How to use disable_debug_info method in Slash

Best Python code snippet using slash

CMakeCompiler.py

Source:CMakeCompiler.py

1# ----------------------------------------------------------------------2# |3# |  CMakeCompiler.py4# |5# |  David Brownell <db@DavidBrownell.com>6# |      2019-04-04 08:55:247# |8# ----------------------------------------------------------------------9# |10# |  Copyright David Brownell 2019-2211# |  Distributed under the Boost Software License, Version 1.0. See12# |  accompanying file LICENSE_1_0.txt or copy at13# |  http://www.boost.org/LICENSE_1_0.txt.14# |15# ----------------------------------------------------------------------16"""Contains the Compiler object and entry point"""1718import os19import re20import shutil21import sys22import textwrap2324import CommonEnvironment25from CommonEnvironment.CallOnExit import CallOnExit26from CommonEnvironment import CommandLine27from CommonEnvironment import FileSystem28from CommonEnvironment import Interface29from CommonEnvironment import Process30from CommonEnvironment.Shell.All import CurrentShell31from CommonEnvironment.StreamDecorator import StreamDecorator32from CommonEnvironment import StringHelpers3334from CommonEnvironment.CompilerImpl import Compiler as CompilerMod35from CommonEnvironment.CompilerImpl.InputProcessingMixin.IndividualInputProcessingMixin import (36    IndividualInputProcessingMixin,37)38from CommonEnvironment.CompilerImpl.InvocationMixin.CommandLineInvocationMixin import (39    CommandLineInvocationMixin,40)41from CommonEnvironment.CompilerImpl.InvocationQueryMixin.ConditionalInvocationQueryMixin import (42    ConditionalInvocationQueryMixin,43)44from CommonEnvironment.CompilerImpl.OutputMixin.MultipleOutputMixin import (45    MultipleOutputMixin,46)4748from CommonEnvironment.TypeInfo.FundamentalTypes.DirectoryTypeInfo import (49    DirectoryTypeInfo,50)5152# ----------------------------------------------------------------------53_script_fullpath                            = CommonEnvironment.ThisFullpath()54_script_dir, _script_name                   = os.path.split(_script_fullpath)55# ----------------------------------------------------------------------5657# ----------------------------------------------------------------------58@Interface.staticderived59class Compiler(60    IndividualInputProcessingMixin,61    CommandLineInvocationMixin,62    ConditionalInvocationQueryMixin,63    MultipleOutputMixin,64    CompilerMod.Compiler,65):66    """Compiles a CMAKE directory"""6768    # ----------------------------------------------------------------------69    # |  Properties70    Name                                    = Interface.DerivedProperty("CMake")71    Description                             = Interface.DerivedProperty("Compiles a CMake directory.")72    InputTypeInfo                           = Interface.DerivedProperty(DirectoryTypeInfo())7374    # ----------------------------------------------------------------------75    # |  Methods76    @staticmethod77    @Interface.override78    def IsSupportedContent(filename):79        return os.path.isfile(os.path.join(filename, "CMakeLists.txt"))8081    # ----------------------------------------------------------------------82    @staticmethod83    @Interface.override84    def IsSupportedTestItem(item):85        return item.endswith("Tests")8687    # ----------------------------------------------------------------------88    @staticmethod89    @Interface.override90    def CreateInvokeCommandLine(context, verbose_stream):91        return 'cmake --build "{build}"'.format(92            build=context["output_dir"],93        )9495    # ----------------------------------------------------------------------96    @staticmethod97    @Interface.override98    def RemoveTemporaryArtifacts(context):99        output_dir = context["output_dir"]100101        # Move GCC-generated profile data to the output dir102        for filename in FileSystem.WalkFiles(103            output_dir,104            include_file_extensions=[".gcno", ".gcda"],105        ):106            dest_filename = os.path.join(output_dir, os.path.basename(filename))107            if dest_filename == filename:108                continue109110            if not os.path.isfile(dest_filename):111                shutil.copyfile(filename, dest_filename)112113        for potential_dir in ["CMakeFiles", "Testing"]:114            potential_dir = os.path.join(output_dir, potential_dir)115            FileSystem.RemoveTree(potential_dir)116117        for potential_file in ["CMakeCache.txt", "cmake_install.cmake", "Makefile"]:118            potential_file = os.path.join(output_dir, potential_file)119            FileSystem.RemoveFile(potential_file)120121        remove_extensions = set([".ilk"])122123        for item in os.listdir(output_dir):124            if os.path.splitext(item)[1] not in remove_extensions:125                continue126127            fullpath = os.path.join(output_dir, item)128            FileSystem.RemoveFile(fullpath)129130    # ----------------------------------------------------------------------131    @staticmethod132    @Interface.override133    def ExecuteExclusively(context):134        # Don't allow the parallel execution of cmake files, as each of them135        # internally will compile on all available threads.136        return True137138    # ----------------------------------------------------------------------139    # ----------------------------------------------------------------------140    # ----------------------------------------------------------------------141    @classmethod142    @Interface.override143    def _GetOptionalMetadata(cls):144        return [145            (146                "generator",147                None148                if os.getenv("DEVELOPMENT_ENVIRONMENT_CPP_USE_DEFAULT_CMAKE_GENERATOR")149                else "Ninja",150            ),151            ("is_debug", True),152            ("cmake_debug_output", False),153            ("use_unicode", False),154            (155                "static_crt",156                False if os.getenv("DEVELOPMENT_ENVIRONMENT_CPP_NO_STATIC_CRT") else True,157            ),158            ("is_profile", False),159            ("is_benchmark", False),160            ("disable_debug_info", False),161            ("disable_aslr", False),162        ] + super(Compiler, cls)._GetOptionalMetadata()163164    # ----------------------------------------------------------------------165    @classmethod166    @Interface.override167    def _GetRequiredContextNames(cls):168        return ["output_dir"] + super(Compiler, cls)._GetRequiredContextNames()169170    # ----------------------------------------------------------------------171    @classmethod172    @Interface.override173    def _CreateContext(cls, metadata, status_stream):174        if "output_dir" not in metadata:175            return metadata176177        # Invoke cmake to get a list of the generated files. The best way that I have found178        # to do this is to parse generated dot files (as we don't want to get into the business179        # of parsing cmake files).180        temp_directory = CurrentShell.CreateTempDirectory()181        with CallOnExit(lambda: FileSystem.RemoveTree(temp_directory)):182            dot_filename = os.path.join(temp_directory, "generated.dot")183184            command_line_options = [185                '-S "{}"'.format(metadata["input"]),186                '-B "{}"'.format(metadata["output_dir"]),187                '"--graphviz={}"'.format(dot_filename),188                "-DCMAKE_BUILD_TYPE={}".format(189                    "Debug" if metadata["is_debug"] else "Release",190                ),191                "-DCppCommon_CMAKE_DEBUG_OUTPUT={}".format(192                    "ON" if metadata["cmake_debug_output"] else "OFF",193                ),194                "-DCppCommon_UNICODE={}".format(195                    "ON" if metadata["use_unicode"] else "OFF",196                ),197                "-DCppCommon_STATIC_CRT={}".format(198                    "ON" if metadata["static_crt"] else "OFF",199                ),200                "-DCppCommon_CODE_COVERAGE={}".format(201                    "ON" if metadata["is_profile"] else "OFF",202                ),203                "-DCppCommon_NO_DEBUG_INFO={}".format(204                    "ON" if metadata["disable_debug_info"] else "OFF",205                ),206                "-DCppCommon_NO_ADDRESS_SPACE_LAYOUT_RANDOMIZATION={}".format(207                    "ON" if metadata["disable_aslr"] else "OFF",208                ),209            ]210211            if metadata["generator"]:212                command_line_options.append('-G "{}"'.format(metadata["generator"]))213214            result, output = Process.Execute(215                "cmake {}".format(" ".join(command_line_options)),216            )217            if result != 0:218                raise Exception(219                    textwrap.dedent(220                        """\221                        cmake failed ({}):222                            {}223                        """,224                    ).format(result, StringHelpers.LeftJustify(output, 4)),225                )226227            # Parse the dot file. This regular expression has been configured to work228            # with dot files generated by:229            #230            #   - CMake 3.13.4231            #   - CMake 3.17.2232            #233            regex = re.compile(234                r"""(?#235                node                        )\"node\d+\"\s*(?#236                lbracket                    )\[\s*(?#237                label key                   )label\s*=\s*(?#238                name                        )\"(?P<name>.+?)\"(?#239                [optional] comma delimiter  ),?\s+(?#240                shape key                   )shape\s*=\s*(?#241                value                       )(?:\"house\"|egg)\s*(?#242                rbracket                    )\](?#243                terminator                  );(?#244                )""",245            )246247            with open(dot_filename) as f:248                content = f.read()249250            output_filenames = []251252            for match in regex.finditer(content):253                output_filenames.append(254                    os.path.join(255                        metadata["output_dir"],256                        CurrentShell.CreateExecutableName(match.group("name")),257                    ),258                )259260            metadata["output_filenames"] = output_filenames261262        return super(Compiler, cls)._CreateContext(metadata, status_stream)263264265# ----------------------------------------------------------------------266@CommandLine.EntryPoint267@CommandLine.Constraints(268    output_dir=CommandLine.DirectoryTypeInfo(269        ensure_exists=False,270    ),271    input=CommandLine.DirectoryTypeInfo(272        arity="+",273    ),274    generator=CommandLine.StringTypeInfo(275        arity="?",276    ),277    output_stream=None,278)279def Compile(280    output_dir,281    input,282    generator=None,283    release=False,284    cmake_debug_output=False,285    unicode=False,286    no_static_crt=False,287    profile=False,288    benchmark=False,289    disable_debug_info=False,290    disable_aslr=False,291    output_stream=sys.stdout,292    verbose=False,293):294    """Compiles a CMake directory"""295296    inputs = input297    del input298299    return CompilerMod.CommandLineCompile(300        Compiler,301        inputs,302        StreamDecorator(output_stream),303        verbose,304        output_dir=output_dir,305        is_debug=not release,306        cmake_debug_output=cmake_debug_output,307        use_unicode=unicode,308        static_crt=not no_static_crt,309        is_profile=profile,310        is_benchmark=benchmark,311        disable_debug_info=disable_debug_info,312        disable_aslr=disable_aslr,313    )314315316# ----------------------------------------------------------------------317# ----------------------------------------------------------------------318# ----------------------------------------------------------------------319if __name__ == "__main__":320    try:321        sys.exit(CommandLine.Main())322    except KeyboardInterrupt:
...

fine tuning.py

Source:fine tuning.py

1import tensorflow as tf2import numpy as np3from images import load_images45X_train, y_train, X_test, y_test, X_val, y_val = load_images(shuffle_before=True)6input_shape = X_train.shape[1:]78disable_debug_info = False9if disable_debug_info:10    import os11    os.environ['TF_CPP_MIN_LOG_LEVEL'] = '1'121314def new_design_va80(augment_data=False, tune_hypers=False, learning_rate=3.4e-4, batch_size=26, return_results=False):1516    initializer = tf.keras.initializers.VarianceScaling(2)1718    layers = [19        tf.keras.layers.Conv2D(64, 3, 1, 'same', input_shape=input_shape, kernel_initializer=initializer, activation='relu'),20        tf.keras.layers.BatchNormalization(),21        tf.keras.layers.Conv2D(64, 3, 1, 