FashionMNIST.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "FashionMNIST.ipynb",
      "version": "0.3.2",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "display_name": "Python 3",
      "language": "python",
      "name": "python3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "metadata": {
        "id": "NrXBHeOT5tKA",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "# Imports:"
      ]
    },
    {
      "metadata": {
        "id": "XVevzz7Z5tKD",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "from torch.utils.data import Dataset, DataLoader\n",
        "from torchvision import transforms, utils\n",
        "from torch.autograd import Variable\n",
        "import matplotlib.pyplot as plt\n",
        "import torch.optim as optim\n",
        "import torch.nn as nn\n",
        "import urllib.request\n",
        "import numpy as np\n",
        "import random\n",
        "import struct\n",
        "import torch\n",
        "import errno\n",
        "import math\n",
        "import gzip\n",
        "import io\n",
        "import os"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "NS3d_4i_5tKQ",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "# Dataset Class:"
      ]
    },
    {
      "metadata": {
        "id": "q7t2Hj4n5tKS",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "class Fashion(Dataset):\n",
        "\t\"\"\"Dataset: https://github.com/zalandoresearch/fashion-mnist\n",
        "\tArgs:\n",
        "\t\troot (string): Root directory of dataset where ``processed/training.pt``\n",
        "\t\t\tand  ``processed/test.pt`` exist.\n",
        "\t\ttrain (bool, optional): If True, creates dataset from ``training.pt``,\n",
        "\t\t\totherwise from ``test.pt``.\n",
        "\t\tdownload (bool, optional): If true, downloads the dataset from the internet and\n",
        "\t\t\tputs it in root directory. If dataset is already downloaded, it is not\n",
        "\t\t\tdownloaded again.\n",
        "\t\ttransform (callable, optional): A function/transform that takes in a numpy image\n",
        "\t\t\tand may return a horizontally flipped image.\"\"\"\n",
        "\n",
        "\turls = \t[\n",
        "\t'http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/train-images-idx3-ubyte.gz',\n",
        "\t'http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/train-labels-idx1-ubyte.gz',\n",
        "\t'http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/t10k-images-idx3-ubyte.gz',\n",
        "\t'http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/t10k-labels-idx1-ubyte.gz'\n",
        "\t\t\t]\n",
        "\n",
        "\tfile_name =\t[\n",
        "\t\t\t'train-images-idx3-ubyte',\n",
        "\t\t\t'train-labels-idx1-ubyte',\n",
        "\t\t\t't10k-images-idx3-ubyte',\n",
        "\t\t\t't10k-labels-idx1-ubyte'\n",
        "\t\t\t]\n",
        "\n",
        "\traw = \"raw\"\n",
        "\tprocesssed = \"processsed\"\n",
        "\n",
        "\n",
        "\tdef __init__(self, root, train=True, transform=True, download=False):\n",
        "\t\tsuper(Fashion, self).__init__()\n",
        "\t\tself.root = root\n",
        "\t\tself.transform = transform\n",
        "\t\tself.train = train\n",
        "\t\tself.tensor_transform = transforms.ToTensor()\n",
        "\n",
        "\t\traw_path = os.path.join(self.root,self.raw)\n",
        "\t\tif download and (os.path.exists(raw_path) == False):\n",
        "\t\t\tself.download(self.root)\n",
        "\n",
        "\t\tif self.train:\n",
        "\t\t\ttrain_path = os.path.join(self.root,self.processsed,\"training_set.pt\")\n",
        "\t\t\tself.train_images, self.train_labels = torch.load(train_path)\n",
        "\t\telse:\n",
        "\t\t\ttest_path = os.path.join(self.root,self.processsed,\"testing_set.pt\")\n",
        "\t\t\tself.test_images, self.test_labels = torch.load(test_path)\n",
        "\n",
        "\n",
        "\n",
        "\t'''\n",
        "\t__getitem__(index) -> Will return the image and label at the specified index\n",
        "\t\n",
