VGG16_CIFAR10.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "VGG-16_CIFAR10.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "id": "Ky9SzYr7xHPr",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import numpy as np\n",
        "import matplotlib.pyplot as plt"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "cyl7oBEtPYzP",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import torch\n",
        "import torchvision\n",
        "import torchvision.transforms as transforms\n",
        "import torch.nn as nn\n",
        "import torch.optim as optim"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "FLRh6HhOPmit",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nTE4_GT4P1Lx",
        "colab_type": "code",
        "outputId": "cf0e0755-b802-432b-8637-bb7f883a8c59",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "print(device)"
      ],
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "cuda:0\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0KT5xDEYP-dm",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "FRnjsDALrz2G",
        "colab_type": "text"
      },
      "source": [
        "<h3> Loading Dataset and Applying Transforms </h3>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "CBdQQIe5P-uj",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "transform_train = transforms.Compose([\n",
        "    transforms.RandomResizedCrop(224),\n",
        "    transforms.ToTensor(),\n",
        "    transforms.Normalize((0.5,0.5,0.5),(0.5,0.5,0.5)),\n",
        "])\n",
        "\n",
        "transform_test = transforms.Compose([\n",
        "    transforms.RandomResizedCrop(224),\n",
        "    transforms.ToTensor(),\n",
        "    transforms.Normalize((0.5,0.5,0.5),(0.5,0.5,0.5)),\n",
        "])\n",
        "\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "mTgDDvGZphdF",
        "colab_type": "text"
      },
      "source": [
        "<b>Note : </b>transforms.Normalize does the following for each of the 3 channels: image = (image - mean) / std The parameters mean, std are passed as 0.5, 0.5 in your case. This will normalize the image in the range [-1,1]. For example, the minimum value 0 will be converted to (0-0.5)/0.5=-1, the maximum value of 1 will be converted to 1\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "_r9QERResEZw",
        "colab_type": "text"
      },
      "source": [
        "<b> Steps in Image Preprocessing </b>\n",
        "\n",
        "<ul>\n",
        "  <li>Uniform aspect ratio : Most of the neural network models assume a square shape input image, which means that each image needs to be checked if it is a square or not, and cropped appropriately </li>\n",
        "  <li> Image scaling </li>\n",
        "  <li> Computing mean, std dev and normalizing inputs </li>\n",
        "  <li> Data augmentation </li>\n",
        "\n",
        "</ul>\n",
        "\n",
        "https://becominghuman.ai/image-data-pre-processing-for-neural-networks-498289068258"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "YhV4Foh5sB9t",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "psboGfB5f3nh",
        "colab_type": "code",
        "outputId": "d9ece938-36e3-426a-eebe-29d0cb5b54cc",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 67
        }
      },
      "source": [
        "trainset = torchvision.datasets.CIFAR10(root = './data' , train = True, download = True , transform = transform_train)\n",
        "testset = torchvision.datasets.CIFAR10(root= './data', train = False, download = True, transform = transform_test)"
      ],
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "\r0it [00:00, ?it/s]"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Downloading https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz to ./data/cifar-10-python.tar.gz\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            " 99%|█████████▉| 169287680/170498071 [00:12<00:00, 17258974.15it/s]"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Extracting ./data/cifar-10-python.tar.gz to ./data\n",
            "Files already downloaded and verified\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "IGGDBJ2bhB1y",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "num_classes = 10"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "J2VB6b3ShOdE",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "batch_size = 4"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "wufApf5_hP8C",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "trainloader = torch.utils.data.DataLoader(trainset, batch_size=batch_size, shuffle=True)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6EYJY3uHhSNQ",
        "colab_type": "code",
        "outputId": "776d83f1-deed-4ffc-e5f5-7dc7039a0406",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 67
