Assignment-4-CNN-CIFAR-10_Abdullah_Karagøz.ipynb

{"nbformat":4,"nbformat_minor":0,"metadata":{"accelerator":"GPU","colab":{"name":"Assignment-4-CNN-CIFAR-10_Abdullah_Karagøz.ipynb","provenance":[],"collapsed_sections":["kAlKQ38zC3pv","at11vCUTe2SN","tib_cQVAuMnV","8n0womPPf4Ud","dinIwj7QuUqZ","VD7FGA8UgcGe","gw0E69dduePT","tHG_DVGohRwh","09uuCyVn57_8","UXHWaRBphYe8","RfkIM5ihjqdF","ldiy2OFWulTA","6PKI76wd6fJs","sa39bLRXkQAN","xHBV9C-Cuttp"],"authorship_tag":"ABX9TyPWFoj/MCdaLConoyNR3iQu"},"kernelspec":{"display_name":"Python 3","name":"python3"},"language_info":{"name":"python"},"widgets":{"application/vnd.jupyter.widget-state+json":{"59a6161b168946b782cf9ddf8964c89b":{"model_module":"@jupyter-widgets/controls","model_name":"HBoxModel","model_module_version":"1.5.0","state":{"_view_name":"HBoxView","_dom_classes":[],"_model_name":"HBoxModel","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.5.0","box_style":"","layout":"IPY_MODEL_04d97b53f0b84c96aaea15a45154963c","_model_module":"@jupyter-widgets/controls","children":["IPY_MODEL_6f3ae6f26f1e442ca17c15ccb4f9c76b","IPY_MODEL_05576de6269a4f9d85345301b3031d64","IPY_MODEL_4de88f0a4e1c4d6f8cd742b937799ae3"]}},"04d97b53f0b84c96aaea15a45154963c":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"6f3ae6f26f1e442ca17c15ccb4f9c76b":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_8f5cc96cb4184e3186d23b279bf8db1b","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":"","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_0c102031597c471b8e1d3f3af33b1946"}},"05576de6269a4f9d85345301b3031d64":{"model_module":"@jupyter-widgets/controls","model_name":"FloatProgressModel","model_module_version":"1.5.0","state":{"_view_name":"ProgressView","style":"IPY_MODEL_ebcca2b5988d41ea9a4baadde3c95b2a","_dom_classes":[],"description":"","_model_name":"FloatProgressModel","bar_style":"success","max":170498071,"_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":170498071,"_view_count":null,"_view_module_version":"1.5.0","orientation":"horizontal","min":0,"description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_0597d16d14c84e01a1ecb6da900405b1"}},"4de88f0a4e1c4d6f8cd742b937799ae3":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_24ee78a6d8014dd9bd4c955646300c77","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":" 170499072/? [00:03&lt;00:00, 56315649.74it/s]","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_67d2ff6628ec420eaf287b17dd7abda1"}},"8f5cc96cb4184e3186d23b279bf8db1b":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"0c102031597c471b8e1d3f3af33b1946":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"ebcca2b5988d41ea9a4baadde3c95b2a":{"model_module":"@jupyter-widgets/controls","model_name":"ProgressStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"ProgressStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","bar_color":null,"_model_module":"@jupyter-widgets/controls"}},"0597d16d14c84e01a1ecb6da900405b1":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"24ee78a6d8014dd9bd4c955646300c77":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"67d2ff6628ec420eaf287b17dd7abda1":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}}}}},"cells":[{"cell_type":"markdown","metadata":{"id":"tUz8eDT3CqwP"},"source":["# Mandatory assignment 4 by Abdullah Karagøz"]},{"cell_type":"code","metadata":{"id":"cMdqaimhjOI4"},"source":["%matplotlib inline\n","import matplotlib.pyplot as plt\n","\n","import numpy as np\n","\n","import torch\n","import torch.nn.functional as F\n","from torch import nn\n","from torch import optim\n","from torchvision import datasets, transforms\n","from torchvision.utils import make_grid"],"execution_count":null,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"kAlKQ38zC3pv"},"source":["## Introduction"]},{"cell_type":"markdown","metadata":{"id":"uJRbuZfUsWIN"},"source":["In this assignment we'll just experiment with Convolutional Neural Network with CIFAR-10 dataset.\n","\n","I'll do the following as assignment asks:\n","1. Download the CIFAR-10 dataset and preprocess it.\n","2. Print size of the train and test datasets, shape of the images and different classes of the dataset.\n","3. Visualize 2 or 3 random images and their class labels from the training and testing datasets, respectively.\n","4. Split the training dataset into training and validation sets.\n","5. Use data loaders for the three datasets.\n","6. Implementing convolutional neural network class to classify CIFAR-10 data using nn.Module.\n","7. Change the runtime type to a GPU to speed up the training.Then, plot the training and validation losses with respect to the epoch numbers.\n","8. Test CNN using the test data. Report the accuracy of your CNN classification overall and per-class.\n","9. Make a new CNN class that is a copy of original CNN. Then add a convolutional layer to the new network architecture. Train the new network, the same way as the first network (same hyperparameters, optimizer and number of epochs ...) and compare the accuracy results (overall and perclass) with the old network.\n","\n","### Goal\n","The goal is to see how CNN works and how one more convolution layer affects.\n","\n","### Method\n","We will build a CNN model using PyTorch, train and test the model using CIFAR-10 image set. Then make a new model with one more convolution layer to the model, train and test it again. And compare the results. We will use mini-batch SDG (batch size = 32) with Momentum (Momentum rate 0.4) in optimization. We'll measure results with prediction accuracy and Cross Entropy Loss.\n","\n","### Assignment outline\n","\n","1. Download the data from CIFAR-10 using torchvision.datasets\n","2. Data exploration (size, content, label classes)\n","3. Visualize data instances (show some images from the set)\n","4. Split the original training dataset between training and validation\n","5. Preparing dataloaders for training, validation and test sets.\n","6. Building first CNN model class.\n","7. Training the dataset\n","> 7.1 Functions for training (with validation) and testing routine\n",">\n","> 7.2 Training the dataset with the model using a GPU, plot the training curve\n","8. Report the test set performance\n","9. Second CNN model\n","> 9.1 Build the second CNN model with an additional convolution stage, training the dataset with the model. \n",">\n","> 9.2 Compare the results from the first CNN model.