Demo.ipynb

{
 "cells": [
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# CNN from scratch\n",
    "### Demo: Building LeNet\n",
    "\n",
    "(Training this model for 1 epoch takes roughly 5 minutes on Apple M1 chip. You can opt to skip the first section and load a pre-trained model in the next section.)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loading datasets...\n",
      "Training data: 60000 28x28 images\n",
      "Testing data: 10000 28x28 images\n",
      "==================================================\n",
      "Iteration: 0 | Loss: 2.388\n",
      "Iteration: 1000 | Loss: 0.248\n",
      "Iteration: 2000 | Loss: 0.040\n",
      "Iteration: 3000 | Loss: 1.202\n",
      "Iteration: 4000 | Loss: 0.215\n",
      "Iteration: 5000 | Loss: 0.036\n",
      "Iteration: 6000 | Loss: 0.036\n",
      "Traninig completed!\n",
      "Total training time:307.4896s\n",
      "Training iteration : 6000\n",
      "Model saved.\n"
     ]
    }
   ],
   "source": [
    "from Get_Datasets import get_data_from_torchvision_datasets\n",
    "from CNN_lib import Linear, Conv2d, MaxPool2d, ReLU, Flatten, Sequential, Cross_entropy, SGD, DataLoader\n",
    "import numpy as np\n",
    "import time\n",
    "\n",
    "# download dataset from torchvision datasets, then convert to numpy arrays\n",
    "print('Loading datasets...')\n",
    "train_data, train_label, test_data, test_label = get_data_from_torchvision_datasets()\n",
    "print(f'Training data: {train_data.shape[0]} {train_data.shape[1]}x{train_data.shape[2]} images')\n",
    "print(f'Testing data: {test_data.shape[0]} {test_data.shape[1]}x{test_data.shape[2]} images')\n",
    "\n",
    "# define model: Using a LeNet-like structure, mostly the same, except using ReLU instead of sigmoid\n",
    "# and maxpooling rather than average pooling\n",
    "model = Sequential(\n",
    "    Conv2d(in_channels=1, out_channels=6, kernel_size=5, stride=1, padding=2),\n",
    "    ReLU(),\n",
    "    MaxPool2d(kernel_size=2, stride=2),\n",
    "    Conv2d(in_channels=6, out_channels=16, kernel_size=5, stride=1, padding=0),\n",
    "    ReLU(),\n",
    "    MaxPool2d(kernel_size=2, stride=2),\n",
    "    Flatten(),\n",
    "    Linear(400, 120),\n",
    "    ReLU(),\n",
    "    Linear(120, 84),\n",
    "    ReLU(),\n",
    "    Linear(84, 10),\n",
    "    )\n",
    "print('='*50)\n",
    "\n",
    "# use SGD optimizer\n",
    "optimizer = SGD(model, lr=1e-3)\n",
    "\n",
    "epochs = 1\n",
    "iteration = 0\n",
    "starting_time = time.time()\n",
    "\n",
    "for _ in range(epochs):\n",
    "\n",
    "    # define training dataloader\n",
    "    trainloader = DataLoader(train_data, train_label, batch_size=10, shuffle=True)\n",
    "    \n",
    "    for data in trainloader:\n",
    "        images = data[0]\n",
    "        labels = data[1]\n",
    "        # forward\n",
    "        loss = Cross_entropy(model, images, labels)\n",
    "        # backward\n",
    "        loss.backward()\n",
    "        # update\n",
    "        optimizer.step()\n",
    "\n",
    "        if iteration % 1000 == 0:\n",
    "            print(f'Iteration: {iteration} | Loss: {loss.item():.3f}')\n",
    "        iteration += 1\n",
    "    print(f'Iteration: {iteration} | Loss: {loss.item():.3f}')\n",
    "\n",
    "print('Traninig completed!')\n",
    "print(f'Total training time:{time.time() - starting_time:.4f}s')\n",
    "print(f'Training iteration :', iteration)\n",
    "\n",
    "# save trained model to 'pretrained_model.npy'\n",
    "model.save_model('pretrained_model')\n",
    "print('Model saved.')\n",
    "\n",
    "# 1 epoch of training takes about 5 minutes on Apple M1 chip"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### (Optional) Load pre-trained model\n",
    "To save your time."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from Get_Datasets import get_data_from_torchvision_datasets\n",
    "from CNN_lib import Linear, Conv2d, MaxPool2d, ReLU, Flatten, Sequential, Cross_entropy, SGD, DataLoader\n",
    "import numpy as np\n",
    "import time\n",
    "# get data\n",
    "train_data, train_label, test_data, test_label = get_data_from_torchvision_datasets()\n",
    "\n",
    "model = Sequential(\n",
    "    Conv2d(in_channels=1, out_channels=6, kernel_size=5, stride=1, padding=2),\n",
    "    ReLU(),\n",
    "    MaxPool2d(kernel_size=2, stride=2),\n",
