GAN/GAN(2014).ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "GAN(2014).ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "machine_shape": "hm",
      "authorship_tag": "ABX9TyO4B8L7fNte1OjXoB4ltTxS",
      "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/Seonghoon-Yu/Paper_Review_and_Implementation_in_PyTorch/blob/master/GAN/GAN(2014).ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "QFaVqwFDCB-B"
      },
      "source": [
        "import torch\n",
        "import torch.nn as nn\n",
        "import torch.nn.functional as F\n",
        "\n",
        "from torchvision import datasets\n",
        "import torchvision.transforms as transforms\n",
        "from torch.utils.data import DataLoader\n",
        "from torchvision.transforms.functional import to_pil_image\n",
        "import matplotlib.pylab as plt\n",
        "%matplotlib inline\n",
        "\n",
        "import os\n",
        "import numpy as np\n",
        "import time\n",
        "\n",
        "device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')"
      ],
      "execution_count": 2,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "d77l7D7EC4i7"
      },
      "source": [
        "# 1. 데이터셋 불러오기"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "bntTCkWpCe_r"
      },
      "source": [
        "# 데이터 경로 지정\n",
        "path2data = './data'\n",
        "os.makedirs(path2data, exist_ok=True) # 폴더 생성"
      ],
      "execution_count": 40,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Bw66vqNSEZNW"
      },
      "source": [
        "# MNIST dataset 불러오기\n",
        "train_ds = datasets.MNIST(path2data, train=True, transform=transforms.Compose([transforms.ToTensor(),transforms.Normalize([0.5],[0.5])]), download=True)"
      ],
      "execution_count": 41,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 283
        },
        "id": "sVRHwXtbEyfw",
        "outputId": "d554d58d-d8f3-4dc8-ac86-c2bcef944271"
      },
      "source": [
        "# 샘플 이미지 확인\n",
        "img, label = train_ds[0]\n",
        "plt.imshow(to_pil_image(0.5*img+0.5),cmap='gray')"
      ],
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.image.AxesImage at 0x7f9831742390>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 5
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "rDXUwtGNFJSG"
      },
      "source": [
        "# 데이터 로더 생성\n",
        "train_dl = DataLoader(train_ds, batch_size=32, shuffle=True)"
      ],
      "execution_count": 6,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "qO02zq-BIyG9",
        "outputId": "0b91878e-30c4-48d1-c4b2-e4d8e85fc051"
      },
      "source": [
        "# check\n",
        "for x, y in train_dl:\n",
        "    print(x.shape, y.shape)\n",
        "    break"
      ],
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "torch.Size([32, 1, 28, 28]) torch.Size([32])\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "4uY95ka7Jfr6"
      },
      "source": [
        "# 2. 모델 구축하기"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "M9LUeRvII25r",
        "outputId": "a0efa064-edf1-4f12-bb0c-4e0a2c1835a1"
      },
      "source": [
        "# generator: noise를 입력받아 이미지를 생성합니다.\n",
        "class Generator(nn.Module):\n",
        "    def __init__(self, params):\n",
        "        super().__init__()\n",
        "        self.nz = params['nz'] # 입력 노이즈 벡터 수, 100\n",
        "        self.img_size = params['img_size'] # 이미지 크기, 1x28x28\n",
        "\n",
