lenet.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "nbgrader": {
     "cell_type": "markdown",
     "checksum": "93518ec5c15a747f0361fb99b24b3953",
     "grade": false,
     "grade_id": "cell-440df6cfa709812f",
     "locked": true,
     "schema_version": 3,
     "solution": false
    }
   },
   "source": [
    "# LeNet-5"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "nbgrader": {
     "cell_type": "code",
     "checksum": "30d471f40f64f2218f1cd61e5ff1a6fb",
     "grade": false,
     "grade_id": "cell-7c0b5567d0681262",
     "locked": true,
     "schema_version": 3,
     "solution": false
    }
   },
   "outputs": [],
   "source": [
    "import os\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "\n",
    "import torch\n",
    "from torch.utils import data\n",
    "import torchvision\n",
    "import torchvision.transforms as transforms\n",
    "from torchvision import datasets\n",
    "\n",
    "\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "import torch.optim as optim\n",
    "\n",
    "import pandas as pd\n",
    "from training import compute_accuracy\n",
    "from tools import save_net, show_predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "skip_training = False"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "device = torch.device('cuda:0') #device = torch.device('cpu')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "nbgrader": {
     "cell_type": "code",
     "checksum": "48d33ffe246f5459117f53cac15b370d",
     "grade": false,
     "grade_id": "cell-fe95dcf02c6b9c5e",
     "locked": true,
     "schema_version": 3,
     "solution": false
    }
   },
   "outputs": [],
   "source": [
    "if skip_training:\n",
    "    device = torch.device(\"cpu\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "nbgrader": {
     "cell_type": "markdown",
     "checksum": "2f2b11aa8f0d0377563333bd78493751",
     "grade": false,
     "grade_id": "cell-e5b565cc4aae8e7f",
     "locked": true,
     "schema_version": 3,
     "solution": false
    }
   },
   "source": [
    "## Load data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "nbgrader": {
     "cell_type": "code",
     "checksum": "05dae4d3d681cf9666b90ac940628a83",
     "grade": false,
     "grade_id": "cell-8b0fded08998282c",
     "locked": true,
     "schema_version": 3,
     "solution": false
    }
   },
   "outputs": [],
   "source": [
    "transform = transforms.Compose([\n",
    "    transforms.Resize((28,28)),\n",
    "    transforms.Grayscale(num_output_channels=1),\n",
    "    transforms.ToTensor(),  # Transform to tensor\n",
    "    transforms.Normalize((0.5,), (0.5,))  # Min-max scaling to [-1, 1]\n",
    "])\n",
    "\n",
    "data_dir = 'tiny-imagenet-200/'\n",
    "train_dataset = datasets.ImageFolder(os.path.join(data_dir, 'train'), transform)\n",
    "test_dataset = datasets.ImageFolder(os.path.join(data_dir, 'test'), transform)\n",
    "val_dataset = datasets.ImageFolder(os.path.join(data_dir, 'val'), transform)\n",
    "\n",
    "classes = pd.read_csv('wnids.txt',header=None).values\n",
    "words = pd.read_csv('words.txt',sep=\"\\t\",header=None,names=['class','words'])\n",
    "words = words.set_index('class')\n",
    "\n",
    "trainloader = data.DataLoader(train_dataset, batch_size=32, shuffle=True)\n",
    "valloader = data.DataLoader(val_dataset, batch_size=5, shuffle=True)\n",
    "testloader = data.DataLoader(test_dataset, batch_size=5, shuffle=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "nbgrader": {
