Repository that gathers code to train StyleGAN.

The default parameters employs progressive training from 8x8 to 128x128, but this may be modified by passing arguments in train.py

Based on https://github.com/rosinality/style-based-gan-pytorch

First, create a new virtual environment

virtualenv venv -p python3
source venv/bin/activate

You might need to make sure your python3 link is ready by typing

which python3

Then install the development requirements

pip install -r requirements.txt

Download the CelebA data (if you're not planning to use your own)

python helper.py

python prepare_data.py --out data PATH_TO_DIRECTORY

python train.py --mixing data

python generate.py checkpoint/full-latest.model