ComputerVision_(chapter_14).ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": [],
      "gpuType": "T4",
      "authorship_tag": "ABX9TyMtlqvFIpHrhBq6870k7AZP",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/S4vyss/machine-learning/blob/main/ComputerVision_(chapter_14).ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# **Exercise 9**"
      ],
      "metadata": {
        "id": "-0VH62pkP-R2"
      }
    },
    {
      "cell_type": "code",
      "execution_count": 48,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Z-qyXajxwcUf",
        "outputId": "978ffff6-e372-4e49-f9b7-5621e71618dc"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Requirement already satisfied: tensorflow-datasets in /usr/local/lib/python3.10/dist-packages (4.9.3)\n",
            "Requirement already satisfied: absl-py in /usr/local/lib/python3.10/dist-packages (from tensorflow-datasets) (1.4.0)\n",
            "Requirement already satisfied: array-record in /usr/local/lib/python3.10/dist-packages (from tensorflow-datasets) (0.4.1)\n",
            "Requirement already satisfied: click in /usr/local/lib/python3.10/dist-packages (from tensorflow-datasets) (8.1.7)\n",
            "Requirement already satisfied: dm-tree in /usr/local/lib/python3.10/dist-packages (from tensorflow-datasets) (0.1.8)\n",
            "Requirement already satisfied: etils[enp,epath,etree]>=0.9.0 in /usr/local/lib/python3.10/dist-packages (from tensorflow-datasets) (1.5.0)\n",
            "Requirement already satisfied: numpy in /usr/local/lib/python3.10/dist-packages (from tensorflow-datasets) (1.23.5)\n",
            "Requirement already satisfied: promise in /usr/local/lib/python3.10/dist-packages (from tensorflow-datasets) (2.3)\n",
            "Requirement already satisfied: protobuf>=3.20 in /usr/local/lib/python3.10/dist-packages (from tensorflow-datasets) (3.20.3)\n",
            "Requirement already satisfied: psutil in /usr/local/lib/python3.10/dist-packages (from tensorflow-datasets) (5.9.5)\n",
            "Requirement already satisfied: requests>=2.19.0 in /usr/local/lib/python3.10/dist-packages (from tensorflow-datasets) (2.31.0)\n",
            "Requirement already satisfied: tensorflow-metadata in /usr/local/lib/python3.10/dist-packages (from tensorflow-datasets) (1.14.0)\n",
            "Requirement already satisfied: termcolor in /usr/local/lib/python3.10/dist-packages (from tensorflow-datasets) (2.3.0)\n",
            "Requirement already satisfied: toml in /usr/local/lib/python3.10/dist-packages (from tensorflow-datasets) (0.10.2)\n",
            "Requirement already satisfied: tqdm in /usr/local/lib/python3.10/dist-packages (from tensorflow-datasets) (4.66.1)\n",
            "Requirement already satisfied: wrapt in /usr/local/lib/python3.10/dist-packages (from tensorflow-datasets) (1.15.0)\n",
            "Requirement already satisfied: fsspec in /usr/local/lib/python3.10/dist-packages (from etils[enp,epath,etree]>=0.9.0->tensorflow-datasets) (2023.6.0)\n",
            "Requirement already satisfied: importlib_resources in /usr/local/lib/python3.10/dist-packages (from etils[enp,epath,etree]>=0.9.0->tensorflow-datasets) (6.1.0)\n",
            "Requirement already satisfied: typing_extensions in /usr/local/lib/python3.10/dist-packages (from etils[enp,epath,etree]>=0.9.0->tensorflow-datasets) (4.5.0)\n",
            "Requirement already satisfied: zipp in /usr/local/lib/python3.10/dist-packages (from etils[enp,epath,etree]>=0.9.0->tensorflow-datasets) (3.17.0)\n",
            "Requirement already satisfied: charset-normalizer<4,>=2 in /usr/local/lib/python3.10/dist-packages (from requests>=2.19.0->tensorflow-datasets) (3.3.0)\n",
            "Requirement already satisfied: idna<4,>=2.5 in /usr/local/lib/python3.10/dist-packages (from requests>=2.19.0->tensorflow-datasets) (3.4)\n",
            "Requirement already satisfied: urllib3<3,>=1.21.1 in /usr/local/lib/python3.10/dist-packages (from requests>=2.19.0->tensorflow-datasets) (2.0.6)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.10/dist-packages (from requests>=2.19.0->tensorflow-datasets) (2023.7.22)\n",
            "Requirement already satisfied: six in /usr/local/lib/python3.10/dist-packages (from promise->tensorflow-datasets) (1.16.0)\n",
            "Requirement already satisfied: googleapis-common-protos<2,>=1.52.0 in /usr/local/lib/python3.10/dist-packages (from tensorflow-metadata->tensorflow-datasets) (1.60.0)\n"
          ]
        }
      ],
      "source": [
        "pip install tensorflow-datasets"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import tensorflow as tf\n",
        "from tensorflow import keras\n",
        "import numpy as np\n",
        "\n",
        "(X_train_full, y_train_full), (X_test, y_test) = keras.datasets.mnist.load_data()\n",
        "X_train_full = X_train_full / 255.\n",
        "X_test = X_test / 255.\n",
        "X_train, X_valid = X_train_full[:-5000], X_train_full[-5000:]\n",
        "y_train, y_valid = y_train_full[:-5000], y_train_full[-5000:]\n",
        "\n",
        "X_train = X_train[..., np.newaxis]\n",
        "X_valid = X_valid[..., np.newaxis]\n",
        "X_test = X_test[..., np.newaxis]"
      ],
      "metadata": {
        "id": "gcHMXqfZF5H5"
      },
      "execution_count": 49,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "len(X_train)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ghyQ1jEmJS1f",
        "outputId": "9cbc1917-9691-48e3-e5b2-2e8e7b29cff1"
      },
      "execution_count": 50,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "55000"
            ]
          },
          "metadata": {},
          "execution_count": 50
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "len(X_valid)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "DBXJAO9UJXq-",
        "outputId": "9ef3a15d-37f5-411e-8e31-a6a7f692d688"
      },
      "execution_count": 51,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "5000"
            ]
          },
          "metadata": {},
          "execution_count": 51
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "data_augmentation = tf.keras.Sequential([\n",
        "    keras.layers.RandomFlip(\"horizontal_and_vertical\"),\n",
        "    keras.layers.RandomRotation(0.2),\n",
        "])\n",
        "\n",
        "X_train_augumented = data_augmentation(X_train)"
      ],
      "metadata": {
        "id": "FHORsb2bT5w6"
      },
      "execution_count": 52,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "import tensorflow as tf\n",
        "from tensorflow import keras\n",
        "\n",
        "model = keras.Sequential([\n",
        "    keras.layers.Conv2D(64, 7, activation='relu', padding='same', input_shape=(28, 28, 1)),\n",
        "\n",
        "    keras.layers.Conv2D(64, 3, padding='same'),\n",
        "    keras.layers.Dropout(0.2),\n",
        "    keras.layers.BatchNormalization(),\n",
        "    keras.layers.Activation('relu'),\n",
        "    keras.layers.MaxPooling2D(),\n",
        "\n",
        "    keras.layers.Conv2D(128, 3, padding='same'),\n",
        "    keras.layers.Dropout(0.2),\n",
        "    keras.layers.BatchNormalization(),\n",
        "    keras.layers.Activation('relu'),\n",
        "    keras.layers.MaxPooling2D(),\n",
        "\n",
        "    keras.layers.Conv2D(256, 3, padding='same'),\n",
        "    keras.layers.Dropout(0.2),\n",
        "    keras.layers.BatchNormalization(),\n",
        "    keras.layers.Activation('relu'),\n",
        "    keras.layers.MaxPooling2D(),\n",
        "\n",
        "    keras.layers.Conv2D(512, 3, padding='same'),\n",
        "    keras.layers.Dropout(0.2),\n",
        "    keras.layers.BatchNormalization(),\n",
        "    keras.layers.Activation('relu'),\n",
        "    keras.layers.MaxPooling2D(),\n",
        "\n",
        "    keras.layers.Flatten(),\n",
        "    keras.layers.Dense(1024, activation='relu'),\n",
        "    keras.layers.Dense(10, activation='softmax')\n",
        "])\n",
        "\n",
        "model.compile(loss=keras.losses.SparseCategoricalCrossentropy(), optimizer=tf.keras.optimizers.Lion(), metrics=['accuracy'])\n",
        "\n",
        "model.summary()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "uy-gZpVSJd7s",
        "outputId": "b55c8370-201f-42df-c039-5452613c729b"
      },
      "execution_count": 53,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential_5\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " conv2d_18 (Conv2D)          (None, 28, 28, 64)        3200      \n",
