Image_classification_with_CNN.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "89319537",
   "metadata": {},
   "source": [
    "# Image classification \n",
    "\n",
    "Image classification is a computer vision task that involves categorizing images into predefined classes or categories. The goal is to develop algorithms or models that can accurately identify and assign labels to images based on their visual features and content. This task is commonly used in various applications, such as object recognition, facial recognition, medical imaging, and autonomous driving, to enable machines to understand and interpret visual information in a similar way to humans. The output of an image classification task is a prediction or probability distribution indicating the likelihood of each class for a given image."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "349fb80c",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "import time, copy\n",
    "\n",
    "import torch\n",
    "from torch import nn\n",
    "import torch.nn.functional as F\n",
    "from torch.utils.data import DataLoader\n",
    "from torchvision import datasets, transforms\n",
    "from torch.utils.data.dataset import random_split\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "dcbdabb7",
   "metadata": {},
   "source": [
    "## Load dataset and transform\n",
    "\n",
    "The \"Modified National Institute of Standards and Technology\" (__MNIST__) dataset consists of a large collection of 60,000 handwritten digits for training and an additional 10,000 handwritten digits for testing.\n",
    "It dataset has served as a benchmark for evaluating and comparing the performance of various machine learning algorithms, particularly in the field of image classification. It has played a crucial role in the development and advancement of deep learning models, especially convolutional neural networks (CNNs), and has been used extensively for educational purposes and as a baseline for assessing new algorithms and techniques in the field.\n",
    "\n",
    "\n",
    "_Image Format_: Each image in the MNIST dataset is a grayscale image with a resolution of 28x28 pixels. This results in a total of 784 pixels per image.\n",
    "\n",
    "_Digit Classes_: The dataset covers ten classes representing the digits from 0 to 9. Each image is labeled with the corresponding digit class, providing the ground truth for training and evaluation.\n",
    "\n",
    "_Data Distribution_: The dataset is balanced, meaning that it contains an equal number of samples for each digit class. This balance ensures that the model is exposed to an equal representation of each digit during training.\n",
    "\n",
    "\n",
    "-------------------\n",
    "\n",
    "__FashionMNIST__ is intended to be a more challenging dataset compared to MNIST, as it requires models to recognize and classify images of various clothing items accurately. FashionMNIST has the same Image Format, Data Distribution, and size of train- and test-set as of MNIST.\n",
    "\n",
    "_Clothing Categories_: The dataset covers ten different clothing categories, including T-shirts/tops, trousers, pullovers, dresses, coats, sandals, shirts, sneakers, bags, and ankle boots. Each image is labeled with the corresponding clothing category, providing the ground truth for training and evaluation."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "0b35907a",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "FashionMNIST \n",
      "\n",
      "number of (train, test) examples = (60000, 10000)\n"
     ]
    }
   ],
   "source": [
    "# Load the dataset and transform it into tensors and normalized between -1 and 1.\n",
    "\n",
    "\n",
    "transform = transforms.Compose([       \n",
    "    transforms.ToTensor(),\n",
    "    transforms.Normalize(mean=(0.5,), std=(0.5,)) \n",
    "])\n",
    "\n",
    "\n",
    "def load_dataset(name):\n",
    "    print(name,'\\n')\n",
    "    if name == \"FashionMNIST\":\n",
    "        train_dataset = datasets.FashionMNIST(root=\"FashionMNIST\", train=True, download=True, transform=transform)\n",
    "        test_dataset = datasets.FashionMNIST(root=\"FashionMNIST\", train=False, download=True, transform=transform)\n",
    "    elif name == \"MNIST\":\n",
