StockMarketPredictor.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Make sure that you have all these libaries available to run the code successfully\n",
    "import os\n",
    "import datetime as dt\n",
    "\n",
    "import pandas as pd\n",
    "pd.core.common.is_list_like = pd.api.types.is_list_like\n",
    "from pandas_datareader import data\n",
    "import matplotlib.pyplot as plt\n",
    "import urllib.request, json\n",
    "import numpy as np\n",
    "import tensorflow as tf # This code has been tested with TensorFlow 1.6\n",
    "from sklearn.preprocessing import MinMaxScaler"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loaded data from the Kaggle repository\n"
     ]
    }
   ],
   "source": [
    "data_source = 'kaggle' # alphavantage or kaggle\n",
    "\n",
    "if data_source == 'alphavantage':\n",
    "    # ====================== Loading Data from Alpha Vantage ==================================\n",
    "\n",
    "    api_key = '<your API key>'\n",
    "\n",
    "    # American Airlines stock market prices\n",
    "    ticker = \"HPQ\"\n",
    "\n",
    "    # JSON file with all the stock market data for AAL from the last 20 years\n",
    "    url_string = \"https://www.alphavantage.co/query?function=TIME_SERIES_DAILY&symbol=%s&outputsize=full&apikey=%s\"%(ticker,api_key)\n",
    "\n",
    "    # Save data to this file\n",
    "    file_to_save = 'stock_market_data-%s.csv'%ticker\n",
    "\n",
    "    # If you haven't already saved data,\n",
    "    # Go ahead and grab the data from the url\n",
    "    # And store date, low, high, volume, close, open values to a Pandas DataFrame\n",
    "    if not os.path.exists(file_to_save):\n",
    "        with urllib.request.urlopen(url_string) as url:\n",
    "            data = json.loads(url.read().decode())\n",
    "            # extract stock market data\n",
    "            data = data['Time Series (Daily)']\n",
    "            df = pd.DataFrame(columns=['Date','Low','High','Close','Open'])\n",
    "            for k,v in data.items():\n",
    "                date = dt.datetime.strptime(k, '%Y-%m-%d')\n",
    "                data_row = [date.date(),float(v['3. low']),float(v['2. high']),\n",
    "                            float(v['4. close']),float(v['1. open'])]\n",
    "                df.loc[-1,:] = data_row\n",
    "                df.index = df.index + 1\n",
    "        print('Data saved to : %s'%file_to_save)        \n",
    "        df.to_csv(file_to_save)\n",
    "\n",
    "    # If the data is already there, just load it from the CSV\n",
    "    else:\n",
    "        print('File already exists. Loading data from CSV')\n",
    "        df = pd.read_csv(file_to_save)\n",
    "\n",
    "else:\n",
    "\n",
    "    # ====================== Loading Data from Kaggle ==================================\n",
    "    # You will be using HP's data. Feel free to experiment with other data.\n",
    "    # But while doing so, be careful to have a large enough dataset and also pay attention to the data normalization\n",
    "    df = pd.read_csv(os.path.join('Stocks','hpq.us.txt'),delimiter=',',usecols=['Date','Open','High','Low','Close'])\n",
    "    print('Loaded data from the Kaggle repository')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "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>Date</th>\n",
       "      <th>Open</th>\n",
       "      <th>High</th>\n",
       "      <th>Low</th>\n",
       "      <th>Close</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1970-01-02</td>\n",
       "      <td>0.30627</td>\n",
       "      <td>0.30627</td>\n",
       "      <td>0.30627</td>\n",
       "      <td>0.30627</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1970-01-05</td>\n",
       "      <td>0.30627</td>\n",
       "      <td>0.31768</td>\n",
       "      <td>0.30627</td>\n",
       "      <td>0.31385</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1970-01-06</td>\n",
       "      <td>0.31385</td>\n",
       "      <td>0.31385</td>\n",
       "      <td>0.30996</td>\n",
       "      <td>0.30996</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1970-01-07</td>\n",
       "      <td>0.31385</td>\n",
       "      <td>0.31385</td>\n",
       "      <td>0.31385</td>\n",
       "      <td>0.31385</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1970-01-08</td>\n",
       "      <td>0.31385</td>\n",
       "      <td>0.31768</td>\n",
       "      <td>0.31385</td>\n",
       "      <td>0.31385</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         Date     Open     High      Low    Close\n",
       "0  1970-01-02  0.30627  0.30627  0.30627  0.30627\n",
       "1  1970-01-05  0.30627  0.31768  0.30627  0.31385\n",
       "2  1970-01-06  0.31385  0.31385  0.30996  0.30996\n",
       "3  1970-01-07  0.31385  0.31385  0.31385  0.31385\n",
       "4  1970-01-08  0.31385  0.31768  0.31385  0.31385"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Sort DataFrame by date\n",
    "df = df.sort_values('Date')\n",
    "\n",
    "# Double check the result\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABCgAAAJICAYAAACjTNWcAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMi4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvhp/UCwAAIABJREFUeJzs3Xe8XHWd//H3d8q9Nze9kgqB0AIICUQRFSmiKHFFV5Rld+2usqsurv52Fzu4u4oVCy6WtWBZREXXQpEOUoQkgJQkQBJuAiG93dw+5fv7Y+bMPXOmz5yZM+X1fDzyYM6ZM2e+uTlzme/nfL6fj7HWCgAAAAAAIEihoAcAAAAAAABAgAIAAAAAAASOAAUAAAAAAAgcAQoAAAAAABA4AhQAAAAAACBwBCgAAAAAAEDgCFAAAAAAAIDAEaAAAAAAAACBI0ABAAAAAAACFwl6AH6YNWuWXbx4cdDDAAAAAAAALmvWrNltrZ1dzrFtEaBYvHixVq9eHfQwAAAAAACAizFmc7nHssQDAAAAAAAELrAAhTGmxxjzkDHmL8aYJ40xl6f3H26MedAY84wx5jpjTFdQYwQAAAAAAI0RZAbFqKSzrbUnSVom6bXGmJdK+oKkK621R0naJ+k9AY4RAAAAAAA0QGABCpsykN6Mpv9YSWdL+lV6/zWS3hjA8AAAAAAAQAMFWoPCGBM2xjwqaaekWyVtlLTfWhtPH/K8pAUFXvs+Y8xqY8zqXbt2NWbAAAAAAACgLgINUFhrE9baZZIWSnqJpKX5Divw2u9aa1dYa1fMnl1WxxIAAAAAANCkmqKLh7V2v6S7JL1U0jRjjNP+dKGkF4IaFwAAAAAAaIwgu3jMNsZMSz+eIOkcSesk3SnpgvRh75D022BGCAAAAAAAGiVS+pC6mSfpGmNMWKlAyS+stX8wxqyV9HNjzH9KekTS9wMcIwAAAAAAaIDAAhTW2sckLc+zf5NS9SgAAAAAAECHaIoaFAAAAAAAoLMRoAAAAAAAAIEjQAEAAAAAAAJHgAIAAAAAAASOAAUAAAAAAAgcAQoAAAAAABA4AhQAAAAAACBwBCgAAAAAAEDgCFAAAAAAAIDAEaAAAABAQ/WPxDQSSwQ9DABAkyFAAQAAgIY68bJb9Kqv3B30MAAATSYS9AAAAADQGQ6OxPR/j2yVJG3dPxzwaAAAzYYABQAAABri0799Ur9JBygAAPBiiQcAAAAaYs/gWNBDAAA0MQIUAAAAAAAgcAQoAAAAAABA4AhQAAAAoCGstUEPAQDQxAhQAAAAAACAwBGgAAAAAAAAgSNAAQAAAAAAAkeAAgAAAAAABI4ABQAAAAAACBwBCgAAADQETTwAAMUQoAAAAEBDxBLJoIcAAGhiBCgAAADQEA8+uzfoIQAAmhgBCgAAAAAAEDgCFAAAAAAAIHAEKAAAAAAAQOAIUAAAAAAAgMARoAAAAAAAAIEjQAEAAAAAAAJHgAIAAAAAAASOAAUAAAAAAAgcAQoAAAAAABA4AhQAAAAAACBwBCgAAAAAAEDgCFAAAAAAAIDAEaAAAAAAAACBI0ABAAAAAAACR4ACAAAADTFrUnfWdjJpAxoJAKAZEaAAAABAg2QHJD547cMBjQMA0IwIUAAAACAQNz6+PeghAACaCAEKAAAAAAAQOAIUAAAAaAhLyQkAQBEEKAAAABCY137tHv38oS1BDwMA0AQIUAAAAKAh8iVQrN9+UJf++vGGjwUA0HwIUAAAAAAAgMARoAAAAEBDWIpQAACKIEABAAAAAAACR4ACAAAAAAAEjgAFAAAAAAAIHAEKAAAANAQVKAAAxRCgAAAAAAAAgSNAAQAAAAAAAkeAAgAAAA1Bl1EAQDEEKAAAAAAAQOAIUAAAAKAhLCkUAIAiCFAAAAAAAIDAEaAAAABAQ/SPxIMeAgCgiRGgAAAAQODe/5PV+sgvHg16GACAABGgAAAAQOD++OQO/frhrUEPAwAQIAIUAAAAAAAgcAQoAAAAAABA4AhQAAAAAACAwBGgAAAAAAAAgSNAAQAAAAAAAkeAAgAAAAAABI4ABQAAAAAACFxgAQpjzCJjzJ3GmHXGmCeNMZek919mjNlqjHk0/ee8oMYIAAAAAAAaIxLge8clfdRa+7AxZrKkNcaYW9PPXWmt/XKAYwMAAAAAAA0UWIDCWrtN0rb044PGmHWSFgQ1HgAAAAAAEJymqEFhjFksabmkB9O7PmiMecwY8wNjzPTABgYAAIC6MyboEQAAmkHgAQpjzCRJ10v6sLW2X9LVkpZIWqZUhsVXCrzufcaY1caY1bt27WrYeAEAAFA/8UQy6CEAAAISaIDCGBNVKjjxM2vtryXJWrvDWpuw1iYlfU/SS/K91lr7XWvtCmvtitmzZzdu0AAAAKibhLVBDwEAEJAgu3gYSd+XtM5a+1XX/nmuw94k6YlGjw0AAACN417hQXwCADpXkF08Xi7pbZIeN8Y8mt73cUkXGWOWSbKS+iS9P5jhAQAAwC/JZOHIg/spAhQA0LmC7OJxr7ID5o4bGz0WAAAA1Nc//ezhso5LEqEAgI4VeJFMAAAAtL+bn9xe1nGEJwCgcxGgAAAAQENd9lfHFXyODAoA6FwEKAAAANBQ73z54frBO1fkfc7SZRQAOhYBCgAAADRcskAgwrLIAwA6FgEKAAAANNysyd159xdp9gEAaHMEKAAAANAw03qjkqRli6aptyuc87ylBgUAdCwCFAAAAGiYsBnvMn/s3Mk5z5NBAQCdiwAFAAAAGsYdfzCuYEXmeTIoAKBjEaAAAABA3U3uiUhKLe1w5IYnRIlMAOhgBCgAAABQd+ceP1eS9M2Llhc9LkkGBQB0LAIUAAAAqLv/e2SrIiGjid2RosdRgwIAOlfx/0MAAAAANRqLJxXPE3nIU4KCGhQA0MHIoAAAAEBdVbJsg/gEAHQuAhQAAAAIhMlTJjNprX5037N6Yf9wACMCAASJAAUAAADqqmAGRZ4lHtsPjOiy36/Vu3+0qr6DAgA0HQIUAAAAqKtChS/ztRkdSyQlSQdH4vUbEACgKRGgAAAAQF1VUvgykY5mDI4RoACATkOAAgAAAHVVSetQZznI/qFYnUYDAGhWBCgAAABQV4UyKPK1GU2v8AAAdCACFAAAAKgrJz5xymHTs/bn6+KRqCTdAgDQVghQAAAAoK6cZRvnL5tf8th4khQKAOhUBCgAAABQV05ShDdfIt8SjxhrPACgYxGgAAAAQF1ZpSIUJl9EwiOWYIkHAHQqAhQAAACoK6cGRaisAAUZFADQqQhQAAAAoK6cGhTe+ES+eIXTXjQcKh3MAAC0FwIUAAAAqKvxDIrs/fm6eHzpj09JopsHAHQiAhQAAACoq/EMiuyARBkrPgAAHYQABQAAAOrKFujiAQCAGwEKAAAA1FUlRTIBAJ2LAAUAAADqylniEfJ888xXZ+LMY2ZLkl5y+Iy6jwsA0FwIUAAAAKCuMjUoPIs87t+4J+fY0ViqzWiELh4A0HEIUAAAAKCunDyJclZ4xBLJuo4FANC8CFAAAACgrmyBLh75xNPLPixdRgGg4xCgAAAAQF0lM0UySx8bT6YyKKyIUABApyFAAQAAgKps2Dmgx57fX/K48Taj2RGKnmjuV9F4ggwKAOhUBCgAAABQlXO+erfecNV9JY9zimSGPd88r7ro5MxjpyhmgiUeANCxCFAAAACgrpIFalCcc9wh+vTrj5MkLZ03RZKrBgVLPACg4xCgAAAAQF2MxZNafOkNWvmNeyVJoTxFMp1doXQGRaYGBfEJAJAk7Rsc0+JLb9Dv//JC0EOpOwIUAAAAqIsd/SNZ28VqZDoFNBNODYo6jQkAWs2m3YOSpO/f+2zAI6k/AhQAAACoi5CnbUcozzfPs46ZI0l688kLJUmx9BKPJCkUACBpPIBrO+D3YiToAQAAAKA9eduKemtQSNLiWRPVd8VKPbH1gCRp18FRSSzxAABH2Cki3AG/GMmgAAAAQF1424rmq0GROdbzVPt/DQeA8ji/S9MletoaAQoAAADUhXeZhjejIvs5b4SCEAUASNJ9G3dLyq3r044IUAAAAKAucgMUZFAAQKW+9MenJEl7BscCHkn9EaAAAABAXXjTkYvEJ3KCFxTJBICURLJzfh8SoAAAAEBNRmIJxRPJnC/RlWRQeJd/EJ8AgGxL500Jegh1RxcPAAAA1OTYT90sY6QjZ0/SrR85I7PfW3G+WIBCnoKaBCgAIOWEBVP0xNZ+TewKBz2UuiODAgAAADWzVnpm50DWvmSykiKZnvP5NTAAaHFnHj1HkrR6876AR1J/BCgAAABQF95l06boEg9vBgUhCgCQspfLtfvvRgIUAAAAqAtvTYpiKzzoMgoA+bl/lbZ7FgUBCgAAANRFZUUyPRkULPIAAEnZWRP7h2IBjqT+CFAAAACgLnIDFOW/lgwKAEhx/zociSUCG0cjEKAAAABAXfQPx7O2i2ZQeKIX3uAGAHQqd8HhvYNjAY6k/mgzCgAAgLr4+aotWdvFalDQxQMA8ktaqbcrrLecslBHHzI56OHUFQEKAAAA+O6JrQf0h8e2Ze0rlkFhRIQCAPJJWqtIyOjy808Ieih1xxIPAAAA+G4skczZV7xIZvY28QkASLHWFm3T3E4IUAAAAMB34TxfposVyfR++aYGBQCkWFVWZLiVEaAAAACA7/IFGIrdAfQ+RXwCAFKS1hbNQGsnBCgAAADgu3zxhWJ3AL1fvi2LPNBmrLWK5Vn6BJSStMUDvO2EAAUAAAB8ZyvMoHAHL85fNp8MCrSdy3+/Vkd94ibdv3F30ENBi7HWssQDAAAA8FOxuhLuLh6RUIgABdrOj+7vkyT97fceDHYgaDnJZPE2ze2ENqMAAACoWL4MiVKSySIBCtdtM2OqOz8AtKPrVj8X9BAahgwKAAAAVKxIrEGSlG+pfbHXGM9jwhMA0HkIUAAAAKBi8WTxYn/5lnMkikQonCKZxjgZFLWND2gmQ2PxoIcAtAQCFAAAAKhYsWCDlH85R7EaFE6AImyMQsYoQYQCbeTrtz8T9BCAlkCAAgAAABWLlwpQ5Hn6sJm9BY93CsCFjNHU3qj2D43VMjygqVy3qnNqCAC1IEABAACAiiUSxQMU3gyIvitWanJPtODxzuHGSN3hUMkACDrXum392rBzIOhhVGTe1AlBDwEtaixefDlduyFAAQAAgIoVWoLhdN8o1rEjn65I6mvp/3vNMYqEU21GKz0HOsPrvv4nnfPVu4MeRkWWHzot6CGgRR0YjkmSPnv+8QGPpDECC1AYYxYZY+40xqwzxjxpjLkkvX+GMeZWY8wz6f9OD2qMAAAAyM8JHkzvzc6KcOIWxepN5BMOGfVdsVL/8MojFA6l1nt8+56NtQ8UCMB1q7Zo3bb+zPaOAyNaOm+KLjhloeZP7QlwZGg1ToBi6oTCGWjtJMgMirikj1prl0p6qaQPGGOOk3SppNuttUdJuj29DQAAgCbiZFC8/4wlWfudwESpIprFRNIBii/e/FTV50B7un/D7qztZuyOMTyW0L9f/7he9/U/qW/3oCRpe/+I5k3t0YRoWEOxRMAjRCtxAhRTCFDUl7V2m7X24fTjg5LWSVog6XxJ16QPu0bSG4MZIQAAAApx4g9d4VDe/ZVmULjdtm5H1a9F+xqJJfS3//NgZvut335Ax336j7ptbXNdL3tdBV7f9oPUeLcfGNEhU3o0uSeigyPNF1RB8+ong6LxjDGLJS2X9KCkQ6y126RUEEPSnOBGBgAAgHycJR7RsMnab5WuQVFD+YhVffuqfzHa1i/XPJ+1/VDfXknSvZ6siqCNuDIknts7rNF4QnsGxzR3So8i4ZASSZup1QKUsvPgiCQCFA1jjJkk6XpJH7bW9pc63vW69xljVhtjVu/atat+AwQAAECORCZAkf110pl31bLEA8in0KTeGyQL2ohnCcdze4clSbMmd2WWL/HxQLn+/frHJUnTCFDUnzEmqlRw4mfW2l+nd+8wxsxLPz9P0s58r7XWftdau8Jau2L27NmNGTAAAAAkjdegiBQIUNSyxAPIxymemrs/8HuuWUZdbSFfceSsTMeRrnAo83eIJzurdSRqRw2KOjPGGEnfl7TOWvtV11O/k/SO9ON3SPpto8cGAACA4got8XACE7UEKC5csaj6gaFthU2hAEWDB1KCO4Nick8k83jznqFMgIL4BMp1/rL5WjRjQk62WrsK8m/5cklvk3S2MebR9J/zJF0h6dXGmGckvTq9DQAAgCbiZFB4vzRv2pXqWpBwTcBu/OfTKzr3G5bNr21waEuFMigizZZBERu/+MfiSf318gWSpH88c0kmyEIGBco1NJbQxK5I6QPbRGB/U2vtvZIKLRh7VSPHAgAAgMo4NSZCnrvaW/cP60ULp2ZlUBw3f0pF5y5woxwdznutOQoFLoLiZFB0RUJKWKu7n07Vy5vYHSGDAhUbGourtysc9DAaprnCjQAAAGgJzgTLOzmcPbk7/Xz1SzwKpfKjsyUKLBtqtgDFN+7YIEma2BVWImm1Z3C87Sg1KFCp+zbsUSzROTV9CFAAAACgYmPpNRzeZdEToqk7fbV0KQg12YQTzSGWyD+pryUYVg/rtqUaE/Z2RRT3TCydAEWhYAvgFk9f849vPRDwSBqncxazAAAAwDdfvHm9JOngSDxrv7O0o5YJGOEJ5PNEgUnaWIHARaNZa/XZP6zNbE/qjihhreZP7dHLjpwlyRWgaLKgCprTSLw5ru1GIoMCAAAAFXvw2b2SpN0DY1n74+mJVy13tZm6wctaq2sfei7vc/EmmezvPDiqH97Xl9nuiYYUSyT1woGRTDcPAhSoxMGRmCTpklcdFfBIGocABQAAAKoWDRt9/W+W6dOvP07S+MTLyaQ4ZEp3xeck+x1em3YPZh6/+eSFmcfdkVDTLPHwlk6Z2tulR7bslyQ9lA7oOfVVCFCglOf2Dum0z98hSVq/vT/g0TQOAQoAAABUzUg6f9kCHX3IZEnjEy/nv3/88CsrPqclQgGP/UPjmTqj8UTmcSRkmmay7y5keOLCqYq4aqk4j52OqFziKObO9Tt1+hfvzGxTJBMAAACogDPx8mZQRLxVNMvQJPNNNJFfrdmaebx1/3DmcShkmmaJR8xVL6AnEs7qLnLFm0+UNN4qNUmEAkW860ersrZPPXxGQCNpPAIUAAAAqJozzYqkIxTOxMuZM1bTkIPJG9wOjsR07UNbMttRV9ArHDJNc724W4d2R0O6de2OzPbMSV2SJJMJUDR2bGhtb1q+IOghNAwBCgAAAFRtUrdT/C+1Hfcs8Qh5F+aXoVlS9tEc3OntxqTqnjjCpnmWeDz07L7M4550u12HE8BzAnYsY4Lb//xpk8756t0Fn49WkYnWqjrnbwoAAADf/N2ph0qS3rgsdWcvk7qeniw6E7BwFSkUzXJHHM0h5mojetMlp2c910wZFFfe9nTmcU80nPmMSOOfAyMyKJDrP29Ypw07BwoGrrqjnTNt75y/KQAAAHwTDhlN640qlJ54OXeIxzMoUsdVk0HRLBNONIcxV22H7khY923Yk9kOh4ziTVJA8KIXL8o8/v1fXtCUCdHMtpP18dy+IUnS9+/d1NjBoSXEkzZvkKK3KxLAaIJBgAIAAAAVSyRtpmWilFskM2GdJR6Vn/v4+VNrHh/aw47+Ed3zzK7Mdncke/oSMiZzrQXNO4qo6+J3Mii27ksV+PzF6ucbNSy0kETS5nTs+Pn7XhrQaIJBgAIAAAAVS1qbyZ6QcotkWmsVMuNFAStxyJQevXXFQs2b2uPPYNGyVn7jXn3iN09ktrs8AYpI2GSWFW3ZM5RpQfqnZ3Zp8aU36Lm9Qw0b66gr0yM1tvGxOp+PSLiKiB06xt7Bsaxiq5I0sYOyJyQCFAAAAKiCN4MiX5HMapZ3OEKmeWoLIDi7B0aztr0BirBJtRkdHkvolV+6U//2q8ckSdevSWUo/O4vLzRmoMpeiiJlFzZ0YnmdVOwQlXvZFXfkZFB0WlCLTwgAAAAqlkhmF8B0ghF70hPKpFVWhkWljDEaiycz5wMkqSsc0oyJXZntULpI5nAslTlx99Op5SDhdMbC8FiiIeOy1mrDzgFJqXoTN11yela