Design_ResNet_by_using_CIFAR_10_+_Data_Augmentation.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Design CNN by using CIFAR-10 + Data Augmentation.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "authorship_tag": "ABX9TyORSTRMUzWPhlORCwrMLYLp",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "widgets": {
      "application/vnd.jupyter.widget-state+json": {
        "27f67571692145dc9fdb4a3679fa5e86": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HBoxModel",
          "state": {
            "_view_name": "HBoxView",
            "_dom_classes": [],
            "_model_name": "HBoxModel",
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "box_style": "",
            "layout": "IPY_MODEL_6238706b15d44342acd3096c2c12f30d",
            "_model_module": "@jupyter-widgets/controls",
            "children": [
              "IPY_MODEL_28dd28413f504ccebb2d70ec3ceb0aae",
              "IPY_MODEL_d29984d6421e49fb89b2df97fc0930b3"
            ]
          }
        },
        "6238706b15d44342acd3096c2c12f30d": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "state": {
            "_view_name": "LayoutView",
            "grid_template_rows": null,
            "right": null,
            "justify_content": null,
            "_view_module": "@jupyter-widgets/base",
            "overflow": null,
            "_model_module_version": "1.2.0",
            "_view_count": null,
            "flex_flow": null,
            "width": null,
            "min_width": null,
            "border": null,
            "align_items": null,
            "bottom": null,
            "_model_module": "@jupyter-widgets/base",
            "top": null,
            "grid_column": null,
            "overflow_y": null,
            "overflow_x": null,
            "grid_auto_flow": null,
            "grid_area": null,
            "grid_template_columns": null,
            "flex": null,
            "_model_name": "LayoutModel",
            "justify_items": null,
            "grid_row": null,
            "max_height": null,
            "align_content": null,
            "visibility": null,
            "align_self": null,
            "height": null,
            "min_height": null,
            "padding": null,
            "grid_auto_rows": null,
            "grid_gap": null,
            "max_width": null,
            "order": null,
            "_view_module_version": "1.2.0",
            "grid_template_areas": null,
            "object_position": null,
            "object_fit": null,
            "grid_auto_columns": null,
            "margin": null,
            "display": null,
            "left": null
          }
        },
        "28dd28413f504ccebb2d70ec3ceb0aae": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "state": {
            "_view_name": "ProgressView",
            "style": "IPY_MODEL_68c9e6a9a8a7413091ddc91d42d8b66b",
            "_dom_classes": [],
            "description": "",
            "_model_name": "FloatProgressModel",
            "bar_style": "success",
            "max": 170498071,
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "value": 170498071,
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "orientation": "horizontal",
            "min": 0,
            "description_tooltip": null,
            "_model_module": "@jupyter-widgets/controls",
            "layout": "IPY_MODEL_46ed3a3a436b484aa4c06146a7225b95"
          }
        },
        "d29984d6421e49fb89b2df97fc0930b3": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "state": {
            "_view_name": "HTMLView",
            "style": "IPY_MODEL_7663954fb62a42deac8c50424ffd2c5f",
            "_dom_classes": [],
            "description": "",
            "_model_name": "HTMLModel",
            "placeholder": "​",
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "value": " 170499072/? [00:08&lt;00:00, 21310329.46it/s]",
