/ode-gan-tf

Primary LanguagePythonGNU General Public License v3.0GPL-3.0

Tensorflow 1.* implementation of Training Generative Adversarial Networks by Solving Ordinary Differential Equations.

Usage

Just drop ode_gan.py in your project directory, import ode_gan, and instantiate your choice of optimizer (RK2 [Heun's method] or the slower and more accurate RK4), passing the generator and discriminator parameters and losses.

Call the step() method in your training loop to execute a single training step.