Mnist test network.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 143,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "from torch.autograd import Variable\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "import torch.optim as optim\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# User defined functions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def CalcPooling(input_dim,kernel_dim,stride,dilation=1,padding=0):\n",
    "    result = int(((input_dim+2*padding-dilation*(kernel_dim-1)-1)/stride)+1)\n",
    "    return result"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def CalcDimensionality(input_dim,kernel_dim,stride,padding=0):\n",
    "    result = int(((input_dim-kernel_dim+2*padding)/stride)+1)\n",
    "    return result"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Load the data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Information on dataset\n",
      "x_train (1000, 1, 28, 28)\n",
      "targets_train (1000,)\n",
      "x_valid (500, 1, 28, 28)\n",
      "targets_valid (500,)\n",
      "x_test (500, 1, 28, 28)\n",
      "targets_test (500,)\n"
     ]
    }
   ],
   "source": [
    "## LOAD the mnist data. \n",
    "\n",
    "# To speed up training we'll only work on a subset of the data.\n",
    "# Note that we reshape the data from \n",
    "#  (nsamples, num_features) = (nsamples, nchannels*rows*cols)\n",
    "#    -> (nsamples, nchannels, rows, cols)\n",
    "# in order to retain the spatial arrangements of the pixels\n",
    "data = np.load('mnist.npz')\n",
    "num_classes = 10\n",
    "nchannels, rows, cols = 1, 28, 28\n",
    "\n",
    "x_train = data['X_train'][:1000].astype('float32')\n",
    "x_train = x_train.reshape((-1, nchannels, rows, cols))\n",
    "targets_train = data['y_train'][:1000].astype('int32')\n",
    "\n",
    "x_valid = data['X_valid'][:500].astype('float32')\n",
    "x_valid = x_valid.reshape((-1, nchannels, rows, cols))\n",
    "targets_valid = data['y_valid'][:500].astype('int32')\n",
    "\n",
    "x_test = data['X_test'][:500].astype('float32')\n",
    "x_test = x_test.reshape((-1, nchannels, rows, cols))\n",
    "targets_test = data['y_test'][:500].astype('int32')\n",
    "\n",
    "print(\"Information on dataset\")\n",
    "print(\"x_train\", x_train.shape)\n",
    "print(\"targets_train\", targets_train.shape)\n",
    "print(\"x_valid\", x_valid.shape)\n",
    "print(\"targets_valid\", targets_valid.shape)\n",
    "print(\"x_test\", x_test.shape)\n",
    "print(\"targets_test\", targets_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXIAAAF1CAYAAAD1IWGxAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nDTmHw+EYONyQczgcjoHzf1owm3f9Cym3AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f672a4d8f60>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "## plot a few MNIST examples\n",
    "\n",
    "idx, dim, classes = 0, 28, 10\n",
    "# create empty canvas\n",
    "canvas = np.zeros((dim*classes, classes*dim))\n",
    "\n",
    "# fill with tensors\n",
    "for i in range(classes):\n",
    "    for j in range(classes):\n",
    "        canvas[i*dim:(i+1)*dim, j*dim:(j+1)*dim] = x_train[idx].reshape((dim, dim))\n",
    "        idx += 1\n",
    "\n",
    "# visualize matrix of tensors as gray scale image\n",
    "plt.figure(figsize=(6, 6))\n",
    "plt.axis('off')\n",
    "plt.imshow(canvas, cmap='gray')\n",
    "plt.title('MNIST handwritten digits')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Define network"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Net(\n",
      "  (fc1): Linear(in_features=784, 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",
      ")\n"
     ]
    }
   ],
   "source": [
    "class Net(nn.Module):\n",
    "\n",
    "    def __init__(self):\n",
    "        super(Net, self).__init__()\n",
    "        \n",
    "        # an affine operation: y = Wx + b\n",
    "        self.fc1 = nn.Linear(28*28, 120)\n",
    "        self.fc2 = nn.Linear(120, 84)\n",
    "        self.fc3 = nn.Linear(84, 10)\n",
    "\n",
    "    def forward(self, x):\n",
    "        x = x.view(-1, self.num_flat_features(x))\n",
    "        x = F.relu(self.fc1(x))\n",
    "        x = F.relu(self.fc2(x))\n",
    "        x = F.relu(self.fc3(x))\n",
    "        return x\n",
    "\n",
    "    def num_flat_features(self, x):\n",
    "        size = x.size()[1:]  # all dimensions except the batch dimension\n",
    "        num_features = 1\n",
    "        for s in size:\n",
    "            num_features *= s\n",
    "        return num_features\n",
    "\n",
    "\n",
    "net = Net()\n",
    "print(net)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Test the network with an input"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAEICAYAAACQ6CLfAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMS4wLCBo\ndHRwOi8vbWF0cGxvdGxpYi5vcmcvpW3flQAAEJhJREFUeJzt3XusHPV5xvHvExtEMQbjUoxr7Dh2\nkamh1KmMiQgKIOpwEQjMTbGK6gqKUwm3REqtIEsVpq0pCpcmFiiyEy4GEi4KUAxFAYQBp0F1OdyC\nY0qCEBfDqQ0xBxtzi33e/rHj6GDO/vac3dmd9fk9H2l1dved2Xm98rMzszOzP0UEZpafL1TdgJlV\nw+E3y5TDb5Yph98sUw6/WaYcfrNMOfyZkPSEpL8te15JSyT9qLXurAoO/x5G0muS/rLqPnaJiCsj\nYtgfKsUHyseSPihuL7ejP6vP4bcqLYqI/YrbjKqbyY3DP0JIOlDSg5LekfRecf/Q3SabLul/JL0v\n6X5J4wfM/xVJT0nqk/SCpBOGuNylkm4v7u8j6XZJvy1e52lJE8r7V1qZHP6R4wvAzcAXgSnAR8D1\nu03z18CFwB8DO4DlAJImAf8J/CswHvhH4B5JfzTMHhYABwCTgT8E/q7oo55/k/SupF8M9cPGyuPw\njxAR8duIuCciPoyIbcAy4PjdJrstItZHxHbgn4DzJY0CLgAeioiHIqI/Ih4FeoDThtnG76iF/k8i\nYmdEPBMRW+tM+x1gGjAJWAk8IGn6MJdnLXD4RwhJ+0paIel1SVuBtcC4Ity7vDng/uvAXsBB1LYW\nzis21fsk9QHHAROH2cZtwMPAnZLelvRdSXsNNmFErIuIbRHxSUSsAn7B8D9srAUO/8jxbWAGcExE\n7A98rXheA6aZPOD+FGpr6nepfSjcFhHjBtzGRMRVw2kgIn4XEVdExEzgWOB0arsaQ5p9t16tzRz+\nPdNexZdru26jgbHU9q/7ii/yLh9kvgskzZS0L/DPwE8jYidwO3CGpJMljSpe84RBvjBMknSipD8r\ntja2Uvtw2TnIdOOKZe0jabSkv6L2YfXwcJZnrXH490wPUQv6rttS4HvAH1Bbk/838LNB5rsNuAX4\nP2Af4B8AIuJN4ExgCfAOtS2BxQz//8chwE+pBf8l4ElqHyy724val4vvFP3+PXBWRPhYfwfJP+Zh\nliev+c0y5fCbZcrhN8uUw2+WqdGdXJgkf7to1mYRMaTzJVpa80s6RdLLkl6RdFkrr2VmndX0ob7i\nRI5fA3OBjcDTwPyI2JCYx2t+szbrxJp/DvBKRLwaEZ8Cd1I7UcTM9gCthH8Sn71QZGPx3GdIWiip\nR1JPC8sys5K18oXfYJsWn9usj4iV1C7Z9Ga/WRdpZc2/kc9eJXYo8HZr7ZhZp7QS/qeBwyR9SdLe\nwDeA1eW0ZWbt1vRmf0TskLSI2mWYo4CbIuJXpXVmZm3V0av6vM9v1n4dOcnHzPZcDr9Zphx+s0w5\n/GaZcvjNMuXwm2XK4TfLlMNvlimH3yxTDr9Zphx+s0w5/GaZcvjNMuXwm2XK4TfLlMNvlimH3yxT\nDr9Zphx+s0w5/GaZcvjNMuXwm2XK4TfLlMNvlimH3yxTDr9Zphx+s0w5/GaZcvjNMtX0EN22Zxg1\nalSyfsABB7R1+YsWLapb23fffZPzzpgxI1m/5JJLkvVrrrmmbm3+/PnJeT/++ONk/aqrrkrWr7ji\nimS9G7QUfkmvAduAncCOiJhdRlNm1n5lrPlPjIh3S3gdM+sg7/ObZarV8AfwiKRnJC0cbAJJCyX1\nSOppcVlmVqJWN/u/GhFvSzoYeFTS/0bE2oETRMRKYCWApGhxeWZWkpbW/BHxdvF3M3AfMKeMpsys\n/ZoOv6Qxksbuug98HVhfVmNm1l6tbPZPAO6TtOt1fhIRPyulqxFmypQpyfree++drB977LHJ+nHH\nHVe3Nm7cuOS855xzTrJepY0bNybry5cvT9bnzZtXt7Zt27bkvC+88EKy/uSTTybre4Kmwx8RrwJ/\nXmIvZtZBPtRnlimH3yxTDr9Zphx+s0w5/GaZUkTnTrobqWf4zZo1K1lfs2ZNst7uy2q7VX9/f7J+\n4YUXJusffPBB08vu7e1N1t97771k/eWXX2562e0WERrKdF7zm2XK4TfLlMNvlimH3yxTDr9Zphx+\ns0w5/GaZ8nH+EowfPz5ZX7duXbI+bdq0MtspVaPe+/r6kvUTTzyxbu3TTz9Nzpvr+Q+t8nF+M0ty\n+M0y5fCbZcrhN8uUw2+WKYffLFMOv1mmPER3CbZs2ZKsL168OFk//fTTk/XnnnsuWW/0E9Ypzz//\nfLI+d+7cZH379u3J+hFHHFG3dumllybntfbymt8sUw6/WaYcfrNMOfxmmXL4zTLl8