Training GAN is hard. Models may never converge and mode collapses are common. To move forward, we can make incremental improvements or…