\n",">\n","> 9.3 Results\n","\n","\n","\n","\n","\n"]},{"cell_type":"markdown","metadata":{"id":"at11vCUTe2SN"},"source":["## 1. Download the CIFAR-10 data set"]},{"cell_type":"code","metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"7rJN1D86nq2C","executionInfo":{"status":"ok","timestamp":1633470192271,"user_tz":-120,"elapsed":8,"user":{"displayName":"Abdullah Karagoz","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"06337825384430712267"}},"outputId":"2f9de876-5c80-4c15-ad6d-a7017a65f1f6"},"source":["# This is to make it more reproductible, taken from https://pytorch.org/docs/stable/notes/randomness.html\n","seed = 3\n","\n","import random\n","random.seed(seed)\n","np.random.seed(seed)\n","torch.manual_seed(seed)\n","if torch.cuda.is_available():\n","  torch.cuda.manual_seed(seed)\n","  torch.cuda.manual_seed_all(seed)\n","  torch.backends.cudnn.deterministic = True\n","  torch.backends.cudnn.benchmark = False\n","\n","# The function and generator will be used when generating dataloaders as parameters\n","def seed_worker(worker_id):\n","    worker_seed = torch.initial_seed() % 2**32\n","    np.random.seed(worker_seed)\n","    random.seed(worker_seed)\n","\n","g = torch.Generator()\n","g.manual_seed(seed)"],"execution_count":null,"outputs":[{"output_type":"execute_result","data":{"text/plain":["<torch._C.Generator at 0x7fc74444f450>"]},"metadata":{},"execution_count":2}]},{"cell_type":"code","metadata":{"colab":{"base_uri":"https://localhost:8080/","height":103,"referenced_widgets":["59a6161b168946b782cf9ddf8964c89b","04d97b53f0b84c96aaea15a45154963c","6f3ae6f26f1e442ca17c15ccb4f9c76b","05576de6269a4f9d85345301b3031d64","4de88f0a4e1c4d6f8cd742b937799ae3","8f5cc96cb4184e3186d23b279bf8db1b","0c102031597c471b8e1d3f3af33b1946","ebcca2b5988d41ea9a4baadde3c95b2a","0597d16d14c84e01a1ecb6da900405b1","24ee78a6d8014dd9bd4c955646300c77","67d2ff6628ec420eaf287b17dd7abda1"]},"id":"U_iNW6qZoNfd","executionInfo":{"status":"ok","timestamp":1633470199860,"user_tz":-120,"elapsed":7592,"user":{"displayName":"Abdullah Karagoz","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"06337825384430712267"}},"outputId":"cdb9a1ae-bdb2-4453-e4b0-d3a47922f4cb"},"source":["# Downloading data, normalization using values from https://github.com/kuangliu/pytorch-cifar/issues/19\n","transform_data = transforms.Compose([transforms.ToTensor(),\n","                                     transforms.Normalize((0.4914, 0.4822, 0.4465), (0.247, 0.243, 0.261))\n","                                     ])\n","\n","train = datasets.CIFAR10(root='cifar10/', download=True, train=True, transform=transform_data)\n","test = datasets.CIFAR10(root='cifar10/', download=True, train=False, transform=transform_data)"],"execution_count":null,"outputs":[{"output_type":"stream","name":"stdout","text":["Downloading https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz to cifar10/cifar-10-python.tar.gz\n"]},{"output_type":"display_data","data":{"application/vnd.jupyter.widget-view+json":{"model_id":"59a6161b168946b782cf9ddf8964c89b","version_minor":0,"version_major":2},"text/plain":["  0%|          | 0/170498071 [00:00<?, ?it/s]"]},"metadata":{}},{"output_type":"stream","name":"stdout","text":["Extracting cifar10/cifar-10-python.tar.gz to cifar10/\n","Files already downloaded and verified\n"]}]},{"cell_type":"markdown","metadata":{"id":"tib_cQVAuMnV"},"source":["## 2. Data exploration (size, content, label classes)"]},{"cell_type":"code","metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"-1nDY8WPjnWK","executionInfo":{"status":"ok","timestamp":1633470199861,"user_tz":-120,"elapsed":4,"user":{"displayName":"Abdullah Karagoz","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"06337825384430712267"}},"outputId":"02ca49aa-c664-469b-b33e-901b8d98ba45"},"source":["# Printing length and classes\n","print(\"Length of training set:\", len(train))\n","print(\"Length of testing set:\", len(test))\n","image, label = next(iter(train))\n","print(\"Shape of an image: \", image.shape)\n","\n","classes_dict = {c: l for c,l in zip(train.classes, range(10))}\n","\n","print(\"\\nThe classes of the dataset are: \", classes_dict)\n","\n","print(\"\\n\\nFirst 10 labels: \")\n","j = 10\n","for image, label in train:\n","  j -= 1\n","  print(\" {} ({}), \".format(label, train.classes[label]), end=\"\")\n","  if j == 0:\n","    break"],"execution_count":null,"outputs":[{"output_type":"stream","name":"stdout","text":["Length of training set: 50000\n","Length of testing set: 10000\n","Shape of an image:  torch.Size([3, 32, 32])\n","\n","The classes of the dataset are:  {'airplane': 0, 'automobile': 1, 'bird': 2, 'cat': 3, 'deer': 4, 'dog': 5, 'frog': 6, 'horse': 7, 'ship': 8, 'truck': 9}\n","\n","\n","First 10 labels: \n"," 6 (frog),  9 (truck),  9 (truck),  4 (deer),  1 (automobile),  1 (automobile),  2 (bird),  7 (horse),  8 (ship),  3 (cat), "]}]},{"cell_type":"markdown","metadata":{"id":"8n0womPPf4Ud"},"source":["## 3. Visualize data instances (show some images from the set)"]},{"cell_type":"code","metadata":{"colab":{"base_uri":"https://localhost:8080/","height":576},"id":"2PIEyw4IsNfK","executionInfo":{"status":"ok","timestamp":1633470201381,"user_tz":-120,"elapsed":1191,"user":{"displayName":"Abdullah Karagoz","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"06337825384430712267"}},"outputId":"c8d57c3a-78e4-4459-cf12-1adbfe0f07e6"},"source":["# Showing some random images from the training set and testing set\n","indexes = np.random.randint(0, 50000, 3)\n","\n","#This function is taken and partly edited from https://pytorch.org/tutorials/beginner/blitz/cifar10_tutorial.html?highlight=cifar\n","def imshow(img, lbl, ax):\n","    img2 = img / 2 + 0.5     # unnormalize\n","    npimg = img2.numpy()\n","    ax.set_title(train.classes[lbl])\n","    ax.imshow(np.transpose(npimg, (1, 2, 0)))\n","\n","fig, axes = plt.subplots(1, 3)\n","for i in range(3):\n","  image, label = train[indexes[i]]\n","  imshow(make_grid(image), label, axes[i])\n","fig.tight_layout(pad=1.0)\n","plt.suptitle(\"3 random images from the train set\")\n","plt.show()\n","\n","indexes = np.random.randint(0, 10000, 3)\n","fig, axes = plt.subplots(1, 3)\n","for i in range(3):\n","  image, label = test[indexes[i]]\n","  imshow(make_grid(image), label, axes[i])\n","fig.tight_layout(pad=1.0)\n","plt.suptitle(\"3 random images from the test set\")\n","plt.show()"],"execution_count":null,"outputs":[{"output_type":"stream","name":"stderr","text":["Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n","Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n","Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n"]},{"output_type":"display_data","data":{"image/png":"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\n","text/plain":["<Figure size 432x288 with 3 Axes>"]},"metadata":{"needs_background":"light"}},{"output_type":"stream","name":"stderr","text":["Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n","Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n","Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n"]},{"output_type":"display_data","data":{"image/png":"iVBORw0KGgoAAAANSUhEUgAAAaoAAADhCAYAAACZZ/HtAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nO2deXwc1ZXvf6fVarVaLVnWYlmWbdnYxtjgjZ0ECGtYAgkhCSGZF5aQbSYLeUkm25uXR+Zln5dJZktmyAsvZGBYEkhYwxLAgMNmzG6Md8uyJMuWZVlqt1rtVt/3R5X7nnOtbi22WyX7fD8ffXSrTy23bp2qW/ecU+eSMQaKoiiKElRC410BRVEURSmEdlSKoihKoNGOSlEURQk02lEpiqIogUY7KkVRFCXQaEelKIqiBBrtqJSDhogMEc0twnH+RETXHu7jHC6IaD4RvUZEfUT0pQDU5xwi2jbe9VCU4dCOagJCRLcRUQcR9RLROiL61HjXqRgYYy4xxtw63vU4CL4O4CljTKUx5p+LffDD+UJBRNcR0YpDtK8tRHTBodiXs9+biOi2Q71f5fCjHdXE5IcAZhljqgC8H8D3iOikkWxIROHDWjOlEM0AVucTElFJEeuiKBMHY4z+TeA/APMBdAC4Ko/8OgB/AfAzALsAfA/AHABP+stdAG4HUM222QLgawDeALAHwF0Aokz+t/4x2wF8EoABMNeXTQLwWwA7AbQA+DsAoSHq0gNgE4B3+b+3AtgB4NoC57ocwKfGsi8A7wPwKoBeX36Ts+9r/PruAvA//Ta4wJeFAHwTwEZffjeAGl8WBXCb/3sPgJUAGoao+5MABgGkACQAHAvgNwB+CeBhAHsBXABggX+ePfA6tfezffwGwC8A/Mnfx18ATAXwcwC7AbwDYFmetnvGv057/W0/CuAcANsAfNVvrw4A17NtygD8HwBbAXQC+HcA5UPse4F/XoP+vnuG2x5AHYAH/fPsBvCs387/CSALoN/f19eHON6Q2/qyaQDugad/mwF8yf/9YgBpAPv8/b4+3veu/o38b9wroH9jvHDeAyvpP3