Adjoint variational solver #3723
Open
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Ig-dolci
commented
Aug 5, 2024
Ig-dolci
commented
Aug 6, 2024
Ig-dolci
commented
Aug 6, 2024
Ig-dolci
commented
Aug 6, 2024
Ig-dolci
commented
Aug 6, 2024
| @@ -60,7 +60,8 @@ def evaluate_adj_component(self, inputs, adj_inputs, block_variable, idx, | |||
| adj_value = firedrake.Function(self.collapsed_space, vec.dat) | |||
|
|
|||
| if adj_value.ufl_shape == () or adj_value.ufl_shape[0] <= 1: | |||
| r = adj_value.dat.data_ro.sum() | |||
| R = firedrake.FunctionSpace(self.parent_space.mesh(), "R", 0) | |||
| r = firedrake.Function(R.dual(), val=adj_value.dat.data_ro.sum()) | |||
There was a problem hiding this comment.
Fix: We need a Cofunction instead of a float.
Ig-dolci
commented
Sep 4, 2024
| @@ -48,7 +49,7 @@ def __init__(self, lhs, rhs, func, bcs, *args, **kwargs): | |||
| # Equation RHS | |||
| self.rhs = rhs | |||
| # Solution function | |||
| self.func = func | |||
| self._func = weakref.ref(func) | |||
| @@ -72,6 +73,10 @@ def __init__(self, lhs, rhs, func, bcs, *args, **kwargs): | |||
| self.add_dependency(mesh) | |||
| self._init_solver_parameters(args, kwargs) | |||
|
|
|||
| @property | |||
| def func(self): | |||
| dFdu_form, dJdu, compute_bdy | ||
| ) | ||
|
|
||
| if self._ad_nlvs._problem._constant_jacobian: |
| @@ -667,31 +721,33 @@ def _assemble_dFdu_adj(self, dFdu_adj_form, **kwargs): | |||
|
|
|||
| def prepare_evaluate_adj(self, inputs, adj_inputs, relevant_dependencies): | |||
| dJdu = adj_inputs[0] | |||
|
|
|||
| F_form = self._create_F_form() | |||
| @@ -701,7 +757,7 @@ def evaluate_adj_component(self, inputs, adj_inputs, block_variable, idx, | |||
| if not self.linear and self.func == block_variable.output: | |||
| # We are not able to calculate derivatives wrt initial guess. | |||
| return None | |||
| F_form = prepared["form"] | |||
There was a problem hiding this comment.
Make this in prepare
| J_hat = ReducedFunctional(J, [Control(a), Control(b)]) | ||
|
|
||
| assert taylor_test( | ||
| J_hat, [a, b], [Function(R, val=rand(1)), Function(R, val=rand(1))] |
There was a problem hiding this comment.
See this random form. Keep the same random values.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description
This PR updates the adjoint solver implementation when the Jacobian is not constant, changing it from
firedrake.solvetoLinearVariationalSolver. The existing test suites in pyadjoint and firedrake cover various adjoint-solving scenarios. Additionally, I have included a new test to check the adjoint-based gradient when adjoint boundary conditions need to be computed.Note: The adjoint solver for cases where the Jacobian is constant may have improvements, but this requires more thought.