simple_cnn_mnist.ipynb

simple_cnn_mnist.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "simple_cnn_mnist.ipynb",
      "version": "0.3.2",
      "provenance": [],
      "collapsed_sections": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/gist/mmsamiei/e047234bdd65636eb65228eecfe5e359/simple_cnn_mnist.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VPQzc5ayY7PT",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import torch\n",
        "import torchvision.datasets"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XtpTyrRtZePs",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import torchvision.transforms as transforms\n",
        "trans = transforms.Compose([transforms.ToTensor(), transforms.Normalize((0,), (1.0,))])\n",
        "MNIST_dataset = torchvision.datasets.MNIST('./mnist', transform=trans)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6i4BcXkkZpFc",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from torch.utils.data import DataLoader\n",
        "train_dl = DataLoader(MNIST_dataset, batch_size=128)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "dhQWM-x4bDdr",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from torch import nn\n",
        "class ImanNet(nn.Module):\n",
        "    def __init__(self):\n",
        "        super().__init__()\n",
        "        self.conv1 = nn.Conv2d(1, 3, 3)\n",
        "        self.conv2 = nn.Conv2d(3, 6, 3)\n",
        "        self.pool1 = nn.MaxPool2d(2)\n",
        "        self.conv3 = nn.Conv2d(6, 9, 5)\n",
        "        self.lin = nn.Linear(576, 10)\n",
        "        self.softmax = nn.Softmax(1)\n",
        "\n",
        "    def forward(self, x):\n",
        "        temp = self.conv1(x)\n",
        "        temp = nn.functional.relu(temp)\n",
        "        temp = self.conv2(temp)\n",
        "        temp = nn.functional.relu(temp)\n",
        "        temp = self.pool1(temp)\n",
        "        temp = self.conv3(temp)\n",
        "        temp = nn.functional.relu(temp)\n",
        "        temp = temp.view(-1, 576)\n",
        "        temp = self.lin(temp)\n",
        "        out = self.softmax(temp)\n",
        "        return out"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-WxMLSbMdBGD",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "test = torch.randn(1,1,28,28)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VubQsjlDdavc",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 53
        },
        "outputId": "047da79c-4d35-41dc-ff68-5166980ec75a"
      },
      "source": [
        "model = ImanNet()\n",
        "model(test)"
      ],
      "execution_count": 122,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "tensor([[0.1055, 0.0977, 0.1070, 0.0914, 0.0856, 0.1202, 0.0983, 0.0903, 0.0961,\n",
              "         0.1079]], grad_fn=<SoftmaxBackward>)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 122
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zYRkP6dxe4ot",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "opt = torch.optim.Adam(model.parameters())"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "uW4am5pSeMyW",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 70
        },
        "outputId": "a870b32c-4df1-4fb6-f590-37c54eb27f66"
      },
      "source": [
        "loss_arr = []\n",
        "for epoch in range(3):\n",
        "  for xb,yb in train_dl:\n",
        "    out = model(xb)\n",
        "    loss = nn.functional.cross_entropy(out, yb)\n",
        "    loss.backward()\n",
        "    opt.step()\n",
        "    opt.zero_grad()\n",
        "    loss_arr.append(loss.item())\n",
        "  print(\"epoch \",epoch,\", loss = \",loss)"
      ],
      "execution_count": 124,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "epoch  0 , loss =  tensor(1.6238, grad_fn=<NllLossBackward>)\n",
            "epoch  1 , loss =  tensor(1.6036, grad_fn=<NllLossBackward>)\n",
            "epoch  2 , loss =  tensor(1.5770, grad_fn=<NllLossBackward>)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "UwJkZ-81r-Z1",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        },
        "outputId": "8131f829-7818-46f0-cef2-5f54c0ba114a"
      },
      "source": [
        "correct = 0\n",
        "total = 0\n",
        "with torch.no_grad():\n",
        "    for data in train_dl:\n",
        "        images, labels = data\n",
        "        out = model(images)\n",
        "        _, predicted = torch.max(out.data, 1)\n",
        "        total += labels.size(0)\n",
        "        correct += (predicted == labels).sum().item()\n",
        "\n",
        "print('Accuracy of the network on the 60000 test images: %d %%' % (\n",
        "    100 * correct / total))\n"
      ],
      "execution_count": 125,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Accuracy of the network on the 60000 test images: 86 %\n"
          ],
          "name": "stdout"
        }
      ]
    }
  ]
}