chemtrain-deploy: A parallel and scalable framework for machine learning potentials in million-atom MD simulations

TL;DR

chemtrain-deploy is a versatile, scalable framework that enables large-scale, GPU-accelerated molecular dynamics simulations using machine learning potentials within LAMMPS, supporting multiple architectures and complex systems.

Contribution

It introduces a model-agnostic, parallel framework for deploying machine learning potentials in LAMMPS, supporting any JAX-defined semi-local potential for large-scale MD simulations.

Findings

Supports multi-GPU, million-atom MD simulations.

Validates with graph neural network architectures like MACE, Allegro, PaiNN.

Achieves state-of-the-art efficiency and scalability.

Abstract

Machine learning potentials (MLPs) have advanced rapidly and show great promise to transform molecular dynamics (MD) simulations. However, most existing software tools are tied to specific MLP architectures, lack integration with standard MD packages, or are not parallelizable across GPUs. To address these challenges, we present chemtrain-deploy, a framework that enables model-agnostic deployment of MLPs in LAMMPS. chemtrain-deploy supports any JAX-defined semi-local potential, allowing users to exploit the functionality of LAMMPS and perform large-scale MLP-based MD simulations on multiple GPUs. It achieves state-of-the-art efficiency and scales to systems containing millions of atoms. We validate its performance and scalability using graph neural network architectures, including MACE, Allegro, and PaiNN, applied to a variety of systems, such as liquid-vapor interfaces, crystalline materials, and solvated peptides. Our results highlight the practical utility of chemtrain-deploy for real-world, high-performance simulations and provide guidance for MLP architecture selection and future design.