xeARDPs951ADIAvgggDKAcwFwAF/oPknr/IfZzts0WAC/5N30NgDUAPufLLvYfOicAqADwX5Ad1W8B3AegEsAsAOsA3ODU5XoAJfA6za0A/s2vy3sB9BU4l+WQHdWI9wXvobzIfxgu9s/hCl+20H94nQkg4j9c98F2VDcCeAHAdH/f/wHgDl/2WQAPAIj59TgJQNVw9feXfwPvReDdfr0qAWwA8G2/Huf55zCfrd/lHyMKr/PbDK+T3d8GTxXQmdx1Ym2SAfD3AEoBXApPpyb78p8BuN/XgUr/PH9YQM9WOL/l3R6eVeDf/eOWAjgLADH9u6DAeQy5rd+GqwB8x2+/Y+C9wFzkb3cTgNvG+97Vv9H/jXsF9O8gLp73cDoT3qilNM861wHYOsx+rgDwKlveAuC/seWfAPh3v3wLgB8x2bH7H4B+fdIAFjL5ZwEsZ3VZz2SL/G0b2G+7ACzNU8/lkB3Vwezr5wB+5pe/A7/j8Zdj/nns76jWADifyRvhdWRheCPK5wAsHsH1ytXfX/4NgN+y5bMAbIc/OvB/uwP+6M9f/1dM9kUAa5w26Clw/KE6qn4AYfbbDgCn+w/+vQDmMNkZADYX0LMVbLng9vA6x/t4fRz9K9RRDbktgNPg6DqAbwH4f375JmhHNSH/1Ec1gTHGDBpjVsB70//rAqu28gUiaiCiO4mojYh64Zmu6pxttrNyEkDcL09z9tfCynXw3nBbHHkTW+5k5X7/PNzf4hgZI94XEZ1GRE8R0U4i2gPgc7DnLM7JGJOE18ntpxnAH4ioh4h64HVcgwAa4JmqHgVwJxG1E9FPiKh0hPUHZFtOA9BqjMmy34Zrv7G23X52GWMybHn/ta6H12GvYuf9iP/7SBhu+3+AN3p8jIg2EdE3R1HnfNs2A5i2/3j+Mb8N7zopExjtqI4MwvD8TvlwU+T/wP9tkfECMv4bvDfgkdABYAZbnsnKXfBGGs2OvG2E+z6c/Bc8M9QMY8wkeKaj/efcAa+zBwAQUTmAWrZtK4BLjDHV7C9qjGkzxuwzxnzXGLMQno/sMnimuJHCr007gBlExO/L8Wq/Lngd3/HsnCcZY/J1hK6OFdzeGNNnjPmqMeYYeAFBXyGi8/PsSx4o/7at8EZs/DpVGmMuHcl+leCiHdUEg4imENHVRBQnohIiugjAxwA8MYrdVMLzyewhoiZ4wREj5W4A1xHRQiKKAfhf+wXGmEFf/n0iqiSiZgBfgTdiG28qAXQbY1JEdCqAjzPZ7wFcTkTvIqIIPBMR77j/Hd45NQMAEdUT0Qf88rlEtMiP2OuF11HzEdFoeBHeiObrRFRKROcAuBzAnWPcn0snPL/NsPijul8B+BkRTQEAImry9S3fvqf77Tfs9kR0GRHNJSKC56cbhG23gvUssO1LAPqI6BtEVO7fHycQ0Slsv7OcFwFlAqAXbOJh4Jn5tsGL9Po/AL5sjLl/FPv4LoAT4d3kDwG4d8QHN+ZP8Pw7T8IzvzzprPJFeL6JTQBWwBvJ3DKKuh0u/gbA3xNRHzyf1N37BcaY1fDqfSe80VUCnq9mwF/ln+CNxh7zt38Bnj8E8KLufg+vk1oD4Gl45sBRY4xJw+uYLoE3IvkFgGuMMe+MZX9DcBOAW32z2FUjWP8b8K7xC76J+M/wokyH4kl4UYrbiahrBNvP85cTAJ4H8AtjzFO+7IcA/s6v59eGONaQ2/ovSpcBWAovyKQLwP+FF4kKAL/z/+8ioldGcP5KQNgfZaMoig8RxeGFPs8zxmwe7/ooytGOjqgUBQARXU5EMSKqgDdKfRNe9JmiKOOMdlSK4vEBeMEM7fBMS1cbNTcoSiBQ05+iKIoSaHREpSiKogQa7agURVGUQKMdlaIoihJotKNSFEVRAo12VIqiKEqg0Y5KURRFCTTaUSmKoiiBRjsqRVEUJdBoR6UoiqIEGu2oFEVRlECjHZWiKIoSaLSjUhRFUQKNdlSKoihKoNGOSlEURQk02lEpiqIogUY7KkVRFCXQaEelKIqiBBrtqBRFUZRAox2VoiiKEmi0o1IURVECjXZUiqIoSqDRjkpRFEUJNNpRKYqiKIFGOypFURQl0GhHpSiKogQa7agURVGUQKMdlaIoihJotKNSFEVRAo12VIqiKEqg0Y5KURRFCTTaUSmKoiiBRjsqRVEUJdBoR6UoiqIEGu2oFEVRlECjHZWiKIoSaLSjUhRFUQKNdlSKoihKoNGOSlEURQk02lEpiqIogUY7KkVRFCXQaEelKIqiBBrtqBRFUZRAox2VoiiKEmi0o1IURVECjXZUiqIoSqDRjkpRFEUJNNpRKYqiKIFGOypFURQl0GhHpSiKogQa7agURVGUQKMdlaIoihJotKNSFEVRAo12VIqiKEqg0Y5qGIhoPhG9RkR9RPSl8a6PMvEgot8Q0ffGux7K+EJEf0VEjx3E9tcR0YpDWaeJQni8KzAB+DqAp4wxS8e7IoqiTFyMMbcDuH286zER0RHV8DQDWD2UgIhKilwXRVGOQIhIBw0F0I6qAET0JIBzAfwrESWI6L+I6JdE9DAR7QVwLhEtIKLlRNRDRKuJ6P1s+1oieoCIeoloJRF972gduh9NENEyInrFNxffBSDKZJ8mog1E1E1E9xPRNCZ7LxGtJaI9RPQLInqaiD41LiehjBki+iYRbfSv/9tE9EH/d2G6IyJDRJ8novUA1rPfvkREm4ioi4j+gYiGfE4T0T8RUav/fFlFRGcx2U1EdDcR/davx2oiOpnJpxHRPUS0k4g2B92toR1VAYwx5wF4FsAXjDFxAGkAHwfwfQCVAF4E8ACAxwBMAfBFALcT0Xx/F/8GYC+AqQCu9f+UIxgiigD4I4D/BFAD4HcAPuTLzgPwQwBXAWgE0ALgTl9WB+D3AL4FoBbAWgDvKnL1lUPDRgBnAZgE4LsAbiOixjzrXgHgNAAL2W8fBHAygBMBfADAJ/NsuxLAUnh69l8AfkdEUSZ/Pzz9qgZwP4B/BQC/43sAwOsAmgCcD+DLRHTRqM6yiGhHNXruM8b8xRiThackcQA/MsakjTFPAngQwMd8s+CHAPwvY0zSGPM2gFvHr9pKkTgdQCmAnxtj9hljfg/vgQIAfwXgFmPMK8aYAXid0hlENAvApQBWG2PuNcZkAPwzgO1Fr71y0BhjfmeMaTfGZI0xd8EbLZ2aZ/UfGmO6jTH97Lcf+79tBfBzAB/Lc5zbjDG7jDEZY8xPAZQBmM9WWWGMedgYMwjvxWmJ//spAOqNMX/vP7c2AfgVgKvHfNKHGbWLjp5WVp4GoNXvtPbTAu8tpR5e+7bm2VY5MpkGoM0YY9hvLUz2yv4fjTEJItoFT1+mgemHMcYQ0bYi1Fc5xBDRNQC+AmCW/1McQB2AwSFWH+qZwH9rgacbQx3nawBu8OUGQJV/nP3wF50kgKjvC2sGMI2Iepi8BJ71KJDoiGr08AdQO4AZjg15JoA2ADsBZABMZ7IZh796yjjTAaCJiIj9NtP/3w7vIQEAIKIKeGa+Nn+76UxGkLqjTACIqBne6OQLAGqNMdUA3gJAeTYxQ/zGnxMz4emNe5yz4EUkXwVgsn+cPQWOw2kFsNkYU83+Ko0xl45g23FBO6qD40V4bypfJ6JSIjoHwOUA7vSH2/cCuImIYkR0HIBrxq+qSpF4Ht4Lypd8nbgS1uxzB4DriWgpEZUB+AGAF40xWwA8BGAREV3hv/V+Hp5vU5lYVMDrfHYCABFdD+CEUe7jb4loMhHNAHAjgLuGWKcSnp7tBBAmou/AG1GNhJcA9BHRN4ionIhKiOgEIjpllPUsGtpRHQTGmDS8jukSAF0AfgHgGmPMO/4qX4DnUN0Oz0Z8B4CBcaiqUiR8nbgSwHUAugF8FN4LC4wxfwbwPwHcA28ENQe+X8AY0wXgIwB+AmAXPOf6y1B9mVD4vuifwnth6QSwCMBfRrmb+wCsAvAavBeYXw+xzqMAHgGwDp55MIURuhb8l+jL4PnYN8N7dv1feM+qQELSlK4cTojoxwCmGmM0+k8piG9O3gbgr4wxT413fZTiQEQGwDxjzIbxrkuQ0BHVYYSIjiOixeRxKjzH5x/Gu15KMCGii4io2jcLfhuev+GFca6Woow7GvV3eKmEZ+6bBs8M8FN4w3pFGYoz4H0PEwHwNoArnLBlRTkqUdOfoiiKEmgOyvRHRBf7KV82ENE3D1WllCMX1RlltKjOKGMeUfmZF9YBuBCe03clgI/5US+KcgCqM8poUZ1RgIPzUZ0KYIOffgNEdCe8vFR5FagsHDYVZREAQIic79LYcklIDvRKS6yMQnI7vh9ythsctB+C79u3T8oyVhbx62Srwo4hkk4AJSGbML0kXJJXlt6XFrLkXhtl7L4chFi9S0tLhYwnvchmnbqUsOOlM0LGzy9U4n4DyI5P+duT17MvvQ+pzOBIPiYsxKh0xo+AAgCEIa/RlNL6XDkajQpZKmPbflf/biEbQDJXLoHcrqIkliv3DyaFLFZijx8KyevXuy+RKw8OmXwgH/z2k7pUGarMlSdPmuxsZy9DT1+fkPRnrJ5lIHXegJ+TvPXDTlvkwz0/YnUpC5XJumS7uowx9Tg4xqwzSjEZeVdC7DnJn7XZ7D6YbGbIZ8zBdFRNkHH72+AlV5SVIvoMgM8AQCxSiouOPw4AEInIB3I4bKtSVSFvmilV9iFRFpFVjkbtzeF2OL177E3cuWOHkHXvsg+wWbObhayU1WUwnRKyyVX2AVI9uVrI4vGKXLm9rUPIVq200aaptOzE4jH7gJzaJHNXZjP2+Mm98qEUr7SfPWxr3SVk3Tvt+cUnybbOsgdYOCxl0Yhtwwx74N+77pBk8xlWZ7i+cCajSSx/rv5zufL8+ccJ2foue4hb3/y9lOGlXLna+Q7ztEo75dibPauEbEnlzFw5FpcvBY9vs5lndqMHI4frj+yMTomdmyt/9PwPClmGddr3PfOMkL25w+pZN6QODuA1tlQrZLUsRZx8HZL0IiGWw+wRMjd6jJC9nry5BQfPmHVGGQuFZi5yX8L4unXIj7xfojGr6/y5v7dvc949HPaoP2PMzQBuBoCaipjZPypIJmUwE38rTodltUrCtgOIVcaFjHdGrWs3Cdm21rZcOVQiR1vV1fYhXx6rEDLeeFnHjTdzhu3U6hrkzc47h3RKXtRk0r7Ndu2UnUqP6BzkYyKdtg+GUEjuM1bB2iUq32YTrAOKROQ58A415LR1tNxeh8Qe+9ANFelDBq4vRCHj5dkE3o3TxXpNDfZFPdEj23PDmjdy5Xa4n6PY6zAL8kXjI6ecmSsvXSNljXOsvoRr5TWa3mhv0l+tvEPI+sBfdORDoJF1FmfMkvNynn/68bnywgXyBezl1etz5c07nheyDrSJJUm3qBmnk31XXOp0mhGU58oZpxurQkOuXB6V9ybkoPSwIXVGR1SHFn693cEOf+a4F5vrrHzG9CfzBLJm81sjDubx0waZk2q6/5ui5EN1RhktqjPKQXVUKwHMI6LZ/hw8V8Ob80RR8qE6o4wW1Rll7KY/Y0yGiL4AL+dUCbx5doacsl1RANUZZfSozijAQfqojDEPA3h4pOsPDg6it9fze6T6ZURSvNL6TdIpGcAQr7DV7EvL7V552fojWrfKbPjlUeunWXi8dLjPP/7YXLm6VvqauN8rndnrnIUdhGaljxA7dlhfScaRzZxpgwFS/TLPKA8sqamReSEzGeszSqV6kY8YC8gAgHCZPYdMRrZnOGLXrXfOPc6CRXpj1gZdusb1d4yN0elMBOTPirHeS0ad44FXn86Vs5Dn1wLrw98r/DKSLsiIwB52/U5asFDIItYVgwde/JOQbWm1frC5qBGy7Sy4Iu34dzKsnlVlMqAnHrX1fu1VGTDRutPqZDvceAXeFoWcRPucZRs4E3Lak9c7ikrkY1f37ryyg2G0zxnlUMJdfm6gRaEIV67rTjeTze+/yofm+lMURVECjXZUiqIoSqApalJaY4DMvqGjR7k5rHf3HiHjodFuOHVLuzUJRR3z17QZ1txWVd8gZPuydj9tO+XxssxuVxGWYd+plJVtbZHmsPXrrBlmsvONVWOj/T6qq0uaSFLM1Bl2zq9qUoytl//bs2RWfpuV6rdmn+Ruab5M99theSgrQ/ORtceLRW0IeCg0PvmLjW9CeBud4vdNzBQ4xQmnDoHrgRMyzd7N3C+eHl9jXR/xzfIdLsPMhGt63xCySvadyMevkJOkrl3zTq786tpXhOySMy7Klbu7ZMju00/a77guvPxqIdudtSbK7AFfPfFrnUJ+8n8vk4bUpShiedYEouwRUlsjr8P6/FZXJbAU+orOHddwF4arT/x54W6XyVPOf2wdUSmKoiiBRjsqRVEUJdBoR6UoiqIEmiI7Hgz22yGrJskwV+4X2tnVJTfrsP1p1WS5XcNU6/uZ1jRVyCIsL2DXLumHau/YydZzEp7W25Dt8li5kKWZj6rb8aX19NiQ8Gi59I2Ew9b+6h6P+6i2b5e+mGg0xMrycvEQ/64dMgcbr0syKX0VPV0sf2BCvqtMabDnV1Nj/Wz70oVs14eLDADPH2PcNCws9LrFSYUkbd17HJnVkYSTg2xVemuuHHJyPJ4y1YaPn7/gXCE7bZHVwbmNMk/l3o02d+qsJScK2fnnnJwrv90iw+/vevC5XPnW3z0kZDt7bN36Dzg/597JixtmbvXcwPXP8fB0SYjJdvUcmk8YlPHEjSHgaZPcTxo4bqh6odRL3AfKn4X5s1/piEpRFEUJNNpRKYqiKIGm+DHHfvb0qGteGLBDxbKkHGJmsjYjQywqw6lrKq0JIx6VJrVEwppFdrTIPJa7e+zxuPkQAKpZNvU+Z0qO7rg16XU6GTQSLAOE+3V/iM3V41QTqLY/tPdI01+aRRtPneRkkQjbY6T7pQkoHLHnV40qIcuwuauiTmb1mjpr3Kmtt+agcOnBTkU1Vvafh5v1gOtIwpFxA1XakXGVl+H+SXbN3IDsj157Va788c9fJoWbbdb+Z+64TYgaqu01i9fKvVY1Wt193znSLHj29Vfmyn+8Q0699PNb+DGcFCgF4SHErhJG8sr4DAJpx/yzHXb6l1BW33snPu59znVhACOH60Kh7bjJUE1/iqIoygRFOypFURQl0GhHpSiKogSaovqoyAAh3+WQ2iNnGE11Wz/UYNbJrJ6x/anMTQ2E2Yy4CSd7czJpU81UOSmAInGbpTxW5nokrN0/7TTR9AU2C3uNM2twb9r6iVKOjyqbZTPnZmRW66p6e/yLrnifPAfmfml7SU7VXFNibb/RsAwfj09m5+f4qLgHp75Zpr2ZMtv6VKpqrK8uukL6c4oBIYYoPN9N2Hmn6sM7bMkNi+Zn6F5bez0HHZt4H5v1NuGkc+lJs2smXZoA7BTs8y6W08Y3n3NWrlztRIRPOmWRXaiXU3nzhF8ntsrrF77l92zJDRjny27IcGme9YASsBmMHZ2PMT9FlbNdAva+df1XykRhpP5n9xlQ6HqH8pSH2274vSmKoihK4NCOSlEURQk0RTX9ZY1BKuWFC/f0SvPXbmZacad7q2TdaaRcmh5izJ4Sisp+N8tCr8PVcrtk0pp2ks4REyEuc7KLsyzsV3/iBiE68T0208Avf/oP8njMEpiukmHmM+fYrAcf/exnhWzSZGtn+sn13xayeNqaNsujcvge2m2H15PL5WSMMZaxo6pO2qOiUdtO0YgtExU/PL0yXI3Tq6/w6hKTqtqTtJkcOrvlpwchZgYNO+H3mYzdD8+EDwAJlnnczXURr2K/9DtCFsF7yxPLhWj5Qw9a2a9+JGSTHHOfYJc12z36yBNClGL62gA5K0BVbGauvDUpJxIdEPni5awAIXYSGcfsmWam8IR7rzBz6cABnwkoE4P8k2GKsUzImYkguw354Rl93G6Gp9Xnn0zk/9RCR1SKoihKoNGOSlEURQk02lEpiqIogaa4PqrBLPYmPB9Vd1raunewspv0ppnZzLMRWeU4m0k3Xif9Sakddt3ePdKxkIyyDOJZx+6esL6KqJPl/Wvf/7dcOVzfLGTvuWBprnzhtdcI2aN/fsnuMyFD1+edYtPnPLRczh6L5Gu54spVMpXOJafNzZWzqV4hC2XtMbL7pGzPgA3LzkZka8cbbCh0OMx9VOPwThMKIxzzPkgIl8nrPrfW+u5OnH+ykEX5LMmOhifTLB2Qo2jZrNWzWFzqxHHzFtsF16Sftv7ABfOmCNHsGz6cK884/ljkpU1mPW99c32uHHf8bH99zUdy5U99Ws4o3LRgTq789p/ljMI//d+/zZU3bJT3w6609Vn1ZOWnI/yOzDifXUTYu27IaZh+yM8plMMN9yEV8hfKzx3K49bP2Z/odNZlqe2i0q854AYTjIn8aZM4OqJSFEVRAo12VIqiKEqgKarpb9AYJPzwdHfUuPfA1XP0MdNKOitDGDMhPiGhPJ1E1po3khFnKMwitqMhJ3NDvD5Xrm5aLGTPv/ZUrvzjn9wiZGec++tcuWq2NPNs22NNf9MqZUjx4tPtRHwPL39cyFYufzFXXt8iMzAsnW1PoiwtTTmVLFQ/XCK/BN/bZ9uiPCu/Nq+qsuabWLkNMQ2Fiv9OEyJCtNwzN1RXydDYumpr5m2oldcvFrXnFHJMC6GIDbPPZmSW8FDYLodKpb7MqJfXTDDJHu/KT14pZVxf3Sbcac19e1vWyuOdtjBX/vrUeUKWYFn7a0+fnrdaCz96ili+fofNQvLsii1ClszYa+2aydNsEslUSt65yVQ/k0mz4P3rn4Myekoi1qUwmHZNcfwzG/ksLI3a59a+VCHTX6HHvptVnx0tM5pM/fmPcfq7bfaWr/7tF3Plb3z1c3m30RGVoiiKEmiG7aiI6BYi2kFEb7HfaojocSJa7/+fXGgfytGF6owyWlRnlEKMZET1GwAXO799E8ATxph5AJ7wlxVlP7+B6owyOn4D1RklD8P6qIwxzxDRLOfnDwA4xy/fCmA5gG8Mt68MDLr88NbsMOtyutnKvX3S9hrvszbztJuCgyXOXnzGHCHa0tFq65WVL2pTG2y4+JTmpUJ2440sFDrhxjdbO3+P43Q79+IrbF2aZgvZsXNm5cpvrZFpSfYusP6ltU+uELLly5/PlU+fL30os5ut7yLtZtFmmeRjcZlaituhe/fYkxgcHPkVO1Q6Ew6HUVfnpZuaUi/z5k+O21DZ6knyHOIsDVQm3S1kmay1+YfL5HahkFWY6hrpu2ucUyCFFL+Lwm6G6QJZ5yusvlQUSKdU5owjytZszb9Pxjv3SV3a2mH1taFprpAl+q2sMi59moNZe+6umyKRsBn8e3b1SOF6jJhD+ZyZ6HB/8OABmca5z3U0T1GOnA28XzxS3egBrtxOOjkRDu/WxW7XfIzUtbvvuTVXntFgfeI/+t/5UzmN1UfVYIzZ79nfDqCAp1lRAKjOKKNHdUYBcAiCKYwxBgW+2iKizxDRy0T08sg+7VKOdArpDNeXVGbPUKsoRyEj1ZkiV0spEmMNT+8kokZjTAcRNUImlhAYY24GcDMAhIjM/lHmaA7Mc0BvT8iHVzhlzTVuFoljFkzNlecvklkkNnXYjNvnnnOJkL3rXVfZfcblS9zcWmuuGXSeoyUsYfBli2VoMGSUe16uvOw8sdwyz4a5L7/zLiFb1W6//J+6WWbKXjR/Vq6cdSNOQzybgAwp7t7RwVazprHBfQc9Kd6IdIbry5Sq+SbiZ2UIh+U7VTTKTsoxS25rtdd2INMqZOdeYEN4T7/gDCHr77Smzq5OmZEd0nI8viyYmV9mLXHY3i4VNM50ORorlzL2KUciIc3ryT5rFnQi0JHJWLNyPO5k1z54Rq0zRBTQ9+ESZ9nac0uj8rm1bJl9WLS11QtZe/v2XNlkZHPsSxX6yIfjjk+4C0M+mcvjNgPMrNnyGZpO2+X0gHSDRMrs/fngn34nZNzc507tmY+xjqjuB3CtX74WwH1j3I9y9KA6o4wW1RkFwMjC0+8A8DyA+US0jYhuAPAjABcS0XoAF/jLilv+gCQAACAASURBVAJAdUYZPaozSiFGEvX3sTyi8w9xXZQjBNUZZbSoziiFKGoKJWIHHI3NkRudN/bKENiaSXa23Gmz5bys9U025DfpZBSZXn9CrnzJOR8WshmNs0ZUr5JJw6+Tg4f1Oq3e2WFtyxtfkhmve7Zbn1EsLDfkrqdUSIZPR5lPJ+qEbyfTNjt2Ji0zZUdh2zPCwthD42D6Ly8LY8Ec7xrGKmRobCZjbeJtHTLNTDpl/WmXf/gsITv1ejm7sjjePHuMGSiQ6TzADLxq/VKZPqkTVVF7flknVU6GKWU0LLNkJ8LWf5VOSX9gmIczZ8YaLn2k4H7CwJ9yrjfGXqd9KZk5/6UXrf66GctNhj//3Kcob3/3swi+H9dpzZZD8hnTn7TX/p0164Qsyurmple68b//Ta48uVo+KJ95YXWuPLXR+k1TaZ3hV1EURZmgaEelKIqiBJqimv5KANTkkblTteUj6YwO5x6/IFc+ftkJQlZda0NAV732jpBVxZty5ZqaJuTj5ZdkCOh999rAo0xShv82sIzeYSfb+MaN1oS3I7lbyLa12jDz9StfFbKpEXvC5WG53Wl11qR37GwZRp9ml7YqJlMbhMN2n8nkgJD17LJD/XjcmgSy2eKb/sLhEkxpmOTXRYbw9vTYts86X+/X1tl1mxodUx9XNFf7yzHhWbvG6tnWzdIkGq+0J+ya/vikkW6m/EzaytL98gYMsezwkdDRbvpz7xHejm7b8HD1/Io3kJL3Z2W1fVbFYjEh49cwm5XPtBj7HKG3Vz5t93TvypXJufbVzGzn6kUkYnWouVlm8f+HH3wtV375jU1C9pcVL+TKCxZaE3uq3/n2gaEjKkVRFCXQaEelKIqiBBrtqBRFUZRAU1QfVSmFMMXPWF0ZkbOyhnp7c+WWAzL4WmbNnyqWZzQfkytPaZRZeuum2PQfG1rkPhfOPzVX7uqWsopGG3aZhfSNPPec9XV1tL4tZHNmWHvurOZGIevqsuHU61ukzTbdZ8/9pJOWCVm0d0OuXNW/S8iYq+mAgNPerL20NWHpv0qlrd07FpdhrMmk9fdkWEz94GDxfVQGWaQz3ncFCTdTPfOHzJ8v7ePclv78cxuELPHYzlw5GpE28as/bmdaHmnKq3HHce7uaLPXL+vkzmrvtNnUY07Yc4Slywo5769p5qPKOOHp6WSayZxrNGFxU0EVmi23EIXSjnHdc8YLWR6CLj8tYY8KpNOyvbnPKh6Xn3NcdLH9HI37dwHgzTftc2xSlXze8X1yHxgg77Pb7vgV8rF9u/SV8lD2LZvtTAADA/n1R0dUiqIoSqDRjkpRFEUJNEU1/YXDIdTWeMPqmskyK3A4bUOoI3tkGHY0bvvT911xkZAtXGQnOayqmSX3WWqH8Ndf90Uhm1yWP0MB59RTZejopz//6Vz50Qd/L2RdndbM9PbGzUK2p9cOmyucyfU+9MEP5MoXXXqukD161z/lyun10s6TaLGTAmZS0jSXYGbBdS3SZJhK2ePPbJZhrL1haxbY0WWH7JlRTJx4SPFNfOlMv/iZmyGijhkrHGHZ4R3zV1XMTlrZ29MhZPfe/Zdc+cryd8t6zBtFnYtI/yppXkr02nPPZuTtvafXXs++XrldY401qXMzIACE2aXPuOafrDXXhA4+w35AGKupb6wMOMs8i4WTEZ01/4DjIRlI2mdaT48MXf/DvQ/myjEnc37VpComk9tx3OwTH/qIfW411buTKlrOd2Yp4Pfus8/YyV/3FdAfHVEpiqIogUY7KkVRFCXQaEelKIqiBJriplAqKUH1ZC/8MRaVvoOpNTa0+7wPXyhkzfOtP6s3Jf001XU2DDxUWidk8+Yfnyvnt6COjsWnHpcrT2/+jJD1dNjZZDdt3Chku3qt360pKu3Acxvt+TVOk6HkS8629t0/rHpcyCIp+54RgxNuzMNMs/J4Ydb2Wzukn6a61q67cIkN9y9f+TqKzeBgFolez34fizl+qDBLQeNG93K/iZPWJ15tM+xX18uZcjMDVreWP7ZayOa32vZsPNuZYbeodxGw87ktufLzKxz/I+x9lHHS9kTYJyHpVK+QbW6xM8dGymRbV8asr3fA8VMk+qw/J5EY6QyziqTQPLeFPguJOsv2WpiMdGDt2mFTwe064Hj8BnKflFa2aNnJQvKVL36iQN0s5c79+f6Lrf/3vj88nCu7PrB8NVQURVGUwKEdlaIoihJoipuZIlKKaX7Ghq0bpclp4RKb+fxbf/8tuWHYhs6+tf4RIYrEbEhmRWyakB0Kc98G+VE1wsxiGYlIE0k6Y4fiexJyyD6JmZxiTqT34w9ak970+dKsdOn1n8+VH73/j0K2Ze2KXPmU844TslkL5+fKoZCcuOyNl96w9eyWnwLUNNgTrKu3Jp9weHzeabK+lSKdluaKRJ81bUQc2wKvazwuzSOxStv41dUyq3wkYnVp+44tQtbWaTMJNMpIeTjJSw47e1g2iFDUyXYdtdcsFpcZFuIs6rpnl9xuN+ynDj27pU60M5N2VZXMKBPmk3mGJ05mig9f/QWx/Nef/1Su7GZ14LhZHTastyb+559bKWSrVr2WK3ftlCbanh6bfaK6Wk74yicTdCcd5NTUSv2tq7P6y0POAaCmxq5bW+vqvb3no1F5v/DsE1d/6IK8dRkrn7j26lx5xdN/zLuejqgURVGUQKMdlaIoihJotKNSFEVRAk2RfVRhTPHDr9NxaZf9wvd+wGolM6QLYseKxdIqa1Ntqpzsrj0m9jIfRK0TAcobLJSSPoB2NhvvopNlPdFvbdLP/vF+IVq3/q1cORPKf0lOuuwKsfzQ/U/kyufH5HZTmG27IiJt7tkFNm1SJikdLCGW8bm3xaaBGky7aV4OP0SEcKl3XnEn7Qvi1j8Ydc6du00cNyK4Cb6yUm7HorCRhUzxlUrbrOsHZNg53D4q1yfGMuPPbHZSgWXtCSecNEnVGbtuaopszxRLv5NKytDmHZ323JNJtzKW+CQ363iweP+V9nOS393xL4dkn+8956Rc+W8+fdUh2eeRwAuvrMuV3c8WuE/s0UfsM6y3V34ywdERlaIoihJotKNSFEVRAk1xTX+hEJoqvGHfu885W8gaGqcNtckBvPTki2J58ZlLcuXZlSPbBwD077QmjEhUmkFCLAQ96rRQhs13luyTIafVk+zK7W2tQrb5TWvey2al7SidtTt9buWrQrbwdGtCXHrmmUKWZea9th4ZUpxN233uczIUzJltw19DaWkW7Gyzk+v1sInwqOAX8oeHEIUQ9a9N1MnqHK+0Jq6IM2tkNBZlsowjs20Rccy6fD+NDbJdVq+1n1O84UzGuPhMNmGnTCwydnbZcPwNK2WWkwzLij651sk6EuIy2TBRbgJ2PpHIpOz9sNMJpY5X2fdZNzw7k7Xtm9orJ6Icb6755NfF8q2//vE41WTicv8jdkaBiy6WMwp077I688D9fxKy2//z7lx5Z5fUJx4C395m76tdXd3Ih46oFEVRlEAzbEdFRDOI6CkiepuIVhPRjf7vNUT0OBGt9/8fmkgGZcKjOqOMBtUXZThGMqLKAPiqMWYhgNMBfJ6IFgL4JoAnjDHzADzhLysKoDqjjA7VF6Ugw/qojDEdADr8ch8RrQHQBOADAM7xV7sVwHIA3yi0r1IMYtp+/8xmmZ267xk7A2Vlw1wh29Nv7eI9q18QsvqzFw93CgCAASe6ev36Tblye2ebkCVD1v+RTsnUPdtbre/p9VdfEbKtLXafu7t2CNk1V9rZMC/88PuE7FYW8nvXHfcI2ae++NFcefH8JUJ20XttlvmqiExf8/Zq60cJhaTvYOoUG4pcXSn9GOmsDU2urrb+j5KSkVuJD5XOGDLIhHwfiKOpLEIbmbDjP2Mz/Iad88uEbVukstJ/lUnYmY9jbmLqjN3nto1SX5prbWUmdTgZoCfznFtOG6ato+ilP78kRG+92Z4rn3zWZUJW22A/L3BTW0UibPZoxw/FJ+d1M2KFmcOuPizT9kRjduUd22V7blpr/WfphJzZYKQcymdMuDSG2voF3gYFfFKux5XfPV07ZTi162PhZNnM1+m0vAe5Py/phPzz1EQumX1Wh0IF7jt3HxHXWcuIs3Ra7ozYfD/3/fEhIevttdd05UvyedfN0q/17pHXnu8zlZLPn+Relv6M6SsR5a3/qHxURDQLwDIALwJo8BUMALbj0LmRlSMI1RllNKi+KEMx4o6KiOIA7gHwZWOM+DLLGGOQZ+IUIvoMEb1MRC/v6Z84SSuVg2csOsP1pS+52xUrRzCH4hmTzeaf00iZuIwoPJ2ISuEp0O3GmHv9nzuJqNEY00FEjQB2DLWtMeZmADcDwDFVUdP1ztsAgI0rnhbrbXrdmqqyEWl6CLNEwNOcr9/bXns5V25uPl7IuBXmC1dLi8Hal+0X0VnHNJZgIb7btu0UMpNpx0j43JUfE8sXnLk0V37DGUK//ZrNZr5865