WGAN_practices

This is code of WGAN-GP using celebA and CIFAR10 datasets.

Explanation of file

sang_gan.py    : includes model of generator & discriminator and gradient_penalty
sang_main.py   : main file of the code 
sang_plot.py   : functions of plotting generator & discriminator loss, generate images, animation of generate images  
sang_utils.py  : dataloader of CelebA & Cifar10
wrapper.py     : training with parameters

Requirments

1. Python(3.6.9)
2. pytorch (1.1.0)
3. matplotlib(3.1.0)
4. numpy (1.16.4)

Usage

In 'sang_main.py', Some hyperparameter which can changed by users included in parser.

1. 'dataset' : select the dataset between Cifar10 and CelebA.
2. 'latent_dim' : input of latent vector in Generator, normally defaults at 100.
3. 'b1, b2' : Adam optimizer parameters (beta 1,2).
4. 'img_size' : training image size.
5. 'n_critic' : number of discriminator iteration after that, generator updates one time
6. 'lambda_gp' : gradient descent parameter

User runs the 'sang_main.py' with changing the list in 'iter_list'. Function 'wrapper' runs with the set of 'iter_list' automatically.

Generated images