3_Training_FashionMNIST.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "mI57AoYm0Qpl"
   },
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "import torch\n",
    "import torchvision\n",
    "import torchvision.transforms as transforms\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "import torch.optim as optim\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "import sys\n",
    "sys.path.insert(0, '/home/sprakashdash/Documents/DLFA/')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "C_VM9tdh1mpn"
   },
   "outputs": [],
   "source": [
    "# Hyperparameters\n",
    "batch_size = 64\n",
    "input_size = 784\n",
    "hidden_size_1 = 512 # 4096\n",
    "hidden_size_2 = 256 # 2048\n",
    "hidden_size_3 = 128 # 1024\n",
    "num_classes = 10\n",
    "\n",
    "# Learning rate\n",
    "lr = 0.0001\n",
    "# Number of training epochs\n",
    "num_epochs = 20\n",
    "# Device configuration\n",
    "device = torch.device(\"cuda:0\" if torch.cuda.is_available() else \"cpu\")\n",
    "\n",
    "# Path to dataset\n",
    "dataset_root_path = '/home/sprakashdash/Documents/DLFA/'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "eheeSXs7yY_f",
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "0it [00:00, ?it/s]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Downloading http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/train-images-idx3-ubyte.gz to /home/sprakashdash/Documents/DLFA/FashionMNIST/raw/train-images-idx3-ubyte.gz\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "  0%|          | 0/26421880 [00:01<?, ?it/s]\u001b[A\n",
      "  0%|          | 90112/26421880 [00:01<00:54, 485049.19it/s]\u001b[A\n",
      "  1%|          | 180224/26421880 [00:01<00:52, 498666.93it/s]\u001b[A\n",
      "  1%|▏         | 335872/26421880 [00:01<00:45, 569730.48it/s]\u001b[A\n",
      "  2%|▏         | 524288/26421880 [00:01<00:39, 659227.63it/s]\u001b[A\n",
      "  3%|▎         | 835584/26421880 [00:01<00:29, 859211.23it/s]\u001b[A\n",
      "  4%|▍         | 1138688/26421880 [00:01<00:23, 1089145.02it/s]\u001b[A\n",
      "  6%|▌         | 1540096/26421880 [00:02<00:18, 1365407.35it/s]\u001b[A\n",
      "  7%|▋         | 1769472/26421880 [00:02<00:16, 1533281.49it/s]\u001b[A\n",
      "  8%|▊         | 2097152/26421880 [00:02<00:13, 1820794.16it/s]\u001b[A\n",
      "  9%|▉         | 2506752/26421880 [00:02<00:11, 2161548.85it/s]\u001b[A\n",
      " 11%|█         | 2801664/26421880 [00:02<00:11, 2062351.92it/s]\u001b[A\n",
      " 12%|█▏        | 3088384/26421880 [00:02<00:10, 2200230.44it/s]\u001b[A\n",
      " 14%|█▍        | 3751936/26421880 [00:02<00:08, 2747861.71it/s]\u001b[A\n",
      " 17%|█▋        | 4382720/26421880 [00:02<00:06, 3306056.84it/s]\u001b[A\n",
      " 20%|█▉        | 5177344/26421880 [00:02<00:05, 4007502.18it/s]\u001b[A\n",
      " 23%|██▎       | 6029312/26421880 [00:03<00:04, 4342645.75it/s]\u001b[A\n",
      " 25%|██▌       | 6660096/26421880 [00:03<00:04, 4789918.25it/s]\u001b[A\n",
      " 27%|██▋       | 7241728/26421880 [00:03<00:03, 4887027.50it/s]\u001b[A\n",
      " 30%|██▉       | 7798784/26421880 [00:03<00:04, 4473344.99it/s]\u001b[A\n",
      " 32%|███▏      | 8527872/26421880 [00:03<00:03, 5058212.83it/s]\u001b[A\n",
      " 36%|███▌      | 9379840/26421880 [00:03<00:02, 5753365.10it/s]\u001b[A\n",
      " 39%|███▊      | 10207232/26421880 [00:03<00:02, 6330694.72it/s]\u001b[A\n",
      " 41%|████▏     | 10919936/26421880 [00:03<00:02, 6117897.55it/s]\u001b[A\n",