        "        self.model = nn.Sequential(\n",
        "            *self._fc_layer(self.nz, 128, normalize=False),\n",
        "            *self._fc_layer(128,256),\n",
        "            *self._fc_layer(256,512),\n",
        "            *self._fc_layer(512,1024),\n",
        "            nn.Linear(1024,int(np.prod(self.img_size))),\n",
        "            nn.Tanh()\n",
        "        )\n",
        "\n",
        "    def forward(self, z):\n",
        "        img = self.model(z)\n",
        "        img = img.view(img.size(0), *self.img_size)\n",
        "        return img\n",
        "\n",
        "    # fc layer\n",
        "    def _fc_layer(self, in_channels, out_channels, normalize=True):\n",
        "        layers = []\n",
        "        layers.append(nn.Linear(in_channels, out_channels)) # fc layer\n",
        "        if normalize:\n",
        "            layers.append(nn.BatchNorm1d(out_channels, 0.8)) # BN\n",
        "        layers.append(nn.LeakyReLU(0.2)) # LeakyReLU\n",
        "        return layers\n",
        "\n",
        "# check\n",
        "params = {'nz':100,\n",
        "          'img_size':(1,28,28)}\n",
        "x = torch.randn(16,100).to(device) # random noise\n",
        "model_gen = Generator(params).to(device)\n",
        "output = model_gen(x) # noise를 입력받아 이미지 생성\n",
        "print(output.shape)"
      ],
      "execution_count": 45,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "torch.Size([16, 1, 28, 28])\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "lge1B1VJMKUi",
        "outputId": "d49ad253-ecd4-4056-df72-9ad95fd9061f"
      },
      "source": [
        "# discriminator: 진짜 이미지와 가짜 이미지를 분류합니다.\n",
        "class Discriminator(nn.Module):\n",
        "    def __init__(self,params):\n",
        "        super().__init__()\n",
        "        self.img_size = params['img_size'] # 이미지 크기, 1x28x28\n",
        "\n",
        "        self.model = nn.Sequential(\n",
        "            nn.Linear(int(np.prod(self.img_size)), 512),\n",
        "            nn.LeakyReLU(0.2),\n",
        "            nn.Linear(512,256),\n",
        "            nn.LeakyReLU(0.2),\n",
        "            nn.Linear(256,1),\n",
        "            nn.Sigmoid()\n",
        "        )\n",
        "\n",
        "    def forward(self, x):\n",
        "        x = x.view(x.size(0),-1)\n",
        "        x = self.model(x)\n",
        "        return x\n",
        "\n",
        "# check\n",
        "x = torch.randn(16,1,28,28).to(device)\n",
        "model_dis = Discriminator(params).to(device)\n",
        "output = model_dis(x)\n",
        "print(output.shape)"
      ],
      "execution_count": 46,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "torch.Size([16, 1])\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hR7CZhmyqUYc"
      },
      "source": [
        "# 가중치 초기화\n",
        "def initialize_weights(model):\n",
        "    classname = model.__class__.__name__\n",
        "    # fc layer\n",
        "    if classname.find('Linear') != -1:\n",
        "        nn.init.normal_(model.weight.data, 0.0, 0.02)\n",
        "        nn.init.constant_(model.bias.data, 0)\n",
        "    # batchnorm\n",
        "    elif classname.find('BatchNorm') != -1:\n",
        "        nn.init.normal_(model.weight.data, 1.0, 0.02)\n",
        "        nn.init.constant_(model.bias.data, 0)\n",
        "\n",
        "# 가중치 초기화 적용\n",
        "model_gen.apply(initialize_weights);\n",
