Update fine-tuning API

[branislav-burdiliak]

Update fine-tuning API

Affected endpoints

1. Create fine-tuning - changed

   • supports 2 new parameters (settings)
   • integrations
   • currently supported only Weight and Biases integrations ("type": "wandb")
   • default - no integrations
   • seed - controls reproducibility of the job, in rare cases may return different results for the same seed
   • default - no seed will be passed to OpenAI which will in turn generate one for you The response has been enhanced with the same integrations object.

2. List fine-tuning checkpoints - added

   • OpenAI now produces a full model checkpoint at the end of every training epoch
   • these are full models that can be used with completions and chat-completions
   • useful to access a model before overfitting happened
   • for more refer to: https://platform.openai.com/docs/guides/fine-tuning/use-a-checkpointed-model