vllm
https://github.com/vllm-project/vllm
Python
A high-throughput and memory-efficient inference and serving engine for LLMs
Triage Issues!
When you volunteer to triage issues, you'll receive an email each day with a link to an open issue that needs help in this project. You'll also receive instructions on how to triage issues.
Triage Docs!
Receive a documented method or class from your favorite GitHub repos in your inbox every day. If you're really pro, receive undocumented methods or classes and supercharge your commit history.
Python not yet supported37 Subscribers
View all SubscribersAdd a CodeTriage badge to vllm
Help out
- Issues
- [Bug]: Qwen/Qwen3.6-35B-A3B and Qwen/Qwen3.6-35B-A3B-FP8 issue while inferencing on Intel XPU (4xB70)
- [Bug][Spec Decode] DSpark speculative decoding broken on nightly
- [RFC]: Balance request admission across Model Runner V2 pipeline-parallel batches
- fix: release committed encoder data for resumable sessions
- fix: don't ship uninitialized prompt logprobs after remote prefill
- [Bug]: DSpark speculative decoding triggers FlashInfer MNNVL allreduce "buffer size insufficient" via draft model's embed_input_ids (TP8, GB200 NVL72)
- [Docs] Add a readiness checklist for new tool-calling models
- [Bug]: Gemma4 tool parser (non-streaming): finish_reason=tool_calls with empty tool_calls array — model's move silently lost under concurrent load
- [Feature]: Populate completion_tokens_details in streaming and non-streaming usage responses
- [Bugfix] Transformers backend: tokenize multimodal prompts once
- Docs
- Python not yet supported