     "cell_type": "code",
     "checksum": "b6f82b52a8e640030852a0dca3bdb84a",
     "grade": false,
     "grade_id": "cell-a3a0c0ed4bbc2d05",
     "locked": false,
     "schema_version": 3,
     "solution": true
    }
   },
   "outputs": [],
   "source": [
    "class Net(nn.Module):\n",
    "    def __init__(self):\n",
    "        super(Net, self).__init__()\n",
    "        # YOUR CODE HERE\n",
    "        self.conv1 = nn.Conv2d(1, 16, 5)\n",
    "        self.conv2 = nn.Conv2d(16, 32, 5)\n",
    "        self.fc1 = nn.Linear(32*16, 120)\n",
    "        self.fc2 = nn.Linear(120, 84)\n",
    "        self.fc3 = nn.Linear(84, 200)\n",
    "\n",
    "    def forward(self, x):\n",
    "        x = F.max_pool2d(F.relu(self.conv1(x)), kernel_size=2, stride=2)\n",
    "        x = F.max_pool2d(F.relu(self.conv2(x)), kernel_size=2, stride=2)\n",
    "        x = F.relu(self.fc1(x.reshape(-1,32*16)))\n",
    "        x = F.relu(self.fc2(x))\n",
    "        x = self.fc3(x)\n",
    "        return x\n",
    "    \n",
    "    def num_flat_features(self, x):\n",
    "        size = x.size()[1:]  # all dimensions except the batch dimension\n",
    "        num_features = 1\n",
    "        for s in size:\n",
    "            num_features *= s\n",
    "        return num_features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "nbgrader": {
     "cell_type": "code",
     "checksum": "a2ef196862fad8696a3161d25b54f04b",
     "grade": true,
     "grade_id": "architecture",
     "locked": true,
     "points": 1,
     "schema_version": 3,
     "solution": false
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Shape of the input tensor: torch.Size([32, 1, 28, 28])\n",
      "The shapes seem to be ok.\n"
     ]
    }
   ],
   "source": [
    "net = Net()\n",
    "net.to(device)\n",
    "\n",
    "with torch.no_grad():\n",
    "    dataiter = iter(trainloader)\n",
    "    images, labels = dataiter.next()\n",
    "    images = images.to(device)\n",
    "    print('Shape of the input tensor:', images.shape)\n",
    "\n",
    "    y = net(images)\n",
    "    assert y.shape == torch.Size([32, 200]), \"Bad shape of y: y.shape={}\".format(y.shape)\n",
    "\n",
    "print('The shapes seem to be ok.')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "nbgrader": {
     "cell_type": "code",
     "checksum": "e35202382f6f8e1913ce7127a411663a",
     "grade": false,
     "grade_id": "cell-863d60f5f7458388",
     "locked": true,
     "schema_version": 3,
     "solution": false
    }
   },
   "outputs": [],
   "source": [
    "# Define the loss and the optimizer\n",
    "criterion = nn.CrossEntropyLoss()\n",
    "optimizer = optim.SGD(net.parameters(), lr=0.001, momentum=0.9)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "n_epochs = 30"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "nbgrader": {
     "cell_type": "code",
     "checksum": "19c277c5c5987808b22c9adfeffebecb",
     "grade": false,
     "grade_id": "training_loop",
     "locked": true,
     "schema_version": 3,
     "solution": false
    },
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1,   200] loss: 5.300\n",
      "[1,   400] loss: 5.299\n",
      "[1,   600] loss: 5.299\n",
      "[1,   800] loss: 5.301\n",
      "[1,  1000] loss: 5.300\n",
      "[1,  1200] loss: 5.300\n",
      "[1,  1400] loss: 5.301\n",
      "[1,  1600] loss: 5.301\n",