JtlyuE3y5Sv\n5+8C+++/f7LeaDjpFStW1K1ddNFFyXkvuOCCZP2OO+5I1q37lHY9v6SbJG2WtH7Ac+MlPSrpN8Xf\nA1tp1sw6byib/bcAp+z23GXAYxFxGPBY8djM9iANwx8Ra4Hdz189E1hV3F8FnFVyX2bWZs2e2z8h\nInoBIqJX0sH1JpS0EFjY5HLMrE3afmFPRKwEVoK/8DPrJs0e6tskaSJA8XdzeS2ZWSc0G/7VwILi\n/gLg/nLaMbNOabjZL+kO4ATgIEkbgcuBq4C7JV0EvAGc184mR7qtW7e2NP/777/f9LwXX3xxsn7X\nXXcl6/39/U0v26rVMPwRMb9O6aSSezGzDvLpvWaZcvjNMuXwm2XK4TfLlMNvlilf0jsCjBkzpm7t\ngQceSM57/PHHJ+unnnpqsv7II48k69Z5HqLbzJIcfrNMOfxmmXL4zTLl8JtlyuE3y5TDb5YpH+cf\n4aZPn56sP/vss8l6X19fsv74448n6z09PXVrN9xwQ3LeTv7fHEl8nN/Mkhx+s0w5/GaZcvjNMuXw\nm2XK4TfLlMNvlikf58/cvHnzkvWbb745WR87dmzTy16yZEmyfuuttybrvb29TS97JPNxfjNLcvjN\nMuXwm2XK4TfLlMNvlimH3yxTDr9Zpnyc35KOPPLIZP26665L1k86qfnBnFesWJGsL1u2LFl/6623\nml72nqy04/ySbpK0WdL6Ac8tlfSWpOeL22mtNGtmnTeUzf5bgFMGef7fI2JWcXuo3LbMrN0ahj8i\n1gJbOtCLmXVQK1/4LZL0y2K34MB6E0laKKlHUv0fczOzjms2/D8ApgOzgF7g2noTRsTKiJgdEbOb\nXJaZtUFT4Y+ITRGxMyL6gR8Cc8pty8zaranwS5o44OE8YH29ac2sOzU8zi/pDuAE4CBgE3B58XgW\nEMBrwDcjouHF1T7OP/KMGzcuWT/jjDPq1hr9VoCUPly9Zs2aZH3u3LnJ+kg11OP8o4fwQvMHefrG\nYXdkZl3Fp/eaZcrhN8uUw2+WKYffLFMOv1mmfEmvVeaTTz5J1kePTh+M2rFjR7J+8skn16098cQT\nyXn3ZP7pbjNLcvjNMuXwm2XK4TfLlMNvlimH3yxTDr9Zphpe1Wd5O+qoo5L1c889N1k/+uij69Ya\nHcdvZMOGDcn62rVrW3r9kc5rfrNMOfxmmXL4zTLl8JtlyuE3y5TDb5Yph98sUz7OP8LNmDEjWV+0\naFGyfvbZZyfrhxxyyLB7GqqdO3cm67296V+L7+/vL7OdEcdrfrNMOfxmmXL4zTLl8JtlyuE3y5TD\nb5Yph98sUw2P80uaDNwKHAL0Aysj4vuSxgN3AVOpDdN9fkS8175W89XoWPr8+YMNpFzT6Dj+1KlT\nm2mpFD09Pcn6smXLkvXVq1eX2U52hrLm3wF8OyL+FPgKcImkmcBlwGMRcRjwWPHYzPYQDcMfEb0R\n8WxxfxvwEjAJOBNYVUy2CjirXU2aWfmGtc8vaSrwZWAdMCEieqH2AQEcXHZzZtY+Qz63X9J+wD3A\ntyJiqzSk4cCQtBBY2Fx7ZtYuQ1rzS9qLWvB/HBH3Fk9vkjSxqE8ENg82b0SsjIjZETG7jIbNrBwN\nw6/aKv5G4KWIuG5AaTWwoLi/ALi//PbMrF0aDtEt6Tjg58CL1A71ASyhtt9/NzAFeAM4LyK2NHit\nLIfonjBhQrI+c+bMZP36669P1g8//PBh91SWdevWJetXX3113dr996fXF74ktzlDHaK74T5/RPwX\nUO/FThpOU2bWPXyGn1mmHH6zTDn8Zply+M0y5fCbZcrhN8uUf7p7iMaPH1+3tmLFiuS8s2bNStan\nTZvWVE9leOqpp5L1a6+9Nll/+OGHk/WPPvpo2D1ZZ3jNb5Yph98sUw6/WaYcfrNMOfxmmXL4zTLl\n8JtlKpvj/Mccc0yyvnjx4mR9zpw5dWuTJk1qqqeyfPjhh3Vry5cvT8575ZVXJuvbt29vqifrfl7z\nm2XK4TfLlMNvlimH3yxTDr9Zphx+s0w5/GaZyuY4/7x581qqt2LDhg3J+oMPPpis79ixI1lPXXPf\n19eXnNfy5TW/WaYcfrNMOfxmmXL4zTLl8JtlyuE3y5TDb5YpRUR6AmkycCtwCNAPrIyI70taClwM\nvFNMuiQiHmrwWumFmVnLIkJDmW4o4Z8ITIyIZyWNBZ4BzgLOBz6IiGuG2pTDb9Z+Qw1/wzP8IqIX\n6C3ub5P0ElDtT9eYWcuGtc8vaSrwZWBd8dQiSb+UdJOkA+vMs1BSj6Seljo1s1I13Oz//YTSfsCT\nwLKIuFfSBOBdIIB/obZrcGGD1/Bmv1mblbbPDyBpL+BB4OGIuG6Q+lTgwYg4ssHrOPxmbTbU8Dfc\n7Jck4EbgpYHBL74I3GUesH64TZpZdYbybf9xwM+BF6kd6gNYAswHZlHb7H8N+Gbx5WDqtbzmN2uz\nUjf7y+Lwm7VfaZv9ZjYyOfxmmXL4zTLl8JtlyuE3y5TDb5Yph98sUw6/WaYcfrNMOfxmmXL4zTLl\n8JtlyuE3y5TDb5apTg/R/S7w+oDHBxXPdaNu7a1b+wL31qwye/viUCfs6PX8n1u41BMRsytrIKFb\ne+vWvsC9Nauq3rzZb5Yph98sU1WHf2XFy0/p1t66tS9wb82qpLdK9/nNrDpVr/nNrCIOv1mmKgm/\npFMkvSzpFUmXVdFDPZJek/SipOerHl+wGANxs6T1A54bL+lRSb8p/g46RmJFvS2V9Fbx3j0v6bSK\nepss6XFJL0n6laRLi+crfe8SfVXyvnV8n1/SKODXwFxgI/A0MD8iNnS0kTokvQbMjojKTwiR9DXg\nA+DWXUOhSfousCUirio+OA+MiO90SW9LGeaw7W3qrd6w8n9Dhe9dmcPdl6GKNf8c4JWIeDUiPgXu\nBM6soI+uFxFrgS27PX0msKq4v4raf56Oq9NbV4iI3oh4tri/Ddg1rHyl712ir0pUEf5JwJsDHm+k\nwjdgEAE8IukZSQurbmYQE3YNi1b8PbjifnbXcNj2TtptWPmuee+aGe6+bFWEf7ChhLrpeONXI+Iv\ngFOBS4rNWxuaHwDTqY3h2AtcW2UzxbDy9wDfioitVfYy0CB9VfK+VRH+jcDkAY8PBd6uoI9BRcTb\nxd/NwH3UdlO6yaZdIyQXfzdX3M/vRcSmiNgZEf3AD6nwvSuGlb8H+HFE3Fs8Xfl7N1hfVb1vVYT/\naeAwSV+StDfwDWB1BX18jqQxxRcxSBoDfJ3uG3p8NbCguL8AuL/CXj6jW4ZtrzesPBW/d9023H0l\nZ/gVhzK+B4wCboqIZR1vYhCSplFb20PtcuefVNmbpDuAE6hd8rkJuBz4D+BuYArwBnBeRHT8i7c6\nvZ3AMIdtb1Nv9YaVX0eF712Zw92X0o9P7zXLk8/wM8uUw2+WKYffLFMOv1mmHH6zTDn8Zply+M0y\n9f92qBJS5AjBuwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f675c01b0b8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Reshape the array into 28 x 28 array (2-dimensional array)\n",
    "pixels = x_train[0].reshape((28, 28))\n",
    "\n",
    "# Plot\n",
    "plt.title('Label is {label}'.format(label=targets_train[0]))\n",
    "plt.imshow(pixels, cmap='gray')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[0.0000, 0.0091, 0.0000, 0.0783, 0.0107, 0.0000, 0.1668, 0.0826, 0.0631,\n",
       "         0.0000]], grad_fn=<ReluBackward>)"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# change the numpy array data to tensor format\n",
    "x = torch.from_numpy(x_train[0])\n",
    "net(x)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Initialize the optimizer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 118,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "optimizer = optim.SGD(net.parameters(), lr=0.01)\n",
    "criterion = nn.CrossEntropyLoss()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Prepare input data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 149,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "inputs = torch.from_numpy(x_train)\n",
    "labels = torch.from_numpy(targets_train).long()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Train the network"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 120,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "get_slice = lambda i, size: range(i * size, (i + 1) * size)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 121,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "range(0, 100)"
      ]
     },
     "execution_count": 121,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "get_slice(0,100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 150,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1,    20] loss: 0.022\n",
      "[1,    40] loss: 0.019\n",
      "[1,    60] loss: 0.011\n",
      "[1,    80] loss: 0.008\n",
      "[1,   100] loss: 0.006\n",
      "[2,    20] loss: 0.008\n",
      "[2,    40] loss: 0.007\n",
      "[2,    60] loss: 0.007\n",
      "[2,    80] loss: 0.007\n",
      "[2,   100] loss: 0.006\n",
      "[3,    20] loss: 0.008\n",
      "[3,    40] loss: 0.007\n",
      "[3,    60] loss: 0.007\n",
      "[3,    80] loss: 0.007\n",
      "[3,   100] loss: 0.006\n",
      "[4,    20] loss: 0.008\n",
      "[4,    40] loss: 0.007\n",
      "[4,    60] loss: 0.007\n",
      "[4,    80] loss: 0.007\n",
      "[4,   100] loss: 0.006\n",
      "[5,    20] loss: 0.008\n",
      "[5,    40] loss: 0.007\n",
      "[5,    60] loss: 0.007\n",
      "[5,    80] loss: 0.007\n",
      "[5,   100] loss: 0.006\n",
      "[6,    20] loss: 0.008\n",
      "[6,    40] loss: 0.007\n",
      "[6,    60] loss: 0.007\n",
      "[6,    80] loss: 0.007\n",
      "[6,   100] loss: 0.006\n",
      "[7,    20] loss: 0.008\n",
      "[7,    40] loss: 0.007\n",
      "[7,    60] loss: 0.007\n",
      "[7,    80] loss: 0.007\n",
      "[7,   100] loss: 0.006\n",
      "[8,    20] loss: 0.008\n",
      "[8,    40] loss: 0.007\n",
      "[8,    60] loss: 0.007\n",
      "[8,    80] loss: 0.007\n",
      "[8,   100] loss: 0.006\n",
      "[9,    20] loss: 0.008\n",
      "[9,    40] loss: 0.007\n",
      "[9,    60] loss: 0.007\n",
      "[9,    80] loss: 0.007\n",
      "[9,   100] loss: 0.006\n",
      "[10,    20] loss: 0.008\n",
      "[10,    40] loss: 0.007\n",
      "[10,    60] loss: 0.007\n",
      "[10,    80] loss: 0.007\n",
      "[10,   100] loss: 0.006\n",
      "[11,    20] loss: 0.008\n",
      "[11,    40] loss: 0.007\n",
