PyTorch's torch.nn.functional.interpolate behaving strangely

Question

I\'m having issues with PyTorch\'s tensor-resizing options. In the following code, "x" is a dataset of 888 64x64 RGB images of pokemon. "xs" is to be a d