Serpent interface by cahowar9 · Pull Request #74 · ardorlab/MIDAS

cahowar9 · 2025-10-01T14:34:38Z

I updated the optimizer.py file to handle the new serpent code interface, as well as the addition of the interface itself. The bayesian optimization file can now handle continuous variables as well. Serpent interface can handle base case, doppler temperature coefficient with automatic temperature updates, and automatic depletion runs. If any one of multiple required runs fails, it will consider the evaluation as a whole to have failed and result in unrealistic results to be thrown to drive the optimization away.

Added a serpent interface and initial solution generation for continuous variables

JakeMikouchi · 2025-10-01T15:00:35Z

midas/algorithms/bayesian_optimization.py

    def __init__(self, input, noise=1e-6, random_state=None):
        self.input = input
-        self.dimensions = self.parse_dimensions()  # List of length D, each an array of categories
+        #self.dimensions = self.parse_dimensions()  # List of length D, each an array of categories


Did you forget to un-comment or is this purposeful?

JakeMikouchi · 2025-10-01T15:01:17Z

midas/algorithms/bayesian_optimization.py

-        self.binary_dims = [self.calculate_binary_length(len(dim_cats)) for dim_cats in self.dimensions]
-        self.total_features = sum(self.binary_dims)  # Total number of binary features
+        # Precompute the total number of binary features for LP problem
+        if input.calculation_type in ['single_cycle','eq_cycle']:


should lattice opts be added here?

JakeMikouchi · 2025-10-01T15:01:42Z

midas/algorithms/bayesian_optimization.py

-        #Encode population in binary and scale in preparation to fit data
-        encoded_population = self.binary_encode_population(self.population)
-        scaled_population = self.scaler_x.fit_transform(encoded_population)
+        if self.input.calculation_type in ['single_cycle','eq_cycle']:


same question

JakeMikouchi · 2025-10-01T15:06:10Z

midas/codes/serpent.py

please change file name to include version number

JakeMikouchi · 2025-10-01T15:06:42Z

midas/optimizer.py

        core_parameters = [self.input.nrow, self.input.ncol, self.input.num_assemblies,
                            self.input.symmetry, self.input.calculation_type]
-        if chromosome:
+        if chromosome is not None:


Why the change?

JakeMikouchi · 2025-10-01T15:09:57Z

midas/optimizer.py


    ## Initialize beginning population
            self.population.current = []
+            for i in range(self.population.size):


why was this added? the same function is performed 3 lines down?

JakeMikouchi · 2025-10-01T15:11:24Z

midas/optimizer.py

                ## Execute and parse objective/constraint values
                self.population.current = pool.starmap(self.eval_func, zip(self.population.current, repeat(self.input)))
-                if 'cost_fuelcycle' in self.input.objectives.keys():
+                if 'cost_fuelcycle' in self.input.objectives.keys() and self.input.code_interface not in ['serpent','function']:


I don;t think that the conditions after the and are necessary?

JakeMikouchi · 2025-10-01T15:12:58Z

midas/utils/optimizer_tools.py

+            if "normalized_continuous_range" in value:
+                chromosome.append(random.uniform(0.0, 1.0))
+            elif "normalized_discrete_range" in value:
+                chromosome.append(random.choice(value["normalized_discrete_range"]))


I don't think "narmalized" is required but I'll accept it

JakeMikouchi

some minor changes and questions. My biggest question is why does BO not use the contraceptive check. You will have to tell me why when we get the chance

JakeMikouchi · 2025-10-02T12:18:26Z

midas/codes/serpent.py

+    logger.debug(f"Changing to new working directory: {indv_dir}")
+    os.chdir(indv_dir)
+    #Create subdirectories for base, DTC, and depletion runs
+    if not base_dir.exists():


I suggest you put all of the file/folder building in its own function to clean up evaluate()

Same applies for the cleaning of the files after the calculation. One function for creating the folders/files, one function to delete files.

JakeMikouchi · 2025-10-02T12:19:13Z

midas/codes/serpent.py

+        dep_process = subprocess.Popen(dep_cmd)
+
+    #Unrealistically bad results to return if code fails to drive optimization away from this region
+    fail_results = {'doppler_temperature_coefficient': 5, 'cycle_length': 0.01, 'cost_fuelcycle': 1000000000000,'total_mass': 1000000000,'fdeltah': 8,'max_doserate':50000}


results parsing and handling should be put in serpent_get_results

JakeMikouchi · 2025-10-02T12:19:26Z

midas/codes/serpent.py

+
+        else:
+            peaking_results = [] #!TODO: Add parsing logic for going through detector results 
+        peaking_results = peaking_results[peaking_results != 0]


same comment

gkdelipei · 2025-10-09T15:19:09Z

midas/algorithms/bayesian_optimization.py

+            #Scale the candidates to fit the data
+            scaled_candidates = self.scaler_x.transform(candidates)
+        else:
+            scaled_candidates = candidates