      " 45%|████▌     | 11935744/26421880 [00:04<00:02, 6941768.82it/s]\u001b[A\n",
      " 50%|████▉     | 13115392/26421880 [00:04<00:01, 7914937.36it/s]\u001b[A\n",
      " 53%|█████▎    | 14008320/26421880 [00:04<00:01, 6401017.27it/s]\u001b[A\n",
      " 57%|█████▋    | 15171584/26421880 [00:04<00:01, 7395902.83it/s]\u001b[A\n",
      " 61%|██████    | 16056320/26421880 [00:04<00:01, 6337750.67it/s]\u001b[A\n",
      " 64%|██████▎   | 16818176/26421880 [00:04<00:01, 5596761.36it/s]\u001b[A\n",
      " 66%|██████▌   | 17481728/26421880 [00:04<00:01, 4669492.57it/s]\u001b[A\n",
      " 68%|██████▊   | 18046976/26421880 [00:05<00:03, 2588993.10it/s]\u001b[A\n",
      " 72%|███████▏  | 19136512/26421880 [00:05<00:02, 3354986.86it/s]\u001b[A\n",
      " 76%|███████▋  | 20152320/26421880 [00:05<00:01, 4197848.89it/s]\u001b[A\n",
      " 80%|███████▉  | 21012480/26421880 [00:05<00:01, 4947121.30it/s]\u001b[A\n",
      " 82%|████████▏ | 21790720/26421880 [00:05<00:00, 5432106.37it/s]\u001b[A\n",
      " 85%|████████▌ | 22552576/26421880 [00:05<00:00, 5280718.78it/s]\u001b[A\n",
      " 88%|████████▊ | 23232512/26421880 [00:06<00:00, 5195569.04it/s]\u001b[A\n",
      " 90%|█████████ | 23904256/26421880 [00:06<00:00, 4989080.65it/s]\u001b[A\n",
      " 93%|█████████▎| 24510464/26421880 [00:06<00:00, 5261508.37it/s]\u001b[A\n",
      " 95%|█████████▍| 25100288/26421880 [00:06<00:00, 4124988.36it/s]\u001b[A\n",
      " 97%|█████████▋| 25591808/26421880 [00:06<00:00, 3634758.82it/s]\u001b[A\n",
      " 99%|█████████▊| 26025984/26421880 [00:06<00:00, 3486031.22it/s]\u001b[A\n",
      "26427392it [00:06, 3794659.76it/s]                              \u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Extracting /home/sprakashdash/Documents/DLFA/FashionMNIST/raw/train-images-idx3-ubyte.gz\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "0it [00:00, ?it/s]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Downloading http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/train-labels-idx1-ubyte.gz to /home/sprakashdash/Documents/DLFA/FashionMNIST/raw/train-labels-idx1-ubyte.gz\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "  0%|          | 0/29515 [00:00<?, ?it/s]\u001b[A\n",
      "32768it [00:00, 48800.85it/s]            \u001b[A\n",
      "0it [00:00, ?it/s]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Extracting /home/sprakashdash/Documents/DLFA/FashionMNIST/raw/train-labels-idx1-ubyte.gz\n",
      "Downloading http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/t10k-images-idx3-ubyte.gz to /home/sprakashdash/Documents/DLFA/FashionMNIST/raw/t10k-images-idx3-ubyte.gz\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "  0%|          | 0/4422102 [00:00<?, ?it/s]\u001b[A\n",
      "  2%|▏         | 81920/4422102 [00:00<00:08, 486823.20it/s]\u001b[A\n",
      "  4%|▍         | 172032/4422102 [00:01<00:08, 496268.63it/s]\u001b[A\n",
      "  8%|▊         | 360448/4422102 [00:01<00:06, 594329.85it/s]\u001b[A\n",
      " 14%|█▍        | 614400/4422102 [00:01<00:05, 722065.13it/s]\u001b[A\n",
      " 18%|█▊        | 786432/4422102 [00:01<00:04, 782048.22it/s]\u001b[A\n",
      " 26%|██▌       | 1146880/4422102 [00:01<00:03, 972307.74it/s]\u001b[A\n",
      " 30%|███       | 1335296/4422102 [00:01<00:03, 979765.27it/s]\u001b[A\n",
      " 38%|███▊      | 1679360/4422102 [00:02<00:02, 1161030.30it/s]\u001b[A\n",
