search

mesh-adaptation / goalie

Goal-oriented error estimation and mesh adaptation for finite element problems solved using Firedrake
Other
1 stars 1 forks source link

Adjoint breaks when `num_timesteps_per_export` in `Time_Partition` is set to 1 instead of 2 #86

Closed acse-ej321 closed 8 months ago

acse-ej321 commented 8 months ago

@jwallwork23 - I have been getting an error when running adjoint_solve: "ValueError: Haven't implemented supermesh projection with boundary conditions yet, sorry!" It seems to trace back to goalie/adjoint.py line 392, in solve_adjoint.

I was able to recreate the issue in most of the goalie demos utilizing adjoint_solve by changing the num_timesteps_per_export from 2 to 1: https://github.com/pyroteus/goalie/blob/25b4b269e4e0423f0627c539847f19a53d53108a/demos/burgers_time_integrated.py#L121

Not sure exactly why the export timestep attribute is impacting the adjoint_solve?