tC9shyu3zxOYuE7PLLLs+Vf/mdvxOy57pt2OfSZccI2bKlNtP6lrWyM+Dh+DOb7AvsaIPTx6ozXF/mTl9sqmq86x2bJD8hiLFsFLG4NHnEKqxpo8w14YXYD1lpauA3g2PFQc9uqwcL5jcL2aRma4rDenltsYPpVrUMJR9MMgXdI7NIdDLzYvQCuR1P4Z/OSPteOmtN1eFQiZBxy1DWNf2FbVtEw/IeC0dsy2Sz8nhRPp2AY2rEbRgxh+oZQ0Sms927Bv/jpn8W682bZ++DqkkynUiiz5r73HPkWcpjzmcSPHu8u93F552SK/c5roe2dnsqvXtkRgaeyUGYcgGkUnZHbkb2VS+/hnwsO9G6SNIDUtcWL7UzV6xbKz+92N5hp48QmfIhTXquGZK3RdO0RiHLZ/bcvWvtkL8DI4v6IwC/BrDGGPOPTHQ/gGv98rUA7htuX8rRgeqMMhpUX5ThGMmI6t0APgHgTSLa32V/G8CPANxNRDcAaAGgia6U/ajOKKNB9UUpyEii/lYAyBeOcf6hrY5yJKA6o4wG1RdlOIqaQikUDiFW69m/ezbKtDDJNda+uuyMs4SspoHZzB3nwTtPPZorVx+7UO4zOiNXvvePP3Nqsw8Hy0nz3iWWr7nShp2fd/bJQvbGc9Yn9+arbwjZ2WefmivXdJ8mZMlk/rQ0vb085NX1R1g/zbrVMp3T5Eobnn7eWacKWUOttd0z1wTKIitQbErCIVTXer6pqio3hZL1jZTHHF+TSHHv7DRk1w05fvcoc7esfn2TFMK29ezFTVLEZ7Z1fEYiDtwJHy5JW9kpS+VnFtu2Wb/Fjp1yNuzmGuvDHXDuh3TG6nXI8VFx10ClM/tvNMZ9IU57Mr9eU7M893Dz9Fw5lTr4e+rQ4HlUf/DdG8WvZbE5ufIJi+SzoqFhSq7shplzn0q1M+MuD0F3fTj/9i/2HrzwvecK2bQm67d58423hWzTxi25cqRM+qjOOss+c1avfkfIeNi56y976IHHcuUlzCcFAAvn2NkqZs2W/tdMxt4kPFQdkLMIh8NS19JpqwvpAdmekTJ7H8TYrAhu+3E0hZKiKIoSaLSjUhRFUQJNUU1/WQJ6/a5xyZkyy0I6Y4etq9fLIe2cEAtv7JHDz87ddlhZCylbONsOaRctOUXI3nz9uVx5wcxlsp4VdtK8pukzhOyCs61prskJVT35OBv+2lAlzTyPbFyfK8cj0rTyuc9/2p7D6dL0x40pboj4We+22dTDO2W2hAdu+62VOdt1dtgQ1+3tMpN7sse2ZyZj1+sfB7MOkbWWxSrkO1U4bE0bromLW9jC+T/WR1RaVcASiKO9Y4uQnXUmmwhztjweOpkNcY4MxUWcz+LovBcmWXb6mJyo8col9lML48wn19NnNSEUkiaeEpbtYp9jhsyy0PWsE0ueYotu9HC0gplLZaIWEZIeCTkNGjAGktbdsOrFPY60a4R7qRJLZey6HbdgvpA1MfNeJiMbjmd5eGfNOiH7/Z0283h5XJoaa2pqcuWtLdKkzzORJxLyvj5hkdWnj338w8jHuedJt8uSJdZMeNbZZwhZVZV1E2xk5koA2LKlxZY3bxWyNlZPfu7ZbP6PYHREpSiKogQa7agURVGUQKMdlaIoihJoiuujApDyjdrTG2R+yTgLU+zs2C5kr7P0Ss1Ns4RsS48N397lpIbjR/jQ+2XG8gV11bny9//xp0JWUmvDWCNhaXefytxS3/vyV4Vs87M28/D5p8gM8HUhlgE+IqfVWb3WpkKJpKR/jmf8WTZf+stefdFmkn/fpRcJWU/rllw5XiHTD6X7bF3aN8sw7OdbWeqeGlvPxN5+FBsKWR/TAb4Y7vqJSJ9RtILL5D5ZdiyUOjIefTtzjtTPxjOmIy8xlo4nWukImW/PvdtYVnI44b1gCa6pVoqqN1ufUaxc+kmzWaswiT2O/zFpfQCZtMzmzX1+opEApPfZ5YyTdoq/6WbHYRZol0hZHE0zvXRlm9fn/6Si1LlOF118Ra68eIlMVXba6SflysuWyc8I6urtfspH8dq/e69tKzcc/hvf+u+5Mk/7BABOJrG8uFcif15yyTFzZotlnhX9X//lV0J2/PE2FRsPtwekL42XAeDkU07Mlbnfa+3b+a+XjqgURVGUQKMdlaIoihJoimr6i5aV4dh5XphvT1enkLV32JDF0xZLs9k9D9tQzpnvPVHI7nvlT7lyT488XkertQWef7b8MvxDF7wnV567+DiMhUhG2hofu/+OXDmzeaqQzZtvh/AtrfL94Pn/uDdXrp4hw1Gv/cQldqFBpgJ/6kF77tglsxccu8Ce06YtMvx1aqPNSrCtW5r+MizLwgevtZM23rFhG4pNKBRCvMI754q4nOwtypqCm/pc2QGvYkzjjSPbnbAm5+Z58vpxUxycTNjgVrSENLp0tllTamen1PnGRmvTq18idV4cz8mgkei016KzU2bQzjArUkWFNG9FwrZh+hLS9Lc3aRujwslagYht4EjYyeTOMn0EYYKN7OBgLhN643T5Scr7LntvrnzjV/5GyE6YNy3vPnlguXuOPbutMvzlTZlhgpv03IzhfCJFV5bqt/vkmdQBYEpDPVtPZq3h2cx5qDoArGVZ0d94/S15Diy7xsyZ0sT9t9/8cq58PntmAkBvr/1uYnaTdGeMhe99J55XpiMqRVEUJdBoR6UoiqIEGu2oFEVRlEBTVB9VaiCDDeu9mVI3OjNJpnqtv+c9Z0l757yZ1paf3LlZyGLMT5TqWCNkjZeeZ8szZPqPQvBA7JWPvShkyRbr72meIUMyj5lnQzsnz5BZpqPNdobf7tYWIdvSame2rO6WVvDnaqzdeV7dx4Vs0Xx7vKf+LGdMXjBvVq7cskWmWsFM6webOf9YIerOWJ9DV2d3rpzZ5+bOOfwQhRDxHU4FEisj42R3SvOob3eG3wJMYilh4k7KJnBXgTP7r7iLqmQgcE3G+hQ6d0t/UpI7lNzZcTlt0im2ebXVwe2dMvVPNGLt/EsWLxWySXU25L4uK0+imzna3ATwMeb0C5XI8+NtHxpxEPTho66+Dp/8zHUAZPg0AMxkmd7feE36aW7/T5u2iM9qC8hURW4oOc9SHq90ZkZmStu9q1vI+HYR9xsKxt6k/CyE+7aiUanck1lmd74eIP1g55x3tpDVshmM9+2Tz59nltuQcTfj/GyWaX1di9TDY5vrcCjREZWiKIoSaLSjUhRFUQJNUU1/vb178SfflPZ0lzRnLGDlGdNlaGV/0pr3tq2XIaAXnmmHo4lWKet44alcufF0GZ5eiO4We7y3lj8u97neTnpY6WTDnj7bmuLqmuUX3lv32HVXb94iZM1N1jy0tFlm0c52WbPDi3+WdenptO20o11mT08lbOjoVDYpHABUTbam1FC1VIEpzXY4v2VLe648kB7fSfGceeDA5nM7wCworoqzXRm3sjhmwcYZNgzbuHHIfKdu8gl+fOd4pTG7z8U18tMKhJmpTH6VIO3PzgyPc06yOl+5XZoTB/htVe2kMai02S9KI1LWEMuf8mCQWZEcy5c43VAh82WRqKyM4z3nvBuAnPQPAJ568tlc+bFHnhCyV9lkpnudkHBkeYYPqTTE0vMvdbJWcNNjTY10Z6xba2dTmNksM87wcHHX1Fhon0lmJpzuuB54BohFjgkvxmaBiDpTCmQy9nMLHo4OAB3brYl0crWrwIcWHVEpiqIogUY7KkVRFCXQaEelKIqiBJqi+qjKolHMPd4Lh1719OtCxifE3ZOS/isevunOTJrJWLtsCNK2vOrJh3Plbfc+LGQfueGGXLl2vgxjbZpij1edlTOBhtmMres3y7DvFE+Z0ibDNbtSdnles7TnnrHM2qhPqJWpgrpZipztm2UqpIxIwyJESKVtOPmMaTItSg+zwW/d3S5kndtsGG00a2czHUi5MdnFwGC/F8SdxZfP3Bt23CssQhslbng6X3Z9KuxuoAaMDffVj6d3qhhb+PagE75cMcemd5o9e6q7eo693fKTglfYbKpxZ3bqRpaaxw2X5m6SUEieQ5S1Wbr4XzAcQCKRwLPPeLN3Nzu+n/p665udO2+OkHEf1TnnyllueSh5rEK2W3KvvQf7HN8W90MNOk7WdezznA4nHH7BQjtT8M4u6YPk18b1iZUxWbRcKj6fZXed82lQPG6VtK5epurnYe08kzoAPL38L7nytm3SR37tdfZTmqUnynryfTZOtf7zwQI+Th1RKYqiKIFGOypFURQl0BTV9BcpK8WM2Z5N5YOpmUJWN9nabybF5bA1zYaEfSmZMZgPW6c0ykwRW9bYoXeqQw7Ln7jHfol+1WevE7K2lXZCwuqszJCeSNqheCwim297tx0ah6PyeGcstRkgsqEqIQtnbdr3KkjTQnWjXXddh5wIr3X3zlzZDXHd1mHr2dG+U8h6M1aWisov2FN7rf0m0WqzgAyk3JThxcBgf77qaEyGzYaYhTTkJPROM8tVufvRP2vCgR5p1i2Lsx0NyOOBWx7dSP0RTmY3Ktg+SzJOXbgV1n3VZMsxZzLGeMw2RiYt76OOVmvGzmbkTidXW3NQLCp1N8UiwLOh8c9MsaNzJ/7ln/4DAHD5+y8RsiVLT8iV+WSIABBmbeVOAsjNZmHnW4hwqd3OzRTB1+XZLQAZSu5mOheh81l5725eb+/DqknyWlxy6YW58jFzZgkZD1d3Q8m5Kc7N5N65wx6fmzkB4NzzrImUTyAJyM9JXLdEmC2nmf5QAfXREZWiKIoSaIbtqIgoSkQvEdHrRLSaiL7r/z6biF4kog1EdBcR5U9YpRxVqM4oo0H1RRmOkYyoBgCcZ4xZAmApgIuJ6HQAPwbwM2PMXAC7AdxQYB/K0YXqjDIaVF+UggzrozLGGFjLfqn/ZwCcB2B/DOKtAG4C8MtC++pPpbBmrZcpvKttq5BVx23qnqoqmUaofbv1E4WcGUYjURuLnMnIEOpIxBpK21iGcgAIxa1NdcMKmU7luYcfsnUplX350iU2/Uhdl/RfVU62damrkfHNyTYbGpzISN9IlL8nzpG+u9pqu5/MZhnGytx66Nwr67Jpo80kH83KVDKnnWlt9c++8oKQTZ1sj3fRB6zN+/nbf4eRcqh0JpvNIrnX8/UlU9L+HyvLb9AWUa7uqxhTn7KstPHz2Woh3YFyalf3vf5w+Kg4bsom7i4sMK0uOad37KJZufI+6W4Q+yl1nwp8VmTnKwU26QGSibH5MQ/lM2ZaUyO+9s3/AQCoq6s9QLafaFR+BnLiSTbTfGoU/lgert/bK8O3efqjePxCIeMZ2rdvl/f1zp3Wh5x1wtrfYrMIb94sZ2G4g92jbgolPjvvqaefLGQzptt1ayulX7OxVj6PRswInUpc1wq5OEe0OyIqIaLXAOwA8DiAjQB6jMllRNsGoCnf9srRh+qMMhpUX5RCjKijMsYMGmOWApgO4FQAxw2zSQ4i+gwRvUxEL/ePw5xGyvgwVp3h+tLTu2v4DZQjgkP1jEkk+obfQJlwjCo83RjTQ0RPATgDQDURhf03nukA2vJsczOAmwGgsoTMW697Jr8pzpEnl1s7RXWlzAocZWaJnqy0u0ytt6aqcEbaMyLs6/sdSRkunlprv0Tv3S6zCR/DJgQLhaTJqarW2nkyrTJTxMIZ1kYTzsjw366IPYkex+ySDtsQ+5QTa72HxW+mncnQIsyEWOWkrj7zXcfnyjOqpQmkjrX9KU0yrH0Wy54+d441wZZHx/Ylw2h1huvLcbMXm0SvZxGKVjjh6REbYutYgwtmTxdmunLH1sAtPvLyyYwWFRhfSvKUXQpcslKnzUQ7ueZEHkLs7DPOGjuVlea0sXCwz5hj5hxr9pv8YhXSJtvdbbOuuGHmblg2h2dhd9fj+3GztWfYJIRuFvQaNlkhLwMyzN2dAPGMd51it6upybvPVL9UYJ6R3c0wkZnK611IocaPkUT91RNRtV8uB3AhgDUAngLwYX+1awHcd7gqqUwsVGeU0aD6ogzHSF6TGwHcSkQl8Dq2u40xDxLR2wDuJKLvAXgVwK8PYz2ViYXqjDIaVF+Ugowk6u8NAMuG+H0TPFuyoghUZ5TRoPqiDAd5kaFFOhjRTgAtAOoAdA2z+tFG0Nuk2RhTP/xqhw7Vl2EJeruozgSLoLdJXn0pakeVOyjRy8aYk4df8+hB2yQ/2jZDo+2SH22bA5nIbaK5/hRFUZRAox2VoiiKEmjGq6O6eZyOG2S0TfKjbTM02i750bY5kAnbJuPio1IURVGUkaKmP0VRFCXQFLWjIqKLiWitP7/MN4t57KBARDOI6Ckietufe+dG//caInqciNb7/ycPt6+jAdUZ1ZnRoPricaTpTNFMf/5X5+vgpUfZBmAlgI8ZY94uuOERBhE1Amg0xrxCRJUAVgG4AsB1ALqNMT/yb7DJxphvjGNVxx3VGQ/VmZGh+mI50nSmYgTr3gAAAXhJREFUmCOqUwFsMMZsMsakAdwJ4ANFPH4gMMZ0GGNe8ct98HKaNcFri1v91W6Fp1RHO6ozUJ0ZBaovPkeazhSzo2oC0MqWj/r5ZYhoFrzUMS8CaDDGdPii7QAa8mx2NKE646A6UxDVlyE4EnRGgynGCSKKA7gHwJeNMb1c5s94quGYikB1RhktR4rOFLOjagPAJz/KO7/MkQ4RlcJTntuNMff6P3f6duX99uUd41W/AKE646M6MyJUXxhHks4Us6NaCWAeEc0mogiAqwHcX8TjBwIiInjTFawxxvwjE90Pb84dQOfe2Y/qDFRnRoHqi8+RpjPFzp5+KYCfw5tG8hZjzPeLdvCAQERnAngWwJuw86p+G579+G4AM+Flf77KGNM95E6OIlRnVGdGg+qLx5GmM5qZQlEURQk0GkyhKIqiBBrtqBRFUZRAox2VoiiKEmi0o1IURVECjXZUiqIoSqDRjkpRFEUJNNpRKYqiKIFGOypFURQl0Px/UyVPIY6/E7IAAAAASUVORK5CYII=\n","text/plain":["<Figure size 432x288 with 3 Axes>"]},"metadata":{"needs_background":"light"}}]},{"cell_type":"markdown","metadata":{"id":"dinIwj7QuUqZ"},"source":["## 4. Split the original training dataset between training and validation"]},{"cell_type":"code","metadata":{"id":"VPjwBd0VHyJU"},"source":["# Split rate, 20% of training data will be used as validation. That is to make validation size equal to testing size. (0.2*50,000 = 10,000)\n","train_valid_split = 0.2\n","\n","\n","# Splitting the data\n","train_set, valid_set = torch.utils.data.random_split(train, [int(round((1-train_valid_split)*len(train))), int(round(train_valid_split*len(train)))])"],"execution_count":null,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"VD7FGA8UgcGe"},"source":["## 5. Preparing dataloaders for training, validation and test sets"]},{"cell_type":"markdown","metadata":{"id":"dUDTdOb-twLN"},"source":["We use batch size 32. Too low batch may cause more volatile learning with higher variance in classification accuracy, but too high batch may slow down the learning process. So I use a batch size that's equal to 32. I tried with 16 too, but the it performed better with 32. I didn't experiemnt much with batch size. I wanted to speed things up so I choose 32 and didn't experiment with higher batch sizes."]