            "                                                                 \n",
            " conv2d_19 (Conv2D)          (None, 28, 28, 64)        36928     \n",
            "                                                                 \n",
            " dropout_8 (Dropout)         (None, 28, 28, 64)        0         \n",
            "                                                                 \n",
            " batch_normalization_16 (Ba  (None, 28, 28, 64)        256       \n",
            " tchNormalization)                                               \n",
            "                                                                 \n",
            " activation_8 (Activation)   (None, 28, 28, 64)        0         \n",
            "                                                                 \n",
            " max_pooling2d_8 (MaxPoolin  (None, 14, 14, 64)        0         \n",
            " g2D)                                                            \n",
            "                                                                 \n",
            " conv2d_20 (Conv2D)          (None, 14, 14, 128)       73856     \n",
            "                                                                 \n",
            " dropout_9 (Dropout)         (None, 14, 14, 128)       0         \n",
            "                                                                 \n",
            " batch_normalization_17 (Ba  (None, 14, 14, 128)       512       \n",
            " tchNormalization)                                               \n",
            "                                                                 \n",
            " activation_9 (Activation)   (None, 14, 14, 128)       0         \n",
            "                                                                 \n",
            " max_pooling2d_9 (MaxPoolin  (None, 7, 7, 128)         0         \n",
            " g2D)                                                            \n",
            "                                                                 \n",
            " conv2d_21 (Conv2D)          (None, 7, 7, 256)         295168    \n",
            "                                                                 \n",
            " dropout_10 (Dropout)        (None, 7, 7, 256)         0         \n",
            "                                                                 \n",
            " batch_normalization_18 (Ba  (None, 7, 7, 256)         1024      \n",
            " tchNormalization)                                               \n",
            "                                                                 \n",
            " activation_10 (Activation)  (None, 7, 7, 256)         0         \n",
            "                                                                 \n",
            " max_pooling2d_10 (MaxPooli  (None, 3, 3, 256)         0         \n",
            " ng2D)                                                           \n",
            "                                                                 \n",
            " conv2d_22 (Conv2D)          (None, 3, 3, 512)         1180160   \n",
            "                                                                 \n",
            " dropout_11 (Dropout)        (None, 3, 3, 512)         0         \n",
            "                                                                 \n",
            " batch_normalization_19 (Ba  (None, 3, 3, 512)         2048      \n",
            " tchNormalization)                                               \n",
            "                                                                 \n",
            " activation_11 (Activation)  (None, 3, 3, 512)         0         \n",
            "                                                                 \n",
            " max_pooling2d_11 (MaxPooli  (None, 1, 1, 512)         0         \n",
            " ng2D)                                                           \n",
            "                                                                 \n",
            " flatten_2 (Flatten)         (None, 512)               0         \n",
            "                                                                 \n",
            " dense_6 (Dense)             (None, 1024)              525312    \n",
            "                                                                 \n",
            " dense_7 (Dense)             (None, 10)                10250     \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 2128714 (8.12 MB)\n",
            "Trainable params: 2126794 (8.11 MB)\n",
            "Non-trainable params: 1920 (7.50 KB)\n",
            "_________________________________________________________________\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "#history = model.fit(X_train_augumented, y_train, epochs=10, validation_data=(X_valid, y_valid))"
      ],
      "metadata": {
        "id": "Vht38qKzNtE5"
      },
      "execution_count": 54,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "model.evaluate(X_test, y_test)"
      ],
      "metadata": {
        "id": "89tPVaNSWSaW",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "8aec4b44-b7a0-4b3d-a874-086876ac5180"
      },
      "execution_count": 55,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "313/313 [==============================] - 2s 5ms/step - loss: 2.3034 - accuracy: 0.1107\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[2.3034462928771973, 0.11069999635219574]"
            ]
          },
          "metadata": {},
          "execution_count": 55
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Exercise 10"
      ],
      "metadata": {
        "id": "fkNLM32VGX6Y"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "(x_train_full, y_train_full), (x_test, y_test) = keras.datasets.cifar100.load_data()\n",
        "\n",
        "x_train, x_valid = x_train_full[:-5000], x_train_full[-5000:]\n",
        "y_train, y_valid = y_train_full[:-5000], y_train_full[-5000:]"
      ],
      "metadata": {
        "id": "fkNLM32VGX6X"
      },
      "execution_count": 56,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "x_train.shape"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "rfmAydNUR9D7",
        "outputId": "0882e97c-9673-44ff-d482-5a70878ee66b"
      },
      "execution_count": 57,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(45000, 32, 32, 3)"
            ]
          },
          "metadata": {},
          "execution_count": 57
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import matplotlib.pyplot as plt\n",
        "\n",
        "image = x_train[1020]\n",
        "\n",
        "plt.imshow(image)\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 430
        },
        "id": "jY1JFbEBRAsy",
        "outputId": "97d0d3aa-f1ac-4453-f7a7-dc6a6d46a2ad"
      },
      "execution_count": 58,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "y_train[:10]"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "gGG2hia3SJ2R",
        "outputId": "d4b8d7a4-cd11-4235-b023-713dc523e3ba"
      },
      "execution_count": 59,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([[19],\n",
              "       [29],\n",
              "       [ 0],\n",
              "       [11],\n",
              "       [ 1],\n",
              "       [86],\n",
              "       [90],\n",
              "       [28],\n",
              "       [23],\n",
              "       [31]])"
            ]
          },
          "metadata": {},
          "execution_count": 59
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "def preprocess(image):\n",
        "  image = tf.image.random_flip_left_right(image)\n",
        "  image = tf.image.random_flip_up_down(image)\n",
        "  resized_image = tf.image.resize(image, [224, 224])\n",
        "  final_image = keras.applications.xception.preprocess_input(resized_image)\n",
        "  return final_image\n",
        "\n",
        "train_dataset = tf.data.Dataset.from_tensor_slices((x_train, y_train)).shuffle(buffer_size=10000).map(lambda x, y: (preprocess(x), y)).batch(batch_size=32).prefetch(buffer_size=tf.data.experimental.AUTOTUNE)\n",
        "valid_dataset = tf.data.Dataset.from_tensor_slices((x_valid, y_valid)).map(lambda x, y: (preprocess(x), y)).batch(batch_size=32).prefetch(buffer_size=tf.data.experimental.AUTOTUNE)\n",
        "test_dataset = tf.data.Dataset.from_tensor_slices((x_test, y_test))\n",
        "test_dataset = test_dataset.map(lambda x, y: (preprocess(x), y))\n",
        "test_dataset = test_dataset.batch(batch_size=32)\n",
