Skip to content
Open access

ASSYST: A Python package for automated generation of training data for machine learning interatomic potentials

Nov 2026 · Journal of Open Source Software · 0 citations · 25 references

Abstract

The accuracy of machine-learning interatomic potentials (MLIPs) depends on high-quality training data, yet existing automation tools are often monolithic and tied to specific electronic-structure codes. ASSYST is a lightweight, pure Python library that implements the Automated Small Symmetric Structure Training method. It has a modular design: small, orthogonal classes communicate via abstract interfaces and the Atomic Simulation Environment (ASE) Calculator API. The package can be easily integrated with workflow managers, e.g., Executorlib (Janssen et al., 2025) and pyiron (Janssen et al., 2019), command-line interfaces (CLI) or graphical user interfaces (GUI), enabling heterogeneous high-performance computing (HPC) deployment. Released under the BSD 3-Clause license with full documentation and continuous integration (CI) testing, ASSYST enables the flexible and extensible generation of MLIP training data across diverse research workflows.

Read PDF

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.