Closed SouthEndMusic closed 2 months ago
Since the sensible gradients don't fix the Crystal model with BackTracking, I guess we should not enable that yet?
How do these gradients compare with Shampine's advice in https://github.com/Deltares/Ribasim/issues/1705#issuecomment-2283312301, should we bring that back as well in this PR?
I added Shampine's advice, but it doesn't seem to change anything
BackTracking
as relaxation is now enabled again, with a thin wrapper to reject it when the residual gets worse. Upstream issue:https://github.com/SciML/OrdinaryDiffEq.jl/issues/2442