},{"cell_type":"code","metadata":{"id":"ub0B4tejgZ0P"},"source":["batch_size = 32\n","\n","# Lodaing data with batches\n","# We set pin_memory=True to allocate samples in page-locked memory to speed up\n","train_loader = torch.utils.data.DataLoader(train_set, batch_size=batch_size, shuffle=True, num_workers=2, pin_memory=True, worker_init_fn=seed_worker, generator=g)\n","valid_loader = torch.utils.data.DataLoader(valid_set, batch_size=batch_size, shuffle=False, num_workers=2, pin_memory=True, worker_init_fn=seed_worker, generator=g)\n","test_loader = torch.utils.data.DataLoader(test, batch_size=batch_size, shuffle=False, num_workers=2, pin_memory=True)"],"execution_count":null,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"gw0E69dduePT"},"source":["## 6. Building the first CNN model class"]},{"cell_type":"markdown","metadata":{"id":"PTiC996jsi-u"},"source":["Here I almost copied what was shown in the lectures. I added one more convolution layer because convolution layers are used to build features out of features. In theory it works like this: first convolution layer decets features like edges. Second detects features like circles by using features detected in the first convolution layer. Thrid layer detects features like faces, legs etc. also more high-level features are detected.\n","\n","The pictures are small and difficult to classify. So I thought one more convolution layer would make the model predict better and it did in my experience.\n","\n","I didn't made much tuning because the training process is slow and it takes time. But I still tried a couple of things.\n","\n","Dropout rate prevents over-learning. I tried dropout of 0.2, 0.3 and 0.6. I got better results when drop-out rate is 0.6.\n","\n","I have experimented with different learning rates and found 0.02 to be good enough. I set the Momentum rate to 0.9 because this is what's usually set (from lecture notes). I experimented with Momentum rate 0, 0.4, 0.5, and 0.6 too. But I got best results when I set it to 0.5 or 0.6. So I set it to 0.5. I didn't try other values.\n","\n","I started with small epoch like 20. The dataset is big, it takes a lot of time to run the code. And in the plots it seems after 20 iterations the mean CEL on validation doesn't change much. After tuning other parameters like dropout and Momentum I saw that CEL is still moving down after epoch 20. Thus I increased the epoch to 30 and 40. I observed it works better with 30 (at least not worst than 40, so no need to iterate too many times).\n","\n"]},{"cell_type":"code","metadata":{"id":"OPKhm73jSJmw","colab":{"base_uri":"https://localhost:8080/","height":241},"executionInfo":{"status":"error","timestamp":1636310857500,"user_tz":-60,"elapsed":278,"user":{"displayName":"Abdullah Karagoz","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"06337825384430712267"}},"outputId":"36a9c257-6a0f-45d8-e0aa-8c488ae633b9"},"source":["# Class for the first CNN model\n","class CNN(nn.Module):\n","  def __init__(self):\n","    super(CNN, self).__init__()\n","    self.conv1 = nn.Conv2d(3, 16, 3, stride=1, padding=1)\n","    self.conv2 = nn.Conv2d(16, 32, 3, stride=1, padding=1)\n","    self.conv3 = nn.Conv2d(32, 64, 3, stride=1, padding=1)\n","    self.pool = nn.MaxPool2d(2,2)\n","    self.dropout = nn.Dropout(0.6)\n","    self.fc1 = nn.Linear(64*4*4, 256)\n","    self.output = nn.Linear(256, 10)\n","  \n","  def forward(self, x):\n","    x = self.pool(F.relu(self.conv1(x)))\n","    x = self.pool(F.relu(self.conv2(x)))\n","    x = self.pool(F.relu(self.conv3(x)))\n","    x = x.view(x.size(0), -1)\n","    x = self.dropout(x)\n","    x = self.fc1(x)\n","    return self.output(x)\n"],"execution_count":null,"outputs":[{"output_type":"error","ename":"NameError","evalue":"ignored","traceback":["\u001b[0;31m---------------------------------------------------------------------------\u001b[0m","\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)","\u001b[0;32m<ipython-input-1-7c3f086a5394>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      1\u001b[0m \u001b[0;31m# Class for the first CNN model\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 2\u001b[0;31m \u001b[0;32mclass\u001b[0m \u001b[0mCNN\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnn\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mModule\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      3\u001b[0m   \u001b[0;32mdef\u001b[0m \u001b[0m__init__\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      4\u001b[0m     \u001b[0msuper\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mCNN\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m__init__\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      5\u001b[0m     \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mconv1\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnn\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mConv2d\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m3\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;36m16\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;36m3\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mstride\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mpadding\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;31mNameError\u001b[0m: name 'nn' is not defined"]}]},{"cell_type":"code","metadata":{"id":"WLN9BNMrP9Z7"},"source":["# Setting the hyper parameters\n","epochs = 30\n","lr = 0.02\n","decay_rate = 0.5\n","\n","# Use the NN model with GPU\n","cnn = CNN()\n","device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')\n","cnn.to(device)\n","\n","# Set loss criterion and optimization method\n","criterion = nn.CrossEntropyLoss()\n","optimizer = optim.SGD(cnn.parameters(), lr=lr, momentum=decay_rate)"],"execution_count":null,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"HMcnyh_xQAAS"},"source":["The settings and hyper-paremeters we have used are:\n","* Epoch = 30\n","* Learning rate = 0.02\n","* Batch size: 32\n","* Dataset size: 60,000\n","* Dataset: CIFAR-10\n","* Split rate between training, validation and testing: 4:1:1\n","* Optimizer: mini-Batch SDG with Momentum, Momentum rate: 0.5\n","* Loss: Cross Entropy Loss\n","* First model: 3 convolution layers with pooling using ReLU and dropout\n","* Second model: 4 convolution layers with pooling using ReLU and dropout\n","* Dropout rate: 0.6"]},{"cell_type":"markdown","metadata":{"id":"tHG_DVGohRwh"},"source":["## 7. Training the dataset"]},{"cell_type":"markdown","metadata":{"id":"09uuCyVn57_8"},"source":["### 7.1 Functions for training (with validation) and testing routine"]},{"cell_type":"code","metadata":{"id":"gWpaevFF51j_"},"source":["# Functions for training with validation routine and testing routine. We define them first to make the code more clean later.