        "model_dis.apply(initialize_weights);"
      ],
      "execution_count": 78,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "aJdMBc0wqKzw"
      },
      "source": [
        "# 3. 학습하기"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "uykVj7nhpq7u"
      },
      "source": [
        "# 손실 함수\n",
        "loss_func = nn.BCELoss()"
      ],
      "execution_count": 79,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "AC4r8RVHsF8c"
      },
      "source": [
        "from torch import optim\n",
        "\n",
        "# 최적화 파라미터\n",
        "lr = 2e-4\n",
        "beta1 = 0.5\n",
        "\n",
        "opt_dis = optim.Adam(model_dis.parameters(),lr=lr,betas=(beta1,0.999))\n",
        "opt_gen = optim.Adam(model_gen.parameters(),lr=lr,betas=(beta1,0.999))"
      ],
      "execution_count": 80,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ofBe2ekcsRgy"
      },
      "source": [
        "real_label = 1.\n",
        "fake_label = 0.\n",
        "nz = params['nz']\n",
        "num_epochs = 100\n",
        "\n",
        "loss_history={'gen':[],\n",
        "              'dis':[]}"
      ],
      "execution_count": 81,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "5clROy9DLqdZ",
        "outputId": "505f24a7-4363-4b11-b9eb-bdcc1416f5c8"
      },
      "source": [
        "batch_count = 0\n",
        "start_time = time.time()\n",
        "model_dis.train()\n",
        "model_gen.train()\n",
        "\n",
        "for epoch in range(num_epochs):\n",
        "    for xb, yb in train_dl:\n",
        "        ba_si = xb.size(0)\n",
        "\n",
        "        xb = xb.to(device)\n",
        "        yb_real = torch.Tensor(ba_si,1).fill_(1.0).to(device)\n",
        "        yb_fake = torch.Tensor(ba_si,1).fill_(0.0).to(device)\n",
        "\n",
        "        # Generator\n",
        "        model_gen.zero_grad()\n",
        "        noise = torch.randn(ba_si,nz, device=device) # 노이즈 생성\n",
        "        out_gen = model_gen(noise) # 가짜 이미지 생성\n",
        "        out_dis = model_dis(out_gen) # 가짜 이미지 판별\n",
        "\n",
        "        loss_gen = loss_func(out_dis, yb_real)\n",
        "        loss_gen.backward()\n",
        "        opt_gen.step()\n",
        "\n",
        "        # Discriminator\n",
        "        model_dis.zero_grad()\n",
        "\n",
        "        out_real = model_dis(xb) # 진짜 이미지 판별\n",
        "        out_fake = model_dis(out_gen.detach()) # 가짜 이미지 판별\n",
        "        loss_real = loss_func(out_real, yb_real)\n",
        "        loss_fake = loss_func(out_fake, yb_fake)\n",
        "        loss_dis = (loss_real + loss_fake) / 2\n",
        "\n",
        "        loss_dis.backward()\n",
        "        opt_dis.step()\n",
        "\n",
        "        loss_history['gen'].append(loss_gen.item())\n",
        "        loss_history['dis'].append(loss_dis.item())\n",
        "\n",
        "        batch_count += 1\n",
        "        if batch_count % 1000 == 0:\n",
        "            print('Epoch: %.0f, G_Loss: %.6f, D_Loss: %.6f, time: %.2f min' %(epoch, loss_gen.item(), loss_dis.item(), (time.time()-start_time)/60))"
      ],
      "execution_count": 82,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch: 0, G_Loss: 3.498638, D_Loss: 0.375694, time: 0.23 min\n",
            "Epoch: 1, G_Loss: 1.850641, D_Loss: 0.299748, time: 0.45 min\n",
            "Epoch: 1, G_Loss: 2.619798, D_Loss: 0.536759, time: 0.68 min\n",