        "\tIf transform parametr of class is set as True the function would or would not\n",
        "\tperform a random horizontal flip of the image.\n",
        "\t'''\n",
        "\tdef __getitem__(self,index):\n",
        "\t\tif self.train:\n",
        "\t\t\timage, label = self.train_images[index], self.train_labels[index]\n",
        "\t\telse:\n",
        "\t\t\timage, label = self.test_images[index], self.test_labels[index]\n",
        "\n",
        "\t\timage = image.numpy()\n",
        "\t\timage = np.rot90(image,axes = (1,2)).copy()\n",
        "\n",
        "\t\tif self.transform and self.train:\n",
        "\t\t\timage = self.transform_process(image)\n",
        "\n",
        "\t\timage = self.tensor_transform(image)\n",
        "\t\timage = image.contiguous()\n",
        "\t\timage = image.view(1,28,28)\n",
        "\t\t\t\n",
        "\t\treturn image,label\n",
        "\n",
        "\n",
        "\tdef __len__(self):\n",
        "\t\tif self.train:\n",
        "\t\t\treturn(len(self.train_images))\n",
        "\t\telse:\n",
        "\t\t\treturn(len(self.test_images))\n",
        "\n",
        "\n",
        "\tdef transform_process(self, image): #Would or would not return a flipped image\n",
        "\t\tself.rotate = random.getrandbits(1)\n",
        "\t\timage = np.flip(image,self.rotate).copy()\n",
        "\t\treturn image\n",
        "\n",
        "\t'''\n",
        "\tdownload(root) -> The function will download and save the MNIST images in raw\n",
        "\tformat under the 'raw' folder under the user specified root directory\n",
        "\t'''\n",
        "\n",
        "\tdef download(self, root):\n",
        "\t\traw_path = os.path.join(self.root,self.raw)\n",
        "\t\tprocesssed_path = os.path.join(self.root,self.processsed)\n",
        "\n",
        "\t\ttry:\n",
        "\t\t\tos.makedirs(raw_path)\n",
        "\t\t\tos.makedirs(processsed_path)\n",
        "\t\texcept OSError as exc:\n",
        "\t\t\tif exc.errno != errno.EEXIST:\n",
        "\t\t\t\traise\n",
        "\t\t\tpass\n",
        "\n",
        "\t\tfor file_index in range(len(self.file_name)):\n",
        "\t\t\tprint(\"Downloading:\",self.urls[file_index])\n",
        "\t\t\turllib.request.urlretrieve(self.urls[file_index],(self.file_name[file_index]+'.gz'))\n",
        "\t\t\tprint(\"Extracting:\",self.file_name[file_index]+\".gz\")\n",
        "\t\t\tf = gzip.open(self.file_name[file_index]+'.gz', 'rb')\n",
        "\t\t\twith open(raw_path+\"/\"+self.file_name[file_index],'wb') as w:\n",
        "\t\t\t\tfor line in f.readlines():\n",
        "\t\t\t\t\tw.write(line)\t\n",
        "\t\t\tf.close()\n",
        "\t\t\tos.remove(self.file_name[file_index]+\".gz\")\n",
        "\n",
        "\t\tprint()\n",
        "\t\tprint(\"Raw data downloaded and extracted in your specified root directory under /raw\")\n",
        "\t\tprint()\n",
        "\t\tself.process(self.root)\n",
        "\n",
        "\t'''\n",
        "\tprocess(root) -> Will process the raw downloaded files into a usable format\n",
        "\tand store them into the a 'processed' folder under user specified root\n",
        "\tdirectory.\n",
        "\t'''\n",
        "\n",
        "\tdef process(self, root):\n",
        "\t\traw_path = os.path.join(self.root,self.raw)\n",
        "\t\tprocesssed_path = os.path.join(self.root,self.processsed)\n",
        "\n",
        "\t\tprint(\"Processing training data\")\n",
        "\t\ttrain_image = self.readimg(self.root, self.file_name[0], 2051)\n",
        "\t\ttrain_label\t= self.readlab(self.root, self.file_name[1], 2049)\n",
        "\t\ttrain_data = (train_image,train_label)\n",
        "\n",
        "\t\tprint(\"Processing testing data\")\n",
        "\t\ttest_image = self.readimg(self.root, self.file_name[2], 2051)\n",
        "\t\ttest_label = self.readlab(self.root, self.file_name[3], 2049)\n",
        "\t\ttest_data = (test_image,test_label)\n",
        "\n",
        "\t\ttrain_path = os.path.join(self.root,self.processsed,\"training_set.pt\")\n",
