notebooks/SoftwareAnalysis.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# PyTorch software benchmark analysis\n",
    "\n",
    "Single node throughput comparisons between installation versions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import re\n",
    "import glob\n",
    "\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "plt.rcParams.update({'font.size': 14})"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load the results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[0m\u001b[01;34mgpu-ngc_20.03_v0-nccl-n1\u001b[0m/  \u001b[01;34mgpu-v1.5.0-gloo-n2\u001b[0m/   \u001b[01;34mgpu-v1.5.0-nccl-n2\u001b[0m/\n",
      "\u001b[01;34mgpu-v1.1.0-nccl-n1\u001b[0m/        \u001b[01;34mgpu-v1.5.0-gloo-n32\u001b[0m/  \u001b[01;34mgpu-v1.5.0-nccl-n32\u001b[0m/\n",
      "\u001b[01;34mgpu-v1.2.0-nccl-n1\u001b[0m/        \u001b[01;34mgpu-v1.5.0-gloo-n4\u001b[0m/   \u001b[01;34mgpu-v1.5.0-nccl-n4\u001b[0m/\n",
      "\u001b[01;34mgpu-v1.3.1-nccl-n1\u001b[0m/        \u001b[01;34mgpu-v1.5.0-gloo-n64\u001b[0m/  \u001b[01;34mgpu-v1.5.0-nccl-n64\u001b[0m/\n",
      "\u001b[01;34mgpu-v1.4.0-nccl-n1\u001b[0m/        \u001b[01;34mgpu-v1.5.0-gloo-n8\u001b[0m/   \u001b[01;34mgpu-v1.5.0-nccl-n8\u001b[0m/\n",
      "\u001b[01;34mgpu-v1.5.0-gloo-n1\u001b[0m/        \u001b[01;34mgpu-v1.5.0-nccl-n1\u001b[0m/   \u001b[01;34mhsw-v1.5.0-mpi-n1\u001b[0m/\n",
      "\u001b[01;34mgpu-v1.5.0-gloo-n16\u001b[0m/       \u001b[01;34mgpu-v1.5.0-nccl-n16\u001b[0m/  \u001b[01;34mknl-v1.5.0-mpi-n1\u001b[0m/\n"
     ]
    }
   ],
   "source": [
    "ls /global/cscratch1/sd/sfarrell/pytorch-benchmarks/results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['/global/cscratch1/sd/sfarrell/pytorch-benchmarks/results/hsw-v1.5.0-mpi-n1/results.txt']"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "results_files = glob.glob('/global/cscratch1/sd/sfarrell/pytorch-benchmarks/results/hsw-*-n1/results.txt')\n",
    "\n",
    "results_files"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>train_rate</th>\n",
       "      <th>inference_rate</th>\n",
       "      <th>ranks</th>\n",
       "      <th>hardware</th>\n",
       "      <th>version</th>\n",
       "      <th>backend</th>\n",
       "      <th>model</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>627.266843</td>\n",
       "      <td>1936.923361</td>\n",
       "      <td>1</td>\n",
       "      <td>hsw</td>\n",
       "      <td>v1.4.0</td>\n",
       "      <td>mpi</td>\n",
       "      <td>lstm</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>203.579070</td>\n",
       "      <td>849.929892</td>\n",
       "      <td>1</td>\n",
       "      <td>hsw</td>\n",
       "      <td>v1.4.0</td>\n",
       "      <td>mpi</td>\n",
       "      <td>alexnet</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>14.873055</td>\n",
       "      <td>34.675494</td>\n",
       "      <td>1</td>\n",
       "      <td>hsw</td>\n",
       "      <td>v1.4.0</td>\n",
       "      <td>mpi</td>\n",
       "      <td>resnet50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>115.603731</td>\n",
       "      <td>228.013513</td>\n",
       "      <td>1</td>\n",
       "      <td>hsw</td>\n",
       "      <td>v1.4.0</td>\n",
       "      <td>mpi</td>\n",
       "      <td>cnn3d</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>642.970956</td>\n",
       "      <td>2556.861520</td>\n",
       "      <td>1</td>\n",
       "      <td>hsw</td>\n",
       "      <td>v1.3.1</td>\n",
       "      <td>mpi</td>\n",
       "      <td>lstm</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>5</td>\n",
       "      <td>204.135253</td>\n",
       "      <td>606.233518</td>\n",
       "      <td>1</td>\n",
       "      <td>hsw</td>\n",
       "      <td>v1.3.1</td>\n",
       "      <td>mpi</td>\n",
       "      <td>alexnet</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>6</td>\n",
       "      <td>15.196844</td>\n",
       "      <td>29.728784</td>\n",
       "      <td>1</td>\n",
       "      <td>hsw</td>\n",
       "      <td>v1.3.1</td>\n",
       "      <td>mpi</td>\n",
       "      <td>resnet50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>7</td>\n",
       "      <td>201.667720</td>\n",
       "      <td>407.555616</td>\n",
       "      <td>1</td>\n",
       "      <td>hsw</td>\n",
       "      <td>v1.3.1</td>\n",
       "      <td>mpi</td>\n",
       "      <td>cnn3d</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>8</td>\n",
       "      <td>776.840712</td>\n",
       "      <td>2576.349997</td>\n",
       "      <td>1</td>\n",
       "      <td>hsw</td>\n",
       "      <td>v1.1.0</td>\n",
       "      <td>mpi</td>\n",
       "      <td>lstm</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>9</td>\n",
       "      <td>170.298037</td>\n",
       "      <td>689.007436</td>\n",
       "      <td>1</td>\n",
       "      <td>hsw</td>\n",
       "      <td>v1.1.0</td>\n",
       "      <td>mpi</td>\n",
       "      <td>alexnet</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>10</td>\n",
       "      <td>13.970692</td>\n",
       "      <td>22.921428</td>\n",
       "      <td>1</td>\n",
       "      <td>hsw</td>\n",
       "      <td>v1.1.0</td>\n",
       "      <td>mpi</td>\n",