        }
      },
      "source": [
        "dataiter = iter(trainloader)\n",
        "\n",
        "images, labels = dataiter.next()\n",
        "\n",
        "print(images.shape)\n",
        "\n",
        "print(images[1].shape)\n",
        "print(labels[1].item())"
      ],
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "torch.Size([4, 3, 224, 224])\n",
            "torch.Size([3, 224, 224])\n",
            "6\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "4Ad3q6HEqa01",
        "colab_type": "text"
      },
      "source": [
        "Note: The original size of CIFAR image is 32 x 32, but using the RandomResizedCrop function we changed it to 224 x 224"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "PBpwoq7QgfGJ",
        "colab_type": "code",
        "outputId": "aedd610b-df43-4e98-c7f4-cb0caf615f7c",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "len(trainloader)"
      ],
      "execution_count": 13,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "12500"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 13
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "rZOX77-dhYS5",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def imshow(img, title):\n",
        "    # Unnormalize the image, we multiply by 0.5 that is the std deviation, and add 0.5 which is the mean\n",
        "    npimg = img.numpy() / 2 + 0.5\n",
        "    plt.figure(figsize=(batch_size, 1))\n",
        "    plt.axis('off')\n",
        "    #Convert [3,224,224] to [224,224,3]\n",
        "    plt.imshow(np.transpose(npimg, (1, 2, 0)))\n",
        "    plt.title(title)\n",
        "    plt.show()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5Zw0p-QAhfhY",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def show_batch_images(dataloader):\n",
        "    images, labels = next(iter(dataloader))\n",
        "    img = torchvision.utils.make_grid(images)\n",
        "    imshow(img, title=[str(x.item()) for x in labels])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "tk9Y7bduhjOP",
        "colab_type": "code",
        "outputId": "62f5889c-7c13-4302-cedc-45deb1edf9f5",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 353
        }
      },
      "source": [
        "for i in range(4):\n",
        "    show_batch_images(trainloader)"
      ],
      "execution_count": 16,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAOYAAABUCAYAAACWaEvFAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nYII=\n",
            "text/plain": [
              "<Figure size 288x72 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAOYAAABUCAYAAACWaEvFAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nY/yG4yGYj/EYv+H4b6Mz/VL9oz2kAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 288x72 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAOYAAABUCAYAAACWaEvFAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nL8B8kRf5BOX/B9D9RGvwyiJTAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 288x72 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAOYAAABUCAYAAACWaEvFAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nEsojMB/lUb6E8gjMR3mUL6H8vzSlTfoxb74PAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 288x72 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "uxa8Nd3vhnI7",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "iNegzbreuD9N",
        "colab_type": "text"
      },
      "source": [
        "<h3>Using inbuilt VGG-16 Model </h3> "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "aPQIZ4vSilwx",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from torchvision import models"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "oTc9isLhinr7",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "vgg = models.vgg16_bn()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "WL31ov2IitqL",
        "colab_type": "code",
        "outputId": "dbad9fed-bcae-4052-fbd7-13cface714a8",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 991
        }
      },
      "source": [
        "print(vgg)"
      ],
      "execution_count": 19,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "VGG(\n",
            "  (features): Sequential(\n",
            "    (0): Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (2): ReLU(inplace=True)\n",
            "    (3): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (4): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (5): ReLU(inplace=True)\n",
            "    (6): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "    (7): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (8): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (9): ReLU(inplace=True)\n",
            "    (10): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (11): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (12): ReLU(inplace=True)\n",
