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