NFFLr documentation

Neural Force Field Learning toolkit (NFFLr) is an experimental library intended to support rapid iteration in research on machine learning for atomistic systems. The main goal is to provide an easily extensible uniform interface for developing, training, and deploying atomistic machine learning models.

Project history

The initial codebase is a fork of ALIGNN, with modified configuration and modeling interfaces to enable usability and performance improvements.

• discuss development on GitHub

• read the documentation

• install with pip install nfflr

The primary components of NFFLr are:

1. a general atomistic representation that is easy to work with, supports efficient batching, and can be easily converted to input formats of various machine learning models.

2. a consistent modeling interface, regardless of backend or material representation.

3. a trainer (based on ignite)that can be used regardless of the model and/or backend.

Correspondence

Please report bugs by filing an issue on usnistgov/nfflr#issues or email to brian.decost@nist.gov.

Code of conduct

Please see our Code of conduct.