    "    Conv2d(in_channels=6, out_channels=16, kernel_size=5, stride=1, padding=0),\n",
    "    ReLU(),\n",
    "    MaxPool2d(kernel_size=2, stride=2),\n",
    "    Flatten(),\n",
    "    Linear(400, 120),\n",
    "    ReLU(),\n",
    "    Linear(120, 84),\n",
    "    ReLU(),\n",
    "    Linear(84, 10),\n",
    "    )\n",
    "\n",
    "model.load_model('pretrained_model_5_epochs.npy')\n",
    "# or \n",
    "# model.load_model('pretrained_model.npy')"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Model information"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Sequential( \n",
      "  Conv2d(1, 6, kernel_size=(5, 5), stride=(1, 1), padding=((2, 2)))\n",
      "  ReLU()\n",
      "  MaxPool2d(kernel_size=2, stride=2, padding=0)\n",
      "  Conv2d(6, 16, kernel_size=(5, 5), stride=(1, 1), padding=((0, 0)))\n",
      "  ReLU()\n",
      "  MaxPool2d(kernel_size=2, stride=2, padding=0)\n",
      "  Flatten(start_dim=1, end_dim=-1)\n",
      "  Linear(in_features=400, out_features=120, bias=True)\n",
      "  ReLU()\n",
      "  Linear(in_features=120, out_features=84, bias=True)\n",
      "  ReLU()\n",
      "  Linear(in_features=84, out_features=10, bias=True)\n",
      ")\n",
      "Number of parameters : 61706\n"
     ]
    }
   ],
   "source": [
    "print(model)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Test accuracy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "------------------------------\n",
      "test_accuracy : 98.58%\n",
      "------------------------------\n"
     ]
    }
   ],
   "source": [
    "# compute accuracy using the whole 10000 test images\n",
    "\n",
    "testloader = DataLoader(test_data, test_label, batch_size=1000, shuffle=False)\n",
    "test_accuracy = []\n",
    "for data in testloader:\n",
    "    images, labels = data\n",
    "    test_accuracy.append(np.mean(np.argmax(model(images), axis=1) == labels))\n",
    "\n",
    "test_accuracy = np.mean(test_accuracy)\n",
    "print('-'*30)\n",
    "print(f'test_accuracy : {test_accuracy*100:.2f}%')\n",
    "print('-'*30)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Classify images with trained model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 864x864 with 36 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "cols, rows = 6, 6,\n",
    "testloader = DataLoader(test_data, test_label, batch_size=cols*rows, shuffle=True)\n",
    "image_data, label_data = testloader[0]\n",
    "\n",
    "figure = plt.figure(figsize=(12, 12))\n",
    "for i in range(1, cols * rows + 1):\n",
    "    predicted_label = np.argmax(model(image_data[i-1].reshape(1, 1, 28, 28)))\n",
    "    image = np.squeeze(image_data[i-1])\n",
    "    # prediction correct -> true; incorrect -> false\n",
    "    if_correct = (predicted_label==label_data[i-1])\n",
    "    figure.add_subplot(rows, cols, i)\n",
    "    plt.imshow(image, cmap='gray')\n",
    "    plt.axis('off')\n",
    "    plt.title(f'Prediction:{predicted_label} | {if_correct}')"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Total time consumption for forward and backward passes in each layer during training."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\t\tforward\t\tbackward\n",
      "Conv2d    |\t19.606s\t\t157.369s\n",
      "ReLU      |\t0.633s\t\t1.039s\n",
      "MaxPool2d |\t11.346s\t\t49.476s\n",
      "Conv2d    |\t10.881s\t\t44.958s\n",
      "ReLU      |\t0.281s\t\t0.852s\n",
      "MaxPool2d |\t3.415s\t\t9.228s\n",
      "Flatten   |\t0.186s\t\t0.026s\n",
      "Linear    |\t0.904s\t\t1.489s\n",
      "ReLU      |\t0.072s\t\t0.142s\n",
      "Linear    |\t0.195s\t\t0.338s\n",
      "ReLU      |\t0.044s\t\t0.131s\n",
      "Linear    |\t0.059s\t\t0.189s\n",
      "----------------------------------------\n",
      "Total forward + backward time : 312.8575s\n",
      "Iterations : 6000\n"
     ]
    }
   ],
   "source": [
    "print('\\t\\tforward\\t\\tbackward')\n",
    "for layer, i in zip(model.layers, model.forward_time):\n",
    "    print(f'{layer.name}|\\t{model.forward_time[i]:.3f}s\\t\\t{model.backward_time[i]:.3f}s')\n",
    "print('-'*40)\n",
    "print(f'Total forward + backward time : {sum(model.forward_time[i]+model.backward_time[i] for i in model.forward_time):.4f}s')\n",
    "print(f'Iterations :', iteration)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "base",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.9"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}