chapter12_part05_gans.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "nVDUXepFk_5a"
      },
      "source": [
        "이 노트북은 [케라스 창시자에게 배우는 딥러닝 2판](https://tensorflow.blog/kerasdl2/)의 예제 코드를 담고 있습니다.\n",
        "\n",
        "<table align=\"left\">\n",
        "    <tr>\n",
        "        <td>\n",
        "            <a href=\"https://colab.research.google.com/github/rickiepark/deep-learning-with-python-2nd/blob/main/chapter12_part05_gans.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>\n",
        "        </td>\n",
        "    </tr>\n",
        "</table>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "IGNOeppck_5c"
      },
      "source": [
        "## 생성적 적대 신경망 소개"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ECLt0piUk_5d"
      },
      "source": [
        "### GAN 구현 방법"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "lKmqFiRVk_5d"
      },
      "source": [
        "### 훈련 방법"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-mZbiQIOk_5d"
      },
      "source": [
        "### CelebA 데이터셋 준비하기"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "BQhBgdvhk_5d"
      },
      "source": [
        "**CelebA 데이터 다운로드**"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "pJ6sKUZfk_5e"
      },
      "source": [
        "로컬에서 실행하는 경우 `pip install gdown` 명령으로 gdown 패키지를 먼저 설치해 주세요."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "id": "JGyhu29Zk_5e",
        "outputId": "2acb2b44-8484-437c-cd9b-d1539f1245ba",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "--2022-11-30 11:45:35--  https://drive.google.com/uc?id=1up5bN8LCE2vHigVY-Z9yY2_aKRW5jN_9&confirm=t\n",
            "Resolving drive.google.com (drive.google.com)... 209.85.234.101, 209.85.234.139, 209.85.234.102, ...\n",
            "Connecting to drive.google.com (drive.google.com)|209.85.234.101|:443... connected.\n",
            "HTTP request sent, awaiting response... 303 See Other\n",
            "Location: https://doc-04-a4-docs.googleusercontent.com/docs/securesc/ha0ro937gcuc7l7deffksulhg5h7mbp1/k6t2393meh4h4dgsjbuppr11jdog03sc/1669808700000/10613826186492816817/*/1up5bN8LCE2vHigVY-Z9yY2_aKRW5jN_9?uuid=71432cf5-11e0-496b-8fbc-d5b47f4eaf8d [following]\n",
            "Warning: wildcards not supported in HTTP.\n",
            "--2022-11-30 11:45:35--  https://doc-04-a4-docs.googleusercontent.com/docs/securesc/ha0ro937gcuc7l7deffksulhg5h7mbp1/k6t2393meh4h4dgsjbuppr11jdog03sc/1669808700000/10613826186492816817/*/1up5bN8LCE2vHigVY-Z9yY2_aKRW5jN_9?uuid=71432cf5-11e0-496b-8fbc-d5b47f4eaf8d\n",
            "Resolving doc-04-a4-docs.googleusercontent.com (doc-04-a4-docs.googleusercontent.com)... 74.125.126.132, 2607:f8b0:4001:c1d::84\n",
            "Connecting to doc-04-a4-docs.googleusercontent.com (doc-04-a4-docs.googleusercontent.com)|74.125.126.132|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 1443490838 (1.3G) [application/zip]\n",
            "Saving to: ‘celeba_gan/data.zip’\n",
            "\n",
            "celeba_gan/data.zip 100%[===================>]   1.34G   104MB/s    in 15s     \n",
            "\n",
            "2022-11-30 11:45:50 (93.2 MB/s) - ‘celeba_gan/data.zip’ saved [1443490838/1443490838]\n",
            "\n"
          ]
        }
      ],
      "source": [
        "!mkdir celeba_gan\n",
        "!wget \"https://drive.google.com/uc?id=1up5bN8LCE2vHigVY-Z9yY2_aKRW5jN_9&confirm=t\" -O celeba_gan/data.zip\n",
        "!unzip -qq celeba_gan/data.zip -d celeba_gan"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "tags": [],
        "id": "pxa9mE2wk_5f"
      },
      "source": [
        "**이미지 디렉토리에서 데이터셋을 만듭니다**"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 2,
      "metadata": {
        "id": "SZUmI5BZk_5g",
        "outputId": "d04ccab1-e362-4423-e0ad-36f99d097209",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Found 202599 files belonging to 1 classes.\n"
          ]
        }
      ],
      "source": [
        "from tensorflow import keras\n",
        "dataset = keras.utils.image_dataset_from_directory(\n",
        "    \"celeba_gan\",\n",
        "    label_mode=None,\n",
        "    image_size=(64, 64),\n",
        "    batch_size=32,\n",
        "    smart_resize=True)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "FiLo3_pNk_5g"
      },
      "source": [
        "**픽셀 값 범위 바꾸기**"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 3,
      "metadata": {
        "id": "HbsrlNzNk_5g"
      },
      "outputs": [],
      "source": [
        "dataset = dataset.map(lambda x: x / 255.)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "gQYfVuAsk_5g"
      },
      "source": [
        "**첫 번째 이미지 출력하기**"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 4,