Do you need to put this if statement in every function ? Seems a bit redundant. Can you put it somewhere else where this is coded only once ? I think a good coding principle is to use decorators for such situations (i.e. when the same if else statement is present in multiple functions). If you are not familiar with decorators, check them and see if you can use them for this.

gkdelipei · 2025-10-09T15:26:33Z

midas/algorithms/bayesian_optimization.py

        def acquisition_function(x):
-            binary_vec = self.real_vector_to_binary(x)
+            if self.input.calculation_type in ['single_cycle','eq_cycle']:
+                vec = self.real_vector_to_binary(x)


might be good to use a more descriptive name for a variable than "vec"

gkdelipei · 2025-10-09T15:28:30Z

midas/algorithms/bayesian_optimization.py


        #Create LWR core parameters for use in checking constraints
-        core_parameters = [self.input.nrow, self.input.ncol, self.input.num_assemblies, self.input.symmetry]
+        #core_parameters = [self.input.nrow, self.input.ncol, self.input.num_assemblies, self.input.symmetry]


If this line is not needed just delete it

gkdelipei · 2025-10-09T15:28:42Z

midas/algorithms/bayesian_optimization.py

-            if optools.Gene_Validity_check.abortive_check(self.input, gene_list, self.input.genome, core_parameters, candidate):
-                candidates.append(candidate)
-        return candidates
+            #if optools.Gene_Validity_check.abortive_check(self.input, gene_list, self.input.genome, core_parameters, candidate):


same as previous comment.

gkdelipei · 2025-10-09T15:35:53Z

midas/input_parser.py

+                    new_item = item
+                    if not isinstance(new_item, bool):
+                        raise ValueError("'apply' flag for input template must be true or false")
+                elif new_key == 'depletion_steps':


Do you need to parametrize all these inputs ? Overall, it is good to parametrize through the input, parameters such as the number of mpis, neutron histories etc. However, very specific aspects of the model, like the detectors or the depletion step might be better left for the input template. The reason is that if this input is too specific for a case only it cannot be practically re-used in other Serpent models.

gkdelipei · 2025-10-09T15:53:55Z

midas/utils/LWR_fuelcyclecost.py

+
+    P_feed = 31.35 # USD/lb_U3O8
+    P_conv = 5.00 # USD/kg_U3O8
+    P_enr = 105 # USD/SWU


Because we will start having multiple cost functions, we can first maybe change the name of the file because it is referencing LWR. Also, it might be better to make the general parameters as global or maybe make this as a class. So we should define some default values for these parameters and then modified if needed for a specific case.

gkdelipei · 2025-10-09T15:56:24Z

midas/utils/problem_preparation.py

+        def get_precision(x):
+            """Return the number of decimal places needed to represent x exactly. Used for rounding in case of floating point errors"""
+            d = Decimal(str(x))
+            return -d.as_tuple().exponent


This needs to be defined within the function ? I do not like nested functions if it is not necessary

gkdelipei

Check our comments. We can discuss them in a separate meeting.

cahowar9 and others added 14 commits September 18, 2025 16:02

Allow continuous vars to pass num gene combination

340a9ca

Add serpent interface and initial solutions

a82b670

Added a serpent interface and initial solution generation for continuous variables

Continue serpent integration

88336e5

Continue serpent interface

ea72731

Serpent interface almost ready to test

c7f23f3

Ready for first serpent code interface test

e8304fc

Update BO to handle continuous variables

b7343e5

Update serpent code interface

8ef68fa

Serpent interface debugging

09d5fb2

Debugging

f04a79c

Merge branch 'ardorlab:devel' into devel

ddaf5c3

Update BO to handle continuous variables

6602072

Add serpent code interface

430f89a

Merge remote-tracking branch 'refs/remotes/origin/devel' into devel

2f73dd8

JakeMikouchi reviewed Oct 1, 2025

View reviewed changes

midas/codes/serpent.py

Copy link
Copy Markdown

Collaborator

JakeMikouchi Oct 1, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

please change file name to include version number

JakeMikouchi reviewed Oct 1, 2025

View reviewed changes

JakeMikouchi reviewed Oct 2, 2025

View reviewed changes

gkdelipei reviewed Oct 9, 2025

View reviewed changes

gkdelipei requested changes Oct 9, 2025

View reviewed changes

Conversation

cahowar9 commented Oct 1, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

JakeMikouchi left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

gkdelipei Oct 9, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

gkdelipei left a comment

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

cahowar9 commented Oct 1, 2025 •

edited

Loading

gkdelipei Oct 9, 2025 •

edited

Loading