I did that but I m not sure where I m

[broken crystal]

#prompt-engineering message
@crimson atlas I'm already using the API, the issue that happened is while using the API

[crimson atlas]

@broken crystal try adding some debug within your file and see if the API captures that new info.

[broken crystal]

ok, I'll try that, although I'm not sure how to debug that

what steps did you do in fine-tuning the model?
@lone umbra

I just followed the tutorial, I uploaded my .jsonl file into the gpt-3, it was uploaded successfully and returned an ID, then I used playground to test it, the model is already identified there, I also tested on CLI and the returns are also wrong

[lone umbra]

Just to confirm, is uploading the .jsonl file all you did to fine tune or was there more additional work?

To be more specific, I know that that didn't actually upload the file 🙂

One thing that might help is removing the END from the completions. Just put in a fullstop.

[lone umbra]

I don't know if this helps

[broken crystal]

before uploading I used CLI data preparation command, that basically just verify and adjust the format of the file, after that I used the the upload/create command, like that:

openai api fine_tunes.create -t my-training-test_prepared.jsonl -m davinci --suffix "EthHackathon event"

[broken crystal]

the END was added by the CLI data preparation command, but I'll test without it

[broken crystal]

I tried removing the END but it didn't work, I also tried in playground passing like you did and it worked in playground, I just don't know how to make it works in a personal model created with fine-tuning like I'm trying to do