API reference

Use the Python API to construct FlashCart models, evaluate trained potentials, and work directly with Cartesian tensor operations. See Quickstart for a complete training and evaluation example.

Package	Description
flashcart.model	FlashCart potential and the base class for atomistic models.
flashcart.calculators	Interfaces to atomistic simulation software, including ASE and LAMMPS.
flashcart.o3	Irreducible Cartesian tensors, tensor products, and equivariant linear maps.
flashcart.nn	Layers and radial functions for equivariant message passing.
flashcart.data	Atomic configurations, datasets, neighbor graphs, and batching.
flashcart.training	Model training and evaluation, loss functions, optimizers, and learning-rate schedules.
flashcart.cli	Command-line entry points for training, evaluation, profiling, and model export.
flashcart.utils	Shared helpers for configuration, geometry, distributed execution, and logging.