Way to share/keep good answer from GPT to train lora

[Stonedge]

Just an idea like that... I'm really interested at running an autogpt over local model (probably using koala and llamaccp python). Most of the local model hasn't be train to answer correctly prompt generated by autogpt. So finding a way to log/share those question+answers could be really useful to train Lora over local model. Since those use case would be specialized for the autogpt query that kind of fine tuning, local model would become a really viable solution.