'same', kernel_initializer=initializer, activation='relu'),2223        tf.keras.layers.MaxPooling2D(),24        tf.keras.layers.BatchNormalization(),2526        tf.keras.layers.Conv2D(128, 3, 1, 'same', kernel_initializer=initializer, activation='relu'),27        tf.keras.layers.BatchNormalization(),28        tf.keras.layers.Conv2D(128, 3, 1, 'same', kernel_initializer=initializer, activation='relu'),2930        tf.keras.layers.MaxPooling2D(),31        tf.keras.layers.BatchNormalization(),3233        tf.keras.layers.Conv2D(256, 3, 1, 'same', kernel_initializer=initializer, activation='relu'),34        tf.keras.layers.BatchNormalization(),35        tf.keras.layers.Conv2D(256, 3, 1, 'same', kernel_initializer=initializer, activation='relu'),3637        tf.keras.layers.MaxPooling2D(),38        tf.keras.layers.BatchNormalization(),3940        tf.keras.layers.Conv2D(128, 3, 1, 'same', kernel_initializer=initializer, activation='relu'),41        tf.keras.layers.BatchNormalization(),42        tf.keras.layers.Conv2D(64, 3, 1, 'same', kernel_initializer=initializer, activation='relu'),4344        tf.keras.layers.MaxPooling2D(),45        tf.keras.layers.BatchNormalization(),4647        tf.keras.layers.Conv2D(8, 3, 1, 'same', kernel_initializer=initializer, activation='relu'),48        tf.keras.layers.BatchNormalization(),4950        # tf.keras.layers.Conv2D(6, 3, 1, 'same', kernel_initializer=initializer, activation='relu'),51        # In kernel_initializer, the 2 is removed so that the initialization returns to Xavier initialization instead of52        # Kaiming initialization. Xaiver initialization should in theory work better since Kaiming initialization53        # is meant for ReLU, but the final activation is softmax. However, in practice, it does not make a difference.54        tf.keras.layers.Conv2D(6, 3, 1, 'same', kernel_initializer=tf.keras.initializers.VarianceScaling(), activation='relu'),55        tf.keras.layers.GlobalAveragePooling2D(),56        tf.keras.layers.Activation('softmax')57    ]5859    callback = tf.keras.callbacks.EarlyStopping(monitor='val_loss', patience=3)6061    model = tf.keras.models.Sequential(layers)6263    train_loss = []64    val_loss = []65    train_acc = []66    val_acc = []67    augmentor = tf.keras.preprocessing.image.ImageDataGenerator(horizontal_flip=True, vertical_flip=True)68    optimizer = tf.keras.optimizers.RMSprop()6970    if not tune_hypers:71        for fit in range(2):72            # validation_accuracies are the result of training 7 models independently and averaging validation_accuracies7374            # 48x48 / 64x64-75            # 48x48 images - 81.305933% accuracy76            # 64x64 images - 82.466024% accuracy77            # model.compile(tf.keras.optimizers.RMSprop(learning_rate=2.5e-4 * 10**-fit), loss=tf.keras.validation_losses.sparse_categorical_crossentropy, metrics=['accuracy'])7879            # 100x10080            # 100x100 images - 85.449123% accuracy81            # model.compile(tf.keras.optimizers.RMSprop(learning_rate=6e-4 * 10**-fit), loss=tf.keras.validation_losses.sparse_categorical_crossentropy, metrics=['accuracy'])8283            # 128x12884            # 128x128 images - 88.067615% accuracy85            # model.compile(tf.keras.optimizers.RMSprop(learning_rate=4.5e-4 * 10**-fit), loss=tf.keras.validation_losses.sparse_categorical_crossentropy, metrics=['accuracy'])86            # model.compile(tf.keras.optimizers.RMSprop(learning_rate=2.5e-4 * 10**-fit), loss=tf.keras.validation_losses.sparse_categorical_crossentropy, metrics=['accuracy'])87            # model.compile(tf.keras.optimizers.RMSprop(learning_rate=3e-4 * 5**-fit), loss=tf.keras.validation_losses.sparse_categorical_crossentropy, metrics=['accuracy'])88            # the good one is below89            # model.compile(tf.keras.optimizers.RMSprop(learning_rate=2.5e-4 * 5**-fit), loss=tf.keras.validation_losses.sparse_categorical_crossentropy, metrics=['accuracy'])90            optimizer.learning_rate = 1e-4 * 10**-fit9192            model.compile(optimizer, loss=tf.keras.losses.sparse_categorical_crossentropy, metrics=['accuracy'])9394            # 156x15695            # 156x156 images - 86.54293% accuracy96            # model.compile(tf.keras.optimizers.RMSprop(learning_rate=4.5e-4 * 10**-fit), loss=tf.keras.validation_losses.sparse_categorical_crossentropy, metrics=['accuracy'])9798            if not augment_data:99                # history = model.fit(X_train, y_train, 26, epochs=25, validation_data=(X_val, y_val), verbose=2, callbacks=[callback])100                history = model.fit(X_train, y_train, 26, epochs=16, validation_data=(X_val, y_val), verbose=2, callbacks=[callback])101102                print()103                train_loss.extend(history.history['loss'])104                val_loss.extend(history.history['val_loss'])105                train_acc.extend(history.history['accuracy'])106                val_acc.extend(history.history['val_accuracy'])107108            else:109                highest_val_loss = np.inf110                count = 0111                patience = 3112                num_epochs = 18113                model.compile(optimizer, loss=tf.keras.losses.sparse_categorical_crossentropy, metrics=['accuracy'])114                for epoch in range(num_epochs):115                    print(f'Epoch: {epoch+1}/{num_epochs}')116                    data = augmentor.flow(X_train, y_train, batch_size=20)117                    history = model.fit_generator(data, validation_data=(X_val, y_val), epochs=1, verbose=2)118                    train_loss.extend(history.history['loss'])119                    val_loss.extend(history.history['val_loss'])120                    current_val_loss = history.history['val_loss'][-1]121                    train_acc.extend(history.history['accuracy'])122                    current_val_acc = history.history['val_accuracy'][-1]123                    val_acc.append(current_val_acc)124                    count += 1125                    if highest_val_loss - current_val_loss > 0:126                        count = 0127                        highest_val_loss = current_val_loss128                    if count == patience:129                        break130131                    # test = input.txt('break?')132                    # if test == 'yes':133                    #     break134135                print()136137        test_results = model.evaluate(X_test, y_test)138        print(input_shape)139140        if return_results:141            return test_results[1]142143    else:144        model.compile(learning_rate=learning_rate, loss=tf.keras.losses.sparse_categorical_crossentropy, metrics=['accuracy'])145        history = model.fit(X_train, y_train, batch_size, epochs=12, validation_data=(X_val, y_val), verbose=0)146        train_acc = model.evaluate(X_train, y_train, verbose=0)[1]147        val_acc = model.evaluate(X_val, y_val, verbose=0)[1]148149        return train_acc, val_acc150151    import matplotlib.pyplot as plt152153    fig, axs = plt.subplots(2, 1, constrained_layout=True)154    # fig.suptitle(f'{material}, {label}', fontsize=18)155156    axs[0].plot(train_loss, label='Train loss')157    axs[0].plot(val_loss, label='Validation loss')158    axs[0].set_title('Train and Validation Losses')159    axs[0].set_xlabel('Epoch')160    axs[0].set_ylabel('Loss')161    axs[0].legend(loc='upper right')162163    axs[1].plot(train_acc, label='Train accuracy')164    axs[1].plot(val_acc, label='Validation Accuracy')165    axs[1].set_title('Train and Validation Accuracies')166    axs[1].set_xlabel('Epoch')167    axs[1].set_ylabel('Accuracy')168    axs[1].legend(loc='lower right')169170    plt.show()171172    # plt.plot(train_loss, label='Train loss')173    # plt.plot(val_loss, label='Validation loss')174    # plt.xlabel('Epoch')175    # plt.ylabel('Loss')176    # plt.legend(loc='upper right')177    # plt.show()178179180def random_search(num_iters):181    for _ in range(num_iters):182        lr = np.random.uniform(2e-4, 6e-4)183184        train_acc, val_acc = new_design_va80(False, True, lr)185186        print(f'LR: {lr:e} | Train Accuracy: {train_acc:.5f} | Val Accuracy: {val_acc:.5f}')187        print()188189190def check_mean_accuracy(num_models=4, augment_data=False):191    accuracies = []192    for _ in range(num_models):193        global X_train, y_train, X_test, y_test, X_val, y_val, input_shape194        X_train, y_train, X_test, y_test, X_val, y_val = load_images(shuffle_before=True)195        input_shape = X_train.shape[1:]196        accuracies.append(new_design_va80(augment_data=augment_data, return_results=True))197198    print(np.mean(accuracies))199200201new_design_va80(augment_data=True)
...