      "[1,  1800] loss: 5.299\n",
      "[1,  2000] loss: 5.300\n",
      "[1,  2200] loss: 5.299\n",
      "[1,  2400] loss: 5.300\n",
      "[1,  2600] loss: 5.300\n",
      "[1,  2800] loss: 5.300\n",
      "[1,  3000] loss: 5.301\n",
      "Accuracy of the network on the test images: 0.007\n",
      "[2,   200] loss: 5.299\n",
      "[2,   400] loss: 5.299\n",
      "[2,   600] loss: 5.300\n",
      "[2,   800] loss: 5.298\n",
      "[2,  1000] loss: 5.299\n",
      "[2,  1200] loss: 5.298\n",
      "[2,  1400] loss: 5.297\n",
      "[2,  1600] loss: 5.297\n",
      "[2,  1800] loss: 5.297\n",
      "[2,  2000] loss: 5.297\n",
      "[2,  2200] loss: 5.297\n",
      "[2,  2400] loss: 5.296\n",
      "[2,  2600] loss: 5.296\n",
      "[2,  2800] loss: 5.295\n",
      "[2,  3000] loss: 5.297\n",
      "Accuracy of the network on the test images: 0.007\n",
      "[3,   200] loss: 5.295\n",
      "[3,   400] loss: 5.293\n",
      "[3,   600] loss: 5.293\n",
      "[3,   800] loss: 5.290\n",
      "[3,  1000] loss: 5.294\n",
      "[3,  1200] loss: 5.293\n",
      "[3,  1400] loss: 5.290\n",
      "[3,  1600] loss: 5.287\n",
      "[3,  1800] loss: 5.287\n",
      "[3,  2000] loss: 5.288\n",
      "[3,  2200] loss: 5.286\n",
      "[3,  2400] loss: 5.285\n",
      "[3,  2600] loss: 5.282\n",
      "[3,  2800] loss: 5.280\n",
      "[3,  3000] loss: 5.279\n",
      "Accuracy of the network on the test images: 0.008\n",
      "[4,   200] loss: 5.270\n",
      "[4,   400] loss: 5.270\n",
      "[4,   600] loss: 5.263\n",
      "[4,   800] loss: 5.266\n",
      "[4,  1000] loss: 5.262\n",
      "[4,  1200] loss: 5.252\n",
      "[4,  1400] loss: 5.248\n",
      "[4,  1600] loss: 5.246\n",
      "[4,  1800] loss: 5.241\n",
      "[4,  2000] loss: 5.239\n",
      "[4,  2200] loss: 5.234\n",
      "[4,  2400] loss: 5.220\n",
      "[4,  2600] loss: 5.219\n",
      "[4,  2800] loss: 5.220\n",
      "[4,  3000] loss: 5.210\n",
      "Accuracy of the network on the test images: 0.015\n",
      "[5,   200] loss: 5.209\n",
      "[5,   400] loss: 5.194\n",
      "[5,   600] loss: 5.185\n",
      "[5,   800] loss: 5.182\n",
      "[5,  1000] loss: 5.185\n",
      "[5,  1200] loss: 5.175\n",
      "[5,  1400] loss: 5.170\n",
      "[5,  1600] loss: 5.171\n",
      "[5,  1800] loss: 5.154\n",
      "[5,  2000] loss: 5.164\n",
      "[5,  2200] loss: 5.167\n",
      "[5,  2400] loss: 5.156\n",
      "[5,  2600] loss: 5.144\n",
      "[5,  2800] loss: 5.157\n",
      "[5,  3000] loss: 5.152\n",
      "Accuracy of the network on the test images: 0.020\n",
      "[6,   200] loss: 5.145\n",
      "[6,   400] loss: 5.127\n",
      "[6,   600] loss: 5.121\n",
      "[6,   800] loss: 5.142\n",
      "[6,  1000] loss: 5.132\n",
      "[6,  1200] loss: 5.128\n",
      "[6,  1400] loss: 5.127\n",
      "[6,  1600] loss: 5.118\n",
      "[6,  1800] loss: 5.111\n",
      "[6,  2000] loss: 5.106\n",
      "[6,  2200] loss: 5.098\n",
      "[6,  2400] loss: 5.107\n",
      "[6,  2600] loss: 5.103\n",
      "[6,  2800] loss: 5.103\n",
      "[6,  3000] loss: 5.093\n",
      "Accuracy of the network on the test images: 0.025\n",
      "[7,   200] loss: 5.073\n",
      "[7,   400] loss: 5.078\n",
      "[7,   600] loss: 5.060\n",
      "[7,   800] loss: 5.064\n",
      "[7,  1000] loss: 5.055\n",
      "[7,  1200] loss: 5.048\n",
      "[7,  1400] loss: 5.044\n",
      "[7,  1600] loss: 5.018\n",
      "[7,  1800] loss: 5.013\n",