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "description_tooltip": null,
            "_model_module": "@jupyter-widgets/controls",
            "layout": "IPY_MODEL_9c2f0e1973f14fb28f58f8ec6012cab9"
          }
        },
        "68c9e6a9a8a7413091ddc91d42d8b66b": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "ProgressStyleModel",
          "state": {
            "_view_name": "StyleView",
            "_model_name": "ProgressStyleModel",
            "description_width": "initial",
            "_view_module": "@jupyter-widgets/base",
            "_model_module_version": "1.5.0",
            "_view_count": null,
            "_view_module_version": "1.2.0",
            "bar_color": null,
            "_model_module": "@jupyter-widgets/controls"
          }
        },
        "46ed3a3a436b484aa4c06146a7225b95": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "state": {
            "_view_name": "LayoutView",
            "grid_template_rows": null,
            "right": null,
            "justify_content": null,
            "_view_module": "@jupyter-widgets/base",
            "overflow": null,
            "_model_module_version": "1.2.0",
            "_view_count": null,
            "flex_flow": null,
            "width": null,
            "min_width": null,
            "border": null,
            "align_items": null,
            "bottom": null,
            "_model_module": "@jupyter-widgets/base",
            "top": null,
            "grid_column": null,
            "overflow_y": null,
            "overflow_x": null,
            "grid_auto_flow": null,
            "grid_area": null,
            "grid_template_columns": null,
            "flex": null,
            "_model_name": "LayoutModel",
            "justify_items": null,
            "grid_row": null,
            "max_height": null,
            "align_content": null,
            "visibility": null,
            "align_self": null,
            "height": null,
            "min_height": null,
            "padding": null,
            "grid_auto_rows": null,
            "grid_gap": null,
            "max_width": null,
            "order": null,
            "_view_module_version": "1.2.0",
            "grid_template_areas": null,
            "object_position": null,
            "object_fit": null,
            "grid_auto_columns": null,
            "margin": null,
            "display": null,
            "left": null
          }
        },
        "7663954fb62a42deac8c50424ffd2c5f": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "state": {
            "_view_name": "StyleView",
            "_model_name": "DescriptionStyleModel",
            "description_width": "",
            "_view_module": "@jupyter-widgets/base",
            "_model_module_version": "1.5.0",
            "_view_count": null,
            "_view_module_version": "1.2.0",
            "_model_module": "@jupyter-widgets/controls"
          }
        },
        "9c2f0e1973f14fb28f58f8ec6012cab9": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "state": {
            "_view_name": "LayoutView",
            "grid_template_rows": null,
            "right": null,
            "justify_content": null,
            "_view_module": "@jupyter-widgets/base",
            "overflow": null,
            "_model_module_version": "1.2.0",
            "_view_count": null,
            "flex_flow": null,
            "width": null,
            "min_width": null,
            "border": null,
            "align_items": null,
            "bottom": null,
            "_model_module": "@jupyter-widgets/base",
            "top": null,
            "grid_column": null,
            "overflow_y": null,
            "overflow_x": null,
            "grid_auto_flow": null,
            "grid_area": null,
            "grid_template_columns": null,
            "flex": null,
            "_model_name": "LayoutModel",
