GAMD

Data and code for Graph neural network Accelerated Molecular Dynamics

Dependencies

• Pytorch 1.7.1
• jax-md (latest), jax and jaxlib 0.1.67 (for gpu support)
• Pytorch-lightning 1.3.0
• DGL 0.7.0 (https://www.dgl.ai/)
• Cupy 9.1.0
• OpenMM (https://openmm.org/) & OpenMMTools (https://openmmtools.readthedocs.io/en/0.18.1/)

Data generation

The data generation scripts using classical MD are in the dataset subfolder. For example, to generate the data for Lennard Jones particles:

cd dataset
python generate_lj_data.py

The generated data can also be downloaded from below links:

• Lennard Jones
• TIP3P
• TIP4P-Ew

The DFT data used in this work is derived from the paper: How van der Waals interactions determine the unique properties of water (pdf) (data link).

We provide our processed data (.npz fromat) at data link.

Training

The training scripts and pretrained models are in the code subfolder. Please refer to the subfolder for more details.