        "test_dataset = test_dataset.prefetch(buffer_size=tf.data.experimental.AUTOTUNE)"
      ],
      "metadata": {
        "id": "0TNDnPTWSbIf"
      },
      "execution_count": 60,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "train_dataset"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "RNQXstBMTrH-",
        "outputId": "f9bb50d6-424e-4626-d341-0ddfad41a9cc"
      },
      "execution_count": 61,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<_PrefetchDataset element_spec=(TensorSpec(shape=(None, 224, 224, 3), dtype=tf.float32, name=None), TensorSpec(shape=(None, 1), dtype=tf.int64, name=None))>"
            ]
          },
          "metadata": {},
          "execution_count": 61
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "base_model = keras.applications.xception.Xception(weights='imagenet', include_top=False)\n",
        "avg = keras.layers.GlobalAveragePooling2D()(base_model.output)\n",
        "output = keras.layers.Dense(100, activation='softmax')(avg)\n",
        "\n",
        "model = keras.Model(inputs=base_model.input, outputs=output)\n",
        "\n",
        "for layer in base_model.layers:\n",
        "  layer.trainable = False\n",
        "\n",
        "model.summary()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "yaJeM44dTzQy",
        "outputId": "4a69dd85-0916-4884-d51b-950b598e170b"
      },
      "execution_count": 62,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"model_2\"\n",
            "__________________________________________________________________________________________________\n",
            " Layer (type)                Output Shape                 Param #   Connected to                  \n",
            "==================================================================================================\n",
            " input_3 (InputLayer)        [(None, None, None, 3)]      0         []                            \n",
            "                                                                                                  \n",
            " block1_conv1 (Conv2D)       (None, None, None, 32)       864       ['input_3[0][0]']             \n",
            "                                                                                                  \n",
            " block1_conv1_bn (BatchNorm  (None, None, None, 32)       128       ['block1_conv1[0][0]']        \n",
            " alization)                                                                                       \n",
            "                                                                                                  \n",
            " block1_conv1_act (Activati  (None, None, None, 32)       0         ['block1_conv1_bn[0][0]']     \n",
            " on)                                                                                              \n",
            "                                                                                                  \n",
            " block1_conv2 (Conv2D)       (None, None, None, 64)       18432     ['block1_conv1_act[0][0]']    \n",
            "                                                                                                  \n",
            " block1_conv2_bn (BatchNorm  (None, None, None, 64)       256       ['block1_conv2[0][0]']        \n",
            " alization)                                                                                       \n",
            "                                                                                                  \n",
            " block1_conv2_act (Activati  (None, None, None, 64)       0         ['block1_conv2_bn[0][0]']     \n",
            " on)                                                                                              \n",
            "                                                                                                  \n",
            " block2_sepconv1 (Separable  (None, None, None, 128)      8768      ['block1_conv2_act[0][0]']    \n",
            " Conv2D)                                                                                          \n",
            "                                                                                                  \n",
            " block2_sepconv1_bn (BatchN  (None, None, None, 128)      512       ['block2_sepconv1[0][0]']     \n",
            " ormalization)                                                                                    \n",
            "                                                                                                  \n",
            " block2_sepconv2_act (Activ  (None, None, None, 128)      0         ['block2_sepconv1_bn[0][0]']  \n",
            " ation)                                                                                           \n",
            "                                                                                                  \n",
            " block2_sepconv2 (Separable  (None, None, None, 128)      17536     ['block2_sepconv2_act[0][0]'] \n",
            " Conv2D)                                                                                          \n",
            "                                                                                                  \n",
            " block2_sepconv2_bn (BatchN  (None, None, None, 128)      512       ['block2_sepconv2[0][0]']     \n",
            " ormalization)                                                                                    \n",
            "                                                                                                  \n",
            " conv2d_23 (Conv2D)          (None, None, None, 128)      8192      ['block1_conv2_act[0][0]']    \n",
            "                                                                                                  \n",
            " block2_pool (MaxPooling2D)  (None, None, None, 128)      0         ['block2_sepconv2_bn[0][0]']  \n",
            "                                                                                                  \n",
            " batch_normalization_20 (Ba  (None, None, None, 128)      512       ['conv2d_23[0][0]']           \n",
            " tchNormalization)                                                                                \n",
            "                                                                                                  \n",
            " add_24 (Add)                (None, None, None, 128)      0         ['block2_pool[0][0]',         \n",
            "                                                                     'batch_normalization_20[0][0]\n",
            "                                                                    ']                            \n",
            "                                                                                                  \n",
            " block3_sepconv1_act (Activ  (None, None, None, 128)      0         ['add_24[0][0]']              \n",
            " ation)                                                                                           \n",
            "                                                                                                  \n",
            " block3_sepconv1 (Separable  (None, None, None, 256)      33920     ['block3_sepconv1_act[0][0]'] \n",
            " Conv2D)                                                                                          \n",
            "                                                                                                  \n",
            " block3_sepconv1_bn (BatchN  (None, None, None, 256)      1024      ['block3_sepconv1[0][0]']     \n",
            " ormalization)                                                                                    \n",
            "                                                                                                  \n",
            " block3_sepconv2_act (Activ  (None, None, None, 256)      0         ['block3_sepconv1_bn[0][0]']  \n",
            " ation)                                                                                           \n",
            "                                                                                                  \n",
            " block3_sepconv2 (Separable  (None, None, None, 256)      67840     ['block3_sepconv2_act[0][0]'] \n",
            " Conv2D)                                                                                          \n",