3ESeTqJap5tM7DuprrkKcaE/uorBSdteaTkCAAgAAABVLLfEY33ZS2C///drx52v4Xh0y0r6hmE75z9sU93xhR2dY1bc3Z19XOJSVWRMJmfS6/dQ14lxyc6Z0S5ImdIW9p6iLD177iO7dsFuS9Mx/nael86YokucDMGtSd9Xv8Zor79HXbntGdz21s+pzoPnc8Ni2rG1v0ddOy7rprL8tAAAAfFGoSKbjh/c9q5FY9YEF9+qQOD0ZO9I7f/BQzr5QOiCR2TbZAQrHxnQ2Qz6j8YSuuGm9Bkfjvo3VO8mUpA27Co+hFj+6v68u50UwnnjhQNa291qOEKAAAAAAikt4imR6601411FXyn0+SlF0pkLdOZyimA987GyF0wELJ2ixbyimWCKpW9buKHjeX6x6Tt++e6O+eccG/wftsn8olrPP/TH54X3Pqm/3YMXnPfqQybUMC00m6QnAegOyLPEAAAAASkh6Mij85g5QUCyzMxUqsnryYdMlSdN7u1IZFTZ7UrfBlT0RzrPMwum2MRKrb32KfO/tdvnv1+rC7z5Q8XlH6zxuNNZ37tmUte29LqPe9LQ211k9SwAAAOCLpLVZEzC/QwjuoEShO+lob94A2E2XnC5JuvrvT9GzuwbVEw0rEkq1GXUv+3jd1/+UeXzdqucUiyf1oVcd1ZhBu5QTwDswnJtlUcoYNVnaypFzJmUF1YbGspceNaqOSrMgQAEAAICKJZKF73BL0qIZE7TisBlVn//HD2zOPLbMxzqS9/KaOSnVvWNSd0QvWjhVktNmNJkVoHB7dvegvnLr01kBCqejxlPbD9Zh1OPyZVAYH7KOJkSZwrWTZYumaWg0rhcOjEiS+kdSAYqvXbhMC6dPUE+0swIUnZUvAgAAAF94Myhmp7sTzJ6c+m88YfN2MagGGRSdydumtjucO1ELhVItbQsFKBzudf7OWR/YtKfmMUrSPenWpl6llnhI0oQKJp+96Tvpi2f1lv0aNL8te4ayghC7+lOtlRfNmKAVi6sP8rYqAhQAAACo2B3rd+rgyHh6elckpNmTu3XO0jmSUkUya6k+/4GzlmQel5p8oj15i0x2RXKvp5AxstaW7PTy3T+Nr/P3u3TK213dRtznPm3JzJKv3TcU0wVX35+T1u+VSNpMbQJvUUW0tof69mqTq1jqnsExSdLMidW3pG1lBCgAAABQtie2HtC37kx1P+jbM5T1XNiMt4CMJ5M1VZ+fNqEr89j6XuECrahQgCLVxaP4OqCePK/1gzdYcMqh0zOPV75oXs7x+T4Rqzfv03Gf/mPR97nwOw/IeSs+De1tYDQVmOuOduZUnQVMAAAAKNvrv3lvwefCIZOZROVrsVgJ951oVnhAyr9kIpS+5uIl2tqaOrWtjXkCI2ceMzvzOH8NiureZ/XmfZnHfB7azztftlhDY3H9YvXzGkjXoOiOdFbtCUdnhmUAAADgO2NSd5R39qeKvd3w2LaqzxWq04QS7SVkJGttyWVATj2LPzz2gj77h7W+vb83MPJPZx6ZeewERaZOiGb2+VGXhY9D+3Cu2+m9XTr72NTyuLvSNU2665T10+w6828NAAAA34VDRglXPYD3nn5E1edyz+PWbeuvdWhoUyGTuuZKFVLtTtdD+eD/PuLr+3sDFN7Cnlf97XL94UOvyGy/ZcWimt/TErFrG06b2ck9EUXT1+jm9NI5AhQAAABADZwaFM70acbEaNHji3Gn5L/rR6tqHBnaVcgYJZMqWSTTuxTDL+7zvuLIWTnPv/7E+Vo0Y7zrRk80rFMOm55zXCWIT7SPnQdT2WaHTOnJBCgctRQZbmWd+bcGAACA70IhkzV5MnlLApZ5Lp87LaA9Ja3V2m39JWtQxOKpQMLimePBghMWTKn5/d3v++5XLC7rNYOjxTt2lELR2PbhXD+RsMkJUHQqfgoAAADwRcik1lT7kYJu/O4FibZ0x/qdkqSbHi9e7ySWngi6r6uwD9eYcwdckg6fNams16zffrCm9ySDon04NSgiIaOuCL/zJAIUAAAA8IlTDyAzgarh+3aIAAUqsPPgaNHnxxKpDIqsq8qHa+wNV92XeexHAcxyEJ9oH07tlHDIaM/AWMCjaQ60GQUAAIAvwiGjpKsWQC3TNZZ4dKbhsYSuuvMZxZNWISOdcfRs3fnULh03r/hyjFiieI2JzFKMrPa1/k71G5WinySFom04GRThkNFovD51UloNAQoAAAD4wuni4cyfalmmQQZFZzr3a/doy96hzPZYicCDo1Sb0b2Do7p17Q5t2jVY9msqFQnXds0mkzanC0g+xCfax8U/WSMp9btzYnc44NE0B5Z4AAAAoCofOGtJ1rYxRsmsIpnVIz7RmdzBCUm6b8MeSaWXNTip8pN78t9/veaBzfr8jeuy9g3HEtUNsoCeaHkTzPNeNDfv/kKZEZv3DObdj9a3ZzC1rCNsjE4/anbAo2kOZFAAAACgKssXZbdLDJvUXWA/ugxQJBOVcJZw3PIvr9T03i6NxpI66bO3ZB0ze3K3Nu0en+wPjfoXoDj72Dma2FVegGLqhPztdxPW5p2cPfTs3qxtv5emIBjuf9dI2DSshkmzI4MCAAAAVQl7UtrDIZPu4pHariXG8Pjz+2sYGVrVmcfkv4tcalIed63l74mG8y63WLZoWta2H7UcTlgwRWcfO0c/eOeLyw6qed/WeVmh4SyeNVGSNGNilyTJ55UpCMhbv/NA5nHIGIKyaQQoAAAAUBVvm8ZMF4/0di3ft//y/IHqX4yWZK3VXU/tytq34rDpBY7OFk/XqoiEUtObfNee02rUUetEf2A0rie29itc4Z1vJxDhLEfpjqTGXKgmxv6hmCTpmxctz3o9Wtc19/dlbRN0GkeAAgAAAFXxpiSHQ0bW2szdblNDFYpKJ31ofS8cGMnZ9+W3nCRJOm5+8S4emW4I6ciE+9qbObFL86b26Af3PZv1mlqXSnzp5vWSpFv8Qi+CAAAgAElEQVTX7qjodW9cvkCS9IojZ0mSuiOppSGJAuO57HdPSpIeTC8J8GMJFYL1mfS/qWOMDh4ZBCgAAABQFW/HgZBJL/FIb9eSQUF8ovN89ZanM49PO2KmpNTyhl9dfJo+96YXFX2tU1vCWXbkvvZeesTMvEs+CgUEynXNA5uret1pS2aq74qVOjy9dMMJxtkCc1Rnacfsyd2p44hPtJ3DZvYGPYSmUVWRTGPMkZIOkfSEtZb8OwAAgA4U87SADIWMEj5NnliP3Xm2HRjOPL7m3S/RaDxVxHLF4hlln8PJ6nEyKiZEw7rh8W15j002SV6901K3UMDkrSsW6vGtB3Tu8YfoU//3BPkTbeYDZy3R/GkTJEkPfvxVGo0lNalAN5pOUFEGhTHm9caYjZKeknSPpFPS++cYYzYYYy6owxgBAADQhGZO7M7aznTx8GEGRXii8zhLHz57/vHqioQ0uSd/t4tinGyECdGwTj9qlr79tlPyHrdoxgTfMhGqzfZxYnDh9IzMW7TTWYLiFADtCodkDF08Stmw86C251ku1Kye3jGQeXzIlB4dOrM3kzXTicoOUBhjzpT0G0l7JV0u1/83rLU7JW2U9Dc+jw8AAABNxEkzl6SJ3bltFft2D0rpe7y1ZEGcXGZxRLQPZ+J9ztJDqj6HU4MiFDL6yXtO1RlHz9bKE+flHHfucXNr7uIxuTt1l3vRjOrS8523d8bszuj477s26PCP3aiRWCKTqRQJh2TEEo9SzvnqPXrp52/Xs66Wss1s486B0gd1kEoyKD4t6S+STpX0rTzPPyDpZD8GBQAAgNZz51O7dHA0rmd3D0mqLQviXS9b7MuY0DrG0uuDouHqy+R566JI0o48d9NDIVNT54Qd/SM6OBqv/gSesUjZSzx+fH+qvsWewbFM95FoONWKkiKZ5fnenzYFPYSyfO8dK4IeQlOp5NO/QtLPrC1UvkXPS5pb+5AAAADQrNzp5YXu5P74gT5JNRbJpEpmx4mlOxl01RCgyGdbngCFMbUVydy6f7xexlfSnUaq5SxLcQdMnOyOf/n5o5kMimgopJAhg6KY7N9PzfuDOnbuZElS3xUrtWT2pIBH01wq+fSHJY0WeX6WpLHahgMAAIBmVs53/ofS7RBrajNKkcyOM+ZMxCP+/tt3R7OnPB84a4lCxtQ0gR2Njd+zraSIZz6hPEs8nLoTD/XtVSyRVDhkFAoZGdWW+dHu4q4fTrKJO3d2R8M64+jZQQ+jKVUSoFgn6fQiz79eqSUgAAAAaFPudfuF5kl+zJ/CedpCor0NjcZljNQTya1tUsgHzlpS8hj3+WZM7NK/nnuswqa2if5Ywr/Zr5Ms5P5sxV3njyesounPw1giqY27qFlQiLuz0HWrnwtwJMVZa2mlXEAlAYrvS7rAGPMe1+usMabXGPMNSadJ+q7fAwQAAEDzcE/qCt2BdvbXkgRBBkXnGRhNaGJXpKLlPf/8qqNKHtPjyqBwrs2Qye2aUYnhMX/qT6TGkt0aVcrOBBhLJBUNjf8dbl27w7f3bjexeGuklyStzfy7I1vZAQpr7dWSrpP0PUnPKBUcv1bSAUkflPQja+3P6jFIAAAANAdbTgZF+olavn6Hub3Ytlb37dXiS2/Qdau2ZO0fGI1pUrozRrkiodLTmYmuc86alOpCY4yRtdXXKbj4pw9X9To3553Ha1DkD1DEE1bRiL91OdqVN7Ml2aTrYZLJ2roctbOKrnRr7d9LerOk2yWtV6rl6I2S3mKtfY//wwMAAECzODAUU/9I6s7xO047TIfPnJj3uIQPGRRoXzc8vk2S9O/XP64DQ7HM/sHRRN7WtcWUE8f62OuWZh5/8YIT06/LLUxZje/70IHBWyRz58ERjcXHJ9qxRFIRz1800aQT76B5AxTxJv05JVniUVDFoThr7W+stW+21h5vrT3OWnu+tfb6egwOAAAAzWPXwHg3hMvPP6FgKv74jeDqv4H73ckBzcOd2v7lW57KPD44GteknmhF5yrnLvTsyd2Zx8sPnS5Jci6vWpZ5SNKpR8ys6fVS7hKPUz93e9bzj289kNN69R9+vFojsUTN791uYvHsAEWzBnKsFUs8Cij7N78xJmKMmVLk+SnGmMpysgAAANAyykmnl6Q3LV9Q83uFQkZ9V6zUh84+kkyMNuP+5/zJnzdreCw10R4cjWtShRkU5YjmKbhqTO6yimpMiFY/3oldqdc6WSPOWLxDevKFfnV5lnjcsX6n7nl6V9Xv3a7Wbz+YtR1r0lYeSWv5vVZAJQGFr0h6naSjCzy/StIfJH201kEBAACg+UTK7KzhFCX04wu4UXmtTdE6vNeFMzEfGIlr1qRe398vkicbx7l7Xe21NaUnogld4Zpqpbz39CNkjNERsybqz5v2KpksfMc/39vUGlxpRxf/dE3WdiLRnD8jimQWVknu3LmSii3luF6pAAYAAADakDMZ++z5xxc9zrlp6cvXb77Etx3vsgynTsDAaDyroKVfvPUbpPEJf7VLACb3RPXyI2fVMiz1RMP6wFlHqjsd0EtYq/f/ZHXeYzfuGpQkvXjx9Mw+JrjZHt6yL2dfs2ZQWPGrrZBKAhSLJG0s8vym9DEAAABoQ85crrtER4HxIpl8A0eu3QdHs7aTrgDF5DoEKLz1G6T8nTMqEfO0/qyFuwbFbet2Fj32XS8/PPOYzh7ZvnXHhpx91KBoPZVc1WOS5hV5fq6k5gxRAQAAoGa2zMCDMynw4+u3c45q20Gi+fz6ka1Z2wlrZa3VYJ0yKPItwzA1dvGIJ62iEX8mmM743Ne4t+aEw1k+JUndFJLNMjSWWzT0a7c+05S/O+jiUVglV/Ujkt5qjOnyPpHed6Gkx/waGAAAAJqL8z2/0Pfqo+ZMkuRvaz8nFtKEcwxU6cxjZmdtJ5NWo/Gk4kmrST2NqbkfylxX1WdQlFs0tvRYsrt4SNKyRdOyjnG2eyLjRTkLddHpVEcdMiln33Wrn9NvH30hgNEURw2Kwir5VH1L0vGSbjDGrDDGdKX/rFCqOOZxkq6qxyABAAAQPGcuV+iL9bXve6kkKZ5IJdX6UyQzfXe59lOhSXjT7hPWamA0LkmaVIcMinxCNWZQxBLJvN1BahlLwhUscbdGlZT5+biPIWiXzQmQen34ukcbPJLSkkmWwBVSdoDCWnu9pM9LepWkByUNSRpMPz5H0hettdfVY5AAAAAIXjKzxCP/81MnRCVJsUTx4yrBd/j286dndqsrHNIbl82XlApYfPb3ayU1MkCR+u9geuJfif6RmEZiyZwgQrUy9TBci+VnT8o+95lHp7JOJvdEM/t+95cX1Ld7UCOx3KUNnWisSTt25GNZ4lFQRXlJ1tpPSDpVqUyJP0q6VdI3JJ1qrf2Y/8MDAABAs3C+/hfKoAhnUtXTGRT+9PFIvTe3i9vCcLpOwFgiqVcclZp0J5JWv/tLKg2/mgDFHR89o+LXOMsjLvrenyt+7b7BMUnSjIl+BShS/40lxiMUE7vDWcdc+rpjJUknLZya2XftQ1t05pfv0j/97GFfxtHq3D8/txMWTGnwSIo7MBTTCwdGNExgKa+KfwNYa1dJWlWHsQAAAKCJlcqgCIWMjHHVoPBliUcK4Yn2MDiWylh418sXZ9p/uosbVlODYt7UCSWPWf8fr826bp0g2/P7hit+P2e8E7vCJY4sjzOWd/1ofIo1sTuiRz71ag3HEtozMKZIOoqRb1nAHeuLd/5oV9fc36eTFk3L1OeIxccDFBefsUTfvjvVgHL5oul5Xx+U79yTGtcfHtumq/424ME0ocbkUAEAAKDllZPEEAkZxRM+dvGgSGZbcTIols6bksliGHAts1gwrXSwwcu5RiJFcuZ7otnBhHANa4eG0kGWXp+Wo+TLSJrYFdH0iV2aLml+FT+TdpdIWn3md09KkvquWCkplZXjWDCtJ/PYNkF4c1XfXu0dHNO5x8/N21UG4wp+qowxn1YqWP1f1tpkersUa639D99GBwAAgEDsPDii+zfs0RuXL8jsc5ZZFKs+Hw4ZxZP+dZ537hg3wyQDtfv1w6kWo/3DsUwAa0f/SOb5CdHKsxK6wiFN7o7o4yuXlv2aWiaJTgZFr08ZFPnGEqmgAOe03mjpg9rMQ8/uzdn32PMHJEnvfNliHTlncma/j02FqvaWbz8gKRVMidIetqhiYb/LlApQfEHSWHq7FCuJAAUAAEATeXjLPp24YGomTbwc7/zBKq3d1q8zj5mtab2pLvPO9/xiN5/DxmSWeFClHl7P7DwoSTpu3hTdsnaHJOmWJ3dknp8yofLJdihk9Pjl51b0mkoCAF6Do/4GKPJ9TG5bu0N/d+phZb0+2Qwz8AYby1Nv4u6nd0mSLnvD8VrVNx7ACDr76qntB7O2uyIEKIop9tM5XNIR1tox13apP0fUb6gAAACo1GPP79df//f9uvK2pyt63c6DqbvaMVdlfGfpRiRU+Ctk2OclHo6gJxnwx5LZqVaQLz1ipi588SJJ2UsYvEsx6qWWDIrhWHqJR5c/SzwGRsaXuDhBj6v//pSCx3/zouVZ25340XCCMlMK1CxxL/cJusDuuV+7J2ub32XFFfxUWWs3F9sGAABA89t2IBVoeGr7QEWvyyytcH2bfnpH6k5g/0is4Osi4VCmmj5tRuHYMzCqq+7coJFYUj3RkEIhk7mT7BQz/P47VjRsPMXqVZTiZFD4VSTTnQ0wNJbQRS9ZVDRQk/BkTIzG/VtS1Sqcv3Oh5BF3ELWZAgIHhmPqiabG9onzyl+S1EnKyi8xxkwyxiSMMZ+q94AAAADgH2cyU+mEzDnaPQH411/9RZJ0Z5GuAeGQybynH21GnXM00yQDlfvlmuf1w/v6dO1DWzKZB1FPJk4jaykUywIqxSn0OcGnAEVOAc8Sn1VvgGKsIwMUqX+DgdG4PnLdo5l/E4d7CU+yiX55DI3FMwHci049NODRNKey8pKstQPGmP2SOrOHDQAAQItyJjPhCtfchzzFKa21Wcs9CgkbkzOBqkWmi0dHJrK3j+0Hxgth7h1MrSD3XpONWt6R770r4bRK9WuJx4rDsttgrn2hv+jxCc+E+9i5kwsc2b7cWSO/fmRrpk6OI2uJR8NGlcu7vCSesJkiq9UUhO0ElYQO75R0Rr0GAgAAAP9VnUGRPtyJNbhjDicunFbwdeGQUSzp4xKP2k+BJpCvDoD3mpzumWQ2q+GxhLojId/aRRpj9IaT5me2H96yv+jxUz2FRDuxbWXMUyTz/o27JUkvWTxDkrIKAgeVQXHS5bfo/T9Zk7VvNJ7U8FhCXT5eP+2mkgDFv0p6hTHmcmPMlFrf2BjzA2PMTmPME659lxljthpjHk3/Oa/W9wEAAOhkTkeNSr8Mhzw1KNytQ197wtyCr4uEKZKJbOu39+uaB3LL2XmvSXexzHq7fd2O0gcVMBpPqtvnTgyVBBBfc9whevfLD89sd2ATj8zvGMf6dKeMh9LdO7J+nAH9fA4MxzKdahwjsYSGxhK+dYBpR5XkJd0uqUfSJyV90hizS9KQ5xhrrV1S5vl+JOkqST/27L/SWvvlCsYFAACAAhLpwEK1GRROYMAVn1C4SGpEOGQyhTn9iFCML/FAq3rL1Q9kbX/9b5ZJyr4mb/7w6Q0d06LpvVW/Np5MVtSy12/GGH1i5VJFwkar+vbm1F/oBN4MCseLFkzN2RfE744X9g/n3R9LJFMBCpZ3FFTJJ2uLpCcl3ZP+s07SZs+fLeWezFp7j6S9JQ8EAABA1arNoHACA84SEfe692JLN9zBC3+LZBKiaFXDsewJ9LnHpzJw3JP8Y+fWnKBdkbOOnVP1axPJ8Qwjv1R6dYdDRh8/b6kOmdzTVEUgGyVeIG3kC28+MWdfED+fq+/amHd/PGk1HIv7VmC1HZWdQWGtPbOO43D7oDHm7ZJWS/qotXZfg94XAACg7SRrXOLhBCYSrpTq7mjhe1x+r6smg6L19UTDGhiNZ7ad5RG1tPoMSiyR1Prt/b6P3T2JfuBjZ5f9ulCoM5c/xQtkUMye3C1JioaDbTO6ZPbEvPtj6RoUfhVYbUfB5Sbld7WkJZKWSdom6SuFDjTGvM8Ys9oYs3rXrl2NGh8AAEBLiWeKZFb2tS/TZjRPBsWcyT0FX+du7+fzTWa0KHdwQkotUZBas7jj529cr0e27Nf2/pHSB1dg067BzGNvEcxijDEdl0GRSFqNxHIDFBO7wpo1KVVodf60CfryW07SlJ5IID+faIEaJbsGRnXnU7vIoCii5P+pjDF/Y4x53BgzbIx5zhjzn8aYugQ2rLU7rLUJa21S0vckvaTIsd+11q6w1q6YPXt2PYYDAADQ8hI+ZVC4i2QWE3YFQvycgHbYHKwjRAOs4+A2Eiu/hsOfN+2pyxge33og87iSYKJR5302Tv3cbbrqzg05WSwzJnVlgl+SdMEpCzV7cncg2Vf5AiiSdMnPH5UkrdnMIoFCil79xpiVkv5X0vGSBiTNk/QxSVfUYzDGmHmuzTdJeqLQsQAAACgtlqitzagT4HDiExe95NCir3MlUPiyTt+wxqPtTeoONt39jd+6r+xjG3E3Phou/3MT6sAMit0DY5Kys7UkaWKeZRPGmEDq15QKeiU6sfVKmUr9NrhE0h5J51prHzbGzJL0f5L+0RjzKWvtaLVvbIy5VtKZkmYZY56X9BlJZxpjlin1v6A+Se+v9vwAAAAY7+IRqjhAkTreCUw4mRTLF00r+rqYq1aFHwkUziksEYqW5J0cfu3CZVnbfVesbORw8nJaVJajERNLU0FgL2Q6N3YXDYU0ovFMhXz/jiETTIbJaIkABcvfCisVoDhF0tXW2oclyVq72xjzcUl3Sloq6dFq39hae1Ge3d+v9nwAAADI5dSSq7aon7O0wymSWSrQ4U5V92OJh7fdKVrLWPoCfOXRs/XPZx+pFYtnBDyiwrbsGdKugVGdctj0gsckmuxC7MQMCsdBT22TI+dMyjnGKPfnMzga18Q6Z+3c8dTOos93F6hRgdI1KKZJetqz7ymlgtnFw+cAAAAInJNBUWmwwPlS7/zXmZhVUjbAlyUeNZ8BQTowHJMknbN0TlMHJyTplV+6U2+++v6ix7iLWfrp0Bm91b3QjGc5dbo3n7wwZ5/xZFDcv3G3jv/MH3Xfht11HcsTW/uLPt/VJPVXmlGpn4yRFPfsc7b5qQIAADQ5p4tHxQUJ01/qnS/a48U2yz+PHwEKz3DQYnb2p1aEF+v8EoQ56XaUlUjWcXnH8kNT936vvPCkil7n52esFRRqLypJPXnaHx8YjmnL3qHM9uq+VHHKBzbWp9hpMe6siS4yKAoq5yez2BhzsvNH0onp/Ue597ueBwAAQJOotouHMxX7zO+ezD5PBRMif5Z4pM4RRKE71G7nwVQ7zkOmVB4QqKdpvV16/yuPqOg19Vze8e6XH65wyOhlS2ZV9LqQaUzhzmaxd3As89j762Xh9NwslG0HRrJqUzivqffP7MWLs5cJ3fIvr9T1//iyzHZ3hDajhZSz+OY/0n+8/rvA8fy0AQAAmsTDW1J3DCsJLEi5AYHxQEf55/C1BkXNZ0IQdjgZFFOaK4NCqvwudj0LZJ60aJo2fu68il/nrUHx+PMHdNisXk3pifo5vMAlk1bHfOomvXHZgsy+85ct0L0bdmvXwdQ11ttVehrq1NCpd63TsXh2pkckZBRyZXjMn9Z8n4dmUSpAcXlDRgEAAIC6WJVOaa40E9z7Bb6aJR4TypgwlJLp4kGEoiU5NShm9HYFPJJchT4SG3cNaMns3IKL7rv3zcKYVBBoy54hLZw+QX911b065bDpWXfr28FwLKFYwuqXa57P7PvCm0/UxT9dozvWpwpSlhNwCjUoI2vUE6CIhkNZ4/va3yyv6/u3sqIBCmstAQoAAIAO5G3rWU2RzB4/1ll7IisjsYQ27hrQ8fOn1n5u1F3f7lRRyXz1AYJWqJ3n8Fj+FpH/fv1j9RxOVZy/wzlX3q3HPvMaSdKazfuCHFJdeDMSbv7w6eqKhLKWeRQrPBlPJBUJjx9f75+Rd7zRcEiHuLKIZk5svoBds2i+3xQAAADwXaU3DL2dAarJoOiO+rfy1wmYXHnr01r5jXu1adeAb+dG/fx81XOSCgcDglRoSIUyf4IorFiKM+Eeiyc1GquuY08r8GYkHDZjoiTpnS87PLOvWAaF0+7WyaBYXccARTJptWl3dreX3u7UNXXq4alONtW2fe4EBCgAAADa2JSeVMKsNyOimCe2HtDW/cOZbWttVUUy/cigyLxbevhOTQ1n3Tma2ymHTS99UJN5zZX35F0CEK934YIquIf5uRvXSaq83kwrGI2PZ7VM6YlkgkivOGq8qGi+AMX7z0gVQh2LJxVPJHXbuh11Hun4siY3pybIde8/TX1XrFSENqMFlVMkEwAAAC3KCSxUkkHx+m/em7V9+7qdmtid+tpYKoHidSfM1U1PbJckX76Ee4tkOndA69lRAf7YtGtAuwdGdezcyUEPJS9vvCHVESP1mRlLJDOdFjbsPKg71+/KHPfdt52iST3NMY2KJ8b/EtetTmWrVJDk1DLcGRShAtkH+ZZ4LEp39hiLJ/Wt+zfqz5v21meALo9vPZC1TbZEZZrjkwUAAIC6cO761nLz99ePPK83LV8oSYqUmP2cctj0TIDCD0ZOUbvUtpO+Xs+OCqidtVZnf+VuSdLx86cEPJoCPEGucMgomZ7wxxNW6Zic3vSt+3VwNJ457jXHz23YEEuJeddiaTyI106c5SuStH8oN0NBkrrzZFA4WRVv+c4D2teAIqexRFJv/8FDkqRjDpmsp3YcLBhQQX4EKAAAANpUImkzdx4rWeLhdePj2zN3Hhudmeyda4Ub1CYQtbl3w+7M4ydf6A9wJIV5r6GYKxvh+oef19tPWyxJmeDEohkTNH/qhEYNryzbD4zk7GvHJR6bdpeuOZNviYcTtNi8Z8j3MeVz1CduyjzuSS9DIT5RmTZMAAIAAICUvRa61hURTovFUndnk+k3Om6ev3fNr75rQ9b5r7m/z9fzw18DI/HSBwXMvUxow86DWc99+rdPSspuR/nc3mFFm6x2gLtWjKMd79hf8vNHSx6TL0BRrLNHvfWmiwS3Y8ConprrEwYAAADf1OOLcaklHs58zl28rhbO3+CaBzZLkmLx1BvcsX6nL+dHfTRj1w6vpCv4cN7X7817jDcQ1rdnMO9xQTnrmDk5+9owPpExdUJU1/7DS/M+ly8YUSygtLbOmT0zJ6Vaibbjkpt6KrjEwxjz9mpOaK39cfXDAQAAgF/cE7B8XQmqUaoAn5M279dX8qd3eFK7+a7fElqi1aXrI+G0ofT6/WPbsraf35ebsdBsWuJnXwZrrazNzgj55cWn6ehD8hddzVeUt1jr0SdfOKDjfKyPkvSsGZo1qVtS4Xa2yK9YDYofKfWxdf9I3T91k2efJBGgAAAAqJNE0urS6x/Te04/XMfOLf7l2h2g8KtmQ56afFkytS58+lL+i3RnArSW61ZtCXoIJSXLCNrlK7zY7Fohe6Uch3/sRknS2s+em9lXKDhRyLypPQWf87sR0I1PZAezli2a5u8bdIhin7izJJ2d/u9Zkl4j6VFJz0n6hKQ3SnqTpE+m9z2SPgYAAAB1su3AsH655nm978drSh7rDkr49WW8fyR/BX3v+xifIhTezI+Hnq1/m0DUxlqr29Y1/xKccoJ2frTKraePvObo3EKybRKgcPzwvj5J0kdefXTFr53aG83Z95LDZ0iSJqSLWPrlg//7SObxhSsWac7kVAaFu/gqSiv4ibPW3u3+I+l0ST2SXmStvcJa+ztr7W+ttZ+XdKKk3vQxAAAAqJN4+stuOW023ZP7Wrp4uJU79fFrjuROV98/VP82gahdO03IIk2+XGJKT1Qff93SrH3tsMTD/bvrS398SpJ01JxJFZ/HW5fip+85VZ89/3hJ0oeufSTfS6ryyJZ9WdvGSJN7UsGRYstMkKuSn9a7JP3QWnvQ+4S1tl/SD9PHAAAAwGdj8aRG44nMWvmt+4dz1jx7JbJqUPgzjlLp42ccPVuS9JrjDvHl/dyTrWWfvTXrOb/qasBf8VLrgJpEOZePd7J/yauOqtNoqjexO3vVfjskUOQLch2sojOMNwPm6LmT1BPxN3NCkvYP5WaW9Xan3qfX50yNdldJgGK2pGI/3bCk3DKyAAAAqNnRn7xJx3zyZo3Fxyd/hQr7ObKXePgzmXfSows