Fine-tuning with Structure output

[pablospe]

It is unclear to me how to create a JSONL file with user/assistant pairs (for fine-tuning a model) if one uses Instructor to get the structured data. What would it be the real answer of the system? Is raw_response from the following link?

Do you need to include response_model schema in user messages in your finetuning jsonl? Or should it just be the system/user/assistant unstructured messages? Or alternatively, should I try to access the prompt given to the LLM ([as discussed here](#350)) and raw_response as user/assistant pairs?

I was also considering this approach: #580 (reply in thread)

call1: prompt -> LLM  -> text_response  
call2: prompt (which is only the text_response)  -> LLM (with response_model) -> response_model_Object

but it is unclear to me if this is the correct approach for creating a jsonl dataset for fine-tuning.

Thanks!

[pablospe]

Hi. Basically, I was wondering if anyone had tried fine-tuning with the parsed results from Instructor? If so, could you please share how it is done?