      "[11,    60] loss: 0.007\n",
      "[11,    80] loss: 0.007\n",
      "[11,   100] loss: 0.006\n",
      "[12,    20] loss: 0.008\n",
      "[12,    40] loss: 0.007\n",
      "[12,    60] loss: 0.007\n",
      "[12,    80] loss: 0.007\n",
      "[12,   100] loss: 0.006\n",
      "[13,    20] loss: 0.008\n",
      "[13,    40] loss: 0.007\n",
      "[13,    60] loss: 0.007\n",
      "[13,    80] loss: 0.007\n",
      "[13,   100] loss: 0.006\n",
      "[14,    20] loss: 0.008\n",
      "[14,    40] loss: 0.007\n",
      "[14,    60] loss: 0.007\n",
      "[14,    80] loss: 0.007\n",
      "[14,   100] loss: 0.006\n",
      "[15,    20] loss: 0.008\n",
      "[15,    40] loss: 0.007\n",
      "[15,    60] loss: 0.007\n",
      "[15,    80] loss: 0.007\n",
      "[15,   100] loss: 0.006\n",
      "[16,    20] loss: 0.008\n",
      "[16,    40] loss: 0.007\n",
      "[16,    60] loss: 0.007\n",
      "[16,    80] loss: 0.007\n",
      "[16,   100] loss: 0.006\n",
      "[17,    20] loss: 0.008\n",
      "[17,    40] loss: 0.007\n",
      "[17,    60] loss: 0.007\n",
      "[17,    80] loss: 0.007\n",
      "[17,   100] loss: 0.006\n",
      "[18,    20] loss: 0.008\n",
      "[18,    40] loss: 0.007\n",
      "[18,    60] loss: 0.007\n",
      "[18,    80] loss: 0.007\n",
      "[18,   100] loss: 0.006\n",
      "[19,    20] loss: 0.008\n",
      "[19,    40] loss: 0.007\n",
      "[19,    60] loss: 0.007\n",
      "[19,    80] loss: 0.007\n",
      "[19,   100] loss: 0.006\n",
      "[20,    20] loss: 0.008\n",
      "[20,    40] loss: 0.007\n",
      "[20,    60] loss: 0.007\n",
      "[20,    80] loss: 0.007\n",
      "[20,   100] loss: 0.006\n",
      "[21,    20] loss: 0.008\n",
      "[21,    40] loss: 0.007\n",
      "[21,    60] loss: 0.007\n",
      "[21,    80] loss: 0.007\n",
      "[21,   100] loss: 0.006\n",
      "[22,    20] loss: 0.008\n",
      "[22,    40] loss: 0.007\n",
      "[22,    60] loss: 0.007\n",
      "[22,    80] loss: 0.007\n",
      "[22,   100] loss: 0.006\n",
      "[23,    20] loss: 0.008\n",
      "[23,    40] loss: 0.007\n",
      "[23,    60] loss: 0.007\n",
      "[23,    80] loss: 0.007\n",
      "[23,   100] loss: 0.006\n",
      "[24,    20] loss: 0.008\n",
      "[24,    40] loss: 0.007\n",
      "[24,    60] loss: 0.007\n",
      "[24,    80] loss: 0.007\n",
      "[24,   100] loss: 0.006\n",
      "[25,    20] loss: 0.008\n",
      "[25,    40] loss: 0.007\n",
      "[25,    60] loss: 0.007\n",
      "[25,    80] loss: 0.007\n",
      "[25,   100] loss: 0.006\n",
      "[26,    20] loss: 0.008\n",
      "[26,    40] loss: 0.007\n",
      "[26,    60] loss: 0.007\n",
      "[26,    80] loss: 0.007\n",
      "[26,   100] loss: 0.006\n",
      "[27,    20] loss: 0.008\n",
      "[27,    40] loss: 0.007\n",
      "[27,    60] loss: 0.007\n",
      "[27,    80] loss: 0.007\n",
      "[27,   100] loss: 0.006\n",
      "[28,    20] loss: 0.008\n",
      "[28,    40] loss: 0.007\n",
      "[28,    60] loss: 0.007\n",
      "[28,    80] loss: 0.007\n",
      "[28,   100] loss: 0.006\n",
      "[29,    20] loss: 0.008\n",
      "[29,    40] loss: 0.007\n",
      "[29,    60] loss: 0.007\n",
      "[29,    80] loss: 0.007\n",
      "[29,   100] loss: 0.006\n",
      "[30,    20] loss: 0.008\n",
      "[30,    40] loss: 0.007\n",
      "[30,    60] loss: 0.007\n",
      "[30,    80] loss: 0.007\n",
      "[30,   100] loss: 0.006\n",
      "[31,    20] loss: 0.008\n",
      "[31,    40] loss: 0.007\n",
      "[31,    60] loss: 0.007\n",
      "[31,    80] loss: 0.007\n",
      "[31,   100] loss: 0.006\n",
      "[32,    20] loss: 0.008\n",
      "[32,    40] loss: 0.007\n",
      "[32,    60] loss: 0.007\n",
      "[32,    80] loss: 0.007\n",
      "[32,   100] loss: 0.006\n",
      "[33,    20] loss: 0.008\n",
      "[33,    40] loss: 0.007\n",
      "[33,    60] loss: 0.007\n",
      "[33,    80] loss: 0.007\n",
      "[33,   100] loss: 0.006\n",
      "[34,    20] loss: 0.008\n",
      "[34,    40] loss: 0.007\n",
      "[34,    60] loss: 0.007\n",
      "[34,    80] loss: 0.007\n",
      "[34,   100] loss: 0.006\n",
      "[35,    20] loss: 0.008\n",
      "[35,    40] loss: 0.007\n",
      "[35,    60] loss: 0.007\n",
      "[35,    80] loss: 0.007\n",
      "[35,   100] loss: 0.006\n",
      "[36,    20] loss: 0.008\n",
      "[36,    40] loss: 0.007\n",
      "[36,    60] loss: 0.007\n",
      "[36,    80] loss: 0.007\n",
      "[36,   100] loss: 0.006\n",
      "[37,    20] loss: 0.008\n",
      "[37,    40] loss: 0.007\n",
      "[37,    60] loss: 0.007\n",
      "[37,    80] loss: 0.007\n",
      "[37,   100] loss: 0.006\n",
      "[38,    20] loss: 0.008\n",
      "[38,    40] loss: 0.007\n",
      "[38,    60] loss: 0.007\n",
      "[38,    80] loss: 0.007\n",
      "[38,   100] loss: 0.006\n",
      "[39,    20] loss: 0.008\n",
      "[39,    40] loss: 0.007\n",
      "[39,    60] loss: 0.007\n",
      "[39,    80] loss: 0.007\n",
      "[39,   100] loss: 0.006\n",
      "[40,    20] loss: 0.008\n",
      "[40,    40] loss: 0.007\n",
      "[40,    60] loss: 0.007\n",
      "[40,    80] loss: 0.007\n",
      "[40,   100] loss: 0.006\n",
      "[41,    20] loss: 0.008\n",
      "[41,    40] loss: 0.007\n",
      "[41,    60] loss: 0.007\n",
      "[41,    80] loss: 0.007\n",
      "[41,   100] loss: 0.006\n",
      "[42,    20] loss: 0.008\n",
      "[42,    40] loss: 0.007\n",
      "[42,    60] loss: 0.007\n",
      "[42,    80] loss: 0.007\n",
      "[42,   100] loss: 0.006\n",
      "[43,    20] loss: 0.008\n",
      "[43,    40] loss: 0.007\n",
      "[43,    60] loss: 0.007\n",
      "[43,    80] loss: 0.007\n",
      "[43,   100] loss: 0.006\n",
      "[44,    20] loss: 0.008\n",
      "[44,    40] loss: 0.007\n",
      "[44,    60] loss: 0.007\n",
      "[44,    80] loss: 0.007\n",
      "[44,   100] loss: 0.006\n",
      "[45,    20] loss: 0.008\n",
      "[45,    40] loss: 0.007\n",
      "[45,    60] loss: 0.007\n",
      "[45,    80] loss: 0.007\n",
      "[45,   100] loss: 0.006\n",
      "[46,    20] loss: 0.008\n",
      "[46,    40] loss: 0.007\n",
      "[46,    60] loss: 0.007\n",
      "[46,    80] loss: 0.007\n",
      "[46,   100] loss: 0.006\n",
      "[47,    20] loss: 0.008\n",
      "[47,    40] loss: 0.007\n",
      "[47,    60] loss: 0.007\n",
      "[47,    80] loss: 0.007\n",
      "[47,   100] loss: 0.006\n",
      "[48,    20] loss: 0.008\n",
      "[48,    40] loss: 0.007\n",
      "[48,    60] loss: 0.007\n",
      "[48,    80] loss: 0.007\n",
      "[48,   100] loss: 0.006\n",
      "[49,    20] loss: 0.008\n",
      "[49,    40] loss: 0.007\n",
      "[49,    60] loss: 0.007\n",
      "[49,    80] loss: 0.007\n",
      "[49,   100] loss: 0.006\n",
      "[50,    20] loss: 0.008\n",
      "[50,    40] loss: 0.007\n",
      "[50,    60] loss: 0.007\n",
      "[50,    80] loss: 0.007\n",
      "[50,   100] loss: 0.006\n",
      "[51,    20] loss: 0.008\n",
      "[51,    40] loss: 0.007\n",
      "[51,    60] loss: 0.007\n",
      "[51,    80] loss: 0.007\n",
      "[51,   100] loss: 0.006\n",
      "[52,    20] loss: 0.008\n",
      "[52,    40] loss: 0.007\n",
      "[52,    60] loss: 0.007\n",
      "[52,    80] loss: 0.007\n",
      "[52,   100] loss: 0.006\n",
      "[53,    20] loss: 0.008\n",
      "[53,    40] loss: 0.007\n",
      "[53,    60] loss: 0.007\n",
      "[53,    80] loss: 0.007\n",
      "[53,   100] loss: 0.006\n",
      "[54,    20] loss: 0.008\n",
      "[54,    40] loss: 0.007\n",
      "[54,    60] loss: 0.007\n",
      "[54,    80] loss: 0.007\n",
      "[54,   100] loss: 0.006\n",
      "[55,    20] loss: 0.008\n",
      "[55,    40] loss: 0.007\n",
      "[55,    60] loss: 0.007\n",
      "[55,    80] loss: 0.007\n",
      "[55,   100] loss: 0.006\n",
      "[56,    20] loss: 0.008\n",
      "[56,    40] loss: 0.007\n",
      "[56,    60] loss: 0.007\n",
      "[56,    80] loss: 0.007\n",
      "[56,   100] loss: 0.006\n",
      "[57,    20] loss: 0.008\n",
      "[57,    40] loss: 0.007\n",
      "[57,    60] loss: 0.007\n",
      "[57,    80] loss: 0.007\n",
      "[57,   100] loss: 0.006\n",
      "[58,    20] loss: 0.008\n",
      "[58,    40] loss: 0.007\n",
      "[58,    60] loss: 0.007\n",
      "[58,    80] loss: 0.007\n",
      "[58,   100] loss: 0.006\n",
      "[59,    20] loss: 0.008\n",
      "[59,    40] loss: 0.007\n",
      "[59,    60] loss: 0.007\n",
      "[59,    80] loss: 0.007\n",
      "[59,   100] loss: 0.006\n",
      "[60,    20] loss: 0.008\n",
      "[60,    40] loss: 0.007\n",
      "[60,    60] loss: 0.007\n",
      "[60,    80] loss: 0.007\n",
      "[60,   100] loss: 0.006\n",
      "[61,    20] loss: 0.008\n",
      "[61,    40] loss: 0.007\n",
      "[61,    60] loss: 0.007\n",
      "[61,    80] loss: 0.007\n",
      "[61,   100] loss: 0.006\n",
      "[62,    20] loss: 0.008\n",
      "[62,    40] loss: 0.007\n",
      "[62,    60] loss: 0.007\n",
      "[62,    80] loss: 0.007\n",
      "[62,   100] loss: 0.006\n",
      "[63,    20] loss: 0.008\n",
      "[63,    40] loss: 0.007\n",
      "[63,    60] loss: 0.007\n",
      "[63,    80] loss: 0.007\n",
      "[63,   100] loss: 0.006\n",
      "[64,    20] loss: 0.008\n",
      "[64,    40] loss: 0.007\n",
      "[64,    60] loss: 0.007\n",
      "[64,    80] loss: 0.007\n",
      "[64,   100] loss: 0.006\n",
      "[65,    20] loss: 0.008\n",
      "[65,    40] loss: 0.007\n",
      "[65,    60] loss: 0.007\n",
      "[65,    80] loss: 0.007\n",
      "[65,   100] loss: 0.006\n",
      "[66,    20] loss: 0.008\n",
      "[66,    40] loss: 0.007\n",
      "[66,    60] loss: 0.007\n",
      "[66,    80] loss: 0.007\n",
      "[66,   100] loss: 0.006\n",
      "[67,    20] loss: 0.008\n",
      "[67,    40] loss: 0.007\n",
      "[67,    60] loss: 0.007\n",
      "[67,    80] loss: 0.007\n",
      "[67,   100] loss: 0.006\n",
      "[68,    20] loss: 0.008\n",
      "[68,    40] loss: 0.007\n",
      "[68,    60] loss: 0.007\n",
      "[68,    80] loss: 0.007\n",
      "[68,   100] loss: 0.006\n",
      "[69,    20] loss: 0.008\n",
      "[69,    40] loss: 0.007\n",
      "[69,    60] loss: 0.007\n",
      "[69,    80] loss: 0.007\n",
      "[69,   100] loss: 0.006\n",
      "[70,    20] loss: 0.008\n",
      "[70,    40] loss: 0.007\n",
      "[70,    60] loss: 0.007\n",
      "[70,    80] loss: 0.007\n",
      "[70,   100] loss: 0.006\n",
      "[71,    20] loss: 0.008\n",
      "[71,    40] loss: 0.007\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[71,    60] loss: 0.007\n",
      "[71,    80] loss: 0.007\n",
      "[71,   100] loss: 0.006\n",
      "[72,    20] loss: 0.008\n",
      "[72,    40] loss: 0.007\n",
      "[72,    60] loss: 0.007\n",
      "[72,    80] loss: 0.007\n",
      "[72,   100] loss: 0.006\n",
      "[73,    20] loss: 0.008\n",
      "[73,    40] loss: 0.007\n",
      "[73,    60] loss: 0.007\n",
      "[73,    80] loss: 0.007\n",
      "[73,   100] loss: 0.006\n",
      "[74,    20] loss: 0.008\n",
      "[74,    40] loss: 0.007\n",
      "[74,    60] loss: 0.007\n",
      "[74,    80] loss: 0.007\n",
      "[74,   100] loss: 0.006\n",
      "[75,    20] loss: 0.008\n",
      "[75,    40] loss: 0.007\n",
      "[75,    60] loss: 0.007\n",
      "[75,    80] loss: 0.007\n",
      "[75,   100] loss: 0.006\n",
      "[76,    20] loss: 0.008\n",
      "[76,    40] loss: 0.007\n",
      "[76,    60] loss: 0.007\n",
      "[76,    80] loss: 0.007\n",
      "[76,   100] loss: 0.006\n",
      "[77,    20] loss: 0.008\n",
      "[77,    40] loss: 0.007\n",
      "[77,    60] loss: 0.007\n",
      "[77,    80] loss: 0.007\n",
      "[77,   100] loss: 0.006\n",
      "[78,    20] loss: 0.008\n",
      "[78,    40] loss: 0.007\n",
      "[78,    60] loss: 0.007\n",
      "[78,    80] loss: 0.007\n",
      "[78,   100] loss: 0.006\n",
      "[79,    20] loss: 0.008\n",
      "[79,    40] loss: 0.007\n",
      "[79,    60] loss: 0.007\n",
      "[79,    80] loss: 0.007\n",
      "[79,   100] loss: 0.006\n",
      "[80,    20] loss: 0.008\n",
      "[80,    40] loss: 0.007\n",
      "[80,    60] loss: 0.007\n",
      "[80,    80] loss: 0.007\n",
      "[80,   100] loss: 0.006\n",
      "[81,    20] loss: 0.008\n",
      "[81,    40] loss: 0.007\n",
      "[81,    60] loss: 0.007\n",
      "[81,    80] loss: 0.007\n",
      "[81,   100] loss: 0.006\n",
      "[82,    20] loss: 0.008\n",
      "[82,    40] loss: 0.007\n",
      "[82,    60] loss: 0.007\n",
      "[82,    80] loss: 0.007\n",
      "[82,   100] loss: 0.006\n",
      "[83,    20] loss: 0.008\n",
      "[83,    40] loss: 0.007\n",
      "[83,    60] loss: 0.007\n",
      "[83,    80] loss: 0.007\n",
      "[83,   100] loss: 0.006\n",
      "[84,    20] loss: 0.008\n",
      "[84,    40] loss: 0.007\n",
      "[84,    60] loss: 0.007\n",
      "[84,    80] loss: 0.007\n",
      "[84,   100] loss: 0.006\n",
      "[85,    20] loss: 0.008\n",
      "[85,    40] loss: 0.007\n",
      "[85,    60] loss: 0.007\n",
      "[85,    80] loss: 0.007\n",
      "[85,   100] loss: 0.006\n",
      "[86,    20] loss: 0.008\n",
      "[86,    40] loss: 0.007\n",
      "[86,    60] loss: 0.007\n",
      "[86,    80] loss: 0.007\n",
      "[86,   100] loss: 0.006\n",
      "[87,    20] loss: 0.008\n",
      "[87,    40] loss: 0.007\n",
      "[87,    60] loss: 0.007\n",
      "[87,    80] loss: 0.007\n",
      "[87,   100] loss: 0.006\n",
      "[88,    20] loss: 0.008\n",
      "[88,    40] loss: 0.007\n",
      "[88,    60] loss: 0.007\n",
      "[88,    80] loss: 0.007\n",
      "[88,   100] loss: 0.006\n",
      "[89,    20] loss: 0.008\n",
      "[89,    40] loss: 0.007\n",
      "[89,    60] loss: 0.007\n",
      "[89,    80] loss: 0.007\n",
      "[89,   100] loss: 0.006\n",
      "[90,    20] loss: 0.008\n",
      "[90,    40] loss: 0.007\n",
      "[90,    60] loss: 0.007\n",
      "[90,    80] loss: 0.007\n",
      "[90,   100] loss: 0.006\n",
      "[91,    20] loss: 0.008\n",
      "[91,    40] loss: 0.007\n",
      "[91,    60] loss: 0.007\n",
      "[91,    80] loss: 0.007\n",
      "[91,   100] loss: 0.006\n",
      "[92,    20] loss: 0.008\n",
      "[92,    40] loss: 0.007\n",
      "[92,    60] loss: 0.007\n",
      "[92,    80] loss: 0.007\n",
      "[92,   100] loss: 0.006\n",