5YcFU9V2xUssPne7L+/V2FS6ZRnyiOcU9gbNj51ZWN6BRyskq8mZQTJ2Qu2QgaJN6ipUVbE0/z1PDpJrMEG8GRXckrO5o6Snwf9+1QYsvvaFkENgRCWeP7a9Omq8F0yboRQum6ksXnFT+gFG0SKbXekn/YIz5rrU2K4fFGDND0j9IWufn4AAAAJBtNJ4YfxxLqida+P6R+8u1X0UyG50+fsKCKdq6P3/nBOITzSnuufvdyksOnvS0ovRORJvBpO7x3wHRsNHEIkG9VnGDp3uKVLio6fffsULb+0fyPtcVCelrFy7Th697VFKq6Gk8UTpA8cWbU8tKEtYqVMbCtqGx8d/LfVeszDz+/YdeUfK1yFbJ1XuZpF9LesoY8wNJTyn1/4WlSi3tmCHpAr8HCAAA0OludlWHH3VlUKSCFYXv6Lq/zydaNN2gWGAlaa3C9B1tOs4Sj7euWKhfrH5eoRZdc2Ct1UsOn6Ete4cy+4ZdE9Fm4QQpX7x4uuZM6dG6bf0lXtH8RmK5P+dCAYpXLS2+nGzRjN7M465wSKOleiW7lPtr88BwaonHu19+eNnnRn5l/+tYa3+rVABiRNK/Sfq+pB9I+tf0vguttf9Xj0ECAAB0sot/+nDm8YBrHbY7WJFPVpvRMlMoBkYrX+cdlBaNubQ9J4PisJkTJUkXnLIwyOEUVOr6iSVsps6JHocAACAASURBVBOD47ombHvrdO2wNjUB92awtKJ9eWo6vGhBdW073Z2HQiGjqb3Rolk9O1zZGOUWF+5PByguOaf5apS0moryf6y1vzHG/FbSKZKOUKqQ80ZJa6y1rVENBwAAoIX9z5+ezTx2L/fIxx2gKKfrhyTtOjha3cDqpNgkstAdVQTLudYOmdKjjZ87T62ywuOoOZP0zM6BzHY8mcwpONmM2SDd6QyKab1RRUJGsRK1aVrBPk/Hnk+uXKrj5k+p6lwnzJ+as+8DZy7RN+7YIGttTuHfN1x1b+ZxOb9ikkmr/UMxhYw0ubv1l9cEreKeJ9bapLV2lbX2Omvtz9OPW/9TAAAA0AKe2zeebv6rNVuLHpusYomHMzFonrveBCGa1VV3PKM1m/fl7HcmyNGwUThkSnZ+aRav9nSeicWtvnXnxqx9b3vpYY0cUllOWjhVn1y5VF9484mKhENt0ebV27HjvacfUfW58mVLOK0/r7hpvfp2D2Y9t6N/PEhb6tfmzU9s0xEfv1GPbT2gqROiCrVKNK6J0ZQVAACghSxOp81L0p3rdxY9tpolHrH0spE3LltQxej8RwZFc/rpnzfry7c8rTdffX/Oc04GRasVx/QG8UYTuRlKF754UaOGUzZjjN57+hGaOalbXWHTMm1eiznrmNm+nWtCOsPk4jOWZPZF0909vnPPJr3jhw8VfG2p3zG/fjgVJF7Tt7doxyGUr+BP0RjzrKSkpGOttTFjzKYyzmettUtKHwYAAIBqTOgar9hfarKUtcSjzMm807rUucMYtI+dd6xuLxCIIT7ReImk1UXf/bMe6ttb8BjnDn6kgmKEzeDPm7L/TvlqOXQ3yeeikEg4lAkytrLe7oiOmD1R3ZFwVpeSakzoCmv9f7w2q+WoO3jmFLjMp9SvGCdbKBwyTdnhpRUVC/NsVurfxPl32SJy7AAAAAJ1h2uyPhxLKJG0+uXq53TBKQsVCWdPntw3Ustdlj6Wntx0R0L69t+fnFWgMwhHzplc8DkyKBrvP29YmxOc6Ns9qMWzxjN7nDv40RabsG3bP6yr/na51mzepx/e15dVy+Hxy16jF/aPNP1ylWg4pJhfPYUDlExaRUJGN11yui/n87ZjfuS5/ZnH+z0FOV+2ZKbu37gnNY4Sv2OcYFwsYVsuY6hZFQxQWGvPLLYNAACAYH3pj09pem+XPv6bx9U/EtP7XrlEA6NxnfCZP+q/3nSCTlo4XvW+3CUeToCiKxLSyYdOr8u4/dL607DWc+Pj23L2nfnlu3TbR87Qzx7crA+fc7TiLbLEw3omn8fMnazXnzhf/cOp+gfuWg6Te6I6Zm7hlr7NIhpujyKZ8aRVuI4ZOMWuTHegt1SlRednPRxLKNLk13uraO4cJQAAABTlFLXcO5i6C7gz3SLvf/70bNYSiNvW7SjrfO4lHk7BtxkTu/wablUm9+S/p0aZ9sYzBaZ2P/3zZv3wvj6ddPktmaUR0XBzTzUKBbicYbfiRD8SCsna8rv2NKtE0qqel483E+aRLePFXt3PlGozGnf9nOsZUOkkVPIAAABoYc73bOeLtLPuP5ZIamB0vBL+nsGxnNfmM+pkUIRDCqdPnrRW37houWYFFKgo1Nqx1OQB/it0k7h/ZDxN/tcPPy+pdZfgOJPXfDUoml00khp7LJFUOFRb7YYgJeqcQeHNntmyd0jL0xlj7l83peI8cVcQiwwKfxQNUBhj7qjwfNZa+6oaxgMAAIAKZO5oe75IWysNjcVzX1DAAxv36KLv/Tmz3R0JZQIDyaTVG06aX/NYq1WoiF2L3yRuSYVqMLgn8z9f9Zwkad9Q4eKDzciZszqBuTWbCxcCbVbOJDne4h+ORLoGRb1M681eruMu6JqVQVFmDQqp+Zc0tYpSGRRnSopJKi/kzlJAAACAhjKe+ETMVRmzkknK/z60JWt7am80k03RrDfCS00e4L9CP/PDZvbm7Btr8m4S3r/KZW84XpLkzFUv+/1aSdIFpyxs5LBq4ky04y24PMUtnkxmAkX10BXOzi4ZiY23lHVnbJXMoEiSQeG3UnkzcaWCSLdJ+jtJU621k4v8mVL3EQMAAHS4T5y3NPPY+UrsTBzdd7K9KerFJvRj8UTWdnckPJ5B0aSBgBa/SdySCnWIyFc883UnzK33cHzzVyfN15FzJknKXVLUShNPp3NK356hgEdSm9QSj/r93N+wLDsjzH39uv/5Sy0jc9dkIYPCH6UCFAskfUzSkZJ+I2mrMeYLxphj6j4yAAAA5DVvWk/msfcutbuwn3N3769PXpB+rliAIveOq/N9O+jWit4v/kelJ5Luu55ojEJ35jfuGszZN7G7dcrdzZ3SnXmcE6BooXapTgeKN37rvoBHUptE0tb1575s0TT1XbFSf/5YqjrBWcfOkSRdfdfGTItRqXT2WMJ1QCtdJ82saIDCWrvLWvsVa+2LJJ0m6beS3idprTHmAWPMe40xhZtTAwAAwHddrvL2X7n1aUnS9/70rF5+xR266o4NkqRtB4YzGRQToql05mJdCfIFLyZEw/qnM5foF+8/zbexV8Ob+eEELE7/4p1BDKejlVs4sivS/B0N3HfH3ZeYNyD2zI6BRg2pZq2U7VHI4GhcD2/Z35ClZc516nQ9+cLN6zU0Nh74LJU9lqCLh+/K/ilaax+y1l4saZ6kt0salPQdSS8YY/6+TuMDAADoaPmWZUQLTP627h/WzU9ul5Ra/vCzBzdLGg9QFJtc5sugMMbo3157rI6bH+wqXu+qAlKpg+PUNXnVsXP0l0+/Rn/40CvyHnfErImNHFbN3NeY9/J68NnWKZbZ6nfxrbVa/tlbJUn3bthd9/dz6lwUqtdTKkhCDQr/VRzmsdaOWGt/Jukzkm6XNFHSEX4PDAAAAPm/IHeHy/sK9/CW/ZKkCV2pAMVAka4eo67siklNlpp/7NzshN0nX+jPPF586Q3aW2YLVdTmyluf1nB6Wc0Zx8zW1N6oTlgwNe+x67cfbOTQaubOpijU1rYVRFr8Lv7TOwY01sACn+HweKeifAplUGzeM6j+kZgSdPHwXUVXsDFmvjHmUmPMekn3SFoq6fOSfliPwQEAAHS6fF+QK02f70lnULz8isId5N0ZFP/zjhUVnb/efnlx8SUmJ//HrVrrClrAf9Zaff32ZySlMnLe9tLDAh5R7dwfLZuVQZE90fzkyqVqFdEWzqBYfOkNOvdr9zT0PZ0Mit0Do3mfL5RBccaX7tJbrn4gK/OCDAp/lPy/mzEmaoy5wBhzg6TNki6T9Jik10s6zFr7CWvtc/UdJgAAQGdyvv++4shZmX1dkZA++uqjyz5Hb1e45DGjri4ezpKQZjG5J1rymA/+78MNGEnnct/VHo4lAi+c6oeRWP479c6d8NOOmClJeu/prZMs3srZH0Fw/q2/c8+mvM/nC1A42RZP7TiYVYPipie2+z/ADlQ0QGGM+YakbZKukzRf0kclzbfWvtVae5O1trUb7AIAADQ5J4Pi5MOmZ/Z1RUKqZB5SzgTfnUGxvX+k/JM3ic17W7utYrNb3bcv6CH47r2nH5557K714ny2Yolk0wXrSmmHwFEjubMe8tX7yddm1L1UrlDtClSv1ALDD0oalnStpIfTx7+zyIVvrbVX+jc8AACAzuYEKCZ1j0+UouGQ1m4rf0lDORkUz+8bzjw++dDpRY5sTvkmF/DPnet3VnT8xWcsqdNI/LN03hRd/obj9ZnfPZk1DXXuqo8lki2Xtt9aow1eKGT0/9m77/g26vMP4J/TsLzt2EmcxBnO3oQkzobsHVahQCmbttCyCrRQNpRR0rJayipltgUKP3YJBJKQQUJ2QnbiOImznDje25r3+0O60510kiVb0/q8Xy9e3J1Ouq/ls6J77vk+T3aqEZP65aqyISRa8Ye6Zqu83Gj2XdeH2iaQCkgpAH7u+q81IgAGKIiIiIhCRPrSrEzdTtLr8NWuwNOJjQEW1ZTkpiUFtX8kXFbYEx9uOeHz8cvH9YrgaBLPyJ7axTB9mTqoc+s7xQDpz0pZ60X6W7PYHHHXFSPOa2SqnBXkOddWPbJSsO1YtWZhV60aQHXNzKAIp9YCFDMiMgoiIiIi0iR9/1VmsJqCLJIZbOE8XQzeNbYqquVnJhtQ16K+c5mdGntBlY5EOQXIs4XorTMG4MWVxaptyXEyNUIKADYozid1gCK+rvg7yhSPdffORHZK61PTQqG6yYKyOjPO+/tar8e0ErMa/XRDovbzG6AQRXF1pAZCRERERN6kgmzKGEOwXTySFBdZK/aVYdbQvJCMLZKsiiKNSQY9APVFwv5TdXjnhxJcO7kgsgNLEC2KAEVWqvrC8ffzBnsFKIINokXLkp2nAACf/ViKv/5sNABAis+ZbZziES352SkRO9apWt81d5otdq9tVpt2GcZ4OedjHd9FIiIiohgmpRgrsxqCvTudqbgT+Yt3tqC60eJz37xMU5AjjAybXZmC7/34ygPlzloCrEURFi2KCzVjAPMI4uVizaxxsSnVoGi22uNuigeF1jPfHvDapuxoo/TI+cPDPZyEEB+fHEREREQJyi66a1Ac/tNC7HhkLpKNemSYAikl5jS0e6ZqffTjy1TrDsU8aiFG78EqMyj8tVLknPDwaLa6AxRa8/Il2a7sCpMhPqZ4aJ1K0jSJqkYLDHFW1CE1KfDPhViibHMcS2wOdTBCFEVc99bmKI0mMcTXXxwRERFRgpGuBXWCAJ1OQJYrG+Llq8YE/Bp6neC3q8LxaneLzlidwt6iuIAZ0j3D534WH+nX1D4tigCF3U+A4rWrC/Ho+cPQKyc1EsNqN63TXZmhE29TPMb3zQEADOnm+28k1nyxoxSDH1wa7WFoWldcqfpMOV7V7GdvCgUGKIiIiIhimDzFw+M66dyBXXw+58Kze2DRWd0BuNPV/dWt0CtePFYvx+aPcP48L/18DG6bOcDnflYf6dfUPsoMCn9TPHpkJ+O6KX0jMaSwUYZf4q1IJgDMHNK11akppTXNKLh3CTaXVEVoVNpOVDfh9ve3R3UMrTlS0Sgv3/AOsyfCLT5zgIiIiIgShNxm1M+d3E9vnozHv9yLu+cNwdLdp/DHC0cAAO6d34TcdGd3C381AezKKR4xmkJx1YTeuKywJ0wGPdYerPC5397SOkweEB8tLuNBaU0z1hVXoLrRggyTAT1zUvHgeUN97q+Ps4yDThotdZUJIvFSS0MpkN/AD4cqAQDvbzyGcQU54R2QH1oZTzMG+w6+hsMlY3ri422+WxiLipBViSJYQeHBAAURERFRDFNO8fAlv1MKPrl5CgBgUv9cebsyzb7B7Ls1nvIiwR6jNRwEQZDrGljsvuerrzlYwQBFCE1e/J283DsnFV//9ly/+xvjLOPgd3MGY9WBclw8Ol/epiy0Go8BCkC7PaaSVHcm1gKSPxmdj2cuHRXRY2rXIXG/h8rPR381bkTE5mdnvInPvzgiIiKiBCEFDLSu+1KTnBfsgRTye2vdEZ+PmVVfwGN/ioS/n3fl/jMRHElikc43LXfPGwwAyE4x+twnFo3smYU3ri3Ekz8ZKW9TXoMG2zEnFgQSc/jz0v0AtD9XIskzQJJuMkQ8C8fhEXR47rJR2PzAbHldK8tj5pCu8nJStN/EDobvJhEREVEMqG+xor7F6rXdoeji4enRC4ZDJwBpptYvov54ge8WeMq6DfFQZNLfBdiBsvrIDaSD82xHa/JzsX7LjAEoWbwoLms2zBqahxRF8EXZpWR1UXk0hhR2la7fbaxNyYlGDRnPoq+pSXpVzR5lVtk5ruysC8/uIW8zshVtSMXfJwgRERFRB+NwiBj56LcY+ei33o+JvlOxLyvshcNPLQqopWN2qvdce4kyKGG1x36acry1foxXysKYAJBiTIz3vSNcbrY2xUPib+pYJIgeA53YL9fHnuHjOa1NJwjITDbKhYZLKt11J3rlpKBLhkk1fc4Yp9OAYhXfTSIiIqIo+3LXKZ+PVTc5syq0siuC4e+CZcNhdyX/qyf1addxImFC3+gV9UsEoijCZnd4XbilxOF0h7YojGLRyNAIPOgQ7Sksx6vdbTtvmdEfFylqgUSKZ9aG1AHlhikFAIA/fLxLfsxsdcBk0Kk62Vw8uicAYELfyAdXOiIGKIiIiIiirK7Zd/BhXbGzY8X2YzXtOobDI0KhvHP5yupiAMD9C4fgvgVD2nWcSPDX0aRfl7QIjqRjuu6tzSh8cjnMNnUGRSCZOh2Bqu1unKZT+EugOKrICEgzRbdnwpc7SuXlzummqIyhokE9lUnvCj7oNTK1zHZXgMLgPjHOHdQZJYsXYUDX9PAONEEwQEFEREQUZZ5F1posNhTcuwQvryrG8B5ZAIBr2pnZ4BmgeHfjMVS55qH/ZtoAAMDl43rHXFX/YBk5/aNd9pbWYXVROWqarPjPhmOqx5buOR2lUUXPK1eOjfYQgtban/Dc59fIy55TLMLF4RBx3ye7sOtErWr7rKHuYpPKwpORdMGoHqp1vesNNGgEQpfsPIVD5Y3IUhaDjf1ZcXGFn+BEREREUWZQFFlbsa8Mla47eu9uOCbfxW5ver1nuv6Dn+3GLe9ucz7mukjJiPLd1FAoqWzE31cc9KrMT4F5aVWxvPz2DyUAgIvHRD7tPlbEawFEf4EHc4BtM0OpptmK9zcdw8//uUG1XTm7Ii8zOSJj8XTNpD4ofnKBvC5l0HgWEFW+p8psD3McFBaOJwxQEBEREYXImqJy2NpQhV7Z+eAX72xBg9kGAKhrsaLF6ny99s4V/06j/WZprXP+d5PZhtQkvd+pE/HCbHPg2WVF2HqsOtpDiUta5+9lhb0AAFMHdYn0cKIuHv8mtEa8t7QOt7y3DZtLqlTbPQOX4fL7/9sBALB4nF/KDhqmKBWbFARB9RksBYw9M1GUwRyjKuuNwdBQiv8wOREREVEMWF1Ujmvf3ITfzRmE22YNDOq5Ro+LICmYUN9iQ4urk0J7v7zft2AoPv+xFEa9IHfqsLn+32ixIzUpvr4WbnlwNmqarJj93GrNx1s8OlBQYGwaXVxsdhEb75+FzGSjxjM6Nn2cT3kCgIJ7l8jLS3aqC/JGKkAhfaZ5ZhsoM51iZXqZ1NlE2d3I7hC9PlM+u2UKrn1zE8azOGZIxde/REREREQx6lSNMxvheHVTq/uKflqHAsContkAgJH5WfIXelM7Myi6ZSWjZPEinKlvwfgnVwBw381sNNuQboqvAoitFdSL0NT6Dkcr5d/qcEQt/T7aPNP8O5pIBSh8OeH6vFx99/SojkNJCgZ3UrRmttodcjab5Oxe2djxyNyIji0RcIoHERERUQhIF3YGfetfr15aWYy+932FZovzjtyHW46rHpfSnrNTjSHLoJB0zXBfaEqv3Wy1R73dYFvo/NxxtTNC0SaexVQB7yKuiSA71Zkt4u8ci1WtDVkZc7E5Ils/QXkubT1ahWe+LXKNKXbeZylTqFdOKm6a2g+AM5gr1QO6aVq/qI0tESTepw0RERFRGEhz97Uqv3t6YYWzEGFNs7MY5soD5arHpbRnnSC4MyjCMD9bSmG22R1IitL87/bwl34fO5c78UMURXx/sMJr+4S+OVEYTXRJF6nxmkHhLz53zsAuyEw2oEuGCW0omRO0Gc+skpeVnTr2n66Xl2PpfU42uj8L8zulAACsNgdOVjuz5PKzU6IyrkQRf/8SEREREcWYFqsdj/5vLwBn3YjWSFMr/rvJmTlxuasIoUTKxtDrBGw4VAkgPPOzRRF46LPdWHmg3KPoW3wQ/Aw5lu7IxostR70Li94zf3BAWUEdjXT6xOOPLrQSnrPY7BjSPRMGnQB7BDIojlQ0ysstNncdBymDDIitAIXJ4M4mkz4XrXYR76wvAQAs21sWhVEljjj8kyMiIiKKLSv2uTtkfLr9ZMDPk4IZUjq55E9f7QPgvMje5FF1P5Qcooh/bzgKID7bKfoLQtz+3+0RHEnHYLZ6X6wadYl9uRCvgS7RT2cJs80Bk0EHvU6IWJtRSU2TVV5uNLsDFLH0PiuzyaSMOKvdgbnDugEA7pk3JCrjShSJ/YlDREREFAJtvcu691Qthj28VO6qAQCdUo3yHccGs9XXU9tFulvpu21efPB3SaO8EKLASOfFnGF5XtsSVTz+/K1d628/VgPA+bM5IhygaLK4M8yeX14kL8fS26wMUEjLFrtDrtmTm56k+TwKjfj7l4iIiIgoxrT17t+Gw1Vostix5WgVMkwGXDe5QBU0UKZAh5JWnYx4DFDE0l3XjkC6eExNcqe4G+IwsyYUpJ+6o55j3x+siEgGRXWjRbWuzJpQioVA0G9nDUTvnFTVWNxTPBxocn0eK/8+KPTi718iIiIiohjjeRFnDbLynNnqgCDANSfcfcFwpt7sfP0Qf3nXDlBE/wIhWCaDDhnJBlw9sU+0h9Ih/HnpfgBAueu8AwBDgk7xkGq+xMKFc1tIRTJFP9UyPT9vwmH048tU62abdoBCFwPv851zBmHNPTNU26QAhc0uKgJ4hoiPLZEk5icOERERUQh53mV9aWWxz321LhgOlNVDrxNg0OtgU0z3OFXbAgB49rJRIRqpk1bRw7jMoNAJ2PXoPFw7mQGKUCgqawAALBzZXd4W6uBYvJB+6nj8+ZUfR77iD9dNLoBepwt7gEKpS4ZJ1V1kvKI7TDi6FIWCFLi12B1otNhh0Alx2fEonkTt3RUE4U1BEM4IgrBbsS1HEIRlgiAcdP2/U7TGR0RERBQozw4bnm1DlRp9TNuwO0QYdILc4UPphKu9Xaho3RVOisMAhURZdZ/a76dje2LhSGdBQJMxfs+LdnH9icTrDA8pDuArAPHI+cNgMujQbA3PNDIAOFbZJC9fN7kAC0Z0g0MRoVAGf2L1b1j6XLTaHGi22Dm9IwKi+YnzNoD5HtvuBbBCFMWBAFa41omIiIhimmehuR3Ha3zu+39bjmtut9gdOFTeoPnYvOHd2j44DVrXXPGYQSHxdffVX3o7+ZZs1OP+hUNx4dk9Qn7uxQvpbyQeTyFlm9F9p+q09xEEZKYYURdAW+S2mvr0Snl5XEEOBKgzOqRssfsWxG5XDKPrs6Ws3oxGsw1pJk7vCLeo/UskiuIaAJ59sy4E8I5r+R0AF0V0UERERERtEEzNiUaz9gVBi9WBigaz5mMDuqa3aVy+eGZ8APFdDNFk1L6rWeVRoI9ad9O0fgCAnp1S8befjUayj/e2o4vDuITM7hBRfKYB7286hgtfWudzv2SDDuYwZVDYPD4TM1MMEARBFTS0Ohw4d2Bn3DStf1jGEAq1rm5At7+/Hc1We8L+PURSrIXK80RRPAUArv93jfJ4iIiIiFoVzDzu4T2yfD6Wm2YKxXBapRVQ6SgZFBeM6iEvf+AjW4W8SVlAKbwAAwBcOCofANAtKznKIwmeNG3jgU93+d0v2aiXW2eGmtmm/ozJSjFCJwiqjBSbXYz5Gh+ZKUZ5ecPhyrC9X+QWt/8SCYJwoyAIWwRB2FJe7nueJxEREVG4WYMIUPib06+PUBaDViX9A6frI3LscFAGKJ651F1Q9C9LD0RjOO22rrgCm454JhqHl9115aiP16ILIXb7rAHY99h8ZCQbW985xlhcwYHWOpAkG3VosQbXcSjYMUicAQqoalBY7Q7Ngr2xZFyBsyTi/OHdUNFgkQsXU/jE2hlRJghCdwBw/f+Mrx1FUXxNFMVCURQLu3TpErEBEhEREQHA7pO1KLh3CYrP1KvSmecMy8Ow7pk+nxfJqvm+eHYdAYD1hyujMJLQUE5Zicd2qZ6ufH0jLvvH+ogeUzovIxUki3WCICAlTgsiSoV2W2sRm2zUh61IZotHEDQrxQidTlDVoGix2mM+Y0cQBHTLTEaCdtuNilh7q78AcK1r+VoAn0dxLEREREQ+fbT1BADgrXUlqtagJoPO68u5knJfT1qXhulhKMqmdZxbZsTuPPBgaNXXoNbJAQq+f3FPK4Mi3WRAyeJF+N2cQfjH1WMBqKd4NJpteGPtEa+Cv211vErdeSgj2egqkul+/aY46Yqh1wlhyzQhb1ErQyoIwvsApgPoLAjCCQCPAFgM4ENBEH4B4BiAS6M1PiIiIiJ/3v6hBADw7sZj8vSI9381ER9tPQGzny+zZXWBpQj/6ty+GNYjE4tG9mh95yBpXcT7q40RL7pkRKaGRzhFa6qNPMUjxmsCUOukAIXyN/n3K0YDAG6bNVDelmzQwWxzQBRF/OmrfXh34zH07JQSks4tnkV39TrBWSTTtb6mqBxn6s34345SLL7krHYfL5ycAQrWnoiUaHbxuEIUxe6iKBpFUewpiuIboihWiqI4SxTFga7/R3byHREREVGACnJT5eUtR6sBACN7ZsFk1OFkTTNueXeb5vOOVjWp1r+5Y6q8rAwcWGwO/GR0TyT5aKHZHlo3yeP9wnTF76bJ7+WLPx8d5dG03driiqgcV7pzrjX9h+KLNMVDGaHITPG+Ly11v6lttuLdjccAAM2W0FyIa7X41Qnu7c8uKwIApCTFfttOBigiK9ameBARERHFhT65aap1nQCkJenlgo1Ldp3SfF6nVHXRvUF52i1E31l/NASj1KZ1ERrvtRv6d0lHTloSAGBUz2x5+7VvbgpZ2nokaF3YRYI8xSPOA1UEuXWo8u9c67SSWmY+9uVexbbQXB5q/cnpBHcNivPP6g4AWHbnVO8dY4xOAKd4RFDsh6yIiIiIYpDn9+90kwGCIMBk8D+n2uqqQfH6NYXolpWsypqI1KWh1k3yRnPHuUOovMheXVSOykZL3Ez/iEZbyw+3HMfRykYAgI4BirhX0WAB0HrQUfpVf7LtZEiOK4oi1h+uxPiCHM2goLKLR5MrUyM9OfYvR/U6Qe58dP2UgugOJgHE/hlBREREFAfqWmwA1C0vtUhFMmcM6SpfSK+7dyZyUpOwcQyodgAAIABJREFU80QNvthRGt6BArjo7Hy5hobEEaU79+Fg8LjIFr3CSbErGr+Gez7aKS+zSGb8k6Z45KQlycGKYT28Owst31fmtc1sa3umwKoD5bj+7c147MLhGNg1AwBw07R+7q5GgiCf341mG0wGHYwx3mYUAPQ6HRrNVgDAxH65UR5Nxxf7ZwQRERFRDPKVim9qJUXaandAJ6jv8udnpyAlSY8Jii+/MwaHr436Q+cNw46H52Kon3ao8cxzmoLVT+eUWBPpue7/Xl+iWo+D60UKkJTN9dTFI5GqUevhN9MGeG1rT4Bi76k6AMDJ6mb583Hm4K648Ox8AO6MDVEUUW+2haVDUTjsO1WHkzXOriTxPhUuHvAjiIiIiKgNfN3pTtK4wmuy2LDrRC0AoNFiQ5qfwnADu6YjPzsFr1w1NiTj1KLXCchKNeLr354btmNEk2eAwhxHBe7uVmQz1DZbw3osq92Bhz7fo9pm51T7DmPXSednjq+6IucM7Oy1rT0Bim/2nAbgLL4pzfBQThmSamI4RGcGRTxM7/AUDxkf8Y7vMBEREVEI7XZdFCgNe/gbnP/iWhSfqUej2YY0P3cOl901DevunSkXsIsUg67jfC30vCCzxOlV92Wvrg/r62tdjLJbQcfjOeVJ6cKz1W2M2xPM2+kKwpoMOnnKmPLQ0rJDFPH5j6U4Wtnk+RIxjwGK8OM7TERERNQGvuoa+GsYMfu5NWg025FqimzwwZ/fTO8PADirZ1aURxI6noVKLe24KxxpygvGA2X1YT2W1sVovAZzyDd/nVlWF5Wr1kPx+1cGKFRFgOUMiviZcuWJUzzCjwEKIiIiojbw9R07Ncl/8KHREltzr/8wfwiOPLUQvXJSoz2UkEky6PC/W8+R1+MpQGGLYEvUVQfKvbZ1j0IXEQovwU/h01SPTK1KV1HN9jDo3MUwdaoAhfP/0mP52SntPlakMYMi/PgOExEREflRWtOMz3882Wo9gH//YjwA7S4eI/Od2QlDumXgdG0LkltpRRpp/i5g4tVwRdeCeApQVDaYI3asZ789IC8/duFwvH5NIS4Y1cPPMyge+SroCwCZKUbV+htrj7T7eHYRmlM8dhyvAQAMeWgpDDrBa3pJPOhIU+FiVeyE74mIiIhi0DVvbkLxmQYAQMniRfJ2m0dniHMHOrtu3D5rIN5Zf1T1WIorq2L/aWfK/pQBbFUXbsrifGaNtHWHwzlJx1/6ezRsOFylWhdFEXaHCEMY7twqz+CpA7ugoHNayI9B0edvSkWTJXQ1R5L0OljsDtgdDneRTEXw85s97ramdlGMub+9QCQZ4m/M8YYhICIiIiI/pOCEp2QfUzly0024dlIfZCimcdg8LpCZJhwZt890tlFUZlCIooilu09j1nOrMfqxbyM2FlEU8cf/7cFBjboStU1W1LdoZ+iU1Zkx4IGv8eHm4yEf08Vj8uVlmyN+skzIv/F9c1Tr/spKvHLVGGQmG3DrjAGYNqgLRrWjFo3UleNPX+1X1KBwP678TBTF+Mzc0jODIuz4DhMRERH54BlYUHIoagW8drW6JWiyUa8qNudZV4ABisi