/EvoGAN

EvoGAN: Evolutionary Algorithm based Neural Architecture Search for Generative Adversarial Networks

Primary LanguagePythonMIT LicenseMIT

EvoGAN: Evolutionary Algorithm based Neural Architecture Search for Generative Adversarial Networks

Installation

The conda package manager is required to install the dependencies for this project.

To get the environment setup, you can run the following command:

source setup.sh

The environment is stable on Ubuntu 18.04, with the Amazon DeepLearning AMI in particular.

Acknowledgements

  1. Bootstraps AutoGAN for training and is heavily inspired by AutoGAN
  2. Inception Score code from OpenAI's Improved GAN (official).
  3. FID code and CIFAR-10 statistics file from https://github.com/bioinf-jku/TTUR (official).