Transfer_Learning_by_using_hymenoptera_data.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Transfer Learning by using hymenoptera_data.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "mount_file_id": "1mY_q5a9-qO6bO5mlxiRyiq4Xl3Z9Q2iy",
      "authorship_tag": "ABX9TyM8Guziiz1J3DiiPZs14Oam",
      "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/JINU6497/JINUrepo/blob/master/Transfer_Learning_by_using_hymenoptera_data.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Wk0ikdkizE20"
      },
      "source": [
        "'''1. 모듈 임포트하기'''\n",
        "\n",
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "import torch\n",
        "import torch.nn as nn\n",
        "import torch.nn.functional as F\n",
        "from torchvision import transforms, datasets"
      ],
      "execution_count": 24,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Qi2zxuEp1Bun"
      },
      "source": [
        "'''2. 딥러닝 모델 설계에 활용하는 장비 확인'''\n",
        "\n",
        "if torch.cuda.is_available():\n",
        "  DEVICE=torch.device(\"cuda\")\n",
        "else:\n",
        "  DEVICE=torch.device(\"cpu\")\n"
      ],
      "execution_count": 25,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "CAcenKTk1aMG"
      },
      "source": [
        "BATCH_SIZE=32\n",
        "EPOCHS=10"
      ],
      "execution_count": 26,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4kxWeYyB1euT"
      },
      "source": [
        "'''3. 데이터 다운(Train과 Test Set 분리)'''\n",
        "\n",
        "data_transforms={\n",
        "    'train':transforms.Compose([\n",
        "                                transforms.RandomResizedCrop(224),\n",
        "                                transforms.RandomHorizontalFlip(),\n",
        "                                transforms.ToTensor(),\n",
        "                                transforms.Normalize([0.5,0.5,0.5],[0.5,0.5,0.5])\n",
        "                                ]), \n",
        "    'val':transforms.Compose([\n",
        "                                transforms.RandomResizedCrop(224),\n",
        "                                transforms.Resize(256),\n",
        "                                transforms.RandomHorizontalFlip(),\n",
        "                                transforms.ToTensor(),\n",
        "                                transforms.Normalize([0.5,0.5,0.5],[0.5,0.5,0.5])\n",
        "                                ]),}\n",
        "\n",
        "image_datasets={x:datasets.ImageFolder(\"/content/drive/MyDrive/hymenoptera_data/hymenoptera_data\",data_transforms[x]) for x in ['train','val']}\n",
        "dataloaders={x:torch.utils.data.DataLoader(image_datasets[x],batch_size=BATCH_SIZE,num_workers=0,shuffle=True) for x in ['train','val']}"
      ],
      "execution_count": 27,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "c1Xzj1gK-bx-"
      },
      "source": [
        "RandomResizedCrop(224)는 해당 이미지 중앙을 기준으로 224*224 크기로 이미지를 잘라내어 사이즈를 변경하는 것을 의미.\n",
        "\n",
        "맨 아래 줄은 이미지 데이터를 불러오면서 기존에 정의한 data_transforms의 학습 데이터셋에 이용되는 전처리 과정, 검증 데이터셋에 이용되는 전처리 과정을 각각 적용하는 것을 의미. 이때 학습 데이터셋과 검증 데이터셋에 각각 적용하기 위하여 Dictionary 구조를 사용.\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "oOu8bB-Z1hAR",
        "outputId": "992d78ee-8cad-4e0e-8b69-2142a35f4910"
      },
      "source": [
        "'''4. 데이터 확인'''\n",
        "\n",
        "for (X_train, Y_train) in dataloaders['train']:\n",
        "  print('X_train:', X_train.size(), 'type:', X_train.type())\n",
        "  print('Y_train:', Y_train.size(), 'type:', Y_train.type())  "
      ],
      "execution_count": 28,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "X_train: torch.Size([32, 3, 224, 224]) type: torch.FloatTensor\n",
            "Y_train: torch.Size([32]) type: torch.LongTensor\n",
            "X_train: torch.Size([32, 3, 224, 224]) type: torch.FloatTensor\n",
            "Y_train: torch.Size([32]) type: torch.LongTensor\n",
            "X_train: torch.Size([32, 3, 224, 224]) type: torch.FloatTensor\n",
            "Y_train: torch.Size([32]) type: torch.LongTensor\n",
            "X_train: torch.Size([32, 3, 224, 224]) type: torch.FloatTensor\n",
            "Y_train: torch.Size([32]) type: torch.LongTensor\n",
            "X_train: torch.Size([32, 3, 224, 224]) type: torch.FloatTensor\n",
            "Y_train: torch.Size([32]) type: torch.LongTensor\n",
            "X_train: torch.Size([32, 3, 224, 224]) type: torch.FloatTensor\n",
            "Y_train: torch.Size([32]) type: torch.LongTensor\n",
            "X_train: torch.Size([32, 3, 224, 224]) type: torch.FloatTensor\n",
            "Y_train: torch.Size([32]) type: torch.LongTensor\n",
            "X_train: torch.Size([32, 3, 224, 224]) type: torch.FloatTensor\n",
            "Y_train: torch.Size([32]) type: torch.LongTensor\n",
            "X_train: torch.Size([32, 3, 224, 224]) type: torch.FloatTensor\n",
            "Y_train: torch.Size([32]) type: torch.LongTensor\n",
            "X_train: torch.Size([32, 3, 224, 224]) type: torch.FloatTensor\n",
            "Y_train: torch.Size([32]) type: torch.LongTensor\n",
            "X_train: torch.Size([32, 3, 224, 224]) type: torch.FloatTensor\n",
            "Y_train: torch.Size([32]) type: torch.LongTensor\n",
