Unable to Use Fine-Tuned Model – Access Issue Despite Correct Setup

API
, , ,
1

Hello everyone,

I’m facing an issue with fine-tuning, and I can’t find a solution anywhere. I’ve seen that other users have encountered the same problem before, but I haven’t found a clear resolution in any case.

Some suggest checking if the model was created within the same project you’re trying to use, while others recommend verifying if it is properly enabled. I have confirmed both, but the issue persists.

Issue Description:

I created a fine-tuned model, uploaded a JSONL file, and the entire process was completed successfully via the API. After receiving the confirmation that training was completed, I attempted to use the model by making a request. However, I encountered the following error:

Project `proj_kmr5Wmb6s4BMDz2MDHm9WlMo` does not have access to model `ft:gpt-3.5-t (truncated...)"```

I have double-checked everything—API keys, project settings, and even attempted to generate a response directly through the OpenAI Dashboard—but no matter which model I train, I always receive this error. You can see the attached image for reference.

Has anyone experienced this before and found a solution? Any help would be greatly appreciated! (Apologies if this
![The image shows a screenshot of the OpenAI Playground with a notification about an unavailable model, defaulting to a standard model instead. (Captioned by AI)|690x407](upload://u5MvkjX6Aye9OuEaH1vjpU1gimZ.png)
 is a basic question, haha.)

Thanks in advance! 🚀
![The image shows a pop-up notification on a user interface indicating that a specified AI model is not available and the default model is being used instead. (Captioned by AI)|690x149](upload://8bvSnfHH6JcfnhbWVsF70lBZrMQ.png)
2

I have noticed that my organization owner has to go manually add them individually into the limits tab under settings for the project it is under before I can use it in the playground or API. I don’t know why they do that as they allow me to create the model but not use it. They for some reason are not automatically allowed when there are not restrictions set.

It would be great if OpenAI looked into that policy and made it easier for developers in an org to build and test without bothering the admins in their org.