Davout 0.1.1.dev148__tar.gz → 0.1.1.dev150__tar.gz

{davout-0.1.1.dev148/source/Davout.egg-info → davout-0.1.1.dev150}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: Davout
-Version: 0.1.1.dev148
+Version: 0.1.1.dev150
 Summary: Scientific computing utilities combining FEM, ANN and tooling
 Author: Matheus Janczkowski
 Author-email: Matheus Janczkowski <matheusj2009@hotmail.com>

{davout-0.1.1.dev148 → davout-0.1.1.dev150}/pyproject.toml

@@ -63,7 +63,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "Davout"
-version = "0.1.1.dev148"
+version = "0.1.1.dev150"
 description = "Scientific computing utilities combining FEM, ANN and tooling"
 readme = "README.md"
 authors = [

{davout-0.1.1.dev148 → davout-0.1.1.dev150}/source/Davout/DeepMech/aa_tests/tests_tool_box/test_SVD_based_NNs.py

@@ -569,7 +569,8 @@ class TestSVDArchitecture:
 
             evaluate_function_performance(evaluation_function, 
             n_warm_up_runs=n_warm_up_runs, n_evaluation_runs=
-            n_evaluation_runs, n_evaluations_to_show_memory_data=10)
+            n_evaluation_runs, n_evaluations_to_show_memory_data=10,
+            evaluate_gpu_memory=True)
 
             print("\nThe architecture of the whole network of the last"+
             " experiment is:\n"+string_of_neurons_auxiliar_network+

{davout-0.1.1.dev148 → davout-0.1.1.dev150}/source/Davout/DeepMech/custom_architectures/SVD_quotient_space.py

@@ -1333,9 +1333,9 @@ class SVDQuotientSpace:
     def get_householder_vector_from_parameters(self, 
     householder_first_index, householder_length, 
     householder_number_of_leading_zeros, 
-    #split_householder_DOFs,
+    split_householder_DOFs,
     #householder_parameters, 
-    householder_dofs_tensor_array,
+    #householder_dofs_tensor_array,
     householder_reflector_index, input_dimensionality):
         
         # Gets the first and last indices of the Householder vector, 
@@ -1353,10 +1353,10 @@ class SVDQuotientSpace:
         raw_vector = tf.slice(householder_parameters, [initial_index], [
         length])#"""
 
-        #raw_vector = split_householder_DOFs[householder_reflector_index]
+        raw_vector = split_householder_DOFs[householder_reflector_index]
 
         # TensorArray
-        #"""
+        """
         raw_vector = householder_dofs_tensor_array.read(
         householder_reflector_index) #"""
 
@@ -1386,9 +1386,9 @@ class SVDQuotientSpace:
     def multiply_input_vector_by_householder_reflector(self, 
     input_vector, householder_reflector_index, householder_first_index, 
     householder_length, householder_number_of_leading_zeros, 
-    #split_householder_DOFs,
+    split_householder_DOFs,
     #householder_parameters_orthogonal_matrix, 
-    householder_dofs_tensor_array,
+    #householder_dofs_tensor_array,
     input_dimensionality):
         
         # Gets the Householder vector from the Householder parameters of
@@ -1400,9 +1400,9 @@ class SVDQuotientSpace:
         ) = self.method_for_householder_vector_from_parameters(
         householder_first_index, householder_length, 
         householder_number_of_leading_zeros, 
-        #split_householder_DOFs,
+        split_householder_DOFs,
         #householder_parameters_orthogonal_matrix, 
-        householder_dofs_tensor_array,
+        #householder_dofs_tensor_array,
         householder_reflector_index, input_dimensionality)
 
         # Multiplies the input vector by the Householder reflector (the
@@ -1432,12 +1432,12 @@ class SVDQuotientSpace:
         #split_householder_DOFs = tf.split(
         #householder_parameters_orthogonal_matrix, householder_length)
 
-        """
+        #"""
         split_householder_DOFs = tf.RaggedTensor.from_row_lengths(
         householder_parameters_orthogonal_matrix,
         householder_length)#"""
 
-        #"""
+        """
         householder_dofs_tensor_array = self.split_flat_tensor_into_householder_dofs(
         householder_parameters_orthogonal_matrix, householder_length)#"""
 
@@ -1450,9 +1450,9 @@ class SVDQuotientSpace:
             accumulator_vector, householder_reflector_index, 
             householder_first_index, householder_length, 
             householder_number_of_leading_zeros, 
-            #split_householder_DOFs,
+            split_householder_DOFs,
             #householder_parameters_orthogonal_matrix,
-            householder_dofs_tensor_array,
+            #householder_dofs_tensor_array,
             input_dimensionality)
 
         # Uses foldl to perform the chain multiplication

{davout-0.1.1.dev148 → davout-0.1.1.dev150}/source/Davout/PythonicUtilities/testing_tools.py

@@ -153,7 +153,8 @@ sort_methods_alphabetically=True):
 
 def evaluate_function_performance(function_object, n_warm_up_runs=1,
 n_evaluation_runs=10, confidence_level=0.95, 
-n_evaluations_to_show_memory_data=None, ten_notation=" E"):
+n_evaluations_to_show_memory_data=None, ten_notation=" E", gpu_device=
+"GPU:0", evaluate_gpu_memory=False):
 
