prompt format

#3
by underlines - opened

Thanks for the dataset release and the great model!

Is there any specific prompt format to be followed?

I used the default template in fastchat to fine-tune the model, so it should be the same as vicuna, etc., something like:

A chat between a curious user and an artificial intelligence assistant.  The assistant gives helpful, detailed and polite answers to the user's instructions.

USER: [prompt]
<\s>

ASSISTANT:

Ok, because in your Repo, the example was:

Human: Write an email introducing a new instruction-tuned AI model named airoboros.  The key features about airoboros are: 1. it was fine-tuned with entirely machine-generated synthetic instruction/response pairs  2. the 7 billion parameter model performs very well in comparison to 13b models  3. the synthetic prompt generation tool is free, open-source, and highly customizable.
Assistant: Subject: Introducing airoboros - a new instruction-tuned AI model

Dear [Recipient],

So I guess the model is smart enough to deduce the prefixes, no matter if it's USER/ASSISTANT or HUMAN/ASSISTANT etc...

Ah, yeah that's the older version that I trained using a previous version of FastChat (original vicuna, not 1.1 with updated stop token). I'll add it to the model card.

@jondurbin To clarify, was this model trained with 1.1 of vicuna token format or 1.0? Thanks.

Sign up or log in to comment