API Reference¶
This section contains the selected public API reference for AIMNet2. Lower-level kernels, tensor operators, and training internals are documented through source code and focused guides where they are intended for extension.
Package Structure¶
- Calculators - Calculator interfaces for molecular simulations
- Modules - Neural network modules and model components
- Data - Dataset handling and data loading utilities
- Config - Configuration and model building utilities
Supported Public Imports¶
| Import path | Public surface |
|---|---|
aimnet.calculators |
AIMNet2Calculator, AIMNet2ASE, AIMNet2Pysis, AIMNet2TorchSim |
aimnet.data |
DataGroup, SizeGroupedDataset, SizeGroupedSampler |
aimnet.models |
AIMNet2, model loading helpers |
aimnet.modules |
Model building blocks intended for configuration and extension |
aimnet.config |
YAML/module construction helpers |
The aimnet.ops, aimnet.nbops, aimnet.kernels, aimnet.train, and aimnet.cli modules are importable but are primarily advanced or internal extension points. Their contracts may be narrower than the calculator and model APIs.
Quick Links¶
Calculators¶
The main entry points for using AIMNet2:
AIMNet2Calculator- Core calculator for inferenceAIMNet2ASE- ASE calculator interface (requiresaimnet[ase])AIMNet2Pysis- PySisyphus calculator interface (requiresaimnet[pysis])AIMNet2TorchSim- TorchSim model interface (requires Python 3.12+ andaimnet[torchsim])
Command Line Interface¶
aimnet --help