How to fine-tune for second time, alredy fine-tuned model (by me) without data loss?

API
1

Hello guys,

I fine-tuned the babbage OpenAI model with my custom data (100K prompts). I achieved almost satisfy results but I want more, how to fine-tune the same model for second time with more 100K prompts without data loss and overtraining?

Thanks in advance!

1 Like
2

There currently is no continuing fine-tune on the new endpoint for the just-released models. “coming soon”.

1 Like
3

It’s currently not available per se as a feature. Why not fine-tune it from basic with 200k prompts. Personally, I’ve never been too satisfied with the results of fine-tuning an already fine-tuned model.

1 Like
4

because 100k is the maximum (50000000 tokens)