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 supported62 Subscribers
View all SubscribersAdd a CodeTriage badge to vllm
Help out
- Issues
- [Bugfix][Rust Frontend] Open a new assistant turn after a final assistant message in the DeepSeek V4/V4.1 renderer
- [Bugfix][Frontend] Sync DeepSeek V4.1 encoder with the 1017 reference (tool-result turns, content-block tool results)
- [Bugfix][ROCm] Fix DeepSeek-V4.1 decode on gfx11/gfx12 (missing inverse RoPE)
- [ROCm] Enable RDNA3 MXFP4 kernels on gfx1151 and for DeepSeek-V4
- [Feature][MRV2] Allow sparse prefill CUDA graph captures
- [Bugfix] Treat negative decimal CLI values as values, not dotted args
- [Bug]: DeepSeek V4.1 rejects reasoning_effort "medium" and "minimal" with 400
- [Bugfix] Fix ShardedStateLoader round-trip for attention/KV-cache scale params
- [Perf] Optimize packed encoder FlexAttention masks
- [DeepSeek-V4.1] Enable Engram disk offload for memory-constrained nodes
- Docs
- Python not yet supported