add validated receiver-local adapter staging
#57225 opened Sep 16, 2026 by hershg • Draft
1
Cap multimodal content-part uuid length (cache_salt parity) frontend
#57221 opened Sep 16, 2026 by AUTHENSOR Loading…
1
1
[Perf][DSV4.1] Prefer MegaMoE for supported local expert parallelism DSv4.1 Related to DeepSeek-V4.1 models
#57220 opened Sep 16, 2026 by WoosukKwon Collaborator Loading…
feat(mla-prefill): let FlashInfer serve (256,256) MLA prefill via backend="auto" nvidia
#57217 opened Sep 16, 2026 by Anerudhan Contributor • Draft
2
ProTip! Add no:assignee to see everything that’s not assigned.