        "\t\twith open(train_path,\"wb\") as f:\n",
        "\t\t\ttorch.save(train_data,f)\n",
        "\n",
        "\t\ttest_path = os.path.join(self.root,self.processsed,\"testing_set.pt\")\n",
        "\t\twith open(test_path,\"wb\") as f:\n",
        "\t\t\ttorch.save(test_data,f)\n",
        "\t\tprint()\n",
        "\t\tprint(\"Processed data has been stored in your specified root directory under /processsed\")\n",
        "\t\tprint()\n",
        "\n",
        "\n",
        "\tdef readimg(self, root, file, magic):\n",
        "\t\timage = []\n",
        "\t\tpath = os.path.join(self.root,self.raw,file)\n",
        "\t\twith open(path,'rb') as f:\n",
        "\t\t\tmagic_number, size, row, col = struct.unpack('>IIII',f.read(16))\n",
        "\t\t\tassert (magic_number == magic)\n",
        "\t\t\tfor run in range(size*row*col):\n",
        "\t\t\t\timage.append(list(struct.unpack('B',f.read(1)))[0])\n",
        "\t\t\timage = np.asarray(image, dtype = np.float32)\n",
        "\t\t\treturn (torch.from_numpy(image).view(size,1,row,col))\n",
        "\n",
        "\n",
        "\tdef readlab(self, root, file, magic):\n",
        "\t\tlabel = []\n",
        "\t\tpath = os.path.join(self.root,self.raw,file)\n",
        "\t\twith open(path,'rb') as f:\n",
        "\t\t\tmagic_number, size = struct.unpack(\">II\",f.read(8))\n",
        "\t\t\tassert (magic_number == magic)\n",
        "\t\t\tfor run in range(size):\n",
        "\t\t\t\tlabel.append(list(struct.unpack('b',f.read(1)))[0])\n",
        "\t\t\tlabel = np.asarray(label)\n",
        "\t\t\treturn (torch.from_numpy(label))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "78LYxft05tKZ",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "# Initialising Dataset Class"
      ]
    },
    {
      "metadata": {
        "id": "6y7BE7IP5tKb",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "outputId": "38c4c2b7-b45d-48f1-c739-e240076904f3"
      },
      "cell_type": "code",
      "source": [
        "train_dataset = Fashion(root = \"./FashionMNIS\", train = True, transform = True, download = True)\n",
        "test_dataset = Fashion(root = \"./FashionMNIS\", train = False, transform = False, download = True)"
      ],
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Downloading: http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/train-images-idx3-ubyte.gz\n",
            "Extracting: train-images-idx3-ubyte.gz\n",
            "Downloading: http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/train-labels-idx1-ubyte.gz\n",
            "Extracting: train-labels-idx1-ubyte.gz\n",
            "Downloading: http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/t10k-images-idx3-ubyte.gz\n",
            "Extracting: t10k-images-idx3-ubyte.gz\n",
            "Downloading: http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/t10k-labels-idx1-ubyte.gz\n",
            "Extracting: t10k-labels-idx1-ubyte.gz\n",
            "\n",
            "Raw data downloaded and extracted in your specified root directory under /raw\n",
            "\n",
            "Processing training data\n",
            "Processing testing data\n",
            "\n",
            "Processed data has been stored in your specified root directory under /processsed\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "i5pWBGNB5tKm",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "# Defining Batch Size and Total Iterations"
      ]
    },
    {
      "metadata": {
        "id": "pbeidJ-x5tKo",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "'''\n",
        "Calculation of total epochs using a defined batch size(batch_size)\n",
        "and total iterations(n_ters)\n",
        "'''\n",
        "\n",
        "batch_size = 100\n",
        "n_iters = 18000\n",
        "epoch_size = n_iters/(len(train_dataset)/batch_size)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "daUTWj3S5tKt",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "# Making the Test Dataset Iterable"
      ]
    },
    {
      "metadata": {
