lebeli opened this issue 2 years ago · 1 comments
Whenever I try to provide my own randomly sampled latent codes as source latent, I get images like this:
The sampling was done with: torch.randn(1, 512, device=device)
Do you have any idea why the image generation fails?
Kind regards!
I found the mistake: the statement above generates the z-code for the mapping network. I had to set "input_is_latent=false" otherwise it would not be fed to the mapper. The issue can be closed!