            "justify_items": null,
            "grid_row": null,
            "max_height": null,
            "align_content": null,
            "visibility": null,
            "align_self": null,
            "height": null,
            "min_height": null,
            "padding": null,
            "grid_auto_rows": null,
            "grid_gap": null,
            "max_width": null,
            "order": null,
            "_view_module_version": "1.2.0",
            "grid_template_areas": null,
            "object_position": null,
            "object_fit": null,
            "grid_auto_columns": null,
            "margin": null,
            "display": null,
            "left": null
          }
        }
      }
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/JINU6497/JINUrepo/blob/master/Design_ResNet_by_using_CIFAR_10_%2B_Data_Augmentation.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "qiSJDlgQa4PG"
      },
      "source": [
        "'''1. Module Import'''\n",
        "import numpy as np  # 선형대수 관련 모듈\n",
        "import matplotlib.pyplot as plt  # 결과 산출물을 시각화하는 모듈\n",
        " \n",
        "import torch\n",
        "import torch.nn as nn  # 인공 신경망 모델 설계에 필요한 함수들 모아 놓은 모듈 \n",
        "import torch.nn.functional as F # torch.nn 중에서도 자주 쓰이는 functional 함수들을 따로 F로 지정 \n",
        "from torchvision import transforms, datasets  # Computer vision 분야에서 자주 이용되는 torchvision 내의 함수 임포트"
      ],
      "execution_count": 1,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "eO5fvbkJ45En",
        "outputId": "96e9225c-3cfe-49a1-a851-c0f57d079000"
      },
      "source": [
        "'''2. 딥러닝 모델 설계에 활용하는 장비 확인'''\n",
        "if torch.cuda.is_available():\n",
        "  DEVICE=torch.device('cuda')\n",
        "else:\n",
        "  DEVICE=torch.device('cpu')\n",
        " \n",
        "print(\"Using pytorch version:\",torch.__version__,'Device:',DEVICE)"
      ],
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Using pytorch version: 1.9.0+cu102 Device: cpu\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "yiem0Cuc48Q7"
      },
      "source": [
        "BATCH_SIZE = 32\n",
        "EPOCHS = 10"
      ],
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 103,
          "referenced_widgets": [
            "27f67571692145dc9fdb4a3679fa5e86",
            "6238706b15d44342acd3096c2c12f30d",
            "28dd28413f504ccebb2d70ec3ceb0aae",
            "d29984d6421e49fb89b2df97fc0930b3",
            "68c9e6a9a8a7413091ddc91d42d8b66b",
            "46ed3a3a436b484aa4c06146a7225b95",
            "7663954fb62a42deac8c50424ffd2c5f",
            "9c2f0e1973f14fb28f58f8ec6012cab9"
          ]
        },
        "id": "k6ZDbdtW49-D",
        "outputId": "2bd59ef4-a619-4fc9-c421-38aa2b170b9a"
      },
      "source": [
        "'''3. Data Augmentation 적용한 CIFAR-10 데이터 다운로드(Train과 Test set 분리)'''\n",
        "train_dataset=datasets.CIFAR10(root=\"../data/Fashion,MNIST\",train=True,download=True,\n",
        "                               transform=transforms.Compose([transforms.RandomHorizontalFlip(),\n",
        "                                                             transforms.ToTensor(),\n",
        "                                                             transforms.Normalize((0.5,0.5,0.5),\n",
        "                                                             (0.5,0.5,0.5))]))\n",
        "test_dataset=datasets.CIFAR10(root=\"../data/Fashion,MNIST\",train=False,\n",
        "                               transform=transforms.Compose([transforms.RandomHorizontalFlip(),\n",
        "                                                             transforms.ToTensor(),\n",
        "                                                             transforms.Normalize((0.5,0.5,0.5),\n",
        "                                                             (0.5,0.5,0.5))]))\n",
        " \n",