      " 47%|████▋     | 2072576/4422102 [00:02<00:01, 1385228.94it/s]\u001b[A\n",
      " 53%|█████▎    | 2326528/4422102 [00:02<00:01, 1374804.21it/s]\u001b[A\n",
      " 60%|██████    | 2654208/4422102 [00:02<00:01, 1540737.81it/s]\u001b[A\n",
      " 66%|██████▌   | 2908160/4422102 [00:02<00:00, 1528186.19it/s]\u001b[A\n",
      " 72%|███████▏  | 3170304/4422102 [00:02<00:00, 1528195.11it/s]\u001b[A\n",
      " 78%|███████▊  | 3440640/4422102 [00:03<00:00, 1495643.30it/s]\u001b[A\n",
      " 82%|████████▏ | 3604480/4422102 [00:03<00:00, 1419983.97it/s]\u001b[A\n",
      "\n",
      "0it [00:00, ?it/s]\u001b[A\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Extracting /home/sprakashdash/Documents/DLFA/FashionMNIST/raw/t10k-images-idx3-ubyte.gz\n",
      "Downloading http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/t10k-labels-idx1-ubyte.gz to /home/sprakashdash/Documents/DLFA/FashionMNIST/raw/t10k-labels-idx1-ubyte.gz\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "  0%|          | 0/5148 [00:00<?, ?it/s]\u001b[A\u001b[A\n",
      "\n",
      "8192it [00:00, 15954.16it/s]            \u001b[A\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Extracting /home/sprakashdash/Documents/DLFA/FashionMNIST/raw/t10k-labels-idx1-ubyte.gz\n",
      "Processing...\n",
      "Done!\n"
     ]
    }
   ],
   "source": [
    "trainset = torchvision.datasets.FashionMNIST(root=dataset_root_path, train=True,\\\n",
    "                                             download=True, transform=transforms.ToTensor())\n",
    "testset = torchvision.datasets.FashionMNIST(root=dataset_root_path, train=False,\\\n",
    "                                            download=True, transform=transforms.ToTensor())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "Rk5kD8PAzyvT"
   },
   "outputs": [],
   "source": [
    "trainloader = torch.utils.data.DataLoader(trainset, batch_size=batch_size, shuffle=True, num_workers=2)\n",
    "testloader = torch.utils.data.DataLoader(testset, batch_size=batch_size, shuffle=False, num_workers=2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 51
    },
    "colab_type": "code",
    "id": "SdmzLIj52tXh",
    "outputId": "ead885ff-e855-4748-a651-1eed110d371c"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "torch.Size([60000, 28, 28])\n",
      "torch.Size([60000])\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/sprakashdash/anaconda3/lib/python3.7/site-packages/torchvision/datasets/mnist.py:53: UserWarning: train_data has been renamed data\n",
      "  warnings.warn(\"train_data has been renamed data\")\n",
      "/home/sprakashdash/anaconda3/lib/python3.7/site-packages/torchvision/datasets/mnist.py:43: UserWarning: train_labels has been renamed targets\n",
      "  warnings.warn(\"train_labels has been renamed targets\")\n",
      "\n",
      "4423680it [00:21, 1419983.97it/s]                             \u001b[A"
     ]
    }
   ],
   "source": [
    "print(trainloader.dataset.train_data.shape)\n",
    "print(trainloader.dataset.train_labels.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### The classes are defined in the main repo in [Github](https://github.com/zalandoresearch/fashion-mnist#Labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "aNy0D0Mx61Nd"
   },
   "outputs": [],
   "source": [
    "classes = ('T-shirt/top', 'Trouser', 'Pullover', 'Dress', 'Coat', 'Sandal', 'Shirt', 'Sneaker', 'Bag', 'Ankle-boot')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "TRdIjxre2Gdw"
