Fine tuning checkpoint endpoint does not return validation metrics

API
1

Hi

Previously I have used the fine-tuning checkpoint endpoint to return train_loss, train_mean_token_accuracy, valid_loss and valid_mean_token_accuracy for a 4 epoch model.

I just used it (same function) and only received train_loss. I couldn’t find a parameter that ensures these metrics are returned. The dashboard online has valid_loss and full_valid_loss listed.

Has anyone addressed this problem or have openai changed the return and not updated their API documentation?

2

Hi there,

We have same problems,

We use: https://api.openai.com/v1/fine_tuning/jobs/{fine_tuning_job_id}/events

U can find more at: https://platform.openai.com/docs/api-reference/fine-tuning/list-events