Skip to content
Code for reproducing experiments in "Improved Training of Wasserstein GANs"
Branch: master
Clone or download
Pull request Compare This branch is 1 commit ahead of igul222:master.
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.

Improved Training of Wasserstein GANs

Code for reproducing experiments in "Improved Training of Wasserstein GANs".


  • Python, NumPy, TensorFlow, SciPy, Matplotlib
  • A recent NVIDIA GPU


Configuration for all models is specified in a list of constants at the top of the file. Two models should work "out of the box":

  • python Toy datasets (8 Gaussians, 25 Gaussians, Swiss Roll).
  • python MNIST

For the other models, edit the file to specify the path to the dataset in DATA_DIR before running. Each model's dataset is publicly available; the download URL is in the file.

  • python 64x64 architectures (this code trains on ImageNet instead of LSUN bedrooms in the paper)
  • python Character-level language model
  • python CIFAR-10
You can’t perform that action at this time.