            "Epoch: 2, G_Loss: 1.812477, D_Loss: 0.254422, time: 0.90 min\n",
            "Epoch: 2, G_Loss: 2.712598, D_Loss: 0.236421, time: 1.12 min\n",
            "Epoch: 3, G_Loss: 1.607915, D_Loss: 0.181307, time: 1.35 min\n",
            "Epoch: 3, G_Loss: 1.966413, D_Loss: 0.210961, time: 1.57 min\n",
            "Epoch: 4, G_Loss: 2.309307, D_Loss: 0.189411, time: 1.79 min\n",
            "Epoch: 4, G_Loss: 0.466723, D_Loss: 0.722811, time: 2.01 min\n",
            "Epoch: 5, G_Loss: 1.647252, D_Loss: 0.169069, time: 2.24 min\n",
            "Epoch: 5, G_Loss: 2.407640, D_Loss: 0.215403, time: 2.46 min\n",
            "Epoch: 6, G_Loss: 2.970379, D_Loss: 0.280537, time: 2.68 min\n",
            "Epoch: 6, G_Loss: 1.146987, D_Loss: 0.241861, time: 2.90 min\n",
            "Epoch: 7, G_Loss: 2.779574, D_Loss: 0.309728, time: 3.12 min\n",
            "Epoch: 7, G_Loss: 3.040692, D_Loss: 0.074625, time: 3.35 min\n",
            "Epoch: 8, G_Loss: 2.921538, D_Loss: 0.106340, time: 3.57 min\n",
            "Epoch: 9, G_Loss: 1.533596, D_Loss: 0.294615, time: 3.79 min\n",
            "Epoch: 9, G_Loss: 2.948410, D_Loss: 0.105762, time: 4.01 min\n",
            "Epoch: 10, G_Loss: 2.249388, D_Loss: 0.159334, time: 4.24 min\n",
            "Epoch: 10, G_Loss: 2.100729, D_Loss: 0.213161, time: 4.46 min\n",
            "Epoch: 11, G_Loss: 2.914240, D_Loss: 0.143784, time: 4.68 min\n",
            "Epoch: 11, G_Loss: 2.811936, D_Loss: 0.091328, time: 4.90 min\n",
            "Epoch: 12, G_Loss: 2.489695, D_Loss: 0.226623, time: 5.12 min\n",
            "Epoch: 12, G_Loss: 0.566901, D_Loss: 0.491195, time: 5.35 min\n",
            "Epoch: 13, G_Loss: 2.287121, D_Loss: 0.141648, time: 5.57 min\n",
            "Epoch: 13, G_Loss: 2.620127, D_Loss: 0.125651, time: 5.79 min\n",
            "Epoch: 14, G_Loss: 3.327636, D_Loss: 0.226355, time: 6.01 min\n",
            "Epoch: 14, G_Loss: 2.338894, D_Loss: 0.161854, time: 6.23 min\n",
            "Epoch: 15, G_Loss: 2.297944, D_Loss: 0.328029, time: 6.45 min\n",
            "Epoch: 15, G_Loss: 2.307637, D_Loss: 0.179546, time: 6.68 min\n",
            "Epoch: 16, G_Loss: 3.503555, D_Loss: 0.129540, time: 6.90 min\n",
            "Epoch: 17, G_Loss: 2.588130, D_Loss: 0.243826, time: 7.12 min\n",
            "Epoch: 17, G_Loss: 4.544392, D_Loss: 0.203891, time: 7.34 min\n",
            "Epoch: 18, G_Loss: 4.020932, D_Loss: 0.083406, time: 7.56 min\n",
            "Epoch: 18, G_Loss: 3.488378, D_Loss: 0.293025, time: 7.79 min\n",
            "Epoch: 19, G_Loss: 1.528384, D_Loss: 0.191433, time: 8.01 min\n",
            "Epoch: 19, G_Loss: 3.293623, D_Loss: 0.068279, time: 8.23 min\n",
            "Epoch: 20, G_Loss: 1.772479, D_Loss: 0.166918, time: 8.46 min\n",
            "Epoch: 20, G_Loss: 3.238154, D_Loss: 0.056369, time: 8.68 min\n",
            "Epoch: 21, G_Loss: 1.705830, D_Loss: 0.177880, time: 8.90 min\n",
            "Epoch: 21, G_Loss: 2.763665, D_Loss: 0.085219, time: 9.12 min\n",
            "Epoch: 22, G_Loss: 1.471455, D_Loss: 0.221638, time: 9.35 min\n",
            "Epoch: 22, G_Loss: 2.986885, D_Loss: 0.399185, time: 9.57 min\n",
            "Epoch: 23, G_Loss: 5.712301, D_Loss: 0.215026, time: 9.79 min\n",
            "Epoch: 23, G_Loss: 4.285053, D_Loss: 0.066787, time: 10.01 min\n",
            "Epoch: 24, G_Loss: 2.016886, D_Loss: 0.138151, time: 10.23 min\n",
            "Epoch: 25, G_Loss: 1.412758, D_Loss: 0.234239, time: 10.45 min\n",