    "        train_dataset = datasets.MNIST(root=\"MNIST\", train=True, download=True, transform=transform)\n",
    "        test_dataset = datasets.MNIST(root=\"MNIST\", train=False, download=True, transform=transform)\n",
    "    return train_dataset, test_dataset \n",
    "\n",
    "\n",
    "# Choose either \"MNIST\" or \"FashionMNIST\" \n",
    "train_dataset, test_dataset = load_dataset(\"FashionMNIST\")\n",
    "\n",
    "\n",
    "#train_dataset, val_dataset = random_split(train_dataset, [len(dataset) - 5000, 5000])\n",
    "\n",
    "print(f'number of (train, test) examples = {len(train_dataset), len(test_dataset)}')\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "09367beb",
   "metadata": {},
   "source": [
    "### view a training example:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "0eaa7b82",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "x of torch.Size([1, 28, 28]) :\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "true label = y = 4\n",
      "\n",
      "(x_min, x_max) =  (-1.0, 1.0)\n"
     ]
    }
   ],
   "source": [
    "idx = np.random.choice(len(train_dataset))\n",
    "\n",
    "x = train_dataset[idx][0]\n",
    "print(f'x of {x.shape} :')\n",
    "plt.imshow(x[0], cmap='gray')\n",
    "plt.show()\n",
    "\n",
    "print(f'true label = y = {train_dataset[idx][1]}\\n')\n",
    "\n",
    "print(f'(x_min, x_max) =  {x.min().item(), round(x.max().item(),3)}')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "519ac1b9",
   "metadata": {},
   "source": [
    "# Define model (CNN), training and test loops"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "88315e95",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Using cpu device\n",
      "\n"
     ]
    }
   ],
   "source": [
    "device = \"cuda\" if torch.cuda.is_available() else \"cpu\"     # Get gpu or cpu device for training\n",
    "print(f\"Using {device} device\\n\")"
   ]
  },
  {
   "attachments": {
    "LeNET5.png": {
     "image/png": "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"
    }
   },
   "cell_type": "markdown",
   "id": "fe0bf006",
   "metadata": {},
   "source": [
    "# [LeNet-5][1]\n",
    "\n",
    "[1]:http://vision.stanford.edu/cs598_spring07/papers/Lecun98.pdf\n",
    "\n",
    "$ $\n",
    "\n",
    "![LeNET5.png](attachment:LeNET5.png)\n",
    "\n",
    "\n",
    "\n",
    "$ $\n",
    "\n",
    "LeNet-5 is a convolutional neural network (CNN) architecture developed by Yann LeCun and his colleagues in the early 1990s. It was designed for handwritten digit recognition tasks and played a crucial role in the development of deep learning. Here is an overview of the LeNet-5 architecture:\n",
    "\n",
    "__Input Layer__: LeNet-5 takes grayscale images of size 32x32 = 1x32x32 (in our case, it is 1x28x28) pixels as input.\n",
    "\n",
    "\n",
    "__Convolutional Layer 1__: The first convolutional layer consists of 6 feature maps with a 5x5 kernel size and a stride of 1.\n",
    "\n",
    "__Pooling Layer 1__: Following each convolutional layer, a pooling (subsampling) layer is applied to reduce the spatial dimensions. The first subsampling layer performs 2x2 max pooling with a stride of 2.\n",
    "\n",
    "\n",
    "__Convolutional Layer 2__: The second convolutional layer consists of 16 feature maps with a 5x5 kernel size and a stride of 1.\n",
    "\n",
    "__Pooling Layer 2__: The second subsampling layer performs 2x2 max pooling with a stride of 2. \n",
    "\n",
    "    \n",
    "-----------------    \n",
    "In the original case: the size decreases as 32 $\\rightarrow$ 32-4=28 $\\rightarrow$ 28/2=14 $\\rightarrow$ 14-4=10 $\\rightarrow$ 10/2=5 $\\rightarrow$\n",
    "\n",
    "In our case: the size decreases as 28 $\\rightarrow$ 28-4=24 $\\rightarrow$ 24/2=12 $\\rightarrow$ 12-4=8  $\\rightarrow$  8/2=4 $\\rightarrow$\n",
    "\n",
    "$\\text{size_out} = \\lfloor\\frac{\\text{size_in} \\ +\\ 2\\; \\text{padding}\\ -\\ \\text{kernel_size}}{\\text{stride}}\\rfloor +1$. Here, $\\text{padding}=0$.\n",
    "    \n",
    "\n",
    "----------------------------\n",
    "\n",
    "__Fully Connected Layer 1__: After the convolutional and subsampling layers, the output is flattened and fed into a fully connected layer with 120 neurons.\n",
    "\n",
    "__Fully Connected Layer 2__: The second fully connected layer has 84 neurons.\n",
    "\n",
    "\n",
    "__Output Layer__: The final output layer consists of 10 neurons, corresponding to the 10 possible digit classes (0 to 9).\n",
    "    \n",
    "Activation Functions: LeNet-5 uses a sigmoid activation function for the fully connected layers and a tanh activation function for the convolutional layers.\n",
    "\n",
    "Training: The LeNet-5 architecture is trained using the backpropagation algorithm with gradient descent optimization.\n",
    "    \n",
    "LeNet-5 was an influential model and helped establish the effectiveness of CNNs in various computer vision tasks. While it may seem relatively simple compared to modern CNN architectures, it laid the foundation for more complex and powerful networks that have been developed in the field of deep learning.\n",
    "    \n",
    "\n",
    "    \n",
    "__Choose one of the following two CNNs.__"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "84093be9",
   "metadata": {},
   "outputs": [],
   "source": [
    "'''This CNN is LeNet-5'''\n",
    "\n",
    "class CNN(nn.Module):                                                 # Define model\n",
    "    def __init__(self):\n",
    "        super().__init__()\n",
    "        \n",
    "        self.layers_stack = nn.Sequential(\n",
    "            nn.Conv2d(in_channels=1, out_channels=6, kernel_size=5),  # stride's default value is 1\n",
    "            nn.Tanh(),                                                # shape = [batch_size, 6, 24, 24]\n",
    "            nn.MaxPool2d(kernel_size=2), # stride's default value is kernel_size, shape = [batch_size, 6, 12, 12]\n",
    "            nn.Conv2d(in_channels=6, out_channels=16, kernel_size=5), # shape = [batch_size, 16, 8, 8]\n",
    "            nn.Tanh(),\n",
    "            nn.MaxPool2d(kernel_size=2),                              # shape = [batch_size, 16, 4, 4]\n",
    "            nn.Flatten(),\n",
    "            nn.Linear(16*4*4, 120),                                   # shape = [batch_size, 120]\n",
    "            nn.Sigmoid(),\n",
    "            nn.Linear(120, 84),\n",
    "            nn.Sigmoid(),\n",
    "            nn.Linear(84, 10),\n",
    "            nn.Softmax(dim=1)\n",
    "        )\n",
    "        \n",
    "    def forward(self, x):\n",
    "        x = self.layers_stack(x)\n",
    "        return x"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "2695c3a7",
   "metadata": {},
   "outputs": [],
   "source": [
    "'''This is a simple CNN with 1 convolution, 1 pooling and 1 fully connected layer'''\n",
    "\n",
    "class CNN(nn.Module):                                                 # Define model\n",
    "    def __init__(self):\n",
    "        super().__init__()\n",
    "        \n",
    "        self.layers_stack = nn.Sequential(\n",
    "            nn.Conv2d(in_channels=1, out_channels=32, kernel_size=3), # shape = [batch_size, 32, 26, 26]\n",
    "            nn.LeakyReLU(negative_slope=0.0),\n",
    "            #nn.Dropout(p=0.5),\n",
    "            nn.MaxPool2d(kernel_size=2),                              # shape = [batch_size, 32, 13, 13]\n",
    "            nn.Flatten(),\n",
    "            nn.Linear(32*13*13, 10)\n",
    "        )\n",
    "       \n",
    "    def forward(self, x):\n",
    "        x = self.layers_stack(x)\n",
    "        return x"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "eec9caac",
   "metadata": {},
   "outputs": [],
   "source": [
    "def performance_estimate(dataset, model, loss_fn, train_or_test):\n",
    "    '''this function computes accuracy and loss of a model on the training or test set'''\n",
    "    data_size = len(dataset)\n",
    "    \n",
    "    dataloader = DataLoader(dataset=dataset, batch_size=batch_size, shuffle=True)\n",
    "    num_batches = len(dataloader)\n",
    "    \n",
    "    model.eval()\n",
    "    loss, accuracy = 0, 0\n",
    "    with torch.no_grad():\n",
    "        for X, y in dataloader:\n",
    "            X, y = X.to(device), y.to(device)\n",
    "            pred = model(X)\n",
    "            accuracy += (pred.argmax(1) == y).sum().item()\n",
    "            loss += loss_fn(pred, y).item()\n",
    "    accuracy /= data_size                                            # accuracy lies in the interval [0, 1]  \n",
    "    loss /= num_batches\n",
