I am trying to implement GANimorph (https://arxiv.org/abs/1808.04325) on Pytorch. If possible, did everything as indicated in the article. However, I ran into a problem: losses in the learning process change very weakly, noise is generated all the time. What could be the problem? I didn’t train for long, maybe I need to wait? When implementing other GANs, the result was noticeable almost immediately.
Sorry for my bad English, I’m russian student.