How to train the model on just 500 Questions or less to run it on a CPU?

I want to reduce the size of data enormously to just 500 Question from something around 200 passages or less. I do not care about the accuracy or F1 score, I just want to replicate this project in python and see neural net work.

Initially, I thought to edit read_data.py files but there are a number of them and I quite don''t understand where and what to edit.