            "Epoch: 25, G_Loss: 3.199535, D_Loss: 0.155740, time: 10.67 min\n",
            "Epoch: 26, G_Loss: 2.502598, D_Loss: 0.226552, time: 10.89 min\n",
            "Epoch: 26, G_Loss: 3.815098, D_Loss: 0.081462, time: 11.12 min\n",
            "Epoch: 27, G_Loss: 3.310400, D_Loss: 0.059737, time: 11.34 min\n",
            "Epoch: 27, G_Loss: 3.064906, D_Loss: 0.121215, time: 11.56 min\n",
            "Epoch: 28, G_Loss: 2.739565, D_Loss: 0.124082, time: 11.78 min\n",
            "Epoch: 28, G_Loss: 4.595774, D_Loss: 0.124876, time: 12.00 min\n",
            "Epoch: 29, G_Loss: 2.489260, D_Loss: 0.129450, time: 12.22 min\n",
            "Epoch: 29, G_Loss: 1.125296, D_Loss: 0.621487, time: 12.44 min\n",
            "Epoch: 30, G_Loss: 3.691485, D_Loss: 0.144749, time: 12.66 min\n",
            "Epoch: 30, G_Loss: 4.027232, D_Loss: 0.121965, time: 12.88 min\n",
            "Epoch: 31, G_Loss: 4.212004, D_Loss: 0.055662, time: 13.10 min\n",
            "Epoch: 31, G_Loss: 3.212385, D_Loss: 0.054350, time: 13.32 min\n",
            "Epoch: 32, G_Loss: 4.163943, D_Loss: 0.419858, time: 13.54 min\n",
            "Epoch: 33, G_Loss: 1.329105, D_Loss: 0.229900, time: 13.76 min\n",
            "Epoch: 33, G_Loss: 4.954628, D_Loss: 0.209757, time: 13.98 min\n",
            "Epoch: 34, G_Loss: 0.670434, D_Loss: 0.590529, time: 14.20 min\n",
            "Epoch: 34, G_Loss: 3.799446, D_Loss: 0.226425, time: 14.42 min\n",
            "Epoch: 35, G_Loss: 3.362128, D_Loss: 0.132861, time: 14.63 min\n",
            "Epoch: 35, G_Loss: 3.128906, D_Loss: 0.193885, time: 14.85 min\n",
            "Epoch: 36, G_Loss: 2.437550, D_Loss: 0.138790, time: 15.07 min\n",
            "Epoch: 36, G_Loss: 2.623072, D_Loss: 0.215258, time: 15.29 min\n",
            "Epoch: 37, G_Loss: 3.232531, D_Loss: 0.127286, time: 15.51 min\n",
            "Epoch: 37, G_Loss: 1.369317, D_Loss: 0.311535, time: 15.73 min\n",
            "Epoch: 38, G_Loss: 3.157478, D_Loss: 0.254002, time: 15.94 min\n",
            "Epoch: 38, G_Loss: 2.653631, D_Loss: 0.085755, time: 16.16 min\n",
            "Epoch: 39, G_Loss: 3.115479, D_Loss: 0.103593, time: 16.38 min\n",
            "Epoch: 39, G_Loss: 2.409679, D_Loss: 0.235843, time: 16.60 min\n",
            "Epoch: 40, G_Loss: 7.776254, D_Loss: 0.041929, time: 16.82 min\n",
            "Epoch: 41, G_Loss: 3.388594, D_Loss: 0.091750, time: 17.03 min\n",
            "Epoch: 41, G_Loss: 3.891364, D_Loss: 0.084286, time: 17.25 min\n",
            "Epoch: 42, G_Loss: 3.618611, D_Loss: 0.091433, time: 17.47 min\n",
            "Epoch: 42, G_Loss: 2.503089, D_Loss: 0.100597, time: 17.68 min\n",
            "Epoch: 43, G_Loss: 3.659994, D_Loss: 0.229176, time: 17.90 min\n",
            "Epoch: 43, G_Loss: 3.667628, D_Loss: 0.057874, time: 18.12 min\n",
            "Epoch: 44, G_Loss: 1.782672, D_Loss: 0.181885, time: 18.34 min\n",
            "Epoch: 44, G_Loss: 3.963471, D_Loss: 0.075346, time: 18.55 min\n",
            "Epoch: 45, G_Loss: 3.312521, D_Loss: 0.132464, time: 18.77 min\n",
            "Epoch: 45, G_Loss: 2.431857, D_Loss: 0.181363, time: 18.99 min\n",
            "Epoch: 46, G_Loss: 3.375199, D_Loss: 0.048896, time: 19.20 min\n",
            "Epoch: 46, G_Loss: 3.230518, D_Loss: 0.061733, time: 19.42 min\n",
            "Epoch: 47, G_Loss: 3.114158, D_Loss: 0.109792, time: 19.63 min\n",
            "Epoch: 47, G_Loss: 2.695733, D_Loss: 0.179052, time: 19.85 min\n",
