Installing LAMMPS with MACE on ALCF HPC Systems

This repo details the step-by-step process to install LAMMPS with MACE forcefield (and some other optional packages) on ALCF HPC systems.

Running MACE finetuning using multigpu setting on ALCF systems

To use multigpu for training, you have to write the DistributedEnvironment class (mace/tools/slurm_distributed) for PBS scheduler. Then edit the relevant import in run_train.py (amce/cli/run_train.py, line 56, from mace.tools.slurm_distributed import DistributedEnvironment)