【BUG？】About openai finetuning...

[Fay-why]

Thank you for your work. I have some questions regarding OpenAI's fine-tuning process and the code involved:

1. Validation Data Impact: During the fine-tuning process, does the validation data affect the model's gradient updates, or is it solely used for monitoring/metrics (e.g., validation loss)?

2. Test Data Clarification: In the provided figure, the test.jsonl file is passed to OpenAI as validation data. Could you please confirm whether this file represents the final test set used for evaluation? Specifically, I want to ensure that the test data is not inadvertently introduced into the training process, and whether this is an additional test file just for training.

I would greatly appreciate your clarification on these points.

Best regards,
Fay.