imagenet15-scratch.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# ImageNet in 15 minutes notes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def downloadFile(URL=None):\n",
    "    import httplib2\n",
    "    h = httplib2.Http(\".cache\")\n",
    "    resp, content = h.request(URL, \"GET\")\n",
    "    return content\n",
    "log_url=\"https://s3.amazonaws.com/yaroslavvb/logs/data/training/nv/2018-07-27_07-19-baseline-single-ami12-w8/full.log\"\n",
    "log=downloadFile(log_url).decode('ascii')\n",
    "import re\n",
    "times_ms = [1000*float(x) for x in re.findall(\"Time ([.0-9]+)\", log)]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Timing data for original run"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "python launch_nv.py --name $name --ami-name=\"$ami\" --zone=$zone  --num-tasks 1 --attach-volume imagenet_high_perf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.text.Text at 0x112339630>"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAjQAAAFkCAYAAAA37aFpAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11225c668>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "import matplotlib.pyplot as plt\n",
    "plt.plot(times_ms)\n",
    "plt.ylabel('ms')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "Batch-size 192, nvidia-smi around 85"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.text.Text at 0x11df554e0>"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAicAAAFkCAYAAAD7dJuCAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAAPYQAAD2EBqD+naQAAIABJREFUeJzt3XmcHFW5//HPk32BhARNIiqK4uXiRcHEXxCVTZRVAcWF\nAWW7XlBANOpV4aogCAoqQRQUFRAFoxjkggiETbnsaIKoGBYlGMISCEsy2WYyM8/vj9PF1HS6e3q6\nq7vr9Hzfr1e/Zqa6pvtMTXfXt55zTpW5OyIiIiJ5MaLVDRARERFJUzgRERGRXFE4ERERkVxROBER\nEZFcUTgRERGRXFE4ERERkVxROBEREZFcUTgRERGRXFE4ERERkVxROBEREZFcaXk4MbNPmNn9Zray\ncLvTzPZO3X+xmfUV3a4teoyxZnaema0ws04zm29m05r/14iIiEi9Wh5OgMeBLwIzgVnALcBVZrZt\nap3rgOnAjMKto+gxzgH2Aw4CdgG2AK5obLNFRESkESyPF/4zs+eAz7v7xWZ2MTDZ3T9QZt1JwLPA\nwe5+ZWHZNsBi4G3ufm+z2i0iIiL1y0Pl5CVmNsLMDgYmAHem7trNzJab2YNmdr6ZTU3dNwsYBdyc\nLHD3h4ClwE7NaLeIiIhkZ1SrGwBgZtsBdwHjgE7g/YWAAaFL5wpgCfB64BvAtWa2k4eyzwyg291X\nFT3s8sJ95Z5zc2Av4DFgfXZ/jYiISNsbB7wWWODuz2X94LkIJ8CDwPbAZOCDwM/MbBd3f9DdL0+t\n94CZ/RX4J7Ab8Ps6nnMv4LI6fl9ERGS4OxT4RdYPmotw4u49wKOFH+8zs9nAp4FPllh3iZmtALYm\nhJOngTFmNqmoejK9cF85jwFceumlbLvtthVWkyzNmTOHuXPntroZw4q2efNpmzeftnlzLV68mI9+\n9KNQ2JdmLRfhpIQRwNhSd5jZq4DNgacKixYCPcAeQHpA7JaErqJy1gNsu+22zJw5M5tWy6AmT56c\n++19880wZgzsvHOrW5KNGLZ5u9E2bz5t85ZpyLCIlocTMzuDMK5kKbApoUS0K7CnmU0ETiaMOXma\nUC05E3gYWADg7qvM7ELgbDN7gTBm5VzgDs3UkVp885swaVL7hBMRkdjkYbbONOASwriTmwizb/Z0\n91uAXuDNwFXAQ8CPgT8Cu7j7htRjzAGuAeYDfwCeJJzzRGTIurrCTYa3BQvgi19sdStEhqeWV07c\n/eMV7lsP7F3u/tR6XcCnCjeRunR3K5wI3HQT/PKXcOaZrW6JyPCTh8qJDCMdHcUn982frq4QUNpF\nDNs8j+qpoGmbN5+2eXtROJGmiuEDpN0qJzFs8zxav772kKpt3nza5u1F4USkSLtVTlrt9tth3bpW\nt2Lo1q9vr5AqEhOFE5Ei7VY5aaXubthtN/jNb1rdkqGrp3IiIvVROBEp0t2tnVJW1q+H3l5YvbrV\nLRm6ri7o64Oenla3RGT4UTgRKaKpxNlJtmOM23N94dRSCqoizadwIlJElZPstEM4ibHtIrFr+XlO\nRPKmqyucvl7ql4S8GHfwSZsVVEWaT5UTkZTe3nDTDikbMe/gVTkRaR2FE5GUmI/080jdOiJSC4UT\nkZQknCQVFKlPO4STGKs+IrFTOBFJSe+ItFOqX8yVqJiDlUjsFE5EUtI7IoWT+rXDmJMY2y4SO4UT\naZrrroNvf7vVragsvSPSEXP9Yq4+aMyJSOsonEjTXH01XHRRq1tRmSon2Yo1nLjHXfURiZ3CiTRN\nV1f/0WheqXKSrVjHnGzYEAIKxNd2kXagcCJNE8Np4dPty3tbYxBr5SQdolU5EWk+hRNpmhguQa/Z\nOtmKtWskHU7y/poVaUcKJ9I06tYZfmKtnKiCJtJaCifSNLF168R2tJ9HsY45UbeOSGspnEjTdHVB\nT0++z7yqykm2Yq2cqFtHpLUUTqRpYthRqXKSrRj+56WociLSWgon0jQx7KhUOclWrANiNeZEpLUU\nTqRpkg/5PA+KVeUkWxpzIiK1UDiRponhdOCqnGQrhmpZKclrdfz4+Nou0g4UTqRpYthRdXWFHRLo\niDkLMfzPS0naO2mSXgcirTCq1Q2Q4SOGbp3ubhg3Lswoim2HmkexjjlJXqOTJ+t1INIKqpxI08Rw\nFN3dDWPGhFtsO9Q8SrZhb2++p5AXS8LJppvm+/Uq0q4UTqRpYqicdHXB2LHhpp1S/WKd9bJ+fQio\n48YppIq0gsKJNEVfXzgBG+R7J6XKSbZiDSddXSGYKKSKtIbCiTRFLDspVU6y1dUFEyf2fx+L9evD\na0AhVaQ1FE6kKdJdOXnu1kkqJ2PHaqeUhe7uMOMl+T4W69erciLSSgon0hQxVU6Sbp08tzMWXV39\n4SSm7ZmEE1VORFqj5eHEzD5hZveb2crC7U4z27tonVPN7EkzW2tmN5rZ1kX3jzWz88xshZl1mtl8\nM5vW3L9EKoklnHR393fraKdUv66uMOMl+T4W6t4Taa2WhxPgceCLwExgFnALcJWZbQtgZl8EjgeO\nBmYDa4AFZjYm9RjnAPsBBwG7AFsAVzTrD5DBpT/gY+jWUeUkG7GGE1VORFqr5Sdhc/ffFS36spl9\nEngbsBj4NHCau18DYGaHAcuBA4HLzWwScBRwsLvfWljnSGCxmc1293ub9KdIBbFUTpIjZp2ELRsa\ncyIitchD5eQlZjbCzA4GJgB3mtlWwAzg5mQdd18F3APsVFj0VkLISq/zELA0tY60WIyVk5h2pnkV\n65gTTSUWaa2WV04AzGw74C5gHNAJvN/dHzKznQAnVErSlhNCC8B0oLsQWsqtIy2WDiR5/rDv6gqn\nLFflJBsxd+toKrFI6+QinAAPAtsDk4EPAj8zs11a2yTJUrJjMsv3TiqpnPT0wKriuCtDFmvlZP16\n2GQTVU5EWiUX4cTde4BHCz/eZ2azCWNNzgKMUB1JV0+mA/cVvn8aGGNmk4qqJ9ML91U0Z84cJk+e\nPGBZR0cHHR0dtfwpUkb6Kq957tZJphKrcpKN7u54Kycve5kqJyIA8+bNY968eQOWrVy5sqHPmYtw\nUsIIYKy7LzGzp4E9gL8AFAbA7gicV1h3IdBTWOfKwjrbAFsSuooqmjt3LjNnzsz8D5CB0uEkzzup\nZCpxb692SvXq64MNG+IcEKupxCL9Sh2wL1q0iFmzZjXsOVseTszsDOA6wgDWTYFDgV2BPQurnEOY\nwfMP4DHgNGAZcBWEAbJmdiFwtpm9QBizci5wh2bq5EfyAT95cr4rJ0m3jion9UvCSKyVE00lFmmd\nlocTYBpwCfAKYCWhQrKnu98C4O5nmdkE4AJgM+A2YB93T39kzAF6gfnAWOB64Lim/QUyqFgqJ+mp\nxNop1Sf5P48fDyNH5vv/XkxTiUVaq+XhxN0/XsU6pwCnVLi/C/hU4SY51NUVBsNuskm+P+xVOclO\nEu5i7B5JphKPGRO6p3p7Q8ASkebI1XlOpH0lUzPHjct3t05SOVE5v35JGEm2Z0zhJHm9jh0bfo6p\n7SLtQOFEmiLZ6Y8bl+8P+vRVifPczhikw0ls1ypKd+uAXgsizdbybh0ZHmKZ/ZCeShzTzjSPkv9z\njGEvPSAW9FoQaTaFE2mKdDjJc7dOMpW4ry+unWkexTrmxH3g6eshnraLtAuFE2mKGLp1entDKFHl\nJBvF3Tp5/b8X6+kJr4NkrAzotSDSbAon0hQxVE7SO9PkBGJ9fTBCI7NqUjwgNpYdfPL6VOVEpHX0\nsStNkZTJ81w5SXaeyVWJIQQUqU2sY07S4USVE5HWUDiRpkhPzczrTqq4GyK9TIYu1jEneh2ItJ66\ndaQpYujWSVdOensHLpOhi3XMiSonIq2ncCJNEcOA2HQ3RF/fwGUydO0QTlQ5EWkNdetIUxRXTtxb\n3aKNFXdDgHZK9UiHvZgGxCbtToeTWNou0i4UTqQp0uHEPUzXzJvinSlop1SP7u5wPaVRo+KsnKSn\nEsfSdpF2oW4daYp0t07y8+jRrW1TsXTlJKnsaKdUu+R/bhZnOFG3jkjrqHIiTVF8xs08DootNZVY\nlZPaJeEE4g0neh2ItIYqJ9IU6asSQz53VOkBnKqc1C8dTmK6KnH6dZBU92Jpu0i7UDiRpkiPOUl+\nzpt05SSZraMj5tolV3iGuK5KnK6cmMU1mFekXSicSFMUh5M8duukB8SqclK/2Lt1Ymy7SLtQOJGm\nKDUgNm9KDYjVEXPtYg0nyWDtkSPDz6qciDSfwok0hSonw0+s4SQZH5WIqe0i7ULhRJoifeG/5Oe8\nSY85UeWkfl1d/WNOYqo+rF/f/zqFuMbLiLQLTSWWpohlQOzo0f2DICGf7YxFd/fGlZM8nhm4WHE4\niWmmkUi7UDiRhuvpCRfSi6FbJ2nfiBHhzKY6Yq5dcbdOXs8MXCyp8iXUrSPSfAon0nDp80bkvVsn\nqZiAdkr1Kg4nybK8Kx5zElOXlEi7UDiRhiu+Oi3kt3KSDifaKdUnvT1jCyeqnIi0lsKJNFw6nOR5\nLEd6jARop1Sv9PaM6TTwpcacxNBukXaicCINlw4neb4InCon2Yq1Wyfdbsjv61WknSmcSMMlH+zJ\n0ejYsfns1lHlJFuxhhNVTkRaT+FEGi5dOYHwwZ/HnVTxgFjtlOqjMSciUiuFE2m4WK5VonJ+tkqN\nOYlhe5aaSqyQKtJcCifScMWVkzx36xRXTmLYmeZVqW6dGHbypaYS63Ug0lwKJ9JwsXTrFA+I1RFz\nfdplzIkqaCLNp3AiDRdT5URHzNlplzEnGnsk0nwKJ9JwxbN1VDkZHoqvrQP5/L8X0+nrRVpP4UQa\nrlTlJI8f9ppKnK10t05sJ2HT6etFWqvl4cTMTjSze81slZktN7Mrzezfita52Mz6im7XFq0z1szO\nM7MVZtZpZvPNbFpz/xoppdRsnbx262gqcTb6+sJF/mKsnGjMiUjrtTycADsD3wN2BN4NjAZuMLPx\nRetdB0wHZhRuHUX3nwPsBxwE7AJsAVzRuGZLtbq6+q/yC/nu1lHlJBvJdkvCXsxTiRVSRZpvVKsb\n4O77pn82syOAZ4BZwO2pu7rc/dlSj2Fmk4CjgIPd/dbCsiOBxWY2293vbUTbpTqldvovvti69pSj\nykl2ku2W/N9HjIDRo+MIJ8XdOgqpIs2Xh8pJsc0AB54vWr5bodvnQTM738ympu6bRQhaNycL3P0h\nYCmwU6MbLJXFUpEoNSA2j+2MQfE4o+T7vG/Pnh7o7dVJ2ERareWVkzQzM0L3zO3u/vfUXdcRumiW\nAK8HvgFca2Y7ubsTunm63X1V0UMuL9wnLVQcTvLarVNqKrF2SrUpFU5i2J7JWKjibp3e3nAbObI1\n7RIZbnIVToDzgTcC70gvdPfLUz8+YGZ/Bf4J7Ab8vp4nnDNnDpMnTx6wrKOjg46O4iEtUqtSUzPz\nOCBWlZPsFI85gTi2Z/G0dxg4mHfChOa3SaTV5s2bx7x58wYsW7lyZUOfMzfhxMy+D+wL7OzuT1Va\n192XmNkKYGtCOHkaGGNmk4qqJ9ML95U1d+5cZs6cWV/jpSJVToaf4jEnyfd5/L+nFc8sg4HToBVO\nZDgqdcC+aNEiZs2a1bDnzMWYk0IwOQDY3d2XVrH+q4DNgSTELAR6gD1S62wDbAnclXmDZUhiGWBY\nPCA2r+2MQaxjTkp168Q0DVr67bwz/OpXrW6F1KrllRMzO58wLXh/YI2ZTS/ctdLd15vZROBkwpiT\npwnVkjOBh4EFAO6+yswuBM42sxeATuBc4A7N1Gm9UgNi89qto8pJNsqNOcn7Dr7cmBPQayE2f/oT\nLF7c6lZIrVoeToBPEGbn/KFo+ZHAz4Be4M3AYYSZPE8SQslX3X1Dav05hXXnA2OB64HjGtlwqU4M\n3TruqpxkqdyYk7zv4MtVfNL3Sf51d4eguXZtq1sitWp5OHH3il1L7r4e2LuKx+kCPlW4SY7EUDnp\n6QkBpVTlxB3MWte2GMU+5kSVk7h1doav69a1th1Su1yMOZH2Vqpykuz08yLZ8RQf6QNs2LDx+lKZ\nxpxIK60qTItQ5SReCifScKWmEkO+jkRLdUPoiLl2sYaTSlOJ9TqIR1I5UTiJl8KJNFypbh3IV9dO\nuW4IyP8ONY/Khb287+ArTSXW6yAeqpzET+FEGq54KnFyVJqnD/tKlZM8tTMW7TTmRCE1Pkk40ZiT\neCmcSMPFXjnJ+9F+HhVfiRriCiflTsImcVDlJH4KJ9JwpQbEJsvzotSAWFVOalf8P4c4wklXVwhU\nxaEquU/ioDEn8VM4kYYrVznJ04d9pfNb6Ih56IqvUwRxhJPiLkhQ5SRGqpzET+FEGq7cbJ08duuo\ncpKN4usUQTwDYtOvVchnmJbKNOYkfgon0nAxdOuUO6Mp5H+HmkexduuUCiejR4eveh3EQ9068VM4\nkYaLoVun1IBYVU5qF2s4KdVusziuCyT91K0TP4UTabhSVyVOlueFKifZinnMSXHlBOK4LpD0S4eT\nPJ2JWqqncCIN5R5Ht45OwpatcmNO8r4ty4WTGNou/ZJwAvq/xUrhRBqq1AX18lg5qTQgVkfMQ1eu\nWyfv27J48HYihqqP9OvshEmTwvfq2omTwok0VKkpunmsnFSaSpyndsaiUjjJc5m91FRiiGOmkfRb\ntQqmTw/fK5zESeFEGqrUhdRGjQpnD83TTj/Z8aRPvqXKSe3KjTmBfG/PSmNO8vR6lcpWrYIZM8L3\nCidxUjiRhipVkUh+zlO3TrIzNetfNnJkuGmnNHSlxpzEUImqNOYkz6FKBurs7K+c6FwncVI4kYYq\nF07GjcvXTqrUzhS0U6pVqW6dGCpRpdoNqpzExF2Vk3agcCINVepCasnPefqwL9UNAflrZyzKjTlJ\n7ssrVU7it24d9PZqzEnsFE6koSpVTvLUraPKSbYqjTmJMZwopMYjmUasykncFE6koSqNOcnTh313\ntyonWYp1zEmlqcQKqXFITl2vMSdxUziRhio1WwfyOSBWlZPsxNytU+51kOd2S7+kcqJunbgpnEhD\nxTQgVpWT7MQ6IFbdOvFLwsnmm4dTAyicxEnhRBoqlm6dcgNidcRcm3Ybc6IKWjyScLLppjBhgsJJ\nrBROpKFiOc9JuQGxGmtQm3Ycc5Lndku/ZMzJpEkhnGjMSZwUTqShyk0lzlu3jion2WrHMScKqXFY\ntQpGjw7/R1VO4qVwIg2lysnwFOOYk56ecFPlJG6rVoUuHTMYP17hJFYKJ9JQXV39p4FPy1vlpNyA\nWFVOahPjmJNyM8tAlZOYpK9IrG6deCmcSEPF0odf6bTl2ikNTW9vuMXWrVOuypcsy2u7ZaBVqwaG\nE1VO4qRwIg1Vaaeft24dVU6ykYS54v/76NHha163Z/J61EnY4qZw0h4UTqShyoWTvHXrVLq2jnZK\nQ1OuAmGW7wpEpXCikBqPZMwJaMxJzBROpKHKzX7I206q0rV18tTOGCTbq1wlKq9hr9KYk7y9XqU8\njTlpDwon0lCVKid56tZR5SQ75bp1kmV53cmXm/YO+Q5VMpC6ddqDwok0VKUxJ3naSalykp1YB5YO\nNuYkGegr+Zbu1lE4iVfLw4mZnWhm95rZKjNbbmZXmtm/lVjvVDN70szWmtmNZrZ10f1jzew8M1th\nZp1mNt/MpjXvL5FSKs3WUeWkPbVjOMn7OVqkX7pbR2NO4tXycALsDHwP2BF4NzAauMHMxicrmNkX\ngeOBo4HZwBpggZmldyfnAPsBBwG7AFsAVzTjD5DyYhkQW+kkbHlqZwwqjTnJ8/YcLFSl15H8Ku7W\n0ZiTOI1qdQPcfd/0z2Z2BPAMMAu4vbD408Bp7n5NYZ3DgOXAgcDlZjYJOAo42N1vLaxzJLDYzGa7\n+73N+FtkY5W6dXp6Qpm8+ARtrVBpKrGOloem0piTPG9PVU7i19sLq1erW6cd5KFyUmwzwIHnAcxs\nK2AGcHOygruvAu4BdioseishaKXXeQhYmlpHWqBSOEnuz4NK3Tp5aWMs2rFbJ2+vVylt9erwtbhb\nx711bZLa5CqcmJkRumdud/e/FxbPIISV5UWrLy/cBzAd6C6ElnLrSAuUm0qc7ADy8mFfaUCsjpaH\nJtZwMthUYtBrIe/SVySGUDnp7YUNG1rXJqlNy7t1ipwPvBF4R7OecM6cOUyePHnAso6ODjo6OprV\nhLbW1QVTp268PPmwz8OgWPfy3TrJztQ9nERMBhfrmJP162HECBhV4lMx+Vvy2nYJVhUOT9PhBMK4\nk1KvR6nOvHnzmDdv3oBlK1eubOhz5iacmNn3gX2Bnd39qdRdTwNGqI6kqyfTgftS64wxs0lF1ZPp\nhfvKmjt3LjNnzqy3+VJGpQGxyf2tlhxVlaucuIfxMcnp16WymMeclKqagLp1YpGEk/SYEwhdO0XH\noDIEpQ7YFy1axKxZsxr2nLno1ikEkwOA3d19afo+d19CCBh7pNafRJjdc2dh0UKgp2idbYAtgbsa\n2nipqNJU4uT+VhvsSB/yu0PNo1i7dSqFEw2IjUNxt874wpxPDYqNT8srJ2Z2PtAB7A+sMbPphbtW\nuntS9D8H+LKZ/QN4DDgNWAZcBWGArJldCJxtZi8AncC5wB2aqdNagw2IzUO3zmBH+hD+jokTm9em\nmA0WTl58sbntqVa5IA35CtNSXrluHYWT+LQ8nACfIAx4/UPR8iOBnwG4+1lmNgG4gDCb5zZgH3dP\nH8fMAXqB+cBY4HrguIa2XAYVQ7dOEk5UOclGV1cYu1FqinjeKyelXqugykksynXr6Fwn8Wl5OHH3\nqrqW3P0U4JQK93cBnyrcJCdimEo82IXq0uvI4MrNfIL8hxNVTuK2alXoykkGNatyEq9cjDmR9jXY\nVOK8d+uocjJ05QIpxDsgVpWTOKRPXQ8acxIzhRNpKFVOhp9K4STPlZPB2p2sI/mVvugfqHISM4UT\naajYB8SqcjJ05c62C/kOJ9V06+h1kG/p6+qAxpzETOFEGsa9/AyIPA2IVeUkW+045iQ5x01e2y5B\ncTgZMyYMzlblJD4KJ9IwlU5ulqcyuSon2Yq5W6dcODELO7q8tl2C4jEnZv3X15G4KJxIw1Q638WI\nEeFoNE/dOqqcZCPmAbHl2g35brsExWNOQFcmjpXCiTRMpXCSLM/DTl9niM1WO445gXy3XYLibh0I\n4URjTuKjcCINk1RFKoWTPFVOBjtDrFRnsDEnPT3Q19fcNlVjsHCiykn+FXfrgConsVI4kYYZrHIy\nblw+dvqqnGQr1im5ldoNqpzEoFS3jsacxEnhRBom+SCvND0zDx/2qpxkq5pwksewp8pJ/Mp16yic\nxKemcGJmh5vZfqmfzzKzF83sTjN7TXbNk5hVM+YkD906XV1hVH+pa8GMGhXu006pepXGnOQ57GnM\nSdy6u8P/R2NO2kOtlZOTgHUAZrYT4QJ7XwBWAHOzaZrELpZunWSMhNnG92kK6dANNuYE8rk9qwkn\nCqn51dkZvqpy0h5qvfDfq4F/FL4/ELjC3X9kZnew8dWFZZiKZbZOd3f5I33QTmmo2nXMiUJqvhVf\nkTgxfjy8+GLz2yP1qbVyshrYvPD9nsCNhe/XA+PrbZS0h2oqJ3np1hksnGinVL1Yw4kqJ3FLwokq\nJ+2h1srJjcBPzOw+4N+AawvL/wP4VxYNk/hVM5U4DzupSt0QoCPmoRrsJGyQv518X184o/FgA2L1\nOsivSt06GnMSn1orJ8cBdwEvBw5y9+cKy2cBv8iiYRK/mAbEqlsnO4OdhC1ZJ08Gm1kG+QnTUlq5\nbh1VTuJUU+XE3V80s88Dbwammdn+hbsWZtYyid5gH/jjxsHzzzevPeWocpKtGAfEDlblg/A6WLmy\nOe2RoSvXraPznMSppnBiZnsDPyOMOyme4+BAiUmZMtxUOrkZ5OdIVJWTbMU45iQJJ6qcxGvVqjC7\nbuLEgcvVrROnWrt1vgf8GtjC3UcU3RRMBAgf5KNHh4v8lZKXAbGqnGSrXcOJTsKWb52dsMkmG3/e\nqFsnTrWGk+nA2e6+PMvGSHuJ5XTgqpxkq5qTsOVtew42Piq5Lw+vVymt1NlhIXTrdHeHazpJPGoN\nJ/OB3TJsh7ShasKJKiftJ8YxJ888E75uvnn5dVQ5ybdy4WTChPBVXTtxqXUq8fHAr81sZ+CvwIb0\nne5+br0Nk/itX195p5+nM8SqcpKN3t5wiy2cLFkSvr72teXXUeUk36oJJ8UzeSS/ag0nHYSTr60n\nVFA8dZ8DCidCV1ccAwwH69ZR5aR6g3WPjBoVxgTkbXsuWQIzZoQugHIUUvOts7N0+EjCicadxKXW\ncHI6cDLwTXfvy7A90kZi6tbZbLPy948dC2vWNK89MRtshhbkcye/ZEnlqgkopOZdpTEnoHASm1rH\nnIwBfqVgIpUMFk7y0q2jykl2ktAR2xieJUtgq60qr5OXSp+UNli3jsJJXGoNJ5cAH8myIdJ+qp2t\n415+nWYYbEBsHo/08yrWWS+PPTZ4ONGA2HwbrFtHA2LjUmu3zkjgC2a2F/AXNh4Q+9l6Gybxq6Zy\n0tcXpviNHt28dhVT5SQ7MYaT7m5YtkyVk9ipctJeag0nbwLuK3y/XdF9LT4OlrwYbLZOeuZGK8OJ\nKifZqXbMSZ528kuXhupdNZWTZDbSSJ1qMnc05qS91Hptnd2zboi0n2q6dSCEmE02aU6bShlsKrEq\nJ9WrdsxJnsJeNdOIof9v6u6uPKtHms89hBPN1mkftY45ERnUYFOJk/tavePXGWKzE2O3zpIlYXrz\nlltWXi+vZ7eVMJ6kr6905ST5nNGYk7gonEjDVFs5afWOSmeIzU6M4eSxx+BVrxq8azEvr1fZWLkr\nEkMInroycXwUTqRhqhkQC60/14kqJ9mJccxJNdOIYWC3juRLpXACCicxUjiRhlHlZPipZsxJrOEk\nCVx5arsEnZ3ha7nT0+vKxPHJRTgxs53N7Goze8LM+sxs/6L7Ly4sT9+uLVpnrJmdZ2YrzKzTzOab\n2bTm/iWSNpQBsa1UTeVEO6TqVNOtk8cBsYMNhoXGh+mFC+Gmmxrz2O1usMrJhAkacxKbXIQTYCLw\nZ+BYyk9RK3A6AAAgAElEQVRFvg6YDswo3DqK7j8H2A84CNgF2AK4ohGNlepUc+E/aO2OPznPymA7\n076+MIVUKottzMmaNeGKxEOpnDQqWJ1wAhxzTGMeu91VE05UOYlLLsKJu1/v7l9196sAK7Nal7s/\n6+7PFG4rkzvMbBJwFDDH3W919/uAI4F3mNnsxv8FUko1F/5L1muVDYXTBw5WOYH87FDzLLYxJ//6\nV/g6lDEnjWj7E0/AnXfCo4/Cs89m//jt5Nln4dBDYdGi/mVJOCnXraMxJ/HJRTip0m5mttzMHjSz\n881sauq+WYRzttycLHD3h4ClwE7VPLg7fOADcPXVldc791z4zGeG3PZhKYZunWp3pjDwiPm662Dv\nvVVNKdbdHU5QVukkZVmEk85OuPBC2H13OPvs2h8nOcdJqysnv/kNWOGw7J57sn/8dnLFFfCLX8Db\n3w4/+lH47O7sDFe8LncwpMpJfGIJJ9cBhwHvAr4A7Apca5a8nZkBdLv7qqLfW164b1B33QVXXgnf\n/375ddzh29+G730PHn98iH/BMFTtbJ1WHkVXe9Iw6G9nXx984QuwYAHcfHP53xuOBvufQ+2zn9zh\n1lvhiCNgxgz4r/8K4eLUU2H16pqay5IlYQrxFlsMvm4jKyfz54ewO3063H139o/fThYsgB13hKOO\nCt1ghx0GTz0VunSsTN1dY07iE0U4cffL3f0ad3/A3a8G3gvMBnbL6jl++tPw9ZZb4LnnSq/zxz+G\nUOIOP/lJVs/cvmKYrVNL5eR3v4O//Q2mTOl/3UhQTTipdfbT974Hu+0Gt98OJ50UumT+7/9CMLnk\nkpqay5Il8JrXhHNhDKZRlZOnn4bbboMPfQje9rZswsmyZe1Z1duwIRwQvPe9cP75cNll4aDy9NPL\nd+mAKicxqvXaOi3l7kvMbAWwNfB74GlgjJlNKqqeTC/cV9acOXPYZJPJ3HBDOEPk0qVw0kkdXHBB\n8XjbcHTz8pfDAQeEcPKVr4RSomzMvboputDabp2hVk7cwwfhO98J73sfnHwyvPgibLZZ49sag8Fm\nPkFt3Trr18M3vgGHHw4XXzzwCPmgg+C734VPfrK6kJFW7TRiaFyYvvLK0A12wAGwfDmccUb56/f0\n9oaxFm99a/kqwbXXwv77hx34r341eFiMyT33hC6cPfcMPx9yCLzlLfDBD8K0CnMzNeakvM7OEGa3\n3bb8OvPmzWPevHkDlq1cubLM2tmIonJSzMxeBWwOPFVYtBDoAfZIrbMNsCVwV6XHmjt3LoceejU9\nPVdzyy1Xs+uuV/P44xsHE/cQTj7wATj2WHjySbjmmqz+ovZTzU7frPWDI4daOfnDH8IH5EknwUc/\nGpZdfnnDmxmNwQIp1PY/v/jiMKvmf/5n453yZz4DjzwSxgEN1WOPDT2cpCsnXV0hXNRTpZg/H971\nLpg6NVROOjth8eLS6158McyeDSeeGD6Tii1cCB/+cOj2uP76EFIavVPu64Pf/rY57+MFC8J2mjWr\nf9m228L994eKZjmqnJR3/PGwww6hy7Scjo4Orr766gG3uXPnNrRduQgnZjbRzLY3sx0Ki15X+PnV\nhfvOMrMdzew1ZrYH8L/Aw8ACgEK15ELgbDPbzcxmARcBd7j7vYM9/09/CjvvDK9/fUjgN90EL7ww\ncJ377gtHWR/8YEjqs2fDBRdktglyp6srHKFddFE4kiveHoNJqiHV7Kha2Rc81MrJGWeEN/Lee4dx\nCnvtpa6dtGrHnAxlR7ZhA5x1VtjpvuENG9//treF9+M55wytrTC0ykmpk7DNmRMOWC6+eOjPDWHm\nyR/+ED5XIFRERowo37Vz5ZWhenvmmfCpT4VgkHjssVAt+Y//gBtvDBWUO+4Ir9VVxaPxMnTJJSEE\n1bL9i91xB/zyl+XvX7AA3vOejatKo0b1X+CvFI05Ke2550J1beJEOPBA+PvfW92iFHdv+Y0wwLUP\n6C26XQSMA64ndM+sBx4FfgC8vOgxxgLfA1YAncCvgWkVnnMm4L/73UI3c7/wQnd39yeecAf3Sy7x\nAU480X3zzd27u8PPF17obub+6KPeVs4/3/3Nb3YfNSpsBzP3MWPcd9vNff366h9n+fLw+1ddVXm9\nt73N/f/9P/fe3vraXY3eXvevftX9b3/rX/bHP4Z23ndf+d/7y1/COueeG75efnn/fb/6VVj24ION\na3crzJ/vfsYZ7hs2DO33TjjBfbvtKq9z5pnum23W/3Nnp/upp7r/9Kel1//Zz8I2vv/+8o/5i1+E\ndf761+rb+sIL4Xd++cvq1u/rC+v/6Efh51/+Mvy81Vbur3qV+7p11T934kc/ch8xwv2ZZ/qXbb+9\n+3/+58brrloV3otz54bfM3M/4ojwP3r+efdtt3V/3evCey9x553ukye7v/Wt7s89N/T2DebFF92n\nTXMfPz58Xbu2tsfp7Q2vtxEj3EeOdF+6dON1nn02/M0XXTT0x//KV8L/SAY6+2z30aPdH3nE/U1v\ncn/Na9yffLK63124cKETzks20xuRCxrxoDHcknBy7LELfcKE8MZPvOMd7u97X//PfX3uW2898ANj\n9erwpj/xxDL/uQitXRs+ZN7znhBS7rwz/J233eY+dqz7IYeEbVGNpUvDq+v66yuv93//F9ar5QNn\nqH7/+/BcW2/tvnJlWHb77WHZAw+U/70HHwzrbLGF+zbbuPf09N+3bl3Y0Z50UkOb3lSLFoWdILjv\nvHMI7NU65hj3mTMrr3POOeF11tvrfvHF7q94Rdgpmbn/+tcD1+3tDTvd9PuxlO7u8P/5+Merb+ui\nReFvvPvu6n9n9Gj373/f/eGH3Tfd1L2jw/2hh8IO9eyzq3+cxJ57ur/rXQOXHXOM+3/8x8br/vrX\nob3JAdFll4Xn/dCH3Hfd1X3q1NCWYosWub/sZSH0JAdXWfnc59wnTnS/9dbwP/z+94f+GM89577f\nfuFv++//Dp+rX/zixuslYfDxx4f+HN/4Rtg+0q+vL3yeHXxw+Pnxx91f+crw/u3sHPz3FU4aHE5e\n/eqF/rGPDdzoc+eGD+dkB3b//WFLXXfdwPWOP959+nT3rq4S/7mCzs6w49111xB6LrggHG1koafH\n/dprw4fT298ejgTrce215XfUl18e7qt2J/zII2H93/9+8HU7OsJ2TLZ3oxx+eDh62nRT9498JLw5\nb7kltPMf/yj/e0uWhHUg7EyLffKT4XHToaUef/qT+447tqYas3Kl++tfHz6gbrop7PCnTXO/+ebq\nfv/II9132qnyOj/4QQgis2aFbfqRj4QdbkdHCMG3396/7hVXhHXuumvw5z7jjPD76SpEJb/5TXjs\ndKVhMJts4n766e477OD+hjf0H9T813+FAJA+yBnMihUhXJx//sDlF18ctk/x58RHPxqObtOuvDJ8\nVo0dGw4iyrn77vC3/uY31bdvMIsXhwrr6aeHnw85xH3LLYcWgO65JxytT50aPn/cQ+CZMiUcGKUd\neWTp0FaN737Xfdy42n63Vnfe6T57dunAmAfJwVr6M/rPfw6fj/vuO3jVVOGkweEEFm70wfuvf4Ut\nc9ll4eevfCUcHReHkL/+1Tcq87uHnd5tt4U308SJ4YPm3e9233vvcHQxfrz7xz4WXhSDVSK6usKO\nM327775QsXnlK8Pzb7ddONr4wAcqP95gJdfjjnN/7WvLP8a3vx2e74ILKj+Oe+g6gfAGHczjj7tP\nmOD++c8Pvm6tOjvD/+LrX+/vivnhD0PghNJl5ETS1Vfugzf54L/hhsHb0ddX+ahkyRL3GTPC4xWH\n5kbr63P/8IfdJ03qD2vLl4fX7ogR7qedNnj32yGHhC7ASi67LPx9s2YN3KGuXz+wAtDXF9bZfffq\n2r9iRdgBff3rA5evW1c6+H7nO+F1V2010D20bcqUEAb+/Of+5Y8/HpZ97WvVP9ZFF4XPhqeeGrh8\n8eKwfW68sX9Zd3d43i9/eePHufvuysEkseOO7nvtVf7+J58MYfSwwwavlvX1hc+z172uvzsr+Tws\nFeBLuffeEKx23DF85iaWLAmvtx/8YODzbbGF+2c/W91jF/vxj0PbmtF9nPjBD8JzTp8euoaz1Ntb\n/8HowQeHyknx63/BghA63/SmUOV89tnSv69w0uBwMmPGwpIv2B13dH//+8P3224bjrpLecc73PfY\nI3y/bFk4envDG/ylvuhTTx34xkvW2XrrsM4555R+XPfwAnz72/2lo/b0bfLkcMT+xz+GF9eVV/pL\n4yJK+clPwgsu/YGX1tcXgsmxx5ZvT19fCDAjR4YSa6U+9ltvDe1ZuLD8OmmnnRba16hqQXI0mvwv\njj22f2cy2NHzqlWhbeedV/r+vj73f//3sGOuZN260D2xySbu8+ZtfH963MD//E/YzpXGNC1dOvg4\ngmefdX/66crrJM4/v3TY7ulxP+WUsP323rv8h9WiRaGClLxvylm3zv0Pfyi9o0hvg5//fOOd9GCO\nPjp0E519dgh3220XtuP06RuHwuOPH/qR+CteUT6gf/az4aiz1PZ55JHwv0zvCPbd132XXTZet7c3\nHAyddlr/sqTC98c/Dq29acl74J//LH3/f/93eG2+7GUhtJ12WvkDmt/+NrTnyisHLj/ggI27PktZ\nvz5s+1mzSleeDzoovKeS10gSfAbrJi7n0kvD769ZU9vv1+Jb3wrbcYcdQqit53/3xBPh/3fCCaGr\nddNNw9/zne/U9njLl4cuynJdkbfeGv4Ho0eH20EHuV9zzcBqisJJg8PJxz9eeu/5rW+Fo7B77glb\n6be/LbnaS4P1kqPLpCpyyy2VU3pfX+hbnjSp/M7jwgv9pSORW27pv916a+kPjc98JryQit8E3/1u\neJyxY90/+MHSz/X3v4d1fve78m12Dx86BxwQ1k3S9WGHhTfJV77i/t739ld0zNwfe6zy4yXWrg3h\naJ99qlt/qHbdNfyPEuvWub/lLeF/BoMfhfzjH5WPsL/5zfB6Kddlt3p1CLHjxoWdEoSAlAwyLq4a\nrFkTdhKf/GTpx1u2LATUMWNCteO66/p3CBs2uF99tfuBB4b/0bhx/QO+y0nGmRx3XPl1brghtOlV\nrxpYEevrCwM0x44N3UH1DhJfsiSECQiDpYdS2XjggfA3jx8fDjCOOSZ0044atfEH8X77hdfrULzl\nLe6HHlq6Tc88E3bu6QrgsmXh8yA5qNhss1BZSt6r5Q4m9tortC/x6U+H99VQtkWxNWvCa+ZLX9r4\nvueeC20/8cTwXvjc50L7Xv3qUHV46KH+z7P168PB1Xves3F7kipi8dihYl/9anj8chWF227zAV3p\n3/lOeB3XOuA26cIrF6wb4eSTw//shRfCwP9NNw1j7Ibq6af7Q/Eb3hDe72ecEV7bpQ4mqnHmmeH9\nWs3BzTnnhEkSxQebCicNDif/+7+lw8mjj4ats/324UVVbqbKunXhKO9tbxv6eJIVK0KpttTI/GQU\nfEdH9Y/X1RU+zLfaqn9ne/rp/tJAs+98J+yAVqzY+He//e3q3/w9PSG0/fCH4Q0ye3b43enTw5H1\nl74Uuk6WLKm+7e79HyDXXDO03xvMP/8ZHvfnPx+4/JFH+o9Aav3QSyxbFoLOKads3PXz4ouhwrbJ\nJqFi0NcXSr5jxoRZFP/8Z+nxFqefHpYVj57v6ws71Ve8wv2ss8IRKIQPwsMP79+xv+UtYef38Y+H\nn484ovSR4/33948zGWzGybJl4W8ZNSrs9Fev7t/5fuITtc1YKeVPfwp/z4IFQ//dFSs27i8/4ojQ\nLZB+H7/xjaF6MhSrV1c+6PjqV8N74eGHQ9V0woQQ6M4/PwT/r389HCC8/vUhqJTrPjn55DA7sK+v\nv6pZLqgOxac+FT5XiqsVp5wSAl26gvjww/0HIhBev+94RwglI0eWH0S+xx7htVcuSP35z+H1c/LJ\n5duZdOkl3VB77hlutbr++vA3pKvYjfa5z4Xqj3uo2u2+e9jGP/tZ9TMfe3rCQdW0aRsPBO7tLf25\nMZje3vD6G0q3cV9fCJLp/6nCSYPDycIK/Q7JgL1DDy27St3OOy9UGIqrHZ/9bPhgG+rI9EcfDR96\nH/hAOAqC8CHZ1xc+eEaNKj2ifvfdwxF9reo5oks/xh57hHD1wx+GADRYaHj++VBNOvvssDMv9aY/\n+eQQQooH2LmHsvTOO2fTF52EgGnTwgfT3/4Wjjxmzgwh9J57Bq7/pz+Fv3X0aC95BPTCC6GyVjwW\nJylRJ9O0+/pC//0nPxk+DE84YeOp0ZdcEj4Yt9sudJ0991x4Hcyc2R9sHnmkur+zuzv8fUklYMKE\n/vFZWcriNZV48MHwPkumAff1hXbXWhYvZ+XKUP0aNSr8Xz//+fIHLJX+vmQs1COP9A/Ir7VLIy0Z\nC/arX/UvW7UqvD5POKH07zzzTKianXlmGKew7bYhzJRz881ecgKBe3jtvOUt4XVYaSKBe3+33sKF\nIfDV879KZgU2c5D5MceEg4/E2rWhewTCa+SEEwaOWyrl1FPD67Zc12ZxxbUaN9wQ2jCUQFOKwkkL\nw8k3vuGZj3AvtmFD6BrZaaf+D6tkFHzxwL5qJeNPSvVJ7r//wDeMe/jwrDSmopkefDDsMEeODO0f\nMSIc4e6zTwhPyW2vvcLRZPJ3jhsX/obi6c69vWG9UtWpRrj//lCyf9nLQrsmTQphpdw5Ol54IQyc\n/uEPS99/0klhIG9Sfn3qqfBBNJSKWuKvfw3jASZMCEdbI0eG18P//m9tU0yvvDJUcCpNw86TD30o\nHDFu2NB/Hp5GvLcvvTTMrKk27JXy3HP+UrXv1FMrV2+H6p3vHDh9+dvfDu+dSoPCh6KvL3Spbbdd\nqHylx5+cfnp4T1cz/qKrKwwO33bbsC2Gcg6bYsn5jBYtqv0xhurQQ0sPDv/730MlO13h/M1vNg6r\nt9wSttVXv1r5ecqd46acgw4K1dZ6w7/CSQvDyfLloa83q1J1OcmUrp//vH8U/FZb1fe83/9+6aPZ\npOsk/UafPz8sG2o3TCOtWxc+UH70o1AROPDAgbf3vz8clV56adg5btjQPwsnPd052bbVzGbIUldX\nmAb7n/8Zwmatli8PFY/kSPWgg9xf/vLa+85XrQoVtbPO2niWSLtLzmty2WX9YyMqnXyv1bbZJowB\nmjUrjDPISlKReOih8D6bMcP9qKOye3z3UCVMQsWrXx1mGV17bejKLHUOk3JOO81fquzVszN94AHP\npFowFAccUHlMU3d3GBv2nveEtu2+e38l5emnw/9lt92qO0VBMk5rq61CxeYHPwiv8TVrQjfnTTeF\ncZSHHhqC6Pe+V//fp3DSwnDSTB/6UBhDkJzpsngUfFa6ukJfdrqr4MgjQ3WiHXzrWz5gNsXhh4ej\n5Sy7CJrthBNC2f2ii3yjkrwMzT77hCP65H1W73TMRjr88DB9PX1agyysWxeqb5/7XNiJjRgRxpdk\nra8v7CCPPjpUECEErqGM73rmmVDlO/LI+tqSnKuomun+Wdljj/4TnA3m2mtDl+yIEWF7vetdoeJa\n7dla3cOYkPQMtWRSQlJdnjAhVOg/97lsZi0pnAyTcPLYY6FrYsSIMACqkTvTE04IKXvDhtDtMWNG\nKDO2g/R058svD10i6SmZMVq6tP9yAoOdy0YqS2aBvP3tA0+hn0fJeTJGjco+RH32s+Eg5bWvDSfB\na7Q1a8L7sZYTkt1yS/1dTtVeTiNLs2cP7YzF3d1hZuWUKZXHmVQjqTz/+MchiC9enN1JIhMKJ8Mk\nnLiHvuXRoxvfh5+Ut6+5JgzKhDCLpF309ITziSRHDs0cod8oRx8djnaHW1dMI+yyi7/U159n990X\n2pmcRylLySUZYPBBme1g1arwt5Y6v1CjvPGNYfzZUK1YUd85UZql0eEkF1clluDLX4Zly+CNb2zs\n8+ywA7z5zeFqur/7HUyeDG9/e2Ofs5lGjoR582CnncLVUrfcstUtqt9558HDD8OMGa1uSfxOOil8\nrfZqxK2y3Xbh/33IIdk/9jbbwD77wPvfD9tvn/3j58348eHr2rXNe87Vq2GTTYb+e5tvHq5OPdyN\nanUDpJ8ZTJvWnOc54gj40pfggQdgzz1h9OjGP28zTZwYLr/e09PqlmRj1KjwoSX123NPeM97YOed\nW92SykaNgscegzFjGvP4V13VmMfNo1GjwnZct655z1lrOJFAlZNh6tBDoa8PFi+G/fZrdWsaw6z9\nQpfUzwxuuAE+85lWt2RwY8eG9jbC6NHD6/0xYULzKyebbtq852s3CifD1LRpsO++4fu9925tW0RE\nGm38+OaFk+7ucFPlpHbq1hnGvva1UNqePr3VLRERaaxmVk7WrAlfFU5qp3AyjO2wQ7iJiLS7CROa\nN+Zk9erwVeGkdurWERGRttfMyonCSf0UTkREpO01c8xJZ2f4qnBSO4UTERFpe6qcxEXhRERE2p7G\nnMRF4URERNqeKidxUTgREZG218wxJ6tXw4gR/afNl6FTOBERkbbX7MrJJps07uy+w4HCiYiItL1m\njzlRl059FE5ERKTtNbtbR+GkPgonIiLS9lrRrSO1UzgREZG2l4QT98Y/V2enwkm9FE5ERKTtTZgQ\ngkl3d+OfS5WT+unCfyIi0vaSab3PPw/LlsGf/wz33QezZ8MRR2T7XKtXw9Sp2T7mcKNwIiIibW/C\nhPD1la8MFZQRI8Kyu+9uTDjZcstsH3O4UbeOiIi0vXe+E770JfjBD+Cee0KAOOGEUEnJ2urVsOmm\n2T/ucKLKiYiItL1Jk+Ab3xi4bOpUeOGF7J9LY07ql4vKiZntbGZXm9kTZtZnZvuXWOdUM3vSzNaa\n2Y1mtnXR/WPN7DwzW2FmnWY238ymNe+vEBGRmEyZAqtWwYYN2T6uwkn9chFOgInAn4FjgY0mepnZ\nF4HjgaOB2cAaYIGZjUmtdg6wH3AQsAuwBXBFY5stIiKxSgatvvhito+rcFK/XHTruPv1wPUAZiWv\nRvBp4DR3v6awzmHAcuBA4HIzmwQcBRzs7rcW1jkSWGxms9393ib8GSIiEpEknLzwArz85dk8Zk8P\nrF+vcFKvvFROyjKzrYAZwM3JMndfBdwD7FRY9FZC0Eqv8xCwNLWOiIjIS6ZMCV+zHBS7enX4qnBS\nn9yHE0IwcUKlJG154T6A6UB3IbSUW0dEROQlSeVE4SR/YggnIiIimUt362RF4SQbuRhzMoinASNU\nR9LVk+nAfal1xpjZpKLqyfTCfWXNmTOHyZMnD1jW0dFBR0dHve0WEZEcGz8exo5V5WQw8+bNY968\neQOWrVy5sqHPmftw4u5LzOxpYA/gLwCFAbA7AucVVlsI9BTWubKwzjbAlsBdlR5/7ty5zJw5szGN\nFxGRXJs6tTHhpJ1OwlbqgH3RokXMmjWrYc+Zi3BiZhOBrQkVEoDXmdn2wPPu/jhhmvCXzewfwGPA\nacAy4CoIA2TN7ELgbDN7AegEzgXu0EwdEREpJ+sTsbVj5aQVchFOCLNtfk8Y+OrAdwrLLwGOcvez\nzGwCcAGwGXAbsI+7p68vOQfoBeYDYwlTk49rTvNFRCRGU6aoWyePchFOCucmqTg4191PAU6pcH8X\n8KnCTUREZFCN6NYx678KstRGs3VERGTYyjqcdHbCxInhqsdSO20+EREZtqZMyX7Mibp06qdwIiIi\nw1YjunUUTuqncCIiIsNWEk58o0vO1kbhJBsKJyIiMmxNmRIu1rdmTTaPp3CSDYUTEREZtrK+vo7C\nSTYUTkREZNhqRDhpp7PDtorCiYiIDFtTpoSvWc3YUeUkGwonIiIybKlbJ58UTkREZNiaPDmc0TWr\ncNLZqXCSBYUTEREZtkaODAFF3Tr5onAiIiLDWpYnYlM4yYbCiYiIDGtZhZPeXli3TuEkCwonIiIy\nrGV1fZ3kRG4KJ/VTOBERkWEtq8rJ6tXhq8JJ/RRORERkWMs6nOgkbPVTOBERkWFt6tRsunVUOcmO\nwomIiAxrU6aoWydvFE5ERGRYmzo1nDxtw4b6HqezM3xVOKmfwomIiAxrySnsX3yxvsdR5SQ7Cici\nIjKsJRf/q7drJwknEybU9ziicCIiIsNcVhf/W706BJORI+tv03CncCIiIsNaEk7qnbGjU9dnR+FE\nRESGtSy7dRROsqFwIiIiw9r48TBuXDbhRCdgy4bCiYiIDHtZnIhNlZPsKJyIiMiwl8WJ2BROsqNw\nIiIiw14W19fp7FQ4yYrCiYiIDHvq1skXhRMRERn21K2TLwonIiIy7GXRraNwkh2FExERGfbUrZMv\nCiciIjLsJd067rU/hsJJdhRORERk2Js6FXp6+i/eN1R9fbBmjU7ClpUowomZnWxmfUW3vxetc6qZ\nPWlma83sRjPbulXtFRGRuNR7fZ21a8NXVU6yEUU4KfgbMB2YUbi9M7nDzL4IHA8cDcwG1gALzGxM\nC9opIiKRqff6Op2d4avCSTZGtboBQ9Dj7s+Wue/TwGnufg2AmR0GLAcOBC5vUvtERCRSSeWk1nCS\ndAcpnGQjpsrJG8zsCTP7p5ldamavBjCzrQiVlJuTFd19FXAPsFNrmioiIjFROMmXWMLJ3cARwF7A\nJ4CtgP8zs4mEYOKESkna8sJ9IiIiFU2eDGa1jzlROMlWFN067r4g9ePfzOxe4F/Ah4EH63nsOXPm\nMHny5AHLOjo66OjoqOdhRUQkIiNGwGabqXJSyrx585g3b96AZStXrmzoc0YRToq5+0ozexjYGvgD\nYITBsunqyXTgvsEea+7cucycObMRzRQRkYjUc5bYdg4npQ7YFy1axKxZsxr2nLF06wxgZpsQgsmT\n7r4EeBrYI3X/JGBH4M7WtFBERGIzZUr93ToTJ2bXnuEsisqJmX0L+C2hK+eVwNeADcAvC6ucA3zZ\nzP4BPAacBiwDrmp6Y0VEJEr1Vk7GjYNRUexV8y+Wzfgq4BfA5sCzwO3A29z9OQB3P8vMJgAXAJsB\nt5PY78kAAAe4SURBVAH7uHt3i9orIiKRmToVlhdPrajS6tU6O2yWoggn7j7o6FR3PwU4peGNERGR\ntjRlCjxY4xSLzs72HG/SKlGOOREREclavd06CifZUTgRERFB4SRPFE5EREQI3TqrV8OGDUP/XYWT\nbCmciIiIUN+ViRVOsqVwIiIiQn3X11E4yZbCiYiICKFbB1Q5yQOFExEREVQ5yROFExEREforJ088\nMfTf1UnYsqVwIiIiAowfD699LRxzDOyxB1x6KaxdW93v6iRs2YriDLEiIiLN8Le/wfz5cPHF8LGP\nwXHHwXvfO3hVZM0ahZMsKZyIiIgUTJwIhx8ebv/8J/z0p3DDDdDbW/n3dtwx3CQbCiciIiIlvP71\ncNpp4SbNpTEnIiIikisKJyIiIpIrCiciIiKSKwonIiIikisKJyIiIpIrCiciIiKSKwonIiIikisK\nJyIiIpIrCiciIiKSKwonIiIikisKJyIiIpIrCiciIiKSKwonIiIikisKJyIiIpIrCiciIiKSKwon\nIiIikisKJyIiIpIrCiciIiKSKwonIiIikisKJyIiIpIrCifSVPPmzWt1E4YdbfPm0zZvPm3z9tJ2\n4cTMjjOzJWa2zszuNrP/1+o2ST99gDSftnnzaZs3n7Z5e2mrcGJmHwG+A5wMvAW4H1hgZi9racNE\nRESkam0VToA5wAXu/jN3fxD4BLAWOKq1zRIREZFqtU04MbPRwCzg5mSZuztwE7BTq9olIiIiQzOq\n1Q3I0MuAkcDyouXLgW1KrD8OYPHixQ1ulqStXLmSRYsWtboZw4q2efNpmzeftnlzpfad4xrx+BaK\nC/Ezs1cATwA7ufs9qeVnAru4+05F6x8CXNbcVoqIiLSVQ939F1k/aDtVTlYAvcD0ouXTgadLrL8A\nOBR4DFjf0JaJiIi0l3HAawn70sy1TeUEwMzuBu5x908XfjZgKXCuu3+rpY0TERGRqrRT5QTgbOCn\nZrYQuJcwe2cC8NNWNkpERESq11bhxN0vL5zT5FRCd86fgb3c/dnWtkxERESq1VbdOiIiIhK/tjnP\niYiIiLQHhRMRERHJlWEbTnSBwMYwsxPN7F4zW2Vmy83sSjP7txLrnWpmT5rZWjO70cy2bkV725GZ\nfcnM+szs7KLl2uYZMrMtzOznZraisE3vN7OZRetom2fEzEaY2Wlm9mhhe/7DzL5cYj1t8xqZ2c5m\ndrWZPVH4DNm/xDoVt6+ZjTWz8wrvi04zm29m04balmEZTnSBwIbaGfgesCPwbmA0cIOZjU9WMLMv\nAscDRwOzgTWE7T+m+c1tL4WQfTThNZ1erm2eITPbDLgD6AL2ArYFPge8kFpH2zxbXwKOAY4F/h34\nAvAFMzs+WUHbvG4TCRNJjgU2GpBa5fY9B9gPOAjYBdgCuGLILXH3YXcD7ga+m/rZgGXAF1rdtna7\nES4r0Ae8M7XsSWBO6udJwDrgw61ub8w3YBPgIeBdwO+Bs7XNG7atvwncOsg62ubZbvPfAj8uWjYf\n+Jm2eUO2dx+wf9Gyitu38HMX8P7UOtsUHmv2UJ5/2FVOdIHAptuMkMCfBzCzrYAZDNz+q4B70Pav\n13nAb939lvRCbfOGeB/wJzO7vNB9ucjMPp7cqW3eEHcCe5jZGwDMbHvgHcC1hZ+1zRuoyu37VsIp\nStLrPEQ4GeqQ/gdtdZ6TKg31AoFSo8IZes8Bbnf3vxcWzyCElVLbf0YTm9dWzOxgYAfCh0MxbfPs\nvQ74JKF7+HRCiftcM+ty95+jbd4I3yQcmT9oZr2EYQn/4+6/LNyvbd5Y1Wzf6UB3IbSUW6cqwzGc\nSPOcD7yRcHQjDWJmryKEwHe7+4ZWt2eYGAHc6+5fKfx8v5ltB3wC+HnrmtXWPgIcAhwM/J0Qxr9r\nZk8WAqG0kWHXrcPQLxAoNTCz7wP7Aru5+1Opu54mjPHR9s/OLODlwCIz22BmG4BdgU+bWTfhqEXb\nPFtPAYuLli0Gtix8r9d59s4Cvunuv3b3B9z9MmAucGLhfm3zxqpm+z4NjDGzSRXWqcqwCyeFI8uF\nwB7JskL3wx6EPk2pUyGYHADs7u5L0/e5+xLCizS9/ScRZvdo+9fmJuBNhCPJ7Qu3PwGXAtu7+6No\nm2ftDjbuBt4G+Bfodd4gEwgHlml9FPZj2uaNVeX2XQj0FK2zDSG03zWU5xuu3Tq6QGCDmNn5QAew\nP7DGzJKUvdLd1xe+Pwf4spn9A