There are no specific details available about fine-tuning GPT-4o Mini with image data in the https://platform.openai.com/docs/guides/fine-tuning, if it is a multimodal model that supports both text and vision tasks
Correct, at this time vision fine-tuning isn’t publicly available for any OpenAI models. Hopefully this will change soon!
Trying to finetune gpt-4o-mini with images in the training data currently gives the following error
The job failed due to an invalid training file. Invalid file format. Please remove all mentions of ‘image_url’ from your file and try again.
So, it seems using Vision is currently not supported. OpenAI should document this somewhere.
It seems like they’re working on this in general, but haven’t mentioned images yet:
https://openai.com/index/gpt-4o-fine-tuning/