Pytorch WGAN-GP

This is a pytorch implementation of Improved Training of Wasserstein GANs. Most of the code was inspired by this repository by EmilienDupont.

Training

To train on the MNIST dataset, run

python main.py --dataset mnist --epochs 200

For the FashionMNIST dataset, run

python main.py --dataset fashion --epochs 200

You cans also set up a generator and discriminator pair and use the WGANGP class:

wgan = WGANGP(generator, discriminator,
              g_optimizer, d_optimizer,
              latent_shape, dataset_name)
wgan.train(data_loader, n_epochs)

The argument latent_shape is the shape whatever the generator's forward function accepts as input.

The training process is monitored by tensorboardX.

Results

Here is the training history for both datasets:

Two gifs of the training process:

Interpolation in latent space

We can generate samples going smoothly from one class to another by interpolating points on the latent space (done in this notebook):

The weights of the models are on the saved_models folder.