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 supported43 Subscribers
View all SubscribersAdd a CodeTriage badge to vllm
Help out
- Issues
- Zero padded model inputs before execution
- [Model] Qwen3-VL-MoE: remap ModelOpt NVFP4 per-expert weight names (#…
- [Bug]: Significant accuracy discrepancies across different vLLM versions.
- fix(v1/spec-decode): apply min_p to verified target tokens
- [Bug] MiMoV2 load_weights: fused qkv_proj path uses naive chunk(tp,dim=0)[rank], misplaces Q values into K/V slots
- [RFC]: Split Flashattn Forward for Prefill/Decode Separation
- Revert "[Kernel][UX] Add `--linear-backend` arg for linear kernel selection" (#39538)
- [Bug]: kv_cache_offloadig crashes on 0.21.0 - KeyError in self._block_id_to_pending_jobs[bid]
- [Bug]: 这次崩溃的直接原因是 TurboQuant 注意力后端与 MTP 推测解码在 vLLM 0.21.0 版本中的兼容性问题,具体表现为工作区(workspace)预留不足,导致引擎在处理第一个请求时触发断言错误
- [RFC]: Changes in vLLM Model Development
- Docs
- Python not yet supported