        "train_loader=torch.utils.data.DataLoader(dataset=train_dataset,batch_size=BATCH_SIZE,shuffle=True)\n",
        "test_loader=torch.utils.data.DataLoader(dataset=test_dataset,batch_size=BATCH_SIZE,shuffle=False)"
      ],
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Downloading https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz to ../data/Fashion,MNIST/cifar-10-python.tar.gz\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "27f67571692145dc9fdb4a3679fa5e86",
              "version_minor": 0,
              "version_major": 2
            },
            "text/plain": [
              "HBox(children=(FloatProgress(value=0.0, max=170498071.0), HTML(value='')))"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "\n",
            "Extracting ../data/Fashion,MNIST/cifar-10-python.tar.gz to ../data/Fashion,MNIST\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "BgoaktPURmS8"
      },
      "source": [
        "transforms.Compose()는 불러오는 이미지 데이터 전처리 및 Augmentation을 다양하게 적용할 때 이용하는 method. 즉, Compose의 괄호 안에 있는 처리 과정을 거친 데이터를 불러오는 것을 의미.\n",
        "RandomHorizontalFlip은 50%확률로 좌우 반전하는 것을 의미\n",
        "Normalize는 ToTensor()로 전환된 이미지에 대해 또 다른 정규화를 진행하는 것을 의미. 정규화를 진행할 때 평균과 표준편차가 필요한데, RGB순으로 평균을 0.5씩 적용하고, 다음 괄호에서는 표준편차를 0.5씩 적용하는 것"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "TeNyspGA5U5L",
        "outputId": "2d375eda-5299-4956-a584-48f704f0ac8b"
      },
      "source": [
        "'''4. 데이터 확인'''\n",
        "for (X_train, Y_train) in train_loader:\n",
        "  print('X_train:',X_train.size(),'type:',X_train.type())\n",
        "  print('Y_train:',Y_train.size(),'type:',Y_train.type())\n",
        "  break"
      ],
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "X_train: torch.Size([32, 3, 32, 32]) type: torch.FloatTensor\n",
            "Y_train: torch.Size([32]) type: torch.LongTensor\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 274
        },
        "id": "UFvzEC3y5a6E",
        "outputId": "31a91df9-83d9-4538-d498-dbd84e4b79fd"
      },
      "source": [
        "pltsize=1\n",
        "plt.figure(figsize=(10*pltsize,pltsize))\n",
        "for i in range(10):\n",
        "  plt.subplot(1,10,i+1)\n",
        "  plt.axis('off')\n",
        "  plt.imshow(np.transpose(X_train[i],(1,2,0)))\n",
        "  plt.title('Class:'+str(Y_train[i].item()))"
      ],
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x72 with 10 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jRh42qvz5uwI"
      },
      "source": [
        "'''5. Convolutional Neural Network(CNN) 모델 설계하기'''\n",
        "\n",
        "class BasicBlock(nn.Module):\n",
        "  def __init__(self, in_planes, planes, stride=1):\n",
        "    super(BasicBlock, self).__init__()\n",
        "    self.conv1=nn.Conv2d(in_planes,planes,kernel_size=3,stride=stride,padding=1,bias=False)\n",
        "    self.bn1=nn.BatchNorm2d(planes)\n",
        "    self.conv2=nn.Conv2d(planes,planes,kernel_size=3,stride=1,padding=1,bias=False)\n",
        "    self.bn2=nn.BatchNorm2d(planes)\n",
        "\n",
        "    self.shortcut=nn.Sequential()\n",
        "    if stride!=1 or in_planes!=planes:\n",
        "      self.shortcut=nn.Sequential(\n",
        "          nn.Conv2d(in_planes,planes,kernel_size=1,stride=stride,bias=False),\n",
        "          nn.BatchNorm2d(planes))\n",
        "      \n",
        "  def forward(self,x):\n",
        "    out=F.relu(self.bn1(self.conv1(x)))\n",
        "    out=self.bn2(self.conv2(out))\n",
        "    out+=self.shortcut(x)\n",
        "    out=F.relu(out)\n",
        "    return out\n",
        "\n",
        "class ResNet(nn.Module):\n",
        "  def __init__(self, num_classes=10):\n",
        "    super(ResNet,self).__init__()\n",
        "    self.in_planes=16\n",
        "\n",
        "    self.conv1=nn.Conv2d(3,16,kernel_size=3,stride=1,padding=1,bias=False)\n",
