4.심층신경망.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "8.심층신경망.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "vaU2pRqXqESs",
        "colab_type": "text"
      },
      "source": [
        "# Deep Neural Network (DNNs)\n",
        "MNIST Dataset Classficiation with DNN"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "sZs_JS5GqD3v",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import torch\n",
        "import torchvision.datasets as dsets\n",
        "import torchvision.transforms as transforms\n",
        "from torch.utils.data import DataLoader\n",
        "import matplotlib.pyplot as plt\n",
        "import random"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vSu0b7UZqSWa",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "GPU = torch.cuda.is_available() \n",
        "device = torch.device(\"cuda\" if GPU else \"cpu\") # GPU 사용 가능하면 사용하고 아니면 CPU 사용"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "iFlS9aB3qdS_",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "torch.manual_seed(1)\n",
        "if device == 'cuda':\n",
        "    torch.cuda.manual_seed_all(1)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "v3TrMkiPqn8B",
        "colab_type": "text"
      },
      "source": [
        "## Load MNIST Dataset"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DY1b__-fqmPI",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "mnist_train = dsets.MNIST(root='data/',\n",
        "                          train=True,\n",
        "                          transform=transforms.ToTensor(),\n",
        "                          download=True)\n",
        "mnist_test = dsets.MNIST(root='data/',\n",
        "                         train=False,\n",
        "                         transform=transforms.ToTensor(),\n",
        "                         download=True)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0SnG76TJqxUj",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "data_loader = DataLoader(dataset=mnist_train,\n",
        "                         batch_size=100,\n",
        "                         shuffle=True,\n",
        "                         drop_last=True) # Drop Last Batch"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "FRIkMK9nrFCC",
        "colab_type": "text"
      },
      "source": [
        "## DNN Model\n",
        "| Input | 784 |\n",
        "|--|--|\n",
        "| Dense | 512 |\n",
        "| ReLU |  |\n",
        "| Dense | 512 |\n",
        "| Output | 10 |"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jat07t9IrHfv",
        "colab_type": "code",
        "outputId": "ee78e709-4f28-4243-dc97-43afe72888fc",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 111
        }
      },
      "source": [
        "l1 = torch.nn.Linear((28*28), 512).to(device)\n",
        "l2 = torch.nn.Linear(512, 10).to(device)\n",
        "relu = torch.nn.ReLU()\n",
        "\n",
        "model = torch.nn.Sequential(l1, relu, l2)\n",
        "model"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "Sequential(\n",
              "  (0): Linear(in_features=784, out_features=512, bias=True)\n",
              "  (1): ReLU()\n",
              "  (2): Linear(in_features=512, out_features=10, bias=True)\n",
              ")"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 6
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "sCPzWgSLsucr",
        "colab_type": "text"
      },
      "source": [
        "## Train Model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Nd4Pa7gZsQjY",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "cost = torch.nn.CrossEntropyLoss().to(device) # Built in Softmax Function\n",
        "optimizer = torch.optim.SGD(model.parameters(), lr=0.1)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "SU3xQ2R0tGQJ",
        "colab_type": "code",
        "outputId": "12fd388a-5169-4d6b-da60-97e09ca3619b",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 111
        }
      },
      "source": [
        "epochs = 20\n",
        "for epoch in range(1, epochs+1):\n",
        "    avg_cost = 0\n",
        "    total_batch = len(data_loader)\n",
        "\n",
        "    for x, y in data_loader:    # batch loop\n",
        "        x = x.view(-1, 28*28).to(device)\n",
        "        y = y.to(device)\n",
        "\n",
        "        optimizer.zero_grad()\n",
        "        hypothesis = model(x)\n",
        "        cost_val = cost(hypothesis, y)\n",
        "        cost_val.backward()\n",
        "        optimizer.step()\n",
        "\n",
        "        avg_cost += cost_val\n",
        "    \n",
        "    avg_cost /= total_batch\n",
        "\n",
        "    if epoch % 5 == 1 or epoch == epochs:\n",
        "        print('Epoch {:4d}/{} Cost: {:.6f}'.format(epoch, epochs, avg_cost.item()))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch    1/20 Cost: 0.503713\n",
            "Epoch    6/20 Cost: 0.121553\n",
            "Epoch   11/20 Cost: 0.070358\n",
            "Epoch   16/20 Cost: 0.047347\n",
            "Epoch   20/20 Cost: 0.035963\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "tbGC7MbHulXu",
        "colab_type": "text"
      },
      "source": [
        "## Evaluate Model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "K4VM9AxauV9K",
        "colab_type": "code",
        "outputId": "97731ca7-e06c-46fb-b14e-a07e834a1856",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 396
        }
      },
      "source": [
        "with torch.no_grad(): # Don't Calculate Gradient\n",
        "    x_test = mnist_test.test_data.view(-1, 28 * 28).float().to(device)\n",
        "    y_test = mnist_test.test_labels.to(device)\n",
        "\n",
        "    pred = model(x_test)\n",
        "    correct_pred = torch.argmax(pred, 1) == y_test\n",
        "    acc = correct_pred.float().mean()\n",
        "    print('Accuracy:',acc.item())\n",
        "\n",
        "    r = random.randint(0, len(mnist_test) - 1)\n",
        "    X_single_data = mnist_test.test_data[r].view(-1, 28 * 28).float().to(device)\n",
        "    Y_single_data = mnist_test.test_labels[r].to(device)\n",
        "\n",
        "    print('Label: ', Y_single_data.item())\n",
        "    single_prediction = model(X_single_data)\n",
        "    print('Prediction: ', torch.argmax(single_prediction, 1).item())\n",
        "\n",
        "    plt.imshow(mnist_test.test_data[r:r + 1].view(28, 28), cmap='Greys', interpolation='nearest')\n",
        "    plt.show()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torchvision/datasets/mnist.py:60: UserWarning: test_data has been renamed data\n",
            "  warnings.warn(\"test_data has been renamed data\")\n",
            "/usr/local/lib/python3.6/dist-packages/torchvision/datasets/mnist.py:50: UserWarning: test_labels has been renamed targets\n",
            "  warnings.warn(\"test_labels has been renamed targets\")\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Accuracy: 0.977400004863739\n",
            "Label:  5\n",
            "Prediction:  5\n"
          ],
          "name": "stdout"
        },
        {
          "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": "markdown",
      "metadata": {
        "id": "rGeA7JLcvnjL",
        "colab_type": "text"
      },
      "source": [
        "### Reference\n",
        "- [PyTorch로 시작하는 딥러닝 입문](https://wikidocs.net/60324)"
      ]
    }
  ]
}