   },
   "source": [
    "#### Let us show some of the training images, for fun"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "0wue3bFm2ESp"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "torch.Size([64, 1, 28, 28])\n",
      "torch.Size([64])\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Sneaker Shirt Shirt Dress Pullover\n"
     ]
    }
   ],
   "source": [
    "# Function to show an image\n",
    "def imshow(img):\n",
    "    npimg = img.numpy()\n",
    "    plt.imshow(np.transpose(npimg, (1,2,0)), interpolation='nearest')\n",
    "    plt.axis('off')\n",
    "    plt.show()\n",
    "    \n",
    "# Get some random training images\n",
    "dataiter = iter(trainloader)\n",
    "images, labels = dataiter.next()\n",
    "print(images.shape)\n",
    "print(labels.shape)\n",
    "\n",
    "# Show images\n",
    "imshow(torchvision.utils.make_grid(images[:5],nrow=5))\n",
    "# Print their labels\n",
    "print(' '.join('%5s' % classes[labels[j]] for j in range(5)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "MeFkBraQ3b3T"
   },
   "outputs": [],
   "source": [
    "# Fully connected neural network with three hidden layer\n",
    "class NeuralNetThreeLayer(nn.Module):\n",
    "    def __init__(self, input_size, hidden_size_1, hidden_size_2, hidden_size_3, num_classes):\n",
    "        super(NeuralNetThreeLayer, self).__init__()\n",
    "        self.fc1 = nn.Linear(input_size, hidden_size_1)\n",
    "        self.fc2 = nn.Linear(hidden_size_1, hidden_size_2)\n",
    "        self.fc3 = nn.Linear(hidden_size_2, hidden_size_3)\n",
    "        self.fc4 = nn.Linear(hidden_size_3, num_classes)\n",
    "        \n",
    "        self.relu = nn.ReLU()\n",
    "        \n",
    "    \n",
    "    def forward(self, x):\n",
    "        x = self.fc1(x)\n",
    "        x = self.relu(x)\n",
    "        x = self.fc2(x)\n",
    "        x = self.relu(x)\n",
    "        x = self.fc3(x)\n",
    "        x = self.relu(x)\n",
    "        x = self.fc4(x)\n",
    "        return x\n",
    "\n",
    "net = NeuralNetThreeLayer(input_size, hidden_size_1, hidden_size_2, hidden_size_3, num_classes)\n",
    "net = net.to(device)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "YhqWgSC--o1z"
   },
   "source": [
    "Define a Loss function and optimizer\n",
    "\n",
    "---\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "vxgiHPSI-Sdy"
   },
   "outputs": [],
   "source": [
    "criterion = nn.CrossEntropyLoss()\n",
    "# optimizer = optim.SGD(net.parameters(), lr=lr, momentum=0.9)\n",
    "optimizer = optim.Adam(net.parameters(), lr=lr)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "Pmz-13VvG-Qg"
   },
   "outputs": [],
   "source": [
    "# Helper function to test the model\n",
    "# In test phase, we don't need to compute gradients (for memory efficiency)\n",
    "def eval(testloader, epoch):\n",
    "    # print('here')\n",
    "    with torch.no_grad():\n",
    "        running_loss_test = 0.0\n",
    "        test_correct = 0\n",
    "        test_total = 0\n",
    "        test_i = 0\n",
    "        for test_images, test_labels in testloader:\n",
    "            # get the inputs\n",
    "            test_images = test_images.reshape(-1, 28*28).to(device)\n",
    "            test_labels = test_labels.to(device)\n",
    "            test_outputs = net(test_images)\n",
    "            test_loss = criterion(test_outputs, test_labels)\n",
