How to use training output?

Hi, it's not a bug, but question.
I've run train.py script and received bunch of data/index/meta files. How can I use them? For example, instead of using yolo_net. Can they somehow increase recognition speed?