Stochastic Generative Flow Networks

This repository is the implementation of Stochastic Generative Flow Networks in UAI 2023 (Spotlight). This codebase is based on the open-source gflownet implementation and BioSeq-GFN-AL implementation, and please refer to those repos for more documentation.

Citing

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

@inproceedings{
	pan2023stochastic,
	title={Stochastic Generative Flow Networks},
	author={Ling Pan and Dinghuai Zhang and Moksh Jain and Longbo Huang and Yoshua Bengio},
	booktitle={Proceedings of the Thirty-Ninth Conference on Uncertainty in Artificial Intelligence}ce on Learning Representations},
	year={2023},
	url={https://proceedings.mlr.press/v216/pan23a.html}
}

Requirements

Grid

• python: 3.6
• torch: 1.3.0
• scipy: 1.5.4
• numpy: 1.19.5
• tdqm

Sequence

Please check the BioSeq-GFN-AL repo for more details about the environment.

Usage

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

• Grid (in the grid folder)

python main.py --stick <STICK> --horizon <HORIZON> --seed <SEED>

• Sequence (in the tfb folder)

python run_tfbind.py --stick <STICK> --seed <SEED>