xnn.gnn.models#

Graph-network models: SchNet, DimeNet, DimeNet++, GemNet and PaiNN (plain PyTorch) and the E(3)-equivariant NequIP / MACE / Allegro / CACE / AIMNet2 (e3nn).

Modules

`aimnet2`	AIMNet2 (Anstine, Zubatyuk and Isayev, 2025).
`aimnet2_foundation`	Load the published AIMNet2 models into the xnn `AIMNet2`.
`allegro`	Allegro (Musaelian et al. 2023): strictly local equivariant potential.
`base`	Shared bases for the GNN potentials (NequIP / MACE / Allegro / CACE).
`blocks`	Shared E(3)-equivariant building blocks (e3nn) for the GNN potentials.
`cace`	CACE (Cheng 2024): Cartesian Atomic Cluster Expansion potential.
`dimenet`	DimeNet and DimeNet++ (Gasteiger et al.) -- directional message passing.
`gemnet`	GemNet (Gasteiger, Becker, Guennemann, NeurIPS 2021, arXiv:2106.08903).
`mace`	MACE (Batatia et al. 2022): higher body-order equivariant message passing.
`mace_foundation`	Load pretrained MACE foundation models into the xnn `MACE`.
`nequip`	NequIP (Batzner et al. 2022): E(3)-equivariant message-passing potential.
`painn`	PaiNN (Schuett, Unke, Gastegger, ICML 2021) -- equivariant message passing.
`schnet`	SchNet (Schuett et al., NIPS 2017) -- continuous-filter convolutional network.