GAN test (64x64)

My fourth GAN after the previous one. Trained on with PyTorch and manually exported to WebGL 2.0.

This is a larger model with regularization and hyperparameter tuning consideration. In addition, there is a feedback of generator results into VAE encoder.

==== Generator (271k params) ====
Latent - 32
Dense layer - 1024x32+1024
 => 1024, LeakyReLU(0.1)
 => 64x4x4
Conv2d - 64x64x3x3
 => 64x4x4, BatchNorm2d, LeakyReLU(0.1)
ConvTranspose2d - 64x64x4x4
 => 64x8x8, BatchNorm2d, LeakyReLU(0.1)
ConvTranspose2d - 64x64x4x4
 => 64x16x16, BatchNorm2d, LeakyReLU(0.1)
ConvTranspose2d - 64x64x4x4
 => 64x32x32, BatchNorm2d, LeakyReLU(0.1)
ConvTranspose2d - 64x3x4x4
 => 3x64x64, BatchNorm2d, Tanh*0.5+0.5