hw3_1.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "name": "hw3.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.7.3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/wallykop/iad-deep-learning/blob/master/hw3_1.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "gO2BchskQWoH"
      },
      "source": [
        "# Домашнее задание 3.\n",
        "\n",
        "`Keras` и сверточные нейронные сети."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "WmIuIaLEATMn",
        "outputId": "333d279f-fef2-427c-8ed1-61d63a09bf9f",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 118
        }
      },
      "source": [
        "import tensorflow as tf\n",
        "import keras\n",
        "from keras import backend as K\n",
        "import numpy as np\n",
        "%matplotlib inline\n",
        "import matplotlib.pyplot as plt\n",
        "print(tf.__version__)\n",
        "print(keras.__version__)"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<p style=\"color: red;\">\n",
              "The default version of TensorFlow in Colab will soon switch to TensorFlow 2.x.<br>\n",
              "We recommend you <a href=\"https://www.tensorflow.org/guide/migrate\" target=\"_blank\">upgrade</a> now \n",
              "or ensure your notebook will continue to use TensorFlow 1.x via the <code>%tensorflow_version 1.x</code> magic:\n",
              "<a href=\"https://colab.research.google.com/notebooks/tensorflow_version.ipynb\" target=\"_blank\">more info</a>.</p>\n"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "1.15.0\n",
            "2.2.5\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "Using TensorFlow backend.\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "efsrp4vQfls5",
        "colab": {}
      },
      "source": [
        "def reset_tf_session():\n",
        "    curr_session = tf.get_default_session()\n",
        "    if curr_session is not None:\n",
        "        curr_session.close()\n",
        "    K.clear_session()\n",
        "    config = tf.ConfigProto()\n",
        "    config.gpu_options.allow_growth = True\n",
        "    s = tf.InteractiveSession(config=config)\n",
        "    K.set_session(s)\n",
        "    return s"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "HE_SpMQMQm8Y"
      },
      "source": [
        "## Задание 1 — инициализация весов CNN (3 балла).\n",
        "\n",
        "В этом задании нужно будет исследовать, как выбор функции инициализации весов влияет на обучение CNN.\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "oIoAOXkJSwbs"
      },
      "source": [
        "Продолжим работать с датасетом CIFAR-10."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "IsuA4kiHA4ff",
        "outputId": "0f89dc1e-0a30-4b4d-9915-1e81a800f902",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 89
        }
      },
      "source": [
        "from keras.datasets import cifar10\n",
        "(x_train, y_train), (x_test, y_test) = cifar10.load_data()\n",
        "\n",
        "NUM_CLASSES = 10\n",
        "cifar10_classes = [\"airplane\", \"automobile\", \"bird\", \"cat\", \"deer\", \n",
        "                   \"dog\", \"frog\", \"horse\", \"ship\", \"truck\"]\n",
        "\n",
        "print(\"Train samples:\", x_train.shape, y_train.shape)\n",
        "print(\"Test samples:\", x_test.shape, y_test.shape)\n",
        "\n",
        "# нормализуем входные данные\n",
        "x_train = x_train / 255 - 0.5\n",
        "x_test = x_test / 255 - 0.5\n",
        "\n",
        "y_train = keras.utils.to_categorical(y_train, 10)\n",
        "y_test = keras.utils.to_categorical(y_test, 10)"
      ],
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Downloading data from https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz\n",
            "170500096/170498071 [==============================] - 11s 0us/step\n",
            "Train samples: (50000, 32, 32, 3) (50000, 1)\n",
            "Test samples: (10000, 32, 32, 3) (10000, 1)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "xNl2e8LeftcC",
        "colab": {}
      },
      "source": [
        "from keras.models import Sequential\n",
        "from keras.layers import Conv2D, MaxPooling2D, Flatten, Dense, Activation, Dropout\n",
        "from keras.layers.advanced_activations import LeakyReLU\n",
        "from keras.models import load_model"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "f2vQIUE3UeiE"
      },
      "source": [
