search

christiancosgrove / pytorch-spectral-normalization-gan

Paper by Miyato et al. https://openreview.net/forum?id=B1QRgziT-
MIT License
676 stars 110 forks source link

spectral_normalization_nondiff.py #16

Closed hellopipu closed 5 years ago

hellopipu commented 5 years ago

Thanks for your code . I'm quite confused with the nondiff version, what does it for , and what's the difference with the diff version. Thanks

christiancosgrove commented 5 years ago

nondiff stands for nondifferentiable. Rather than creating a new tensor to represent the normalized weights (which is what the other version does), the nondiff version simply replaces w with a normalized weight matrix.

In my experiments (long time ago) I found there wasn’t much of a difference between these two implementations.

On May 26, 2019, at 8:48 PM, Bingyu Xin notifications@github.com wrote:

Thanks for your code . I'm quite confused with the nondiff version, what does it for , and what's the difference with the diff version. Thanks

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread.