atomistic-machine-learning/schnetpack

SchNetPack - Deep Neural Networks for Atomistic Systems

What it solves

SchNetPack is a toolbox for developing and applying deep neural networks to predict quantum-chemical properties of molecules and materials, such as potential energy surfaces, dipole moments, and polarizability. It simplifies the process of implementing and evaluating new atomistic neural network models by providing essential building blocks, training management, and access to benchmark datasets.

How it works

The library leverages PyTorch Lightning and Hydra for flexible configuration and training. It implements several specialized architectures, including SchNet (a continuous-filter CNN) and PaiNN (equivariant message-passing), and includes modules for electrostatics, Ewald summation, and ZBL repulsion. It also provides a GPU-accelerated molecular dynamics code that supports thermostats, barostats, and path-integral MD.

Who it’s for

Researchers and scientists in the field of atomistic machine learning, quantum chemistry, and materials science who need to build, train, and train models to predict molecular and material properties.

Highlights

  • Pre-built Architectures: Includes SchNet and PaiNN for modeling quantum interactions.
  • Benchmark Integration: Simple CLI access to common benchmark datasets like QM9 and MD17.
  • Advanced Physics Modules: Built-in support for dipole moments, stress tensors, and electrostatics.
  • MD Integration: GPU-accelerated molecular dynamics and an interface to LAMMPS.
  • Flexible Configuration: Uses Hydra for hierarchical configuration of models and data.

Related

  • Project
    materialyzeai/matgl

    A graph deep learning library for materials science that provides various GNN architectures to predict material properties and simulate potential energy surfaces.

  • Project
    deepmodeling/deepmd-kit

    A deep learning package that turns quantum-mechanical reference data into fast, scalable interatomic potentials for molecular and materials science simulations.

  • Project
    microsoft/mattersim

    A deep learning atomistic model for simulating bulk materials across various elements, temperatures, and pressures to predict energy, forces, and stress.

  • Project
    qiskit-community/qiskit-machine-learning

    A library providing fundamental building blocks for quantum machine learning, enabling the implementation of quantum kernels, neural networks, and hybrid quantum-classical models.

  • Project
    torch-uncertainty/torch-uncertainty

    A deep learning framework for uncertainty quantification that provides tools to make neural networks more reliable through uncertainty-aware training, evaluation, and post-processing.