single_kanji_cnn_training.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# DaKanjiRecognizer - Single Kanji CNN : Training\n",
    "## Setup\n",
    "\n",
    "Import the needed libraries."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "#std lib\n",
    "import sys\n",
    "import os\n",
    "import time\n",
    "import datetime\n",
    "import tempfile\n",
    "import importlib\n",
    "\n",
    "#ML\n",
    "import tensorflow as tf\n",
    "import numpy as np\n",
    "\n",
    "from PIL import Image\n",
    "\n",
    "# data augmentation\n",
    "from elasticdeform import tf as tfedf\n",
    "import elasticdeform \n",
    "import RandomErasing\n",
    "\n",
    "#plotting/showing graphics\n",
    "%matplotlib inline\n",
    "import matplotlib.pyplot as plt\n",
    "from IPython.display import Image"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Because the data sets are big let's make sure that the GPU is available to speed up training."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "GPUs Available:  /device:GPU:0\n"
     ]
    }
   ],
   "source": [
    "print(\"GPUs Available: \", tf.test.gpu_device_name())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "enable memory growth"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "physical_devices = tf.config.list_physical_devices('GPU') \n",
    "tf.config.experimental.set_memory_growth(physical_devices[0], True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Enable mixed precision (for RTX, etc.)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Mixed precision compatibility check (mixed_float16): OK\n",
      "Your GPU will likely run quickly with dtype policy mixed_float16 as it has compute capability of at least 7.0. Your GPU: NVIDIA GeForce RTX 2070 SUPER, compute capability 7.5\n"
     ]
    }
   ],
   "source": [
    "policy = tf.keras.mixed_precision.Policy('mixed_float16')\n",
    "tf.keras.mixed_precision.set_global_policy(policy)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Setting up the Data pipeline\n",
    "\n",
    "Read the labels for each class from file."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "labels_1, labels_2, labels = [], [], []\n",
    "\n",
    "# load labels from file\n",
    "with open(r'Z:\\data_sets\\dakanji_single_kanji_cnn\\encoding.txt', mode=\"r\", encoding=\"utf-8\") as f:\n",
    "    labels_1 = list(eval(f.read()).keys())\n",
    "\n",
    "ls = labels_1\n",
    "# order the labels\n",
    "indexs = sorted([str(i) for i in range(0, len(ls))])\n",
    "ordered_labels = [ls[int(i)] for i in indexs]\n",
    "\n",
    "# save the labels to text files\n",
    "with open(r'E:\\projects\\DaKanjiRecognizerML\\single_kanji_cnn\\labels.txt', mode=\"w+\", encoding=\"utf-8\") as f:  \n",
    "    f.write(''.join(ordered_labels))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Create a `tf.keras.dataset` from the saved files for test and validation."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "#batch size\n",
    "bs=1024\n",
    "val_split = 0.2\n",
    "h, w = 64, 64\n",
    "# class names\n",
    "classes = [str(i) for i in range(len(ls))]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "#define a font to show japanese characters in matplotlib figures\n",
    "import matplotlib.font_manager as fm\n",
    "show_sample_font = fm.FontProperties(fname=os.path.join(\"..\", \"fonts\", \"NotoSerifCJKjp-Regular.otf\"), size=20)\n",
    "\n",
    "def show_image(img : np.array, label : str):\n",
    "    plt.figure(figsize=(2.2, 2.2))\n",
    "    plt.title(label=label, font=show_sample_font)\n",
    "    plt.axis(\"off\")\n",
    "    plt.imshow(img.astype(np.float64), cmap=\"gray\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## create train and validation dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 13014000 files belonging to 6507 classes.\n",
      "Using 10411200 files for training.\n"
     ]
    }
   ],
   "source": [
    "t_dataset = tf.keras.preprocessing.image_dataset_from_directory(\n",
    "    #\"Z:\\data_sets\\kanji_cnn_test\",\n",
    "    \"Z:\\data_sets\\dakanji_single_kanji_cnn\",\n",
    "    labels=\"inferred\",\n",
    "    label_mode=\"categorical\",\n",
    "    interpolation=\"lanczos3\",\n",
    "    #class_names=classes,\n",
    "    color_mode=\"grayscale\",\n",
    "    batch_size=bs,\n",
    "    image_size=(h, w),\n",
    "    shuffle=True,\n",
    "    seed=123,\n",
    "    validation_split=val_split,\n",
    "    subset=\"training\"\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_dataset = t_dataset.map(lambda x, y : (tf.cast(x, dtype=tf.float16), tf.cast(y, dtype=tf.float16)))\n",
    "train_dataset = train_dataset.cache(r\"Z:\\data_sets\\dakanji_single_kanji_cnn_cache\\train\")\n",
    "train_dataset = train_dataset.prefetch(buffer_size=tf.data.AUTOTUNE)\n",
    "\n",
    "#print(train_dataset.take(1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 13014000 files belonging to 6507 classes.\n",
