{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"pygments_lexer":"ipython3","nbconvert_exporter":"python","version":"3.6.4","file_extension":".py","codemirror_mode":{"name":"ipython","version":3},"name":"python","mimetype":"text/x-python"}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"markdown","source":"# TF 2.2 блокнот","metadata":{"papermill":{"duration":0.014347,"end_time":"2022-11-20T10:59:40.628486","exception":false,"start_time":"2022-11-20T10:59:40.614139","status":"completed"},"tags":[]}},{"cell_type":"code","source":"# This Python 3 environment comes with many helpful analytics libraries installed\n# It is defined by the kaggle/python Docker image: https://github.com/kaggle/docker-python\n# For example, here's several helpful packages to load\n\nimport numpy as np # linear algebra\nimport pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\n\n# Input data files are available in the read-only \"../input/\" directory\n# For example, running this (by clicking run or pressing Shift+Enter) will list all files under the input directory\n\nimport os\nfor dirname, _, filenames in os.walk('/kaggle/input'):\n    for filename in filenames:\n        print(os.path.join(dirname, filename))\n\n# You can write up to 20GB to the current directory (/kaggle/working/) that gets preserved as output when you create a version using \"Save & Run All\" \n# You can also write temporary files to /kaggle/temp/, but they won't be saved outside of the current session","metadata":{"execution":{"iopub.status.busy":"2022-12-20T07:01:18.711708Z","iopub.status.idle":"2022-12-20T07:01:18.712482Z","shell.execute_reply.started":"2022-12-20T07:01:18.712256Z","shell.execute_reply":"2022-12-20T07:01:18.712275Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"!pip install -q efficientnet","metadata":{"papermill":{"duration":10.406519,"end_time":"2022-11-20T10:59:51.048328","exception":false,"start_time":"2022-11-20T10:59:40.641809","status":"completed"},"tags":[],"execution":{"iopub.status.busy":"2022-12-20T07:01:21.702986Z","iopub.execute_input":"2022-12-20T07:01:21.703509Z","iopub.status.idle":"2022-12-20T07:01:31.787881Z","shell.execute_reply.started":"2022-12-20T07:01:21.703456Z","shell.execute_reply":"2022-12-20T07:01:31.787103Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"from tensorflow.keras.applications.vgg16 import VGG16\nfrom tensorflow.keras.applications.vgg19 import VGG19\nfrom tensorflow.keras.applications.inception_v3 import InceptionV3\nfrom tensorflow.keras.applications.inception_resnet_v2 import InceptionResNetV2\nfrom tensorflow.keras.applications.densenet import DenseNet121, DenseNet169, DenseNet201 \nfrom tensorflow.keras.applications.xception import Xception\nfrom tensorflow.keras.applications.resnet50 import ResNet50\nfrom tensorflow.keras.applications.resnet_v2 import ResNet50V2, ResNet101V2, ResNet152V2\nfrom tensorflow.keras.applications.nasnet import NASNetLarge\nfrom efficientnet.tfkeras import EfficientNetB7, EfficientNetL2, EfficientNetB0, EfficientNetB1\n\nimport tensorflow as tf\nfrom tensorflow.keras.models import Model\nfrom tensorflow.keras import layers\nfrom tensorflow.keras.layers import Dense, Flatten\nfrom tensorflow.keras.layers import Dropout, BatchNormalization, GaussianDropout\nfrom tensorflow.keras.layers import GlobalAveragePooling2D\nfrom tensorflow.keras.callbacks import ModelCheckpoint\n\nimport tensorflow as tf\nfrom tensorflow.keras.applications.densenet import DenseNet121\nfrom efficientnet.tfkeras import EfficientNetB7\nfrom tensorflow.keras.models import Model\nfrom tensorflow.keras.layers import Dense\n#  библиотека для работы с наборами данных на Kaggle\nfrom kaggle_datasets import KaggleDatasets\nimport re\nimport numpy as np\nimport random\nimport matplotlib.pyplot as plt\n%matplotlib inline \nprint(\"Tensorflow version \" + tf.