            "    (13): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "    (14): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (15): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (16): ReLU(inplace=True)\n",
            "    (17): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (18): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (19): ReLU(inplace=True)\n",
            "    (20): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (21): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (22): ReLU(inplace=True)\n",
            "    (23): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "    (24): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (25): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (26): ReLU(inplace=True)\n",
            "    (27): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (28): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (29): ReLU(inplace=True)\n",
            "    (30): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (31): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (32): ReLU(inplace=True)\n",
            "    (33): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "    (34): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (35): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (36): ReLU(inplace=True)\n",
            "    (37): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (38): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (39): ReLU(inplace=True)\n",
            "    (40): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (41): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (42): ReLU(inplace=True)\n",
            "    (43): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "  )\n",
            "  (avgpool): AdaptiveAvgPool2d(output_size=(7, 7))\n",
            "  (classifier): Sequential(\n",
            "    (0): Linear(in_features=25088, out_features=4096, bias=True)\n",
            "    (1): ReLU(inplace=True)\n",
            "    (2): Dropout(p=0.5, inplace=False)\n",
            "    (3): Linear(in_features=4096, out_features=4096, bias=True)\n",
            "    (4): ReLU(inplace=True)\n",
            "    (5): Dropout(p=0.5, inplace=False)\n",
            "    (6): Linear(in_features=4096, out_features=1000, bias=True)\n",
            "  )\n",
            ")\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XcGEVvnLiw-F",
        "colab_type": "code",
        "outputId": "5c699d12-12a6-4f78-99ca-995a60a41799",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "print(vgg.features[0])"
      ],
      "execution_count": 20,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "U0uZo6eSi-9q",
        "colab_type": "code",
        "outputId": "469b3602-cef9-4c2e-c7c7-5763423c8ffd",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "# out_features is 1000, but we have only 10 output features, so we should modify last layer such that number of output \n",
        "# classes is 10\n",
        "print(vgg.classifier[6])"
      ],
      "execution_count": 21,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Linear(in_features=4096, out_features=1000, bias=True)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "CKkfcQACjB_7",
        "colab_type": "code",
        "outputId": "96883e23-c874-4c8d-b479-8c5749a18528",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 54
        }
      },
      "source": [
        "# get number of input features in the last layer, as we do not want to change that, but only change the number of output features\n",
        "final_in_features = vgg.classifier[6].in_features\n",
        "\n",
        "# take all layers of vgg.classifier except the last layer\n",
        "mod_classifier = list(vgg.classifier.children())[:-1]\n",
        "\n",
        "# add our own layer to vgg.classier such that number of output_features equals number of classes\n",
        "# number of input features in same as the removed last layer\n",
        "mod_classifier.extend([nn.Linear(final_in_features, num_classes)])\n",
        "print(mod_classifier)"
      ],
      "execution_count": 22,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[Linear(in_features=25088, out_features=4096, bias=True), ReLU(inplace=True), Dropout(p=0.5, inplace=False), Linear(in_features=4096, out_features=4096, bias=True), ReLU(inplace=True), Dropout(p=0.5, inplace=False), Linear(in_features=4096, out_features=10, bias=True)]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "fKR_lf64jXTk",
        "colab_type": "code",
        "outputId": "102e4d15-424d-4714-854d-b87e8d3720ff",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 991
        }
      },
      "source": [
        "vgg.classifier = nn.Sequential(*mod_classifier)\n",
        "print(vgg)"
      ],