3gMOI0wW+qqJje3Lbj7GkKZ+yVmtgZ4zt2To3tt82zNBe4wsxOB\nywkf0B8H/iu1jrZ5tn5L2J7LgAeAmYTP7p+k1tE2r4OZTQS2JlRIAF5XGHj8vLs/ziDb191XmdmF\nwNlm9gLQCZwL3OHu9w6pMa2ertTCaVLHFjbuOkKie2ur29QON8KRTG+J22FF651CmJa2FlgAbN3q\ntrfTDbiF1FRibfOGbON9gb8UtucDwFEl1tE2z257TyQcWC4hnF/jEeBrwCht88y28a5lPsMvqnb7\nAmMJ57paUQgnvwamDbUtuvCfiIiI5MqwG3MiIiIi+aZwIiIiIrmicCIiIiK5onAiIiIiuaJwIiIi\nIrmicCIiIiK5onAiIiIiuaJwIiIiIrmicCIiIiK5onAiIiIiuaJwIiIiIrny/wFq9cr3l/1cugAA\nAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x108942e48>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(times_ms[:100])\n",
    "plt.ylabel('ms')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# scratch"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Get in/out bytes from /proc/net/dev\n",
    "def network_bytes():\n",
    "  import subprocess\n",
    "  proc = subprocess.Popen(['cat', '/proc/net/dev'], stdout=subprocess.PIPE)\n",
    "  stdout,stderr = proc.communicate()\n",
    "  stdout=stdout.decode('ascii')\n",
    "\n",
    "  recv_bytes = 0\n",
    "  transmit_bytes = 0\n",
    "  lines=stdout.strip().split('\\n')\n",
    "  lines = lines[2:]  # strip header\n",
    "  for line in lines[2:]:\n",
    "    line = line.strip()\n",
    "    # ignore loopback interface\n",
    "    if line.startswith('lo'):\n",
    "      continue\n",
    "    toks = line.split()\n",
    "\n",
    "    recv_bytes += int(toks[1])\n",
    "    transmit_bytes += int(toks[9])\n",
    "  return recv_bytes, transmit_bytes\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0, 0)"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "network_bytes()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "yaro = [\n",
    "  '--lr-sched', '0.14,0.47,0.78,0.95',\n",
    "  '--epochs', 45,\n",
    "  '--lr', 0.4,\n",
    "  '--dist-url', 'file:///home/ubuntu/data/file.sync', # single instances are faster with file sync\n",
    "  '--init-bn0',\n",
    "  '--batch-sched', 64,\n",
    "  '--num-tasks', 1\n",
    "]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['--num-tasks']"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "[v for v in yaro if v=='--num-tasks']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['--lr-sched',\n",
       " '0.14,0.47,0.78,0.95',\n",
       " '--epochs',\n",
       " 45,\n",
       " '--lr',\n",
       " 0.4,\n",
       " '--dist-url',\n",
       " 'file:///home/ubuntu/data/file.sync',\n",
       " '--init-bn0',\n",
       " '--batch-sched',\n",
       " 64,\n",
       " '--num-tasks',\n",
       " 1]"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "yaro"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "12"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(yaro)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "del yaro[12], yaro[11]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['--lr-sched',\n",
       " '0.14,0.47,0.78,0.95',\n",
       " '--epochs',\n",
       " 45,\n",
       " '--lr',\n",
       " 0.4,\n",
       " '--dist-url',\n",
       " 'file:///home/ubuntu/data/file.sync',\n",
       " '--init-bn0',\n",
       " '--batch-sched',\n",
       " 64]"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "yaro"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "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": "3.6.5"
  },
  "toc": {
   "colors": {
    "hover_highlight": "#DAA520",
    "running_highlight": "#FF0000",
    "selected_highlight": "#FFD700"
   },
   "moveMenuLeft": true,
   "nav_menu": {
    "height": "30px",
    "width": "252px"
   },
   "navigate_menu": true,
   "number_sections": false,
   "sideBar": true,
   "threshold": 4,
   "toc_cell": false,
   "toc_section_display": "block",
   "toc_window_display": true
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}