Open wlsdzyzl opened 1 year ago
When I change the nn.Conv2d(channels, channels * 2, 5, 4, 2)
to nn.Conv2d(channels, channels * 2, 5, 2, 2)
(as long as the tranposed convolution layer), the results become corrrect. But I'm still confused about how this happened.
When I run the AAE tutorial, the output fake image is as following. For each image, only left top part is clear. Is there any problem related to this outcome?