      "[7,  2000] loss: 5.019\n",
      "[7,  2200] loss: 5.023\n",
      "[7,  2400] loss: 4.997\n",
      "[7,  2600] loss: 4.989\n",
      "[7,  2800] loss: 4.984\n",
      "[7,  3000] loss: 4.952\n",
      "Accuracy of the network on the test images: 0.034\n",
      "[8,   200] loss: 4.955\n",
      "[8,   400] loss: 4.940\n",
      "[8,   600] loss: 4.934\n",
      "[8,   800] loss: 4.936\n",
      "[8,  1000] loss: 4.921\n",
      "[8,  1200] loss: 4.915\n",
      "[8,  1400] loss: 4.923\n",
      "[8,  1600] loss: 4.905\n",
      "[8,  1800] loss: 4.896\n",
      "[8,  2000] loss: 4.906\n",
      "[8,  2200] loss: 4.884\n",
      "[8,  2400] loss: 4.884\n",
      "[8,  2600] loss: 4.857\n",
      "[8,  2800] loss: 4.877\n",
      "[8,  3000] loss: 4.847\n",
      "Accuracy of the network on the test images: 0.042\n",
      "[9,   200] loss: 4.826\n",
      "[9,   400] loss: 4.827\n",
      "[9,   600] loss: 4.839\n",
      "[9,   800] loss: 4.814\n",
      "[9,  1000] loss: 4.802\n",
      "[9,  1200] loss: 4.801\n",
      "[9,  1400] loss: 4.806\n",
      "[9,  1600] loss: 4.779\n",
      "[9,  1800] loss: 4.773\n",
      "[9,  2000] loss: 4.783\n",
      "[9,  2200] loss: 4.762\n",
      "[9,  2400] loss: 4.765\n",
      "[9,  2600] loss: 4.742\n",
      "[9,  2800] loss: 4.739\n",
      "[9,  3000] loss: 4.754\n",
      "Accuracy of the network on the test images: 0.055\n",
      "[10,   200] loss: 4.692\n",
      "[10,   400] loss: 4.710\n",
      "[10,   600] loss: 4.731\n",
      "[10,   800] loss: 4.689\n",
      "[10,  1000] loss: 4.715\n",
      "[10,  1200] loss: 4.708\n",
      "[10,  1400] loss: 4.692\n",
      "[10,  1600] loss: 4.687\n",
      "[10,  1800] loss: 4.677\n",
      "[10,  2000] loss: 4.683\n",
      "[10,  2200] loss: 4.682\n",
      "[10,  2400] loss: 4.704\n",
      "[10,  2600] loss: 4.685\n",
      "[10,  2800] loss: 4.683\n",
      "[10,  3000] loss: 4.671\n",
      "Accuracy of the network on the test images: 0.061\n",
      "[11,   200] loss: 4.631\n",
      "[11,   400] loss: 4.645\n",
      "[11,   600] loss: 4.638\n",
      "[11,   800] loss: 4.618\n",
      "[11,  1000] loss: 4.639\n",
      "[11,  1200] loss: 4.617\n",
      "[11,  1400] loss: 4.620\n",
      "[11,  1600] loss: 4.612\n",
      "[11,  1800] loss: 4.625\n",
      "[11,  2000] loss: 4.599\n",
      "[11,  2200] loss: 4.623\n",
      "[11,  2400] loss: 4.588\n",
      "[11,  2600] loss: 4.613\n",
      "[11,  2800] loss: 4.597\n",
      "[11,  3000] loss: 4.603\n",
      "Accuracy of the network on the test images: 0.071\n",
      "[12,   200] loss: 4.570\n",
      "[12,   400] loss: 4.555\n",
      "[12,   600] loss: 4.577\n",
      "[12,   800] loss: 4.529\n",
      "[12,  1000] loss: 4.556\n",
      "[12,  1200] loss: 4.547\n",
      "[12,  1400] loss: 4.562\n",
      "[12,  1600] loss: 4.540\n",
      "[12,  1800] loss: 4.576\n",
      "[12,  2000] loss: 4.559\n",
      "[12,  2200] loss: 4.548\n",
      "[12,  2400] loss: 4.544\n",
      "[12,  2600] loss: 4.529\n",
      "[12,  2800] loss: 4.540\n",
      "[12,  3000] loss: 4.542\n",
      "Accuracy of the network on the test images: 0.076\n",
      "[13,   200] loss: 4.500\n",
      "[13,   400] loss: 4.512\n",
      "[13,   600] loss: 4.533\n",
      "[13,   800] loss: 4.487\n",
      "[13,  1000] loss: 4.506\n",
      "[13,  1200] loss: 4.501\n",
      "[13,  1400] loss: 4.492\n",
