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 supported36 Subscribers
View all SubscribersAdd a CodeTriage badge to vllm
Help out
- Issues
- Rename flashinfer_utils.py → flashinfer_moe.py for clarity
- [Bug]: Dramatic KV cache size increase (~40%) for Gemma4 from v0.25.1 to v0.26
- [Bugfix] Reject custom tools on non-Harmony Responses routes
- SM120 NVFP4 KV cache support + MTP cudagraph fix + KV offload crash fix
- [Bugfix][Quantization] Match draft model quant targets under its root prefix
- [Feature]: DeepGEMM kernels are never warmed — only 2 of 24 entry points, so fp8_einsum JIT-loads during serving
- [Bugfix] Fix llama3_json parser dropping parallel and prose-adjacent tool calls
- [WIP] Switch to the Rock, Keep Python 3.12 and Ubuntu 22.04
- [Perf] #48137 costs ~10.6% spec-decode acceptance and #48660 shifts output distributions on DeepSeek-V4-Flash — isolated via #48660-only arm on a production 2-node deployment
- [Bugfix][Kernel] Keep DeepGEMM FP8 quant inside opaque op for TMA scales
- Docs
- Python not yet supported