It is almost impossible to tune gpt-4 models.
The job failed due to an invalid training file. This training file was blocked by our moderation system because it contains too many examples that violate OpenAI’s usage policies, or because it attempts to create model outputs that violate OpenAI’s usage policies.
But nowhere do you tell me what the reason is. Which line or anything else.
I am currently solving the problem by copying each of my entries 10 times and putting them into a single jsonl & uploading them to see if they are approved or rejected, but that can’t be the solution.
Especially which things are flagged is a mystery to me, in the normal moderation API they have values of 0.000064