      "[13,  1600] loss: 4.487\n",
      "[13,  1800] loss: 4.473\n",
      "[13,  2000] loss: 4.495\n",
      "[13,  2200] loss: 4.486\n",
      "[13,  2400] loss: 4.452\n",
      "[13,  2600] loss: 4.476\n",
      "[13,  2800] loss: 4.458\n",
      "[13,  3000] loss: 4.466\n",
      "Accuracy of the network on the test images: 0.080\n",
      "[14,   200] loss: 4.430\n",
      "[14,   400] loss: 4.439\n",
      "[14,   600] loss: 4.429\n",
      "[14,   800] loss: 4.422\n",
      "[14,  1000] loss: 4.453\n",
      "[14,  1200] loss: 4.435\n",
      "[14,  1400] loss: 4.406\n",
      "[14,  1600] loss: 4.431\n",
      "[14,  1800] loss: 4.413\n",
      "[14,  2000] loss: 4.382\n",
      "[14,  2200] loss: 4.425\n",
      "[14,  2400] loss: 4.418\n",
      "[14,  2600] loss: 4.384\n",
      "[14,  2800] loss: 4.407\n",
      "[14,  3000] loss: 4.432\n",
      "Accuracy of the network on the test images: 0.090\n",
      "[15,   200] loss: 4.363\n",
      "[15,   400] loss: 4.374\n",
      "[15,   600] loss: 4.350\n",
      "[15,   800] loss: 4.380\n",
      "[15,  1000] loss: 4.348\n",
      "[15,  1200] loss: 4.381\n",
      "[15,  1400] loss: 4.357\n",
      "[15,  1600] loss: 4.335\n",
      "[15,  1800] loss: 4.359\n",
      "[15,  2000] loss: 4.369\n",
      "[15,  2200] loss: 4.379\n",
      "[15,  2400] loss: 4.348\n",
      "[15,  2600] loss: 4.334\n",
      "[15,  2800] loss: 4.331\n",
      "[15,  3000] loss: 4.347\n",
      "Accuracy of the network on the test images: 0.098\n",
      "[16,   200] loss: 4.310\n",
      "[16,   400] loss: 4.295\n",
      "[16,   600] loss: 4.345\n",
      "[16,   800] loss: 4.294\n",
      "[16,  1000] loss: 4.283\n",
      "[16,  1200] loss: 4.319\n",
      "[16,  1400] loss: 4.331\n",
      "[16,  1600] loss: 4.291\n",
      "[16,  1800] loss: 4.311\n",
      "[16,  2000] loss: 4.272\n",
      "[16,  2200] loss: 4.307\n",
      "[16,  2400] loss: 4.274\n",
      "[16,  2600] loss: 4.276\n",
      "[16,  2800] loss: 4.281\n",
      "[16,  3000] loss: 4.264\n",
      "Accuracy of the network on the test images: 0.105\n",
      "[17,   200] loss: 4.239\n",
      "[17,   400] loss: 4.242\n",
      "[17,   600] loss: 4.238\n",
      "[17,   800] loss: 4.251\n",
      "[17,  1000] loss: 4.222\n",
      "[17,  1200] loss: 4.227\n",
      "[17,  1400] loss: 4.257\n",
      "[17,  1600] loss: 4.239\n",
      "[17,  1800] loss: 4.260\n",
      "[17,  2000] loss: 4.265\n",
      "[17,  2200] loss: 4.250\n",
      "[17,  2400] loss: 4.228\n",
      "[17,  2600] loss: 4.250\n",
      "[17,  2800] loss: 4.259\n",
      "[17,  3000] loss: 4.257\n",
      "Accuracy of the network on the test images: 0.104\n",
      "[18,   200] loss: 4.211\n",
      "[18,   400] loss: 4.202\n",
      "[18,   600] loss: 4.201\n",
      "[18,   800] loss: 4.197\n",
      "[18,  1000] loss: 4.236\n",
      "[18,  1200] loss: 4.205\n",
      "[18,  1400] loss: 4.209\n",
      "[18,  1600] loss: 4.177\n",
      "[18,  1800] loss: 4.194\n",
      "[18,  2000] loss: 4.210\n",
      "[18,  2200] loss: 4.161\n",
      "[18,  2400] loss: 4.178\n",
      "[18,  2600] loss: 4.222\n",
      "[18,  2800] loss: 4.196\n",
      "[18,  3000] loss: 4.174\n",
      "Accuracy of the network on the test images: 0.110\n",
      "[19,   200] loss: 4.151\n",
      "[19,   400] loss: 4.156\n",
      "[19,   600] loss: 4.150\n",
      "[19,   800] loss: 4.100\n",
      "[19,  1000] loss: 4.166\n",
