check out https://github.com/tatsu-lab/stanford_alpaca

i didn't want to have to convert it to use the full prompts every single time when training a causal language model so this is easier