Excited state neural network potential using M3GNet model

Ex_MatGL is a MatGL-based neural network potential that computes excited state energies and forces.

Installation

1. Install Pytorch. This package is tested on

   • CUDA==11.7
   • Python=3.9.17
   • torch==2.0.0

   python -m pip install torch==2.0.0+cu117 --index-url https://download.pytorch.org/whl/cu117
   

2. Install DGL.

   python -m pip install dgl==1.0.1+cu117 -f https://data.dgl.ai/wheels/cu117/repo.html
   python -m pip install dglgo -f https://data.dgl.ai/wheels-test/repo.html
   

3. Install Matgl

   Clone matgl repository and install latest version.¹

   git clone [email protected]:materialsvirtuallab/matgl.git
   python -m pip install ./matgl
   

4. Install this package

   python -m pip install .
   

Usage

Sample code is here. Please read instructuion.

References

• C. Chen, S. P. Ong, A universal graph deep learning interatomic potential for the periodic table. Nature Computational Science. 2, 718–728 (2022).
• https://github.com/materialsvirtuallab/matgl
• Y. Shi, S. Zheng, G. Ke, Y. Shen, J. You, J. He, S. Luo, C. Liu, D. He, T.-Y. Liu, Benchmarking Graphormer on Large-Scale Molecular Modeling Datasets. arXiv [cs.LG] (2022).
• https://github.com/microsoft/Graphormer

Footnotes

1. If you install matgl 0.8.5 version via pip and try to run sample code using GPU, you see the following error.TypeError: can't convert cuda:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first. ↩