    "            _, test_predicted = torch.max(test_outputs.data, 1)\n",
    "            \n",
    "            running_loss_test += test_loss.item()\n",
    "            test_total += test_labels.size(0)\n",
    "            test_correct += (test_predicted == test_labels).sum().item()\n",
    "            test_i += 1\n",
    "\n",
    "            if test_total >= testloader.dataset.test_data.shape[0]-batch_size:\n",
    "                print('[{:d}], test accuracy: {:4.2f} %, test_loss {:.5f}'.format(epoch+1, \\\n",
    "                                                100*test_correct/test_total, running_loss_test/(test_i+1)))\n",
    "                break\n",
    "        return 100*test_correct/test_total, running_loss_test/(test_i+1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "VEUnl-1gNnOU"
   },
   "outputs": [],
   "source": [
    "# Eval on training data. This is to be run with initial model only\n",
    "# In eval phase, we don't need to compute gradients (for memory efficiency)\n",
    "def eval_initial_model(trainloader):\n",
    "    with torch.no_grad():\n",
    "        running_loss_train = 0.0\n",
    "        train_correct = 0\n",
    "        train_total = 0\n",
    "        train_i = 0\n",
    "        for train_images, train_labels in trainloader:\n",
    "            # get the inputs\n",
    "            train_images = train_images.reshape(-1, 28*28).to(device)\n",
    "            train_labels = train_labels.to(device)\n",
    "            train_outputs = net(train_images)\n",
    "            train_loss = criterion(train_outputs, train_labels)\n",
    "            _, train_predicted = torch.max(train_outputs.data, 1)\n",
    "            \n",
    "            running_loss_train += train_loss.item()\n",
    "            train_total += train_labels.size(0)\n",
    "            train_correct += (train_predicted == train_labels).sum().item()\n",
    "            train_i += 1\n",
    "        \n",
    "        print('[0], train accuracy: {:4.2f} %, train_loss {:.5f}'.format(100*train_correct/train_total, \\\n",
    "                                                                        running_loss_train/(train_i+1)))\n",
    "        return 100*train_correct/train_total, running_loss_train/(train_i+1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 748
    },
    "colab_type": "code",
    "id": "7mB27O1Z-27r",
    "outputId": "2a2f38e7-5825-452d-85b5-b1373989b007"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0], train accuracy: 11.30 %, train_loss 2.30089\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/sprakashdash/anaconda3/lib/python3.7/site-packages/torchvision/datasets/mnist.py:58: UserWarning: test_data has been renamed data\n",
      "  warnings.warn(\"test_data has been renamed data\")\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1], test accuracy: 11.26 %, test_loss 2.28860\n",
      "[1] train accuracy: 73.79 % train loss: 0.75173\n",
      "[1], test accuracy: 79.84 %, test_loss 0.55185\n",
      "[2] train accuracy: 83.71 % train loss: 0.46546\n",
      "[2], test accuracy: 83.66 %, test_loss 0.46679\n",
      "[3] train accuracy: 85.31 % train loss: 0.42034\n",
      "[3], test accuracy: 84.31 %, test_loss 0.43979\n",
      "[4] train accuracy: 86.20 % train loss: 0.39004\n",
      "[4], test accuracy: 85.40 %, test_loss 0.40890\n",
      "[5] train accuracy: 86.91 % train loss: 0.37047\n",
      "[5], test accuracy: 86.13 %, test_loss 0.38870\n",
