BatchNorm_Keras.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Untitled1.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "machine_shape": "hm"
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "id": "Yi4cOIuV05rV",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "from matplotlib.pyplot import imshow\n",
        "import sklearn as skt\n",
        "from sklearn.preprocessing import OneHotEncoder #not used yet, learn \n",
        "from sklearn.model_selection import train_test_split\n",
        "from sklearn.metrics import classification_report\n",
        "import numpy as np\n",
        "import keras\n",
        "from keras import layers\n",
        "from keras.layers import Input, Add, Dense, Activation, ZeroPadding2D, BatchNormalization, Flatten, Conv2D, AveragePooling2D, MaxPooling2D, GlobalMaxPooling2D, Dropout\n",
        "from keras.models import Model\n",
        "from keras.initializers import glorot_uniform\n",
        "from keras.utils import layer_utils\n",
        "import keras.backend as K\n",
        "from keras.optimizers import SGD\n",
        "from keras.utils import np_utils\n",
        "K.set_image_data_format('channels_last')\n",
        "from keras.datasets import fashion_mnist\n",
        "from imutils import build_montages # for visualization\n",
        "%matplotlib inline"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "YrOcQCjL2P79",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def data_loader():\n",
        "  # load data \n",
        "  print(\"[INFO] data loading...\")\n",
        "  fashion_mnist = keras.datasets.fashion_mnist\n",
        "  ((X_train, y_train), (X_test, y_test)) = fashion_mnist.load_data()\n",
        "  _, height, width= X_train.shape\n",
        "  depth = 1 # fashion mnsit is greyscale \n",
        "  #channelDim = -1 \n",
        "  nb_classes = len(np.unique(y_train))\n",
        "  # preprocess data \n",
        "  X_train = X_train.reshape((X_train.shape[0], height, width, depth))\n",
        "  X_test = X_test.reshape((X_test.shape[0], height, width, depth))\n",
        "  \n",
        "  # normalization \n",
        "  X_train = X_train.astype(\"float32\") / 255.0 \n",
        "  X_test = X_test.astype(\"float32\") / 255.0 \n",
        "  \n",
        "  # one hot encoding\n",
        "  y_train = np_utils.to_categorical(y_train, nb_classes)\n",
        "  y_test = np_utils.to_categorical(y_test, nb_classes)\n",
        "\n",
        "  X_train, X_validation, y_train, y_validation = train_test_split(X_train, y_train, test_size = 0.2, random_state=42)\n",
        "  labelNames = [\"top\", \"trousers\", \"pullover\", \"dress\", \"coat\", \"sandal\", \"shirt\", \"sneakers\", \"bag\", \"ankle boot\"]\n",
        "\n",
        "  return\n",
        "\n",
        "def model_keras(input_shape, classes, BN):\n",
        "  X_input = Input(input_shape)\n",
        "  \n",
        "  X = Conv2D(filters = 32, kernel_size = (3, 3), strides = (1,1), padding = 'same', kernel_initializer = glorot_uniform(seed=0))(X_input)\n",
        "  if BN == True:  \n",
        "    X = BatchNormalization(axis = 3)(X)\n",
        "  X = Activation('relu')(X)\n",
        "\n",
        "  X = Conv2D(filters = 48, kernel_size = (3, 3), strides = (1,1), padding = 'valid', kernel_initializer = glorot_uniform(seed=0))(X)\n",
        "  if BN == True:  \n",
        "    X = BatchNormalization(axis = 3)(X)\n",
        "  X = Activation('relu')(X)\n",
        "  X = MaxPooling2D((2, 2), strides=(2, 2))(X)\n",
        "\n",
        "  X = Conv2D(filters = 64, kernel_size = (3, 3), strides = (1,1), padding = 'valid', kernel_initializer = glorot_uniform(seed=0))(X)\n",
        "  if BN == True:  \n",
        "    X = BatchNormalization(axis = 3)(X)\n",
        "  X = Activation('relu')(X)\n",
        "\n",
        "  X = Conv2D(filters = 64, kernel_size = (3, 3), strides = (1,1), padding = 'same', kernel_initializer = glorot_uniform(seed=0))(X)\n",
        "  if BN == True:  \n",
        "    X = BatchNormalization(axis = 3)(X)\n",
        "  X = Activation('relu')(X)\n",
        "  X = MaxPooling2D((2, 2), strides=(2, 2))(X)\n",
        "\n",
        "  X = Flatten()(X)\n",
        "  X = Dense(256, activation='relu', kernel_initializer = glorot_uniform(seed=0), use_bias=False)(X)\n",
        "  if BN == True:  \n",
        "    X = BatchNormalization()(X)\n",
        "  X = Dense(classes, activation='softmax', kernel_initializer = glorot_uniform(seed=0))(X)\n",
        "\n",
        "  model = Model(inputs = X_input, outputs = X)\n",
        "\n",
        "  return model\n",
        "\n",
