Skip to content

A reinforcement learning library for material and molecule optimization

License

Notifications You must be signed in to change notification settings

NREL/rlmolecule

Repository files navigation

DOI PyPI version

rlmolecule

About

A library for general-purpose material and molecular optimization using AlphaZero-style reinforcement learning.

Code under development as part of the End-to-End Optimization for Battery Materials and Molecules by Combining Graph Neural Networks and Reinforcement Learning project at the National Renewable Energy Laboratory (NREL), Colorado School of Mines (CSU), and Colorado State University (CSU). Funding provided by the Advanced Research Projects Agency–Energy (ARPA-E)'s DIFFERENTIATE program.

For more information, see our publication:

S. V., S. S., Law, J. N., Tripp, C. E., Duplyakin, D., Skordilis, E., Biagioni, D., Paton, R. S., & St. John, P. C. (2022). Multi-objective goal-directed optimization of de novo stable organic radicals for aqueous redox flow batteries. Nature Machine Intelligence. 10.1038/s42256-022-00506-3

Installation

pip install rlmolecule

Running in AWS Ray clusters

Refer to the instructions available in: devtools/aws/README.md