Train/test/dev splits with torchdata

hi there, I’m playing with torchdata which I find quite nice, but I’m wondering: is there some convenient receipe you may know for doing train/dev/test random splits in there ?

My datapipe starts with some dp.iter.FileLister class, and so maybe there’s a way to easily split that one in three in a reproductible manner for the splits ? A random split would do fine

thanks for your help