Can't access fine tuned model

i can’t access my fine tuned model, i do fine tune through the UI using model gpt-3.5-turbo-0125. The playground button below fine tune doesn’t direct to my fine tuned model.
this is my API response:
{‘error’: {‘message’: ‘Project proj_IR9shwKpjWSru345cjWAuQ3m does not have access to model ft:gpt-3.5-turbo-0125:gegeh:pertama:9Qrmxlk1’, ‘type’: ‘invalid_request_error’, ‘param’: None, ‘code’: ‘model_not_found’}}

i do fine tuned without suffix before and the model show up but this time i can’t access my model

image

image
this playground button can’t find my model instead choose the model to gpt4

As per the stats I wouldn’t recommend using that model it’s happening for good probably these stats are terrible the loss closer to 0 is good


i fine tuned another model but still can’t access it, your argument doesn’t seems valid because i’ve done it before and the model shows up.

The playground is full of bugs since the recode released last week.

Try this python script to see that you can list your fine-tuning models via API:

Then you can continue with an example chatbot, changing the model in code to your exact fine-tuning model name (and in that script at the bottom of the post, optionally send an image).

I want to add that I was just trying to reproduce your error on my end. However, using also gpt-3.5-0125 as a baseline model as well as including a suffix, I was able to access the fine-tuned model both via the playground and via Python script.

Did you create the fine-tuned model under a different project perhaps?

i only have 2 project which is “Default Project” and 1 other project, i tried to access it through both project and still can’t find the model in the playground. im accessing it as “owner” which have access to all project but still can’t find it

i think i have to contact support

I think that’s the right way forward. I also can’t think of another plausible explanation as to what might be causing this.

I assume though you have tried j’s suggestion of getting the list of all your fine-tuned models via API?!

I am also finetuning through the API. Finetuned gpt-4-0613

for model in client.models.list():
    if model.id.startswith('ft:gpt-4'):
        print(model)

returns Model(id=‘ft:gpt-4-0613::::ckpt-step-360’, created=, object=‘model’, owned_by='')

and when I immediately try to prompt the model I get:
Error code: 404 - {‘error’: {‘message’: ‘The model ft:gpt-4-0613:***::***:ckpt-step-360 does not exist or you do not have access to it.’,

still can’t find it, been trying to use multiple API accross projects but still can’t find my models. live chat can’t help at all

Why can’t support help? Have you been able to submit a message detailing the problem? It can take some time to get a response.

it might be best to reach out for further assistance. Since you’re experiencing this issue with the OpenAI API and fine-tuned models, I recommend submitting a detailed bug report through the OpenAI support form available in the Help Center.

idk how to submit a report because it hard to find email or anything to contact human. and yes, thats answer is provided by human customer support

Ok, here’s some guidance for you regarding the support.

Go to the Help Center page. On the bottom right, access the chat widget. Select the option Messages > Ask a question.

Then replicate the interaction as shown on the following three screenshots. This will enable you to submit a detailed message of the issue you are facing along with screenshots. Support staff will then pick it up. I have no estimation how long it will take for them to respond but at least the matter will be reviewed by someone.

image

image

image

Good luck! I hope it gets resolved.

1 Like

Has anyone else solved this? I’m facing this issue, today. I fine-tuned a model. The fine tuning job completed, but the fine-tuned model doesn’t show up in the /model endpoint. I can’t retrieve it, and I can’t delete it. Yet, I know it exists, because I can create assistants in the playground for it…

https://community.openai.com/t/finetuned-models-missing-from-api/956582

Looks like many of us have the same issue. The above link has some details regarding this latest issue.

1 Like

Thanks for the link @fluxgill ! I’ll keep an eye on it for a solution.

1 Like

I had this issue because I had set up “limits” for which models a project could use, here:
https://platform.openai.com/settings/organization/limits

Finetuned models do show up here after they are created but they are not enabled by default (or were not for me at least), so I had to check the checkbox next to them

1 Like