      "metadata": {
        "id": "T1QqXmV6k_5h",
        "outputId": "dc53ec02-07d6-44f6-dab1-13133ffc6a58",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 248
        }
      },
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ],
      "source": [
        "import matplotlib.pyplot as plt\n",
        "for x in dataset:\n",
        "    plt.axis(\"off\")\n",
        "    plt.imshow((x.numpy() * 255).astype(\"int32\")[0])\n",
        "    break"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "HcOE7K-7k_5h"
      },
      "source": [
        "### 판별자"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "pq5ugYZOk_5h"
      },
      "source": [
        "**GAN 판별자 네트워크**"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 5,
      "metadata": {
        "id": "dZNhbeHjk_5h"
      },
      "outputs": [],
      "source": [
        "from tensorflow.keras import layers\n",
        "\n",
        "discriminator = keras.Sequential(\n",
        "    [\n",
        "        keras.Input(shape=(64, 64, 3)),\n",
        "        layers.Conv2D(64, kernel_size=4, strides=2, padding=\"same\"),\n",
        "        layers.LeakyReLU(alpha=0.2),\n",
        "        layers.Conv2D(128, kernel_size=4, strides=2, padding=\"same\"),\n",
        "        layers.LeakyReLU(alpha=0.2),\n",
        "        layers.Conv2D(128, kernel_size=4, strides=2, padding=\"same\"),\n",
        "        layers.LeakyReLU(alpha=0.2),\n",
        "        layers.Flatten(),\n",
        "        layers.Dropout(0.2),\n",
        "        layers.Dense(1, activation=\"sigmoid\"),\n",
        "    ],\n",
        "    name=\"discriminator\",\n",
        ")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 6,
      "metadata": {
        "id": "_lqkry-Lk_5i",
        "outputId": "ba7e434a-d06a-49ca-e04d-913d506197ab",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"discriminator\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " conv2d (Conv2D)             (None, 32, 32, 64)        3136      \n",
            "                                                                 \n",
            " leaky_re_lu (LeakyReLU)     (None, 32, 32, 64)        0         \n",
            "                                                                 \n",
            " conv2d_1 (Conv2D)           (None, 16, 16, 128)       131200    \n",
            "                                                                 \n",
            " leaky_re_lu_1 (LeakyReLU)   (None, 16, 16, 128)       0         \n",
            "                                                                 \n",
            " conv2d_2 (Conv2D)           (None, 8, 8, 128)         262272    \n",
            "                                                                 \n",
            " leaky_re_lu_2 (LeakyReLU)   (None, 8, 8, 128)         0         \n",
            "                                                                 \n",
            " flatten (Flatten)           (None, 8192)              0         \n",
            "                                                                 \n",
            " dropout (Dropout)           (None, 8192)              0         \n",
            "                                                                 \n",
            " dense (Dense)               (None, 1)                 8193      \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 404,801\n",
            "Trainable params: 404,801\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ]
        }
      ],
      "source": [
        "discriminator.summary()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "w2pW40o6k_5i"
      },
      "source": [
        "### 생성자"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "TRQAI509k_5i"
      },
      "source": [
        "**GAN 생성자 네트워크**"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 7,
      "metadata": {
        "id": "wn4U502Nk_5i"
      },
      "outputs": [],
      "source": [
        "latent_dim = 128\n",
        "\n",
        "generator = keras.Sequential(\n",
        "    [\n",
        "        keras.Input(shape=(latent_dim,)),\n",
        "        layers.Dense(8 * 8 * 128),\n",
        "        layers.Reshape((8, 8, 128)),\n",
        "        layers.Conv2DTranspose(128, kernel_size=4, strides=2, padding=\"same\"),\n",
        "        layers.LeakyReLU(alpha=0.2),\n",
        "        layers.Conv2DTranspose(256, kernel_size=4, strides=2, padding=\"same\"),\n",
        "        layers.LeakyReLU(alpha=0.2),\n",
        "        layers.Conv2DTranspose(512, kernel_size=4, strides=2, padding=\"same\"),\n",
        "        layers.LeakyReLU(alpha=0.2),\n",