    "    print(f\"{train_or_test} accuracy: {round(accuracy, 3)},  {train_or_test} loss: {round(loss,3)}\")\n",
    "    return accuracy, loss\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "def one_epoch(model, loss_fn, optimizer, dataset, batch_size):\n",
    "    \n",
    "    A_train, L_train, A_test, L_test = [], [], [], []\n",
    "\n",
    "    dataloader = DataLoader(dataset=dataset, batch_size=batch_size, shuffle=True)\n",
    "    \n",
    "    model.train()\n",
    "    for batch, (X, y) in enumerate(dataloader):\n",
    "        X, y = X.to(device), y.to(device)\n",
    "\n",
    "        out = model(X)                             # Perform a single forward pass \n",
    "        loss = loss_fn(out, y)          \n",
    "    \n",
    "        optimizer.zero_grad()                      # Clear gradients\n",
    "        loss.backward()                            # Derive gradients, backpropagation\n",
    "        optimizer.step()                           # Update parameters based on gradients\n",
    "        \n",
    "    \n",
    "        if batch % batch_size == 0: \n",
    "            #As training progress, computing and appending loss and accuracy of the model on train and test set\n",
    "            accuracy_train, loss_train = performance_estimate(train_dataset, model, loss_fn, 'train')\n",
    "            accuracy_test, loss_test = performance_estimate(test_dataset, model, loss_fn, 'test ')\n",
    "            print()\n",
    "            \n",
    "            A_train.append(accuracy_train)        # \n",
    "            L_train.append(loss_train)\n",
    "            A_test.append(accuracy_test)\n",
    "            L_test.append(loss_test)\n",
    "            \n",
    "            #print(f\"train loss: {round(loss,3)}\")\n",
    "    \n",
    "    return A_train, L_train, A_test, L_test \n",
    "            \n",
    "\n",
    "    \n",
    "def training(dataset, batch_size, lr_, weight_decay_, epochs):\n",
    "    \n",
    "    model = CNN().to(device)\n",
    "    loss_fn = nn.CrossEntropyLoss()\n",
    "    optimizer = torch.optim.Adam(model.parameters(), lr=lr_, weight_decay=weight_decay_)\n",
    "    \n",
    "    A_Train, L_Train, A_Test, L_Test = [], [], [], []\n",
    "    for t in range(epochs): \n",
    "        print(f\"Epoch {t+1} ---------------------------------- \\n\")\n",
    "        #As training progress, computing and appending loss and accuracy of the model on train and test set\n",
    "        A_train, L_train, A_test, L_test = one_epoch(model, loss_fn, optimizer, dataset, batch_size)\n",
    "        A_Train += A_train\n",
    "        L_Train += L_train \n",
    "        A_Test += A_test\n",
    "        L_Test += L_test\n",
    "        \n",
    "        #accuracy, loss = performance_estimate(test_dataset, model, loss_fn, 'test ')\n",
    "    \n",
    "    model_state_dict = model.state_dict()           # for saving or loading the trained model\n",
    "    \n",
    "    return A_Train, L_Train, A_Test, L_Test, model_state_dict\n",
    "    "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "834b1b08",
   "metadata": {},
   "source": [
    "# training..."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "aa679743",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "batch_size = 64\n",
      "\n",
      "Epoch 1 ---------------------------------- \n",
      "\n",
      "train accuracy: 0.261,  train loss: 2.301\n",
      "test  accuracy: 0.258,  test  loss: 2.303\n",
      "\n",
      "train accuracy: 0.797,  train loss: 0.561\n",
      "test  accuracy: 0.79,  test  loss: 0.578\n",
      "\n",
      "train accuracy: 0.822,  train loss: 0.491\n",
      "test  accuracy: 0.808,  test  loss: 0.515\n",
      "\n",
      "train accuracy: 0.852,  train loss: 0.425\n",
      "test  accuracy: 0.84,  test  loss: 0.451\n",
      "\n",
      "train accuracy: 0.846,  train loss: 0.442\n",
      "test  accuracy: 0.837,  test  loss: 0.471\n",
      "\n",
      "train accuracy: 0.86,  train loss: 0.401\n",
      "test  accuracy: 0.847,  test  loss: 0.431\n",
      "\n",
      "train accuracy: 0.862,  train loss: 0.391\n",
      "test  accuracy: 0.853,  test  loss: 0.42\n",
      "\n",
      "train accuracy: 0.865,  train loss: 0.38\n",
      "test  accuracy: 0.852,  test  loss: 0.412\n",
      "\n",
      "train accuracy: 0.874,  train loss: 0.357\n",
