In addition to saving the processed image arrays to disk, shouldn't we also save their classes/labels?

Maybe I’m missing something.

But if the point of saving these arrays to disk is so we don’t have to use batches, shouldn’t we also be saving their classes/labels to disk as well since they will be required as well in training our models?

It seems that while the example notebooks recommend and do persist the image arrays to disk … they still rely on batches to get the class/label information.