Why flip the third dimension y = torch.flip(x, (3,)) for the invertible toy problem?

[zuoshifan]

In Tutorial 1: Training iUNets in Pytorch, for the An invertible toy problem, when Training the model, the loss was computed by comparing the model output output = model(x) and y = torch.flip(x, (3,)) instead of x. Why the third dimension should be flipped? Thank you!

[tasptz]

To mirror the image:

In the following, we will train an iUNet to mirror the input image as a warm-up.