Skip to content

rfjd/GNS-DispersedSystems

Repository files navigation

Aref Hashemi & Aliakbar Izadkhah (2024)

This repository includes python codes, data, and results for a project on the use of a Graph Neural Network (GNN) for simulation of the dymamics of a multidisperse suspension of partices in a box of fluid.

Sample Rollout GIF

This work is inspired by the following prior works:

Note: parts of the codes in this repo have been borrowed from the repository https://github.com/geoelements/gns and heavily modified for our own problem. In particular, we use the data_loader script and the corresponding parts of the interface from https://github.com/geoelements/gns.

Installation (These specific versions work):

  1. Install CUDA 12.4 and the appropriate NVIDIA driver.
    (CUDA 11.7 would also work with PyTorch 1.13.1 and related dependencies.)

  2. Install the required Python packages with the following commands:

How to run the code:

Use bash script run.bash to perform a full training along with rendering and generating animations. Use bash script rollout_render.bash to generate outputs from a pretrained model.

How to cite:

Aref Hashemi & Aliakbar Izadkhah, A graph neural network simulation of dispersed systems,
Mach. Learn.: Sci. Technol. 6 015044 (2025)
DOI: 10.1088/2632-2153/adb0a0

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published