Issue with fine-tuning model name

API Bugs
,
1

We fine-tuned GPT models(gpt-4o-2024-08-06, gpt-3.5-turbo-0125) on a simple dataset from the documentation. The training was successful(ft:gpt-3.5-turbo-0125:trinity-audio:test3:*****) but we can’t use this model because of this error This model requires that either input content or output modality contain audio.

It appears that the problem was related to our organization name, “TRINITY-AUDIO.” When we created an organization name without “AUDIO” in it, the issue was resolved.

2 Likes
2

Seems like a dumb API. “If ‘audio’ in model_name: deny”.

Or have amusement employing gpt-3.5-turbo with modalities=["text", "audio"]

Ask “help” for a credit refund for model training. See if that or the API fix comes first.

2 Likes