        "def train_model(optimizer,BN,base):\n",
        "  print(\"[INFO] compiling model...\")\n",
        "  if BN == True:\n",
        "    model_BN = model_keras(input_shape=(28, 28, 1), classes=10, BN=True)\n",
        "    model_BN.summary()\n",
        "    model_BN.compile(loss=\"categorical_crossentropy\", optimizer=optimizer, metrics=[\"accuracy\"])\n",
        "    print(\"[INFO] training model...\")\n",
        "    trained_model_BN = model_BN.fit(X_train, y_train, validation_data=(X_validation, y_validation), batch_size= mini_batch_size, epochs=nb_epochs)\n",
        "  if base == True:\n",
        "    model_base = model_keras(input_shape=(28, 28, 1), classes=10, BN=False)\n",
        "    model_base.summary()\n",
        "    model_base.compile(loss=\"categorical_crossentropy\", optimizer=optimizer, metrics=[\"accuracy\"])\n",
        "    print(\"[INFO] training model...\")\n",
        "    trained_model_base = model_base.fit(X_train, y_train, validation_data=(X_validation, y_validation), batch_size= mini_batch_size, epochs=nb_epochs)\n",
        "  return"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7VDbQyzr_yRG",
        "colab_type": "code",
        "outputId": "a6ec2ecb-e7fb-4901-ee28-af5489502f5b",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 2193
        }
      },
      "source": [
        "optimizer='adam'\n",
        "\n",
        "#optimizer = SGD(learning_rate=0.1, momentum = 0.9, decay = 1/np.exp(10))\n",
        "\n",
        "mini_batch_size = 256            # change, small to begin \n",
        "nb_epochs = 10                   # change, small to begin \n",
        "\n",
        "data_loader()\n",
        "base = True\n",
        "BN = True\n",
        "\n",
        "train_model(optimizer,BN,base)\n",
        "\n",
        "#models.save_model(model, filepath, overwrite=True, include_optimizer=True, save_format=None, signatures=None, options=None)\n",
        "#models.save_model(model, filepath, overwrite=True, include_optimizer=True, save_format=None, signatures=None, options=None)"
      ],
      "execution_count": 207,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[INFO] data loading...\n",
            "[INFO] compiling model...\n",
            "Model: \"model_48\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "input_50 (InputLayer)        (None, 28, 28, 1)         0         \n",
            "_________________________________________________________________\n",
            "conv2d_197 (Conv2D)          (None, 28, 28, 32)        320       \n",
            "_________________________________________________________________\n",
            "batch_normalization_134 (Bat (None, 28, 28, 32)        128       \n",
            "_________________________________________________________________\n",
            "activation_213 (Activation)  (None, 28, 28, 32)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_198 (Conv2D)          (None, 26, 26, 48)        13872     \n",
            "_________________________________________________________________\n",
            "batch_normalization_135 (Bat (None, 26, 26, 48)        192       \n",
            "_________________________________________________________________\n",
            "activation_214 (Activation)  (None, 26, 26, 48)        0         \n",
            "_________________________________________________________________\n",
            "max_pooling2d_99 (MaxPooling (None, 13, 13, 48)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_199 (Conv2D)          (None, 11, 11, 64)        27712     \n",
            "_________________________________________________________________\n",
            "batch_normalization_136 (Bat (None, 11, 11, 64)        256       \n",
            "_________________________________________________________________\n",
            "activation_215 (Activation)  (None, 11, 11, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_200 (Conv2D)          (None, 11, 11, 64)        36928     \n",
            "_________________________________________________________________\n",
            "batch_normalization_137 (Bat (None, 11, 11, 64)        256       \n",
            "_________________________________________________________________\n",
            "activation_216 (Activation)  (None, 11, 11, 64)        0         \n",
            "_________________________________________________________________\n",
            "max_pooling2d_100 (MaxPoolin (None, 5, 5, 64)          0         \n",
            "_________________________________________________________________\n",
            "flatten_50 (Flatten)         (None, 1600)              0         \n",
