Some questions about replicating results from the paper

HI, I just downloaded clothing 100k dataset from your link here "Download and untar the file here for all 5 datasets: https://drive.google.com/file/d/1rr2nvnnBMsbo1qcU3i3urJsDw86PJ9tR/view?usp=sharing". But, i found that there's a make_dataset.py in clothing 100k, so i wonder if extra noise has been added to these data, or it's only used to divided 100k from 1M?