I'm working on something based off of this example: https://sgpp.sparsegrids.org/docs/example_predictiveRefinement_py.html

I have a ModBSplineGrid and I've refined it to conform to a set error tolerance, with the coefficients from solving the hierarchisation in a separate DataVector. What is the simplest way to save this grid and load it from file to reuse later? There does not seem to be a generic (whatever grid you happen to be working with).save(filename) or .load(filename).

A GridDataBase (my best guess right now)? Or a Checkpoint? Or break everything up into DataMatrix's and DataVector's and save those? But how do I reinstate my grid from these?
I'm probably just not "seeing the forest for the trees", but I also feel like it shouldn't be this difficult to find a clear answer for this for any given grid.

In a next step I need to port over a uniform sparse grid from a non-SGpp-implementation for refinement, so I definitely need a way to convert from numpy-array to DataMatrix (easy enough), to ModBSplineGrid (no clue right now).

I would appreciate some guidance on this. Saving and loading grids should be covered by the examples.