chainer-LSGAN

An implementation of _Mao et al., "Least Squares Generative Adversarial Networks" 2017_ using the Chainer framework.

Disclaimer: PFN provides no warranty or support for this implementation. Use it at your own risk. See license for details.

Results

CIFAR10 & MNIST for 100 epochs

Usage

Tested using python 3.5.1. Install the requirements first:

pip install -r requirements.txt

Trains on the CIFAR10 dataset by default, and will generate an image of a sample batch from the network after each epoch. Run the following:

python train.py --device_id 0

to train. By default, an output folder will be created in your current working directory. Setting --device_id to -1 will run in CPU mode, whereas 0 will run on GPU number 0 etc. To train on MNIST, use the flag --mnist.

License

MIT License. Please see the LICENSE file for details.