        "        layers.Conv2D(3, kernel_size=5, padding=\"same\", activation=\"sigmoid\"),\n",
        "    ],\n",
        "    name=\"generator\",\n",
        ")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 8,
      "metadata": {
        "id": "cr63oFSTk_5j",
        "outputId": "7d7c4000-f108-4470-ee49-5a0b515e9316",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"generator\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " dense_1 (Dense)             (None, 8192)              1056768   \n",
            "                                                                 \n",
            " reshape (Reshape)           (None, 8, 8, 128)         0         \n",
            "                                                                 \n",
            " conv2d_transpose (Conv2DTra  (None, 16, 16, 128)      262272    \n",
            " nspose)                                                         \n",
            "                                                                 \n",
            " leaky_re_lu_3 (LeakyReLU)   (None, 16, 16, 128)       0         \n",
            "                                                                 \n",
            " conv2d_transpose_1 (Conv2DT  (None, 32, 32, 256)      524544    \n",
            " ranspose)                                                       \n",
            "                                                                 \n",
            " leaky_re_lu_4 (LeakyReLU)   (None, 32, 32, 256)       0         \n",
            "                                                                 \n",
            " conv2d_transpose_2 (Conv2DT  (None, 64, 64, 512)      2097664   \n",
            " ranspose)                                                       \n",
            "                                                                 \n",
            " leaky_re_lu_5 (LeakyReLU)   (None, 64, 64, 512)       0         \n",
            "                                                                 \n",
            " conv2d_3 (Conv2D)           (None, 64, 64, 3)         38403     \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 3,979,651\n",
            "Trainable params: 3,979,651\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ]
        }
      ],
      "source": [
        "generator.summary()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "BKO6zT4Tk_5j"
      },
      "source": [
        "### 적대 네트워크"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "AZT3GvUnk_5j"
      },
      "source": [
        "**GAN 모델**"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 9,
      "metadata": {
        "id": "DQEeSL-7k_5j"
      },
      "outputs": [],
      "source": [
        "import tensorflow as tf\n",
        "class GAN(keras.Model):\n",
        "    def __init__(self, discriminator, generator, latent_dim):\n",
        "        super().__init__()\n",
        "        self.discriminator = discriminator\n",
        "        self.generator = generator\n",
        "        self.latent_dim = latent_dim\n",
        "        self.d_loss_metric = keras.metrics.Mean(name=\"d_loss\")\n",
        "        self.g_loss_metric = keras.metrics.Mean(name=\"g_loss\")\n",
        "\n",
        "    def compile(self, d_optimizer, g_optimizer, loss_fn):\n",
        "        super(GAN, self).compile()\n",
        "        self.d_optimizer = d_optimizer\n",
        "        self.g_optimizer = g_optimizer\n",
        "        self.loss_fn = loss_fn\n",
        "\n",
        "    @property\n",
        "    def metrics(self):\n",
        "        return [self.d_loss_metric, self.g_loss_metric]\n",
        "\n",
        "    def train_step(self, real_images):\n",
        "        batch_size = tf.shape(real_images)[0]\n",
        "        random_latent_vectors = tf.random.normal(\n",
        "            shape=(batch_size, self.latent_dim))\n",
        "        generated_images = self.generator(random_latent_vectors)\n",
        "        combined_images = tf.concat([generated_images, real_images], axis=0)\n",
        "        labels = tf.concat(\n",
        "            [tf.ones((batch_size, 1)), tf.zeros((batch_size, 1))],\n",
        "            axis=0\n",
        "        )\n",
        "        labels += 0.05 * tf.random.uniform(tf.shape(labels))\n",
        "\n",
        "        with tf.GradientTape() as tape:\n",
        "            predictions = self.discriminator(combined_images)\n",
        "            d_loss = self.loss_fn(labels, predictions)\n",
        "        grads = tape.gradient(d_loss, self.discriminator.trainable_weights)\n",
        "        self.d_optimizer.apply_gradients(\n",
