Davout 0.1.1.dev157__tar.gz → 0.1.1.dev159__tar.gz

{davout-0.1.1.dev157/source/Davout.egg-info → davout-0.1.1.dev159}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: Davout
-Version: 0.1.1.dev157
+Version: 0.1.1.dev159
 Summary: Scientific computing utilities combining FEM, ANN and tooling
 Author: Matheus Janczkowski
 Author-email: Matheus Janczkowski <matheusj2009@hotmail.com>

{davout-0.1.1.dev157 → davout-0.1.1.dev159}/pyproject.toml

@@ -63,7 +63,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "Davout"
-version = "0.1.1.dev157"
+version = "0.1.1.dev159"
 description = "Scientific computing utilities combining FEM, ANN and tooling"
 readme = "README.md"
 authors = [

{davout-0.1.1.dev157 → davout-0.1.1.dev159}/source/Davout/DeepMech/aa_tests/tests_tool_box/test_SVD_based_NNs.py

@@ -526,7 +526,8 @@ class TestSVDArchitecture:
             modulating_function_performance, "Householder epsilon": 1.0, 
             "activations accessory layer list": 
             accessory_activation_list_performance, "householder vector"+
-            " builder method": householder_vector_builder_method}  
+            " builder method": householder_vector_builder_method,
+            "hardware device": "CPU"}  
 
             # Assembles the model
 

{davout-0.1.1.dev157 → davout-0.1.1.dev159}/source/Davout/DeepMech/aa_tests/tests_tool_box/test_quotient_space_NNs.py

@@ -11,9 +11,11 @@ import os
 
 import tensorflow as tf
 
-from ...tool_box import ANN_tools
+from .....Davout.DeepMech.tool_box import ANN_tools
 
-from ...tool_box import training_tools
+from .....Davout.DeepMech.tool_box import training_tools
+
+from .....Davout.PythonicUtilities.path_tools import get_parent_path_of_file
 
 # Defines a function to test the ANN tools methods
 
@@ -104,7 +106,60 @@ class TestANNTools(unittest.TestCase):
 
         self.save_model_file = "saved_model.keras"
 
-    # Defines a function to test the gated architecture
+        self.parameters_dtype = "float64"
+    
+    # Defines a function to test saving and reloading the model
+    
+    def test_saving_and_loading(self):
+
+        print("\n#####################################################"+
+        "###################\n#                       Tests saving and"+
+        " loading                       #\n###########################"+
+        "#############################################\n")
+        
+        # Sets the path to save the model
+
+        model_path = os.path.join(get_parent_path_of_file(), "gated_mo"+
+        "del.keras")
+
+        print("\nSaves the model at: "+str(model_path)+"\n")
+
+        # Saves the model
+
+        # Tests now with custom layers
+        
+        ANN_class = ANN_tools.MultiLayerModel(
+        self.input_dimension_gradient_tests, 
+        self.activation_list_gradient_tests, enforce_customLayers=True, 
+        verbose=True, parameters_dtype=self.parameters_dtype, 
+        custom_architecture={"name": "GatedQuotientSpace", 
+        "quotient space dimension": self.quotient_space_dimension})
+
+        custom_model = ANN_class()
+
+        custom_model.save(model_path)
+
+        # Loads it back
+
+        loaded_model = tf.keras.models.load_model(model_path)
+
+        # Gets the training data as a tensorflow array
+
+        self.training_input_constant = tf.constant(self.training_data, 
+        dtype=tf.as_dtype(self.parameters_dtype))
+
+        # Tests the loaded model
+
+        output_original = custom_model(self.training_input_constant)
+
+        output_loaded = loaded_model(self.training_input_constant)
+
+        print("Output from saved model:\n"+str(output_original))
+
+        print("\nOutput from loaded model:\n"+str(output_loaded)+"\n\n"+
+        "The maximum absolute difference between components of the ori"+
+        "ginal and loaded models is: "+str(tf.reduce_max(tf.abs(
+        output_original-output_loaded)).numpy())+"\n")
 
     def test_gated_nn(self):
 

{davout-0.1.1.dev157 → davout-0.1.1.dev159}/source/Davout/DeepMech/tool_box/ANN_tools.py

@@ -4,6 +4,8 @@ import tensorflow as tf
 
 import numpy as np
 
+import inspect
+
 from ...DeepMech import custom_architectures
 
 from ..tool_box import differentiation_tools as diff_tools
@@ -679,8 +681,23 @@ class MixedActivationLayer(tf.keras.layers.Layer):
 
         code_info_dictionary = config.pop("code_given_info_class")
 
+        # Gets all the necessary arguments of CodeGivenLayerInfo
+
+        arguments_of_layer_info = inspect.signature(CodeGivenLayerInfo
+        ).parameters.keys()
+
+        # Completes the dictionary of code-given information to allow
+        # for the inclusion of keys that were not incorporated in older
+        # models. Notice that, for missing keys, the default value will
+        # be None
+
+        complete_dictionary = {argument_name: code_info_dictionary.get(
+        argument_name, None) for argument_name in arguments_of_layer_info}
+
+        # Instantiates the class of code-given information
+
         config["code_given_info_class"] = CodeGivenLayerInfo(
-        **code_info_dictionary)
+        **complete_dictionary)
 
         return cls(**config)
 

{davout-0.1.1.dev157 → davout-0.1.1.dev159}/source/Davout/MultiMech/tool_box/binary_tools.py

@@ -534,7 +534,8 @@ def write_visualization_copy(functional_data_dictionary, file_name,
 mesh_file, code_given_field_name, time=0.0, time_step=0, 
 visualization_copy_file=None, close_file=True, 
 code_given_mesh_data_class=None, comm_object=None, original_function=
-None, verbose=True):
+None, verbose=True, reading_is_ready=False, 
+return_visualization_file_name=False):
     
     if verbose:
     