            "_________________________________________________________________\n",
            "dense_97 (Dense)             (None, 256)               409600    \n",
            "_________________________________________________________________\n",
            "batch_normalization_138 (Bat (None, 256)               1024      \n",
            "_________________________________________________________________\n",
            "dense_98 (Dense)             (None, 10)                2570      \n",
            "=================================================================\n",
            "Total params: 492,858\n",
            "Trainable params: 491,930\n",
            "Non-trainable params: 928\n",
            "_________________________________________________________________\n",
            "[INFO] training model...\n",
            "Train on 48000 samples, validate on 12000 samples\n",
            "Epoch 1/10\n",
            "48000/48000 [==============================] - 109s 2ms/step - loss: 0.4027 - accuracy: 0.8553 - val_loss: 0.2818 - val_accuracy: 0.8972\n",
            "Epoch 2/10\n",
            "48000/48000 [==============================] - 109s 2ms/step - loss: 0.2512 - accuracy: 0.9090 - val_loss: 0.2594 - val_accuracy: 0.9054\n",
            "Epoch 3/10\n",
            "48000/48000 [==============================] - 109s 2ms/step - loss: 0.2081 - accuracy: 0.9247 - val_loss: 0.2495 - val_accuracy: 0.9103\n",
            "Epoch 4/10\n",
            "48000/48000 [==============================] - 109s 2ms/step - loss: 0.1750 - accuracy: 0.9363 - val_loss: 0.2454 - val_accuracy: 0.9122\n",
            "Epoch 5/10\n",
            "48000/48000 [==============================] - 108s 2ms/step - loss: 0.1546 - accuracy: 0.9442 - val_loss: 0.2421 - val_accuracy: 0.9154\n",
            "Epoch 6/10\n",
            "48000/48000 [==============================] - 108s 2ms/step - loss: 0.1313 - accuracy: 0.9528 - val_loss: 0.2452 - val_accuracy: 0.9157\n",
            "Epoch 7/10\n",
            "48000/48000 [==============================] - 109s 2ms/step - loss: 0.1139 - accuracy: 0.9588 - val_loss: 0.2720 - val_accuracy: 0.9090\n",
            "Epoch 8/10\n",
            "48000/48000 [==============================] - 109s 2ms/step - loss: 0.1018 - accuracy: 0.9634 - val_loss: 0.2549 - val_accuracy: 0.9180\n",
            "Epoch 9/10\n",
            "48000/48000 [==============================] - 109s 2ms/step - loss: 0.0831 - accuracy: 0.9700 - val_loss: 0.2806 - val_accuracy: 0.9136\n",
            "Epoch 10/10\n",
            "48000/48000 [==============================] - 109s 2ms/step - loss: 0.0754 - accuracy: 0.9731 - val_loss: 0.2963 - val_accuracy: 0.9169\n",
            "Model: \"model_49\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "input_51 (InputLayer)        (None, 28, 28, 1)         0         \n",
            "_________________________________________________________________\n",
            "conv2d_201 (Conv2D)          (None, 28, 28, 32)        320       \n",
            "_________________________________________________________________\n",
            "activation_217 (Activation)  (None, 28, 28, 32)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_202 (Conv2D)          (None, 26, 26, 48)        13872     \n",
            "_________________________________________________________________\n",
            "activation_218 (Activation)  (None, 26, 26, 48)        0         \n",
            "_________________________________________________________________\n",
            "max_pooling2d_101 (MaxPoolin (None, 13, 13, 48)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_203 (Conv2D)          (None, 11, 11, 64)        27712     \n",
            "_________________________________________________________________\n",
            "activation_219 (Activation)  (None, 11, 11, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_204 (Conv2D)          (None, 11, 11, 64)        36928     \n",
            "_________________________________________________________________\n",
            "activation_220 (Activation)  (None, 11, 11, 64)        0         \n",
            "_________________________________________________________________\n",
            "max_pooling2d_102 (MaxPoolin (None, 5, 5, 64)          0         \n",
            "_________________________________________________________________\n",
            "flatten_51 (Flatten)         (None, 1600)              0         \n",
            "_________________________________________________________________\n",
            "dense_99 (Dense)             (None, 256)               409600    \n",