       "      <td>resnet50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>11</td>\n",
       "      <td>206.463203</td>\n",
       "      <td>438.406375</td>\n",
       "      <td>1</td>\n",
       "      <td>hsw</td>\n",
       "      <td>v1.1.0</td>\n",
       "      <td>mpi</td>\n",
       "      <td>cnn3d</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>12</td>\n",
       "      <td>652.895608</td>\n",
       "      <td>2706.192794</td>\n",
       "      <td>1</td>\n",
       "      <td>hsw</td>\n",
       "      <td>v1.2.0</td>\n",
       "      <td>mpi</td>\n",
       "      <td>lstm</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>13</td>\n",
       "      <td>214.485323</td>\n",
       "      <td>610.464312</td>\n",
       "      <td>1</td>\n",
       "      <td>hsw</td>\n",
       "      <td>v1.2.0</td>\n",
       "      <td>mpi</td>\n",
       "      <td>alexnet</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>14</td>\n",
       "      <td>14.350628</td>\n",
       "      <td>43.360747</td>\n",
       "      <td>1</td>\n",
       "      <td>hsw</td>\n",
       "      <td>v1.2.0</td>\n",
       "      <td>mpi</td>\n",
       "      <td>resnet50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>15</td>\n",
       "      <td>197.393156</td>\n",
       "      <td>410.678840</td>\n",
       "      <td>1</td>\n",
       "      <td>hsw</td>\n",
       "      <td>v1.2.0</td>\n",
       "      <td>mpi</td>\n",
       "      <td>cnn3d</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    train_rate  inference_rate  ranks hardware version backend     model\n",
       "0   627.266843     1936.923361      1      hsw  v1.4.0     mpi      lstm\n",
       "1   203.579070      849.929892      1      hsw  v1.4.0     mpi   alexnet\n",
       "2    14.873055       34.675494      1      hsw  v1.4.0     mpi  resnet50\n",
       "3   115.603731      228.013513      1      hsw  v1.4.0     mpi     cnn3d\n",
       "4   642.970956     2556.861520      1      hsw  v1.3.1     mpi      lstm\n",
       "5   204.135253      606.233518      1      hsw  v1.3.1     mpi   alexnet\n",
       "6    15.196844       29.728784      1      hsw  v1.3.1     mpi  resnet50\n",
       "7   201.667720      407.555616      1      hsw  v1.3.1     mpi     cnn3d\n",
       "8   776.840712     2576.349997      1      hsw  v1.1.0     mpi      lstm\n",
       "9   170.298037      689.007436      1      hsw  v1.1.0     mpi   alexnet\n",
       "10   13.970692       22.921428      1      hsw  v1.1.0     mpi  resnet50\n",
       "11  206.463203      438.406375      1      hsw  v1.1.0     mpi     cnn3d\n",
       "12  652.895608     2706.192794      1      hsw  v1.2.0     mpi      lstm\n",
       "13  214.485323      610.464312      1      hsw  v1.2.0     mpi   alexnet\n",
       "14   14.350628       43.360747      1      hsw  v1.2.0     mpi  resnet50\n",
       "15  197.393156      410.678840      1      hsw  v1.2.0     mpi     cnn3d"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "results = [pd.read_csv(f, delim_whitespace=True) for f in results_files]\n",
    "results = pd.concat(results, ignore_index=True)\n",
    "\n",
    "results"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Training throughput comparisons"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "models = results.model.unique()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x720 with 4 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, axs = plt.subplots(nrows=2, ncols=2, figsize=(16, 10))\n",
    "axs = axs.flatten()\n",
    "\n",
    "for i, model in enumerate(models):\n",
    "    ax = axs[i]\n",
    "    model_results = results[results.model == model].sort_values('version')\n",
    "    ax.bar(model_results.version, model_results.train_rate)\n",
    "    ax.set_title(model)\n",
    "    ax.set_ylabel('samples / s')\n",
    "\n",
    "plt.tight_layout()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Inference throughput comparisons"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABHAAAALACAYAAAD7bIWVAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy8QZhcZAAAgAElEQVR4nOzdf7hlZV3//+dLUEBAJZlpUC8cTEMC+ioO3xwCNWwQJM20PqaAYAYkQpIYYZCOlkmoCBTW4C9sjLSiElD5lR+EGESHNEH5UQr+AGbmIIiAg+b0/vyx1mE2m5lhH9hn73XOeT6ua117rft+r7XuNZuZ6+a973XfqSokSZIkSZLUXY8ZdwMkSZIkSZK0aSZwJEmSJEmSOs4EjiRJkiRJUseZwJEkSZIkSeo4EziSJEmSJEkdZwJHkiRJkiSp40zgSJoxkhyWpJIsHHdbJEmSBpFkYdt/OWzcbZE0s5nAkTSrJTkpySvG3Q5JkqSZJMk2SZYmedG42yKpYQJH0mx3EmACR5IkaWq2Ad4BvGjM7ZDUMoEjSZIkSZLUcSZwJM1YSZ6Z5B+S3J7k/iTfTfKPSZ7a1hewBXBo++55JbmsrZucT+eFSU5NsibJ3Uk+mmTLJFsnWZbkjrb8zCSbj/FxJUlShyR5epIPJrkhyY+S/CDJ+Ul2HfD8HZKcleS2JD9O8l9Jjk+Snph3tv2V/fvOPTnJ/yb5lfZ4cp6dE5K8rm3Tj5N8LcmvTvXe7XyDt7fh7+jpR539iP6wJA2F/zMiaUZK8ljgImAr4EyaTsYOwEuApwG3AocAHwW+CJzVnrq671KnARPAO4HnAa8H7gN+DvgR8CfAvsBRwH+18ZIkSXsCLwT+GbgFeArwe8DlSXatqlUbOzHJfJr+yWNp+ii3A/sAf9Fe59g29E+BA4GPJNmtqu5Kshh4K3BGVf3fvkv/JrA9sAxY217nX5I8varunMK9J4Cjgb8C/qV9RoBvTu2PSNIwparG3QZJGki7esPHgJ2AJwJfBX6rqv5pE+fcD3yyqg7byLUuBfar9h/DJJ+nedf776rqkJ74bwDrqmr34T2RJEmaqZI8vqp+1Ff2TOA64E+r6t1t2ULgZuD1VXV2W3YW8OvAL1bV6p7zTwGOA36uqm5py34BuIYmiXI48J/AOuC5VbW27x4/AH6+qiba8ucAXwGOrqozp3LvJAtokjvvrKqlj/bPS9Kj5ytUkmaqH7afL0my9aO4zkfrwZnsq4EAH+mLu5pmVI4kSRK9yZskj0/yZJoEyk00o3o3qH1N6TeBC4B1Sbaf3GhGFz+GZmTP5H2+AZwIvBb4PLAQeN1k8qbPP04mb9pzv0rTZ3rGI7m3pG4xgSNpRqqqm4HTgd8F7khyaZJj2w7IVHyn7/ju9vO7GyjfKskWU2