@@ -551,7 +552,7 @@ None, verbose=True):
 
     time_points = time*1.0
 
-    if comm_object is None:
+    if (comm_object is None) and (not reading_is_ready):
 
         read_function, function_space_info, time_points = read_field_from_binary(
         file_name, mesh_file, functional_data_dictionary, time_step=
@@ -593,6 +594,14 @@ None, verbose=True):
 
         visualization_copy_file.close()
 
+    # Verifies if the name of the visualization file is to be returned
+
+    if return_visualization_file_name:
+
+        return visualization_copy_file, copy_file_name
+
+    # Otherwise, returns the visualization copy plainly
+
     return visualization_copy_file
 
 ########################################################################
@@ -607,7 +616,9 @@ None, verbose=True):
 def read_field_from_binary(field_file, mesh_file, function_space_info,
 directory_path=None, code_given_field_name=None, comm_object=None,
 code_given_mesh_data_class=None, time_step=0, rename_function=True,
-return_functional_data_class=False, verbose=True):
+return_functional_data_class=False, verbose=True, 
+data_matrix_has_time_point_per_row=True, save_to_xdmf=False,
+return_visualization_file_name=False):
     
     if comm_object is not None:
 
@@ -833,15 +844,34 @@ return_functional_data_class=False, verbose=True):
 
         data = np.load(field_file)
 
+        # Sets a variable that off-sets the measure of the number of 
+        # columns of the data matrix
+
+        columns_number_offset = 0
+
+        # If the data matrix contains the time point value for each row,
+        # the off-set of the number of columns must be 1 and, thus, the
+        # number of DOFs must be equal to the number of columns of the
+        # data matrix minus one
+
+        if data_matrix_has_time_point_per_row:
+
+            columns_number_offset = 1
+
         # Verifies if the data has the same number of DOFs
 
-        if (data.shape[1]-1)!=DOFs_number:
+        if (data.shape[1]-columns_number_offset)!=DOFs_number:
 
             raise IndexError("The read data from '"+str(field_file)+"'"+
             " has "+str(data.shape[1]-1)+" DOFs, whereas the request"+
             "ed function space has "+str(DOFs_number)+" DOFs. Thus, re"+
             "ading the former onto the latter is impossible using 'rea"+
-            "d_field_from_binary'")
+            "d_field_from_binary'.\nThe flag 'data_matrix_has_time_poi"+
+            "nt_per_row' is "+str(data_matrix_has_time_point_per_row)+
+            ". If this flag is True, the data matrix is supposed to ha"+
+            "ve a number of columns equal to the number of DOFs plus o"+
+            "ne. The shift by one is due to the addition of the time p"+
+            "oints values in the first column")
 
         # If the time step is a list with multiple time steps
 
@@ -854,7 +884,7 @@ return_functional_data_class=False, verbose=True):
                 # Gets the current time step (integer index)
 
                 function_space_info.monolithic_solution.vector(
-                ).set_local(data[step, 1:])
+                ).set_local(data[step, columns_number_offset:])
 
                 # Finalizes the vector with the method to insert a new 
                 # data point instead of accumulating it
@@ -880,7 +910,7 @@ return_functional_data_class=False, verbose=True):
         else:
 
             function_space_info.monolithic_solution.vector().set_local(
-            data[time_step,1:])
+            data[time_step,columns_number_offset:])
 
             # Finalizes the vector with the method to insert a new data
             # point instead of accumulating it
@@ -915,15 +945,52 @@ return_functional_data_class=False, verbose=True):
         # And the list of time points the proper time value
 
         time_points = time_points[0]
+
+    # Initializes the tuple that will be returned as a list
+
+    return_tuple = []
+
+    # If the solution is to be saved as a xdmf file as well (which can 
+    # be used later for taking snapshots, for example)
+
+    xdmf_field_file = None
+
+    if save_to_xdmf:
+
+        # Creates a name for the xdmf file by taking the .npy extension
+        # out and putting in the .xdmf extension
+
+        xdmf_field_file = take_outFileNameTermination(field_file)+".xdmf"
+
+        # Saves a visualization copy. Sets the flag 'reading_is_ready' 
+        # to True since the file has already been read and the function
+        # space is ready for plotting in a xdmf file
+
+        _, xdmf_field_file = write_visualization_copy(None, 
+        xdmf_field_file, mesh_file, code_given_field_name, time_step=
+        time_points[-1], comm_object=comm_object, original_function=
+        solutions_across_time_steps, verbose=verbose, reading_is_ready=
+        True, return_visualization_file_name=True)
     
     # If the functional data class is to be spit out too
 
     if return_functional_data_class:
 
-        return solutions_across_time_steps, function_space_info, time_points
+        return_tuple.extend([solutions_across_time_steps, 
+        function_space_info, time_points])
 
     # Returns the function
 
     else:
 
-        return solutions_across_time_steps, time_points
+        return_tuple.extend([solutions_across_time_steps, time_points])
+
+    # If the visualization file name is to be returned
+
+    if return_visualization_file_name:
+
+        return_tuple.append(xdmf_field_file)
+
+    # Returns the tuple as a tuple
+
+    return tuple(return_tuple)

{davout-0.1.1.dev157 → davout-0.1.1.dev159/source/Davout.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: Davout
-Version: 0.1.1.dev157
+Version: 0.1.1.dev159
 Summary: Scientific computing utilities combining FEM, ANN and tooling
 Author: Matheus Janczkowski
 Author-email: Matheus Janczkowski <matheusj2009@hotmail.com>