Error when computing MS-SSIM in images with size different than 144x144

[Warvito]

When computing the ms-ssim for images with different size than 144 x144, for example in:

from generative.metrics import MSSSIM
import torch

metric = MSSSIM(data_range=torch.tensor(1.0))

metric(torch.ones([3, 1, 256, 256]), torch.ones([3, 1, 256, 256]))

returns

RuntimeError: stack expects each tensor to be equal size, but got [3, 1] at entry 0 and [1, 1] at entry 4

[Warvito]

After investigating more, I found that the implementation from MONAI is wrong. Working to fix it.