Using linear GANs to generate MNIST images in PyTorch

A simple application of Ian Goodfellow's original GAN article on the MNIST dataset.

Repository overview

• models.py defines a linear Generator model and a linear Discriminator model.
• train.py initialises and trains these models on MNIST examples.
• sample_generation.py exports generated examples as a png image.

Using this repository

Start by installing the required packages if you don't have them yet:

pip install -r requirements.txt

You can run training by executing:

python train.py

You can export generated images as a png file (after training) by running:

python sample_generation.py