      "test  accuracy: 0.86,  test  loss: 0.393\n",
      "\n",
      "train accuracy: 0.882,  train loss: 0.341\n",
      "test  accuracy: 0.87,  test  loss: 0.37\n",
      "\n",
      "train accuracy: 0.881,  train loss: 0.337\n",
      "test  accuracy: 0.868,  test  loss: 0.369\n",
      "\n",
      "train accuracy: 0.885,  train loss: 0.333\n",
      "test  accuracy: 0.874,  test  loss: 0.364\n",
      "\n",
      "train accuracy: 0.888,  train loss: 0.318\n",
      "test  accuracy: 0.876,  test  loss: 0.352\n",
      "\n",
      "train accuracy: 0.891,  train loss: 0.315\n",
      "test  accuracy: 0.879,  test  loss: 0.347\n",
      "\n",
      "train accuracy: 0.892,  train loss: 0.308\n",
      "test  accuracy: 0.881,  test  loss: 0.342\n",
      "\n",
      " ~~~~~ training is done ~~~~~\n",
      "\n",
      "CPU times: user 18min 59s, sys: 7.44 s, total: 19min 6s\n",
      "Wall time: 4min 47s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "\n",
    "batch_size = 64\n",
    "print(f'batch_size = {batch_size}\\n')\n",
    "\n",
    "#----------------------------------------------------------------------------------\n",
    "\n",
    "\n",
    "A_Train, L_Train, A_Test, L_Test, model_state_dict = training(train_dataset, batch_size=batch_size, \n",
    "                                                              lr_=1e-3, weight_decay_=1e-8, epochs=1)\n",
    "\n",
    "\n",
    "print(f' ~~~~~ training is done ~~~~~\\n')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "234381d4",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(A_Train, label='train set')\n",
    "plt.plot(A_Test, label='test set')\n",
    "plt.ylabel('accuracy')\n",
    "plt.legend()\n",
    "plt.show()\n",
    "\n",
    "plt.plot(L_Train, label='train set')\n",
    "plt.plot(L_Test, label='test set')\n",
    "plt.ylabel('loss')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "416b30e5",
   "metadata": {},
   "source": [
    "## save model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "988bcfc3",
   "metadata": {},
   "outputs": [],
   "source": [
    "torch.save(model_state_dict, \"model_MNIST_CNN.pth\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "bdf78969",
   "metadata": {},
   "source": [
    "## load model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "3eb4ede1",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<All keys matched successfully>"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model = CNN().to(device)\n",
    "model.load_state_dict(model_state_dict)\n",
    "\n",
    "#model.load_state_dict(torch.load(\"model_MNIST_CNN.pth\"))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "52a30c42",
   "metadata": {},
   "source": [
    "## predict test examples"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "b5be82d3",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "x of torch.Size([1, 28, 28]) :\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "true label = y = 5\n",
      "\n",
      "predicted label = 5\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYIAAAEGCAYAAABo25JHAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjUuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/YYfK9AAAACXBIWXMAAAsTAAALEwEAmpwYAAAUf0lEQVR4nO3dfbRddX3n8feHABqfiELWtCRhQIcBg4LoFbDMshanw9MUqFULaF2iDrJGhE5HRmirtqMdmUW16IjELMrgSEd0JItSm8I4FdHxqSSEB4FJm0GFJDiG2qjVrELgO3+cnXpyc5N7Qu6+J8nv/VrrrHv2b//23t+7k3s/Zz/c305VIUlq1z7jLkCSNF4GgSQ1ziCQpMYZBJLUOINAkhq377gL2FkHHXRQHXrooeMuQ5L2KCtXrny0quZPNW+PC4JDDz2UFStWjLsMSdqjJPnu9uZ5akiSGmcQSFLjDAJJapxBIEmNMwgkqXG9BUGSa5N8P8m3tjM/ST6aZE2Se5K8tK9apD3BTavWceLlX+SwS/+cEy//IjetWjfuktSIPo8IrgNO2cH8U4HDu9f5wNU91iLt1m5atY7Llt3Luo2bKGDdxk1ctuxew0CzorcgqKovAz/YQZczgf9WA98A5iX5+b7qkXZnV9y6mk2PP7FV26bHn+CKW1ePqSK1ZJzXCBYADw9Nr+3atpHk/CQrkqzYsGHDrBQnzab1GzftVLs0k8YZBJmibcqn5FTV0qqaqKqJ+fOn/AtpaY928Ly5O9UuzaRxBsFaYNHQ9EJg/ZhqkcbqkpOPYO5+c7Zqm7vfHC45+YgxVaSWjDMIbgbe1N09dALww6p6ZIz1SGNz1rEL+OBrXsz+cwY/kgvmzeWDr3kxZx075dlSaUb1Nuhckk8DrwIOSrIWeB+wH0BVLQGWA6cBa4CfAuf1VYu0Jzjr2AV8+q8eAuAzb3/FmKtRS3oLgqo6Z5r5Bbyjr+1LkkbjXxZLUuMMAklqnEEgSY0zCCSpcQaBJDXOIJCkxhkEktQ4g0CSGmcQSFLjDAJJapxBIEmNMwgkqXEGgSQ1ziCQpMYZBJLUOINAkhpnEEhS4wwCSWqcQSBJjTMIJKlxBoEkNc4gkKTGGQSS1DiDQJIaZxBIUuMMAklqnEEgSY0zCCSpcQaBJDXOIJCkxvUaBElOSbI6yZokl04x/4Akf5bk7iT3JTmvz3okSdvqLQiSzAGuAk4FFgPnJFk8qds7gPur6hjgVcCHkuzfV02SpG31eURwHLCmqh6sqseAG4AzJ/Up