MACE Inference¶

High-level Python interface for MACE machine learning interatomic potentials

What is MACE Inference?¶

MACE Inference is a user-friendly Python library that provides a high-level interface for performing atomistic simulations using MACE machine learning interatomic potentials.

Instead of writing complex setup code, you can perform common computational chemistry tasks with just a few lines:

from mace_inference import MACEInference
from ase.io import read

# Initialize calculator
calc = MACEInference(model="medium", device="auto")

# Load structure and calculate
atoms = read("structure.cif")
result = calc.single_point(atoms)

print(f"Energy: {result['energy']:.4f} eV")
print(f"Max force: {result['max_force']:.4f} eV/Å")

Key Features¶

Easy to Use

Simple, intuitive API for common atomistic simulation tasks
Comprehensive Tasks

Single-point, optimization, MD, phonons, mechanics, adsorption
GPU Accelerated

Automatic device detection with CUDA/MPS support
D3 Correction

Optional DFT-D3 dispersion correction for van der Waals systems

Supported Tasks¶

Task	Method	Description
Single-point	`single_point()`	Calculate energy, forces, stress
Optimization	`optimize()`	Geometry relaxation with BFGS/FIRE
NVT MD	`run_md()`	Constant temperature dynamics
NPT MD	`run_md()`	Constant pressure dynamics
Phonons	`phonon()`	Phonon spectrum & thermal properties
Bulk Modulus	`bulk_modulus()`	Elastic properties via EOS fitting
Adsorption	`adsorption_energy()`	Gas-framework binding energy
Coordination	`coordination()`	Coordination number analysis

Quick Links¶

Installation

Get started by installing MACE Inference
Quick Start

Learn the basics with a simple example
User Guide

Detailed guides for each feature
API Reference

Complete API documentation

Requirements¶

Python 3.9+
ASE (Atomic Simulation Environment)
mace-torch
phonopy (for phonon calculations)
torch-dftd (optional, for D3 correction)

License¶

MACE Inference is released under the MIT License. See LICENSE for details.