Shengyu Liu

My homepage

No description.

FlashInfer: Kernel Library for LLM Serving

A tiny yet powerful LLM inference system tailored for researching purpose. vLLM-equivalent performance with only 2k lines of code (2% of vLLM).