     # Instantiates the class of colors in the terminal
 
@@ -212,6 +213,14 @@ n_evaluations_to_show_memory_data=None, ten_notation=" E"):
 
     memory_peak_during_evaluation = []
 
+    # Initializes lists to measure GPU memory usage 
+
+    GPU_memory_usage_before_evaluation = []
+
+    GPU_memory_usage_after_evaluation = []
+
+    GPU_memory_peak_during_evaluation = []
+
     # Evaluates the running time
 
     start_general_time = time.perf_counter()
@@ -229,6 +238,21 @@ n_evaluations_to_show_memory_data=None, ten_notation=" E"):
         memory_usage_before_evaluation.append(process_data.memory_info(
         ).rss/(1024**2))
 
+        # If GPU memory is to be tracked
+
+        if evaluate_gpu_memory:
+
+            tf.config.experimental.reset_memory_stats(gpu_device)
+
+            gpu_memory_before = tf.config.experimental.get_memory_info(
+            gpu_device)
+
+            # Converts the GPU memory assessed before the function eval-
+            # uation to MB
+
+            GPU_memory_usage_before_evaluation.append(gpu_memory_before[
+            "current"]/(1024**2))
+
         # Starts to count this step's time interval
 
         start_interval = time.perf_counter()
@@ -259,6 +283,22 @@ n_evaluations_to_show_memory_data=None, ten_notation=" E"):
 
         time_intervals_list.append(time.perf_counter()-start_interval)
 
+        # If GPU memory is to be tracked
+        
+        if evaluate_gpu_memory:
+
+            gpu_memory_after = tf.config.experimental.get_memory_info(
+            gpu_device)
+
+            # Converts the GPU memory assessed after the function eval-
+            # uation and the peak memory usage to MB
+
+            GPU_memory_usage_after_evaluation.append(gpu_memory_after[
+            "current"]/(1024**2))
+
+            GPU_memory_peak_during_evaluation.append(gpu_memory_after[
+            "peak"]/(1024**2))
+
     # Gets the final general time
 
     general_time_interval = time.perf_counter()-start_general_time
@@ -357,6 +397,58 @@ n_evaluations_to_show_memory_data=None, ten_notation=" E"):
     memory_peak_during_evaluation)), decimal_places=5, ten_notation=
     ten_notation, n_digits_for_power_of_ten=2)+" MB")    
 
+    # If GPU memory data is to be shown
+
+    if evaluate_gpu_memory:
+
+        # Cleans the string that prints the result
+
+        string_result = ""
+
+        # Iterates over the evaluation runs
+        
+        for evaluation_number in range(
+        n_evaluations_to_show_memory_data):
+    
+            memory_before = GPU_memory_usage_before_evaluation[
+            evaluation_number]
+    
+            memory_after = GPU_memory_usage_after_evaluation[
+            evaluation_number]
+    
+            memory_peak = GPU_memory_peak_during_evaluation[
+            evaluation_number]
+    
+            string_result += ("\n         "+float_to_scientific_notation(
+            memory_before, decimal_places=5, ten_notation=ten_notation, 
+            n_digits_for_power_of_ten=2)+" MB        |          "+
+            float_to_scientific_notation(memory_after, decimal_places=5, 
+            ten_notation=ten_notation, n_digits_for_power_of_ten=2)+" MB  "+
+            "      |         "+float_to_scientific_notation(memory_peak, 
+            decimal_places=5, ten_notation=ten_notation, 
+            n_digits_for_power_of_ten=2)+" MB")
+    
+        print("\n########################### GPU Memory usage ########"+
+        "###################\n")   
+    
+        print("Function '"+function_name+"' was evaluated for "+str(
+        n_warm_up_runs)+" warm-up runs and\nfor "+str(n_evaluation_runs
+        )+" evaluation runs.\n\n  Memory before evaluation    |    Mem"+
+        "ory after evaluation    | Peak memory during evaluation"+
+        string_result+"\n\n    Average memory before     |      Averag"+
+        "e memory after     |    Average peak memory during\n         "+
+        float_to_scientific_notation(np.mean(np.asarray(
+        GPU_memory_usage_before_evaluation)), decimal_places=5, 
+        ten_notation=ten_notation, n_digits_for_power_of_ten=2)+" MB  "+
+        "      |          "+
+        float_to_scientific_notation(np.mean(np.asarray(
+        GPU_memory_usage_after_evaluation)), decimal_places=5, 
+        ten_notation=ten_notation, n_digits_for_power_of_ten=2)+" MB  "+
+        "      |         "+
+        float_to_scientific_notation(np.mean(np.asarray(
+        GPU_memory_peak_during_evaluation)), decimal_places=5, 
+        ten_notation=ten_notation, n_digits_for_power_of_ten=2)+" MB")    
+
     # Collects all memory garbage to avoid any information leakage that
     # might affect future computations
 

{davout-0.1.1.dev148 → davout-0.1.1.dev150/source/Davout.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: Davout
-Version: 0.1.1.dev148
+Version: 0.1.1.dev150
 Summary: Scientific computing utilities combining FEM, ANN and tooling
 Author: Matheus Janczkowski
 Author-email: Matheus Janczkowski <matheusj2009@hotmail.com>