[실습04]_심층신경망_훈련.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "[실습04] 심층신경망 훈련.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "authorship_tag": "ABX9TyPbUKZEvqTuS8JgLMSgcknX",
      "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/yeahxne/Deep_Learning_2022/blob/main/%5B%EC%8B%A4%EC%8A%B504%5D_%EC%8B%AC%EC%B8%B5%EC%8B%A0%EA%B2%BD%EB%A7%9D_%ED%9B%88%EB%A0%A8.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### CIFAR10을 CNN으로 학습하기. 여러 학습 방법으로 비교"
      ],
      "metadata": {
        "id": "w_cGXvFwsfC-"
      }
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "id": "I_jhgtCTr4sT"
      },
      "outputs": [],
      "source": [
        "import numpy as np\n",
        "import torch\n",
        "import torch.nn as nn\n",
        "import torch.optim as optim\n",
        "import torch.nn.init as init\n",
        "import torchvision.datasets as dset\n",
        "import torchvision.transforms as transforms\n",
        "from torch.utils.data import DataLoader\n",
        "from torch.autograd import Variable\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "# (8)learning rate decay\n",
        "from torch.optim import lr_scheduler\n",
        "\n",
        "batch_size=16\n",
        "learning_rate=0.002\n",
        "num_epoch=1\n",
        "# num_epoch=60"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### 1. CIFAR10 train,test dataset 가져오기(163MB)\n",
        "MNIST: 11MB"
      ],
      "metadata": {
        "id": "kCI2amqGsrYU"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "cifar_train=dset.CIFAR10(\"CIFAR10/\",train=True,transform=transforms.ToTensor(),target_transform=None,download=True)\n",
        "cifar_test=dset.CIFAR10(\"CIFAR10/\",train=False,transform=transforms.ToTensor(),target_transform=None,download=True)\n",
        "\n",
        "# # (2) Data augmentation\n",
        "# cifar_train=dset.CIFAR10(\"CIFAR10/\", train=True,\n",
        "#                         transform=transforms.Compose([\n",
        "#                             transforms.Resize(36),    # torchvision.transforms.Scale was deprecated in favor of torchvision.transforms.Resize so use the latter transformation in current releases.\n",
        "#                             transforms.CenterCrop(32),\n",
        "#                             transforms.RandomHorizontalFlip(),\n",
        "#                             transforms.Lambda(lambda x: x.rotate(90)),\n",
        "#                             transforms.ToTensor()\n",
        "#                         ]))\n",
        "\n",
        "# # (4) Data Normalization\n",
        "# cifar_train=dset.CIFAR10(\"CIFAR10/\", train=True,\n",
        "#                         transform=transforms.Compose([\n",
        "#                             transforms.ToTensor(),\n",
        "#                             transforms.Normalize(mean=(0.5,0.5,0.5), std=(0.5,0.5,0.5)),\n",
        "#                         ])\n",
        "#                         , target_transform=None, download=False)\n",
        "\n",
        "# cifar_test=dset.CIFAR10(\"CIFAR10/\", train=False,\n",
        "#                         transform=transforms.Compose([\n",
        "#                             transforms.ToTensor(),\n",
        "#                             transforms.Normalize(mean=(0.5,0.5,0.5), std=(0.5,0.5,0.5)),\n",
        "#                         ])\n",
        "#                         , target_transform=None, download=False)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "jiTO9k-gssiV",
        "outputId": "0d01f280-0775-4b4f-a2ea-366e43026ef6"
      },
      "execution_count": 25,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Files already downloaded and verified\n",
            "Files already downloaded and verified\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### 2. 대략적인 데이터 형태"
      ],
      "metadata": {
        "id": "0_3jX8_lsyv_"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "print(\"cifar_train 길이:\", len(cifar_train))\n",
        "print(\"cifar_test 길이:\", len(cifar_test))\n",
        "\n",
        "# 데이터 하나 형태\n",
        "image, label = cifar_train.__getitem__(1)  # 1번째 데이터\n",
        "print(\"image data 형태:\", image.size())\n",
        "print(\"label:\", label)\n",
        "\n",
        "# 그리기\n",
        "img = image.numpy()  # image 타입을 numpy로 변환(3,32,32)\n",
        "\n",
        "# (3,32,32) -> (32,32,3)\n",
        "r, g, b = img[0,:,:], img[1,:,:], img[2,:,:]\n",
        "# img = img.reshape(img.shape[1], img.shape[2], img.shape[0])\n",
        "img2 = np.zeros((img.shape[1], img.shape[2], img.shape[0]))\n",
        "img2[:,:,0], img2[:,:,1], img2[:,:,2] = r, g, b\n",
        "\n",
        "plt.title(\"label: %d\" %label)\n",
        "plt.imshow(img2,interpolation='bicubic')\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 354
        },
        "id": "dMDNvVnLsv78",
        "outputId": "e0117084-a9e1-4673-91f8-9f66e61819fe"
      },
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "cifar_train 길이: 50000\n",
            "cifar_test 길이: 10000\n",
            "image data 형태: torch.Size([3, 32, 32])\n",
            "label: 9\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "def ComputeAccr(dloader, imodel):\n",
        "    correct = 0\n",
        "    total = 0\n",
        "    \n",
