mamoeed/apebench-pde-simulation-data

This dataset is generated from APEBench.

It contains 128x128 grid 2D data of various PDEs with periodic initial conditions. Data is already partitioned into train/val/test sets randomly. Train and val set have been rolled out for 50 time steps. Test set is rolled out for 200 time steps to testing generality over time.

The PDE cases included are: Advection-Diffusion, Pure advection, Burgers, Hyper-Diffusion, Fisher-KPP, Gray Scott, Kolmogorov Flow and Kuramoto Sivashinsky.

For generating the datasets and configs can be found in the generate_datasets.ipynb. The configs not specified are default from APEBENCH.