#fine tuning text-davinci-003

8 messages · Page 1 of 1 (latest)

drowsy panther
#

The API reference makes it difficult to be certain, can I fine-tune the text-davinci-003 model and is it the same pricing as working with the base davinci model?

fickle lion
drowsy panther
#

Yeah that's the thing it seemed a little ambiguous to me because it said that only the base models could be used. Then it said there's the other models that have had additional training like text-davinci-003 and then you can actually fine tune further something that's already been fine tuned as if text - davinci-003 might actually be fine tuned and you can further fine tune it...

but maybe it only meant you can only further fine tune what you have already fine-tuned yourself, still based only on The DaVinci base model, which I'm guessing is the case but I just wanted to make sure.

fickle lion
drowsy panther
#

ah nice, thanks for that, I had called models already to populate a list but overlooked the allow_fine_tuning. Now I just have to get this to work, it seems to hate my json and jsonl and the validation tool only seems to validate it will hate my json. I have queried chatgpt to all ends on this and it just doesnt seem to understand how to create and update fine-tuning on itself enough to connect the dots unfortunately.

But this is helpful though and definitely answered my question, I may be back with a new question about the tuning if tomorrow if I dont have any luck in the morning.

Thanks much

drowsy panther
#

Actually looks like it doesnt really let it happen. I think I would have to have my org own text-davinvi-003 to use it apparently.

"[organization=user-xxxxxxxxx] Error: Invalid base model: text-davinci-003 (model must be one of ada, babbage, curie, davinci) or a fine-tuned model created by your organization: org-Yxxxxxxxx (HTTP status code: 400)
"

zealous estuary
drowsy panther
#

sad day.