I\'m working with a dataset of size over 30 Gb(30k+ images), where images aren\'t divided into train and validation folders. Here\'s my implementation:
train_