Getting started.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "import numpy as np\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "import torch\n",
    "from torch.autograd import Variable\n",
    "import torch.nn as nn"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(1000,) (1000,)\n"
     ]
    }
   ],
   "source": [
    "X = np.arange(-5,5,0.01)\n",
    "y = 4*X**3 + 2**X + 3\n",
    "print(X.shape, y.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x7f252f9b9320>]"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYMAAAD8CAYAAACVZ8iyAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\noDIQERFUBiIiAvwfbZn0c5WznEcAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f25800df198>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(X,y)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Creating variable from numpy arrays\n",
    "Similar to tensorflow placeholders"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Variable containing:\n",
       "-5.0000\n",
       "-4.9900\n",
       "-4.9800\n",
       "-4.9700\n",
       "[torch.DoubleTensor of size 4x1]"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "batch = Variable(torch.from_numpy(X[:4, np.newaxis])) \n",
    "batch"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Concatanating along axis"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Variable containing:\n",
       "-5.0000 -5.0000\n",
       "-4.9900 -4.9900\n",
       "-4.9800 -4.9800\n",
       "-4.9700 -4.9700\n",
       "[torch.DoubleTensor of size 4x2]"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "torch.cat((batch, batch), 1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Convert tensors to type float for easy usage in nn layers"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Variable containing:\n",
       "-5.0000\n",
       "-4.9900\n",
       "-4.9800\n",
       "-4.9700\n",
       "[torch.FloatTensor of size 4x1]"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "batch = Variable(torch.from_numpy(X[:4, np.newaxis])).float() ## Converting to float is important for GPU\n",
    "batch"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Matix multiplication as Layer operation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Variable containing:\n",
       "-2.5177 -1.7919 -2.4977\n",
       "-2.5135 -1.7896 -2.4947\n",
       "-2.5094 -1.7872 -2.4916\n",
       "-2.5052 -1.7849 -2.4886\n",
       "[torch.FloatTensor of size 4x3]"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "nn.Linear(1,3)(batch)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Do the same for target"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Variable containing:\n",
       "-496.9688\n",
       "-493.9745\n",
       "-490.9923\n",
       "-488.0220\n",
       "[torch.FloatTensor of size 4x1]"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "target = Variable(torch.from_numpy(y[:4, np.newaxis])).float()\n",
    "target"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Broadcasting surrogate"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Variable containing:\n",
       " 0  0  0\n",
       "[torch.FloatTensor of size 1x3]"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "hidden = Variable(torch.zeros(1,3))\n",
    "hidden"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Variable containing:\n",
       " 0.1454  0.2194 -0.2428\n",
       "[torch.FloatTensor of size 1x3]"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "h = nn.Linear(3,3)(hidden)\n",
    "h"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Variable containing:\n",
       "-2.4399  2.8048 -3.9366\n",
       "-2.4362  2.7988 -3.9279\n",
       "-2.4325  2.7927 -3.9192\n",
       "-2.4288  2.7866 -3.9105\n",
       "[torch.FloatTensor of size 4x3]"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x = nn.Linear(1,3)(batch)\n",
    "x"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "inconsistent tensor size at /data/users/soumith/miniconda2/conda-bld/pytorch-cuda80-0.1.6_1485185046107/work/torch/lib/TH/generic/THTensorMath.c:601\n"
     ]
    }
   ],
   "source": [
    "try:\n",
    "    x + h ## Will give error\n",
    "except RuntimeError as e:\n",
    "    print(e)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Variable containing:\n",
       " 0.1454  0.2194 -0.2428\n",
       " 0.1454  0.2194 -0.2428\n",
       " 0.1454  0.2194 -0.2428\n",
       " 0.1454  0.2194 -0.2428\n",
       "[torch.FloatTensor of size 4x3]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "h.expand_as(x) ## This makes h same size as x and compatible for addition"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Variable containing:\n",