4wJVCruxW8dHWE/j1f7biSEUj6lpsERtLRYMFb60rwZWvb/R6bNRj32LSU98BADqlqlPuV+x33nX+aNuJkI+pxep8Xx49fxj6d0kP+etTdPz3VxNV6/4yKIb3yMLOR+fh9/MGI82kb1dGhdUjEAioMyjW/mGman99HAYo/HVEodDgv45EREREPqwtrvD5mFURgJg7vJvqsSSDM9X57XVH8NLKYlg9poMkMUAREdLdTmUwqbSmRbVPpNpqSunsdT4yJRrMzmBJo8WOucPy5O0PfLobALDpSJXm89qq2WLH5z+WIictCddN6RuXd7NJm87jItoRYOHV1CQDDp5pwPubjrXpuMqgrNYUj6xUI965Yby8Hi8fg89dNkpeNrCLR9jFyWlBREREFHn+Ltr8dVtI0usgisCj/9uLp785ALvDc4oHv+RGgvQuK+8ge/4u3v6hJCJjsbvOF7PNgQ83H0d1o3e3hCaLDRabA6N6ZauCFOHw5Fd7UdFgRpXGOCj+JRvdl3n2ANt6prmmrd33ya6gj+dwiKpMJa0imQDQv4u7zolnICVWXTymp7zMIpnhx3eYiIiIyAdf6bw2uwNbj1b7fF6SRyePojJ1HYsIdpJMaNLdW+X1mdXjzV+xrwyRIF2wiSJwz8c7cccHP7rW3eMpq3N28OiSYYLR4xya2E9dV6C9TlQ3h/T1KLYoz/lAMyhSktw1Ik7XtvjZ05syOJGVYpQ/4zyDvF0yTPJya52RYhGneIQfAxREREREPnhWmZcuJpcrLmr3PjbP63mt1ZiwM0IREdK1kTqDQv3eby7xHWgKpc+2n1StV7jaiSov7GY8swoA0GS2qVtswFnXJJRWHSgP6etRbBnSLUNe9pyC5osysFrTHFxmjTRVSq8TYHeI+OJH5/nueT1vUrRYPlbZFNQxokl6bzjFI/wYoCAiIiLywTNAIV3bKmtKpCZ5d233zKDwxI4JkSGlkEvxibUHK/DamsMRH8cn207gXx6tZ6XsDqn2hNLMIXn4zfT+qm1N5vDUyhiUx+KYHdHb17trPeRlJgf0nCTFxfe+U3VBHa/F5jw/05L0sNodWL7vDAD/0+Ss/tqLxChO8Qg/vsNEREREPnh+gZbWW6sn2GqAws4MikjwrEFx1RveHTQAoKwuuHT2YN314Q6crFFPqZDOoXqNTiJdMkwYkZ+FBxcNlbc1WkLbcWR8QQ4m9svBt3dOC+nrUmzolJYEAEg3eQdQfVHWhLjzgx1BHa/Z1f0jI9moygry7EoDAA+dNwwAUNccuS467TV9UBcA3kFrCr3Az1giIiKiBOM5HUBab+1udmtdOvwV2KTQkbIUWnu7o/n7qGo0e22TChz2ykmVt0kXgLtO1MIhihjVK7tdx7U6HEg38lKgI9v20JygCvLuP1Xf5mNJLWs9j5edmuS179m9sgA4z8F48cIVo3G6tqXV4DO1H99hIiIiIh+kTIdxBZ0AAPtPO9OepVaRk/vnaj6vtS+xx6riZ+51PJNudoqugg6+fi+BFhEMJekyrtmivkhLMujktPhGxfQPKYPi/BfX4sKX1rXpmEt3n0bBvUuw9WgVrHZHq7VSKL7lpCUhI9k7g8GXaa4sAQBYODKwuhWSSlegrURRV+KxC4dr7itNi8tOCXxs0ZZs1KOgc1rrO1K78VOJiIiIyAdpSke/zs55+pe8sh57S+tQ12yFIAD/+cUEzedpVXqfP7wb3v2lc/80U2gLHpI2wSODwuTjgtwSxrnwvoIfep2AFqvda9pJ0RML5GXlFKOyOrPqtcQAW0cCwN7SOtQ0WfDU1/sAOM9jm11kRwJSaW3qmj9SR45zBnSWtx04rZ2RMbR7Jp78yQg8e9nZbT8gdVgMUBARERH5IKX+K+dmP/rFHtS12JBhMqi2K2nNU57YLwdTBnTGw+cNw8s/HxueAZOKnEHhuphPT9ae0mCxhS9AYfcRSDAZ9Djl0crRMz3+vLN6qNafX14kLzdZAi+aufCF7/HTV9fjssJe8jZmUJAnZSwtmPMLcE97u2RsvrztmkkFPve/ckIf5KR5T/8g4qcSERERkQdRFPHlzlK5dV55vftCclNJFeqarcj0k56sFbiQvqzfcE5f9M5N9XqcQk/OoHBdeaUmaWeuhDVA4SODIsmg8zqulKkjSfMocPjx1hPyco3rjrWvY64uKldlWRSfaYBJMcXFaheDqk9AHZ+yHW+wAQppClJumkneNqArO8RQ8FgZh4iIiMjDqqJy3Prednnds9NCXYsVmX7mdus1cqV9ZVtQ+LhrUDj5ahEYzikevgIULVY7Pt7mDjg8uGgoLji7h+a+klJFxkVpTTPys1M09/vX+hL88X97AQDL73J36ZAygkwGHWx2BwzMoCCF0YrCq2ZrkBkUroCGMiuCHS+oLfipRERERKRwvKoJf/56v2rbiWp1i8jKRgsyU3zf5/H8Ys471dHhWYPC1xz7JTtPobrREpYx+OoQsvFIFV5bc1hev2h0PrpmJHvt98WtU/DzCb29tl/66nqfx1Ser7OfWy0vS51AzDYHSmtbOMWDVAbmZaBk8SLMG54nd+UI1IbDlQCcxSSJ2oOfSkREREQKi5fux36P4m5P/GSEan3niVq/GRQ6jyvhtX+YGboBUsA8a1B4TqmQOhW8/UMJRj++rM3HsTtE+eJf67FA+OowclbPbDx2gXY3BF9MPl7rbysOqtabLTbN/SixJRv1OFBWjzN1La3v7PL9wQoA2gWCiYLBAAURERERgCaLDVe+vgHbjlZ7PTYyP0u1bneI2HGixudrSRkUndNNKFm8CHmZ3nfGKfx0cgaFK0DhMZVDK2OhLR74dBeGPrxUs7OGzeE85hMXjcD4ghyfr+ErqAAABr0OPbICG2uTxYaXVx0KaN9g75JTYpBCDNe9tTno53JaB7UXAxREREREAN7fdBzriiu9OisAgFGjdkFZndnna0mZ8zlpvrMsKPykRBYpicGz5eeCEd28niOKIpbuPhVw5gMA/HfzcQDOqROerHbn6xh0AuYOz/P5GkmtTLcwuVLnk406XDOpDzqlGvGnr/bhH6vVwYhXVx9WrftrHWm2BVdngBJDnavmTmltM5bvLcO1b25q9TnDe2Ri1pCu8nrndHbooLZhgIKIiIgI7ukAWgwaNSQePm+Yn9dy7i+AdxOjSfo9SIkNw3qoM2G0Mlv+t/MUfv2fbXhz7ZGgj6fV+UCaVmIy6vCLc/pqPm9ivxy5XoYvUobFzCFdoRMEVDdZ8dqaw3jKo16K3eEOkvzinL7w0eUUAFgkkzRJwTm9IOCX/9qC1UXlsLVSSLbBbEN6sgFdMkzIz07B4ovPisRQqQNiFw8iIiIiAEcrm3w+ZtALGN4jE3tK6+RtKT5aVgJMc44V7gwK5wVXp1Qjumcl4/1fTUSz1Y6Czmmq/a12ByrqnZkxJ2vUhVED0Wi2qboYAO4ARZJe7zMI8d8bJ7X62lLL26pGC77addrnfsrCl2Wt1BC4b8GQVo9LicuuiG5Z7SIMfupf1rfYkJFsQLJRj3X3suYOtR3DpkRERERwFkr0xaDT4dWrxqq2JRt9f42SMygYp4gqdwaF80JLdG0r6JyGod0zvfZvttrR5CocqVVPwlOD2YaKBvdUn0aNopNygMJPjYlAlLgCaBsOV/ndr3dOqrz85c5T8rJnDYvpg7ugX5f0do2JOibpc6umySpva60Vb4PZhjQT731T+zFAQURERATtegQSvU5AqkfGRIqfdnqeXTwoOnQebUYdougVNEpT/F5bLHY8820RAGcr2dbMe34NCp9YLq83mr2neEh1HvwVwQzW/sfn+3zso60nADgDEoPzMuTtjyg6gex4eC7euHZcyMZDHUuj2TvQZm0lQGF3iJq1eoiCxbOIiIiICM7Wev543gE3+dlfmlLQWl0BCi/p3Zd+H6LoHTzSKabjNFvdAYYfDlW2+vqe00CawphBoQyQ+TtXpXG/df14fHLzZK9xAEBWqpHTkMinayYVeG3zbNHrye4Q/dbxIQoUAxREREREaD2F2ehZUNDPDADpGjiQaQIUPu7fg/P/n24/iWNV6lojyoCFMkBRFUAGhae/Lj/otc1sD02AYvej81Tr0wZ1kZc9u5MAzk4yypT7ucPz8P09M1gfgFp1/qgeXtv8ZVBI55+OEQoKAQYoiIiIiND6HULPNpBanT0kuWkmAEBhQaf2D4zaTBAECIL/QJHymqrJYkdhn7b/zrYerfbaZrZKRTKd588dswcCUNeKCIRO5/xZbp7eHwDwzg3jcdecQQCcGSJWu0N1DqckqesBmAx69MpJRX52SlDHpcRk8Ag2+A1QiO6uH0TtxUomRERElPDqW6xYtrfM7z6edwcNfuZbd8tKxrd3TkVBbprPfSgydIIAjQQD2Z1zBuHhz/cAcNag6N8lHVtcgYYWq73VqT+tqWx0FtHs5OruccfsQbhjtjOw8NqaQ5jcv3PAr3XkqUWqdemMFAHMeW61XEgTALqkm9o+aEp484Z3w5Jd7iKrZj8BXKnbBzMoKBSYQUFEREQJb+Sj32pu75ap7nzw1vXj0LOT8w60vwwKABiUl9HutH5qPwHuO7xarplUgC9unQLAOcVD2T72ix2lQR1rZH6W17Zmi3PaSEay933BG6f2xwiN5wRKOYVFGZzITDbI597SO87Fs5eOavMxKDE9dN4w1fpraw7LrW49OVyxCxYHplBgBgURERElvCS9zqsGxdI7zkX3zBTUNrtb7c0Y3BVvXz8ezy8rwqie2ZEeJrVBaxkUgLsjS7PVrkpl33CoEpcV9tJ8jlanA60ghHTnOZRdPCRSEdZmjwvHu+cPkZeHdMvEkG7eLVWJ/EkzqTOHPv+xFN/sOY39jy/w2lfKoPAs00PUFjyNiIiIKOGdM1CdZt85PQlDumUiK9WI3rnqWgEDuqbjpSvHMDsiTggCIPqraArIWRPf7inDuxuPyds/2X7S53OaLN53k7U6f0gBCs8aJqH0xY/qcWalGMN2LEoMWm2UW6za0zykDCVmUFAo8F9WIiIiSnjpJvWd78xkXuB1FDpBgCgC5fXOWhCzh+Z57SNdjAUzpSPQDi1mmx0mgy4sLWell3zIVUNDwgAFtZdBI6A2qV+u5r5yFw8GKCgEGKAgIiKihCfVdvvnNYUAgAxe4HUYOgGoqDdj3JPLAQD9u3oXLlXWnQAQUKeLQBvI1rfYVO0+Q0mA9gVhpsZUE6L20vsogml3iH4fJwoGAxRERESU8Mw2BwblpcvTNjLCdEFJkScIAkprm+V1rbu8yQZ1gEK64OqcnuTzdX0V3iy4d4lqvaLe7Pd12sPXDesMZgBRCEgtcSW+AhBSjRfGJygUGKAgIiKihNditcNk0CMn1XkhOaZPpyiPiEJFEICqRou8rte4qvdsj2hziJg5pCu6Z/nOpPBXeFNZWLWq0YLUpPAEvJTDHpyXIS9npjDARu13x+xBWPn76fL66qJyVRFZABj+8FIsfOF7AGwzSqHBAAUREREltD2ltSipbEKyUYeRPbPw2S1T8NtZA1t/IsWF+hYbisoagnqO3eGAyaCD2abdVhFw16CYPbSrxjGtrtcRseVoNX48XhPU8QOlnOLRLSsZBa6CrqyhQqFi9Gin7HkuN1rscn0XreAfUbAYXiUiIqKEtuiFtQDcdQfO7sX2oR2ZZ0tOydg+nbD1aDUAZwbF17tPAwBqm6zISvW+4JdmeJgM3t0OTte2IC8z2eexQsXqcN/NLuzTCU9dPBKbjlQhWaMDA1FbeHaf8ZckUXwmuEAgkRZmUBAREVHCemllsbysrFNAHceEvjmq9UazTXO/glx38Uyb3T1/o6y+RXP/f60vAQDNdrM/fXU9bntvO2qbrF6PhVKzq9XpiPxM3DxjAHpkp+Ci0flhPSYlFs/z26gIWByvalI9xnOPQoEZFERERJSwnv7mgLx8uLwxiiOhcDF4pKg3+AhQpCS5L7wsinn2Ro12i6dqm/HP748A8N0xY+me0zh4pj7o8QajxZWhcd5ZPdhBgcLCMxvHrii+Yrap61GMyM+KyJioY2MGBRERERF1WHqd+uuurwyKLEVrWbtDxFhXoVSbR1FAQN0JZHTvTnjiohGar3kozEGvFqtzbMkaWRxEoZBs1OPxi0ZgzrA8AP6LwxKFAj/NiIiIKGEp51e/c8P4KI6EwsUjgcJnBkXPTqmq9Run9gOgzqaQKGsBbjxShasm9sFdcwb5HMPb148LcLTBkTIoWHOCwunqiX1wzaQ+ANTtdZVFZK+fUhDpYVEHxQAFERERJSzlxef4ghw/e1K88sygGNItU3O/hSO7q9al4JXVrnHLWLHpULmzMODUQV28dps5pCuyUoyYPti700cotLhS7BmgoHCTsoYcihQKi+v8e+u6cXjk/OFRGRd1PKxBQURERATAxDT5DklZQuK9X01AYR/tQJRnDQejHKDwzqBQprnfv3AoAMCgUQOivsWKgV3Tgx1ywGqaLACAdBO/0lN4yQEKxbkvBSi0CsUStRU/zYiIiCghna5Vd2fQschgh2RwZVCYDDpM7t/Z537pJgP+8tOzYLE5MKFvDqoanRf/FptWgMJ9ldYjKxmAdjHNzSXVmpkVoXLH7IEorzdjUv/csB2DCHC3F1VP8XD+bTC4S6HEAAURERElpMmLV0R7CBQBUmaEZ8cBLZcV9pKXtx2rBqBdg0J5kWYyOKdX5KYnab7mmqLywAcbpLF9crD0jqlhe30iifR3pDz3mUFB4cCziYiIiBJSp1TnBWVakh5f//bcKI+GwqWt7TelGhTbj1Z7PeZQxCxMRud+ndNNeOGK0V77nndWd69tRPFGcE3xULYZlYJ3DFBQKPFsIiIiooRU6Urhf+GK0RjaXbtwIsW/VQfOtOl50kXXC98Vo/hMveqxtcUV7v0UUzvys1O8Xqdfl/DVoCCKFCnQp0igwM3vbgOgXX+FqK0YoCAiIqKENmtoXrSHQGGk0YMjIMqaEhUNFtVjfXLdLUmVtUuUbRcle0tr2zgCotghnebKDAqicGCAgoiIiBJOi9V5ITmkW0aUR0LhJk3lCZZR7w48ODwuyg67Wot60iqo+ZvpA9p0fKJY4u7i4f5bmDHYWQB2QFd+jlLoMEBBRERECaXZYseQh5YCAKYM8N3VgToGoY3Z58qpG3ZRHaB46PM9ms9psXoHKHp28p72QRRvpABFeYNZ3pZs1Ie1jS4lJgYoiIiIKKGcrGmSl7NTjFEcCUVCW2fHK6d42OyBpbUPzPO+WMtIZtM8in8nqp2fmw98uhsAcLCsHl/vPg2DRntdovbgGUVEREQJ5ZVVh+XlzhmmKI6EIkHqPvDHC4YH9TxlZ4IWqx3n/uU7rFMUx9TSv0u6XJ/iy9vOwf7H5yM1iQEKin86RSpS8ZkGXPLKDwC0664QtQcDFERERJRQPt52Ql428u5fhycV9xtXkBPU85TnxvJ9Z3C8qhlXvr4RAHDF+F4AgH2Pzfd6nlTzwqAXkGzUt2XIRDFnuqveBABUN1lQ12IDABwub4zWkKiDYkiXiIiIEkpqkh5NFjtmD83DhWf3iPZwKMwE1ySPYGtRKItkltW1qB7rnuWsK6HMspC8dOUYfLD5OAbnsXAgdRzKqRxPfLk3iiOhjo63DYiIiChhHCyrR5PFjgcWDsXr1xYygyIBSIEJhxhce0RBEdFY6zG1Q2q1qNMIeuRnp+CuOYNUzyfqSHacqEXfzmnRHgZ1UPxXmYiIiBLGnOfXANC+800dkzR3Psj4hE8fbj4OURQhCGAQghJK8ZML5GWzlbUnKDz4rzMRERElnIvOzo/2EChCpBhCWwIUG+6bpXoNALjn452wi6KqaCBRIlBO8yitdU57umlav2gNhzooBiiIiIgo4WSlsr1oopADFAg+QtEtKxl5mSakehS7bLE6YGIWDiWgJy4aIS//bs4g3LdgaBRHQx0RP1mJiIgoYXRON+GK8b2jPQyKoNlD8wAAXdrYUtao16HRok5nL6trQed0tqilxJOXmSwvpySxSw2FHrt4EBERUcKwORxI0jM1P5HcPnMgrpzQp80BCqkgplKj2YZsZuFQAspMdl8+pibxUpJCLybPKkEQSgDUA7ADsImiWBjdEREREVFHYLU52Lkjweh0QpuDEwBw2qPFKMApHpS4lH9LaSZmUFDoxWSAwmWGKIoVre9GRERE5J/N7sDx6mZY7aKq0BtRa/SCAJtHhU2zzc67x5SQemSnyMvpJv4NUOjxrCIiIqIO7+6PduLT7ScBAGKo+k1SQjDoBdg8pnmUVDZhWPfMKI2IKHqUmUMM9lI4xOpZJQL4VhCErYIg3BjtwRAREVF8k4ITAFDXYo3iSCjeGHTeX5erGi1YW8xEX0o8gqK97pBuGVEcCXVUsZpBMUUUxVJBELoCWCYIwn5RFNcod3AFLm4EgN69WY2biIiIAnP1xIJoD4HiCEuqEmljPR8Kh5g8q0RRLHX9/wyATwGM19jnNVEUC0VRLOzSpUukh0hERERxovhMvWo9Nz0pSiMhIuo4ctL4WUqhF3MZFIIgpAHQiaJY71qeC+CxKA+LiIiI4szv/28HvthRitQkdaX5vMzkKI2I4pIihWJEfiZ2n6wDADx6/rAoDYiIqOOKuQAFgDwAn7rmNxkAvCeK4tLoDomIiIjiicXmwEdbT8jLADB9cBf88YLh0RwWxSHlFI85Q7vJAYr0ZGN0BkRE1IHFXIBCFMXDAEZFexxEREQUvwY9+LXXtsUXn4VuWcyeoOBIRQHnDsvD+aO64/nlRQCAZGNMzpQmIopr/GQlIiKiDu+cAZ0ZnKA2kZoW3DZzIPrkpsnbU4x6H88gIqK2irkMCiIiIqJQY0tIaitpiodBL0Cvc0/4SGaAghLUN3dM9artQxQqDFAQERFRhyKKote2D26cGIWRUEdQ3WQFABj16oajDFBQohrcLSPaQ6AOjAEKIiIiimlldS0oKqvHuQMDayv+1+UHAQAZyQZ8dfu5+OFQBSb0yw3nECkBmAzqgASneBARhR5rUBAREVFMu/fjnbj6jU34dPsJv/s1WWy447/b8bcVzgDF0z89C71yUnH5uN6RGCZ1UFkpzm4dPbJTAAB9OzvrUCQZ+DWaiCjUmEFBREREMW3lgXIAwJ0f7MDCkd297mRLhj38jWp96qDAMi6I/Pnk5snYeLhKrj8hTSFS1qMgIqLQYOiXiIiIIkIURby0shgnqpuCeo5SRYNFc7/yerNqfWK/HKQm8T4MtV//Lun4+QR3Fs6k/s7pQhnJPL+IiEKNn6xEREQUdrXNVlz9xkbsPFGLz7afxLK7prX6nNve345le0+rtpXXm5HvSrVXmvvnW6taAAAgAElEQVT8atX6hsNV7RswkQ9/vGAEfnFOP3RON0V7KEREHQ4zKIiIiCjsPtp6AjtP1AIADp5pCOg5/9tRiharAwAwf3g3AECFR6aEZFSvbADArCFd2ztUIr+SDDoM6Joe7WEQEXVIDFAQERGR7FhlE1787qBmq872qG50T83ITjUG/fyTNc0AgIoG7QBFj+wU5KYl4dWrxwIArhjfqw2jJCIiomhigIKIiIhkt72/Dc98W4TjVc344VAFbHZHSF63S4Y7Hb6myYodx2v87u953KcvPQsAcO8nu9BitaP4TD0azTb5cYvNgSSDDka9Dvsfn48nLxoZknETERFR5LAGBREREcmkIpRTn14JAPjtrIG4c86gdr+uxaYOOGwuqZKnZWipabaq1gd0cafUD3loKQBg+uAuePv68QCcU0gkyUbtLh9EREQU2xigICIiIlmdR2Bgxwn/mQ6BarLYVeuHyv3Xoah1jeOh84ZhWPdMGPTeSZ+rXO1Haz3GTERERPGJUzyIiIhIptcLqnXPwEIwHvl8NzYergQAPL+8CABw7aQ+AID3Nx33qnNhttlRcO8SvLn2CG7691YAQL/OaXJbx8cvGuF1jElPrcDmI+zYQURE1BEwQEFEREQAgDN1LahpUmcjbDpShfqW4DMU6luseGf9UVz+2gY4HO5AxKMXDJeXKxWFMwGg0jW95LEv96LY1emjX5c0+fGrJ/bxOs6p2hb839bjAICtD84OepxEREQUOxigICIiIgDAhKdWaG4f+ei3Qb/W/21x14SY9sxKeVkQ3BkahU8sx4Q/LYfVVRDTMzgCAH1y07y2efpmTxkAICctKehxEhERUexggIKIiIhwsqYZ0oyLcwZ01tzH4RDRZLFpPiax2Bz4etcpPPblXnnb8Spni9BnLx0FAEhRFLEsqzPj5ZWHAAALX/he9Vo7Hpnr9fpf3nYOPrl5Mt771QTkegQklMEPIiIiij8MUBARUdzadaIWdW2YfkDeqhrc0y3euK4QJYsX4ZHzh8nbRFFEv/u/wrCHv4HZ5rsuxaAHv8Zv3t2m+djAPGcnjh/unana/vzyIjmLQpKfnYKsFKPXa4zIz8KY3p0wuX9nbH1ojrzdV1CFiIiI4gcDFEREFJdEUcT5L67FxS//EO2hdAhWhzNAcOnYnjAZnBkO10/piySD86vCCyuK5X3P1Jk1X6NWY4qG0qC8DABAJ42pGDuOq7uFvHnduABH7vSfX04Ian8iIiKKPQxQEBFRXJK6S0jFFKl9bHbn/I4Lz85XbR/VMwsA8N2BM/K2BrP2NI+LXl6nWn/m0lHY//h8jCvoBABIVkztuLywl2rfL3aUAgCGdHMGMXrnpAY07q0PzsbmB1gck4iIqCMwRHsAREREwThd24Kvd5/ymhJA7WNzvZ8Gjzaj98wfgktfXa/KcLA71O1BJUcqGlXrGckGJBv1eO9XE2G2qX9fj14wHIUFnXD3RzsBAP9afxQA8MnNkyFAQEqSHoHITTcFtB8RERHFPgYoiIgorpz397WoaFBPMRBFkQUS28nqCjoYPQIUg7pmeO1r8xGg8HS6tsX1mjoY9eqkzZQkPS4t7IUV+85g6Z7T8vbUJH41ISIiSlSc4kFERDHNYnOoaht4BicAoNR1IUxtJ2dQ6LwDCZ7sDnU2hMXmwMxnVgEALivsiT9fMtK13MvzqV7+/vPR8vKsIV2DGjMRERF1LLxNQUREMW3GM6twsqYZR55a6DNLosWq3VXC4RBxuKIBAzSyAEjN6qpB4TnFQyqSqbWv5O0fjuCwa3pH/y7puHxcb1w+rndAxzXqdZg1pCumD+6CqycVtGHkRERE1FEwQEFERDFr6e7TOFnTDAD4385TGNM7GyaDzqueQX2Ld9FGh8PZFlPy50tG4sKz81WFGsmtttnZZjTD5N3a88AT8zH5qe/QOzcV24/VeNWgqGx0tyid0C836GO/EWTHDiIiIuqYGKAgIqKYdd8nO+Xl29/f7vV4RrIB9S02NFu8Myiqmiyq9T98vAt/+HgXXr1qDOaP6B76wcYpURTR9z53ICe/U4rXPiaDHlsfmoOtR6txySs/eNWgyEpxBjWMekHu+kFEREQULNagICKimHKqthnz/7oGx6uaUK2oPaE0uX8unr98FN65YTwAoMXmDlA0mG0QRRHf7inTfK7UzjIY6w9VwmzTnkYS75o8gjt6ne9iowbXY98oiloCzgwWnQDsf3wBi5USERFRmzFAQUREMeXjrSew/3Q9/rPhqLxt/+PzVftcNDofPxndE0ZXQUeH647+8aomjHjkG3yw+Tj+tb5E8/W7ZabAYgu8RenJmmZc8c8NGPzg0uB+kDhRpZieUdink999KxudBUrf23gMQx76GptLqlDbZMUrqw4hJ83kN7hBRERE1BoGKIiIKGaUVDSi3uysJ/GPNYeRpNfhl+f0RbJRj8N/Woi5w/LQKdWIS8f2BABIDSekKQdriysAAEt2nYLF5kDfzmmYOywP0wd3QfGTCwAAb647gmve3BjwmLYfq5aXrXb/gQ1RDKz9Zix5Y+0Refln4/0Xtjx3YBd5ucXqwKWvrseox74FoN1dhYiIiCgYrEFBREQxY7qrVaXEYndgmytAoNMJeO2aQtXj0h17KYPivk92AQC+P1gh77Py99O9jrPhcFXAY/ps+0l5ubrRgkaLHf/dfAzDe2Rh0cju8hh++c4WrC46g4NPLgz4tWPBjhM1AIC1f5iBnp1S/e5r1OswqV8u1h+ujMTQiIiIKMEwg4KIiGKCZ2cIyb5T9T6fI9VE8CzaKBnTO9vncz/YfCygcR2vapaXx/9pBf7+3UH8Y/Vh3P7+dlz+j/VycGT5vjJY7aLPnyMaXl5VjIJ7l+DPS/f73Gf7MWeAIi8zOaDXzElPCsnYiIiIiDwxQEFERDHhx+M1mtvfuK5QczsA6FwFGR2iiBardxHLaycX+HzuHz7ehSMVjX7HtKe0FgfK6jE4L0Pe9sk2d0bFlqPV6Hf/V9h61D0NpNqje0gkvfjdQfx4vAa3vrcNr39/GH9ZegAA8MqqQ3hyyV6/zzXqA/tK8MLPRmNM72wsGNFN3pZs1OEttgolIiKidmKAgoiIou5weQMueeUHeX1cgbtYY/cs77aXEml6hd0hygGODJN79qLU/lLy8W8mq9b/urzI52u/sOIgFr2wFgBw07R+fsevHPvDn+8OqghnqHy96xSe+bYIF720Dl/uPIUnluxTPf7P749go8fUDKmmxp2zBwV8HL1OwCc3T8FYRUHNnY/Mw4whXdsxeiIiIiIGKIiIKAbMfHa1av3DmybJy/nZrQco7vpwB3722gYAwHOXnw2j3rl9rEdXirF9OuGPFwyX1z//sRSnapsV6yfx3sZjOF7VhOeWuYMXfXLTVJ1EPF9X6atdpzHowa9x8cvrUBPBbIrb/7u91X3WuYqI/mt9CU7WNKO22dnGtVOa0c+ztPXJTZOXkwz8OkFERETtxyKZREQUdfnZKThZ4wwUvH39OAiCgI33z0JWitHvxa9WW8vJ/XP9Fqq8dnIBrp1cgIJ7lwAALnxxHTY9MBu/eHszVuw/AwDo29l98f2/W8/ByJ5ZAIDsVCNqmqx45PxhSDcZ0Ck1CaMfXwYAmDWkq/x8ANh2rAZnP7YMR55aCEFovf3mmfoWpBj1yEgOLlhwuLwBN7+7DVa7du2Lu+cNxtPfOKd6vPBdMbYcrcYPhyrx8Od78LefnQ3AO9MkEHOG5eGt68dhQt+coJ9LREREpIUBCiIiioj6FiuKyuoxto/3Ba0UnACA6YOdUwUCKdoowPvCP9moD2pcZ+rNcrBCItWm+O2sgXJwAgCW3zUNRafrcVZPd/HNksWL5OU31x6B0aDDQ5/tlreV1ZnRLSsZ7/xQgrN7ZWNUL+/CnUVl9Zj7/BrnsZ9aCFF0di0JxFWvb0RpbYvX9otH56O4vAE3Te0nBygA4IdD7mkev/3vjwCCf88kMwZzWgcRERGFDnMyiYgizGp3YOazq/DKqkPRHkrI7TpRiz8v3Q+HQ4Qoqu/oX//WZlzyynpUNfqe9rDm7hlBHc8uemcNaGVVaPn+Hu9jKetXAMCdc9S1GTqnmzB5QGefr3nDOX1x9cQ++J3iecVnGlDRYMYjX+zBhS+tw09eXoenvtqHb/ecxnPfHkBdi1UOTgBA3/u+wuznV2u9vJeaJosqOKGs3fHc5Wfji1vPgUGvw7u/nOB3qkyvVtqLEhEREUUCAxRERBF2oroZh8sb8eel+/Hm2iPyhfyn20+grsUa5dG1z90f7cArqw6h8MnlmLz4O9VjW1ydLn48Xu31vB5ZybhkTE/0zo3chXKvnFS8cuUYef2xC4dj1x/nheS1b5s1EOvvmwkAeH55EQqfWC4/tv1YDf6x5jBu/PdWvPBdMW55d5vX8w+X++8uItlbWqda79c5XXO/KQM6Y+7wPHm9QPE+b39oDob1yAzoeEREREThxCkeREQR9uJ3xfLyY1/uxfrDlZgzLA/3fLQT4/vm4P1fTVRlAdz/6S6IIvDUxSOjMdxWmW12FJ9pQIvVgZomZ4DFX5bEDW9vkadFzHt+DS4b1wtn6s3ommkK+tg9slqfBuKPcmrDRaPzAQArfz8dr646hEcuGNau1+6UmgQAqhakWr4/WCEv3zVnEN5YewS1zVYcr2pCrxz/AZvqJnVA67xR3fHBluOa+94yYwDys1PQPSsF4/vm4IsdpRjbpxM6pSUF8uMQERERhR0DFEREYVJeb8afl+7HsO6ZuOGcvgCA41VN+HjbCdV+y/aWYdneMgDApiNV6H//V/IFfGlNM97beAxA6wEKURTx/LIiLBjZHUO7R+aO+O6TtTjv72t9Pn7Zq+vxwU0TUdds83ps36k6HCirx+Nf7gUA5GUEH6BQFp88/KeFcGhM+fDnnIHO6RpvXTcOma7ilH07p+HPPz0r6LF48qzrMCgvHUVlDT73v3/hENw4tT8cooi/Lj+Iz388iVtnDvR7DLPNDgB449pCDOuRie5ZKdj9x3lyFxOlzukm/PJcd7vUX7jOSSIiIqJYwSkeRERhMu3plfho6wk89uVeHKtsAgD88p0t8uP9FJ0itBypaFRNk2ix2v3uX9lowQvfFWPB375vx6iD8+/1R/0+vqmkCv/ZcBR3f7RDtV0URVz/1mbVtsHd2hdU0ekEGPTB/bNm1OtQsngRZgwJb7HHBxYOxZLbz8Uf5g9Rbb9pmjtgcOPU/gCA30x3/r/J4v/3DQBmmwMAMLxHFrpnOWtMpJsMMBnaVvSSiIiIKJoYoCAiChGb3YGnvt6H8noz3tt4THWBOfXpldh5ogYHyuoBOC9Mv/v9dOT5mNbw6upDmPHMKtW2dcUVmvtK1hSVy8uetQnCJT3ZOxEvNy0JB59cIK8/9PkefOvKELlucgEAYFVROU7XOYs7/nbWQCwc2Q1j+nh3t4h3w121HQoLOsGo1+E30/vjuctGAQCundQH9y0Yih2PzFV1AjEZ9OjbOQ1HXUEtf+77ZJfrOfznnIiIiOIfp3gQEYXItmM1+Mfqw1h/qBI7T9R6PX7Bi+vk5fsWDAUAbLx/ttzi8oMbJ+KxL/diT2kdFn+9X973/349CZe+uh4PfrYbs4bmwRepNSYALHzhe3z060k4WdOMecO7tbmNpD9vrTuCN9YeAQCsv28mGs02/GP1YTx18UgY9Dp8f88MnPuXlarnSDUVpOyJFKPeq1NGsP70k5EQAmvcEXGf3TIFe0rrcLaitehPRudj4cju8u8kK8Xo9bwjFY04UtGIl/y8trLORzh+v0RERESRxgAFEVGI2BzOdHtlcOKt68fBbLXj1//x7tQg+eyWKfjL0v0Y1Ssb54/qgT2K7IcHFw3FuIIcAMCp2haMeXwZRuRn4V83jFe9RoPZBs/yCz99db28LN2hr2+x4tJX12P/6XrVdiWLzYFBD34NANj0wCx0zVAXojxU3oB7P96JzSXu4o/S9IKnLx0lb+uVk4r7FgzBU65gywWjeuDAaXVmx97H2t814+cTerf7NcLFqNepghOAs25GoAEFm93hc9rKq6vdbWpTkhigICIiovjHnFAiohCpaPDuXDFjcFfMH9Edz1/uvnBfc/cM1T5n98rGe7+aiGSjHspEgIfPGyYXNZS6elQ1WrCmqBwF9y5Bwb1LUNFgxi/e3owRj3yDF1c6u4OMzM/yGsfOEzXOxx79Vg5OAMDWo1Ve+z6/vEheHv/kChSV1asev++TXargxB4/rTmtdoe8vPiSkejVyd2V4rrJBaoil+R2wageAIA7P9zhc5/X1hwGAIzq1fGmxhAREVFiYoCCiCgE1hSV4/b3t6u2XT2xj7z8k9E98epVY/Hraf3RKyfF5+uM65sjL1c3uQMeK383XXP/wieWY8X+M6ptX9w6BffMH6zaVtdsw+vfH/Z6/vubjmP3yVrUNFnw1Nf7cLyqCa+sOqTaZ+7za+TlU7XN2HTEHdS4dGxPpJl8J+NN7JcLAPj4N5OQmmTAr6f3x+vXFOLueYNx19z2Te3oyP54wXAAwP92lOKWd31n3wDAe7+cEIkhEREREYWdIAbZki0WFRYWilu2bGl9RyKiMJHqSADA3GF5+PX0/hjTu1ObXuvjrSfw3LIifPybyeiW5Zxe4XCI6Hf/V60+NzvViB8fngur3YGBD3ztc78PbpyIy1/b4Pe1/nH1WNz0760AgE9vnozRvTthwd++x75TdZgyIBf3LxyK4T28szU8We0OGIPsrkHqc0qaiiOKIr7ceQqzh+Zh6MNLMb4gBx/+elK0hkhERETUKkEQtoqiWBjIvvzGSETUTu9udLfavGBUD7x2TWGbgxMAcMnYnlh370w5OAE4W2ge+tNCTBvURfM5j1/ouuN+6zkAnLUP9j42T3P6xU1T+2G8IlNDy5GnFmLe8G544qIRAICfvPwDfjxeg32nnDUk/n3DhICCE9JYqH3WHnR2cFldVI7b3t+OoQ8vBQAsGNktmsMiIiIiCil+ayQiaoftx6rxwKe7AQCT++fihStGh+1Yep2Ad24Yj1evGoNRPbNQsngRXr+mEN/fMwNXTypAyeJFcpcMAEhNMiDNZMCds91TKbY/NAf3LRwKQRCw9cHZ6JRqxJBuGV7HkmpDXDHeXYDyopfcXUh0OtaOCLe75w1GfrZzOtBVb2wEANQ0WVX7aHUAISIiIopXnOJBRNQO3x8sx9VvbAIA/OWSs3DZuF5RHlHwbHYHVheVY/rgrmhosQGC+sLXZndggGK6yJYHZ6NzuikaQ004DWYbRjzyDQBnwMLhEPHsMncR01evGov5I5hFQURERLErmCkebDNKRNQOTRY7AOCJi0bEZXACAAx6HWYNzQMAZKV635H3bHOZm5YUkXERkK4oQPr0Nwfk5YNPLsDXu09j7rC8aAyLiIiIKCwYoCCiDsnhELHhSCUm9csNSytLURRxwYvrsOtkLQBg6kDt2hAdxY5H5qKq0YK+ndOiPZSEs/yuaZj93GrVNqNeJ7ciJSIiIuooGKAgog6lxWrHbe9vx7K9ZQCAjGQDdj3qXSiyvfaU1snBCQDITe/YWQVZKUbWO4iSnp3UbWmfu2xUlEZCREREFF4skklEcaeorB4Oh3b9nCeW7JWDEwBQ32LDy6uK8cjnuzHikW/QZLHJjx0ub4DZZm/TGP663F0H4MoJvZFmYryXwiPZqMfOR+dieI9MTBmQi4vH9Iz2kIiIiIjCgt+oiSiubDhciZ+9tgEA8M4N4zGlfy5abA6kJenx/cEK/GfDMXnft64fh+vf2oy/LHXP3f/nmiO4fdYAHKloxMxnnWnzJYsXaR5LFEUcKm9AXmYyaput6NkpFU98uRfvbTomBySKnliAJANjvRRemclGLLn93GgPg4iIiCisGKAgorhSfKZBXr72zU2YMiAX64orVfssv2sq+nVOh04n4ImLRuDBz3bLjz2/vAjPK7IfAOBoZSP65KbhpZXFciHCl34+Bre8t021X9cME87UmwE4i2NmJBsYnCAiIiIiChEGKIgoLpyobsJVr29ESWWTartncOLueYMxoGuGvH7lhN5YdeAMTEY9Oqcl4Z31R71ee9rTq7y2eQYnAMjBCUl9i81rHyIiIiIiahsGKIgoJu0prcWiF9YCAP5yyVm45+OdqsdLFi/CV7tO4eZ3nYGEZy4dhdKaZtwyY4BqP0EQ8Pq14+T1Cf1y5ees+N00zHpW3R3B0zs3jMfUgZ1x47+3yrUthnTLwP7T9Xj4vGHt+yGJiIiIiEgmiKJ2obl4UlhYKG7ZsiXawyCiAFU3WvDq6kMYkZ+F8zVaJYqiiL73feXz+TdP74975g8B4JzyUZCbCoM+8KkWx6ua0C0rGUbXc45UNOKFFQexYl8Zvrj1HPx7w1G8sfYIXrhitKqVo8XmQJJBhyaLDR9sPo4rxvdGslEf8HGJiIiIiBKNIAhbRVEsDGhfBiiIKBKsdgcMuv9v77zDrKquPvwuZobO0DuIUgRFQREVO0XAghVsscYaNfaoMZ/dWGKJMbYUe+81Go0l1hgjsRvF3rGB0qTP+v5Y+8wcLneGGfXucwbW+zz3mdNm7m/23meXtddeW1hcpYy+4Ek+nm5LNY7cfADbrdWTVTq1AuDY21/h9v9+Wv17O63Tq/r8tVPH8dLH37FBv47VxoVSMWPuQt9W03Ecx3Ecx3F+JG6gcJwfwXPvTeNf733DxY+/y27r9eaM7dZY5uz8u1/NonWzCrq1bc6UL2bx5tSZbLpqZzq0aso3s+fz7Lvf0LyijPGDu0X6L7Jl4eIqXv9sBvtdO5n9Nl6FzVbtzISLn6nzd/p2asVKHVvyxJSvq69NPnFzOrVuxhufz6BPx1a09q08HcdxHMdxHKdR4QYKx2kgqsqEi5/hjc9nFr3/wOEbM7hHW777fgFNy5tQUdakegb/xHteW2JryzT9Orfiva/nVJ/3bNeCK/cZzqBulT/9P5ETJn84nUl/em6Zzz14+CZ0aNWUEWc/tsT1bpXNGTmwM7us25u1V2pfKpmO4ziO4ziO40TADRRObvj8u7m0a1lBi4oyFlVpvdzyVZVjbn+Fe1/+nF+O6s+Rmw9AREqmcXGVct1zH3La/f+rvta0vAlHbj6Acx+aUn1tnw1X5pp/fVh9PrBrG5qWN+G1z2Ys9TdH9O3Av9+fXn3ev0vrJbbHvGKv4bRvVcE6fTqwuEp54cPprNOnfcmXLfxQps2ez6x5i1i5UytUlfte+Zwrn/mATQZ0Ykivdjz+5lf8439fsPO6vXnwtal8Mn3uUn9jULc27D6iD8+99w2HjxlQbaRZtLiKbS95lv9NncnoQV347fZr0KNdi9j/ouM4juM4juM4JcANFE7mLFxcxe5XPM9/Ppi+1L2dh/fiuC0G8cWMefRq34L5i6pY/6zHWKVTK3YJA9xXP11y0P/0caNo27KCb+csYPb8RdW7O4we1IXH3/qq+rlB3dpw8W5r079L62UaNVSVy554j/MerjFC/GKzfowc2JkRfTsCFodgw7MfY86CxbX+nTbNynnsmM3oUtkcgHkLF9O8oox5Cxdz2+RPGNi1DeuHv/fLm17kb69Orf7dDft1ZHGV8nxIp3VXbs+luw+jS5vmtWoGeOPzmXRs3ZSvZ81nSK92df6fxf7GzLmLuPmFj9m4fyde+HB6tXGmTbNyZs1fxGrdK5k4rCfvfjWbW174pPp3+3ZqxfvfzKntT1fzm60GceCm/QCYv2gxH037npU7tqJpeXEDTFWVMmveItq29JgPjuM4juM4jrM84QYKp2SoKu99PZt5C6v4ZvZ8npjyNRv268jaK7XnmXe/ZnifDrzy6Xf88qaXfvR3nbj1avz2gTd/0O+WNRFOnrA6e23Qh7kLFzPli1lLLRe4/rkPOeneN6rPDx8zgKPHrrrU35o5byFb/uFphq/cnnN2HEKLpmUsWlzFh9Pm8J8PvmXUoM50b1u/Gf93v5rNLn9+jh7tWrBSh5Y88NrUWp9NjC/bDu3BoO5tePzNr5j80bdLPdehVVN6tGvOpgM607yijOlzFtC7Q0t6t2/BiH4dqWxeM+h/YspX7HP1C7V+55o92zJ1xjy+mT1/qXud2zTj61l2vVf7Fpw7cQgKvPDhdDbu34mhvdvx5tSZLFhUxTp92pfU68VxHMdxHMdxnMaBGyicn5TFVco3s+fz5cx5bHvJsw363XfP3JLysiZUVSmffjuX6//9IbPnL+Lm/3yyxHMHbdqXPz/1PmAxCB4+alPatqjgvx9NZ+LlS8czOHfSECYO68W9L39Gm+YVbLpqJ2bPW8Q6v320Tj1TfrsFn0yfy6uffsdp9/+PGXMX8rfDNmaNnm0b9H/9VDz33jSmzpjL+MHd+G7uQm5+/mMu+ee7tT7frmUF8xYuZtMBnfl8hi2j6Ne5NW98PnOJJSQJZU2E4X3a07NdC2bNX8Qj//uy+t7gHpVs3L8Tw/q0Z6P+nZYIQHnRo+9w4aNvc/DIfhy1+apLeD5UVSlNmrjxwXEcx3Ecx3GcZeMGipwzf9Fi7nnpM55/fzpzFixiqzW7s82QHpkP+hZXKdPnLEBVefC1qZzz0Ft0rWzOwkVVfD5jXvVzQ3q15a2psxjSqy2/mzSE8Rc+RdsWFbRrWUHH1s1YXKVcvkftyxQKUVVEhHkLbRlF84qypZ5ZsKiKt7+cRZfKZnw07XvWXblD0b81b+FimpU34Z6XP+OoW19Z5ndnaZyojXkLF/O3V6fSulkZowZ14auZ87nrxc+YNLwXXdo0KxqnQlV5YsrXzJq/iME9KpnyxSzmL1rMc+9N49l3p/HFzHksrlLaNC/nj7utzaiBXTL4zxzHcRzHcRzHWdFwA0XOeXPqTLa86Omlrvds14K1erfjqLED6N+lTYP/7uIq5YNvZnPhI+/QpcqhsYsAACAASURBVLIZq3evpGl5E2bPX8TDb3zJRv060rJZOS98MJ32LSt4+8vZTJsznzGrdaV/59Zc/uR7S83Cd2zVlDV6tqV9ywoeeuMLzps0lG2G9vjB/3tsnpjyFSfd+zr3Hroxm533T2bNW1R97+p91mXUoBVnoJ4Yb3zpheM4juM4juM4sXADRSPgjc9n0LWyOX956n2aiHDni59Wr+9PuPPgDVinj3kKqCqq8OG0Obz95WzW6FlJ2xYVXPH0BzQtb8LDb3zBB1/PYdb8RcW+rlYGdWvDlC9noWoBEgf3rKRl03LW7NmWrYd0Z9WuDTeUOI7jOI7jOI7jOA4sBwYKEdkCuAgoA65Q1XPqer4xGiiKsWhxFTf952NOTgVuFIFVu7RhXtgJoTbaNC9n/OBuVDavYHCPStq3ssCI3y9YTGXzCgZ0bc3seYt49t1v6FLZnIWLqxi2Unt6d2jJR9Pm8OqnM9hkQCfatWxa8v/TcRzHcRzHcRzHWTFo1AYKESkD3gbGAp8CLwC7qer/avud5cVAkfD6ZzN48LWp3PLCJ0yfs4C+nVrx3dyFrNa9DWv3bk/fzq2Ys2Ax7301m35dWrPtkB5Utih3133HcRzHcRzHcRwnVzTEQFG+7Eeisx7wrqq+DyAitwDbAbUaKJY31ujZljV6tuW4LQZlLcVxHMdxHMdxHMdxorD0dgDZ0xNI70H5abi2BCJyoIhMFpHJX3/9dTRxjuM4juM4juM4juP89OTRQFFsncJS61BU9S+qOlxVh3fu3DmCLMdxHMdxHMdxHMdxSkUeDRSfAr1T572AzzPS4jiO4ziO4ziO4zhOBPJooHgBGCAiq4hIU2BX4L6MNTmO4ziO4ziO4ziOU0JyFyRTVReJyC+Bh7FtRq9S1TeW8WuO4ziO4ziO4ziO4zRicmegAFDVB4EHs9bhOI7jOI7jOI7jOE4c8rjEw3Ecx3Ecx3Ecx3GcFQw3UDiO4ziO4ziO4ziOkzluoHAcx3Ecx3Ecx3EcJ3PcQOE4juM4juM4juM4Tua4gcJxHMdxHMdxHMdxnMxxA4XjOI7jOI7jOI7jOJnjBgrHcRzHcRzHcRzHcTLHDRSO4ziO4ziO4ziO42SOGygcx3Ecx3Ecx3Ecx8kcN1A4juM4juM4juM4jpM5bqBwHMdxHMdxHMdxHCdz3EDhOI7jOI7jOI7jOE7muIHCcRzHcRzHcRzHcZzMcQOF4ziO4ziO4ziO4ziZ4wYKx3Ecx3Ecx3Ecx3EyR1Q1aw0/GhH5Gvgoax0/gE7AN1mLKCCPmsB1NRTXVX/yqAlcV0NxXQ0jj7ryqAlcV0NxXfUnj5rAdTUU11V/8qgJXFcM+qhq5/o8uFwYKBorIjJZVYdnrSNNHjWB62oorqv+5FETuK6G4roaRh515VETuK6G4rrqTx41getqKK6r/uRRE7iuvOFLPBzHcRzHcRzHcRzHyRw3UDiO4ziO4ziO4ziOkzluoMiWv2QtoAh51ASuq6G4rvqTR03guhqK62oYedSVR03guhqK66o/edQErquhuK76k0dN4LpyhcegcBzHcRzHcRzHcRwnc9yDwnEcx3Ecx3Ecx3GczHEDheM4juM4juM0IkREstbQmEinVx7TLo+a8oyn1/KNGygcx3GWY7wRb5x4vjlOfsjp+9gifZIXjSKyoYjskLWOIlQmB6qqeUgvEekqIp3ypCnPiEgHEWkG+UovEemdtYZlkZe0qi9uoHCWe0TEy3kjR0RWylpDMUSkX9JY5gUR2UBEdhSR0WCNeNaaGhM5asTbZy1gWeQorXKNt0E/jDyUrzDY7pWnwRCAiIwDbhGR00RkN8hHXS8i2wFXALMLrmeadiIyHrhHRC4Ukd9A9uklIlsCDwGXiMifEk1ZpxXks84Kefgg8EcR+TVkn4cAIrIa8JGI7J21lkJCH3U45Kds1ZfcFUCnbkRksIisKiIDs9aSRkQ2E5EjRWTnrLWkEZGRwCkismfWWtKIyKYicpCIbCMizVPXs27E1xORrURkVJY60ojINsCjIjJQRCqy1pMgIlsA7wB7Za0lIXRarwWGAneLyMYZS6pGRFpnraEYocxvLiKbQG46PFsBfxOR7llrSSMim4jIwSKyLeQjrQBEZLiIrCMiw7LWkkZEeolImapW5anDn8d6HvJXvkRkDPAM8HCejBShnj8fuBFYDAwruJ+JRhFpC+wL7K+qj4hIcxFpCtnmpYiMxdLrIiw/+xTcj/5uisj6wAXAscCJdsl0hHKWWX0RJjd+JiK5MZKHvswfgLMwI0XHgvtZ1q9lwGfA/4nIQRnqWAIR2Rq4HzhPRP4NjctIkZsG01k2wdp6C1ahXRMq3cwJA7U/AVXAFSKSiwFbSJ8rgG+BP4jIpIwlARA6hXcDbYCTsEptHGTeiG8JXA9sCDyYB2uwiKyFNUgHq+oUVV1YcD+rjtiWwP8B5wI7isjKWehIIyKdgTOAI1T1FOC3QBMR6Z+tMhCR7YGrg3dHbhrHUHddi5X5h8KAJFNCR+wi4HRVnZq1noRQR10FtMNmIsdnLAmo1nUzsDnWLv5SRMozlpXMJD8AHC8i5XkxUuSxnof8lS8RaQmsD2wH3AHclTJSZJKPYnQE9geOUtVbgSeA9URk91DPZtmPmBd+ThORXsA9wHUicp+ItIC4bXZIr7bABOBwVb0H+BjYSEQOFZEjAMK7GbtdagncoaqPYgPcLYGzReTqDDUhIhsBjwJ7A+NyZKToDlyhqvcBnwJbicivReR0yC69wne/jhnADgR+JSKTRGQVEalcxq+WDBEZjPUB91DVUcBXItI36M3FxMKyyLwRd+qHmIvOhVil8UL4ubWIPIptF1uVka6+mBX4MFV9VERmAN1FZB3g1cIBZSRNglX+hwAnqOrtIvIt0CvMkr6oqnNi60qxFnCiql4uIncBOwNjRaQqNFbREZFB2GD7QFV9UkT+AVwgIg8C32RYoVUB96jqY8EIsCfwJfCRqj6cha4wS3secIiqPiUi12MzMh+G2dLFkfVISIdZwGRgfmicTsAGSJuLyKmq+ueYulL6VsXqiE+ArbE8fT4LLWlEZAhWpx6sqk+IyHygIgxCPs1QWhfgUlV9SER6AusAC4FnVHVWbDGhPm2LladjVfWeUM+3FZG1VPXl2JpSuioxg/0RqvqgiDyNDdhaisiFWbQ/QVt34NSgpQ1wjIhcoKqLRKRJhu117ur5vJYvVf1eRK4DvlPV+8U8wO4SkZ1U9aOMNCk2+D9MVb8UkS7A7zGvgK7AMBHpq6q/z0jffBGZDKwNrAvchxnDrgLuAraMWcbCd80QkdNUdbqIdMCM+H/H2qP9RGQVVT0yhq5UWw3m+TJRRBZjBqe/YBOQfxKRO1V1Yuz3MRje2mN9UjDjXBMReUhVvy3yP8RkIeYNvRD4DfBX4FXgYBH5Q6w8LIaIlAGbYmOzUVj/pjv2HryShSasP/gS8HUwam4EnBkMTseo6hsZ5mW9yNya79RNyiLYFThbVZ8PnZv3gCFkaJwIfAzsGIwTvYBLgf6YK9YJoQGNihpzsJczcd2+GBgcfh4rIl1j60oxHXOf66qq7wPXhWubxLYAp75vEXBy6LSWAy8DM7DylWUF1h6zlK8JXAY0w/JxBxGZmJGmcmAXVX0qnL+HudCVxzZOBFoBqOo84CvMeHkHcJGq7glsj83MbBBTVKpsLcDcfnfADIc7i8j6oVFPGvcsdCk2u/CEiPTA3Gx3xZZWHBBTUwFlWL01APO0GoV5ER0duz5NOjCq+h3W+VopGJ/Pxmb87hORY2JqKtA1A3gTmBuu/Qu4HetgR1/Wl5St4PmyPzYYehbohRkpmmY504cNik7MSz2fx/IltlRot2DM+UxVZwOo6jHA08DtItJEbHnmHhF1DRWRHULfJTFUzgV+o6pHYl5Xj2HGnmiIyJoismUwpgK8CBwPrAL8XVVnqepOwMIwUIqpa6uga0a4PBc4VVWPDTPx5wEx38Xq/z/0H47A6ofHVfUMVX0H2BGrz1rU8jd+clL1VhXwFPCAqt6BxcfYEuuDdQjPRKsnQpnfPvSV7wF2AqYFfSep6oOYITjqZLuIjBCRA4K+HqHfdx3WF1uA9S0+BdaIqauABVh9fxZWxs5R1d2A/2ETM/n3pFBV/+T4A7RJHbcDmiTXsco/udc1sq7OQPPUeRk2cNwrnPfFrIiTMky7McDvsFnkc8K1/lgFvHNkLWWp40rgHOBIoH241g94DRgbWVdXoDwcl2Md1eTe34CVw/GADMvWqcC9wFnhvD22vOLorMpWkTy9IVX2JaKGLbHZoLOAvVPXfwWMSp1fDGwWOX3apo6TMlaJeVP8HhgRrnWPrKtlwXlTYA9gv3C+PjATWD+yrqRu74jNcp8HHB+u9QeejF2fJvVTON4NW8r3HHBGuLYmZqDbIqs8DHXBNcBhoZz/EVu+8CzQI/L72LbItXLMc+iSVH4OAJrFTLOUnk4F55nU83ksX8AWWAf+CuC/wJBwvSL1zInYstGPgEGRdG2LTQbdjHklnAKsXuS5o4CrKWjLS6hra+At4DbMW2izcH33cH338A7uhBmglno/Iuh6nFraPszL9i5s4qOk6RXy8DNg34LrnbA4AauF870xj5jWMdIqfGf3gvN0P3BPbPA9FjgOmyiNoSld5h8IZb4f1v97MqmrgF8AjwAtIpatD4HLseWhV2Ne0WsAnwNfAJsBq2LGuvYxdAVtGwFHY0ua2mF96QEh/1ZNPfeP2HX9D/m4B0WOEVsrepuInCkiB6vqd2qzL2VYA9QrWPL3wWI/tIwxMyO2NvQ+4FIRuRJAVRer6huqel04fx+rNFqWWk9K1+hgya8IGh5T1eOxzutCEWmpqu9iBopoHhRisTAOTLw2VHUm8G+gN7BPsA6/hwX+6RRR1zbYjMtlInIrUKWqKiLloRx1BsrDLNHtItIugqZ02bo6XL4e+AbYRUT6qLkazgMGikhZpDKflK1qS70Gb4ngFvkqsEG4HsUqLSJDgT9jniWvA6NE5PZwexpwgIgMEJH9MGNdNLdksfXkt4nIOSKyn9a4ts/E1kUqtr7198BjItImUj6Ow9ZEnyMhcK6qLgDuVtUrw4zu81jHI4pnmliA4dU1eMKp6jTgfazTM1REOoR66wni1g9bA38XkT+IyG9V9WZV/QXWIfsoeAy9hnXwowVATeXhuSKyg6qeidWnzbEZ0uPUPCneAGZFfB/HAXeIyMUiclJyXVUXYfXsQ0Cl2LKKf5La8rDEugrbxW/C9aTujF7Ph+/PVfkSkfUwD9ADVHV/bIJlqFgsinRf+XWs/tpSVd+KoEuA0cBBajOgF2Izo4eILZ8j9AP3xwI2n6uqi0pd7kVkbWzd/d6qujPWfzkLQFVvxNzw18G8iA7DBuYzavlzpdT1ENbmpJ+pCO3iAZhH0fxSppeI9MMmWq4CDhKRfZN74X18AnhGRC7HJhcO0uC5U2rEYuV8JiK/WvJytVfF9Vgf7HwsH2+LoKmwzJ+PtcdHYZOhtwEviMj5wOHYEr+5pdYVWBcrMwdj5f1FrHxVYZ5f+6nqk6r6NrBR6K+WHBGZgPUHe2PeqocDs9W8ct4G1haRTmLb/3bFjKz5JmsLiX+Kf7Ao/B8D2wA/A67EAuok91thAYiOxzpngyPpGg1MAcYDq2Ed+QOLPLcX5hHQP5KujbEK4mXM0pue8VgPm706Entp3yaS9RAYEXQ9gVnGu6fubY95UkzGGq9vIupaGVsCsxnmjXMTZpXunHrmCqzCewpYM4OydR1htgEYBJwZ0upE4F3izV4Vlq3yIs+0xWZH9oyhKZUml4TjCmzAcwNwdbh2TXg/n6XIbFsJdY3DDDbbY9b8i2p57t+Y0WRoJF1bhPI1ETgGm21ftchzu4W8XimCplGhbH1cWIdjy2Iuxjq1x2CzNrHqh0EhrcZhOwTcBDwX7m0WytbWwMHhuX4Z5eHlQM8iz+0D/Ctdn5VY19igaxts5u8vBG84gmdMOP4r8AFhZj6CrlrrLmo8dq6MWc/ntXxh7WDi1dULM/LegS0ZOhrzbGqLeeysESOdUtquBi5MnQ/B2sETMA+wtYH/FNYhJdbUF9it4NrfSHnoYP3Uyljv4TJ0pb11BmKz8rH6zWXAduF4THgfCz0pNsLag1UiplUvzLB0PLYk4djUvSYETwpswDuLuP2IYmX+ZCyuHFj/YiKRxhgpHWcBf02dd8IMN38GeqTTDuJ472FeEpOB9VJl7FGgQzjfLuTzHZgnU5R6/kf/X1kL8E8tGVN88HE9cGvqmf9g7oirRdLUDPg1FnMiuXYkwR0znLfBOmmvRaz8yzEjzoTw8zFskJvujB2MGQNuj1XJhgpqTKgctsAGj/uypJGiObbufX9gYMTy1RoLyNQvde1S4KnU+fWYy1rJDQF1lK0zC54biy1riGX4qk/Zahp+7gD0iZiHg7DZvI1T19piRolJ4bwFqWViETRVYoalMeF8ODYTeRDws9RzQ7Fgp7EGRG3Cu79j6vxBUp3EoH3vWHUXNrA4KpTnw7AZ/zUKnumHLT85PsZ7mPreblgHMT24vgVbKw02OLoUcxWNVZ/Wlof7pZ5pHuqI94lnBGiJrf/fIZWvl2CByJJnyrGB0xcRdS2z7grPXRerns9r+UrrCOe7YJ44AFuxpAt+01jplNKzMja42CN1bRzmHZAMQqIsn0inWdK2hONybJA0MFzrQ8RlCg3RFX62jKWnyLVRmJEiWVa4OlCZUVqNT2n4mpSRIvXc+Fj1fOo7i5X58aFO6BA7rVIa2mHG3F+lrq0Z6q+lJjwiaSrH+gnppdH3AKNT5z2xuDA9skq7Bv9fWQvwTy0ZU/vg40psj2kw17log9rwnb3CJ7GsjsZ2WUg/0zX2S4CtS6sMx/th1sMt0y9suNc8sq5WQKtwvC026N+fOtb8RdCURL+/jNSgMdy7BbgxHG9DfGt+nWUri08dZaui4LmyDLTthXmTrB/OyzG31VOyTK/wsxO2nvxczEB4RaILcyuPXUesGspXWTg/Egswl35mTOQy3xVoF45/Her8pQawFOnkllhXD2zwP77g+i3Ar1PnUTr5DczDdoX1awRdqwRdiVfCLsDlRZ6L2rGuT92FDZb6evmqU+8t2JKOqG116vubhzJ1BSkvPSyI7visdKV0NMW8BB7H+qk7A3eSwaC7HrruzlpX0DYa8/K6Axvwdoz8/VJ4jHmuVhspsHhM3TJKn7rKfNRYbanvTur3EUHHcal7N1LQn46pKXWetI13AduE4/Uo4mmY949vM5pTVPUtETkX29d9d7V10XMwt+ge4ZmT6vobJdJVuP3eImw7HUIsjB6qelYGur5NHV8Z1rAdA3whIusCvVT1ZGB+ZF1zUsf3BV07YVv/DMbcyH8RWZMCM0XkbuB8EZmlqveH20dibnSkrsXSVVfZ2gub+Tgjpqagq66ytQ7QW1VP0Yg7eKTWh14nFu37FhHZV1X/GdaaDxGRZsCCkN/RSKXXd1hQwKeC3p2xNcmo6tcxNYXvfLvg0gJs5oOwLnOhqv4tsqYvU3l5Tji+WUQ2xdzze6rqZdia95i6PheRq4AbRWR7VX0m3LoUM1wmz30fWdey8lDVor1/F1nXBwWXFmMzyIjILtjg4zIir/utR7vYKaP2OlflK8Se0eRnwb1dsNnl14OmqO9i0DRPRB7AluvsIiIjgp61MM+rqLoK00stjg8iMgUL7jsEW/o7M5amvOqqrWyp6uNiW7Luig24p8XSFL4/rUVFpEJV3wxtzyMiMhqbZNg2pq5U3tVV5t+MqSnoqt4eWlX/LSJnAb8TkdWwZeMjMM+vTDSlaIK1P58AU8VizR2FBaxtVLiBIofUY/AxNPbgo7ByFZGyMCD7BHhZRLbFIiL/PKaWIrrK1QJEXSEiX2GumVXYetZojXgduu4VkamYJ0xbbPlHdF3h5yMicjpwhoh0wlzx1wPWE5FKIgaYK9BYrGz9EltXnhnLKlsRvr+6TIU8LAMWq+qfReRzzNj0CraGe1tVLbkxrliHPlxvohYc8KlUmesE9I5Rd9WmK9wrD9q+BT4VC2J7EmY4LBm1aQppk5Sts0VkGlb2p2Fu3CWtH4rVp+HnHSLSCrhbLKDcI5hn37pigQPnrmh5WA9dSd31OTBFLOjvsdgSi2j5WOReZnVXLYP+TMuXiIzEdnr5CHhIVaclnX2x7UU7YMHvJmGDjomq+kkptBToGoQNCidjdftCrJ++EPM6+waLo3Uc5rGzvS5t2C+FrpHUnV6dVfVpbCe3NTCPvndWRF310NRRVZ8VkfUxI/QYtWCwUUlpSgwCCwGCkeIqrD8/Sm3L5FLqGMnS6ZXUo1mW+bHYcsHpwA2q+mmiS0TWwLzTtgUOxerS7XVpQ3WptK0FfKGqX6SuLZGPWP/hr5gRf/9S52NJ0By4cfhnafc8ltzCcBtsy6ursC23YsV2qCzQUUZqK7zwswv2ArwSUVeLgvO0rvTWhlti+1/HWiNdV3p1KNAVLeBQLbrSMRTWxLZ9vAVzyS/5GmmKr8vMvGzVpTWrsoVZ5ruEYyn4uSHWeIKtce+DeQvFSptWBeflKW2tU+XtIKzDH6vM16UrWZ+cbCf6NBGC3tUnrcLxVlinKFZaVW9fndKVrh82w9xZb8Dic0QJEEjBNpyhHGWah3XkY1JHJMt1BmDeev/Jia4s28ViujIrX1h/6hUsJtUNpLYwxdzuXwfWCedrE2n5CzZZ8Tq2reJl2KxnhwJdUbc+bkB6rZ86j7LsOI+66qlpeDgvJ+JSNGywfS62lLBXuJbUD0OoCeI5FDMWxugH1jsPY34ww+1LmLfZZaQ2AsCWgb4GjIytK3z/OGwC41fhPN0uDsd2IQLb+WQ+kQJYl+R/zVrAiv4hp4MPLBjNP7Fo5GcX3NsEc7HqgQVeeYRIwWFCJXsf8Dvg0IJ7G2GBwDpjwQF3Il4A0fqkVzcs4OmGxNuBoi5dm2F7hbfAgmY2I8I6SGAktrYw3WlOynyWZWtjzIOk2NrMTMoWsDlmnX8pqSdS91YPabV1jPQpom2rkCZnEiJrp+6NwOKttMNmRZ8h3kBtWbquw4x26wHfx3gX65lW7YOuo4ln7N0Wi+p9GXBxwb1NqAla2CXoi7IeOZT7G7DYJYVxcjLJw/Dd47Ho/yeQipUQ7iV1RCdsl4ApxBus1UdXVu1ibbqil6/w/99MzcD1NGxmdjC2nPAAYKcYaVOgqx1wLyFoMOZx899QX3QFjsC8OCBuvKq8plfudDVEE5HjVVG/AfdG4bwJqZ1YcpJeMct8GbYjx7hwfhA23hiJ9Ul3BHaOrSt83zhsoudK4KaCexsQdmsK583IKGjnT/b/Zi1gRf6Q08EHZgR4C7PoJ8FgkopiJcw9dPvU87GiIY8PabIrsCe2FVmyLdgAbDvFCannY23x06D0ylk+JpHnowThwzqlVVj08YksaaTonWHZSrZ7fBqzQqeNFP2zKFvY7gAnY4PIc4OGLqn7ralp3GM3lEOBd4K2MZgh6Y5wr0+4t2Xq+VY50rVV6vmSB0v7AWm11Da2JdK1KtYxHYNFTH8S21++KdYRex5bKhS1fGGGwvcwI+ZEbOB2Vri3CrbeN2oehu8ZgRkdJmBeZ08A14Z73UN6bZN6PsoOBvXUlUW7WB9dUcsX1ml/BDgFcxd/FwvmeDW2Q8xKMfWkdLXCJhLSu1jdCvwB21lIMtKV1/TKna48agrf15ABdzTDSY7TqxwznJyHxbuYgm2DfCnmEZeVro2wvkLigfM8cFLq/u6EdpGCAO6N9ZO5gBX1Q04HH0HXmdREf22Cre09LJxXW+WIuNcvNlt2FcFgg20xdw1hC6Kge0jO02uF1hU0bI8NPH6GzYLuRDBSYFGbB6aejVW2mmKN9kTMrfZvpDwpQloOjZ2H4ftWJixpCg3ksxSZZcxA10Dg6oJrTwM3h+MkvRIX0lhlrL66ymLpaoCm2HnYHesQdkpduxO4JRz3z0jXrsBR4bgpcCrWSfxtqBcS9/toeRi+Zyxwaup8b2wpzhXhfFD4GbvM11dX7HzMjS4s3lOy/GZwqEf/TvAqxAz2lxLc3COmUTOCER7bkel6bL392dggaW9SW8tH1JXX9MqdrjxqKtCXqwF3XtML83DuGo77YBN6twMXpZ45jwLP7Yj6ugLDUucTQjq1y0JPjE8TnExQi059PfCoqh6HuebcLSLdwv3Zajt3RCEVmPN7bOA4OQRdqcJmrDYM9+drTSR1TSi1PrWoy38AXgzBfWZhgaSGJrpV9dVS60j4gem1QupKaarCZonuVdWbwvEEYAsR6aiq81R1Svi1kpetlK4F2Cztg6p6IRaH4yRgfbHI1t+r6iul0lFEV2VyrKofAvPC8aHYes07w3MTxCLME+MdTNJLRJpg3ibdRGTNlNZNgK4icqiqvpIqdyXV9wN1LS6lrh+oKWYeJoH3vgWGpXRNxAKZnqOq74ZrJddVQBlwhIj0Ce9mE6zu74ctE/pvjDwsQhNgfxHpH857Yy7TPURkW7Wdt6KU+R+oK3Y+5kKXiEzCBhoPichBwHRV3Qjb5vQzAFX9GFt+2bOUWgp0TcS8JO4TCwT9EjaQHIgZuXZT1WuBFiLSPqKuvKZX7nTlUVNKWzcR6aoWTPjXmCfo/2GBKH8W+hNPEnbNidT+5DK9wrt4O3CviPwfZrTfAXs/08ElFTOwRCPkY3dV/VJVX0zdegObSNsipp6Y+C4ekRGRIZiF8F/AZ6EDhqoeKiKXYYOPjURkAuYWfWukjkVn4KugZUrBvblYVGtEZA/MdfVPkSq0Dqo6PegqNEDMJ2y5GiqYMlW9bUVOr5zqSmuakTIMXB0GcFsBX4rIcKzMnxYprdpjM3qo6hcpXWeKyIlYY36o2HZbrVT10gid6a2B40TkbFV9KHW9TFUXwN0Q4wAAIABJREFUq+ohInKmiHyNbTs8vpR6CmiPdSiqgHdE5BHgARHZOHQqwAaSQyDqAC2PuvKoCcK7GDqt34jIU8CZIvKtqr4QnjkA2D+SHmDJdlFVbwyG+udCuvXB1lAPxYwUr8ZKLxFZD5vle15VHxaRs4H7ReRFbAvYkSLSAegIUXdjcl3109MT81TdGxtY7ACsLCK3YcsMLw5t0FfYFsjnlVJPSld/4HRgP6wPsx+2BOZ2VX0w9dze4X6ULaxznF6505VHTSltE7Et4ytE5H7MKLFDMBD0Tz0abcCd1/QSkY7YTj37Ykb7scBBId2eBE4S223oIyxw5x4xdAVt6Xy8C3hZVf8BtrW1iPwOOExE/pXqVyw/aA7cOFaUDxYL4EPMUncP9lL0LXjmTODr8FysAFvbY7N8h6SuVbvXY/EwLsAGki8QL/r3tlilsFuBrsSFdiy2fm1s0DVgBU+v3OmqQ1N614AJmGfAJ8BaeShb4fwXwPvhuRi7PKwJfAFcgrkXblFwP3Fpn4gZVqLtbFIsvcL1k8P1xO3+l9gSmaZJuVvRdOVRU/i+pd7FcP0wzNV3ArbsY3/gKQp2SyqhrnS7eB9mGGyFxcgYQlhPiy3xOCCGpvB9E8L7fwVwI7ZlW2dscN2XECsE87Y6ORzHyEfXVX9N/TCPuKTuXBXrd52FGbzGY+7lNxFh14KUrnWAp1Lnw7E4AcdggzfBlny8SQicGUlXXtMrd7ryqCno6Ih546yNba16FBYkfZvwPr4CnIHV85OJF2A4r+nVBTNEJMtOumMx7v4CrIsZXG8FLo/8LhbLxz8Bu6Se6YxNao+KpSvmJ3MBK8onNDiXUhOoJhlc/xFYJfVc1MEHNmv1MNZ5/oLU+ipqDAEDqAnmGUvXAGzZyznYYHq3IrqGYxbP54m3Q0Be0yt3upahKW04mYRtFRjLkFNX2UobwCYRdzu+5ligo87Az7Ho91sUaGuHRbuO2VDWml7h/hHYEplbMLfDWO9i7nTlUVP43lrfxXB/d6x9ugfrtMbq5BdrF08HLmLJdnE/zLgTxQgdvvPMJP8w4+6vMAPKmqln9sUMq1EmE1zXD9J1OfAboHk4Hxiu7RXOm1KwpW0ETYIZcHYDmoZrw7EBWhI4tDORtjjNe3rlVVdONeVywJ3X9ArfexFmWG0Vzntgy2KOC+fVWzbnIB8vA0annjsG6BM7zaKkQdYCVpRPaJBuJEQjD9eGYUaK48KL2RELCBZz8FFBTeTXdTDjSGHndSXg30TcsgYLHrVjON4Ks/oWdvbXxjo7MXXlNb1yp6s+msK9CcT1BqhP2WoGHEi8AWSTgvO2mCvkA4QdHqjZuzxqhOZ6pld3LKBnlG2Q86orj5qSMlOP+qEFNijqGlFXsXZxbWraxXJs544/EclQmNJxHqmtV7EO41FYh7odFrTsZNeVT13UGMDHA7/HjCNJwOFxwKNE2mmlmDbMQ+/32C46zcL1nbEZ0eZZaMpjeuVRVx41FejL1YA7r+lFzWTUqkHX6ak0WxszEHTIYT4enZWmqP9/1gJWpE94Ce4ibOkTrm2FDUIqw3k062Gq0khvqTiMVOcVs7hWxKw8KBIFHdgSeBX4WThPIoBHi2Cb4/TKna56alqPCPtt/4CytRo2WIsa9b6I1g7APthsxxVYIKnYHbH6plfUSNJ51JVHTWk99XgXu8fUldJSn3YxynbDBbq6YrGiDktdG4oFMuwbzrOY7XNdDdPVFJt5PB8bgFRghoD7iGgIKHwPMY+5k7Fllz8P1/YM6RVzu8dC43jm6VWsnORBV941kbMBNwV9qLykV5EyXwZsEnRdD3TCdpt7jEjbWP+AfGwfW1fsT1JROhEQkRbYnsOjgEdU9dZw/X7gQlV9PGN95aq6SESGYZ3D57FIutuo6hdZagMQkS2xmbUPMEviDhoCaGakJ/P0KhYFPQ+6iugspqk7tpVU3spWN2BilmUrjYjciAVn2lJVX85aDyyVXl2BSbHSq67I/1nqqo28aarlXewGbJ/Fu1ifdjH2LhRiO0VVicgm2FKFx1X1onDvVuBOVb0toh5RVc2brpS+3OhKtITjpKx3wzr2E7BYP62weCYv1vGnfio9lcBCVZ2bulahqgtFpBMWG2Z9zN29LbCPqr4UQdd6wPuq+k3qWh7SaxTmUXlJEhw6a12hjlqsIah9uNZUVRdkmVa1ISJl2I5t22FtzlGYp86BWL95Zom/vxsWxHt2Um9nnYdBw6bA2+l2LlXGemGTQntjS9Q6AAfHykcRaa2qswuuZZqPmZO1hWRF+LDkzFVbbN3h3Zi17iDgPSzCdTQdy7oO/A74hojBagr11KLrcizK79DYeZen9MIsq5V501XPPIymqYFlPlrZwhrmMZiRJkmrxGreF+gfjkcDHxNpqUlD0jGDd7E3BbMsedCVt7RqbPV8OM6sXawrzcK9DYGnQx4eD7wLrBxJ11L1fB501ZV2WegCNgX2TZ2ngx2PBq4EuoTzXkTyYsIGYfcBjwBj09qAzYELgbbhfHWgcyRd44DX0vVRqh0ak2F6jcM8u+ZTs6QxSa9M8hEbHN6EeXmNI7XeP+RhJmlVoLGs8DhoGYJ56PwdM0QPi6Bla+BxzPi9L+adkORh1mXrA2D91LWkzI/GgjSvFM7bEjwWImkbhbXF5am0yjQf8/DJXMDy+MHcZ7fDtkdrGa4lgZBWCY3pylhwsHOJ18FPGsJ04528DGtSs0Z57VDBxBpsrwvsFL63eYGugUmFgnV8XiZSjA4s6vFvWbIjnYf0Ggf8D9g6nJdlrQtYC3PL7pUq8xVZagrf1zr8LJaHmZQtzO3/NWyt8T+A3ql7I7ElAYPDeQsiDjpCmT+u4Foe3sXx2E4Tq4bzsqx1he8dgbmpJjsUVGSpKXxfY6rnk05Ylu1iXfX8YKzz2AHbXeQ3EfOxrno+S13F6vqmWekK6fQt5vZ8TMG97qHemBgjbQq+e2vgxVBHHIDtYJC4aa8UdO2Uga5tg65hSblK3euKDYAmZaBrQtC1OnACFmg4KVeZ5CO2ROk1bIC4I2aoOB8YhMVVySStgrZxwAmp83R9P4oMBtyhzL8U6vqtQvvSPtzrFNIrizI/HosHNSKcN0vVpW2w2GxZ5eN4YCowi5ql64m2TPIxL5/MBSxvn1D5vw3cBlwDXExY24tZW18F1stA1w7hBUgC7qUrs5Hh5d0wnFcQz5KfDNauxGYaNkzdGxV0JdvyNQN6RNI1PlS0mxe5NzrD9BoXdP0TuLrgXib5GMr8a1ichPOxWaFO4d6YDNNqm6Brk3CeHoBslkXZCt/7DjWD19sJO3WEhvL0pKEkDHpjfVJla2yRe1m+i4mud4A/FUnPLPJxu1C2bsO2hT2cmngJozNMq8ZYz48mu3axrno+qbs2y0BXfer5LHTVVddvnoUu4BBsQLs2cAcpIwXQnrAbDAXrzkusqSKkT1K398IMKOeFdNqAsCMNkWMdAdcBX4bjSmxm9rpQdwwiGMxj6gr5dBUwJpzvgMXiSAIptiMY6iPr2ha4NXV+IOYddCzQn5rYKrHzcFPMI+8t4PzU9TLMOPg8IVhzZF0nAxPCcTfgv9juIT/DYjx0yyi9zsOWdoAFgv4rFtvrUGwLz3YZ6doGM8r1w7b7voOaSbWOWeVjXj6ZC1jePli08e3C8TBswHFraKCOpGbwEbOS7Yfta38pMI2aGbTESnc0NZHnYwZnGgG8Ts1g7XIsgE4LbF3aUQRra2RdQ7Et7bYP5+2xzs/K2GBjv4x0jQY+JMwsYjMKe6buH02YYYilK+TVrdTMxowMuq4NFez+qTIfM60GY3vI/zk0AJuG6xIa8VMyKvMDgJHhuDu25eM92MDtQMKAMYOGcgTwdepdbIfNaleGMn9sRvk4BnP1XyOcP0TYYiurfMQ6gH+jxstlL2xrztOwWaLjYr+H4buWh3o+Zrs4NNSn9annY+oajbki16eej6mrBbY9bn3q+tj1V1vMKDASM/oel7oXPTBn+N7Em6oNNtD4XSj31yb5GTudUtruxAa3T2JbHx+GBQf8WRa6sHa5ZcG1fwKXZJE+KQ2rADek8utkbMB9IxkYVFO6dsW8cjpigRwvKLifTIpG3Roz9f0tMY+h0zFPirMxI6JkqOmv2Nbf/8IME7uEd/K4LHRhQXIvIhjHQ9tzDdAv9UwmBp28fDxI5k9EKtDKX7DAQ+eE672xhrslcLqqzsog4FclZpm+W0R2wBrI3VT1gWL/Q0Rd3bBZhKdFpCvWyX8V+AyzDl+oqtMySK/uWAN0U9B0CdbhX4R1tM9R1bkZ6BqEWXr/Hc6PwCIyn1LwXBRdIiJYVOaHgatU9bpw/S9YWn2qqmfF1JTS1hkYp6o3isjPsUHQYar6ZLifvK9Rynwo36qqX6Wu/QrrPJ8pInti1vQjVfXzUuspoq8v5o55ArZe+k5gNrAAG4zfoqpzMsjHMcD3qvqciLTFjAAfquofwv0k8FY0XaE+vR84N6lDReQ24HOsA3STqmoGadUOM97clbN6vjsWUyVv9XxPbFB2M5ZveannV8eW6TwXzjOt51Pf1xwzEF6tqteGa5nU9SIyEtvGtKmq3lCgcQPMq+JBrA5rDtyoIXBmBF1dMQPgTanrg1T1rXC8GzajvKOqLiy1ppSuLphHQpJ3NwMfqOpvwvnPsIHvjqq6KKKurtjg8OZwLQm4ujpwInCGqr4ZQ09KU5eg6RYR2R/YAjOmVqnq1iJyCDBQVY+IpauIzvaq+q2IrIx5NL2uqkeGe+1U9btIOnoBs1R1RvrdF5FeqvppOB6FGVYnqer82LpS1y4GpqbqqjFY33BiLF0FGpsl3xv607cBC1R199hacknWFpLl4QOMxSIvg3lNvExq5gVrMK8nWMMi60q2r2qeur495gY8IaU55raYYwkBragJUrM/cHg43hQzDqydQXrtF477YQPv97AIw2Du5DcSP+BkdT6G82RGdBg2Azgqpp4iZWsbbLbjBMyt9Q7MjfWmDHR1pSYAU3nq+j6Ya/LIcN6fSDNr2HKAJ7B1jrvW8dz9RJ6VCenVNRwPxJZRTE2V+b2wRjN23dUV6Jg6T2IVbAZ8CaybUdlKPFz2w2bWDgDOCvXVfsANGeiqSMoyYc12OM66nq9g6aCmB+agnq/Whc2QPgS8n4N6vjofw3nSNmZWz6d0JbEAkgB4v8mqrseWm03FdguZDFxGKpgq5v2yGuYhMINIAYaL6Lqc1NKuVH7uGuqOWO1Poa6/kgpMmNK1e9DVNCNdlxWkV9fQ9hyQUdl6MZTxxJNwADVeMYdjE45RdIXv7EWIL1TkXl/gUczDYxIWnLYigqbtw3t2NDVLvZpg4570ktpJ2NK+WNvcL6UrdS8dc2Vn4N5YulL52K7gWtKn74gZVsfFLFt5/WQuoLF/QuM8Dfges6gmL8cDLLmv+/0UWdsdSVeyI4CkXoQdsOjtf8UGT7HWIqd19avjuduA8RnnY29gj4Ln7sggH6cXpleqU3EYNlMaba/mYnmIdeovAE5MPfcAqQCQEXSlDQG7Fbn/c8zV9pLQWMaIAL4uZrBcA+vcPwO0KfLcJKxD1D2D9HqeGhfW/sAvCp57ABgeWdeT6XzEOj1J3XVS+FRfi1y2dsLctncO9Wd6DfD91NKJLKGuG0O9NI6aOBhJ/ZBVPZ/WNba2+imDej6tK4nT0Y3g2p56LnY9X5hebQruR6/na9HVCQv6ej5wUuq5KHU91oc5FzgqnDcP+i4iGFrD9YOw3Y9Wj5ROten6I8FgHq7vjw3GYxlNatN1SYGuA4KuWIFW65uPuwBvYMuLSurmXoumm0Mepo3lv8JirQyOkVbhO4sZAgp3z2mKeaV9GyMfsVgOjwFXA2cAv6TAGBCeOwSLRRGrbNVX16Exy3xt+VhQ/lpgEx7HxdKU50/mAhrzBxtwvIytsT0WOCJcb0NNMKkTsdgTbxIisWaoq0nBz6uIuzVmbbrKC57bEQsQlll6FWoKz00MuqIMupeVj+E4icTfJyNNh1NkkIittX2JEME5gq76GgJuxmItxNohYCJwbThui63LvBLbfmsVbFupfbFo/dG2Eq0lvZbKK8xw8hKRPCgKdG1VLB+xQdPzFMzOR9I0AQuSttSMC+Zt8nyxclciXcODrmFB15PYEpgk4ndipIhdz9ema+WC52LX84W6nsJ27yjsKMau5+vMx/BM1Hq+Dl2nF9YTGdT1u2KxvhLPr5ZYbIzLUs8cScSBR310YR5qV+RQV49wP1e6Us9FKVcNKFunE3ELa+o/4J6EeVpFMZxgBpE1MY+liZhx6TBSHqyhbF0Us2zVQ1cTaoLDxtRV33wcj3n0tWIFjT1RnRZZC2isH8z97BpqXMYPwFys0oPHwdhWSefHqtBq0fUINUaJpNO6HhYsJlandZm6MFfSA7DBWqxKtr7p9XPM8pkXXelydhEhUnkGmh4tklY7YgO1aC7S1G4I+Dk1kbZHYWvLYzZKg4G7sUHi+5jBcruQjoeExnI8wWsnB+m1b1KWsGUxb8Yq8/XIx7QH0XUxynwtmv4dNO2XSqu9Q1rFLFsTgWtS57/ABor7U+Nmuz4R6/l66mqGLfWIVs/XU1eTUP6j1fP10JV2R45Sz9eh6+Cg64BUek2MUddj3ozNsBnGPths+1hqdnlogc3SbhcrfRqoa6uQXi1ypmvrUE/EWm5SX13b5jAPo5atlL5lDrjDz72B1SLoWSloKgxqOhHzNjksnA8JP6PsRtYAXUnw4dhBMeuVj+E4qqdcXj9NcH4o07AZ7ScAVPWvWLCo30F1oKg3VPXXqvorVX0lQ12LMUMJGko/MAULhvRqXnSpBYz6D9YQvJEXXeG5Z4BtcqSrSkSahOMjVPWDjDQtYum0egoLhhSrbIENKipF5CqsE/0AtoxjM2yLQzBX0a1U9bVSChGR9URkIxFZK5SXk7AYHa+p6m9V9V5sgD0J6wQ9rKpTSqmpCLWl16ZYZxos/sr2Ect8XbrS+Yiq7hWpzBfT9LegaRNq0upRLNZDSctWAR8BVSIyNpx3wJYJ7QNsEN7Ht4hbzy9L1wi1oGCx6/ll6dpALYhi7Hp+WbpGJA9FrOdr09U+6NqbmvQqeV0vIlsDf8e2bL8KC9x7M+YlsYmIdFfVudjsZJTgjj9AVxNVrQrnedIloe9V8gCBDdQVJYhvXstW0LaSiDTFBvivqep8Vb0Te+cGYMtfwHYjQlWv1RIHEw3p9SC2ROjqELid8P13Yl5WnUXkHuBfItJDIwRcbaCuZ4KukgfNDdrqm49DUppnxtCWd3wXjwYiIuthFrA5qvpiuNZCLdr3KKywnaAWXTdmZPmG6IoWxb2BuppErDQafXrF0NNQTRmkVUXQ9bKIrIG5Qk9S1e3CM6Mxr4UdNUJU69BQnoUNrFcCZqjqoeHeH4F7VfUxEdkWWye9i6rOLrWu8P0NSa/tVHXWiqorj2WrQNdsVX1FRI7FvOFaYTbCrUXkQCxo4CkxNLku11UiLYIFlHsQm2l8EzOO/BILPL42FgcGbO39ntguNm+7LtfV2DQV6Nsam+z8F+a1d4qGXWDC/YmYYWIIthX3IFX9rIR6iqXXHsAxWKyeN1LP3gBsjBl6Sz0ZlEtdqe9saD4O1Ax2ccsr5VkLaEwUDD56ichcVT0oZRWfglVsuwKXRzRONFRXTOt0Q3TFMk4sF+mVR00ZpdVKIpIYAl4XkXVFZIyqPga0BuYSYfZDRFpiHZyjgxFiJWCyiLRU1Z8DnwL7i8hhmPHi5xGNEw1NryzqrlzoymPZKqKrj4hMVdVfiUgbbNehpDPWPpamH6Ar5vamy4OuvOZjyXWpqorI59gyr3eAr1T1XBFZiHX4R2AeHetinfwxMQaQrqvx68qjJlhiwH0O1pdIBtyPi0j1gFtV7xTbSnotYMNSGifC9xVLrwtCev1DREap6ttiW0oPwzwvS24EyKuuH5GPbpxIozlYZ9IYPliwnL9jFRXYAOMr4KqC5yZi21O2JEKAE9flulYkTcvQdXU4Pw5z07wH62TEiv/SCtuVYI3UtfOwtfanYmuQh2M7K8RcS57X9MqdrjxqqkPXNODKgucOx+KslHwtsutyXSXU0x8bHHYEbqUgqj22pfVVRIqd4LqWH1151FTw/WXAX4Ce1Hi5H455cqwazrtj/Yq1cpBex2ExtZKYHbG2Es2lrrzmY2P8eAyK+iPYnvJfAqjqx9i2X+uLyAWp5x7Ftsb7XkMJdF2uq5HqyqOmZek6FQtKe0G4NlFLHP9FRFYNOuZgjc1DIrKTiPwZC7A1Aevwt1DVyap6t8ZdS56r9Mq5rjxqqk3XVcCI5F0UkRZY/ILdtcRrkV2X6yqZEJEJwF3Yu3YaFrTwEBE5IfXYzZgXx4JS6XBdy5+uPGpKaesvIusC7bDlALsn/SlV/