            "                                                                                                  \n",
            " block3_sepconv2_bn (BatchN  (None, None, None, 256)      1024      ['block3_sepconv2[0][0]']     \n",
            " ormalization)                                                                                    \n",
            "                                                                                                  \n",
            " conv2d_24 (Conv2D)          (None, None, None, 256)      32768     ['add_24[0][0]']              \n",
            "                                                                                                  \n",
            " block3_pool (MaxPooling2D)  (None, None, None, 256)      0         ['block3_sepconv2_bn[0][0]']  \n",
            "                                                                                                  \n",
            " batch_normalization_21 (Ba  (None, None, None, 256)      1024      ['conv2d_24[0][0]']           \n",
            " tchNormalization)                                                                                \n",
            "                                                                                                  \n",
            " add_25 (Add)                (None, None, None, 256)      0         ['block3_pool[0][0]',         \n",
            "                                                                     'batch_normalization_21[0][0]\n",
            "                                                                    ']                            \n",
            "                                                                                                  \n",
            " block4_sepconv1_act (Activ  (None, None, None, 256)      0         ['add_25[0][0]']              \n",
            " ation)                                                                                           \n",
            "                                                                                                  \n",
            " block4_sepconv1 (Separable  (None, None, None, 728)      188672    ['block4_sepconv1_act[0][0]'] \n",
            " Conv2D)                                                                                          \n",
            "                                                                                                  \n",
            " block4_sepconv1_bn (BatchN  (None, None, None, 728)      2912      ['block4_sepconv1[0][0]']     \n",
            " ormalization)                                                                                    \n",
            "                                                                                                  \n",
            " block4_sepconv2_act (Activ  (None, None, None, 728)      0         ['block4_sepconv1_bn[0][0]']  \n",
            " ation)                                                                                           \n",
            "                                                                                                  \n",
            " block4_sepconv2 (Separable  (None, None, None, 728)      536536    ['block4_sepconv2_act[0][0]'] \n",
            " Conv2D)                                                                                          \n",
            "                                                                                                  \n",
            " block4_sepconv2_bn (BatchN  (None, None, None, 728)      2912      ['block4_sepconv2[0][0]']     \n",
            " ormalization)                                                                                    \n",
            "                                                                                                  \n",
            " conv2d_25 (Conv2D)          (None, None, None, 728)      186368    ['add_25[0][0]']              \n",
            "                                                                                                  \n",
            " block4_pool (MaxPooling2D)  (None, None, None, 728)      0         ['block4_sepconv2_bn[0][0]']  \n",
            "                                                                                                  \n",
            " batch_normalization_22 (Ba  (None, None, None, 728)      2912      ['conv2d_25[0][0]']           \n",
            " tchNormalization)                                                                                \n",
            "                                                                                                  \n",
            " add_26 (Add)                (None, None, None, 728)      0         ['block4_pool[0][0]',         \n",
            "                                                                     'batch_normalization_22[0][0]\n",
            "                                                                    ']                            \n",
            "                                                                                                  \n",
            " block5_sepconv1_act (Activ  (None, None, None, 728)      0         ['add_26[0][0]']              \n",
            " ation)                                                                                           \n",
            "                                                                                                  \n",
            " block5_sepconv1 (Separable  (None, None, None, 728)      536536    ['block5_sepconv1_act[0][0]'] \n",
            " Conv2D)                                                                                          \n",
            "                                                                                                  \n",
            " block5_sepconv1_bn (BatchN  (None, None, None, 728)      2912      ['block5_sepconv1[0][0]']     \n",
            " ormalization)                                                                                    \n",
            "                                                                                                  \n",
            " block5_sepconv2_act (Activ  (None, None, None, 728)      0         ['block5_sepconv1_bn[0][0]']  \n",
            " ation)                                                                                           \n",
            "                                                                                                  \n",
            " block5_sepconv2 (Separable  (None, None, None, 728)      536536    ['block5_sepconv2_act[0][0]'] \n",
            " Conv2D)                                                                                          \n",
            "                                                                                                  \n",
            " block5_sepconv2_bn (BatchN  (None, None, None, 728)      2912      ['block5_sepconv2[0][0]']     \n",
            " ormalization)                                                                                    \n",
            "                                                                                                  \n",
            " block5_sepconv3_act (Activ  (None, None, None, 728)      0         ['block5_sepconv2_bn[0][0]']  \n",
            " ation)                                                                                           \n",
            "                                                                                                  \n",
            " block5_sepconv3 (Separable  (None, None, None, 728)      536536    ['block5_sepconv3_act[0][0]'] \n",
            " Conv2D)                                                                                          \n",
            "                                                                                                  \n",
            " block5_sepconv3_bn (BatchN  (None, None, None, 728)      2912      ['block5_sepconv3[0][0]']     \n",
            " ormalization)                                                                                    \n",
            "                                                                                                  \n",
            " add_27 (Add)                (None, None, None, 728)      0         ['block5_sepconv3_bn[0][0]',  \n",
            "                                                                     'add_26[0][0]']              \n",
            "                                                                                                  \n",