        "Определим функцию `fit_model` с архитектурой архитектура CNN.\n",
        "\n",
        "Метод `model.fit` возвращает объект класса `keras.callbacks.History()` — это колбэк, который автоматически применяется ко всем моделям и логирует много чего полезного. В частности логируются значения функции потерь на каждой итерации."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "dDslR6vNZry8"
      },
      "source": [
        "**Задание 1.1** (0.5 балла) Добавьте в архитектуру модели инициализацию весов для тех слоев, где она необходима.\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "2_Lu7SO3CMid",
        "colab": {}
      },
      "source": [
        "def fit_model(initializer='glorot_normal'):\n",
        "    s = reset_tf_session()  \n",
        "    INIT_LR = 5e-3\n",
        "    BATCH_SIZE = 32\n",
        "    EPOCHS = 10\n",
        "   \n",
        "    def lr_scheduler(epoch):\n",
        "        return INIT_LR * 0.9 ** epoch  \n",
        "\n",
        "    ### YOUR CODE HERE \n",
        "    # kernel_initializer=initializer для тех слоев, которым нужна инициализация весов\n",
        "    model = Sequential()\n",
        "    model.add(Conv2D(filters=16, padding='same', kernel_size=(3,3), input_shape=(32,32,3), kernel_initializer=initializer))\n",
        "    model.add(LeakyReLU(0.1))\n",
        "    model.add(Conv2D(filters=64, padding='same', kernel_size=(3,3), kernel_initializer=initializer))  \n",
        "    model.add(LeakyReLU(0.1))\n",
        "    model.add(MaxPooling2D(pool_size=(2,2), padding='same'))\n",
        "    model.add(Dropout(0.25))\n",
        "    model.add(Flatten())\n",
        "    model.add(Dense(40, kernel_initializer=initializer))\n",
        "    model.add(LeakyReLU(0.1))\n",
        "    model.add(Dropout(0.5))    \n",
        "    model.add(Dense(10))\n",
        "    model.add(Activation(\"softmax\"))\n",
        "\n",
        "\n",
        "    model.compile(\n",
        "        loss='categorical_crossentropy',\n",
        "        optimizer=keras.optimizers.adamax(lr=INIT_LR),\n",
        "        metrics=['accuracy']\n",
        "    )\n",
        "  \n",
        "    history = model.fit(\n",
        "        x_train, y_train,\n",
        "        batch_size=BATCH_SIZE,\n",
        "        epochs=EPOCHS,\n",
        "        callbacks=[keras.callbacks.LearningRateScheduler(lr_scheduler)],                   \n",
        "        shuffle=True,\n",
        "        verbose=0,\n",
        "        initial_epoch=0    \n",
        "    )\n",
        "\n",
        "    # Возвращаем объект класса keras.callbacks.History\n",
        "    return history\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "bYkZkghfT4uN"
      },
      "source": [
        "**Задание 1.2** (1.5 балла). Обучите модель с разными [функциями инициализации](https://keras.io/initializers/) весов:\n",
        "* `Zeros` — веса инициализируются нулями\n",
        "* `Constant=0.05` — веса инициализируются константой 0.05\n",
        "* `RandomUniform` — веса генерируются равномерно из отрезка [-0.05, 0.05]\n",
        "* `glorot_normal` — Xavier initializer из лекций\n",
        "* `lecun_uniform`\n",
        "\n",
        "Добавьте в список `losses` значения функции потерь для каждой функции инициализации, их можно достать из `History`\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "dTieYmgUmQ8G",
        "colab": {}
      },
      "source": [
        "losses = []\n",
        "\n",
        "losses.append(fit_model(keras.initializers.Zeros()).history['loss'])\n",
        "losses.append(fit_model(keras.initializers.constant(value = 0.05)).history['loss'])\n",
        "losses.append(fit_model(keras.initializers.random_uniform(-0.05, 0.05)).history['loss'])\n",
        "losses.append(fit_model().history['loss'])\n",
        "losses.append(fit_model(keras.initializers.lecun_uniform()).history['loss'])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "opnAMTLafEod",
        "colab_type": "code",
        "outputId": "fa5bcb72-cd24-47a2-e4fe-491fbf12da9b",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 917
        }
      },
      "source": [
        "losses"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[[2.3030023341369628,\n",
              "  2.3029882539367676,\n",
              "  2.3029243270111084,\n",
              "  2.302886666717529,\n",
              "  2.302938383026123,\n",
              "  2.3027837757873537,\n",
              "  2.3028022551727294,\n",
