Pytorch implementation of OpenAI's generative model GLOW. This repo provides a modular approach for stacking invertible transformations.
python train.py <args>
e.g.
CUDA_VISIBLE_DEVICES=0 python train.py --depth 10 --coupling affine --batch_size 64 --print_every 100 --permutation conv
- Multi-GPU support. If performance is an issue for you, I encourage you to checkout this pytorch implementation.
- Support for more datasets
- LU-decomposed invertible convolution.
This repository is no longer maintained. Feel free to file an issue if need be, however response may be slow.