TorchGAN: Model Zoo

Collection of Generative Adversarial Networks developed using TorchGAN

Models Present

1. Generative MultiAdversarial Networks (GMAN)
   Link to Paper
   Requires the torchgan master.

2. Training GANs with Binary Neurons by End-to-end Backpropagation
   Link to Paper
   Requires the torchgan master.

Contribution Guidelines

We are open to accepting any model that you have built. The only things to keep in mind are the following:

1. Keep the models simple and reuse features of torchgan if possible.
2. Have enough command line options for users to play with.
3. Once you are done run isort and yapf (in this order only) for formatting the code properly.

FAQ

How to run the Model?

To run these models you need to have torchgan installed.

Then simply move into the directory.

$ python3 <model name.py> --help

This will show you the configurable options that are available.

Why do the models mostly use MNIST or CIFAR10?

The aim of this repository is to demonstrate the usage of torchgan. We believe this is best done if users can simply download the script and run it without having to download hundreds of GBs of dataset. However, we shall definitely add more models in the future which are specifically designed for high resolution data.