[OptApp] Integration of NLOPT libraray #11349
Conversation
There was a problem hiding this comment.
@RezaNajian I am happy to see that your and @sunethwarna design allow this clean implementation of the external opt. library. In general, the PR is nice and clean. Still, I think there are places for further improvement.
| @kratos_unittest.skipIf(not nlopt_available, "Missing nlopt python libraries ") | ||
| class TestNLOPTOptimizers(kratos_unittest.TestCase): | ||
| # @classmethod | ||
| # def setUpClass(cls): |
|
|
||
| # nlopt algorithm | ||
| self.algorithm_name = NLOPT_settings["algorithm_name"].GetString() | ||
| if(len(self.__constraints)==0): |
There was a problem hiding this comment.
can you update the CheckOptimizerSupport to take _optimization_problem and do all the checks inside to keep init cleaner?
applications/OptimizationApplication/python_scripts/algorithms/NLOPT_algorithms.py
Show resolved
Hide resolved
| @@ -0,0 +1,124 @@ | |||
| { | |||
There was a problem hiding this comment.
please move NLOPT_tests folder into algorithm_test.
applications/OptimizationApplication/python_scripts/controls/master_control.py
Show resolved
Hide resolved
applications/OptimizationApplication/python_scripts/responses/response_routine.py
Show resolved
Hide resolved
| @@ -203,7 +203,7 @@ def Update(self, new_control_field: ContainerExpressionTypes) -> bool: | |||
| if not IsSameContainerExpression(new_control_field, self.GetEmptyField()): | |||
| raise RuntimeError(f"Updates for the required element container not found for control \"{self.GetName()}\". [ required model part name: {self.model_part.FullName()}, given model part name: {new_control_field.GetModelPart().FullName()} ]") | |||
|
|
|||
| if KratosOA.ExpressionUtils.NormL2(self.control_field - new_control_field) > 1e-9: | |||
| if KratosOA.ExpressionUtils.NormL2(self.control_field - new_control_field) > 1e-15: | |||
There was a problem hiding this comment.
Should we move this tolerance into JSON file to allow changing it?
There was a problem hiding this comment.
I do not know honestly, I am open to suggestions
applications/OptimizationApplication/python_scripts/algorithms/standardized_NLOPT_objective.py
Show resolved
Hide resolved
applications/OptimizationApplication/python_scripts/algorithms/standardized_NLOPT_objective.py
Show resolved
Hide resolved
applications/OptimizationApplication/python_scripts/algorithms/NLOPT_algorithms.py
Show resolved
Hide resolved
|
this PR and the corresponding reviews are moved and addressed in PR #11480. |
This PR introduces an interface to the NLOPT optimizers. Currently, it supports three gradient-based optimizers: MMA, SLSQP, and LBFGS. The interface is designed in a flexible manner to ensure gradient-free optimizers can also be integrated with ease. However, our current focus is on gradient-based methods.
The interface is implemented at the Python level, and the NLOPT Python library needs to be in the Python path.
A shell thickness optimization case has been added and tested using the MMA algorithm.