Captcha_Solver_numerical.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Captcha_Solver_numerical.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "id": "6Tzheb9xoLiD",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "ccb08657-9331-49e3-bb2c-9cce0fda919f"
      },
      "source": [
        "%pip install captcha \n",
        "%pip install opencv-python\n",
        "!apt update && apt install -y libsm6 libxext6\n",
        "!pip install opencv-python-headless\n",
        "!pip install opencv-contrib-python-headless\n",
        "%pip install keras\n",
        "\n",
        "\n",
        "import argparse\n",
        "import json\n",
        "import string\n",
        "import os\n",
        "import shutil\n",
        "import uuid\n",
        "from captcha.image import ImageCaptcha\n",
        "\n",
        "import itertools\n",
        "\n",
        "import os\n",
        "import cv2\n",
        "import numpy as np\n",
        "from random import random, randint, choices\n",
        "\n",
        "import keras\n",
        "from keras.models import Sequential, Model\n",
        "from keras.layers import Dense, Dropout, Activation, Flatten\n",
        "from keras.layers import Conv2D, MaxPooling2D, Input\n",
        "import matplotlib.pyplot as plt"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Collecting captcha\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/90/fe/d4ddf1e6576073b5eaea76e9b2afa022c626212a30c871968480be3ccb7b/captcha-0.3-py3-none-any.whl (101kB)\n",
            "\r\u001b[K     |███▎                            | 10kB 22.3MB/s eta 0:00:01\r\u001b[K     |██████▌                         | 20kB 9.3MB/s eta 0:00:01\r\u001b[K     |█████████▊                      | 30kB 7.6MB/s eta 0:00:01\r\u001b[K     |█████████████                   | 40kB 8.5MB/s eta 0:00:01\r\u001b[K     |████████████████▏               | 51kB 7.3MB/s eta 0:00:01\r\u001b[K     |███████████████████▍            | 61kB 7.6MB/s eta 0:00:01\r\u001b[K     |██████████████████████▋         | 71kB 7.3MB/s eta 0:00:01\r\u001b[K     |█████████████████████████▉      | 81kB 7.6MB/s eta 0:00:01\r\u001b[K     |█████████████████████████████   | 92kB 8.0MB/s eta 0:00:01\r\u001b[K     |████████████████████████████████| 102kB 5.2MB/s \n",
            "\u001b[?25hRequirement already satisfied: Pillow in /usr/local/lib/python3.6/dist-packages (from captcha) (7.0.0)\n",
            "Installing collected packages: captcha\n",
            "Successfully installed captcha-0.3\n",
            "Requirement already satisfied: opencv-python in /usr/local/lib/python3.6/dist-packages (4.1.2.30)\n",
            "Requirement already satisfied: numpy>=1.11.3 in /usr/local/lib/python3.6/dist-packages (from opencv-python) (1.18.5)\n",
            "Get:1 https://cloud.r-project.org/bin/linux/ubuntu bionic-cran35/ InRelease [3,626 B]\n",
            "Ign:2 https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64  InRelease\n",
            "Ign:3 https://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1804/x86_64  InRelease\n",
            "Hit:4 https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64  Release\n",
            "Hit:5 http://ppa.launchpad.net/graphics-drivers/ppa/ubuntu bionic InRelease\n",
            "Hit:6 https://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1804/x86_64  Release\n",
            "Get:7 http://security.ubuntu.com/ubuntu bionic-security InRelease [88.7 kB]\n",
            "Hit:9 http://archive.ubuntu.com/ubuntu bionic InRelease\n",
            "Get:10 http://archive.ubuntu.com/ubuntu bionic-updates InRelease [88.7 kB]\n",
            "Get:11 http://ppa.launchpad.net/marutter/c2d4u3.5/ubuntu bionic InRelease [15.4 kB]\n",
            "Get:13 http://archive.ubuntu.com/ubuntu bionic-backports InRelease [74.6 kB]\n",
            "Get:14 http://ppa.launchpad.net/marutter/c2d4u3.5/ubuntu bionic/main Sources [1,853 kB]\n",
            "Get:15 http://archive.ubuntu.com/ubuntu bionic-updates/restricted amd64 Packages [116 kB]\n",
            "Get:16 http://archive.ubuntu.com/ubuntu bionic-updates/universe amd64 Packages [1,413 kB]\n",
            "Get:17 http://archive.ubuntu.com/ubuntu bionic-updates/multiverse amd64 Packages [27.1 kB]\n",
            "Get:18 http://archive.ubuntu.com/ubuntu bionic-updates/main amd64 Packages [1,334 kB]\n",
            "Get:19 http://ppa.launchpad.net/marutter/c2d4u3.5/ubuntu bionic/main amd64 Packages [894 kB]\n",
            "Fetched 5,909 kB in 5s (1,134 kB/s)\n",
            "Reading package lists... Done\n",
            "Building dependency tree       \n",
            "Reading state information... Done\n",
            "38 packages can be upgraded. Run 'apt list --upgradable' to see them.\n",
            "Reading package lists... Done\n",
            "Building dependency tree       \n",
            "Reading state information... Done\n",
            "libsm6 is already the newest version (2:1.2.2-1).\n",
            "libxext6 is already the newest version (2:1.3.3-1).\n",
            "The following package was automatically installed and is no longer required:\n",
            "  libnvidia-common-440\n",
            "Use 'apt autoremove' to remove it.\n",
            "0 upgraded, 0 newly installed, 0 to remove and 38 not upgraded.\n",
            "Collecting opencv-python-headless\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/17/e4/a98a3c3098ea55b6ae193a1cd19a221dc3c1bde87a36db5550addc879d36/opencv_python_headless-4.3.0.36-cp36-cp36m-manylinux2014_x86_64.whl (36.4MB)\n",
            "\u001b[K     |████████████████████████████████| 36.4MB 77kB/s \n",
            "\u001b[?25hRequirement already satisfied: numpy>=1.11.3 in /usr/local/lib/python3.6/dist-packages (from opencv-python-headless) (1.18.5)\n",
            "Installing collected packages: opencv-python-headless\n",
            "Successfully installed opencv-python-headless-4.3.0.36\n",
            "Collecting opencv-contrib-python-headless\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/a1/90/b2982491b86fc8f39153162da5775081361ce731156df9a248c55396c5c8/opencv_contrib_python_headless-4.3.0.36-cp36-cp36m-manylinux2014_x86_64.whl (42.6MB)\n",
            "\u001b[K     |████████████████████████████████| 42.6MB 72kB/s \n",
            "\u001b[?25hRequirement already satisfied: numpy>=1.11.3 in /usr/local/lib/python3.6/dist-packages (from opencv-contrib-python-headless) (1.18.5)\n",
            "Installing collected packages: opencv-contrib-python-headless\n",
            "Successfully installed opencv-contrib-python-headless-4.3.0.36\n",
            "Requirement already satisfied: keras in /usr/local/lib/python3.6/dist-packages (2.3.1)\n",
            "Requirement already satisfied: six>=1.9.0 in /usr/local/lib/python3.6/dist-packages (from keras) (1.15.0)\n",
            "Requirement already satisfied: numpy>=1.9.1 in /usr/local/lib/python3.6/dist-packages (from keras) (1.18.5)\n",
            "Requirement already satisfied: keras-applications>=1.0.6 in /usr/local/lib/python3.6/dist-packages (from keras) (1.0.8)\n",
            "Requirement already satisfied: pyyaml in /usr/local/lib/python3.6/dist-packages (from keras) (3.13)\n",
            "Requirement already satisfied: keras-preprocessing>=1.0.5 in /usr/local/lib/python3.6/dist-packages (from keras) (1.1.2)\n",
            "Requirement already satisfied: scipy>=0.14 in /usr/local/lib/python3.6/dist-packages (from keras) (1.4.1)\n",
            "Requirement already satisfied: h5py in /usr/local/lib/python3.6/dist-packages (from keras) (2.10.0)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "Using TensorFlow backend.\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "yuX06GVdok0-",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def _gen_captcha(img_dir, num_of_letters, num_of_repetition, width, height):\n",
        "    if os.path.exists(img_dir):\n",
        "        shutil.rmtree(img_dir)\n",
        "    if not os.path.exists(img_dir):\n",
        "        os.makedirs(img_dir)\n",
        "\n",
        "    image = ImageCaptcha(width=width, height=height)\n",
        "\n",
        "    for counter in range(num_of_repetition):\n",
        "        print('generating %d/%d' % (counter + 1, num_of_repetition))\n",
        "        for i in itertools.permutations([str(c) for c in range(10)], num_of_letters):\n",
        "            captcha = ''.join(i)\n",
        "            fn = os.path.join(img_dir, '%s_%s.png' % (captcha, uuid.uuid4()))\n",
        "            image.write(captcha, fn)\n",
        "\n",
        "\n",
        "def gen_dataset(path, num_of_repetition, num_of_letters, width, height):\n",
        "    _gen_captcha(os.path.join(path, 'data'), num_of_letters, num_of_repetition, width, height)\n",
        "    print('Finished Data Generation')"
      ],
      "execution_count": 2,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "MQYOsJyioy5g",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "BATCH_SIZE = 128\n",
        "NUM_OF_LETTERS = 5\n",
        "EPOCHS = 50\n",
        "IMG_ROW, IMG_COLS = 50, 135\n",
        "\n",
        "\n",
        "# Non-configs\n",
        "PATH = os.getcwd()\n",
        "DATA_PATH = os.path.join(PATH, 'train')"
      ],
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ItSQ3uTvo5YW",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def load_data(path, test_split=0.1):\n",
        "    print ('loading dataset...')\n",
        "    y_train = []\n",
        "    y_test = []\n",
        "    x_train = []\n",
        "    x_test = []\n",
        "\n",
        "    # r=root, d=directories, f = files\n",
        "    counter = 0\n",
        "    for r, d, f in os.walk(path):\n",
        "        for fl in f:\n",
        "            if '.png' in fl:\n",
        "                flr = fl.split('_')[0]\n",
        "                counter += 1\n",
        "                label = np.zeros((NUM_OF_LETTERS, 10))\n",
        "                for i in range(NUM_OF_LETTERS):\n",
        "                    label[i, int(flr[i])] = 1\n",
        "\n",
        "                img = cv2.imread(os.path.join(r, fl))\n",
        "                img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)\n",
        "                img = cv2.resize(img, (int(135/2), int(50/2)), interpolation=cv2.INTER_AREA)\n",
        "                img = np.reshape(img, (img.shape[0], img.shape[1], 1))\n",
        "\n",
        "                if random() < test_split:\n",
        "                    y_test.append(label)\n",
        "                    x_test.append(img)\n",
        "                else:\n",
        "                    y_train.append(label)\n",
        "                    x_train.append(img)\n",
        "\n",
        "    print('dataset size:', counter, '(train=%d, test=%d)' % (len(y_train), len(y_test)))\n",
