Similar to issue #43, I trained the VQ-VAE for about 37 epochs with a dataset of about 30k images. But the result is very blurry.  Do I need more training or I need to train PixelSnail in order to get a higher quality result?
