API Reference

This section provides detailed API documentation for the aenet-python package.

Reference Energies

Reference Energies: Helper API for constructing atomic reference energies from regression and other supported workflows.

Training Set Management: Training set file handling and data loading. Includes support for neighbor information required for force training with PyTorch autograd.

Structure Transformations: Structure transformation framework for generating structural variations. Includes deterministic transformations (displacement, volume, strain) and stochastic transformations.

torch_featurize API Reference: PyTorch-based atomic environment descriptors and neighbor lists. Provides GPU-accelerated featurization with automatic differentiation support.

Builders: Network and optimizer builder utilities for constructing training components from configuration specifications.
Committee Training: Committee-training orchestration for training multiple seeded PyTorch members with shared settings, reloadable metadata, PyTorch-side aggregation, and ASCII export interoperability.
Training Utilities: Core training loop components including checkpoint management, metrics tracking, normalization, and epoch execution.
Inference: Inference and prediction utilities for trained models.