        "    return np.array(x_train), np.array(y_train), np.array(x_test), np.array(y_test)\n"
      ],
      "execution_count": 4,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "bCqtzWWvo798",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 70
        },
        "outputId": "d668ac3e-c0c0-435f-d688-789aa49a5212"
      },
      "source": [
        "if not os.path.exists(DATA_PATH):\n",
        "    print('Generating Dataset')\n",
        "    gen_dataset(DATA_PATH, 8 , NUM_OF_LETTERS, IMG_COLS, IMG_ROW)"
      ],
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Generating Dataset\n",
            "generating 1/1\n",
            "Finished Data Generation\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "AenCMkmco-jv",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 52
        },
        "outputId": "25cfddd7-6be1-421f-ba54-fe524c90c115"
      },
      "source": [
        "x_train, y_train, x_test, y_test = load_data(DATA_PATH)\n",
        "x_train = x_train.astype('float32')\n",
        "x_test = x_test.astype('float32')\n",
        "x_train /= 255\n",
        "x_test /= 255"
      ],
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "loading dataset...\n",
            "dataset size: 30240 (train=27133, test=3107)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zqV-rssUuhIy",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 88
        },
        "outputId": "578dcd3f-95af-45f9-9601-e8d86cca9cd6"
      },
      "source": [
        "print(x_train.shape)\n",
        "print(y_train.shape)\n",
        "print(x_test.shape)\n",
        "print(y_test.shape)"
      ],
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(27133, 25, 67, 1)\n",
            "(27133, 5, 10)\n",
            "(3107, 25, 67, 1)\n",
            "(3107, 5, 10)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "eASawvTyvMaL",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "s_train = []\n",
        "s_test = []\n",
        "for i in range(NUM_OF_LETTERS):\n",
        "    s_train.append(y_train[:, i, :])\n",
        "    s_test.append(y_test[:, i, :])"
      ],
      "execution_count": 8,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DCom2NFKvVF0",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "save_dir = os.path.join(PATH, 'saved_models')\n",
        "model_name = 'keras_cifar10_trained_model.h5'"
      ],
      "execution_count": 9,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XS0vWhEuvXgo",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "input_layer = Input((25, 67, 1))\n",
        "x = Conv2D(filters=32, kernel_size=(5, 5), padding='same', activation='relu')(input_layer)\n",
        "x = MaxPooling2D(pool_size=(2, 2))(x)\n",
        "\n",
        "x = Conv2D(filters=48, kernel_size=(5, 5), padding='same', activation='relu')(x)\n",
        "x = MaxPooling2D(pool_size=(2, 2))(x)\n",
        "\n",
        "x = Conv2D(filters=64, kernel_size=(5, 5), padding='same', activation='relu')(x)\n",
        "x = MaxPooling2D(pool_size=(2, 2))(x)\n",
        "\n",
        "x = Dropout(0.3)(x)\n",
        "x = Flatten()(x)\n",
        "x = Dense(512, activation='relu')(x)\n",
        "x = Dropout(0.3)(x)\n",
        "\n",
        "out = [Dense(10, name='digit%d' % i, activation='softmax')(x) for i in range(NUM_OF_LETTERS)]\n",
        "model = Model(inputs=input_layer, outputs=out)"
      ],
      "execution_count": 10,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_1aTZ32pvZpl",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Uncomment these if you want to load and continue an already trained network\n",
        "\n",
        "# model_path = os.path.join(save_dir, model_name)\n",
        "# model = keras.models.load_model(model_path)"
      ],
      "execution_count": 11,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "mvyFxZTlvfKR",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 728
        },
        "outputId": "a11d69b7-98fb-4f62-e951-a5f0fbb843db"
      },
      "source": [
        "# initiate Adam optimizer\n",
        "\n",
        "model.compile(loss='binary_crossentropy',\n",
        "              optimizer='adam',\n",
        "              metrics=['accuracy'])\n",
        "\n",
        "\n",
        "model.summary() "
      ],
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"model_1\"\n",
            "__________________________________________________________________________________________________\n",
            "Layer (type)                    Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            "input_1 (InputLayer)            (None, 25, 67, 1)    0                                            \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_1 (Conv2D)               (None, 25, 67, 32)   832         input_1[0][0]                    \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d_1 (MaxPooling2D)  (None, 12, 33, 32)   0           conv2d_1[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_2 (Conv2D)               (None, 12, 33, 48)   38448       max_pooling2d_1[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d_2 (MaxPooling2D)  (None, 6, 16, 48)    0           conv2d_2[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_3 (Conv2D)               (None, 6, 16, 64)    76864       max_pooling2d_2[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d_3 (MaxPooling2D)  (None, 3, 8, 64)     0           conv2d_3[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "dropout_1 (Dropout)             (None, 3, 8, 64)     0           max_pooling2d_3[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "flatten_1 (Flatten)             (None, 1536)         0           dropout_1[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "dense_1 (Dense)                 (None, 512)          786944      flatten_1[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "dropout_2 (Dropout)             (None, 512)          0           dense_1[0][0]                    \n",
            "__________________________________________________________________________________________________\n",
            "digit0 (Dense)                  (None, 10)           5130        dropout_2[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "digit1 (Dense)                  (None, 10)           5130        dropout_2[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "digit2 (Dense)                  (None, 10)           5130        dropout_2[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "digit3 (Dense)                  (None, 10)           5130        dropout_2[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "digit4 (Dense)                  (None, 10)           5130        dropout_2[0][0]                  \n",
            "==================================================================================================\n",
            "Total params: 928,738\n",
            "Trainable params: 928,738\n",
            "Non-trainable params: 0\n",
            "__________________________________________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xNW40AXivcK7",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "hist_train_loss_digit = {i:[] for i in range(5)}\n",
        "hist_test_loss_digit = {i:[] for i in range(5)}\n",
        "\n",
        "hist_train_acc_digit = {i:[] for i in range(5)}\n",
        "hist_test_acc_digit = {i:[] for i in range(5)}\n",
        "\n",
        "hist_train_loss = []\n",
        "hist_test_loss = []\n",
        "\n",
        "hist_train_acc = []\n",
        "hist_test_acc = []"
      ],
      "execution_count": 13,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VFeIxaatvhap",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "digit_acc = [[] for _ in range(NUM_OF_LETTERS)]\n",
        "val_digit_acc = [[] for _ in range(NUM_OF_LETTERS)]\n",
        "loss = []\n",
        "val_loss = []"
      ],
      "execution_count": 14,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "IQTy-mCRvjIz",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "eac887df-defc-4167-d7ab-70789638feda"
      },
      "source": [
        "history = model.fit(x_train, s_train,\n",
        "                    batch_size=BATCH_SIZE,\n",
        "                    epochs=EPOCHS,\n",
        "                    verbose=1,\n",
        "                    validation_data=(x_test, s_test)\n",
        "                   )"
      ],
      "execution_count": 15,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 27133 samples, validate on 3107 samples\n",
            "Epoch 1/50\n",
            "27133/27133 [==============================] - 12s 450us/step - loss: 1.5333 - digit0_loss: 0.2999 - digit1_loss: 0.3131 - digit2_loss: 0.3125 - digit3_loss: 0.3098 - digit4_loss: 0.2981 - digit0_accuracy: 0.9023 - digit1_accuracy: 0.9005 - digit2_accuracy: 0.9005 - digit3_accuracy: 0.9008 - digit4_accuracy: 0.9025 - val_loss: 1.1149 - val_digit0_loss: 0.1908 - val_digit1_loss: 0.2526 - val_digit2_loss: 0.2525 - val_digit3_loss: 0.2368 - val_digit4_loss: 0.1819 - val_digit0_accuracy: 0.9217 - val_digit1_accuracy: 0.9073 - val_digit2_accuracy: 0.9058 - val_digit3_accuracy: 0.9102 - val_digit4_accuracy: 0.9255\n",
            "Epoch 2/50\n",
            "27133/27133 [==============================] - 4s 156us/step - loss: 0.8478 - digit0_loss: 0.1164 - digit1_loss: 0.2026 - digit2_loss: 0.2203 - digit3_loss: 0.1911 - digit4_loss: 0.1173 - digit0_accuracy: 0.9551 - digit1_accuracy: 0.9238 - digit2_accuracy: 0.9176 - digit3_accuracy: 0.9279 - digit4_accuracy: 0.9549 - val_loss: 0.5127 - val_digit0_loss: 0.0489 - val_digit1_loss: 0.1342 - val_digit2_loss: 0.1576 - val_digit3_loss: 0.1202 - val_digit4_loss: 0.0516 - val_digit0_accuracy: 0.9841 - val_digit1_accuracy: 0.9504 - val_digit2_accuracy: 0.9405 - val_digit3_accuracy: 0.9574 - val_digit4_accuracy: 0.9818\n",
            "Epoch 3/50\n",
            "27133/27133 [==============================] - 4s 156us/step - loss: 0.5337 - digit0_loss: 0.0542 - digit1_loss: 0.1342 - digit2_loss: 0.1639 - digit3_loss: 0.1227 - digit4_loss: 0.0587 - digit0_accuracy: 0.9808 - digit1_accuracy: 0.9509 - digit2_accuracy: 0.9392 - digit3_accuracy: 0.9549 - digit4_accuracy: 0.9788 - val_loss: 0.3570 - val_digit0_loss: 0.0314 - val_digit1_loss: 0.0925 - val_digit2_loss: 0.1180 - val_digit3_loss: 0.0827 - val_digit4_loss: 0.0319 - val_digit0_accuracy: 0.9902 - val_digit1_accuracy: 0.9672 - val_digit2_accuracy: 0.9572 - val_digit3_accuracy: 0.9715 - val_digit4_accuracy: 0.9892\n",
            "Epoch 4/50\n",
            "27133/27133 [==============================] - 4s 156us/step - loss: 0.3964 - digit0_loss: 0.0395 - digit1_loss: 0.0995 - digit2_loss: 0.1240 - digit3_loss: 0.0910 - digit4_loss: 0.0425 - digit0_accuracy: 0.9860 - digit1_accuracy: 0.9639 - digit2_accuracy: 0.9544 - digit3_accuracy: 0.9669 - digit4_accuracy: 0.9849 - val_loss: 0.2556 - val_digit0_loss: 0.0219 - val_digit1_loss: 0.0651 - val_digit2_loss: 0.0839 - val_digit3_loss: 0.0604 - val_digit4_loss: 0.0233 - val_digit0_accuracy: 0.9924 - val_digit1_accuracy: 0.9771 - val_digit2_accuracy: 0.9697 - val_digit3_accuracy: 0.9787 - val_digit4_accuracy: 0.9918\n",