\n","\n","#Training with validation\n","def training(num_epochs, model, optimizer, criterion, train_loader, valid_loader, device):\n","  train_results = list()\n","  valid_results = list()\n","  for ep in range(num_epochs):\n","    running_loss = 0\n","    valid_loss = 0\n","    model.train() # Set model into training mode\n","    for images, labels in train_loader:\n","      images, labels = images.to(device), labels.to(device)\n","      # This will prevent the gradient descents from the previous batches to accumulate. Without this the weight will get updated with the sum of\n","      # all previos gradient descents, instead of the gradient descents on the current batch.\n","      optimizer.zero_grad() \n","      output = model(images) # prediction / output from the model\n","      loss = criterion(output, labels)  # We calculate the loss here\n","      loss.backward() # Computes the derivative of the loss using backpropagation.\n","      optimizer.step() # We update the weights\n","      running_loss += loss.item() # Sum the loss here\n","    else:\n","      train_results.append(running_loss / len(train_loader))  # We append the mean loss into a list\n","    \n","    # Validation part\n","    model.eval()\n","    with torch.no_grad(): # This will prevent calculation of gradient descents\n","      for images, labels in valid_loader:\n","        images, labels = images.to(device), labels.to(device)\n","        valid_output = model(images) # Prediciton\n","        loss = criterion(valid_output, labels) # Calculation of loss\n","        valid_loss += loss.item() # Sum the loss\n","      else:\n","        valid_results.append(valid_loss / len(valid_loader)) # Here we calculate the mean loss\n","  return train_results, valid_results\n","\n","# Testing\n","def testing(model, criterion, test_loader, classes, device):\n","  model.eval()\n","  test_loss = 0\n","  total = 0\n","  correct = 0\n","  # Accuracy calculation codes below are taken from https://pytorch.org/tutorials/beginner/blitz/cifar10_tutorial.html\n","  # prepare to count predictions for each class\n","  correct_pred = {classname: 0 for classname in classes}\n","  total_pred = {classname: 0 for classname in classes}\n","\n","  with torch.no_grad(): # This will prevent calculation of gradient descents\n","    for images, labels in valid_loader:\n","      images, labels = images.to(device), labels.to(device)\n","      test_output = model(images) # Prediciton\n","      loss = criterion(test_output, labels) # Calculation of loss\n","      test_loss += loss.item() # Sum the loss\n","\n","      \n","      # calculate accuracy\n","      _, predicted = torch.max(test_output.data, 1)\n","      total += labels.size(0)\n","      correct += (predicted == labels).sum().item()\n","\n","      # collect the correct predictions for each class\n","      _, predictions = torch.max(test_output, 1)\n","      for label, prediction in zip(labels, predictions):\n","          if label == prediction:\n","              correct_pred[classes[label]] += 1\n","          total_pred[classes[label]] += 1\n","    else:\n","      mean_test_loss = test_loss / len(test_loader) # Here we calculate the mean loss\n","      accuracy = 100 * correct / total\n","      class_accuracies = {classname: 100 * float(correct_pred[classname]) / total_pred[classname] for classname in classes}\n","\n","  return mean_test_loss, accuracy, class_accuracies\n"],"execution_count":null,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"UXHWaRBphYe8"},"source":["### 7.2 Training the dataset with the model using a GPU, plot the training curve"]},{"cell_type":"code","metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"LNawf5O9Mi1l","executionInfo":{"status":"ok","timestamp":1633470853885,"user_tz":-120,"elapsed":643775,"user":{"displayName":"Abdullah Karagoz","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"06337825384430712267"}},"outputId":"8264e35d-acc7-4919-c9ca-ae01cf8572d5"},"source":["train_results, valid_results = training(epochs, cnn, optimizer, criterion, train_loader, valid_loader, device)"],"execution_count":null,"outputs":[{"output_type":"stream","name":"stderr","text":["/usr/local/lib/python3.7/dist-packages/torch/nn/functional.py:718: UserWarning: Named tensors and all their associated APIs are an experimental feature and subject to change. Please do not use them for anything important until they are released as stable. (Triggered internally at  /pytorch/c10/core/TensorImpl.h:1156.)\n","  return torch.max_pool2d(input, kernel_size, stride, padding, dilation, ceil_mode)\n"]}]},{"cell_type":"code","metadata":{"id":"iRYBl0cWNFAp","colab":{"base_uri":"https://localhost:8080/","height":350},"executionInfo":{"status":"ok","timestamp":1633470854470,"user_tz":-120,"elapsed":604,"user":{"displayName":"Abdullah Karagoz","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"06337825384430712267"}},"outputId":"c9dc2277-33e5-4117-a13e-be9e7390dfa3"},"source":["# Plotting the results function\n","x = np.linspace(0, epochs, epochs)\n","\n","fig, ax = plt.subplots(figsize=(10,5))\n","\n","ax.plot(x, train_results, label='Training')\n","ax.plot(x, valid_results, label='Validation')\n","ax.set_xlabel('epochs')\n","ax.set_ylabel('mean CEL')\n","ax.set_title('Cross Entropy Loss first model')\n","ax.grid()\n","\n","plt.legend()\n","plt.show()"],"execution_count":null,"outputs":[{"output_type":"display_data","data":{"image/png":"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\n","text/plain":["<Figure size 720x360 with 1 Axes>"]},"metadata":{"needs_background":"light"}}]},{"cell_type":"markdown","metadata":{"id":"RfkIM5ihjqdF"},"source":["## 8. Report the test set performance"]},{"cell_type":"code","metadata":{"id":"ADpWRDoLSyem"},"source":["#Testing\n","mean_cel, total_accuracy, class_accuracy = testing(cnn, criterion, test_loader, train.classes, device)"],"execution_count":null,"outputs":[]},{"cell_type":"code","metadata":{"id":"KqMpdYPpbFCN","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"status":"ok","timestamp":1633470858777,"user_tz":-120,"elapsed":8,"user":{"displayName":"Abdullah Karagoz","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"06337825384430712267"}},"outputId":"0f087235-1dd2-4bd5-aff1-cbe1187e8027"},"source":["# Showing test results\n","from tabulate import tabulate\n","print(\"Mean Cross Entropy Loss from test is\", mean_cel)\n","\n","print('Total accuracy using the first model is %.2f %%' % total_accuracy)\n","\n","print(\"\\nAccuracy\")\n","\n","table = list()\n","table.append([\"Total accuracy\", total_accuracy])\n","\n","[table.append([k, v]) for k,v in class_accuracy.items()]\n","print(tabulate(table, headers=[\"Class\", \"Accuracy