        "    for j, [imgs, labels] in enumerate(dloader):  # batch_size 만큼\n",
        "        img = Variable(imgs,volatile=True).cuda()  # x\n",
        "        # label = Variable(labels)  # y\n",
        "        label = Variable(labels).cuda()\n",
        "        \n",
        "        output = imodel.forward(img)  # forward prop.\n",
        "        _, output_index = torch.max(output, 1)\n",
        "        \n",
        "        total += label.size(0)\n",
        "        correct += (output_index == label).sum().float()\n",
        "    print(\"Accuracy of Test Data: {}\".format(100*correct/total))"
      ],
      "metadata": {
        "id": "xlyAeaLCs0Rj"
      },
      "execution_count": 4,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# === 3. 데이터 로드함수 ===\n",
        "train_loader=torch.utils.data.DataLoader(list(cifar_train)[:], batch_size=batch_size, shuffle=True, num_workers=2, drop_last=True)\n",
        "test_loader=torch.utils.data.DataLoader(cifar_test, batch_size=batch_size, shuffle=False, num_workers=2, drop_last=True)\n",
        "\n",
        "# === 4. 모델 선언 ===\n",
        "class CNN(nn.Module):\n",
        "    def __init__(self):\n",
        "        super(CNN,self).__init__()\n",
        "        self.layer=nn.Sequential(\n",
        "            nn.Conv2d(3,16,3,padding=1),\n",
        "            nn.ReLU(),\n",
        "            # nn.Dropout2d(0.2),  # (1) drop out\n",
        "            # nn.BatchNorm2d(16),  # (5) Batch normalization\n",
        "            nn.Conv2d(16,32,3,padding=1),\n",
        "            nn.ReLU(),\n",
        "            # nn.Dropout2d(0.2),\n",
        "            # nn.BatchNorm2d(32),\n",
        "            nn.MaxPool2d(2,2),\n",
        "            nn.Conv2d(32,64,3,padding=1),\n",
        "            nn.ReLU(),\n",
        "            # nn.Dropout2d(0.2),\n",
        "            # nn.BatchNorm2d(64),\n",
        "            nn.MaxPool2d(2,2)\n",
        "        )\n",
        "        self.fc_layer=nn.Sequential(\n",
        "            nn.Linear(64*8*8, 100),\n",
        "            nn.ReLU(),\n",
        "            # nn.Dropout2d(0.2),\n",
        "            # nn.BatchNorm1d(100),\n",
        "            nn.Linear(100,10)\n",
        "        )\n",
        "        \n",
        "        # # (3) weight initialization\n",
        "        # for m in self.modules():\n",
        "        #     if isinstance(m, nn.Conv2d):\n",
        "        #         init.kaiming_normal(m.weight.data)  # ReLU 일 때\n",
        "        #         m.bias.data.fill_(0)\n",
        "        #     if isinstance(m, nn.Linear):\n",
        "        #         init.kaiming_normal(m.weight.data)\n",
        "        #         m.bias.data.fill_(0)\n",
        "                \n",
        "    def forward(self,x):\n",
        "        out=self.layer(x)\n",
        "        out=out.view(batch_size,-1)\n",
        "        out=self.fc_layer(out)\n",
        "        \n",
        "        return out\n",
        "\n",
        "model=CNN().cuda()"
      ],
      "metadata": {
        "id": "ueXMFG1PtX_u"
      },
      "execution_count": 32,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# === 5. loss, optimizer ===\n",
        "loss_func=nn.CrossEntropyLoss()\n",
        "optimizer=torch.optim.SGD(model.parameters(), lr=learning_rate)\n",
        "\n",
        "# # (6) Adam optimizer\n",
        "# optimizer=torch.optim.Adam(model.parameters(), lr=learning_rate)\n",
        "\n",
        "# # (7) learning rate decay\n",
        "# scheduler=lr_scheduler.StepLR(optimizer, step_size=20, gamma=0.2)\n",
        "\n",
        "# ===6. 학습 ===\n",
        "for i in range(num_epoch):\n",
        "    for j,[image, label] in enumerate(train_loader):\n",
        "        x=Variable(image).cuda()\n",
        "        y_=Variable(label).cuda()\n",
        "        \n",
        "        optimizer.zero_grad()\n",
        "        output=model.forward(x)\n",
        "        loss=loss_func(output,y_)\n",
        "        loss.backward()\n",
        "        optimizer.step()\n",
        "        \n",
        "        if j%1000==0:\n",
        "            print(j,loss)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "KMnBTwaktdMt",
        "outputId": "111503f9-81d8-4a71-bef6-d9492c50cb86"
      },
      "execution_count": 35,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "0 tensor(1.0727, device='cuda:0', grad_fn=<NllLossBackward0>)\n",
            "1000 tensor(1.1369, device='cuda:0', grad_fn=<NllLossBackward0>)\n",
            "2000 tensor(0.6121, device='cuda:0', grad_fn=<NllLossBackward0>)\n",
            "3000 tensor(1.4179, device='cuda:0', grad_fn=<NllLossBackward0>)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### (0) Naive Test"
      ],
      "metadata": {
        "id": "1lyslc2kuT2w"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "ComputeAccr(test_loader, model)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "pv_3uDwJuok6",
        "outputId": "021fee64-9241-41ae-88ec-f4febb360c85"
      },
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/ipykernel_launcher.py:6: UserWarning: volatile was removed and now has no effect. Use `with torch.no_grad():` instead.\n",
            "  \n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Accuracy of Test Data: 16.03999900817871\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### (1) drop out"