            "Epoch 5/50\n",
            "27133/27133 [==============================] - 4s 155us/step - loss: 0.3083 - digit0_loss: 0.0317 - digit1_loss: 0.0761 - digit2_loss: 0.0937 - digit3_loss: 0.0721 - digit4_loss: 0.0347 - digit0_accuracy: 0.9887 - digit1_accuracy: 0.9726 - digit2_accuracy: 0.9660 - digit3_accuracy: 0.9735 - digit4_accuracy: 0.9874 - val_loss: 0.1984 - val_digit0_loss: 0.0189 - val_digit1_loss: 0.0486 - val_digit2_loss: 0.0618 - val_digit3_loss: 0.0480 - val_digit4_loss: 0.0213 - val_digit0_accuracy: 0.9936 - val_digit1_accuracy: 0.9836 - val_digit2_accuracy: 0.9793 - val_digit3_accuracy: 0.9827 - val_digit4_accuracy: 0.9933\n",
            "Epoch 6/50\n",
            "27133/27133 [==============================] - 4s 155us/step - loss: 0.2482 - digit0_loss: 0.0262 - digit1_loss: 0.0603 - digit2_loss: 0.0737 - digit3_loss: 0.0587 - digit4_loss: 0.0292 - digit0_accuracy: 0.9907 - digit1_accuracy: 0.9787 - digit2_accuracy: 0.9736 - digit3_accuracy: 0.9789 - digit4_accuracy: 0.9895 - val_loss: 0.1459 - val_digit0_loss: 0.0141 - val_digit1_loss: 0.0360 - val_digit2_loss: 0.0439 - val_digit3_loss: 0.0355 - val_digit4_loss: 0.0159 - val_digit0_accuracy: 0.9952 - val_digit1_accuracy: 0.9878 - val_digit2_accuracy: 0.9854 - val_digit3_accuracy: 0.9874 - val_digit4_accuracy: 0.9946\n",
            "Epoch 7/50\n",
            "27133/27133 [==============================] - 4s 158us/step - loss: 0.2105 - digit0_loss: 0.0230 - digit1_loss: 0.0509 - digit2_loss: 0.0612 - digit3_loss: 0.0502 - digit4_loss: 0.0252 - digit0_accuracy: 0.9918 - digit1_accuracy: 0.9820 - digit2_accuracy: 0.9782 - digit3_accuracy: 0.9816 - digit4_accuracy: 0.9910 - val_loss: 0.1288 - val_digit0_loss: 0.0125 - val_digit1_loss: 0.0295 - val_digit2_loss: 0.0375 - val_digit3_loss: 0.0333 - val_digit4_loss: 0.0153 - val_digit0_accuracy: 0.9959 - val_digit1_accuracy: 0.9902 - val_digit2_accuracy: 0.9871 - val_digit3_accuracy: 0.9887 - val_digit4_accuracy: 0.9946\n",
            "Epoch 8/50\n",
            "27133/27133 [==============================] - 4s 162us/step - loss: 0.1811 - digit0_loss: 0.0193 - digit1_loss: 0.0447 - digit2_loss: 0.0520 - digit3_loss: 0.0425 - digit4_loss: 0.0226 - digit0_accuracy: 0.9932 - digit1_accuracy: 0.9841 - digit2_accuracy: 0.9817 - digit3_accuracy: 0.9845 - digit4_accuracy: 0.9919 - val_loss: 0.1113 - val_digit0_loss: 0.0109 - val_digit1_loss: 0.0269 - val_digit2_loss: 0.0322 - val_digit3_loss: 0.0284 - val_digit4_loss: 0.0127 - val_digit0_accuracy: 0.9958 - val_digit1_accuracy: 0.9913 - val_digit2_accuracy: 0.9894 - val_digit3_accuracy: 0.9904 - val_digit4_accuracy: 0.9958\n",
            "Epoch 9/50\n",
            "27133/27133 [==============================] - 4s 162us/step - loss: 0.1595 - digit0_loss: 0.0170 - digit1_loss: 0.0383 - digit2_loss: 0.0468 - digit3_loss: 0.0376 - digit4_loss: 0.0197 - digit0_accuracy: 0.9939 - digit1_accuracy: 0.9864 - digit2_accuracy: 0.9833 - digit3_accuracy: 0.9867 - digit4_accuracy: 0.9929 - val_loss: 0.0997 - val_digit0_loss: 0.0099 - val_digit1_loss: 0.0234 - val_digit2_loss: 0.0272 - val_digit3_loss: 0.0258 - val_digit4_loss: 0.0128 - val_digit0_accuracy: 0.9968 - val_digit1_accuracy: 0.9921 - val_digit2_accuracy: 0.9910 - val_digit3_accuracy: 0.9908 - val_digit4_accuracy: 0.9958\n",
            "Epoch 10/50\n",
            "27133/27133 [==============================] - 4s 157us/step - loss: 0.1402 - digit0_loss: 0.0165 - digit1_loss: 0.0327 - digit2_loss: 0.0401 - digit3_loss: 0.0331 - digit4_loss: 0.0178 - digit0_accuracy: 0.9940 - digit1_accuracy: 0.9884 - digit2_accuracy: 0.9858 - digit3_accuracy: 0.9884 - digit4_accuracy: 0.9937 - val_loss: 0.0835 - val_digit0_loss: 0.0085 - val_digit1_loss: 0.0195 - val_digit2_loss: 0.0226 - val_digit3_loss: 0.0215 - val_digit4_loss: 0.0108 - val_digit0_accuracy: 0.9971 - val_digit1_accuracy: 0.9933 - val_digit2_accuracy: 0.9927 - val_digit3_accuracy: 0.9920 - val_digit4_accuracy: 0.9963\n",
            "Epoch 11/50\n",
            "27133/27133 [==============================] - 4s 157us/step - loss: 0.1274 - digit0_loss: 0.0142 - digit1_loss: 0.0301 - digit2_loss: 0.0364 - digit3_loss: 0.0302 - digit4_loss: 0.0165 - digit0_accuracy: 0.9949 - digit1_accuracy: 0.9894 - digit2_accuracy: 0.9872 - digit3_accuracy: 0.9890 - digit4_accuracy: 0.9938 - val_loss: 0.0781 - val_digit0_loss: 0.0078 - val_digit1_loss: 0.0188 - val_digit2_loss: 0.0231 - val_digit3_loss: 0.0179 - val_digit4_loss: 0.0097 - val_digit0_accuracy: 0.9973 - val_digit1_accuracy: 0.9937 - val_digit2_accuracy: 0.9925 - val_digit3_accuracy: 0.9935 - val_digit4_accuracy: 0.9966\n",
            "Epoch 12/50\n",
            "27133/27133 [==============================] - 4s 155us/step - loss: 0.1171 - digit0_loss: 0.0128 - digit1_loss: 0.0291 - digit2_loss: 0.0325 - digit3_loss: 0.0278 - digit4_loss: 0.0148 - digit0_accuracy: 0.9953 - digit1_accuracy: 0.9895 - digit2_accuracy: 0.9885 - digit3_accuracy: 0.9903 - digit4_accuracy: 0.9946 - val_loss: 0.0751 - val_digit0_loss: 0.0072 - val_digit1_loss: 0.0173 - val_digit2_loss: 0.0208 - val_digit3_loss: 0.0195 - val_digit4_loss: 0.0098 - val_digit0_accuracy: 0.9976 - val_digit1_accuracy: 0.9939 - val_digit2_accuracy: 0.9930 - val_digit3_accuracy: 0.9939 - val_digit4_accuracy: 0.9969\n",
            "Epoch 13/50\n",
            "27133/27133 [==============================] - 4s 156us/step - loss: 0.1067 - digit0_loss: 0.0118 - digit1_loss: 0.0254 - digit2_loss: 0.0296 - digit3_loss: 0.0257 - digit4_loss: 0.0142 - digit0_accuracy: 0.9957 - digit1_accuracy: 0.9912 - digit2_accuracy: 0.9895 - digit3_accuracy: 0.9908 - digit4_accuracy: 0.9949 - val_loss: 0.0770 - val_digit0_loss: 0.0081 - val_digit1_loss: 0.0180 - val_digit2_loss: 0.0204 - val_digit3_loss: 0.0199 - val_digit4_loss: 0.0098 - val_digit0_accuracy: 0.9968 - val_digit1_accuracy: 0.9938 - val_digit2_accuracy: 0.9931 - val_digit3_accuracy: 0.9927 - val_digit4_accuracy: 0.9967\n",
            "Epoch 14/50\n",
            "27133/27133 [==============================] - 4s 156us/step - loss: 0.0982 - digit0_loss: 0.0108 - digit1_loss: 0.0228 - digit2_loss: 0.0277 - digit3_loss: 0.0233 - digit4_loss: 0.0136 - digit0_accuracy: 0.9961 - digit1_accuracy: 0.9918 - digit2_accuracy: 0.9902 - digit3_accuracy: 0.9914 - digit4_accuracy: 0.9953 - val_loss: 0.0708 - val_digit0_loss: 0.0079 - val_digit1_loss: 0.0150 - val_digit2_loss: 0.0200 - val_digit3_loss: 0.0184 - val_digit4_loss: 0.0086 - val_digit0_accuracy: 0.9974 - val_digit1_accuracy: 0.9948 - val_digit2_accuracy: 0.9930 - val_digit3_accuracy: 0.9932 - val_digit4_accuracy: 0.9971\n",
            "Epoch 15/50\n",
            "27133/27133 [==============================] - 4s 156us/step - loss: 0.0907 - digit0_loss: 0.0100 - digit1_loss: 0.0220 - digit2_loss: 0.0248 - digit3_loss: 0.0216 - digit4_loss: 0.0122 - digit0_accuracy: 0.9964 - digit1_accuracy: 0.9923 - digit2_accuracy: 0.9913 - digit3_accuracy: 0.9921 - digit4_accuracy: 0.9956 - val_loss: 0.0597 - val_digit0_loss: 0.0053 - val_digit1_loss: 0.0134 - val_digit2_loss: 0.0159 - val_digit3_loss: 0.0153 - val_digit4_loss: 0.0091 - val_digit0_accuracy: 0.9982 - val_digit1_accuracy: 0.9957 - val_digit2_accuracy: 0.9945 - val_digit3_accuracy: 0.9943 - val_digit4_accuracy: 0.9972\n",
            "Epoch 16/50\n",
            "27133/27133 [==============================] - 4s 156us/step - loss: 0.0856 - digit0_loss: 0.0092 - digit1_loss: 0.0215 - digit2_loss: 0.0232 - digit3_loss: 0.0203 - digit4_loss: 0.0114 - digit0_accuracy: 0.9967 - digit1_accuracy: 0.9923 - digit2_accuracy: 0.9920 - digit3_accuracy: 0.9926 - digit4_accuracy: 0.9960 - val_loss: 0.0634 - val_digit0_loss: 0.0064 - val_digit1_loss: 0.0145 - val_digit2_loss: 0.0171 - val_digit3_loss: 0.0154 - val_digit4_loss: 0.0092 - val_digit0_accuracy: 0.9980 - val_digit1_accuracy: 0.9953 - val_digit2_accuracy: 0.9940 - val_digit3_accuracy: 0.9943 - val_digit4_accuracy: 0.9970\n",
            "Epoch 17/50\n",
            "27133/27133 [==============================] - 4s 156us/step - loss: 0.0795 - digit0_loss: 0.0088 - digit1_loss: 0.0189 - digit2_loss: 0.0221 - digit3_loss: 0.0190 - digit4_loss: 0.0107 - digit0_accuracy: 0.9967 - digit1_accuracy: 0.9933 - digit2_accuracy: 0.9924 - digit3_accuracy: 0.9934 - digit4_accuracy: 0.9961 - val_loss: 0.0544 - val_digit0_loss: 0.0054 - val_digit1_loss: 0.0123 - val_digit2_loss: 0.0150 - val_digit3_loss: 0.0140 - val_digit4_loss: 0.0073 - val_digit0_accuracy: 0.9983 - val_digit1_accuracy: 0.9958 - val_digit2_accuracy: 0.9949 - val_digit3_accuracy: 0.9953 - val_digit4_accuracy: 0.9977\n",
            "Epoch 18/50\n",
            "27133/27133 [==============================] - 4s 155us/step - loss: 0.0750 - digit0_loss: 0.0084 - digit1_loss: 0.0181 - digit2_loss: 0.0207 - digit3_loss: 0.0177 - digit4_loss: 0.0103 - digit0_accuracy: 0.9970 - digit1_accuracy: 0.9936 - digit2_accuracy: 0.9927 - digit3_accuracy: 0.9938 - digit4_accuracy: 0.9963 - val_loss: 0.0558 - val_digit0_loss: 0.0058 - val_digit1_loss: 0.0122 - val_digit2_loss: 0.0157 - val_digit3_loss: 0.0140 - val_digit4_loss: 0.0073 - val_digit0_accuracy: 0.9981 - val_digit1_accuracy: 0.9961 - val_digit2_accuracy: 0.9948 - val_digit3_accuracy: 0.9953 - val_digit4_accuracy: 0.9973\n",
            "Epoch 19/50\n",
            "27133/27133 [==============================] - 4s 156us/step - loss: 0.0695 - digit0_loss: 0.0079 - digit1_loss: 0.0166 - digit2_loss: 0.0191 - digit3_loss: 0.0167 - digit4_loss: 0.0092 - digit0_accuracy: 0.9972 - digit1_accuracy: 0.9941 - digit2_accuracy: 0.9932 - digit3_accuracy: 0.9940 - digit4_accuracy: 0.9968 - val_loss: 0.0543 - val_digit0_loss: 0.0049 - val_digit1_loss: 0.0124 - val_digit2_loss: 0.0143 - val_digit3_loss: 0.0144 - val_digit4_loss: 0.0077 - val_digit0_accuracy: 0.9980 - val_digit1_accuracy: 0.9954 - val_digit2_accuracy: 0.9955 - val_digit3_accuracy: 0.9952 - val_digit4_accuracy: 0.9976\n",
            "Epoch 20/50\n",
            "27133/27133 [==============================] - 4s 155us/step - loss: 0.0665 - digit0_loss: 0.0079 - digit1_loss: 0.0163 - digit2_loss: 0.0180 - digit3_loss: 0.0155 - digit4_loss: 0.0087 - digit0_accuracy: 0.9971 - digit1_accuracy: 0.9943 - digit2_accuracy: 0.9936 - digit3_accuracy: 0.9944 - digit4_accuracy: 0.9970 - val_loss: 0.0542 - val_digit0_loss: 0.0061 - val_digit1_loss: 0.0122 - val_digit2_loss: 0.0156 - val_digit3_loss: 0.0131 - val_digit4_loss: 0.0068 - val_digit0_accuracy: 0.9978 - val_digit1_accuracy: 0.9961 - val_digit2_accuracy: 0.9951 - val_digit3_accuracy: 0.9953 - val_digit4_accuracy: 0.9975\n",
            "Epoch 21/50\n",
