After September 24, 204, fine-tuning the model prompt: the model does not exist or you do not have access to it

API
1

The models fine-tuned on September 23rd can be used normally. Today, the model fine-tuned on September 24th prompted that the model does not exist or has no permission. Confirm that my account has not been updated in the recent period. And today’s fine-tuning fee is displayed as particularly high. Why is this? I am using chatgpt4 mini 2024-07-28

2

Welcome to the Community!

Can you share details on this? That is, how many tokens did you use for training (consider both the number of the tokens in your training set and the number of epochs - see example calculation here). Were you only charged the overage exceeding the 2M free tokens or were you charged for the full amount of training? OpenAI has seemingly extended the free training period up until the end of October. So technically, you should still only see charges for gpt-4o-mini (a) if the number of trained token exceeds 2M and (b) when you consume the fine-tuned model.

Are you using the proper model name? Are you using an API key associated with the project under which you created the model?

3

I’m having the same issue. I fine tuned the model and an hour later I checked the status it said “succeeded”. But still when I use it for my use case I’m unable to, it says:

Error code: 404 - {‘error’: {‘message’: ‘The model finetuned_model_name does not exist or you do not have access to it.’, ‘type’: ‘invalid_request_error’, ‘param’: None, ‘code’: ‘model_not_found’}} .

4

Thanks for flagging. I will try to reproduce this on my end now and see if I face the same issue.

EDIT: I have no trouble accessing the gpt-4o-mini fine-tune I just created.

As per my earlier message, I would again encourage you to (a) double check that you are using the correct and full model name when calling the model as well as (b) ensure that you an API key associated with the project under which you created the fine-tune.

Please also check if you can access the model in the playground.

If the problem still persists, feel free to let us know here.

5

I ran into this issue with a model I fine-tuned today.
However, the GPT4o-mini model that was fine-tuned on 21 September is working fine.