        "id": "ZSeysw_85tKw",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "\n",
        "'''\n",
        "Loading the test dataset (test_loader)\n",
        "'''\n",
        "\n",
        "test_loader = torch.utils.data.DataLoader(dataset = test_dataset, batch_size = batch_size, shuffle = False)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "Q5rbbH38CXfV",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "#MNIST Image Examples:"
      ]
    },
    {
      "metadata": {
        "id": "anmPrn-xCWeY",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 2231
        },
        "outputId": "afd16afd-b5e6-4e27-8c74-d907c9e17ed2"
      },
      "cell_type": "code",
      "source": [
        "image_dict = {0:'T-shirt/Top', 1:'Trouser', 2:'Pullover', 3:'Dress',\n",
        "\t\t\t  4:'Coat', 5:'Sandal', 6:'Shirt', 7:'Sneaker',\n",
        "\t\t\t  8:'Bag', 9:'Ankle Boot'}\n",
        "for index, (images,labels) in enumerate(test_loader):\n",
        "\timage=images[index][0]\n",
        "\tlabel = labels[index]\n",
        "\tplt.imshow(image)\n",
        "\tplt.suptitle(image_dict[label]+\" - \"+str(label))\n",
        "\tplt.show()\n",
        "\tif index == 5:\n",
        "\t\tbreak"
      ],
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAUsAAAFxCAYAAAD6VopVAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nABj8H50ymxjuJpyLAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7f7a334c8198>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAUsAAAFxCAYAAAD6VopVAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nBAADwhIADAhLADD4P2oUofLCWq99AAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7f79d9178780>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAUsAAAFxCAYAAAD6VopVAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nGPwfqEyvsEk0L0IAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7f79d98ef7f0>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAUsAAAFxCAYAAAD6VopVAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMS4yLCBo\ndHRwOi8vbWF0cGxvdGxpYi5vcmcvNQv5yAAAHnZJREFUeJzt3X9wFPX9x/FXJAK5JJgQSCijUUtD\nTSW0oqiBoiZQKFBHsT+oaXTsaAdHQSjjUGQAW+lXJCCjYKf8GNCO1JKZjLXWoSZSUSkDYUClJNYG\n0ApSEgOmgZCQhJDvH46Z3HG3eX+Ou1yCz8df3u6Hz372dvNyb/fe94lrb29vFwDA0yWxHgAA9AaE\nJQAYEJYAYEBYAoABYQkABoQlABgQlgBgEB/rAeDikJ+fr5qaGl1yyRf///X5fMrOztbMmTN1ww03\nxGxcp06d0vLly7Vt2zadPn1aV111lR566CFNmDAhZmNC78SVJSJm7ty52r9/v/bv36+tW7cqKytL\nDzzwgJqammI2pieffFIffPCBSkpKtHv3bhUWFmr27Nn66KOPYjYm9E6EJaIiOTlZP/7xj9XU1KTq\n6mpJUnt7u1avXq38/Hxdd911mjRpkv785z93/Ju2tjYtXbpUo0aN0pgxY7RhwwY9+OCDeuKJJ8Ie\nxz//+U/l5eUpIyND8fHxuuuuuyRJBw4cuLAdxFcOYYmo+Pzzz/XCCy/ouuuu05VXXilJeu211/SH\nP/xBGzZs0Lvvvqs5c+ZowYIF+vjjjyVJL774okpKSvTCCy9o69at+uijj/Tuu+9e0Djy8/NVVlam\nTz/9VG1tbSopKVFiYqJGjx59wfuIr5Y4asMRCYH3LFtaWpSZmamnn35aI0eOlPTFlWNjY6OSk5Ml\nfXGlmZOTo2XLlmnq1Kn6+c9/royMDD311FOSpIaGBo0bN07Tpk3T4sWLwxrXuXPn9Mgjj+iNN95Q\nXFyckpOTtXLlSo0bNy4Ce42vEh7wIGLmzp2r+++/X5LU1NSksrIy3XfffVq3bp1uuOEGNTc3q6io\nSNu2bVN9fb0kqbW1Vc3NzZKk2tpaXX/99R39JSUl6Rvf+MYFjWnJkiU6ceKE3njjDQ0aNEhlZWV6\n5JFHtHnzZn3zm9+8oL7x1cLHcERFQkKC7rjjDn33u9/Vpk2bJElPPPGEdu/erY0bN2rfvn3av3+/\n+vfv3/Fvzp07p759+/r18+WVajCTJk1STk6OcnJytHDhwvPWNzU1afPmzXr44YeVmZkpn8+nO++8\nU9nZ2frLX/4SoT3FVwVXloi6M2fOSJLef/99TZ06VcOHD5ckHTx4sGOdJKWlpenIkSMdrxsbG3Xw\n4EFde+21QfstLS313G5bW5vOnTuntrY2v+UtLS1h7Qe+2riyRFScPXtW27Zt09tvv93xBPryyy/X\n/v371dzcrAMHDuiZZ55RWlqaampqJEk333yzysrK9OGHH6qpqUlFRUXnXWm6SEpK0pgxY7Ru3Tod\nO3ZMLS0t2rJliyorK/meJZxxZYmIWblypZ555hlJUnx8vK666ir93//9nyZOnChJevTRR/WrX/1K\nN954o7KysvTb3/5Wr7/+un7/+98rMTFRDzzwgA4fPqzp06crNTVVDz/8sA4ePOj5UbwrK1as0IoV\nKzR9+nSdOnVKV155pVauXKlRo0ZFZJ/x1cHTcPQozc3N6tevX8fr733ve/rJT36iX/ziFzEcFcDH\ncPQgr732msaOHasPP/xQbW1tevnll3X06FHdeuutsR4awMdw9BxTp07Vf/7zHz344IOqr6/X5Zdf\nrqeffrrjgRAQS3wMBwADPoYDgAFhCQAGhCUAGBCWAGBAWAKAAWEJAAaEJQAYEJYAYEBYAoABYQkA\nBoQlABgQlgBgQFgCgAFhCQAGhCUAGBCWAGBAWAKAAWEJAAaEJQAYEJYAYEBYAoABYQkABoQlABgQ\nlgBgQFgCgAFhCQAG8eH+wyeffFL79u1TXFycFixYoJEjR0ZyXADQo4QVlrt379Ynn3yi4uJiHTp0\nSAsWLFBxcXGkxwYAPUZYH8N37typCRMmSJKGDRum+vp6NTQ0RHRgANCThBWWx48fV2pqasfrgQMH\nqra2NmKDAoCeJiIPeNrb2yPRDQD0WGGFZXp6uo4fP97x+rPPPtPgwYMjNigA6GnCCsuxY8eqtLRU\nklRZWan09HQlJSVFdGAA0JOE9TR81KhRuvbaa/XTn/5UcXFxevzxxyM9LgDoUeLaueEIAF2iggcA\nDAhLADAgLAHAgLAEAAPCEgAMCEsAMCAsAcCAsAQAA8ISAAwISwAwICwBwICwBAADwhIADAhLADAg\nLAHAgLAEAAPCEgAMCEsAMCAsAcCAsAQAA8ISAAzCmgq3J3CZlDIuLi6KIwEi56233jK3veKKK85b\nNmzYMB06dMhvmcv5X1lZaW67b98+U7uFCxea+zx9+nTQ5YmJieetS0xMNPcbCVxZAoABYQkABoQl\nABgQlgBgQFgCgAFhCQAGhCUAGBCWAGBAWAKAAWEJAAZx7S51g2Gqrq4OunzIkCHnrfvRj34U8e33\n7dvX3Pbjjz++oG19/PHHuvrqq/2WNTY2mv/9uXPnTO369Olj7tOl3C05OTno8qqqKg0fPrzjtXWc\nktTa2hrxti77dNlllwVd/sEHH+hb3/qW37J+/fqZ+gz1PgXT0NBgbvvee++Z21566aXnLWtpaTnv\nfHd5/9PS0sxtA8/zUAYNGmTuM9T+V1dXa8iQIX7Lvv/975v6fPzxx83b99onriwBwICwBAADwhIA\nDAhLADAgLAHAgLAEAAPCEgAMCEsAMCAsAcCgWyYsKykpCbp85syZ562rqKgw9Tl06FDz9l2KlJKS\nksxtrX34fD7zvz179qypXbQmbIuPD31KdK4YCTWxVDAuY7VWhWRmZpr7/Pe//x1yXf/+/f1eW/fr\nzJkz5u1bq4Ik6aabbjK3DVXFNXr0aL/X1nNKchtrU1OTqV1tba25z85VYl2te/311019BlZpeZk3\nb17IdVxZAoABYQkABmF9DC8vL9fs2bOVlZUl6YvL40WLFkV0YADQk4R9z/LGG2/UqlWrIjkWAOix\n+BgOAAZh/Z5leXm5fvOb3ygzM1P19fWaOXOmxo4dG7L9iRMnnH4nDwB6mrDCsqamRnv37tXkyZN1\n5MgR3XvvvSorKwv5I7vPPfdc0OUzZ848b93ChQtNY4jWV4e8vjpjsX//fuXk5Pgtc/mh3J761aHK\nykpde+21Ha9dvjrksv9XXHGFqV0kvjr07rvvatSoUX7LrPt1ySX2D2UuX8cJ/CqTl2BfHdqxY8d5\nFy6x/uqQy7ka6mt277zzjm655Ra/ZVVVVaY+586da95+xL86lJGRoSlTpiguLk6ZmZkaNGiQampq\nwukKAHqFsMLy1Vdf1YYNGyR98YXTEydOKCMjI6IDA4CeJKzPnPn5+Xr00Uf197//Xa2trfr1r3/t\nNM8NAPQ2YYVlUlKS1qxZY24/btw48zpraeDXvvY18/ZdStOsE4F53YcLnCDL5Z6RtW007oNK3veX\nOt+nS0xMNPfpcn/Pei/Sem9T8p6ELtz/yaempprbNjc3m9u6vFehJiILPDdc7sO73N+2PrQNNrFa\nKF6TqwVmg/W5xTXXXGPevhe+OgQABoQlABgQlgBgQFgCgAFhCQAGhCUAGBCWAGBAWAKAAWEJAAaE\nJQAYdMvsjl7lVoHrjh07ZupzxIgREdl+IGv5W1tbm7kPl5I6r347cymLs/Ypef9E15AhQzr++8SJ\nE+Y+6+rqzG3feOMNUzuXn4jzKosLnHnw61//uqlPl59SS0hIMLd1Eeq4pqSk+L12mV3U5X21ljG6\nnKuDBw82r9uxY4epz8bGRvP2vXBlCQAGhCUAGBCWAGBAWAKAAWEJAAaEJQAYEJYAYEBYAoABYQkA\nBt1SweNVwRLuhFEuVQkuE0ZZq128JvZymVQ+kMvkYlYu1SbWaiuXfXSZ3Ovqq682tXPZvlcFUXJy\nst/rgwcPmvp0qSBzqaByOa9DTVoXeLxdJrdzqXZraGgwtXM5/7zeq8B11skFXSZh88KVJQAYEJYA\nYEBYAoABYQkABoQlABgQlgBgQFgCgAFhCQAGhCUAGBCWAGDQLeWOXmVJ1pKlQJEqYQpkLY30Gndg\nWZZ1YifJe8KwzlpaWsx9ukwYlZSUZFr36aefmvucNWuWue3evXtN7aqrq8193n333eZ1v/vd70x9\nRqOEVnIrN7S2PXPmjLlPlzJS69+gSwmvV5+B66zntcv579lPRHoBgIscYQkABoQlABgQlgBgQFgC\ngAFhCQAGhCUAGBCWAGBAWAKAAWEJAAbdUu7oMmOblUuZpEsJV1NTk6mdVwlVY2Oj3+vLLrvMvH3r\nbJcuJWQubevr68Na58WlNNP6XrmcN16zGwau++9//2vq8/rrrzdv33pOSdLJkyfNbRMTE4MuDyzF\nPHXqlLlPl9LcAQMGmNq5bN9l1lTKHQGgBzKFZVVVlSZMmKBNmzZJko4dO6Z77rlHBQUFmj17ttOV\nAwD0Rl2GZWNjo5YsWaLc3NyOZatWrVJBQYFeeuklXXnllSopKYnqIAEg1roMy759+2r9+vVKT0/v\nWFZeXq7x48dLkvLy8rRz587ojRAAeoC4duPTj9WrVys1NVWFhYXKzc3tCMjDhw9r3rx52rx5c8h/\n29zcbP6dRgDoiS74abglaw8fPhx0eVZWlg4cOOC3bPjw4abt/vCHPzS1k6TPPvvM3LahocHULtQT\ntj179uiGG27wW+byNNz6NDLwibsXl28OhNp+WVmZJk6c2PH6vffeM/c5Y8YMc9tQ50qgzz//3Nzn\n2LFjgy5/7LHHtHTpUr9lixYtMvV5++23m7fv8jTcpW2wp+FbtmzRlClT/Ja5PGGP9dPwIUOGBF3+\n0ksvqaCgwG9ZWVmZqc+1a9eat++VK2E9Dff5fB2/vlxTU+P3ER0ALkZhheWYMWNUWloq6Yt0Hzdu\nXEQHBQA9TZcfwysqKrRs2TIdPXpU8fHxKi0t1YoVKzR//nwVFxdr6NChuvPOO7tjrAAQM12G5YgR\nI/Tiiy+et/z55583b8TrPojLPZLO6urqzG1dvgdqHY9XVUjgxErWqhzJXm3kMmFba2urua1XZUzn\nia9uuukmc59/+tOfzG0D70uF8p3vfMfc5/Lly4Muf+yxx7Rq1Sq/ZTk5OeZ+raz3wV35fD7Tcuu9\nRcn7vA5k/VtxqbZymbDMWpkWqe+BU8EDAAaEJQAYEJYAYEBYAoABYQkABoQlABgQlgBgQFgCgAFh\nCQAGhCUAGHTLhGUuJUxWLpNwBU7g5CU5OdnULj4+9FsX2IfL9qPB5SfavParc3mbV7tAo0ePNrd9\n5ZVXTO1cStiuueYa87qEhARTny7nX6iyxGBcJtcLNYbA5S5/Yy77ZS2NdPk5Qa/3P/CYW/+uXN5T\nL1xZAoABYQkABoQlABgQlgBgQFgCgAFhCQAGhCUAGBCWAGBAWAKAAWEJAAbdUu7oVRblMptcZy4l\nfP379ze3tY7HqywscJ1LaaC1NC1a+3/y5MmQ6zqXl7ls36XcbNiwYaZ2LiWkXrNrBs58aC2jdJmx\n0KWE0OW9CvUeBC7vPCtnuH0G43JeW3mdV4HrrOdgYmLiBY3pS1xZAoABYQkABoQlABgQlgBgQFgC\ngAFhCQAGhCUAGBCWAGBAWAKAQbdU8ESDSwWJVwVHIGtlhsskbC6VDtYKDpfKp84TjXUlKSnJtK6p\nqcncp0sFRX19vamdS6WLVwXTJZf4Xy9Yq51czj+Xti4VNKH6DVzuUm3kcq5a20bqbzVwnbXazeX8\n98KVJQAYEJYAYEBYAoABYQkABoQlABgQlgBgQFgCgAFhCQAGhCUAGBCWAGDQLeWOXmVR4U565FLC\nZZ2ESrKXu3m18/l8fq9dJqxy2S+r1tZWc1vrsQrcRy8upZnW99+l3NFlcjlruWG0JiyLRhlvtMYa\njQnLXCY3jMbfiheuLAHAwBSWVVVVmjBhgjZt2iRJmj9/vm6//Xbdc889uueee/TWW29Fc4wAEHNd\nXkc3NjZqyZIlys3N9Vs+d+5c5eXlRW1gANCTdHll