mnist-demo-cnn.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "from torch.utils.data import DataLoader\n",
    "from torchvision import datasets, transforms\n",
    "from torchvision.utils import make_grid\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "from sklearn.metrics import confusion_matrix\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "transform = transforms.ToTensor()\n",
    "\n",
    "train_data = datasets.MNIST(root='./Data', train=True, download=True, transform=transform)\n",
    "test_data = datasets.MNIST(root='./Data', train=False, download=True, transform=transform)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Dataset MNIST\n",
       "    Number of datapoints: 60000\n",
       "    Root location: ./Data\n",
       "    Split: Train\n",
       "    StandardTransform\n",
       "Transform: ToTensor()"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Dataset MNIST\n",
       "    Number of datapoints: 10000\n",
       "    Root location: ./Data\n",
       "    Split: Test\n",
       "    StandardTransform\n",
       "Transform: ToTensor()"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test_data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_loader = DataLoader(train_data, batch_size=10, shuffle=True)\n",
    "test_loader = DataLoader(test_data, batch_size=10, shuffle=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define layers\n",
    "# 1 color channel(grayscale images), 6 filters(output channels), 3x3 kernel, stride=1\n",
    "conv1 = nn.Conv2d(1, 6, 3, 1)\n",
    "\n",
    "# 6 input channel(output of conv1), 16 filters(output channels), 3x3 kernel, stride=1\n",
    "conv2 = nn.Conv2d(6, 16, 3, 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Grab the first MNIST record\n",
    "for i, (X_train, y_train) in enumerate(train_data):\n",
    "    break"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "torch.Size([1, 1, 28, 28])\n"
     ]
    }
   ],
   "source": [
    "# Create a rank-4 tensor to be passed into the model (batch of 1 image)\n",
    "# (train_loader will have done this already)\n",
    "x = X_train.view(1,1,28,28)\n",
    "print(x.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "torch.Size([1, 6, 26, 26])\n"
     ]
    }
   ],
   "source": [
    "# Perform the first convolution/activation\n",
    "# resulting image is 26x26 because we are losing some border information. Not as critical for MNIST data is centered.\n",
    "x = F.relu(conv1(x))\n",
    "print(x.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "torch.Size([1, 6, 13, 13])\n"
     ]
    }
   ],
   "source": [
    "# Run the first pooling layer\n",
    "x = F.max_pool2d(x, 2, 2)\n",
    "print(x.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "torch.Size([1, 16, 11, 11])\n"
     ]
    }
   ],
   "source": [
    "# Perform the second convolution/activation\n",
    "x = F.relu(conv2(x))\n",
    "print(x.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "torch.Size([1, 16, 5, 5])\n"
     ]
    }
   ],
   "source": [
    "# Run the second pooling layer\n",
    "# 11/2 = 5.5 which is rounded down to 5\n",
    "x = F.max_pool2d(x, 2, 2)\n",
    "print(x.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "torch.Size([1, 400])\n"
     ]
    }
   ],
   "source": [
    "# Flatten the data\n",
    "x = x.view(-1, 5*5*16)\n",
    "print(x.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "class ConvolutionalNetwork(nn.Module):\n",
    "    def __init__(self):\n",
    "        super().__init__()\n",
    "        self.conv1 = nn.Conv2d(1, 6, 3, 1)\n",
    "        self.conv2 = nn.Conv2d(6, 16, 3, 1)\n",
    "        self.fc1 = nn.Linear(5*5*16, 120)\n",
    "        self.fc2 = nn.Linear(120, 84)\n",
    "        self.fc3 = nn.Linear(84,10)\n",
    "\n",
    "    def forward(self, X):\n",
    "        X = F.relu(self.conv1(X))\n",
    "        X = F.max_pool2d(X, 2, 2)\n",
    "        X = F.relu(self.conv2(X))\n",
    "        X = F.max_pool2d(X, 2, 2)\n",
    "        X = X.view(-1, 5*5*16)\n",
    "        X = F.relu(self.fc1(X))\n",
    "        X = F.relu(self.fc2(X))\n",
    "        X = self.fc3(X)\n",
    "        return F.log_softmax(X, dim=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "ConvolutionalNetwork(\n",
       "  (conv1): Conv2d(1, 6, kernel_size=(3, 3), stride=(1, 1))\n",