            "X_train: torch.Size([32, 3, 224, 224]) type: torch.FloatTensor\n",
            "Y_train: torch.Size([32]) type: torch.LongTensor\n",
            "X_train: torch.Size([13, 3, 224, 224]) type: torch.FloatTensor\n",
            "Y_train: torch.Size([13]) type: torch.LongTensor\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "QY0qhRbo_t5e",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 274
        },
        "outputId": "1b16b299-1de5-46d7-e90d-7437ed94a318"
      },
      "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": 29,
      "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": "0D0xFNyK_7-1"
      },
      "source": [
        "'''5. 불러온 특정 모델에 대해 학습을 진행하며 학습 데이터에 대한 모델 성능을 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}%)]\\tTraint Loss: {:.6f}\".format(Epoch,batch_idx*len(image), len(train_loader.dataset),100.*batch_idx/len(train_loader),loss.item()))\n",
        "    "
      ],
      "execution_count": 30,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "q_qaRsBLPq2D"
      },
      "source": [
        "'''6. 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",
        "\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",
        "      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",
        "  return test_loss, test_accuracy"
      ],
      "execution_count": 31,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "QhBolpRh8jw7"
      },
      "source": [
        "'''7. PyTorch 내에서 제공하는 미리 학습되지 않은 ResNet18 모델을 볼러온 후 Output의 크기 설정'''\n",
        "\n",
        "import torchvision.models as models\n",
        "model=models.resnet18(pretrained=False).cuda()\n",
        "num_ftrs=model.fc.in_features\n",
        "model.fc=nn.Linear(num_ftrs,2)\n",
        "model=model.cuda()"
      ],
      "execution_count": 32,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "V8AFrifvEnud"
      },
      "source": [
        "torchvision내에 있는 models.resnet18을 불러온다. 이 모델에 대해 Fully Connected Layer를 구성하고 있는 부분에 접근. 이때 in_features는 resnet18 모델의 Fully Connected Layer의 input에 해당하는 노드 수를 num_ftrs로 저장. 이때 output은 벌과 개미 이미지 2개이기 때문에 2로 지정"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "jWSr7K80F0h6",
        "outputId": "6ee15973-8891-4c5f-aed3-5a662d7f27fd"
      },
      "source": [
        "'''8. Optimizer, Objective Function 설정'''\n",
        "\n",
        "optimizer=torch.optim.Adam(model.parameters(),lr=0.0001)\n",
        "criterion=nn.CrossEntropyLoss()\n",
        "\n",
        "print(model)"
      ],
      "execution_count": 33,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "ResNet(\n",
            "  (conv1): Conv2d(3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False)\n",
            "  (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "  (relu): ReLU(inplace=True)\n",
            "  (maxpool): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)\n",
            "  (layer1): Sequential(\n",
            "    (0): 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",
            "      (relu): ReLU(inplace=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",
            "    )\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",
            "      (relu): ReLU(inplace=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",
            "    )\n",
            "  )\n",
            "  (layer2): Sequential(\n",
            "    (0): BasicBlock(\n",
            "      (conv1): Conv2d(64, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
            "      (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "      (relu): ReLU(inplace=True)\n",
            "      (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
            "      (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "      (downsample): Sequential(\n",
            "        (0): Conv2d(64, 128, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
            "        (1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "      )\n",
            "    )\n",
            "    (1): BasicBlock(\n",
            "      (conv1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
            "      (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "      (relu): ReLU(inplace=True)\n",
            "      (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
            "      (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    )\n",
            "  )\n",
            "  (layer3): Sequential(\n",
            "    (0): BasicBlock(\n",
            "      (conv1): Conv2d(128, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
            "      (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "      (relu): ReLU(inplace=True)\n",
            "      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
            "      (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "      (downsample): Sequential(\n",
            "        (0): Conv2d(128, 256, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
            "        (1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "      )\n",
            "    )\n",
            "    (1): BasicBlock(\n",
            "      (conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
