mnist_cnn.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Using matplotlib backend: Qt5Agg\n",
      "Populating the interactive namespace from numpy and matplotlib\n"
     ]
    }
   ],
   "source": [
    "%pylab\n",
    "%matplotlib inline\n",
    "import numpy,torch,torchvision"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "import torch.optim as optim\n",
    "from torchvision import datasets, transforms\n",
    "from torch.autograd import Variable"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "MnistModel (\n",
       "  (conv1): Conv2d(1, 32, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))\n",
       "  (conv2): Conv2d(32, 64, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))\n",
       "  (fc1): Linear (3136 -> 1024)\n",
       "  (fc2): Linear (1024 -> 10)\n",
       ")"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "class MnistModel(nn.Module):\n",
    "    def __init__(self):\n",
    "        super(MnistModel, self).__init__()\n",
    "        # input is 28x28\n",
    "        # padding=2 for same padding\n",
    "        self.conv1 = nn.Conv2d(1, 32, 5, padding=2)\n",
    "        # feature map size is 14*14 by pooling\n",
    "        # padding=2 for same padding\n",
    "        self.conv2 = nn.Conv2d(32, 64, 5, padding=2)\n",
    "        # feature map size is 7*7 by pooling\n",
    "        self.fc1 = nn.Linear(64*7*7, 1024)\n",
    "        self.fc2 = nn.Linear(1024, 10)\n",
    "        \n",
    "    def forward(self, x):\n",
    "        x = F.max_pool2d(F.relu(self.conv1(x)), 2)\n",
    "        x = F.max_pool2d(F.relu(self.conv2(x)), 2)\n",
    "        x = x.view(-1, 64*7*7)   # reshape Variable\n",
    "        x = F.relu(self.fc1(x))\n",
    "        x = F.dropout(x, training=self.training)\n",
    "        x = self.fc2(x)\n",
    "        return F.log_softmax(x)\n",
    "    \n",
    "model = MnistModel()\n",
    "model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "batch_size = 50\n",
    "train_loader = torch.utils.data.DataLoader(\n",
    "    datasets.MNIST('data', train=True, download=True, transform=transforms.ToTensor()),\n",
    "    batch_size=batch_size, shuffle=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "test_loader = torch.utils.data.DataLoader(\n",
    "    datasets.MNIST('data', train=False, transform=transforms.ToTensor()),\n",
    "    batch_size=1000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "torch.Size([32, 1, 5, 5])\n",
      "torch.Size([32])\n",
      "torch.Size([64, 32, 5, 5])\n",
      "torch.Size([64])\n",
      "torch.Size([1024, 3136])\n",
      "torch.Size([1024])\n",
      "torch.Size([10, 1024])\n",
      "torch.Size([10])\n"
     ]
    }
   ],
   "source": [
    "for p in model.parameters():\n",
    "    print(p.size())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "optimizer = optim.Adam(model.parameters(), lr=0.0001)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train Step: 0\tLoss: 0.167\tAccuracy: 96.000\n",
      "Train Step: 1000\tLoss: 0.119\tAccuracy: 98.000\n",
      "Train Step: 2000\tLoss: 0.148\tAccuracy: 94.000\n"
     ]
    }
   ],
   "source": [
    "model.train()\n",
    "train_loss = []\n",
    "train_accu = []\n",
    "i = 0\n",
    "for epoch in range(2):\n",
    "    for data, target in train_loader:\n",
    "        data, target = Variable(data), Variable(target)\n",
    "        optimizer.zero_grad()\n",
    "        output = model(data)\n",
    "        loss = F.nll_loss(output, target)\n",
    "        loss.backward()    # calc gradients\n",
    "        train_loss.append(loss.data[0])\n",
    "        optimizer.step()   # update gradients\n",
    "        prediction = output.data.max(1)[1]   # first column has actual prob.\n",
    "        accuracy = prediction.eq(target.data).sum()/batch_size*100\n",
    "        train_accu.append(accuracy)\n",
    "        if i % 1000 == 0:\n",
    "            print('Train Step: {}\\tLoss: {:.3f}\\tAccuracy: {:.3f}'.format(i, loss.data[0], accuracy))\n",
    "        i += 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x7f1779baacc0>]"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX4AAAD8CAYAAABw1c+bAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nacb/A4Nku+hImztxAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f17d58eeeb8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(np.arange(len(train_loss)), train_loss)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x7f17757cd908>]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAD8CAYAAACGsIhGAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1779bce128>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(np.arange(len(train_accu)), train_accu)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Test set: Accuracy: 98.66%\n"
     ]
    }
   ],
   "source": [
    "model.eval()\n",
    "correct = 0\n",
    "for data, target in test_loader:\n",
    "    data, target = Variable(data, volatile=True), Variable(target)\n",
    "    output = model(data)\n",
    "    prediction = output.data.max(1)[1]\n",
    "    correct += prediction.eq(target.data).sum()\n",
    "\n",
    "print('\\nTest set: Accuracy: {:.2f}%'.format(100. * correct / len(test_loader.dataset)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python [default]",
   "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": "3.6.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}