            "_________________________________________________________________\n",
            "dense_100 (Dense)            (None, 10)                2570      \n",
            "=================================================================\n",
            "Total params: 491,002\n",
            "Trainable params: 491,002\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n",
            "[INFO] training model...\n",
            "Train on 48000 samples, validate on 12000 samples\n",
            "Epoch 1/10\n",
            "48000/48000 [==============================] - 83s 2ms/step - loss: 0.6133 - accuracy: 0.7750 - val_loss: 0.3867 - val_accuracy: 0.8592\n",
            "Epoch 2/10\n",
            "48000/48000 [==============================] - 82s 2ms/step - loss: 0.3498 - accuracy: 0.8736 - val_loss: 0.3286 - val_accuracy: 0.8767\n",
            "Epoch 3/10\n",
            "48000/48000 [==============================] - 83s 2ms/step - loss: 0.2897 - accuracy: 0.8944 - val_loss: 0.2947 - val_accuracy: 0.8906\n",
            "Epoch 4/10\n",
            "48000/48000 [==============================] - 81s 2ms/step - loss: 0.2467 - accuracy: 0.9101 - val_loss: 0.2754 - val_accuracy: 0.9007\n",
            "Epoch 5/10\n",
            "48000/48000 [==============================] - 81s 2ms/step - loss: 0.2255 - accuracy: 0.9172 - val_loss: 0.2533 - val_accuracy: 0.9093\n",
            "Epoch 6/10\n",
            "48000/48000 [==============================] - 81s 2ms/step - loss: 0.2010 - accuracy: 0.9256 - val_loss: 0.2316 - val_accuracy: 0.9177\n",
            "Epoch 7/10\n",
            "48000/48000 [==============================] - 81s 2ms/step - loss: 0.1860 - accuracy: 0.9317 - val_loss: 0.2438 - val_accuracy: 0.9115\n",
            "Epoch 8/10\n",
            "48000/48000 [==============================] - 81s 2ms/step - loss: 0.1661 - accuracy: 0.9385 - val_loss: 0.2285 - val_accuracy: 0.9173\n",
            "Epoch 9/10\n",
            "48000/48000 [==============================] - 81s 2ms/step - loss: 0.1524 - accuracy: 0.9435 - val_loss: 0.2365 - val_accuracy: 0.9151\n",
            "Epoch 10/10\n",
            "48000/48000 [==============================] - 81s 2ms/step - loss: 0.1331 - accuracy: 0.9514 - val_loss: 0.2312 - val_accuracy: 0.9190\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "--pXtB39AWhW",
        "colab_type": "code",
        "outputId": "52d8f5a5-3644-4f39-a895-bd574e481a9f",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 703
        }
      },
      "source": [
        "print(\"Base model\")\n",
        "preds = model_base.evaluate(X_test, y_test)\n",
        "print (\"Loss = \" + str(preds[0]))\n",
        "print (\"Test Accuracy = \" + str(preds[1]))\n",
        " \n",
        "# make predictions on the test set \n",
        "predictions_base = model_base.predict(X_test)\n",
        "\n",
        "# classification report\n",
        "print(\"[INFO] evaluating network...\")\n",
        "print(\"Base model\")\n",
        "print(classification_report(y_test.argmax(axis=1), predictions_base.argmax(axis=1), target_names = labelNames))\n",
        "\n",
        "# ploting results\n",
        "plt.figure()\n",
        "#Base\n",
        "plt.plot(np.arange(0, len(trained_model_base.history[\"loss\"])), trained_model_base.history[\"loss\"], label=\"train_loss_Base\")\n",
        "plt.plot(np.arange(0, len(trained_model_base.history[\"val_loss\"])), trained_model_base.history[\"val_loss\"], label=\"val_loss_Base\")\n",
        "plt.plot(np.arange(0, len(trained_model_base.history[\"accuracy\"])), trained_model_base.history[\"accuracy\"], label=\"train_acc_Base\")\n",
        "plt.plot(np.arange(0, len(trained_model_base.history[\"val_accuracy\"])), trained_model_base.history[\"val_accuracy\"], label=\"val_acc_Base\")\n",
        "plt.title(\"Training Loss and accurracy on Dataset : Fshion MNIST\")\n",
        "plt.xlabel(\"Nb of Epoch\")\n",
        "plt.ylabel(\"Loss & accuracy\")\n",
        "plt.legend(loc=\"lower left\")\n",
        "\n"
      ],
      "execution_count": 225,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Base model\n",
            "10000/10000 [==============================] - 6s 592us/step\n",
            "Loss = 0.243783204382658\n",
            "Test Accuracy = 0.9133999943733215\n",
            "[INFO] evaluating network...\n",
            "Base model\n",
            "              precision    recall  f1-score   support\n",
            "\n",
            "         top       0.88      0.84      0.86      1000\n",
