Exercise_10_Optimization.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "from torch.utils.data import DataLoader, TensorDataset\n",
    "import torch.optim as optim\n",
    "from sklearn.preprocessing import MinMaxScaler\n",
    "from utils.util import data_shift\n",
    "import hiplot as hip\n",
    "from matplotlib import pyplot as plt\n",
    "import sys\n",
    "from copy import copy\n",
    "import warnings\n",
    "warnings.filterwarnings(\"ignore\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Before starting this exercise, you need to install the library firstly:\n",
    "\n",
    "- hiplot: conda install -c conda-forge hiplot"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Task 1: Neural Network Implementation and Optimization\n",
    "\n",
    "In this exercise you need to implement a fully-connected neural network using PyTorch to forecast daily minimum temperatures."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**a)** Import temperature time-series dataset **daily-minimum-temperatures.csv** in the data folder. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "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>Temp</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Date</th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1981-01-01</th>\n",
       "      <td>20.7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1981-01-02</th>\n",
       "      <td>17.9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1981-01-03</th>\n",
       "      <td>18.8</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1981-01-04</th>\n",
       "      <td>14.6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1981-01-05</th>\n",
       "      <td>15.8</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1981-01-06</th>\n",
       "      <td>15.8</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1981-01-07</th>\n",
       "      <td>15.8</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1981-01-08</th>\n",
       "      <td>17.4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1981-01-09</th>\n",
       "      <td>21.8</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1981-01-10</th>\n",
       "      <td>20.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "            Temp\n",
       "Date            \n",
       "1981-01-01  20.7\n",
       "1981-01-02  17.9\n",
       "1981-01-03  18.8\n",
       "1981-01-04  14.6\n",
       "1981-01-05  15.8\n",
       "1981-01-06  15.8\n",
       "1981-01-07  15.8\n",
       "1981-01-08  17.4\n",
       "1981-01-09  21.8\n",
       "1981-01-10  20.0"
      ]
     },
     "execution_count": 69,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data = pd.read_csv('./data/daily-minimum-temperatures.csv',header=0, index_col=0)\n",
    "data.head(10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAzYAAAH+CAYAAABOa5uxAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjguMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8g+/7EAAAACXBIWXMAAA9hAAAPYQGoP6dpAADroklEQVR4nOydd5zUZP7HP5nZDuwuvS69gxTpIggCCqJi770r6tlP7PXw7mynnr3ws6Nn70hVkCIgTaS3pdctlG0z+f2xzGySSc+T5MnM930vzp1MJvnmydO+z7c8giiKIgiCIAiCIAiCIAJMyG8BCIIgCIIgCIIgnEKKDUEQBEEQBEEQgYcUG4IgCIIgCIIgAg8pNgRBEARBEARBBB5SbAiCIAiCIAiCCDyk2BAEQRAEQRAEEXhIsSEIgiAIgiAIIvCk+S2Akmg0iu3bt6NOnToQBMFvcQiCIAiCIAiC8AlRFFFaWopmzZohFNK3yXCn2Gzfvh0FBQV+i0EQBEEQBEEQBCcUFhaiRYsWuudwp9jUqVMHQLXwubm5PktDEARBEARBEIRflJSUoKCgIK4j6MGdYhNzP8vNzSXFhiAIgiAIgiAIUyEqlDyAIAiCIAiCIIjAQ4oNQRAEQRAEQRCBhxQbgiAIgiAIgiACDyk2BEEQBEEQBEEEHlJsCIIgCIIgCIIIPKTYEARBEARBEAQReEixIQiCIAiCIAgi8JBiQxAEQRAEQRBE4CHFhiAIgiAIgiCIwEOKDUEQBEEQBEEQgYcUG4IgCIIgCIIgAg8pNgRBEARBEARBBB5SbAiCIAiCIAiCCDyk2BCuEImKeG/eZqzZVeq3KARBEARBEEQKkOa3AERy8tGCLXjwyxUAgE1PjfVZGoIgCIIgCCLZIYsN4QpLC4v8FoEgCIIgCIJIIUixIQiCIAiCIAgi8JBiQ7iC6LcABEEQBEEQREpBig1BEARBEARBEIGHFBvCFUQy2RAEQRAEQRAeQooN4QoiOaMRBEEQBEEQHkKKDUEQBEEQBEEQgYcUG8IdyGBDEARBEARBeAgpNgRBEARBEARBBB5SbAhXIIMNQRAEQRAE4SWk2BCuIFJaNIIgCIIgCMJDSLEhXIHUGoIgCIIgCMJLSLEhCIIgCIIgCCLwWFJsJk6ciH79+qFOnTpo1KgRzjjjDKxevVp2zrBhwyAIguzfDTfcwFRogi/2HizHmS/PweTft8SPkScaQRAEQRAE4SWWFJtZs2Zh/PjxmDdvHn7++WdUVlbipJNOwqFDh2TnXXvttdixY0f837/+9S+mQhN88cyU1fhjSxH+/tlyv0UhCIIgCIIgUpQ0Kyf/+OOPss+TJk1Co0aNsGjRIgwdOjR+PCcnB02aNGEjISdURqLYXVqO5vnZ8WOHyqtwuCKChnUyfZTMfw6WRxKOkcGGIAiCIAiC8BJHMTbFxcUAgHr16smOf/DBB2jQoAG6d++OCRMm4PDhw5rXKC8vR0lJiewfj1zw+jwMfmo65m3YFz/W89Ep6PfkVOw7WO6jZHxCWdEIgiAIgiAIL7Gt2ESjUdx2220YPHgwunfvHj9+0UUX4f3338eMGTMwYcIEvPfee7jkkks0rzNx4kTk5eXF/xUUFNgVyVUWbT4AAJj8e2H8WFW0evK+bFuxLzLxguC3AARBEARBEETKY8kVTcr48eOxYsUKzJ49W3b8uuuui/99zDHHoGnTphgxYgTWr1+Pdu3aJVxnwoQJuOOOO+KfS0pKuFVuAMgsNkQ1arYZstcQBEEQBEEQXmJLsbn55pvx7bff4pdffkGLFi10zx0wYAAAYN26daqKTWZmJjIzgxOjsqO4zG8RCIIgCIIgCIJQYEmxEUURt9xyC7744gvMnDkTbdq0MfzNkiVLAABNmza1JSARUMhkQxAEQRAEQXiIJcVm/Pjx+PDDD/HVV1+hTp062LlzJwAgLy8P2dnZWL9+PT788EOccsopqF+/PpYtW4bbb78dQ4cORY8ePVx5AMJ/KMaGIAiCIAiC8BtLis0rr7wCoHoTTinvvPMOrrjiCmRkZGDq1Kl4/vnncejQIRQUFODss8/GAw88wExggj/UY2zIZEMQBEEQBEF4h2VXND0KCgowa9YsRwIRBEEQBEEQBEFYxdE+NgQBqLui0TY2BEEQBEEQhJeQYkMQBOEyu0vLcOqLv+LD+Vv8FoUgCIIgkhZSbBhAwfOJkMWGIGr494+rsWJbCe77YrnfohAEQRBE0kKKDeEKlDyAIGo4XBnxWwSCCDyiKGL9noOIRGl8IQhCHVJsCIIg3IbmYQThmHfnbsaIZ2bhrk+X+i0KQRCcQooN4Zivl25POEauaARBEARLXpy+FgDwxR/bfJaEIAheIcWGcAXSawhCAgXiEYRjaMHMGtuLjuA/U9di78Fyv0UhCM+wtI8NQZiFBiCCkEDtgSAIj7n4zfnYuPcQ5qzfi0+uH+S3OAThCWSxIQiCIAiCSDI27j0EAFiwcb/PkhCEd5BiQ7gELVETBEEQ7KBRhSAII0ixYYAgkAM9QRAEQRAEQfgJKTYMECmgJAEqEoIgCIIgCMJLSLEhCIIgCIIgCCLwkGLDAHJFS4QMNgRBEAQP/LZ+L0Y8MxPzN+zzWxSCIFyGFBvCFcg9jyAIgmCJ3XHlojfmY/2eQzj/9XmMJSIIgjdIsSEIgiAIgiAIIvCQYkO4AtlrCIIgCIIgCC8hxYYgCMJlRFL1CcIx1IrsQWHARCpBig3hChRiQxAEQThlxbZiPPfzGpRVRvwWhSCIAJDmtwAEQRAEQRBqnPribABAlFbLbEMGGyKVIIsN4Qo0BBEEQRCs+GtHqd8iEAQRAEixYQCthhAEQRCEewgCuTjbhfbaI1IJUmwI5qzdVYpf1uzxWwyCIAiCSHlIrSFSCVJsCOaMeu4Xv0UgCIIgCAKUFY1ILUixIQiCIAiCe0TyRSMIwgBSbAjCI8oqIzj9pdl47JuVfotCeAzNxwiC8AuBnNGIFIIUG4LwiB9X7MSyrcV4e85Gv0UhkoDJv2/B/xZt9VsMgvAEq1Pz39btxfNT1yAapVUFgkglaB8bm5BJnLBKZSTqtwhEknDgUAX+/tlyAMCpPZoiKz3ss0QEwRcXvTkfAFBQN8dnSTiADDZECkEWG4IgiIBxqKIq/jdtXEikAnYD4DfuPcRWkABCeg2RSpBiYxOaS1Qz8fu/LJ3/0YItOOHfM1JysKEqQ7BC2v+EKOURkSLY6UMjNFhTVjQipSDFhgFPT1mN8qqI32L4wmu/bLB0/oTPl2PzvsN46KsVLkkULLYeOIyJ3/+F7UVH/BaFCBBkpSEIc0QoxoYwQVUkimenrMa8Dfv8FoVwCCk2NpF2lcu2FuOzRdt8kyWIVFRRvAkAnPribLz2ywZcNel3v0UhAoR0rkY6DkFoUxWhBmInK9qWfYexs7jMBWn45OPfC/HC9HW44PV5fotCOISSBzBi/6Fyv0UgeEdlfC06XAkAWLWz1GNhiCBDq9BEqiFAsOWLFonSIppVio9UYui/ZwAANj011mdpvCEV3eOTFbLY2CAaFXH+a3NlxygrkTVS0edXpCgbghHSrIxUrwhCm0paBLA83m47QK7RRHAhxcYEP67YIfu8amcpFm4+IDtGK6jWoA3DCMI+0t5G6Yr25/ZiTPz+L5SUVXoqE0HwSIRc0SyPttLFEtraggga5IpmghveXyz7rBa4S5lXrJGKFhsidWHdPUgXUpT90dgXZgOodnP85zk92N6YIHxCsOeJhipJW0nVcUew+OCiIoYvFcqNpnDJA1lsbKDWyGlViDCCOk6CFVJl5r15m1XP+WtniVfiEAS3UIyNMygDIxE0SLGxgZobFVlsrJEKK0BKqIakLqzru7S7eXv2RrYXJ4gkQmqxSdU9nyy7okktNkwl8Yd35mzE8f+cjsL9hzXPoVjF5IEUGxuoWmwoxkaXX9fuwb9/WuW3GNzwxLcrse+gPJNequ6FlAqwXveQrqIeKlevN6k5hSOSGTvxHtKxOWXbhIMHTwaLzaPfrMTWA0fw+Lcr/RaF8ACKsbEBKTbWufStBbLPqZg8QDo+vDl7IzYrVo/embMJN5zQzmOpiCAi7W6OVKauQhyNiliytQhdmuQiO4MyUyYzdo0tVTQ2O0oesGRLEfq3qWc5TodH9OZpSaC/EUchiw0jyBXNGknQRzpmxbZi2efN+yiPPmEOylRUzXvzNuOsl3/DZW/P91sUglMilDzAMtLu5fzX5+G75Tu0TyYIziDFxgZq1oYorQoRDgmHaNQlzGHKPSQFZnEfzt8CAPh90wGDM4mgI0BwnhUtBT0FABtZ0RSfv1m6nZ0wBOEypNjYQK2PqKiK4qkfVmH22r3eC0QEAmVwonJumhai5kiYg9ZRCMIcVRFJVrTU1Gsso7QIp2rSBSKYUIwNIz5csAWVERGvzlqPTU+N9Vsc7kkGf12rKBUZ5ar7Fp2MLQQhhTzRCMIcFGPj3HhLig0RJGiJ2AZqTbyS9rGxBHWTiavu01ftRintFk+YwEyyEt7bGMUJJeJGmaR6OVNWNDvJAxxegCB8hBQbG9DiBcGGxAnH7tJylfMIQk7QJ6sb9hzEoInTMWkO7cET46sl2zDgH9OwpLCI2TXHf7gYZ/x3Tkpn7ayi5AHWY2wU1SUVLDZB71OJGkixIXwhBfrJBJTdZgrPNYgU5+Gv/8TOkjI88g3tKxHjbx8vwe7Sctz4/iJm1/xu2Q4s3VqMpVuLmF3TNwR7LpiRaE2MTaomD7COMsbGJzEIwgak2BC+QP2kemYrWjQiWMHz4kEVue5qQjEhbJHWNZ7bhJtYdkVLUouN3mNQq0seKHmADqIoqq6q0+STsIWi4lCK8NRBmRHP+fWIZMWN8SUZpqXJ8AxBQTk0JYleQ3O3FIEsNjq8PHM92t33fcJxahvOmbF6D2au3u23GJ6irDdUjwi70ABNWCFZslA6XSBIjlKwjtXXr/QmSAUXPupTkwdSbHTQ6gxMbY5HGHLFO7/7LYK/UDUibMLaAuQ1QZc/aCT/tNQcyaLgWUf9uSsjUTz53UrMWrNHdlw5x6EYG33mrNuLx79difKqiN+iECDFRhetVQpJLCJB2IYUZMJNUmEukrLzVELG0sIi7D1onFEyVatLmoZm8sG8zXjj1424/O0FAKrdo+dv2IfSsirZeckSY2OW4sOVWLR5v+lMaRe/OR9vzd6I//ttk7uCuci2oiNYub3EbzGYQDE2OmitUtCElGBBqoTYiKIIQRDi/yWcQ10QYYR0UpbMzW7xlgM46+XfAEB1c2xZW0nictAjrDGZ2bj3kOzz23M24onv/ko4L5QkS+Bm28Go52Zhd2k5Xr3kWIzu3tT09Qv3H7Epmf8Mfmo6AGDOvSeieX62z9I4I0mqqzsk82BAeI9yMnqkMvnN1hv2HET/f0zDbR//gX5PTsUnCwv9FikpMKPX8KxEkmKmB5vCkZZxMsRIaNXnOWv36v6O3B61FZOKiNz9ZPLv6v0zz30JK6T1JLaf3E9/7rJ0jWQopuVbi/0WwTGk2Oig6YpGozJhA3Nm7eSqW498sxJ7Ssvx5ZLt2HuwAvf8b5nfIiUFZupSEoyxhAOiSWixUav2VnrMJCkGy6RpaDYVVfLSi2j0KxRjY45kKKYjlVXGJ3EOKTY6aA0GpNcQhDloN2d3oFJNHaoiUcxdvw+HK6xNOMzWkUWb92P/oQrrgnkMi0ljKlgeYmwvqnGL0lJMlBabiIZ/dDJY/IxQVZotjl/JUL8OlQffk8SSYjNx4kT069cPderUQaNGjXDGGWdg9erVsnPKysowfvx41K9fH7Vr18bZZ5+NXbusmfN4QauSksVGG5rIakMlk7owbxYBr0wBF99TXpqxDhe+MQ/XvrvQ0u/M1Llf1uzB2a/MjfvXBxGj55S55AV/3mma4yTvVCvGprJKrthozW1SwWJDfVI1RypSTLGZNWsWxo8fj3nz5uHnn39GZWUlTjrpJBw6VBOAdvvtt+Obb77Bp59+ilmzZmH79u0466yzmAvuBVptmRqANqkSEE+YIxlWsHjETNwAFX1y8P68LQCAOev2WfqdGVe06auq9xILSryfWq23EkOTqk1CK6tZpcJio5XxNVX78VR87sokSPtrKSvajz/+KPs8adIkNGrUCIsWLcLQoUNRXFyMt956Cx9++CFOPPFEAMA777yDLl26YN68eRg4cGDCNcvLy1FeXpOmsaSEn3RzWqsUZJWoRq0ctEzZhDmSrWql3rBAeImyvWzaewiv/7oBNwxth5b1c/wRijnOOwUtV6JUmLclWZdqCy2LTZVivNYav1Mt3XMM6RxHFEU8M2UNOjSujXG9mquenwzFFE2COZyjGJvi4ursCfXq1QMALFq0CJWVlRg5cmT8nM6dO6Nly5aYO3eu6jUmTpyIvLy8+L+CggInIjFF2xXNY0E4RW0STm562qRi0SRDR88jga9LLsl/0Rvz8OH8Lbj07fnu3CBABL6OKGAR85qKK/CAtmKjRKnoxEgFVzQj5m3Yj5dmrMPfPl6ieU4yxCJFgm+wsa/YRKNR3HbbbRg8eDC6d+8OANi5cycyMjKQn58vO7dx48bYuXOn6nUmTJiA4uLi+L/CQn7SwWr1gcmg0bJArRTIYqNNKpZM8Lt5PjEzmUuGQdYq24vLAACb9x32WRJ22FVQpC5ayTyfp6xoxoSO7iNmhNY5yVx/YhgVjzTBRjJ77USSwBXNtmIzfvx4rFixAh9//LEjATIzM5Gbmyv7xwtaqzvJW6WtoeqKlsQNnrBOqq6Qug21suSFVRcqXWPSaoZJofwaFFiqbFSqx5LCIpzz6lzDCbl2uucULTgJ0iI47qnpePirFf4J4yLJMIezpdjcfPPN+PbbbzFjxgy0aNEifrxJkyaoqKhAUVGR7Pxdu3ahSZMmjgT1A62mTO5W1aiVAlmzCCnkwmCN12atx6Q5Gw3PC/qKoV+bJk6asxGvzVrvy73tYrekZBP6ZFBgAFuFEeyWwo5Fmw9gV0m57jma6Z5TQrExX1N2FJfh/+ZuTjieDMWk5Y4YJCwpNqIo4uabb8YXX3yB6dOno02bNrLv+/Tpg/T0dEybNi1+bPXq1diyZQsGDRrERmIPoX1s9FErB3JF0ybok1F7JEFP7xF7Sssx8YdVeOSblSgzyFKVijXJKWWVETzyzUpM/GEVdpeW+S2O60jrSDJMuLSwuNOIS1IEA6N6oLUwSQtU5kiGYkqGxWlLWdHGjx+PDz/8EF999RXq1KkTj5vJy8tDdnY28vLycPXVV+OOO+5AvXr1kJubi1tuuQWDBg1SzYjGO1rm15Scn6qgZrlyYsYURTFFVoZSB6evM5XqhFSZYWIVTo1iM420TMsrg+NHbndBRJQ8olZVCFLT0tx+wah4UnQfGzuksiuaUT0yUwLJUEwplzzglVdeQXFxMYYNG4amTZvG/02ePDl+znPPPYdTTz0VZ599NoYOHYomTZrg888/Zy64F5ArmnXsxp3d9vEfGPb0zKTYHMoJyVaz1NqQ2YnanHV70fPRKfh22Xa2QgUAK5sOSpGttnFcmVh1ockwkTDCtiuamb2ObF7bL5y6MAbteVlj9Pza+9gwF8UnrD0Ix10oU6RjcjIkD7BksTEzIcnKysJ///tf/Pe//7UtFC9oNea/f7bMW0E4RdUVzeaM5csl1ZPXqX/twmk9mzkRi+AItTZUdLgSdWtlGP724jerU/be/OEfOLVHsOuEmVYhLSsT+YtUj27eX5MNLBk2WiPsIyaZpUI7mY9B8gDZNRgKFESOPv+sNXvw88qdCS6vFRrL9aliNdfDTBEEtZykfUUyJA+wpNikGlqVdEdx8vtnm0FtQHHqnxnQfsEUSdBfWEYtaDmZ37FXaNUl8oUPPspXaz/ds5TkrRip2K/aJdYfX/72Aku/S9V+xeixI1FRtkdQUItJ2oSSIU7a0QadyU5QK6lXqA0oyZBRg2CHmhKTNBmaLMD6ibVamdQXnucgUL8lS4XJsBmXaZ4WGYzqKwtRU7HvMYte+StjbHjuW+yi1lyMnrKiSmHhCkj1Ur4/uTdWQB5CB1JsdAiqWdEr1Bq908xfNPAkF6pNKAVfsTlXtJqCMWpHWl9LyzsJ5x4pie3kAQF6/+/M2Yiej03Bn9uLNc8RYW/ySfvY1KD3/HqLklKLzfRVu9Dj0Sn4cYX6puupRBDjrTfsOYhjn/gZ/52xLn4s2cYKUmx0SFXzq1nUBlyn7SOZBx4zga8B7Cd1UVVUk+wZeUK6spoMLgUEq+QBfNeFR79ZidKyKtz72XLLv7XSZybx8GIKvec3219cNWkhDpZX4Yb3F7ERihOMx+fE0uO7Vanz+LcrUXS4Ev/+aXX8mF/7irkFKTY6JPMkmwVq/WCyTcxZkpJlo6rXpEZBHCqvwozVuxPdFUwQK6G1u0pVV7G1ylBuseG3nFNzTyeb2I2xMfE73rwS9PoGJq5onD0vT1TpJBuh5qo+H1T2sUHwOIkYzNuSoYlQ8gAdglBJfUW1s6MekKhBrQUZLQxGoyJCSWAuvfbdhfht/T5cO6SN8cmQl5UoVpfDqOd+AQAsffgk5GWny75Xv4YkxoZmI4FEqfTZttiI6n/zjGGac9Vj5rOiEdroWWxSwfhrp40ofxMEpSAZ46OUkMVGhyBUUj9RG1CcDqDJXOTJ350korZCqrda/+f2YvR8dAre+GWDm2J5wm/r9wEAPl5QaOp8WbpnUZSl3dxTWi4710xdIle01CbpLKOaO3Tq/ywoSp0X6BWFXoxN0tUlRijHsiDMX9TGhWRrI6TY6EBma33sBHIaQUWeXOhZbPYeLMfdny7F4i0H4t89+d1fKC2vwpPf/+WNgF5go06Lov7PlAPqxO//giiKMisNz3oNx6L5jnLcYZE8QOsKvHW3rCZY0jKTTspTfXzRK1+9hZBkm/iyIojlombJTzbrPik2Omw7cMRvEbhGPSua52L4zu+b9qNQsjEiUYPaRCI20XjwyxX4dNFWnPXyb/HvcjLIOxY4mgHKwvmv/bIBi7cUyX6TbIOVU4JSHKzij7Te/4Y9B2sWEzib6Os9uQDBdMKaXSXlKkdJsdGzvOhbbJKHykgUM1btRmlZpaXfqS/SKSw2Aahfav1CMr1fgGJsdDlUXuW3CFyjPsgkWxPRZ9XOEpz76lwAwKanxuqeG5SJFUvU+nlRrF4d3LDnUMJ30jiSZEGAyUBuSWnZmdweqYjIfkeKTXLAOsbmxGdmAQDmTRhhXyiOUGsrAydOUz03VeJm7fQfurEXSdSXPPfzGrw8cz36tKqLz248Ln7c6AlVkyUpPgehfqkpsMmWzIUUGx3sTgxKyiqRm5V8EzQl7mRF479jkLJsq/a+C0pMpXtOMsVQzZ1zyp878eT3f6GsMjELT5hsyADU9uxQBJSrVJPM9JDsuE6SI9/xYxwNSstyU05phr5dJWUu3skebkywkmzOZgrNZ9YpCz2LDc9urVb5dNFWAMCizQdkx/XqSXlVRDW9dSAtNmqKjQ9yuAlNI3Sw25i/X7aDrSCc4krygAB0DFICJq7nqL3PB7/6U1WpAVJzEqKGKMrbl7Jc1NpehkIrTPXkAd8t24H7v1iOqgjHGp4BkaiIwxURW79Va0vFR2rcb2pnpXG3wswiM5UeQRtf7KJVJPd+vhzT/tql+t2dnyzRuV5y9CWCYG9/wt83HlD/QqVYPlu0FY9+8ye32cciap42kkPJ0ETIYqOD3dWjUAr3nvsPVXgvB0eUVUYwe+1eHNe+PsWLwLppns+hwBlmk5DIFRlRd8Km9l16WGGxSXEtcfyHiwEAPVvk47x+BYFxt5DWlq+XbrN9Hen7j9WtgxL36mQZpay8VeUzL9q8H3VzMtC2YW2WIvmOVl2fvmo3pq/arfrd4i1FOtdjIZX/VCdlUa/5dpQ3pe4iALjz06UAgN4t6+L0ns0sX9MtRFHEr2v3YkeRiqU2Sd5vDLLY6GD3XSfDHhxmUCufS96a77kcPPHAlytwzbsLcfvkJQnfmRkckmUAiWFVx0+257eCXharhM8a5SQdnHlWbLyUbHcpfy5XekjLZm+p/YUitTKWKzv8obtBp2AvYY1888GaDmnLvsM4+5W58ZgjQhse64o9REv9YuxUrXqZcFxSv2796A/L0rnJ9FW7cdnbC7BPZfGZ57HCDrSkrINdV44U0WtcmYQGreiUq/H/O+q/+9Of6ub