       "-2.2945  3.0242 -4.1794\n",
       "-2.2908  3.0182 -4.1707\n",
       "-2.2871  3.0121 -4.1620\n",
       "-2.2834  3.0060 -4.1533\n",
       "[torch.FloatTensor of size 4x3]"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x + h.expand_as(x) ## Finally"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Getting the size of the tensor of variable"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "torch.Size([4, 3])"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x.size()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "4"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x.size(0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(4, 3)"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x.size()[0], x.size()[1], "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "True"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "isinstance(x.size(), tuple)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Variable containing:\n",
       "-2.4399  2.8048 -3.9366\n",
       "-2.4362  2.7988 -3.9279\n",
       "-2.4325  2.7927 -3.9192\n",
       "-2.4288  2.7866 -3.9105\n",
       "[torch.FloatTensor of size 4x3]"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Get data-tensor inside a variable"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "\n",
       "-2.4399  2.8048 -3.9366\n",
       "-2.4362  2.7988 -3.9279\n",
       "-2.4325  2.7927 -3.9192\n",
       "-2.4288  2.7866 -3.9105\n",
       "[torch.FloatTensor of size 4x3]"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x.data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Numpy conversion happens only on tensors and not variables.\n",
      "numpy\n"
     ]
    }
   ],
   "source": [
    "try:\n",
    "    x.numpy()\n",
    "except AttributeError as e:\n",
    "    print(\"Numpy conversion happens only on tensors and not variables.\")\n",
    "    print(e)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[-2.43992019,  2.80484176, -3.93660831],\n",
       "       [-2.43620563,  2.79876256, -3.9279108 ],\n",
       "       [-2.43249106,  2.79268384, -3.91921329],\n",
       "       [-2.4287765 ,  2.78660488, -3.91051579]], dtype=float32)"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x.data.numpy() ## succeeds"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Simple linear regression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "np.random.seed(1337)\n",
    "X = np.random.randn(1000,1)*4\n",
    "W = np.array([0.5,])\n",
    "bias = -1.68"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "y_true = np.dot(X, W) + bias\n",
    "y = y_true + np.random.randn(X.shape[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x7f252c83b518>"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXwAAAD8CAYAAAB0IB+mAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nH8C7TLprxgpYAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f252f967be0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.scatter(X, y, s=1, label=\"data\")\n",
    "plt.scatter(X, y_true, s=1, color='r', label=\"true\")\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def get_variable_from_np(X):\n",
    "    return Variable(torch.from_numpy(X)).float()\n",
    "\n",
    "class LinearRegression(nn.Module):\n",
    "    def __init__(self, input_size, output_size):\n",
    "        super(LinearRegression, self).__init__()\n",
    "        self.x2o = nn.Linear(input_size, output_size)\n",
    "        \n",
    "    def forward(self, X):\n",
    "        return self.x2o(X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Variable containing:\n",
       "-2.8127\n",
       "-1.9611\n",
       "-1.2873\n",
       "-7.0203\n",
       " 0.8267\n",
       "-8.0451\n",
       "-2.2290\n",
       " 1.3489\n",
       " 6.1953\n",
       "-5.4829\n",
       "[torch.FloatTensor of size 10x1]"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "batch_size = 10\n",
    "batch = get_variable_from_np(X[:batch_size])\n",
    "batch"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "model = LinearRegression(1, 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Variable containing:\n",
       "-0.3634\n",
       "-0.2134\n",
       "-0.0947\n",
       "-1.1047\n",
       " 0.2777\n",
       "-1.2852\n",
       "-0.2606\n",
       " 0.3697\n",
       " 1.2235\n",
       "-0.8338\n",
       "[torch.FloatTensor of size 10x1]"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_pred = model.forward(batch)\n",
