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 supported60 Subscribers
Add a CodeTriage badge to vllm
Help out
- Issues
- [Performance]: MiniMax-M2.7 throughput regression on TP4+EP in vLLM 0.29 0.31 0.32
- [ROCm][Perf][MoE] Fuse SiLU-and-mul with per-token FP8 quant in TritonExperts and skip remote EP routes
- [ROCm][Perf][Muse Spark] Enable the bf16x3 router GEMM on gfx942
- [ROCm][Perf][Muse Spark] Run the pooler and unpooler through the default ROCm GEMM dispatch on gfx942 and gfx950
- [ROCm][Perf][Muse Spark] Add gfx942 FP8 W8A8 fused-MoE configs for E=32,N=4096
- [ROCm][Perf] Add AITER FlashKDA prefill backend for GLM KDA
- [Bugfix][Parser] Keep streamed tool-call indices contiguous when a call is rejected
- [ROCm][Perf] Use AITER single-launch KDA decode in GLM-5.3-Flash
- [Perf] Tuned batch-invariant fused MoE configs on Blackwell
- [Feature][KV Connector] Pass typed cache axes at registration
- Docs
- Python not yet supported