            "27133/27133 [==============================] - 4s 156us/step - loss: 0.0650 - digit0_loss: 0.0073 - digit1_loss: 0.0156 - digit2_loss: 0.0181 - digit3_loss: 0.0158 - digit4_loss: 0.0082 - digit0_accuracy: 0.9974 - digit1_accuracy: 0.9942 - digit2_accuracy: 0.9937 - digit3_accuracy: 0.9942 - digit4_accuracy: 0.9970 - val_loss: 0.0554 - val_digit0_loss: 0.0046 - val_digit1_loss: 0.0128 - val_digit2_loss: 0.0161 - val_digit3_loss: 0.0145 - val_digit4_loss: 0.0073 - val_digit0_accuracy: 0.9983 - val_digit1_accuracy: 0.9958 - val_digit2_accuracy: 0.9944 - val_digit3_accuracy: 0.9954 - val_digit4_accuracy: 0.9972\n",
            "Epoch 22/50\n",
            "27133/27133 [==============================] - 4s 156us/step - loss: 0.0600 - digit0_loss: 0.0070 - digit1_loss: 0.0145 - digit2_loss: 0.0163 - digit3_loss: 0.0142 - digit4_loss: 0.0079 - digit0_accuracy: 0.9975 - digit1_accuracy: 0.9948 - digit2_accuracy: 0.9943 - digit3_accuracy: 0.9949 - digit4_accuracy: 0.9972 - val_loss: 0.0502 - val_digit0_loss: 0.0041 - val_digit1_loss: 0.0115 - val_digit2_loss: 0.0140 - val_digit3_loss: 0.0132 - val_digit4_loss: 0.0071 - val_digit0_accuracy: 0.9986 - val_digit1_accuracy: 0.9958 - val_digit2_accuracy: 0.9950 - val_digit3_accuracy: 0.9955 - val_digit4_accuracy: 0.9978\n",
            "Epoch 23/50\n",
            "27133/27133 [==============================] - 4s 159us/step - loss: 0.0578 - digit0_loss: 0.0069 - digit1_loss: 0.0135 - digit2_loss: 0.0163 - digit3_loss: 0.0135 - digit4_loss: 0.0075 - digit0_accuracy: 0.9976 - digit1_accuracy: 0.9952 - digit2_accuracy: 0.9943 - digit3_accuracy: 0.9950 - digit4_accuracy: 0.9973 - val_loss: 0.0526 - val_digit0_loss: 0.0053 - val_digit1_loss: 0.0118 - val_digit2_loss: 0.0142 - val_digit3_loss: 0.0145 - val_digit4_loss: 0.0067 - val_digit0_accuracy: 0.9979 - val_digit1_accuracy: 0.9959 - val_digit2_accuracy: 0.9951 - val_digit3_accuracy: 0.9954 - val_digit4_accuracy: 0.9977\n",
            "Epoch 24/50\n",
            "27133/27133 [==============================] - 4s 157us/step - loss: 0.0569 - digit0_loss: 0.0069 - digit1_loss: 0.0136 - digit2_loss: 0.0159 - digit3_loss: 0.0127 - digit4_loss: 0.0078 - digit0_accuracy: 0.9975 - digit1_accuracy: 0.9951 - digit2_accuracy: 0.9944 - digit3_accuracy: 0.9954 - digit4_accuracy: 0.9972 - val_loss: 0.0549 - val_digit0_loss: 0.0057 - val_digit1_loss: 0.0133 - val_digit2_loss: 0.0135 - val_digit3_loss: 0.0144 - val_digit4_loss: 0.0074 - val_digit0_accuracy: 0.9980 - val_digit1_accuracy: 0.9955 - val_digit2_accuracy: 0.9955 - val_digit3_accuracy: 0.9956 - val_digit4_accuracy: 0.9977\n",
            "Epoch 25/50\n",
            "27133/27133 [==============================] - 4s 157us/step - loss: 0.0536 - digit0_loss: 0.0061 - digit1_loss: 0.0131 - digit2_loss: 0.0143 - digit3_loss: 0.0128 - digit4_loss: 0.0074 - digit0_accuracy: 0.9978 - digit1_accuracy: 0.9954 - digit2_accuracy: 0.9950 - digit3_accuracy: 0.9954 - digit4_accuracy: 0.9974 - val_loss: 0.0509 - val_digit0_loss: 0.0040 - val_digit1_loss: 0.0126 - val_digit2_loss: 0.0141 - val_digit3_loss: 0.0131 - val_digit4_loss: 0.0066 - val_digit0_accuracy: 0.9986 - val_digit1_accuracy: 0.9956 - val_digit2_accuracy: 0.9949 - val_digit3_accuracy: 0.9959 - val_digit4_accuracy: 0.9980\n",
            "Epoch 26/50\n",
            "27133/27133 [==============================] - 4s 155us/step - loss: 0.0510 - digit0_loss: 0.0061 - digit1_loss: 0.0121 - digit2_loss: 0.0133 - digit3_loss: 0.0125 - digit4_loss: 0.0070 - digit0_accuracy: 0.9978 - digit1_accuracy: 0.9957 - digit2_accuracy: 0.9953 - digit3_accuracy: 0.9956 - digit4_accuracy: 0.9974 - val_loss: 0.0461 - val_digit0_loss: 0.0041 - val_digit1_loss: 0.0110 - val_digit2_loss: 0.0116 - val_digit3_loss: 0.0125 - val_digit4_loss: 0.0062 - val_digit0_accuracy: 0.9986 - val_digit1_accuracy: 0.9964 - val_digit2_accuracy: 0.9958 - val_digit3_accuracy: 0.9959 - val_digit4_accuracy: 0.9981\n",
            "Epoch 27/50\n",
            "27133/27133 [==============================] - 4s 156us/step - loss: 0.0483 - digit0_loss: 0.0051 - digit1_loss: 0.0121 - digit2_loss: 0.0132 - digit3_loss: 0.0116 - digit4_loss: 0.0062 - digit0_accuracy: 0.9982 - digit1_accuracy: 0.9959 - digit2_accuracy: 0.9954 - digit3_accuracy: 0.9959 - digit4_accuracy: 0.9979 - val_loss: 0.0494 - val_digit0_loss: 0.0047 - val_digit1_loss: 0.0119 - val_digit2_loss: 0.0128 - val_digit3_loss: 0.0127 - val_digit4_loss: 0.0068 - val_digit0_accuracy: 0.9985 - val_digit1_accuracy: 0.9963 - val_digit2_accuracy: 0.9956 - val_digit3_accuracy: 0.9955 - val_digit4_accuracy: 0.9977\n",
            "Epoch 28/50\n",
            "27133/27133 [==============================] - 4s 157us/step - loss: 0.0468 - digit0_loss: 0.0056 - digit1_loss: 0.0118 - digit2_loss: 0.0131 - digit3_loss: 0.0104 - digit4_loss: 0.0060 - digit0_accuracy: 0.9980 - digit1_accuracy: 0.9959 - digit2_accuracy: 0.9955 - digit3_accuracy: 0.9963 - digit4_accuracy: 0.9979 - val_loss: 0.0456 - val_digit0_loss: 0.0037 - val_digit1_loss: 0.0105 - val_digit2_loss: 0.0125 - val_digit3_loss: 0.0123 - val_digit4_loss: 0.0060 - val_digit0_accuracy: 0.9985 - val_digit1_accuracy: 0.9971 - val_digit2_accuracy: 0.9963 - val_digit3_accuracy: 0.9960 - val_digit4_accuracy: 0.9983\n",
            "Epoch 29/50\n",
            "27133/27133 [==============================] - 4s 157us/step - loss: 0.0486 - digit0_loss: 0.0061 - digit1_loss: 0.0113 - digit2_loss: 0.0131 - digit3_loss: 0.0115 - digit4_loss: 0.0065 - digit0_accuracy: 0.9978 - digit1_accuracy: 0.9961 - digit2_accuracy: 0.9954 - digit3_accuracy: 0.9960 - digit4_accuracy: 0.9978 - val_loss: 0.0503 - val_digit0_loss: 0.0038 - val_digit1_loss: 0.0111 - val_digit2_loss: 0.0146 - val_digit3_loss: 0.0134 - val_digit4_loss: 0.0067 - val_digit0_accuracy: 0.9985 - val_digit1_accuracy: 0.9960 - val_digit2_accuracy: 0.9952 - val_digit3_accuracy: 0.9957 - val_digit4_accuracy: 0.9977\n",
            "Epoch 30/50\n",
            "27133/27133 [==============================] - 4s 156us/step - loss: 0.0462 - digit0_loss: 0.0050 - digit1_loss: 0.0107 - digit2_loss: 0.0134 - digit3_loss: 0.0106 - digit4_loss: 0.0065 - digit0_accuracy: 0.9982 - digit1_accuracy: 0.9962 - digit2_accuracy: 0.9955 - digit3_accuracy: 0.9963 - digit4_accuracy: 0.9977 - val_loss: 0.0484 - val_digit0_loss: 0.0042 - val_digit1_loss: 0.0106 - val_digit2_loss: 0.0133 - val_digit3_loss: 0.0134 - val_digit4_loss: 0.0063 - val_digit0_accuracy: 0.9984 - val_digit1_accuracy: 0.9965 - val_digit2_accuracy: 0.9956 - val_digit3_accuracy: 0.9953 - val_digit4_accuracy: 0.9977\n",
            "Epoch 31/50\n",
            "27133/27133 [==============================] - 4s 158us/step - loss: 0.0459 - digit0_loss: 0.0055 - digit1_loss: 0.0104 - digit2_loss: 0.0124 - digit3_loss: 0.0112 - digit4_loss: 0.0065 - digit0_accuracy: 0.9981 - digit1_accuracy: 0.9964 - digit2_accuracy: 0.9955 - digit3_accuracy: 0.9960 - digit4_accuracy: 0.9976 - val_loss: 0.0458 - val_digit0_loss: 0.0036 - val_digit1_loss: 0.0108 - val_digit2_loss: 0.0129 - val_digit3_loss: 0.0118 - val_digit4_loss: 0.0062 - val_digit0_accuracy: 0.9986 - val_digit1_accuracy: 0.9965 - val_digit2_accuracy: 0.9959 - val_digit3_accuracy: 0.9959 - val_digit4_accuracy: 0.9982\n",
            "Epoch 32/50\n",
            "27133/27133 [==============================] - 4s 157us/step - loss: 0.0433 - digit0_loss: 0.0054 - digit1_loss: 0.0108 - digit2_loss: 0.0111 - digit3_loss: 0.0096 - digit4_loss: 0.0064 - digit0_accuracy: 0.9981 - digit1_accuracy: 0.9962 - digit2_accuracy: 0.9961 - digit3_accuracy: 0.9966 - digit4_accuracy: 0.9977 - val_loss: 0.0444 - val_digit0_loss: 0.0030 - val_digit1_loss: 0.0106 - val_digit2_loss: 0.0134 - val_digit3_loss: 0.0116 - val_digit4_loss: 0.0057 - val_digit0_accuracy: 0.9988 - val_digit1_accuracy: 0.9969 - val_digit2_accuracy: 0.9957 - val_digit3_accuracy: 0.9966 - val_digit4_accuracy: 0.9983\n",
            "Epoch 33/50\n",
            "27133/27133 [==============================] - 4s 155us/step - loss: 0.0408 - digit0_loss: 0.0047 - digit1_loss: 0.0100 - digit2_loss: 0.0107 - digit3_loss: 0.0096 - digit4_loss: 0.0058 - digit0_accuracy: 0.9984 - digit1_accuracy: 0.9965 - digit2_accuracy: 0.9963 - digit3_accuracy: 0.9965 - digit4_accuracy: 0.9979 - val_loss: 0.0409 - val_digit0_loss: 0.0027 - val_digit1_loss: 0.0092 - val_digit2_loss: 0.0126 - val_digit3_loss: 0.0107 - val_digit4_loss: 0.0055 - val_digit0_accuracy: 0.9992 - val_digit1_accuracy: 0.9971 - val_digit2_accuracy: 0.9958 - val_digit3_accuracy: 0.9969 - val_digit4_accuracy: 0.9981\n",
            "Epoch 34/50\n",
            "27133/27133 [==============================] - 4s 155us/step - loss: 0.0389 - digit0_loss: 0.0041 - digit1_loss: 0.0098 - digit2_loss: 0.0101 - digit3_loss: 0.0100 - digit4_loss: 0.0050 - digit0_accuracy: 0.9985 - digit1_accuracy: 0.9965 - digit2_accuracy: 0.9965 - digit3_accuracy: 0.9965 - digit4_accuracy: 0.9982 - val_loss: 0.0445 - val_digit0_loss: 0.0035 - val_digit1_loss: 0.0096 - val_digit2_loss: 0.0126 - val_digit3_loss: 0.0126 - val_digit4_loss: 0.0058 - val_digit0_accuracy: 0.9989 - val_digit1_accuracy: 0.9969 - val_digit2_accuracy: 0.9959 - val_digit3_accuracy: 0.9958 - val_digit4_accuracy: 0.9981\n",
            "Epoch 35/50\n",
            "27133/27133 [==============================] - 4s 156us/step - loss: 0.0407 - digit0_loss: 0.0044 - digit1_loss: 0.0093 - digit2_loss: 0.0114 - digit3_loss: 0.0101 - digit4_loss: 0.0055 - digit0_accuracy: 0.9985 - digit1_accuracy: 0.9967 - digit2_accuracy: 0.9961 - digit3_accuracy: 0.9964 - digit4_accuracy: 0.9980 - val_loss: 0.0426 - val_digit0_loss: 0.0033 - val_digit1_loss: 0.0089 - val_digit2_loss: 0.0121 - val_digit3_loss: 0.0112 - val_digit4_loss: 0.0064 - val_digit0_accuracy: 0.9987 - val_digit1_accuracy: 0.9972 - val_digit2_accuracy: 0.9961 - val_digit3_accuracy: 0.9963 - val_digit4_accuracy: 0.9982\n",
            "Epoch 36/50\n",
            "27133/27133 [==============================] - 4s 157us/step - loss: 0.0383 - digit0_loss: 0.0043 - digit1_loss: 0.0095 - digit2_loss: 0.0104 - digit3_loss: 0.0090 - digit4_loss: 0.0051 - digit0_accuracy: 0.9985 - digit1_accuracy: 0.9968 - digit2_accuracy: 0.9965 - digit3_accuracy: 0.9969 - digit4_accuracy: 0.9982 - val_loss: 0.0493 - val_digit0_loss: 0.0039 - val_digit1_loss: 0.0118 - val_digit2_loss: 0.0141 - val_digit3_loss: 0.0122 - val_digit4_loss: 0.0068 - val_digit0_accuracy: 0.9986 - val_digit1_accuracy: 0.9965 - val_digit2_accuracy: 0.9958 - val_digit3_accuracy: 0.9962 - val_digit4_accuracy: 0.9978\n",
            "Epoch 37/50\n",
            "27133/27133 [==============================] - 4s 156us/step - loss: 0.0376 - digit0_loss: 0.0045 - digit1_loss: 0.0093 - digit2_loss: 0.0104 - digit3_loss: 0.0082 - digit4_loss: 0.0052 - digit0_accuracy: 0.9984 - digit1_accuracy: 0.9967 - digit2_accuracy: 0.9965 - digit3_accuracy: 0.9972 - digit4_accuracy: 0.9981 - val_loss: 0.0467 - val_digit0_loss: 0.0036 - val_digit1_loss: 0.0107 - val_digit2_loss: 0.0135 - val_digit3_loss: 0.0114 - val_digit4_loss: 0.0070 - val_digit0_accuracy: 0.9986 - val_digit1_accuracy: 0.9970 - val_digit2_accuracy: 0.9956 - val_digit3_accuracy: 0.9965 - val_digit4_accuracy: 0.9980\n",