      "[6] train accuracy: 87.41 % train loss: 0.35183\n",
      "[6], test accuracy: 86.35 %, test_loss 0.37662\n",
      "[7] train accuracy: 87.88 % train loss: 0.33632\n",
      "[7], test accuracy: 86.64 %, test_loss 0.36974\n",
      "[8] train accuracy: 88.43 % train loss: 0.32196\n",
      "[8], test accuracy: 86.66 %, test_loss 0.37621\n",
      "[9] train accuracy: 88.89 % train loss: 0.30853\n",
      "[9], test accuracy: 87.46 %, test_loss 0.34828\n",
      "[10] train accuracy: 89.14 % train loss: 0.29946\n",
      "[10], test accuracy: 87.76 %, test_loss 0.34300\n",
      "[11] train accuracy: 89.45 % train loss: 0.28744\n",
      "[11], test accuracy: 87.88 %, test_loss 0.33867\n",
      "[12] train accuracy: 89.78 % train loss: 0.27751\n",
      "[12], test accuracy: 87.66 %, test_loss 0.34239\n",
      "[13] train accuracy: 90.11 % train loss: 0.26880\n",
      "[13], test accuracy: 88.19 %, test_loss 0.33107\n",
      "[14] train accuracy: 90.34 % train loss: 0.26371\n",
      "[14], test accuracy: 88.35 %, test_loss 0.32251\n",
      "[15] train accuracy: 90.58 % train loss: 0.25345\n",
      "[15], test accuracy: 88.04 %, test_loss 0.33770\n",
      "[16] train accuracy: 90.83 % train loss: 0.24720\n",
      "[16], test accuracy: 88.57 %, test_loss 0.31914\n",
      "[17] train accuracy: 91.19 % train loss: 0.24117\n",
      "[17], test accuracy: 88.70 %, test_loss 0.31754\n",
      "[18] train accuracy: 91.34 % train loss: 0.23326\n",
      "[18], test accuracy: 88.34 %, test_loss 0.32696\n",
      "[19] train accuracy: 91.78 % train loss: 0.22607\n",
      "[19], test accuracy: 88.59 %, test_loss 0.32206\n",
      "[20] train accuracy: 91.83 % train loss: 0.22114\n",
      "[20], test accuracy: 89.02 %, test_loss 0.31865\n",
      "Finished Training\n"
     ]
    }
   ],
   "source": [
    "test_acc = [0.0]*(num_epochs+1)\n",
    "test_loss = [0.0]*(num_epochs+1)\n",
    "train_acc = [0.0]*(num_epochs+1)\n",
    "train_loss = [0.0]*(num_epochs+1)\n",
    "lst_counter = 0\n",
    "\n",
    "train_acc[lst_counter], train_loss[lst_counter] = eval_initial_model(trainloader)\n",
    "test_acc[lst_counter], test_loss[lst_counter] = eval(testloader, 0)\n",
    "lst_counter += 1\n",
    "\n",
    "for epoch in range(num_epochs):  # loop over the dataset multiple times\n",
    "\n",
    "    running_loss_training = 0.0\n",
    "    correct = 0\n",
    "    total = 0\n",
    "    i = 0\n",
    "    # for i, data in enumerate(trainloader, 0):\n",
    "    for inputs, labels in trainloader:\n",
    "        # get the inputs\n",
    "        # inputs, labels = data\n",
    "        inputs = inputs.reshape(-1, 28*28).to(device); labels = labels.to(device)\n",
    "\n",
    "        # zero the parameter gradients\n",
    "        optimizer.zero_grad()\n",
    "\n",
    "        # forward + backward + optimize\n",
    "        outputs = net(inputs)\n",
    "        training_loss = criterion(outputs, labels)\n",
    "        _, predicted = torch.max(outputs.data, 1)\n",
    "        \n",
    "        training_loss.backward()\n",
    "        optimizer.step()\n",
    "\n",
    "        # print statistics\n",
    "        running_loss_training += training_loss.item()\n",
    "        total += labels.size(0)\n",
    "        correct += (predicted == labels).sum().item()\n",
    "        i += 1\n",
    "        if total >= trainloader.dataset.train_data.shape[0]: # or (epoch == 0 and i == 0):\n",
