Closed Molugan closed 5 years ago
TODO:
1) refactor: move the BASEGAN.optimizeParameters function into the GAN trainer 2) make the hubconf file
There's an typo here: https://github.com/fairinternal/pytorch_GAN_zoo/blob/master/hubconf.py#L75 (additional :
in the end).
It would be great if we could have pretrained weights available somewhere.
Besides that, it works great with torch.hub
. Thanks!
In [1]: import torch.hub
In [2]: torch.hub.load('fairinternal/pytorch_GAN_zoo', 'DCGAN', useGPU=True, pretrained=False)
Out[2]: <models.DCGAN.DCGAN at 0x7fee97740f60>
The weights will be added soon, I just need to do the procedure for s3.
Arrrg the typo ><. I'm on it !
FYI checkpoint for PGAN
works great!
In [6]: torch.hub.load('fairinternal/pytorch_GAN_zoo', 'PGAN', useGPU=True, pretrained=True)
Using cache found in /private/home/ailzhang/.torch/hub/pytorch_GAN_zoo_masterDownloading: "https://dl.fbaipublicfiles.com/gan_zoo/celebaHQ_s5_i96000-5fe58222.pth" to /private/home/ailzhang/.torch/models/celebaHQ_s5_i96000-5fe58222.pth
275319082it [00:04, 62441877.81it/s]
Average network found !
Out[6]: <models.progressive_gan.ProgressiveGAN at 0x7f496f721eb8>
Refactor the code in order to make the training script compatible with torchub : https://github.com/pytorch/vision/blob/master/hubconf.py