        "            zip(grads, self.discriminator.trainable_weights)\n",
        "        )\n",
        "\n",
        "        random_latent_vectors = tf.random.normal(\n",
        "            shape=(batch_size, self.latent_dim))\n",
        "\n",
        "        misleading_labels = tf.zeros((batch_size, 1))\n",
        "\n",
        "        with tf.GradientTape() as tape:\n",
        "            predictions = self.discriminator(\n",
        "                self.generator(random_latent_vectors))\n",
        "            g_loss = self.loss_fn(misleading_labels, predictions)\n",
        "        grads = tape.gradient(g_loss, self.generator.trainable_weights)\n",
        "        self.g_optimizer.apply_gradients(\n",
        "            zip(grads, self.generator.trainable_weights))\n",
        "\n",
        "        self.d_loss_metric.update_state(d_loss)\n",
        "        self.g_loss_metric.update_state(g_loss)\n",
        "        return {\"d_loss\": self.d_loss_metric.result(),\n",
        "                \"g_loss\": self.g_loss_metric.result()}"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "2bDiQ35xk_5j"
      },
      "source": [
        "**훈련 과정 동안에 이미지를 생성하기 위한 콜백**"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 10,
      "metadata": {
        "id": "ye-2NQshk_5k"
      },
      "outputs": [],
      "source": [
        "class GANMonitor(keras.callbacks.Callback):\n",
        "    def __init__(self, num_img=3, latent_dim=128):\n",
        "        self.num_img = num_img\n",
        "        self.latent_dim = latent_dim\n",
        "\n",
        "    def on_epoch_end(self, epoch, logs=None):\n",
        "        random_latent_vectors = tf.random.normal(shape=(self.num_img, self.latent_dim))\n",
        "        generated_images = self.model.generator(random_latent_vectors)\n",
        "        generated_images *= 255\n",
        "        generated_images.numpy()\n",
        "        for i in range(self.num_img):\n",
        "            img = keras.utils.array_to_img(generated_images[i])\n",
        "            img.save(f\"generated_img_{epoch:03d}_{i}.png\")"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Wa2vU_BIk_5k"
      },
      "source": [
        "**GAN 모델 컴파일하고 훈련하기**"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 11,
      "metadata": {
        "id": "QakL8zD3k_5k",
        "outputId": "b2bee7df-eb43-496b-b645-dc6fd1c82939",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/100\n",
            "6332/6332 [==============================] - 198s 29ms/step - d_loss: 0.6091 - g_loss: 1.2006\n",
            "Epoch 2/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6584 - g_loss: 1.0391\n",
            "Epoch 3/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6702 - g_loss: 1.0154\n",
            "Epoch 4/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6629 - g_loss: 1.0196\n",
            "Epoch 5/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6502 - g_loss: 1.0357\n",
            "Epoch 6/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6408 - g_loss: 1.0269\n",
            "Epoch 7/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6625 - g_loss: 0.9492\n",
            "Epoch 8/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6773 - g_loss: 0.9211\n",
            "Epoch 9/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6729 - g_loss: 0.9119\n",
            "Epoch 10/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6825 - g_loss: 0.8961\n",
            "Epoch 11/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6798 - g_loss: 0.8837\n",
            "Epoch 12/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6798 - g_loss: 0.8822\n",
            "Epoch 13/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6764 - g_loss: 0.8949\n",
            "Epoch 14/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6709 - g_loss: 0.9117\n",
            "Epoch 15/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6673 - g_loss: 0.8966\n",
            "Epoch 16/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6662 - g_loss: 0.9212\n",
            "Epoch 17/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6573 - g_loss: 0.9218\n",
            "Epoch 18/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6616 - g_loss: 0.9185\n",
            "Epoch 19/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6631 - g_loss: 0.9301\n",
            "Epoch 20/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6586 - g_loss: 0.9277\n",
            "Epoch 21/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6578 - g_loss: 0.9231\n",
            "Epoch 22/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6592 - g_loss: 0.9302\n",
