refactor(NeuralNetwork): remove unused multiprocessing parameter

- Removed multiprocessing parameter from NeuralNetwork class. - Updated related documentation and files to reflect this change.
frankkramer-lab · Aug 12, 2024 · 23c78e8 · 23c78e8
1 parent 8e5c66e
commit 23c78e8
Show file tree

Hide file tree

Showing 7 changed files with 4 additions and 20 deletions.
diff --git a/aucmedi/automl/block_pred.py b/aucmedi/automl/block_pred.py
@@ -75,7 +75,6 @@ def block_predict(config):
     nn_paras = {"n_labels": 1,                                  # placeholder
                 "channels": 1,                                  # placeholder
                 "batch_queue_size": 4,
-                "multiprocessing": False,
     }
     # Select input shape for 3D
     if meta_training["three_dim"]:

diff --git a/aucmedi/automl/block_train.py b/aucmedi/automl/block_train.py
@@ -140,7 +140,6 @@ def block_train(config):
                 "loss": loss,
                 "metrics": [AUC(100), F1Score(average="macro")],
                 "pretrained_weights": True,
-                "multiprocessing": False,
     }
     # Select input shape for 3D
     if config["three_dim"] : nn_paras["input_shape"] = config["shape_3D"]

diff --git a/aucmedi/ensemble/bagging.py b/aucmedi/ensemble/bagging.py
@@ -187,7 +187,6 @@ def train(self, training_generator, epochs=20, iterations=None,
                 "meta_variables": self.model_template.meta_variables,
                 "learning_rate": self.model_template.learning_rate,
                 "batch_queue_size": self.model_template.batch_queue_size,
-                "multiprocessing": self.model_template.multiprocessing,
             }
 
             # Gather DataGenerator parameters
@@ -324,7 +323,6 @@ def predict(self, prediction_generator, aggregate="mean",
                 "meta_variables": self.model_template.meta_variables,
                 "learning_rate": self.model_template.learning_rate,
                 "batch_queue_size": self.model_template.batch_queue_size,
-                "multiprocessing": self.model_template.multiprocessing,
             }
 
             # Start inference process for fold i

diff --git a/aucmedi/ensemble/composite.py b/aucmedi/ensemble/composite.py
@@ -244,7 +244,6 @@ def train(self, training_generator, epochs=20, iterations=None,
                 "meta_variables": self.model_list[i].meta_variables,
                 "learning_rate": self.model_list[i].learning_rate,
                 "batch_queue_size": self.model_list[i].batch_queue_size,
-                "multiprocessing": self.model_list[i].multiprocessing,
             }
 
             # Gather DataGenerator parameters
@@ -354,7 +353,6 @@ def train_metalearner(self, training_generator):
                 "meta_variables": self.model_list[i].meta_variables,
                 "learning_rate": self.model_list[i].learning_rate,
                 "batch_queue_size": self.model_list[i].batch_queue_size,
-                "multiprocessing": self.model_list[i].multiprocessing,
             }
 
             # Gather DataGenerator parameters
@@ -467,7 +465,6 @@ def predict(self, prediction_generator, return_ensemble=False):
                 "meta_variables": self.model_list[i].meta_variables,
                 "learning_rate": self.model_list[i].learning_rate,
                 "batch_queue_size": self.model_list[i].batch_queue_size,
-                "multiprocessing": self.model_list[i].multiprocessing,
             }
 
             # Gather DataGenerator parameters

diff --git a/aucmedi/ensemble/stacking.py b/aucmedi/ensemble/stacking.py
@@ -232,7 +232,6 @@ def train(self, training_generator, epochs=20, iterations=None,
                 "meta_variables": self.model_list[i].meta_variables,
                 "learning_rate": self.model_list[i].learning_rate,
                 "batch_queue_size": self.model_list[i].batch_queue_size,
-                "multiprocessing": self.model_list[i].multiprocessing,
             }
 
             # Gather DataGenerator parameters
@@ -342,7 +341,6 @@ def train_metalearner(self, training_generator):
                 "meta_variables": self.model_list[i].meta_variables,
                 "learning_rate": self.model_list[i].learning_rate,
                 "batch_queue_size": self.model_list[i].batch_queue_size,
-                "multiprocessing": self.model_list[i].multiprocessing,
             }
 
             # Gather DataGenerator parameters
@@ -454,7 +452,6 @@ def predict(self, prediction_generator, return_ensemble=False):
                 "meta_variables": self.model_list[i].meta_variables,
                 "learning_rate": self.model_list[i].learning_rate,
                 "batch_queue_size": self.model_list[i].batch_queue_size,
-                "multiprocessing": self.model_list[i].multiprocessing,
             }
 
             # Gather DataGenerator parameters

diff --git a/aucmedi/neural_network/model.py b/aucmedi/neural_network/model.py
@@ -138,7 +138,7 @@ def __init__(self, n_labels, channels, input_shape=None, architecture=None,
                  pretrained_weights=False, loss="categorical_crossentropy",
                  metrics=["categorical_accuracy"], activation_output="softmax",
                  fcl_dropout=True, meta_variables=None, learning_rate=0.0001,
-                 batch_queue_size=10, multiprocessing=False,
+                 batch_queue_size=10,
                  verbose=1):
         """ Initialization function for creating a Neural Network (model) object.
 
