Training loop for generator uses fake output from generator pre-update

tamarott / SinGAN

Official pytorch implementation of the paper: "SinGAN: Learning a Generative Model from a Single Natural Image"

https://tamarott.github.io/SinGAN.htm

Other

3.31k stars 611 forks source link

Training loop for generator uses fake output from generator pre-update #140

Closed AtheMathmo closed 3 years ago

AtheMathmo commented 3 years ago

https://github.com/tamarott/SinGAN/blob/1b4906b9afbd30bf32ea3d3f29ca6c5f4a2a223c/SinGAN/training.py#L175

I'd appreciate confirmation that this is indeed a bug. It looks to me like the generator output is not updated after the first step of the generator update loop.

AtheMathmo commented 3 years ago

I see now there is already some discussion of this issue in pull requests #126 and #117 . I'll close this issue for now, but feel free to reopen.