      "[93,    20] loss: 0.008\n",
      "[93,    40] loss: 0.007\n",
      "[93,    60] loss: 0.007\n",
      "[93,    80] loss: 0.007\n",
      "[93,   100] loss: 0.006\n",
      "[94,    20] loss: 0.008\n",
      "[94,    40] loss: 0.007\n",
      "[94,    60] loss: 0.007\n",
      "[94,    80] loss: 0.007\n",
      "[94,   100] loss: 0.006\n",
      "[95,    20] loss: 0.008\n",
      "[95,    40] loss: 0.007\n",
      "[95,    60] loss: 0.007\n",
      "[95,    80] loss: 0.007\n",
      "[95,   100] loss: 0.006\n",
      "[96,    20] loss: 0.008\n",
      "[96,    40] loss: 0.007\n",
      "[96,    60] loss: 0.007\n",
      "[96,    80] loss: 0.007\n",
      "[96,   100] loss: 0.006\n",
      "[97,    20] loss: 0.008\n",
      "[97,    40] loss: 0.007\n",
      "[97,    60] loss: 0.007\n",
      "[97,    80] loss: 0.007\n",
      "[97,   100] loss: 0.006\n",
      "[98,    20] loss: 0.008\n",
      "[98,    40] loss: 0.007\n",
      "[98,    60] loss: 0.007\n",
      "[98,    80] loss: 0.007\n",
      "[98,   100] loss: 0.006\n",
      "[99,    20] loss: 0.008\n",
      "[99,    40] loss: 0.007\n",
      "[99,    60] loss: 0.007\n",
      "[99,    80] loss: 0.007\n",
      "[99,   100] loss: 0.006\n",
      "[100,    20] loss: 0.008\n",
      "[100,    40] loss: 0.007\n",
      "[100,    60] loss: 0.007\n",
      "[100,    80] loss: 0.007\n",
      "[100,   100] loss: 0.006\n",
      "[101,    20] loss: 0.008\n",
      "[101,    40] loss: 0.007\n",
      "[101,    60] loss: 0.007\n",
      "[101,    80] loss: 0.007\n",
      "[101,   100] loss: 0.006\n",
      "[102,    20] loss: 0.008\n",
      "[102,    40] loss: 0.007\n",
      "[102,    60] loss: 0.007\n",
      "[102,    80] loss: 0.007\n",
      "[102,   100] loss: 0.006\n",
      "[103,    20] loss: 0.008\n",
      "[103,    40] loss: 0.007\n",
      "[103,    60] loss: 0.007\n",
      "[103,    80] loss: 0.007\n",
      "[103,   100] loss: 0.006\n",
      "[104,    20] loss: 0.008\n",
      "[104,    40] loss: 0.007\n",
      "[104,    60] loss: 0.007\n",
      "[104,    80] loss: 0.007\n",
      "[104,   100] loss: 0.006\n",
      "[105,    20] loss: 0.008\n",
      "[105,    40] loss: 0.007\n",
      "[105,    60] loss: 0.007\n",
      "[105,    80] loss: 0.007\n",
      "[105,   100] loss: 0.006\n",
      "[106,    20] loss: 0.008\n",
      "[106,    40] loss: 0.007\n",
      "[106,    60] loss: 0.007\n",
      "[106,    80] loss: 0.007\n",
      "[106,   100] loss: 0.006\n",
      "[107,    20] loss: 0.008\n",
      "[107,    40] loss: 0.007\n",
      "[107,    60] loss: 0.007\n",
      "[107,    80] loss: 0.007\n",
      "[107,   100] loss: 0.006\n",
      "[108,    20] loss: 0.008\n",
      "[108,    40] loss: 0.007\n",
      "[108,    60] loss: 0.007\n",
      "[108,    80] loss: 0.007\n",
      "[108,   100] loss: 0.006\n",
      "[109,    20] loss: 0.008\n",
      "[109,    40] loss: 0.007\n",
      "[109,    60] loss: 0.007\n",
      "[109,    80] loss: 0.007\n",
      "[109,   100] loss: 0.006\n",
      "[110,    20] loss: 0.008\n",
      "[110,    40] loss: 0.007\n",
      "[110,    60] loss: 0.007\n",
      "[110,    80] loss: 0.007\n",
      "[110,   100] loss: 0.006\n",
      "[111,    20] loss: 0.008\n",
      "[111,    40] loss: 0.007\n",
      "[111,    60] loss: 0.007\n",
      "[111,    80] loss: 0.007\n",
      "[111,   100] loss: 0.006\n",
      "[112,    20] loss: 0.008\n",
      "[112,    40] loss: 0.007\n",
      "[112,    60] loss: 0.007\n",
      "[112,    80] loss: 0.007\n",
      "[112,   100] loss: 0.006\n",
      "[113,    20] loss: 0.008\n",
      "[113,    40] loss: 0.007\n",
      "[113,    60] loss: 0.007\n",
      "[113,    80] loss: 0.007\n",
      "[113,   100] loss: 0.006\n",
      "[114,    20] loss: 0.008\n",
      "[114,    40] loss: 0.007\n",
      "[114,    60] loss: 0.007\n",
      "[114,    80] loss: 0.007\n",
      "[114,   100] loss: 0.006\n",
      "[115,    20] loss: 0.008\n",
      "[115,    40] loss: 0.007\n",
      "[115,    60] loss: 0.007\n",
      "[115,    80] loss: 0.007\n",
      "[115,   100] loss: 0.006\n",
      "[116,    20] loss: 0.008\n",
      "[116,    40] loss: 0.007\n",
      "[116,    60] loss: 0.007\n",
      "[116,    80] loss: 0.007\n",
      "[116,   100] loss: 0.006\n",
      "[117,    20] loss: 0.008\n",
      "[117,    40] loss: 0.007\n",
      "[117,    60] loss: 0.007\n",
      "[117,    80] loss: 0.007\n",
      "[117,   100] loss: 0.006\n",
      "[118,    20] loss: 0.008\n",
      "[118,    40] loss: 0.007\n",
      "[118,    60] loss: 0.007\n",
      "[118,    80] loss: 0.007\n",
      "[118,   100] loss: 0.006\n",
      "[119,    20] loss: 0.008\n",
      "[119,    40] loss: 0.007\n",
      "[119,    60] loss: 0.007\n",
      "[119,    80] loss: 0.007\n",
      "[119,   100] loss: 0.006\n",
      "[120,    20] loss: 0.008\n",
      "[120,    40] loss: 0.007\n",
      "[120,    60] loss: 0.007\n",
      "[120,    80] loss: 0.007\n",
      "[120,   100] loss: 0.006\n",
      "[121,    20] loss: 0.008\n",
      "[121,    40] loss: 0.007\n",
      "[121,    60] loss: 0.007\n",
      "[121,    80] loss: 0.007\n",
      "[121,   100] loss: 0.006\n",
      "[122,    20] loss: 0.008\n",
      "[122,    40] loss: 0.007\n",
      "[122,    60] loss: 0.007\n",
      "[122,    80] loss: 0.007\n",
      "[122,   100] loss: 0.006\n",
      "[123,    20] loss: 0.008\n",
      "[123,    40] loss: 0.007\n",
      "[123,    60] loss: 0.007\n",
      "[123,    80] loss: 0.007\n",
      "[123,   100] loss: 0.006\n",
      "[124,    20] loss: 0.008\n",
      "[124,    40] loss: 0.007\n",
      "[124,    60] loss: 0.007\n",
      "[124,    80] loss: 0.007\n",
      "[124,   100] loss: 0.006\n",
      "[125,    20] loss: 0.008\n",
      "[125,    40] loss: 0.007\n",
      "[125,    60] loss: 0.007\n",
      "[125,    80] loss: 0.007\n",
      "[125,   100] loss: 0.006\n",
      "[126,    20] loss: 0.008\n",
      "[126,    40] loss: 0.007\n",
      "[126,    60] loss: 0.007\n",
      "[126,    80] loss: 0.007\n",
      "[126,   100] loss: 0.006\n",
      "[127,    20] loss: 0.008\n",
      "[127,    40] loss: 0.007\n",
      "[127,    60] loss: 0.007\n",
      "[127,    80] loss: 0.007\n",
      "[127,   100] loss: 0.006\n",
      "[128,    20] loss: 0.008\n",
      "[128,    40] loss: 0.007\n",
      "[128,    60] loss: 0.007\n",
      "[128,    80] loss: 0.007\n",
      "[128,   100] loss: 0.006\n",
      "[129,    20] loss: 0.008\n",
      "[129,    40] loss: 0.007\n",
      "[129,    60] loss: 0.007\n",
      "[129,    80] loss: 0.007\n",
      "[129,   100] loss: 0.006\n",
      "[130,    20] loss: 0.008\n",
      "[130,    40] loss: 0.007\n",
      "[130,    60] loss: 0.007\n",
      "[130,    80] loss: 0.007\n",
      "[130,   100] loss: 0.006\n",
      "[131,    20] loss: 0.008\n",
      "[131,    40] loss: 0.007\n",
      "[131,    60] loss: 0.007\n",
      "[131,    80] loss: 0.007\n",
      "[131,   100] loss: 0.006\n",
      "[132,    20] loss: 0.008\n",
      "[132,    40] loss: 0.007\n",
      "[132,    60] loss: 0.007\n",
      "[132,    80] loss: 0.007\n",
