Maybe worth adding the -q option to the unzips as in colab with the dataset you showed will run out of memory trying to display them. I just ran into that problem.