ML potential GPU benchmarks

This project contains fits and LAMMPS input scripts for benchmarking accuracy and performance of ML potentials.

Organization

The following directories for fits are organized like:

<potential>
    <system>

E.g. <potential> can be SNAP, ACE, etc. and <system> can be Si, Li, Mo, etc.

Training data is stored in the data directory.

Instructions on fitting and benchmarking

First activate your Python environment, e.g.

source ~/venv-fitsnap-pace/bin/activate

Proceed into a <potential>/<system> directory and run a fit with:

mpirun -np 4 python -m fitsnap3 input.in --overwrite

Obtain CPU benchmarks on a 10x10x10 unit cell system with:

mpirun -np P lmp -in in.run -v nrep 10

Obtain GPU/Kokkos benchmarks on a 10x10x10 unit cell system with (e.g. using 2 GPUs):

mpirun -np 2 lmp -k on g 2 -sf kk -pk kokkos newton on neigh half -in in.run -v nrep 10