      "execution_count": 23,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "VGG(\n",
            "  (features): Sequential(\n",
            "    (0): Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (2): ReLU(inplace=True)\n",
            "    (3): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (4): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (5): ReLU(inplace=True)\n",
            "    (6): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "    (7): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (8): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (9): ReLU(inplace=True)\n",
            "    (10): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (11): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (12): ReLU(inplace=True)\n",
            "    (13): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "    (14): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (15): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (16): ReLU(inplace=True)\n",
            "    (17): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (18): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (19): ReLU(inplace=True)\n",
            "    (20): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (21): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (22): ReLU(inplace=True)\n",
            "    (23): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "    (24): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (25): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (26): ReLU(inplace=True)\n",
            "    (27): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (28): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (29): ReLU(inplace=True)\n",
            "    (30): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (31): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (32): ReLU(inplace=True)\n",
            "    (33): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "    (34): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (35): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (36): ReLU(inplace=True)\n",
            "    (37): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (38): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (39): ReLU(inplace=True)\n",
            "    (40): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (41): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (42): ReLU(inplace=True)\n",
            "    (43): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "  )\n",
            "  (avgpool): AdaptiveAvgPool2d(output_size=(7, 7))\n",
            "  (classifier): Sequential(\n",
            "    (0): Linear(in_features=25088, out_features=4096, bias=True)\n",
            "    (1): ReLU(inplace=True)\n",
            "    (2): Dropout(p=0.5, inplace=False)\n",
            "    (3): Linear(in_features=4096, out_features=4096, bias=True)\n",
            "    (4): ReLU(inplace=True)\n",
            "    (5): Dropout(p=0.5, inplace=False)\n",
            "    (6): Linear(in_features=4096, out_features=10, bias=True)\n",
            "  )\n",
            ")\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kQoir7HdjdAJ",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Wv83MTPVjpnb",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "batch_size = 16\n",
        "# working with very large batch sizes will cause error because it occupies too much memory, 16 itself is reasonbly large\n",
        "\n",
        "trainloader = torch.utils.data.DataLoader(trainset, batch_size=batch_size, shuffle=True)\n",
        "testloader = torch.utils.data.DataLoader(testset, batch_size=batch_size, shuffle=False)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "T9OHdcF0lx4J",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def evaluation(dataloader, model):\n",
        "    total, correct = 0, 0\n",
        "    for data in dataloader:\n",
        "        inputs, labels = data\n",
        "        inputs, labels = inputs.to(device), labels.to(device)\n",
        "        outputs = model(inputs)\n",
        "        _, pred = torch.max(outputs.data, 1)\n",
        "        total += labels.size(0)\n",
        "        correct += (pred == labels).sum().item()\n",
        "    return 100 * correct / total"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "saPVQykMl43e",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "vgg = vgg.to(device)\n",
        "loss_fn = nn.CrossEntropyLoss()\n",
        "opt = optim.SGD(vgg.parameters(), lr=0.05)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XgEx6k08l7VW",
        "colab_type": "code",
        "outputId": "d950e34c-b510-4d50-9e6b-b64d4f55af2d",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 380
        }
      },
      "source": [
        "loss_epoch_arr = []\n",