              "  2.302804115600586,\n",
              "  2.302805783691406,\n",
              "  2.302728477020264],\n",
              " [2.7258832611083985,\n",
              "  2.0088204623413084,\n",
              "  1.7361099803161621,\n",
              "  1.5653727067184449,\n",
              "  1.4638671482849122,\n",
              "  1.3755998947143555,\n",
              "  1.303909610557556,\n",
              "  1.2419388565444947,\n",
              "  1.1909045807266236,\n",
              "  1.153439089164734],\n",
              " [1.5568914690017701,\n",
              "  1.2443780068969728,\n",
              "  1.1335640732192993,\n",
              "  1.0541191687965392,\n",
              "  0.989559116153717,\n",
              "  0.9445323751068115,\n",
              "  0.9039822071456909,\n",
              "  0.8679016119003296,\n",
              "  0.8382337683868408,\n",
              "  0.8090157961082458],\n",
              " [1.4983693269729614,\n",
              "  1.1775794541549682,\n",
              "  1.0668438763427734,\n",
              "  0.9908705131149292,\n",
              "  0.929420674533844,\n",
              "  0.8807885860824585,\n",
              "  0.8365703752136231,\n",
              "  0.8044158706665039,\n",
              "  0.768353724899292,\n",
              "  0.7457570077705383],\n",
              " [1.482921727142334,\n",
              "  1.1658392739868164,\n",
              "  1.0522454459381103,\n",
              "  0.9689599716186523,\n",
              "  0.9069279641532898,\n",
              "  0.8613294857025147,\n",
              "  0.8138451818275452,\n",
              "  0.7794717638206482,\n",
              "  0.7504262161636353,\n",
              "  0.7195395533370972]]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 92
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "wsk8bCEPcOav"
      },
      "source": [
        "**Задание 1.3** (1 балла). Постройте графики зависимости функций потерь от номера итерации, подпишите их. Прокомментируйте результат."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "3yqBZIYEAPWZ",
        "outputId": "1f56e474-ed89-47b4-82de-5dca17ee7324",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 881
        }
      },
      "source": [
        "label = ['zeros', 'constant', 'randomuniform', 'glorot_normal', 'lecun_uniform']\n",
        "\n",
        "plt.figure(figsize=(15,15))\n",
        "\n",
        "for i in range(0, len(losses)):\n",
        "  plt.plot(range(1, len(losses[i]) + 1), losses[i], label=label[i])\n",
        "  \n",
        "plt.legend()\n"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.legend.Legend at 0x7f626fbda710>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 94
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAA3AAAANOCAYAAABQmu4KAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nhAVOkiRJkjphgZMkSZKkTvwDalNGxUMKPXkAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 1080x1080 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "VF0Sk8HPwdTY",
        "colab_type": "text"
      },
      "source": [
        "Мы видим, что если задавать нулевую инициализацию на веса, то наш лосс не уменьшается, также мы видим, что если задавать в качестве инициализации константу, то лосс уменьшается, однако при применении остальных 3 методов мы можем заметить явное уменьшение лосса и более точные прогнозы."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "nellCcusBAZ8"
      },
      "source": [
        "## Задание 2 — CNN для CIFAR-10 с сохранением весов модели (7 баллов)\n",
        "\n",
        "В этом задании мы модифицируем нейросеть с семинара, чтобы она достигала большего значения `accuracy` и научимся сохранять веса модели в файл во время обучения. Можно использовать только те же слои, которые использовались на семинаре: `Conv2D, MaxPooling2D,  LeakyReLU, Dropout, Flatten, Dense`."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "xuVF3c2qJwyx"
      },
      "source": [
        "**Задание 2.1** (4 балла). Подберите архитектуру модели так, чтобы значение `accuracy` на тестовой выборке было не менее 85."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "syKDJPUdCtoI",
        "colab": {}
      },
      "source": [
        "def make_model():    \n",
        "\n",
        "    model = Sequential()\n",
        "    model.add(Conv2D(32, padding='same', kernel_size=(3,3), input_shape=(32,32,3), kernel_initializer='glorot_normal'))\n",
