I follow the tutorial and install by running
pip install git+https://github.com/lukemelas/pytorch-pretrained-gans

However, I receive the following error:

from pytorch_pretrained_gans import make_gan
Traceback (most recent call last):
File "", line 1, in
File "/home/zeyuy/miniconda3/lib/python3.8/site-packages/pytorch_pretrained_gans/init.py", line 5, in
from .StudioGAN import make_studiogan
File "/home/zeyuy/miniconda3/lib/python3.8/site-packages/pytorch_pretrained_gans/StudioGAN/init.py", line 9, in
from .models import resnet
ModuleNotFoundError: No module named 'pytorch_pretrained_gans.StudioGAN.models'

Hi，can you provided the URL of pre-trained model of BigBIGAN-256 and BigBIGAN-512. By default, we can only use the bigbigan-128.

Hi,

Thank you for making this repository. I tried setting it up as instructed with the following code, but the code fails to import make_gan

pip install git+https://github.com/lukemelas/pytorch-pretrained-gans

import torch
from pytorch_pretrained_gans import make_gan

Sample a class-conditional image from BigGAN with default resolution 256

G = make_gan(gan_type='biggan') # -> nn.Module
y = G.sample_class(batch_size=1) # -> torch.Size([1, 1000])
z = G.sample_latent(batch_size=1) # -> torch.Size([1, 128])
x = G(z=z, y=y) # -> torch.Size([1, 3, 256, 256])

Hi,
Thank you for the repo, it is very useful!

For my current work, I would like to get also the pre-trained Discriminators from the GAN models that I load.
Is it possible for you to share them as well?

Thanks

Hi,

Thanks very much for compiling all of these models. I am curious if the encoder for BigBiGAN would be difficult to implement in this repository? Its quite uncommon for a GAN to also train an encoder that maps real images into the latent space. So it would be useful to have a pytorch implementation if it.

Best,
Cory

Hi,
According to the architecture of BigBiGAN, there are joint, and unary terms in the discriminator. Can you help me in finding that in your BigBiGAN model arch?