      "[132,   100] loss: 0.006\n",
      "[133,    20] loss: 0.008\n",
      "[133,    40] loss: 0.007\n",
      "[133,    60] loss: 0.007\n",
      "[133,    80] loss: 0.007\n",
      "[133,   100] loss: 0.006\n",
      "[134,    20] loss: 0.008\n",
      "[134,    40] loss: 0.007\n",
      "[134,    60] loss: 0.007\n",
      "[134,    80] loss: 0.007\n",
      "[134,   100] loss: 0.006\n",
      "[135,    20] loss: 0.008\n",
      "[135,    40] loss: 0.007\n",
      "[135,    60] loss: 0.007\n",
      "[135,    80] loss: 0.007\n",
      "[135,   100] loss: 0.006\n",
      "[136,    20] loss: 0.008\n",
      "[136,    40] loss: 0.007\n",
      "[136,    60] loss: 0.007\n",
      "[136,    80] loss: 0.007\n",
      "[136,   100] loss: 0.006\n",
      "[137,    20] loss: 0.008\n",
      "[137,    40] loss: 0.007\n",
      "[137,    60] loss: 0.007\n",
      "[137,    80] loss: 0.007\n",
      "[137,   100] loss: 0.006\n",
      "[138,    20] loss: 0.008\n",
      "[138,    40] loss: 0.007\n",
      "[138,    60] loss: 0.007\n",
      "[138,    80] loss: 0.007\n",
      "[138,   100] loss: 0.006\n",
      "[139,    20] loss: 0.008\n",
      "[139,    40] loss: 0.007\n",
      "[139,    60] loss: 0.007\n",
      "[139,    80] loss: 0.007\n",
      "[139,   100] loss: 0.006\n",
      "[140,    20] loss: 0.008\n",
      "[140,    40] loss: 0.007\n",
      "[140,    60] loss: 0.007\n",
      "[140,    80] loss: 0.007\n",
      "[140,   100] loss: 0.006\n",
      "[141,    20] loss: 0.008\n",
      "[141,    40] loss: 0.007\n",
      "[141,    60] loss: 0.007\n",
      "[141,    80] loss: 0.007\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[141,   100] loss: 0.006\n",
      "[142,    20] loss: 0.008\n",
      "[142,    40] loss: 0.007\n",
      "[142,    60] loss: 0.007\n",
      "[142,    80] loss: 0.007\n",
      "[142,   100] loss: 0.006\n",
      "[143,    20] loss: 0.008\n",
      "[143,    40] loss: 0.007\n",
      "[143,    60] loss: 0.007\n",
      "[143,    80] loss: 0.007\n",
      "[143,   100] loss: 0.006\n",
      "[144,    20] loss: 0.008\n",
      "[144,    40] loss: 0.007\n",
      "[144,    60] loss: 0.007\n",
      "[144,    80] loss: 0.007\n",
      "[144,   100] loss: 0.006\n",
      "[145,    20] loss: 0.008\n",
      "[145,    40] loss: 0.007\n",
      "[145,    60] loss: 0.007\n",
      "[145,    80] loss: 0.007\n",
      "[145,   100] loss: 0.006\n",
      "[146,    20] loss: 0.008\n",
      "[146,    40] loss: 0.007\n",
      "[146,    60] loss: 0.007\n",
      "[146,    80] loss: 0.007\n",
      "[146,   100] loss: 0.006\n",
      "[147,    20] loss: 0.008\n",
      "[147,    40] loss: 0.007\n",
      "[147,    60] loss: 0.007\n",
      "[147,    80] loss: 0.007\n",
      "[147,   100] loss: 0.006\n",
      "[148,    20] loss: 0.008\n",
      "[148,    40] loss: 0.007\n",
      "[148,    60] loss: 0.007\n",
      "[148,    80] loss: 0.007\n",
      "[148,   100] loss: 0.006\n",
      "[149,    20] loss: 0.008\n",
      "[149,    40] loss: 0.007\n",
      "[149,    60] loss: 0.007\n",
      "[149,    80] loss: 0.007\n",
      "[149,   100] loss: 0.006\n",
      "[150,    20] loss: 0.008\n",
      "[150,    40] loss: 0.007\n",
      "[150,    60] loss: 0.007\n",
      "[150,    80] loss: 0.007\n",
      "[150,   100] loss: 0.006\n",
      "[151,    20] loss: 0.008\n",
      "[151,    40] loss: 0.007\n",
      "[151,    60] loss: 0.007\n",
      "[151,    80] loss: 0.007\n",
      "[151,   100] loss: 0.006\n",
      "[152,    20] loss: 0.008\n",
      "[152,    40] loss: 0.007\n",
      "[152,    60] loss: 0.007\n",
      "[152,    80] loss: 0.007\n",
      "[152,   100] loss: 0.006\n",
      "[153,    20] loss: 0.008\n",
      "[153,    40] loss: 0.007\n",
      "[153,    60] loss: 0.007\n",
      "[153,    80] loss: 0.007\n",
      "[153,   100] loss: 0.006\n",
      "[154,    20] loss: 0.008\n",
      "[154,    40] loss: 0.007\n",
      "[154,    60] loss: 0.007\n",
      "[154,    80] loss: 0.007\n",
      "[154,   100] loss: 0.006\n",
      "[155,    20] loss: 0.008\n",
      "[155,    40] loss: 0.007\n",
      "[155,    60] loss: 0.007\n",
      "[155,    80] loss: 0.007\n",
      "[155,   100] loss: 0.006\n",
      "[156,    20] loss: 0.008\n",
      "[156,    40] loss: 0.007\n",
      "[156,    60] loss: 0.007\n",
      "[156,    80] loss: 0.007\n",
      "[156,   100] loss: 0.006\n",
      "[157,    20] loss: 0.008\n",
      "[157,    40] loss: 0.007\n",
      "[157,    60] loss: 0.007\n",
      "[157,    80] loss: 0.007\n",
      "[157,   100] loss: 0.006\n",
      "[158,    20] loss: 0.008\n",
      "[158,    40] loss: 0.007\n",
      "[158,    60] loss: 0.007\n",
      "[158,    80] loss: 0.007\n",
      "[158,   100] loss: 0.006\n",
      "[159,    20] loss: 0.008\n",
      "[159,    40] loss: 0.007\n",
      "[159,    60] loss: 0.007\n",
      "[159,    80] loss: 0.007\n",
      "[159,   100] loss: 0.006\n",
      "[160,    20] loss: 0.008\n",
      "[160,    40] loss: 0.007\n",
      "[160,    60] loss: 0.007\n",
      "[160,    80] loss: 0.007\n",
      "[160,   100] loss: 0.006\n",
      "[161,    20] loss: 0.008\n",
      "[161,    40] loss: 0.007\n",
      "[161,    60] loss: 0.007\n",
      "[161,    80] loss: 0.007\n",
      "[161,   100] loss: 0.006\n",
      "[162,    20] loss: 0.008\n",
      "[162,    40] loss: 0.007\n",
      "[162,    60] loss: 0.007\n",
      "[162,    80] loss: 0.007\n",
      "[162,   100] loss: 0.006\n",
      "[163,    20] loss: 0.008\n",
      "[163,    40] loss: 0.007\n",
      "[163,    60] loss: 0.007\n",
      "[163,    80] loss: 0.007\n",
      "[163,   100] loss: 0.006\n",
      "[164,    20] loss: 0.008\n",
      "[164,    40] loss: 0.007\n",
      "[164,    60] loss: 0.007\n",
      "[164,    80] loss: 0.007\n",
      "[164,   100] loss: 0.006\n",
      "[165,    20] loss: 0.008\n",
      "[165,    40] loss: 0.007\n",
      "[165,    60] loss: 0.007\n",
      "[165,    80] loss: 0.007\n",
      "[165,   100] loss: 0.006\n",
      "[166,    20] loss: 0.008\n",
      "[166,    40] loss: 0.007\n",
      "[166,    60] loss: 0.007\n",
      "[166,    80] loss: 0.007\n",
      "[166,   100] loss: 0.006\n",
      "[167,    20] loss: 0.008\n",
      "[167,    40] loss: 0.007\n",
      "[167,    60] loss: 0.007\n",
      "[167,    80] loss: 0.007\n",
      "[167,   100] loss: 0.006\n",
      "[168,    20] loss: 0.008\n",
      "[168,    40] loss: 0.007\n",
      "[168,    60] loss: 0.007\n",
      "[168,    80] loss: 0.007\n",
      "[168,   100] loss: 0.006\n",
      "[169,    20] loss: 0.008\n",
      "[169,    40] loss: 0.007\n",
      "[169,    60] loss: 0.007\n",
      "[169,    80] loss: 0.007\n",
      "[169,   100] loss: 0.006\n",
      "[170,    20] loss: 0.008\n",
      "[170,    40] loss: 0.007\n",
      "[170,    60] loss: 0.007\n",
      "[170,    80] loss: 0.007\n",
      "[170,   100] loss: 0.006\n",
      "[171,    20] loss: 0.008\n",
      "[171,    40] loss: 0.007\n",
      "[171,    60] loss: 0.007\n",
      "[171,    80] loss: 0.007\n",
      "[171,   100] loss: 0.006\n",
      "[172,    20] loss: 0.008\n",
      "[172,    40] loss: 0.007\n",