2bdvX61fv17p6endMR4A6JHi2o13ylevXq3U\n1FQVFhZq/vz5qq2tVWtrq9LS0rRo0SINHDgw5L9tbW2N2G/KAUAshPU464477lBKSoqys7O1bt06\nPffcc1q8eHHI9tXV1UGXX3HFFTpy5IjfsszMTNMYXG4BuPxQ7YU+Df/b3/6myZMn+y2L9dNw6z5J\n5/8Y7pdee+01/eAHP+h47TJOlx9ftj65d3kaHuqp7V//+lfdfvvtfst609PwYN9ICDxOktu3QVy+\nORGNby4MGjQo6PJNmzapsLDQb9nLL79s6rOkpMS8/SlTpoRcF9bT8NzcXGVnZ0uS8vPzVVVVFU43\nANBrhBWWs2bN6rgiLC8vV1ZWVkQHBQA9TZcfwysqKrRs2TIdPXpU8fHxKi0tVWFhoebMmaOEhAT5\nfD4tXbq0O8YKADHTZViOGDFCL7744nnLJ02aFJUBAUBP1C3lji4ztkVDqIcWwVhvhnuVZTU2Nvq9\n7tevn3n70ZgxLxqlgdaZ9SS3GQtjzfr+u7ynkXgYFYx1dkeXc+X06dPmttZvuLhs36Xc0XoOJiQk\nmLfvhXJHADAgLAHAgLAEAAPCEgAMCEsAMCAsAcCAsAQAA8ISAAwISwAwICwBwKBbyh29ypJcyuY6\nc/k9QZdtWEsjvUq4Ate5lLtZdffMdpLbLI2duZTwWbfhMhav32gMXBfu+RgpLudKqHPgQs6NcI+x\nF5f31Gv/A9dZxxqpvz+uLAHAgLAEAAPCEgAMCEsAMCAsAcCAsAQAA8ISAAwISwAwICwBwKBbKni8\nqmJcJhMLl8uESdEQjaoIl/ctUtuPRdVQJHhNQhe4zjqBXrQqfVz6DXVco3G+BWOtjHE5V6NRwROp\nY8WVJQAYEJYAYEBYAoABYQkABoQlABgQlgBgQFgCgAFhCQAGhCUAGBCWAGDQaycscylhdGl79uzZ\ncIbjyaXcy1pS6DJOl/fYa/udt+myTy7vv7XfaJX0XXrppRHv06VMNBKlgYHLo3X+W/crUudf4Dpr\nuSXljgDQjQhLADAgLAHAgLAEAAPCEgAMCEsAMCAsAcCAsAQAA8ISAAwISwAw6LWzO7qUMDU3N5vb\nWkvDvMrCAtdZy7IkexlfpGbMc+m387pIlbAFso7VpdzRpYTuzJkzpj5jPWOoFHq/Ape7nP8+n8/c\n1noOtLa2mvscMGBAyHWBM29aS1MjNSspV5YAYGC6siwqKtLevXt19uxZzZgxQzk5OZo3b57a2to0\nePBgLV++3DzfMgD0Rl2G5a5du3TgwAEVFxerrq5O06ZNU25urgoKCjR58mStXLlSJSUlKigo6I7x\nAkBMdPkxfPTo0Xr22WclfXE/oampSeXl5Ro/frwkKS8vTzt37ozuKAEgxuLaHe7+FxcXa8+ePfrH\nP/7REZCHDx/WvHnztHnz5pD/rq2trUfcEAeAcJmfhm/dulUlJSXauHGjJk6c2LHckrUnT54Mujw1\nNVV1dXV+ywYOHGgaT35+vqmd5PY0zhrqoZ6wvfPOO7rlllvC6lOKztPwSNi2bZvy8vI6Xrs8DXdp\nG40f/w31zYW3335bt956q98y6713l2Pqcv65/PhwsPegrKzM7+9Tis4P+kpSfLwtPlz2KS0tLejy\nP/7xj/rZz37mt+yVV14x9fnqq6+at//lJ+ZgTGfm9u3btWbNGq1fv17Jycny+XwdX7GoqalRenq6\neTAA0Bt1GZanTp1SUVGR1q5dq5SUFEnSmDFjVFpaKumL/5ONGzcuuqMEgBjr8jp6y5Ytqqur05w5\nczqWPfXUU1q4cKGKi4s1dOhQ3XnnnVEdJADEWpdhOX36dE2fPv285c8//7x5I173NQPXZWZmmvp0\nuWflcs8oGlzu2VnvhblU5URDpCaBCpfLPdt+/fqFtc6Ly/kXjQouyV7B47J9lwoe67na0tJi7rOh\nocG8zlptFalJ6KjgAQADwhIADAhLADAgLAHAgLAEAAPCEgAMCEsAMCAsAcCAsAQAA8ISAAy6ZcIy\nl3JHaxmdS7mby5QXkZgwK3Dd6dOnzdu37pfLT4S5lHt5lYZ2XufyU17R+Dk5l3JLr/0PtxQ2Ej8R\nF4zL+xrqHAg8h13G6jK5mZXL35/Xz74Frvv2t79t6tPl/ffClSUAGBCWAGBAWAKAAWEJAAaEJQAY\nEJYAYEBYAoABYQkABoQlABgQlgBg0C3ljseOHQu6PC0t7bx1x48fN/XpUkLn0vbQoUPmtqHs2LHD\n77VLaZ61rVdZWCCX0kiv2Q4rKyvD6tNFd88aWVFR4ffaeq64lBBGYsbGYEKVapaXl/u9djn/XY6r\ndawu5YYjRowIue7w4cN+r6urq019vv/+++bt5+fnh1zHlSUAGBCWAGBAWAKAAWEJAAaEJQAYEJYA\nYEBYAoABYQkABoQlABjEtVtn6Oom//rXv0ztBg0aZO6zf//+EW8bqoImLi4u7EnYgIuRy6RwLS0t\nQZcnJiaeN/Hf//73P1OfQ4YMMW/fq4KJK0sAMCAsAcCAsAQAA8ISAAwISwAwICwBwICwBAADwhIA\nDAhLADAgLAHAoMeVOwJAT8SVJQAYmOZTLSoq0t69e3X27FnNmDFDb775piorK5WSkiJJuv/++3Xb\nbbdFc5wAEFNdhuWuXbt04MABFRcXq66uTtOmTdPNN9+suXPnKi8vrzvGCAAx12VYjh49WiNHjpQk\nDRgwQE1NTU4TwQPAxcDpAU9xcbH27NmjPn36qLa2Vq2trUpLS9OiRYs0cODAaI4TAGLKHJZbt27V\n2rVrtXHjRlVUVCglJUXZ2dlat26dqqurtXjx4miPFQBixvQ0fPv27VqzZo3Wr1+v5ORk5ebmKjs7\nW5KUn5+vqqqqqA4SAGKty7A8deqUioqKtHbt2o6n37NmzdKRI0ckSeXl5crKyoruKAEgxrp8wLNl\nyxbV1dVpzpw5HcvuuusuzZkzRwkJCfL5fFq6dGlUBwkAsUYFDwAYUMEDAAaEJQAYEJYAYEBYAoAB\nYQkABoQlABgQlgBgQFgCgAFhCQAGhCUAGBCWAGBAWAKAAWEJAAaEJQAYEJYAYEBYAoABYQkABoQl\nABgQlgBgQFgCgAFhCQAGhCUAGBCWAGBAWAKAAWEJAAaEJQAYEJYAYEBYAoABYQkABvGx2OiTTz6p\nffv2KS4uTgsWLNDIkSNjMYyIKi8v1+zZs5WVlSVJGj58uBYtWhTjUYWvqqpKDz30kO677z4VFhbq\n2LFjmjdvntra2jR48GAtX75cffv2jfUwnQTu0/z581VZWamUlBRJ0v3336/bbrsttoN0VFRUpL17\n9+rs2bOaMWOGcnJyev1xks7frzfffDPmx6rbw3L37t365JNPVFxcrEOHDmnBggUqLi7u7mFExY03\n3qhVq1bFehgXrLGxUUuWLFFubm7HslWrVqmgoECTJ0/WypUrVVJSooKCghiO0k2wfZKkuXPnKi8v\nL0ajujC7du3SgQMHVFxcrLq6Ok2bNk25ubm9+jhJwffr5ptvjvmx6vaP4Tt37tSECRMkScOGDVN9\nfb0aGhq6exjw0LdvX61fv17p6ekdy8rLyzV+/HhJUl5ennbu3Bmr4YUl2D71dqNHj9azzz4rSRow\nYICampp6/XGSgu9XW1tbjEcVg7A8fvy4UlNTO14PHDhQtbW13T2MqDh48KAefPBB3X333dqxY0es\nhxO2+Ph49e/f329ZU1NTx8e5tLS0XnfMgu2TJG3atEn33nuvfvnLX+rzzz+PwcjC16dPH/l8PklS\nSUmJbrnlll5/nKTg+9WnT5+YH6uY3LPsrL29PdZDiIirrrpKM2fO1OTJk3XkyBHde++9Kisr65X3\ni7pysRyzO+64QykpKcrOzta6dev03HPPafHixbEelrOtW7eqpKREGzdu1MSJEzuW9/bj1Hm/Kioq\nYn6suv3KMj09XcePH+94/dlnn2nw4MHdPYyIy8jI0JQpUxQXF6fMzEwNGjRINTU1sR5WxPh8Pp05\nc0aSVFNTc1F8nM3NzVV2drYkKT8/X1VVVTEekbvt27drzZo1Wr9+vZKTky+a4xS4Xz3hWHV7WI4d\nO1alpaWSpMrKSqWnpyspKam7hxFxr776qjZs2CBJqq2t1YkTJ5SRkRHjUUXOmDFjOo5bWVmZxo0b\nF+MRXbhZs2bpyJEjkr64J/vlNxl6i1OnTqmoqEhr167teEp8MRynYPvVE45VXHsMrtVXrFihPXv2\nKC4uTo8//riuueaa7h5CxDU0NOjRRx/VyZMn1draqpkzZ+rWW2+N9bDCUlFRoWXLluno0aOKj49X\nRkaGVqxYofnz56u5uVlDhw7V0qVLdemll8Z6qGbB9qmwsFDr1q1TQkKCfD6fli5dqrS0tFgP1ay4\nuFirV6/W1Vdf3bHsqaee0sKFC3vtcZKC79ddd92lTZs2xfRYxSQsAaC3oYIHAAwISwAwICwBwICw\nBAADwhIADAhLADAgLAHA4P8BJuHjxRl/sDUAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7f79d9444c18>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAUsAAAFxCAYAAAD6VopVAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7f79d9916240>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAUsAAAFxCAYAAAD6VopVAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAAaEJQAYEJYAYPD/4YsW0EIts/wAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7f79d9171438>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "id": "wU40eaaj5tK3",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "# Neurel Network Model"
      ]
    },
    {
      "metadata": {
        "id": "UFqFQipc5tK5",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "'''\n",
        "Model Details:\n",
        "\n",
        "Two Convolutional Layers:\n",
        "      - Using ReLU activation\n",
        "      - Batch Normalisation\n",
        "      - Uniform Xavier Weigths\n",
        "      - Max Pooling\n",
        "      \n",
        "One Fully Connected Layer:\n",
        "      - Using ReLU activation\n",
        "      \n",
        "One Fully Connected Layer:\n",
        "      - Output Layer\n",
        "\n",
        "'''\n",
        "\n",
        "\n",
        "class CNNModel(nn.Module):\n",
        "\tdef __init__(self):\n",
        "\t\tsuper (CNNModel, self).__init__()\n",
        "\n",
        "\t\tself.cnn1 = nn.Conv2d(in_channels = 1, out_channels = 32, kernel_size = 5, stride = 1, padding = 2)\n",
        "\t\tself.relu1 = nn.ReLU()\n",
        "\t\tself.norm1 = nn.BatchNorm2d(32)\n",
        "\t\tnn.init.xavier_uniform(self.cnn1.weight)\n",
        "\n",
        "\t\tself.maxpool1 = nn.MaxPool2d(kernel_size=2)\n",
        "\n",
        "\t\tself.cnn2 = nn.Conv2d(in_channels = 32, out_channels = 64, kernel_size = 3, stride = 1, padding = 2)\n",
        "\t\tself.relu2 = nn.ReLU()\n",
        "\t\tself.norm2 = nn.BatchNorm2d(64)\n",
        "\t\tnn.init.xavier_uniform(self.cnn2.weight)\n",
        "\n",
        "\t\tself.maxpool2 = nn.MaxPool2d(kernel_size=2)\n",
        "\n",
        "\t\tself.fc1 = nn.Linear(4096, 4096)\n",
        "\t\tself.fcrelu = nn.ReLU()\n",
        "\n",
        "\t\tself.fc2 = nn.Linear(4096, 10)\n",
        "\n",
        "\tdef forward(self, x):\n",
        "\t\tout = self.cnn1(x)\n",
        "\t\tout = self.relu1(out)\n",
        "\t\tout = self.norm1(out)\n",
        "\n",
        "\t\tout = self.maxpool1(out)\n",
        "\n",
        "\t\tout = self.cnn2(out)\n",
        "\t\tout = self.relu2(out)\n",
        "\t\tout = self.norm2(out)\n",
        "\n",
        "\t\tout = self.maxpool2(out)\n",
        "\n",
        "\t\tout = out.view(out.size(0),-1)\n",
        "\n",
        "\t\tout = self.fc1(out)\n",
        "\t\tout = self.fcrelu(out)\n",
        "\n",
        "\t\tout = self.fc2(out)\n",
        "\t\treturn out"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "G4gocgl15tLA",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "# Instantiate Model Class"
      ]
    },
    {
      "metadata": {
        "id": "ouuRSYNG5tLC",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "'''\n",
        "Instantiating the model class\n",
        "'''\n",
        "\n",
        "model = CNNModel()\n",
        "if torch.cuda.is_available():\n",
        "\tmodel.cuda()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "6IJ2ioiL5tLL",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "# Instantiate Loss and Optimizer Class"
      ]
    },
    {
      "metadata": {
