I had finetuned various models in all the possible different ways, to get the desired response, below are the ways I did, but went unsuccessfully
- I wanted a JSON response, so prepared a dataset with a JSON response
- I also tried preparing a dataset with YAML and Markdown to get responses in YAML and Markdown respectively, which I will convert in my code side to JSON.
Doing this the fine0tunned model I got gave an irrelevant response
It was observed that even though I gave Dataset for JSON response strictly, I got responses in YAML, sometimes in Markdown, and sometimes in JSON.
- Is there any co-relation of trained models? As I am getting a fuzzy match of data from different models fine-tuned.
- I also tried deleting all the older models and creating a new model but did not get the response as desired.
Can anyone just help figure out why this is happening…