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 supported21 Subscribers
Add a CodeTriage badge to vllm
Help out
- Issues
- [Perf] Async logits proc and sampling metadata state update
- Allow generated tokens to exceed max_model_len by one
- [Usage]: Multiple vLLM processes with different ports on a Ray cluster
- [Performance]: Multi-Modal Benchmark on NVIDIA A100 – Qwen2.5-VL / MiniCPM-V-4 / InternVL3_5-4B / InternVL3_5-2B
- [Performance]: Fuse padding onto GEMM by making the GEMM out-of-place
- [Bugfix] Fix ShardedStateLoader support for DeepSeek models with MLA scaling parameters
- [Feature]: Enable Automatic Prefix Caching for Qwen3-Next (hybrid attention)
- [Bug]: Mismatched number of arguments
- [Tests] Varun/marlin experts mk tests
- [RFC]: To Inductor partition or to not Inductor partition (by default in v0.11.1)
- Docs
- Python not yet supported