replace buggy `max_norm=1` arg with norm for embedding

akanimax / pro_gan_pytorch

Unofficial PyTorch implementation of the paper titled "Progressive growing of GANs for improved Quality, Stability, and Variation"

MIT License

537 stars 99 forks source link

replace buggy `max_norm=1` arg with norm for embedding #37

Closed arvindmvepa closed 4 years ago

arvindmvepa commented 4 years ago

resolves #21 basically I am replacing the buggy max_norm=1 arg in torch.nn.Embedding with code to manually calculate the norm. the functionality is slightly different, though, as stated by @akanimax, the original paper never mentioned normalizing the embedding vectors to begin with.

akanimax commented 4 years ago

closing due to inactivity. Please feel free to reopen in the light of new evidence :smile:!