Fine-tune unstructured text data and create summary table: Prompt - Completion Format

I want to fine tune the plain text to get the summary as a table.
For example Chat GPT response for a give text is as follows:

Here what is the training set ({prompt:<>\n###\n, completion: <>}) look like?


1 Like

You can ask that ChatGPT :slightly_smiling_face:

Have you tested without fine-tuning on text-davinci-003? You might not need to fine-tune at all.

PS - welcome to the community! Hope you stick around.

Yes I have tested using text-davinci-003, with my custom text data. But it didn’t give the desired number of columns. So, I want to train on my custom text data. Please help me to get the jsonl format to finetune chatGPT (prompt: , completion: <col1:val1, col2:val2, col3:val3>)