Skip to content
master
Switch branches/tags
Code

Latest commit

 

Git stats

Files

Permalink
Failed to load latest commit information.
Type
Name
Latest commit message
Commit time
 
 
 
 
 
 
 
 
 
 
 
 
 
 

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

About

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

Resources

License

Releases

No releases published

Packages

No packages published

Languages