I\'m working on a neural network taking 32x32 size images as input. I need to write a custom loss function and I have to create a mask from a size tensor img = (batch, 32, 3
