Skip to content

peastman/mlipbenchmarks

Repository files navigation

MLIP Benchmarks

This repository contains scripts and data files from the paper

Peter Eastman, Evan Pretti, Thomas E. Markland. "Accuracy and Efficiency Benchmarks of Pretrained Machine Learning Potentials for Molecular Simulations." (https://arxiv.org/abs/2601.16331)

computeSpice.py

This script computes the accuracy of a model on the SPICE test set. Run it as

python computeSpice.py <model>

where <model> is one of the model names defined in models.py. It calculates the MAE for every molecule or dimer and saves the results to a file in the spice directory.

simulateSpice.py

This script runs a short simulation of a molecule from the SPICE test set and reports the speed and memory use. Run it as

python simulateSpice.py <molecule> <model>

simulateWater.py

This script runs a short simulation of a water box and reports the speed and memory use. Run it was

python simulateSpice.py <width> <model>

where <width> is the width of the water box in nm. PDB files for water boxes of width 2, 3, 4, 5, and 6 are found in the water directory.

simulateSolvated.py

This script runs a simulation of a solvated molecule and checks for problems in simulation stability. Run it as

python simulateSolvated.py <model>

models.py

This file is used by the other scripts listed above. It defines the supported MLIPs, including code to create an ASE calculator for each one and to set the total charge and spin multiplicity.

About

Benchmarks of Machine Learning Potentials

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages