How to update and replace data in the fine-tuning of the latest GPT-3.5-turbo model?

Let me explain my problem.

I uploaded a JSONL file for fine-tuning data using this API: https://api.openai.com/v1/files. After uploading the file, I created the fine-tuning job using the API: https://api.openai.com/v1/fine_tuning/jobs. This process successfully created the fine-tuning model.

First Requirement:
I have more training data to further train this model, but I am uncertain about the exact procedure for doing so. Additionally, I would like to know if I need to pay for the previous data in this process.

Second Requirement:
In this scenario, I would like to replace all the old training data with the new training data for this model and then proceed with the training. What is the correct process for achieving this?

1 Like

About the first requirement:

https://platform.openai.com/docs/guides/fine-tuning/can-i-continue-fine-tuning-a-model-that-has-already-been-fine-tuned

No, we do not currently support continuing the fine-tuning process once a job has finished. We plan to support this in the near future.

2 Likes

1: Continuing is currently unsupported. You have to run a new fine-tune.
2: Run a new fine tune on the new data, get a new model name.

2 Likes

Is there any other way to do this below requirement?

because it’s west of money and time to create a new model every time. As per our requirement we continue the train the existing fine-tuning model over time to get a better result.

First Requirement:
I have more training data to further train this model, but I am uncertain about the exact procedure for doing so. Additionally, I would like to know if I need to pay for the previous data in this process.

Other way?

  • use legacy base completion engine that will be going away at end of year
  • wait for more functionality of new endpoint
1 Like

I found the solution about the Update the fine tune model you should pass the your previous model to update the fine tune model but this create the new model not update in existing model.

1 Like

yes now it supports,