            "      (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "      (relu): ReLU(inplace=True)\n",
            "      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
            "      (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    )\n",
            "  )\n",
            "  (layer4): Sequential(\n",
            "    (0): BasicBlock(\n",
            "      (conv1): Conv2d(256, 512, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
            "      (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "      (relu): ReLU(inplace=True)\n",
            "      (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
            "      (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "      (downsample): Sequential(\n",
            "        (0): Conv2d(256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
            "        (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "      )\n",
            "    )\n",
            "    (1): BasicBlock(\n",
            "      (conv1): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
            "      (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "      (relu): ReLU(inplace=True)\n",
            "      (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
            "      (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    )\n",
            "  )\n",
            "  (avgpool): AdaptiveAvgPool2d(output_size=(1, 1))\n",
            "  (fc): Linear(in_features=512, out_features=2, bias=True)\n",
            ")\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "VVqFPwBvGDok",
        "outputId": "45e8a09d-64e5-4962-b363-00accca2b0b4"
      },
      "source": [
        "'''9. 미리 학습되지 않은 ResNet18 학습 실행 후 결과 확인'''\n",
        "\n",
        "for Epoch in range(1, EPOCHS+1):\n",
        "  train(model, dataloaders[\"train\"],optimizer,log_interval=5)\n",
        "  test_loss, test_accuracy=evaluate(model,dataloaders[\"val\"])\n",
        "  print(\"\\n[EPOCH:{}],\\tTest Loss:{:.4f}, \\tTest Accuracy:{:.2f}%\\n\".format(Epoch,test_loss,test_accuracy))"
      ],
      "execution_count": 34,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train Epoch:1[0/397(0%)]\tTraint Loss: 0.746408\n",
            "Train Epoch:1[160/397(38%)]\tTraint Loss: 0.793705\n",
            "Train Epoch:1[320/397(77%)]\tTraint Loss: 0.694827\n",
            "\n",
            "[EPOCH:1],\tTest Loss:0.0286, \tTest Accuracy:38.54%\n",
            "\n",
            "Train Epoch:2[0/397(0%)]\tTraint Loss: 0.638515\n",
            "Train Epoch:2[160/397(38%)]\tTraint Loss: 0.661966\n",
            "Train Epoch:2[320/397(77%)]\tTraint Loss: 0.731717\n",
            "\n",
            "[EPOCH:2],\tTest Loss:0.0251, \tTest Accuracy:61.21%\n",
            "\n",
            "Train Epoch:3[0/397(0%)]\tTraint Loss: 0.621762\n",
            "Train Epoch:3[160/397(38%)]\tTraint Loss: 0.734293\n",
            "Train Epoch:3[320/397(77%)]\tTraint Loss: 0.672936\n",
            "\n",
            "[EPOCH:3],\tTest Loss:0.0220, \tTest Accuracy:61.71%\n",
            "\n",
            "Train Epoch:4[0/397(0%)]\tTraint Loss: 0.646585\n",
            "Train Epoch:4[160/397(38%)]\tTraint Loss: 0.595757\n",
            "Train Epoch:4[320/397(77%)]\tTraint Loss: 0.693303\n",
            "\n",
            "[EPOCH:4],\tTest Loss:0.0210, \tTest Accuracy:64.48%\n",
            "\n",
            "Train Epoch:5[0/397(0%)]\tTraint Loss: 0.723764\n",
            "Train Epoch:5[160/397(38%)]\tTraint Loss: 0.532669\n",
            "Train Epoch:5[320/397(77%)]\tTraint Loss: 0.592906\n",
            "\n",
            "[EPOCH:5],\tTest Loss:0.0207, \tTest Accuracy:63.98%\n",
            "\n",
            "Train Epoch:6[0/397(0%)]\tTraint Loss: 0.654743\n",
            "Train Epoch:6[160/397(38%)]\tTraint Loss: 0.661027\n",
            "Train Epoch:6[320/397(77%)]\tTraint Loss: 0.743798\n",
            "\n",
            "[EPOCH:6],\tTest Loss:0.0228, \tTest Accuracy:61.46%\n",
            "\n",
            "Train Epoch:7[0/397(0%)]\tTraint Loss: 0.602964\n",
            "Train Epoch:7[160/397(38%)]\tTraint Loss: 0.680964\n",
            "Train Epoch:7[320/397(77%)]\tTraint Loss: 0.686007\n",
            "\n",
            "[EPOCH:7],\tTest Loss:0.0207, \tTest Accuracy:64.23%\n",
            "\n",
            "Train Epoch:8[0/397(0%)]\tTraint Loss: 0.633259\n",
            "Train Epoch:8[160/397(38%)]\tTraint Loss: 0.573716\n",
            "Train Epoch:8[320/397(77%)]\tTraint Loss: 0.603246\n",
            "\n",
            "[EPOCH:8],\tTest Loss:0.0197, \tTest Accuracy:67.00%\n",
            "\n",
            "Train Epoch:9[0/397(0%)]\tTraint Loss: 0.637994\n",
            "Train Epoch:9[160/397(38%)]\tTraint Loss: 0.723377\n",
            "Train Epoch:9[320/397(77%)]\tTraint Loss: 0.577720\n",
            "\n",
            "[EPOCH:9],\tTest Loss:0.0200, \tTest Accuracy:65.74%\n",
            "\n",
            "Train Epoch:10[0/397(0%)]\tTraint Loss: 0.547676\n",
            "Train Epoch:10[160/397(38%)]\tTraint Loss: 0.722639\n",
            "Train Epoch:10[320/397(77%)]\tTraint Loss: 0.612876\n",
            "\n",
            "[EPOCH:10],\tTest Loss:0.0199, \tTest Accuracy:67.25%\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    }
  ]
}