Consider moving the data stored in demos/nfem_heat_equation/data to a proper data storage platform. Having to download 500MB of data just to get the main firedrake-to-pytorch linking code is pretty excessive and slow. The demo script can then check to see if the data is on disk before fetching it from a server, similar to how pytorch and tensorflow handle their datasets.
Consider moving the data stored in
demos/nfem_heat_equation/data
to a proper data storage platform. Having to download 500MB of data just to get the main firedrake-to-pytorch linking code is pretty excessive and slow. The demo script can then check to see if the data is on disk before fetching it from a server, similar to how pytorch and tensorflow handle their datasets.