            "Epoch 23/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6684 - g_loss: 0.9187\n",
            "Epoch 24/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6594 - g_loss: 0.9234\n",
            "Epoch 25/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6565 - g_loss: 0.9256\n",
            "Epoch 26/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6584 - g_loss: 0.9264\n",
            "Epoch 27/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6560 - g_loss: 0.9151\n",
            "Epoch 28/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6599 - g_loss: 0.9267\n",
            "Epoch 29/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6524 - g_loss: 0.9198\n",
            "Epoch 30/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6548 - g_loss: 0.9217\n",
            "Epoch 31/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6552 - g_loss: 0.9238\n",
            "Epoch 32/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6471 - g_loss: 0.9262\n",
            "Epoch 33/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6452 - g_loss: 0.9490\n",
            "Epoch 34/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6421 - g_loss: 0.9450\n",
            "Epoch 35/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6402 - g_loss: 0.9530\n",
            "Epoch 36/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6390 - g_loss: 0.9554\n",
            "Epoch 37/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6382 - g_loss: 0.9570\n",
            "Epoch 38/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6371 - g_loss: 0.9607\n",
            "Epoch 39/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6373 - g_loss: 0.9603\n",
            "Epoch 40/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6372 - g_loss: 0.9567\n",
            "Epoch 41/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6371 - g_loss: 0.9620\n",
            "Epoch 42/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6388 - g_loss: 0.9531\n",
            "Epoch 43/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6396 - g_loss: 0.9547\n",
            "Epoch 44/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6392 - g_loss: 0.9533\n",
            "Epoch 45/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6402 - g_loss: 0.9558\n",
            "Epoch 46/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6381 - g_loss: 0.9569\n",
            "Epoch 47/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6394 - g_loss: 0.9537\n",
            "Epoch 48/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6394 - g_loss: 0.9551\n",
            "Epoch 49/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6349 - g_loss: 0.9627\n",
            "Epoch 50/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6343 - g_loss: 0.9714\n",
            "Epoch 51/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6341 - g_loss: 0.9724\n",
            "Epoch 52/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6324 - g_loss: 0.9642\n",
            "Epoch 53/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6320 - g_loss: 0.9780\n",
            "Epoch 54/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6308 - g_loss: 0.9719\n",
            "Epoch 55/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6305 - g_loss: 0.9760\n",
            "Epoch 56/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6287 - g_loss: 0.9816\n",
            "Epoch 57/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6300 - g_loss: 0.9797\n",
            "Epoch 58/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6268 - g_loss: 0.9915\n",
            "Epoch 59/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6260 - g_loss: 0.9884\n",
            "Epoch 60/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6236 - g_loss: 1.0002\n",
            "Epoch 61/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6222 - g_loss: 1.0085\n",
            "Epoch 62/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6208 - g_loss: 1.0030\n",
            "Epoch 63/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6192 - g_loss: 1.0194\n",
            "Epoch 64/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6189 - g_loss: 1.0062\n",
            "Epoch 65/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6156 - g_loss: 1.0116\n",
            "Epoch 66/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6142 - g_loss: 1.0262\n",
            "Epoch 67/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6134 - g_loss: 1.0309\n",
            "Epoch 68/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6129 - g_loss: 1.0352\n",