        "    self.bn1=nn.BatchNorm2d(16)\n",
        "    self.layer1=self._make_layer(16,2,stride=1)\n",
        "    self.layer2=self._make_layer(32,2,stride=2)\n",
        "    self.layer3=self._make_layer(64,2,stride=2)\n",
        "    self.linear=nn.Linear(64,num_classes)\n",
        "\n",
        "  def _make_layer(self, planes, num_blocks,stride):\n",
        "    strides=[stride]+[1]*(num_blocks-1)\n",
        "    layers=[]\n",
        "    for stride in strides:\n",
        "      layers.append(BasicBlock(self.in_planes,planes,stride))\n",
        "      self.in_planes=planes\n",
        "    return nn.Sequential(*layers)\n",
        "\n",
        "  def forward(self,x):\n",
        "    out=F.relu(self.bn1(self.conv1(x)))\n",
        "    out=self.layer1(out)\n",
        "    out=self.layer2(out)\n",
        "    out=self.layer3(out)\n",
        "    out=F.avg_pool2d(out,8)\n",
        "    out=out.view(out.size(0),-1)\n",
        "    out=self.linear(out)\n",
        "    return out"
      ],
      "execution_count": 48,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "kymDVj_B9z0y"
      },
      "source": [
        "self.conv1=nn.conv2d(in_channels=3, out_channels=8, kernel_size=3, padding=1): \n",
        "\n",
        "in_channel: 이미지에 직접 Convolution 연산을 진행하는 Filter의 크기는 상관 없지만, 채널 수를 이미지의 채널 수와 동일하게 맞춰야 한다. 동일한 위치에 있으며 앞뒤로 존재하는 RGB 픽셀에 대해 동시에 Convolution을 진행해야 하기 때문. 따라서 이미지의 채널 수와 동일하게 채널 수를 3으로 설정.\n",
        "\n",
        "out_channel은 8이므로 해당 Convolution 연산 결과는 Depth가 8인 Feature map이 생성 될 것.\n",
        "\n",
        "kernel_size는 filter의 크기를 설정해주는 부분.\n",
        "\n",
        "log_softmax()를 이용하여 0부터 9까지 총 10가지 경우의 수 중 하나로 분류하는 일을 확률 값 계산.일반적인 softmax가 아닌 log_softmax()를 이용하는 이유는 MLP모델이 Back propagation을 진행할 때 loss 값에 대한 gradient값을 좀 더 원활하게 계산할 수 있기 때문에. log 함수 그래프의 기울기가 부드럽게 변화하는 것을 생각해보면 직관적으로 이해할 수 있다.\n",
        "\n",
        "\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "W_wTbz99Ha35",
        "outputId": "5c3212d7-8242-49f8-d419-38c0b431751e"
      },
      "source": [
        "'''6. Optimizer, Objective Fundtion 설정'''\n",
        " \n",
        "model=ResNet().to(DEVICE)\n",
        "optimizer = torch.optim.Adam(model.parameters(),lr=0.001)\n",
        "criterion=nn.CrossEntropyLoss()\n",
        " \n",
        "print(model)"
      ],
      "execution_count": 49,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "ResNet(\n",
            "  (conv1): Conv2d(3, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
            "  (bn1): BatchNorm2d(16, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "  (layer1): Sequential(\n",
            "    (0): BasicBlock(\n",
            "      (conv1): Conv2d(16, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
            "      (bn1): BatchNorm2d(16, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "      (conv2): Conv2d(16, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
            "      (bn2): BatchNorm2d(16, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "      (shortcut): Sequential()\n",
            "    )\n",
            "    (1): BasicBlock(\n",
            "      (conv1): Conv2d(16, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
            "      (bn1): BatchNorm2d(16, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "      (conv2): Conv2d(16, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
            "      (bn2): BatchNorm2d(16, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "      (shortcut): Sequential()\n",
            "    )\n",
            "  )\n",
            "  (layer2): Sequential(\n",
            "    (0): BasicBlock(\n",
            "      (conv1): Conv2d(16, 32, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
            "      (bn1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "      (conv2): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