            " block6_sepconv1_act (Activ  (None, None, None, 728)      0         ['add_27[0][0]']              \n",
            " ation)                                                                                           \n",
            "                                                                                                  \n",
            " block6_sepconv1 (Separable  (None, None, None, 728)      536536    ['block6_sepconv1_act[0][0]'] \n",
            " Conv2D)                                                                                          \n",
            "                                                                                                  \n",
            " block6_sepconv1_bn (BatchN  (None, None, None, 728)      2912      ['block6_sepconv1[0][0]']     \n",
            " ormalization)                                                                                    \n",
            "                                                                                                  \n",
            " block6_sepconv2_act (Activ  (None, None, None, 728)      0         ['block6_sepconv1_bn[0][0]']  \n",
            " ation)                                                                                           \n",
            "                                                                                                  \n",
            " block6_sepconv2 (Separable  (None, None, None, 728)      536536    ['block6_sepconv2_act[0][0]'] \n",
            " Conv2D)                                                                                          \n",
            "                                                                                                  \n",
            " block6_sepconv2_bn (BatchN  (None, None, None, 728)      2912      ['block6_sepconv2[0][0]']     \n",
            " ormalization)                                                                                    \n",
            "                                                                                                  \n",
            " block6_sepconv3_act (Activ  (None, None, None, 728)      0         ['block6_sepconv2_bn[0][0]']  \n",
            " ation)                                                                                           \n",
            "                                                                                                  \n",
            " block6_sepconv3 (Separable  (None, None, None, 728)      536536    ['block6_sepconv3_act[0][0]'] \n",
            " Conv2D)                                                                                          \n",
            "                                                                                                  \n",
            " block6_sepconv3_bn (BatchN  (None, None, None, 728)      2912      ['block6_sepconv3[0][0]']     \n",
            " ormalization)                                                                                    \n",
            "                                                                                                  \n",
            " add_28 (Add)                (None, None, None, 728)      0         ['block6_sepconv3_bn[0][0]',  \n",
            "                                                                     'add_27[0][0]']              \n",
            "                                                                                                  \n",
            " block7_sepconv1_act (Activ  (None, None, None, 728)      0         ['add_28[0][0]']              \n",
            " ation)                                                                                           \n",
            "                                                                                                  \n",
            " block7_sepconv1 (Separable  (None, None, None, 728)      536536    ['block7_sepconv1_act[0][0]'] \n",
            " Conv2D)                                                                                          \n",
            "                                                                                                  \n",
            " block7_sepconv1_bn (BatchN  (None, None, None, 728)      2912      ['block7_sepconv1[0][0]']     \n",
            " ormalization)                                                                                    \n",
            "                                                                                                  \n",
            " block7_sepconv2_act (Activ  (None, None, None, 728)      0         ['block7_sepconv1_bn[0][0]']  \n",
            " ation)                                                                                           \n",
            "                                                                                                  \n",
            " block7_sepconv2 (Separable  (None, None, None, 728)      536536    ['block7_sepconv2_act[0][0]'] \n",
            " Conv2D)                                                                                          \n",
            "                                                                                                  \n",
            " block7_sepconv2_bn (BatchN  (None, None, None, 728)      2912      ['block7_sepconv2[0][0]']     \n",
            " ormalization)                                                                                    \n",
            "                                                                                                  \n",
            " block7_sepconv3_act (Activ  (None, None, None, 728)      0         ['block7_sepconv2_bn[0][0]']  \n",
            " ation)                                                                                           \n",
            "                                                                                                  \n",
            " block7_sepconv3 (Separable  (None, None, None, 728)      536536    ['block7_sepconv3_act[0][0]'] \n",
            " Conv2D)                                                                                          \n",
            "                                                                                                  \n",
            " block7_sepconv3_bn (BatchN  (None, None, None, 728)      2912      ['block7_sepconv3[0][0]']     \n",
            " ormalization)                                                                                    \n",
            "                                                                                                  \n",
            " add_29 (Add)                (None, None, None, 728)      0         ['block7_sepconv3_bn[0][0]',  \n",
            "                                                                     'add_28[0][0]']              \n",
            "                                                                                                  \n",
            " block8_sepconv1_act (Activ  (None, None, None, 728)      0         ['add_29[0][0]']              \n",
            " ation)                                                                                           \n",
            "                                                                                                  \n",
            " block8_sepconv1 (Separable  (None, None, None, 728)      536536    ['block8_sepconv1_act[0][0]'] \n",
            " Conv2D)                                                                                          \n",
            "                                                                                                  \n",
            " block8_sepconv1_bn (BatchN  (None, None, None, 728)      2912      ['block8_sepconv1[0][0]']     \n",
            " ormalization)                                                                                    \n",
            "                                                                                                  \n",
            " block8_sepconv2_act (Activ  (None, None, None, 728)      0         ['block8_sepconv1_bn[0][0]']  \n",
            " ation)                                                                                           \n",
            "                                                                                                  \n",
            " block8_sepconv2 (Separable  (None, None, None, 728)      536536    ['block8_sepconv2_act[0][0]'] \n",
            " Conv2D)                                                                                          \n",
