Welcome to the documentation of MLIP!

mlip is a Python library for Machine Learning Interatomic Potentials (MLIP) in JAX. It contains the following features:

• Multiple model architectures (for now: MACE, NequIP and ViSNet)

• Dataset preprocessing

• Training of MLIP models

• Batched inference with trained MLIP models

• MD simulations with MLIP models using multiple simulation backends

• Energy minimizations with MLIP models using multiple simulation backends

• Fine-tuning of pre-trained MLIP models

As a first step, we recommend that you check out our page on Installation and our User guide which contains several tutorials on how to use the library. Furthermore, we also provide an API reference.

Note

This project is under active development.

Contents:

• Installation
• User guide
  • Getting started
  • Jupyter Notebook Tutorials
  • Deep-dive Tutorials
• API reference
  • Data processing
  • Models
  • Model training
  • Simulations