Custom Image dataset working only for validation data

vision
2

For training you are using a transform ‘Cutout’ from torchtoolbox library which I believe does not have ‘always_apply’, why don’t you use ‘Cutout’ from albumentations itself?