        "max_epochs = 1\n",
        "\n",
        "n_iters = np.ceil(50000/batch_size)\n",
        "\n",
        "for epoch in range(max_epochs):\n",
        "\n",
        "    for i, data in enumerate(trainloader, 0):\n",
        "\n",
        "        inputs, labels = data\n",
        "        inputs, labels = inputs.to(device), labels.to(device)\n",
        "\n",
        "        opt.zero_grad()\n",
        "\n",
        "        outputs = vgg(inputs)\n",
        "        loss = loss_fn(outputs, labels)\n",
        "        loss.backward()\n",
        "        opt.step()\n",
        "        \n",
        "        del inputs, labels, outputs\n",
        "        torch.cuda.empty_cache()\n",
        "        \n",
        "        if i % 100 == 0:\n",
        "            print('Iteration: %d/%d, Loss: %0.2f' % (i, n_iters, loss.item()))\n",
        "        \n",
        "    loss_epoch_arr.append(loss.item())\n",
        "        \n",
        "    print('Epoch: %d/%d, Test acc: %0.2f, Train acc: %0.2f' % (\n",
        "        epoch, max_epochs, \n",
        "        evaluation(testloader, vgg), evaluation(trainloader, vgg)))\n",
        "    \n",
        "plt.plot(loss_epoch_arr)\n",
        "plt.show()"
      ],
      "execution_count": 27,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Iteration: 0/3125, Loss: 2.47\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "\r170500096it [00:30, 17258974.15it/s]                               "
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Iteration: 100/3125, Loss: 2.77\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "error",
          "ename": "KeyboardInterrupt",
          "evalue": "ignored",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-27-e6281ceebae1>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m     19\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     20\u001b[0m         \u001b[0;32mdel\u001b[0m \u001b[0minputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlabels\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0moutputs\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 21\u001b[0;31m         \u001b[0mtorch\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcuda\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mempty_cache\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     22\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     23\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mi\u001b[0m \u001b[0;34m%\u001b[0m \u001b[0;36m100\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;36m0\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/torch/cuda/__init__.py\u001b[0m in \u001b[0;36mempty_cache\u001b[0;34m()\u001b[0m\n\u001b[1;32m    459\u001b[0m     \"\"\"\n\u001b[1;32m    460\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0m_initialized\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 461\u001b[0;31m         \u001b[0mtorch\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_C\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_cuda_emptyCache\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    462\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    463\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "_xemDwKAugsR",
        "colab_type": "text"
      },
      "source": [
        "<b> Note: </b> Training is taking too long because of the huge number of parameters invloved while training the net, approximately 138 million. To overcome this, we use pretrained models and train only the last layer using the images in our dataset. This is known as transfer learning "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Exm2b_vgl-sy",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "6jCHBrqSvE2b",
        "colab_type": "text"
      },
      "source": [
        "<h3> Using Pre trained Model and Transfer Learning </h3>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VtpqcQ23nV3U",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "batch_size = 16\n",
        "\n",
        "trainloader = torch.utils.data.DataLoader(trainset, batch_size = batch_size, shuffle = True)\n",
        "#Why shuffle false for test data\n",
        "testloader = torch.utils.data.DataLoader(testset, batch_size = batch_size, shuffle = False)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xmYIhkw8nVoS",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "#this trained model is trained on imagenet data\n",
        "\n",
        "vgg = models.vgg16_bn(pretrained = True)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "U4IN1vFmnVWD",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "for param in vgg.parameters():\n",
        "  param.requires_grad = False\n",
        "\n",
        "# we have frozen the parameters by making gradients 0\n",
