Fine-tuned model error—"openai: error: unrecognized arguments:"

Hi everyone,

I was wondering if anyone would be kind enough help me with an issue I’m having?

I have recently fine-tuned a GPT-3 model, but every time I enter (openai api completions.create -m <FINE_TUNED_MODEL> -p <YOUR_PROMPT>), the API returns a message (openai: error: unrecognized arguments:).

I have written the prompts in the exact same format/style as the fine-tuning data, but I have only added ~80 prompt/completion examples, so perhaps this just isn’t enough data for the API to understand the prompts.

I’m not the most clued up programmer, so there’s also a chance I’m doing something wrong that’s totally obvious!

Thanks in advance for any help,
Sam

Did you wrote the prompt in quotes?

Yes I wrote all of the prompts and completions in quotes.

Exactly as suggested in this format: {“prompt”: “”, “completion”: “”}.

And then into the OpenAI CLI in this format: openai api completions.create -m <FINE_TUNED_MODEL> -p <YOUR_PROMPT>

{“prompt”: “”, “completion”: “”} - this is JSON format and should not be used in the CLI.

for the CLI it would just look like:

openai api completions.create -m DAVINCI -p "my prompt here"

1 Like

Sorry, I made a mistake in my previous message.

I created a prompt/completion dataset in .csv format and converted it to JSONL using the CLI.

I then used: (openai api completions.create -m <FINE_TUNED_MODEL> -p <YOUR_PROMPT>), in the CLI, but it still seems to be returning: (openai: error: unrecognized arguments:)

:thinking:

It would be helpful if you posted exactly what you’re seeing. Copy/paste or a screenshot.

I think you need

openai.Completion.create

Could be wrong, I haven’t used the CLI or MacBook installation

Update: never mind I see the documentation you’re following: OpenAI API

I’m stumped. Maybe try single quote instead of double quote?

I’ll give that a try, but I think I must have done something wrong earlier in the process!

Thank you for your time anyway, it’s appreciated :slightly_smiling_face:

Were you able to find a solution. I am currently struggling with the exact same thing