            "Epoch 38/50\n",
            "27133/27133 [==============================] - 4s 156us/step - loss: 0.0354 - digit0_loss: 0.0045 - digit1_loss: 0.0090 - digit2_loss: 0.0090 - digit3_loss: 0.0080 - digit4_loss: 0.0049 - digit0_accuracy: 0.9984 - digit1_accuracy: 0.9969 - digit2_accuracy: 0.9967 - digit3_accuracy: 0.9972 - digit4_accuracy: 0.9983 - val_loss: 0.0450 - val_digit0_loss: 0.0050 - val_digit1_loss: 0.0102 - val_digit2_loss: 0.0127 - val_digit3_loss: 0.0112 - val_digit4_loss: 0.0056 - val_digit0_accuracy: 0.9985 - val_digit1_accuracy: 0.9966 - val_digit2_accuracy: 0.9957 - val_digit3_accuracy: 0.9964 - val_digit4_accuracy: 0.9984\n",
            "Epoch 39/50\n",
            "27133/27133 [==============================] - 4s 156us/step - loss: 0.0370 - digit0_loss: 0.0050 - digit1_loss: 0.0088 - digit2_loss: 0.0100 - digit3_loss: 0.0083 - digit4_loss: 0.0049 - digit0_accuracy: 0.9984 - digit1_accuracy: 0.9969 - digit2_accuracy: 0.9967 - digit3_accuracy: 0.9971 - digit4_accuracy: 0.9984 - val_loss: 0.0424 - val_digit0_loss: 0.0033 - val_digit1_loss: 0.0088 - val_digit2_loss: 0.0117 - val_digit3_loss: 0.0104 - val_digit4_loss: 0.0073 - val_digit0_accuracy: 0.9988 - val_digit1_accuracy: 0.9970 - val_digit2_accuracy: 0.9964 - val_digit3_accuracy: 0.9966 - val_digit4_accuracy: 0.9978\n",
            "Epoch 40/50\n",
            "27133/27133 [==============================] - 4s 156us/step - loss: 0.0343 - digit0_loss: 0.0045 - digit1_loss: 0.0077 - digit2_loss: 0.0091 - digit3_loss: 0.0084 - digit4_loss: 0.0048 - digit0_accuracy: 0.9984 - digit1_accuracy: 0.9973 - digit2_accuracy: 0.9967 - digit3_accuracy: 0.9970 - digit4_accuracy: 0.9982 - val_loss: 0.0455 - val_digit0_loss: 0.0046 - val_digit1_loss: 0.0106 - val_digit2_loss: 0.0132 - val_digit3_loss: 0.0109 - val_digit4_loss: 0.0058 - val_digit0_accuracy: 0.9983 - val_digit1_accuracy: 0.9964 - val_digit2_accuracy: 0.9962 - val_digit3_accuracy: 0.9968 - val_digit4_accuracy: 0.9980\n",
            "Epoch 41/50\n",
            "27133/27133 [==============================] - 4s 156us/step - loss: 0.0357 - digit0_loss: 0.0044 - digit1_loss: 0.0087 - digit2_loss: 0.0095 - digit3_loss: 0.0084 - digit4_loss: 0.0047 - digit0_accuracy: 0.9984 - digit1_accuracy: 0.9971 - digit2_accuracy: 0.9968 - digit3_accuracy: 0.9969 - digit4_accuracy: 0.9984 - val_loss: 0.0434 - val_digit0_loss: 0.0037 - val_digit1_loss: 0.0099 - val_digit2_loss: 0.0109 - val_digit3_loss: 0.0114 - val_digit4_loss: 0.0068 - val_digit0_accuracy: 0.9988 - val_digit1_accuracy: 0.9969 - val_digit2_accuracy: 0.9967 - val_digit3_accuracy: 0.9965 - val_digit4_accuracy: 0.9978\n",
            "Epoch 42/50\n",
            "27133/27133 [==============================] - 4s 159us/step - loss: 0.0356 - digit0_loss: 0.0043 - digit1_loss: 0.0087 - digit2_loss: 0.0093 - digit3_loss: 0.0086 - digit4_loss: 0.0047 - digit0_accuracy: 0.9985 - digit1_accuracy: 0.9970 - digit2_accuracy: 0.9967 - digit3_accuracy: 0.9970 - digit4_accuracy: 0.9984 - val_loss: 0.0440 - val_digit0_loss: 0.0048 - val_digit1_loss: 0.0099 - val_digit2_loss: 0.0112 - val_digit3_loss: 0.0100 - val_digit4_loss: 0.0069 - val_digit0_accuracy: 0.9986 - val_digit1_accuracy: 0.9968 - val_digit2_accuracy: 0.9965 - val_digit3_accuracy: 0.9969 - val_digit4_accuracy: 0.9981\n",
            "Epoch 43/50\n",
            "27133/27133 [==============================] - 4s 156us/step - loss: 0.0358 - digit0_loss: 0.0045 - digit1_loss: 0.0086 - digit2_loss: 0.0094 - digit3_loss: 0.0085 - digit4_loss: 0.0048 - digit0_accuracy: 0.9984 - digit1_accuracy: 0.9969 - digit2_accuracy: 0.9966 - digit3_accuracy: 0.9971 - digit4_accuracy: 0.9983 - val_loss: 0.0453 - val_digit0_loss: 0.0029 - val_digit1_loss: 0.0102 - val_digit2_loss: 0.0133 - val_digit3_loss: 0.0122 - val_digit4_loss: 0.0061 - val_digit0_accuracy: 0.9990 - val_digit1_accuracy: 0.9967 - val_digit2_accuracy: 0.9961 - val_digit3_accuracy: 0.9960 - val_digit4_accuracy: 0.9980\n",
            "Epoch 44/50\n",
            "27133/27133 [==============================] - 4s 156us/step - loss: 0.0313 - digit0_loss: 0.0040 - digit1_loss: 0.0076 - digit2_loss: 0.0087 - digit3_loss: 0.0070 - digit4_loss: 0.0041 - digit0_accuracy: 0.9986 - digit1_accuracy: 0.9975 - digit2_accuracy: 0.9970 - digit3_accuracy: 0.9977 - digit4_accuracy: 0.9987 - val_loss: 0.0428 - val_digit0_loss: 0.0044 - val_digit1_loss: 0.0088 - val_digit2_loss: 0.0117 - val_digit3_loss: 0.0118 - val_digit4_loss: 0.0061 - val_digit0_accuracy: 0.9986 - val_digit1_accuracy: 0.9970 - val_digit2_accuracy: 0.9964 - val_digit3_accuracy: 0.9961 - val_digit4_accuracy: 0.9985\n",
            "Epoch 45/50\n",
            "27133/27133 [==============================] - 4s 156us/step - loss: 0.0331 - digit0_loss: 0.0040 - digit1_loss: 0.0080 - digit2_loss: 0.0087 - digit3_loss: 0.0083 - digit4_loss: 0.0041 - digit0_accuracy: 0.9986 - digit1_accuracy: 0.9972 - digit2_accuracy: 0.9970 - digit3_accuracy: 0.9973 - digit4_accuracy: 0.9984 - val_loss: 0.0403 - val_digit0_loss: 0.0035 - val_digit1_loss: 0.0087 - val_digit2_loss: 0.0115 - val_digit3_loss: 0.0116 - val_digit4_loss: 0.0046 - val_digit0_accuracy: 0.9988 - val_digit1_accuracy: 0.9973 - val_digit2_accuracy: 0.9964 - val_digit3_accuracy: 0.9965 - val_digit4_accuracy: 0.9987\n",
            "Epoch 46/50\n",
            "27133/27133 [==============================] - 4s 156us/step - loss: 0.0286 - digit0_loss: 0.0036 - digit1_loss: 0.0070 - digit2_loss: 0.0074 - digit3_loss: 0.0068 - digit4_loss: 0.0038 - digit0_accuracy: 0.9987 - digit1_accuracy: 0.9976 - digit2_accuracy: 0.9974 - digit3_accuracy: 0.9976 - digit4_accuracy: 0.9986 - val_loss: 0.0414 - val_digit0_loss: 0.0037 - val_digit1_loss: 0.0087 - val_digit2_loss: 0.0129 - val_digit3_loss: 0.0099 - val_digit4_loss: 0.0058 - val_digit0_accuracy: 0.9986 - val_digit1_accuracy: 0.9976 - val_digit2_accuracy: 0.9958 - val_digit3_accuracy: 0.9967 - val_digit4_accuracy: 0.9983\n",
            "Epoch 47/50\n",
            "27133/27133 [==============================] - 4s 158us/step - loss: 0.0313 - digit0_loss: 0.0037 - digit1_loss: 0.0076 - digit2_loss: 0.0082 - digit3_loss: 0.0076 - digit4_loss: 0.0042 - digit0_accuracy: 0.9988 - digit1_accuracy: 0.9974 - digit2_accuracy: 0.9972 - digit3_accuracy: 0.9975 - digit4_accuracy: 0.9986 - val_loss: 0.0459 - val_digit0_loss: 0.0038 - val_digit1_loss: 0.0106 - val_digit2_loss: 0.0133 - val_digit3_loss: 0.0115 - val_digit4_loss: 0.0061 - val_digit0_accuracy: 0.9987 - val_digit1_accuracy: 0.9968 - val_digit2_accuracy: 0.9956 - val_digit3_accuracy: 0.9965 - val_digit4_accuracy: 0.9986\n",
            "Epoch 48/50\n",
            "27133/27133 [==============================] - 4s 156us/step - loss: 0.0303 - digit0_loss: 0.0040 - digit1_loss: 0.0077 - digit2_loss: 0.0075 - digit3_loss: 0.0068 - digit4_loss: 0.0042 - digit0_accuracy: 0.9986 - digit1_accuracy: 0.9974 - digit2_accuracy: 0.9974 - digit3_accuracy: 0.9976 - digit4_accuracy: 0.9984 - val_loss: 0.0467 - val_digit0_loss: 0.0041 - val_digit1_loss: 0.0088 - val_digit2_loss: 0.0134 - val_digit3_loss: 0.0123 - val_digit4_loss: 0.0073 - val_digit0_accuracy: 0.9989 - val_digit1_accuracy: 0.9974 - val_digit2_accuracy: 0.9963 - val_digit3_accuracy: 0.9966 - val_digit4_accuracy: 0.9981\n",
            "Epoch 49/50\n",
            "27133/27133 [==============================] - 4s 156us/step - loss: 0.0293 - digit0_loss: 0.0038 - digit1_loss: 0.0071 - digit2_loss: 0.0077 - digit3_loss: 0.0062 - digit4_loss: 0.0046 - digit0_accuracy: 0.9987 - digit1_accuracy: 0.9976 - digit2_accuracy: 0.9974 - digit3_accuracy: 0.9978 - digit4_accuracy: 0.9985 - val_loss: 0.0490 - val_digit0_loss: 0.0047 - val_digit1_loss: 0.0109 - val_digit2_loss: 0.0126 - val_digit3_loss: 0.0139 - val_digit4_loss: 0.0068 - val_digit0_accuracy: 0.9984 - val_digit1_accuracy: 0.9967 - val_digit2_accuracy: 0.9959 - val_digit3_accuracy: 0.9959 - val_digit4_accuracy: 0.9978\n",
            "Epoch 50/50\n",
            "27133/27133 [==============================] - 4s 157us/step - loss: 0.0307 - digit0_loss: 0.0040 - digit1_loss: 0.0073 - digit2_loss: 0.0083 - digit3_loss: 0.0070 - digit4_loss: 0.0041 - digit0_accuracy: 0.9986 - digit1_accuracy: 0.9975 - digit2_accuracy: 0.9971 - digit3_accuracy: 0.9976 - digit4_accuracy: 0.9986 - val_loss: 0.0401 - val_digit0_loss: 0.0028 - val_digit1_loss: 0.0085 - val_digit2_loss: 0.0114 - val_digit3_loss: 0.0101 - val_digit4_loss: 0.0065 - val_digit0_accuracy: 0.9989 - val_digit1_accuracy: 0.9974 - val_digit2_accuracy: 0.9966 - val_digit3_accuracy: 0.9974 - val_digit4_accuracy: 0.9982\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "PK5FjnNp63eL",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "digit_acc = [[] for _ in range(NUM_OF_LETTERS)]\n",
        "val_digit_acc = [[] for _ in range(NUM_OF_LETTERS)]\n",
        "loss = []\n",
        "val_loss = []\n",
        "\n",
        "\n",
        "def plot_diagram(digit_acc_now, val_digit_acc_now, loss_now, val_loss_now):\n",
        "    global digit_acc, val_digit_acc, loss, val_loss\n",
        "    \n",
        "    for i in range(NUM_OF_LETTERS):\n",
        "        digit_acc[i].extend(digit_acc_now[i])\n",
        "        val_digit_acc[i].extend(val_digit_acc_now[i])\n",
        "    loss.extend(loss_now)\n",
        "    val_loss.extend(val_loss_now)\n",
        "    \n",
        "    for i in range(NUM_OF_LETTERS):\n",
        "        s = {0:'First', 1:'Second', 2:'Third', 3:'Fourth', 4:'Fifth'}[i]\n",
        "        # plt.plot(val_digit_acc[i], label='%s Digit Train' % s)\n",
        "        plt.plot(digit_acc[i], label='%s Digit Test' % s)\n",
        "\n",
        "    plt.title('Model accuracy')\n",
        "    plt.ylabel('Accuracy')\n",
        "    plt.xlabel('Epoch')\n",
        "    plt.legend()\n",
        "    plt.show()\n",
        "\n",
        "    for i in range(NUM_OF_LETTERS):\n",
        "        s = {0:'First', 1:'Second', 2:'Third', 3:'Fourth', 4:'Fifth'}[i]\n",
        "        plt.plot(val_digit_acc[i], label='%s Digit Train' % s)\n",
        "        # plt.plot(digit_acc[i], label='%s Digit Test' % s)\n",
        "\n",
        "    plt.title('Model accuracy')\n",
        "    plt.ylabel('Accuracy')\n",
        "    plt.xlabel('Epoch')\n",
        "    plt.legend()\n",
        "    plt.show()\n",
        "\n",
        "\n",
        "\n",
        "    # Plot training & validation loss values\n",
        "    plt.plot(val_loss, label='Train')\n",
        "    plt.plot(loss, label='Test')\n",
        "    plt.title('Model loss')\n",
        "    plt.ylabel('Loss')\n",