        "# because weights are already pretrained on imagenet dataset"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "qOU6o_rOoUiT",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "final_in_features = vgg.classifier[6].in_features\n",
        "vgg.classifier[6] = nn.Linear(final_in_features, num_classes)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ERS2fVtboURn",
        "colab_type": "code",
        "outputId": "b8adb6a5-aebc-4bc4-d518-01be58791d6f",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 50
        }
      },
      "source": [
        "for param in vgg.parameters():\n",
        "    if param.requires_grad:\n",
        "        print(param.shape) # only last layer should be print as gradients has been set to zero for all other layers"
      ],
      "execution_count": 50,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "torch.Size([10, 4096])\n",
            "torch.Size([10])\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ojNS8GCpoUDP",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "vgg = vgg.to(device)\n",
        "loss_fn = nn.CrossEntropyLoss()\n",
        "opt = optim.SGD(vgg.parameters(), lr=0.05)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "JFkm9c91oT0j",
        "colab_type": "code",
        "outputId": "0831d0b5-088b-46b3-9dda-ecc9cacb5796",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 819
        }
      },
      "source": [
        "loss_epoch_arr = []\n",
        "max_epochs = 1\n",
        "\n",
        "n_iters = np.ceil(50000/batch_size)\n",
        "\n",
        "for epoch in range(max_epochs):\n",
        "\n",
        "    for i, data in enumerate(trainloader, 0):\n",
        "\n",
        "        inputs, labels = data\n",
        "        inputs, labels = inputs.to(device), labels.to(device)\n",
        "\n",
        "        opt.zero_grad()\n",
        "\n",
        "        outputs = vgg(inputs)\n",
        "        loss = loss_fn(outputs, labels)\n",
        "        loss.backward()\n",
        "        opt.step()\n",
        "        \n",
        "        if i % 100 == 0:\n",
        "            print('Iteration: %d/%d, Loss: %0.2f' % (i, n_iters, loss.item()))\n",
        "            \n",
        "        del inputs, labels, outputs\n",
        "        torch.cuda.empty_cache()\n",
        "        \n",
        "    loss_epoch_arr.append(loss.item())\n",
        "        \n",
        "    print('Epoch: %d/%d, Test acc: %0.2f, Train acc: %0.2f' % (\n",
        "        epoch, max_epochs, \n",
        "        evaluation(testloader, vgg), evaluation(trainloader, vgg)))\n",
        "    \n",
        "    \n",
        "plt.plot(loss_epoch_arr)\n",
        "plt.show()"
      ],
      "execution_count": 34,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Iteration: 0/3125, Loss: 2.32\n",
            "Iteration: 100/3125, Loss: 1.58\n",
            "Iteration: 200/3125, Loss: 1.99\n",
            "Iteration: 300/3125, Loss: 1.89\n",
            "Iteration: 400/3125, Loss: 1.76\n",
            "Iteration: 500/3125, Loss: 1.71\n",
            "Iteration: 600/3125, Loss: 3.02\n",
            "Iteration: 700/3125, Loss: 1.20\n",
            "Iteration: 800/3125, Loss: 1.16\n",
            "Iteration: 900/3125, Loss: 1.41\n",
            "Iteration: 1000/3125, Loss: 2.47\n",
            "Iteration: 1100/3125, Loss: 2.11\n",
            "Iteration: 1200/3125, Loss: 1.87\n",
            "Iteration: 1300/3125, Loss: 2.05\n",
            "Iteration: 1400/3125, Loss: 1.51\n",
            "Iteration: 1500/3125, Loss: 3.60\n",
            "Iteration: 1600/3125, Loss: 1.80\n",
            "Iteration: 1700/3125, Loss: 2.71\n",
            "Iteration: 1800/3125, Loss: 3.33\n",
            "Iteration: 1900/3125, Loss: 2.37\n",
            "Iteration: 2000/3125, Loss: 2.22\n",
            "Iteration: 2100/3125, Loss: 1.41\n",
            "Iteration: 2200/3125, Loss: 1.51\n",
            "Iteration: 2300/3125, Loss: 2.30\n",
            "Iteration: 2400/3125, Loss: 1.31\n",
            "Iteration: 2500/3125, Loss: 3.72\n",
            "Iteration: 2600/3125, Loss: 3.07\n",
            "Iteration: 2700/3125, Loss: 2.46\n",
            "Iteration: 2800/3125, Loss: 2.09\n",
            "Iteration: 2900/3125, Loss: 1.92\n",
            "Iteration: 3000/3125, Loss: 1.90\n",
            "Iteration: 3100/3125, Loss: 2.05\n",
            "Epoch: 0/1, Test acc: 46.06, Train acc: 46.37\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXoAAAD4CAYAAADiry33AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nUuMMeklq3P8DAihk5+eBn6sAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xlGjaKj3oTmT",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6-BcHgKaoTWP",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "OtH8vNhDwEIA",
        "colab_type": "text"