      "[19,  1200] loss: 4.146\n",
      "[19,  1400] loss: 4.141\n",
      "[19,  1600] loss: 4.194\n",
      "[19,  1800] loss: 4.189\n",
      "[19,  2000] loss: 4.147\n",
      "[19,  2200] loss: 4.193\n",
      "[19,  2400] loss: 4.174\n",
      "[19,  2600] loss: 4.123\n",
      "[19,  2800] loss: 4.141\n",
      "[19,  3000] loss: 4.125\n",
      "Accuracy of the network on the test images: 0.116\n",
      "[20,   200] loss: 4.109\n",
      "[20,   400] loss: 4.076\n",
      "[20,   600] loss: 4.094\n",
      "[20,   800] loss: 4.140\n",
      "[20,  1000] loss: 4.156\n",
      "[20,  1200] loss: 4.126\n",
      "[20,  1400] loss: 4.099\n",
      "[20,  1600] loss: 4.085\n",
      "[20,  1800] loss: 4.105\n",
      "[20,  2000] loss: 4.121\n",
      "[20,  2200] loss: 4.118\n",
      "[20,  2400] loss: 4.129\n",
      "[20,  2600] loss: 4.103\n",
      "[20,  2800] loss: 4.124\n",
      "[20,  3000] loss: 4.104\n",
      "Accuracy of the network on the test images: 0.115\n",
      "[21,   200] loss: 4.069\n",
      "[21,   400] loss: 4.059\n",
      "[21,   600] loss: 4.029\n",
      "[21,   800] loss: 4.098\n",
      "[21,  1000] loss: 4.116\n",
      "[21,  1200] loss: 4.040\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[21,  1400] loss: 4.111\n",
      "[21,  1600] loss: 4.104\n",
      "[21,  1800] loss: 4.084\n",
      "[21,  2000] loss: 4.086\n",
      "[21,  2200] loss: 4.090\n",
      "[21,  2400] loss: 4.063\n",
      "[21,  2600] loss: 4.061\n",
      "[21,  2800] loss: 4.076\n",
      "[21,  3000] loss: 4.074\n",
      "Accuracy of the network on the test images: 0.121\n",
      "[22,   200] loss: 4.021\n",
      "[22,   400] loss: 4.051\n",
      "[22,   600] loss: 4.005\n",
      "[22,   800] loss: 4.043\n",
      "[22,  1000] loss: 4.073\n",
      "[22,  1200] loss: 4.027\n",
      "[22,  1400] loss: 4.036\n",
      "[22,  1600] loss: 4.054\n",
      "[22,  1800] loss: 4.054\n",
      "[22,  2000] loss: 4.052\n",
      "[22,  2200] loss: 4.045\n",
      "[22,  2400] loss: 4.040\n",
      "[22,  2600] loss: 4.068\n",
      "[22,  2800] loss: 4.020\n",
      "[22,  3000] loss: 4.048\n",
      "Accuracy of the network on the test images: 0.124\n",
      "[23,   200] loss: 3.987\n",
      "[23,   400] loss: 3.976\n",
      "[23,   600] loss: 4.024\n",
      "[23,   800] loss: 3.986\n",
      "[23,  1000] loss: 4.001\n",
      "[23,  1200] loss: 3.998\n",
      "[23,  1400] loss: 4.028\n",
      "[23,  1600] loss: 4.030\n",
      "[23,  1800] loss: 4.031\n",
      "[23,  2000] loss: 4.030\n",
      "[23,  2200] loss: 4.030\n",
      "[23,  2400] loss: 4.019\n",
      "[23,  2600] loss: 4.012\n",
      "[23,  2800] loss: 4.007\n",
      "[23,  3000] loss: 4.012\n",
      "Accuracy of the network on the test images: 0.121\n",
      "[24,   200] loss: 3.985\n",
      "[24,   400] loss: 3.944\n",
      "[24,   600] loss: 3.979\n",
      "[24,   800] loss: 3.996\n",
      "[24,  1000] loss: 4.001\n",
      "[24,  1200] loss: 3.978\n",
      "[24,  1400] loss: 3.994\n",
      "[24,  1600] loss: 3.967\n",
      "[24,  1800] loss: 3.958\n",
      "[24,  2000] loss: 4.007\n",
      "[24,  2200] loss: 3.968\n",
      "[24,  2400] loss: 3.986\n",
      "[24,  2600] loss: 4.003\n",
      "[24,  2800] loss: 4.022\n",
      "[24,  3000] loss: 3.990\n",
      "Accuracy of the network on the test images: 0.128\n",
      "[25,   200] loss: 3.932\n",