        "    model.add(LeakyReLU(0.1))\n",
        "    model.add(Conv2D(32, padding='same', kernel_size=(3,3), kernel_initializer='glorot_normal'))\n",
        "    model.add(LeakyReLU(0.1))\n",
        "    model.add(MaxPooling2D((2,2)))\n",
        "    model.add(Dropout(0.2))\n",
        "    \n",
        "    model.add(Conv2D(64, padding='same', kernel_size=(3,3), kernel_initializer='glorot_normal'))\n",
        "    model.add(LeakyReLU(0.1))\n",
        "    model.add(Conv2D(64, padding='same', kernel_size=(3,3), kernel_initializer='glorot_normal'))\n",
        "    model.add(LeakyReLU(0.1))\n",
        "    model.add(MaxPooling2D((2,2)))\n",
        "    model.add(Dropout(0.3))\n",
        "\n",
        "    model.add(Conv2D(128, padding='same', kernel_size=(3,3), kernel_initializer='glorot_normal'))\n",
        "    model.add(LeakyReLU(0.1))\n",
        "    model.add(Conv2D(128, padding='same', kernel_size=(3,3), kernel_initializer='glorot_normal'))\n",
        "    model.add(LeakyReLU(0.1))\n",
        "    model.add(MaxPooling2D((2,2)))\n",
        "    model.add(Dropout(0.4))\n",
        "\n",
        "    model.add(Flatten())\n",
        "    model.add(Dense(1024, kernel_initializer='glorot_normal'))\n",
        "    model.add(LeakyReLU(0.1))\n",
        "    model.add(Dropout(0.5))\n",
        "    model.add(Dense(10))\n",
        "    model.add(Activation('softmax'))\n",
        "\n",
        "    return model\n",
        "        "
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "MoZifjkgFdcg",
        "outputId": "282edd88-8a5c-4875-bbc5-fe4eae0d7862",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        }
      },
      "source": [
        "s = reset_tf_session()\n",
        "model = make_model()\n",
        "model.summary()"
      ],
      "execution_count": 37,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_1\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv2d_1 (Conv2D)            (None, 32, 32, 32)        896       \n",
            "_________________________________________________________________\n",
            "leaky_re_lu_1 (LeakyReLU)    (None, 32, 32, 32)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_2 (Conv2D)            (None, 32, 32, 32)        9248      \n",
            "_________________________________________________________________\n",
            "leaky_re_lu_2 (LeakyReLU)    (None, 32, 32, 32)        0         \n",
            "_________________________________________________________________\n",
            "max_pooling2d_1 (MaxPooling2 (None, 16, 16, 32)        0         \n",
            "_________________________________________________________________\n",
            "dropout_1 (Dropout)          (None, 16, 16, 32)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_3 (Conv2D)            (None, 16, 16, 64)        18496     \n",
            "_________________________________________________________________\n",
            "leaky_re_lu_3 (LeakyReLU)    (None, 16, 16, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_4 (Conv2D)            (None, 16, 16, 64)        36928     \n",
            "_________________________________________________________________\n",
            "leaky_re_lu_4 (LeakyReLU)    (None, 16, 16, 64)        0         \n",
            "_________________________________________________________________\n",
            "max_pooling2d_2 (MaxPooling2 (None, 8, 8, 64)          0         \n",
            "_________________________________________________________________\n",
            "dropout_2 (Dropout)          (None, 8, 8, 64)          0         \n",
            "_________________________________________________________________\n",
            "conv2d_5 (Conv2D)            (None, 8, 8, 128)         73856     \n",
            "_________________________________________________________________\n",
            "leaky_re_lu_5 (LeakyReLU)    (None, 8, 8, 128)         0         \n",
            "_________________________________________________________________\n",
            "conv2d_6 (Conv2D)            (None, 8, 8, 128)         147584    \n",
            "_________________________________________________________________\n",
            "leaky_re_lu_6 (LeakyReLU)    (None, 8, 8, 128)         0         \n",
            "_________________________________________________________________\n",
            "max_pooling2d_3 (MaxPooling2 (None, 4, 4, 128)         0         \n",
