Welcome to CrabNet!
This is the documentation for crabnet.
Compositionally-restricted attention-based Network (CrabNet) is a
composition-based regression and classification algorithm for materials properties.
CrabNet leverages a transformer architecture inspired by recent advances in natural
language processing to learn chemical relationships and improve property prediction.
CrabNet was originally proposed by Wang, Kauwe, Murdock, and Sparks (npj Computational
Materials, ChemRxiv). The
library crabnet provides a GPU-accelerated (and CPU-compatible) Python
implementation. crabnet will regress or classify materials properties for a custom training dataset of chemical formula and target
properties in just a few lines of code! Get started learning about CrabNet.