            "      (bn2): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "      (shortcut): Sequential(\n",
            "        (0): Conv2d(16, 32, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
            "        (1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "      )\n",
            "    )\n",
            "    (1): BasicBlock(\n",
            "      (conv1): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
            "      (bn1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "      (conv2): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
            "      (bn2): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "      (shortcut): Sequential()\n",
            "    )\n",
            "  )\n",
            "  (layer3): Sequential(\n",
            "    (0): BasicBlock(\n",
            "      (conv1): Conv2d(32, 64, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
            "      (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "      (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
            "      (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "      (shortcut): Sequential(\n",
            "        (0): Conv2d(32, 64, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
            "        (1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "      )\n",
            "    )\n",
            "    (1): BasicBlock(\n",
            "      (conv1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
            "      (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "      (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
            "      (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "      (shortcut): Sequential()\n",
            "    )\n",
            "  )\n",
            "  (linear): Linear(in_features=64, out_features=10, bias=True)\n",
            ")\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "gjofLjfdIfX8"
      },
      "source": [
        "'''7. CNN모델 학습 진행하면서 Train데이터에 대한 모델의 성능 확인하는 함수 정의'''\n",
        "def train(model, train_loader, optimizer, log_interval):\n",
        "  model.train()\n",
        "  for batch_idx,(image,label) in enumerate(train_loader):\n",
        "    image=image.to(DEVICE)\n",
        "    label=label.to(DEVICE)\n",
        "    optimizer.zero_grad()\n",
        "    output= model(image)\n",
        "    loss = criterion(output, label)\n",
        "    loss.backward()\n",
        "    optimizer.step()\n",
        " \n",
        "    if batch_idx %log_interval ==0:\n",
        "      print(\"Train Epoch:{}[{}/{}({:.0f}%)]\\tTrain Loss:{:.6f}\".format(Epoch, batch_idx*len(image),\n",
        "                                                                       len(train_loader.dataset),100.*batch_idx/len(train_loader),loss.item()))"
      ],
      "execution_count": 50,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "KGrixKBdJFWL"
      },
      "source": [
        "'''8. 학습 과정 속에서 Test 데이터에 대한 모델 성능 확인하는 함수 정의'''\n",
        " \n",
        "def evaluate(model, test_loader):\n",
        "  model.eval()\n",
        "  test_loss=0\n",
        "  correct=0\n",
        "  with torch.no_grad():\n",
        "    for image, label in test_loader:\n",
        "      image=image.to(DEVICE)\n",
        "      label=label.to(DEVICE)\n",
        "      output = model(image)\n",
        " \n",
        "      test_loss += criterion(output, label).item()\n",
        "      prediction=output.max(1,keepdim=True)[1]\n",
        "      correct += prediction.eq(label.view_as(prediction)).sum().item()\n",
        "\n",
        "  test_loss/=len(test_loader.dataset)\n",
        "  test_accuracy=100.*correct/len(test_loader.dataset)\n",
        "\n",
        "  return test_loss, test_accuracy"
      ],
      "execution_count": 51,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "83v0IHM7LnJN",