            "                                                                                                  \n",
            " block8_sepconv2_bn (BatchN  (None, None, None, 728)      2912      ['block8_sepconv2[0][0]']     \n",
            " ormalization)                                                                                    \n",
            "                                                                                                  \n",
            " block8_sepconv3_act (Activ  (None, None, None, 728)      0         ['block8_sepconv2_bn[0][0]']  \n",
            " ation)                                                                                           \n",
            "                                                                                                  \n",
            " block8_sepconv3 (Separable  (None, None, None, 728)      536536    ['block8_sepconv3_act[0][0]'] \n",
            " Conv2D)                                                                                          \n",
            "                                                                                                  \n",
            " block8_sepconv3_bn (BatchN  (None, None, None, 728)      2912      ['block8_sepconv3[0][0]']     \n",
            " ormalization)                                                                                    \n",
            "                                                                                                  \n",
            " add_30 (Add)                (None, None, None, 728)      0         ['block8_sepconv3_bn[0][0]',  \n",
            "                                                                     'add_29[0][0]']              \n",
            "                                                                                                  \n",
            " block9_sepconv1_act (Activ  (None, None, None, 728)      0         ['add_30[0][0]']              \n",
            " ation)                                                                                           \n",
            "                                                                                                  \n",
            " block9_sepconv1 (Separable  (None, None, None, 728)      536536    ['block9_sepconv1_act[0][0]'] \n",
            " Conv2D)                                                                                          \n",
            "                                                                                                  \n",
            " block9_sepconv1_bn (BatchN  (None, None, None, 728)      2912      ['block9_sepconv1[0][0]']     \n",
            " ormalization)                                                                                    \n",
            "                                                                                                  \n",
            " block9_sepconv2_act (Activ  (None, None, None, 728)      0         ['block9_sepconv1_bn[0][0]']  \n",
            " ation)                                                                                           \n",
            "                                                                                                  \n",
            " block9_sepconv2 (Separable  (None, None, None, 728)      536536    ['block9_sepconv2_act[0][0]'] \n",
            " Conv2D)                                                                                          \n",
            "                                                                                                  \n",
            " block9_sepconv2_bn (BatchN  (None, None, None, 728)      2912      ['block9_sepconv2[0][0]']     \n",
            " ormalization)                                                                                    \n",
            "                                                                                                  \n",
            " block9_sepconv3_act (Activ  (None, None, None, 728)      0         ['block9_sepconv2_bn[0][0]']  \n",
            " ation)                                                                                           \n",
            "                                                                                                  \n",
            " block9_sepconv3 (Separable  (None, None, None, 728)      536536    ['block9_sepconv3_act[0][0]'] \n",
            " Conv2D)                                                                                          \n",
            "                                                                                                  \n",
            " block9_sepconv3_bn (BatchN  (None, None, None, 728)      2912      ['block9_sepconv3[0][0]']     \n",
            " ormalization)                                                                                    \n",
            "                                                                                                  \n",
            " add_31 (Add)                (None, None, None, 728)      0         ['block9_sepconv3_bn[0][0]',  \n",
            "                                                                     'add_30[0][0]']              \n",
            "                                                                                                  \n",
            " block10_sepconv1_act (Acti  (None, None, None, 728)      0         ['add_31[0][0]']              \n",
            " vation)                                                                                          \n",
            "                                                                                                  \n",
            " block10_sepconv1 (Separabl  (None, None, None, 728)      536536    ['block10_sepconv1_act[0][0]']\n",
            " eConv2D)                                                                                         \n",
            "                                                                                                  \n",
            " block10_sepconv1_bn (Batch  (None, None, None, 728)      2912      ['block10_sepconv1[0][0]']    \n",
            " Normalization)                                                                                   \n",
            "                                                                                                  \n",
            " block10_sepconv2_act (Acti  (None, None, None, 728)      0         ['block10_sepconv1_bn[0][0]'] \n",
            " vation)                                                                                          \n",
            "                                                                                                  \n",
            " block10_sepconv2 (Separabl  (None, None, None, 728)      536536    ['block10_sepconv2_act[0][0]']\n",
            " eConv2D)                                                                                         \n",
            "                                                                                                  \n",
            " block10_sepconv2_bn (Batch  (None, None, None, 728)      2912      ['block10_sepconv2[0][0]']    \n",
            " Normalization)                                                                                   \n",
            "                                                                                                  \n",
            " block10_sepconv3_act (Acti  (None, None, None, 728)      0         ['block10_sepconv2_bn[0][0]'] \n",
            " vation)                                                                                          \n",
            "                                                                                                  \n",
            " block10_sepconv3 (Separabl  (None, None, None, 728)      536536    ['block10_sepconv3_act[0][0]']\n",
            " eConv2D)                                                                                         \n",
            "                                                                                                  \n",
            " block10_sepconv3_bn (Batch  (None, None, None, 728)      2912      ['block10_sepconv3[0][0]']    \n",
            " Normalization)                                                                                   \n",
            "                                                                                                  \n",
            " add_32 (Add)                (None, None, None, 728)      0         ['block10_sepconv3_bn[0][0]', \n",
