Autograd.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import torch\n",
    "from torch.autograd import Variable\n",
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "x = Variable(torch.from_numpy(np.zeros(10)).float(),\n",
    "             requires_grad=True)\n",
    "\n",
    "y = Variable(torch.from_numpy(np.ones(10)).float(), \n",
    "            requires_grad=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Compute z as a function of x and y.\n",
    "z = x.dot(y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Variable containing:\n",
      " 0\n",
      "[torch.FloatTensor of size 1]\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Compute the forward pass.\n",
    "print(z)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Variable containing:\n",
      " 1\n",
      " 1\n",
      " 1\n",
      " 1\n",
      " 1\n",
      " 1\n",
      " 1\n",
      " 1\n",
      " 1\n",
      " 1\n",
      "[torch.FloatTensor of size 10]\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(x.grad)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Compute the backward pass (reverse-mode autodiff)\n",
    "z.backward()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Variable containing:\n",
       " 1\n",
       " 1\n",
       " 1\n",
       " 1\n",
       " 1\n",
       " 1\n",
       " 1\n",
       " 1\n",
       " 1\n",
       " 1\n",
       "[torch.FloatTensor of size 10]"
      ]
     },
     "execution_count": 80,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x.grad"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Building logistic regression from scratch. \n",
    "m = 784\n",
    "\n",
    "# parameters\n",
    "w = Variable(torch.from_numpy(np.zeros(m)).float(), requires_grad=True)\n",
    "b = Variable(torch.from_numpy(np.zeros(1)).float(), requires_grad=True)\n",
    "\n",
    "def logistic_regression(x):\n",
    "    \"Write the function like you would in standard method\"\n",
    "    return torch.sigmoid(w.dot(x) + b) \n",
    "\n",
    "def negative_likelihood(xs, ys):\n",
    "    ll = 0\n",
    "    for n in range(xs.size(0)):\n",
    "        # Can even have if statements \n",
    "    \n",
    "        if ys[n] == 5:\n",
    "            ll = ll + torch.log(logistic_regression(xs[n]))\n",
    "        else:\n",
    "            ll = ll + torch.log(1 - logistic_regression(xs[n]))\n",
    "    return -ll"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Load image data.\n",
    "from data_util import load_mnist\n",
    "trX, teX, trY, teY = load_mnist(onehot=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "5\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x7f8d28d0d5c0>"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRfiBoAg/EBThB4Ii/EBQhB8IivADQf0/sEWOix6VKakAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f8d4c52b4a8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "print(trY[0])\n",
    "plt.imshow(trX[0].reshape(28, 28))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.5"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "xs = Variable(torch.from_numpy(trX[:10]).float(), requires_grad=True)\n",
    "logistic_regression(xs[0]).data[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Variable containing:\n",
      " 6.9315\n",
      "[torch.FloatTensor of size 1]\n",
      "\n"
     ]
    }
   ],
   "source": [
    "ys = torch.from_numpy(trY[:10]).long()\n",
    "L = negative_likelihood(xs, ys)\n",
    "print(L)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "L.backward()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x7f8d28937ba8>"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nQYoAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f8d28cde400>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(w.grad.data.numpy().reshape(28, 28))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 225,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(60000, 784)"
      ]
     },
     "execution_count": 225,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "trX.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "304.41976741483086\n",
      "186.578247347079\n",
      "143.98104433847948\n",
      "109.42108591646684\n",
      "87.42873766692355\n",
      "72.18946863081874\n",
      "60.8571748311424\n",
      "49.934328361894586\n",
      "41.826126895002744\n",
      "35.904716126919084\n",
      "31.470956345334343\n",
      "28.175670119810093\n",
      "25.52304454682053\n",
      "23.189465945504708\n",
      "21.181663327374736\n",
      "19.47156900009213\n",
      "17.965678690309687\n",
      "16.621669217478484\n",
      "15.43870530051663\n",
      "14.410063948577772\n"
     ]
    }
   ],
   "source": [
    "import torch.optim as optim\n",
    "w.data.zero_()\n",
    "b.data.zero_()\n",
    "opt = optim.SGD([w, b], lr=0.1)\n",
    "\n",
    "for epoch in range(20):\n",
    "    total = 0\n",
    "    for i in range(0, 1000, 5):\n",
    "        # Reset param grads\n",
    "        opt.zero_grad()\n",
    "\n",
    "        # Find minibatch\n",
    "        xs = Variable(torch.from_numpy(trX[i:i+5]).float(), requires_grad=True)\n",
    "        ys = torch.from_numpy(trY[i:i+5]).long()\n",
    "\n",
    "        # Compute minibatch likelihood and gradients.\n",
    "        L = negative_likelihood(xs, ys)\n",
    "        L.backward()\n",
    "        #print(xs.grad)\n",
    "        #print(L.data[0])\n",
    "        total += L.data[0]\n",
    "        # Update with a gradient step.\n",
    "        #print(w.grad.data)\n",
    "        opt.step()\n",
    "    print(total)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x7f8d288a5828>"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nIqXkF4nU/wHKRVaeIh5cgAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f8d28c0aeb8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(w.data.numpy().reshape(28, 28))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Everything else in pytorch is just fancy extensions beyond the core autograd mechanism"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# The module or blocks approach.\n",
    "from torch import nn\n",
    "class Linear(nn.Module):\n",
    "    def __init__(self, m, n):\n",
    "        super() \n",
    "        self.W = torch.FloatTensor(m, n)\n",
    "        self.b = torhc.FloatTensor(n)\n",
    "        self.register_parameter()\n",
    "        self.register_parameter()\n",
    "\n",
    "    def forward(self, input):\n",
    "        return self.W.dot(input) + self.b\n",
    "\n",
    "\n",
    "layer = Linear(784, 10)\n",
    "layer.forward(x)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.5.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}