Welcome to AmpTorch’s documentation!

AmpTorch is a PyTorch implementation of the Atomistic Machine-learning Package (AMP) code that seeks to provide users with improved performance and flexibility as compared to the original code. The implementation does so by benefiting from state-of-the-art machine learning methods and techniques to be optimized in conjunction with high-throughput supercomputers. AmpTorch is built on top of PyTorch Geometric and Skorch.

Manual:

• Home
• Installation
• Usage
  • Training and Predicting with 2D Water Trajectory
  • Configs
  • Train model
  • Load checkpoints
  • Make predictions
  • Construct AmpTorch-ASE calculator
• Examples
  • Gaussian Multiple (GMP) Descriptors
  • SingleNN Atomistic Neural Network Structures
  • lmdb as Database Management Solution for Large Dataset
  • Uncertainty Quantification (UQ) via Conformal Prediction (CP)
• amptorch
  • API

Development notes

Reporting issues

Regarding bugs, issues or suggested feature improvements related to the software, please use the issue tracker of the GitHub project.

Contributing

If you want to contribute to this project, please use the fork and pull following the guidelines and pertaining to the overall objective of this project as described above.