SMWVPE3YstrpwLrqerLJdZUn/S6DUureUFryb1C86oraMtdPjZW3ECxDOox+NgGC6DWLjw3Q1Wnuy7X1Vh15VFTPXVFNwSEhvJlEbk1aDsV85oYCHyH7an+PmblryylliLacpdeedWVR0311FX9LqotuzothtHEdbmuEunZEBt07K2qm2GR79cDNgQOFpETRaQ/MBJz2W5XKi2ua/nSlUdNKW25G3DnNb3yqitoy10+Nmq0xC4ajfmDdUq/B25NXTsCC4z2O6AiXLsX6O66XFdj15VHTXnVhS3peAjbMvEa4OZantsD2x1gqT2vV6T0yquuPGpyXa5rRdSFDTL2SZ13Bh4Ix30xr47LgMnE3dLXdTVyXXnUlNL1FrB2OP8L8FugB/BxeB/7Y7v6TAbae3rlVlfu8rExfzIXkNcPOR18uC7XtSJpyrOu8J09sGCJnYA7gJtS98qBLbAtFaOtMcxreuVRVx41uS7XtQLrKgMqU8e9sK19u4drfUK92jaGHte1/OjKo6bwvXkdcOc1vfKqK5f52Jg/mQvI84ccDj5cl+ta0TTlWVeBxo7AncAN4XwNYFugWwZacpleedSVR02uy3WtqLoKNLQGHgvnewB/JgS9y+rjuhq/rjxpIqcD7rymV151NYZ8bGyfJLKoswxEpCPmsjNXVfcQkTUwq9h/VPUL1+W6ljddedSUZ11BWycsBsWGWIyfzTTjraPyml551JVHTa7Lda2ouoK2a4CpwDhshrLk2wTWB9fVMPKoK2+aRKQcaA7cq6pjRGQPYBPgSK3Z2j0z8pZeCXnTlfd8bCy4gaIB5HHwAa6robiuxq0J8qsLQESOAo4HxmbdUCbkNb3yqCuPmsB1NRTX1TDypktEBKgA3gw/x6jqO1npSXBdDSOPuvKoKU0OB9y5TK+86krIWz42NnwXjwagqt8Ar2Jbx+yYh04FuK6G4rrqTx41QX51iUh7YCtgXJ4ao7ymVx515VETuK6G4roaRt50qbEAOAPYMi8DD9fVMPKoK4+awAbcItIUm23fHdg1D/2IvKZXXnXlNR8bG+VZC2hM5HXw4boahuuqP3nUBPnVparfisg2qjovay1p8ppeedSVR03guhqK62oYedUFXKuaS1df19Uw8qgrV5qClgUicgbwQl4G3ClylV4pcqWrEeRjo8CXeDQQEWmet8EHuK6G4rrqTx41QX515ZW8plcedeVRE7iuhuK6GkZedTnOioSISJ4G3M4Pw/Pxx+EGCsdxHMdxHMdxHMdxMsdjUDiO4ziO4ziO4ziOkzluoHAcx3Ecx3Ecx3EcJ3PcQOE4juM4juM4juM4Tua4gcJxHMdxHMdxHMdxnMxxA4XjOI7jOD8KERkpIpr6LBaRb0XkdRG5VkS2EBH5EX9/LRE5VURW/ulUO47jOI6TN8qzFuA4juM4znLDzcCDgABtgIHA9sBewKMispOqfvcD/u5awCnAE8CHP4lSx3Ecx3FyhxsoHMdxHMf5qXhRVW9IXxCRo4FzgaMxA8aWWQhzHMdxHCf/+BIPx3Ecx3FKhqouVtVjgGeALURkYwAR6SEiF4jIy2E5yDwR+Z+IHC8iZcnvi8ipwNXh9J+pZSTXpJ5pJiK/EZE3wt/5TkTuF5G14/2njuM4juP8WNyDwnEcx3GcGFwJbAxsjRkrhgA7AncD7wEVmHfFOUBf4KDwe3cB3YEDgbOAN8P19wBEpAJ4CNgQuB64BGgLHAA8KyKbqurkEv9vjuM4juP8BLiBwnEcx3GcGLwafq4afj4J9FVVTT3zBxG5HthfRE5V1amq+qqIPIcZKB5R1ScK/u4vgZHAFqr6cHJRRC4DXgfOD/cdx3Ecx8k5vsTDcRzHcZwYzAw/KwFUdW5inBCRpiLSQUQ6AQ9j/ZPh9fy7ewBvAf8VkU7JB2gKPAJsLCItfsp/xHEcx3Gc0uAeFI7jOI7jxKAy/JwJICLlwK+xHT76Yzt/pGlfz7+7GtAC+LqOZzoBn9RbqeM4juM4meAGCsdxHMdxYjAk/JwSfv4eOAy4FTgT+ApYCAwDfkf9vTwFeA3bJaQ26jJeOI7jOI6TE9xA4TiO4zhODPYLPx8IP/cEnlLVXdMPiUj/Ir+rRa4lvAN0Bh5X1aofrdJxHMdxnMzwGBSO4ziO45QMESkTkfOxHTweVNVnw63FFCzrEJFWwFFF/szs8LNDkXvXAd2oxYNCRLr+EN2O4ziO48THPSgcx3Ecx/mpGCYie4TjNsBAYHugD/AP4GepZ+8ADhKRW4FHga7AvsC0In/3BaAK+D8RaQ/MAT5Q1eeBi4CxwHkiMhp4HItzsRIwBpgHjPop/0nHcRzHcUqDLLm7l+M4juM4TsMQkZHAP1OXqjCvh0+BycDNqvpQwe+0BE4DdsaME58AV2LGiEeBn6vqNann9waOxwJqVgDXquo+4V45cAi2bGT18CufA/8Jz/3jJ/tnHcdxHMcpGW6gcBzHcRzHcRzHcRwnczwGheM4juM4juM4juP1/hn/AAAAkUlEQVQ4meMGCsdxHMdxHMdxHMdxMscNFI7jOI7jOI7jOI7jZI4bKBzHcRzHcRzHcRzHyRw3UDiO4ziO4ziO4ziOkzluoHAcx3Ecx3Ecx3EcJ3PcQOE4juM4juM4juM4Tua4gcJxHMdxHMdxHMdxnMxxA4XjOI7jOI7jOI7jOJnjBgrHcRzHcRzHcRzHcTLn/wGichbT/o87IgAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 1296x648 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize = (18,9))\n",
    "plt.plot(range(df.shape[0]),(df['Low']+df['High'])/2.0)\n",
    "plt.xticks(range(0,df.shape[0],500),df['Date'].loc[::500],rotation=45)\n",
    "plt.xlabel('Date',fontsize=18)\n",
    "plt.ylabel('Mid Price',fontsize=18)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "c:\\users\\leomr\\anaconda3\\envs\\stockmarketpredictor\\lib\\site-packages\\ipykernel_launcher.py:2: FutureWarning: Method .as_matrix will be removed in a future version. Use .values instead.\n",
      "  \n",
      "c:\\users\\leomr\\anaconda3\\envs\\stockmarketpredictor\\lib\\site-packages\\ipykernel_launcher.py:3: FutureWarning: Method .as_matrix will be removed in a future version. Use .values instead.\n",
      "  This is separate from the ipykernel package so we can avoid doing imports until\n"
     ]
    }
   ],
   "source": [
    "# First calculate the mid prices from the highest and lowest\n",
    "high_prices = df.loc[:,'High'].as_matrix()\n",
    "low_prices = df.loc[:,'Low'].as_matrix()\n",
    "mid_prices = (high_prices+low_prices)/2.0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_data = mid_prices[:11000]\n",
    "test_data = mid_prices[11000:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Scale the data to be between 0 and 1\n",
    "# When scaling remember! You normalize both test and train data with respect to training data\n",
    "# Because you are not supposed to have access to test data\n",
    "scaler = MinMaxScaler()\n",
    "train_data = train_data.reshape(-1,1)\n",
    "test_data = test_data.reshape(-1,1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Train the Scaler with training data and smooth data\n",
    "smoothing_window_size = 2500\n",
    "for di in range(0,10000,smoothing_window_size):\n",
    "    scaler.fit(train_data[di:di+smoothing_window_size,:])\n",
    "    train_data[di:di+smoothing_window_size,:] = scaler.transform(train_data[di:di+smoothing_window_size,:])\n",
    "\n",
    "# You normalize the last bit of remaining data\n",
    "scaler.fit(train_data[di+smoothing_window_size:,:])\n",
    "train_data[di+smoothing_window_size:,:] = scaler.transform(train_data[di+smoothing_window_size:,:])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Reshape both train and test data\n",
    "train_data = train_data.reshape(-1)\n",
    "\n",
    "# Normalize test data\n",
    "test_data = scaler.transform(test_data).reshape(-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Now perform exponential moving average smoothing\n",
    "# So the data will have a smoother curve than the original ragged data\n",
    "EMA = 0.0\n",
    "gamma = 0.1\n",
    "for ti in range(11000):\n",
    "  EMA = gamma*train_data[ti] + (1-gamma)*EMA\n",
    "  train_data[ti] = EMA\n",
    "\n",
    "# Used for visualization and test purposes\n",
    "all_mid_data = np.concatenate([train_data,test_data],axis=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "Unrolled index 0\n",
      "\tInputs:  [0.03143791 0.6904868  0.82829314 0.32585657 0.11600105]\n",
      "\n",
      "\tOutput: [0.03143791 0.6848606  0.833369   0.33078218 0.11600105]\n",
      "\n",
      "\n",
      "Unrolled index 1\n",
      "\tInputs:  [0.06067836 0.6890754  0.8325337  0.32857886 0.11785509]\n",
      "\n",
      "\tOutput: [0.132895   0.6858036  0.833369   0.33355275 0.12158521]\n",
      "\n",
      "\n",
      "Unrolled index 2\n",
      "\tInputs:  [0.08698314 0.68685144 0.8329321  0.33078218 0.11946969]\n",
      "\n",
      "\tOutput: [0.11098009 0.6836884  0.83387965 0.33355275 0.12230608]\n",
      "\n",
      "\n",
      "Unrolled index 3\n",
      "\tInputs:  [0.11098009 0.6858036  0.83294916 0.33219692 0.12106793]\n",
      "\n",
      "\tOutput: [0.11098009 0.6820074  0.83420646 0.33355275 0.12106793]\n",
      "\n",
      "\n",
      "Unrolled index 4\n",
      "\tInputs:  [0.132895   0.6848606  0.833369   0.33355275 0.12158521]\n",
      "\n",
      "\tOutput: [0.132895   0.6848606  0.833369   0.33355275 0.12192084]\n"
     ]
    }
   ],
   "source": [
    "class DataGeneratorSeq(object):\n",
    "\n",
    "    def __init__(self,prices,batch_size,num_unroll):\n",
    "        self._prices = prices\n",
    "        self._prices_length = len(self._prices) - num_unroll\n",
    "        self._batch_size = batch_size\n",
    "        self._num_unroll = num_unroll\n",
    "        self._segments = self._prices_length //self._batch_size\n",
    "        self._cursor = [offset * self._segments for offset in range(self._batch_size)]\n",
    "\n",
    "    def next_batch(self):\n",
    "\n",
    "        batch_data = np.zeros((self._batch_size),dtype=np.float32)\n",
    "        batch_labels = np.zeros((self._batch_size),dtype=np.float32)\n",
    "\n",
    "        for b in range(self._batch_size):\n",
    "            if self._cursor[b]+1>=self._prices_length:\n",
    "                #self._cursor[b] = b * self._segments\n",
    "                self._cursor[b] = np.random.randint(0,(b+1)*self._segments)\n",
    "\n",
    "            batch_data[b] = self._prices[self._cursor[b]]\n",
    "            batch_labels[b]= self._prices[self._cursor[b]+np.random.randint(0,5)]\n",
    "\n",
    "            self._cursor[b] = (self._cursor[b]+1)%self._prices_length\n",
    "\n",
    "        return batch_data,batch_labels\n",
    "\n",
    "    def unroll_batches(self):\n",
    "\n",
    "        unroll_data,unroll_labels = [],[]\n",
    "        init_data, init_label = None,None\n",
    "        for ui in range(self._num_unroll):\n",
    "\n",
    "            data, labels = self.next_batch()    \n",
    "\n",
    "            unroll_data.append(data)\n",
    "            unroll_labels.append(labels)\n",
    "\n",
    "        return unroll_data, unroll_labels\n",
    "\n",
    "    def reset_indices(self):\n",
    "        for b in range(self._batch_size):\n",
    "            self._cursor[b] = np.random.randint(0,min((b+1)*self._segments,self._prices_length-1))\n",
    "\n",
    "\n",
    "\n",
    "dg = DataGeneratorSeq(train_data,5,5)\n",
    "u_data, u_labels = dg.unroll_batches()\n",
    "\n",
    "for ui,(dat,lbl) in enumerate(zip(u_data,u_labels)):   \n",
    "    print('\\n\\nUnrolled index %d'%ui)\n",
    "    dat_ind = dat\n",
    "    lbl_ind = lbl\n",
    "    print('\\tInputs: ',dat )\n",
    "    print('\\n\\tOutput:',lbl)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "D = 1 # Dimensionality of the data. Since your data is 1-D this would be 1\n",
    "num_unrollings = 50 # Number of time steps you look into the future.\n",
    "batch_size = 500 # Number of samples in a batch\n",
    "num_nodes = [200,200,150] # Number of hidden nodes in each layer of the deep LSTM stack we're using\n",
    "n_layers = len(num_nodes) # number of layers\n",
    "dropout = 0.2 # dropout amount\n",
    "\n",
    "tf.reset_default_graph() # This is important in case you run this multiple times"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Input data.\n",
    "train_inputs, train_outputs = [],[]\n",
    "\n",
    "# You unroll the input over time defining placeholders for each time step\n",
    "for ui in range(num_unrollings):\n",
    "    train_inputs.append(tf.placeholder(tf.float32, shape=[batch_size,D],name='train_inputs_%d'%ui))\n",
    "    train_outputs.append(tf.placeholder(tf.float32, shape=[batch_size,1], name = 'train_outputs_%d'%ui))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "lstm_cells = [\n",
    "    tf.contrib.rnn.LSTMCell(num_units=num_nodes[li],\n",
    "                            state_is_tuple=True,\n",
    "                            initializer= tf.contrib.layers.xavier_initializer()\n",
    "                           )\n",
    " for li in range(n_layers)]\n",
    "\n",
    "drop_lstm_cells = [tf.contrib.rnn.DropoutWrapper(\n",
    "    lstm, input_keep_prob=1.0,output_keep_prob=1.0-dropout, state_keep_prob=1.0-dropout\n",
    ") for lstm in lstm_cells]\n",
    "drop_multi_cell = tf.contrib.rnn.MultiRNNCell(drop_lstm_cells)\n",
    "multi_cell = tf.contrib.rnn.MultiRNNCell(lstm_cells)\n",
    "\n",
    "w = tf.get_variable('w',shape=[num_nodes[-1], 1], initializer=tf.contrib.layers.xavier_initializer())\n",
    "b = tf.get_variable('b',initializer=tf.random_uniform([1],-0.1,0.1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Create cell state and hidden state variables to maintain the state of the LSTM\n",
    "c, h = [],[]\n",
    "initial_state = []\n",
    "for li in range(n_layers):\n",
    "  c.append(tf.Variable(tf.zeros([batch_size, num_nodes[li]]), trainable=False))\n",
    "  h.append(tf.Variable(tf.zeros([batch_size, num_nodes[li]]), trainable=False))\n",
    "  initial_state.append(tf.contrib.rnn.LSTMStateTuple(c[li], h[li]))\n",
    "\n",
    "# Do several tensor transofmations, because the function dynamic_rnn requires the output to be of\n",
    "# a specific format. Read more at: https://www.tensorflow.org/api_docs/python/tf/nn/dynamic_rnn\n",
    "all_inputs = tf.concat([tf.expand_dims(t,0) for t in train_inputs],axis=0)\n",
    "\n",
    "# all_outputs is [seq_length, batch_size, num_nodes]\n",
    "all_lstm_outputs, state = tf.nn.dynamic_rnn(\n",
    "    drop_multi_cell, all_inputs, initial_state=tuple(initial_state),\n",
    "    time_major = True, dtype=tf.float32)\n",
    "\n",
    "all_lstm_outputs = tf.reshape(all_lstm_outputs, [batch_size*num_unrollings,num_nodes[-1]])\n",
    "\n",
    "all_outputs = tf.nn.xw_plus_b(all_lstm_outputs,w,b)\n",
    "\n",
    "split_outputs = tf.split(all_outputs,num_unrollings,axis=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Defining training Loss\n",
      "Learning rate decay operations\n",
      "TF Optimization operations\n",
      "\tAll done\n"
     ]
    }
   ],
   "source": [
    "# When calculating the loss you need to be careful about the exact form, because you calculate\n",
    "# loss of all the unrolled steps at the same time\n",
    "# Therefore, take the mean error or each batch and get the sum of that over all the unrolled steps\n",
    "\n",
    "print('Defining training Loss')\n",
    "loss = 0.0\n",
    "with tf.control_dependencies([tf.assign(c[li], state[li][0]) for li in range(n_layers)]+\n",
    "                             [tf.assign(h[li], state[li][1]) for li in range(n_layers)]):\n",
    "  for ui in range(num_unrollings):\n",