final.py

Source:final.py

1import tensorflow as tf2import numpy as np34from images import load_images5tf.config.experimental.set_memory_growth(tf.config.list_physical_devices('GPU')[0], True)6X_train, y_train, X_test, y_test, X_val, y_val = load_images(shuffle_before=True, num_augments=3)  # 4789input_shape = X_train.shape[1:]1011disable_debug_info = False12if disable_debug_info:13    import os14    os.environ['TF_CPP_MIN_LOG_LEVEL'] = '1'151617def create_and_train_TARnet(return_results=False, save_history=False, save_file='TARnet.hdf5'):1819    initializer = tf.keras.initializers.VarianceScaling(2)2021    layers = [22        tf.keras.layers.Conv2D(64, 3, 1, 'same', input_shape=input_shape, kernel_initializer=initializer, activation='relu'),23        tf.keras.layers.BatchNormalization(),24        tf.keras.layers.Conv2D(64, 3, 1, 'same', kernel_initializer=initializer, activation='relu'),2526        tf.keras.layers.MaxPooling2D(),27        tf.keras.layers.BatchNormalization(),2829        tf.keras.layers.Conv2D(128, 3, 1, 'same', kernel_initializer=initializer, activation='relu'),30        tf.keras.layers.BatchNormalization(),31        tf.keras.layers.Conv2D(128, 3, 1, 'same', kernel_initializer=initializer, activation='relu'),3233        tf.keras.layers.MaxPooling2D(),34        tf.keras.layers.BatchNormalization(),3536        tf.keras.layers.Conv2D(128, 3, 1, 'same', kernel_initializer=initializer, activation='relu'),37        tf.keras.layers.BatchNormalization(),38        tf.keras.layers.Conv2D(128, 3, 1, 'same', kernel_initializer=initializer, activation='relu'),3940        tf.keras.layers.MaxPooling2D(),41        tf.keras.layers.BatchNormalization(),4243        tf.keras.layers.Conv2D(128, 3, 1, 'same', kernel_initializer=initializer, activation='relu'),44        tf.keras.layers.BatchNormalization(),45        tf.keras.layers.Conv2D(128, 3, 1, 'same', kernel_initializer=initializer, activation='relu'),4647        # tf.keras.layers.MaxPooling2D(),48        # tf.keras.layers.BatchNormalization(),49        #50        # tf.keras.layers.Conv2D(128, 3, 1, 'same', kernel_initializer=initializer, activation='relu'),51        # tf.keras.layers.BatchNormalization(),52        # tf.keras.layers.Conv2D(128, 3, 1, 'same', kernel_initializer=initializer, activation='relu'),5354        # tf.keras.layers.MaxPooling2D(),55        # tf.keras.layers.BatchNormalization(),56        #57        # tf.keras.layers.Conv2D(128, 3, 1, 'same', kernel_initializer=initializer, activation='relu'),58        # tf.keras.layers.BatchNormalization(),59        # tf.keras.layers.Conv2D(64, 3, 1, 'same', kernel_initializer=initializer, activation='relu'),6061        tf.keras.layers.MaxPooling2D(),62        tf.keras.layers.BatchNormalization(),6364        tf.keras.layers.Conv2D(6, 3, 1, 'same', kernel_initializer=initializer),65        tf.keras.layers.GlobalAveragePooling2D(),66        # tf.keras.layers.Flatten(), 67        # tf.keras.layers.Dense(6),68        tf.keras.layers.Activation('softmax')69    ]7071    model = tf.keras.models.Sequential(layers)7273    train_loss = []74    val_loss = []75    train_acc = []76    val_acc = []7778    early_stopping = tf.keras.callbacks.EarlyStopping(monitor='val_loss', patience=3)  79    save_best_model = tf.keras.callbacks.ModelCheckpoint(save_file, monitor='val_loss', verbose=0, save_best_only=True, save_weights_only=True)  8081    first_learning_rate = 2.26e-482    second_learning_rate = 3.35e-58384    optimizer = tf.keras.optimizers.RMSprop()85    