Hi,
I am trying to fine-tune gpt-4o mini with a custom dataset. After the model is fine-tune I get this error: Error while running moderation eval refusals_v3 for snapshot ####: Error while running eval for category hate for snapshot ###. Please try again.
Does anyone know why this happens?
Even I have been facing similar refusals_v3 errors in last few days for different model trainings.
Few other category provided in this message were illicit, sexual etc.,
I would really appreciate any help in this regard.
I have the same issue (didn’t have it last week). I am fine-tuning gpt-4.1-nano on data that is not linked to this topics.
Is there a workaround to this?
Same issues. And the monitoring takes a lot of time
Hi, and thanks for flagging this!
OpenAI has rolled out an update to prevent this from happening again.
You may need to wait a day or two before retrying a previously rejected job, but new jobs should work fine.
If the issue persists, please let us know here.
Thanks @vb for acknowledging this. Can you please share any release note or change log reference that talks about this issue and the fix that the team have worked on.
Sorry, I have no further information to share besides what has already been said.
OpenAI has rolled out an update to prevent this from happening again.
You may need to wait a day or two before retrying a previously rejected job, but new jobs should work fine.
If the issue persists, please let us know here.
We are still seeing this issue with new finetuning models. When the job finishes, it hangs for a long time, and then errors out with Error while running moderation eval refusals_v3
Hi Nick, are you able to provide send a note with the ft-job id to support@openai.com? Let me know on this thread once you do, we can look into this more on our end.
