API Reference#

Complete reference documentation, generated from the docstrings in src/xnn. Start from the subpackage matching what you need:

  • xnn.common: data pipeline, featurizer base, configuration, model registry and outputs, training, benchmarking, deployment, CLI

  • xnn.gnn: E(3)-equivariant models (NequIP, MACE, Allegro, CACE) and featurizers (requires e3nn)

  • xnn.cnn: continuous-filter convolution models (SchNet)

  • xnn.dnn: descriptor models (HDNNP, ANI, PhysNet) and symmetry-function / AEV featurizers

  • xnn.ffnn: learnable classical force fields (ReaxFF / ReaxFF-nn) and their parameter-library I/O

  • xnn.transformer: shared graph-transformer building blocks (multi-head edge attention, exponential-normal radial basis)

  • xnn.hybrid: GNN + transformer models with a physics energy split (BAMBOO)

xnn.common

Shared abstractions used across all model families (cnn / dnn / gnn).

xnn.gnn

E(3)-equivariant GNN family (NequIP / MACE / Allegro / CACE).

xnn.cnn

Convolutional (continuous-filter) family (SchNet).

xnn.dnn

Descriptor + per-element-network family (HDNNP / ANI).

xnn.ffnn

ffnn: learnable classical force fields.

xnn.transformer

Transformer building blocks shared by attention-based potentials.

xnn.hybrid

Hybrid GNN + transformer interatomic potentials.
Show Source