      "[25,   400] loss: 3.935\n",
      "[25,   600] loss: 3.942\n",
      "[25,   800] loss: 3.995\n",
      "[25,  1000] loss: 3.947\n",
      "[25,  1200] loss: 3.957\n",
      "[25,  1400] loss: 3.962\n",
      "[25,  1600] loss: 3.958\n",
      "[25,  1800] loss: 3.925\n",
      "[25,  2000] loss: 3.955\n",
      "[25,  2200] loss: 3.938\n",
      "[25,  2400] loss: 3.934\n",
      "[25,  2600] loss: 3.973\n",
      "[25,  2800] loss: 3.939\n",
      "[25,  3000] loss: 3.971\n",
      "Accuracy of the network on the test images: 0.130\n",
      "[26,   200] loss: 3.916\n",
      "[26,   400] loss: 3.929\n",
      "[26,   600] loss: 3.921\n",
      "[26,   800] loss: 3.909\n",
      "[26,  1000] loss: 3.940\n",
      "[26,  1200] loss: 3.940\n",
      "[26,  1400] loss: 3.948\n",
      "[26,  1600] loss: 3.924\n",
      "[26,  1800] loss: 3.903\n",
      "[26,  2000] loss: 3.902\n",
      "[26,  2200] loss: 3.953\n",
      "[26,  2400] loss: 3.920\n",
      "[26,  2600] loss: 3.940\n",
      "[26,  2800] loss: 3.917\n",
      "[26,  3000] loss: 3.954\n",
      "Accuracy of the network on the test images: 0.132\n",
      "[27,   200] loss: 3.890\n",
      "[27,   400] loss: 3.888\n",
      "[27,   600] loss: 3.928\n",
      "[27,   800] loss: 3.897\n",
      "[27,  1000] loss: 3.916\n",
      "[27,  1200] loss: 3.897\n",
      "[27,  1400] loss: 3.879\n",
      "[27,  1600] loss: 3.925\n",
      "[27,  1800] loss: 3.927\n",
      "[27,  2000] loss: 3.891\n",
      "[27,  2200] loss: 3.858\n",
      "[27,  2400] loss: 3.899\n",
      "[27,  2600] loss: 3.937\n",
      "[27,  2800] loss: 3.898\n",
      "[27,  3000] loss: 3.893\n",
      "Accuracy of the network on the test images: 0.128\n",
      "[28,   200] loss: 3.857\n",
      "[28,   400] loss: 3.836\n",
      "[28,   600] loss: 3.871\n",
      "[28,   800] loss: 3.903\n",
      "[28,  1000] loss: 3.835\n",
      "[28,  1200] loss: 3.874\n",
      "[28,  1400] loss: 3.901\n",
      "[28,  1600] loss: 3.880\n",
      "[28,  1800] loss: 3.871\n",
      "[28,  2000] loss: 3.876\n",
      "[28,  2200] loss: 3.884\n",
      "[28,  2400] loss: 3.901\n",
      "[28,  2600] loss: 3.911\n",
      "[28,  2800] loss: 3.883\n",
      "[28,  3000] loss: 3.861\n",
      "Accuracy of the network on the test images: 0.131\n",
      "[29,   200] loss: 3.846\n",
      "[29,   400] loss: 3.819\n",
      "[29,   600] loss: 3.827\n",
      "[29,   800] loss: 3.824\n",
      "[29,  1000] loss: 3.865\n",
      "[29,  1200] loss: 3.861\n",
      "[29,  1400] loss: 3.846\n",
      "[29,  1600] loss: 3.856\n",
      "[29,  1800] loss: 3.863\n",
      "[29,  2000] loss: 3.895\n",
      "[29,  2200] loss: 3.808\n",
      "[29,  2400] loss: 3.858\n",
      "[29,  2600] loss: 3.845\n",
      "[29,  2800] loss: 3.885\n",
      "[29,  3000] loss: 3.864\n",
      "Accuracy of the network on the test images: 0.141\n",
      "[30,   200] loss: 3.807\n",
      "[30,   400] loss: 3.786\n",
      "[30,   600] loss: 3.842\n",
      "[30,   800] loss: 3.800\n",
      "[30,  1000] loss: 3.818\n",
      "[30,  1200] loss: 3.785\n",
      "[30,  1400] loss: 3.822\n",
      "[30,  1600] loss: 3.811\n",
      "[30,  1800] loss: 3.897\n",
      "[30,  2000] loss: 3.825\n",
      "[30,  2200] loss: 3.814\n",
      "[30,  2400] loss: 3.850\n",
      "[30,  2600] loss: 3.872\n",
      "[30,  2800] loss: 3.825\n",
      "[30,  3000] loss: 3.806\n",
      "Accuracy of the network on the test images: 0.139\n",
