#How to do fine-tuning on text-davinci-003?

8 messages · Page 1 of 1 (latest)

stiff hamlet
#

I'm trying to do some fine-tuning, but the cli only allows to use the davinci model, which doesn't give the same results as I get in the Playground using the more advanced text-davinci-003.

Is this model currently available for fine-tuning?

Thanks!

crystal pebble
# 
{
  "id": "text-davinci-003",
  "object": "model",
  "created": 1669599635,
  "owned_by": "openai-internal",
  "permission": [
    {
      "id": "modelperm-IdOatecWAhWDJ8kxOlsqPnJk",
      "object": "model_permission",
      "created": 1669843334,
      "allow_create_engine": false,
      "allow_sampling": true,
      "allow_logprobs": true,
      "allow_search_indices": false,
      "allow_view": true,
      "allow_fine_tuning": false,
      "organization": "*",
      "group": null,
      "is_blocking": false
    }
  ],
  "root": "text-davinci-003",
  "parent": null
}
stiff hamlet
#

Ah thank you, that's very unlucky!!!
I'm trying to train GPT to do some code refactoring of a specific JS library, but I can't get the result I'm looking for, it spits out other random stuff after applying some of the refactor I want to the original code

crystal pebble
#

I'd say use Text DaVinci 1 and fine tune that model, since it allows it.

stiff hamlet
#

What would be a big enough dataset if I have roughly 6 lines to replace with?

crystal pebble
stiff hamlet
#

aaah I got 4 ahahahah that might be why it's not working ahahah