Vision Fine Tuning - Great News but I'm still unable to upload images to the Fine tuned model?

API
1

I saw the news about fine tuning the GPT-4o model with images which was amasing.

I’ve fine tuned a model, but when trying to upload an image to the fine tuned model in the playground I’m met with “Error: this model does not support vision.”

So…we can fine tune the model with images, but still can’t upload images to the model when we go to use it? Am I missing the point here?

2 Likes
2

Hi @maxm1 and welcome to the Forum!

Is the fine-tuned model you are referring to a previously fine-tuned model or did you fine-tune a model with images, which then did not work properly at inference?

2 Likes
3

Hi @jr.2509 thank you for the reply. It was a previously fine tuned model, so maybe I need to fine tune again with images then try to upload an image?

2 Likes
4

Yes, I think so.

Previously it was not possible to use a fine-tuned gpt-4o with images.

2 Likes
5

thanks @jr.2509 , trying to retune my model but I’m getting the following error message. I’ve set detail:low in my images, and do not violate any policies (it’s photos of houses).

Seems like it’s going in circles, it took example 2 OK when ‘example 8’ was erroring, then when I fixed ‘example 8’ it went back to example 2 having an issue…

“The job failed due to an invalid training file. Images at example 2 violate our usage policy, are inaccessible, or are too large. Please remove them and try again.”