clinicalnlplab

/

me-llama

like 20

Follow

Clinical NLP Lab 51

Hi there

I was wondering which chat template did you employ for fine-tuning? I noticed in your How to use-Section in the README on Github, you don't mention any chat template.

Looking at the ME-LLaMA-70B-chat model repositories, there is a chat template defined in the tokenizer_config.json, namely the same template as for its backbone Llama-2-70b-chat. However, there is no chat template defined in the tokenizer_config.jsonfor the ME-LLaMA-13B-chat model.

Could you clarify whether ME-LLaMA-13B/17B-chat requires a chat template for proper interaction, and if so, which one? Without this information, it’s difficult to use the model effectively.

Thanks