How to determine the data format when creating a custom dataset for a given task?

Trying to fine tune a DialoGPT model on the cornell dataset.
It needs to be preprocessed.

Where do I find the format that the model expects the data to be in? For this, and for other models?