            "                                                                     'add_31[0][0]']              \n",
            "                                                                                                  \n",
            " block11_sepconv1_act (Acti  (None, None, None, 728)      0         ['add_32[0][0]']              \n",
            " vation)                                                                                          \n",
            "                                                                                                  \n",
            " block11_sepconv1 (Separabl  (None, None, None, 728)      536536    ['block11_sepconv1_act[0][0]']\n",
            " eConv2D)                                                                                         \n",
            "                                                                                                  \n",
            " block11_sepconv1_bn (Batch  (None, None, None, 728)      2912      ['block11_sepconv1[0][0]']    \n",
            " Normalization)                                                                                   \n",
            "                                                                                                  \n",
            " block11_sepconv2_act (Acti  (None, None, None, 728)      0         ['block11_sepconv1_bn[0][0]'] \n",
            " vation)                                                                                          \n",
            "                                                                                                  \n",
            " block11_sepconv2 (Separabl  (None, None, None, 728)      536536    ['block11_sepconv2_act[0][0]']\n",
            " eConv2D)                                                                                         \n",
            "                                                                                                  \n",
            " block11_sepconv2_bn (Batch  (None, None, None, 728)      2912      ['block11_sepconv2[0][0]']    \n",
            " Normalization)                                                                                   \n",
            "                                                                                                  \n",
            " block11_sepconv3_act (Acti  (None, None, None, 728)      0         ['block11_sepconv2_bn[0][0]'] \n",
            " vation)                                                                                          \n",
            "                                                                                                  \n",
            " block11_sepconv3 (Separabl  (None, None, None, 728)      536536    ['block11_sepconv3_act[0][0]']\n",
            " eConv2D)                                                                                         \n",
            "                                                                                                  \n",
            " block11_sepconv3_bn (Batch  (None, None, None, 728)      2912      ['block11_sepconv3[0][0]']    \n",
            " Normalization)                                                                                   \n",
            "                                                                                                  \n",
            " add_33 (Add)                (None, None, None, 728)      0         ['block11_sepconv3_bn[0][0]', \n",
            "                                                                     'add_32[0][0]']              \n",
            "                                                                                                  \n",
            " block12_sepconv1_act (Acti  (None, None, None, 728)      0         ['add_33[0][0]']              \n",
            " vation)                                                                                          \n",
            "                                                                                                  \n",
            " block12_sepconv1 (Separabl  (None, None, None, 728)      536536    ['block12_sepconv1_act[0][0]']\n",
            " eConv2D)                                                                                         \n",
            "                                                                                                  \n",
            " block12_sepconv1_bn (Batch  (None, None, None, 728)      2912      ['block12_sepconv1[0][0]']    \n",
            " Normalization)                                                                                   \n",
            "                                                                                                  \n",
            " block12_sepconv2_act (Acti  (None, None, None, 728)      0         ['block12_sepconv1_bn[0][0]'] \n",
            " vation)                                                                                          \n",
            "                                                                                                  \n",
            " block12_sepconv2 (Separabl  (None, None, None, 728)      536536    ['block12_sepconv2_act[0][0]']\n",
            " eConv2D)                                                                                         \n",
            "                                                                                                  \n",
            " block12_sepconv2_bn (Batch  (None, None, None, 728)      2912      ['block12_sepconv2[0][0]']    \n",
            " Normalization)                                                                                   \n",
            "                                                                                                  \n",
            " block12_sepconv3_act (Acti  (None, None, None, 728)      0         ['block12_sepconv2_bn[0][0]'] \n",
            " vation)                                                                                          \n",
            "                                                                                                  \n",
            " block12_sepconv3 (Separabl  (None, None, None, 728)      536536    ['block12_sepconv3_act[0][0]']\n",
            " eConv2D)                                                                                         \n",
            "                                                                                                  \n",
            " block12_sepconv3_bn (Batch  (None, None, None, 728)      2912      ['block12_sepconv3[0][0]']    \n",
            " Normalization)                                                                                   \n",
            "                                                                                                  \n",
            " add_34 (Add)                (None, None, None, 728)      0         ['block12_sepconv3_bn[0][0]', \n",
            "                                                                     'add_33[0][0]']              \n",
            "                                                                                                  \n",
            " block13_sepconv1_act (Acti  (None, None, None, 728)      0         ['add_34[0][0]']              \n",
            " vation)                                                                                          \n",
            "                                                                                                  \n",
            " block13_sepconv1 (Separabl  (None, None, None, 728)      536536    ['block13_sepconv1_act[0][0]']\n",
            " eConv2D)                                                                                         \n",
            "                                                                                                  \n",
            " block13_sepconv1_bn (Batch  (None, None, None, 728)      2912      ['block13_sepconv1[0][0]']    \n",
            " Normalization)                                                                                   \n",
            "                                                                                                  \n",
            " block13_sepconv2_act (Acti  (None, None, None, 728)      0         ['block13_sepconv1_bn[0][0]'] \n",
            " vation)                                                                                          \n",
            "                                                                                                  \n",