            "    trousers       1.00      0.98      0.99      1000\n",
            "    pullover       0.84      0.89      0.87      1000\n",
            "       dress       0.89      0.92      0.91      1000\n",
            "        coat       0.85      0.85      0.85      1000\n",
            "      sandal       0.98      0.98      0.98      1000\n",
            "       shirt       0.77      0.74      0.76      1000\n",
            "    sneakers       0.96      0.97      0.97      1000\n",
            "         bag       0.98      0.98      0.98      1000\n",
            "  ankle boot       0.98      0.97      0.97      1000\n",
            "\n",
            "    accuracy                           0.91     10000\n",
            "   macro avg       0.91      0.91      0.91     10000\n",
            "weighted avg       0.91      0.91      0.91     10000\n",
            "\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.legend.Legend at 0x7f5463d90978>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 225
        },
        {
          "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": "QO60wyWvPAdi",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 703
        },
        "outputId": "5a6634a1-9125-41df-8af9-805da0e07e75"
      },
      "source": [
        "\n",
        "#BN\n",
        "print(\"BN model\")\n",
        "preds = model_BN.evaluate(X_test, y_test)\n",
        "print (\"Loss = \" + str(preds[0]))\n",
        "print (\"Test Accuracy = \" + str(preds[1]))\n",
        " \n",
        "predictions_BN = model_BN.predict(X_test)\n",
        "print(\"[INFO] evaluating network...\")\n",
        "print(\"BN model\")\n",
        "print(classification_report(y_test.argmax(axis=1), predictions_BN.argmax(axis=1), target_names = labelNames))\n",
        "plt.figure()\n",
        "plt.plot(np.arange(0, len(trained_model_BN.history[\"loss\"])), trained_model_BN.history[\"loss\"], label=\"train_loss_BN\")\n",
        "plt.plot(np.arange(0, len(trained_model_BN.history[\"val_loss\"])), trained_model_BN.history[\"val_loss\"], label=\"val_loss_BN\")\n",
        "plt.plot(np.arange(0, len(trained_model_BN.history[\"accuracy\"])), trained_model_BN.history[\"accuracy\"], label=\"train_acc_BN\")\n",
        "plt.plot(np.arange(0, len(trained_model_BN.history[\"val_accuracy\"])), trained_model_BN.history[\"val_accuracy\"], label=\"val_acc_BN\")\n",
        "\n",
        "plt.title(\"Training Loss and accurracy on Dataset : Fshion MNIST\")\n",
        "plt.xlabel(\"Nb of Epoch\")\n",
        "plt.ylabel(\"Loss & accuracy\")\n",
        "plt.legend(loc=\"lower left\")"
      ],
      "execution_count": 223,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "BN model\n",
            "10000/10000 [==============================] - 8s 835us/step\n",
            "Loss = 0.27714633579850195\n",
            "Test Accuracy = 0.9041000008583069\n",
            "[INFO] evaluating network...\n",
            "BN model\n",
            "              precision    recall  f1-score   support\n",
            "\n",
            "         top       0.85      0.88      0.87      1000\n",
            "    trousers       0.99      0.98      0.99      1000\n",
            "    pullover       0.86      0.87      0.87      1000\n",
            "       dress       0.94      0.89      0.91      1000\n",
            "        coat       0.74      0.96      0.83      1000\n",
            "      sandal       0.97      1.00      0.98      1000\n",
            "       shirt       0.84      0.57      0.68      1000\n",
            "    sneakers       0.98      0.92      0.95      1000\n",
            "         bag       0.96      0.99      0.97      1000\n",
            "  ankle boot       0.94      0.98      0.96      1000\n",
            "\n",
            "    accuracy                           0.90     10000\n",
            "   macro avg       0.91      0.90      0.90     10000\n",
            "weighted avg       0.91      0.90      0.90     10000\n",
            "\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.legend.Legend at 0x7f5463dfa6d8>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 223
        },
        {
          "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": "7qEDFJEZLtUP",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "c346c656-e96a-4798-9f49-e6648abca381"
      },
      "source": [
        "print (len(trained_model_BN.history[\"loss\"]))\n"
      ],
      "execution_count": 222,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "5\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "18ayJpHIN9Y4",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}