            "_________________________________________________________________\n",
            "dropout_3 (Dropout)          (None, 4, 4, 128)         0         \n",
            "_________________________________________________________________\n",
            "flatten_1 (Flatten)          (None, 2048)              0         \n",
            "_________________________________________________________________\n",
            "dense_1 (Dense)              (None, 1024)              2098176   \n",
            "_________________________________________________________________\n",
            "leaky_re_lu_7 (LeakyReLU)    (None, 1024)              0         \n",
            "_________________________________________________________________\n",
            "dropout_4 (Dropout)          (None, 1024)              0         \n",
            "_________________________________________________________________\n",
            "dense_2 (Dense)              (None, 10)                10250     \n",
            "_________________________________________________________________\n",
            "activation_1 (Activation)    (None, 10)                0         \n",
            "=================================================================\n",
            "Total params: 2,395,434\n",
            "Trainable params: 2,395,434\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "JJLpZQhJGBqL"
      },
      "source": [
        "**Задание 2.2** (2 балла). Реализуйте колбэк, который сохраняет модель в `.hdf5` файл и печатает имя файла, в который была сохранена модель. Используйте функцию `model_save`. Строка с именем файла имеет вид `<name>_{0:02d}.hdf5`, отформатируйте ее так, чтобы в имени строки фигурировал номер эпохи."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "xYjUeuDXGBII",
        "colab": {}
      },
      "source": [
        "from keras.models import save_model, load_model\n",
        "\n",
        "class ModelSaveCallback(keras.callbacks.Callback):\n",
        "\n",
        "    def __init__(self, file_name):\n",
        "        super(ModelSaveCallback, self).__init__()\n",
        "        self.file_name = file_name\n",
        "\n",
        "    def on_epoch_end(self, epoch, logs=None): \n",
        "      epoch += 1\n",
        "      save_model(model,\n",
        "                 f\"weights_{epoch:02d}.hdf5\")\n",
        "      print(f\"Saved weights_{epoch:02d}.hdf5\")"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "Svrwkh8ALpHa"
      },
      "source": [
        "**Задание 2.3** (1 балл). Реализуйте функцию, которая с помощью `load_model` будет загружать модель из файла."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "fVD0P5V7M00C",
        "colab": {}
      },
      "source": [
        "def load_from_file(model_filename, last_epoch):\n",
        "  model = load_model(f\"weights_{last_epoch:02d}.hdf5\")\n",
        "  return model\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "zOLb6flQFjCw",
        "outputId": "ae7330a4-85ef-4863-ea4f-191bc5ca2bfe",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        }
      },
      "source": [
        "INIT_LR = 5e-3\n",
        "BATCH_SIZE = 32\n",
        "EPOCHS = 20\n",
        "\n",
        "model_filename = 'weights_{0:02d}.hdf5'\n",
        "\n",
        "s = reset_tf_session()\n",
        "\n",
        "model = make_model()\n",
        "\n",
        "model.compile(\n",
        "    loss='categorical_crossentropy',\n",
        "    optimizer=keras.optimizers.adamax(lr=INIT_LR),\n",
        "    metrics=['accuracy']\n",
        ")\n",
        "\n",
        "def lr_scheduler(epoch):\n",
        "    return INIT_LR * 0.9 ** epoch \n",
        "\n",
        "# в случае, если обучение было прервано, можно загрузить модель из файла, \n",
        "# соответствующего последней эпохе, за которую есть сохраненные веса\n",
        "# model = load_from_file(model_filename, 2)\n",
        "  \n",
        "history = model.fit(\n",
        "    x_train, y_train,\n",
        "    batch_size=BATCH_SIZE,\n",
        "    epochs=EPOCHS,\n",
        "    callbacks=[keras.callbacks.LearningRateScheduler(lr_scheduler), \n",
        "               # не забудьте передать сюда ModelSaveCallback\n",
        "               ModelSaveCallback(model_filename)\n",
        "              ],\n",
        "    validation_data=(x_test, y_test),\n",
        "    shuffle=True,\n",
        "    verbose=1,\n",
        "    initial_epoch=0    \n",
        ")\n",
