zhairuizhi/graphs4cfd

Graph Neural Networks (GNN) based solvers for Computational Fluid Dynamics (CFD)

Graphs4CFD

Graphs4CFD is a library built upon PyTorch and Pytorch Geometric (PyG) to code and train Graph Neural Networks (GNNs) based solvers for Computational Fluid Dynamics (CFD) applications.

Contents

• Implemented GNN models
• Installation
• Examples
• Cite

Implemented GNN models

To date, Graphs4CFD supports the following GNN models:

• MuS-GNN - Lino et al. 2022 (https://doi.org/10.1063/5.0097679)
• REMuS-GNN - Lino et al. 2022 (https://doi.org/10.1063/5.0097679)
• Mult-scale GNNs with low-resolution graphs obtained by Guillard's coarsening - Appendix C.3 in Lino et al. 2022 (https://doi.org/10.1063/5.0097679)

Installation

Graphs4CFD requires Python 3.7 or higher and a version of PyTorch compatible with your CUDA version. We recomend installing Graphs4CFD and its dependencies in a virtual envioroment, e.g., using Anaconda or Miniconda. To install Graphs4CFD and its dependecies (except PyTorch), run the following commands:

git clone git@github.com:mario-linov/graphs4cfd.git
cd graphs4cfd
pip install -e .

This also installs PyG and compiles and installs PyTorch Cluster, so it may take a while. Once Graphs4CFD has been installed, it can be imported in Python as follows:

import graphs4cfd as gfd

Examples

There are examples of how to use Graphs4CFD for training and inference in the examples/ folder.

⚠️To clone the jupyter notebooks in the inference/ folder you need to have Git LFS installed. On Ubuntu, Git LFS can be installed painlessly by running:

curl -s https://packagecloud.io/install/repositories/github/git-lfs/script.deb.sh | sudo bash
sudo apt-get install git-lfs

Cite

To cite Graphs4CFD, please use the following reference:

Mario Lino, Stathi Fotiadis, Anil A. Bharath, and Chris Cantwell. "Multi-scale rotation-equivariant graph neural networks for unsteady Eulerian fluid dynamics". Physics of Fluids, 34 (2022).

@article{lino2022multi,
    author = {Lino, Mario and Fotiadis, Stathi and Bharath, Anil A. and Cantwell, Chris},
    title = {{Multi-scale rotation-equivariant graph neural networks for unsteady Eulerian fluid dynamics}},
    journal = {Physics of Fluids},
    volume = {34},
    year = {2022},
    url = {https://doi.org/10.1063/5.0097679},
}