    "y_pred"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x7f252c749fd0>"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXwAAAD8CAYAAAB0IB+mAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f252f919f60>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "batch = get_variable_from_np(X[:])\n",
    "y_pred = model.forward(batch)\n",
    "y_pred_np = y_pred.squeeze().data.numpy()\n",
    "plt.scatter(X, y, s=1, label=\"data\")\n",
    "plt.scatter(X, y_true, s=1, color='r', label=\"true\")\n",
    "plt.scatter(X, y_pred_np, s=1, color='k', alpha=0.5, label=\"fit\")\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Define loss criterion and optimizer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "criterion = nn.MSELoss()\n",
    "optimizer = torch.optim.SGD(model.parameters(), lr=0.001, momentum=0.9)\n",
    "losses = []"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Train the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss at epoch [9]: 4.496\n",
      "Loss at epoch [19]: 3.292\n",
      "Loss at epoch [29]: 1.980\n",
      "Loss at epoch [39]: 1.496\n",
      "Loss at epoch [49]: 1.491\n",
      "Loss at epoch [59]: 1.008\n",
      "Loss at epoch [69]: 1.075\n",
      "Loss at epoch [79]: 0.758\n",
      "Loss at epoch [89]: 1.242\n",
      "Loss at epoch [99]: 1.871\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.text.Text at 0x7f252c7725f8>"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAW4AAAEKCAYAAAAyx7/DAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAIABJREFUeJztnXl0HNWV/79X8qbFlmxZ3mRbMrExtsHYsUwwWxggwWAC\nTAIJ0DCZJBOHLcMyc0KIGbIqYTKZTJg5kKAJW3CTEAgMWYD8CDAHBzBYBgw2XjG2JS9YtvG+Srq/\nP67eVHV1VVd1q0uq7rqfc3Rq6aru96pb37p13733ETNDURRFKRxK+roBiqIoSnaocCuKohQYKtyK\noigFhgq3oihKgaHCrSiKUmCocCuKohQYKtyKoigFhgq3oihKgaHCrSiKUmD0C+NNhw8fzg0NDWG8\ntaIoSlGydOnSHcxcG+TYUIS7oaEBLS0tYby1oihKUUJEG4Meq64SRVGUAkOFW1EUpcDwFW4imkxE\nb9v+9hLRzb3ROEVRFCUdXx83M68GMAMAiKgUwGYAT4XcLkVRFMWDbF0l5wJ4n5kDO9EVRVGU/JKt\ncF8B4NdhNATJJNDQAJSUyDKZDOVjFEVRCp3Awk1EAwBcDOBxj9fnE1ELEbW0t7dn14pkEpg/H9i4\nEWCW5fz5Kt6KoiguUNCpy4joEgA3MPOn/Y5tbGzkrOK4GxpErJ3U1wMbNgR/H0VRlAKFiJYyc2OQ\nY7NxlVyJsNwkmzZlt19RFCXGBBJuIqoA8CkAT4bSivHjs9uvKIoSYwIJNzMfYOYaZt4TSiuamoDy\n8tR95eWyX1EURUkhGpmTiQTQ3Az07y/b9fWynUj0bbsURVEiSChFpnIikQDuvRcoKwP+8pe+bo2i\nKEpkiYbFbaisBA4c6OtWKIqiRJroCff+/X3dCkVRlEgTLeGuqFCLW1EUxYdoCbda3IqiKL5ES7jV\n4lYURfElWsJdWQkcPAh0dfV1SxRFUSJLtIS7okKWBw/2bTsURVEiTLSEu7JSlurnVhRF8SRawm0s\nbvVzK4qieBIt4VaLW1EUxZdoCbexuFW4FUVRPImWcBuLW10liqIonkRLuNXiVhRF8SVawq0Wt6Io\nii/REm61uBVFUXyJlnCrxa0oiuJLtITbTF+mFreiKIon0RLukhIRb7W4FUVRPImWcAPi51aLW1EU\nxZNAwk1E1UT0BBGtIqKVRDQntBY5py9LJoGGBrHGGxpkW1EUJcYEnSz4bgDPMfNlRDQAQHloLbJb\n3MkkMH++VS1w40bZBnQGeEVRYouvxU1EVQDOAnA/ADDzUWbeHVqL7Bb3ggXpJV4PHpT9iqIoMSWI\nq2QCgHYADxLRW0T0SyKqCK1Fdot70yb3Y7z2K4qixIAgwt0PwMcB/JyZZwI4AOCbzoOIaD4RtRBR\nS3t7e+4tslvc48e7H+O1X1EUJQYEEe42AG3M/Hr39hMQIU+BmZuZuZGZG2tra3Nvkd3ibmqyYrsN\n5eWyX1EUJab4CjczbwPQSkSTu3edC+C90Fpkt7gTCeDuu63XysuB5mYdmFQUJdYEjSr5OoBkd0TJ\negBfCq1Fzjjuc86x9g8cCFxxRWgfrSiKUggEiuNm5re73SDTmflSZv4otBYZi5tZtnfulOWllwK7\ndgEtLaF9tKIoSiEQzcxJZuDQIdk2wn3FFZKE89xzfdc2RVGUCBA94XZWCDTCPWkScMopKtyKosSe\n6Am3sya3Ee6aGmDMGGDxYk1/VxQl1kRPuJ0zve/cCRABzz4LPPOM7GO20t9VvBVFiRnRE25jcRtX\nyY4dQHU18C//Ahw+nHqspr8rihJDoifcbhZ3TY2mvyuKonQTPeF2WtxGuDX9XVEUBUAUhdvL4tb0\nd0VRFABRFG4vizuRkHT3YcNkf12dpr8rihJLgqa89x5eFjcgIl1TA1xwAfDYY8Dpp/dNGxVFUfqQ\naFvcR47I0gg3IJY2AGze3PttUxRFiQDRE+5+/aSY1P79qck3BhVuRVFiTvSEGxCr+8ABS7iHD7de\nGzoUGDRIhVtRlNgSTeGurPS2uInE6lbhVhQlpkRTuJ0Wt124AalZsmVL77dLURQlAkRTuDNZ3IBa\n3IqixJroCncmi9sIt5lsQVEUJUZEU7jN9GU7dwJlZfJnp65OCk59FN5EPIqiKFElmsJtt7id1jag\nIYGKosSaaAq33eJW4VYURUkhmsLtZ3GPGSNLFW5FUWJIoFolRLQBwD4AnQA6mLkxzEalWNzTp6e/\nboRbQwIVRYkh2Vjcf8PMM0IXbUAs7s5OEWY3i3vgQMmmNBZ3MilzUOpclIqixIDoVQcErEJT+/a5\nCzdghQQmkzL35MGDst/MRQloyVdFUYqSoBY3A/gLES0lovlhNgiAVdoV8BfuBQss0TboXJSKohQx\nQYX7DGaeAeACADcQ0VnOA4hoPhG1EFFLe3t7z1plLG7AX7i95pzcuFHdJoqiFCWBhJuZN3cvtwN4\nCsApLsc0M3MjMzfW1tb2rFVBLO4xY4Dt24Fx47zfx7hNVLwVRSkifIWbiCqIaLBZB/BpAMtDbVVQ\nixsAbr0V6N/f+73UbaIoSpERxOIeCeCvRLQMwBsA/sTMz4XaqqA+bgCYPRs47rjM4u3lTlEURSlA\nfKNKmHk9gJN7oS0W2Vjcb74JrFkDfOc7wAMPiHvEyfjxeW+ioihKXxHdzElA4rKrq92PMcJ9zz1S\nJfBznwOamoDy8tTjystlv6IoSpEQTeE2FvfQoSLebvz5z7JctUrmqXz7bYnbbm62LOyqKtnWeG5F\nUYqIaCbgGIvby01ikm4MHR2pSTeJBPCxj4n/W0VbUZQiI5oW9+OPy3LNGvdY7CBJN5MmAWvXhtpM\nRVGUviB6wu20pt1isb2iROz7jXDrLDmKohQZ0RPuINa0V5SIff+kSVLrpKdZnIqiKBEjesIdxJoO\nEj0yaZIs1V2iKEqRET3hDmJNm+iR+nqASJbO6JGJE2XpJtxaBlZRlAImesIdNBY7kQA2bAC6umTp\njB5paABKS9OF2/jQN24U/7fWM1EUpcCInnAHsaaD0L8/MGFCunBrGVhFUQqcaMZxm1jsnuIWEhjE\nh64oihJhomdx55OJE4F161JDAoP40BVFUSJMcQv3pEky6fCHH1r7tJ6JoigFTvELN5DqLjE+dFO8\nSuuZKIpSYMRPuAER6WuvlfUrr1TRVhSloChu4a6vl8qBbrHc+/bJcseO3m2ToihKDylu4X7sMVne\ndVd6oo0Rbk2JVxSlwChe4TaJNh0dsu1MtFGLW1GUAqV4hdsv0UYtbkVRCpTiFW6/RBsj3Dt3Stq8\noihKgVC8wu2XaGOEu7MT2L27d9qkKIqSBwILNxGVEtFbRPTHMBuUN/wSbfbvBwYMkHX1cyuKUkBk\nY3HfBGBlWA3JOybRZuRI2R4xIjXRZt8+iTQBgvm5tRSsoigRIZBwE9FYAPMA/DLc5uSZRAJ49llZ\n/8UvLNFmFuGeMEG2/SxuLQWrKEqECGpx/wzANwAU3iieSW23+7GPHJEwweOOk20/i1tLwSqKEiF8\nhZuILgKwnZmX+hw3n4haiKilPUohdlVVstyzx9pnBiaDWtxaClZRlAgRxOI+HcDFRLQBwG8AnENE\nC50HMXMzMzcyc2NtbW2em9kDhgyRpZtwjxghA5Z+NxotBasoSoTwFW5mvp2ZxzJzA4ArALzIzFeH\n3rJ80a8fUFnpLtyDBwO1tf4Wd1MTMGhQ6j4tBasoSh9RvHHcdqqqUn3cduEePtzf4k4kgG98w9rO\ndTq1TGjUiqIoAclq6jJm/l8A/xtKS8KkqiqzxR3EJ/+JT8hy5EiZnDifmKgVMwBqolYALTmrKEoa\n8bC4q6u9hXv48GAJOOb8HTvynyKvUSuKomRBPITby1VSWRnc4jbnd3am3gTygUatKIqSBfER7kwW\n94EDwKFDmd/Dfn6+U+Q1akVRlCxQ4R4+XNb9xNh+fr7j1HUCY0VRsiAewm183MyyvW8fMHAg0L+/\nuEqA7IQ73xa3qatSWSnbzroqiqIoNuIh3FVVwNGjwOHDsr1vn1jbgGVx+1nRu3eHW00wkQAuvljW\n//u/VbQVRfEkPsINWFazXbizsbhNinxYKf1798rywIFw3l9RlKIgHsJtCk0Z4d6/P3uLe88eYMwY\nyaAMq3638b2rcCuKkoF4CLexuE1In93iHjpUshWDWNzV1cHjvnNBhVtRlADES7jdXCWlpcCwYcEs\n7qqq4HHfuWBcJc5kHEVRFBvxEG6nq8Qu3ECwQlO7d4twq8WtKEofEw/hzmRxA/6Fpjo75Rwj3GFZ\n3CrciqIEIF7C7ebjBvwtbiOo1dXpx3pV9cu22l9Hh+UiUeFWFCUDWVUHLFgqK0VATRKOPaoE8Lei\njaVuLO69eyUu/PHH3av6vfIK8PDD7tX+ACketWmTpLQ3NUnM9v791usq3IqiZCAewl1SIjPh7Nkj\nYtrVZWUpAmJF79wp+0tcHkLswn3kiKzv2OFd1a+5Wdwrzv033SQ1UdwE/cwzrWNVuBVFyUA8XCWA\nVSHQXqfEsGGDiHa/fu5uDeNiMVElgAi3V/U+p2gbdu70Lt9q2mX2KYqieBAf4Tb1SpzCnUyKywMQ\nN4qxgu3i7XSVACLcXtX7Skuza9umTanC3VOLW2fTUZSiJj7CbSoEOoV7wQLxV9txTmJghNsk4ADi\nE29qkmJVdsrLRfjd5qisqXFv2/jxVgx3TU3PhNvMprNxo/eNSFGUgkaFO8gkBnaL2+4qSSSASy6x\njjNV/e69F7j+emt/ebnsv/tu70mHTbtGjuyZcMd1Nh19ylBiRLyE283HHWQSA7twDxsm6yYkcMAA\nq5b2nXdaVf1GjJDluecCo0bJ/kQCuPJK631HjbLKt5p2jRrVM+GO42w6+pShxIz4CLeXjzvIJAa7\nd4tLxNTwrq62wgdXrADOOEPeb+VK65yVK4HRo4G5c4H164Ht22W/KS0LAL/6lSX0xlUyenTPhDuO\ns+nE9SlDiS2+wk1Eg4joDSJaRkQriOi7vdGwvGNcJUYgjXCbSQz6dUdG1tenT2Jg6pQYTBJOZ6cI\n9LRpwJQp6cI9ZQpw6qmyvXixWIMvvwxMny77du2yjrdb3D2JKvHyuxfzbDpxfMpQYk0Qi/sIgHOY\n+WQAMwDMJaJTw21WCFRVScjftm2ybQ8HTCSAOXOAs86S0EDnJAamMqDB1CvZsEEsaKdwM1vC/fGP\ny01h8WI5fvNm4NJL5Ti7cO/dK26XoUMli9I5YBqURAL4/OetbbcbUbERx6cMJdb4CjcLJq2vf/cf\nh9qqMDDCu3mzLO0JOAAwdizQ1uZ+rtPiNpmWK1bI9rRpwNSpwNatcuyWLWJBT5ki1u7JJ4twv/yy\nHG8GNJ0W9+DBQEWFbPfEXTJkiCwHDgQ++KC4RRvQOTuV2BHIx01EpUT0NoDtAJ5n5tfDbVYIGOFt\na5N/ames9dixIursck/ycpW8955sT50qIg2IpW0sb7NvzhzgjTeAl16Swc0ZM+TG4RTuIUPyI9yr\nVsnyyBFJ+il2jLvLROyMHFn8TxlKrAkk3MzcycwzAIwFcAoRneg8hojmE1ELEbW0h1U9rycY4W1t\nTXWTGOrqvIXOlHQ1GFfJihUi+EOGZBbuU08VIf7tbyW1vaREBNzpKhk82LIceyLcq1dbTxRbtuT+\nPoVEIiE3UAB46CEVbaWoySqqhJl3A3gJwFyX15qZuZGZG2tNrHOUMK6StjZ34R471nrdiZvFffSo\nuD+mTZN9EyaIa2LlSrF4q6pkoBGwIkoOHRJ3STKZLtz5cpXs3y99+OQnZdu4hsIiSvHTxmAw4ZuK\nUqQEiSqpJaLq7vUyAJ8CsCrshuUde2nXXITbOTgJAOvWWcJdWgocf7xlcU+ZAhCJkN1xh3XuRx9J\njPHRo6nW/d69qa6SXCNL1qyR5TnnyDJM4Y5S/DSzCrcSG4JY3KMBvERE7wBYAvFx/zHcZoWA3WJ2\nDkwC4ioB0oWuo0OsX6erxGCEG7AiS4xwA94xxhs2hGNxr14ty7PPdu9PPolS/PSBA1aMvAq3UuT4\nlnVl5ncAzOyFtoSL3WJ2s7hHjZLHfafFbeK+na4Sg1O4H39crL8TTpB9XrHEBw+GMzi5apX0Y+pU\naWeYPu4oxU/bx1VUuJUiJz6Zk4MGSdYj4C7c/fpJ1qJTuO0lXQ2vvmqtX3655RqYMsWKSjEWt1cs\n8ZAhItzmeDM4mQ+Lu6FB+ltXF67FHaX4aRVuJUbER7iJLPF1E27AXejsdUoAEWm7K6C11fLrrl9v\n7b/2WtnnFWN84YXAsWMi0F1dssxHVMnq1Za1P2ZMuMIdpfhpFW4lRsRHuAF/4XZLwnEK94IFEh1i\nx8xu84MfWPu2bLFmt2lulgxGIiuT8bzz5LVdu6xpy3o6ONnVJcI9ebJs19WF6ypJJKQSomHYsL6L\nnzbCPXiwCrdS9MRj6jKD8XNnEu7nn0/dZ6/FDXj7b93iv81AnVsa/VNPyXLXLmu6tJ66Stra5KZi\nLO66OglFPHpU0unDwMROAxI901fx00a4J05U4VaKHrW47dTVySChGZAE0i3ubP23XkJvysPu2pVa\nsbBfPxHZXITbZEzaLW5AUvHDYvFia91+3Xqb9naJo6+rU+FWih4VbjsmltvuF3YOTnr5dTPNbuOG\nXbiN4JkaIxUVuQm3CQU0wj1mjCzD9HMvXixiWVHR98JdW2tVgVSUIiZewh3EVQKk+rmdFrepi+H0\nWd99d3YDdUbonRa3Oc9LuL0yFZNJ4FvfkvVPfEK2jcUdpp978WJJ6R8yJHXezN5GhVuJEfHycQdx\nlQCpFuqePUBZmRVKCFiz2bixYIG4R8aPF9H2Om7oUFnu2mXFhftZ3CZT0QxcmkzFV14BHn7Y2r9p\nk+z/939P708+2b5dImmuuw5YvjxaFjez3FgVpQiJj8WdTAIPPijrX/2qe1q2EW6nxW1P3slEIiED\nkV1d7gOSdsrK5M/uKjE3lIoK96gSr0zF5mb3/T/6kfjLwxJu4982FndUhLuzs2eTUShKxImHcBtL\n1TxCf/ihe02NQYMknd0p3Pbkm3xiCk05XSVeFrfXQGdnp/v+1lbxc4flKlm8WAZTZ82StkdFuAF1\nlyhFTTyEO5uaGqYut8FZ0jWfGOEOOjjpNdDprC1uPz6M7EnjZ//Rj8TX/uST2Vnc+a4oePiwxMKr\ncCsxIR7CnU1Njbq6vrG4+/Wz5or0Eu6mpnS/bXm5PD14zTOZb+G2VwQEJEZ8/nypTx5kcDKMioI7\ndshShVuJCfEQ7mxqajizJ3tLuAcPtkTZK6rkjDNSZ+gxES333gt88Yuyzx7pkkhYae9uM/sYsrGA\nvZ5eli0LZnGHUVHQJN+ocCsxIR5RJU1NqdEYgHeo3tixYsEdPgz87nfA2rVW4aZMUSK5YHeVGDcJ\n4G1x//nPspw4UY55++3UdgOSOWm3vuvqpN9797rfgLwiVQD3vno9vZinBr9ojjAqCqpwKzEjHha3\nV+x1JmEqKwOuuUYiRIBwJglwWtwGr6iS554Dxo0DPvUpGXi009oKjBiR7jLxqjNuyNYC9np6qa6W\n2uWmJrYXYVQUNMI9fLgKtxIL4iHcQLBQvWQS+NWvrG2neyHfkwQMGyYW8vbt6Rb3wYOpn3/sGPDC\nC8D554t479qVKritrbLfiV/2pJelu3FjeoJPQ4Pl27ZTXm7NXO/nLmlq8vbH54pa3ErMiI9wB2HB\nApkwOBP5nCTApL1v3JhucTOnViF8/XURRSPcQKrV7SXcS5fK8tOfdvdfZ7J0zVPG9denDkgCljvE\nWe3Qb4AykQCuuir183taUbC9XSJrqqtldiMiFW6lqFHhthNElPM5SYAR7i1bUoXbWZM7mQTmzZP1\nW26xiknZB1HdhNs536Wbu8et9oodrwQfZhFt8/RinhiCDFCapwAAeOONno8btLeLm6SkRP6GDFHh\nVooaFW47fqKc70kCjHAzp7tKABFuM3hoBLGtzUplNxb33r3y5xRur9rhdndPIgH8679mbqdXgo/9\nRmduPEGE2/6k4OZ6yRaTfGPQeiVKkaPCbcfN+nS6BPIdVWJwukoAEW63wUMzAGgE0Cydwh00gmP2\nbFnaxc9OpgQfQzYWd2urdbwKt6JkjQq3Hbfok0ceEYvYr/ZILvgJ98GDmd03fsIdNILDnH/LLe4V\nDufP9698mK1wn3aarKtwK0rW+Ao3EY0jopeI6D0iWkFEN/VGw/qMbApF9RS7cHu5SrzEt39/y8ft\nJdxB54Q053/ta3LjGjlStkeMsBJ8mpslThtwf/ow7fcbnGSWdk+bJgKrwq0oWRPE4u4A8E/MPBXA\nqQBuIKKpPucoQaiocJ953i7cTU2WYBrKy4GTTkq1uEtKUgf9gPQnCECOu+aa1AiT1lZ5z6FD5ZzX\nXpP9d91liXMiIUW4brrJ/YYW1OI2yU3jxlmDmz3h2DHgo496X7jzXW+l0NHr0av4Cjczb2XmN7vX\n9wFYCaAu7IbFAiLL6rZb3PaokkQCOPlkEXh78tBpp6UK9+jR6QIPWE8Qjzwi5+/fn14jxESkGHF3\ni/3et0/OrfP46gcNEl+4n3Dbnw7q63tucZu5PntTuMOot1LI6PXodbLycRNRA4CZAF4PozGxxAi3\nl8UNiFU5d26q+2bsWBGnffu8Y7jtLFjgnVDkPH/gQAmvswu3WXda9QaiYBUC8y3c9uQbg30yhTAI\no95KIaPXo9cJLNxEVAngdwBuZua0/04imk9ELUTU0m7+mRR//ISbWWaZOe641POM0La1BRPuTBEm\nbuc7qwqamt5eFjcQbPoyp3Dv3WvN65kLXsLd0ZEeCpkvwqi3Usjo9eh1Agk3EfWHiHaSmZ90O4aZ\nm5m5kZkba73CypR03Fwl9qiSHTvERTFhQup59uzJIMLtNcg5bpzMAu8n3H4Wt+lDEIu7f38Z+Gxo\nkH25Wt3JJPCFL8j6VVdZj+Zhp72HUW+lkNHr0es+/iBRJQTgfgArmfmnobYmbiSTwIsvyvpnPmN9\n2WVlsjxwQKxtIN3iNtUAly0Ty9JPuL0iTG69Vaz6oBZ3JuEOMgtOa6u0vaRELG4gtwFK41c1tbi3\nbrX8qmELt1dd9HwmZxUSXgPocbkefeDjD2Jxnw7gGgDnENHb3X8XhtaiuGC+bOPH3rLF+rJLSqya\n3B98IK87hbuuTsTj1Vdl20+4TYSJM9Rv1iz38+vqpPjVsWOyvXmzWNSVld6fEdTiNp9lhDsXizuT\nXzWocOdqJY0fL/+g5mmpvDz/yVmFRCIBnHiitR1GslqU6QMff5Cokr8yMzHzdGae0f33TGgtigt+\nX7YRbmNxO10lAwaICJvQPT/hBuQfafVqWb/pJtn2igGvqxNx2rpVtrdsyWxtA9kLd22tPF3kItyZ\n/KpBhDsXK8kI/Vlnidj/5Ccy8XS/fsDll2ffh2LCjCece274+Q9RIVPFTCBUH79mTvYVfgM6ZjKF\n9euBUaPcC0GNGycTH5v1IFRVifVuJmHIJNyAleSzZUvmgUnAf3Cys1Msd/NZRGK95iLcmfyqQYQ7\nWyvJOWVbVxdw442We2jRouzaX0x0dFgGRlwCE5y/BzdC9PGrcPcVfgM6duF2ukkMxs/dv7/lAgnC\njBnAW2/JemurCJ09qgVIn4Bh8+aeW9wffij/5PabRK4hgZmyQoMId7aREF5C//jjEj75xz8Ga3cx\nsmmTuNQGDoyPcLv9HuyE7ONX4e4r/NLRzWQKmYTbCGBdnTy6B2XmTGDduswx4Hbh7uoSl4mfxT14\nsETAeFUTdLPuGxqswclsfM6JBHDffda23a8aRLizjYTwEvS2NnEP/OEP4cWNR521a2XZ2CiDxXG4\nDpncIL3g41fh7iv8plOrqJD45tZWf+EO6iYxzJghy2XLvIV72DCxoDZvln/GY8eCWdyAiLcbbsJd\nXy/v/8AD2fuczz1Xlvfck+pXNRMvZxJut5l4iNJn/jFkEvoRI4D335fM0TimexvhPu00+Z3EoU6M\n1+/BXqM+RFS4+5JMBa3Ky4GVK+U158CkwViqixZlJxgzZ8ryrbdETN1+hERWSGCQ5BvAv16Jl3AD\nMuFDtiPzpv8mHtxQUiLinUlAEgng6qtT9xlL0WvCCbcp1y68EHjsMev8fIaCFUr9j7VrxdA46STZ\njoO7pKkp/Sm3F0MgVbijSkWFJTxuFncyCfzyl9Z2NoIxZoyktL/2mvyTeVnsTuEOanF7DVC2tkpN\nE3tVRBPlYqJXnGR6JDW+cSP+doLUK5k4UZZu/XebcMJEjtifkJ55xn+yilwopPof69bJtTSJd3EQ\n7rPOEqNqyBD/CchDQIU7qpjsScBduN3mxwwqGERidT/3nGz7CXeQrEnAfxactrbUYlbJpITUZSLT\nyLyxuHMV7p075UZinwLOjvOmYWq4dHZaT0hhpXvnEhvcVxb62rXApEnxEu6HHpLl22/3TgloByrc\nUcUI94AB7oLZU8GYOVPKoQLBhXv06MzvmclVkkwCTz8t/+RGVNymVrPj9+i5cSNQU+OeFGSEO5OY\n7dwp5wcdqFy+XNwB9qzJsNK9s/1++8pC7+iQJDEv4S4Udw8QrK3JpBgKd94pN32TANfbMHPe/2bN\nmsVKD/nGN5gB5smT3V+vr5fXnX/19cHe/9FHrXPWrnU/5qc/ldcvu4x5xAj/93znHTn+8cdT9y9c\nyFxentpO57bzr7RUzjPn19czE8nS7J87l9nrt3bhhcwNDe6fa86/+GLm6dO922eOY2bu7GSuqGD+\nx38M1jf7uU68+mMn2++3p7+HXFm7Vj7ngQeYDx6U9R/+UF7L5dr0FUHaGnJ/ALRwQI1V4Y4q3/2u\nfD0XXOD+ek9/RD/+sXXeuHHu5z32mLw+fjzzzJn+77lhgxx///2p+71EpbTUfX9NjSxbWjL3c8oU\n5s9+1r0tV17J3K9fZjE7/XTmv/kb63oOHGi97rwe778vrzU3p3/WwoXMdXXy+rBh/qId5HtbuDD9\n+mT6fonc+0rk3ZZ88Mwz8jmLFsl2RQXzLbfIel/dTHIhSFtD7o8Kd6GzcCFzdbV8PZWV3v+sQSw3\nr/OCiMcUhRkZAAAWq0lEQVRf/2q9Pm+e//vu2iXH/uxnqfu9RMXN8i4vZ/7FL0S0br/d+59l/Hjm\nsjLmW291b8u113p/phGzKVOYP/c565zLLmM+4QT393v6aTn3tdfcX+/qku/q61/PfI2y+edvaLBe\nr6vL/P2GJSp+v7G775bP2bbNavPVV8t6X91MciFIW0Pujwp3IdMbj5dB/8k/+MB6bf58//c9dkyO\n/d73gn+elzBMm+ZtMZt/FkCEw8nChcxDhnifa/o5YkRqv268kXnoUPe+/eAHcu7evd79nz2b+bzz\nMl+joP/85kYwZ468/tvfZn5f+xNDvn43QX6LN97IPHiwtJdZrsH558u6WtxZkY1w6+Bk1OiNSmNB\nB77sg6J+ESWAFFsqK0sfnGxqsubWNJiBR7dY9mRSBjE7Orw/y6T4O2O4zSCdV2SL+VxmYNcuGZw0\njBolA7bOaB0AePdd+SxnaQA7U6YA773n/ToQfDBz2zZJZLrsMhmgXrIk8/smElKT3FBT0/PwtCC/\nRRNRYgZsa2utwcmmJmm7naiWe/3qV9P3Odva1GSVXPY6ppdQ4Y4avTGbSFDxGDDAihQIItyAe72S\nRELKx5aWBot5XbAAOHrU+zPKy4HPf17WnaGAmWpIVFVZn7t3r9wY7MJtbgamcJedd9+1Eky8mDpV\nYt4zhSF+//vBanmb+PaTTpI5R/2EG5DJnsvKJE7+0kt7Hp7m91tMJoEXXgDefNOKwrALdyIBXHKJ\ndV5JiWS5Bm1Xb0akvPWWXDvzexg5Mv03mkjI92foy/K1QU3zbP7UVdIDeuPxMpsBsv795fXa2mCP\n3ZMmycCgkxNPZL7oomDt8/OJL1zI/G//JtsffRT83GuvtY4zg40PPmjt+8