Energy-Based Generative Adversarial Networks in Tensorflow
Below show some non-cherry picked results after ~40,000 iterations.
The losses for G and D were not as informative as I'd hope they would be.
Loss for the discriminator
Loss for the generator