            "Epoch: 48, G_Loss: 2.953986, D_Loss: 0.059182, time: 20.07 min\n",
            "Epoch: 49, G_Loss: 2.542557, D_Loss: 0.264080, time: 20.28 min\n",
            "Epoch: 49, G_Loss: 3.334543, D_Loss: 0.079255, time: 20.50 min\n",
            "Epoch: 50, G_Loss: 3.707491, D_Loss: 0.056738, time: 20.72 min\n",
            "Epoch: 50, G_Loss: 3.582634, D_Loss: 0.131147, time: 20.93 min\n",
            "Epoch: 51, G_Loss: 3.309745, D_Loss: 0.088996, time: 21.15 min\n",
            "Epoch: 51, G_Loss: 2.142919, D_Loss: 0.178596, time: 21.37 min\n",
            "Epoch: 52, G_Loss: 3.496007, D_Loss: 0.052681, time: 21.58 min\n",
            "Epoch: 52, G_Loss: 2.633413, D_Loss: 0.212295, time: 21.80 min\n",
            "Epoch: 53, G_Loss: 2.670349, D_Loss: 0.238003, time: 22.02 min\n",
            "Epoch: 53, G_Loss: 7.047523, D_Loss: 0.026169, time: 22.24 min\n",
            "Epoch: 54, G_Loss: 3.731299, D_Loss: 0.130555, time: 22.46 min\n",
            "Epoch: 54, G_Loss: 4.044308, D_Loss: 0.324557, time: 22.67 min\n",
            "Epoch: 55, G_Loss: 1.694305, D_Loss: 0.266185, time: 22.89 min\n",
            "Epoch: 55, G_Loss: 3.372304, D_Loss: 0.077528, time: 23.11 min\n",
            "Epoch: 56, G_Loss: 2.822736, D_Loss: 0.141111, time: 23.33 min\n",
            "Epoch: 57, G_Loss: 2.895681, D_Loss: 0.239868, time: 23.54 min\n",
            "Epoch: 57, G_Loss: 3.423336, D_Loss: 0.052131, time: 23.76 min\n",
            "Epoch: 58, G_Loss: 3.397103, D_Loss: 0.172800, time: 23.98 min\n",
            "Epoch: 58, G_Loss: 3.273804, D_Loss: 0.105899, time: 24.19 min\n",
            "Epoch: 59, G_Loss: 2.555472, D_Loss: 0.069924, time: 24.41 min\n",
            "Epoch: 59, G_Loss: 5.207808, D_Loss: 0.112289, time: 24.63 min\n",
            "Epoch: 60, G_Loss: 4.430760, D_Loss: 0.155407, time: 24.84 min\n",
            "Epoch: 60, G_Loss: 4.233754, D_Loss: 0.155034, time: 25.06 min\n",
            "Epoch: 61, G_Loss: 3.565111, D_Loss: 0.107371, time: 25.27 min\n",
            "Epoch: 61, G_Loss: 4.813363, D_Loss: 0.189577, time: 25.49 min\n",
            "Epoch: 62, G_Loss: 3.459729, D_Loss: 0.107057, time: 25.71 min\n",
            "Epoch: 62, G_Loss: 2.792338, D_Loss: 0.074543, time: 25.93 min\n",
            "Epoch: 63, G_Loss: 4.096992, D_Loss: 0.058835, time: 26.14 min\n",
            "Epoch: 63, G_Loss: 4.945120, D_Loss: 0.195930, time: 26.36 min\n",
            "Epoch: 64, G_Loss: 2.673388, D_Loss: 0.246955, time: 26.58 min\n",
            "Epoch: 65, G_Loss: 2.682917, D_Loss: 0.076982, time: 26.80 min\n",
            "Epoch: 65, G_Loss: 2.847306, D_Loss: 0.155299, time: 27.01 min\n",
            "Epoch: 66, G_Loss: 2.159664, D_Loss: 0.252792, time: 27.23 min\n",
            "Epoch: 66, G_Loss: 2.622749, D_Loss: 0.097166, time: 27.45 min\n",
            "Epoch: 67, G_Loss: 6.378613, D_Loss: 0.153081, time: 27.67 min\n",
            "Epoch: 67, G_Loss: 4.294859, D_Loss: 0.048351, time: 27.88 min\n",
            "Epoch: 68, G_Loss: 2.721730, D_Loss: 0.179621, time: 28.10 min\n",
            "Epoch: 68, G_Loss: 3.335936, D_Loss: 0.122942, time: 28.32 min\n",
            "Epoch: 69, G_Loss: 4.437977, D_Loss: 0.260216, time: 28.54 min\n",
            "Epoch: 69, G_Loss: 4.383181, D_Loss: 0.097826, time: 28.76 min\n",
            "Epoch: 70, G_Loss: 4.243635, D_Loss: 0.110845, time: 28.97 min\n",
            "Epoch: 70, G_Loss: 1.770964, D_Loss: 0.216282, time: 29.19 min\n",
            "Epoch: 71, G_Loss: 3.223286, D_Loss: 0.061769, time: 29.41 min\n",