in %\"], tablefmt='orgtbl', floatfmt=\".2f\"))"],"execution_count":null,"outputs":[{"output_type":"stream","name":"stdout","text":["Mean Cross Entropy Loss from test is 0.7630577220703466\n","Total accuracy using the first model is 73.81 %\n","\n","Accuracy\n","| Class          |   Accuracy in % |\n","|----------------+-----------------|\n","| Total accuracy |           73.81 |\n","| airplane       |           81.82 |\n","| automobile     |           84.58 |\n","| bird           |           62.80 |\n","| cat            |           62.91 |\n","| deer           |           67.60 |\n","| dog            |           63.16 |\n","| frog           |           67.73 |\n","| horse          |           82.01 |\n","| ship           |           79.88 |\n","| truck          |           86.59 |\n"]}]},{"cell_type":"markdown","metadata":{"id":"ldiy2OFWulTA"},"source":["## 9. Second CNN model\n"]},{"cell_type":"markdown","metadata":{"id":"6PKI76wd6fJs"},"source":["### 9.1 Build the second CNN model with an additional convolution stage, training the dataset with the model. "]},{"cell_type":"code","metadata":{"id":"KkqFgdmBjAWO"},"source":["# Class for the second CNN model\n","class CNN2(nn.Module):\n","  def __init__(self):\n","    super(CNN2, self).__init__()\n","    self.conv1 = nn.Conv2d(3, 16, 3, stride=1, padding=1)\n","    self.conv2 = nn.Conv2d(16, 32, 3, stride=1, padding=1)\n","    self.conv3 = nn.Conv2d(32, 64, 3, stride=1, padding=1)\n","    self.conv4 = nn.Conv2d(64, 128, 3, stride=1, padding=1)\n","    self.pool = nn.MaxPool2d(2,2)\n","    self.dropout = nn.Dropout(0.6)\n","    self.fc1 = nn.Linear(128*2*2, 256)\n","    self.output = nn.Linear(256, 10)\n","  \n","  def forward(self, x):\n","    x = self.pool(F.relu(self.conv1(x)))\n","    x = self.pool(F.relu(self.conv2(x)))\n","    x = self.pool(F.relu(self.conv3(x)))\n","    x = self.pool(F.relu(self.conv4(x)))\n","    x = x.view(x.size(0), -1)\n","    x = self.dropout(x)\n","    x = self.fc1(x)\n","    return self.output(x)"],"execution_count":null,"outputs":[]},{"cell_type":"code","metadata":{"id":"Qz6EhIfgeSps"},"source":["# Use the NN model with GPU\n","cnn2 = CNN2()\n","device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')\n","cnn2.to(device)\n","optimizer_2 = optim.SGD(cnn2.parameters(), lr=lr, momentum=decay_rate)\n"],"execution_count":null,"outputs":[]},{"cell_type":"code","metadata":{"id":"TzgMknTNjIzl"},"source":["train_results_2, valid_results_2 = training(epochs, cnn2, optimizer_2, criterion, train_loader, valid_loader, device)"],"execution_count":null,"outputs":[]},{"cell_type":"code","metadata":{"id":"Vzr0FYHbkM96"},"source":["#Testing\n","mean_cel_2, total_accuracy_2, class_accuracy_2 = testing(cnn2, criterion, test_loader, train.classes, device)"],"execution_count":null,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"sa39bLRXkQAN"},"source":["### 9.2 Compare results from the first CNN model."]},{"cell_type":"code","metadata":{"id":"3INDiV_pkc8M","colab":{"base_uri":"https://localhost:8080/","height":859},"executionInfo":{"status":"ok","timestamp":1633471534574,"user_tz":-120,"elapsed":497,"user":{"displayName":"Abdullah Karagoz","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"06337825384430712267"}},"outputId":"0fea933e-4c05-4070-887e-483c06dfb4f5"},"source":["# Compare first and second model\n","x = np.linspace(0, epochs, epochs)\n","\n","fig, ax = plt.subplots(figsize=(10,8))\n","\n","ax.plot(x, train_results, color='#1f77b4', linestyle = 'dashdot', label='Training 1st model')\n","ax.plot(x, valid_results, color='#ff7f0e', linestyle = 'dashdot', label='Validation 1st model')\n","ax.plot(x, train_results_2, color='#1f77b4', label='Training 2nd model')\n","ax.plot(x, valid_results_2, color='#ff7f0e', label='Validation 2nd model')\n","ax.set_xlabel('epochs')\n","ax.set_ylabel('mean CEL')\n","ax.set_title('Cross Entropy Loss')\n","ax.grid()\n","\n","plt.legend()\n","plt.show()\n","\n","# Showing test results\n","\n","results = ['Cross Entropy Loss']\n","\n","print(\"Mean Cross Entropy Loss from test on first model is\", mean_cel)\n","\n","print(\"Mean Cross Entropy Loss from test on second model is\", mean_cel_2)\n","\n","print('Total accuracy using the fist model is %.2f %%' % total_accuracy)\n","\n","print('Total accuracy using the second model is %.2f %%' % total_accuracy_2)\n","\n","print(\"\\nAccuracy comparison in %\")\n","\n","table2 = list()\n","table2.append([\"Total accuracy\", total_accuracy, total_accuracy_2])\n","\n","[table2.append([c, a, b]) for c, a, b in zip(train.classes, class_accuracy.values(), class_accuracy_2.values())]\n","print(tabulate(table2, headers=[\"Class\", \"Model 1\", \"Model2\"], tablefmt='orgtbl', floatfmt=\".2f\"))"],"execution_count":null,"outputs":[{"output_type":"display_data","data":{"image/png":"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\n","text/plain":["<Figure size 720x576 with 1 Axes>"]},"metadata":{"needs_background":"light"}},{"output_type":"stream","name":"stdout","text":["Mean Cross Entropy Loss from test on first model is 0.7630577220703466\n","Mean Cross Entropy Loss from test on second model is 0.7733145008643214\n","Total accuracy using the fist model is 73.81 %\n","Total accuracy using the second model is 75.24 %\n","\n","Accuracy comparison in %\n","| Class          |   Model 1 |   Model2 |\n","|----------------+-----------+----------|\n","| Total accuracy |     73.81 |    75.24 |\n","| airplane       |     81.82 |    81.61 |\n","| automobile     |     84.58 |    86.07 |\n","| bird           |     62.80 |    61.80 |\n","| cat            |     62.91 |    69.93 |\n","| deer           |     67.60 |    70.50 |\n","| dog            |     63.16 |    65.54 |\n","| frog           |     67.73 |    75.40 |\n","| horse          |     82.01 |    82.63 |\n","| ship           |     79.88 |    82.06 |\n","| truck          |     86.59 |    77.44 |\n"]}]},{"cell_type":"markdown","metadata":{"id":"xHBV9C-Cuttp"},"source":["### 9.3 Results\n","One more convolutional layer means the model finds more high-level features out of the features it found until the previous convolutional layer.\n","\n","The second model having alittle higher accuracy than the first may be a just coincidence. I think that because mean CEL of second model is a little higher, the difference of accuracy is not that big, and when we look at the validation lines of both models in the plot, we see that they are on the same range and oscillating\n","\n","In the plot we observe that in the training the second model performs better with lower mean Cross Entropy Loss. But in validtion and testing the difference is not that big. We also observe mean CEL of validation is reduced faster with second model than with first model. Also the learning is quicker in second model. Here we just used the model with additional layer with same hyper-parameters that's (somewhat) optimized for the first model. Maybe the second model could perform better if we fine-tuned the hyper-parameters.\n","\n"]},{"cell_type":"code","metadata":{"id":"oKXFkcLsk0Vb"},"source":[""],"execution_count":null,"outputs":[]},{"cell_type":"code","metadata":{"id":"Iu26A367Ochr"},"source":[""],"execution_count":null,"outputs":[]}]}