        "    plt.xlabel('Epoch')\n",
        "    plt.legend()\n",
        "    plt.show()"
      ],
      "execution_count": 17,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "gmz3c1Ve8xG1",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 851
        },
        "outputId": "b480f317-e5d7-4c9e-b2b6-bf9a30bbe85b"
      },
      "source": [
        "plot_diagram(\n",
        "    \n",
        "    [history.history['digit%d_accuracy' % i] for i in range(NUM_OF_LETTERS)],\n",
        "    [history.history['val_digit%d_accuracy' % i] for i in range(NUM_OF_LETTERS)],\n",
        "    history.history['loss'],\n",
        "    history.history['val_loss'],\n",
        ")"
      ],
      "execution_count": 18,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYgAAAEWCAYAAAB8LwAVAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nOzdeZwcZZ348c+3qvqcnnsmmUkmFznInQAhyCWXF4IiIKAruuCBqKisgsJvESQr6nq74ktXBYEVRA4VURDlyLIIiYQQIBNISCDH3PfR01cdz++P6kx6JpOkQ6YzyeR5v15NV9dTVf10T6hvP7copdA0TdO04YyxzoCmaZp2aNIBQtM0TRuRDhCapmnaiHSA0DRN00akA4SmaZo2Ih0gNE3TtBHpAKEd8URkuogoEbHyOPYyEXnmYORL08aaDhDaYUVEtopIRkSqhu1/MXuTnz42OdO08UcHCO1w9Cbw4Z0vRGQREB277Bwa8ikBadr+0AFCOxz9D/CxnNf/CtyVe4CIlIrIXSLSLiLbROQGETGyaaaIfE9EOkTkDeCcEc69TUSaRaRRRL4hImY+GROR+0WkRUR6ReRpEVmQkxYRke9n89MrIs+ISCSbdoqIPCsiPSKyQ0Quy+5fKSKfzLnGkCqubKnpcyLyOvB6dt+Ps9foE5EXROTUnONNEfl/IrJFRPqz6VNE5Kci8v1hn+VPIvJv+XxubXzSAUI7HK0CSkRkXvbG/SHgN8OO+QlQChwFnIYfUC7Ppn0KOBc4BlgGfHDYuXcADjAre8y7gE+Sn0eB2cAEYC1wd07a94DjgJOACuArgCci07Ln/QSoBpYC6/J8P4APACcA87Ovn89eowK4B7hfRMLZtC/hl77eC5QAHwcSwJ3Ah3OCaBXwjuz52pFKKaUf+nHYPICt+DeuG4BvAe8B/g5YgAKmAyaQAebnnPdpYGV2+0ngypy0d2XPtYCJQBqI5KR/GHgqu30Z8EyeeS3LXrcU/8dYElgywnHXA3/YwzVWAp/MeT3k/bPXP3Mf+eje+b7ARuC8PRz3KvDO7PZVwCNj/ffWj7F96DpL7XD1P8DTwAyGVS8BVUAA2JazbxswObs9CdgxLG2nadlzm0Vk5z5j2PEjypZmbgEuwi8JeDn5CQFhYMsIp07Zw/58DcmbiFwDfAL/cyr8ksLORv29vdedwKX4AfdS4McHkCdtHNBVTNphSSm1Db+x+r3A74cldwA2/s1+p6lAY3a7Gf9GmZu20w78EkSVUqos+yhRSi1g3/4FOA+/hFOKX5oBkGyeUsDMEc7bsYf9AAMMbYCvGeGYwSmZs+0NXwEuBsqVUmVAbzYP+3qv3wDnicgSYB7wxz0cpx0hdIDQDmefwK9eGcjdqZRygfuAW0SkOFvH/yV2tVPcB3xBROpEpBy4LufcZuBvwPdFpEREDBGZKSKn5ZGfYvzg0ol/U/9mznU94HbgByIyKdtYfKKIhPDbKd4hIheLiCUilSKyNHvqOuACEYmKyKzsZ95XHhygHbBE5Eb8EsROvwL+Q0Rmi2+xiFRm89iA337xP8CDSqlkHp9ZG8d0gNAOW0qpLUqpNXtI/jz+r+83gGfwG1tvz6b9EngMeAm/IXl4CeRjQBDYgF9//wBQm0eW7sKvrmrMnrtqWPo1wCv4N+Eu4D8BQym1Hb8k9OXs/nXAkuw5P8RvT2nFrwK6m717DPgrsCmblxRDq6B+gB8g/wb0AbcBkZz0O4FF+EFCO8KJUnrBIE3TfCLydvyS1jSlbw5HPF2C0DQNABEJAF8EfqWDgwY6QGiaBojIPKAHvyrtR2OcHe0QoauYNE3TtBHpEoSmaZo2onEzUK6qqkpNnz59rLOhaZp2WHnhhRc6lFLVI6WNmwAxffp01qzZU49HTdM0bSQism1PabqKSdM0TRuRDhCapmnaiHSA0DRN00ZUsAAhIreLSJuIrN9DuojIf4nIZhF5WUSOzUn7VxF5Pfv410LlUdM0TduzQpYg7sCfq39PzsZfWGU2cAXwMwARqQBuwl8AZTlwU3ZCNU3TNO0gKliAUEo9jT/x2J6cB9ylfKuAMhGpBd4N/F0p1aWU6safm35vgUbTNE0rgLFsg5jM0FkmG7L79rR/NyJyhYisEZE17e3tBcuopmnakeiwHgehlPoF8AuAZcuW6TlDNO0gcVwPEcE0ZK/HKaVwPYXtKkQgaBoY+zjnUKOUorc3Q3NrnI72JMqAUNQiGLEIFwUIhkwsQ/AU2K5HOuOSStqkEi7ppE067ZJxPdKOR8bzSLuKjOthex5FRQFiRUFisQDF4QDFYYuikIXjKlK2S8r2SDv+c8p2EQFLBAuwEAzAVFASCXDcvBHHuh2QsQwQjQxd1asuu68ROH3Y/pUHLVeadhDZrkdTT5IdXUn6UzbRkEUsZFIUsigKWsRCFrGwRcDcd2E/7bi09qZp6UvheopQwCBkGYQDJiHLIGT5NzLYtQTdzrnYkrZLc2+Kpp4kzb0pmnuSNPWmaOtLkci4pBwXO+VhpV2CaUXUBU8gLQrXBMc08ExwLUEpEMfDcEBcRUBBUIGB4KEQUzAMwbQMTNMgYAoBD0IeBHMeAQ9MBFP5VR0mYCj/oQzBMcExxX9/w38GMF0wPTA9hemB5SrEy35Wlfuc/SZEwAARAUMQQxAByXgEMoqwo7DYc1CzUaTF/05DCoJ7OTZXAH+hjy6gE4UNZAQyopDsZzWzQcDI+Q7MEa7fExGO++EZeb3v/hjLAPEn4CoRuRe/QbpXKdUsIo8B38xpmH4X/qLumlYwSinSjv8rDwUKRfY+glIKESFoGQSzN7Sc9aqxXY+2/jQt2RtqS1+Ktv40tuMvSS3C4PECdA1kaOwYoK8thduXocwVyl2hSA27eec8iymYpoEZEAKWQSBoYlkGCdej13boSTv02g6O+DcsT/wFsVXOw8u+v7HzoXZu+zfhIBBUQlBBzDSotCymGgYRWxHMCKa988zCc03wLMETwRNQ4gckT8Az/AAQTCssV2G6Cssbdr6RfZiCa4BnCEjOzT/7DAKeQimF8hS4gO3531fAQFWa2EUWVnGAMitD+UArEgiRjpaRDhTh2IKTcnBTLiKCFTKxQiaBiEnQ8ogm2gm5cYxoCVasDCtajBUKETANRMFAwibelyLRmyQVT5EeyGAnbQzXwVIOpudguhlMO4VpJzENhYSCEA5BOISKhvHCRURrSgvydyhYgBCR3+KXBKpEpAG/Z1IAQCn1c+AR/FW0NgMJ4PJsWpeI/Af+qlsAK5RSe2vs1o4QnqdI2i6JjEsy4zKQcUhkXNK2O1jdYRpgGgam+DeA3qRNRzxNe3/uI8XAgE0m5eKkXNy0i8p4WJ7/S9cCLOUX403lbxuAKwoHcAQwBEz/l6eX8Qh7EFZCRAlhBRElmOy6OZPzXI0w3d0ZYIL+f0sChEuCeJ5fJeN6anDb8xSuq/BcD5VSKM9FPBfxFDEl2UWqjcFrHSgxhKBpErBMQhGLWHmYWEWY4ooQsfIwxRVhispCKE+RTjpkso+d22IIgZBJMGwRCJvZbRPDMPA8DzfjYnd1Y7d3kunoxovHCQY8gkEhGDIIBME0DTBMJBTECIWQUBgjFETCYSQYQoJBRFwk2YGk2lEDbTh97RgGWLFSjKIyiJRDJPscjIFhgphgWIPbCgHPQ7kuynHBc/3tRILU+pdIPf80yRdeJrWlETee2e27skoCBCpiBCaU4+CQ6OzG7k0icQcrtefv2LNAWULEVhS5+/6beKLIBP3SUcAZoYRSE4APvLwff+X8FCxAKKU+vI90BXxuD2m3s2t5SG2cSDsuPQmbnoRNdyJDxvHwlP9L3VMKL/vcn3Jo6U3S0peipTdFc2+K1r4UnQMZds5OH/agwjOocIVyTwgp8Yv32V/AIfxnM1tMN4GIghkIM4fkStjr/wYGfiAQAVehXIXs1tplggFWxCIUtYjGghQVBzFNYWcthsp+TpQiVBSgbGKUsglRyiZGKZ0QIRA03/L36roebsbDzrg4GRcn4+G5fmDxfxmDym4DGKaBka3mMUwBO43q6iQYDRCMBjDDQSQQQCwLMU3EssCyEMPw62U8FzwbUv1g9ILZ4z+sHgj0gpNGeQ5OQx+pHR2kG7tIN3Zjt/dj9wzg9CbBG/olZrKPA+UZCs8C11K4FrgWOJYCAcsByxFMBwxHMBww3BFutjmUKOLliqapwut1QV6qFTzHo6pXUd2jqO51qOrtZsJr3X7pMAY9FUL3VCEeM5GyKBINY6eSuMk0XsYhmFGEbQg4kA4KyaDgBsAMGgQtCAWEdMQiEQ2QiFjEoxaJsIltGMTdDAPpBO5AilgK/5FU1EVL+cYofH/DHdaN1NrYc1yP5t4UDd1J2of9Ut/5y70nkaE7YZO0d/+pJNn66WD2hh7MucmXBS0qQwEWBy1KAsUUlYI54CL9NiqVU6dggBU2MYImRtDACBhI0EQCQihkEQ1bFEUsQkET0/JvjoGQ/+s4GLEGn4MRi2DYxAqYmAEDMzByg6rneji2h+t4eI4iGLGwgsaQaqe3SimF29lJZvt2Mtu2k9m2FbezE6u6Gqu2lkBNLYFJtQQmTMAIgkrFUZ1tqPZWaG+DznakqwsLBysKVsjFCtuYVhLD6cfL2KS7DJKtNqnmBKkdvaSbeiCPdWEUCi9bdeMZ4FigAgoshWkpgqZH2PRwkiZer4Vh7/o+OouhsVLoroOeuYreYqG3WNEXg/4o9BoGKMFQ/r8JyQb3gANBB4KO2rVtk21byD5ytoOOImhDyIGQDWFbCDlgeJAMKDIWpIKQsSAdANvy2y+87Gdys1VYtgkN1cLWCaDCQSYX11EXq2NebDLRQNT/t4uQAhpFaPQ8ikPFTC6ewrJYHZNjkykNle72b8L1XHozvfSkeojbccpD5VRGKgevmS/Xc+nP9NOT7qE304tRoKo/HSC0fepL2WzvTLCjK8H2YY/G7iTOsF+DAVOoKgpRWxRkeiBARWmYkmIh6iiCaYWRcPEGHNyUg+fs5caUAL/mPINIhlBRgPKaKOVzKiirKfK3a4oorgwf1J4xhum3ReRFKVSiC9WxHbflTdzWBtz2ZpyODtyuLpyePpzeOG5/CrvPxu718DI534mAGTFwk96uOqqd+bA8PDdbQZ8HCYJy8L9SIBVWbK+B+hOFhgpjsAHYGmzkhagnhJRJ2DMIKYOg8p8DSvzSSioDaZtwBsIZRTgJvVFhxzTYXi3E68oJzJxB7cRZVIQrcDwby7Mp8xyKPJsJroMhQkV0ApXRKirDlVRGKqkMV1IeLsdVLrZrk/EyQ55NwyRoBAma/iNgBAafTcPEFBNDhv6NlFJkvAwpJ0XaTZN20iTdJLZr+6/dNBk3Q8bL4HgOE6ITqIvVUR2t3u1ab5VpmFSEK6gIVxzwdcrCZZSFy0YlX3uiA4SGUorOgQxvdgzwZscAWzsG2Na1KyD0JOwhx5dFA0wvj3BsRYz3TyinUhlEbTDTHm7CId1vk2hO49g2sOtcwxQi5SGKK8MUzyglXBwkOFhHbREImQTCZnafld32n63A6PxC388vBuJt0P0mqqcRuzuO3daH3dGL3dqF3d6N3dGNstPg2OBm/Idjo1wbO2XjpDzICLKHqgxPIBmFRJFBvNigc1qAznKhrRxay4W2UsiYYLpCxYBQ0Q8VfUJ5n6I4rkiGhJ4iobsIuqOKzohHZ8TFdDxKB6AsrihNQFl2OxWEN2qELbWCWTORWWUzmRmr4/SiWipKplAaqaQ8VE5ZuIzSYCkBM7DPr8l2bVoGWmiIN9AUbyIWiPKekmlML5m+37+MC0lECJkhQmZorLNy2Bg3S44uW7ZM6fUg8tMZT/PM5g6e3tTBptZ+tnYM0J92BtNLEGZHw0yOhpgYClBhWRSLQVgJRsploDNFf1fa7/WRZQUMispDFJWGKCrLPkqDFJWFKK7wGzajJUHkEOsDrzyPzJZNpFY9Tmrd89gNDXj9/XiJBF4yg2srPMfAs4f9UheFFXEJRD3EUKRFiJsGcdMibggDAgMhiIchHoF4WOiPgIpF8EqKSJSESBaHyEQtTDOAKab/q3jnr2EjSMAMEDACWIaFpzxc5eJ67uC2UoqA6f9yDpkhgkb22QxiGRaC39vKEGPwF3B5qJyZZTOZWTaT4mDxGH3r2qFERF5QSi0bKU2XII4AjuuxbkcP/7upnac3tfNyYy9KQXk0wMLJpZy/eBJTHINYr4PdmKC/NekvX08aSOMB8YCBGwsQLQkycXoJs4+PUFrtN7CWVkf8m//B/oW/UyaB19NKZstrZLa8TmbrVjLbG8m0tCOGgRENYxaFMYqimEURjKIoTksTqde3kmrow8sWcjxT0V8CTtjAjlg4FcU4kRBONEImFiFeHqS3IkhfWYC+EsE2FQNOitcGGum2+wEImSHmV85nYdVCphZPZUq0muqI/6iKVOX1i1zTDhU6QBzmehM2a7d3s2ZbFxtb+hlIuyRtl5TtP5N0icVdxFYEBKaXRXlHXQ3Ty6KUhyzat/XTsraDjKvotQxqZ5Wy8KRJVE2JEYkFCccChGOBA+pl81YopfAGBnBaW3Gam3B2bMZp3Irb3ojb1Ybb043X1487kMRJeDgJE3IGEJkRF1Xs+mMBWgywxa/qsf1f0p6p6KoWXl9osa7W440aobFKKI/5o1H9cRAKhQIcoB9LLEzDxDIsLM/CVCYhM8Tp097BwqqFLK5ezMyymQQMHQS08UEHiMNM10CGlRvbWLOtmxe2drOprR+lwDSE2RNilAUtpqSFyrhBSa8iNKAYMrgp4UBTL9vppcEUKiYVseSsKUyZW0HtrFKsgxwIdso0NDLwxKMM/P1PpLc2YvemULa3+4GiMIMeZtiEohDJijL6poboqArRUGmyucKlvjhBk/SN+D7iKSIZvwdLcbScxdWLWVy1mA9XL2ZR1SJiwViBP6mmHT50gDhMrG/s5Y5nt/Knl5rIOB7FIYtjppVzzuJaFlcWEWu3aazvonlDL67jYVhC7cwyps6vYPLR5USKA373TSs7xYFljGl7gBuPk3j+eQaeepyBp58i09INgBV1iEw0iE0twqoow6yuIj2hmv6JVbRWFvNq0OHVVCuvdW9ie/92/K5OCYJGkNpYLZOKpnJibBK1RbXUxmqJWBEMMbDEwhDD791iGNREa5hWMm3sqsU07TCgA8QhzHY9Hqtv4c5nt/L81m4iAZOLl9VxybKp1AUDbF3XwZY1bdS/6U9+W14TZeHpk5kyr4JJs8sOerXQICcNXW+iOl7Hfv0VMo1tpFt6ybT2kmnpIdPai9OT8I81FU6tTedpUd48dhabZ9TSqjJ0pbvoSnbRk96Eq16FfvwHMDk2mXkV83j/zPczr3Iec8rnMCE6YdS6Imqa5tMB4hDU3JvkgTUN3L16Oy19KaZWRPn3s+dyxoQyujb3suHOjTy9Iw5A9dRiTjjvKGYeU015TdHBzajrQMdGaHoRml/Ga9lE6rXNJLb1kOywSHYEcdM5QSrokSjzaK4RNi00eKHOZMMUwbH8OnthCxP74oP9zxdXLaYiXEFltuvlhOgE5lTMoSRYcnA/p6YdoXSAOERkHI8nXm3ld2t28PSmdjwFZ02t4NrZk4l22TQ+1MhjiW0gUDOjhJMumMXMY6spqYocvEwmu+H1v0PjWlTDWtIb60m1uaS6giS7w6S6zewgrGKCtZXIyTPYMr2IF0t6ecbaSpPVDyKUh8pYUrWYEyqO5sLiOibFJjG5aDI1RTW6l4+mHUJ0gBhjm9vi3PvP7fzhxUY6BzLUFIf5/NzJTGjI0PNynBaSFJUGmbG0mqnzK5gyr4Jw0cG9ibo9nWQe/iHpp+4h1W6T6g6T6gmgHP+XvBGNEF6wkMqlS3EXzuaZsjYe6nyKl9rXAXBU6VGcOOHdLKlewjETjtF1/5p2mNABYozE0w7fe2wjdz23FdMQ3jF3AmdXlpF8sYuO57pwK8OceP5Mpi2spGJS0UG7oSrXZeDZZ0k8v4b0po2k61/Cbu/JpoaQcAnh+QsoP2ch4YULCS9YSN/EIta0vcBf3vgL/2j8H5wWh1lls7j62Kt574z3UhurPSh51zRtdOkAMQYeq2/hpofqae1PcenyqVxYW8nGJxrY8Y/tlFRHOPNjc5lzQo0/5fFBYjc30/Pg7+l58EGc5mYwDUJlQqSon7KTqgid+VFCp56PPbGCV3s2sr5jPa90rGT92ltpGmgCYEJ0Ah+d/1HOOeoc5pTP0aUETTvM6QBxEDX3JrnpoXr+tqGVuTXFfOcdc2l5vIl/PPYapdURzvrXecxZPhGj0IEhHYfX/4ba8Ajxl9+kZ2038TeToKBoqsnEdxrEyhowKqbAmd+hfebpPLzjCR7bcCPr/ncdnvLHJ0yOTWZx9WL+Zd6/sKR6CYuqFmEaY9RzStO0UacDxEHgeYq7ntvKdx/biKsUX33X0SztF9bdsYlwcYCzLpvHnOMLGxhUsgf7mXtJPf0nUhs2kOowSHYH8TKCFTOpPKmSsmOrCVYXgxmko3oOf6+s4bEdj7J23S0oFLPKZvHJRZ9kSfUSFlYtPOAZKTVNO7TpAFFgKdvlmvtf4s8vN/P2OdVce/wM6h96k7UNcY4+oYZTLp5d0Ebngb89SOeP/5Pkjl68TDYAGVFCM6ZQfMoyis88k9hppyGWheu5PNnwv9z72r2sevM3qDf9oPCZpZ/h3dPezVFlRxUsn5qmHXp0gCig7oEMn7prDWu2dfPVdx/NcXGDZ362nnAswHs/s4gZS6oL9t6ZN7bQ9u+fo//FbVhRRcmyowgffzrhk88mNHcuRnDX8pTdqW5+/+rvuW/jfTQNNDExOpFPL/k075n+HmaWzdzLu2iaNp7pAFEg2zoHuOzXz9PYk+SH5ywg83Qbz++IM2f5RE69ZE7BSg1uXx8d31tB1wN/QQyP6rOmUnHz7RhVU3Y7dkPnBu559R4effNRMl6G5TXLufb4azl9yulYhv6noWlHOn0XKIAXtnXzqbvWoJTiZ++Yz5t/8Ae4nX3lIo5aWphSg3Icun97Dx0//gFuPEXpHEX1V28icPK/DD1OKZ5pfIY76u/gny3/JGJFOH/2+Xzo6A8xq3xWQfKmadrhSQeIUfboK81c/bt11JSGuXn+NDbc9wZlE6Oc89nFlFYXZtRzZts2mq6+iuSrm4lOSDPx8pMJf/y/IFI+eIzt2jy69VF+vf7XbO7ZzIToBL503Je4cM6FeuoKTdNGpAPEKHrklWY+d89ajqkr5dOxCuof3sr0RZW88+MLCEZG/6tWStFz7720fuubCBkmnWlQ8m//jcx+5+Axjudw96t3c9eGu2hLtDGrbBbfOPkbvHfGe/W0Fpqm7ZUOEKOkP2Vz05/qOXZiCR9JR9n8SgvHvGsqb/vATIwCTKttt7XRfP1XGfjHKopqUtR+6FgCH/slRHd1Pe1N9/Ll//0yq5tXc3zN8dx04k2cOvlUPYBN07S86AAxSn70+OvYvRneG4/Q2t/LWZfNY+7bCjPFRN9f/0rL127ASw4wcVk/5Z+9Djnxc5Bz49/Ss4XPP/l5WgZa+I+T/4MPzPpAQfKiadr4pQPEKNjY0s8dz27lM6ESvH6H8790LDVHlY76+yjPo/Ub36D7nt8SrrCZdE6Y0Ccfgrrjhhy3csdKrvu/6wibYW5/9+0snbB01POiadr4pwPEAVJKcdOf1jPDsIi0Zzjm/TMKExxcl+brrqH34b9ScXScCZe8HTn/pxApG5KX29bfxn+t/S/mVc7jx2f8mJqimlHPi6ZpRwYdIA7Qwy83s+qNLq4vqsAqclh85u7jDQ6UchyarvwIfc+8TNWiJFXX3ogc/8khVUoD9gA3P3czj775KGfPOJsVJ60gbIVHPS+aph05dIA4APG0wy1/2cAp5TGcN5Mcf/5MguHR/UpVbxuNHz+f/vouqk8poeo//wyVu0Y3J+wE97x2D3fW30lvupcvHvtFPrHwE7ohWtO0A6YDxAH4yROv09qX5qrSEuziNItOrxvV63sbHqHx6i8R3y5MuGg5lV+/DUz/T5awE9y78V7uWH8H3eluTpl8Cp9b+jkWVi0c1Txomnbk0gHiLdrc1s9tz7zJR4+aSN/aPk65aDaB0ChNda0U3sPX0fDD+xhoDjPxC5dT8dmvAJByUvxu4++4ff3tdKW6OHnSyXxm