FFLAME: A Fragment-to-Framework Learning Approach for MOF Potentials

A toolkit for fine-tuning MACE

Functions

• Extract ligands from MOFs and add hydrogen atoms and optimize the ligands.

• Run molecular dynamics simulations with various ase calculators.

• Select configurations for training using KMeans clustering method.

• Fine-tune MACE.

Installation

We recommend installing fflame in a clean conda environment. Follow these steps:

1. Create and activate the environment

conda create -n fflame python=3.10 -y
conda activate fflame

2. Install PyTorch and MACE

3. Install dependencies and fflame

conda install xtb-python -c conda-forge
pip install .
# install mofchecker
pip install git+https://github.com/Au-4/mofchecker_2.0

How to use

The example scripts are put in experiments.

References

If you use this code, please cite our paper: