15.Pytorch Robust Deep learning Neural Network with Adding Noise.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "source": [
    "**Adding Noise for Robust Deep Neural Network Models**\r\n",
    "---\r\n",
    "* In this notebook, We will explore how to add noise in the image.\r\n",
    "* Train a model with noise based images.\r\n",
    "* Noise Types: Guassian, Salt and Papper and Speckle Noise\r\n",
    "* **Dataset** : CIFAR10"
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "**1.Load Modules**\r\n",
    "---"
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "source": [
    "# torch loaded..!!!\r\n",
    "import torch\r\n",
    "import torch.nn as nn\r\n",
    "import torch.nn.functional as F\r\n",
    "import torch.optim as optim\r\n",
    "torch.manual_seed(42)\r\n",
    "\r\n",
    "# torchvision loaded..!!!\r\n",
    "import torchvision\r\n",
    "from torchvision import datasets, transforms\r\n",
    "import torchvision.models as models\r\n",
    "from torchsummary import summary\r\n",
    "\r\n",
    "# other module loaded..!!!\r\n",
    "import numpy as np\r\n",
    "import matplotlib.pyplot as plt\r\n",
    "import time\r\n",
    "import os\r\n",
    "plt.style.use('ggplot')"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "Going over all the important imports:\r\n",
    "\r\n",
    "* `torch`: as we will be implementing everything using the PyTorch deep learning library, so we import torch first.\r\n",
    "* `torchvision`: this module will help us download the CIFAR10 dataset, pre-trained PyTorch models, and also define the transforms that we will apply to the images.\r\n",
    "* `torch.nn`: we will get access to all the neural network layers through torch.nn.\r\n",
    "* `torch.optim`: this will help us define the optimizer of our choice.\r\n",
    "* `argparse`: for defining our argument parsers.\r\n",
    "* `matplotlib`: for saving the plots after training and validation."
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "**2.Define the Constants**\r\n",
    "---"
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "source": [
    "EPOCHS = 10\r\n",
    "TRAIN_NOISE = True\r\n",
    "TEST_NOISE = False\r\n",
    "BATCH_SIZE = 32\r\n",
    "NUM_CLASSES = 10\r\n",
    "pretrained=True\r\n",
    "requires_grad=True"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "**3.Defining Transform and Helper Function**\r\n",
    "---"
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "source": [
    "transform = transforms.Compose([transforms.ToTensor(),transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))])"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "source": [
    "device = torch.device(\"cuda:0\" if torch.cuda.is_available() else \"cpu\")"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "**4.Preparing Data**\r\n",
    "---"
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "source": [
    "# Data Downloading\r\n",
    "trainset = datasets.CIFAR10(root='./data_cifar10/', train=True, download=True, transform=transform)\r\n",
    "testset = datasets.CIFAR10(root='./data_cifar10', train=False, download=True, transform=transform)\r\n",
    "\r\n",
    "# Data Loader\r\n",
    "trainLoader = torch.utils.data.DataLoader(trainset, batch_size=BATCH_SIZE, shuffle=True)\r\n",
    "testLoader = torch.utils.data.DataLoader(testset, batch_size=BATCH_SIZE, shuffle=False)"
   ],
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "Files already downloaded and verified\n",
      "Files already downloaded and verified\n"
     ]
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "**5.Prepare the Neural Network**\r\n",
    "---"
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "source": [
    "def model():\r\n",
    "    model = models.resnet18(progress=True, pretrained=pretrained)\r\n",
    "    # freeze hidden layers\r\n",
    "    if requires_grad == False:\r\n",
    "        for param in model.parameters():\r\n",
    "            param.requires_grad = False\r\n",
    "    # train the hidden layers\r\n",
    "    elif requires_grad == True:\r\n",
    "        for param in model.parameters():\r\n",
    "            param.requires_grad = True\r\n",
    "    # make the classification layers learnable\r\n",
    "    model.fc = nn.Linear(512, NUM_CLASSES)\r\n",
    "    \r\n",
    "    model = model.to(device)\r\n",
    "    \r\n",
    "    return model\r\n",
    "model = model()"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "source": [
    "print(summary(model, (3, 32, 32)))"
   ],
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "==========================================================================================\n",
      "Layer (type:depth-idx)                   Output Shape              Param #\n",
      "==========================================================================================\n",
      "├─Conv2d: 1-1                            [-1, 64, 16, 16]          9,408\n",
      "├─BatchNorm2d: 1-2                       [-1, 64, 16, 16]          128\n",
      "├─ReLU: 1-3                              [-1, 64, 16, 16]          --\n",
      "├─MaxPool2d: 1-4                         [-1, 64, 8, 8]            --\n",
      "├─Sequential: 1-5                        [-1, 64, 8, 8]            --\n",
      "|    └─BasicBlock: 2-1                   [-1, 64, 8, 8]            --\n",
      "|    |    └─Conv2d: 3-1                  [-1, 64, 8, 8]            36,864\n",
      "|    |    └─BatchNorm2d: 3-2             [-1, 64, 8, 8]            128\n",
      "|    |    └─ReLU: 3-3                    [-1, 64, 8, 8]            --\n",
      "|    |    └─Conv2d: 3-4                  [-1, 64, 8, 8]            36,864\n",
      "|    |    └─BatchNorm2d: 3-5             [-1, 64, 8, 8]            128\n",
      "|    |    └─ReLU: 3-6                    [-1, 64, 8, 8]            --\n",
      "|    └─BasicBlock: 2-2                   [-1, 64, 8, 8]            --\n",
      "|    |    └─Conv2d: 3-7                  [-1, 64, 8, 8]            36,864\n",
      "|    |    └─BatchNorm2d: 3-8             [-1, 64, 8, 8]            128\n",
      "|    |    └─ReLU: 3-9                    [-1, 64, 8, 8]            --\n",
      "|    |    └─Conv2d: 3-10                 [-1, 64, 8, 8]            36,864\n",
      "|    |    └─BatchNorm2d: 3-11            [-1, 64, 8, 8]            128\n",
      "|    |    └─ReLU: 3-12                   [-1, 64, 8, 8]            --\n",
      "├─Sequential: 1-6                        [-1, 128, 4, 4]           --\n",
      "|    └─BasicBlock: 2-3                   [-1, 128, 4, 4]           --\n",
      "|    |    └─Conv2d: 3-13                 [-1, 128, 4, 4]           73,728\n",
      "|    |    └─BatchNorm2d: 3-14            [-1, 128, 4, 4]           256\n",
      "|    |    └─ReLU: 3-15                   [-1, 128, 4, 4]           --\n",
      "|    |    └─Conv2d: 3-16                 [-1, 128, 4, 4]           147,456\n",
      "|    |    └─BatchNorm2d: 3-17            [-1, 128, 4, 4]           256\n",
      "|    |    └─Sequential: 3-18             [-1, 128, 4, 4]           8,448\n",
      "|    |    └─ReLU: 3-19                   [-1, 128, 4, 4]           --\n",
      "|    └─BasicBlock: 2-4                   [-1, 128, 4, 4]           --\n",
      "|    |    └─Conv2d: 3-20                 [-1, 128, 4, 4]           147,456\n",
      "|    |    └─BatchNorm2d: 3-21            [-1, 128, 4, 4]           256\n",
      "|    |    └─ReLU: 3-22                   [-1, 128, 4, 4]           --\n",
      "|    |    └─Conv2d: 3-23                 [-1, 128, 4, 4]           147,456\n",
      "|    |    └─BatchNorm2d: 3-24            [-1, 128, 4, 4]           256\n",
      "|    |    └─ReLU: 3-25                   [-1, 128, 4, 4]           --\n",
      "├─Sequential: 1-7                        [-1, 256, 2, 2]           --\n",
      "|    └─BasicBlock: 2-5                   [-1, 256, 2, 2]           --\n",
      "|    |    └─Conv2d: 3-26                 [-1, 256, 2, 2]           294,912\n",
      "|    |    └─BatchNorm2d: 3-27            [-1, 256, 2, 2]           512\n",
      "|    |    └─ReLU: 3-28                   [-1, 256, 2, 2]           --\n",
      "|    |    └─Conv2d: 3-29                 [-1, 256, 2, 2]           589,824\n",
      "|    |    └─BatchNorm2d: 3-30            [-1, 256, 2, 2]           512\n",
      "|    |    └─Sequential: 3-31             [-1, 256, 2, 2]           33,280\n",
      "|    |    └─ReLU: 3-32                   [-1, 256, 2, 2]           --\n",
      "|    └─BasicBlock: 2-6                   [-1, 256, 2, 2]           --\n",
      "|    |    └─Conv2d: 3-33                 [-1, 256, 2, 2]           589,824\n",
      "|    |    └─BatchNorm2d: 3-34            [-1, 256, 2, 2]           512\n",
      "|    |    └─ReLU: 3-35                   [-1, 256, 2, 2]           --\n",
      "|    |    └─Conv2d: 3-36                 [-1, 256, 2, 2]           589,824\n",
      "|    |    └─BatchNorm2d: 3-37            [-1, 256, 2, 2]           512\n",
      "|    |    └─ReLU: 3-38                   [-1, 256, 2, 2]           --\n",
      "├─Sequential: 1-8                        [-1, 512, 1, 1]           --\n",
      "|    └─BasicBlock: 2-7                   [-1, 512, 1, 1]           --\n",
      "|    |    └─Conv2d: 3-39                 [-1, 512, 1, 1]           1,179,648\n",
      "|    |    └─BatchNorm2d: 3-40            [-1, 512, 1, 1]           1,024\n",
      "|    |    └─ReLU: 3-41                   [-1, 512, 1, 1]           --\n",
      "|    |    └─Conv2d: 3-42                 [-1, 512, 1, 1]           2,359,296\n",
      "|    |    └─BatchNorm2d: 3-43            [-1, 512, 1, 1]           1,024\n",
      "|    |    └─Sequential: 3-44             [-1, 512, 1, 1]           132,096\n",
      "|    |    └─ReLU: 3-45                   [-1, 512, 1, 1]           --\n",
      "|    └─BasicBlock: 2-8                   [-1, 512, 1, 1]           --\n",
      "|    |    └─Conv2d: 3-46                 [-1, 512, 1, 1]           2,359,296\n",
      "|    |    └─BatchNorm2d: 3-47            [-1, 512, 1, 1]           1,024\n",
      "|    |    └─ReLU: 3-48                   [-1, 512, 1, 1]           --\n",
      "|    |    └─Conv2d: 3-49                 [-1, 512, 1, 1]           2,359,296\n",
      "|    |    └─BatchNorm2d: 3-50            [-1, 512, 1, 1]           1,024\n",
      "|    |    └─ReLU: 3-51                   [-1, 512, 1, 1]           --\n",
      "├─AdaptiveAvgPool2d: 1-9                 [-1, 512, 1, 1]           --\n",
      "├─Linear: 1-10                           [-1, 10]                  5,130\n",
      "==========================================================================================\n",
      "Total params: 11,181,642\n",
      "Trainable params: 11,181,642\n",
      "Non-trainable params: 0\n",
      "Total mult-adds (M): 59.52\n",
      "==========================================================================================\n",
      "Input size (MB): 0.01\n",
      "Forward/backward pass size (MB): 0.77\n",
      "Params size (MB): 42.65\n",
      "Estimated Total Size (MB): 43.44\n",
      "==========================================================================================\n",
      "==========================================================================================\n",
      "Layer (type:depth-idx)                   Output Shape              Param #\n",
      "==========================================================================================\n",
      "├─Conv2d: 1-1                            [-1, 64, 16, 16]          9,408\n",
      "├─BatchNorm2d: 1-2                       [-1, 64, 16, 16]          128\n",
      "├─ReLU: 1-3                              [-1, 64, 16, 16]          --\n",
      "├─MaxPool2d: 1-4                         [-1, 64, 8, 8]            --\n",
      "├─Sequential: 1-5                        [-1, 64, 8, 8]            --\n",
      "|    └─BasicBlock: 2-1                   [-1, 64, 8, 8]            --\n",
      "|    |    └─Conv2d: 3-1                  [-1, 64, 8, 8]            36,864\n",
      "|    |    └─BatchNorm2d: 3-2             [-1, 64, 8, 8]            128\n",
      "|    |    └─ReLU: 3-3                    [-1, 64, 8, 8]            --\n",
      "|    |    └─Conv2d: 3-4                  [-1, 64, 8, 8]            36,864\n",
      "|    |    └─BatchNorm2d: 3-5             [-1, 64, 8, 8]            128\n",
      "|    |    └─ReLU: 3-6                    [-1, 64, 8, 8]            --\n",
      "|    └─BasicBlock: 2-2                   [-1, 64, 8, 8]            --\n",
      "|    |    └─Conv2d: 3-7                  [-1, 64, 8, 8]            36,864\n",
      "|    |    └─BatchNorm2d: 3-8             [-1, 64, 8, 8]            128\n",
      "|    |    └─ReLU: 3-9                    [-1, 64, 8, 8]            --\n",
      "|    |    └─Conv2d: 3-10                 [-1, 64, 8, 8]            36,864\n",
      "|    |    └─BatchNorm2d: 3-11            [-1, 64, 8, 8]            128\n",
      "|    |    └─ReLU: 3-12                   [-1, 64, 8, 8]            --\n",
      "├─Sequential: 1-6                        [-1, 128, 4, 4]           --\n",
      "|    └─BasicBlock: 2-3                   [-1, 128, 4, 4]           --\n",
      "|    |    └─Conv2d: 3-13                 [-1, 128, 4, 4]           73,728\n",
      "|    |    └─BatchNorm2d: 3-14            [-1, 128, 4, 4]           256\n",
      "|    |    └─ReLU: 3-15                   [-1, 128, 4, 4]           --\n",
      "|    |    └─Conv2d: 3-16                 [-1, 128, 4, 4]           147,456\n",
      "|    |    └─BatchNorm2d: 3-17            [-1, 128, 4, 4]           256\n",
      "|    |    └─Sequential: 3-18             [-1, 128, 4, 4]           8,448\n",
      "|    |    └─ReLU: 3-19                   [-1, 128, 4, 4]           --\n",
      "|    └─BasicBlock: 2-4                   [-1, 128, 4, 4]           --\n",
      "|    |    └─Conv2d: 3-20                 [-1, 128, 4, 4]           147,456\n",
      "|    |    └─BatchNorm2d: 3-21            [-1, 128, 4, 4]           256\n",
      "|    |    └─ReLU: 3-22                   [-1, 128, 4, 4]           --\n",
      "|    |    └─Conv2d: 3-23                 [-1, 128, 4, 4]           147,456\n",
      "|    |    └─BatchNorm2d: 3-24            [-1, 128, 4, 4]           256\n",
      "|    |    └─ReLU: 3-25                   [-1, 128, 4, 4]           --\n",
      "├─Sequential: 1-7                        [-1, 256, 2, 2]           --\n",
      "|    └─BasicBlock: 2-5                   [-1, 256, 2, 2]           --\n",
      "|    |    └─Conv2d: 3-26                 [-1, 256, 2, 2]           294,912\n",
      "|    |    └─BatchNorm2d: 3-27            [-1, 256, 2, 2]           512\n",
      "|    |    └─ReLU: 3-28                   [-1, 256, 2, 2]           --\n",
      "|    |    └─Conv2d: 3-29                 [-1, 256, 2, 2]           589,824\n",
      "|    |    └─BatchNorm2d: 3-30            [-1, 256, 2, 2]           512\n",
      "|    |    └─Sequential: 3-31             [-1, 256, 2, 2]           33,280\n",
      "|    |    └─ReLU: 3-32                   [-1, 256, 2, 2]           --\n",
      "|    └─BasicBlock: 2-6                   [-1, 256, 2, 2]           --\n",
      "|    |    └─Conv2d: 3-33                 [-1, 256, 2, 2]           589,824\n",
      "|    |    └─BatchNorm2d: 3-34            [-1, 256, 2, 2]           512\n",
      "|    |    └─ReLU: 3-35                   [-1, 256, 2, 2]           --\n",
      "|    |    └─Conv2d: 3-36                 [-1, 256, 2, 2]           589,824\n",
      "|    |    └─BatchNorm2d: 3-37            [-1, 256, 2, 2]           512\n",
      "|    |    └─ReLU: 3-38                   [-1, 256, 2, 2]           --\n",
      "├─Sequential: 1-8                        [-1, 512, 1, 1]           --\n",
      "|    └─BasicBlock: 2-7                   [-1, 512, 1, 1]           --\n",
      "|    |    └─Conv2d: 3-39                 [-1, 512, 1, 1]           1,179,648\n",
      "|    |    └─BatchNorm2d: 3-40            [-1, 512, 1, 1]           1,024\n",
      "|    |    └─ReLU: 3-41                   [-1, 512, 1, 1]           --\n",
      "|    |    └─Conv2d: 3-42                 [-1, 512, 1, 1]           2,359,296\n",
      "|    |    └─BatchNorm2d: 3-43            [-1, 512, 1, 1]           1,024\n",
      "|    |    └─Sequential: 3-44             [-1, 512, 1, 1]           132,096\n",
      "|    |    └─ReLU: 3-45                   [-1, 512, 1, 1]           --\n",
      "|    └─BasicBlock: 2-8                   [-1, 512, 1, 1]           --\n",
      "|    |    └─Conv2d: 3-46                 [-1, 512, 1, 1]           2,359,296\n",
      "|    |    └─BatchNorm2d: 3-47            [-1, 512, 1, 1]           1,024\n",
      "|    |    └─ReLU: 3-48                   [-1, 512, 1, 1]           --\n",
      "|    |    └─Conv2d: 3-49                 [-1, 512, 1, 1]           2,359,296\n",
      "|    |    └─BatchNorm2d: 3-50            [-1, 512, 1, 1]           1,024\n",
      "|    |    └─ReLU: 3-51                   [-1, 512, 1, 1]           --\n",
      "├─AdaptiveAvgPool2d: 1-9                 [-1, 512, 1, 1]           --\n",
      "├─Linear: 1-10                           [-1, 10]                  5,130\n",
      "==========================================================================================\n",
      "Total params: 11,181,642\n",
      "Trainable params: 11,181,642\n",
      "Non-trainable params: 0\n",
      "Total mult-adds (M): 59.52\n",
      "==========================================================================================\n",
      "Input size (MB): 0.01\n",
      "Forward/backward pass size (MB): 0.77\n",
      "Params size (MB): 42.65\n",
      "Estimated Total Size (MB): 43.44\n",
      "==========================================================================================\n"
     ]
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "**6. Define The Loss Function**\r\n",
    "---"
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "source": [
    "# loss function\r\n",
    "criterion = nn.CrossEntropyLoss()\r\n",
    "\r\n",
    "# optimizer\r\n",
    "optimizer = optim.SGD(model.parameters(), lr = 0.001, momentum=0.99)"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "**7.The Training Function**\r\n",
    "---"
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "source": [
    "# The training function\r\n",
    "from tqdm import tqdm\r\n",
    "print(f'Training with Noise:{TRAIN_NOISE}')\r\n",
    "print(f'Testing with Noise:{TEST_NOISE}')\r\n",
    "print(f'Pretrained: {pretrained}, Required Gradient: {requires_grad}')\r\n",
    "\r\n",
    "if pretrained and requires_grad:\r\n",
    "    print('Training with ImageNet Weights and updating hidden layer weights')\r\n",
    "elif pretrained and not requires_grad:\r\n",
    "    print('Training with ImageNet Weights but Freezing hidden layer weights')\r\n",
    "elif not pretrained and requires_grad:\r\n",
    "    print('Training with random weights and updating hidden layer weights')\r\n",
    "elif not pretrained and not requires_grad:\r\n",
    "    print('Training with random weights but Freezing hidden layer weights')\r\n",
    "\r\n",
    "def train(NUM_EPOCHs, epoch, model, dataloader):\r\n",
    "    model.train()\r\n",
    "    loss = 0\r\n",
    "    acc = 0\r\n",
    "    running_loss = 0\r\n",
    "    running_correct = 0\r\n",
    "    with tqdm(trainLoader, unit=\"batch\") as tepoch:\r\n",
    "        for i,data in enumerate(tepoch):\r\n",
    "            tepoch.set_description(f\"Epoch {epoch+1}\")\r\n",
    "            # get the inputs\r\n",
    "            inputs, labels = data[0].to(device), data[1].to(device)\r\n",
    "            if TRAIN_NOISE:\r\n",
    "                noise = torch.randn(inputs.shape).to(device)\r\n",
    "                new_img = inputs + noise\r\n",
    "            elif TRAIN_NOISE == False:\r\n",
    "                new_img = inputs\r\n",
    "                \r\n",
    "            new_img = new_img.to(device)\r\n",
    "            \r\n",
    "            # zero the parameter gradients\r\n",
    "            optimizer.zero_grad()\r\n",
    "            \r\n",
    "            # forward + backward + optimize\r\n",
    "            outputs = model(new_img)\r\n",
    "            loss = criterion(outputs, labels)\r\n",
    "            _, preds = torch.max(outputs.data, 1)\r\n",
    "            \r\n",
    "            running_loss += loss.item()\r\n",
    "            running_correct += torch.sum(preds == labels.data).sum().item()\r\n",
    "            \r\n",
    "            # backward + optimize only if in training phase\r\n",
    "            loss.backward()\r\n",
    "            optimizer.step()\r\n",
    "            # print statistics\r\n",
    "            tepoch.set_postfix(loss=running_loss/(i+1), acc= 100. * running_correct / (BATCH_SIZE * (i+1)))\r\n",
    "            \r\n",
    "    loss = running_loss / len(trainLoader)\r\n",
    "    acc = 100. * running_correct / len(trainLoader.dataset)\r\n",
    "    print(f'Epoch {epoch+1} Loss: {loss:.4f} Acc: {acc:.4f}')\r\n",
    "    return loss, acc"
   ],
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "Training with Noise:True\n",
      "Testing with Noise:False\n",
      "Pretrained: True, Required Gradient: True\n",
      "Training with ImageNet Weights and updating hidden layer weights\n"
     ]
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "**8.The valildation Function**\r\n",
    "---"
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "source": [
    "# the validation function\r\n",
    "def validate(NUM_EPOCHS, epoch, model, testloader):\r\n",
    "    model.eval()\r\n",
    "    loss = 0.0\r\n",
    "    acc = 0\r\n",
    "    running_loss = 0.0\r\n",
    "    running_correct = 0\r\n",
    "    with torch.no_grad():\r\n",
    "        for i, data in enumerate(testloader):\r\n",
    "            img, labels = data[0].to(device), data[1].to(device)\r\n",
    "            # add noise to the image data\r\n",
    "            if TEST_NOISE:\r\n",
    "                noise = torch.randn(img.shape).to(device)\r\n",
    "                new_img = img + noise\r\n",
    "            elif TEST_NOISE == False:\r\n",
    "                new_img = img\r\n",
    "            new_img = new_img.to(device)\r\n",
    "            outputs = model(new_img)\r\n",
    "            loss = criterion(outputs, labels)\r\n",
    "            _, preds = torch.max(outputs.data, 1)\r\n",
    "            running_loss += loss.item()\r\n",
    "            running_correct += (preds == labels).sum().item()\r\n",
    "    loss = running_loss / len(testset)\r\n",
    "    acc = 100. * running_correct / len(testset)\r\n",
    "    print(f\"Epoch {epoch+1} of {NUM_EPOCHS}, val loss: {loss:.3f}, val acc: {acc:.3f}\")\r\n",
    "    return loss, acc "
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "source": [
    "train_loss, train_acc = [], []\r\n",
    "val_loss, val_acc = [], []\r\n",
    "start = time.time()\r\n",
    "for epoch in range(EPOCHS):\r\n",
    "    e_start = time.time()\r\n",
    "    train_epoch_loss, train_epoch_acc = train(EPOCHS, epoch, model, trainLoader)\r\n",
    "    train_loss.append(train_epoch_loss)\r\n",
    "    train_acc.append(train_epoch_acc)\r\n",
    "    val_epoch_loss, val_epoch_acc = validate(EPOCHS, epoch, model, testLoader)\r\n",
    "    val_loss.append(val_epoch_loss)\r\n",
    "    val_acc.append(val_epoch_acc)\r\n",
    "    e_end = time.time()\r\n",
    "    print(f\"Took {(e_end-e_start)/60:.3f} minutes for epoch {epoch+1}\")\r\n",
    "end = time.time()\r\n",
    "print(f\"Took {(end-start)/60:.3f} minutes to train\")"
   ],
   "outputs": [
    {
     "output_type": "stream",
     "name": "stderr",
     "text": [
      "Epoch 1: 100%|██████████| 1563/1563 [01:07<00:00, 23.20batch/s, acc=32, loss=1.99]\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "Epoch 1 Loss: 1.9949 Acc: 32.0180\n",
      "Epoch 1 of 10, val loss: 0.073, val acc: 30.370\n",
      "Took 1.207 minutes for epoch 1\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stderr",
     "text": [
      "Epoch 2: 100%|██████████| 1563/1563 [01:07<00:00, 23.16batch/s, acc=39.9, loss=1.71]\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "Epoch 2 Loss: 1.7139 Acc: 39.8840\n",
      "Epoch 2 of 10, val loss: 0.097, val acc: 26.600\n",
      "Took 1.207 minutes for epoch 2\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stderr",
     "text": [
      "Epoch 3: 100%|██████████| 1563/1563 [01:08<00:00, 22.83batch/s, acc=44, loss=1.58]\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "Epoch 3 Loss: 1.5783 Acc: 43.9740\n",
      "Epoch 3 of 10, val loss: 0.105, val acc: 26.970\n",
      "Took 1.224 minutes for epoch 3\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stderr",
     "text": [
      "Epoch 4: 100%|██████████| 1563/1563 [01:07<00:00, 23.26batch/s, acc=47.1, loss=1.48]\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "Epoch 4 Loss: 1.4815 Acc: 47.1320\n",
      "Epoch 4 of 10, val loss: 0.074, val acc: 32.890\n",
      "Took 1.201 minutes for epoch 4\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stderr",
     "text": [
      "Epoch 5: 100%|██████████| 1563/1563 [01:07<00:00, 23.09batch/s, acc=49.3, loss=1.43]\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "Epoch 5 Loss: 1.4307 Acc: 49.2880\n",
      "Epoch 5 of 10, val loss: 0.124, val acc: 24.470\n",
      "Took 1.210 minutes for epoch 5\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stderr",
     "text": [
      "Epoch 6: 100%|██████████| 1563/1563 [01:08<00:00, 22.66batch/s, acc=52.2, loss=1.35]\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "Epoch 6 Loss: 1.3495 Acc: 52.1680\n",
      "Epoch 6 of 10, val loss: 0.074, val acc: 33.030\n",
      "Took 1.226 minutes for epoch 6\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stderr",
     "text": [
      "Epoch 7: 100%|██████████| 1563/1563 [01:00<00:00, 25.72batch/s, acc=53.9, loss=1.3]\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "Epoch 7 Loss: 1.2987 Acc: 53.9180\n",
      "Epoch 7 of 10, val loss: 0.077, val acc: 31.160\n",
      "Took 1.100 minutes for epoch 7\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stderr",
     "text": [
      "Epoch 8: 100%|██████████| 1563/1563 [01:03<00:00, 24.66batch/s, acc=54.6, loss=1.28]\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "Epoch 8 Loss: 1.2794 Acc: 54.5740\n",
      "Epoch 8 of 10, val loss: 0.065, val acc: 35.730\n",
      "Took 1.133 minutes for epoch 8\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stderr",
     "text": [
      "Epoch 9: 100%|██████████| 1563/1563 [01:17<00:00, 20.22batch/s, acc=56.3, loss=1.23]\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "Epoch 9 Loss: 1.2277 Acc: 56.3420\n",
      "Epoch 9 of 10, val loss: 0.082, val acc: 33.280\n",
      "Took 1.370 minutes for epoch 9\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stderr",
     "text": [
      "Epoch 10: 100%|██████████| 1563/1563 [01:16<00:00, 20.46batch/s, acc=56.9, loss=1.22]\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "Epoch 10 Loss: 1.2160 Acc: 56.8800\n",
      "Epoch 10 of 10, val loss: 0.064, val acc: 38.480\n",
      "Took 1.350 minutes for epoch 10\n",
      "Took 12.226 minutes to train\n"
     ]
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "**9.Model Evaluation**\r\n",
    "---"
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "source": [
    "torch.save(model, f\"{TRAIN_NOISE}_{TEST_NOISE}.pth\")\r\n",
    "plt.figure(figsize=(20,6))\r\n",
    "plt.subplot(121)\r\n",
    "plt.plot(train_acc, label='training accuracy')\r\n",
    "plt.plot(val_acc, label='validation accuracy')\r\n",
    "plt.title('Accuracy Plots')\r\n",
    "plt.xlabel('Epochs')\r\n",
    "plt.ylabel('Accuracy')\r\n",
    "plt.legend()\r\n",
    "# plt.savefig(f\"outputs/plots/{TRAIN_NOISE}_{TEST_NOISE}_acc.png\")\r\n",
    "plt.subplot(122)\r\n",
    "plt.plot(train_loss, label='training loss')\r\n",
    "plt.plot(val_loss, label='validation loss')\r\n",
    "plt.title('Loss Plots')\r\n",
    "plt.xlabel('Epochs')\r\n",
    "plt.ylabel('Loss')\r\n",
    "plt.legend()\r\n",
    "# plt.savefig(f\"outputs/plots/{args['train_noise']}_{args['test_noise']}_loss.png\")\r\n",
    "plt.show()"
   ],
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1440x432 with 2 Axes>"
      ]
     },
     "metadata": {}
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "source": [],
   "outputs": [],
   "metadata": {}
  }
 ],
 "metadata": {
  "orig_nbformat": 4,
  "language_info": {
   "name": "python",
   "version": "3.8.0",
   "mimetype": "text/x-python",
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "pygments_lexer": "ipython3",
   "nbconvert_exporter": "python",
   "file_extension": ".py"
  },
  "kernelspec": {
   "name": "python3",
   "display_name": "Python 3.8.0 64-bit ('pytorch19': conda)"
  },
  "interpreter": {
   "hash": "884a8f81666a19c0851426c83cd6eaa7b212468ad852fb3caa21591c98d4369f"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}