        "id": "P-U2_VqD5tLN",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "\n",
        "\n",
        "'''\n",
        "Loss Function: Cross Entropy Loss\n",
        "Optimizer: Stochastic Gradient Descent (Nesterov Momentum is enabled).\n",
        "\n",
        "Variable: learning_rate -> Stores the learning rate for the optimizer function.\n",
        "          moment -> Stores the momentum for the optimizer function.\n",
        "'''\n",
        "\n",
        "criterion = nn.CrossEntropyLoss()\n",
        "\n",
        "learning_rate = 0.015\n",
        "moment = 0.9\n",
        "optimizer = optim.SGD(model.parameters(),lr = learning_rate, momentum = moment, nesterov = True)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "oLMso3jA5tLY",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "# Training the Model\n",
        "Loading the training dataset after every epoch which will load it from the Fashion Dataset Class making a new train loader for every new epoch causing random flips and random shuffling of above exampled Fashion MNIST images. "
      ]
    },
    {
      "metadata": {
        "id": "0e1fhDnd5tLY",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 121
        },
        "outputId": "d8d542a5-ff70-4f02-e125-771052de6f5c"
      },
      "cell_type": "code",
      "source": [
        "iter = 0\n",
        "for epoch in range(int(math.ceil(epoch_size))):\n",
        "\n",
        "\t'''\n",
        "\tLoading the training dataset after every epoch which will\n",
        "\tload it from the Fashion Dataset Class making a new train\n",
        "\tloader for every new epoch causing random flips and random\n",
        "\tshuffling of above exampled Fashion MNIST images. \n",
        "\t'''\n",
        "\n",
        "\ttrain_loader = torch.utils.data.DataLoader(dataset = train_dataset, batch_size = batch_size, shuffle = True)\n",
        "\tfor i, (images, labels) in enumerate(train_loader):\n",
        "\t\tif torch.cuda.is_available():\n",
        "\t\t\timages = Variable(images.cuda())\n",
        "\t\t\tlabels = Variable(labels.cuda())\n",
        "\t\telse:\n",
        "\t\t\timages = Variable(images)\n",
        "\t\t\tlabels = Variable(labels)\n",
        "\t\toptimizer.zero_grad()\n",
        "\t\toutputs = model(images)\n",
        "\t\tloss = criterion(outputs,labels)\n",
        "\t\tloss.backward()\n",
        "\t\toptimizer.step()\n",
        "\t\titer += 1\n",
        "\n",
        "\t\t'''\n",
        "\t\tAt every 3000th epoch a test on the above initialised test dataset\n",
        "\t\t(test_loader) would be performed and an accuracy would be provided.\n",
        "\t\t'''\n",
        "\n",
        "\t\tif iter%3000 == 0:\n",
        "\t\t\tcorrect = 0\n",
        "\t\t\ttotal = 0\n",
        "\t\t\tfor image,label in test_loader:\n",
        "\t\t\t\tif torch.cuda.is_available():\n",
        "\t\t\t\t\timage = Variable(image.cuda())\n",
        "\t\t\t\telse:\n",
        "\t\t\t\t\timage = Variable(image)\n",
        "\t\t\t\toutput = model(image)\n",
        "\t\t\t\t_, predicted = torch.max(output.data,1)\n",
        "\t\t\t\ttotal += label.size(0)\n",
        "\t\t\t\tif torch.cuda.is_available():\n",
        "\t\t\t\t\tcorrect += (predicted.cpu() == label.cpu()).sum()\n",
        "\t\t\t\telse:\n",
        "\t\t\t\t\tcorrect += (predicted == label).sum()\n",
        "\n",
        "\t\t\taccuracy = 100 * (correct/total)\n",
        "\t\t\tprint('Iteration: {} Loss: {} Accuracy: {}'.format(iter, round(loss.data[0],5), round(accuracy,2)))"
      ],
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Iteration: 3000 Loss: 0.09267 Accuracy: 91.71\n",
            "Iteration: 6000 Loss: 0.17845 Accuracy: 92.2\n",
            "Iteration: 9000 Loss: 0.04771 Accuracy: 92.66\n",
            "Iteration: 12000 Loss: 0.00568 Accuracy: 93.12\n",
            "Iteration: 15000 Loss: 0.00323 Accuracy: 93.29\n",
            "Iteration: 18000 Loss: 0.00092 Accuracy: 93.42\n"
          ],
          "name": "stdout"
        }
      ]
    }
  ]
}