4NlJAjwL+AGwuceaJEmT9BkEC4CHh6bXdm3DPga8EFgP3AtcXFVPTl5RkvOTrEiyYsOGDX3VK0lN6jMIMkVbTZo+GbgLOBh4CfCxJM/ZZqGqpVU1UVUT8+fPn+k6JalpfQbBWmDR0PRCBp/8h50HLKuBNcC3gSN7rEmSNEmfQXAHcHiSw7oLwGcDN0/q8xDwaoAk/wQ4Aniwx5okSZPs29eKq2pzkguBW4E5wLVVdV+SC7r5S4D3A9cluZfBqaR3V9WjfdUkSdpWb0EAUFXLgeWT2pYMvV8P/Ks+a5Ak7Zh/WSxJjTMIJKlxBoEkNc4gkKTGGQSS1DiDQJIaZxBIUuMMAklqnEEgSY0zCCSpcQaBJDXOIJCkxhkEktQ4g0CSGmcQSFLjDAJJapxBIEmNMwgkqXEGgSQ1ziCQpMYZBJLUOINAkhpnEEhS4wwCSWqcQSBJjTMIJKlxBoEkNc4gkKTGGQSS1DiDQJIaN1IQJLkxyelJDA5J2suM+ov9auBc4G+SXJ7kyFEWSnJKktVJ1iS5dDt9XpXkriT3Jbl9xHokSTNkpCCoqv9VVW8AXgp8B/hCkq8lOS/JflMtk2QOcBVwKrAYOCfJ4kl95gEfB86oqqOA1z3Vb0SS9NSMfKonyYHAm4G3AauAjzAIhi9sZ5HjgDVV9WBVPQbcAJw5qc+5wLKqegigqr6/U9VLknbZqNcIlgFfAZ4B/EpVnVFVn6mqdwLP2s5iC4CHh6bXdm3D/jnw3CRfSrIyyZu2s/3zk6xIsmLDhg2jlCxJGtG+I/a7pqqWDzckeVpV/UNVTWxnmUzRVlNs/2XAq4G5wNeTfKOq/nqrhaqWAksBJiYmJq9DkrQLRj019IEp2r4+zTJrgUVD0wuB9VP0uaWqflJVjwJfBo4ZsSZJ0gzY4RFBkp9jcDpnbpJj+dmn/OcwOE20I3cAhyc5DFgHnM3gmsCwPwU+lmRfYH/geOCPduo7kCTtkulODZ3M4ALxQuDDQ+0/Bn57RwtW1eYkFwK3AnOAa6vqviQXdPOXVNUDSW4B7gGeZHAK6ltP6TuRJD0lOwyCqvok8Mkkv1ZVN+7syrvrCssntS2ZNH0FcMXOrluSNDOmOzX0xqq6Hjg0yW9Nnl9VH55iMUnSHmS6U0PP7L5u7xZRSdIebrpTQ5/ovv7+7JQjSZpt050a+uiO5lfVRTNbjiRptk13amjlrFQhSRqbUe4akiTtxaY7NXRlVf1mkj9j2+EhqKozeqtMkjQrpjs19Knu6x/2XYgkaTymOzW0svt6e5L9gSMZHBms7oaWliTt4UYafTTJ6cAS4P8yGG/osCRvr6q/6LM4SVL/Rh2G+kPAL1XVGoAkLwD+HDAIJGkPN+ow1N/fEgKdBwGfJiZJe4Hp7hp6Tff2viTLgc8yuEbwOgbDTEuS9nDTnRr6laH3/w/4xe79BuC5vVQkSZpV0901dN5sFSJJGo9R7xp6OvBW4Cjg6Vvaq+otPdUlSZolo14s/hTwcwyeWHY7gyeW/bivoiRJs2fUIPhnVfUe4Cfd+EOnAy/uryxJ0mwZNQge775uTPIi4ADg0F4qkiTNqlH/oGxpkucC7wFuZvDEsvf0VpUkadaMFARVdU339nbg+f2VI0mabSOdGkpyYJL/kuTOJCuTXJnkwL6LkyT1b9RrBDcwGFLi14DXAo8Cn+mrKEnS7Bn1GsHzqur9Q9MfSHJWD/VIkmbZqEcEtyU5O8k+3ev1DEYflSTt4aYbdO7HDAaZC/BbwPXdrH2Avwfe12t1kqTeTTfW0LNnqxBJ0niMeo2AJGcAr+wmv1RVn++nJEnSbBr19tHLgYuB+7vXxV2bJGkPN+oRwWnAS6rqSYAknwRWAZf2VZgkaXaMetcQwLyh9weMskCSU5KsTrImyXZDI8nLkzyR5LU7UY8kaQaMekTwn4BVSW5jcAfRK4HLdrRAkjnAVcAvA2uBO5LcXFX3T9HvPwO37mTtkqQZMG0QJNkHeBI4AXg5gyB4d1V9b5pFjwPWVNWD3XpuAM5kcI1h2DuBG7t1S5Jm2bRBUFVPJrmwqj7LYOTRUS0AHh6aXgscP9whyQLgV4GTMAgkaSxGvUbwhSTvSrIoyfO2vKZZJlO01aTpKxkcXTyxwxUl5ydZkWTFhg0bRixZkjSKUa8RvIXBL/F/O6l9R0NSrwUWDU0vBNZP6jMB3JAE4CDgtCSbq+qm4U5VtRRYCjAxMTE5TCRJu2DUIFjMIAT+BYNA+AqwZJpl7gAOT3IYsA44Gzh3uENVHbblfZLrgM9PDgFJUr9GDYJPAj8CPtpNn9O1vX57C1TV5iQXMrgbaA5wbVXdl+SCbv50QSJJmgWjBsERVXXM0PRtSe6ebqGqWg4sn9Q2ZQBU1ZtHrEWSNINGvVi8KskJWyaSHA98tZ+SJEmzadQjguOBNyV5qJs+BHggyb1AVdXRvVQnSerdqEFwSq9VSJLGZqQgqKrv9l2IJGk8dmbQOUnSXsggkKTGGQSS1DiDQJIaZxBIUuMMAklqnEEgSY0zCCSpcQaBJDXOIJCkxhkEktQ4g0CSGmcQSFLjDAJJapxBIEmNMwgkqXEGgSQ1ziCQpMYZBJLUOINAkhpnEEhS4wwCSWqcQSBJjTMIJKlxBoEkNc4gkKTGGQSS1LhegyDJKUlWJ1mT5NIp5r8hyT3d62tJjumzHknStnoLgiRzgKuAU4HFwDlJFk/q9m3gF6vqaOD9wNK+6pEkTa3PI4LjgDVV9WBVPQbcAJw53KGqvlZVf9dNfgNY2GM9kqQp9BkEC4CHh6bXdm3b81bgL6aakeT8JCuSrNiwYcMMlihJ6jMIMkVbTdkx+SUGQfDuqeZX1dKqmqiqifnz589giZKkfXtc91pg0dD0QmD95E5JjgauAU6tqr/tsR5J0hT6PCK4Azg8yWFJ9gfOBm4e7pDkEGAZ8BtV9dc91iJJ2o7ejgiqanOSC4FbgTnAtVV1X5ILuvlLgPcCBwIfTwKwuaom+qpJkrStPk8NUVXLgeWT2pYMvX8b8LY+a5Ak7Zh/WSxJjTMIJKlxBoEkNc4gkKTGGQSS1DiDQJIaZxBIUuMMAklqnEEgSY0zCCSpcQaBJDXOIJCkxhkEktQ4g0CSGmcQSFLjDAJJapxBIEmNMwgkqXEGgSQ1ziCQpMYZBJLUOINAkhpnEEhS4wwCSWqcQSBJjTMIJKlxBoEkNc4gkKTGGQSS1DiDQJIaZxBIUuP27XPlSU4BPgLMAa6pqssnzU83/zTgp8Cbq+rOma7jplXruOLW1azfuImD583lkpOP4KxjF8z0ZqzDOvZ4u8u+sI7ZraO3IEgyB7gK+GVgLXBHkpur6v6hbqcCh3ev44Gru68z5qZV67hs2b1sevwJANZt3MRly+4FmNV/UOuwjt3d7rIvrGP260hVzciKtllx8grg96rq5G76MoCq+uBQn08AX6qqT3fTq4FXVdUj21vvxMRErVixYuQ6Trz8i6zbuIm33/OnPP+H6/6x/Wn7zuHYQ+bt3De1C1Y9tJF/2PzENu3WYR3D7n/kRwAs/vnnzPq2d5d9YR3br+PBAxbwiaPPBGDBvLl89dKTRl5PkpVVNTHVvD5PDS0AHh6aXsu2n/an6rMA2CoIkpwPnA9wyCGH7FQR6zdumrJ9qn/gPm1ve9ZhHcPGEQBb7C77wjpG2972frc9FX0GQaZom3z4MUofqmopsBQGRwQ7U8TB8+aybuOmf0zRLRbMm8uv70Sa7qpzuyOTyazDOnYXu8u+sI7R6jh43twZ20afdw2tBRYNTS8E1j+FPrvkkpOPYO5+c7Zqm7vfHC45+YiZ3Ix1WMceb3fZF9Yx+3X0eURwB3B4ksOAdcDZwLmT+twMXJjkBganjX64o+sDT8WWiynjvvJvHdaxu9td9oV1zH4dvV0sBkhyGnAlg9tHr62qP0hyAUBVLeluH/0YcAqD20fPq6odXgne2YvFkqTxXSymqpYDyye1LRl6X8A7+qxBkrRj/mWxJDXOIJCkxhkEktQ4g0CSGtfrXUN9SLIB+O5TXPwg4NEZLGdP5/7YmvvjZ9wXW9sb9sc/rar5U83Y44JgVyRZsb3bp1rk/tia++Nn3Bdb29v3h6eGJKlxBoEkNa61IFg67gJ2M+6Prbk/fsZ9sbW9en80dY1AkrSt1o4IJEmTGASS1LhmgiDJKUlWJ1mT5NJx1zNOSRYluS3JA0nuS3LxuGsatyRzkqxK8vlx1zJuSeYl+VyS/9P9H3nFuGsalyT/rvsZ+VaSTyd5+rhr6kMTQZBkDnAVcCqwGDgnyeLxVjVWm4F/X1UvBE4A3tH4/gC4GHhg3EXsJj4C3FJVRwLH0Oh+SbIAuAiYqKoXMRhO/+zxVtWPJoIAOA5YU1UPVtVjwA3AmdMss9eqqkeq6s7u/Y8Z/KC39ySWTpKFwOnANeOuZdySPAd4JfDHAFX1WFVtHGtR47UvMDfJvsAzmOEnKO4uWgmCBcDDQ9NrafgX37AkhwLHAt8ccynjdCXwH4Anx1zH7uD5wAbgv3anyq5J8sxxFzUOVbUO+EPgIeARBk9Q/J/jraofrQRBpmhr/r7ZJM8CbgR+s6p+NO56xiHJvwa+X1Urx13LbmJf4KXA1VV1LPAToMlrakmey+DMwWHAwcAzk7xxvFX1o5UgWAssGppeyF56iDeqJPsxCIE/qapl465njE4EzkjyHQanDE9Kcv14SxqrtcDaqtpyhPg5BsHQon8JfLuqNlTV48Ay4BfGXFMvWgmCO4DDkxyWZH8GF3xuHnNNY9M9K/qPgQeq6sPjrmecquqyqlpYVYcy+H/xxaraKz/1jaKqvgc8nOSIrunVwP1jLGmcHgJOSPKM7mfm1eylF857fWbx7qKqNie5ELiVwZX/a6vqvjGXNU4nAr8B3Jvkrq7tt7tnTEvvBP6k+9D0IHDemOsZi6r6ZpLPAXcyuNNuFXvpUBMOMSFJjWvl1JAkaTsMAklqnEEgSY0zCCSpcQaBJDXOIFBzkvxeknf1tO7vJDlomj5/v5Pr7K1eCQwCSWqeQaC9WpI3Jbknyd1JPjXF/H+T5I5u/o1JntG1v64bg/7uJF/u2o5K8ldJ7urWefg0274pycpuPPvzJ837UJI7k/xlkvld2wuS3NIt85UkR06xzouS3N9t/4Zd2TfSFgaB9lpJjgJ+Bzipqo5h8MyByZZV1cu7+Q8Ab+3a3wuc3LWf0bVdAHykql4CTDAYl2dH3lJVL+v6XpTkwK79mcCdVfVS4HbgfV37UuCd3TLvAj4+xTovBY6tqqO7eqRd1sQQE2rWScDnqupRgKr6wRR9XpTkA8A84FkMhiEB+CpwXZLPMhhsDODrwO90zy9YVlV/M832L0ryq937RcDhwN8yGO76M1379cCybiTYXwD+x2BYGwCeNsU672Ew/MNNwE3TbF8aiUcE2puF6Ycbvw64sKpeDPw+8HSAqroA+F0Gv8DvSnJgVf13BkcHm4Bbk5y03Q0nr2IweuUruqOKVVvWPYVi8LO4sapeMvR64RR9T2fwtL2XASu7B6ZIu8Qg0N7sL4HXbzklk+R5U/R5NvBINyz3G7Y0JnlBVX2zqt4LPAosSvJ84MGq+iiD0WuP3sG2DwD+rqp+2p3rP2Fo3j7Aa7v35wL/u3sexLeTvK7bfpIcM7zCJPsAi6rqNgYP0pnH4ChG2iV+mtBeq6ruS/IHwO1JnmDwqfzNk7q9h8HT2b4L3MsgGACu6C4Gh0Gg3M3g/PwbkzwOfA/4jzvY/C3ABUnuAVYD3xia9xPgqCQrgR8Cv961vwG4OsnvAvsxeD7C3UPLzQGuT3JAV9cfNf4YSc0QRx+VpMZ5akiSGmcQSFLjDAJJapxBIEmNMwgkqXEGgSQ1ziCQpMb9fxSzyZdhsdh3AAAAAElFTkSuQmCC",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "idx = np.random.choice(len(test_dataset))\n",
    "\n",
    "x = test_dataset[idx][0]\n",
    "print(f'x of {x.shape} :')\n",
    "plt.imshow(x[0], cmap='gray')\n",
    "plt.show()\n",
    "\n",
    "print(f'true label = y = {test_dataset[idx][1]}\\n')\n",
    "\n",
    "\n",
    "out_ = model(x.view(1, 1, 28, 28)).detach().flatten()\n",
    "prob = F.softmax(out_, dim=0)\n",
    "pred = prob.argmax().item()\n",
    "print(f'predicted label = {pred}\\n')\n",
    "\n",
    "plt.stem(np.arange(10), prob)\n",
    "plt.ylabel('probability')\n",
    "plt.xlabel('class labels')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8b298922",
   "metadata": {},
   "source": [
    "$ $\n",
    "\n",
    "$ $\n",
    "\n",
    "## simplified training loop"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "b3dcb3aa",
   "metadata": {},
   "outputs": [],
   "source": [
    "def one_epoch_(model, loss_fn, optimizer, dataset, batch_size):\n",
    "    \n",
    "    dataloader = DataLoader(dataset=dataset, batch_size=batch_size, shuffle=True)\n",
    "    \n",
    "    model.train()\n",
    "    for batch, (X, y) in enumerate(dataloader):\n",
    "        X, y = X.to(device), y.to(device)\n",
    "\n",
    "        out = model(X)                             # Perform a single forward pass\n",
    "        loss = loss_fn(out, y)          \n",
    "    \n",
    "        optimizer.zero_grad()                      # Clear gradients\n",
    "        loss.backward()                            # Derive gradients, backpropagation\n",
    "        optimizer.step()                           # Update parameters based on gradients\n",
    "    \n",
    "        if batch % batch_size == 0: \n",
    "            print(f\"train loss: {round(loss.item(),3)}\")\n",
    "\n",
    "\n",
    "    \n",
    "    \n",
    "def training_(dataset, batch_size, lr_, weight_decay_, epochs):\n",
    "    \n",
    "    model = CNN().to(device)\n",
    "    loss_fn = nn.CrossEntropyLoss()\n",
    "    optimizer = torch.optim.Adam(model.parameters(), lr=lr_, weight_decay=weight_decay_)\n",
    "    \n",
    "    for t in range(epochs): \n",
    "        print(f\"Epoch {t+1} ---------------------------------- \\n \")\n",
    "        one_epoch_(model, loss_fn, optimizer, dataset, batch_size)\n",
    "        \n",
    "        #accuracy_train, loss_train = performance_estimate(train_dataset, model, loss_fn, 'train')\n",
    "        #accuracy_test, loss_test = performance_estimate(test_dataset, model, loss_fn, 'test ')\n",
    "        #print()\n",
    "\n",
    "    \n",
    "    model_state_dict = model.state_dict()\n",
    "    return model_state_dict\n",