      "[172,    60] loss: 0.007\n",
      "[172,    80] loss: 0.007\n",
      "[172,   100] loss: 0.006\n",
      "[173,    20] loss: 0.008\n",
      "[173,    40] loss: 0.007\n",
      "[173,    60] loss: 0.007\n",
      "[173,    80] loss: 0.007\n",
      "[173,   100] loss: 0.006\n",
      "[174,    20] loss: 0.008\n",
      "[174,    40] loss: 0.007\n",
      "[174,    60] loss: 0.007\n",
      "[174,    80] loss: 0.007\n",
      "[174,   100] loss: 0.006\n",
      "[175,    20] loss: 0.008\n",
      "[175,    40] loss: 0.007\n",
      "[175,    60] loss: 0.007\n",
      "[175,    80] loss: 0.007\n",
      "[175,   100] loss: 0.006\n",
      "[176,    20] loss: 0.008\n",
      "[176,    40] loss: 0.007\n",
      "[176,    60] loss: 0.007\n",
      "[176,    80] loss: 0.007\n",
      "[176,   100] loss: 0.006\n",
      "[177,    20] loss: 0.008\n",
      "[177,    40] loss: 0.007\n",
      "[177,    60] loss: 0.007\n",
      "[177,    80] loss: 0.007\n",
      "[177,   100] loss: 0.006\n",
      "[178,    20] loss: 0.008\n",
      "[178,    40] loss: 0.007\n",
      "[178,    60] loss: 0.007\n",
      "[178,    80] loss: 0.007\n",
      "[178,   100] loss: 0.006\n",
      "[179,    20] loss: 0.008\n",
      "[179,    40] loss: 0.007\n",
      "[179,    60] loss: 0.007\n",
      "[179,    80] loss: 0.007\n",
      "[179,   100] loss: 0.006\n",
      "[180,    20] loss: 0.008\n",
      "[180,    40] loss: 0.007\n",
      "[180,    60] loss: 0.007\n",
      "[180,    80] loss: 0.007\n",
      "[180,   100] loss: 0.006\n",
      "[181,    20] loss: 0.008\n",
      "[181,    40] loss: 0.007\n",
      "[181,    60] loss: 0.007\n",
      "[181,    80] loss: 0.007\n",
      "[181,   100] loss: 0.006\n",
      "[182,    20] loss: 0.008\n",
      "[182,    40] loss: 0.007\n",
      "[182,    60] loss: 0.007\n",
      "[182,    80] loss: 0.007\n",
      "[182,   100] loss: 0.006\n",
      "[183,    20] loss: 0.008\n",
      "[183,    40] loss: 0.007\n",
      "[183,    60] loss: 0.007\n",
      "[183,    80] loss: 0.007\n",
      "[183,   100] loss: 0.006\n",
      "[184,    20] loss: 0.008\n",
      "[184,    40] loss: 0.007\n",
      "[184,    60] loss: 0.007\n",
      "[184,    80] loss: 0.007\n",
      "[184,   100] loss: 0.006\n",
      "[185,    20] loss: 0.008\n",
      "[185,    40] loss: 0.007\n",
      "[185,    60] loss: 0.007\n",
      "[185,    80] loss: 0.007\n",
      "[185,   100] loss: 0.006\n",
      "[186,    20] loss: 0.008\n",
      "[186,    40] loss: 0.007\n",
      "[186,    60] loss: 0.007\n",
      "[186,    80] loss: 0.007\n",
      "[186,   100] loss: 0.006\n",
      "[187,    20] loss: 0.008\n",
      "[187,    40] loss: 0.007\n",
      "[187,    60] loss: 0.007\n",
      "[187,    80] loss: 0.007\n",
      "[187,   100] loss: 0.006\n",
      "[188,    20] loss: 0.008\n",
      "[188,    40] loss: 0.007\n",
      "[188,    60] loss: 0.007\n",
      "[188,    80] loss: 0.007\n",
      "[188,   100] loss: 0.006\n",
      "[189,    20] loss: 0.008\n",
      "[189,    40] loss: 0.007\n",
      "[189,    60] loss: 0.007\n",
      "[189,    80] loss: 0.007\n",
      "[189,   100] loss: 0.006\n",
      "[190,    20] loss: 0.008\n",
      "[190,    40] loss: 0.007\n",
      "[190,    60] loss: 0.007\n",
      "[190,    80] loss: 0.007\n",
      "[190,   100] loss: 0.006\n",
      "[191,    20] loss: 0.008\n",
      "[191,    40] loss: 0.007\n",
      "[191,    60] loss: 0.007\n",
      "[191,    80] loss: 0.007\n",
      "[191,   100] loss: 0.006\n",
      "[192,    20] loss: 0.008\n",
      "[192,    40] loss: 0.007\n",
      "[192,    60] loss: 0.007\n",
      "[192,    80] loss: 0.007\n",
      "[192,   100] loss: 0.006\n",
      "[193,    20] loss: 0.008\n",
      "[193,    40] loss: 0.007\n",
      "[193,    60] loss: 0.007\n",
      "[193,    80] loss: 0.007\n",
      "[193,   100] loss: 0.006\n",
      "[194,    20] loss: 0.008\n",
      "[194,    40] loss: 0.007\n",
      "[194,    60] loss: 0.007\n",
      "[194,    80] loss: 0.007\n",
      "[194,   100] loss: 0.006\n",
      "[195,    20] loss: 0.008\n",
      "[195,    40] loss: 0.007\n",
      "[195,    60] loss: 0.007\n",
      "[195,    80] loss: 0.007\n",
      "[195,   100] loss: 0.006\n",
      "[196,    20] loss: 0.008\n",
      "[196,    40] loss: 0.007\n",
      "[196,    60] loss: 0.007\n",
      "[196,    80] loss: 0.007\n",
      "[196,   100] loss: 0.006\n",
      "[197,    20] loss: 0.008\n",
      "[197,    40] loss: 0.007\n",
      "[197,    60] loss: 0.007\n",
      "[197,    80] loss: 0.007\n",
      "[197,   100] loss: 0.006\n",
      "[198,    20] loss: 0.008\n",
      "[198,    40] loss: 0.007\n",
      "[198,    60] loss: 0.007\n",
      "[198,    80] loss: 0.007\n",
      "[198,   100] loss: 0.006\n",
      "[199,    20] loss: 0.008\n",
      "[199,    40] loss: 0.007\n",
      "[199,    60] loss: 0.007\n",
      "[199,    80] loss: 0.007\n",
      "[199,   100] loss: 0.006\n",
      "[200,    20] loss: 0.008\n",
      "[200,    40] loss: 0.007\n",
      "[200,    60] loss: 0.007\n",
      "[200,    80] loss: 0.007\n",
      "[200,   100] loss: 0.006\n",
      "Finished Training\n"
     ]
    }
   ],
   "source": [
    "epochs = 200\n",
    "batch_size = 10\n",
    "\n",
    "\n",
    "num_samples_train = x_train.shape[0]\n",
    "num_batches_train = num_samples_train // batch_size\n",
    "\n",
    "\n",
    "for epoch in range(epochs):  # loop over the dataset multiple times\n",
    "\n",
    "    running_loss = 0.0\n",
    "    for i in range(num_batches_train):\n",
    "        \n",
    "        #get slice\n",
    "        slice_range = get_slice(i,batch_size)\n",
    "        \n",
    "        # get the inputs\n",
    "        x = inputs[slice_range]\n",
    "        y = labels[slice_range]\n",
    "        # zero the parameter gradients\n",
    "        optimizer.zero_grad()\n",
    "\n",
    "        # forward + backward + optimize\n",
    "        outputs = net(x)\n",
    "        _, pred = torch.max(outputs,1)\n",
    "        loss = criterion(outputs,y)\n",
    "        loss.backward()\n",
    "        optimizer.step()\n",
    "\n",
    "        # print statistics\n",
    "        running_loss += loss.item()\n",
    "        if i % 20 == 19:    # print every 2000 mini-batches\n",
    "            print('[%d, %5d] loss: %.3f' %\n",
    "                  (epoch + 1, i + 1, running_loss / 2000))\n",
    "            running_loss = 0.0\n",
    "\n",
    "print('Finished Training')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 108,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "range(0, 10)"
      ]
     },
     "execution_count": 108,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "labels[slice_range].type(long)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 125,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor([4, 4, 3, 1, 7, 7, 6, 0, 3, 6])\n"
     ]
    }
   ],
   "source": [
    "_, pred = torch.max(outputs,1,)\n",
    "print(pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 126,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([4, 4, 3, 1, 7, 7, 6, 0, 3, 6])"
