Some datasets are availble in .npy format (e.g. https://laion.ai/blog/laion-400-open-dataset/ )
This facilitates loading those datasets directly without having to first transform them into another format.
Thanks for the contribution! Looks okay to me. Don't mind the CI failures, they are unrelated. Give me a bit of time to fix CI then I'm happy to merge.
Some datasets are availble in .npy format (e.g. https://laion.ai/blog/laion-400-open-dataset/ ) This facilitates loading those datasets directly without having to first transform them into another format.