    "    loss += tf.reduce_mean(0.5*(split_outputs[ui]-train_outputs[ui])**2)\n",
    "\n",
    "print('Learning rate decay operations')\n",
    "global_step = tf.Variable(0, trainable=False)\n",
    "inc_gstep = tf.assign(global_step,global_step + 1)\n",
    "tf_learning_rate = tf.placeholder(shape=None,dtype=tf.float32)\n",
    "tf_min_learning_rate = tf.placeholder(shape=None,dtype=tf.float32)\n",
    "\n",
    "learning_rate = tf.maximum(\n",
    "    tf.train.exponential_decay(tf_learning_rate, global_step, decay_steps=1, decay_rate=0.5, staircase=True),\n",
    "    tf_min_learning_rate)\n",
    "\n",
    "# Optimizer.\n",
    "print('TF Optimization operations')\n",
    "optimizer = tf.train.AdamOptimizer(learning_rate)\n",
    "gradients, v = zip(*optimizer.compute_gradients(loss))\n",
    "gradients, _ = tf.clip_by_global_norm(gradients, 5.0)\n",
    "optimizer = optimizer.apply_gradients(\n",
    "    zip(gradients, v))\n",
    "\n",
    "print('\\tAll done')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Defining prediction related TF functions\n",
      "\tAll done\n"
     ]
    }
   ],
   "source": [
    "print('Defining prediction related TF functions')\n",
    "\n",
    "sample_inputs = tf.placeholder(tf.float32, shape=[1,D])\n",
    "\n",
    "# Maintaining LSTM state for prediction stage\n",
    "sample_c, sample_h, initial_sample_state = [],[],[]\n",
    "for li in range(n_layers):\n",
    "  sample_c.append(tf.Variable(tf.zeros([1, num_nodes[li]]), trainable=False))\n",
    "  sample_h.append(tf.Variable(tf.zeros([1, num_nodes[li]]), trainable=False))\n",
    "  initial_sample_state.append(tf.contrib.rnn.LSTMStateTuple(sample_c[li],sample_h[li]))\n",
    "\n",
    "reset_sample_states = tf.group(*[tf.assign(sample_c[li],tf.zeros([1, num_nodes[li]])) for li in range(n_layers)],\n",
    "                               *[tf.assign(sample_h[li],tf.zeros([1, num_nodes[li]])) for li in range(n_layers)])\n",
    "\n",
    "sample_outputs, sample_state = tf.nn.dynamic_rnn(multi_cell, tf.expand_dims(sample_inputs,0),\n",
    "                                   initial_state=tuple(initial_sample_state),\n",
    "                                   time_major = True,\n",
    "                                   dtype=tf.float32)\n",
    "\n",
    "with tf.control_dependencies([tf.assign(sample_c[li],sample_state[li][0]) for li in range(n_layers)]+\n",
    "                              [tf.assign(sample_h[li],sample_state[li][1]) for li in range(n_layers)]):  \n",
    "  sample_prediction = tf.nn.xw_plus_b(tf.reshape(sample_outputs,[1,-1]), w, b)\n",
    "\n",
    "print('\\tAll done')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Initialized\n",
      "Average loss at step 1: 2.781574\n",
      "\tTest MSE: 0.01085\n",
      "\tFinished Predictions\n",
      "Average loss at step 2: 0.142638\n",
      "\tTest MSE: 0.00551\n",
      "\tFinished Predictions\n",
      "Average loss at step 3: 0.081698\n",
      "\tTest MSE: 0.00409\n",
      "\tFinished Predictions\n",
      "Average loss at step 4: 0.070348\n",
      "\tTest MSE: 0.00331\n",
      "\tFinished Predictions\n",
      "Average loss at step 5: 0.060920\n",
      "\tTest MSE: 0.00296\n",
      "\tFinished Predictions\n",
      "Average loss at step 6: 0.057505\n",
      "\tTest MSE: 0.00245\n",
      "\tFinished Predictions\n",
      "Average loss at step 7: 0.057221\n",
      "\tTest MSE: 0.00237\n",
      "\tFinished Predictions\n",
      "Average loss at step 8: 0.053376\n",
      "\tTest MSE: 0.00255\n",
      "\tFinished Predictions\n",
      "Average loss at step 9: 0.053780\n",
      "\tTest MSE: 0.00241\n",
      "\tFinished Predictions\n",
      "Average loss at step 10: 0.051061\n",
      "\tDecreasing learning rate by 0.5\n",
      "\tTest MSE: 0.00273\n",
      "\tFinished Predictions\n",
      "Average loss at step 11: 0.046651\n",
      "\tTest MSE: 0.00236\n",
      "\tFinished Predictions\n",
      "Average loss at step 12: 0.045406\n",
      "\tTest MSE: 0.00259\n",
      "\tFinished Predictions\n",
      "Average loss at step 13: 0.040161\n",
      "\tTest MSE: 0.00266\n",
      "\tFinished Predictions\n",
      "Average loss at step 14: 0.041807\n",
      "\tTest MSE: 0.00232\n",
      "\tFinished Predictions\n",
      "Average loss at step 15: 0.042201\n",
      "\tTest MSE: 0.00251\n",
      "\tFinished Predictions\n",
      "Average loss at step 16: 0.041839\n",
      "\tTest MSE: 0.00240\n",
      "\tFinished Predictions\n",
      "Average loss at step 17: 0.043388\n",
      "\tDecreasing learning rate by 0.5\n",
      "\tTest MSE: 0.00233\n",
      "\tFinished Predictions\n",
      "Average loss at step 18: 0.042720\n",
      "\tTest MSE: 0.00245\n",
      "\tFinished Predictions\n",
      "Average loss at step 19: 0.041457\n",
      "\tTest MSE: 0.00252\n",
      "\tFinished Predictions\n",
      "Average loss at step 20: 0.040291\n",
      "\tDecreasing learning rate by 0.5\n",
      "\tTest MSE: 0.00234\n",
      "\tFinished Predictions\n",
      "Average loss at step 21: 0.040911\n",
      "\tTest MSE: 0.00235\n",
      "\tFinished Predictions\n",
      "Average loss at step 22: 0.040177\n",
      "\tTest MSE: 0.00241\n",
      "\tFinished Predictions\n",
      "Average loss at step 23: 0.040117\n",
      "\tDecreasing learning rate by 0.5\n",
      "\tTest MSE: 0.00249\n",
      "\tFinished Predictions\n",
      "Average loss at step 24: 0.038937\n",
      "\tTest MSE: 0.00244\n",
      "\tFinished Predictions\n",
      "Average loss at step 25: 0.039151\n",
      "\tTest MSE: 0.00241\n",
      "\tFinished Predictions\n",
      "Average loss at step 26: 0.039702\n",
      "\tDecreasing learning rate by 0.5\n",
      "\tTest MSE: 0.00242\n",
      "\tFinished Predictions\n",
      "Average loss at step 27: 0.039485\n",
      "\tTest MSE: 0.00238\n",
      "\tFinished Predictions\n",
      "Average loss at step 28: 0.041876\n",
      "\tTest MSE: 0.00248\n",
      "\tFinished Predictions\n",
      "Average loss at step 29: 0.041091\n",
      "\tDecreasing learning rate by 0.5\n",
      "\tTest MSE: 0.00244\n",
      "\tFinished Predictions\n",
      "Average loss at step 30: 0.040064\n",
      "\tTest MSE: 0.00238\n",
      "\tFinished Predictions\n"
     ]
    }
   ],
   "source": [
    "epochs = 30\n",
    "valid_summary = 1 # Interval you make test predictions\n",
    "\n",
    "n_predict_once = 50 # Number of steps you continously predict for\n",
    "\n",
    "train_seq_length = train_data.size # Full length of the training data\n",
    "\n",
    "train_mse_ot = [] # Accumulate Train losses\n",
    "test_mse_ot = [] # Accumulate Test loss\n",
    "predictions_over_time = [] # Accumulate predictions\n",
    "\n",
    "session = tf.InteractiveSession()\n",
    "\n",
    "tf.global_variables_initializer().run()\n",
    "\n",
    "# Used for decaying learning rate\n",
    "loss_nondecrease_count = 0\n",
    "loss_nondecrease_threshold = 2 # If the test error hasn't increased in this many steps, decrease learning rate\n",
    "\n",
    "print('Initialized')\n",
    "average_loss = 0\n",
    "\n",
    "# Define data generator\n",
    "data_gen = DataGeneratorSeq(train_data,batch_size,num_unrollings)\n",
    "\n",
    "x_axis_seq = []\n",
    "\n",
    "# Points you start your test predictions from\n",
    "test_points_seq = np.arange(11000,12000,50).tolist()\n",
    "\n",
    "for ep in range(epochs):       \n",
    "\n",
    "    # ========================= Training =====================================\n",
    "    for step in range(train_seq_length//batch_size):\n",
    "\n",
    "        u_data, u_labels = data_gen.unroll_batches()\n",
    "\n",
    "        feed_dict = {}\n",
    "        for ui,(dat,lbl) in enumerate(zip(u_data,u_labels)):            \n",
    "            feed_dict[train_inputs[ui]] = dat.reshape(-1,1)\n",
    "            feed_dict[train_outputs[ui]] = lbl.reshape(-1,1)\n",
    "\n",
    "        feed_dict.update({tf_learning_rate: 0.0001, tf_min_learning_rate:0.000001})\n",
    "\n",
    "        _, l = session.run([optimizer, loss], feed_dict=feed_dict)\n",
    "\n",
    "        average_loss += l\n",
    "\n",
    "    # ============================ Validation ==============================\n",
    "    if (ep+1) % valid_summary == 0:\n",
    "\n",
    "      average_loss = average_loss/(valid_summary*(train_seq_length//batch_size))\n",
    "\n",
    "      # The average loss\n",
    "      if (ep+1)%valid_summary==0:\n",
    "        print('Average loss at step %d: %f' % (ep+1, average_loss))\n",
    "\n",
    "      train_mse_ot.append(average_loss)\n",
    "\n",
    "      average_loss = 0 # reset loss\n",
    "\n",
    "      predictions_seq = []\n",
    "\n",
    "      mse_test_loss_seq = []\n",
    "\n",
    "      # ===================== Updating State and Making Predicitons ========================\n",
    "      for w_i in test_points_seq:\n",
    "        mse_test_loss = 0.0\n",
    "        our_predictions = []\n",
    "\n",
    "        if (ep+1)-valid_summary==0:\n",
    "          # Only calculate x_axis values in the first validation epoch\n",
    "          x_axis=[]\n",
    "\n",
    "        # Feed in the recent past behavior of stock prices\n",
    "        # to make predictions from that point onwards\n",
    "        for tr_i in range(w_i-num_unrollings+1,w_i-1):\n",
    "          current_price = all_mid_data[tr_i]\n",
    "          feed_dict[sample_inputs] = np.array(current_price).reshape(1,1)    \n",
    "          _ = session.run(sample_prediction,feed_dict=feed_dict)\n",
    "\n",
    "        feed_dict = {}\n",
    "\n",
    "        current_price = all_mid_data[w_i-1]\n",
    "\n",
    "        feed_dict[sample_inputs] = np.array(current_price).reshape(1,1)\n",
    "\n",
    "        # Make predictions for this many steps\n",
    "        # Each prediction uses previous prediciton as it's current input\n",
    "        for pred_i in range(n_predict_once):\n",
    "\n",
    "          pred = session.run(sample_prediction,feed_dict=feed_dict)\n",
    "\n",
    "          our_predictions.append(np.asscalar(pred))\n",
    "\n",
    "          feed_dict[sample_inputs] = np.asarray(pred).reshape(-1,1)\n",
    "\n",
    "          if (ep+1)-valid_summary==0:\n",
    "            # Only calculate x_axis values in the first validation epoch\n",
    "            x_axis.append(w_i+pred_i)\n",
    "\n",
    "          mse_test_loss += 0.5*(pred-all_mid_data[w_i+pred_i])**2\n",
    "\n",
    "        session.run(reset_sample_states)\n",
    "\n",
    "        predictions_seq.append(np.array(our_predictions))\n",
    "\n",
    "        mse_test_loss /= n_predict_once\n",
    "        mse_test_loss_seq.append(mse_test_loss)\n",
    "\n",
    "        if (ep+1)-valid_summary==0:\n",
    "          x_axis_seq.append(x_axis)\n",
    "\n",
    "      current_test_mse = np.mean(mse_test_loss_seq)\n",
    "\n",
    "      # Learning rate decay logic\n",
    "      if len(test_mse_ot)>0 and current_test_mse > min(test_mse_ot):\n",
    "          loss_nondecrease_count += 1\n",
    "      else:\n",
    "          loss_nondecrease_count = 0\n",
    "\n",
    "      if loss_nondecrease_count > loss_nondecrease_threshold :\n",
    "            session.run(inc_gstep)\n",
    "            loss_nondecrease_count = 0\n",
    "            print('\\tDecreasing learning rate by 0.5')\n",
    "\n",
    "      test_mse_ot.append(current_test_mse)\n",
    "      print('\\tTest MSE: %.5f'%np.mean(mse_test_loss_seq))\n",
    "      predictions_over_time.append(predictions_seq)\n",
    "      print('\\tFinished Predictions')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "best_prediction_epoch = 28 # replace this with the epoch that you got the best results when running the plotting code\n",
    "\n",
    "plt.figure(figsize = (18,18))\n",
    "plt.subplot(2,1,1)\n",
    "plt.plot(range(df.shape[0]),all_mid_data,color='b')\n",
    "\n",
    "# Plotting how the predictions change over time\n",
    "# Plot older predictions with low alpha and newer predictions with high alpha\n",
    "start_alpha = 0.25\n",
    "alpha  = np.arange(start_alpha,1.1,(1.0-start_alpha)/len(predictions_over_time[::3]))\n",
    "for p_i,p in enumerate(predictions_over_time[::3]):\n",
    "    for xval,yval in zip(x_axis_seq,p):\n",
    "        plt.plot(xval,yval,color='r',alpha=alpha[p_i])\n",
    "\n",
    "plt.title('Evolution of Test Predictions Over Time',fontsize=18)\n",
    "plt.xlabel('Date',fontsize=18)\n",
    "plt.ylabel('Mid Price',fontsize=18)\n",
    "plt.xlim(11000,12500)\n",
    "\n",
    "plt.subplot(2,1,2)\n",
    "\n",
    "# Predicting the best test prediction you got\n",
    "plt.plot(range(df.shape[0]),all_mid_data,color='b')\n",
    "for xval,yval in zip(x_axis_seq,predictions_over_time[best_prediction_epoch]):\n",
    "    plt.plot(xval,yval,color='r')\n",
    "\n",
    "plt.title('Best Test Predictions Over Time',fontsize=18)\n",
    "plt.xlabel('Date',fontsize=18)\n",
    "plt.ylabel('Mid Price',fontsize=18)\n",
    "plt.xlim(11000,12500)\n",
    "plt.show()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.5.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}