MfZN8bb6S+5+HD4rb5\n1rcytzuTH9zuEgLEvQAwDxrkfl1/8Qt5feNG5uuuk+M7OzN//kUXSZTMvHniv+8pfi4ut9/QhRdK\nn4zr5MtfZh45kvkvf5FjamqCjcn0VkTKwoXMo0fL+1dVMf/Xf8n6f/6n+/Evviivv/hiftvB6iop\nbPyKT+UDvzopgOVyMBMptLcHiwt2mwXn2DGxIKdNC9Y+ryeC8nJJgEkkJE65qgqorg52bv/+4n4w\nmNnhna4SIPU4QNre0ZE6WYAbU6fKcuXK1P3OGGtAknjmzJHYeDeLbc0asQDHjgVmz5Zs1DVrMn/+\nmjXitjjtNGnDrl2Zj/cj02/Ry43y6qvA4cPWBCEmq3LbNvmt7dwp18AeZ+5mWfdGApL5XkzW7p49\nwG23yVNjS4v7OSZZy6/8Q9gEVfhs/tTi7iG5Rovkk1wt/7PPZj7zzNR9K1bIub/6VbDP9rK2rrlG\n1jdsEOvy5JODn3viicynnmod9+yz8torr1j7Nm2Sfc6Qv4ULZf+772Zud0eHDBD+8z+n7ve6llVV\n8h0fPJj+XvPmifXMLJ/rd/2OHZOngttvZ37pJTn+T3/K3N4gGAsUkIFb81vM9GQDMK9fL8fV1TF/\n8Yve16CmJrsYf68IjlwsdK82lZXJ4LgbP/yhHLN/f44X1BuoxV3gZCo+1Vvk6mt383GvWCHLoBa3\n1xPBrbfK64sWicXmlurude7Mmf4W94gRsrT7uJNJ4PrrZX3evMxWXGkpMHlyusXtdc327hWpcLOk\n16wBjj9e1qdMkUzaTH7uDRvkqWDSJLHQS0vzk9Vnz5b92tes36LXk409e/LgQcm6nTjR+xrs3Olu\nWZeWuh/v9bm5WOhebTp0SL5D89Rgp61NxhLsJSn6ABVuxZ1cU7m9hJtIBCgobjevk06S91+0SPY5\nI0oynTtqlAi3cVW4CffAgfJPaQTeGaGyaZO/u2jq1PTIEq9rZlwzq1al7j96VOqwT54s26Wl4jK5\n7z5vN4AprXr88SIqM2bkR7hffVWuy8iRqYXAvCKFbrhB1tvbrVlxJk7MvgRAZ2d6REpZmbfLMBdD\nI9PNp6vLmiXKzubN1gQmfYgKt+JOrr52t+nLli8HPvax9FCqbCktBc44Q2ab2bfP3eL2YtQo8b0a\nEd65U24mQ4emH2eEOxcrbsoUuVnYz8sUDkmULtzr14twGYs7mZR630ePpvuHDcZqN+ecdhrw+uuZ\nQyqD8NprMkHC+PGpwp1IyHdRUpL6ZHPNNfJ6e7v4twH57r1+T/Ybp536enlCM5FIgJTQ9Xr6zMXQ\naGpKt+zLy6UOCeDu525r63v/NlS4FS+CDGC64TY4uWJFcDeJH2eeaZWZ9bK43XAOPO7cKQObzn/c\nUaMsV0kuVtzUqSKuJpwPAL7wBSmENWhQ6rX80pek1rpTuI0IG4t7wYJ0AXbeQNaskcHa4cNlu6tL\njhkwIPdQuiNHgKVL5SYwenR66d3ycmD69NQnG7ur5P33ZX3ixNTfk6F/f+vJx/m+d94p7orrr5f3\nnzEDeOop7ycOtxtDJgsdkDaNGCHH2b+XG2+U/i5dmn5OW1thWNxE9AARbSei5b3RICVC5OJrHzJE\nLEOTxHLkiDzG+0VkBMVeTfCGG4ILklO4d+1KrQtuGDnSOiYXK864g+x+7j/9SRJ7Hn00/VqecEK6\ncBvRN9ZzkBvI2rVyPJFckwcekP1eFnoQ3nxTvksj3M5oGze3QWWlNffkunVyjc1Tjfk93X+/bNvj\n3e2x7d/5jpx3+LDEgSeT1qQiXv0xNwb7jTiRyPyb3b9f+nTbbenfy6xZ6cJ99Kjc1AtBuAE8BGBu\nyO1QigVnadc1a+SxPx8Wt7N+97ZtwQXJzeJ2e0y3W9y5ZP6Zx+tEQqzf4cMlGaa01H1KtxNOEKHu\n6rL2rVkjlqsRvCA3EPtgplu5XGOhe4XMue03PnITttjeboWHAu5uAyIrCceEAjr53vfS9zFbc6fu\n2SMlgKurZcKCoLXnP/tZ+a1997tybbZvT/8cO8uWyefOmpX+2oABMlZhvx7miSMCwh0o9ARAA4Dl\nQUNVNBwwxnzta1bYVn098/XXy/ayZT1/754kJ+3cKceaAlizZrlXXrzrrtRwr4svTu1PtkkjfuFp\nzc3y2gcfWPvOPJP5jDMyv6/9vQ4dkvZ95zuynW1Rr+uuc9/f2Mh83HHynvfdJ/tbW2X78GHZdtal\nYZZKkvPmMU+YwHzVVemvZ6rXctJJVlVE08eg9V2WL5f9jz4qxb4GDcoctmcKZG3enP49utV9ufNO\nWX/2We/37AHQcEClT0gmrZlBzCNtc7NYYcZf2xN6Ug5g6NDUJBwvi9uZ9l5SIlZxEHdRpnR7wN1K\nPOEEWa5aZVm9ixZJCraxho0bwDzNjBuXOt7w/vtyvY3F7WWhl5a6D7Y2N7vvb2mR65VMWmGBxuo0\n4wxu1mdtrbhRNm50t7i92jdsmDx9dHZabZg/392l5fY+ZjB04kRxsRw+DDz/vPu5gLiCRo1KL+fg\nZeHfe6+sR8DizptwE9F8Imohopb2OExdpKTj9oPv6BBRmTy557UmejLbDFGq/zqTqwSwhHvlyuBh\njEFuIM5jzHsvXGhlVwISQ2x3AyUSlt/6d79LvYGYwcxJk2TpFcFhBNGJ137AEk8TGmeEO1MGYW2t\nRBJ1dUlEiROv9gHpNWrMDcV5/KBB6S4ru3Bv3izf+d/+rffg7JtvAh//ePp+r+9xxw5ZFpNwM3Mz\nMzcyc2OtGVlW4oXfxL49nUqrp+UATKjf0aMSTpjJ4jbHrVsXXLiD3ECcxwwfLu14+mn/0MPGRlk6\nw9Scwu0WwXHrrd7hk17JLvZ23HefrBvhNtPZeQm3iYJxs7i9Ipa8UvR37Uo9HgDOPjv96WfdOrmW\nzzwDXHedFbPv9ts7dEh82G7C7fU9Dh4sMfJVVe6v9yZB/ClQH7cSBC8fdLb+6Ez0pBzARRcxz5jB\nvHWrtOWee9KP2bJFXrv3XitV/5FHgrctWx+3mz/Vy4/b1cU8fDjzV76Sen5FhXVtne+/f7/4er3a\n5eXjdmsHEfO3vy3v+5OfyP7du9OvQ1OTdd6HHwa7dszBxzBOPdVqj73P553H/IlPBHufxYtl31NP\npbfDa0zhlFO8pxLMA8jnRAoAfg1gK4BjANoAfMXvHBXumOInXG4DSr3JP/wD86hR1iDWb36Tfsyx\nY5ZAPfGEHLdkSfDPsN9YamoyV8MLcr2cojV3rlWjJeg8iV4TUlRXW8d+//v+7aittSaeuPlmuWGY\nKoB2zICrfYKFoNcuSH+8JoxoaGBOJDIPzprv4u//XrY3bvRui6kaWFMj23PmMJ9zTvD+ZElehTuX\nPxXuGGOEKyyLuyfccQdzSYlVmvP5592Pq62V6BgjZvv2hdMevycUNwv9jjsk6uLgwZ7N2lJaKlO1\nGUwZ2f/4D2/xnD6d+TOfkeMvv5z5+OPd+3XzzaltyeapyO+JKtNUdiUlcsMN8uRnP8+rfZ2dUljL\nPOGMG8f8d38XvC9Zko1wa1SJkl9MksXCheGXp82WUaNkwMwkuHilWxtf+MqVEsFRWRlOezKNCXhl\nqjY2ymDismXBomy8junsTC1atWSJXI+bbvLOmLVnT3rV7EgmgZ//3NrOdmzDL+nLqz+trXLOxInu\nYyFeZKo/U1IimbovvyzXa8uWSAxMApryroRFrinzYWIiRky1wkzC/eGH2UWU5ILXIFh9vXfooX2A\nMkiUjdcx1dUiqiYCrKVF3pvIWzztwu1VsyNoskyuePXHpPo70+udsw25kal9Z54pWanvvCPircKt\nFD1RKE9rxwi3qd7nFR9sKuGtXh2ucOcSJTNmjPSjpQX4ylfSX3ebJ9HtM0wVvyVLJMJi+XLrpuDF\n6NFyQ+voEOvTTbjDnnrPqz/nny/rJorF/tsLUozMq31nnSXLX/9alhEoMAWocCtxwm5xDxjgXVN5\n1CixRg8eDFe4c3kqIRKBbWkRK7CsTNw5Xud7fcZtt8n2kiXidunsDCbcHR3yJNLR4W599iTWPghu\noY7f/rY8QVRVuT9FBXGdeLVv5kz5nRjhVotbUXoZe1ZkTY33Y7Q5DghXuIHcnkoGDJCbzxNPSDbo\nj36U+Xy3zxg8WLI2lyyx4sKDCDdgHe9mffbW1HsbNlix5EeOWHVR3L5Tp9gHmazZ0L+/1GoxCUcq\n3IrSy1RWWgONXv5twLLMgfCFO1uSSak2aNi7N/fEptmzRbiXLJGblZ8bwAi3GdR0O743xzbGjAFO\nP11uYF4Frezt2rBBYkkeeSS79tkTbmbPDne2+aAEDT/J5k/DAZXIMnGihIF98pPex3zzm7mHs4VN\nTwptOTHzSQ4fLslJfrz/vhw/e7Yst2zJ/jPzTSJhXYMhQ8KZBd4rbjzPQMMBFcUDY017WdzJJPCz\nn1nb+UjVzyf5HPwzkxjs2CGFrfz6aCzuZcskTd7M0dlXJJPAk09a2z15+vAi7CiZHFHhVuKFn3Av\nWCBV5exE4B/1/8jX4F8yCfz4x9b2nj3+oldWJm6Do0dFxP1qnIRNprrj+SLsKJkcUeFW4oWfcEf0\nH/X/yNfgXy7zaQKW1R2FQbre+K7CjpLJERVuJV6Ysq533eVe7jOi/6j/R74G/3IVPSPcUYhn7o3v\nqjeiZHJAhVuJD8kk8PvfW9tu/uuI/qOmkI/EplxFL0oWd2+FHkYtAxjQqBIlRgSNyOhJ6dhCIUgl\nPrdzBg/mtMqCfUkRfVfIIqqE5Pj80tjYyC3OYu+K0teUlIhEOSFKnaw3LiST4tPetEks7aYmb0sy\nmZSnE7tfvLw8GtZnkUBES5nZJwuq+1gVbiU2NDRYU4PZMUWdFG/02oVONsKtPm4lPhSC/zqqRD3a\nJmaocCvxIaoDTYVA1KNtYoYKtxIvolZqtlDQp5VIocKtKIo/+rQSKfr1dQMURSkQEgkV6oigFrei\nKEqBEUi4iWguEa0monVE9M2wG6UoiqJ44yvcRFQK4B4AFwCYCuBKIpoadsMURVEUd4JY3KcAWMfM\n65n5KIDfALgk3GYpiqIoXgQR7joArbbttu59iqIoSh+Qt6gSIpoPYH735n4iWp3jWw0HsCM/rSoY\n4thnIJ79jmOfgXj2O9s+1/sfIgQR7s0Axtm2x3bvS4GZmwE0B/1gL4ioJWi+frEQxz4D8ex3HPsM\nxLPfYfY5iKtkCYBJRDSBiAYAuALA733OURRFUULC1+Jm5g4iuhHAnwGUAniAmVeE3jJFURTFlUA+\nbmZ+BsAzIbfF0GN3SwESxz4D8ex3HPsMxLPfofU5lHrciqIoSnhoyruiKEqBERnhjktaPRGNI6KX\niOg9IlpBRDd17x9GRM8T0dru5dC+bmu+IaJSInqLiP7YvR2HPlcT0RNEtIqIVhLRnGLvNxHd0v3b\nXk5EvyaiQcXYZyJ6gIi2E9Fy2z7PfhLR7d36tpqIzu/JZ0dCuGOWVt8B4J+YeSqAUwHc0N3XbwJ4\ngZknAXihe7vYuAnAStt2HPp8N4DnmPkEACdD+l+0/SaiOgD/CKCRmU+EBDRcgeLs80MA5jr2ufaz\n+3/8CgDTus+5t1v3ciISwo0YpdUz81ZmfrN7fR/kH7kO0t+Huw97GMClfdPCcCCisQDmAfilbXex\n97kKwFkA7gcAZj7KzLtR5P2GBD2UEVE/AOUAtqAI+8zMLwPY5djt1c9LAPyGmY8w8wcA1kF0Lyei\nItyxTKsnogYAMwG8DmAkM2/tfmkbgJF91Kyw+BmAbwCwT6de7H2eAKAdwIPdLqJfElEFirjfzLwZ\nwE8AbAKwFcAeZv5/KOI+O/DqZ141LirCHTuIqBLA7wDczMx77a+xhPoUTbgPEV0EYDszL/U6ptj6\n3E0/AB8H8HNmngngABwugmLrd7dP9xLITWsMgAoiutp+TLH12Ysw+xkV4Q6UVl8sEFF/iGgnmfnJ\n7t0fEtHo7tdHA9jeV+0LgdMBXExEGyBusHOIaCGKu8+AWFVtzPx69/YTECEv5n6fB+ADZm5n5mMA\nngRwGoq7z3a8+plXjYuKcMcmrZ6ICOLzXMnMP7W99HsAX+xe/yKAp3u7bWHBzLcz81hmboB8ty8y\n89Uo4j4DADNvA9BKRJO7d50L4D0Ud783ATiViMq7f+vnQsZxirnPdrz6+XsAVxDRQCKaAGASgDdy\n/hRmjsQfgAsBrAHwPoAFfd2eEPt5BuTx6R0Ab3f/XQigBjIKvRbAXwAM6+u2htT/swH8sXu96PsM\nYAaAlu7v+38ADC32fgP4LoBVAJYDeATAwGLsM4BfQ/z4xyBPV1/J1E8AC7r1bTWAC3ry2Zo5qSiK\nUmBExVWiKIqiBESFW1EUpcBQ4VYURSkwVLgVRVEKDBVuRVGUAkOFWykKiGh/X7dBUXoLFW5FUZQC\nQ4VbKVqIqIGIXiSid4joBSIa373/8u5a0cuI6OXufdOI6A0ierv7+El923pF8UYTcJSigIj2M3Ol\nY98fADzBzA8T0ZcBXMzMlxLRuwDmMvNmIqpm5t1E9F8AFjNzsrvsQikzH+qDriiKL2pxK8XMHACP\ndq8/Aik3AACvAHiIiL4KKfQPAK8B+BYR3QagXkVbiTIq3ErsYOZrAdwBqda2lIhqmPlRABcDOATg\nGSI6py/bqCiZUOFWiplXIdUIASABYBEAENHHmPl1Zr4TMtHBOCI6DsB6Zv5PSEW36X3RYEUJgvq4\nlaKAiLogU2QZfgqpef4ggOEQgf4SM28ioichZTUJUsntZgC3AbgGUultG4CrmNk5LZWiRAIVbkVR\nlAJDXSWKoigFhgq3oihKgaHCrSiKUmCocCuKohQYKtyKoigFhgq3oihKgaHCrSiKUmCocCuKohQY\n/x+UIdTuvKPDdgAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f252c791160>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "batch_size = 10\n",
    "epochs = 100\n",
    "print_every = 10\n",
    "\n",
    "for i in range(epochs):\n",
    "    loss = 0\n",
    "    optimizer.zero_grad() # Important for each epoch\n",
    "    idx = np.random.randint(X.shape[0], size=batch_size)\n",
    "    batch = get_variable_from_np(X[idx])\n",
    "    target = get_variable_from_np(y[idx])\n",
    "    output = model.forward(batch)\n",
    "    loss += criterion(output, target)\n",
    "    loss.backward() # Calculate the gradients\n",
    "    optimizer.step() # Updates the parameters of the model\n",
    "    if (i+1) % print_every == 0:\n",
    "        print(\"Loss at epoch [%s]: %.3f\" % (i, loss.data[0]))\n",
    "    losses.append(loss.data[0])\n",
    "    \n",
    "plt.plot(losses, '-or')\n",
    "plt.xlabel(\"Epoch\")\n",
    "plt.xlabel(\"Loss\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x7f252c100ac8>"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXwAAAD8CAYAAAB0IB+mAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nEP4fnGzP9EiNRwAAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f252f931048>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "batch = get_variable_from_np(X[:])\n",
    "y_pred = model.forward(batch)\n",
    "y_pred_np = y_pred.squeeze().data.numpy()\n",
    "plt.scatter(X, y, s=1, label=\"data\")\n",
    "plt.scatter(X, y_true, s=1, color='r', label=\"true\")\n",
    "plt.scatter(X, y_pred_np, s=1, color='k', alpha=0.5, label=\"fit\")\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[Parameter containing:\n",
       "  0.5226\n",
       " [torch.FloatTensor of size 1x1], Parameter containing:\n",
       " -1.5220\n",
       " [torch.FloatTensor of size 1]]"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "list(model.x2o.parameters())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Parameter containing:\n",
       " 0.5226\n",