6WdYUr1kdN5b0zQtSweIt8BvmK4nGjBZ2A3J0iAL3j5ptC6O+8ev0PDDB0m0h6m56WuUf9ifLiPlpPj03z/N2ra1nFh7Ip9d+lndQ0nTtIIp6Mo0IvIeEdkoIptF5LoR0qeJyBMi8rKIrBSRupy074hIvYi8KiL/JYdQpfqj61v4x+ZOvrhoCh1v9nHc2dOxAqNTesj8/ia2fusPJDrDTPr2tweDg+u5XP9/17O2bS3fPvXb/OJdv9DBQdO0gipYgBARE/gpcDYwH/iwiMwfdtj3gLuUUouBFcC3sueeBJwMLAYWAscDpxUqr/vr92sbqSuLUPL6ALGKEPNPHp3SQ/Ker7N1xb04mQhTf3UbpeedB/gllu+u+S6Pb3+ca5ddyzlHnTMq76dpmrY3hSxBLAc2K6XeUEplgHuB84YdMx94Mrv9VE66AsJAEAgBAaC1gHndL409SY4PRWjb1s/x752BGTjwr7HvFzey7ZbfYkQiTL/vQYpOPHEw7c76O7n71bv56PyP8rEFHzvg99I0TctHIQPEZGBHzuuG7L5cLwEXZLfPB4pFpFIp9Rx+wGjOPh5TSr06/A1E5AoRWSMia9rb20f9A+xJY1eCac02JVVhjj7xwAaiKaXo/NY1NP7gPkITo0x/6K+E5swZTH/kjUf4/gvf593T3801y6450KxrmqblraBtEHm4BjhNRF7Er0JqBFwRmQXMA+rwg8qZInLq8JOVUr9QSi1TSi2rri7c6my5+lM2E/s9An0Ox587A/MA1pFWStH6lc/QdudfKJ4dZdofH8eauGv+pn82/5N//8e/c9zE47jllFswZKz/XJqmHUkK2YupEcgdVlyX3TdIKdVEtgQhIjHgQqVUj4h8ClillIpn0x4FTgT+r4D5zUtTT4r5tkMVRhsAACAASURBVIUZs5hz/MQDulb8L3+g++H/pXxpmIm3PY4U7ZqJdVP3Jr741BeZXjKdH5/xY0Jm6ECzrmmatl8K+ZP0eWC2iMwQkSDwIeBPuQeISJXI4M/i64Hbs9vb8UsWlogE8EsXu1UxjYXGngRlrlBSG8U4gNKDl8nQ+u1vEixxmPjje4cEh4Sd4AtPfoGoFeVn7/gZpaHRn9tJ0zRtXwoWIJRSDnAV8Bj+zf0+pVS9iKwQkfdnDzsd2Cgim4CJwC3Z/Q8AW4BX8NspXlJKPVyovO6Phq4EpZ5QOaHogK7T/cufYncMMPGDxyETjx6S9oMXfkBTvInvnvZdPdmepmljpqAD5ZRSjwCPDNt3Y872A/jBYPh5LvDpQubtrWpqT1KMMKH2rQcIp6ODjl/cTmxSmtin/nNI2urm1fxu4+/46PyPcuzEYw80u5qmaW+ZbvXcT52tAwCUVr319aXbv/NNvIzNhI+cBeXTBvcn7AQ3PXsT00qm8fljPn/AedU0TTsQeqqN/dTfkQSg5C0GiNSGDfQ8/CgVR6cIffCmIWk7q5buPPtOItZbD0CapmmjQZcg9lOmLwNASdX+r7WglKJ1xU2YQY+qj34ASnfN/rqzaunS+ZdyzIRjRi2/mqZpb5UOEPsh43iYCQ8VNN7Sug/9jz1GYt16qpckMd/51cH9umpJ07RDka5i2g8tvSlKPSFQFtjvc71UirZvf4tQmU3Zhz4CJbsGxOmqJU3TDkW6BLEfGnr8Lq6xiv2vXuq64w7sljYmLkshb/+3wf25vZZ01ZKmaYcSHSD2Q2N3khJPqJgY3a/z7LY2Ov77vymuS1H0/sshNsHf79qDVUtXHXNVIbKsaZr2lukqpv3Q1BTHQqidFNuv87ruvBOVSjFhmQ0nXz24/8kdT9IYb+TWM2/VVUuaph1ydAliP3Rkx0BUTMi/BOH29dHz299SMiVJ8J1XQFHlYNr9G+9ncmwyp0w+ZdTzqmmadqB0gNgP/Z0pYP/GQHTf+zu8RJLKeQk44TOD+7f2bmV1y2ounH0hpjFKa1lrmqaNIh0g9kOmz0YBxXk2UnvpNF133klRHYSPfzvEdk1J/sCmB7DE4vzZ5xcot5qmaQdGB4g8eZ5CBhxU2Mh7BbnePz6E29lJ5ewOWHzx4P60m+aPW/7ImVPPpCpSVagsa5qmHRAdIPLUOZCh2AGrJJjX8cp16bz9NsJ1JUQnW3D0ewfT/rb1b/Sme7n46Iv3cgVN07SxpQNEnhp7ktkxEPkt3NP/979jb9tO5ex2ZN45ENrV8+mBTQ8wrWQay2uWFyq7mqZpB0wHiDw1dCQoVpJXDyalFJ2//BXBSVUUV3fCol0lhde7X2dt21oumnMRIlLILGuaph0QHSDy1NjUhyDUTi7e57GJVatI1ddTsawYiVXCzDMG0+7fdD9BI8j7Z75/L1fQNE0bezpA5KmjJQFAzaR9LxTU+ctfYVZWUFr0Aiy4AEx/7qaEneDhLQ/zzunvpDxcXtD8apqmHSgdIPLUlx0DUVy59zEQyfp6Bp59lop3LcVQqSG9lx7b+hhxO87Fc3TjtKZphz491UaeMr0ZPIGisr03UnfddhtGURHltdsgOR3qjh9Mu2/jfcwqm6Un5dM07bCgSxB5kgEHL2JiGHtuWM5s307fXx+j7IJzMZv/AYsugmxD9IbODazvXM8H53xQN05rmnZY0AEiD/G0Q8QGq2Tv60D0/vEhACqOiYDyhvReun/T/YTNMO+b+b6C5lXTNG206ACRh8buJGWeULSPKTYGVq8mvHAhgR2PQO1SqJ4DQDwT5y9v/IWzZ5xNSbDkYGRZ0zTtgOkAkYeGtjhRJVRM2HMDtZdIkHzpJYoWz4bmdUMap59pfIakk9TzLmmadljRASIPOxr7AZg0ac9jIBIvrAXHIVreA2LAwgsH01Y1r6I4UMyiqkUFz6umadpo0QEiD+3N/hiIyXV7CRCrV0HAIpp8Gma8HYprAH9U9XNNz7G8djmWoTuNaZp2+NABIg/9nUkAyqr3XMU0sGo10bkzMeLbhjRO7+jfQdNAE2+rfVvB86lpmjaadIDIQ6Y3g2tAODZyLya3t5fUhg1EpwTAsGDerp5Kq5pXAegAoWnaYUcHiHwMOLgRc4/jFxJr1oDnUVTaAbVLILyrp9Kq5lXUFNUwrWTawcqtpmnaqNABYh9s1yOcUZh7GQMxsGo1Eg4TUfUw9cTB/a7nsrp5NSfWnqgHx2madtjRAWIfmnuSlLpCrHzPYyASq54jumAmQhqmnTS4/7Wu1+jL9OnqJU3TDks6QOzDtqY4QYTyPYyBcDo6SL++mei07DoRU3YFg+eanwNgea1eGEjTtMNPQQOEiLxHRDaKyGYRuW6E9Gki8oSIvCwiK0WkLidtqoj8TUReFZENIjK9kHndk4bsGIjaybER0wdWrwagqLwTqo6GosrBtFVNq5hTPkevO61p2mGpYAFCREzgp8DZwHzgwyIyf9hh3wPuUkotBlYA38pJuwv4rlJqHrAcaCtUXvemo2UAgKlTR54iI7FqNUZxMeHMKzB1V+kh6SRZ27aWE2tPHPE8TdO0Q10hSxDLgc1KqTeUUhngXuC8YcfMB57Mbj+1Mz0bSCyl1N8BlFJxpVSigHndo94OfwxE1cSRFwoaWL2a6OKjEbt3SPvDi20vYns2b5uk2x80TTs8FTJATAZ25LxuyO7L9RJwQXb7fKBYRCqBOUCPiPxeRF4Uke9mSyRDiMgVIrJGRNa0t7cX4CNAujdDxoRgePdR0HZjI/b27RRNz1Y/5fRgWtW0ioAR4NgJxxYkX5qmaYW2zwAhIu8TkUIFkmuA00TkReA0oBFw8RcyOjWbfjxwFHDZ8JOVUr9QSi1TSi2rrq4uSAYl7o+BGMnA6n8CEC3vguJJUDZ1MG1V8yqWTlhKNBAtSL40TdMKLZ8b/yXA6yLyHRGZux/XbgSm5Lyuy+4bpJRqUkpdoJQ6Bvj37L4e/NLGumz1lAP8ETjoP8WVUgTT3h7HQCRWr8KsqCCUehmmnTi4OFB3qptXu17V3Vs1TTus7TNAKKUuBY4BtgB3iMhz2aqdPc9c53semC0iM0QkCHwI+FPuASJSlVM6uR64PefcMhHZWSw4E9iQ1ycaRW19aYo9oah892VGlVL+/EvHLETiTUOql1a3+D2bdIDQNO1wllfVkVKqD3gAv6G5Fr+9YK2IfH4v5zjAVcBjwKvAfUqpehFZISLvzx52OrBRRDYBE4Fbsue6+NVLT4jIK4AAv9z/j3dgtjX0YSGUT9i9miizdStOaytFM7Jxclj7Q3GgmPmVwzttaZqmHT72Of909mZ+OTALv+vpcqVUm4hE8X/V/2RP5yqlHgEeGbbvxpztB/ADz0jn/h1YnMdnKJgdDf4YiJpJu4+BSOwc/1DRA3YpTJgH+CWLVc2r9PTemqYd9vK5g10I/FAp9XTuTqVUQkQ+UZhsHRram/0xENOnlu6WNrBqNVZNDYGBl2DqCWD4DdkN/Q00xhu5bMFlBzOrmqZpoy6fKqavA//c+UJEIjtHNSulnihIrg4RfR1JFIra2qFjIJTnkVi9mqLjliKdm4ZUL+2cXkO3P2iadrjLJ0DcD3g5r93svnEv3ZshZQlWcGg31/Trr+N2dxOdmS1Z5LY/6Om9NU0bJ/IJEFZ2JDQA2e1g4bJ06FBxByey+1eUWOUvAlRU3gtmCCb7PXD19N6apo0n+QSI9pxeR4jIeUBH4bJ06AhmPIyi3cdAJF5cR2DyZALxl/3gYPndYPX03pqmjSf5BIgrgf8nIttFZAfwVeDThc3WoSHogRnefRS13dREcGodNK8bUr30zxa/qUZP761p2niwz15MSqktwNtEJJZ9HS94rg4BGdslqEBFdv+KnOZmQsfMAc8ZMkHf+o711MXq9PTemqaNC3l11BeRc4AFQHhn3bpSakUB8zXmevrSCEJgWIBQmQxORweBwCRAoO74wbT6znoWVi08yDnVNE0rjHwm6/s5/nxMn8cf0XwRMO676HT3pAGIRIcGCLutDZQioFpg4kKIlAHQk+qhMd7IgsoFBz2vmqZphZBPG8RJSqmPAd1KqZuBE/Gn4x7Xevr8ABGNDe2w5TQ3A2Bl3hiyQFB9Zz2ADhCapo0b+QSIVPY5ISKTABt/PqZxra/P79kbiw3txWS3tAAQCA74M7hm7QwQev4lTdPGi3zaIB4WkTLgu8BaQDEGE+cdbPG4X4IoLh46k6vd5JcgAlF3SA+m+o56ppdMJxYcee1qTdO0w81eA0R2Ku4nsms0PCgifwbCSqneg5K7MTQQtwEoLRlaxWS3NGNGLIzKKVAyaXD/+s71HF9zPJqmaePFXquYlFIe8NOc1+kjITgAJAf8AFFWOrQE4TS3YMUEKmYM7mtPtNOWaNPtD5qmjSv5tEE8ISIXyhE2d0Qq6QBQURYest9ubiYQsaG0bnDfhk5/LSMdIDRNG0/yCRCfxp+cLy0ifSLSLyJ9Bc7XmMskHVwUkWHjIOzmZgKhASiZPLivvrMeQwzmVuzPiqyapmmHtnxGUu9radFxyUm5YDBk0j0vkcDr68Oa7g4pQazvWM9RpUcRDey+8pymadrhKp8V5d4+0v7hCwiNN27aBXNordpgF9eoC6V+CUIpRX1nPadOPvWg51HTNK2Q8unmem3OdhhYDrwAnFmQHB0ivIyHDA8QuV1cS/wSRGuila5UFwuqdPuDpmnjSz5VTO/LfS0iU4AfFSxHhwrbQwWGBginJTuKOqcEUd/hD5BbWKnnYNI0bXzJp5F6uAZg3mhn5FBjOAoJDP167OYWEAhUxCDkN82s71yPJRZzKsb97COaph1h8mmD+An+6GnwA8pS/BHV45rpKggNXQvCbm7GKrKQsl0N1PUd9cwun03IDA2/hKZp2mEtnzaINTnbDvBbpdQ/CpSfQ0bAAzUsQDgtzVgxdmugfue0d45BDjVN0wornwDxAJBSSrkAImKKSFQplShs1saOnV0syIsML0G0EAqnB8dANPQ30Jfp0w3UmqaNS3mNpAYiOa8jwOOFyc6hoaffXywoFNk1k6tSCru5iUAoOTgGYucMrrqBWtO08SifABHOXWY0uz2uR4R1ZRcLCucsFuT29KBS6ewYiF0BImgEmVU2a0zyqWmaVkj5BIgBETl25wsROQ5IFi5LY6+n1w8QRTlrQTjZQXJW1B2sYqrvrOfoiqMJmIHdL6JpmnaYy6cN4mrgfhFpwl9ytAZ/CdJxq7/fDxCxnNXk7OacQXKlk/GUx4bODZx71LljkkdN07RCy2eg3PMiMhc4Ortro1LKLmy2xlZ/v//xiot3DxA7SxBb+7YyYA+wsEq3P2iaNj7ts4pJRD4HFCml1iul1gMxEfls4bM2dhID/nKjJSW7xjY4LS1gClZlFVihwRHUeopvTdPGq3zaID6VXVEOAKVUN/CpwmVp7CUH/LUgchcLsptbCBRbSNmu9oeIFWFG6YwRr6Fpmna4yydAmLmLBYmICQT3cvwgEXmPiGwUkc0ict0I6dNE5AkReVlEVopI3bD0EhFpEJFb83m/0ZJK+FVMFWW5AaKZQNTb1UDdUc/cirlYRj7NOJqmaYeffALEX4HfichZInIW8Fvg0X2dlA0kPwXOBuYDHxaR+cMO+x5wl1JqMbAC+Naw9P8ADvq04jsXC4rmjINwmpuxwkkonYLjObzW9ZquXtI0bVzLJ0B8FXgSuDL7eIWhA+f2ZDmwWSn1hlIqA9wLnDfsmPnZawM8lZue7U47EfhbHu81quyUi52zWJByXey2VgKhNJRO5o3eN0i5KT2CWtO0cW2fAUIp5QGrga34N/0zgVfzuPZkYEfO64bsvlwvARdkt88HikWkUkQM4PvANXt7AxG5QkTWiMia9vb2PLKUHzft4uSsBeF0dIDjZteBmKwbqDVNOyLsMUCIyBwRuUlEXgN+AmwHUEqdoZQarTaBa4DTRORF4DSgEXCBzwKPKKUa9nayUuoXSqllSqll1dXVo5QlUBkPNzdA5HZxLa2jvrOeokAR00qmjdp7apqmHWr21sL6GvB/wLlKqc0AIvJv+3HtRmBKzuu67L5BSqkmsiUIEYkBFyqlekTkRODUbHfaGBAUkbhSareG7oKwPchZLGhwqdEiP0Bsqd/CrLJZGPJWltPQNE07POztDncB0Aw8JSK/zDZQy16OH+55YLaIzBCRIPAh4E+5B4hIVbY6CeB64HYApdRHlFJTlVLT8UsZdx204ACIoyBnsaDBpUZjArGJbO/frksPmqaNe3sMEEqpPyqlPgTMxW9AvhqYICI/E5F37evCSikHuAp4DL/N4j6lVL2IrBCR92cPOx3YKCKb8BukbzmgTzNKLFdh5KwFYbc0I0EDo2IiSS9DW6KNqcVTxzCHmqZphZfPVBsDwD3APSJSDlyE37Npn72LlFKPAI8M23djzvYD+OtN7O0adwB37Ou9RpPlgZcTIJzmFgLFJlJay45+v919aokOEJqmjW/7VYmulOrONgyfVagMjTXH8QgqCEZySxAtBKIOlNaxo08HCE3Tjgx6GPAw3b3+YkHByK6vxm5qIlSWhNLJbOvfBqCrmLRxxbZtGhoaSKVSY50VrUDC4TB1dXUEAvkvT6ADxDBdvf7/IJGo/yV6mQxuZyeB2gyU1LG9bzsV4QqKg8VjmU1NG1UNDQ0UFxczffr0wQGi2vihlKKzs5OGhgZmzMh//jjdT3OY3uxiQdEiP0A4ra3ArnUgtvdvZ0rxlD2er2mHo1QqRWVlpQ4O45SIUFlZud8lRB0ghunrH7qa3GAX1+wgue1923X1kjYu6eAwvr2Vv68OEMPEBxcL8mdydVp2jaJORqtoTbTqBmpN044IOkAMMzC4WJA/o7ndnB1FXRKgwR0AdAO1phWCaZosXbp08LF161ZOOumk/brGj370IxKJxIhpp59+OkcffTSLFy9m7ty5XHXVVfT0DC51k9d7ffKTn2TDhg0AfPOb3xzxmBNOOIGlS5cydepUqqurh3yefKxbt45HHnlk3wceBDpADLNzsaDy7GJBdnMzZtTCqKxje3YMhB5FrWmjLxKJsG7dusHH9OnTefbZZ3c7znGcPV5jbwEC4O677+bll1/m5ZdfJhQKcd55uyaYHum9hvvVr37F/Pn+qgV7ChCrV69m3bp1rFixgksuuWTI58nHoRQgdC+mYdKDiwWFAX8UtRUTKPEbqAGmlOhGam38uvnhejY09Y3qNedPKuGm9+3/7MexWIx4PM7KlSv52te+Rnl5Oa+99hovvvgiF198MQ0NDbiuy9e+9jVaW1tpamrijDPOoKqqiqeeemqP1w0Gg3znO99h1qxZvPTSSyxZsmTwvTzP46qrruLJJ59kypQpBAIBPv7xj/PBD36Q008/ne9973s88MADJJNJli5dyoIFC7j77rv3+jm2bNnC5z73Odrb24lGo/zyl79k7ty53H///dx8882YpklpaSmPP/44N954I8lkkmeeeYbrr7+eSy65ZL+/t9GiA8Qw6aSDiSKaHQfhNLcQiNhQWse2vm2Uh8opCZaMcS41bfzZecMFmDFjBn/4wx+GpK9du5b169czY8YMHnzwQSZNmsRf/vIXAHp7eyktLeUHP/gBTz31FFVVVft8P9M0WbJkCa+99hpLliwZ3P/73/+erVu3smHDBtra2pg3bx4f//jHh5z77W9/m1tvvZV169bl9dmuuOIKfv7znzN79mxWr17NZz/7WZ588klWrFjBY489xuTJk+np6SEYDLJixQrWrFnDrbce1IU0R6QDxDBO2sUzwDD82je7pZnoxAEomcyO/jd06UEb997KL/3RsLOKaU+WL18+2Id/0aJFfPnLX+arX/0q5557Lqeeeupbek+l1G77nnnmGS666CIMw6CmpoYzzjjjLV17p3g8zrPPPstFF100uC+d9ntLnnzyyVx22WVcfPHFXHDBBXu6xJjRbRDD5C4W5MYH8Pr6B9eB2N6/nWnFuv1B08ZCUVHR4PacOXNYu3YtixYt4oYbbmDFihX7fT3XdXnllVeYN2/eaGZzN57nUVZWNqR95dVX/TXXfv7zn/ONb3yDHTt2cNxxx9HZ2VnQvOwvHSCG8TIeXjZA7OziGoi6pGLVtAy06BKEph0CmpqaiEajXHrppVx77bWsXbsWgOLiYvr7+/d5vm3bXH/99UyZMoXFixcPSTv55JN58MEH8TyP1tZWVq5cOeI1AoEAtm3v871KSkqYMWMG999/P+CXWl566SXAb5s44YQTWLFiBdXV1ezYsSPvz3Aw6AAxXMZDZRcLGuziGnVpsPyBc7oEoWlj75VXXmH58uUsXbqUm2++mRtuuAHw6/rf85737LFa6CMf+QiLFy9m4cKFDAwM8NBDD+12zIUXXkhdXR3z58/n0ksv5dhjj6W0tHS346644goWL17MRz7ykX3m9+677+a2225jyZIlLFiwYPB9r732WhYtWsTChQs56aSTWLJkCWeccQYbNmxg6dKl/O53v9ufr2XUyUh1cIejZcuWqTVr1hzwdb71hSfxYhb//s23033ffbTceBOz3tfK01fexdXPXM9vz/ktC6sWjkKONe3Q8eqrrxa8quVwEo/HicVidHZ2snz5cv7xj39QU1Mz1tk6YCP9nUXkBaXUspGO143Uw5iuQoL+VN9OSwsIWOUxdiQ7APQ8TJp2BDj33HPp6ekhk8nwta99bVwEh7dCB4hhLA/csB8g7OYWrFgAKatjW/82ykJllIZ2L2pqmja+7Knd4UijA0QON7tY0K4A0UygSEHpZHb07dBzMGmadkTRjdQ5uvv9xYJCg4PkmrHCaSjxFwrSczBpmnYk0QEiR3ePP3glXBRAKeUvNRpKkiquoWWgRZcgNE07ougAkaMnu5pctCiA29ODSqf9Lq5hf4COLkFomnYk0QEiR3+/P9V3LBbEbmoC/HUgtmcHzulZXDWtcG655RYWLFjA4sWLWbp0KatXrz7oeVi5ciXnnnvuiPtLS0s55phjOProo3n729/On//858H0n//859x11117vfaaNWv4whe+MHi9kWaP/fWvfz04PXgwGGTRokUsXbqU6667Lu/PsK8ZbfeHbqTOMRggigN+F1f8QXLb8auedBdXTSuM5557jj//+c+sXbuWUChER0cHmUxmrLM1xKmnnjoYFNatW8cHPvABIpEIZ511FldeeeU+z1+2bBnLlvnDDVauXEksFtttDYrLL7+cyy+/HIDp06fnPfFgrh/96EdceumlRKPR/TpvJDpA5EjE/WHzpSUh7M05ASLTp7u4akeOR6+DlldG95o1i+Dsb+8xubm5maqqKkIhfx2W3JviCy+8wJe+9CXi8ThVVVXccccd1NbWsnnzZq688kra29sxTZP777+fo446iq985Ss8+uijiAg33HADl1xyCStXruTrX/86VVVVrF+/nuOOO47f/OY3iAh//etfufrqq4lGo5xyyil5fZylS5dy4403cuutt3LWWWfx9a9/nVgsxjXXXMPzzz/PJz7xCQzj/7d379FRlefix79PJgkkJCTcbwFCBeFwCQmw8FBPSmgWyPkVsVEkVTgFtAhHkNZVU01VTsFfBYFlUahQEAoCKoJYsKBGKwgIP5SECSKgAU8UkgABDJIrmcn7+2PvxAADuZBhWJnnsxaLmX19XhznmXe/ez9vAMOHD+e9997j0KFD7NixgwULFrB48WKWLl2Kw+Fg7dq1LFq0qMZig/Pnz+ett96irKyMpKQkZs2aRVFR0Q2VPK8NvcRUTbE9F0SLiCa4TuUhDsHRqjXfFeXo+INSXjRixAhOnDjB7bffzqOPPsonn3wCWDWTHnvsMTZu3Eh6ejoPPfQQTz/9NGCVzZg2bRqZmZns2bOHDh06sGnTJpxOJ5mZmXz00UekpKSQl2fVVDtw4AALFy7k8OHDfPPNN3z66