      ]
     },
     "execution_count": 126,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 110,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'torch.IntTensor'"
      ]
     },
     "execution_count": 110,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "labels.type()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 153,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "images = torch.from_numpy(x_test)\n",
    "labels = torch.from_numpy(targets_test).long()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 154,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([7, 6, 1, 0, 4, 1, 4, 7, 6, 0, 0, 6, 0, 0, 1, 5, 7, 7, 6, 4, 0, 6, 6, 5,\n",
       "        4, 0, 7, 4, 0, 1, 3, 1, 3, 4, 7, 0, 7, 1, 3, 1, 1, 7, 4, 8, 5, 5, 1, 6,\n",
       "        4, 4, 6, 3, 4, 5, 6, 8, 4, 1, 7, 5, 7, 8, 4, 3, 3, 4, 6, 4, 3, 5, 7, 0,\n",
       "        0, 7, 1, 7, 3, 7, 8, 7, 7, 6, 0, 7, 8, 4, 7, 5, 6, 1, 3, 6, 4, 3, 1, 4,\n",
       "        1, 3, 6, 7, 6, 0, 5, 4, 0, 0, 0, 1, 5, 4, 8, 7, 3, 0, 7, 4, 4, 4, 0, 0,\n",
       "        5, 4, 7, 6, 4, 4, 0, 5, 8, 5, 6, 6, 5, 7, 8, 1, 0, 1, 6, 4, 6, 7, 3, 1,\n",
       "        7, 1, 8, 0, 0, 4, 4, 0, 5, 5, 1, 5, 6, 0, 3, 4, 4, 6, 5, 4, 6, 5, 4, 5,\n",
       "        1, 4, 4, 7, 3, 3, 0, 7, 1, 8, 1, 8, 1, 8, 5, 0, 8, 4, 0, 5, 0, 1, 1, 1,\n",
       "        0, 3, 0, 3, 1, 6, 4, 0, 3, 6, 1, 1, 1, 3, 8, 5, 6, 5, 4, 7, 3, 3, 0, 0,\n",
       "        3, 5, 5, 5, 7, 0, 0, 7, 1, 0, 8, 4, 1, 7, 5, 3, 8, 7, 7, 7, 0, 6, 4, 1,\n",
       "        5, 5, 8, 4, 7, 5, 0, 8, 4, 1, 4, 1, 4, 5, 7, 7, 1, 8, 6, 6, 8, 5, 7, 7,\n",
       "        5, 1, 8, 1, 4, 0, 3, 0, 1, 0, 7, 4, 1, 8, 0, 1, 8, 0, 8, 5, 7, 0, 6, 4,\n",
       "        1, 6, 4, 0, 7, 6, 0, 4, 0, 5, 0, 8, 6, 7, 1, 7, 4, 5, 0, 7, 4, 5, 3, 0,\n",
       "        0, 5, 1, 7, 6, 5, 0, 5, 7, 7, 0, 3, 0, 4, 6, 0, 7, 1, 1, 1, 1, 5, 3, 3,\n",
       "        0, 7, 8, 6, 5, 6, 1, 3, 8, 1, 0, 8, 1, 7, 1, 5, 0, 6, 1, 8, 5, 1, 3, 8,\n",
       "        4, 6, 7, 1, 5, 5, 6, 5, 6, 3, 7, 0, 0, 8, 8, 5, 4, 1, 1, 4, 5, 7, 3, 7,\n",
       "        6, 1, 5, 7, 1, 8, 0, 8, 6, 1, 0, 5, 6, 5, 4, 4, 0, 8, 5, 0, 0, 4, 5, 7,\n",
       "        3, 1, 7, 7, 3, 7, 0, 7, 1, 7, 0, 1, 4, 0, 3, 0, 0, 4, 3, 1, 4, 8, 1, 8,\n",
       "        4, 8, 7, 8, 8, 3, 7, 6, 0, 0, 3, 8, 8, 7, 6, 4, 5, 5, 5, 3, 1, 3, 7, 1,\n",
       "        8, 5, 8, 0, 5, 4, 6, 6, 3, 8, 8, 7, 7, 5, 8, 7, 6, 1, 8, 4, 1, 0, 8, 5,\n",
       "        1, 0, 7, 7, 4, 0, 5, 4, 7, 1, 6, 5, 8, 3, 7, 0, 0, 4, 8, 6])"
      ]
     },
     "execution_count": 154,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x_batch = Variable(torch.from_numpy(x_test))\n",
    "output = net(x_batch)\n",
    "preds = torch.max(output, 1)[1]\n",
    "preds"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 155,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([7, 2, 1, 0, 4, 1, 4, 9, 5, 9, 0, 6, 9, 0, 1, 5, 9, 7, 3, 4, 9, 6, 6, 5,\n",
       "        4, 0, 7, 4, 0, 1, 3, 1, 3, 4, 7, 2, 7, 1, 2, 1, 1, 7, 4, 2, 3, 5, 1, 2,\n",
       "        4, 4, 6, 3, 5, 5, 6, 0, 4, 1, 9, 5, 7, 8, 9, 3, 7, 4, 6, 4, 3, 0, 7, 0,\n",
       "        2, 9, 1, 7, 3, 2, 9, 7, 7, 6, 2, 7, 8, 4, 7, 3, 6, 1, 3, 6, 9, 3, 1, 4,\n",
       "        1, 7, 6, 9, 6, 0, 5, 4, 9, 9, 2, 1, 9, 4, 8, 7, 3, 9, 7, 4, 4, 4, 9, 2,\n",
       "        5, 4, 7, 6, 7, 9, 0, 5, 8, 5, 6, 6, 5, 7, 8, 1, 0, 1, 6, 4, 6, 7, 3, 1,\n",
       "        7, 1, 8, 2, 0, 2, 9, 9, 5, 5, 1, 5, 6, 0, 3, 4, 4, 6, 5, 4, 6, 5, 4, 5,\n",
       "        1, 4, 4, 7, 2, 3, 2, 7, 1, 8, 1, 8, 1, 8, 5, 0, 8, 9, 2, 5, 0, 1, 1, 1,\n",
       "        0, 9, 0, 3, 1, 6, 4, 2, 3, 6, 1, 1, 1, 3, 9, 5, 2, 9, 4, 5, 9, 3, 9, 0,\n",
       "        3, 6, 5, 5, 7, 2, 2, 7, 1, 2, 8, 4, 1, 7, 3, 3, 8, 8, 7, 9, 2, 2, 4, 1,\n",
       "        5, 9, 8, 7, 2, 3, 0, 4, 4, 2, 4, 1, 9, 5, 7, 7, 2, 8, 2, 6, 8, 5, 7, 7,\n",
       "        9, 1, 8, 1, 8, 0, 3, 0, 1, 9, 9, 4, 1, 8, 2, 1, 2, 9, 7, 5, 9, 2, 6, 4,\n",
       "        1, 5, 8, 2, 9, 2, 0, 4, 0, 0, 2, 8, 4, 7, 1, 2, 4, 0, 2, 7, 4, 3, 3, 0,\n",
       "        0, 3, 1, 9, 6, 5, 2, 5, 9, 2, 9, 3, 0, 4, 2, 0, 7, 1, 1, 2, 1, 5, 3, 3,\n",
       "        9, 7, 8, 6, 5, 6, 1, 3, 8, 1, 0, 5, 1, 3, 1, 5, 5, 6, 1, 8, 5, 1, 7, 9,\n",
       "        4, 6, 2, 2, 5, 0, 6, 5, 6, 3, 7, 2, 0, 8, 8, 5, 4, 1, 1, 4, 0, 3, 3, 7,\n",
       "        6, 1, 6, 2, 1, 9, 2, 8, 6, 1, 9, 5, 2, 5, 4, 4, 2, 8, 3, 8, 2, 4, 5, 0,\n",
       "        3, 1, 7, 7, 5, 7, 9, 7, 1, 9, 2, 1, 4, 2, 9, 2, 0, 4, 9, 1, 4, 8, 1, 8,\n",
       "        4, 5, 9, 8, 8, 3, 7, 6, 0, 0, 3, 0, 2, 6, 6, 4, 9, 3, 3, 3, 2, 3, 9, 1,\n",
       "        2, 6, 8, 0, 5, 6, 6, 6, 3, 8, 8, 2, 7, 5, 8, 9, 6, 1, 8, 4, 1, 2, 5, 9,\n",
       "        1, 9, 7, 5, 4, 0, 8, 9, 9, 1, 0, 5, 2, 3, 7, 8, 9, 4, 0, 6])"
      ]
     },
     "execution_count": 155,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 156,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([1, 0, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 0, 1, 1, 1, 0, 1, 0, 1, 0, 1, 1, 1,\n",
       "        1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 0, 1, 1, 1, 1, 0, 0, 1, 1, 0,\n",
       "        1, 1, 1, 1, 0, 1, 1, 0, 1, 1, 0, 1, 1, 1, 0, 1, 0, 1, 1, 1, 1, 0, 1, 1,\n",
       "        0, 0, 1, 1, 1, 0, 0, 1, 1, 1, 0, 1, 1, 1, 1, 0, 1, 1, 1, 1, 0, 1, 1, 1,\n",
       "        1, 0, 1, 0, 1, 1, 1, 1, 0, 0, 0, 1, 0, 1, 1, 1, 1, 0, 1, 1, 1, 1, 0, 0,\n",
       "        1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,\n",
       "        1, 1, 1, 0, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,\n",
       "        1, 1, 1, 1, 0, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1,\n",
       "        1, 0, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 0, 1, 0, 0, 1, 0, 0, 1, 0, 1,\n",
       "        1, 0, 1, 1, 1, 0, 0, 1, 1, 0, 1, 1, 1, 1, 0, 1, 1, 0, 1, 0, 0, 0, 1, 1,\n",
       "        1, 0, 1, 0, 0, 0, 1, 0, 1, 0, 1, 1, 0, 1, 1, 1, 0, 1, 0, 1, 1, 1, 1, 1,\n",
       "        0, 1, 1, 1, 0, 1, 1, 1, 1, 0, 0, 1, 1, 1, 0, 1, 0, 0, 0, 1, 0, 0, 1, 1,\n",
       "        1, 0, 0, 0, 0, 0, 1, 1, 1, 0, 0, 1, 0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 1, 1,\n",
       "        1, 0, 1, 0, 1, 1, 0, 1, 0, 0, 0, 1, 1, 1, 0, 1, 1, 1, 1, 0, 1, 1, 1, 1,\n",
       "        0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 0, 1, 1, 0, 1, 1, 1, 1, 1, 0, 0,\n",
       "        1, 1, 0, 0, 1, 0, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1,\n",
       "        1, 1, 0, 0, 1, 0, 0, 1, 1, 1, 0, 1, 0, 1, 1, 1, 0, 1, 0, 0, 0, 1, 1, 0,\n",
       "        1, 1, 1, 1, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 0, 0, 1, 1, 0, 1, 1, 1, 1, 1,\n",
       "        1, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 1, 0, 1,\n",
       "        0, 0, 1, 1, 1, 0, 1, 1, 1, 1, 1, 0, 1, 1, 1, 0, 1, 1, 1, 1, 1, 0, 0, 0,\n",
       "        1, 0, 1, 0, 1, 1, 0, 0, 0, 1, 0, 1, 0, 1, 1, 0, 0, 1, 0, 1],\n",
       "       dtype=torch.uint8)"
      ]
     },
     "execution_count": 156,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "preds == labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.6.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}