MOF Generation on HPC

Create new MOFs by combining generative AI and simulation on HPC.

Installation

The requirements for this project are defined using Anaconda.

Install the environment file appropriate for your system with a command similar to:

conda env create --file envs/environment-cpu.yml --force

If solving is slow try updating to the newest version of conda and using the libmamba solver:

conda update -n base conda
conda install -n base conda-libmamba-solver
conda config --set solver libmamba
conda env create --file envs/environment-cpu.yml

Running MOFA

The run_parallel_workflow.py script defines an HPC workflow using MOFA.

First set up the required input files by running assemble-inputs.ipynb in input_files/zn-paddle-pillar. and get-macemp-0a.sh in inputs-files/mace.

The run scripts available in the root directory include input argument configurations appropriate for different systems at different scales. For example, run-polaris-test.sh is configured for a short run on Polaris using a small number of nodes.

Each run will produce a run directory in run named using the start time and a hash of the run parameters.

The run directory contains the following files:

• run.log: The log messages produced during execution
• params.json: The arguments provided to the run script
• all-ligands.csv: A CSV file with the geometries of the generated ligands in XYZ format, if they passed all validation screens, and the SMILES string (if available).
• db: A MongoDB database folder. Convert to JSON format using ./bin/dump_data.sh
• *-results.json: Summaries of different types of computations. See visualizations in scripts for examples on reading them.