      ],
      "metadata": {
        "id": "dde7DrtDuUCq"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "ComputeAccr(test_loader, model)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "MP49MJr0upEU",
        "outputId": "222f6f2b-7363-4931-eed6-7835219d542a"
      },
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/ipykernel_launcher.py:6: UserWarning: volatile was removed and now has no effect. Use `with torch.no_grad():` instead.\n",
            "  \n",
            "/usr/local/lib/python3.7/dist-packages/torch/nn/functional.py:1320: UserWarning: dropout2d: Received a 2-D input to dropout2d, which is deprecated and will result in an error in a future release. To retain the behavior and silence this warning, please use dropout instead. Note that dropout2d exists to provide channel-wise dropout on inputs with 2 spatial dimensions, a channel dimension, and an optional batch dimension (i.e. 3D or 4D inputs).\n",
            "  warnings.warn(warn_msg)\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Accuracy of Test Data: 17.739999771118164\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### (2) Data augmentation"
      ],
      "metadata": {
        "id": "SMSXXVt6uUGS"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "ComputeAccr(test_loader, model)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Ngp1t25mupjy",
        "outputId": "a9c62f59-31d4-402b-cd8f-7e884fd6bcaa"
      },
      "execution_count": 14,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/ipykernel_launcher.py:6: UserWarning: volatile was removed and now has no effect. Use `with torch.no_grad():` instead.\n",
            "  \n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Accuracy of Test Data: 11.279999732971191\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### (3) Weight initialization"
      ],
      "metadata": {
        "id": "li2mgrL7uUJ6"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "ComputeAccr(test_loader, model)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "yUhh7ZULuqJe",
        "outputId": "e75daa4b-ad3c-4426-c150-212e96119b1d"
      },
      "execution_count": 18,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/ipykernel_launcher.py:6: UserWarning: volatile was removed and now has no effect. Use `with torch.no_grad():` instead.\n",
            "  \n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Accuracy of Test Data: 45.53999710083008\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### (4) Data Normalization"
      ],
      "metadata": {
        "id": "e17XHTxCuUPo"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "ComputeAccr(test_loader, model)"
      ],
      "metadata": {
        "id": "zHaBxf-TuqkY",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "c18bce97-8573-459f-cfe0-38f6a6254f7a"
      },
      "execution_count": 24,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/ipykernel_launcher.py:6: UserWarning: volatile was removed and now has no effect. Use `with torch.no_grad():` instead.\n",
            "  \n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Accuracy of Test Data: 29.670000076293945\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### (5) Batch normalization"
      ],
      "metadata": {
        "id": "v-ZbkOItue85"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "ComputeAccr(test_loader, model)"
      ],
      "metadata": {
        "id": "ke_BXAn8urAE",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "7fbf35c6-16ea-4d38-e788-c1fd3ed41896"
      },
      "execution_count": 31,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/ipykernel_launcher.py:6: UserWarning: volatile was removed and now has no effect. Use `with torch.no_grad():` instead.\n",
            "  \n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Accuracy of Test Data: 61.90999984741211\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### (6) Adam optimizer"
      ],
      "metadata": {
        "id": "2dPjEP2yufNC"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "ComputeAccr(test_loader, model)"
      ],
      "metadata": {
        "id": "GvDNOR-iuroS",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "7b1c7e9c-54b0-47bd-9e9f-1659b5303ba8"
      },
      "execution_count": 34,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/ipykernel_launcher.py:6: UserWarning: volatile was removed and now has no effect. Use `with torch.no_grad():` instead.\n",
            "  \n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Accuracy of Test Data: 59.57999801635742\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### (7) learning rate decay"
      ],
      "metadata": {
        "id": "G8kwS9oMufWf"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "ComputeAccr(test_loader, model)"
      ],
      "metadata": {
        "id": "2Zo4aBtAunJH",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "881a4ba6-187b-4a31-af4b-1ee7de6b0d9a"
      },
      "execution_count": 36,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/ipykernel_launcher.py:6: UserWarning: volatile was removed and now has no effect. Use `with torch.no_grad():` instead.\n",
            "  \n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Accuracy of Test Data: 63.56999969482422\n"
          ]
        }
      ]
    }
  ]
}