Home

Awesome

Generative Augmented Flow Networks

This repository is the implementation of Generative Augmented Flow Networks in ICLR 2023 (Spotlight). This codebase is based on the open-source gflownet implementation, and please refer to that repo for more documentation.

Citing

If you used this code in your research or found it helpful, please consider citing our paper:

@inproceedings{
	pan2023generative,
	title={Generative Augmented Flow Networks},
	author={Ling Pan and Dinghuai Zhang and Aaron Courville and Longbo Huang and Yoshua Bengio},
	booktitle={International Conference on Learning Representations},
	year={2023},
	url={https://openreview.net/forum?id=urF_CBK5XC0}
}

Requirements

Grid

Molecule discovery

Please check the gflownet repo for more details about the environment

Usage

Please follow the instructions below to replicate the results in the paper.

python toy_grid_dag.py --augmented 1 --seed <SEED> --horizon <HORIZON>
python gflownet.py --w_ri 1