            "Epoch: 71, G_Loss: 4.212081, D_Loss: 0.162298, time: 29.63 min\n",
            "Epoch: 72, G_Loss: 1.573482, D_Loss: 0.297443, time: 29.84 min\n",
            "Epoch: 73, G_Loss: 2.771781, D_Loss: 0.145966, time: 30.06 min\n",
            "Epoch: 73, G_Loss: 2.761238, D_Loss: 0.233548, time: 30.27 min\n",
            "Epoch: 74, G_Loss: 2.651371, D_Loss: 0.151971, time: 30.49 min\n",
            "Epoch: 74, G_Loss: 2.525054, D_Loss: 0.361976, time: 30.70 min\n",
            "Epoch: 75, G_Loss: 2.191387, D_Loss: 0.147214, time: 30.92 min\n",
            "Epoch: 75, G_Loss: 3.098619, D_Loss: 0.071997, time: 31.13 min\n",
            "Epoch: 76, G_Loss: 3.567944, D_Loss: 0.210863, time: 31.35 min\n",
            "Epoch: 76, G_Loss: 4.392158, D_Loss: 0.132867, time: 31.56 min\n",
            "Epoch: 77, G_Loss: 3.886850, D_Loss: 0.131642, time: 31.78 min\n",
            "Epoch: 77, G_Loss: 2.422560, D_Loss: 0.110232, time: 31.99 min\n",
            "Epoch: 78, G_Loss: 4.566884, D_Loss: 0.116173, time: 32.21 min\n",
            "Epoch: 78, G_Loss: 3.709624, D_Loss: 0.067735, time: 32.43 min\n",
            "Epoch: 79, G_Loss: 2.056180, D_Loss: 0.193841, time: 32.64 min\n",
            "Epoch: 79, G_Loss: 6.566374, D_Loss: 0.252467, time: 32.86 min\n",
            "Epoch: 80, G_Loss: 4.255255, D_Loss: 0.151197, time: 33.08 min\n",
            "Epoch: 81, G_Loss: 3.901773, D_Loss: 0.118810, time: 33.29 min\n",
            "Epoch: 81, G_Loss: 3.504569, D_Loss: 0.148592, time: 33.51 min\n",
            "Epoch: 82, G_Loss: 2.723624, D_Loss: 0.151581, time: 33.72 min\n",
            "Epoch: 82, G_Loss: 4.110572, D_Loss: 0.080868, time: 33.94 min\n",
            "Epoch: 83, G_Loss: 4.270828, D_Loss: 0.221109, time: 34.15 min\n",
            "Epoch: 83, G_Loss: 4.076939, D_Loss: 0.216016, time: 34.37 min\n",
            "Epoch: 84, G_Loss: 2.609418, D_Loss: 0.191630, time: 34.58 min\n",
            "Epoch: 84, G_Loss: 3.919292, D_Loss: 0.212500, time: 34.79 min\n",
            "Epoch: 85, G_Loss: 2.810794, D_Loss: 0.161165, time: 35.01 min\n",
            "Epoch: 85, G_Loss: 2.676283, D_Loss: 0.145690, time: 35.22 min\n",
            "Epoch: 86, G_Loss: 3.338372, D_Loss: 0.117736, time: 35.44 min\n",
            "Epoch: 86, G_Loss: 3.398826, D_Loss: 0.263057, time: 35.65 min\n",
            "Epoch: 87, G_Loss: 4.305058, D_Loss: 0.124101, time: 35.87 min\n",
            "Epoch: 87, G_Loss: 3.608358, D_Loss: 0.167141, time: 36.08 min\n",
            "Epoch: 88, G_Loss: 2.659655, D_Loss: 0.155571, time: 36.30 min\n",
            "Epoch: 89, G_Loss: 4.427001, D_Loss: 0.240419, time: 36.51 min\n",
            "Epoch: 89, G_Loss: 4.414146, D_Loss: 0.144233, time: 36.73 min\n",
            "Epoch: 90, G_Loss: 2.749214, D_Loss: 0.229142, time: 36.94 min\n",
            "Epoch: 90, G_Loss: 2.223974, D_Loss: 0.100308, time: 37.15 min\n",
            "Epoch: 91, G_Loss: 4.395050, D_Loss: 0.094896, time: 37.37 min\n",
            "Epoch: 91, G_Loss: 2.309773, D_Loss: 0.142970, time: 37.58 min\n",
            "Epoch: 92, G_Loss: 3.382685, D_Loss: 0.131216, time: 37.80 min\n",
            "Epoch: 92, G_Loss: 2.120972, D_Loss: 0.201700, time: 38.02 min\n",
            "Epoch: 93, G_Loss: 3.691568, D_Loss: 0.288425, time: 38.23 min\n",
            "Epoch: 93, G_Loss: 2.069589, D_Loss: 0.183553, time: 38.44 min\n",
            "Epoch: 94, G_Loss: 2.642028, D_Loss: 0.265432, time: 38.66 min\n",
