Simple questions regarding library

Hello,

I have two very simple questions regarding this library:

For the annotations in the dataset, are they supposed to be in PASCAL XML format or YOLO(some_version) format?
When training is completed, I see that 3 files are generated. Two .pt model files and a weight file, .json. When trying to implement object detection from live camera, I want to use this generated model in a different code. Does it matter which .pt model I use? what about the .json file?

Thanks in advance.

OlafenwaMoses / ImageAI