search

meta-llama / codellama

Inference code for CodeLlama models
Other
15.4k stars 1.78k forks source link

Does codellama 13B/34B/70B support function calling and Lora fine tuning on multi-run chat with function calling? #201

Closed zlfeng83 closed 4 months ago

zlfeng83 commented 5 months ago

Does codellama 13B/34B/70B support function calling and Lora fine tuning on multi-run chat with function calling? Are there any instruction pages about this? Thanks a lot.

jgehring commented 5 months ago

Hi @zlfeng83, Code Llama models have not be specifically trained or evaluated for function calling. However, I would think that it might be worth experimenting with it and maybe try to fine-tune it for these use-cases.

This repository does not support LoRA (or any fine-tuning for that matter), but there are several open-source projects offering this functionality.

Regarding multi-run chat: all instruction models support multi-turn prompts (i.e., user -> assistant -> user -> assistant -> ... -> user) out of the box, while the other models can either be few-shot prompted or fine-tuned to support multi-turn chat given sufficient example data.