aeUlpYyefJk3n33XdLT0zllXzmojQEDBnD06NGrlk+aNIm//e1vOJ1OHA7HVeujo6OZOnUqjz/+OE6ns8bkkJaWRlZWFp999hlOp5P09HR27tzJ+++/T8eOHcnMzOTQoUOMHDmSGTNm0LFjR7Zv337DyQG0B3GZ0soEERnCD3mnCAwPRCLb8t0P3zGw3UAfR6fUTXKdX/reEhYWRnp6Ort27WL79u0kJyczd+5cBg0axKFDhxg+fDhgVWDt0KEDFy9eJCcnh6SkJACaNrUm+Nq9ezcPPPAADoeDdu3aMXToUD7//HOaN2/O4MGDiYqKAqiaAjQsLIxu3brRo0cPAMaPH8+yZctqFbOnMkUFBQVcvHiRIUOGAPDggw9eNlZRH2lpaaSlpREXFwdYZUCysrKIj49vkJLn16MJoppLJW4CMISFBnIuL4+gZhWUhXfgVNGX2oNQysscDgcJCQkkJCTQr18/Vq9ezcCBA+nTpw979+69bNv6VDutvHxVea7rTV1aGwcOHLgp9auMMaSmpjJlypSr1mVkZLBt2zaeeeYZEhMTmTlzZoOeWy8xVVNe6qI8AESE8rw8AoOLORneGoPRZyCU8qKvvvqKrKysqvdOp5OuXbvSs2dP8vPzqxJEeXk5X375JeHh4URFRfGPf/wDsCbgKS4uJj4+nvXr1+N2u8nPz2fnzp0MHjz4muft1asX2dnZHD9+HIA33nijVvEePHiQ5557jmnTpl22PDIykvDw8Ko7sN58802P+9elpPddd93FypUrKSwsBCAnJ4czZ87ccMnz2tAeRDXuMjcVAYJxu3GdOU3Q7Zf4Wp+BUMrrCgsLeeyxxygoKCAwMJDu3buzbNkygoOD2bhxIzNmzODChQu4XC5+97vf0adPH9asWcOUKVOYOXMmQUFBbNiwgaSkJPbu3Uv//v0REebNm0f79u09jhWAdWlq2bJl/OIXvyA0NJT4O8IzPQAAFm5JREFU+Phrfrnu2rWLuLg4iouLadu2LS+//DKJiYlXbbdixQomT55MQEAAQ4cO9Vgq/O6772bMmDFs3ry5xkHqESNGcOTIkarLVmFhYaxdu5Zjx46RkpJCQEAAQUFBLFmyBPix5HnlWMSN0HLf1Tz/1CdQWkFKam+ODU2g/aACtvxmEgu+fZfdv9qtdzGpRkvLfTecylLhYM1dnZeXx0svveTjqCxa7vtGlFuTBbnsux4CQ918Z8qIaBKhyUEpVStbt25lzpw5uFwuunbtyqpVq3wdUr15NUGIyEjgJcABvGqMmXvF+q7ASqANcB4Yb4w5KSKxwBKgOeAG/myM8frUSlJuMGEOyvN+nGr027Lv9fKSUqrWkpOTSU5O9nUYDcJrg9Qi4gD+Cvwn0Bt4QER6X7HZAuA1Y0wMMBuYYy8vBn5tjOkDjAQWikikt2Kt5HBDQHDAj1ONtm3LicIcHaBWSvklb97FNBg4Zoz5xhhzCXgTuOeKbXoDH9uvt1euN8Z8bYzJsl/nAmewehleFVRhCGwaSPmpPAKChfJ2XcgrytMehFLKL3kzQXQCTlR7f9JeVl0mcK/9OgkIF5FW1TcQkcFAMHDcS3EC4HJVEGyE4KYOXHmnCAw15Ea011tclVJ+y9fPQTwBDBWRA8BQIAdrzAEAEekArAEmGWMqrtxZRB4Rkf0isj8/P/+GArnwg1WQLzg0kPK8XIKalvFtaDigt7gqpfyTNxNEDlC9/GmUvayKMSbXGHOvMSYOeNpeVgAgIs2BrcDTxpj/5+kExphlxphBxphBbdrc2BWo7y9YCSIkNIjy3BwCQ118FxQEaJlvpbzt3LlzVWWu27dvT6dOnYiNjSUyMpLeva8curTMnDmTjz76qMZjawnv+vPmXUyfAz1EpBtWYvgV8GD1DUSkNXDe7h2kYt3RhIgEA+9gDWBv9GKMVQoqE0RTcJ8vIKijm+8op3lwc73FVSkva9WqFU6nE+CyyqjZ2dkev9wBZs+e7XG52+32WCTPE38s4V0XXksQxhiXiEwHPsC6zXWlMeZLEZkN7DfGbAESgDkiYoCdQOVz62OBnwGtRGSivWyiMcbprXh/uGgliPCKEsAu8+26qL0H5Xde+OwFjp73/ORxffVq2YsnBz9Zr33dbjeTJ09mz549dOrUic2bNxMSEsLEiRMZNWoUY8aMITo6muTkZD788EP+8Ic/EBkZqSW8G4BXn4MwxmwDtl2xbGa11xuBq3oIxpi1wFpvxnalQnuyoPBL1mP2QeEOvi3KI65t3M0MQyl1haysLN544w2WL1/O2LFjefvttxk/fvxV27Vq1YqMjAxKS0vp0aMHH3/8Md27d6/TMwkDBgxg/vz5Vy2fNGkSy5cvZ8iQIR4vDVWW8K5MKDWpXsLbGMPo0aPZuXMn+fn5dOzYka1btwJw4cIFIiIiePHFF+vV87hR+iS1rcieLCis5DwAZR06kFeUx9gWY30ZllI3XX1/6XtLt27diI2NBWDgwIFkZ2d73K4yERw9elRLeDcQTRC2ysmCQn84SwXwv53aAie4vcXtPo1LKX93ZZnukpISj9s1a9bshs/lDyW868LXt7neMsqKrNrwTb4/hSPY8HVz6xbXni16+jIspVQdaQnvhqM9CFtZqYsADAGnviMw1MXXDohoEkHb0La+Dk0pVQdawrvhaLlv2/w/70FySrg3ex5BpV/z1LTeNAlrx8q7VjZglErdmrTc9/XdyiW866Ku5b71EpPNXeam3CGUn8knMNRNVslpHX9QSgFWCe/Y2Fj69u3Lrl27eOaZZ3wd0k2hl5hsFZcqIMBQUVhCUZihxF2m4w9KKaBxlfCuC+1BVCo3OLAGqk+3tp5W1B6EUsqfaYKwSXkFwVgPy/1vm3ACJICfRP7Ex1EppZTv6CUmm8MNTQKschtHWzWhS3gHQgJDfByVUkr5jvYgbEEVhhBXIWA4EO6iZ0sdf1BK+TdNEIDbbU0W1KzsexwhFXxHsY4/KHWTORyOqrLYsbGx1yypURfZ2dm8/vrrVe9XrVrF9OnTa9wvISGBnj17EhMTQ69evZg+fToFBQVV66+s2OrJb37zGw4fPgzA888/73GbO+64g9jYWLp06UKbNm3q3Han08m2bdtq3rCeNEEAF36wxh5Ci89T3syal0gThFI3V0hICE6ns+pPdHT0DR3P5XJdlSDqYt26dRw8eJCDBw/SpEkT7rnnxxmTPc37cKVXX321ai6LayWIffv24XQ6mT17NsnJyXVuu7cThI5BAN9fKAWgSeF5LloVNjRBKL916vnnKTvSsOW+m/xbL9r/8Y913s/pdDJ16lSKi4u57bbbWLlyJS1atCAhIYEFCxYwaNAgzp49y6BBg8jOzmbVqlVs2rSJwsJC3G43ZWVlHDlyhNjYWCZMmECLFi3Izc1l5MiRHD9+nKSkJObNm3fdGIKDg5k3bx7du3cnMzOT/v37ExYWRmFhIRUVFUyfPp2PP/6Yzp07ExQUxEMPPcSYMWOqYty4cSMlJSXExsbSp08f1q1bd93zHT9+nGnTppGfn09oaCjLly+nV69ebNiwgVmzZuFwOIiIiOCjjz5i5syZlJSUsHv3blJTUxv8VlztQfDjZEHBhQWcaR5AeFAYHZp18HFUSvmXyi/R2NhYkpKSAPj1r3/NCy+8wMGDB+nXrx+zZs2q8TgZGRls3LiRTz75hLlz5xIfH4/T6eTxxx8HrKSzfv16vvjiC9avX8+JEydqPKbD4aB///4cPXp54ty0aRPZ2dkcPnyYNWvWsHfv3qv2nTt3blXvqKbkAFZZjUWLFpGens6CBQt49NFHAWuCpA8++IDMzEy2bNlCcHDwZT0PbzynoT0I4Ad7PurAsiK+bRlEjxa3IyI+jkop36jPL/2GUPklWunChQsUFBQwdOhQACZMmMD9999f43GGDx9Oy5Ytr7k+MTGxqpZS7969+fbbb+ncufM1t6/kqSzR7t27uf/++wkICKB9+/YMGzasxuNcT2FhIXv27LmsnWVl1vfTnXfeycSJExk7diz33nvvDZ2ntjRBABftyYICXSV8FRmgl5eUusUFBgZSUWGNF5aWll62rqay31eWD3e5XDWez+1288UXX3i9XlVFRQWRkZGXJcpKS5cuZd++fWzdupWBAweSnp7u1VhALzEBUFRkzQUR6CrhZHOjt7gqdQuIiIigRYsW7Nq1C4A1a9ZU9Saio6OrviA3brz2tPUNUSa7vLyc1NRUOnfuTExMzGXr7rzzTt5++20qKio4ffo0O3bs8HiMoKAgysvLazxX8+bN6datGxs2bACsXktmZiZgjU3ccccdzJ49mzZt2nDixAmvlwHXBAGUVEsQ55rrALVSt4rVq1eTkpJCTEwMTqezavKcJ554giVLlhAXF8fZs2evuX9MTEzV+MFf/vKXOp173LhxxMTE0LdvX4qKiti8efNV29x3331ERUXRu3dvxo8fz4ABAzyWAn/kkUeIiYlh3LhxNZ533bp1rFixgv79+9OnT5+q86akpNCvXz/69u3LT3/6U/r378+wYcM4fPgwsbGxrF+/vk7tqw0t9w288koG5mABP9s1g/Epwt5x+wgNCm3gCJW6dWm57/qrLAV+7tw5Bg8ezKeffkr79u19HZZHdS33rWMQwKUSF0EVLkqauYhqFqXJQSlVa6NGjaKgoIBLly7x7LPP3rLJoT40QQDlpS5CKko50xx6ttRfUUqp2rvWuENjoGMQgLusgkBXMbkRAfRo3dvX4Sil1C1BEwRQcclNUHkJZ3WAWimlqmiCAByl5dYdTOGiCUIppWyaIIDASy4CXSUURjjoFNbJ1+EopdQtQRME1lwQga5Swtq0IkD0n0QpX/BU7rt6We2UlBT69OlDSkoKq1atIjc3t2pddHT0dZ+HAGswOSIigri4OHr27MnPfvYz/vnPf1atX7p0Ka+99tp1j7F//35mzJhRdTxPVV3//ve/V7UhODiYfv36ERsby1NPPVWrfweAhQsXUlxcXOvtvUXvYgIcJoBAVwltom7zdShK+a0razHB5WW1ly1bxvnz53E4HCQkJNC3b186duxYp3PEx8dXJQWn08kvf/lLQkJCSExMZOrUqTXuP2jQIAYNsh4Z2LFjB2FhYVfNDTFp0iQmTZoEWIlr+/bttG7duk5xLly4kPHjxxMa6ttb7v0+QbjdFSBBYEro1nWAr8NRyud2vfU1Z08UNugxW3cOI35s3cf3Kstqjx49msLCQgYOHEhqair79+9n3LhxhISEVFVQXbRoEe+++y7l5eVs2LCBXr16XffYsbGxzJw5k8WLF5OYmMif/vQnwsLCeOKJJ/j88895+OGHCQgIYPjw4bz33nscOnSIHTt2sGDBAhYvXszSpUtxOBysXbuWRYsWER8ff93zzZ8/n7feeouysjKSkpKYNWsWRUVFjB07lpMnT+J2u3n22Wc5ffo0ubm5DBs2jNatW7N9+/Y6/7s1FL9PEBfsSq6lgSX07DTYx9Eo5b8qy30DdOvWjXfeeadq3ZYtWwgLC6vqYSxZsqRqPohKrVu3JiMjg1deeYUFCxbw6quv1njOAQMGMH/+/KuWT5o0ieXLlzNkyBCPl4aio6OZOnVqVUKpSVpaGllZWXz22WcYYxg9ejQ7d+4kPz+fjh07snXrVsCqYBsREcGLL75Yr55HQ/P7BFFc7qbF93soC8ihe8vr/+JQyh/U55d+Q/B0iakuKktgDxw4kE2bNtVqH0+lhgoKCrh48SJDhgwB4MEHH7xsrKI+0tLSSEtLIy4uDrDKc2RlZREfH8/vf/97nnzySUaNGlVjL+Rm8/sEEdU2jNuPv84XPwkgLDjM1+Eopeqpsox3bUt4Axw4cOCm1KAyxpCamsqUKVOuWpeRkcG2bdt45plnSExMrCpIeCvw6i07IjJSRL4SkWMiclU/TUS6isi/ROSgiOwQkahq6yaISJb9Z4K3YjSXLhFSaHBENvXWKZRSDawhylwfPHiQ5557jmnTpl22PDIykvDwcPbt2wfAm2++ecMx3HXXXaxcuZLCQmtsJycnhzNnzpCbm0toaCjjx48nJSWFjIyMOh/bm7zWgxARB/BXYDhwEvhcRLYYYw5X22wB8JoxZrWI/ByYA/yXiLQE/gcYBBgg3d73+4aOsyj3GwpDIKxdm4Y+tFLKSyZOnMjUqVMvG6SujV27dhEXF0dxcTFt27bl5ZdfJjEx8artVqxYweTJkwkICGDo0KEeS3jffffdjBkzhs2bN9c4SD1ixAiOHDlSddkqLCyMtWvXcuzYMVJSUggICCAoKIglS5YAVnnwkSNH0rFjR58OUnut3LeIDAH+ZIy5y36fCmCMmVNtmy+BkcaYE2LN8XnBGNNcRB4AEowxU+zt/gbsMMa8ca3z1bfc97nzx3nhw2kk3T6WIXEP1Xl/pRoDLfd9ucoS3mDNKZ2Xl8dLL73k46hu3K1U7rsTUH028JPAHVdskwncC7wEJAHhItLqGvte9YiziDwCPALQpUuXegXZquVtzEt+v177KqUap61btzJnzhxcLhddu3Zl1apVvg7JJ3w9SP0EsFhEJgI7gRzAXdudjTHLgGVg9SC8EaBSyv8kJyeTnJzs6zB8zpsJIgfoXO19lL2sijEmF6sHgYiEAfcZYwpEJAdIuGLfHV6MVSm/Z4zButKrGqP6DCd48y6mz4EeItJNRIKBXwFbqm8gIq1FqoofpQIr7dcfACNEpIWItABG2MuUUl7QtGlTzp07V68vEXXrM8Zw7tw5mjat292aXutBGGNcIjId64vdAaw0xnwpIrOB/caYLVi9hDkiYrAuMU2z9z0vIs9hJRmA2caY896KVSl/FxUVxcmTJ8nPz/d1KMpLmjZtSlRUVM0bVuO1u5hutvrexaSUUv7sencxaW1rpZRSHmmCUEop5ZEmCKWUUh41mjEIEckHvr2BQ7QGrj8lVeOk7fYv2m7/Upt2dzXGeKw11GgSxI0Skf3XGqhpzLTd/kXb7V9utN16iUkppZRHmiCUUkp5pAniR8t8HYCPaLv9i7bbv9xQu3UMQimllEfag1BKKeWRJgillFIe+X2CqGne7MZERFaKyBkROVRtWUsR+dCe+/tDu3puoyEinUVku4gcFpEvReS39vLG3u6mIvKZiGTa7Z5lL+8mIvvsz/t6u9JyoyMiDhE5ICL/tN/7S7uzReQLEXGKyH57Wb0/636dIKrNm/2fQG/gARHp7duovGoVMPKKZU8B/zLG9AD+Zb9vTFzA740xvYF/B6bZ/40be7vLgJ8bY/oDscBIEfl34AXgL8aY7sD3wMM+jNGbfgscqfbeX9oNMMwYE1vt+Yd6f9b9OkEAg4FjxphvjDGXgDeBe3wck9cYY3YCV5ZNvwdYbb9eDfzypgblZcaYPGNMhv36ItaXRicaf7uNMabQfhtk/zHAz4GN9vJG124AEYkCfgG8ar8X/KDd11Hvz7q/J4hazX3dyLUzxuTZr08B7XwZjDeJSDQQB+zDD9ptX2ZxAmeAD4HjQIExxmVv0lg/7wuBPwAV9vtW+Ee7wfoRkCYi6SLyiL2s3p91X89JrW4hxhhjT97U6NhT2r4N/M4Y80P1qTUba7uNMW4gVkQigXeAXj4OyetEZBRwxhiTLiIJvo7HB/7DGJMjIm2BD0XkaPWVdf2s+3sPosZ5s/3AaRHpAGD/fcbH8TQ4EQnCSg7rjDGb7MWNvt2VjDEFwHZgCBApIpU/DBvj5/1OYLSIZGNdMv458BKNv90AGGNy7L/PYP0oGMwNfNb9PUHUOG+2H9gCTLBfTwA2+zCWBmdff14BHDHGvFhtVWNvdxu754CIhADDscZftgNj7M0aXbuNManGmChjTDTW/88fG2PG0cjbDSAizUQkvPI1MAI4xA181v3+SWoR+T9Y1ywr583+s49D8hoReQNrHvDWwGngf4B/AG8BXbDKpY9tTPN/i8h/ALuAL/jxmvQfscYhGnO7Y7AGJB1YPwTfMsbMFpGfYP2ybgkcAMYbY8p8F6n32JeYnjDGjPKHdtttfMd+Gwi8boz5s4i0op6fdb9PEEoppTzz90tMSimlrkEThFJKKY80QSillPJIE4RSSimPNEEopZTySBOEUnUgIm67UmblnwYr8ici0dUr7Srla1pqQ6m6KTHGxPo6CKVuBu1BKNUA7Dr88+xa/J+JSHd7ebSIfCwiB0XkXyLSxV7eTkTesedryBSRn9qHcojIcnsOhzT7KWilfEIThFJ1E3LFJabkausuGGP6AYuxns4HWASsNsbEAOuAl+3lLwOf2PM1DAC+tJf3AP5qjOkDFAD3ebk9Sl2TPkmtVB2ISKExJszD8mysCXq+sYsDnjLGtBKRs0AHY0y5vTzPGNNaRPKBqOrlHuxy5B/aE7sgIk8CQcaY/+v9lil1Ne1BKNVwzDVe10X1+kBudJxQ+ZAmCKUaTnK1v/far/dgVRUFGIdVOBCsqR//G6om9om4WUEqVVv660SpugmxZ2mr9L4xpvJW1xYichCrF/CAvewx4O8ikgLkA5Ps5b8FlonIw1g9hf8G8lDqFqJjEEo1AHsMYpAx5qyvY1GqoeglJqWUUh5pD0IppZRH2oNQSinlkSYIpZRSHmmCUEop5ZEmCKWUUh5pglBKKeXR/weGVKvrKBmWQwAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "MQIbUVBY6-SG",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        },
        "outputId": "0596423a-7d03-46d0-b108-a0f126218d04"
      },
      "source": [
        "# Save model and weights\n",
        "if not os.path.isdir(save_dir):\n",
        "    os.makedirs(save_dir)\n",
        "model_path = os.path.join(save_dir, model_name)\n",
        "model.save(model_path)\n",
        "print('Saved trained model at %s ' % model_path)"
      ],
      "execution_count": 20,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Saved trained model at /content/saved_models/keras_cifar10_trained_model.h5 \n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "x8cshRKD7Bak",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 70
        },
        "outputId": "75117368-9ece-4ef2-965d-d0f76781c72c"
      },
      "source": [
        "# Score trained model.\n",
        "scores = model.evaluate(x_train, s_train, verbose=1)\n",
        "print('Train loss:     %f' % np.mean(scores[0:5]))\n",
        "acc = 1.\n",
        "for i in range(5):\n",
        "    acc *= scores[6+i]\n",
        "print('Train accuracy: %.2f' % (acc * 100.))"
      ],
      "execution_count": 21,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "27133/27133 [==============================] - 5s 179us/step\n",
            "Train loss:     0.000731\n",
            "Train accuracy: 99.96\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "yfQ_5Ago7EWh",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 70
        },
        "outputId": "dc90d08c-c0cb-4a7d-e498-85a0c9d9fbbf"
      },
      "source": [
        "scores = model.evaluate(x_test, s_test, verbose=1)\n",
        "print('Test loss:     %f' % np.mean(scores[0:5]))\n",
        "acc = 1.\n",
        "for i in range(5):\n",
        "    acc *= scores[6+i]\n",
        "print('Test accuracy: %.2f' % (acc * 100.))"
      ],
      "execution_count": 22,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "3107/3107 [==============================] - 1s 187us/step\n",
            "Test loss:     0.014718\n",
            "Test accuracy: 98.84\n"
          ],
          "name": "stdout"
        }
      ]
    }
  ]
}