BLACK images generated after changing networks. Is that because of the instability of GAN?

[Oranging1]

When I try to change the network，reducing the number of resnet layers. The results become a BLACK image, which means all the output is nan, Is that because of the instability of GAN? How can I fix it? @menyifang

[BadourAlBahar]

I have the same issue. I, however, am not changing the model. I am using pytorch 1.2.0.