+tJEmabdo5805JchvN69d30Lx6tDvwpE2cOg/YDvidNr53u7SNmd93zmnAl4FfAt5bVV/ayLW/vYGyu4CfeRT3ltQRzoEjacaqqmOTfAh4ObAf8D7gpCQvrKqvD3iZdVMsz0bKJUnS3DL5Q9JfAlfS/NjzvzTJlk39UD5Z9/c0c/VtyH/1HT8d2KXd320T1364/ssjubekjjCBI2lGaxM1Xwfek+QXad4RP5bmHXEAJ/qSJEnT4dXA31bVsb2FSbajGY2zMRM0rzVtXlWXbiJu8nqhmbfvJ8A7gHcmeX1VfewRtHkq97YPJXWMr1BJmpGSPGEDy3pfT7PiwnY9Zff1HUuSJA3DOvpG5iZ5Dc1KThtVVeuAfwJ+I8ke/fVJntiutjnpWJp5aY6qqncBFwKnJdlxqg2e4r3vaz/tR0kd4QgcSTPVvsCZSf4JuJGmA/VqYFvgkz1xK4FfTfJW4HvAmqr6/KgbK0mSZp3zgNcl+SHNylPPoemLfGuAc0+gScpcmeQjwLXAE2hej3oV8ExgVZKdgXcD/1BVn2rP/d32fh9NsqRvMYZBDHTvqro3yY3Abye5Cfg+cHNVXT3F+0kaEhM4kmaq/wQ+B7yU5nWp+2lepXpFVX26J+7NwF8DS4GtgS/QrOAgSZL0aLwZ+B+apM0baH402h9478OdWFUTSX4JOAl4BXAE61ewWgrcmWQz4G9p5tY5qufcW5McDXwCeBPwV1Np9CD37gl/Pc2cPu8HtgA+TrMyp6QxyNQTtpIkSZIkSRol58CRJEmSJEnqOBM4kiRJkiRJHWcCR5IkSZIkqeNM4EiSJEmSJHWcCRxJkiRJkqSOcxnxHttvv30tXLhw3M2QJEmPwDXXXHNHVc0bdzvGzf6MJEkz16b6MyZweixcuJCVK1eOuxmSJOkRSPLtcbehC+zPSJI0c22qP+MrVJIkSZIkSR1nAkeSJEmSJKnjTOBIkiRJkiR1nAkcSZIkSZKkjjOBI0mSJEmS1HEmcCRJkiRJkjrOBI4kSZIkSVLHmcCRJEmSJEnqOBM4kiRJkiRJHWcCR5IkSZIkqeNM4EiSJEmSJHWcCRxJkiRJkqSO23zcDZBmuoUnfGbcTZizbjn5wHE3QZIkSRoL/z9kfMb1/yGOwJEkSZIkSeo4EziSJEmSJEkdZwJHkiRJkiSp40zgSJIkSZIkdZwJHEmSJEmSpI4zgSNJkiRJktRxJnAkSZIkSZI6zgSOJEmSJElSx5nAkSRJkiRJ6jgTOJIkSZIkSR1nAkeSJEmSJKnjTOBIkiQNUZLNkvxpkpuT3N9+/lmSzXtikmRpktuSrE1yWZJd+66zXZLlSe5ut+VJnjT6J5IkSV1gAkeSJGm4/gh4E/D7wLOBN7fHb+uJOR44DjgG2BNYA1ySZNuemHOAPYADgP3b/eXT3XhJktRNmz98iIZh4QmfGXcT5qxbTj5w3E2QJM0tewHnV9X57fEtSc4Dfgma0TfAscDJVXVuW3YoTRLntcCyJLvQJG32rqoVbcyRwBVJdq6qG0f6RJIkaewcgSNJkjRc/w78SpJnAyT5BWBf4LNt/U7AAuDiyROqai1wOU3yB2AxcC+woue6VwL39cRIkqQ5xBE4krQRjpwbH0fOaYb7C2Bb4BtJ1tH0t95dVR9s6xe0n6v7zlsNPLUnZqKqarKyqirJmp7zJUnSHDKyEThJ3pbky0l+mGQiyflJduuLOTtJ9W1f7IvZIslfJrkjyX1JzkvytL6YHdvr39fGnZHkcaN4TkmSNOe9GngdzetQe7T7RyV5Q19c9R2nr6y/fkMxTWFyRJKVSVZOTEw84oZLkqTuGuUrVC8CPkgz7Hdf4KfApUl+pi/uUmCHnu2lffWnAa8CXgPsAzwBuCDJZtCs/AB8huaXr33auN8E3j/0J5IkSXqo9wLvq6pPVtW1VbUcOJX1kxivaj/7R9LMZ/2onFXA/Ha+HOCBuXPm8dCRO1TVWVW1qKoWzZs3b4iPIkmSumJkr1BV1Ut6j5McAtwN/DJwfk/Vj6tqFRuQ5InAG4DXV9UlPdf5NvCrwEXAfsCuwNOr6rttzPHAh5OcWFU/HOqDSZIkPdjjgXV9ZetY/8PZzTQJmiXAlwGSbEnzw9MftjFXAdvQzIUzOQ/OYmBrHjwvjiRJmiPGOYnxtu397+or3zvJmiQ3JflQkvk9dc8DHsuDJ/37LnA9D5707/rJ5E3rImCL9nxJkqTpdD5wQpIDkyxM8hvAW4B/gWYuG5oRxSckeWX7SvnZNJMWn9PGXA9cSLMi1fOTLAaWARe4ApUkSXPTOCcxPh34Ks0vTJMuBP6Z5pephcCfAZ9P8ryq+jHNUON1wB1911rN+mHIC3jo0OI72vOc9E+SJE23Y4A/pXl1fD5wO/Ah4F09MacAWwFnAtsBVwP7VdU9PTEHAWew/oer84Cjp7XlkiSps8aSwElyKrA3sHdVPTDEuKo+2RN2bZJraF6POpAmsbPRS/Lwk/5tsDzJEcARADvuuONA7ZckSdqYNglzbLttLKaApe22sZg7gYOH3DxJkjRDjfwVqiQfoJlYeN+q+tamYqvqNuB7wLPaolXAZsD2faH9k/71j7TZvj3PSf8kSZIkSdKMM9IETpLTaZbU3LeqbhggfnvgqTRDjwGuAf6HZtK/yZinAbuwfkK/q4Bd+pYWXwL8uD1fkiRJkiRpRhnZK1RJzgQOAV4B3JVkcpTMvVV1b5JtaIYRn0uTsFkIvAdYw/pJ/+5O8hHgvUnWAN+nWZbzazTLj0PznvjXgb9NchzwZJrlPD/kClSSJEmSJGkmGuUInKNoVp76N5oEzeT21rZ+HbA78GngJuDjwI3A4r4J/f6AZj6cTwFX0qzY8LLJuXTazwOBH7X1n2rj34okSZIkSdIMNLIROFWVh6lfC7xkgOvcT7O6wzGbiPkO8GtTbaMkSZIkSVIXjXwSY0mSJEmSJE2NCRxJkiRJkqSOM4EjSZIkSZLUcSZwJEmSJEmSOs4EjiRJkiRJUseZwJEkSZIkSeo4EziSJEmSJEkdZwJHkiRJkiSp40zgSJIkSZIkdZwJHEmSJEmSpI4zgSNJkiRJktRxJnAkSZIkSZI6zgSOJEmSJElSx5nAkSRJkiRJ6jgTOJIkSZIkSR1nAkeSJEmSJKnjTOBIkiRJkiR1nAkcSZIkSZKkjjOBI0mSNERJbklSG9g+0xNzVJKbk9yf5Jok+/RdY4skf5nkjiT3JTkvydNG/zSSJKkrTOBIkiQN157ADj3bHkAB/wCQ5NXA6cCfA88FVgCfS7JjzzVOA14FvAbYB3gCcEGSzUb0DJIkqWNM4EiSJA1RVU1U1arJDXgp8EPgH9uQtwBnV9WHqur6qjoGuB14I0CSJwJvAP6wqi6pqv8ADgF+EfjVUT+PJEnqBhM4kiRJ0yRJaJIxn6iqHyV5HPA84OK+0IuBvdr95wGP7Y2pqu8C1/fESJKkOcYEjiRJ0vRZAuwEfLg93h7YDFjdF7caWNDuLwDWAXdsIuZBkhyRZGWSlRMTE8NotyRJ6hgTOJIkSdPncODLVfXVvvLqO84GyvptNKaqzqqqRVW1aN68eY+spZIkqdNM4EiSJE2DJPOBXwc+1FN8B83omv6RNPNZPypnFc0one03ESNJkuYYEziSJEnT4zDgx8AnJwuq6ifANTSvVvVaQrMaFW39//TGtEuI79ITI0mS5pjNx90ASZKk2aadvPh3gU9W1T191acCy5N8CbgS+D3gKcDfAFTV3Uk+Arw3yRrg++05XwMuHdEjbNDCEz4zztvPabecfOC4myBJGjMTOJIkScP3IuBZwMH9FVX1qSRPBk4CdgCuA15aVd/uCfsD4KfAp4CtgH8DXldV66a53ZIkqaNM4EiSJA1ZVf1fmkmHN1b/QeCDm6i/Hzim3SRJkpwDR5IkSZIkqetM4EiSJEmSJHWcCRxJkiRJkqSOM4EjSZIkSZLUcSZwJEmSJEmSOs4EjiRJkiRJUseZwJEkSZIkSeo4EziSJEmSJEkdZwJHkiRJkiSp40zgSJIkSZIkdZwJHEmSJEmSpI4zgSNJkiRJktRxI0vgJHlbki8n+WGSiSTnJ9mtLyZJlia5LcnaJJcl2bUvZrsky5Pc3W7LkzypL2b3JF9or3FrkrcnySieU5IkSZIkadhGOQLnRcAHgb2AfYGfApcm+ZmemOOB44BjgD2BNcAlSbbtiTkH2AM4ANi/3V8+WZnkCcAlwOr2Gr8P/CHwlul4KEmSJEmSpOm2+ahuVFUv6T1OcghwN/DLwPntCJljgZOr6tw25lCaJM5rgWVJdqFJ2uxdVSvamCOBK5LsXFU3AgcBjwcOraq1wHXteW9JcmpV1SieV5IkSZIkaVjGOQfOtu3972qPdwIWABdPBrQJmMtpRu0ALAbuBVb0XOdK4L6+mCvacyddBDwFWDjUJ5AkSZIkSRqBcSZwTge+ClzVHi9oP1f3xa3uqVsATPSOomn31/TFbOgavfeQJEmSJEmaMUb2ClWvJKcCe9O8CrWur7r/Faf0lW3oFaiHi8lGyklyBHAEwI477rjphkuSJEmSJI3ByEfgJPkA8Bpg36r6Vk/Vqvazf5TMfNaPoFkFzO9dUardn9cXs6FrwENH5lBVZ1XVoqpaNG/evKk+jiRJkiRJ0rQbaQInyek0ExLvW1U39FXfTJN8WdITvyWwD+vnvLkK2IZmnptJi4Gt+2L2ac+dtAS4DbhlKA8iSZIkSZI0QiNL4CQ5E3g9zeibu5IsaLdt4IG5bE4DTkjyyiS7AWfTTFp8ThtzPXAhzYpUz0+yGFgGXNCuQEUb+yPg7CS7JXklcALgClSSJEmSJGlGGuUcOEe1n//WV/5OYGm7fwqwFXAmsB1wNbBfVd3TE38QcAbrV6s6Dzh6srKq7k6ypL3GSppVrt4PnDqsB5EkSZIkSRqlkSVwqioDxBRNMmfpJmLuBA5+mOtcC7xgai2UJEkajiQ7ACcDLwW2Bb4FvLGqvtDWB3gHzUIKkz9avamqvt5zje1ofrR6eVt0HnBMVf1gVM8hSZK6YyyrUEmSNE4LT/jMuJswZ91y8oHjbsK0S/Ik4Erg34EDgQngGcCanrDjgeOAw4AbgbcDlyTZuWfk8TnAjsABNCtpfhhYDrxs+p9Cc43/Lo7PXPh3UdJwmMCRJEkaruOB26vqdT1lN0/utKNvjgVOrqpz27JDaRI8r6WZ628XYH9g76pa0cYcCVzRJnluRJIGYHJufEzOadhGvoy4JEnSLPcK4Ookn0qyJslXkxzdJm4AdgIWsH4+P6pqLXA5sFdbtJhmIYcVPde9ErivJ0aSJM0hJnAkSZKG6xk0izd8C3gJcDrNfDhvausXtJ+r+85b3VO3AJjoXUGz3V/TE/OAJEckWZlk5cTExLCeQ5IkdYgJHEmSpOF6DPAfVfW2qvpKVX2MZjLiN/XFVd9x+sr66zcU0wRWnVVVi6pq0bx58x5F0yVJUleZwJEkSRqu24Fv9JVdTzMhMcCq9rN/JM181o/KWQXM73ntanLunHk8dOSOJEmaA0zgSJIkDdeVwM59ZT8PfLvdv5kmQbNksjLJlsA+rJ/z5ipgG5q5cCYtBrbmwfPiSJKkOcJVqCRJkobrA8CKJCcCnwKeC/w+8MfQzGWT5DTgxCQ3ADcBJ9FMWnxOG3N9kgtpVqQ6nObVqWXABa5AJUnS3GQCR5IkaYiq6stJXgH8OfAnwHfazw/2hJ0CbAWcCWwHXA3sV1X39MQcRDN3zuRqVecBR09v6yVJUleZwJEkSRqyqvoM8JlN1BewtN02FnMncPCw2yZJkmYm58CRJEmSJEnqOBM4kiRJkiRJHWcCR5IkSZIkqeNM4EiSJEmSJHWcCRxJkiRJkqSOM4EjSZIkSZLUcQMlcJL8nyT79Ry/Pcn3klyUZIfpa54kSdLo2OeRJEldNegInKWTO0n2AP4YOAN4LPD+4TdLkiRpLJZO7tjnkSRJXbL5gHFPB25s938D+NeqOiXJxcBF09IySZKk0bPPI0mSOmnQETj3A9u2+y8GLm337+4plyRJmuns80iSpE4adATOFcD7k/w7sAj4zbb854HvTkfDJEmSxsA+jyRJ6qRBR+AcDfyEphPze1V1W1t+AA4nliRJs4d9HkmS1EkDjcCpqu8BL9tA+bFDb5EkSdKY2OeRJEldNegIHEmSJEmSJI2JCRxJkiRJkqSOM4EjSZIkSZLUcSZwJEmSJEmSOm6TCZwk70+yTxITPZIkadayzyNJkrru4Topjwf+Hlid5Owkr0iy1QjaJUmSNEr2eSRJUqdtMoFTVW+sqqcBBwK3An8G3JHkvCS/k2TeKBopSZI0nezzSJKkrhtomHBVfamqTqyq3YD/D/gCcBjwvST/nuStSZ46je2UJEmadvZ5JElSV035Pe+q+u+qen9VvQB4KvBRYG/gNcNunCRJ0rjY55EkSV3yqCbqq6o7quqjVfWKqnrfsBolSZLUJVPp8yRZmqT6tlU99WljbkuyNsllSXbtu8Z2SZYnubvdlid50nQ9nyRJ6j5XWpAkSRq+G4Ederbde+qOB44DjgH2BNYAlyTZtifmHGAP4ABg/3Z/+fQ3W5IkddXm426AJEnSLPTTqlrVX5gkwLHAyVV1blt2KE0S57XAsiS70CRt9q6qFW3MkcAVSXauqhtH9RCSJKk7HIEjSZI0fM9IcmuSm5N8Mskz2vKdgAXAxZOBVbUWuBzYqy1aDNwLrOi53pXAfT0xkiRpjjGBI0mSNFxX06xcdQBwOE3CZkWSJ7f7AKv7zlndU7cAmKiqmqxs99f0xDxIkiOSrEyycmJiYljPIUmSOmSgBE6SFyb5pZ7jw9qlNJcl2Wb6midJkjQ6w+jzVNXnquofquprVXUp8Gs0fa5De8P6b91X1l+/oZjee55VVYuqatG8efMGaaYkSZphBh2BcxrtLz5JdgaWAV+jGeL73ulpmiRJ0sgNvc9TVfcCXweeBUzOi9M/kmY+60flrALmt/Pl0LYlwDweOnJHkiTNEYMmcH4OuLbdfxVwSVUdRTMs+GXT0TBJkqQxGHqfJ8mWwLOB24GbaRI0S/rq92H9nDdXAdvQJI0mLQa25sHz4kiSpDlk0AROAZu1+y8GLmz3VwFPHnajJEmSxuRR93mSvK99FWun9nWsf6JJvny8ncvmNOCEJK9MshtwNs2kxecAVNX17X2XJXl+ksU0I4EucAUqSZLmrkGXEf8y8CdJLqH5heiItnwh64cCS5IkzXTD6PM8Dfh7YHtgAvgi8Pyq+nZbfwqwFXAmsB3NpMf7VdU9Pdc4CDiD9atVnQcc/QieR5IkzRKDjsA5FngO8FfAu6vqm235bzGFobxJXpDkvHZZzUpyWF/92W157/bFvpgtkvxlkjuS3Nde72l9MTsmOb+tvyPJGUkeN2g7JUnSnPWo+zxV9dtV9ZSqelxVPbWqXlVV3+ipr6paWlU7VNWWVfXCqrqu7xp3VtXBVfWEdju4qn4wpGeUJEkz0EAjcNpOxS9uoOqtwLop3G8b4Drgb9ttQy4FDuk5/klf/WnArwOvAb4PnApckOR5VbUuyWbAZ9q6fWiGO3+cZuWGY6bQVkmSNMcMsc8jSZI0VIO+QgVAkkU0k/tdUFX30bwj/tNBz6+qzwKfba919kbCflxVGxyinOSJwBuA11fVJW3ZIcC3gV8FLgL2A3YFnl5V321jjgc+nOTEqvrhoO2VJElz06Pt80iSJA3bQK9QJfnZJFcDX6KZYO9n26pTgfcPuU17J1mT5KYkH0oyv6fuecBjWf8+OG2S5npgr7ZoMXD9ZPKmdRGwRXu+JEnSBo24zyNJkjSwQefA+QDrV1/4UU/5P9KMeBmWC4HX0az6cBzw/wOfT7JFW7+AZvjyHX3nrW7rJmNW99Xf0Z63oK+cJEckWZlk5cTExFAeQpIkzVij6vNIkiRNyaCvUL0YeHFV3ZWkt/ybwI7DakxVfbLn8Nok19C8HnUg8M+bODU0y34+cKmN3WID9zwLOAtg0aJFGztPkiTNDSPp80iSJE3VoCNwtuKhkwkDzAPuH15zHqyqbgO+BzyrLVpF8w769n2h81k/6mYVDx1ps317Xv/IHEmSpF5j6fNIkiQ9nEETOJcDh/UcV7va0x8B/zbsRk1Ksj3wVOD2tuga4H+AJT0xTwN2Yf3SnlcBu/QtLb4E+HF7viRJ0saMpc8jSZL0cAZ9hep44AtJ9qSZDPj9NCs9PRH45UFvlmQb4Jnt4WOAHZM8B7iz3ZYC59IkbBYC7wHWAP8CUFV3J/kI8N4ka1i/jPjXaJYfh2aC468Df5vkOJp32N8LfMgVqCRJ0sMYSp9HkiRp2AYagVNV3wB2pxnlcjGwJc1kfs+tqm9O4X6LgK+021bAO9v9d9FMMrw78GngJuDjwI3A4qq6p+caf0AzH86ngCuBe4GXVdW6tq3raObM+VFb/6k2/q1TaKckSZqDhtjnkSRJGqpBR+BQVauAdzyam1XVZTQTDm/MSwa4xv3AMe22sZjvAL821fZJkiQNo88jSZI0bBtN4CTZY9CLVNV/DKc5kiRJo2WfR5IkzQSbGoGzkmbZ7U2NmKGN2WxoLZIkSRot+zySJKnzNpXA2WlkrZAkSRof+zySJKnzNprAqapvj7IhkiRJ42CfR5IkzQQDT2KcZAfgjcAvtEXXA39dVbdNR8MkSZLGwT6PJEnqooGWEU+yBPgm8Gqa5bl/BPwW8N9J9pu+5kmSJI2OfR5JktRVg47AOQP4MPDmqqrJwiSnA6cDu0xD2yRJkkbNPo8kSeqkgUbgAAuBv+rtyLTOBJ4+1BZJkiSNz0Ls80iSpA4aNIGzEth9A+W7A18ZXnMkSZLGyj6PJEnqpEFfofog8IEkzwK+2JY9n2aCvxOS7DEZWFX/MdwmSpIkjYx9HkmS1EmDJnD+rv38803UARSw2aNqkSRJ0vjY55EkSZ00aAJnp2lthSRJUjfY55EkSZ00UAKnqr493Q2RJEkat+no8yT5Y+DdwJlVdXRbFuAdwBHAdsDVwJuq6us9521HsyrWy9ui84BjquoHw26jJEnqvkFH4JBkAbAXMJ++yY+r6oNDbpckSdJYDLPPk+T5wOHA1/qqjgeOAw4DbgTeDlySZOequqeNOQfYETiA5pWtDwPLgZdNpQ2SJGl2GCiBk+Rgmk5DgLtoOhGTimbCP0mSpBltmH2eJE+kmTfnDTQJmsnyAMcCJ1fVuW3ZocAa4LXAsiS7APsDe1fVijbmSOCKNslz4yN+SEmSNCMNuoz4u4FTgK2rakFV7dCzPWUa2ydJkjRKw+zznAX8U1V9vq98J2ABcPFkQVWtBS6nGfkDsBi4F1jRc96VwH09MZIkaQ4Z9BWqJwBnV9VPp7MxkiRJYzaUPk+Sw4FnAodsoHpB+7m6r3w18NSemImqemAEUFVVkjU950uSpDlk0BE4fwccOJ0NkSRJ6oBH3edJsjPNMuQHVdVPNhFafcfhoa9sPeTyGypPckSSlUlWTkxMTLXJkiRpBhh0BM5bgH9N8mLgWuB/eiur6l3DbpgkSdIYDKPPsxjYHriume4GgM2AFyT5PWDXtmwB8N2e8+azflTOKmB+kkyOwmnnzpnHQ0fuUFVn0byyxaJFizaU+JEkSTPcoAmcI2km0ruDZjhw/69DJnAkSdJsMIw+z78CK/vKPiX2CuEAACAASURBVAb8F83InJtoEjRLgC8DJNkS2Af4wzb+KmAbmmTQ5Dw4i4GtefC8OJIkaY4YNIHzJ8BxVfWB6WyMJEnSmD3qPk9V/QD4QW9ZkvuAO6vquvb4NODEJDfQJHROopm0+Jz2GtcnuZBmRarDaV6dWgZc4ApUkiTNTYMmcDYDzpvOhkiSJHXAqPo8pwBbAWcC2wFXA/tV1T09MQcBZ7B+tarzgKNH0DZJktRBgyZwPkbTifBVKUmSNJtNS5+nql7Ud1zA0nbb2Dl3AgcPsx2SJGnmGjSB83jgd5O8BPgaD53Q7/eH3TBJkqQxsM8jSZI6adAEzi7AV9r9Z/fVudKBJEmaLezzSJKkThoogVNVvzLdDZEkSRo3+zySJKmrHjPuBkiSJEmSJGnTBn2FiiS/ArwG2BF4XG9dVe075HZJkiSNhX0eSZLURQONwElyGPA5YFvgRcAEzZKXewDfmKa2SZIkjZR9HkmS1FWDvkL1VuDoqnoNzWoMb6uq5wKfAO6drsZJkiSNmH0eSZLUSYMmcJ4BXNru/xjYpt3/K+CwIbdJkiRpXOzzSJKkTho0gfN9mqHEALcCu7X7Twa2GnajJEmSxsQ+jyRJ6qRBJzG+AtgPuBb4B+CMJEuAFwOXTFPbJEmSRs0+jyRJ6qRBEzhHA1u2++8Bfgr8Mk3H5s+moV2SJEnjYJ9HkiR10kAJnKq6s2f/f4G/mLYWSZIkjYl9HkmS1FWDLiP+C0l27jlekuQTSd6WZLPpa54kSdLo2OeRJEldNegkxh8BnguQ5GnAp4GfAd6Ew4klSdLsYZ9HkiR10qAJnF2A/2j3fwu4uqpeChwCvGY6GiZJkjQG9nkkSVInDZrA2Qz4Sbv/YuCz7f43gZ8ddqMkSZLGxD6PJEnqpEETONcBb0yyD01n5sK2/KnAHdPRMEmSpDGwzyNJkjpp0ATOHwGHA5cBf19V17blLwe+NA3tkiRJGgf7PJIkqZMGSuBU1eXAPGD7qvqdnqplwBsHvVmSFyQ5L8mtSSrJYX31SbI0yW1J1ia5LMmufTHbJVme5O52W57kSX0xuyf5QnuNW5O8PUkGbackSZqbhtXnkSRJGrZBR+BQVeuq6q6+sluqas0U7rcNzdDkNwNrN1B/PHAccAywJ7AGuCTJtj0x5wB7AAcA+7f7yycrkzwBuARY3V7j94E/BN4yhXZKkqQ5akh9HkmSpKHafJQ3q6rP0k4GmOTs3rp2hMyxwMlVdW5bdihNEue1wLIku9AkbfauqhVtzJHAFUl2rqobgYOAxwOHVtVa4Lr2vLckObWqagSPKkmSJEmSNDQDj8AZgZ2ABcDFkwVtAuZyYK+2aDFwL7Ci57wrgfv6Yq5oz510EfAUYOF0NFySJEmSJGk6dSmBs6D9XN1XvrqnbgEw0TuKpt1f0xezoWv03uMBSY5IsjLJyomJiUfRfEmSJEjypiRfS/LDdrsqyYE99UOZ80+SJM0tXUrgTOp/xSl9ZRt6BerhYrKRcqrqrKpaVFWL5s2bN9W2SpIk9fsezWpWewCLgM8D/5rkF9v6Rz3nnyRJmnu6lMBZ1X72j5KZz/oRNKuA+b0rSrX78/piNnQNeOjIHEmSpKGqqk9X1eeq6r+r6qaqOhG4B1jcP+dfVV0HHApsSzPnHz1z/h1RVSuq6irgSODXkuw8loeSJElj16UEzs00yZclkwVJtgT2Yf2cN1fRrGS1uOe8xcDWfTH7tOdOWgLcBtwyHQ2XJEnakCSbJfltmv7LCoY3558kSZpjRprASbJNkuckeU577x3b4x3buWxOA05I8sokuwFn03RgzgGoquuBC2lWpHp+ksXAMuCCdgUq2tgfAWcn2S3JK4ETAFegkiRJI5Fk9yT3Aj8G/gb4jaq6luHN+SdJkuaYUY/AWQR8pd22At7Z7r+rrT8FOBU4E1gJ7ADsV1X39FzjIOA/aX65uqjdP2Sysqruphlx85T2GmcC72+vK0mSNAo3As8Bng/8NfDx9sepScOY8299hYsySJI0620+yptV1WWsn1B4Q/UFLG23jcXcCRz8MPe5FnjBI2mjJEnSo1VVPwH+uz1cmWRP4A+Ad7dlC4Dv9pyywTn/JkfhbGDOv/77nQWcBbBo0SJHHEuSNAt1aQ4cSZKk2eoxwBYMb84/SZI0x4x0BI4kSdJsl+Rk4DM0I2wmV5d6EXBgVVWS04ATk9wA3AScRN+cf0km5/w7nGb0cv+cf5IkaY4xgSNJkjRcC4BPtJ93A18DDqiqi9r6U2jmAjwT2A64mg3P+XcG61erOg84evqbLkmSusoEjiRJ0hBV1WEPUz+UOf8kSdLc4hw4kiRJkiRJHWcCR5IkSZIkqeNM4EiSJEmSJHWcCRxJkiRJkqSOM4EjSZIkSZLUcSZwJEmSJEmSOs4EjiRJkiRJUseZwJEkSZIkSeo4EziSJEmSJEkdZwJHkiRJkiSp40zgSJIkSZIkdZwJHEmSJEmSpI4zgSNJkiRJktRxJnAkSZIkSZI6zgSOJEmSJElSx5nAkSRJkiRJ6jgTOJIkSZIkSR1nAkeSJEmSJKnjTOBIkiRJkiR1nAkcSZIkSZKkjjOBI0mSJEmS1HEmcCRJkiRJkjrOBI4kSZIkSVLHmcCRJEkaoiRvS/LlJD9MMpHk/CS79cUkydIktyVZm+SyJLv2xWyXZHmSu9tteZInjfZpJElSV5jAkSRJGq4XAR8E9gL2BX4KXJrkZ3pijgeOA44B9gTWAJck2bYn5hxgD+AAYP92f/l0N16SJHXT5uNugCRJ0mxSVS/pPU5yCHA38MvA+UkCHAucXFXntjGH0iRxXgssS7ILTdJm76pa0cYcCVyRZOequnFkDyRJkjrBETiSJEnTa1uaPtdd7fFOwALg4smAqloLXE4zagdgMXAvsKLnOlcC9/XESJKkOcQEjiRJ0vQ6HfgqcFV7vKD9XN0Xt7qnbgEwUVU1Wdnur+mJeUCSI5KsTLJyYmJimG2XJEkdYQJHkiRpmiQ5FdgbeFVVreurrv7wvrL++g3FNIFVZ1XVoqpaNG/evEfTZEmS1FEmcCRJkqZBkg8ArwH2rapv9VStaj/7R9LMZ/2onFXA/Ha+nMnrBZjHQ0fuSJKkOcAEjiRJ0pAlOZ1mQuJ9q+qGvuqbaRI0S3ritwT2Yf2cN1cB29DMhTNpMbA1D54XR5IkzRGuQiVJkjRESc4EDgFeAdyVZHKkzb1VdW9VVZLTgBOT3ADcBJxEM2nxOQBVdX2SC2lWpDqc5tWpZcAFrkAlSdLcZAJHkiRpuI5qP/+tr/ydwNJ2/xRgK+BMYDvgamC/qrqnJ/4g4AzWr1Z1HnD0NLRXkiTNACZwJEmShqiqMkBM0SRzlm4i5k7g4KE1TJIkzWjOgSNJkiRJktRxJnAkSZIkSZI6zgSOJEmSJElSx3UqgZNkaZLq21b11KeNuS3J2iSXJdm17xrbJVme5O52W57kSaN/GkmSJEmSpOHoVAKndSOwQ8+2e0/d8cBxwDHAnsAa4JIk2/bEnAPsARwA7N/uL5/+ZkuSJEmSJE2PLq5C9dOqWtVfmCTAscDJVXVuW3YoTRLntcCyJLvQJG32rqoVbcyRwBVJdq6qG0f1EJIkSZIkScPSxRE4z0hya5Kbk3wyyTPa8p2ABcDFk4FVtRa4HNirLVoM3Aus6LnelcB9PTGSJEmSJEkzStcSOFcDh9G8/nQ4TcJmRZInt/sAq/vOWd1TtwCYqKqarGz31/TEPEiSI5KsTLJyYmJiWM8hSZIkSZI0NJ16haqqPtd7nOSLwLeAQ4EvTob1nZa+sv76DcX03vMs4CyARYsWbTBGkiRJkiRpnLo2AudBqupe4OvAs4DJeXH6R9LMZ/2onFXA/Ha+HOCBuXPm8dCRO5IkSZIkSTNCpxM4SbYEng3cDtxMk6BZ0le/D+vnvLkK2IZmLpxJi4GtefC8OJIkSZIkSTNGp16hSvI+4HzgOzQja/6EJvny8aqqJKcBJya5AbgJOIlm0uJzAKrq+iQX0qxIdTjNq1PLgAtcgUqSJEmSJM1UnUrgAE8D/h7YHpigmffm+VX17bb+FGAr4ExgO5pJj/erqnt6rnEQcAbrV6s6Dzh6+psuSZIkSZI0PTqVwKmq336Y+gKWttvGYu4EDh5qwyRJkiRJksao03PgSJIkSZIkyQSOJEmSJElS55nAkSRJkiRJ6jgTOJIkSZIkSR1nAkeSJEmSJKnjTOBIkiRJkiR1nAkcSZIkSZKkjjOBI0mSJEmS1HEmcCRJkoYsyQuSnJfk1iSV5LC++iRZmuS2JGuTXJZk176Y7ZIsT3J3uy1P8qSRPogkSeoMEziSJEnDtw1wHfBmYO0G6o8HjgOOAfYE1gCXJNm2J+YcYA/gAGD/dn/5NLZZkiR12ObjboAkSdJsU1WfBT4LkOTs3rokAY4FTq6qc9uyQ2mSOK8FliXZhSZps3dVrWhjjgSuSLJzVd04qmeRJEnd4AgcSZKk0doJWABcPFlQVWuBy4G92qLFwL3Aip7zrgTu64mRJElziAkcSZKk0VrQfq7uK1/dU7cAmKiqmqxs99f0xDwgyRFJViZZOTExMQ1NliRJ42YCR5IkaTyq7zh9Zf31G4ppAqvOqqpFVbVo3rx5Q2yiJEnqChM4kiRJo7Wq/ewfSTOf9aNyVgHz2/lygAfmzpnHQ0fuSJKkOcAEjiRJ0mjdTJOgWTJZkGRLYB/Wz3lzFc1KVot7zlsMbM2D58WRJElzhKtQSZIkDVmSbYBntoePAXZM8hzgzqr6TpLTgBOT3ADcBJxEM2nxOQBVdX2SC2lWpDqc5tWpZcAFrkAlSdLc5AgcSZKk4VsEfKXdtgLe2e6/q60/BTgVOBNYCewA7FdV9/Rc4yDgP2lWq7qo3T9kFI2XJEnd4wgcSZKkIauqy2hGzWysvoCl7baxmDuBg4fcNEmSNEM5AkeSJEmSJKnjTOBIkiRJkiR1nAkcSZIkSZKkjjOBI0mSJEmS1HEmcCRJkiRJkjrOBI4kSZIkSVLHmcCRJEmSJEnqOBM4kiRJkiRJHWcCR5IkSZIkqeNM4EiSJEmSJHWcCRxJkiRJkqSOM4EjSZIkSZLUcSZwJEmSJEmSOs4EjiRJkiRJUseZwJEkSZIkSeo4EziSJEmSJEkdZwJHkiRJkiSp40zgSJIkSZIkdZwJHEmSJEmSpI4zgSNJkiRJktRxJnAkSZIkSZI6zgSOJEmSJElSx83qBE6So5LcnOT+JNck2WfcbZIkSZoK+zOSJAlmcQInyauB04E/B54LrAA+l2THsTZMkiRpQPZnJEnSpFmbwAHeApxdVR+qquur6hjgduCNY26XJEnSoOzPSJIkYJYmcJI8DngecHFf1cXAXqNvkSRJ0tTYn5EkSb1SVeNuw9AleQpwK/DCqrq8p/ztwEFVtXNP2RHAEe3hzsCNo2zrDLI9cMe4G6Fp4Xc7e/ndzl5+txv29KqaN+5GDIv9mWnh353Zy+929vK7nb38bjdso/2ZzUfdkhHrz06lv6yqzgLOGlmLZqgkK6tq0bjboeHzu529/G5nL7/bOcf+zJD4d2f28rudvfxuZy+/26mbla9Q0WTx1gEL+srnA6tH3xxJkqQpsz8jSZIeMCsTOFX1E+AaYElf1RKa1RskSZI6zf6MJEnqNZtfoToVWJ7kS8CVwO8BTwH+Zqytmrkclj17+d3OXn63s5ff7dxhf2a4/Lsze/ndzl5+t7OX3+0UzcpJjCclOQo4HtgBuA74g95JACVJkrrO/owkSYJZnsCRJEmSJEmaDWblHDiSJEmSJEmziQkcSZIkSZKkjjOBM8ck2SHJOUluSLIuydkDnndikiuT3Jdk4PfukhyV5OYk9ye5Jsk+j7jx2qRH8t0mWZjkI0m+lWRt+/meJFsNcK7f7Yg8wu/2MUnOS/Kd9ju6Pcknkjz1Yc57QXverUkqyWHDeg491CP9N7nn/C2T/Gf7XS0aIN6/t5oV7M/MXvZnZi/7M7OX/ZnRMYEz92wB3AGcDFw9xfP+GTht0BOSvBo4Hfhz4Lk0S55+LsmOU7ivBvdIvttnA5sBbwR2BY4BXkfzvW2U3+3IPdK/t58H/g+wM/Aq4BnAvzzMOdvQTJL6ZmDtlFuqqXqk3+2k9wHfGyTQv7eaZezPzF72Z2Yv+zOzl/2ZUakqt1m0AUcCq4HN+8rPAT7dV3YBcPYUr/+bzX82A8VeDXyor+y/gPeM+89pJm7T/d32nHsU8H2/21n53b4cKGDLAePvBQ4b95/PTN6m87sFfh