CURATOR

CURATOR is a toolbox for constructing machine learning interatomic potentials via active learning. It contains basic building blocks of graph neural networks, the scripts for training and deploy of ML potentials.

Getting start

• Installation
  • Requirements
  • Install via pip
  • Install from source
  • Installing PyTorch with CUDA

Tutorials

• Training Machine Learning Potentials
  • Preparing Data
  • Using Hydra’s Defaults List
  • Instantiating Objects in Python
  • Saving model and restart training
  • Multi-GPU Training
  • cuEquivariance acceleration

Interface

• LAMMPS interface
  • Requirements
  • Build LAMMPS on Gbar
  • Download LAMMPS
  • Patch curator into LAMMPS
  • Configure LAMMPS
  • Build LAMMPS
  • Deploy trained model
  • Deploy trained model ensemble
• Using CURATOR with LAMMPS ML-IAP package
  • Install LAMMPS with ML-IAP package
  • Preparing the model for ML-IAP
  • Using CURATOR with ML-IAP in LAMMPS
• OpenMM interface
  • Intallation
  • Deploy trained model