Skip to content

Pull requests: vllm-project/vllm

New pull request New
1,120 Open 12,704 Closed
1,120 Open 12,704 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

[V0 Deprecation][KVConnector] Remove KVConnector v1/v0 differentiation ci/build documentation Improvements or additions to documentation kv-connector tpu Related to Google TPUs v1
#25376 opened Sep 22, 2025 by NickLucche Loading…
3
[Misc] Remove unused encoder-decoder error strings ready ONLY add when PR is ready to merge/full CI is needed
#25374 opened Sep 22, 2025 by DarkLight1337 Loading…
5 tasks
1
[Model] Support multi-vector retrieval documentation Improvements or additions to documentation qwen Related to Qwen models
#25370 opened Sep 22, 2025 by noooop • Draft
5 tasks
1
1
[Docs] Fix griffe warnings in vllm/lora/ops
#25369 opened Sep 22, 2025 by windsonsea Loading…
1
[Docs] wheel larger than limit documentation Improvements or additions to documentation
#25367 opened Sep 22, 2025 by pfk-beta Loading…
3
[Bugfix] Qwen3-next generate ! always qwen Related to Qwen models
#25365 opened Sep 22, 2025 by yych0745 Loading…
5 tasks
2
[Core] Enable KV cache connector + hybrid allocator kv-connector tpu Related to Google TPUs v1
#25363 opened Sep 22, 2025 by KuntaiDu Loading…
5 tasks
2
[wip] Fix OOM in vLLM replicas by ensuring consistent NCCL memory accounting ready ONLY add when PR is ready to merge/full CI is needed v1
#25359 opened Sep 22, 2025 by kouroshHakha • Draft
Flex attention hybrid allocator compatibility #25256 v1
#25358 opened Sep 21, 2025 by baonudesifeizhai Loading…
1 of 5 tasks
3
Fp8 lm_head drafter with cudagraph capture speculative-decoding v1
#25357 opened Sep 21, 2025 by yugong333 • Draft
5 tasks
1
[Feature] OTEL Tracing for Individual Model Steps documentation Improvements or additions to documentation v1
#25356 opened Sep 21, 2025 by tomasruizt Loading…
6
[Bugfix] Improve GLM4 MoE Reasoning Parser's is_reasoning_end Condition
#25355 opened Sep 21, 2025 by frankwang28 Loading…
5 tasks done
@chaunceyjiang
3
[Bugfix][Model] Fix inference for Hunyuan dense models
#25354 opened Sep 21, 2025 by Anionex Loading…
3 of 5 tasks
6
[V0 Deprecation] Remove MambaCache & Params
#25353 opened Sep 21, 2025 by WoosukKwon • Draft
3
Use macro guard CUDA functions for back compatibility in grouped_topk_kernel.cu ci/build ready ONLY add when PR is ready to merge/full CI is needed
#25346 opened Sep 21, 2025 by minosfuture Loading…
5
[Feature] Allow configuring FlashInfer workspace size (#25342) v1
#25344 opened Sep 21, 2025 by mishra-krishna Loading…
8
[Feature] Improve GGUF loading from HuggingFace user experience needs-rebase
#25340 opened Sep 21, 2025 by simondanielsson • Draft
3 of 9 tasks
1
1
[V0 Deprecation][Models] Remove all V0 condition for mm embeddings merge deepseek Related to DeepSeek models llama Related to Llama models qwen Related to Qwen models
#25331 opened Sep 21, 2025 by Isotr0py Loading…
1 of 5 tasks
2
Add-new-encoder-models new-model Requests to new models
#25322 opened Sep 20, 2025 by hmellor • Draft
Handle triton kernel import exception
#25319 opened Sep 20, 2025 by minosfuture Loading…
8
upgrade flashinfer to v0.4.0rc1 ci/build v1
#25315 opened Sep 20, 2025 by mmangkad Loading…
5 tasks
2
Make mypy behave like a proper pre-commit hook ci/build frontend needs-rebase ready ONLY add when PR is ready to merge/full CI is needed
#25313 opened Sep 20, 2025 by hmellor Loading…
4
[BugFix] Support EP/DP + EPLB with DeepSeek MTP deepseek Related to DeepSeek models needs-rebase speculative-decoding v1
#25311 opened Sep 20, 2025 by ilmarkov • Draft
5 tasks
1
1
[Bugfix] Enable API server headless mode and scale-out using Dockerfile entrypoint frontend ready ONLY add when PR is ready to merge/full CI is needed v1
#25309 opened Sep 20, 2025 by DarkLight1337 Loading…
5 tasks
1
[torch.compile][Minor Fix] Gate cudagraph_unsafe tag for torch>=2.9
#25304 opened Sep 20, 2025 by BoyuanFeng Loading…
7
ProTip! Mix and match filters to narrow down what you’re looking for.