No feature for early stopping in finetuning

I am disappointed to find that there is no feature for “early stopping” in the fine tuning process provided by OpenAI’s API. As a user, I find it crucial to be able to stop the fine tuning process in case of overfitting, as it can lead to poor performance of the model on unseen data. Furthermore, the lack of visibility of validation results after every epoch makes it difficult to monitor the progress of the fine tuning process and make informed decisions about when to stop training. I believe that the addition of these features would greatly enhance the functionality of the API and the user experience. It would be greatly appreciated if the OpenAI team could consider adding them in a future update.

5 Likes