       "  (conv2): Conv2d(6, 16, kernel_size=(3, 3), stride=(1, 1))\n",
       "  (fc1): Linear(in_features=400, out_features=120, bias=True)\n",
       "  (fc2): Linear(in_features=120, out_features=84, bias=True)\n",
       "  (fc3): Linear(in_features=84, out_features=10, bias=True)\n",
       ")"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "torch.manual_seed(42)\n",
    "model = ConvolutionalNetwork()\n",
    "model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "def count_parameters(model):\n",
    "    params = [p.numel() for p in model.parameters() if p.requires_grad]\n",
    "    for item in params:\n",
    "        print(f'{item:>6}')\n",
    "    print(f'______\\n{sum(params):>6}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "    54\n",
      "     6\n",
      "   864\n",
      "    16\n",
      " 48000\n",
      "   120\n",
      " 10080\n",
      "    84\n",
      "   840\n",
      "    10\n",
      "______\n",
      " 60074\n"
     ]
    }
   ],
   "source": [
    "count_parameters(model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "criterion = nn.CrossEntropyLoss()\n",
    "optimizer = torch.optim.Adam(model.parameters(), lr=0.001)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch:  0  batch:  600 [  6000/60000]  loss: 0.04055628  accuracy:  78.417%\n",
      "epoch:  0  batch: 1200 [ 12000/60000]  loss: 0.08253479  accuracy:  85.800%\n",
      "epoch:  0  batch: 1800 [ 18000/60000]  loss: 0.36470532  accuracy:  88.689%\n",
      "epoch:  0  batch: 2400 [ 24000/60000]  loss: 0.01839534  accuracy:  90.525%\n",
      "epoch:  0  batch: 3000 [ 30000/60000]  loss: 0.00862619  accuracy:  91.687%\n",
      "epoch:  0  batch: 3600 [ 36000/60000]  loss: 0.00092614  accuracy:  92.508%\n",
      "epoch:  0  batch: 4200 [ 42000/60000]  loss: 0.52274615  accuracy:  93.145%\n",
      "epoch:  0  batch: 4800 [ 48000/60000]  loss: 0.02997379  accuracy:  93.623%\n",
      "epoch:  0  batch: 5400 [ 54000/60000]  loss: 0.00648251  accuracy:  94.020%\n",
      "epoch:  0  batch: 6000 [ 60000/60000]  loss: 0.06554906  accuracy:  94.317%\n",
      "epoch:  1  batch:  600 [  6000/60000]  loss: 0.01876213  accuracy:  97.817%\n",
      "epoch:  1  batch: 1200 [ 12000/60000]  loss: 0.05343531  accuracy:  97.958%\n",
      "epoch:  1  batch: 1800 [ 18000/60000]  loss: 0.00215177  accuracy:  97.978%\n",
      "epoch:  1  batch: 2400 [ 24000/60000]  loss: 0.03366597  accuracy:  97.958%\n",
      "epoch:  1  batch: 3000 [ 30000/60000]  loss: 0.34748060  accuracy:  97.947%\n",
      "epoch:  1  batch: 3600 [ 36000/60000]  loss: 0.00271200  accuracy:  97.967%\n",
      "epoch:  1  batch: 4200 [ 42000/60000]  loss: 0.00097422  accuracy:  98.029%\n",
      "epoch:  1  batch: 4800 [ 48000/60000]  loss: 0.00100963  accuracy:  98.037%\n",
      "epoch:  1  batch: 5400 [ 54000/60000]  loss: 0.00049988  accuracy:  98.033%\n",
      "epoch:  1  batch: 6000 [ 60000/60000]  loss: 0.07968108  accuracy:  98.027%\n",
      "epoch:  2  batch:  600 [  6000/60000]  loss: 0.00135666  accuracy:  98.783%\n",
      "epoch:  2  batch: 1200 [ 12000/60000]  loss: 0.00098702  accuracy:  98.608%\n",
      "epoch:  2  batch: 1800 [ 18000/60000]  loss: 0.00116520  accuracy:  98.589%\n",
      "epoch:  2  batch: 2400 [ 24000/60000]  loss: 0.02344611  accuracy:  98.588%\n",
      "epoch:  2  batch: 3000 [ 30000/60000]  loss: 0.37386140  accuracy:  98.523%\n",
      "epoch:  2  batch: 3600 [ 36000/60000]  loss: 0.04631621  accuracy:  98.525%\n",
      "epoch:  2  batch: 4200 [ 42000/60000]  loss: 0.01826158  accuracy:  98.517%\n",
      "epoch:  2  batch: 4800 [ 48000/60000]  loss: 0.00106201  accuracy:  98.535%\n",
      "epoch:  2  batch: 5400 [ 54000/60000]  loss: 0.00142397  accuracy:  98.524%\n",
      "epoch:  2  batch: 6000 [ 60000/60000]  loss: 0.00013551  accuracy:  98.500%\n",
      "epoch:  3  batch:  600 [  6000/60000]  loss: 0.00090608  accuracy:  98.933%\n",
      "epoch:  3  batch: 1200 [ 12000/60000]  loss: 0.00242908  accuracy:  98.925%\n",
      "epoch:  3  batch: 1800 [ 18000/60000]  loss: 0.00029286  accuracy:  98.944%\n",
      "epoch:  3  batch: 2400 [ 24000/60000]  loss: 0.00011522  accuracy:  98.892%\n",
      "epoch:  3  batch: 3000 [ 30000/60000]  loss: 0.01019260  accuracy:  98.877%\n",
      "epoch:  3  batch: 3600 [ 36000/60000]  loss: 0.00442119  accuracy:  98.844%\n",
      "epoch:  3  batch: 4200 [ 42000/60000]  loss: 0.00043636  accuracy:  98.871%\n",
      "epoch:  3  batch: 4800 [ 48000/60000]  loss: 0.01701790  accuracy:  98.896%\n",
      "epoch:  3  batch: 5400 [ 54000/60000]  loss: 0.00709904  accuracy:  98.887%\n",
      "epoch:  3  batch: 6000 [ 60000/60000]  loss: 0.09021780  accuracy:  98.890%\n",
      "epoch:  4  batch:  600 [  6000/60000]  loss: 0.00065222  accuracy:  99.200%\n",
      "epoch:  4  batch: 1200 [ 12000/60000]  loss: 0.22825721  accuracy:  99.167%\n",
      "epoch:  4  batch: 1800 [ 18000/60000]  loss: 0.00001880  accuracy:  99.089%\n",
      "epoch:  4  batch: 2400 [ 24000/60000]  loss: 0.00065590  accuracy:  99.050%\n",
      "epoch:  4  batch: 3000 [ 30000/60000]  loss: 0.00313960  accuracy:  99.030%\n",
      "epoch:  4  batch: 3600 [ 36000/60000]  loss: 0.00348670  accuracy:  99.044%\n",
      "epoch:  4  batch: 4200 [ 42000/60000]  loss: 0.00265265  accuracy:  99.036%\n",
      "epoch:  4  batch: 4800 [ 48000/60000]  loss: 0.00109186  accuracy:  99.040%\n",
      "epoch:  4  batch: 5400 [ 54000/60000]  loss: 0.00024350  accuracy:  99.037%\n",
      "epoch:  4  batch: 6000 [ 60000/60000]  loss: 0.00126942  accuracy:  99.042%\n",
      "\n",
      "Duration: 173 seconds\n"
     ]
    }
   ],
   "source": [
    "import time\n",
    "start_time = time.time()\n",
    "\n",
    "epochs = 5\n",
    "train_losses = []\n",
    "test_losses = []\n",
    "train_correct = []\n",
    "test_correct = []\n",
    "\n",
    "for i in range(epochs):\n",
    "    trn_corr = 0\n",
    "    tst_corr = 0\n",
    "    \n",
    "    # Run the training batches\n",
    "    for b, (X_train, y_train) in enumerate(train_loader):\n",
    "        b+=1\n",
    "        \n",
    "        # Apply the model\n",
    "        y_pred = model(X_train)  # we don't flatten X-train here\n",
    "        loss = criterion(y_pred, y_train)\n",
    " \n",
    "        # Tally the number of correct predictions\n",
    "        predicted = torch.max(y_pred.data, 1)[1]\n",
    "        batch_corr = (predicted == y_train).sum()\n",
    "        trn_corr += batch_corr\n",
    "        \n",
    "        # Update parameters\n",
    "        optimizer.zero_grad()\n",
    "        loss.backward()\n",
    "        optimizer.step()\n",
    "        \n",
    "        # Print interim results\n",
    "        if b%600 == 0:\n",
    "            print(f'epoch: {i:2}  batch: {b:4} [{10*b:6}/60000]  loss: {loss.item():10.8f}  \\\n",
    "accuracy: {trn_corr.item()*100/(10*b):7.3f}%')\n",
    "        \n",
    "    train_losses.append(loss.item())\n",
    "    train_correct.append(trn_corr.item())\n",
    "        \n",
    "    # Run the testing batches\n",
    "    with torch.no_grad():\n",
    "        for b, (X_test, y_test) in enumerate(test_loader):\n",
    "\n",
    "            # Apply the model\n",
    "            y_val = model(X_test)\n",
    "\n",
    "            # Tally the number of correct predictions\n",
    "            predicted = torch.max(y_val.data, 1)[1] \n",
    "            tst_corr += (predicted == y_test).sum()\n",
    "            \n",
    "    loss = criterion(y_val, y_test)\n",
    "    test_losses.append(loss)\n",
    "    test_correct.append(tst_corr)\n",
    "        \n",
    "print(f'\\nDuration: {time.time() - start_time:.0f} seconds') # print the time elapsed            "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "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(train_losses, label='training loss')\n",
    "plt.plot(test_losses, label='validation loss')\n",
    "plt.title('Loss at the end of each epoch')\n",
    "plt.legend();"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[tensor(0.0040),\n",
       " tensor(0.0003),\n",
       " tensor(0.0017),\n",
       " tensor(0.0003),\n",
       " tensor(0.0019)]"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test_losses"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "While there may be some overfitting of the training data, there is far less than we saw with the ANN model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "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([t/600 for t in train_correct], label='training accuracy')\n",
    "plt.plot([t/100 for t in test_correct], label='validation accuracy')\n",
    "plt.title('Accuracy at the end of each epoch')\n",
    "plt.legend();"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Extract the data all at once, not in batches\n",
    "test_load_all = DataLoader(test_data, batch_size=10000, shuffle=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test accuracy: 9869/10000 =  98.690%\n"
     ]
    }
   ],
   "source": [
    "with torch.no_grad():\n",
    "    correct = 0\n",
    "    for X_test, y_test in test_load_all:\n",
    "        y_val = model(X_test)  # we don't flatten the data this time\n",
    "        predicted = torch.max(y_val,1)[1]\n",
    "        correct += (predicted == y_test).sum()\n",
    "print(f'Test accuracy: {correct.item()}/{len(test_data)} = {correct.item()*100/(len(test_data)):7.3f}%')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[   0    1    2    3    4    5    6    7    8    9]]\n",
      "\n",
      "[[ 973    0    0    0    0    1    4    0    2    0]\n",
      " [   0 1134    2    1    1    0    3    3    0    0]\n",
      " [   2    0 1022    0    0    1    2    7    2    1]\n",
      " [   0    1    1 1006    0   24    0    2    2    9]\n",
      " [   0    0    1    0  977    0    5    0    0    5]\n",
      " [   0    0    0    1    0  860    4    0    0    2]\n",
      " [   2    0    0    0    0    1  938    0    0    1]\n",
      " [   1    0    4    1    0    1    0 1015    1    6]\n",
      " [   2    0    2    1    0    2    2    1  965    6]\n",
      " [   0    0    0    0    4    2    0    0    2  979]]\n"
     ]
    }
   ],
   "source": [
    "# print a row of values for reference\n",
    "np.set_printoptions(formatter=dict(int=lambda x: f'{x:4}'))\n",
    "print(np.arange(10).reshape(1,10))\n",
    "print()\n",
    "\n",
    "# print the confusion matrix\n",
    "print(confusion_matrix(predicted.view(-1), y_test.view(-1)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "131"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "misses = np.array([])\n",
    "for i in range(len(predicted.view(-1))):\n",
    "    if predicted[i] != y_test[i]:\n",
    "        misses = np.append(misses,i).astype('int64')\n",
    "        \n",
    "# Display the number of misses\n",
    "len(misses)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([   8,  184,  235,  320,  321,  340,  412,  445,  448,  582])"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Display the first 10 index positions\n",
    "misses[:10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Set up an iterator to feed batched rows\n",
    "r = 12   # row size\n",
    "row = iter(np.array_split(misses,len(misses)//r+1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Index: [   8  184  235  320  321  340  412  445  448  582  625  659]\n",
      "Label: [   5    8    9    9    2    5    5    6    9    8    6    2]\n",
      "Guess: [   9    3    8    8    7    3    3    0    8    2    4    1]\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 720x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "nextrow = next(row)\n",
    "print(\"Index:\", nextrow)\n",
    "print(\"Label:\", y_test.index_select(0,torch.tensor(nextrow)).numpy())\n",
    "print(\"Guess:\", predicted.index_select(0,torch.tensor(nextrow)).numpy())\n",
    "\n",
    "images = X_test.index_select(0,torch.tensor(nextrow))\n",
    "im = make_grid(images, nrow=r)\n",
    "plt.figure(figsize=(10,4))\n",
    "plt.imshow(np.transpose(im.numpy(), (1, 2, 0)));"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 72x72 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "x = 2019\n",
    "plt.figure(figsize=(1,1))\n",
    "plt.imshow(test_data[x][0].reshape((28,28)), cmap=\"gist_yarg\");"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Predicted value: 9\n"
     ]
    }
   ],
   "source": [
    "model.eval()\n",
    "with torch.no_grad():\n",
    "    new_pred = model(test_data[x][0].view(1,1,28,28)).argmax()\n",
    "print(\"Predicted value:\",new_pred.item())"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3.10.4 ('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": "3.10.4"
  },
  "vscode": {
   "interpreter": {
    "hash": "d4d1e4263499bec80672ea0156c357c1ee493ec2b1c70f0acce89fc37c4a6abe"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}