        "outputId": "642083d3-00e4-4ab1-aa30-21fc26cfa704"
      },
      "source": [
        "'''9. CNN학습을 실행하면서 Test set의 Reconstruction Error 확인'''\n",
        "for Epoch in range(1,EPOCHS+1):\n",
        "  train(model, train_loader, optimizer, log_interval = 200)\n",
        "  test_loss, test_accuracy = evaluate(model, test_loader)\n",
        "  print(\"\\n[EPOCH:{}], \\tTest Loss: {:.4f}, \\tTest Accuracy:{:.2f}% \\n\".format(Epoch,test_loss, test_accuracy))"
      ],
      "execution_count": 52,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train Epoch:1[0/50000(0%)]\tTrain Loss:2.506530\n",
            "Train Epoch:1[6400/50000(13%)]\tTrain Loss:1.331616\n",
            "Train Epoch:1[12800/50000(26%)]\tTrain Loss:1.177642\n",
            "Train Epoch:1[19200/50000(38%)]\tTrain Loss:1.179705\n",
            "Train Epoch:1[25600/50000(51%)]\tTrain Loss:1.026020\n",
            "Train Epoch:1[32000/50000(64%)]\tTrain Loss:1.121610\n",
            "Train Epoch:1[38400/50000(77%)]\tTrain Loss:0.909776\n",
            "Train Epoch:1[44800/50000(90%)]\tTrain Loss:1.524731\n",
            "\n",
            "[EPOCH:1], \tTest Loss: 0.0323, \tTest Accuracy:62.46% \n",
            "\n",
            "Train Epoch:2[0/50000(0%)]\tTrain Loss:0.837573\n",
            "Train Epoch:2[6400/50000(13%)]\tTrain Loss:0.811295\n",
            "Train Epoch:2[12800/50000(26%)]\tTrain Loss:0.960691\n",
            "Train Epoch:2[19200/50000(38%)]\tTrain Loss:0.869237\n",
            "Train Epoch:2[25600/50000(51%)]\tTrain Loss:1.061998\n",
            "Train Epoch:2[32000/50000(64%)]\tTrain Loss:0.856571\n",
            "Train Epoch:2[38400/50000(77%)]\tTrain Loss:0.652171\n",
            "Train Epoch:2[44800/50000(90%)]\tTrain Loss:0.753674\n",
            "\n",
            "[EPOCH:2], \tTest Loss: 0.0272, \tTest Accuracy:69.38% \n",
            "\n",
            "Train Epoch:3[0/50000(0%)]\tTrain Loss:1.003039\n",
            "Train Epoch:3[6400/50000(13%)]\tTrain Loss:0.569603\n",
            "Train Epoch:3[12800/50000(26%)]\tTrain Loss:0.545357\n",
            "Train Epoch:3[19200/50000(38%)]\tTrain Loss:0.984425\n",
            "Train Epoch:3[25600/50000(51%)]\tTrain Loss:0.606043\n",
            "Train Epoch:3[32000/50000(64%)]\tTrain Loss:0.770155\n",
            "Train Epoch:3[38400/50000(77%)]\tTrain Loss:0.995988\n",
            "Train Epoch:3[44800/50000(90%)]\tTrain Loss:0.496590\n",
            "\n",
            "[EPOCH:3], \tTest Loss: 0.0274, \tTest Accuracy:70.41% \n",
            "\n",
            "Train Epoch:4[0/50000(0%)]\tTrain Loss:0.603680\n",
            "Train Epoch:4[6400/50000(13%)]\tTrain Loss:0.642444\n",
            "Train Epoch:4[12800/50000(26%)]\tTrain Loss:0.657148\n",
            "Train Epoch:4[19200/50000(38%)]\tTrain Loss:0.865965\n",
            "Train Epoch:4[25600/50000(51%)]\tTrain Loss:0.700997\n",
            "Train Epoch:4[32000/50000(64%)]\tTrain Loss:0.655059\n",
            "Train Epoch:4[38400/50000(77%)]\tTrain Loss:0.746070\n",
            "Train Epoch:4[44800/50000(90%)]\tTrain Loss:0.501734\n",
            "\n",
            "[EPOCH:4], \tTest Loss: 0.0203, \tTest Accuracy:77.48% \n",
            "\n",
            "Train Epoch:5[0/50000(0%)]\tTrain Loss:0.899854\n",
            "Train Epoch:5[6400/50000(13%)]\tTrain Loss:0.540259\n",
            "Train Epoch:5[12800/50000(26%)]\tTrain Loss:0.831625\n",
            "Train Epoch:5[19200/50000(38%)]\tTrain Loss:0.620340\n",
            "Train Epoch:5[25600/50000(51%)]\tTrain Loss:0.617964\n",
            "Train Epoch:5[32000/50000(64%)]\tTrain Loss:0.654510\n",
            "Train Epoch:5[38400/50000(77%)]\tTrain Loss:0.961067\n",
            "Train Epoch:5[44800/50000(90%)]\tTrain Loss:0.561678\n",
            "\n",
            "[EPOCH:5], \tTest Loss: 0.0193, \tTest Accuracy:78.51% \n",
            "\n",