            " block13_sepconv2 (Separabl  (None, None, None, 1024)     752024    ['block13_sepconv2_act[0][0]']\n",
            " eConv2D)                                                                                         \n",
            "                                                                                                  \n",
            " block13_sepconv2_bn (Batch  (None, None, None, 1024)     4096      ['block13_sepconv2[0][0]']    \n",
            " Normalization)                                                                                   \n",
            "                                                                                                  \n",
            " conv2d_26 (Conv2D)          (None, None, None, 1024)     745472    ['add_34[0][0]']              \n",
            "                                                                                                  \n",
            " block13_pool (MaxPooling2D  (None, None, None, 1024)     0         ['block13_sepconv2_bn[0][0]'] \n",
            " )                                                                                                \n",
            "                                                                                                  \n",
            " batch_normalization_23 (Ba  (None, None, None, 1024)     4096      ['conv2d_26[0][0]']           \n",
            " tchNormalization)                                                                                \n",
            "                                                                                                  \n",
            " add_35 (Add)                (None, None, None, 1024)     0         ['block13_pool[0][0]',        \n",
            "                                                                     'batch_normalization_23[0][0]\n",
            "                                                                    ']                            \n",
            "                                                                                                  \n",
            " block14_sepconv1 (Separabl  (None, None, None, 1536)     1582080   ['add_35[0][0]']              \n",
            " eConv2D)                                                                                         \n",
            "                                                                                                  \n",
            " block14_sepconv1_bn (Batch  (None, None, None, 1536)     6144      ['block14_sepconv1[0][0]']    \n",
            " Normalization)                                                                                   \n",
            "                                                                                                  \n",
            " block14_sepconv1_act (Acti  (None, None, None, 1536)     0         ['block14_sepconv1_bn[0][0]'] \n",
            " vation)                                                                                          \n",
            "                                                                                                  \n",
            " block14_sepconv2 (Separabl  (None, None, None, 2048)     3159552   ['block14_sepconv1_act[0][0]']\n",
            " eConv2D)                                                                                         \n",
            "                                                                                                  \n",
            " block14_sepconv2_bn (Batch  (None, None, None, 2048)     8192      ['block14_sepconv2[0][0]']    \n",
            " Normalization)                                                                                   \n",
            "                                                                                                  \n",
            " block14_sepconv2_act (Acti  (None, None, None, 2048)     0         ['block14_sepconv2_bn[0][0]'] \n",
            " vation)                                                                                          \n",
            "                                                                                                  \n",
            " global_average_pooling2d_2  (None, 2048)                 0         ['block14_sepconv2_act[0][0]']\n",
            "  (GlobalAveragePooling2D)                                                                        \n",
            "                                                                                                  \n",
            " dense_8 (Dense)             (None, 100)                  204900    ['global_average_pooling2d_2[0\n",
            "                                                                    ][0]']                        \n",
            "                                                                                                  \n",
            "==================================================================================================\n",
            "Total params: 21066380 (80.36 MB)\n",
            "Trainable params: 204900 (800.39 KB)\n",
            "Non-trainable params: 20861480 (79.58 MB)\n",
            "__________________________________________________________________________________________________\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "optimizer = keras.optimizers.SGD(lr=0.2, momentum=0.9)\n",
        "model.compile(loss='sparse_categorical_crossentropy', optimizer=optimizer, metrics=['accuracy'])\n",
        "history = model.fit(train_dataset, epochs=5, validation_data=valid_dataset)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "h6XVEtjgU1Ir",
        "outputId": "7b63d011-4a27-4264-e11c-3e0975c6fb95"
      },
      "execution_count": 64,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "WARNING:absl:`lr` is deprecated in Keras optimizer, please use `learning_rate` or use the legacy optimizer, e.g.,tf.keras.optimizers.legacy.SGD.\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/5\n",
            "1407/1407 [==============================] - 223s 155ms/step - loss: 2.1791 - accuracy: 0.4611 - val_loss: 1.8167 - val_accuracy: 0.5156\n",
            "Epoch 2/5\n",
            "1407/1407 [==============================] - 217s 154ms/step - loss: 1.6304 - accuracy: 0.5616 - val_loss: 1.6832 - val_accuracy: 0.5474\n",
            "Epoch 3/5\n",
            "1407/1407 [==============================] - 217s 154ms/step - loss: 1.5014 - accuracy: 0.5889 - val_loss: 1.6594 - val_accuracy: 0.5530\n",
            "Epoch 4/5\n",
            "1407/1407 [==============================] - 236s 168ms/step - loss: 1.4200 - accuracy: 0.6072 - val_loss: 1.6135 - val_accuracy: 0.5568\n",
            "Epoch 5/5\n",
            "1407/1407 [==============================] - 217s 154ms/step - loss: 1.3590 - accuracy: 0.6250 - val_loss: 1.5955 - val_accuracy: 0.5660\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "for layer in base_model.layers:\n",
        "  layer.trainable=True\n",
        "\n",
        "optimizer = keras.optimizers.SGD(learning_rate=0.01, momentum=0.9)\n",
        "model.compile(loss='sparse_categorical_crossentropy', optimizer=optimizer, metrics=['accuracy'])\n",
        "history = model.fit(train_dataset, epochs=5, validation_data=valid_dataset)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "IArroZuKb0UN",
        "outputId": "9770f9a9-b5dd-435f-80d0-53a78d1a96a7"
      },
      "execution_count": 65,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/5\n",
            "1407/1407 [==============================] - 754s 526ms/step - loss: 1.1396 - accuracy: 0.6703 - val_loss: 1.0368 - val_accuracy: 0.6944\n",
            "Epoch 2/5\n",
            "1407/1407 [==============================] - 738s 524ms/step - loss: 0.6934 - accuracy: 0.7873 - val_loss: 0.9202 - val_accuracy: 0.7326\n",
            "Epoch 3/5\n",
            "1407/1407 [==============================] - 718s 510ms/step - loss: 0.4901 - accuracy: 0.8462 - val_loss: 0.8870 - val_accuracy: 0.7464\n",
            "Epoch 4/5\n",
            "1407/1407 [==============================] - 717s 509ms/step - loss: 0.3666 - accuracy: 0.8846 - val_loss: 0.8909 - val_accuracy: 0.7570\n",
            "Epoch 5/5\n",
            "1407/1407 [==============================] - 718s 510ms/step - loss: 0.2756 - accuracy: 0.9116 - val_loss: 0.7630 - val_accuracy: 0.7868\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "zU_7XkwqqRys"
      },
      "execution_count": null,
      "outputs": []
    }
  ]
}