@@ -169,7 +169,6 @@ def __init__(self, n_labels, channels, input_shape=None, architecture=None,
                                                     ([Classifier][aucmedi.neural_network.architectures.classifier]).
             learning_rate (float):                  Learning rate in which weights of the neural network will be updated.
             batch_queue_size (int):                 The batch queue size is the number of previously prepared batches in the cache during runtime.
-            multiprocessing (bool):                 Option whether to utilize multi-processing for workers instead of threading .
             verbose (int):                          Option (0/1) how much information should be written to stdout.
 
         ???+ danger
@@ -191,7 +190,6 @@ def __init__(self, n_labels, channels, input_shape=None, architecture=None,
         self.metrics = metrics
         self.learning_rate = learning_rate
         self.batch_queue_size = batch_queue_size
-        self.multiprocessing = multiprocessing
         self.pretrained_weights = pretrained_weights
         self.activation_output = activation_output
         self.fcl_dropout = fcl_dropout
@@ -296,7 +294,6 @@ def train(self, training_generator, validation_generator=None, epochs=20,
                                      callbacks=callbacks, epochs=epochs,
                                      steps_per_epoch=iterations,
                                      class_weight=class_weights,
-                                     use_multiprocessing=self.multiprocessing,
                                      max_queue_size=self.batch_queue_size,
                                      verbose=self.verbose)
             # Return logged history object
@@ -318,7 +315,6 @@ def train(self, training_generator, validation_generator=None, epochs=20,
                                            epochs=self.tf_epochs,
                                            steps_per_epoch=iterations,
                                            class_weight=class_weights,
-                                           use_multiprocessing=self.multiprocessing,
                                            max_queue_size=self.batch_queue_size,
                                            verbose=self.verbose)
             # Unfreeze base model layers again
@@ -334,7 +330,6 @@ def train(self, training_generator, validation_generator=None, epochs=20,
                                          initial_epoch=self.tf_epochs,
                                          steps_per_epoch=iterations,
                                          class_weight=class_weights,
-                                         use_multiprocessing=self.multiprocessing,
                                          max_queue_size=self.batch_queue_size,
                                          verbose=self.verbose)
             # Combine logged history objects
@@ -365,7 +360,6 @@ def predict(self, prediction_generator):
         # Run inference process with the Keras predict function
         preds = self.model.predict(prediction_generator,
                                    max_queue_size=self.batch_queue_size,
-                                   use_multiprocessing=self.multiprocessing,
                                    verbose=self.verbose)
         # Output predictions results
         return preds

diff --git a/examples/applications/xray_pneumonia.ipynb b/examples/applications/xray_pneumonia.ipynb
@@ -5306,7 +5306,7 @@
         "  \n",
         "model = model.NeuralNetwork(n_labels=2, channels=3, input_shape=(224, 224, 3), architecture = architecture, pretrained_weights=True, \n",
         "                             loss=\"categorical_crossentropy\", metrics=[\"categorical_accuracy\"], activation_output=\"softmax\", \n",
-        "                             fcl_dropout=True, learninig_rate=0.0001, batch_queue_size=10, multiprocessing=False, verbose=1)"
+        "                             fcl_dropout=True, learninig_rate=0.0001, batch_queue_size=10, verbose=1)"
       ]
     },
     {
@@ -7419,7 +7419,7 @@
         "architecture = architectures.MobileNetV2(channels = 3)\n",
         "model = model.NeuralNetwork(n_labels=2, channels=3, input_shape=(224, 224, 3), architecture = architecture, pretrained_weights=True, \n",
         "                             loss=\"categorical_crossentropy\", metrics=[\"categorical_accuracy\"], activation_output=\"softmax\", \n",
-        "                             fcl_dropout=True, learninig_rate=0.0001, batch_queue_size=10, multiprocessing=False, verbose=1)\n",
+        "                             fcl_dropout=True, learninig_rate=0.0001, batch_queue_size=10, verbose=1)\n",
         "```\n",
         "\n",
         "The imports needed for this part are:\n",
@@ -10832,7 +10832,7 @@
         "#architecture = architectures.MobileNetV2(channels = 3)\n",
         "#model = model.NeuralNetwork(n_labels=2, channels=3, input_shape=(224, 224, 3), architecture = architecture, pretrained_weights=True, \n",
         "#                             loss=\"categorical_crossentropy\", metrics=[\"categorical_accuracy\"], activation_output=\"softmax\", \n",
-        "#                             fcl_dropout=True, learninig_rate=0.0001, batch_queue_size=10, multiprocessing=False, verbose=1)\n",
+        "#                             fcl_dropout=True, learninig_rate=0.0001, batch_queue_size=10, verbose=1)\n",
         "\n",
         "model.load(file_path='/content/drive/MyDrive/Models/fold1.h5')\n",
         "pred_one_image = model.predict(prediction_generator=img)\n",