      "Finished Training\n"
     ]
    }
   ],
   "source": [
    "net.train()\n",
    "for epoch in range(n_epochs):\n",
    "    running_loss = 0.0\n",
    "    print_every = 200  # mini-batches\n",
    "    for i, (inputs, labels) in enumerate(trainloader, 0):\n",
    "        # Transfer to GPU\n",
    "        inputs, labels = inputs.to(device), labels.to(device)\n",
    "\n",
    "        # zero the parameter gradients\n",
    "        optimizer.zero_grad()\n",
    "\n",
    "        # forward + backward + optimize\n",
    "        outputs = net(inputs)\n",
    "        loss = criterion(outputs, labels)\n",
    "        loss.backward()\n",
    "        optimizer.step()\n",
    "\n",
    "        # print statistics\n",
    "        running_loss += loss.item()\n",
    "        if (i % print_every) == (print_every-1):\n",
    "            print('[%d, %5d] loss: %.3f' % (epoch+1, i+1, running_loss/print_every))\n",
    "            running_loss = 0.0\n",
    "        if skip_training:\n",
    "            break\n",
    "\n",
    "    # Print accuracy after every epoch\n",
    "    accuracy = compute_accuracy(net, device, valloader)\n",
    "    print('Accuracy of the network on the test images: %.3f' % accuracy)\n",
    "\n",
    "    if skip_training:\n",
    "        break\n",
    "\n",
    "print('Finished Training')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "if not skip_training:\n",
    "    filename = 'lenet_epoch_50.pth'\n",
    "    save_net(net, filename)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {
    "nbgrader": {
     "cell_type": "code",
     "checksum": "588833d22121ec23197b12f710e73f9b",
     "grade": false,
     "grade_id": "cell-d786a74f883ad32e",
     "locked": true,
     "schema_version": 3,
     "solution": false
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Prediction 1:\n",
      "Turth: cougar, puma, catamount, mountain lion, painter, panther, Felis concolor\n",
      "Pred: cougar, puma, catamount, mountain lion, painter, panther, Felis concolor\n",
      "\n",
      "Prediction 2:\n",
      "Turth: obelisk\n",
      "Pred: brain coral\n",
      "\n",
      "Prediction 3:\n",
      "Turth: pizza, pizza pie\n",
      "Pred: frying pan, frypan, skillet\n",
      "\n",
      "Prediction 4:\n",
      "Turth: altar\n",
      "Pred: beaker\n",
      "\n",
      "Prediction 5:\n",
      "Turth: fly\n",
      "Pred: fly\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 5 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Show the ground truth labels and the network's predictions\n",
    "show_predictions(net, device, testloader)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {
    "nbgrader": {
     "cell_type": "code",
     "checksum": "f420b734e91bcc49286c70c043995b25",
     "grade": true,
     "grade_id": "accuracy",
     "locked": true,
     "points": 1,
     "schema_version": 3,
     "solution": false
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy of the network on the test images: 0.218\n"
     ]
    }
   ],
   "source": [
    "# Let us compute the accuracy on the test set\n",
    "accuracy = compute_accuracy(net, device, testloader)\n",
    "print('Accuracy of the network on the test images: %.3f' % accuracy)"
   ]
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "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": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}