Images are broken into 1GB and 6GB file chunks so that the download process does not get interrupted when downloading one large file. File chunks are created using:
"tar czpvf - ../../image_data | split -d -b 6000M - malnet-image"

To recombine file chunks after downloading, run:
"cat malnet-image* | tar xzpvf -"

Once the files have been recombined, the data can be uncompressed using:
"tar -xvzf /path/to/file.tar.gz"

The full dataset can be easily loaded using the code in the Github repository. We note that in order to provide classification across "type", "family" and "binary" class labels,
we include code that symlinks to the original data so 3 copies are not needed, which would take up lots of extra space.

MalNet-Tiny contains 61201 training images, 8742 validation images and 17486 test images in a 70-10-20 ratio stratified across 43 types of malware.
The data is stored in 3 folders "train" "val" and "test".