            "Epoch: 94, G_Loss: 3.599956, D_Loss: 0.075526, time: 38.87 min\n",
            "Epoch: 95, G_Loss: 2.997996, D_Loss: 0.074552, time: 39.08 min\n",
            "Epoch: 95, G_Loss: 3.224421, D_Loss: 0.080576, time: 39.30 min\n",
            "Epoch: 96, G_Loss: 4.828950, D_Loss: 0.170278, time: 39.51 min\n",
            "Epoch: 97, G_Loss: 4.668216, D_Loss: 0.145631, time: 39.73 min\n",
            "Epoch: 97, G_Loss: 2.868256, D_Loss: 0.071798, time: 39.94 min\n",
            "Epoch: 98, G_Loss: 3.282309, D_Loss: 0.060444, time: 40.15 min\n",
            "Epoch: 98, G_Loss: 2.445757, D_Loss: 0.101659, time: 40.37 min\n",
            "Epoch: 99, G_Loss: 4.160480, D_Loss: 0.127495, time: 40.58 min\n",
            "Epoch: 99, G_Loss: 4.495906, D_Loss: 0.016227, time: 40.79 min\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 350
        },
        "id": "06z0hdr1y5DZ",
        "outputId": "7473c5c4-8ea6-4d2d-c600-3acd7bc0e43a"
      },
      "source": [
        "# plot loss history\n",
        "plt.figure(figsize=(10,5))\n",
        "plt.title('Loss Progress')\n",
        "plt.plot(loss_history['gen'], label='Gen. Loss')\n",
        "plt.plot(loss_history['dis'], label='Dis. Loss')\n",
        "plt.xlabel('batch count')\n",
        "plt.ylabel('Loss')\n",
        "plt.legend()\n",
        "plt.show()"
      ],
      "execution_count": 83,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x360 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "81QXiRat9HSW"
      },
      "source": [
        "# 가중치 저장\n",
        "path2models = './models/'\n",
        "os.makedirs(path2models, exist_ok=True)\n",
        "path2weights_gen = os.path.join(path2models, 'weights_gen.pt')\n",
        "path2weights_dis = os.path.join(path2models, 'weights_dis.pt')\n",
        "\n",
        "torch.save(model_gen.state_dict(), path2weights_gen)\n",
        "torch.save(model_dis.state_dict(), path2weights_dis)"
      ],
      "execution_count": 84,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Hjbu0upa9Cud"
      },
      "source": [
        "# 4. Generator이 생성한 이미지 확인"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "YQ7LcgQt8-vA",
        "outputId": "a7b8db68-7e88-404e-9682-02ee7df49fe4"
      },
      "source": [
        "# 가중치 불러오기\n",
        "weights = torch.load(path2weights_gen)\n",
        "model_gen.load_state_dict(weights)\n",
        "\n",
        "# evaluation mode\n",
        "model_gen.eval()\n",
        "\n",
        "# 가짜 이미지 생성\n",
        "with torch.no_grad():\n",
        "    fixed_noise = torch.randn(16, 100, device=device)\n",
        "    img_fake = model_gen(fixed_noise).detach().cpu()\n",
        "print(img_fake.shape)"
      ],
      "execution_count": 85,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "torch.Size([16, 1, 28, 28])\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 575
        },
        "id": "UQ9aR5Ku91ct",
        "outputId": "51850595-5748-4cd4-f509-d737106e1465"
      },
      "source": [
        "# 가짜 이미지 시각화\n",
        "plt.figure(figsize=(10,10))\n",
        "for ii in range(16):\n",
        "    plt.subplot(4,4,ii+1)\n",
        "    plt.imshow(to_pil_image(0.5*img_fake[ii]+0.5),cmap='gray')\n",
        "    plt.axis('off')"
      ],
      "execution_count": 86,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x720 with 16 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "LeRy_L_a-QEr"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}