+VCNo79MN7Bhw7WxUGBVF2QpiqruixaiZnAQPJ8Z/WcpjjXhI3pwLbLmiWXiz0sddt6fU+s0CAuu6niwT36l/qVurjNB6/CB1sXPW7dP8LkCPYQqy2OhgtzGHU0SzccPvNtX3DkqS8cM2Qffl/t+irXj4qxW2/KtF2d/6rmiqvxflk9kgDboA8PGCLXj825XMXcYCVgzMME7uErySsVM1amWGaz6YHF7enbsJT/2wyvrNOIH1OBKJBK+ueEGQNujUG1uD4qZrFrLY6GD3XStjbERRxMzVe9CtWS4a5WYxkIwP3NhMLWiY7ciOVEQwZaWxFSfoE3sllvXUgD/+XUf9q4d0aGj5t3or7KJYXYd+XbsHx3dooJlqXXq0b6u6lmXwDBX57/18OQBgRJdGOK5dA+a3CnA3YxO1+iRXfHmbiNl5RZZc0Uxe86Gv/gQAjOvVDF2a5tqQyl9YjyMVAU7AYRq9mEat40nSpyTJY8QhxUYHVskDvl66HX/7eAlyMsJY+dhoFqJxASvzdJAyclRFokizkZP43s+XYWlhkeF5ydJRxjCbPEAUxaRynSqSZJ8yW6eVMTbKycm9ny/DV0u245RjmmBYp0YJv4+K8mt0aFzHgsT8UHzY2sZ5RlRFo9WZ0ZKnepnCzISfNwO53phrV1StK5rpm+xmpPMb1uNIuUYWy1RHWc68tSez6MV3BhFyRdPB7jxLOe+NZSEJaiephdnO00hBDEpDmvz7FrS//wdMX2U9fuarJdtdkIh/QiZ7mGvfXYS+T05FKW2KCyAxK9rsdXvjdej75TtVG031QkNQWpM2rJ/gxenr0P7+H3Dp28FIbMLMEi77W0y4No/WYTckko4/ZlydpW6kqeJWbkQqWGz06p7WHEbZhnhLn24WHvsCJ5Bio4Pdl63sPJNtFd4qRgpiVDbwuCyMA/7+WbWrzA3vLY4f41neIDH1r10oOlyJn0246wUBq5OpxN/LPz/+7Ur596p7SIncT1xj6E8i3Lmnlc10ecFJzKGaRV16hMsNXHXevVZRGC6cWaxP0n0+wgHt4JlbbKqSa1HWKlrFqZzbBFUPlo0b/A4bpiHFRocTOye6e5ghIcaGhTA+cKi8Cj+u2IkjGpYmsw3AyMUocA3Jxc4rcGVhg/q1Mmz9bsW2YlPufLxRXV2MX6xycLE6+U/4TUDrkqUsVwGdeHqB0WSFR8XXVoyNhe/NTDyl41VQqxerd5t2tMAqqnjUgtmiHrdo/BuphS+oW30EPamIElJsdBjU1l4Aq7JuBzU4/rbJS3DD+4sw4fNlqt+b7TyNYnGk1+F1IKnUMMWzlpfHyQZr1HY+NqIyEsWpL87GuP/OkW0ymKxUZ0WzVk5RMTnSslp5hIywegPktR/xEi3lV/q31HUmqOOUETLrqYlVqSrpRDWgFYnVq8xOr84ol2xu9JbRKM+oKB/PAltfVBONBBdSbHSwW0eVvwtqRYm5BX2pER+SKhabtbtK0emBH/wWI2mwkwpZumJYfIRtcLnbmO1HrAwual9HE1zRgokVudMlAY3P/rwm/ncwpxeJOHkOdVc07Tp27qtzHdyNDXaUK6OfWO1upKmNzcYI8gartt8sPxsAUHjgMKMrBotY+9NecJQnvbGRV4gLyBWNMETpu5ysq/Bmn8pohV6eFY2/KclzU9fIBkephGblNWuhvvWjP7CzOFg7pUupikRx72fL8MUfW48eSXxwO0k5ZHFYNmXjncSsaObOjbG0sAjjP1wsOYfffoeVaFLF5oVpa9lcNAkpr4rgrk+XVieeOIpyXFq4+YDXYiVgp1rY9giQ/C1tK1WSAVy6Al90uAK3ffwHZq/da0NKb2HV9rMyqi02B8uS30quhgigcP9h3PrREtXvo6J84ZZni41elRBlf/M7bpiFFBsd7NZRZUe7t7SCgTT8YdbtxWiFnveGpKe8mK0jaSaX/jbtO4x7PlN3/QsCny/eho9/L8Ttk5dqnmMnrbPcT57fwUMd6/IeKq/CD8t3aH6v1mYm/rAKG/Ycsnwv3rAyKctIU29XyRJ74+QxpMX43tzN+N+irTLlj8d0z3po9cNG45CZJAnSS0hd0aTH//H9X/hyyXZc8hZ/2fXW7CrFvA01O8uzGlFjnp58j9Bs0HrG699bpOn+LCpc0YLa7+jtoRZEaB8bF1DO2xZs2u+PIC5SURV1xxWNx36BgUzhkACYdFPevC+4k9MdJqxNdmJsRMnkJKDxmYZIS+W2yUuwbvdB7XNNFGEyDFBaVFRFkZEWkllsCDlS5Xd3aXnC9zzGY9kRyUp2N62JZ1QUETra0UsVG2kZbT1wxLpwHnHSc78AAH69ZzgK6uWgklGwfyzdNYdVxTPW7dHuh6OiKHNd1Aj5455kcGGWQoqNDnbdonh2AWFBSVkljn9qOvJzzGW3MprIpoKbUarsh1BmIi2onRgb2SQsSYtS2m/oKTWAucGH515Iz0pr1H3+uGInbnh/ESaedQwy0jSSBzgRzmdYjR/SZqaeFY2/ctKrF5rpng1qumhifJGWlXSiyqPyp0fh/sMIhQQMfmo6k+vFrOPJPqcB7Clvoggs21aTRj4ZsqIlw6um5S4d7LuisZWDN6au3IWSsips2W8uoNBoRY334lJWAzv1ImUUm0pjxcaWxUbyN49xWHqYTx7AlqAOUEYT1RveXwQAmPD58gC6JVrDydNJJ6OqiQREJFROO4sOLLFTZ432jtR6JOmTS+ucNMZGtp7CaVWTvuf0tBD+77dNzK4da18lZVUY/+FiLNiYXN4n93+xPF7nNauebmyKKNtjLBn6XN5DA8xAio0OdvsxOxM33tm09xBmrN4NQNuvXQu98jhSEVEMHvyNHixksqLYBLn6lJtwgbC3MqYRAMwJy7YWYZGXwddBriQGJPGjMaNw/2FMXblLdyVd1Phb76jfFgq929veoNPEdgLSS0Q0Ymx4RSpvejgU33uGBdLQ0O+W7cB5r/mfOY8lH8zfgplrquc1WvXIyLosLe6gWrakyn9FVRTfL9+BA4eCGxtOrmg62J3QBrVy6zHs6ZkAgE+uH4TMtLCl3+qtAj727Z+4d3QXJ6K5jl4t4FER85PySrliw6p4eE4eUBmJ4vSX5mh+b1ZaK92GOVc0PvshURR1N/xj0X36PUF3myH/mgEAePuKvjixc2PVc2TjkEpxqHXLEVFEpCqCjHCIu75N65UaJg8wYXWRXkMrxoZXKiQmq/SwgDSGcWe89bVucOBQ9fYBWq9aN5uYKC8j/muLOlLl+NtlO/Dtsh3o2Lg2ptx+go9S2YcsNjrYbdJB6AztsmxrETItWmyqdBSbzxZv4z7GRtm3u+0KxeuE1AwRK5G8FuC5TalN0pXSmhPf/DMGOXnAJW/Nx5pd+gG5TjHyqApKGzNSLhZs1LYSyvUaDVc0BXsPVqDXoz/H3f2CQMToXcvC82rKU1q2sqxoAYuxqayqkTEjHNLctJZQJ+ZRotZGvlmqvodfjKgocrcAYAe1eq7XR/MOKTY62I6xcWduxw1WFRu9rGiZaaGATDEIM7j2LqmSyAiyVXjOun2631t5MrulEODiM43MFU0teYAoJizRfLF4K45URvDTn7vcFE0TvXptxtqicVVL15DF2BhcmQekFptwiK3FJhkm7UbE5id25m0i5K5ony3eqnkuz1QZrQ4EDFJsdLDbqKOiiOIjlfhm6XZTwdRuI4oifl65C4Umg/31EATBciC83sCTmRbmPn5CL3mAkbixst8fYH9VK7gVeyzP8BSsTlgQzNVr1q5ogSWpH84YWaIMg3qj971RCle1tpqVbs3NmDWJlk7jymB0jtwVTSvdc83f8hgb/itjpSJ7AssU6KmQ8yYSTx5gNhKthqgoylzRVmwrYSkaU/TqchAsk1agGBsXEEVg/AeLMXvdXlzQr8BvcTDtr9249t2FAIBNT411dC0B1ucdem0mCBYbJ6tWP6/cheveC45bh1Pc6iCD3vGqiV9WGUFmWk0sA+snDMKkTA0v3MSCWTKJKGPapMize6m4oqmUgtQaL3LgZiMXW10WpUeA8lnNpHsuq4wgLzsdgDLGxrSovnG4ombxVASYuqKlgF4TH1u03rWex4koAk3ysrBckvI5iNjZNJtnyGLjAlFRxOx1ewEAH/9e6LM0wO+b2aVoFATrbhy6Fpv0EPeT1gSLjYXf2kmPyXlx6OOS7DwXiZFsajFZhfsPo/ODP+K2yUvs3ZPnArFBsu187RVvz9mIt2dvVP3u8rcXxP9WnbeIiRYfqcXGTIZD1ijfvZmqoHw25WczFrAB/5iGZ39eA0A+yZMmvuExzXx5VQQjn50lO0bJA6wRc8OysxAkiiIGtKnHWiTP0YuDDiKk2LhAktURGQKsr57rnZ4eCslGHi4nNTp9u7TfV8v+ZmfDLp53uDbCLSVVNvF15Q7eEttr4qslNcGpZotOEMxmRQsOtne+tvmQQbFmmek9HpPsoyGlMqLfZtTGKambsdLFiVek/W7x4Urc9MEize/1yvOFaWsBVFvZY/BeS1btKE04xjLds98WOy+Iu6LZeNmizd/xBu+Ly1YhxcYFtCpJToa//sssEATBciMwOl/U+JsXzK7Uqe3Xk/zDghz3FBtXLusav67da/k3Zl2wBJiNPbAsgm9IRT1YVmXhdwF6SBdZYeQKo+GKpuzbonYVTEYo36eZei7tc575eXVC4oMSSX0yM1GfJNng0u8J3/KtxZi3QTvZhvJxRNH6PnN6pIBeEx+37bzraFT0vY6wgFzRCEO0KnoS1P/qjs7icxhtuib9vrwy4vvu11aQTgzc7ByiUZGLRBRGSIsgGtXfr8TadYPlqiRNE2o0ObD6XoO8iqq5CZ7k+JPf/2XhejblsPczbzDjO6Xg1Bdnx/82W59EFVe0aFTezsqr/O2PzRSFVLxdJWW61wsJ1XWtrDJiakLqd19z2kuzccHr87CntNz0b9JCDBUbZlfil5rkAdYRkRwbspNiQxgSpIm5Vapd0az9xngDtZrvr3tvEc5/na/djRP2sdHMrKPynIxGhovenIcuD/2IosN8Z1eTTlAveH0e/reITfpLnpuUU7emYx75CaVlleZd0WBudZFHa4bWe7QraRLMKZjy3xnr0PnBHzFj1W7ZcXVXtMSj0mOHK6rQ45EpOP2/sxPO8woz71cqs1FMiAAB1767CD0enWJKWfhu+Y6a3/o4yzdS2KRYzVqqRyrE2MTTPdvoS6KimBR9ECk2hCHag3fwK8+DX/0py7ZjBqOnVn7/+6YDlq7PC2qdA6uA03kb9kMUgal/7TY+2UekRbBgE7ukFTKLTcDakVoNkD5BZURE78d+RqkFFywz4Q88Drhafcf6PfY2g7NbF3gsmzgOuox//7QaAHDlpN9lx1X3sVH5vaz9btyP8qqorylszbxf2QacRmUnAFP/2oWKqii+XqK/+SIAfDh/i+rxiT+YtyqywLx+ITJVwFJAr8GzP6/Bml2l+GXNHus/FpNjITsZrE5SSLFxga8Mdqv1GtbZXH5YvtPS+Xqry4IgcB/Iq7cAJk8e4L4svOPWu+S1jqzYVmwYTyMIguH0rCoq4uWZ60zdUxDMWmz4Q6uNXPduTcB3raOxiEcqIvhqyTYUH67UvB6n1cIZLjyTan0RE3Uo6QTHj7JNyIomVVo0fiN9NiM3Tem3udmJu12Y7Wdem7XBkhXFDvI01erPpXac5XtLBYsNAIz5z6+2fieCL2+CA4cq8Pbsjfjk98IEl1Q9MSNJtkEn7WPjAksLi1SPJ8sgvM/iZpNGgwXv5aIcPASNv79eJldoo1HR9gZn0ahoK6Oa2xypiCArPaQ6gRBFEYcq3IkDkg0eHNUXaWyDFdTe7AGdCbySIKwSlldFEBbkO6GrrQweqYhgi2Tz4JO7NwEAPPz1CnyycCuObZmPz28arHoP+6XgX/mVVUaQEQ552r7V9RqVhAIBtIxKLeVVBqbMCsn3an2YlbS3bmeNs2SJkv2G4XvjbwhyBbuuWFFR1LR29G6Z70Aie1z69vy4hXXdnoO475Qupn5HFhvCNslSdax26Hp9RnWGJ2fy+Il0wJmvyF5z1iu/2Tblaw2wflou1u0uRZeHfsTtGnuv3P2/Zbb27TFDkOuIWTIt7D9hZiDys8zKqyI49rGfMezpmbLjaiuD3R/5SX7g6ClfHnUVWrylSPM+QasX+w9VoPODP+KCN+ZpnuPGI6m6nal044LB925jK4Bb8iNlRjQl0gWB75btSPjeyt49LGNZ1OChaqeKxcYuoqg9JvtRclK30R9XmPesCcJCmRVIsSEsYzXGxqjR8L4yqNa3z1qzB9e9uxB7DtZYr5SDwJLCIttugDymkHzz140AaiacSlglClBDHmOTnGSmm+uOBQgmByL/Smr97kM4VBFJ2JNJTSFL2Dn+6H+lLWd3SRmue3chfl0r94O3q+j71bx++rN6smF2AYDV5EitnNSKwG/DqJ4rmhZW+kqjM6245Lg96Ze5olm4Fcu6TWqNPlExSdI9J8EzSCFXNC9JkrpTWWXtQfTOVqZ75hHloFJaXhXf1XvqXzUrhCwX8LRM40FO9euEIA8e24qOYFuRfIKv9jSZFvafiFWPK45rja0HDnOXVEK6WLFsaxGKj1RiSIeGplw+YhM66cTxoa/+xJSVuzBlpXxFftoqvp6bBV61cFFMDDSXKsw8tDlpPdLq+lhOynh45hhSScwukIlgO81IhuFm9c7ETUxZYSbGRhRF/LBiJ7o0zUWbBrVck8UJHFV7JpBi4yEVAdnJ2YhKqxYb3eQBfA0mVpF2amoreHYHhmRbQXFKshXHYZVYpHQbrmjVdU69kvFSZqe/NAcAMOfeE03tih632EhO3cE4UNuvorH6TlhNLNXuqzYhEzU/+IM5iw2766mNReVVEYRUNqZ2u32Zub7bikcyuKKd/Pwvrl1bFEVD6/nUv3bjpg8WAwA2PTXWNVmM0KtPHDR1ppBi4zEfL9iCC/q39FsMR1iNsTHqoJOlUalZU+wOC1qdpZ8xNn5OlIO2QaceL05bi48WJKaRtTKHiNWPcEjbUuhnOande+v+w2hZP8f0NbSSdLAg6HXIKuruvmJCnyUtFx5chM1YLVjGB6hdqtMDP6JpXhZ2FJcpznW3fMxYq1R/x9IVLfh6jauIonE9WLT5gEfS2IfXrKN2oRgbj7n38+V+i2AKPfOtVVc0XYsNBO4nGWblm6dIHgDYV9qSbcMspyRTcTzz8xrV46bbgcTKGRKEwKyqRqLmNrP76mgMl/S5OEwQaAtTr8qFZ1XNiqZqxfF7AUFpFTEWwlqMjf65WtdSKjVW72sHM6mu1X7DUiE16wL3yNd/4vmp6v1aMhMVjTcd9nuB4Oul23HDe4twqEJ7nzTe52BWIYtNCmBl3rNg435ERRE3f/iH5jlWXdGMG01ytCplHAVgLX2oFC1XNIqx8X+gcAuzg4uAmg06QyEBIcXyVOcmdbBqZyl35VQVtSiRpKqzrvd+lY1lVzQLWs6mvYc0v1ObhKu6oon63/MIS1c0KwlQvHRFM1v9v1m6HR2b1GEmg5n7btp7CJN+2wQA+NuIDik1RomiaLwI6XM7uvUj7blcDK3+cNaaPcjLTkevgnzGUrkLKTZEnLLKCM57ba7heVUWN3Oq0EmhWR1jY+lynuNkADPaV0EL2uxTTrKZytWwMtmusdgkTn5jaWh5K7Jqi415oaRPlSwWGzdRptaWolbqlZFoQt8srYM8KMZmJGDpivbvn1abPtdLVzQlB8urUDszcfr20ox1GNmlETMZzLQ7aYrsqAiEU6itRsXkGJu0HiGWJMnP2CA7kGJDxDlUrm2qlCLdTM8M17y7UPO7oO9jY0SlzR19KXmAnDs/WRr/O9WLRppwIywICauqvC6YmnVFiyFd+bWbNl2LlKtDKs97m8p+VH5bbNxO98wSt8tH/lg19f/NXzfgie/+wjPn9kQnFesMy73ErLq5RkUR4ZRKEi0aZ0XzRhBHJFt/SIoNEcevus3DyqAeTuSzvaNxVERlJIo7PlmKAW3q1cjiZ/IAH9/Tpn3WlOkgYvbVllVG8e7czQCOuqIpJh+xz362KrVnsaqsh2SuaA4FCijMsqKZPE/WXfnQ1yTc0YQ7lhXFhqUy4r7Fpgbpsz/x3V8AgDs/lY8N8d95nDxAek4kKiI9zO7+vGMmeUAQLDr8S2gNSh5AxPFl5UtInuQBauy0maY2EhXxxeJt+Gbpdjzw5Qr7AjikoiqKzxZtxU6V4Fm/4Ly62MbOc4VULDZxxeboBcsqI/jfoq3YU1ruTEALqCnBkai1zeykFpv5DFehgeRboTTC7ORK7ormP2YWU7QeLScjcYbNcpLp9oRV2la09Au1dsFWKmPNRnpGkLdusEPUlGLjkTAOSLb3RooNEcevup1sjUrKz4oNBc1SFRVVlSKvAzNfmbked366FCc9N8vT++px2KTLZNCw0wzCKhYbZRV56odVuOvTpTjn1d8cSOecKouuaKkcV1NaVlm9gSaj65ktdpkrGgfBj2Yyg2mNHzkZYVx/QltT59qBh31s1H/HTjCjNli9yWvNSamWzbMiEjGMhw1EiQRCSPOQYkPE8UPBCEKMjR/iRUVRdRNHr5m5pnpn95IyfpSJUc/9wpUFyU90LTZHa+6UP3cCADb77M4XiUYttiX3NBue3V9Ly6pwzCNTcNeny5hJabaPlU6K/Sgh5aTcjAx6GSSb52fLjrF1RWN3LVVkbnjm2wJLuYxue98XK2TnpFrSm9snL9Wsf7zPa6Tw3B/awbJi88svv+C0005Ds2bNIAgCvvzyS9n3V1xxBQRBkP0bPXo0K3kDjx++4mZv6cdii1F57D9Ugasm/Y6fjk7OUoVIVMRhlbzzqbpBp5Lvlu/wWwTmTF9l3bqnmhUt1qhE2X88RTXGJmqt/iZjXI2VCcRni7cya3Nm7yvboJOj9q5FZSSKwv2JafaB6rYRVpgcWC7eubEQ+I/v/8KV7yzA5W8vwOItB+LHrTQFL/ex+WjBFtkZqZj0Rm2clhKEIgmCjFawnDzg0KFD6NmzJ6666iqcddZZqueMHj0a77zzTvxzZmamfQmTjLAPo7V5NwR/arfebZ/64S9MX7Ub01ft9jzl4OGKKny7dAf2HvQuNiFGJCqivDLFlr8skIwpRe0sLFRv0Ck/ptBruBm0rFhsWLphqV/fxYszhNUk1exKurQO8uAibDQmfX10M1c1woKAdMUmTxv2aO/1YxXW5VN8uBKv/7Ih/nnWmj3xv61MG1iKZdUdNNVc0QDg++X6i65BsIYYZnZTuBzyjmXFZsyYMRgzZozuOZmZmWjSpIltoZIZnncJ9yV3AIAqnVHX7YDniqooIlER2SqBpk989xc+nL/F1ftroTVoBqlzcRMqh2oEIbFPURaNHwOr2h2rLTbmfl9aXhWA6YD7sOqTS8sqzd1PmjzAoxdwUBIzp7ylkQh7dBadBEFAmosrIKzLR2/jayvpzp2I1bMgH0sLi2rua6KflU6KecoAdrC8CtkupGjr0Kg2wiEBq3aW6p7H28KSFiVllYZjRND2J3IlxmbmzJlo1KgROnXqhBtvvBH79u3TPLe8vBwlJSWyf8kM365o/rTAM1/WDmp2ewHouKemoctDP6KsMjGeZYqP7m+8dIa89mWpHFguRdDNiiYe/a/XUqlPcKoHT3PC9HhkiqdZ3HiF1aszGyMnc0WTKTnuVKJtRUfQ/eGfVO+fKE8iR3TiELcVHUlwRWMJ6/GS2eUcXKdjo9qyz2YWYqV1gxdXtF0lZej+8E846xX2CVNaN6hl6rwgDFFP/bAKPR6Zgp/+1HeD5klhNQNzxWb06NF49913MW3aNPzzn//ErFmzMGbMGEQi6h3QxIkTkZeXF/9XUFDAWiRXOLVHU1u/49li40+MjX55uC3S3oMVAIB1uw9CFEXc9elSvDhtrSnZ3CQqiqpKsNcdDK/dGVlsqgkJiWURm8zFVww9lkkLUeRHYQ8MXrd3afIAD+Jt9FzJAGNro9qClJT0sHv5kbwcL61YXZ0oXMoU2Wa6WWk58OKKFovJlVqfWJFMI8+rs9YDAH6RuD2qwclrNQ3zVn/BBRfg9NNPxzHHHIMzzjgD3377LX7//XfMnDlT9fwJEyaguLg4/q+wsJC1SFzB83zMD63cqDi8lGlJYRH+t2grnvl5DQDv4qFGd0t02+SxH1GT6VB5FT5esMXzOCSeFwi8RICZlKyeiGKIeXuN+/BSJkZ4LaZo4m83aZyriMc1UK4qIvrBQ2kBstiwwolUIoCMtJppoZnSk5YDL1nReBgd4rGOPtUTlkNkEOKEpFiOsbFK27Zt0aBBA6xbtw4jRoxI+D4zMzOlkgtw2hcCCJ5WzpoyRbC+V+5OIZXlBVHDYuM1RiI89NWf+GzxVnT+bZMX4sRRezfFRyqRl53uqRx+IwiJ+9goN+jkRZ0QRRHFR8zFergNLwP14YoqpIVCssmkFOV4UXykEnUy3Ru2ZfvYyKw3IryYLjasI58LSB9fre4YxZ64G2PjXR2KREWUllWiTpZx/+ZErqgiaUfIxCAY5dAVzU3MjstRsbrOJkOJBO21ur6PzdatW7Fv3z40bWrPdSvZ4GVAVcMXi41GJxGTxctVMaUsXrk7qVkftB6bNxes74+mXTYKpGSNssw+XViIno9OibsRegEPfsfVrmjyYwmuaH7E2Kgc+3bZDpz76lzPZeEJ6bs4UhFB14d+wqCJ03TOl5dkz0enYNjTM12STqnMuHYb7fvrrPh/vTTRbc2oO3TTsst6IVBvbnDWK7/hmEemYHuRempr+XUcyKD4sZnik/6GF1c0N1cFzSZyWFpYhJ6PTvF8bHQDDoY6S1hWbA4ePIglS5ZgyZIlAICNGzdiyZIl2LJlCw4ePIi7774b8+bNw6ZNmzBt2jSMGzcO7du3x8knn8xa9kDiZQWpikRx0weL8M6cTabO56VPAmrKyc3yMpqYuhl4anQfLcm8nkwbKVJ+uWMoxbr38+UAEHcjTBlULDY8qL5q1WLBxv3eC6IBDwP1qp3ViXL2HarQPOeRb1YmHNuy372NVjX7HdfuJ+p/NrixUV13VbFhr9loUnS42lr1wwrjhDZO6nZUlPetZibx0jHgyz+22b85Q3joA2N40e/tPViOy99e4Nr1eV6QV8OyYrNw4UL07t0bvXv3BgDccccd6N27Nx566CGEw2EsW7YMp59+Ojp27Iirr74affr0wa+//hpYd7PcLHWz/4X9W9q6npfV47vlO/D98p04YhBgGcOPSarWxDkmiZsyJaxOKb73SrFRG3yNBs1VO0vwxR9bXVd0tIKJ9Y55gVfvRg8eJsdqFhsh7op2NCua10IRmnBmcE1A5ooW1W/7bt7/66Xb8ef2YscTqiBZbMzgfrcnypQZq8kDXpqxzgWZrMN7O2PNUz+sku17xBoexjorWHbWHTZsmO5k6qefftL8LojMu28Eznr5twRz4uD2Dexd0MMKIt0fwAzcmJFR49PtqsVG8regkl3Kq85R7TZGjz36+V8BAHnZ6Tixc2PmMpnFL4uNcsLih1sYD61FOPo/KbFEUDWuaH5IykPpaOOXdLxPEKR1RbY/iUclJgKYt2Efbv3oDwDAb/eeqHu+oSuai872rNuVmau5nTRFTLDYmPkN55WaMTwqTft1rL4s4DVRhhaux9gEnZyMNDTLz2Z2PZ5Nen7UXa3MY17EByg7ZK39QNxGLbNPVCNYV6l8rdzu7r5PVlzRCuqxaydG8Di4+EH1Bp2KY4p6w2+PQ/CGLMbGj41dRRFrdtUsIjqVgCw21khIHmBxg04AKDrs7iTbDFY2NLV87RQce4I2hpBi4wBpBT+2Zb6p33ipPGg17gUb9+OiN+Zh7S65FcqPjCZzN6hv3hqPsXGxSSmvrCwtL7ydamemobxKJWJW47GVypjTVzZ15S5c8uZ87Cg2DkpVQzqoubHLsxZKhcsXmwQHq1ghITFzUXyV2oM4taDCw7vjEa29a7x0RZNOpo3ek9HCi7uKDWOLjZnLuTyrViopZsZA5TviIZbOzWJyU2myi9sSiZyk8TYLKTYOkFobPr9pMM7va7y5KA8mvfNem4vf1u/Dte8ulB3nQbYYMYWG9arYjuIj+GD+ZhypiBgOJF5YbO4f20VVsfFqNfCadxdi9rq9eOTrPw3PNVIyvaw+ygGXo6rrKYmOaJJ0z0ffly9uepy/jw/nbzHc3NFteMtwCPizd43y/lb6XaMzXdyf05fx0u3FtmpXNKkvmnWLDY/12k+8WCB1u8h59jRSgxQbE2jVGWUHXF5lPFB6WT2MKuPuUvmmijytYtZkRWMr0+kvzcH9X6zAP39clfCd9HWKouiJYtMsPxsVKvXG644klnXHCV5KzMMGnVy0FtXYMP+tWbzz6aKtePqn1X6LwR0HpL76HqR+Vl43KoqyiaDhfX1M98y6TMz0+e7H2Chd0Yx/o1Tw/O+Z7ctwZu/mTOUAvEp0437sVZAgxcYByj7GzLvnqYIoO0mOcgfEYS3SnqPK3PRVu1UGEqkLhLuBpzGGdmig6sJl9l2wKh+tDQJ5hYOkaFy05ZAgJJRF7HNcPh/k5KBoDPllrXtZhILK55J0vb4kDxDZTt6D5IpmBtctNoBsjmw13TMv2H3taWYK2OK1ecjg6RQe37EewZrNcIbTTvPlmetw4/uLfMtGxnNzU9vH5oLX52LdbjabXUVFUXdiKkI7sQFLBEHAY+O6J95fFD0NUszQ8NnQEuHThYW4TJE330uLH7k7VCMgsR+Ku6LF49S8J2DjIKGCL8kDII8Z096o+Oh/DUaxICUPMNNm3O73lMkDTKV75jD+wm4cjBkl5IDFDGRexOSovafN+w7j4jfnMbl+0LpzUmxMoNW4nfYx//pxNX5YsRMzVu12diGbJFicOKq9avEB8zbsx/XvLWJzfZVnTXBF82ilpaBeDm49sb3s2LtzN5vqQFm9My2LjfTyhyTpw+/+3zL84mLefCN4UGt48DsOhbSz+fkZYxME/HBn1HoTu0vLPJXDDDKLjVdVSOmKplFiogi8O3eTYQauNg1rsZROBvPkASbOcbvORhUxNmaGQOWGsUFeczKjOP62Xj3hkRZm9xF0gznrrMmqRdCGEMv72BA12H3Zew+Wo0Htmg1Ly0zE5riBcuLO0wTowOFKZIRDCZ29Mi7ILmoWG2lpqCdbdg9lhzrdY2XXjCvaT3/u0v3ey9pDFptqBAiG+y/x06oJKdLXdMmb832TQwtZVjSP7hkV5avmekPSQ18ZJzypneneFIf1eGm0KTPgRfIAuaeAmX6Wy67Ypkxmk01wNFUC4E29CBJksXGAUhM3++77PjEVny3a6oJE1kjY5NAnOdQY/NR0nP3Kb675dkZFMWE1UNqJR0XR08mz34NDmkZAkSWxPKxAPHS0HIiguo+NMsbGDzntvp/0sJdtzt9GJy2hNbsO+iaHFtJ36FV7EyHvd/1qYlcObm14jj/72LgfJG7VFa1SsQ+b32MZYH9RksdUzmZwW24OhjpLkGLDECsv//HvVsb/9qsxSSdEFVVRXPt/C7VP9oGlW4t1rSpOiIryCZ8yba7Xk0GzA9a/f1qNW47uyg2wc4dyMy2qW+woPoILXp+Ln/7c6bcoviEIiQmfa1zRgkdlxMM4Lc/uFEz8SP1cnTzAo5vp0LtlXcNz/NjH5sXpa5neM0EG5WKfiVZSpWizV01a6PrG0W5hVimzO+5+srDQ1u/8hpIHpBB3ndQRAHDHqI6WfyvNvuHXCod0ZeyThYUolcRQ8IJbOfJFkYcIiRrMDua7S8vxzdLt7gpzlN837cfCzQdMn+91eT745QqmcVdBpDp5gOJYgiuaP0HghD4czN918cMqGhXlSVv8ssz2a22s2OwoYhsXZaadrt9ziOk9lUSj8jHWzHBbpZI94LK3/XWtdNMa2zw/2/Zv7/nfMoaS1OD6PjYB69BJsXHA+OHt8es9w3GLIvDbDFI/Yr+0YekAsn4Pf64QQOLAxqoBV1tstK/t9Svx2y1GjXNfneu3CLrst5idhjU8dPaCkBgrF5sy++mKFgS8SOceZGQxNh7VIVHhArz3oPdt/J9nH4MmuVmG5z35/V84XMFuMZCHdqrMimZmwU3NyurHe5Oy/5C9WFwz72D88PZcvCsvCdrjUtduCvXWLQgCCurlxDtiK6tLUqXi5g//0DnTPsapMGv+9ivltFXYuaKp7WIj9e12vzy+Gj+45t42H4ynDtbp6urobk3wwNguuHhAS+N7OboTG3iw+am5MNa0a1Hy/4SSoPrTe4VsSPCwEknHpfNes7+4cnz7BrZ+1zg3y/RC07GP/2zrHmrw4O4jQj4WmXGRVsbY+M2va/fgH98nbsBtBjN9ejjEX5/q9rqomcQWPEGKDUOsvPqwh0GyWsiCNDmttwkxNoxasFpD9dpi07MgP/633aBQnl6bU1k6NqmDa4a0RWZa4oalhDpqtSahKklezG/r9gIAHv92Je78ZKlrrj52L3tGr2ZsBdFBWk6T5mzEpW/Nx5EK/1Kz8oZ0oh3EDTr/c0EvW7+zcv+yyiizNsRDX179LNbKXxlj4xWRqIjr31uI/0ytiTvad7Acl761QOdX+pjdS8jJO7/pg0XMFQVapJFDio1PaGWh8pJtRUewYlux32Lo4tYqlijqd2Jer57xEDDrN7EiMDuv8Nt9j4fFAEEQEspLULqiSaZMF705H6Io4q3ZG/HZ4q3YuNddn32rXDKwlWf3khbbI9+sxK9r9+KD+Zs9uz+RSFQUme3UXl+ypYIVrN6f1RyVh/5EFK2v/ntlsRFFEZ8uLMTSwiIAwC9r9uCnP3fhualr4uc8PWWNxq/Z4VTx/n75TszfuJ+RNEehGBsZ/s+uA0DvlvmmzuvRPM/0Nf2ayCpXGk59cTb2HSzHHkb7w7Am0V3MPtKYjKgoJlxc2kF73Y7tdpblVRHLOyGrwaLjYrXaTTqeedSqTexY7JUq3630s1tZyOyu8LOa1NrlMFls4sjTPXt0T7Bxq+nTyjj4XwurfTG7RTD/Z4/KGBszi0d6fQir8QkAZq/bi7v/twzj/jsHQGJbFUUR63aXOrpHnax0w3NCgvM3VcGZ+54RPLhJWoEUGxNcO6QtHjq1K+rVytA97yITsQEx/LLY/GdaYrrIPk9MxY+cpsxVNii7g972oiMyf2gRiZOvM1/+reZ7j9ux3ZXz12ZtQO/Hf+Zi53Knm6daebc89LMciKDuiqb4rJRT+tm1AcvmZT1VbFQqnOt356HimsSfdM9sLDavXdrH9m+t3n/aX/obF5uFh6pRWzGxz043dgtWy4oWY+i/ZjAbn9Yq9npSjt/3frYcv28yn8VTjXq1ap7/1hEdVM9xey8hO7gtEQdV0xKk2JggIy2Eq45vg3YNa+mel25hMxC/Viafn+puHnzmJLQoe+U2fdVu2eeoKOoOJDd/uNjTVKMfzN/i6Pe/rdvHSBL/iLlQmRs3RN836fT7/kD1IKv0r1YmM1HK6bbcD321An//3F5aUy/7Rb+nJxzOj2TIYmy82qBTdB4vcEzzPDSw6YYGWN/T64Vp62zfS4qf8dn/uaAX+raqiwfGdpHVy7OObW74Wz1XtF0l1Ytd8zYwdr1CoiI4mfEeMR0b11Y9HmJhsmGM227ZZLFJYlh2PGkcJA9wglE7at9IvVNw+75miSZ6osn4de1eLPcw/uhvGqtDZnHa8czfuB9fS/bH8XPSbqaTVvEkTElMuaIpvpf2Y6xf85GKCN6duxmF+4/Y+r2Xlmw1HYp3ZcNLZOmeJcfX7zmISXM2oqKKvTsNC1c0p79n7Yq2bGsRPlqwxbBP9TPL4rhezfG/G4+rzggnUSyzTFhszLizurFe4XZpadUDHuNhXbfYBGywJcXGAkYdk5X+0G9fcqeEFQ/bpkGNNWtM9yb4/tYhrtzXbqkluOOI8hV/tXfn5S7oJ3dr4uj3TpXuLfsP49aP/sCGo/sZ+dGRxd6BmXcswv/Oloe+PmQieYAS6QSK9WRKzy3FDJ5abI4W3K4Sf9w4S47wtyGyFK2aMeKZWXjkm5V4/Zf1zu/hgjXRaQ1irdic/tIcTPh8OWau3qN7nt/9WYyhHavTZJu1elWZiBdxw33L7cU3bcVG4KLv95ZgPTEpNhYwerVWTOhpAVdsMtNqqs4Vx7XGx9cNjH+uXzsDGWl8Vy2lxcarQWVkl0aqx51a8FiZinceneT5YXoWEv7Qhiw2RzFhddBLHsD6NTu9nNf94vRVuzDgH9Pin73MtHfJW/7uzm6ENCWtWj2Zu4G9+6v1ZMMqOHyHbmVFW7dbfxNsXhSbh07rhgfGdsFXNw82Phn+WWzcRktmHp/FzW7rogEtkZetH1/OG3zPPjnDqOOxUuG1VgNWbCvGmS/PwXwXBg2WZEpM1NcNbYvGkp2a3eygbTdghVDKGBuv3AAGtq2vetzpSjWr1auYch7x0xXNxNRGhBhP++kXPExE9JIHaInnhtzPT12Dqyf9jkqH7klex9i8MtO51SFZMdrHprySvSsaiwUVzy02JjUbo7rNw4a/AFA7Mw3XDGmL5vnZps43s7m3GwsGbve/0npw/Qlt43873ccmaNx/Shc0rGM/Zs0PSLGxgKHFxkLj1erkLnt7Af7YUoTzX59nQTIteRxfQpObhrWL/618llg5/X10Z+b3ZbURld4qdvxeLpRfzOWsrSIRhdOVataBp766onG4IqYKB2Ob2iQsFJInD1DihjXu+alrMW3VbkxTJOmwireuaHzsJ8YrMg8jlSpTVsU+NTYLS6zT/sO6xYaRYsNBf2IHM8/PolUr36udfsxK7K+0a7hqcJua4xwOUG7KxOPzGkG9ugUMY2wsXEurk9vPKOe72wzvXONSpaz4sWK6cVg7mcsaC1i2MaMVMjcGmoJ6OVj0wEj8+LehsuNOJ3RmVs3MECtfP7OgmIqxCegkgDWqyQMMfuOmC6bTvYy8dEUTBCGh3QVwDHcNoz7ADYuN0cbJZnD6Cq1mRTtYXr1Xy57ScpTrKHtGVZvXPu2dK/vpfm9mrGA9OY5GRewodjc2TrpQLRW/oF42D2taMtzstoLYJ6b5LUCQMGq/ViqAWkMP0q7XGZLeX6/DbpafzXR3c6YdpE+9k9qO2Ly4osVgpShZITaQmHnFPAwsPLiOJCZ7huEoJxq4GDnBab3x2hXN6yQu/tcY88hd0RJxx2LjvITSrGomCqyOMXsPlqP30T3SWtfPwcy7h6tfNyCuaEpyDDKjmWnzrA2jN32wmPnee1cObi37LMj+FjD5uoHYWVKGzk1ymd6Xd0ixSXKMOh4rrmhqfdyjX6+0KlICr8xcjznr9uLNy/s6vpYe0mD3xI5NP9sYLxitXAuCd6toTidYrPUQP/dUMONueOtHf3ggCf+opiw2yIrmZrpnp5Y+r13DlBYiVq6uauwuKcNDX/3p2vVZEzFIHuAwAZ7qdUU4n+CnO0zE4qQv3rTvsPZ1DQZDXi02RvMaM8lDWcfYuLGheIu6OYhIKrXSYjNAEh/r9F0xj9FxcZ7lZp/oFuSKZgEWHXkMtVWhChNpE43454+rMHvdXnzxxzbH19IjV7JDcX5OdcaMQUcb/gX9Wsa/u3l4e1flMItaN+JH8gAt6uU4yzrCuqNMluDIqkgUb8/eiFU7S5hfm4siMpEVLQEX5Xaq2Hip1wiC8Sq6HeZt2IdPVTYLfPL7v5jfy02Mkge4sWjFok05VY5ZegVI+9Ff1qqne/7yj22YvXavbyPQkA4NdL83atNayROkx4MQp6Ec86QSB0F+t+AxC5wRZLGxAMuOx40BVUpZZcTU5lp2CYcE/PnoyYiKYjy18/vXDMCBwxWy/PcndGzI9L5uxdj4PUlNC4fw56Mno88TP6PMhu+62S13RFHEtqIjaJ6frbuK5ofFxo3kAe/P24zHvq22hG56aiy7C4MPtyL1fWyq0VLWjVyMnGB2bUbLGuqlxUaAkGixYVD3Ljia+KVTkzro0SI/frz4SKXzi3uI1GJTWlaFnPRK5OWk6/zCOSJEx32xFYtNzxZ5WLq1WHaM5dhcJSnD75fvhCiKsn537a5S3DZ5CQDg85uOY3Zfsyx6YCTqZOm/00qDRq3liibtZ4I4OZb2BUrxnS6EsrZguWlVCaJSRxYbC7BcxXZT6fACQQBqZabJOsVwSEjY1It5A7Z5ObVX57cyo6RWZhpqZ9qbOJitm8/9vAbH/3MG/jtjner3seL1JcYGsRgbdnVGOWkJEh1MZPBRK6nYQKS9QWcNrJNEREyatTU3v/PYYuNmjM3WA0dcu7YXSOezJz33C3o+NgVllTVxNW7Md1h0O1aU47eu6IeHT+sqO2bkMmaFKsWKk7K5bdl/WPM7L6hfO9Nwz7kKgxTuWn1IRAyWxQbQfge8y++meJw/uiqk2FiAZceT7vIShtt10S+/S5b3FTX+9hO7nYjZCeoL06sVmqenrFH9/nBlBOe9Nhdv/rrBniAMYFmz3HSpc9tdLzPduHsWhMT0AUZ1SJY8gPEjmLbYqBy7+vg2TCeVhjIIiUo0y7vf9MFi7D1YzvCK3qLWp+wqcTcTFYt0z1Y2O25QOxNXSlL5AmyVa6P9wORKAy+jkBwji43W2CM9zKJZ612CRV+svISsXxX0z/Ubd7OiBU+zIcXGAizjMNxOp1u9iZSb1zd5nk/3VaLW8ckneCo+5PZu5Qi79/x17V4m939v7mYs2Lgfr/3ivWLD0hVt2dYiTJqz0bSLnh3cHtvMKPEhQbu8zCQPYP0Udi02Z/VujgdP7epxVjQhoR9extjC9+zPNQsIvE2GjFAbo6R10p3FLeeFlO4wK5pT5frVWeuxo7jaWqccV5RPVy5RbHitH5UGnaiWdT/COMZGTwo3HAykCp1y2wpe3xVRDSk2Fji+fXW8SJ0s56FJTidc0aiIwv3aGVjcVrLNXt6pHMpiYruar32foMFKsTlYXsXkOnYQ4v91/pZPf2kOHvlmJb5Zut3xtfzCTNtRtTrEXNEgqrqRSBdoGOQrkWG0Qh1H+WxxpdY7xaYyEk2YeH63fAfTexz2sT05xcgd1Y1XVVpWhX0OrVxO90IyUq6Nvn/qh1XxOKuEfKGK+iZtn7yOQUauaFp770kVY7dbNROLjWLpWhoTxzp0gLW1P4BGFVchxcYCd5/cCY+P64YfbxtqfLIBWplEzHL7J0sw5F8z8Pnirarfu+6KZrIlOZ2kerVRpGoMjid35gwOHjoonbTbVdNsC1N63kg/nvPqbwm/kMrNPsbG3HmJek3NkQ+uGcBOIB3mb9zv/juUVGYOmpYlVC02OgHVLCiviuKOT5Y6uobjfWwMFJe6JjJYbj6a9tmofkk39HQ6J7DKzLuGmTrPKAZHK8W11HjLIiGDriua46sH2woTxJTMbkKKjQWyM8K4dFBrNM/Pdnwtp8HZXy2pXol+abp6EDhUsiWxxPSlHcqgHFztrugGpc/ye1LvZ9rruCuabxJwhonKEFJJWRz7mSiqu1a5qdiYvZ7SNUX6cXD7BrhxWDuWYmniukuwq1d3FyOvwk37DuPMl+eg6LD6ir1fON3Hxsht6p6TO5m/mKj7Ue6KZv6qjhnUtj5aN6hl6tzR3ZtgZJfGlu8hTx5g+eeWYNGOlVcYc0wTjOjcCA+e2jXxXM60IL/nDbxBio1PmHbZ4BTTMTZOFRvF4OqWK5rasOJPjI2/PRQPG3MGpZN2Wwk0UwyCICRMxFrWywGgPVFyM825MguUFlopqmNcPqi1rfu3NTlZi+F6fXfZwuEmczfsMzznjy1FeEYjEYmU6at24asl7u6tFsNpyvBYjE3nJnUSvuvRIg/1apnfc0zZRyjbm18xNlb62PRwCG9e3hcXD2hpfLIEubLhbu03U3ZW4/cy08J464p+uPr4NsYnWySIAflBghQbn2BmdtZoH/y4ojkjYWXEdvIAlWMc7WMTw+/+zivXP56pqIrqxq8VHa6oXqXmoKhCgnyFecWjJ9f4g5tIHsA6rbfUtUaPBFc0xQHlHKSuyf1TrLYf990Jg+uKpoZa+a7bfdDwd1dNWoi/fbwkHlTvJlayoun9/u+jO6t+b0VvUtYvpaJTLkmfbZR9jCV2xpmiwzUxJ5cPaoVW9XN0zz8kiy9zt/YfqTDud9JC+suGVvoC3tqy3/MG3iDFxifcttjwUtGdrkxUKSZeXiYP4K3z8gI/LTYxcrPd3QTQiJs/XIwh/5qBOesSEzJURqLo9djP6PXYz6hweSJiKnkABEhDCmpnphm2EeliAWtF9oP5W0ydl+CKlpCyWv5Zuj/WVYO1V1Ct9jduu5Tw0g+7iZWxTBlo7kbpm0kecE6fFvhJI1Y2nlVNc9HQ/Es1er5ySR9y2dsLTF/XKXY8A0oliko4FEKtDP0kSiOemWX5HnYZOHGa4TlpIcEgsxoHgx/BBFJsfMLtDRD9dmmK4VQKVuVkpLg43cCT1QSGxWWWbS3C2Bd+xW8qE3NDfOzcY2V4Xt8C32QAgCkrdwEAXpy+NuE76aplyRF3M16Zc0XTjgnQcpWTKfSKU6av2oWxL/yKVTtLTEppE6UrmoHFRvqMlwzUdomx2n7USqgyEsUlb87Hv39aZfFq+vLw5pfPDIPHcnPfJDXMJA84r28BOqm4mtXOrJmsq9UlAdb6+oR0z0pXtErvrDRS7IxXUkNYVBQNFQHpoqTb773cIHMboO6KZmYTZDWcPg+rvuCl6Wtx3qtzTT1/KkGKjU+wSs/Lh/riHkrzvNX+oKIqitdmrcdfOxInatLO5eWZGkkYTOLlxoJGXPzmfPy5vQQXvTnf8m95sNiwTq1phdmSdnngUGXC91JF2+09V8xYHwQhUQ5p8gA19JIHXDVpIf7cXoIb319sSVar6CUPUPs+LSwg6+iGpc3raidvsdoM1SZnU1fuwux1e/HfGeutXYyBPFZ4bFw39y6ugVqdNIo181qfU25+/fjRcrp0YKv4Ma33Ij2s1f6sWAWNHt3tBU6WSPsZ0YRiI4WHp1RTeKWv0kl4QK0Mf8asp6eswYJN+/H5Ym/i14KC8w1ZCCaoNaryqgj2lJajRV19X1Y1QgIfnYnTgV3Z8Vt14Zv020ZM/EF95VV6pZmr91gVTUb1RIyHEq/eC8IuqW6Ov+StGmVQre5K65/buqyZVb1qV7TE5Mm615XtY6N+j72liXuJHKmI4MDhCjRjkBUyUSeUH1BmegsJApY8dBKioojMNO1JhFVLtVoRH6lU99cXRRGb9h1G6/o5ttLdbz3gfnyJ26g9tVE1jXrYZoDECeylg1pjbI9mOFIZwXvzNgPQydJllOxBECxl+OK1O7XjIi5dbIiK3i+COXVrN3JRjIrm35dSme/RIt9Usg01KiNR7CgqQ0uDmCUWCAK/dZIlZLHhBLUJ5Wkvzsbx/5yBJYVFmr/TXlViJZkznLrEKWNsrDbKP7dru9SwbOAOE/HEcSISk03K/MyKxkul1UG6AMGDEqi2QWcMLemkTUprclKpkut36L9n4LinpmPj3kMWpUwkcVNR+ffKSYggVFvycgz8+hvlZup+r0TtHWope09+9xeGPz0T/51h3robe6656/cxKTfZtZlezT5GrUBanF40md2lZQnH6tXKUEZxGV5HqzsySgctxSgrWpCQKzaiqaQRMXh4brXkAdIjZtzrtHASM33pW/Mx9N8zMH3VLtvXMIuZPZiSAVJsGKOVScUItYaxZld1x/H1kuTePV2PBMXG4tRffxBi19taGez0cDIAsBg8WEzWrziuteE5t43s4Pg+XhGNimh973dofe932FFcM2ma9tduJtfPlrjeZaaFcH7fAowfbm4fl5AgJLhB1riiacXYiKp/S1FL27znqBVnxirnz61cPFWKka5YdTdqXXef3AlDOjTAxLOOsSSH2uPPWqNuvX1z9kYA1e4fZvn490Ks2FaMC9+YZ0kuU3CyEGC0oOL13lhz16uvnMs2FzWZmCPxmEWF0iArGiev0BQyty2L4wQP8WXhsH7yABHm09UrH8eqG5t0YWfehv0AgA/mmUu84gS39xPiBVJsGHPjsHaW91IAgP9qbbQJe50fL8kDnOLUB5mVwuHVfZwoFiyUklU7Sx1f45KBLXGNQe7/20Z2TDjGW42NDT6/b9ofP/bkdyvjf//7p9VM7nNqj6aSewL/PKcH7j7Z3AKJIADHtasPAGhztN8xKkczFhvlgoIUFnEBRulZEzZZNGhfwzo1xHtXD7Dstqv2LN8u22HpGkZc8Y532a78wKg2eB5jo5E8QDommvBE065yVlzRzJ/qKXYUDVlsno1Y9ZKySrw0fS02aVguK6qieHXWevy5vdj6xU2QbuBWIYqibcsLi7FXWd/KKiM455XfZGOO83uYq7zSLJRBhBQbF1D6h5vhBR3FRg/NO3EyS2StV1jtP/ReBcsBl9VjOnJFYySDcwTZhLl/63rmfuVynS06XBG3OqihHOxj4kgn+aXl7DOhyeJ2LO57EhKAurUysPyRkzDldnn6Wu3fS2JsbDQCFqnqDxkoNsoB2K2Vxu2M91VRmzDuPVihcqZzpEXSnEHck12MqoPXis1JXRurHpdWKTMLUdp6jQVXNKXFRvGZk2HaMpYtNgAe/Xolnp6yBmP+86vqOW/P2YinfliFsS/MZiBhIiGDfWyiooiIHY0N1uON1PoJ5aGXpq/Dws0H8MavG7Fln/a+alYw34/yM5uwAyk2LsB6ELZzOcsmc5dgbTmyrtho359l02U1KXdisuch5iOGVJbXLu1j6jfSVKsX9mef8rnXYz+j35NTcVBDOdGyUkhfLbONdSVIrQbSemTudVb/oE5WenylOqYUaP1eHvNg/XncqWf613SrL9tsMGGwWj5+7R7/xmV90UInWxwr7DyetL4YTfRZcN0JbVWPm7HGyBRqlXMa1M5ErUzzGbASYmxM/5I/pM9ifSIPLNhU7SKolZxDLx6WBWrJA+TudYDZrcmUj++Gq93iLQfif+87pL0gZwWesre6CSk2LhBmFUl+FLt1kYdO1O92pPcq2CYPYOWKZv+3nyzcykQGp1RnXql5kLq1MvCvs3sY/m5cr+bxvzs2TtxjghVbD6hPZpX+1fG6K3m1ei5adpFeU6sWjeneRPW4Wvsyqol66Z7N8K8fV+OSN+fHFTIv/OftJpZ496r+ju6r97o/W7QVp/znVxTur6lPXva50kWjerUyZOmM3eJpFfdLQ1c0yd8rthdj9PO/YMZqNvFpSto1rKWdNU+msxgn3VE754GxXVAny/wGwomKnLKPCc5E02m/YbTI6XY/YpSe34nFhoUVW1kV5NkE2dQTQRBwz+hOhueRKxqRgIn9wVxHEATXVg9Hd1OfZHmB1c5Pr0NgGdTKLnmAfZke/HIFExlYoJwQntevAFPvOAEAMLRjw4TzJ551DDLSahqOm2Oc1gCrlglMiRv7TkSjYjyZwhNndlc952qNmCW1emeYPECW7tmKpDXMXrcXs49uAOuFlcLuuoFaXbOC3vu+89OlWLmjBI98/Wf8mJdB0sr+67SezQAAfVrVde2eX/yhsl+GwTNLJ2gTPl+OVTtLceU7v7MWzRKmkgeonNM4N8uSZczPxcVYrO9Dp3Zlfm3rrmii74HrRhu3iqL5havE5AE1f5/YuZFV0VSRXpNV0e09WI6bhrXX/L5L01ykhwUMbFuf0R39gYMpePLBOmBdb3Kut9GYG9loMtJCePniY02f7/d6VCrF2PCCAKB3y/yE4+0b1cayR07CpCv6JXyXsBGrS7LpoWWxkSpCbig2kaiI20Z2xNKHT8KZvVuontNUI4ZCrd7FFRuN+0kHTCduZbHVTRYlYiSG0WqvW8lSzJSP1LXGr/0WBQFolp+N5Y+chE+uH+Tpvf1OHjCySyM0PprmW29SKUseoDNuqv0t/Z1WcgI1lIqul9Xj0kGtsPThk3CVQSIXs0hlt1zPRWOrg9vWq2EGixyiKFpwNZafN6xT9bXTQgLuO8VeZlxljZNagQo1vAysUl6lv5L1xU3HYdnDJ1tyt+QRUmxcQKnYuLFiEkMWbKzYCM2NASUvO92S2xXrzsrqI+n5lLIsH+mmhW9c1hdN87JsXYejMBnbCIKAM3o1x7Pn9cT0O0+QfZebpV5/KhQdrnJCMKRDA4byqR+vUihXXmUWjClLedlyFxdpCTTPz1Z1NVJ3RdOPsfl2WU36eCcWBk8zL/q0QqK2gqssc1lclEtT1/ycRPcntSxfdbLSDV1uWGOcPMDdTi0zPYyvbz4e/zq7B+48SdvNxsjNzAxWy9YweYCLr0oUE/sUp9eLYc8Vzej6BnF2Dstq/HBtSwVQrazZtdjcOqID/nV2D8z++4mmfm9mXiQdj27+8A8s21pk6tpOyEoPIzsjHPh5CCk2LqDs/KxuGqfGwk37cdJzs/DbUfcPNRI6Ucd3TcRq38K637ba4LxyRXvpot4Y2rEhPrp2IEZ1bYweLfJsXcfLBAD/mbrWtWuHQgLOOrYF2jasbep8owHlgn4tWYilS6VW8gCX54lm/bPVVqNVrcNxi436df/vt03xvx1ZGAxc3qygdonHxnVT3spz1Cx0Slnki0vuyNG1aa7+CT6axo36Ub0yeePXDSwEQOPcLJzXrwBZ6dorzWaKSDpeqI0drL0xgjqBtJpERQQc11EnZdUkNwsZaSHdmlodY2PvJlnpYZzXrwBN8rJg5kEvf3sBnpmiv11ApcKD4PvlO23JloqQYuMCCRvmOWzRAoCL3pyPNbsO4qI352ueF1VYbNzoNa3268zTPTPcoJNl8bSqXwvvXtUfg47uKaJ27db1Teyx4eFA99xU8xsNWsHOK1daS/wY8BMsNnFXtBrccDUyO5iarfuxOr9im3qWIWmbcOJa99L0dVi7q9S1KnvZoNbxv5XN+Ly+cpe9tg2t7x1mBrUJnAjg1Vnr45+9iP/Wi6XyG6O2qrdYo5Wh0NL9TdZAM6vkRpnTrBhsnp+6BrtLy+QHDZIJsIT9lWuuaH2DTu/2lFPDzK1FEbj86ObSRnHEek9v9jFfVGzxsXRrEV6euS4+DlUpYj4PHKrAcz+vwbYitinq1Qiovh0nzfgUwiondGqIBZJN/Vi0Z6WrjhrSMViA/i67drGqpDF3RbP4UH4FLKrNF585r5fh78w8XmZayNBXNmh0aya3cLm5W7lWlVCukKmdt/cgm7SbUljH7RhWeanrlKRBKRU7IxZtPoBRz/2Cvx4bbel3VshOD+NIZSQhCcA/z+6BB07tiqy0MCJRUXel3glqlkRRBJ76YVX8s7SPc8viapT9Ttkvx8rNC4wVG3/vH8NIaVEeV4+xMT+gPD91Ld6bu1l2zO30zyGhprz1lCY71VTuimb991bG4n0HyxGJimiUW+PSzWIqMaJLI6zbfVByTXnb7dI0F8sfOUm29YBX7Cktx79+XI287HRcPKBVQt8zeWEhAODzP7bi13vMubzZJaiWxBik2LjAdUPbYs2uUny1ZLvxyWbQadDSxi51MXErxsbvzCaWY2x0BP56KZv38/XNg1WO1kj6/a1DsHnfIVPZisys4OVlp2O3zkaTfmN1AGrfqHY8+DKGshhyMtwPZlSukMUexO2gVieKjeqE10BcLQuU3YnwTR8ssvU7KVqK7NQ7T8Dc9fswrlcz2XFBEJBrIe2uXcwoKiGZouiOHGqr3VJlTvn1tDtPwG/r9+HRr/90ZVNZK7i5SAFYUGwE9b/NnG+XfYfkm7O6vY9PTkYaEyuYEdZd0URLi6J9npgKAFj52MnIyXA+TY3d+faRHdG+YW3c/b9lCefE2rqZdN5647TTarN2V7XipTUuFO73wmITbM2GXNFcID0cwjl9alwlvNIF/vXTKtlnVmZu6aQySHn3AegWvlQRdEKPFvkJx6R9UtdmuRhzTFNT1zIzXvhp0jeD2Yn68+f3wqiujfHV+MG69erM3s1xgsO0vVK0kwcYW2zcQLO8TLRftVMMM4hpWBjKKu1ZAWes3mPrd1KU1rIYzfOzcU6fFpYyUbHETF2WlrZb0wG1BZrOTWv2elJ+2+xouWWmu19uRs/My+qvdZdw93sA1hNIsxLbGUKkktpxRbNzzx3FZcYnmSDW52Wlh3FuX/XNn1lZFq3Mkca+8Kvmd8rxyFM4abN2IcXGA5zOQ812yNLJQXllFC8ofDjt8vqlfZlchwXWXdG0yy7DxcmSXaXSzEDHuV5jWrE5o3dzvHFZX9RSMftLr/Dc+b2YbYBajfq1lCmn42e7VN6XD2oFAJqZnOyOLVaKSvquyiQWG7kVQsTrv6yHm5RXeeM2ZRVTio0Hrmhq71Sevlj9pXuhEBr1dVpfT5qzkc39mc7CpGUq/yYzzXlZKiVlXl1MWg+d3teqEvDOnI1YtbPU2U0ZctXgNgCACWM6xzc/vlwS02eEboyNBTn+3K4eBwloj0eEMeSK5hKCwgPa0bVs/PyVWeuxX2EGZ4EdWSaM6YyJP6wyPtEU1npUvXTPaWH3NAS740YyWGzMpszUw9UNOrUsNgq51ZIHsOTRcd1x9+jOpv25W9VPDJBXfRYDgbVcp6SKTXPJJoRT/9qNf3zPqv2qU27TWuQ2li02LtVbNcXFTDeg1Vf0aJGHZVuLnYplCi1l75FvVjK5vukyt5r4xrIkxri9r41U5m7NDDLpWUQqu1UF3oxV10sjwUOndcXtozqgTlY6hnRogEMVEUtxNWO6N8VHC7a4Ipsoiig+Uumru3nADTbWLTa//PILTjvtNDRr1gyCIODLL7+UfS+KIh566CE0bdoU2dnZGDlyJNaudS+tbBBwOg89WGbdZ3bj3kPObipBuiJm51mMdvy1JAvD5AFOVjNvHdEBz5/fC7PuHqb6vd0Jjpc7l7sFi2B4P3x8/Vgh0xtMG+fK90Jq17A2nju/J84+tsbNlaUrmjQhhfS6m/ex60u0KOPVYmOiPcqUDpeqrdoCjVQ0rTeu1V9PurK/c6FU5FD9ntmdnF3fzD428jgcD1zRGBdOKCTg59uH4uWLj8Vx7dnt/QUYu6LdOKwd0/u5TSyWRhAEy8kC9PYmZFFtLnx9nvOLOCDo8xDLM7tDhw6hZ8+e+O9//6v6/b/+9S+88MILePXVVzF//nzUqlULJ598MsrK2PhKBgWWfeKW/dq7znod88JyU75nz+tp+TeWm5tO+ThRbDo2ro0zejdXXUUH7LukmPlZiHMHUiaKjYv96tgXfsVL0xMXW5Ryx2qOXwayJ87ojpFdGuH/rqqZhJ7ZuwUelezvooaRK5oyecCT363EWS/PkSUPkE2aPSgANy02b1xm35XWnCtazd9uKeRGbV7rFWklT6lXKwNPn9sz7objBKNnthpobvn+NrKimTlHeT6Lp1DJscfgqjUIADo0roNTDGI6ndZTZZ4VAKhfK8PC783dn9U4wLoLy84Iaya0YTFHWrlD20WNMMbyFGnMmDF44okncOaZZyZ8J4oinn/+eTzwwAMYN24cevTogXfffRfbt29PsOzEKC8vR0lJiexfsuG0mhvtdvzTnzvx3txNsmNpDGMS5JMcZpfFWce2MD7JIXripjtwRXNr13Uz/bieex0PJGQX44yyyiienpK4h09C8gCfy7lxbhbevLyfbuIE9axoRhabmr9X7yzBG79uxOItRfhpRc0GcF6v2LlpLRvVtbHtfW52lxi7g2w7cARP/7Qae0rLXUttbPhONfojPbfVc/q0wCuX9HEkF8BDcgBzAlhtz240/60HqjNaFR2uwLNTVmP9HrbWUDf7LHm658QytzLnsGOhdfJkbhSLVr3nffwzg/9t2hlM1343btyInTt3YuTIkfFjeXl5GDBgAObOnav6m4kTJyIvLy/+r6BAPWNFkHHa2RjFVFz/3iI8+NWfsmMsg0alddzvKbXVCZde0TlxkTN6pXY7hnE9mxmew3uMjdKFyg5OJtZrd9kLUtWODeK7vJUYV4+aE76UpKSXPr/X4xqv4+gtHy02PGfljhK8NGMd/vbxH64phIZtXuNrL7oKoyd2K6FC/P42LDbKcmnfqDYAyCxYbixenfda9Tzovi+W44Xp67Bg436DX1jDzdctLWa1MrcyntpZAHBSi9xaiFTjcIX7brWiKOIvF606Zram4Bmmis3OndUrfo0bN5Ydb9y4cfw7JRMmTEBxcXH8X2FhIUuRfENp0n5gbBfb19JbCdH6xok1QglP6Z6tdm5udWhGV7Vr6n/8jO64+2T1LFnxe3M+z25RN8fxNZzMhUY994ut37k9AfMKo+qhVX+kk3Iz8Rsscbvs7T7DgcOVps+dv3G/awqaalY0B8kDWGKkzLm+QafJ8/Q235x83UA8e15PTDjF/jhththG26wVmhhejQ1qsWdW5hzJ0NdqjfFevIPJvxdizH+0U0U7ZXT3Jnjpot6Ycdcw1+7hJr5762dmZiI3N1f2L9kQBKBrU/vPpeeKpuWLmZHGZkPDC/sXoK9Ee/d7Tm21P9TrZKQ7EFvFqPOya42ulZmG8cPbG9zb/FsY8cxMfPJ78BYL/Bj2lHXLzc1oL+zvnmXa2G1JHdmKrMdvwG3vDa8WZLzcoFOKZoyNF4qN0feuW2xMuqLpjF71a2firGNb6G56ypK9B9lnLK3GTVc0/axoVtqYaKO9B8UVrWvTXFzYvyX+PrqzveuaOOelGWy28tBCEASc2qMZ2jSw58LrN0wVmyZNqs24u3btkh3ftWtX/LuUQZZdxdlEzc7qBoud2q84rjUmntVD3mHZ6CBO65EYyNjOps+7lUHyk4WF+GrJNlv3MUa/INycGFqZcK/fcwj3fJa4yzLv+LGgpzTt10wm2QtzYf+Wtn+bkx5GwzqZyM1KQxMVtz+7g3hUy2LjgU6QDCu4AtybxNv0RJP97ry+LsU0GjwyLxYbXpixerdr1zbbVr3ex0bJ4coqPPvzGtzy0R815eFmin8XrqnpuCwImHjWMa5miSu2YElORZgqNm3atEGTJk0wbdq0+LGSkhLMnz8fgwYNYnmrpEaZf97OoJ+bzXaLooJ61ftanNzNuoLaKDcLfz02Ov75+hPaYsrtJ9iSw2xJHDhUgXv+t4x5cKZZ3BzM9VZvi48kR4dnts6f24fdZO0/0+SZ0mLl7PbEzCqhkIDf7j0RCx8YperXbncQlz6n5zE2Lt/QycRme9ERc/dwuIClh7LN11IsXGmtlnvhilZSpt7nRKMiVmwrdj2Y2nSMjaD+t5nzWXLlO7+7c2F451FxYqdGCcesWAdfmr4OL0xbi2+Wbne1PGL47ULPmtJy61uApBKWZ78HDx7EunU1ZrCNGzdiyZIlqFevHlq2bInbbrsNTzzxBDp06IA2bdrgwQcfRLNmzXDGGWewlJt7ZLtCW+xulJMVO4N+TgZbxebzGwfjt/V7MdpmetBsyUCcnR42zPTmlMOVbgfw+bd5g14n3feJn927sYeYLb7YXgRuEEuxy6MxQS85iNEgrqWoSZ/T62d2PcbGQXez9cARNMvPNjxPgMCs3OpkpaFUsn+ZdNJ4bp8WuPOkTijVUCikSNNEuxVvqOVW9fy0tXhh2lrXA5HdqjleBpyzwk1FVlq3bxzWDs9NlWeWtPIeVmxXcaF3sbhduTSH4wJRjeXZ78KFCzF8+PD45zvuuAMAcPnll2PSpEm45557cOjQIVx33XUoKirC8ccfjx9//BFZWc4zJQUWwdpEQTnnt7VizLjRNayTiXG9mjO5lqPO18Rzrd1VilNfnG3/HmbEMHS/8McVrTKSJL2tyfJLY5gkQ0msngZtszKjNYO9B7VSGEtd0Wr+dquE7z65E/7902oAXiQPMH6Ktg1rYYOKhbfcZGpaESKz58hKD8sUG6mCcmrPZmiSlyWzlGg9nZ8ZFF84agFdtPmAo+vcNKwdNuw5hP2HK1SD7k3H2FgsiiAu8nslc0Za4sKKlX5S9Vyjn3P2QtxyNw/YcMMlll3Rhg0bBlEUE/5NmjQJQPVq4WOPPYadO3eirKwMU6dORceOHVnLHSgEWGsEysFI2gmYzRVvZsdsv3BirDHzVPd+vly2i7obGAbMunhvzvp3VzBbfk4sf18v3a77fVyxsX0Hf7C70iyz2DCSRQ9pkgy372emzQztoL5fkNmutDIi4qMFWyxIpY2yn5eOCeqB2+rXeWBs9Q7p15/QFtcObQMAOLM3mwUqKbENF0VRxMsz12EmwziSujkZePXSPnjpwt6OrmO1XQSxn9US+Y5R1XOw2H5Od56kn3lTjaIj+gkPRFRvXG0GW9MTJ3OaAL3L9+Zt9luEwMPWX4mII/fntdihKj5LBzJl/I0WPAfjOvF3NbMq5OZmfzVyGH3vpsUmQL20TcwWn5ONaG/96A+crrNvUEioXq1fsa3Y9j38gE3yAInFxpOUwa7fwhCtx7TSl744nU22ImUbl31UEUdr0t6/TT389djouCvwqsdHI1Nltd0pJWWVyM/JwPRVu/GvH1czvz5QHau56vHR6Pzgj7LjdmJszMx0E8qUgzpqhFZbvXVEB1w7pC2yM8I4UhGRuYabZXinRlixrQSdm9RR/V4URfzwt6Fod9/3htfyen4y2kZssBE89FmEOqTYeIQ1VzR55yR1RTPrlsYi4FnN3MwClz3RmG5Oqi2HviTuWmxSQLFRKUFl3AEANKid6ZoM4ZCAa/5vIX5du9e1e7iB3erhtcVGivsxNsaFohX87Mf8RWmJlE6y1dqG3uNJJ7HSdMYsueWjP/De1QOwcS/7ZC3SZ1OT32x9t9oskq2bjdUDO0oNUG1hbdewNoZ21LZsmrWgq81PDL1aHLyQv43sYPu3WpBewy++72OTrAiKv600gpDirchWUk1eKcpAs2G5yacUNYtDwzrsJqhuyS3FOMbGvXu7nHeBC9TKt3X9xBTh5zpMYatvWRNcU2rcnMfbdkWT/p1kyQPMENJoWH7EWOlNEGNJxnjqBmLtxG0XYDXMLvRYfYs8la9Z3FTGstLDOKN3c9SrlaH6vZXy9bpNZTLa109K0GIvUwlSbDzActAilBYb9f0l9IgwUWzcqR5qK6MsfXO9sdgYncCnK1rr+jkMJXEPs6WX7XAFmrdUzixg74rmVKJEhnRoIL+39/PhBLSe04/5i56LJa9V9scVO/CyCxsHnt5L210UMK+ApIUEFNTLRt2cdLQy0Q8G0WLjp8xW2olqnFjAVMkXLzwWAPDIaV19loRQQq5oLiFdRapOA2p+5E6w2Eh+arbzYLEC6paCoNb5mp2sH6mMQBRFWfmu2VWKxrlZyMuuTv3rhg+5EqPVmlFdG2Pp1mLVDRSd4sRi0615HjbtO8xOGJdQK161KuLULS8qighrDKgsVuTyc9JRpLKZWtM897JEBsEV7crBrRX3dveOZtqMVh/kh2Kj5wasVlY8TMJveH8x82uueny0ofucaVc0QcCMO4chKpod2zgoVIv4qRxYSZBkdkFp9c5SNMrNRK6Laf3tMrZHU4zoYlw/Ce8hxcYDrA46iTE2Ulc0c7BQbDJcUmzq5qibss3y1uyNuGZIWwDA8q3FOO2l2aiVEcafRzcBTVNqhi5QXqmvqF5/Qju0blALA9rUZ35vR5N5Xpd7FajGEbhwn0hUhNa4tHRrEfP7NaidgRcvPBaNXFB4Y9i16IkalmE3yj3RKu3CTaT3M/EQWqf44SanlzxATZqgrXabxcyk0cqTq21om0z46abs1GKj1ueP/3Ax6tfKwKIHRzkRzTVIqeGT5G7lnGA1xkY5cZX2AX/tKMF5r841vAaLiQLrWJV/ndMDZx/bAuNUXAusTNb/b+4mAMDPK3fhtJeq96s5VFGz14Rbq5dS9xmjvS3SwyGc2qMZ09ihGFZ2eA4sJurvsS3zHd/mlP/8igmfL1P9TmvjQSd0aZqLQe3YK7tS7NYOeYyN25qG/KOX+9g8eno31XM0LTauSKSP3gQ1VlSp0A2Ywa1kKkEsXz8Ty1hqJxZO3neouh9+8MsVluQhEunfpp7fIngCKTYukdC/SBryVYPb6P42cYNOeS+wYFPiJmVKnCYPyAiHcJpOKlw7nNe3AM+c11N11cxKdxybpFz77kJGkplj4lnHxP8uM7DYuImdsSu2rwTLTcWUcRIsMSPlVcfrtyMzbNh7CB8tKESVBynCecdLVzSd7tF1BrZVVyy1lAk/goT1Jqi8uqL5hVuPHsQi9VVmC+3E6kLG7ybmPDzy4Kl8xd/0a13XbxE8gRQbL1D0NkaDkJ4rmlmcroAuf/Qk1Hcxla4SKwOzX4O49L2Y3Y3cbTnMcsU7vwMITu79fq1VVpZ0LJlOccM6A/gz0bBbLJpJSlxocLGJe2xfrtN6sF1ESbyfmXP4ibFRKllq29i47S4YFNwaDwKZVt9Hka2spSrPNVo8uPiN+TYk8p+rj2+DVY+P9lsMU/RXG3MDCsXYuIS0f0kPh2QxGUZ9j/J7O8aXiIPROCS4kx5RDyv9sdHE3q2JiPS2dXwMZqyTxUezdXPgH9mlEd64rK9sMzg3V/m92NTVK+zWf5krmss2lNi7/OCaAZizbh9GdGnkyf0A7WfjyxVNJ92zqsUmgJNwZrjkiubKVd3FT5mtWDaVWVuN5jgVAe6feYrD0YrF61WQjzcu7+uxNO5BFhsPSAsJliYKiTE2Niw2AesHrAzMRme6NSkTIOCFC3vjzN7NcUH/AlfuYcSlA1thgAM/WZZKn5uDqCAIGNW1MQrq1aRlVYoeaxdPn9vT8f28Umy8WP23Xf+lrmhuJw84etH8nAyM7dHU/cHfRP+i5Yp20weL8d2yHYwFMpBFL92zWsZAF2XhjX+d0wPnSfavcitgXlll3Fb2WRCUGBvlnIaHfaxSAa06PKZ7k3hW2WSAFBuXkPYvaaGQpQmNsqOusmGycdJR+NE5Woqx8c0VDTi9ZzM8d34vzy1aMR4/o7uj98NycPY6A4+Wgt+3lXO/YRb7PqmhfFf3jO7kyn2k2F3UkPqxS0tj/kb2/u1uZvF68szuAIBXLj5W9XutrlGvWY3/kH0qYzUywiGM7dFUP3nA0f961Q+a3U3eK87rW4B/nVOzmOGaK5qijvJWDmoEJSua8tQKHzZ2TWWk8cJA8u3nRoqNB6QpsotZjbGxM+lypNjY/qV9jrUwOTWa2Lu2+MPJuGY3mHnrAbb713itACsfm2V2qMoI+0pzyjFNZJ9XPzEaPVrkM7+PErvK6+7S8ppriCL2lJZj5fYSfLN0OyvR4rhZdS4e0AqrnxiNMcc0rbmfKZn8b+DLHz0JL13YWyXdc83n2GbGDSQxkG6Kzvt83i0lWVmmtTP5cAHWI+zBVgcxlBm2rPQ6yvnJB/M3Byb+Mxm4sH9LrH6iJvYn2Sxm/LfUJCA9LFiz2Cj6JjsWGyeeNV6O77/cPRzr9xxEfo55M6ixK5o72N0fhBeO/+cMjHQ5lsFNlJ1vbALPYmLD2mLz8sXH4oSODTHkXzPix7yy8rEYo0QR6PfkVOcX0sDtlqQsa9k+MBrlw0P7jsmtJsvUO07ArpIydG5SnXAhPycDn904CJlpYVeVsuprJ9fExw45GfxPl7xUQt+8vC/mrN2LGz+otmZaWXBTnro5AJtGJxvSPtKPzI9uwn9LDSw1PUw4FJINC0aDkPJ77y023vWOLevnoGX9HCwpLDL9G7/mH/5Pe6px0gVVMLRMeL2Sm5hJh921KxkHpZ0isRZ4DRPFxvkl9PG47pi5HU+WCbU+rn2j2mjfqLbsWJ9W7mcy4n3fLPeyosk/52TwEwSuhZfucrlZ6TKrqJMNOkOCkNIpy/0myfQackXzgjSLnY1yta7KxqTL0Qo05x3Mml0H8eK0tZrfu5cVjfOC8RyvXdEUFhuGrmhVLrii+QWLOCq3V/C8XDwBzLVdHiw2MXjqa3hS+NTwKt1zLHaLZ/yow2O6N0FmWghnHtvc9G+U05NwyJpXixXuO6WzOxcOIMp+9/SezZARDuH8fv4kQ3ILsti4hCx5QFiwNFFQDiQRG5Ou4iOVln8Tw49xzKqF6Zmf1+h8604PyfsAbwanG7dK8XoMVVaRLk1zmV37YLn99sIbLF6x6xt0+mixESFibI+mCZnOONIluOpreFL4pNTKCONQRQTDOrnjXit96klX9vPEOuYUPxIcvHzxsaiMiMhIq14nz89JR9Fh/f5UOd67VcV+uXs4WtbPMT4xRVAuev3ngl6yd5csJNfTcIS0naZZDOhLtNh4u5rsx0AWBFOo16vMWjgpK5axJOlhjy02kk75g2sGoGszdorNVZMWMruWFD9qTC0GLjNut0e/W5La/XmykhTU5Wcyppd62k9m3DUMb1zWF+cc28L4ZBtIq4PfE78Rnc0pb364DQqCICsfMwlSlAtsbsmdmU5TXD2U7y5ZSL4n4gRpxpqQIF8BNWrCyjbuVipas/f3ApauL8meFc0JTrOfSOuGlxl4APl7Hdy+QfxvjuajXNChcR3jk3zGayXCzO14mr/fdXInnNnbvGuPm/Ca5rhRbhZGdW3sieKltaiVn5PuyYbJZvd58rhLVuXpc3oYnlNSViX77FYdq1crw5XrEnzDQTNITprnZ6N781x0alwH+TnWGpeyE/XaYuMHLJ/QvaxoLl3YJCO7NAbg7PkOHK5gIwz8SB6Q/O2AFSd3a+y3CLp474oWrBibvOx0PHd+L7/FAOB/v+cX0jqjVTXystPx3Hm9vBDGFDwooY1ysyz/pl2j2swV1MHt6yM9TFPcVIRibFwiFBLw9fjjERXFxMA4g/bru8XG07tVwyL2Y8HG/UfL2pvNFr3mkoEtHV9jza6DDCTxB+2NFf0fzHlDOil77+r+aJKbhVHP/eKjRHI8f2OKdM9qdYaDOaEmflbxVG1f0sfWKwEvrCRm3wBPyrkVwoKATEYuUf+7YRC6N89jdj0ieJBi4yKhkIBQvEsyN9kWhMROzE5WNCf4MZCx0N3Oe22u84vo4PeQwcNqnBSvDShkrzGPtAmHBIG7OAk/kweofQZSdwJvBO/pnt1C+tR+971m66bfctolIorM9vkSBPOue0RyQiotZwhITYsNixS1buPlathtIzskHItNMPzcTMtPbzAtV7RgDuXuwv9clD8Bg7ra7TYBnSs7R7o4oFEIArzpE82+gqAqoVURER8t2MLoasEsA4IdpNh4hNnOT21lxvMYG1+SB3h/T6t4OWaopTDlbdXdazQVm9QuFlWk/Ygo8te+PLfYBCx5AE+kqiVL6s6ppzDw1LaCOkas3FHstwhEEkGKjUd0bFKTqUgvkFVQ+d6rjrN/m+o8/ef28X6zpiAEhns5vqvdijc3A6/fGK9VJBwScPmgVqrfndOnOhVt75b5HkqkZn3gtPB8oDrGJvE4zxYbP1PN89bveIU8A6SGxUYQPGlZZqtmUC02pYosaQThBIqx8Yh2DWvjk+sHoWGdTEz+vVDzPEHwbwX6rcv7Yt6G/RjasYHxyYzhddIqxcvJhfrEy7Pbm6JZvvXsN07QTB7gs+uB3sTvzpM6oV/reujf1tvN/XirK0q8Fs9MHeF5Ttg4N9P4JJfwSrHp36YeFmzcH/985eDWntxXC+lTaym91a5o7g9eZt9AQb1sV+Vwi6+WbGd2LT9dtQk+IMXGQ2IWEV4H0DpZ6RjV1Z80sWSxUdxLZSizutGrW/zfVf3x7dLtuOXEDujWLA/zN+zDB/NZ+Udro1VH/I7P0qsWGWkhjPShTXHaxcTx2r3JTB3h0eXq9Uv7YPqq3bj8uNa+yeBVsdTJlE9H8rLTvbmxBtL6oNf1emOxMX4J5/Rpgb+N7OiBNHxTVultsiWCP/iYKaUYV0oGqYdP6yr7TstNwi7XDW3L7mIuMrBtfTTL89YCYBUvXVXUbhVbOfVbBzyhY0P8+9yeqJ2ZhtN7NsOTZx7jyX3/c0FvZKSF8Ni4bp7czyw8quQ8TtKl+Oky07kp/xuYxjipWxM8dXYPX7M8efWmeGtHZiw2AD8Wm6eP9smpzqEKcmtLdUix8YFGuVlY9+QYrHtyTNwHXw67oSQoHV1Wehi//v1Ev8XQxe+pYqpvNta/TT2sfPRkXDaotey434oewPekDOCjjKRk+LjHRHo45Htb5okL+unHVPqlJPtdZ83uY+OJnFRhTXOYFJuUJ7VnSj6SFg4hzYOJalZ6CN/derzr92EB70GqnrqiqdwrLXzUYsPdNLqGPq3qunp9tTbje2n4LkAiSpF4EzE97G1bb1Wvlqf3CxKPnN4Nb17WV/N7zyw2Cg3B7zordQfWVO4Ej1zRDN5CGudjp5fkZARjMZdwD6oBPqPsMAUBKNx/mOk9ujXLY3q9VMXLlUv1GBv+B696tTI8vycFiwYPry02953SBVFRxNmqFnI+uHFYO1/um5Ue1o0D49yr0RM0s6LBo31sNN7B6T2bIS0s4NohwXA5d4twSMCz5/XE4s0HMKqLP3HCBD+QYuMzav3V7HV7GV6fRqUgorai7YWFL4jwoNfwIIMU3pU9r90q83LS8e9ze+qe43dPea7PStcLF/bGPf9bKgu+Pq5dfew7WOHJ/RNqrN91WLpBp07l8MKCrnX75nWz8ffRnV2/fxAY16s5xvVq7rcYBAfQTMlnaDWMb/zKzKMWLJzGSfIAgn94ryJBsD56jd/76Jzesxn+fHR03HLUt1VdfHDNAN/GKL/rsCBTbHT2sbEh6IQxpIywhHoTQgopNgRX8Kbo/fNsbzJ+KdFTbAh3OaNXM79FYA5vyrDf1seCejkJx7Iz/Ms8Bviv2ADVLj13jOqId67oh3eu7Odrdj2/66z0yfWKwY6YVi2WWvf3u4x4gYqBkEKuaD6jdBWr/py6zfTty/vhykm/+y1GnK5N83DriA5oWNvb2JGs9MSBz+/JIK+wHtxrWcwkyGMyB2mZiEf/xxMZPtflG4e1w96D5RjdvSk27jmIkrIqtKjr7+aGnGxThfRwCMM7N4p/9kq54XmSzjrds9VEOVou5by1a4LgAU660tRF2V+y7qg4WAS0hHRA5QFBAO4Y1RGXKlIMu42exeZ8g/SshDOstpnKCN+Ti9jGwLzQun6OquLuJTkZaZh4Vg+c0LEhrhjcBreO6JBwjteJMHiw2KjBp1TuI23Vynfz2LhuqJ2ZhqfP7elr8gCiGt5jCmPQe/QGUmwIgkPUXBVi6Z7zczJwL/lox2G+GGBjKsfbyqlUmsy0MJrk8rP57eTrB3G/gShQrdj854Jent2PW8XGQKzB7evbvvY7V/SL/52YotzfNiVKNrBXGlguG9Qayx4+Cb0K8m1du7wqYul8zXfAV7dDGGCnjWf7uDlvUCHFhiB04GmuIXXf8XPndt5gvViXDEUbVRRKfk4GOjep45M0coJUvF4qG7yG0LVrWFv3eydlJHXJStjHxudJe1iSmTIzLXFyGYolc/FoJxs1SK8JFnZaCu/7+/EIxdj4jN+TqAFt6qFdo9oY072Jv4Jwit+DqxTpKrff9YYnWL+iZC3ars1ysWpnqep3BfWyUbj/iCdycNSkDPFSseHVivXI6d3w9dLtmt87kZt1UD5Lamem4YGxXRCJisjL0c6OGY2qH2/ToBY27j2k+p1Vq7B28gC/S4kPglIK1f2JNWn57BX4hiw2nOH1vjNZ6WH848xjMKRDQ0/vSziDV7cVP2hVL4fpho92JmrS+cX1J3CwWZ7FkX4k403t6mSpr5n1bpmPRnUymd6LFWpzRC+bGa8Ls/VqZeDaIW00v3cit3S8O65dA7RpUMv+xVzgmiFtcf0J+hunajW1IR0aMJOD06pBWMROf/LCRb3ZC5LkkGLjM1YUmQfGdrF+fYOWFJSVDr/gdUGM10mQH4RCAr6+ebDfYsSZMMZ6O/Ub1htmLn3oJNXjn994HLeWCTW8lJTnxQq9ftCJ1NJ+LD0sYNodJ5i6J09oWU303qfVV03pnuWMU6Tk57flyLHTxnsX5DPLIjm4HTtlm2dIsfEZK/Xczj4msV90a5ar+j2ZsvXxO4BVCzcmh7ec2N7xNfyqTiwnhQ05tShYQW2fFj1Y+3GHNK7Hs1KjZmVyW97hnWos5TwrNno4KiPFT6X1hte+V0mD2ur9Bcv3qbUA2jTf3xTlftFK0b8Fo6bYW5AUBEE2TxzdrYnlRe4F943A5OsG4rj2pNgQnOFk8vHapX1wQb8C3DRMblZXBhnzwLe3HO+3CHE4LB4A7lhsBge402NZHMM7NfJ9nxWn3HJie1wysCU+vHaAqfMpGQXQtmFt3Dmqo+yY28Ui3RRU4LjK6XWDTorIa9drNxjWqSGuG9oWL1wodxnS66OtKoPK09+5oh8uG9QKlw5sZek6ycKNw5wvwvmBnUWAkCB//4Pb10dti3utNcrNwoC29rMXBg2Ou9LUwEo1t+MuEmsQLerm4Kmze+CWEzugpWS1IxLlb+bevXkebjDwa/YKXkrniuNayz5LO8gzFGb5VERtvLhuaFvkZafj4dO6WrpWOCTg7xbTafNST2LUykzDE2ccg+NMuh646dp47ZA2qJuTjgdPtfYe/OAWyX42AtgqzE+c0T3hmHThhGeLja4rmgO5desdb41KA0EQcN8pXXB6T3k/rGW1BKzXK+X5wzs3wmPjujONLQwS0gWBIGGnqQiCIF8AEARURjQyVhAASLHxHSuDgp3GrNytPjsjjFl3D4t/ruJ0c0Fe3BB4cNW7/oS2eOT0brJj0knQPaNpTxs1Lh3YCkseGoUeLfIt/S4kBMdnmxVuulyd27cAix8chauP1w5A5xERbMvljN7NE45JLeY8W830LPt2XKRjSMtXeQv/e15nMHVF47humOWHvw3BxQNaOr6OWlFwMEybwk6dUDYvAUDvlnXZCJSkULpnn7FSzdVy6RuhNlhKO8lKDi02PMFF6Rhka2I1gAZ76EyUPhwSbE0IBEFAo9zgx9lYwQ1rwW/3nojdpeXo2JiP/XPswNKSpdYXSxdegzp3DTlYHpWWLy+LWazQc7Cw+qQt6gY/liYkCFxbJb3ATn8SUsTYANVeLSO7NMLUv3azESzJIItNgLBlsTFoSVW8mjQVPX+r+jm4ZKB8tWdAm3rui8HBWKvmLhgS1P92QpBXBdVEj8WkWX2skACc0r0prh3SBq9eciwD6fjHDVe0ZvnZtndm5wEB5uuOXj97/yld8PBpXVX7b6lFOKiTvrADzUb6yAkWGx46XwfoWeCsPttlg1pzs8GuXQSBTT8TzFZSDYs2HruEGavNrSM64Olzezq+Z9AgxcZnrNTzLBv+tEYJB/h1RZPz2LjuyE6XTwxYB7url5X/5aNmVJNv1hnkrp4NaiVgdxAJCQJCIQH3j+2K0d2b2r5/kNCLB0hlzAa3p0l2qT+meZ7su2uHtsWVg9Xd8KIyxcaGgB6hNxEPO5Jbb/Lv5LocoKvYWLtURloowR05aFQvFHBcyT1A+fi1M9PQs0We+slHCQn2U2zcMaojzunTwuavgwspNj6T0NB1anBmup0YG/0mUam1bbLPKAfS6swgQsIxlqhdr2GdLLY3sYGaf7tUVC8nRCw3nXOb2Cq61eKxoxCl+HidtJh9r+kSq8Xwzo2Qr7NTvRTpogXPFhtpD6TMxOVEbrkrmvY9g4heqSj79Aa1M9wVhgOqLTb81nEvsOcarfgc+GU096EYmwBhJ0gzsBYbhVhqHaLbqz8/3TYUednmJihu0iw/Ublyw2KTrqMEX9i/AKf1aIZeLfN1r9G6vrX9U1ihVgahuCuatfKx41kT9AE76PK7hdlySU8LAeWx35hXpqXdHM+vQNofVykWw8jap47e+1SOus+f3xuCANzw3iKUlle5Kpd/CLpxR6avwnNDMUDpnijAWIGvjrGp+V0y7LPmNmSx4Q0ReFTD5GxnHxujTDs87mMDJDZ2QWWy4Gb/9ukNg9DJZ5/m967ujyuOa43LFamelTiZV+RKNiXMSg/jsXHade+49g2Qk6G+FvLJ9YNw+aBWuE2xD4hXqBWB3X2f7GWuCe5gC+gHOqcytVU27VRDuugUEgTT1gapZZrnCZs0sL9SsRjmJJub1GLF6VBkG72VdeW4m5UewuD2DXQ9LPitHeYIMbLYBLkcEh5fMK73ymFsZJdGTGVKRmg445CWGqveLC02T5/bE/VqZSRsKsYLqhYbQeWYS/Rr7X5iAiOGdGiIR07vZpgNz4lpWrraKgjVQaqZKrFcRvfo36YeHh3X3fLGYaxQTR4geOeKFnTIvUGdHJMJW8IyxQaqbUgNXheWlEjFVO6h4cRioxe7E5Ci0USrGxnXq1nCs8XO1XvkY1vVRecmdXDKMU2YyOc1gmAvS2UyoR7Fq1/RBUWMTaqXoRlIseEQrWprZwW6a7Nc1ePn9GmBRQ+MxLGc5kMf3rmh7LPaZJO1B0SQBlJZR+egFauttqoFOgcxFat9i43xOUp3gKArQwEX3zWa5llPsysIgqYLa71a8liKQQHZDVza+pWKjRNrnyj7W1R8F7w+R4pWk/rPBb1VFLrqsy9TxC8BQLuGtQBUb9D9w9+G4OWL+zCU0jsE8J0gwwvUlBJT844ULzerkGITIKxO1C7sX4AWdbVjHnjW/Id0aIjHJTt1V/utK5MH8Cu/lzgpBVmsztEr3XWSP+5kTlCzONhN92ymXbxxWV9ZNpugD9gs21KnAO9boyQvOx1Tbh9qmLlIvqIKzQWj6XeegP/dMAi/3z8SX9x0HHoGJB22dPLVsl4t2XdOrH1RnX3UgrTQpIae+MrHjjW/W0d0SDj3i/GDJecFt6Oh5AHqc7ig13MeIcWGMwRBu/OyqtgMDMhqoBZ9JJOD6nKRfx/kTp4lrLISxS6TFsCAC7UiiD2b2sRLrymZaWa1MsKok1WzKh/0qshKMbtkYEu8dUVfNhfjhI6N66BV/Vq650j7opAg4L6xXXD5oFb47MbjZOfl52Sgb+t6aFgnE71b1lVN5c4nNYLecmJ72TdO6n5wnp8tCa5oR/+r7Hsb1clEbpb/CWxYIEBgYoULcl+r7GcFmEtWE+BH9gXmM5hHHnkk7ksZ+9e5c2fWt0kqTu7WOP63WpB8jDQnWzwHEEE26XbfFS2oCIK8DlkhJLPYaBPEVSU9xVfvOzOKolLRDtpKZCOlKx2jxvTEGcfoWomDipXXGxKA3Kx0PDquO/q00nf1DcomlFIxa2WmydK+6xXNZYMSXatk14WIa45vgxZ1s3HxAP1zg06T3CzccTS5ijK2StofSTeiDlq/oocgABv3HnJ+nQBP85XjjiAISfWOecGVmXK3bt2wY8eO+L/Zs2e7cZuk4bVLa1Y4BQiag6hVvSYgY6YmyomjslhSuUNQls1rl/ZFKxuplu3GoQQJtWqi99TKenWFQVa66nsEqxzJ+mmM3SKx0i91DKjbntmkB8M76WdwEkXggVO74td7hifEJQVF6TPL3Aknxl3NlM8mrTFPnHFM/O+gds9mY8eePreny5LwhdrrNNP3qsfmJFf7YIkrKYzS0tLQpEkwM3f4jaASSxIj1Sw2UtQ6eObJA9heLhBIq5ReB8t72VidhFZPPtWfSpmM4f6xXSCKIv5v7mbFPSXWroBOQGIEdQLlJmEL/a2RdVmLZvnZ+OFvQ7jYL0sP5RxK+lnfMqp/3ZiClIyKtfKJpM/YLF+elEJLGc4wmV2PN96+oh+WbS3C5N8L8fkf2wCoz21s9TsqvzmmuX4MHC8kLigBBnuoEzZwpdWsXbsWzZo1Q9u2bXHxxRdjy5YtmueWl5ejpKRE9i+VEaA9GFhdXQ9+Vhm53zqtMqsTGxTtLOBIB9QgT24t1wXdGBv5l+nhEAYkrEDKh+iglZ3yGVPZ+qmFnfT6gHV/+C5NcxMmuryhHEvMWmyM6pXeZYI9eunLf06fFhjdrWbxV6uY0gMY7wgA2RlhDGhbX7GdQKKfvZ1+J12lXb52aTAyxanN4cyUQVDrgV8wL60BAwZg0qRJ+PHHH/HKK69g48aNGDJkCEpLS1XPnzhxIvLy8uL/CgoKWIsUKPR9/+Wfrx3SxmVp/EW+Cup+VrSg7CmhxEkxyPffCO7kVk9y665oJu8Z4BgbJUFTzLxANikzOFdeF9yRx0+Ufa/ZoH+1iVzvlvmS6yRXVrTz+rYwdV5aOIQJpxjHHgfVYhNDUPydoZigW+02G9TOxDtX9gcAvHNFPzSonYlJV/bjfmEghnKcEFSOSRnXqxmAasWtYZ1M/OeCXi5KlzwwbzVjxozBueeeix49euDkk0/G999/j6KiInzyySeq50+YMAHFxcXxf4WFhaxFChQCtAdRaQM4r28L3D+2qycy8YDVzFZ2CNJAKp1YxPaisWOhk+5jE/C5uSX0BhO175RHGudmBnrTND03GaIaKxYbmXU5CTWbsNJfRuaKpv07te8+l2SKU+tz6+ZUu+WN7GovIYqfPHp6d9nns3o3BwD8TSWNs7TOJJvFJoZycTLdod/V7/ePQP821ZtnD+/cCL/fPwLDDOK4eCCmzJ/XV75wL6h4okj5zwW9AQB9WtXFgvtGYFyv5m6JmFS4vk14fn4+OnbsiHXr1ql+n5mZiczMTNXvUhIdzSZN0ikEOTOIWWSuPiF1/9RU5UhFVfzv2ETKuStacAvUeoyN3nf6F5t25wmok5UuUwYK6gZjxTCGUpEJ8rt3C7XNa82QjEqisiykiyh6Y1FIEDB3wom48PV52LTvcPX5kmvVVWxYCgAz7xqOjfsOoVdA9viRonz1/zjrGJzfr0A1O55s0q9RhkG32EgJCUKComa131HLKhYEPrhmAFZuL0Gvgnw89u1K2XdmQwyUzxqkhVivcV2xOXjwINavX49LL73U7VslBdXZv9QremZaOP53QNqzI4yyorFW7gQhOJ3FkcpIwjE7q3thmQ+0I5F8RTpAZqSF8Pz5veKf1d6pFZdPJe0a1q6+huTYWce2wPo9hzCwbT0z4nKH2jPXr5WBfYcqTP3+2iFtcGLn4K2w62ElV0uyu6IpJ19WXNGa5mWjU5M6ccUGAF68sDc27VVXXvJy0tErJ/F40BAAZKWHVeLzEknTsGRkpYdVjwcFmVUKQE6G/HmCPOZYIScjDX1b11PNZEaLSuxhvhxw1113YdasWdi0aRN+++03nHnmmQiHw7jwwgtZ3yopUe6PoXeeEUGZpJtBdbLAuD/44JoBaFQnE29cxv8Gg4crEhWbF46ara0QVgZ3asB7XZIODreN7IBTjmka/6yq2Ji8lh7S08IhAfeO6RwItwg11B7Zymrx/WO7YlC7YG8IrESahdJoZVgeS5B8E5VExUZisdG1fsb+Kz/ptJ7NcIuKe1bQkT6nlS5TWb4Pn9YVTfOy8Njp3RhJ5g+CvGHg+hPayb5PtUm9Wj9itwhSrOgswVyx2bp1Ky688EJ06tQJ5513HurXr4958+ahYcOGrG+VlBjV1YZHN9Y7qZtxOu0gmvLlKCbdKoF3LDmuXQMsuH8kRgXAt3tIh+r2VEuyAnZMizxMGGNtM1zpeBrklWY92SMqmo3VuIDuAUknahe1CUZmErnB2MHuHk9BbkdaKMtClu5Z53exiVyqTMKsPKf0XGU815WD22DuhBFo3aAWI8n8R4CABrUzZWNWMrYVKwiwr9wZLTYqrWOpBHNXtI8//pj1JVMKQcXlSsrUO07Apr2H0KOF8USr7VGXmaCi1/FXf2+9Q2hQOxN7D5Y7EYsL+rSqi29uPh4t6prbD0GLVvVrYenWYhO/5dtkozcJjaj4zajVne9vHYLc7DTV7wrq5eCn24bGA5uPXsWWrH7RVjJJ0tv5PEbQ3WCcYil5QJLEqmmRqNjU1B89t7RwXLFJvjJRw+5TpgU8SYAWyuQBgLIupUa90OKYFnmulMA3Nx+PgnrBivtkiesxNoQ1BOgPAnnZ6egZeEuMddLCIZUYG+u0qp+TFIoNUN0pKjGbkemGoy4BJ3RsiK+XbgcQ7CFG2maUK1lqKWXViqlrs1zde3RqIt8lPmhztWuGtEVpWRVGdGmMv338h+w7tUdJdYuNpXTP0r8DVi/MoFTypC1KL2VzTMlLRmVPDSsKnPRcu3sm8Y6aZU++COCtPDzw7Hk9MX/DftTKTMONw9rhwS9X2LqO3lKj2twglSDFhjPMxtikAtJiSAupbdDp7JrJiNmB4vj2DXB8hwZYtHl//FiQV1VZWGysErTSykoPY8IpXQAAtTKNu/5kyshkB63J5o3D2uHThYU4XBFRjXVLxkm88pmkykxVNKr5u9jPkq9E1LHynNJz7bo9Bgk1t8RkbCtGnHVsC5x1bM1+R/eP7YLl24px1fFt8LgkY9odozr6IV5SkNojF5ckY+ipPaSTT7WO306fOEYSVJ6MWA98N7d6xn/ygJq/lZlnzFpsrBLkMfniAS1ln9VjbFLTFS3mshfbHE/J30d3xsIHRqF7M8mqqIrLTTIxvHN1Uoys9Oopg1SXUVs4iBGrV2pZHJMReQpnfaTllqwWGymC4r9AcrYVqxTUy8Gce0/E1ce3iR+7dUQH3GqQXIOKThuy2HAGWWxqkE5I00KJabCtqIA3nNAOY49piuyMMB5nJiF/ONkcUKpILnxgJPYdrMDJz//CQizX0VPo1BeUnTeyDo3q4Kc/dzm+jh9cPqg1WtTNwbXvLgSgvnDQpkEtzF6312vRfOfrW47H+t0HTcUxxpCWXjKuQvcqyMd3tx6P5kd3eJeqMnqKTaxeKWMBkxWZS6zBudJySw2LTey/QsIxwjqcrzX6Cik2nKHVv2UkaXChHsqO34krWk5GGMe0yMO63QcZSccnZsfHGn9n9d82qJ2JBrVrNs7lfQCSpVlV9PhqWdFYzCPGD2+PymgUJ3U1zlDIG6GQgFFdG+Pxcd2QmRZGRlpNgdx9cieUVUYwtGNDvDdvs49S+kPtzLTEOEaV+iLbpDIFJmvdJBYq6SNW6Vpsqv/7txEdkBYScEZv2jk9RkS2cJec47s8xuaoK5rkezMuwbeO6IAXpq1lLBmRzJBiwxnVjT+xsaeiv3tVhF3Hrzf4JhOmV4uPniYbeAK8j410xVMpqporGovJZ3ZGGBPGdHF+IR+5dFBrAMCMVbvjx07o2BDdm+dh8ZYDPkkVbJLRYqPH4XJtN7NYn5Kfk4H7x3b1SqRAEJW6omls0Bl0ZMr/0SHcirseAPQqSO1AeMI6qTdb5hwtV7T0JO349JC5ooUTHc+sBIDHBpEmeVmq39dKkpzvYQeTqiDPx/QsMNEUUWpZo1akPVM8244Urb1cgtyOzCJ9xqIjFZrnpYKLlV2ki21O+u2goJZj0MwiQNemqdfnJH9tcBey2HCGur0GCCepqVoPo47fSuOPXat2Zhp+uXs40sICjntqOgCgY+PamHzdIEey8oJ5g421DDW8W2z00j3Xq5WRcP6ukuRI+c0MleB3tfrw8XWDsHZ3KU5/aY5HgvGBWjyfdOEl1TI9SR+x5EiV5nmprNcYPXptSWZCJ7GRPKPmEWC1rTTJy8K0O09AnSyarhLmoJrCGYIgqFoiameasyhkpoVwzZA2GHp0Z/ogI42xCTmMsZFmnWlZP0f2XacmuairMvkNIlYnVdKBJ1nGVlHhjNa7ZV2fJAkOaq9erS5lZ4TRo0W+6/IEmWRpR2bRS/ecCkqeFkZrQQX1cvDIaV2Rn5McY48RTrKitQv4ZuNuwPtio5+QYhMQXrroWFPnpYdDuPvkzi5L4w1Gu6NbyYp2dp8Wmt8l09Br1rCnNqAkS6Jx6vCdoWbNIxKRVjNp2wnyflBmkT6vnqdnsloiWHHF4DbGJyUJNVnREo8pOb59A+wqKcOoro3dF4xDUqALcRVSbDhDENQn2t2bm/MzVe7hEWRi+0loYaXxt9G5VjJ1Iqb3sVE7pvNTpRWEZ4IjKT+oZfVKpnbhNtI4k1SwUkgfUS+GjfSa1EZN+ZctAmgspl0ysCVGd0/uPefUqJOZhtLyKgzv1MhvUQINKTacQfvY1FC/diZm3jUMORpueKyKqX0SmbmdTKqSZkKWRMq9V6i5hyRNfWCAWlFIq5k0ZisVSk36jGpZB2Okch2ixAlyzFhsMtNC+Pym49C1aa53gnHE7HtPxM7iMnRqUsfW7+8c1RGXDGzFWKrgQYoNZ4SExI0oU5nWEkuLkxgbNT6/6ThM+2sXrh3a1tmFOMK0xUYtGYPOT9MCtI8SqTVs0KtL/7thEKav2g1BAAa2re+hVPyTEnleJHVD1xWNFBtCgbRUlPWjVmaabL+kVCMvOx152em2f9++Ue2kiRd2Aik2nCGALDbmcVZQx7asi2OTLLDcrP6hVsfUJiH3ndIZHy8oxG0jOjiUzDuy0pMjdbeXyPeWqP6gNy/r27oe+rau57JUfKM1n0+NGJsaIuSKpkoqpHA2Qp4VLfZfabpn+flUYs6gKldNKqwtBYpUGBTtorRkGRVVTpLsTWMFJ/VHbRJy3dB2mH7XMDTKVd//h0dyKS2oZeTB7/L/EhoTLsms7bJBNe4fqWalIFc0dVL40VVR80Rh7YWRSqjHvVIBAqTYcIcAZyECqeSGY9SEU7GJKycS7Rupxw+pJw9IjhJrp/HMhDVSce8sLVRjbCR/S92OUsFKIUseoKfYpEJhaECuaHLUF0yUZURlRjiHRi7eEPQHilQmcXWHOkElUle0/q3r4d2r+quel4xF9+ZlfTFhTGcMopgPy8hd0appEiArnR9Iu2npPlmpECMpd0XTPi+V5/ZpqfzwR5FaFWKLbvINOuXnJ+O45CVUftWQYsMhEQuKzRNndJelRU5mnYjWdoyRKnt/H9MZzfKzfZTGW0Z2bYzrT2hHCq8N1LKiZaegK6ddpNatVJjPStuY3hYDqeyKlsrWqjjSGJv4f2vKRWnVohJzBpVfNaTY8IZobS+aSwa2wvS7hrknD0coB0mjMTMVJ7jSMtJ//NQrG8I6D4zt4rcIXKBmhZGuRkstpanQ78gsNjrjVSq7Y1HyADlqrmhWx3RCn1Toe8xAig1niNA37WtRN6c6RWC/NsmbqSg9TJ2gEdLB1Ezx1KPUkASgqCzUsKySahYb6YbReht0pnIfncrWqhiyDTpjrmiSYwmKDfU9jmhUJ9NvEbiA0gdxhiiKuukztfhy/GBM/r0QVx3fxgWp+CA9Ta6HUyeYiNlJVWw8adOgFh4f1w31a1OHmMqoZUUj9NGMsUmBArz75E6ok5WGMd2b4vzX5mqel8qTe3JFU0eW7jmk/M5jYZKENy7riy37D6NnQb7fonABKTacIcJe8oBW9WvhntGd2QvEERnKTVqoE0xAEMxNsKTfXDqotXsCEYGDmpU5pN201OUqFdLM18pMw50ndQJg4IqWwjPVAO1p7CnSGqGsH6lbW5wxqmtjv0XgCmp6nBEVRcqKpkFGgsVGn1QcU8Oy7ExEjIsGtPRbBK7RaiupYH2wi1a65/wc+zuHBxE9D4NUrj59WyWvW7hZVOOFpRkYE2JsUrjCWISmidqQYsMZoqg/UKQy6YolMKNOMMHCkwKEZIOG9nmpNoA8Pq47Pr5uoN9icIs8K1pq1Q0zqG+GV4O0z87NTi3Fpv/RuM4GtTPx021D8eSZ3ePfpWJdWvrQSZh19zAU1MvxWxQukVlsyF2PcAFyReMMUZS7ot0zuhN6tcj3TyCOSFBsDM5PC6depylzRdMpoVQrmXBIQJemuX6LEQgEjb9TGbXVUelqtPTvzLTUWlB57vxemDRnE87rW4CW9XOwtLDIb5F8JS8nHXkpZrWzgizGhvaxIVyAFBsOkWZFu2lYe/8E4YwEVzSDTlCpCKUCtAJG2CEVV9bdItXiShrUzsRdJ3eKfzaybhGpjXSIonTPhBuk3syPc0RRRI8WecYnpiDpCZt5kSuaEvOuaO7LQgQHQaPeUD2pRm2qrqUMpvriAvn+EzFU241k3FZmjqNMpwQLyGLDGSKAxrlZ+PWe4cjNInO2DItm61S02EhXwJSueJcObIX35m0GQAMIoQ3VjUTUJutaLmepbv2iEFEihpGSS65o9qFmpk3qzfw4J9YRFNTLIT9dBYmbeSVyycCa7FfpaanXS8oUm6OjxifXD8I9ozvh7D4t/BKLOx4/o7vxSSmEVktJvRakjpp7lVSxSXVlRgpl9ST0kDYVSvdMuAFZbDiD/JO1kXZ6J3VtnNALXnFcazTJy4p/TkmLjeSRY7uh929TD/3b1MOyrUXx71J9HtaYdmiWQe5nBqhZbNJr9qtRTWubolBJEDGM+hJK92wfKiltUm/mxzk0Pmoj7fSuGdI2wWUmLSQ/clbv5gCA9o1qeyEeF0gtNomrYandFdaSbJxYr1aGj5IEB5poVKPWLada9jOzkJJHxFBLpCEboxJibAizUCvThiw2nFAnKw2lZVXo15o29dJCnk0FiCp6QWUn2a9NPcy4axiaSqw4yY5s0AgrV8O8loYv0sIhLHxgJKKiiN0l5X6LwxkpXjls0KsgHz+v3OW3GNwRpSAb4ijK5ADVxyR/K7+mbsg07RrW8lsEbiHFhhO+veV4fPHHNlxxXGu/ReEWubuMAOWaRSgkyM4JCQLaNEitxi8dKNJ0sjOlqpLToHa1CxopNnLIFU0fNSvEtUPaIhoVMaxTI/yydo8PUvEJqTVEDDXjnXTxLTErGmGW03o0w47iMvRpVddvUbiDFBtOaFW/Fm4b2dFvMbhGubGXshMMC4LM3SoVO0k9Mz9RgzIRBVGDtJ2RW1E1aqWQkRbCLSM6AAApNhLIYEPEUIsZlo/jFGNjl1BIwA0ntPNbDC4hJ2EiMEi7vJAgJHSCSotNqveRehsFpnq8TarXDSWCxt80Sa2G9DvzkDJMxFGpCmGdvpe6ZYIFpNgQgSGkWOlRTk4TguVTcPYqTbWqF2OTgkUjg1LSytFqK1RO1VApmIeqDBFDrf9Q25IgRqqPSwQbSLEhAoNyYp7giqaozanYR0qHEb0Ym1QnGvVbAn6RtjOapFZjZIXIkqR+TnVoywJCD6lik5UexksX9Y5/TnVPAoINpNgQgcHQYhMKyVaeU9FiI51/JabSlJaNVxLxSYRm7DLkrmiSGBuapAIwtthc2L8AvQrycddJFCdJ7otEDLWqoBx7Tu3RTPM7grADKTZEIKlOGSnvBUd2aaQZK5AqSFeWE13zJH+nZOnUEKHZlwwtN0UqJnPkZKThy/GDcfOJHfwWxXfIfZGIYZQVLcaoro0BVGcaJAinUFY0IjDoWWw+vWEQOjSug1/X7lU9P1XIzU6P/01Z0bQhxUYbab2RTkxSsDnVQNXFNG1TLMU+oY1qjI3KcvorFx+LLfsPo23D1NlMm3APUmyIwCDfo0ZukWlRNzvhnFSciDXOzcKz5/VETkZagiteqpeNFFJs5EiLQxqbJZ2Y1M3J8FIkriCXPPOc3K0JHhjbBb0K8v0WhfCZWhmJU0y1Bce0cIiUGoIZpNgQgSGkiJ9Ry4ef4vN1AMBZx7YwPCfVy4ncZeRIFT213cIBYNKV/bwShzuouphHEARcQy5FBIA7T+qIlTtKcGH/lvFjqRj7SngLKTZEYJBbbARZDIkQP0d7V+NUJ9XjaqR0aZrrtwhcIVX0ZBYbicJzTPM8T2XiCVJsCMI6jXKz8M0tx8uO6e1jQxAsoOQBRGAIKVzRMtMl1Vels6T+U5tUXzSrVysDC+4bgVb1c/wWhQukFhtZjI3knFReaSVXNIJgQyrGvhLeQooNESDkrmcZko1rYtYIiiPRRl4eVDiNcrNQO5OM1oDcMiO1hJLLHkEQLEnlBRLCG0ixIQJDSKG0ZKSFZJ8B+XSdVoYIwhzSfX20sqIRBEE4hTzECbchxYYIDMpkAVLFRnJSzZ9eCBUgZHv8UOEQEqSuaNJ2JpJmA4AUPIJgBS04Em5Dig0RGOQxNgLSJa5oVZHqmUckEo0fo31c5NB4Qmih5XJG8/lqqBwIgg1q+9gQBEuoihGBQZDF2ACZEotNzJVmW9GR+LFU3nfDCNJxqqH9bKqRrAfIoBibaupQLBZBMIFibAi3od6aCAzyxAACstLDuLB/AUrLqtAsLwsAUFpWFT+H0j0rke8DRABVpNgA0FbwqHiq+fuYzli/9xAuGdDS+GSCIDQJ09hDuAwpNkQgieksE8/qITteobX0TMigoaWaKM3cAWgrNmSwqaZxbha+Gj/YbzEIIvDQeiPhNuSKRgQSrQDEWKwNkYi0yNLVEi+kILR4WE3zutmqxyl5AEEQLKHkAYTbkMWGCCSaik2ULDZmSKdlMwA0yMboVZCPf559DFrWqyU7TmoNQRAsyc1O91sEIskhxYYIJIKGwaGiiqZiWkhX39PCZLEBSLGRcn6/xPgRctUjCIIlt47ogGVbi3Bu3wK/RSGSFFJsiECiNSGtpBgbTaSB8ulhmtAD5IpmBKk1BEGwpF6tDHx+E8WrEe5By7ZEYJC6+2t5UpErmjbS+KN0stgAoOxwRrRpUO2aRntCEQRBEEHAtdnNf//7X7Ru3RpZWVkYMGAAFixY4NatiBREy2JTm/ab0ERqzUqjiSoAytBjxDl9WuDeMZ3x0bUD/RaFIAiCIAxxRbGZPHky7rjjDjz88MNYvHgxevbsiZNPPhm7d+9243ZEiiCacIx56LRuOLZlPl68sLcHEgULqSsarcBXQwYbfbLSw7jhhHbo36ae36IQBEEQhCGuKDbPPvssrr32Wlx55ZXo2rUrXn31VeTk5ODtt99243YEEad5fjY+v2kwTuvZzG9RuENqsSEXrGpuHt4BAHA61ReCIAiCCDzM/XYqKiqwaNEiTJgwIX4sFAph5MiRmDt3bsL55eXlKC8vj38uKSlhLRKRJORm1aSJJIuDdRrnZvktAneM7t4Ev917IppQ2RAEQRBE4GGu2OzduxeRSASNGzeWHW/cuDFWrVqVcP7EiRPx6KOPshaDSELq1srAq5cci4y0EAW/26Bdw9p4/vxeaFQn029RuKJZvvrmlARBEARBBAvfZ4cTJkxAcXFx/F9hYaHfIhEcM7p7U5zYubHxiYQqZ/RujuPaN/BbDIIgCIIgCOYwt9g0aNAA4XAYu3btkh3ftWsXmjRpknB+ZmYmMjNpBZkgCIIgCIIgCPswt9hkZGSgT58+mDZtWvxYNBrFtGnTMGjQINa3IwiCIAiCIAiCYG+xAYA77rgDl19+Ofr27Yv+/fvj+eefx6FDh3DllVe6cTuCIAiCIAiCIFIcVxSb888/H3v27MFDDz2EnTt3olevXvjxxx8TEgoQBEEQBEEQBEGwQBBF0XjXQw8pKSlBXl4eiouLkZub67c4BEEQBEEQBEH4hBXdwPesaARBEARBEARBEE4hxYYgCIIgCIIgiMBDig1BEARBEARBEIGHFBuCIAiCIAiCIAIPKTYEQRAEQRAEQQQeUmwIgiAIgiAIggg8pNgQBEEQBEEQBBF4SLEhCIIgCIIgCCLwkGJDEARBEARBEETgSfNbACWiKAKo3mWUIAiCIAiCIIjUJaYTxHQEPbhTbEpLSwEABQUFPktCEARBEARBEAQPlJaWIi8vT/ccQTSj/nhINBrF9u3bUadOHQiC4Lc4KCkpQUFBAQoLC5Gbm+u3OCkHlb//0DvwFyp//6F34C9U/v5D78BfUr38RVFEaWkpmjVrhlBIP4qGO4tNKBRCixYt/BYjgdzc3JSsTLxA5e8/9A78hcrff+gd+AuVv//QO/CXVC5/I0tNDEoeQBAEQRAEQRBE4CHFhiAIgiAIgiCIwEOKjQGZmZl4+OGHkZmZ6bcoKQmVv//QO/AXKn//oXfgL1T+/kPvwF+o/M3DXfIAgiAIgiAIgiAIq5DFhiAIgiAIgiCIwEOKDUEQBEEQBEEQgYcUG4IgCIIgCIIgAg8pNgRBEARBEARBBB5SbAiCIAiCIAiCCDyk2BAEQRAEQRAEEXhIsSGSGspmThCE31A/RBAE4Q2k2BBJjSAIfotAEESKQ/2Qf6xcudJvEQgJpOT7QzQa9VsEzyDFxiUikUhKVSTemDx5Mp588kns27cvfow6VG/ZvXs3vv32W3zyySc4cuSI3+KkHNFoFEVFRVizZo3foqQsn3zyCb7++uuE49QXecM///lP9O3bF7/99pvfoqQ0R44cwfbt2wGQku8169evBwCEQiGIopgSfQ8pNowpLy/Hhg0bEA6HEQqFEIlE/BYp5SgpKcH48ePRsGFDZGZmAgDKysogCAIpmx5x8OBBnHPOObjjjjtw8cUX4/jjj0dJSYnfYqUMhw8fxk033YSBAwdi+PDh6N69O1555RUUFhamxMDGA8XFxbjggguQm5sbPxYre5rcuU9JSQkef/xxtGjRAo899hh27doFILVWrv2mrKwMTz75JAYOHIgLLrgAd9xxB4qLi/0WK2UoLS1Fhw4dMHDgQMyePRuCIMT7nmQeB0ixYcx1112H9u3b47bbbsPhw4cRDocBgBQcD7nrrrvQs2dPXHfddThy5AheeOEFDBo0CCeccAJefvllbN68GUByN2y/ueGGG1CnTh189tlnWL16NQBg27ZtmDx5MqZMmYK1a9cCoHfgFldffTX++usvXH/99XjrrbcwfPhw3HbbbTjttNPw008/xc+j8nePa6+9FiNGjMCwYcNw8OBBzJw5E/fffz/uvPNO/PXXX/Hz6B24w1VXXYUhQ4bgww8/xPz583HzzTejvLwcoRBNe7zixhtvxNSpU3HSSSdhxIgRmDt3Lr799lu/xUoJRFHEunXr0LBhQ4TDYQwbNgznnnsuCgsLAQBvvfVW8rppigQz1q1bJzZr1ky88sorxe7du4tNmzYVX3rpJdk5VVVVYmVlpVhWVuaTlMnNoUOHxDPOOEN8//33RVEUxbPPPls87rjjxLvvvlu8/PLLxfr164vnnHOOz1ImNxs2bBCbNm0qLlu2LH5s+PDhYvfu3cWWLVuK2dnZ4pgxY8S9e/f6KGXysnnzZrFhw4biggULZMd37twpnnnmmWI4HBafe+45f4RLEZYsWSKGQiFx3bp1oiiK4vjx48Xu3buLPXv2FLt37y6mpaWJjz76qM9SJi8LFy4UMzIy4uU/depUsW3btuLDDz8sRqNRURTF+H8Jd1i+fLmYm5srLl26NH7slltuEbt16yaWlJRQ+XvEWWedJT711FPiTz/9JHbv3l3MysoSL7zwQrFOnTri/v37RVEUxUgk4rOUbCHFhiGvvvqqOG7cOHHx4sXi0qVLxbvuukts0qSJ2KtXL3HKlCnx8x555BHx1Vdf9VHS5ObCCy8U77zzTnHNmjViixYtxE2bNsW/W7FihZifny8+++yzPkqY3KxatUrs1auXOH/+fFEUqxWdcDgsTpo0SayqqhJ37NghNmjQQLzlllt8ljQ52bVrl9ijRw/xyy+/FEVRFCsrK8WKior4988995zYunVrcfHixX6JmPSMHj1abNCggfjHH3+IH3zwgdisWTNx2rRpYlFRkVhRUSG+8sorYn5+vvjzzz/7LWpS0qZNG/HGG28URbG6/h85ckS87777xFq1aomffvqpz9KlBnfffbd43nnniVVVVWJVVZUoitWLK40aNRLnzZsXP++jjz4S//jjD5+kTF5iiuOUKVPEs846SxRFUdy/f7/4yiuviJmZmUndFsgmy5Bhw4Zh1KhR6N27N3r06IH7778f7777Ljp16oRzzjkHZ511Fn744Qc8+uijKCgo8FvcpGXEiBH45ptvsHjxYgwcODDuelBVVYVu3brh7LPPxpYtW3yWMnlp2LAhSktLceONN+KBBx7A6aefjjFjxuDyyy+HKIpo0qQJrrjiCuzfvx8VFRV+i5t05OfnIy8vD08++SR27NiBtLQ0pKenx8v6nHPOQVpaGgVUu8jNN9+M9u3b44ILLsDtt9+OW265BSeeeCLy8vKQnp6OcePGoX79+tQPuUBJSQmuueYaPPPMMwCAtLQ0ZGVl4cknn8S1116LO+64A1OnTgVALuJuUVVVhTZt2qBVq1YAgHA4DFEU0bhxY/Tp0wezZs0CAOzatQsXX3wx9uzZ46e4SYkgCBBFEcceeyzWrFmDt956C3Xr1kVpaSny8vJwwQUX4OKLL0aPHj1w6NAhv8VlCik2DOnUqRPGjx8PoNq/MT8/H6NGjcKzzz6Ll19+Gfv27cPYsWNxxhln4JRTTvFZ2uTl6quvRo8ePXD11VfLYjzS0tIAVAdWU5Yu96hXrx5++ukntGvXDuvWrcNpp52GRo0aAah5B9u2bUN6ejoyMjL8FDUpycjIwKRJk5CWloZzzjkH7777LkRRjJd1ixYt0LFjR+zYscNnSZOXsWPHYu7cufGYy27dusm+b9KkCTp37kyB1C6Qm5uLO++8E9nZ2fFEAbH/jh8/Hs2bN8fjjz+OkpKSeAwswZa0tDSceOKJuOKKK+JKTSxofcCAAZg5cyYA4G9/+1t8QZhgSzQahSAIqF+/Pu69917MnDkT5eXl+Mc//oEnn3wSb7zxBn788UdcdNFFqFWrlt/iMkUQRYpcdANRFBGNRmUd50svvYS//e1vKCwsRLNmzXyULnmJRqMIhUIoLCzEq6++ivfffx9FRUW4/fbb0bVrV8yfPx9vvvkmli9fjpYtW/otbtIgHbiUx7///ntcddVVePfddzFgwAB89tlnuPPOO7Fs2TJ6By4QawPTp0/Hm2++ibVr16J58+a46KKL0KNHD3z11Vd45JFHsHr1aip/Dzh06BDKy8tRr169+LE5c+bg5JNPxqpVq9CiRQsfpUsNYm0CALZs2YJhw4ahW7du+Pzzz5Genu6zdMmDchwQj6YXliZsmDFjBq6++mo888wzOOecc7B+/Xq0bt3aB2mTD61x+ODBg7jsssuwYsUKNGjQADNnzkzqRcU0vwUIOkeOHMG8efMwefJktGnTBm3btkX//v3RqlUrhMPhuNZcWVmJ9957DzfccAMpNS4RiUTiHWhBQQHuuusuHHfccfjhhx/w+uuvIxQKoV+/fnjllVdoQseYSCSCTZs2YerUqTjmmGMwePBgANXm8LFjx+Lss8/GFVdcgb1796JXr1548MEH6R24QCQSiQ9sJ554Ijp27IjPP/8cc+fOxQ033ICysjL0798fTz/9NJU/YyoqKrBixQp89tln6Nu3L84880wAQK1atWQrotOnT8edd96J66+/npQahuiNxbGtF0RRRMuWLXHnnXciFAqRUsOY2Djw888/o0ePHhg8eHDCRPuEE05AnTp1cPbZZ+P2228npYYhFRUVWLlyJb755hu0adMGXbp0QUFBARo3bhxP+fzmm28mtVIDkMXGMVdccQXmzZuHevXqYefOnahbty6aNGmCU089FZdddll8QNu9ezdefPFFPP744z5LnHy8//77GDVqFBo3bgwAqKysRFpaWrxDjUQiCIfDWL9+Pdq1a+enqEnLAw88gK+//holJSXYunUrLr/8crz00kvIyMhAOBzG1q1bsW7dOhQWFmLgwIHo0KGD3yInFco2UFFRgbS0tLiiv3v3bgiCgC1btqB79+7x/Z0Idtx5552YMmUKKisrsX79eowaNQqTJ09GnTp14uccOHAAH3/8MWbOnInJkyf7KG3yYXYsJtxDaxzIzs6WnXf11Vdj1qxZWLdunU+SJic33HADpk+fjvT0dGzduhVNmzbF0KFDce6552LUqFHYsGED2rRpk/z7aHmdrSCZWLBggZiTkyMuXLgwfuyjjz4Szz//fHHAgAHihAkTKKWty8yfP18UBEGsW7eu+Nhjj8m+Ky8vV/0NpZlky+zZs8X8/Hzxhx9+ENeuXSt+++23YseOHcXp06f7LVpKoNcGKK28N8yZM0fMz88Xf/nlF7GoqEj8888/xfbt24sffvhhwrmlpaViUVGRD1ImL1bG4liGLoItVsaBP/74Q1y0aJEPUiYv8+bNE7Ozs8XffvtNPHz4sBiNRsXnn39eHDhwoNinTx/x9ddf91tEzyDFxgGvvPKKePzxxyd0lPv37xefeOIJsX79+uLjjz8uiiJ1pm5x0UUXiWPGjBEfeughsVWrVmL79u3FyZMnx7+PRCJiJBIRt2/fLlZWVvooafIyatQo8e67745/Li8vF88880zxggsuEEVRjJe7lqJJOMNsG9i2bRu1AZcYO3aseMcdd8iO3XDDDeKwYcNEUazp//fs2UNjgQtYGYsJdzA7DlAf5A7/+Mc/xBEjRiQc37lzp3jdddeJmZmZceUm2Rd3KSuaA3r37o2VK1diypQpAKpdnqLRKOrWrYv7778f//jHP/DCCy9g586dlH3FBQoLCxGJRHD66afj7rvvxgcffIBhw4bhhhtuwPDhw7FkyRKEQiEsW7YMo0ePxr59+/wWOelYv349Dh8+jL59+8aPZWRk4MILL8SSJUtw5MgRpKWlYdOmTRg+fDgOHDjgo7TJh5U2MGbMGGoDLlBYWIjS0lJ07dpVdvzCCy/Eli1bsHfvXoTDYWzatAkjR46k1LYuYGUsJthjZRwYMmQIjQMucOyxx2LlypVYsWIFAKC8vBxVVVVo3LgxXnvtNdx77714+umncfDgwaR3RSPFxgEdO3bEgAEDcO+99+L3339HOBxGKBSKp5Y88cQTUa9ePaxZs8ZnSZOTgoIC3HrrrTj++ONRu3ZtDB48GP/4xz/w9ttvIzMzE0OHDsV1112HW2+9FfXr14/HHxDsaNGiBc4999yEsh0zZgwOHz6MlStXAgBuv/12hMNh1K1b1w8xkxZqA/5TUFCAq666KiEZw+DBgxEOh7FkyRIAwG233RaP+yDYQmOxv1gZB9LT02kccIEuXbqgfv36uPHGG7Fr1y5kZmYiLS0tvn/ZySefDFEU49tfJDV+m4yCzsGDB8VTTz1VTEtLE2+77TZxx44dYlVVlRiJRMQpU6aItWrVEvft2+e3mClFNBoVN27cKL7++utikyZNxMzMTIp1chFlHEdsl/sTTzxR/Oc//ykuXrxYDIfD4q5du/wQLyWhNuAvMbfLs88+W7zjjjvEhQsXUhtwGRqL/YXGAf8pLCwUBwwYINauXVucOHGi7Ltp06aJeXl54oEDB/wRzkMoK5oDYtm2Kioq8H//93946qmnsGfPHowZMwZr167FkSNHcMEFF+Dhhx/2W9SUQBRFRCKR+CaQO3fuxDHHHINbb70VDz74oM/SpQ6xPSMeeOABLFy4EJs3b8aYMWPw7LPP+i1a0kNtgC9efvllvPfee9izZw/GjRuHZ555xm+RkhIai/mDxgFviZX3+vXrMWnSJLz22msAgPPOOw8bNmzA+vXrcf755+Oxxx7zWVL3IcXGAaJkM6Tdu3cjIyMD3377LaZNm4YWLVpgwIABOPXUU32WMrkRNTakAoB33nkHDz30EAoLCz2WKrXQegdz5szBkCFD0Lx5c3oHLkJtwH+k72DPnj1o2LAhAODPP//EMcccQ23AZWgs9h8aB/xFugnt4cOHsWHDBvzwww/48ccf0b59exx//PG49NJLfZbSG0ixcUBslei+++5DRUUF/v3vfyd9UBZvxN7BhAkT0KhRI9x+++3x7+bMmYOMjAz069fPRwmTH+k7yMrKiq+KVlZW4uqrr8Z5551HkwoXoTbgP9KxoLy8HE8//XR8LLjnnnswZMgQnHbaaT5LmbzQWOw/NA74i7T869Spg/vuu+//27v3uBzv/w/gr7uzolTKFJVOhErllFPKaXMmqTm0nKbMmDHD2DCWZvxQM2VDVA41K4dlzKnNsVIOFRIpSREdqO667/vz+8Oj6ytjY+vuo6v387/7um4e797v+7o+9/u+ruvz4R0SN9TY/Es13XFhYSGsrKwQGRmJESNGCB+uv/sVldSN52tgaWmJiIgIjBo1indYjULN57u6uhrq6upCDSIjIzFy5Ehhf0lJCfT09HiHKzo1+a0539AxwE9NLWrGgqioKAwfPpzGgHpCYzF/MpkMampqNA5w8mL+a8aBmmPg+as5jQE1Nm/gZSfIjz76CNnZ2Th06BCnqBqPgoKCl87qNGnSJJSVlSE2Nrb+gyIAgClTpuDhw4fYv38/71AaJV9fX5SWltIxUA9SU1Mhl8vRsWNHaGlpCeNCQEAAcnJyaCyoBzQWvx2kUik0NTWF19OmTUNhYSGNA/WgqqoKJSUlwm2vAOW/hhrvABqKqqoqaGhoAKj9a+no0aPRqVMnztGJX2VlJVxcXPDbb7+hY8eOQg2ePn2K9u3bw8fHh3eIoqdQKHDhwgXs3LkTZmZmaNasGcaNG4cWLVqgf//+6NWrF+8QRY0xhoyMDPzyyy9wcHCAm5sbdHV1UVZWBmtra/j6+vIOUfSkUik8PT2xePFi2NraQktLS/iCPWHCBJibm3OOUPxoLOZLLpcjIiICe/fuhbW1NcaPH4/u3bujsrISvXv3Rr9+/XiHKGoymQyrVq3C4cOHYWRkhMmTJ2P06NGorKxEnz590LdvX94hckdXbF5DSEgIiouLMXbsWLRv3553OI2Sr68vsrOzkZCQAODZl+zq6mqoqKhAXV2dc3SNw5w5c3Du3DnhV+rMzEzo6elh9uzZmDlzJu/wRG/58uXYt28fiouLcffuXQwfPhwxMTHCDGhE+aZPn47r168jISEBVVVVSElJwfnz56Guro5p06YJ5yK6/Uk5aCzmb/bs2Th//jxMTEyQnZ0NU1NTHDhwgD7v9cTf3x9XrlxBly5dkJubi9LSUhw9ehQKhaLWQvCN+hykxKmkReHmzZtMIpEwQ0NDNnz4cBYREcHy8vJ4h9WoXL16lamrq7PU1FTGGGORkZFsxIgRTE9Pj/Xv359999137N69e5yjFLcbN24wHR0d9ueffwrbJkyYwFq0aME6duzIpk2bRusTKFFGRgbT1dVlBw8eZFVVVSw7O5tZWFiw9evX8w6t0Xjw4AHr1KkTO3r0KGOMsU8++YTZ29szW1tbZmFhwXR1dVlkZCTnKMWLxmL+MjIymI6ODjt79ixjjLH09HQ2ZMgQFhgYyLy8vNjo0aOF44PUvRs3brCmTZuy5ORkxhhj+fn5bODAgWzevHmsb9++zMnJicXExHCOkr/G8zTRvxQaGopRo0Zhx44dqK6uxuLFi/Hll1/iyJEjKCkpEd538OBB3Llzh2Ok4hUQEIABAwbA0dER586dw1dffYVmzZrh+++/h7GxMTZs2IAdO3bwDlPU4uLihNvNalYynjlzJuzt7TF+/HjEx8fjwIEDnKMUr6+//hqjR4/G0KFDIZFIYG5uDi8vL+zatQsymQwymQwAcO/ePVRUVHCOVnwYY9DR0YGhoSG0tbWRlpaGyMhIhIWF4eTJkzh27BhmzJiBwMBAZGRk8A5XlGgs5m/ZsmXw9PREjx49AABqamqIj49HUlISDA0NwRjDqlWrcPfuXc6RitPRo0fh6uoKZ2dnAM9u0f/9999RUlKCwYMHo2fPnvjyyy+RlJTEOVLOeHdWbzOpVMpCQ0PZ0qVLhW2RkZHMwcGB2dnZsaVLl7KkpCSWmprKmjdvzk6cOMEvWJG6c+cO69q1K7Ozs2PLly9njo6O7Ouvv2bl5eXCe4KCgpiWlhZdtVGimJgYZmNjw7Kzs4VtM2fOZCNGjGCMMbZ8+XLWrl07VlFRwStE0SooKGBDhgxhoaGhtbanpKQwa2tr4UpZUVER69OnD7t58yaPMEVPJpMxNzc3NnnyZLZ+/Xo2adIkVl1dLey/efMms7CwYDt37uQYpThVVlbSWMxZUVER++yzz9jBgweFbd27d2ejRo0SzvsnT55kWlpadNVGSY4dO8YMDQ1ZeHg4u3fvHuvduzcbPXq0sD89PZ2ZmZmxsLAwjlHyRzdn/w0NDQ2MHTtW+IUaAMaPHw8fHx+sXLkSW7duxalTp4TVvemhubpnZmaG9evX4+jRo4iJiYFMJsOYMWPQpEkTYSpDd3d3REREoLCwEK1ateIdsihZWVmhvLwcc+fOhaenJ65fv44ffvgBZ8+eBQC4uLggOjoaDx48QOvWrRvvvb1KYGxsjGHDhkFfX7/WdkdHR2hoaOCPP/6Ap6cn5s+fj5KSElhZWXGKVNxUVVWxevVqfPbZZygqKkJ+fn6t55usrKzg4uKCvLw8jlGKk6amJry8vCCVSoVtNBbXLwMDA3z44Ydo2rQpgGeTOAwYMABTpkyBlpYWAMDZ2Rlubm548OABz1BFy87ODh4eHvj2228RHBwsjMnP73dyckJBQQHHKN8CvDurt5FCoaj1Wi6XC9uf/4UuLy+PeXh4MIlEwrKysuo1xsamtLSUxcXFsbCwMPbgwYNa+86cOcMsLCzoio2Spaamsq5du7IWLVqwAQMGsO+//17Yd+rUKWZhYUHP2SjR8+elqqoqxhhjfn5+bNKkSSw9PZ2pqKiwa9eu8QpPlF4cC54+fcpWrlzJTExMmEQiYd7e3uzChQusqqqKRUdHMy0tLZaZmckp2sZDoVAwmUwmvKaxWHlePAZefP288+fPMyMjo1pX9kndun//PouNjWV37txhH374IZs8ebJQk/PnzzNtbe1Gf9WeZkV7hfLycty9exfa2tpo3bp1rX0KhQIAoKKigs6dO6N3794ICQnhEaaoSaVSPH36FDKZDMbGxgDwl4W+KioqMHjwYHTo0AGbN2/mFapo1dSgqqoK77zzDgAgMzMT5ubmwpSr+fn58PT0hL29PUJDQ3mGKzo1+ZfL5cJ6Bc+ff6KiohASEoKKigo4ODggPDycZ7iiVF5ejry8PGhpaaFNmzYAgKSkJAQGBiIlJQVVVVUoLS2FjY0Nhg4dihUrVnCOWHxetcAgYwyMMRqLlezvvg/VePz4MUaOHEljsRJIpVKUl5ejqqqq1lp+cXFxmDNnDnx8fHDv3j1kZGSgV69eWL9+Pb9g3wLU2LzExo0bERUVhfv376OyshI9evTA4sWL0a1bNwAQboE6f/48RowYgbt379KUw3Vs27ZtiIqKEk6mnTt3xjfffFProE5LS0NgYCASExORlpZG097Wsedr0KRJEzg5OWHVqlVCg6NQKFBUVIRdu3YhIiICZ86coRrUoRePARcXF6xcuVJo8gEgOzsblpaWMDAwQEFBQa3pPsl/9+JY0KVLFyxZskR4ePrYsWMoKytDSUkJ+vbti7Zt23KOWFwyMjJgbW0tjK+vanBoLFaef/o+BAA3btxAcHAwzp07h7Nnz9I4UIdeHAdqxuGa70JBQUHYtWsXjIyMMHjwYHz66acvPUYaE2psXpCUlAQ3NzesWbMGhoaGkMvlCAsLw+nTpzFt2jR8//33tT40jx8//su97+S/SUpKQr9+/bBgwQI0a9YMFRUV2LNnDzIyMvDFF1/gq6++AgBUV1dj+/btsLa2hru7O+eoxeV1a6BQKHD9+nVoamrC0tKSc9Ti8br5l8vlmDt3Ljp06AB/f3/OUYvL340FH3zwAX788UfeIYra5cuXMW7cOIwePRqenp7o0qWLsI+9ZI0OGovr3ut+H5LJZIiPj4epqakwYxf57/5uHFi0aBGWL18O4Nl3IZlMhiZNmnCO+O1Ajc0LJk6cCIVCgaioKGGbXC7Hzz//jHnz5kFfXx8xMTGwtbVt3AsgKdGYMWOgp6eHbdu2AXj2kGJhYSF27dqFtWvXwsXFBVu2bIGJiQnnSMXrdWrw448/0mQNSkLHAH//NBY0b94cMTExaNeuHY0FSjBs2DAkJyfD3Nwc+vr6GDx4MEaNGgULCwveoTUar/N9KDo6Gu3ateMYpXj90zjg5OSEsLAwtGnT5pVXMxsjysILrK2tUVpaKryuWc113LhxOHToEJo1ayasmUIDmXK8eA+vhoYGWrdujY8//hihoaHIy8vD/v37OUXXOLxODeLi4jhFJ350DPD3T2OBrq4udu7cCYDGgrp29epVFBQU4IcffsC6detgYGCAHTt2YNGiRYiMjBTWrUlPT0dAQECt2dJI3Xmd70M1xwCpe/80DuTn5+PQoUMAQE3NcygTL3BwcMCvv/6KDRs2oLq6utaHxcHBAR4eHjh8+DAePXrEMUpx69SpE/bs2YPo6Ghh4UEA0NLSwsiRI9G1a1fs3r2bFiJUIqoBX5R//mgs4Kd169aYMmUKTE1N0bNnT0RGRuLzzz9HUVERNmzYgC+++AJHjhzBvHnzkJGRAU1NTd4hixIdA3zROPDvUGPzgjFjxmDp0qWIiIjA4sWLkZycXGu/q6srSkpK6AFFJfLz84OPjw+Cg4OxZcsWZGdn19o/YMAAPHr0SJgditQ9qgFflH/+aCzgp3nz5pg+fTq6du0qfMa9vb0RExODsWPHIjU1FbNmzcLRo0cRGRnJOVrxomOALxoH/h16xuYlCgsLsXHjRsTHx8PAwACOjo7w9PRETk4OVqxYgcGDB2PdunW8wxS17OxsLF26FKdOnYKrqyv69OkDd3d3PHz4ELNmzcKgQYOwdu1a3mGKGtWAL8o/fzQWvB0YY5DL5cJsWykpKejWrRuWLFkiTKRBlIOOAb5oHHhz1Ni84PkHsC5fvozQ0FCkpaXh7NmzaNu2Ldzc3GitDiV7vgbHjx/HmjVr8PDhQ1y7dg1GRkbo1q0bdu/ezTlKcaMa8EX554/GgrdPzSQNq1evxooVK1BeXs47JFGjY4AvGgf+HWpsXlCzRk1ERARsbGzQvXt35OTkQEdHB5WVlTA1NeUdoujV1ODAgQPo0KEDrKyskJSUBF1dXcjlclhZWQmLQxLloBrwRfnnj8YCvmryf/DgQRgaGsLV1VXYFxQUhBYtWmDq1KkcIxQ/Ogb4onHg36HG5jk13XFJSQksLCwQERGBoUOH8g6rUXm+BtbW1ti1axcGDBjAO6xGoebXUKoBX5R/fl52DNBYUP+ez7+NjQ327t2Lfv368Q6rUan5Uk3HAB/P55/GgTdDjc1zaga18ePHo6SkRJhGjyjHi2s/yOVyAICqqiq8vLxQUVGBgwcP8gqvUaAa8FdVVVXrV7eamlD+68/Tp0+ho6MjvK6pwfvvv4/S0lIaC5Tsxo0bMDExQdOmTcEYA2MMKioqLz0GaM2guvd3OaVjQPlezP/zr8eOHYvKykoaB95Ao58V7fkp9CQSCUpKSpCXl4fAwECOUTUOEokEMpkMly5dQmlpKVRVVaGqqoqHDx8iKSlJWFWXKI9EIkFhYSH27NmDgoICoQZFRUVITEzEsmXLeIcoaowxrFy5EkVFRcK2mpokJibSMVAPGGOYMGECDhw4gJrf+SQSCYqLi5Gfn4+VK1dyjlDcGGPo2rUrZsyYgerqakgkEqioqODRo0cvHQeoqal7EokEFRUVOHbsGLKyslBWVgaZTIbq6mrcv38fq1at4h2iqEkkEpSXlyMhIQF3796FXC6HXC5HeXk5Ll68SOPwG2q0V2wuXLiAbt26Ca9rLvsB9ItQfYmKikJ4eDgyMzMhkUiwZMkSTJ48GQBoFd16snHjRkRFReHWrVt4/PgxwsPDMX78eABUg/rg5+eH69ev4+zZs3/ZR/mvH7NmzcLFixexY8cOWFtbC1+ua2bgIso1e/ZshIeHw9TUFLNnz4a/v7+wj46B+hEcHIzo6Gjcu3cPt27dgqurK7y8vODt7Y1WrVrxDk/0goODsXPnTuTm5qKgoADu7u6YNGkSvLy8oK2tTd9H31CjbGzu3LmDtm3bwsbGBkFBQRg1ahQACF0yPYylfLdv30bXrl0xd+5cmJmZ4dq1azh27BiioqJgaWlZ6700uCnHrVu34OzsjLVr16Jbt27YuHEjWrZsCX9/f6SkpMDe3h5mZmaQSCR0YlWCrKws2NnZ4Y8//kD37t2RlpaGEydOIDExET169MCkSZOEW3Mo/8qRl5eHTp06Yf/+/ejTpw927tyJ6OhoJCQkoHfv3hg6dCjGjRsHQ0ND3qGKUlZWFmxsbHDy5EnExcUhJCQE0dHRGDFiBJ3368nt27dhb2+P4OBguLm5oaSkBMOGDYNUKkXv3r3x5ZdfwtnZmc5DSnL79m3Y2dlh3bp1cHR0hJqaGlauXInz58+jd+/e+Oqrr+Do6Ej5fxOsEZoxYwZzcXFhkyZNYs2bN2cDBw5kV69eFfbLZDLGGGNlZWVMoVDwClPURo4cyT744APh9e3bt1m7du3YsmXLGGOMVVdXc4qs8Rg5ciTz8/MTXsfGxjJLS0vWsWNHZmBgwPT19dnPP//MMUJx69+/Pxs3bhxjjLGkpCTWq1cvZmZmxoYOHcp0dXWZra1trfMSqXsJCQnMw8ODMcbY2bNnWevWrdn8+fNZeHg48/T0ZDY2NmzTpk2coxQvNze3WuOAn58fc3Z2ZsnJyYyx/43FRHn8/f2Zt7c3Y+x/4+7ChQvZ6NGjma2tLWvTpg27ffs2xwjFbdGiRWzw4MF/2X7kyBHWvn17ZmpqylJTUzlE1nA1up9D7t27h8LCQkyaNAmrV6/GTz/9BIVCAVdXV3z00UcoLy+HqqoqsrKy4Ofnh/v37/MOWXRu3bqFu3fvwsfHR9hmYWGBKVOm4NixYwAANTU1lJWV4fPPP6dVdZUgLS0N6enpmDFjhrBt06ZNaN26NUJCQlBUVARvb28EBATUev6D1I2kpCQcP34ctra2yMnJgb+/P7p06YKEhAQcPHgQV65cgbGxMRYtWkSffyVq2bIlLl++jNTUVISFhcHHxwdr1qyBr68vYmJiMHXqVHz++efIzc3lHaro/Prrr0hISKi1uODs2bNRXl6OTz/9FFKpVLg9nCiPgYGBMHFGze2XpaWlsLOzw/Xr12FmZobt27dzjFDcrK2tkZeXh+LiYgBARUUFAGDgwIHIyMiAo6MjjQNvqNE1Ni1btsT7778PFxcXmJiYYOTIkQgLC8OqVatw6tQpWFlZISQkBMuXL0dqairdX6oETZo0gaur618uq44dOxa5ubnIyckBACxYsABRUVF0O4IS2NraYsmSJWjWrBmAZ7MSPXjwAJs3bxamVR09ejRMTU2FEy6pO/b29pgzZw52794Nd3d3VFdXY8GCBTA3NwcAmJmZ4b333sO9e/dQVlYmPNRO6paFhQX69euHtWvXok2bNmjSpAkACPmeOHEi2rdvj7t37/IMU5Q0NTWxd+9eGBoaCl/anJycEBsbi7y8PPj7++PJkycAQJ9/JTIxMcGOHTvw888/QyqV4uTJkwgLC8OgQYMAPKvJiRMnak20ROqOnZ0dcnJy8N133wF49v2IMSY0OD4+Prh16xb9uPIm+F4w4ufFW8wqKyvZxYsX2cKFC5menh6TSCR0G4iSyGQylpmZWWtbdXU1q6ysZB06dGB79+5l2dnZTE1NjV26dIlTlI3L48eP2cWLFxlj/zs2Tp8+zdq3b8/y8vJ4hiZqaWlpzMfHh02bNo09evSIMcaYXC5njDF26NAh1qVLF/bgwQOeIYpeSkoKa9u2LZNIJMzBwYHdunVL2Jeens4MDAzYzZs3OUYoTjWf8xe3KRQKFhwczPT19ek2wHryySefMHNzc6alpcWsra3ZnDlzhH179uxhffr0YSUlJfwCFLmoqChmbGzM+vfvzxISEmrtS01NZcbGxiw3N5dTdA1Po5o8QKFQgDEGhUIBdXX1l77n/v376NGjB4YNG4aQkJB6jlD8amafe9WDoV5eXmjVqhUyMzPRrFkz7N27l0OU4lZzHDDG/nbmp4EDB8LMzAw//fRTPUYnfgqFAgqFAnK5HJqamgCe3SJrYmJS630DBw6EhYUFtmzZwiNMUVMoFJBIJKiqqoKmpiauXLmCTZs2ITw8HGZmZpgyZQpu376N5ORk2Nvb0zFQx2rGAZlM9spzUFBQEJYsWYL4+HhamFAJaq6SSSQSFBUV4dq1a8jOzoadnR1cXFwAPLtS5uHhAWtrazoPKVlcXBxCQ0Nx9epVODs7Y+bMmbh69Sr27NkDW1tb7Ny5k3eIDUajaWwOHjyI3bt3486dO/D19cX06dNr7a85we7btw8TJkxAUVERtLW1OUUrTjU1yM7Ohp+fH6ZNm/aX98TGxmLMmDHQ0dFBQUEB1aCOveo4YM/NuJKfn4+wsDCEhYUhMzOTalCH/uk8BDxrcn744Qds2bIFt27dovzXsVfVID8/H5cuXcKOHTuQmJiItm3bol+/fpg/fz7NlFmH/ukYqFmwNj8/H19++SVCQkKEHwBI3Xh+LP7ggw9eeh7Kzc1FaGgotm3bRuNAHbtx4wbOnDmDtLQ09OjRA+3bt0fHjh2Rm5uLgwcPYt++fbhw4QKsrKzQs2dPrF27lo6BN9AoGpu0tDR4eHhg1KhRKC4uRnR0NBITE+Hi4oLKykpoaWkBePblLi4uDgqFAmPGjOEctbi8bg3u3LmDgQMHYu7cuQgICOActbi8bg3Onz+PXbt2wdXVFd7e3pyjFo/Xzf+5c+ewbds29O3bFxMmTOActbi8rAYXLlxAly5daq1l9vTpU2hoaLzyyj75d173GKjxd1d0yL/zd8dAeXm50MBUVlZi586dMDU1xZAhQzhHLR7p6enw9PSEkZERZDIZzp07BxsbG3h6emLy5MmwsbEBADx+/BiMMRgYGHCOuOFpFI2Nh4cHHBwcsH79egDAjBkz0K5dOxw4cABGRkaws7PDnDlz6AOkRK9Tg48//hgtWrRAYmIiunbtyjdgEXqdGsydOxc6OjooLy+Hnp4e34BF5k3yX1ZWRucjJXhVDfbv3y/UYNasWTA2Nq7V6JC68SZjMa1joxyvU4PZs2fT2k1K4ubmhs6dO2PFihXQ09PDgQMHMH78eFhaWsLS0hIbNmyAmZkZAFos/t8S/VkjKysLhYWFGDt2rLAtJSUFkZGR6NSpEwwMDBAXF4dffvmFY5Ti9ro1iIuLAwBqapTgdWsQExMDdXV1amrq2Jvmn5qauvd3NbC3t4ehoSH279+PAwcOAAA1NXXsTcdiamrq3uvWIDY2ll+QIpaRkYGCggJMnToVurq6kEqlGDRoENzc3DBq1CgkJSVhypQpKC8vBwBqav4l0V/jNTU1RfPmzXH06FGYmZnh9OnTuHLlCpKTk9GhQwcAzx5Yj4yMxMSJE+k+RiV4kxqMHz9emHKV1J3XrUFUVBQmTpz4l1tCyH9D+eePxgK+KP/8UQ34MjExgZaWFrKysuDg4ABNTU3k5+fjyJEj2L9/PwICAtC1a1ekpKSgV69evMNtsET/k4iGhgZ69eqFNWvW4L333kNwcDDc3d1hZ2eH6upqAMDw4cMhlUqF16RuvUkN5HI552jF6U1qQOsV1D3KP380FvBF+eePasCXtrY2LC0tMX36dOzYsQP79u1D//794ePjAxUVFaipqcHAwAA3btzgHWqD1iiesQGAhIQElJeXQ11dHYsXL8bp06ehpqYGxhj69esHJycn4Z5TohxUA/6oBnxR/vmjGvBF+eePasDP48eP8cUXXyAyMhKGhobo3r07oqKiIJFIIJPJ0Lt3b/j5+cHf3593qA2WqBubmoc/a6aPBIDs7Gy4u7vD1tYWgwYNwpkzZ3DlyhVcu3aN7ulVAqoBf1QDvij//FEN+KL880c14Ksm/zUTAkilUuTl5Qm3p1VUVCAqKgrz5s3Do0ePKP//gSgzd+fOHaxbtw4ODg4YOnQolixZgqNHjwIALCwshIcTAwMDYWxsjK1bt9KHqI5RDfijGvBF+eePasAX5Z8/qgFfz+d/yJAhWLhwIX7//XdoamrC0tJSeJ7y0KFD2Lp1K9atW0f5/49EecXG1dUV+vr6cHd3R0ZGBvbs2QNdXV0MGTIEc+fORadOnQAAxcXFaN68Od9gRYpqwB/VgC/KP39UA74o//xRDfh6Mf+7d++Gnp4ehgwZgnnz5gmTNjx+/BgpKSnw8PDgHHHDJ7pZ0fbt24eCggIcPXoUTZs2BQAYGhrit99+w82bNxEUFISQkBDo6enRQawkVAP+qAZ8Uf75oxrwRfnnj2rA1z/lf/Xq1QgODoaenh709fWpqakjorvedenSJTg7O6NJkyZ48uQJAKBHjx4wNzdHQEAAYmNjERQUBAA0A5eSUA34oxrwRfnnj2rAF+WfP6oBX/+U/19++QXffvstANBsmHVIdI1N+/btkZycjJKSEqFDXrVqFWxtbeHj44NvvvkGv/32GwBagE1ZqAb8UQ34ovzzRzXgi/LPH9WAr9fJ/+HDhwEAamqiu4GKG9Flsm/fvmjWrBns7e0xZMgQZGVloaCgACtXrgTw7GE5iUSCO3fuwNzcnHO04kQ14I9qwBflnz+qAV+Uf/6oBnxR/vkQ3RUbU1NTHD9+HNOmTcPFixfRs2dPHDhwQFjNPjc3F0+fPqUPkRJRDfijGvBF+eePasAX5Z8/qgFflH9OmAg8evSI3b9//2/fo1AoWFJSEjMyMmLbtm2rn8AaEaoBf1QDvij//FEN+KL880c14Ivyz58opnv28PCAvr4+5s2bh86dO0NbW/sv78nNzUV4eDju3r2LzZs3c4hS3KgG/FEN+KL880c14Ivyzx/VgC/K/1uAd2f1X0VERLCmTZsyS0tLpqOjwxYsWMAyMjJYdXX1X95bXV3Nnj59yiFKcaMa8Ec14Ivyzx/VgC/KP39UA74o/2+HBv+MTUJCAqZPn46srCz83//9H0JDQ/Hee+9hy5YtuHfvHgBAoVBg4cKFSExMfGn3TP4bqgF/VAO+KP/8UQ34ovzzRzXgi/L/luDdWf0X1dXVbP/+/SwqKkrYJpfL2axZs5hEImG9e/dm8fHxbNOmTUxdXZ2VlJRwjFacqAb8UQ34ovzzRzXgi/LPH9WAL8r/26NBNzaMPXsIq6qqijHGmFQqFbZnZWWx/v37Mw0NDSaRSNjy5ct5hSh6VAP+qAZ8Uf75oxrwRfnnj2rAF+X/7SCKyQOep1AooFAohMWOvL29cfnyZWRkZHCOrPGgGvBHNeCL8s8f1YAvyj9/VAO+KP98iG6BThUVFaioqEChUCAzMxPR0dHYs2cP77AaFaoBf1QDvij//FEN+KL880c14Ivyz0eDnzzgVVRUVCCVSjF//nx4eXnxDqdRohrwRzXgi/LPH9WAL8o/f1QDvij/9Ut0t6K9SKFQQEVFtP1bg0A14I9qwBflnz+qAV+Uf/6oBnxR/uuH6BsbQgghhBBCiPhR60gIIYQQQghp8KixIYQQQgghhDR41NgQQgghhBBCGjxqbAghhBBCCCENHjU2hBBCCCGEkAaPGhtCCCGEEEJIg0eNDSGEEKXy8/ODRCKBRCKBuro6WrZsiYEDB2Lr1q1QKBSv/f9s374dzZs3V16ghBBCGjRqbAghhCjdu+++i/z8fGRnZyM+Ph7u7u6YM2cOhg0bBplMxjs8QgghIkCNDSGEEKXT1NTEO++8A1NTUzg7O2Px4sWIi4tDfHw8tm/fDgBYt24d7O3toaOjgzZt2mDmzJl48uQJAODkyZOYPHkySkpKhKs/y5YtAwBIpVLMnz8fpqam0NHRQffu3XHy5Ek+fyghhBBuqLEhhBDChYeHBxwdHbFv3z4AgIqKCjZu3Ii0tDSEh4fj+PHjWLBgAQCgZ8+eWL9+PXR1dZGfn4/8/HzMnz8fADBr1iycPXsWu3fvxuXLl+Hl5YV3330XmZmZ3P42Qggh9U/CGGO8gyCEECJefn5+KC4uRmxs7F/2+fj44PLly0hPT//LvpiYGPj7++Phw4cAnj1j88knn6C4uFh4T05ODiwtLZGTkwMTExNh+4ABA9CtWzd88803df73EEIIeTup8Q6AEEJI48UYg0QiAQD8/vvvCAwMxLVr11BaWgqZTIbKykqUl5dDW1v7pf/+ypUrkMvlsLW1rbVdKpXC0NBQ6fETQgh5e1BjQwghhJuMjAy0bdsW2dnZGDZsGAICArBq1SoYGBjgzz//xNSpU1FVVfXKxubJkydQVVVFcnIyVFVVa+1r2rRpffwJhBBC3hLU2BBCCOHi+PHjuHLlCubOnYvk5GQoFAqsXbsWKirPHv/cu3dvrfdraGhALpfX2ubk5AS5XI7CwkL06dOn3mInhBDy9qHGhhBCiNJJpVLcv38fcrkcBQUFOHz4MAIDAzFs2DD4+vri6tWrqK6uRnBwMIYPH47Tp09j8+bNtf4PCwsLPHnyBMeOHYOjoyO0tbVha2uLCRMmwNfXF2vXroWTkxMePHiAY8eOwcHBAUOHDuX0FxNCCKlvNCsaIYQQpTt8+DBatWoFCwsLvPvuuzhx4gQ2btyIuLg4qKqqwtHREevWrUNQUBA6deqEyMhIBAYG1vo/evbsCX9/f3h7e8PIyAjffvstAGDbtm3w9fXFvHnz0K5dO4waNQqJiYkwMzPj8acSQgjhhGZFI4QQQgghhDR4dMWGEEIIIYQQ0uBRY0MIIYQQQghp8KixIYQQQgghhDR41NgQQgghhBBCGjxqbAghhBBCCCENHjU2hBBCCCGEkAaPGhtCCCGEEEJIg0eNDSGEEEIIIaTBo8aGEEIIIYQQ0uBRY0MIIYQQQghp8KixIYQQQgghhDR4/w/IG9KFRyjEyQAAAABJRU5ErkJggg==",
      "text/plain": [
       "<Figure size 1000x500 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Visualize the data\n",
    "plt.figure(figsize=(10,5))\n",
    "data['Temp'].plot(rot=60)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**b)** Generate a **7**-lag input dataset manually, i.e., use the previous 7 days data \\[x(t-7), x(t-6)...x(t-1)\\] as the input to predict the following day X(t) using the given function **data_shift()**."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "       Temp\n",
      "0  0.787072\n",
      "1  0.680608\n",
      "2  0.714829\n",
      "3  0.555133\n",
      "4  0.600760\n",
      "       Temp\n",
      "0  0.479087\n",
      "1  0.524715\n",
      "2  0.555133\n",
      "3  0.501901\n",
      "4  0.467681\n"
     ]
    }
   ],
   "source": [
    "lags = 7\n",
    "train, test = data[0:-50], data[-50:]\n",
    "\n",
    "# Scaling the train/test data using minmaxscaler\n",
    "scaler = MinMaxScaler().fit(train)\n",
    "train_scaled = pd.DataFrame(scaler.transform(train), columns=data.columns)\n",
    "test_scaled = pd.DataFrame(scaler.transform(test), columns=data.columns)\n",
    "\n",
    "print(train_scaled.head(5))\n",
    "print(test_scaled.head(5))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor([[0.7871, 0.6806, 0.7148, 0.5551, 0.6008, 0.6008, 0.6008],\n",
      "        [0.6806, 0.7148, 0.5551, 0.6008, 0.6008, 0.6008, 0.6616],\n",
      "        [0.7148, 0.5551, 0.6008, 0.6008, 0.6008, 0.6616, 0.8289]])\n",
      "tensor([0.6616, 0.8289, 0.7605])\n",
      "tensor([[0.4791, 0.5247, 0.5551, 0.5019, 0.4677, 0.3346, 0.4068],\n",
      "        [0.5247, 0.5551, 0.5019, 0.4677, 0.3346, 0.4068, 0.3764],\n",
      "        [0.5551, 0.5019, 0.4677, 0.3346, 0.4068, 0.3764, 0.3156]])\n",
      "tensor([0.3764, 0.3156, 0.5703])\n"
     ]
    }
   ],
   "source": [
    "# Generate the shifted training/test dataset, using the given function data_shift()\n",
    "train_shifted, test_shifted = data_shift(train=train_scaled, test=test_scaled)\n",
    "\n",
    "# Split the train/test dataset to train_X, train_y / test_X, test_y\n",
    "train_y = train_shifted['t']\n",
    "train_X = train_shifted.drop(columns=['t'])\n",
    "\n",
    "test_y = test_shifted['t']\n",
    "test_X = test_shifted.drop(columns=['t'])\n",
    "\n",
    "# Convert the train/test dataset to tensors, the data type is torch.float32\n",
    "train_X_tensor = torch.tensor(train_X.values, dtype=torch.float32)\n",
    "train_y_tensor = torch.tensor(train_y.values, dtype=torch.float32)\n",
    "test_X_tensor = torch.tensor(test_X.values, dtype=torch.float32)\n",
    "test_y_tensor = torch.tensor(test_y.values, dtype=torch.float32)\n",
    "\n",
    "print(train_X_tensor[:3])\n",
    "print(train_y_tensor[:3])\n",
    "print(test_X_tensor[:3])\n",
    "print(test_y_tensor[:3])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**c)**  Complete the given class **dnn()** using PyTorch to implement a neural network.\n",
    "\n",
    "Hints:\n",
    "\n",
    "- [**nn.Sequential()**](https://pytorch.org/docs/stable/generated/torch.nn.Sequential.html)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [],
   "source": [
    "class dnn(nn.Module):\n",
    "    def __init__(self, input_size=7, network_structure=[16, 4], output_size=1, activation_function='sigmoid', \n",
    "                 dropout_rate=0.2, leaky_relu_slop=0.01, output_activation_function=None):\n",
    "        super().__init__()\n",
    "        \"\"\"\n",
    "        Initialization.\n",
    "        input_size: int, the dimension of the input features.\n",
    "        network_structure: list, the strcture of the hidden layers in the network, each number corresponds to the number \n",
    "                           of neuros in a hidden layer.\n",
    "        output_size: int, the dimension of the prediction target.\n",
    "        activation_function: string, the name of the selected activation function in hidden layers. Implement the function\n",
    "                             str2act() to convert the string to activation function.\n",
    "        dropout_rate: float, the probability of an element to be zeroed.\n",
    "        leaky_relu_slop: float, controls the angle of the negative slope for the function leakyrelu.\n",
    "        output_activation_function: string, the name of the selected activation function in output layer.       \n",
    "        \"\"\"\n",
    "        self.input_size = input_size\n",
    "        self.output_size = output_size\n",
    "        self.leaky_relu_slop = leaky_relu_slop\n",
    "        self.activation_function = self.str2act(activation_function)\n",
    "        self.output_activation_function = self.str2act(output_activation_function)\n",
    "        self.dropout = nn.Dropout(dropout_rate)\n",
    "        self.network_structure = copy(network_structure)\n",
    "        self.network_structure.insert(0, self.input_size)\n",
    "        \n",
    "        # Implement the neural network model using pytorch modules based on the given input/output size \n",
    "        # and the network structure.\n",
    "        \n",
    "        module_list = []\n",
    "\n",
    "        for in_size, out_size in zip(self.network_structure[:-1], self.network_structure[1:]):\n",
    "            module_list.append(nn.Linear(in_size, out_size))\n",
    "            module_list.append(self.dropout)\n",
    "            if self.activation_function is not None:\n",
    "                module_list.append(self.activation_function)\n",
    "\n",
    "        module_list.append(nn.Linear(out_size, self.output_size))\n",
    "        if self.output_activation_function is not None:\n",
    "            module_list.append(self.output_activation_function)\n",
    "\n",
    "        self.model = nn.Sequential(*module_list)\n",
    "\n",
    "        \n",
    "    def forward(self, x):\n",
    "        '''\n",
    "        x: torch.tensor, the input\n",
    "        \n",
    "        Feedforward propagation of the input to the output. \n",
    "        '''\n",
    "        x = self.model(x)   \n",
    "        return x\n",
    "    \n",
    "    def predict(self, x):\n",
    "        '''\n",
    "        x: troch.tensor, the input\n",
    "        \n",
    "        Make a prediction for the input. The mode should be set to evaluation mode there.\n",
    "        '''\n",
    "        training_mode = self.training # store the current mode\n",
    "        self.eval() # to evaluation mode\n",
    "        \n",
    "        # predict\n",
    "        output = self.model(x)\n",
    "        \n",
    "        self.train(training_mode) # reset the saved mode\n",
    "        return output\n",
    "    \n",
    "    def str2act(self, str):\n",
    "        '''\n",
    "        Convert the given string to the corresponding activation function.\n",
    "        '''\n",
    "        \n",
    "        if str == \"sigmoid\":\n",
    "            return nn.Sigmoid()\n",
    "        \n",
    "        elif str == \"tanh\":\n",
    "            return nn.Tanh()     \n",
    "        \n",
    "        elif str == \"relu\":\n",
    "            return nn.ReLU()\n",
    "            \n",
    "        elif str == \"leakyrelu\":\n",
    "            return nn.LeakyReLU()\n",
    "            \n",
    "        elif str == None:\n",
    "            # No activate function is required.\n",
    "            # the function should return the input\n",
    "            return None"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "dnn(\n",
       "  (activation_function): Sigmoid()\n",
       "  (dropout): Dropout(p=0.2, inplace=False)\n",
       "  (model): Sequential(\n",
       "    (0): Linear(in_features=7, out_features=16, bias=True)\n",
       "    (1): Dropout(p=0.2, inplace=False)\n",
       "    (2): Sigmoid()\n",
       "    (3): Linear(in_features=16, out_features=4, bias=True)\n",