    "\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "7217eb93",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "batch_size = 64\n",
      "\n",
      "Epoch 1 ---------------------------------- \n",
      " \n",
      "train loss: 2.347\n",
      "train loss: 0.548\n",
      "train loss: 0.659\n",
      "train loss: 0.437\n",
      "train loss: 0.382\n",
      "train loss: 0.488\n",
      "train loss: 0.368\n",
      "train loss: 0.343\n",
      "train loss: 0.588\n",
      "train loss: 0.183\n",
      "train loss: 0.207\n",
      "train loss: 0.387\n",
      "train loss: 0.236\n",
      "train loss: 0.477\n",
      "train loss: 0.41\n",
      "Epoch 2 ---------------------------------- \n",
      " \n",
      "train loss: 0.507\n",
      "train loss: 0.256\n",
      "train loss: 0.322\n",
      "train loss: 0.385\n",
      "train loss: 0.557\n",
      "train loss: 0.129\n",
      "train loss: 0.353\n",
      "train loss: 0.314\n",
      "train loss: 0.232\n",
      "train loss: 0.333\n",
      "train loss: 0.473\n",
      "train loss: 0.331\n",
      "train loss: 0.394\n",
      "train loss: 0.225\n",
      "train loss: 0.389\n",
      "Epoch 3 ---------------------------------- \n",
      " \n",
      "train loss: 0.307\n",
      "train loss: 0.323\n",
      "train loss: 0.211\n",
      "train loss: 0.186\n",
      "train loss: 0.143\n",
      "train loss: 0.103\n",
      "train loss: 0.195\n",
      "train loss: 0.356\n",
      "train loss: 0.313\n",
      "train loss: 0.265\n",
      "train loss: 0.308\n",
      "train loss: 0.205\n",
      "train loss: 0.216\n",
      "train loss: 0.192\n",
      "train loss: 0.298\n",
      "Epoch 4 ---------------------------------- \n",
      " \n",
      "train loss: 0.376\n",
      "train loss: 0.236\n",
      "train loss: 0.335\n",
      "train loss: 0.319\n",
      "train loss: 0.281\n",
      "train loss: 0.272\n",
      "train loss: 0.069\n",
      "train loss: 0.29\n",
      "train loss: 0.153\n",
      "train loss: 0.19\n",
      "train loss: 0.339\n",
      "train loss: 0.256\n",
      "train loss: 0.276\n",
      "train loss: 0.286\n",
      "train loss: 0.303\n",
      "Epoch 5 ---------------------------------- \n",
      " \n",
      "train loss: 0.17\n",
      "train loss: 0.394\n",
      "train loss: 0.154\n",
      "train loss: 0.177\n",
      "train loss: 0.284\n",
      "train loss: 0.206\n",
      "train loss: 0.304\n",
      "train loss: 0.101\n",
      "train loss: 0.1\n",
      "train loss: 0.213\n",
      "train loss: 0.313\n",
      "train loss: 0.247\n",
      "train loss: 0.273\n",
      "train loss: 0.171\n",
      "train loss: 0.229\n",
      "\n",
      " ~~~~~ training is done ~~~~~\n",
      "\n",
      "CPU times: user 6min 50s, sys: 30.9 s, total: 7min 21s\n",
      "Wall time: 1min 50s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "\n",
    "batch_size = 64\n",
    "print(f'batch_size = {batch_size}\\n')\n",
    "\n",
    "#----------------------------------------------------------------------------------\n",
    "\n",
    "model_state_dict = training_(train_dataset, batch_size=batch_size, lr_=1e-3, weight_decay_=1e-8, epochs=5)\n",
    "\n",
    "print()\n",
    "print(f' ~~~~~ training is done ~~~~~\\n')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "6f5645b9",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train accuracy: 0.931,  train loss: 0.199\n",
      "test  accuracy: 0.9,  test  loss: 0.277\n",
      "\n",
      "CPU times: user 1min 8s, sys: 16 ms, total: 1min 8s\n",
      "Wall time: 17.2 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "\n",
    "model = CNN().to(device)\n",
    "model.load_state_dict(model_state_dict)\n",
    "loss_fn = nn.CrossEntropyLoss()\n",
    "\n",
    "\n",
    "accuracy_train, loss_train = performance_estimate(train_dataset, model, loss_fn, 'train')\n",
    "accuracy_test, loss_test = performance_estimate(test_dataset, model, loss_fn, 'test ')\n",
    "print()\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "22dc309a",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "3a7fbdfc",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "29222ffd",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "id": "2c75a3b1",
   "metadata": {},
   "source": [
    "# "
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.10.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}