    "            print('[{:d}] train accuracy: {:4.2f} % train loss: {:.5f}'.format(epoch+1, \\\n",
    "                                                            100*correct/total, running_loss_training/(i+1)))\n",
    "            train_acc[lst_counter] = 100*correct/total\n",
    "            train_loss[lst_counter] = running_loss_training/(i+1)\n",
    "            \n",
    "            running_loss_training = 0.0\n",
    "            \n",
    "            test_acc[lst_counter], test_loss[lst_counter] = eval(testloader, epoch)\n",
    "            lst_counter += 1\n",
    "\n",
    "print('Finished Training')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "D8q8LGTzDdkZ",
    "outputId": "756a0478-fa55-48c1-edd9-29caebb608aa"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plotting train and test loss vs Epochs\n",
    "fig1 = plt.figure(1)        \n",
    "plt.plot(range(num_epochs+1),train_loss,'r-',label='train_loss')        \n",
    "plt.plot(range(num_epochs+1),test_loss,'g-',label='test_loss')\n",
    "plt.legend(loc='upper right')\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Losses')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "2FpuPCg5BGBt",
    "outputId": "3b21fd9c-3326-4260-82a4-cdd27283656b"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig2 = plt.figure(1)        \n",
    "plt.plot(range(1, num_epochs+1),train_loss[1:],'r-',label='train_loss')        \n",
    "plt.plot(range(1, num_epochs+1),test_loss[1:],'g-',label='test_loss')\n",
    "plt.legend(loc='upper right')\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Losses')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "lCXU3dpCNnOp",
    "outputId": "4c9c0094-0ab3-4732-ace5-abafc15e02d7"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plotting train and test accuracy vs Epochs\n",
    "fig3 = plt.figure(1)        \n",
    "plt.plot(range(num_epochs+1),train_acc,'r-',label='train_acc')        \n",
    "plt.plot(range(num_epochs+1),test_acc,'g-',label='test_acc')\n",
    "plt.legend(loc='lower right')\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Accuracies')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "eVqYKgsENnOt",
    "outputId": "a7661448-c4ee-40b2-b790-1c753845e5b2"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "GroundTruth:  Ankle-boot Pullover Trouser Trouser Shirt\n",
      "Predicted:  Ankle-boot (0.98) Pullover (1.00) Trouser (1.00) Trouser (1.00) Shirt (0.76)\n"
     ]
    }
   ],
   "source": [
    "dataiter = iter(testloader)\n",
    "images, labels = dataiter.next()\n",
    "\n",
    "# print images\n",
    "imshow(torchvision.utils.make_grid(images[:5], nrow=5))\n",
    "print('GroundTruth: ', ' '.join('%5s' % classes[labels[j]] for j in range(5)))\n",
    "\n",
    "# Get the predicted outputs\n",
    "images = images.reshape(-1, 28*28).to(device)\n",
    "labels = labels.to(device)\n",
    "outputs = net(images)\n",
    "outputs_softmax = F.softmax(outputs, dim=1)\n",
    "probabilities, predicted = torch.max(outputs_softmax, 1)\n",
    "probabilities_np = probabilities.cpu().data.numpy()\n",
    "print('Predicted: ', ' '.join('%5s (%.2f)' % (classes[predicted[j]], probabilities[j]) for j in range(5)))"
   ]
  }
 ],
 "metadata": {
  "colab": {
   "collapsed_sections": [],
   "name": "3_FashionMNIST_FullyConnected.ipynb",
   "provenance": [],
   "version": "0.3.2"
  },
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}