model.compile(optimizer, loss=tf.keras.losses.sparse_categorical_crossentropy, metrics=['accuracy'])8687    epoch_zero_train_loss, epoch_zero_train_acc = model.evaluate(X_train, y_train, verbose=0)88    epoch_zero_val_loss, epoch_zero_val_acc = model.evaluate(X_val, y_val, verbose=0)89    train_loss.append(epoch_zero_train_loss)90    val_loss.append(epoch_zero_val_loss)91    train_acc.append(epoch_zero_train_acc)92    val_acc.append(epoch_zero_val_acc)93    print('Before training')94    print(f'{X_train.shape[0]}/{X_train.shape[0]} - ?s - loss: {epoch_zero_train_loss:.4f} - accuracy: {epoch_zero_train_acc:.4f} - val_loss: {epoch_zero_val_loss:.4f} - val_accuracy: {epoch_zero_val_acc:.4f}')9596    print(f'\nFirst Training Session | lr: {first_learning_rate:.2e}')97    optimizer.learning_rate = first_learning_rate98    history = model.fit(X_train, y_train, 26, epochs=16, validation_data=(X_val, y_val), verbose=2, callbacks=[early_stopping, save_best_model])99    train_loss.extend(history.history['loss'])100    val_loss.extend(history.history['val_loss'])101    train_acc.extend(history.history['accuracy'])102    val_acc.extend(history.history['val_accuracy'])103104    print(f'\nSecond Training Session | lr: {second_learning_rate:.2e}')105    model.load_weights(save_file)106    optimizer.learning_rate = second_learning_rate107    history = model.fit(X_train, y_train, 26, epochs=14, validation_data=(X_val, y_val), verbose=2, callbacks=[early_stopping, save_best_model])108    train_loss.extend(history.history['loss'])109    val_loss.extend(history.history['val_loss'])110    train_acc.extend(history.history['accuracy'])111    val_acc.extend(history.history['val_accuracy'])112113    model.load_weights(save_file)114    test_results = model.evaluate(X_test, y_test)115116    if save_history:117        train_loss = np.asarray(train_loss)118        train_acc = np.asarray(train_acc)119        val_loss = np.asarray(val_loss)120        val_acc = np.asarray(val_acc)121        np.save(f'training_losses/TARnet.npy', train_loss)122        np.save(f'training_accuracies/TARnet.npy', train_acc)123        np.save(f'validation_losses/TARnet.npy', val_loss)124        np.save(f'validation_accuracies/TARnet.npy', val_acc)125126    if return_results:127        return test_results[1]128129    import matplotlib.pyplot as plt130131    fig, axs = plt.subplots(2, 1, constrained_layout=True)132133    axs[0].plot(train_loss, label='Train loss')134    axs[0].plot(val_loss, label='Validation loss')135    axs[0].set_title('Train and Validation Losses of TARnet', fontsize=18)136    axs[0].set_xlabel('Epoch')137    axs[0].set_ylabel('Loss')138    axs[0].legend(loc='upper right')139    axs[0].set_xlim(0, len(train_acc))140141    axs[1].plot(train_acc, label='Train accuracy')142    axs[1].plot(val_acc, label='Validation Accuracy')143    axs[1].set_title('Train and Validation Accuracies of TARnet', fontsize=18)144    axs[1].set_xlabel('Epoch')145    axs[1].set_ylabel('Accuracy')146    axs[1].hlines(0.85, 0, len(train_acc), label='85% Accuracy')147    axs[1].legend(loc='lower right')148149    plt.xlim(0, len(train_acc))150151    plt.show()152153
...

Automation Testing Tutorials

Learn to execute automation testing from scratch with LambdaTest Learning Hub. Right from setting up the prerequisites to run your first automation test, to following best practices and diving deeper into advanced test scenarios. LambdaTest Learning Hubs compile a list of step-by-step guides to help you be proficient with different test automation frameworks i.e. Selenium, Cypress, TestNG etc.