            "Train Epoch:6[0/50000(0%)]\tTrain Loss:0.445512\n",
            "Train Epoch:6[6400/50000(13%)]\tTrain Loss:0.652808\n",
            "Train Epoch:6[12800/50000(26%)]\tTrain Loss:0.695730\n",
            "Train Epoch:6[19200/50000(38%)]\tTrain Loss:0.603819\n",
            "Train Epoch:6[25600/50000(51%)]\tTrain Loss:0.313107\n",
            "Train Epoch:6[32000/50000(64%)]\tTrain Loss:0.591481\n",
            "Train Epoch:6[38400/50000(77%)]\tTrain Loss:0.399521\n",
            "Train Epoch:6[44800/50000(90%)]\tTrain Loss:0.643646\n",
            "\n",
            "[EPOCH:6], \tTest Loss: 0.0177, \tTest Accuracy:80.24% \n",
            "\n",
            "Train Epoch:7[0/50000(0%)]\tTrain Loss:0.414257\n",
            "Train Epoch:7[6400/50000(13%)]\tTrain Loss:0.802256\n",
            "Train Epoch:7[12800/50000(26%)]\tTrain Loss:0.463902\n",
            "Train Epoch:7[19200/50000(38%)]\tTrain Loss:0.574553\n",
            "Train Epoch:7[25600/50000(51%)]\tTrain Loss:0.474427\n",
            "Train Epoch:7[32000/50000(64%)]\tTrain Loss:0.289864\n",
            "Train Epoch:7[38400/50000(77%)]\tTrain Loss:0.629136\n",
            "Train Epoch:7[44800/50000(90%)]\tTrain Loss:0.435832\n",
            "\n",
            "[EPOCH:7], \tTest Loss: 0.0177, \tTest Accuracy:80.69% \n",
            "\n",
            "Train Epoch:8[0/50000(0%)]\tTrain Loss:0.475942\n",
            "Train Epoch:8[6400/50000(13%)]\tTrain Loss:0.564668\n",
            "Train Epoch:8[12800/50000(26%)]\tTrain Loss:0.424613\n",
            "Train Epoch:8[19200/50000(38%)]\tTrain Loss:0.274513\n",
            "Train Epoch:8[25600/50000(51%)]\tTrain Loss:0.481067\n",
            "Train Epoch:8[32000/50000(64%)]\tTrain Loss:0.345784\n",
            "Train Epoch:8[38400/50000(77%)]\tTrain Loss:0.583298\n",
            "Train Epoch:8[44800/50000(90%)]\tTrain Loss:0.419711\n",
            "\n",
            "[EPOCH:8], \tTest Loss: 0.0169, \tTest Accuracy:81.36% \n",
            "\n",
            "Train Epoch:9[0/50000(0%)]\tTrain Loss:0.291198\n",
            "Train Epoch:9[6400/50000(13%)]\tTrain Loss:0.454288\n",
            "Train Epoch:9[12800/50000(26%)]\tTrain Loss:0.585595\n",
            "Train Epoch:9[19200/50000(38%)]\tTrain Loss:0.558576\n",
            "Train Epoch:9[25600/50000(51%)]\tTrain Loss:0.474324\n",
            "Train Epoch:9[32000/50000(64%)]\tTrain Loss:0.315912\n",
            "Train Epoch:9[38400/50000(77%)]\tTrain Loss:0.360150\n",
            "Train Epoch:9[44800/50000(90%)]\tTrain Loss:0.594107\n",
            "\n",
            "[EPOCH:9], \tTest Loss: 0.0166, \tTest Accuracy:81.68% \n",
            "\n",
            "Train Epoch:10[0/50000(0%)]\tTrain Loss:0.352208\n",
            "Train Epoch:10[6400/50000(13%)]\tTrain Loss:0.733816\n",
            "Train Epoch:10[12800/50000(26%)]\tTrain Loss:0.478129\n",
            "Train Epoch:10[19200/50000(38%)]\tTrain Loss:0.507405\n",
            "Train Epoch:10[25600/50000(51%)]\tTrain Loss:0.272095\n",
            "Train Epoch:10[32000/50000(64%)]\tTrain Loss:0.420388\n",
            "Train Epoch:10[38400/50000(77%)]\tTrain Loss:0.486020\n",
            "Train Epoch:10[44800/50000(90%)]\tTrain Loss:0.524071\n",
            "\n",
            "[EPOCH:10], \tTest Loss: 0.0166, \tTest Accuracy:82.14% \n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "M-T_Ma2NlQjm"
      },
      "source": [
        "ResNet의 특징은 Shortcut을 정의하는 부분이 있음. Shortcut은 기존의 값과 Convolution 및 Batch Normalization한 결과를 더하는 과정을 의미. 이때 Stride값이 1이 아니거나 in_planes가 planes와 같지 않다면, 즉 두 번째 블록부터 적용되는 Shortcut을 정의. \n",
        "\n",
        "_make_layer 부분에서는 인자 값으로 주어지는 stride를 이용해 stride 범위를 BasicBlock 마다 설정할 수 있도록 정의. 이후, strides의 범위인 stride값을 통해 생성한 BasicBlock의 결과값을 리스트에 추가. 이후, 반복문을 실행하면서 self.in_planes값을 매번 업데이트해 BasicBlock을 새로 생성하기 위해 self.in_planes를 planes 값으로 업데이트함. 이 과정은 Shortcut을 계산하기 위함."
      ]
    }
  ]
}