       "[torch.FloatTensor of size 1x1]"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.x2o.weight"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Parameter containing:\n",
       "-1.5220\n",
       "[torch.FloatTensor of size 1]"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.x2o.bias"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model W: 0.523, True W: 0.500\n",
      "Model bias: -1.522, True bias: -1.680\n"
     ]
    }
   ],
   "source": [
    "print(\"Model W: %.3f, True W: %.3f\" % (model.x2o.weight.data.numpy(), W))\n",
    "print(\"Model bias: %.3f, True bias: %.3f\" % (model.x2o.bias.data.numpy(), bias))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "## Running on CUDA"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "batch = Variable(torch.randn(10,3))\n",
    "target = Variable(torch.randn(10,1))\n",
    "\n",
    "no_gpu_model = LinearRegression(input_size=3, output_size=1)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "torch.Size([10, 1])"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "no_gpu_model.forward(batch).size()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "True"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "torch.cuda.is_available()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "addmm_ received an invalid combination of arguments - got (int, int, torch.FloatTensor, torch.cuda.FloatTensor), but expected one of:\n",
      " * (torch.FloatTensor mat1, torch.FloatTensor mat2)\n",
      " * (torch.SparseFloatTensor mat1, torch.FloatTensor mat2)\n",
      " * (float beta, torch.FloatTensor mat1, torch.FloatTensor mat2)\n",
      " * (float alpha, torch.FloatTensor mat1, torch.FloatTensor mat2)\n",
      " * (float beta, torch.SparseFloatTensor mat1, torch.FloatTensor mat2)\n",
      " * (float alpha, torch.SparseFloatTensor mat1, torch.FloatTensor mat2)\n",
      " * (float beta, float alpha, torch.FloatTensor mat1, torch.FloatTensor mat2)\n",
      "      didn't match because some of the arguments have invalid types: (\u001b[32;1mint\u001b[0m, \u001b[32;1mint\u001b[0m, \u001b[32;1mtorch.FloatTensor\u001b[0m, \u001b[31;1mtorch.cuda.FloatTensor\u001b[0m)\n",
      " * (float beta, float alpha, torch.SparseFloatTensor mat1, torch.FloatTensor mat2)\n",
      "      didn't match because some of the arguments have invalid types: (\u001b[32;1mint\u001b[0m, \u001b[32;1mint\u001b[0m, \u001b[31;1mtorch.FloatTensor\u001b[0m, \u001b[31;1mtorch.cuda.FloatTensor\u001b[0m)\n",
      "\n"
     ]
    }
   ],
   "source": [
    "if torch.cuda.is_available():\n",
    "    gpu_model = no_gpu_model.cuda()\n",
    "    try:\n",
    "        print(gpu_model.forward(batch))\n",
    "    except TypeError as e:\n",
    "        print(e)        "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "I have opened an issue related to the above error at: https://github.com/pytorch/pytorch/issues/584"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "torch.Size([10, 1])\n"
     ]
    }
   ],
   "source": [
    "if torch.cuda.is_available():\n",
    "    gpu_model = no_gpu_model.cuda()\n",
    "    try:\n",
    "        print(gpu_model.forward(batch.cuda()).size())\n",
    "    except TypeError as e:\n",
    "        print(e)   "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "GPU to CPU fallback supported model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "class LinearRegression(nn.Module):\n",
    "    def __init__(self, input_size, output_size):\n",
    "        super(LinearRegression, self).__init__()\n",
    "        self.x2o = nn.Linear(input_size, output_size)\n",
    "        \n",
    "    def forward(self, X):\n",
    "        if next(self.x2o.parameters()).is_cuda:\n",
    "            if not X.is_cuda:\n",
    "                X = X.cuda()\n",
    "        return self.x2o(X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "No GPU model:  torch.Size([10, 1])\n",
      "GPU model:  torch.Size([10, 1])\n"
     ]
    }
   ],
   "source": [
    "batch = Variable(torch.randn(10,3))\n",
    "target = Variable(torch.randn(10,1))\n",
    "\n",
    "no_gpu_model = LinearRegression(input_size=3, output_size=1)\n",
    "print(\"No GPU model: \", no_gpu_model.forward(batch).size())\n",
    "\n",
    "if torch.cuda.is_available():\n",
    "    gpu_model = no_gpu_model.cuda()\n",
    "    try:\n",
    "        print(\"GPU model: \", gpu_model.forward(batch.cuda()).size())\n",
    "    except TypeError as e:\n",
    "        print(e)   "
   ]
  },
  {
   "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": "3.5.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}