        "  "
      ],
      "execution_count": 40,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 50000 samples, validate on 10000 samples\n",
            "Epoch 1/20\n",
            "50000/50000 [==============================] - 45s 890us/step - loss: 1.4266 - acc: 0.4784 - val_loss: 0.9653 - val_acc: 0.6578\n",
            "Saved weights_01.hdf5\n",
            "Epoch 2/20\n",
            "50000/50000 [==============================] - 43s 855us/step - loss: 0.9695 - acc: 0.6569 - val_loss: 0.8126 - val_acc: 0.7183\n",
            "Saved weights_02.hdf5\n",
            "Epoch 3/20\n",
            "50000/50000 [==============================] - 41s 814us/step - loss: 0.8265 - acc: 0.7136 - val_loss: 0.7907 - val_acc: 0.7299\n",
            "Saved weights_03.hdf5\n",
            "Epoch 4/20\n",
            "50000/50000 [==============================] - 41s 811us/step - loss: 0.7360 - acc: 0.7421 - val_loss: 0.6348 - val_acc: 0.7846\n",
            "Saved weights_04.hdf5\n",
            "Epoch 5/20\n",
            "50000/50000 [==============================] - 39s 785us/step - loss: 0.6667 - acc: 0.7677 - val_loss: 0.6346 - val_acc: 0.7804\n",
            "Saved weights_05.hdf5\n",
            "Epoch 6/20\n",
            "50000/50000 [==============================] - 39s 777us/step - loss: 0.6247 - acc: 0.7817 - val_loss: 0.5670 - val_acc: 0.8100\n",
            "Saved weights_06.hdf5\n",
            "Epoch 7/20\n",
            "50000/50000 [==============================] - 39s 779us/step - loss: 0.5675 - acc: 0.8028 - val_loss: 0.5521 - val_acc: 0.8146\n",
            "Saved weights_07.hdf5\n",
            "Epoch 8/20\n",
            "50000/50000 [==============================] - 39s 778us/step - loss: 0.5355 - acc: 0.8156 - val_loss: 0.5511 - val_acc: 0.8228\n",
            "Saved weights_08.hdf5\n",
            "Epoch 9/20\n",
            "50000/50000 [==============================] - 39s 778us/step - loss: 0.5001 - acc: 0.8264 - val_loss: 0.5622 - val_acc: 0.8163\n",
            "Saved weights_09.hdf5\n",
            "Epoch 10/20\n",
            "50000/50000 [==============================] - 39s 782us/step - loss: 0.4648 - acc: 0.8385 - val_loss: 0.5383 - val_acc: 0.8305\n",
            "Saved weights_10.hdf5\n",
            "Epoch 11/20\n",
            "50000/50000 [==============================] - 39s 779us/step - loss: 0.4372 - acc: 0.8476 - val_loss: 0.4982 - val_acc: 0.8388\n",
            "Saved weights_11.hdf5\n",
            "Epoch 12/20\n",
            "50000/50000 [==============================] - 39s 777us/step - loss: 0.4078 - acc: 0.8579 - val_loss: 0.5295 - val_acc: 0.8377\n",
            "Saved weights_12.hdf5\n",
            "Epoch 13/20\n",
            "50000/50000 [==============================] - 39s 777us/step - loss: 0.3865 - acc: 0.8647 - val_loss: 0.5139 - val_acc: 0.8390\n",
            "Saved weights_13.hdf5\n",
            "Epoch 14/20\n",
            "50000/50000 [==============================] - 39s 777us/step - loss: 0.3633 - acc: 0.8719 - val_loss: 0.5129 - val_acc: 0.8478\n",
            "Saved weights_14.hdf5\n",
            "Epoch 15/20\n",
            "50000/50000 [==============================] - 39s 777us/step - loss: 0.3445 - acc: 0.8789 - val_loss: 0.5013 - val_acc: 0.8429\n",
            "Saved weights_15.hdf5\n",
            "Epoch 16/20\n",
            "50000/50000 [==============================] - 39s 778us/step - loss: 0.3272 - acc: 0.8860 - val_loss: 0.4893 - val_acc: 0.8479\n",
            "Saved weights_16.hdf5\n",
            "Epoch 17/20\n",
            "50000/50000 [==============================] - 39s 776us/step - loss: 0.3154 - acc: 0.8883 - val_loss: 0.4734 - val_acc: 0.8516\n",
            "Saved weights_17.hdf5\n",
            "Epoch 18/20\n",
            "50000/50000 [==============================] - 39s 783us/step - loss: 0.3024 - acc: 0.8933 - val_loss: 0.4931 - val_acc: 0.8548\n",
            "Saved weights_18.hdf5\n",
            "Epoch 19/20\n",
            "50000/50000 [==============================] - 39s 781us/step - loss: 0.2901 - acc: 0.8978 - val_loss: 0.4922 - val_acc: 0.8541\n",
            "Saved weights_19.hdf5\n",
            "Epoch 20/20\n",
            "50000/50000 [==============================] - 39s 778us/step - loss: 0.2815 - acc: 0.9004 - val_loss: 0.5094 - val_acc: 0.8556\n",
            "Saved weights_20.hdf5\n"
          ],
          "name": "stdout"
        }
      ]
    }
  ]
}