OpenGenerativeAI / llm-colosseum

Benchmark LLMs by fighting in Street Fighter 3! The new way to evaluate the quality of an LLM

Home Page:https://huggingface.co/spaces/junior-labs/llm-colosseum

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Yi 6b , no action

taozhiyuai opened this issue · comments

截屏2024-04-02 11 17 49

it is yi:6b-fp16 model for two players. why does it not generate any actions?

My guess from my experience trying to get my custom Gonzo-7B-Chat model to battle is... It's expecting ChatML syntax formatting. I re-did my model from an older format when it was just standing there getting it's butt beat to the ChatML format, and now it's beating out Mistral v0.2, Dolphon-Mistral v0.2, andDeepseek-Coder-6.7b-instruct. It's a Mistral v0.1 merged model.