      },
      "source": [
        "<h3> Storing Intermediate Models </h3>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "EiAxsL-1oTHe",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import copy"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "UFrMk6NlwRwM",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "loss_epoch_arr = []"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "N1nkL-Snwmn8",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "max_epochs = 1\n",
        "vgg = vgg.to(device)\n",
        "loss_fn = nn.CrossEntropyLoss()\n",
        "opt = optim.SGD(vgg.parameters(), lr=0.05)\n",
        "n_iters = np.ceil(50000/batch_size)\n",
        "min_loss = 1000\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "IbSD6D5Vwmbq",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 790
        },
        "outputId": "dc73163a-5934-4def-ac01-758fb543bd20"
      },
      "source": [
        "for epoch in range(max_epochs):\n",
        "\n",
        "  for i , data in enumerate(trainloader,0):\n",
        "    inputs , labels = data\n",
        "    inputs , labels = inputs.to(device), labels.to(device)\n",
        "\n",
        "    opt.zero_grad()\n",
        "    outputs = vgg(inputs)\n",
        "    loss = loss_fn(outputs, labels)\n",
        "\n",
        "    loss.backward()\n",
        "    \n",
        "    opt.step()\n",
        "    \n",
        "    if min_loss > loss.item():\n",
        "            min_loss = loss.item()\n",
        "            best_model = copy.deepcopy(vgg.state_dict())\n",
        "            print('Min loss %0.2f' % min_loss)\n",
        "        \n",
        "\n",
        "    if i % 100 == 0:\n",
        "            print('Iteration: %d/%d, Loss: %0.2f' % (i, n_iters, loss.item()))\n",
        "            \n",
        "    del inputs, labels, outputs\n",
        "    torch.cuda.empty_cache()\n",
        "    \n",
        "  loss_epoch_arr.append(loss.item())"
      ],
      "execution_count": 54,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Min loss 2.32\n",
            "Iteration: 0/3125, Loss: 2.32\n",
            "Min loss 2.29\n",
            "Min loss 2.06\n",
            "Min loss 1.77\n",
            "Min loss 1.28\n",
            "Min loss 1.25\n",
            "Min loss 1.17\n",
            "Iteration: 100/3125, Loss: 1.51\n",
            "Min loss 1.00\n",
            "Min loss 0.93\n",
            "Iteration: 200/3125, Loss: 1.76\n",
            "Min loss 0.82\n",
            "Iteration: 300/3125, Loss: 1.52\n",
            "Iteration: 400/3125, Loss: 1.88\n",
            "Min loss 0.65\n",
            "Iteration: 500/3125, Loss: 1.99\n",
            "Min loss 0.62\n",
            "Iteration: 600/3125, Loss: 2.15\n",
            "Iteration: 700/3125, Loss: 1.53\n",
            "Iteration: 800/3125, Loss: 2.68\n",
            "Iteration: 900/3125, Loss: 2.56\n",
            "Iteration: 1000/3125, Loss: 2.31\n",
            "Iteration: 1100/3125, Loss: 2.24\n",
            "Iteration: 1200/3125, Loss: 1.02\n",
            "Iteration: 1300/3125, Loss: 2.17\n",
            "Iteration: 1400/3125, Loss: 2.40\n",
            "Iteration: 1500/3125, Loss: 1.47\n",
            "Iteration: 1600/3125, Loss: 1.59\n",
            "Iteration: 1700/3125, Loss: 1.34\n",
            "Iteration: 1800/3125, Loss: 1.66\n",
            "Min loss 0.61\n",
            "Iteration: 1900/3125, Loss: 1.19\n",
            "Iteration: 2000/3125, Loss: 3.57\n",
            "Iteration: 2100/3125, Loss: 1.90\n",
            "Iteration: 2200/3125, Loss: 2.10\n",
            "Iteration: 2300/3125, Loss: 1.83\n",
            "Iteration: 2400/3125, Loss: 0.88\n",
            "Iteration: 2500/3125, Loss: 2.02\n",
            "Iteration: 2600/3125, Loss: 1.17\n",
            "Iteration: 2700/3125, Loss: 2.13\n",
            "Iteration: 2800/3125, Loss: 2.05\n",
            "Iteration: 2900/3125, Loss: 2.43\n",
            "Iteration: 3000/3125, Loss: 1.14\n",
            "Min loss 0.58\n",
            "Iteration: 3100/3125, Loss: 2.51\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Ql-RAnhu8CDH",
        "colab_type": "text"
      },
      "source": [
        "<b> Note:</b> Without doing opt.zero_grad() gradients will just explode"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3PEcLXJMwl-m",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "a2c6015b-12ae-433c-b335-773c0216bab7"
      },
      "source": [
        "vgg.load_state_dict(best_model)\n",
        "print(evaluation(trainloader, vgg), evaluation(testloader, vgg))"
      ],
      "execution_count": 55,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "46.642 46.44\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vbwS0WZc877p",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}