34OrBL+73+v/buPFqWsr4X/vcnRFFABTl4UBceEhNj1ETJ8UaIKAZBvZgoMYkDeiV5X3DCNyYaxSmSa6KoV1GXE5hEckFyr1MMzmAMDqAoGKOoiBJQEZkEBxBE8Hn/qNrYNPtwpr2ra/f+fNaq1d01Pr0f+/DzW09Vrd/I+n63phUzqWfmd1LPzO+knpnfST0znskInPnzziR3TPKwhRlVtX26H8YJQzWiqm6d5LeTnDy16OQkew/VjjkzVN/ePsmVG1qob5fFsvdtVe2c5OAkZ7TWrl2KfbJJlqVvq+puSd6Srk83embR75YVSD0zv9Qz80s9M7/UMyMhwJkzrbUrk3wo3Y9gwUFJrk/y/gGbsku6oayXTM2/JMnaAdsxN4bo237o4XOTvPkWVtO3S2w5+7aqXllVVyf5fpLdkzxqa/bH5lmOvq2qbZK8I8lrWmtf3MTN/G5ZUdQz80s9M7/UM/NLPTMeApz5dEKSx1TV7frPByd594xS6ukbBNYi89h0y9a3VXXnJB9NckqSozdhE327tJarb1+d7vrgA5LckOSEqqqt3CebZ6n79oVJfpbktVuwrd8tK4l6Zn6pZ+aXemZ+qWdGQIAznz6QLg19dFXtmm6o22DDjXuXp/vHdToJ3TU3T0zZdMvSt1W1Nsm/p7vZ25NbfzHpBujb5bEsfdtau7y1dm5r7ZQkj0/y8CQP2tr9slmWum/3S/LQJD+rquuTfLOf/9mqescGtvG7ZSVSz8wv9cz8Us/ML/XMCGw76waw9FprP62qd6dLRXdJcnGSTwzchuuq6qwk+yd518Si/ZO8Z8i2zJPl6Nuq2i1dsfOVJE9orV2/kTbo22Uw0O92IbS/zRLvl1uwDH37p0m2n/h8l3Rnmw9OctoG2uB3y4qjnplf6pn5pZ6ZX+qZcRDgzK8TknwsyR5JTmyt/XxhQVXdr397+yQ/7z9f11r7ar/8oCSvSLJfa+27/bzdk+ycZN3UPr7ZWruqn3dOkje21t7YL3ttkuOr6nPpfoRPS/fDfOuyfOPVY8n6tqrukuTUJBcleXaSXSZGo17WWruh307fDmMp+3avJHsm+XSSHyT5lSQvS3JBP29hvzfp26raIck9+sW3SrJ7f6wrWmvfXpZvvTosWd+21s6f3HFVXdW/Pa+1duHEfL9b5oF6Zn6pZ+aXemZ+qWdmbdaPwTItz5TuOsAL0l0LeN+pZW2R6YKJ5Yf089ZNzDtuA9vtO7XfI6eO9Yy+HT9NclaSB8/6b7PSp6Xs24nPi03rpvarb1dW394v3ZnI7/d9dH66u/zfbZH9Hjnxed8NHOu4Wf99VvK01P8mT22/Los8dtPv1jQP01L/dqKeGc20xP/NO2QD26hnVn7fqmdGNC31v8lT26+LemajU/V/AAAAAABGyk2MAQAAAEZOgAMAAAAwcgIcAAAAgJET4AAAAACMnAAHAAAAYOQEOAAAAAAjJ8ABAAAAGDkBDgAAAMDICXAAAAAARk6AAwAAADByAhwAAACAkRPgAAAAAIycAAcAAABg5AQ4AAAAACMnwAEAAAAYOQEOAAAAwMgJcAAAAABGToADAACwylXVIVXVqmrdrNsCLE6AAzClqu5TVUcuVsD089sGpm0XWf+AqvpsVV1TVRdX1RuqaochvgcAsHpU1Vur6otVdWVV/aSqvlZVf1NVO866bcDSuNn/2QAg90ny0iSnJrlgA+scnuSHU/NumPxQVb+X5ENJzkjy7CS7J/nLJPeuqoe11trSNRkAWOXWJ/l4kvOTXJtkzyTPT3JAVT2otXbDLW0MjJ8ABxhUVd2utfaTWbdjCbyntXbxRtY5Osm5SR7aWrsuSarqG0nenuTRSd63vE0EAFaL1tr66XlVdW6S1ybZK8mnB28UsKRcQgUsm4nLjX6jqv6pqr6f5Cv9st2q6tiquqiqflpV36iq51VVTe3jT6rq81X1o6r6YVV9uar+emL5wvXa+1TVy6vqe/3lSidX1R6LtGl9VX2gqn7Qr/e5qnrU5P6S/HP/8d8nLo865Oa7qttX1aL/jlbVryf5zSTHLoQ3vROS/CDJ4zbxzwgArDB9nXNMVV3Y1zkXVNXbqmrHzaldqurUqjqnqn61qj5SVVdX1SVV9YoN1SBTLuxf7zC13wdW1elVdW1VfbuqjkhSN98cGBMjcIAhvDPdpUgvSXLrqto1yWeT/FKSY5N8L8k+SV6Z5C7pLjdKVT0syf9JNxz4BekuUbpnkgcvcoyjk1yT5BVJdkny3CTvSLL3wgpV9ZAkH03y5SR/m+Sn6YKUk6rqj1pr703yySRvSvLMJC9P8rV+89Onjndukh2SXFNV70/ynNbahRPL9+xfPze5UWvt+qo6a2I5ADBHqmptuv/+r0nytiRnJ9ktyUFJ7jSx6kZrl94dknwsyQfSjd59RJIj0l0qdezUsX+pX/826U4kvTzJj9LVXQvr/Ea/vx+lq4euS3JYkqu25nsDy0+AAwzh6621xy58qKpjk2yX5Ddba5f0s4+pqouSPKeqXtdauyDJgUl+nOThm3Dd9tXpLlX6eX+MK5IcXVX3bq19pR/Zc0y6IOZhE+u9KclpSV6d5L2ttf+qqk+nC3BOaa2dOnWcK5O8MV0hdG2SB/XrPrCq9mytfb9f7y796/cWaetFSX5nI98HAFiZjkpy1yQPaa19amL+kVMjjW+xdplYb22SQ1trf99/fmtVfTHJoZkKcJI8JMkpE5+/luT3J+qTJHlZklsn2ae1dl5/7Lcn+cYWfFdgQC6hAobwloU3feHyR+nOIt1QVbssTOlGx9wqXfGRdGeGtk9ywCYc45iFAqj3if71l/vX30o3eueEJDtPHHPnJB9O8stVdfeNHaS19vrW2rNaa+9orb2ntfYXSZ6U7gbFfzGx6m37158usptrJ5YDAHOiv6zpoCQfmQpvkiRTDzDYWO2y4Np098/L1LrT6yXJF5Lsn+QPk7wmyc8ycflUVW2TbgTP+xfCm75dl6Ub/QOMmBE4wBDOm3i/JslOSf6snxaza//6liR/kuRD/eicjyV5b5KTFnmC07emPl/Zv+7cv/5a//oPt9DOXRfZz0a11t5ZVa9K8rAkL+5nX9O/3maRTbabWA4AzI81SW6f7nLtjdlY7bLgwkVGIl+5yHpprV2Rrl5Kkn/p7+H3vn6U8H/27btdkq8v0p7F5gEjIsABhjAZViyM/PvnJP+4gfW/kSSttYur6rfSnUl6RD/9jyQfrqoDp0KcDV1itTBUeeG4RyQ5awPrbk3h8u10Q5wXLFw6tVu6a9Qn3SXdZVQAwHxZqDumTzQtZmO1y8bW2xTvSldvPT7Jf+aW2+cmxjByAhxgaJeluzRq29baxza2cv8Epw8m+WB/+dUrkjw/3Q3+TtuM4y6MAvrxJhx3U4quaXukC3EWLIRE/y0TN0Cuqm3T3cD4o1twDABg3C5NV+fcd9YN6d0mXTCzcBnVpUl+kuTXF1n31xaZB4yIe+AAg+qHAL87yUFVdbMnMVXVHfonKKSq7jS1bUvyH/3HnTbz0GelG9nz3Kq6w/TCqloz8fHqDR2jqm42XLmq/izJ3ZJ8ZKKt56R76sRhVXXridWf1O/3XZvZfgBg5Pp72vxLkkdW1fTTpBbuBbjkqmrn/v420w7tX8/s23dDupNIv19VvzKx/ZokT1yOtgFLxwgcYBaOSHej4tOq6h/SXSd++yT3SfLYJPdIcnGSv+9vNPxvSb6T7okOh6e7POkTi+x3g1prP++Dlo8m+WpV/WO6a893S7JXkrsnuXe/+heS/DzJC6rqjukuATujtXZ+kvOq6r1JvpruDNbeSQ5O95SH108d9i/ThTofr6r/ne5Gx89Jcmq64g4AmD8vSHf597/1T978SpI7p7ux8EHLdMw/SPKSvkY5L9399h6S5NHpwpsTJtb96yQPT/Kp/mmcP0v3GPFvJbnjMrUPWAICHGBwrbXLqup30t3w9zHpioYfJDk3yZFJruhXPSHdmaOnpRu1ckmSDyX5m9baj7fguJ+uqv+WrnB5arrhxJcm+WKSl06sd1FVHZYuaHpbkm2S/Gm6e9m8K11o88fpiqNvJzk6yctaaz+aOt4pVXVgkv+ZLtz5UbqbKL9gkZswAwBzoLX2vb7OeVm6e8/cMd29705OcvkyHfbMJJ9N96TPhXvyfTNdDfLq/pL0hfadXVX7p3tK1UvS1UJvTldnbej+hMAIlP8PAQAAADBu7oEDAAAAMHICHAAAAICRE+AAAAAAjJwABwAAAGDkPIVqwi677NLWrVs362YAAFvgrLPOury1tmbW7Zg19QwArFy3VM8IcCasW7cuZ5555qybAQBsgar61qzbMAbqGQBYuW6pnnEJFQAAAMDICXAAAAAARk6AAwAAADByAhwAAACAkRPgAAAAAIycAAcAAABg5AQ4AAAAACMnwAEAAAAYOQEOAAAAwMgJcAAAAABGToADAAAAMHICHAAAAICRE+AAAAAAjNy2s24ArHTrjvjgrJuwal1w1IGzbgLAquK/ebPjv3kAGIEDAAAAMHICHAAAAICRE+AAAAAAjJwABwAAAGDkBDgAAAAAIyfAAQAAABg5AQ4AAADAyAlwAAAAAEZOgAMAAAAwcgIcAAAAgJET4AAAAACMnAAHAAAAYOQEOAAAZZ+NagAAHvpJREFUAAAjJ8ABAAAAGDkBDgAAAMDICXAAAAAARm7bWTdga1TVC5P8XZI3tdYO7+dVkpcmOSzJTknOSPLM1tpXZtZQAAAYsXVHfHDWTVi1LjjqwFk3AVghVuwInKp6YJJDk3xpatHzkjwnybOSPCDJpUlOqaodh20hAAAAwNJYkSNwquoOSd6R5P9J8tcT8yvJs5Mc1Vp7Tz/vKelCnCcmOWb41gIAAMyG0VWzY3QVS22ljsA5Nsm7W2sfn5q/R5K1SU5emNFauybJJ5PsPVzzAAA6VfXCqmpV9caJeVVVR1bVRVV1TVWdWlX3ntpup6o6vqp+2E/HV9Udh/8GAMAYrLgAp6oOTXKPJC9ZZPHa/vWSqfmXTCyb3t9hVXVmVZ152WWXLV1DAYBVbysv+T4xyZ5JHpnkEf3745e7zQDAOK2oAKeq7pnk5UkObq1ddwurtulNF5nXrdjasa219a219WvWrFmilgIAq93UJd9XTsy/ySXfrbWzkzwlyY7pLvlOVd0rXWhzWGvt9NbaZ5I8Ncmj+noIAFhlVlSAk2SvJLskObuqrq+q65M8JMkz+vff79ebHm2za24+KgcAYDltzSXfeyW5KsnpE9udluTquCwcAFallRbgvC/JfZPcb2I6M8n/6d+fm+TiJPsvbFBV2yXZJzctgAAAls0SXPK9NsllrbUbRxD37y/NIpeFuyQcAObfinoKVWvtB0l+MDmvqq5OckU//DhV9bokL6qqc9IFOi9OdwbrxIGbCwCsQhOXfO+zlZd8L3b596KXhbfWjk034ifr169f9LJxAGBlW1EBziZ6VZLbJnlTkp2SnJHkgNbaj2faKgBgtZi85Hth3jZJHlxVT0uy8LSptUm+M7Hd5CXfFyfZtapqYRROf++cNXFZOACsSis+wGmt7Tv1uSU5sp8AAIb2vnSXeE96e5JvpBuZM3nJ9+eTm1zy/Vf9+p9JskO6MGjhMvC9kmwfl4UDwKq04gMcAIAxWYpLvltrX6uqjyQ5pr+fTiU5JskHWmtfH+zLAACjIcABABjeplzyfXCSN+QXT6s6KcnhQzYSABgPAQ4AwDLbkku+W2tXJHnScrYLAFg5VtpjxAEAAABWHQEOAAAAwMgJcAAAAABGToADAAAAMHICHAAAAICRE+AAAAAAjJwABwAAAGDkBDgAAAAAIyfAAQAAABg5AQ4AAADAyAlwAAAAAEZOgAMAAAAwcgIcAAAAgJET4AAAAACMnAAHAAAAYOQEOAAAAAAjJ8ABAAAAGDkBDgAAAMDICXAAAAAARk6AAwAAADByAhwAAACAkRPgAAAAAIycAAcAAABg5AQ4AAAAACMnwAEAAAAYuW1n3QAAGNq6Iz446yasWhccdeCsmwAAsCIZgQMAAAAwcgIcAAAAgJET4AAAAACMnAAHAAAAYOQEOAAAAAAjJ8ABAAAAGDkBDgAAAMDICXAAAAAARk6AAwAAADByAhwAAACAkRPgAAAAAIycAAcAAABg5AQ4AAAAACMnwAEAAAAYOQEOAAAAwMgJcAAAAABGToADAAAAMHICHAAAAICRE+AAAAAAjJwABwAAAGDkBDgAAAAAIyfAAQAAABg5AQ4AAADAyAlwAAAAAEZOgAMAAAAwcisuwKmqZ1bVl6rqR/30mao6cGJ5VdWRVXVRVV1TVadW1b1n2WYAAACArbHiApwkFyZ5fpI9k6xP8vEk76uq3+yXPy/Jc5I8K8kDklya5JSq2nEGbQUAAADYaisuwGmt/Wtr7cOttW+21s5trb0oyY+T7FVVleTZSY5qrb2ntXZ2kqck2THJE2fYbABglViK0cJVtVNVHV9VP+yn46vqjsN/GwBgLFZcgDOpqrapqscn2SHJ6Un2SLI2yckL67TWrknyySR7z6SRAMBqsxSjhU/st39kkkf0748fpPUAwChtO+sGbImqum+SzyTZLslVSQ5qrX25qhZCmkumNrkkyV0HbCIAsEq11v51ataLqurp6UYLfzkTo4WTpKqeki7EeWKSY6rqXulCmwe11k7v13lqkk9V1T1ba18f6rsAAOOxUkfgfD3J/ZI8MMlbkvxTVd1nYnmbWr8WmdctqDqsqs6sqjMvu+yyZWksALA6beFo4b3SnaA6fWJXpyW5OkYUA8CqtSJH4LTWrkvyzf7jmVX1gCR/keTv+nlrk3xnYpNdc/NROQv7OjbJsUmyfv36RUMeYHVad8QHZ92EVeuCow7c+EowYls5WnhtkstaazfWJa21VlWX9ssAgFVopY7AmXarJLdJcn6Si5Psv7CgqrZLsk9uehYLAGA5be1o4cVOKhlRDACr2IobgVNVRyX5YLoRNgtPl9o3yYH92anXpbvW/Jwk5yZ5cbozXyfOpsUAwGqzlaOFL06ya1XVwiic/kmba2JEMQCsWisuwElX8JzQv/4wyZeSPLK19tF++auS3DbJm5LslOSMJAe01n48g7YCACSLjxb+fHKT0cJ/1a/7mXT3zNkrvxhBvFeS7WNEMQCsWisuwGmtHbKR5S3Jkf0EADCorR0t3Fr7WlV9JN0TqQ5Nd+nUMUk+4AlUALB6rbgABwBg5JZitPDBSd6QXzyt6qQkhy9/0wGAsRLgAAAsoaUYLdxauyLJk5ayXQDAyjYvT6ECAAAAmFuDBjhV9SdVdcDE57+uqgur6qNVtduQbQEAmKZWAQDGaugROEcuvKmqPZO8MN313b+U5DUDtwUAYNqRC2/UKgDAmAx9D5y7J1l4esJBSd7XWntVVZ2c5KMb3gwAYBBqFQBglIYegXNtusdpJsl+ST7Wv//hxHwAgFlRqwAAozT0CJxPJXlNVX06yfokf9TP/7Uk3xm4LQAA09QqAMAoDT0C5/Ak16Urhp7WWruon//IGJYMAMyeWgUAGKVBR+C01i5M8vuLzH/2kO0AAFiMWgUAGKuhR+AAAAAAsJkEOAAAAAAjJ8ABAAAAGDkBDgAAAMDIDRLgVNVrqmqfqhIYAQCjo1YBAMZuqCLldkn+OcklVXVcVT2mqm470LEBADZGrQIAjNogAU5r7emttbslOTDJd5P8bZLLq+qkqvqzqlozRDsAABajVgEAxm7QYcKttc+11l7UWrtPkt9K8okkhyS5sKo+XVXPraq7DtkmAIAFahUAYKxmdp13a+2brbXXtNYenOSuSf4xyYOSPGFWbQIAWKBWAQDGZNtZNyBJWmuXpyuK/nHWbQEAmKZWAQBmzZMWAAAAAEZOgAMAAAAwcgIcAAAAgJET4AAAAACM3KABTlU9pKp+Z+LzIf0jOY+pqh2GbAsAwDS1CgAwVkOPwHldkrVJUlX3THJMki8l2SvJqwduCwDANLUKADBKQwc4v5Lky/37xyY5pbX2jCSHJvn9gdsCADBNrQIAjNLQAU5Lsk3/fr8kH+nfX5zkTgO3BQBgmloFABiloQOczyd5SVU9Ock+ST7cz1+XrjACAJgltQoAMEpDBzjPTnK/JG9M8nettfP6+X+c5PSB2wIAME2tAgCM0rZDHqy1dnaS31xk0XOT3DBkWwAApqlVAICxGnoETpKkqtZX1eOqavt+1jbprjkHAJg5tQoAMDaDjsCpqjsnOSnJA9IVQb+a5L+SvDbJtUn+fMj2AABMUqsAAGM19Aico/OLpzj8ZGL+u5IcMHBbAACmqVUAgFEadAROusdx7tdau7KqJuefl2T3gdsCADBNrQIAjNLQI3Bum+S6ReavSTcsGQBgltQqAMAoDR3gfDLJIROfW1Vtk+T5Sf5t4LYAAExTqwAAozT0JVTPS/KJqnpAktskeU2Seye5Q5LfHbgtAADT1CoAwCgNOgKntfbVJPdNcnqSk5Nsl+6mgPdvrZ03ZFsAAKapVQCAsRp6BE5aaxcneenQxwUA2BRqFQBgjJY9wKmqPTd13dbaF5azLbO07ogPzroJq9YFRx046yYAMGJqFQBgJRhiBM6ZSVqS2sh6Lck2y98cAICbUKsAAKM3RICzxwDHAADYUmoVAFYcV3nMzqyu8lj2AKe19q3lPgYAwJZSqwAAK8HgNzGuqt2SPD3Jb/SzvpbkLa21i4ZuCwDANLUKADBGgz5GvKr2T3Jekscl+Uk//XGSb1bVAUO2BQBgmloFABiroUfgvCHJ3yf589ZaW5hZVa9P8vok9xq4PQAAk9QqAMAoDToCJ8m6JG+cLIh6b0py94HbAgAwbV3UKgDACA0d4JyZ5L6LzL9vkv8YuC0AANPUKgDAKA19CdWbkxxdVb+a5LP9vAemu1HgEVW158KKrbUvDNw2AAC1CgAwSkMHOO/oX19+C8uSpCXZZvmbAwBwE2oVAGCUhg5w9hj4eAAAm0OtAgCM0qABTmvtW0MeDwBgc6hVAICxGnoETqpqbZK9k+yaqZsot9bePHR7AAAmqVUAgDEaNMCpqicl+fskleTKdNePL2jpbhy4sX28IMkfJrlnkp+mu8HgC1prZ0+sU0lemuSwJDslOSPJM1trX1mabwIAzKOlqFUAAJbD0I8R/7skr0qyfWttbWttt4npLpu4j33TFU97J/m9JNcn+VhV7TyxzvOSPCfJs5I8IMmlSU6pqh2X6HsAAPNpKWoVAIAlN3SAc/skx7XWrt/SHbTWHt5ae3tr7ezW2peTPDnJmiS/m9w4+ubZSY5qrb2nH5nzlCQ7JnniVn8DAGCebXWtUlUvqKrPV9WPquqyqnp/Vd1nap2qqiOr6qKquqaqTq2qe0+ts1NVHV9VP+yn46vqjlvaLgBgZRs6wHlHkgOXeJ87pvseV/af90iyNsnJCyu01q5J8sl0o3YAADZkKWqVfbM0o4VPTLJnkkcmeUT//vitbBsAsEINfRPjv0zyvqraL8mXk/xscmFr7X9uwT5fn+SLST7Tf17bv14ytd4lSe66BfsHAFaPra5VWmsPn/xcVU9O8sN0o4XfPz1auF/nKelCnCcmOaaq7pUutHlQa+30fp2nJvlUVd2ztfb1rfuaAMBKM3SA89R0xcjlSe6Rm98YcLMCnKp6bZIHpStubpha3KZXX2RequqwdDc7zu677745hwcA5s+S1iq9TRotXFULo4WPSbJXkquSnD6xn9OSXN2vI8ABgFVm6ADnJUme01o7emt3VFVHJ3l8koe21v5rYtHF/evaJN+ZmL9rbj4qJ621Y5McmyTr16+/WcADAKwqS1arTNiS0cJrk1zWWruxNmmttaq6dGL7GzkhBQDzb+h74GyT5KSt3UlVvT7dEOPfa62dM7X4/HQhzv4T62+XZJ/c9CwWAMC0JalVFkyMFn7sFowWXuzE0qIjiltrx7bW1rfW1q9Zs2ZrmgwAjNTQAc7bkxy8NTuoqjcl+dMkT0hyZVWt7acdku7sVJLXJTmiqv6wf+rDcemGIZ+4NccGAObeVtcqC/rRwk9Id8JpQ6OFJ02OFr44ya79/XIW9lfpnrx5sxHFAMD8G/oSqtsl+X+r6uFJvpSb3xjw/9uEfTyjf/23qfl/k+TI/v2rktw2yZuS7JTkjCQHtNZ+vGXNBgBWiaWoVRZGCz8+yb4bGS38+X79hdHCf9Wv85kkO6S7F87CCOK9kmwfI4oBYFUaOsC5V5L/6N//+tSyTbr/TGutNmGdli7MOXIz2gYAsNW1Sj9a+MlJHpN+tHC/6KrW2lX9vWxel+RFVXVOknOTvDgTo4Vba1+rqo+keyLVoekunTomyQc8gQoAVqdBA5zW2kOHPB4AwOZYolplqUYLH5zkDfnF06pOSnL4ErQPAFiBhh6BAwAw15ZqtHBr7YokT1qyhgEAK9rgAU5VPTTdDf12T3LryWWttd8buj0AAJPUKgDAGA36FKqqOiTJh5PsmGTfJJelGza8Z5KvDtkWAIBpahUAYKyGfoz4c5Mc3lp7QrqnOrygtXb/JCeku3EfAMAsqVUAgFEaOsD55SQf69//NN3jMZPkjUkOGbgtAADT1CoAwCgNHeB8P92Q5CT5bpL79O/vlO5JDAAAs6RWAQBGaeibGH8qyQFJvpzknUneUFX7J9kvySkDtwUAYJpaBQAYpaEDnMOTbNe/f0WS65P8broC6W8HbgsAwDS1CgAwSoMGOK21Kybe/zzJK4c8PgDALVGrAABjNfRjxH+jqu458Xn/qjqhql5QVdsM2RYAgGlqFQBgrIa+ifE/JLl/klTV3ZL8a5KdkzwzhiUDALOnVgEARmnoAOdeSb7Qv//jJGe01v57kicnecLAbQEAmKZWAQBGaegAZ5sk1/Xv90vyof79eUnuPHBbAACmqVUAgFEaOsA5O8nTq2qfdEXRR/r5d01y+cBtAQCYplYBAEZp6ADn+UkOTXJqkn9urX25n/8HST43cFsAAKapVQCAURr6MeKfrKo1SW7fWrtyYtExSX4yZFsAAKapVQCAsRo0wEmS1toNSa6cmnfB0O0AAFiMWgUAGKOhL6ECAAAAYDMJcAAAAABGToADAAAAMHICHAAAAICRE+AAAAAAjJwABwAAAGDkBDgAAAAAIyfAAQAAABg5AQ4AAADAyAlwAAAAAEZOgAMAAAAwcgIcAAAAgJET4AAAAACMnAAHAAAAYOQEOAAAAAAjJ8ABAAAAGDkBDgAAAMDICXAAAAAARk6AAwAAADByAhwAAACAkRPgAAAAAIycAAcAAABg5AQ4AAAAACMnwAEAAAAYOQEOAAAAwMgJcAAAAABGToADAAAAMHICHAAAAICRE+AAAAAAjJwABwAAAGDkBDgAAAAAIyfAAQAAABg5AQ4AAADAyAlwAAAAAEZOgAMAAAAwcgIcAAAAgJFbcQFOVT24qk6qqu9WVauqQ6aWV1UdWVUXVdU1VXVqVd17Rs0FAAAA2GorLsBJskOSs5P8eZJrFln+vCTPSfKsJA9IcmmSU6pqx8FaCACsaktxwqmqdqqq46vqh/10fFXdcdAvAgCMxooLcFprH2qtvbC19u4kP59cVlWV5NlJjmqtvae1dnaSpyTZMckTh28tALBKLcUJpxOT7JnkkUke0b8/fhnbDACM2IoLcDZijyRrk5y8MKO1dk2STybZe1aNAgBWl6094VRV90oX2hzWWju9tfaZJE9N8qiquueQ3wUAGId5C3DW9q+XTM2/ZGIZAMAsbcoJp72SXJXk9IntTktydZyUAoBVad4CnAVt6nMtMq9bUHVYVZ1ZVWdedtlly98yAGC125QTTmuTXNZau7F+6d9fmkVOSqlnAGD+zVuAc3H/Ol3Y7JqbF0lJktbasa219a219WvWrFnWxgEATNjYCafFTj4telJKPQMA82/eApzz04U4+y/MqKrtkuyTmw5BBgCYlU054XRxkl37++UkufHeOWuygZNSAMB8W3EBTlXtUFX3q6r7pWv/7v3n3fuhxa9LckRV/WFV3SfJcemuIT9xdq0GALjRppxw+ky6J1ntNbHdXkm2j5NSALAqbTvrBmyB9Un+feLz3/TTPyU5JMmrktw2yZuS7JTkjCQHtNZ+PGwzAYDVqqp2SHKP/uONJ5ySXNFa+3ZVvS7Ji6rqnCTnJnlxJk44tda+VlUfSXJMVR2a7tKpY5J8oLX29YG/DgAwAisuwGmtnZquiNnQ8pbkyH4CAJiFpTjhdHCSN+QXT6s6Kcnhy9pqAGC0VlyAAwAwdktxwqm1dkWSJy1x0wCAFWrF3QMHAAAAYLUR4AAAAACMnAAHAAAAYOQEOAAAAAAjJ8ABAAAAGDkBDgAAAMDICXAAAAAARk6AAwAAADByAhwAAACAkRPgAAAAAIycAAcAAABg5AQ4AAAAACMnwAEAAAAYOQEOAAAAwMgJcAAAAABGToADAAAAMHICHAAAAICRE+AAAAAAjJwABwAAAGDkBDgAAAAAIyfAAQAAABg5AQ4AAADAyAlwAAAAAEZOgAMAAAAwcgIcAAAAgJET4AAAAACMnAAHAAAAYOQEOAAAAAAjJ8ABAAAAGDkBDgAAAMDICXAAAAAARk6AAwAAADByAhwAAACAkRPgAAAAAIycAAcAAABg5AQ4AAAAACMnwAEAAAAYOQEOAAAAwMgJcAAAAABGToADAAAAMHICHAAAAICRE+AAAAAAjJwABwAAAGDkBDgAAAAAIyfAAQAAABg5AQ4AAADAyAlwAAAAAEZOgAMAAAAwcgIcAAAAgJET4AAAAACMnAAHAAAAYOQEOAAAAAAjJ8ABAAAAGLm5DnCq6hlVdX5VXVtVZ1XVPrNuEwDA5lDPAADJHAc4VfW4JK9P8vIk909yepIPV9XuM20YAMAmUs8AAAvmNsBJ8pdJjmutva219rXW2rOSfC/J02fcLgCATaWeAQCSzGmAU1W3TvLbSU6eWnRykr2HbxEAwOZRzwAAk+YywEmyS5JtklwyNf+SJGuHbw4AwGZTzwAAN9p21g1YZm3qc03Pq6rDkhzWf7yqqr4+RMNWoF2SXD7rRmyJeuWsWzB6+nZ+6dv5pW8Xd/dl3fvsqGeWjt/O/NK380vfzi99u7gN1jPzGuBcnuSG3Pzs1K6ZOovVWjs2ybEDtWvFqqozW2vrZ90Olp6+nV/6dn7p21VDPbPE/Hbml76dX/p2funbzTeXl1C11q5LclaS/acW7Z/u6Q0AAKOmngEAJs3rCJwkeW2S46vqc0lOS/K0JHdJ8taZtgoAYNOpZwCAJHMc4LTW/m9V3SnJi5PsluTsJP+9tfat2bZsxTIse37p2/mlb+eXvl0l1DNLzm9nfunb+aVv55e+3UzV2vR98QAAAAAYk7m8Bw4AAADAPBHgAAAAAIycAGeVqardqurEqjqnqm6oquM2cbsXVdVpVXV1VW3ydXdV9YyqOr+qrq2qs6pqny1uPLdoS/q2qtZV1T9U1X9V1TX96yuq6rabsK2+HcgW9u2tquqkqvp230ffq6oTququG9nuwf12362qVlWHLNX34Oa29N/kie23q6r/7Ptqo4/h9LtlXqhn5pd6Zn6pZ+aXemY4ApzV5zZJLk9yVJIzNnO79yZ53aZuUFWPS/L6JC9Pcv90jzz9cFXtvhnHZdNtSd/+epJtkjw9yb2TPCvJ/0jXbxukbwe3pb/bjyf5kyT3TPLYJL+c5F82ss0O6W6S+udJrtnslrK5trRvF/yvJBduyop+t8wZ9cz8Us/ML/XM/FLPDKW1ZpqjKclTk1ySZNup+Scm+depeR9Ictxm7v+Puv/ZbNK6ZyR529S8byR5xaz/TitxWu6+ndj2GUm+r2/nsm//IElLst0mrn9VkkNm/fdZydNy9m2SRyf5SpJ79f26fiPr+92aVsyknpnfST0zv5N6Zn4n9cx4JiNw5s87k9wxycMWZlTV9ul+GCcM1YiqunWS305y8tSik5PsPVQ75sxQfXv7JFduaKG+XRbL3rdVtXOSg5Oc0Vq7din2ySZZlr6tqrsleUu6Pt3omUW/W1Yg9cz8Us/ML/XM/FLPjIQAZ8601q5M8qF0P4IFByW5Psn7B2zKLumGsl4yNf+SJGsHbMfcGKJv+6GHz03y5ltYTd8useXs26p6ZVVdneT7SXZP8qit2R+bZzn6tqq2SfKOJK9prX1xEzfzu2VFUc/ML/XM/FLPzC/1zHgIcObTCUkeU1W36z8fnOTdM0qpp28QWIvMY9MtW99W1Z2TfDTJKUmO3oRN9O3SWq6+fXW664MPSHJDkhOqqrZyn2yepe7bFyb5WZLXbsG2fresJOqZ+aWemV/qmfmlnhkBAc58+kC6NPTRVbVruqFugw037l2e7h/X6SR019w8MWXTLUvfVtXaJP+e7mZvT279xaQboG+Xx7L0bWvt8tbaua21U5I8PsnDkzxoa/fLZlnqvt0vyUOT/Kyqrk/yzX7+Z6vqHRvYxu+WlUg9M7/UM/NLPTO/1DMjsO2sG8DSa639tKrenS4V3SXJxUk+MXAbrquqs5Lsn+RdE4v2T/KeIdsyT5ajb6tqt3TFzleSPKG1dv1G2qBvl8FAv9uF0P42S7xfbsEy9O2fJtl+4vNd0p1tPjjJaRtog98tK456Zn6pZ+aXemZ+qWfGQYAzv05I8rEkeyQ5sbX284UFVXW//u3tk/y8/3xda+2r/fKDkrwiyX6tte/283ZPsnOSdVP7+GZr7ap+3jlJ3thae2O/7LVJjq+qz6X7ET4t3Q/zrcvyjVePJevbqrpLklOTXJTk2Ul2mRiNellr7YZ+O307jKXs272S7Jnk00l+kORXkrwsyQX9vIX93qRvq2qHJPfoF98qye79sa5orX17Wb716rBkfdtaO39yx1V1Vf/2vNbahRPz/W6ZB+qZ+aWemV/qmfmlnpm1WT8Gy7Q8U7rrAC9Idy3gfaeWtUWmCyaWH9LPWzcx77gNbLfv1H6PnDrWM/p2/DTJWUkePOu/zUqflrJvJz4vNq2b2q++XVl9e790ZyK/3/fR+enu8n+3RfZ75MTnfTdwrONm/fdZydNS/5s8tf26LPLYTb9b0zxMS/3biXpmNNMS/zfvkA1so55Z+X2rnhnRtNT/Jk9tvy7qmY1O1f8BAAAAABgpNzEGAAAAGDkBDgAAAMDICXAAAAAARk6AAwAAADByAhwAAACAkRPgAAAAAIycAAcAAABg5AQ4AAAAACMnwAEAAAAYuf8fvHqpx9wYCPUAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 1152x720 with 4 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, axs = plt.subplots(nrows=2, ncols=2, figsize=(16, 10))\n",
    "axs = axs.flatten()\n",
    "\n",
    "for i, model in enumerate(models):\n",
    "    ax = axs[i]\n",
    "    model_results = results[results.model == model].sort_values('version')\n",
    "    ax.bar(model_results.version, model_results.inference_rate)\n",
    "    ax.set_title(model)\n",
    "    ax.set_ylabel('samples / s')\n",
    "\n",
    "plt.tight_layout()"
   ]
  }
 ],
 "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": "3.7.4"
  },
  "toc-autonumbering": false
 },
 "nbformat": 4,
 "nbformat_minor": 4
}