Is this chat or completion finetune? Instruction format?
#1
by
jkolcar1
- opened
What is recommended usage? And instruction format? gpt4 is chat model so i suppose this is chat too.
This is a one-round instruction-tuned model, i.e. it's trained on a single instruction -> response pair per item in the dataset. Some people have used it for chat purposes, but as it's an instruction tuned model your results may vary.
You can find the prompt format and examples in the previous llama version's repo:
https://huggingface.co/jondurbin/airoboros-65b-gpt4-1.4