A styleGAN based on NVIDIA's paper.
- Commander object overview.
- Dataset helper functions.
- Saving model snapshots.
- Logger for metrics.
- Normalize images to [0,255] && decode functions, noise to [-1,1].
- Training loops for Gen and Disc.
- Upsample class.
- Adaptive instance Norm class.
- Pixel Norm class.
- Latent FC linear layers.
- Generator Net.
- Discriminator Net.
- Train the model.