       "    (4): Dropout(p=0.2, inplace=False)\n",
       "    (5): Sigmoid()\n",
       "    (6): Linear(in_features=4, out_features=1, bias=True)\n",
       "  )\n",
       ")"
      ]
     },
     "execution_count": 74,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model = dnn()\n",
    "model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**d)** Train the **model** using mini-batch gradient descent.\n",
    "\n",
    "Hints:\n",
    "\n",
    "- Convert the training tensors to iterable objects using the data loading utility [**DataLoader()**](https://pytorch.org/docs/stable/data.html). \n",
    "\n",
    "- Use [MSE](https://pytorch.org/docs/stable/generated/torch.nn.MSELoss.html) as the loss function\n",
    "\n",
    "- Use SGD with a optimial learning rate as the [optimizer](https://pytorch.org/docs/stable/optim.html)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [],
   "source": [
    "epochs = 256\n",
    "batch_size = 32\n",
    "input_size = train_X_tensor.shape[1]\n",
    "output_size = 1\n",
    "network_structure = [64, 32, 16, 1]\n",
    "activation_function = 'leakyrelu'\n",
    "leaky_relu_slop = 0.01\n",
    "dropout_rate = 0\n",
    "output_activation_function = None\n",
    "\n",
    "# Convert the training tensors to iterable objects\n",
    "dataset = TensorDataset(train_X_tensor, train_y_tensor)\n",
    "train_loader = DataLoader(dataset, batch_size=batch_size, shuffle=False)\n",
    "\n",
    "# Initialize the model\n",
    "model = dnn(\n",
    "    input_size, \n",
    "    network_structure, \n",
    "    output_size, \n",
    "    activation_function, \n",
    "    dropout_rate, \n",
    "    leaky_relu_slop,\n",
    "    output_activation_function\n",
    ")\n",
    "\n",
    "# Define loss function\n",
    "loss_function = nn.MSELoss()\n",
    "\n",
    "# Define optimizer\n",
    "optimizer = optim.SGD(model.parameters(), lr=0.001, momentum=0.999)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1/256]: Train loss: 0.15142955757117113, Test loss: 0.11012813448905945\n",
      "[11/256]: Train loss: 0.07438057250615242, Test loss: 0.16476798057556152\n",
      "[21/256]: Train loss: 0.04573606882260832, Test loss: 0.04333534464240074\n",
      "[31/256]: Train loss: 0.026590377941266098, Test loss: 0.013966716825962067\n",
      "[41/256]: Train loss: 0.015828502685886568, Test loss: 0.017573921009898186\n",
      "[51/256]: Train loss: 0.013117277448733164, Test loss: 0.015532441437244415\n",
      "[61/256]: Train loss: 0.012888302894688286, Test loss: 0.015015969052910805\n",
      "[71/256]: Train loss: 0.012369578259181132, Test loss: 0.013241994194686413\n",
      "[81/256]: Train loss: 0.01250579320636075, Test loss: 0.0132551658898592\n",
      "[91/256]: Train loss: 0.012418690704188384, Test loss: 0.013205341063439846\n",
      "[101/256]: Train loss: 0.012383670628532371, Test loss: 0.013154490850865841\n",
      "[111/256]: Train loss: 0.01239812412645371, Test loss: 0.01333548966795206\n",
      "[121/256]: Train loss: 0.012373880804640002, Test loss: 0.013226479291915894\n",
      "[131/256]: Train loss: 0.012386488394432627, Test loss: 0.013244195841252804\n",
      "[141/256]: Train loss: 0.012379549585892696, Test loss: 0.013238980434834957\n",
      "[151/256]: Train loss: 0.012380182021915648, Test loss: 0.013235630467534065\n",
      "[161/256]: Train loss: 0.012379713931359945, Test loss: 0.013239557854831219\n",
      "[171/256]: Train loss: 0.012379053607170957, Test loss: 0.013236204162240028\n",
      "[181/256]: Train loss: 0.012379078344911733, Test loss: 0.013238980434834957\n",
      "[191/256]: Train loss: 0.012377660606155354, Test loss: 0.013237003237009048\n",
      "[201/256]: Train loss: 0.01237717408695118, Test loss: 0.013233201578259468\n",
      "[211/256]: Train loss: 0.012376759129172537, Test loss: 0.013234701938927174\n",
      "[221/256]: Train loss: 0.012376613300953028, Test loss: 0.013234112411737442\n",
      "[231/256]: Train loss: 0.012376313806569155, Test loss: 0.013232830911874771\n",
      "[241/256]: Train loss: 0.0123758222454483, Test loss: 0.013232678174972534\n",
      "[251/256]: Train loss: 0.012375539970229816, Test loss: 0.013232015073299408\n"
     ]
    }
   ],
   "source": [
    "train_losses = [] # save the average training loss of each iteration\n",
    "test_losses = [] # save the average test loss of each iteration\n",
    "\n",
    "# Training process\n",
    "for epoch in range(epochs):\n",
    "    # set training mode\n",
    "    model.train()\n",
    "    epoch_loss = 0.0\n",
    "    \n",
    "    # Implement the training loop using mini-batch \n",
    "    for idx, batch in enumerate(train_loader):\n",
    "        x, y = batch[0], batch[1]\n",
    "        optimizer.zero_grad()\n",
    "        y_pred = model.forward(x)\n",
    "        loss = loss_function(y_pred, y)\n",
    "        loss.backward()\n",
    "        optimizer.step()\n",
    "\n",
    "        epoch_loss += loss.item()\n",
    "\n",
    "\n",
    "    # compute test error in each epoch and add it to test_losses\n",
    "    # Do the same for training loss as well\n",
    "    # print the test error in each 10 epochs\n",
    "\n",
    "    y_pred = model.predict(test_X_tensor)\n",
    "    test_error = loss_function(y_pred, test_y_tensor).item()\n",
    "    train_error = epoch_loss / (idx+1)\n",
    "    train_losses.append(train_error)\n",
    "    test_losses.append(test_error)\n",
    "    if epoch % 10 == 0:\n",
    "        print(f'[{epoch+1}/{epochs}]: Train loss: {train_error}, Test loss: {test_error}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1000x500 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot the learning curve\n",
    "plt.figure(figsize=(10,5))\n",
    "plt.plot(train_losses, label=\"Training\")\n",
    "plt.plot(test_losses, label=\"Test\")\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Plot the comparison between measurements and predictions of the training data\n",
    "### Your Code Here ### "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Plot the comparison between measurements and predictions of the test data\n",
    "### Your Code Here ### "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# What problem you can observe in this figure? What could cause this problem?"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**e)** Implement [**Early Stopping**](https://en.wikipedia.org/wiki/Early_stopping) in the training process."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "## Re-initialize the model\n",
    "### Your Code Here ### \n",
    "\n",
    "# The training process is monitored and stopped when the loss doesn't decrease within 30 iterations.\n",
    "early_stop_patience = 30\n",
    "\n",
    "train_losses = [] # save the average training loss of each iteration\n",
    "test_losses = [] # save the average test loss of each iteration\n",
    "best_epoch = 0 # the epoch of the lowest test loss\n",
    "lowest_loss = sys.float_info.max # the current lowest test loss\n",
    "epochs_counter = 0 # count how many epochs have been monitored and compare the number to the patience.\n",
    "\n",
    "# Define loss function\n",
    "### Your Code Here ### \n",
    "\n",
    "# Define optimizer SGD\n",
    "### Your Code Here ### \n",
    "\n",
    "# Define optimizer Adam (for subtask f)\n",
    "### Your Code Here ### \n",
    "\n",
    "# Training process\n",
    "for epoch in range(epochs):\n",
    "    # set training mode\n",
    "    model.train()\n",
    "    running_losses = 0.0\n",
    "    \n",
    "     # Implement the training loop using mini-batch\n",
    "    for idx, batch in enumerate(loader):\n",
    "        pass\n",
    "        ### Your Code Here ### \n",
    "    \n",
    "    # compute test error in each epoch and add it to test_losses\n",
    "    # Do the same for training loss as well\n",
    "    # print the test error in each 10 epochs\n",
    "    \n",
    "    ### Your Code Here ###\n",
    "    \n",
    "    # Implment early stopping\n",
    "    # Update the lowest loss, when the newest test loss is smaller.\n",
    "    # Otherwise stop_epochs += 1\n",
    "    # Stop learning and return the lowest test loss, when the stop_epochs is greater than or equal to\n",
    "    # the early_stop_patience\n",
    "    \n",
    "    ### Your Code Here ###"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Plot the comparison between measurements and predictions of the test data\n",
    "### Your Code Here ###"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**f)** Try [**Adam**](https://pytorch.org/docs/stable/optim.html) optimzer based on the code of sub-task **e** and plot the learning curve. What different can you observe between the learning curve of SGD and Adam."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**g)** Search the optimal hyperarameters using **grid search** and visualize the searching results with [HiPlot](https://github.com/facebookresearch/hiplot).\n",
    "\n",
    "- Optimize these hyperparameters: **the structure of neural networks** (i.e., the number of neuros and layers), **activation functions**, **dropout rates**, **optimizers**, **learning rate**.\n",
    "\n",
    "- The searching range can be adjusted depending on your hardware condition."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "nn_structures = [[32, 16, 1], [64, 32, 16, 1]]\n",
    "activation_functions = ['tanh', 'sigmoid']\n",
    "dropout_rates = [0.2, 0.05]\n",
    "learning_rates = [0.05, 0.01]\n",
    "optimizers = ['adam', 'sgd']\n",
    "results = []\n",
    "param_idx = 1\n",
    "\n",
    "# The training process is monitored and stopped when the loss doesn't decrease within 30 iterations.\n",
    "early_stop_patience = 30\n",
    "\n",
    "# Search the optimal parameters using grid search\n",
    "for nn_structure in nn_structures:\n",
    "    for activation_function in activation_functions:\n",
    "        for dropout_rate in dropout_rates:\n",
    "            for learning_rate in learning_rates:\n",
    "                for optimizer_str in optimizers:\n",
    "                    \n",
    "                    # Training settings\n",
    "                    # declare variables, e.g. losses list, lowest loss etc\n",
    "                    # initialize the neural network, optimizer, and the corresponding loss function.\n",
    "                    ### Your Code Here ###\n",
    "                    \n",
    "                    print(f\"The index of the hyparameter set: {param_idx}\")\n",
    "                    \n",
    "                    # Implement the training loop using mini-batch\n",
    "                    for epoch in range(epochs):\n",
    "                        # set training mode\n",
    "                        model.train()\n",
    "                        running_losses = 0.0\n",
    "                        \n",
    "                        # Implement the training loop using mini-batch\n",
    "                        for idx, batch in enumerate(loader):\n",
    "                            pass\n",
    "                            ### Your Code Here ###\n",
    "                        \n",
    "                        # compute test error in each epoch and add it to test_losses\n",
    "                        # Do the same for training loss as well\n",
    "                        # print the test error in each 10 epochs\n",
    "                        \n",
    "                        ### Your Code Here ###\n",
    "                        \n",
    "                        # Implment early stopping\n",
    "                        # Update the lowest loss, when the newest test loss is smaller.\n",
    "                        # Otherwise stop_epochs += 1\n",
    "                        # Stop learning and return the lowest test loss, when the stop_epochs is greater than or equal to\n",
    "                        # the early_stop_patience\n",
    "                        \n",
    "                        ### Your Code Here ###\n",
    "                        \n",
    "                    # Store the parameters and the result for displaying    \n",
    "                    results.append({\"Structure\": nn_structure,\n",
    "                                    \"activation function\": activation_function,\n",
    "                                    \"dropout rate\": dropout_rate,\n",
    "                                    \"learning rate\": learning_rate,\n",
    "                                    \"optimizer\": optimizer_str,\n",
    "                                    \"test error\": lowest_loss\n",
    "                                    })\n",
    "                    param_idx += 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Visualize the results in parallel plots.\n",
    "hip.Experiment.from_iterable(results).display()"
   ]
  }
 ],
 "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": "3.9.18"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}