            "Epoch 69/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6136 - g_loss: 1.0228\n",
            "Epoch 70/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6113 - g_loss: 1.0314\n",
            "Epoch 71/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6092 - g_loss: 1.0377\n",
            "Epoch 72/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6085 - g_loss: 1.0433\n",
            "Epoch 73/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6068 - g_loss: 1.0403\n",
            "Epoch 74/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6062 - g_loss: 1.0443\n",
            "Epoch 75/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6028 - g_loss: 1.0567\n",
            "Epoch 76/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6018 - g_loss: 1.0585\n",
            "Epoch 77/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.6007 - g_loss: 1.0655\n",
            "Epoch 78/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.5987 - g_loss: 1.0841\n",
            "Epoch 79/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.5995 - g_loss: 1.0623\n",
            "Epoch 80/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.5989 - g_loss: 1.0662\n",
            "Epoch 81/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.5975 - g_loss: 1.0707\n",
            "Epoch 82/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.5953 - g_loss: 1.0729\n",
            "Epoch 83/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.5965 - g_loss: 1.0749\n",
            "Epoch 84/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.5958 - g_loss: 1.0864\n",
            "Epoch 85/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.5935 - g_loss: 1.0886\n",
            "Epoch 86/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.5895 - g_loss: 1.1058\n",
            "Epoch 87/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.5894 - g_loss: 1.0947\n",
            "Epoch 88/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.5875 - g_loss: 1.0980\n",
            "Epoch 89/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.5870 - g_loss: 1.1043\n",
            "Epoch 90/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.5842 - g_loss: 1.0996\n",
            "Epoch 91/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.5827 - g_loss: 1.1157\n",
            "Epoch 92/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.5827 - g_loss: 1.1158\n",
            "Epoch 93/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.5816 - g_loss: 1.1179\n",
            "Epoch 94/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.5781 - g_loss: 1.1229\n",
            "Epoch 95/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.5783 - g_loss: 1.1290\n",
            "Epoch 96/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.5759 - g_loss: 1.1402\n",
            "Epoch 97/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.5744 - g_loss: 1.1457\n",
            "Epoch 98/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.5744 - g_loss: 1.1455\n",
            "Epoch 99/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.5724 - g_loss: 1.1482\n",
            "Epoch 100/100\n",
            "6332/6332 [==============================] - 185s 29ms/step - d_loss: 0.5708 - g_loss: 1.1529\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<keras.callbacks.History at 0x7ffa26bdc890>"
            ]
          },
          "metadata": {},
          "execution_count": 11
        }
      ],
      "source": [
        "epochs = 100\n",
        "\n",
        "gan = GAN(discriminator=discriminator, generator=generator, latent_dim=latent_dim)\n",
        "gan.compile(\n",
        "    d_optimizer=keras.optimizers.Adam(learning_rate=0.0001),\n",
        "    g_optimizer=keras.optimizers.Adam(learning_rate=0.0001),\n",
        "    loss_fn=keras.losses.BinaryCrossentropy(),\n",
        ")\n",
        "\n",
        "gan.fit(\n",
        "    dataset, epochs=epochs, callbacks=[GANMonitor(num_img=10, latent_dim=latent_dim)]\n",
        ")"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "RIAF-5u_k_5k"
      },
      "source": [
        "### 정리"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "YhQ9GDfJk_5k"
      },
      "source": [
        "## 요약"
      ]
    }
  ],
  "metadata": {
    "colab": {
      "name": "chapter12_part05_gans.i",
      "provenance": []
    },
    "kernelspec": {
      "display_name": "default:Python",
      "language": "python",
      "name": "conda-env-default-py"
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.9.10"
    },
    "accelerator": "GPU",
    "gpuClass": "premium"
  },
  "nbformat": 4,
  "nbformat_minor": 0
}