__version__)","metadata":{"_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","papermill":{"duration":7.130862,"end_time":"2022-11-20T10:59:58.193103","exception":false,"start_time":"2022-11-20T10:59:51.062241","status":"completed"},"tags":[],"execution":{"iopub.status.busy":"2022-12-20T07:01:41.979986Z","iopub.execute_input":"2022-12-20T07:01:41.980348Z","iopub.status.idle":"2022-12-20T07:01:48.696417Z","shell.execute_reply.started":"2022-12-20T07:01:41.980308Z","shell.execute_reply":"2022-12-20T07:01:48.695476Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Определяем, какой ускоритель можем использовать","metadata":{"papermill":{"duration":0.013431,"end_time":"2022-11-20T10:59:58.221329","exception":false,"start_time":"2022-11-20T10:59:58.207898","status":"completed"},"tags":[]}},{"cell_type":"code","source":"AUTO = tf.data.experimental.AUTOTUNE\n# Обнаружение оборудования, возврат соответствующей стратегии распространения: TPU, GPU, CPU\ntry:\n    tpu = tf.distribute.cluster_resolver.TPUClusterResolver()  # Обнаружение TPU. Параметры среды не требуются, если задана переменная среды TPU_NAME. На Kaggle это всегда так.\n    print('Running on TPU ', tpu.master())\nexcept ValueError:\n    tpu = None\n\nif tpu:\n    tf.config.experimental_connect_to_cluster(tpu)\n    tf.tpu.experimental.initialize_tpu_system(tpu)\n    strategy = tf.distribute.experimental.TPUStrategy(tpu)\nelse:\n    strategy = tf.distribute.get_strategy() # стратегия распространения по умолчанию в Tensorflow. Работает на CPU и одном GPU.\nprint(\"REPLICAS: \", strategy.num_replicas_in_sync)","metadata":{"papermill":{"duration":6.426996,"end_time":"2022-11-20T11:00:04.662689","exception":false,"start_time":"2022-11-20T10:59:58.235693","status":"completed"},"tags":[],"execution":{"iopub.status.busy":"2022-12-20T07:01:54.142724Z","iopub.execute_input":"2022-12-20T07:01:54.143027Z","iopub.status.idle":"2022-12-20T07:02:00.304515Z","shell.execute_reply.started":"2022-12-20T07:01:54.143001Z","shell.execute_reply":"2022-12-20T07:02:00.303505Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Get my data path","metadata":{"papermill":{"duration":0.014973,"end_time":"2022-11-20T11:00:04.694307","exception":false,"start_time":"2022-11-20T11:00:04.679334","status":"completed"},"tags":[]}},{"cell_type":"code","source":"GCS_DS_PATH = KaggleDatasets().get_gcs_path(\"tpu-getting-started\") #получаем путь к наборам данных","metadata":{"papermill":{"duration":0.401337,"end_time":"2022-11-20T11:00:05.112329","exception":false,"start_time":"2022-11-20T11:00:04.710992","status":"completed"},"tags":[],"execution":{"iopub.status.busy":"2022-12-20T07:03:22.898997Z","iopub.execute_input":"2022-12-20T07:03:22.899334Z","iopub.status.idle":"2022-12-20T07:03:23.293839Z","shell.execute_reply.started":"2022-12-20T07:03:22.899300Z","shell.execute_reply":"2022-12-20T07:03:23.292834Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Set some parameters","metadata":{"papermill":{"duration":0.015569,"end_time":"2022-11-20T11:00:05.143607","exception":false,"start_time":"2022-11-20T11:00:05.128038","status":"completed"},"tags":[]}},{"cell_type":"code","source":"IMAGE_SIZE = [512, 512] # при таком размере графическому процессору не хватит памяти. Используйте TPU\nBATCH_SIZE = 16 * strategy.num_replicas_in_sync\n\nGCS_PATH_SELECT = { # available image sizes\n    192: GCS_DS_PATH + '/tfrecords-jpeg-192x192',\n    224: GCS_DS_PATH + '/tfrecords-jpeg-224x224',\n    331: GCS_DS_PATH + '/tfrecords-jpeg-331x331',\n    512: GCS_DS_PATH + '/tfrecords-jpeg-512x512'\n}\nGCS_PATH = GCS_PATH_SELECT[IMAGE_SIZE[0]]\n\nVALIDATION_FILENAMES = tf.io.gfile.glob(GCS_PATH + '/val/*.tfrec')\nTEST_FILENAMES = tf.io.gfile.glob(GCS_PATH + '/test/*.tfrec')\n\nSEED = 2020","metadata":{"papermill":{"duration":0.184127,"end_time":"2022-11-20T11:00:05.342730","exception":false,"start_time":"2022-11-20T11:00:05.158603","status":"completed"},"tags":[],"execution":{"iopub.status.busy":"2022-12-20T07:03:25.863546Z","iopub.execute_input":"2022-12-20T07:03:25.864425Z","iopub.status.idle":"2022-12-20T07:03:26.037459Z","shell.execute_reply.started":"2022-12-20T07:03:25.864385Z","shell.execute_reply":"2022-12-20T07:03:26.036483Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Загружаем данные\n\nЭти данные загружаются из Kaggle и автоматически сегментируются для максимального распараллеливания.","metadata":{"papermill":{"duration":0.015179,"end_time":"2022-11-20T11:00:05.375439","exception":false,"start_time":"2022-11-20T11:00:05.360260","status":"completed"},"tags":[]}},{"cell_type":"code","source":"def decode_image(image_data):\n    \"\"\"Декодирует изображение в vyjujvthye. vfnhbwe (тензор)\n    Нормализует данные и преобразовывает изображения к указанному размеру\"\"\"\n    image = tf.image.decode_jpeg(image_data, channels=3) # Декодирование изображения в формате JPEG в тензор uint8.\n    image = tf.cast(image, tf.float32) / 255.0  # преобразовать изображение в плавающее в диапазоне [0, 1]\n    image = tf.reshape(image, [*IMAGE_SIZE, 3]) # явный размер, необходимый для TPU\n#     image = tf.keras.applications.inception_resnet_v2.preprocess_input(image)\n    return image\n\ndef read_labeled_tfrecord(example):\n    LABELED_TFREC_FORMAT = {\n        \"image\": tf.io.FixedLenFeature([], tf.string), # tf.string означает байтовую строку\n        \"class\": tf.io.FixedLenFeature([], tf.int64),  # [] означает отдельный элемент\n    }\n    example = tf.io.parse_single_example(example, LABELED_TFREC_FORMAT) # парсим отдельный пример в указанном формате\n    image = decode_image(example['image']) # преобразуем изображение к нужному нам формату\n    label = tf.cast(example['class'], tf.int32)\n    return image, label # возвращает набор данных пар (изображение, метка)\n\ndef read_unlabeled_tfrecord(example):\n    UNLABELED_TFREC_FORMAT = {\n        \"image\": tf.io.FixedLenFeature([], tf.string), # tf.string означает байтовую строку\n        \"id\": tf.io.FixedLenFeature([], tf.string),  # [] означает отдельный элемент\n        # класс отсутствует, задача этого конкурса - предсказать классы цветов для тестового набора данных\n    }\n    example = tf.io.parse_single_example(example, UNLABELED_TFREC_FORMAT)\n    image = decode_image(example['image']) # преобразуем изображение к нужному нам формату\n    idnum = example['id']\n    return image, idnum # returns a dataset of image(s)\n\ndef load_dataset(filenames, labeled=True, ordered=False):\n    \"\"\"Читает из TFRecords. Для оптимальной производительности одновременное чтение из нескольких\n    файлов без учета порядка данных. Порядок не имеет значения, поскольку мы все равно будем перетасовывать данные\"\"\"\n\n    ignore_order = tf.data.Options() # Представляет параметры для tf.data.Dataset.\n    if not ordered:\n        ignore_order.experimental_deterministic = False # отключить порядок, увеличить скорость\n\n    dataset = tf.data.TFRecordDataset(filenames, num_parallel_reads=AUTO) # автоматически чередует чтение из нескольких файлов\n    dataset = dataset.with_options(ignore_order) # использует данные сразу после их поступления, а не в исходном порядке\n    dataset = dataset.map(read_labeled_tfrecord if labeled else read_unlabeled_tfrecord, num_parallel_calls=AUTO)\n    # возвращает набор данных пар (изображение, метка), если метка = Истина, или пар (изображение, идентификатор), если метка = Ложь\n    return dataset\n\ndef get_validation_dataset():\n    dataset = load_dataset(VALIDATION_FILENAMES, labeled=True, ordered=True)\n    dataset = dataset.batch(BATCH_SIZE)\n    dataset = dataset.cache() # кешируем набор\n    dataset = dataset.prefetch(AUTO) #готовим следующий набор, пока предыдущий обучается\n    return dataset\n\ndef get_test_dataset(ordered=False):\n    dataset = load_dataset(TEST_FILENAMES, labeled=False, ordered=ordered)\n    dataset = dataset.batch(BATCH_SIZE)\n    dataset = dataset.prefetch(AUTO) #готовим следующий набор, пока предыдущий обучается\n    return dataset\n                               \ndef count_data_items(filenames):\n    # the number of data items is written in the name of the .tfrec files, i.e. flowers00-230.tfrec = 230 data items\n    n = [int(re.compile(r\"-([0-9]*)\\.\").search(filename).group(1)) for filename in filenames]\n    return np.sum(n)\n\nNUM_VALIDATION_IMAGES = count_data_items(VALIDATION_FILENAMES)\nNUM_TEST_IMAGES = count_data_items(TEST_FILENAMES)\nprint('Dataset: {} validation images, {} unlabeled test images'.format(NUM_VALIDATION_IMAGES, NUM_TEST_IMAGES))","metadata":{"papermill":{"duration":0.040261,"end_time":"2022-11-20T11:00:05.431545","exception":false,"start_time":"2022-11-20T11:00:05.391284","status":"completed"},"tags":[],"execution":{"iopub.status.busy":"2022-12-20T07:03:30.151373Z","iopub.execute_input":"2022-12-20T07:03:30.151991Z","iopub.status.idle":"2022-12-20T07:03:30.172141Z","shell.execute_reply.started":"2022-12-20T07:03:30.151955Z","shell.execute_reply":"2022-12-20T07:03:30.170747Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Построить модель на TPU (или GPU, или CPU...) с Tensorflow 2.1!","metadata":{"papermill":{"duration":0.01583,"end_time":"2022-11-20T11:00:05.463815","exception":false,"start_time":"2022-11-20T11:00:05.447985","status":"completed"},"tags":[]}},{"cell_type":"code","source":"def get_model(use_model):\n    # noisy-student\n    base_model = use_model(weights='imagenet', \n                      include_top=False, pooling='avg',\n                      input_shape=(*IMAGE_SIZE, 3))\n#     base_model.trainable = False\n    x = base_model.output\n    predictions = Dense(104, activation='softmax')(x)\n    model = Model(inputs=base_model.input, outputs=predictions)\n    model.compile(\n                    optimizer='nadam',\n                    loss = 'sparse_categorical_crossentropy',\n                    metrics=['sparse_categorical_accuracy']\n                 )\n    return model\nwith strategy.scope():    \n    model1 = get_model(DenseNet121) # тут подставить свою модель\nmodel1.load_weights(\"/kaggle/input/petals-densenet201/my_densenet.h5\")","metadata":{"papermill":{"duration":25.622034,"end_time":"2022-11-20T11:00:31.101636","exception":false,"start_time":"2022-11-20T11:00:05.479602","status":"completed"},"tags":[],"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def get_model(use_model):\n    # noisy-student\n    base_model = use_model(weights='noisy-student', \n                      include_top=False, pooling='avg',\n                      input_shape=(*IMAGE_SIZE, 3))\n#     base_model.trainable = False\n    x = base_model.output\n    x = layers.BatchNormalization(name=\"bn\")(x)\n    # tune\n    x = Dense(512, activation='relu', name=\"dense_512\")(x)\n    x = Dense(208, activation='relu', name=\"dense_208\")(x)\n    predictions = Dense(104, activation='softmax', name=\"dense_104\")(x)\n    model = Model(inputs=base_model.input, outputs=predictions)\n    model.compile(\n                optimizer='nadam',\n                loss = 'sparse_categorical_crossentropy',\n                metrics=['sparse_categorical_accuracy']\n                )\n    return model\n\n\nwith strategy.scope():    \n    model2 = get_model(EfficientNetB7) # тут подставить свою модель\nmodel2.load_weights(\"/kaggle/input/petals-efficientnetb7/my_ef_net_b7.h5\") ","metadata":{"papermill":{"duration":57.045021,"end_time":"2022-11-20T11:01:28.163529","exception":false,"start_time":"2022-11-20T11:00:31.118508","status":"completed"},"tags":[],"execution":{"iopub.status.busy":"2022-12-20T07:05:27.449668Z","iopub.execute_input":"2022-12-20T07:05:27.450165Z","iopub.status.idle":"2022-12-20T07:06:06.251918Z","shell.execute_reply.started":"2022-12-20T07:05:27.450127Z","shell.execute_reply":"2022-12-20T07:06:06.250474Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"from sklearn.metrics import f1_score\nval_dataset = get_validation_dataset()\nimages_ds = val_dataset.map(lambda image, label: image)\nlabels_ds = val_dataset.map(lambda image, label: label).unbatch()\nval_labels = next(iter(labels_ds.batch(NUM_VALIDATION_IMAGES))).numpy() # get everything as one batch\nm1 = model1.predict(images_ds)\nm2 = model2.predict(images_ds)\nscores = []\nfor alpha in np.linspace(0,1,100):\n    val_probabilities = alpha*m1+(1-alpha)*m2\n    val_predictions = np.argmax(val_probabilities, axis=-1)\n    scores.append(f1_score(val_labels, val_predictions, labels=range(104), average='macro'))\n\nbest_alpha = np.argmax(scores)/100\n    \nprint('Best alpha: ' + str(best_alpha))","metadata":{"papermill":{"duration":43.410335,"end_time":"2022-11-20T11:02:11.607025","exception":false,"start_time":"2022-11-20T11:01:28.196690","status":"completed"},"tags":[],"execution":{"iopub.status.busy":"2022-12-20T04:52:58.342928Z","iopub.status.idle":"2022-12-20T04:52:58.343420Z","shell.execute_reply.started":"2022-12-20T04:52:58.343187Z","shell.execute_reply":"2022-12-20T04:52:58.343209Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Вычислите свои прогнозы на тестовом наборе!\n\nCоздадим файл, который можно будет отправить на конкурс.","metadata":{"papermill":{"duration":0.035376,"end_time":"2022-11-20T11:02:11.676760","exception":false,"start_time":"2022-11-20T11:02:11.641384","status":"completed"},"tags":[]}},{"cell_type":"code","source":"# Поскольку мы разделяем набор данных и выполняем итерацию отдельно для изображений и идентификаторов, порядок имеет значение.\ntest_ds = get_test_dataset(ordered=True) \n\nprint('Вычисляем предсказания...')\ntest_images_ds = test_ds.map(lambda image, idnum: image)\n\nprobabilities1 = model1.predict(test_images_ds)\nprobabilities2 = model2.predict(test_images_ds)\nprobabilities = best_alpha * probabilities1 + (1 - best_alpha) * probabilities2\npredictions = np.argmax(probabilities, axis=-1)\nprint(predictions)\n\nprint('Создание файла submission.csv...')\ntest_ids_ds = test_ds.map(lambda image, idnum: idnum).unbatch()\ntest_ids = next(iter(test_ids_ds.batch(NUM_TEST_IMAGES))).numpy().astype('U') # все в одной партии\nnp.savetxt('submission.csv', np.rec.fromarrays([test_ids, predictions]), fmt=['%s', '%d'], delimiter=',', header='id,label', comments='')","metadata":{"papermill":{"duration":43.165354,"end_time":"2022-11-20T11:02:54.875504","exception":false,"start_time":"2022-11-20T11:02:11.710150","status":"completed"},"tags":[],"execution":{"iopub.status.busy":"2022-12-20T04:52:58.345092Z","iopub.status.idle":"2022-12-20T04:52:58.345578Z","shell.execute_reply.started":"2022-12-20T04:52:58.345342Z","shell.execute_reply":"2022-12-20T04:52:58.345363Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"","metadata":{"papermill":{"duration":0.033555,"end_time":"2022-11-20T11:02:54.943488","exception":false,"start_time":"2022-11-20T11:02:54.909933","status":"completed"},"tags":[]},"execution_count":null,"outputs":[]}]}