Variational Discriminator Bottleneck

Code for the image generation experiments in Variational Discriminator Bottleneck: Improving Imitation Learning, Inverse RL, and GANs by Constraining Information Flow.

Code for the iRL experiemnts: https://github.com/qxcv/vdb-irl/

Bibtex

@inproceedings{
  VDBPeng18,
  title={Variational Discriminator Bottleneck: Improving Imitation Learning,
  Inverse RL, and GANs by Constraining Information Flow},
  author = {Peng, Xue Bin and Kanazawa, Angjoo and Toyer, Sam and Abbeel, Pieter
  and Levine, Sergey},
  booktitle={ICLR},
  year={2019}
}

Acknowledgement

Our code is built on the GAN implmentation of Which Training Methods for GANs do actually Converge? [Mescheder et al. ICML 2018]. This repo adds the VGAN and instance noise implementations, along with FID computation.

Usage

First download your data and put it into the ./data folder.

To train a new model, first create a config script similar to the ones provided in the ./configs folder. You can then train you model using

python train.py PATH_TO_CONFIG

You can monitor the training with tensorboard:

tensorboard --logdir output/<MODEL_NAME>/monitoring/

Experiments

To generate samples, use

python test.py PATH_TO_CONIFG

You can also create latent space interpolations using

python interpolate.py PATH_TO_CONFIG

Pre-trained model

A pre-trained model for CelebA-HQ can be found here