How can I re-fine tunning a GPT-3 fine tunned model?

API
1

Hello,

I did a fine tunning of GPT-3 using ‘ada’ model, with instruction:

openai.FineTune.create(training_file=“file-NZ6E***********”, model=‘ada’).

I got a model file name (ft_model) like “ada:ft**********” and I used it for inference with:

res = openai.Completion.create(model=ft_model, prompt="I got fired. ", max_tokens=21, temperature=1)

I whish re fine tunning this model with new data. But I want to use my trained model. Not ‘ada’ model from scratch. How can I do that?

Thank you in advance.

2

Welcome the community!

I don’t believe there’s a way to fine-tune an already fine-tuned model at this time.

1 Like
3

I could be wrong, but I think I read somewhere that you can continue to fine tune a model by using the fine tuned model name as the input “model” argument in the openai.FineTune.create call at the top of your question

You could give that try to see if it works. I don’t know if you need to use the “ada:ft***” or just “ft***” name when you do this. (So maybe try both)

You might need to go back and delete old models if you do this each time and it gives you a new name on each iteration

(I cant find the original post where I read this - but I think it is right)

Edit : here is the link OpenAI API

2 Likes
4

Thanks for the answers!!
I will try that!