      "Using 2602800 files for validation.\n"
     ]
    }
   ],
   "source": [
    "v_dataset = tf.keras.preprocessing.image_dataset_from_directory(\n",
    "    #\"Z:\\data_sets\\kanji_cnn_test\",\n",
    "    \"Z:\\data_sets\\dakanji_single_kanji_cnn\",\n",
    "    labels=\"inferred\",\n",
    "    label_mode=\"categorical\",\n",
    "    interpolation=\"lanczos3\",\n",
    "    #class_names=classes,\n",
    "    color_mode=\"grayscale\",\n",
    "    batch_size=bs,\n",
    "    image_size=(h, w),\n",
    "    shuffle=True,\n",
    "    seed=123,\n",
    "    validation_split=val_split,\n",
    "    subset=\"validation\"\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "val_dataset = v_dataset.map(lambda x, y : (tf.cast(x, dtype=tf.float16), tf.cast(y, dtype=tf.float16)))\n",
    "val_dataset = val_dataset.cache(r\"Z:\\data_sets\\dakanji_single_kanji_cnn_cache\\test\")\n",
    "val_dataset = val_dataset.prefetch(buffer_size=tf.data.AUTOTUNE)\n",
    "\n",
    "#print(val_dataset.take(1))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Defining and training the CNN\n",
    "First define image image preprocessing layers."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "@tf.function\n",
    "def preprocess_tf(x):\n",
    "    \"\"\"\n",
    "    Preprocessing for TF Lite.\n",
    "    \n",
    "    Args:\n",
    "        x : a Tensor(batch_size, height, width, channels) of images to preprocess\n",
    "        \n",
    "    Return: \n",
    "        normalized and resized Tensor of images\n",
    "    \"\"\"\n",
    "    \n",
    "    batch, height, width, channels = x.shape\n",
    "    \n",
    "    # normalize image between [0, 1]\n",
    "    m = tf.math.reduce_max(x)\n",
    "    preprocessed = tf.math.divide(x, m)\n",
    "    \n",
    "    # resize images to (64, 64)\n",
    "    preprocessed = tf.image.resize(preprocessed, (64, 64), method=tf.image.ResizeMethod.BILINEAR)\n",
    "\n",
    "    return preprocessed\n",
    "\n",
    "\n",
    "class PreprocessTFLayer(tf.keras.layers.Layer):\n",
    "    def __init__(self, name=\"preprocess_tf\", **kwargs):\n",
    "        super(PreprocessTFLayer, self).__init__(name=name, **kwargs)\n",
    "        self.preprocess = preprocess_tf\n",
    "\n",
    "    def call(self, input):\n",
    "        return self.preprocess(input)\n",
    "\n",
    "    def get_config(self):\n",
    "        config = super(PreprocessTFLayer, self).get_config()\n",
    "        return config\n",
    "    \n",
    "    def get_prunable_weights(self):\n",
    "        return [] "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now define a Sequential model for image preprocessing and augmentation layers."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "from tensorflow.keras.layers.experimental.preprocessing import Resizing, Rescaling, RandomZoom, RandomRotation, RandomTranslation\n",
    "\n",
    "img_augmentation = tf.keras.Sequential(\n",
    "    [\n",
    "        PreprocessTFLayer(),\n",
    "        RandomErasing.RandomErasing(probability=0.3),\n",
    "        \n",
    "        # random data augmentation\n",
    "        RandomZoom(height_factor=(-0.2, 1.0), width_factor=(-0.2, 1.0), \n",
    "            fill_mode='constant', interpolation='bilinear', fill_value=0.0\n",
    "        ),\n",
    "        RandomTranslation(0.2, 0.2, fill_mode=\"constant\"),\n",
    "        RandomRotation(factor=(-0.1, 0.1) , fill_mode='constant', interpolation='bilinear'),\n",
    "    ],\n",
    "    name = \"img_augmentation\"\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "imgs, lbs = [], []\n",
    "for images, labels in train_dataset.take(1):\n",
    "    for img, label in zip(images, labels):\n",
    "        imgs.append(img.numpy())\n",
    "        lbs.append(label.numpy())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "elapsed timer: 0.8852492999999413\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAjgAAAIuCAYAAACy+nJwAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjQuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8rg+JYAAAACXBIWXMAAAsTAAALEwEAmpwYAADz90lEQVR4nOy9WXMba3alvTAkMpGZmAlOIiUdSefUqcmOdnfZ0f4Dfd+/tq8dHV93u2yXy1WuKp9RI0fMQwKJ8bs4XlsbKVAiJUoiwf1EMEhRJACSOax3v2uvnVoulzAMwzAMw9gk0p/7BRiGYRiGYVw3JnAMwzAMw9g4TOAYhmEYhrFxmMAxDMMwDGPjMIFjGIZhGMbGYQLHMAzDMIyNI/u2/0ylUtZDbnw2lstl6nO/hiR2ThifEzsnDGOVt50TVsExDMMwDGPjMIFjGIZhGMbG8dYtKsMwDMMwPj+p1OudGJtAcDmsgmMYhmEYNxgtbozLYxUcwzAMw7iBaGGTFDlWxXk3VsExDMMwjBtMKpWSN+PyWAXHMAzDMG4ASQFzkaCx6s3lMIFjGIZhGJ+Zi8TMdVRt+Bh3TRiZwDEMwzCMG0LSd/Oh4iT5eHdJ5JjAMQzDMIzPxDojcTqdlvfpdBrL5RLz+VzEydtaxi/a5kqlUlgsFtf/A9xgTOAYhmEYxifgbdtNWojwva7g6OpL8mPg7UJn3WPcBUzgGIZhGMZHIpvNIpvNIpPJIJ/PI51OYzqdYjabYblcSlVlPp9jPp8DeC1uWLlJpVLI5XJIpVJvfB0ALBYLeRwKmHQ6/Yb4WS6Xd0rkmMAxDMMwjI9ENpuF53nI5XIol8twHAfD4RDj8RiLxUK2nuI4FpHCLSoKl2w2C9d1kclkMJ1OMZlM5OtSqRRms5k8Dt+4vcXH0eLmrogcEziGYRiGcY2kUik4joN0Oo2trS3UajW4rotKpQLHcTAajTAajTCbzRBFEWazGUajEaIoQiqVQjabRTqdRi6XQy6XQyaTQTabFTGzWCze+N7RaLTyGih03vYa+XWbigkcwzAMw7hGstksyuUyPM/Db37zG/zX//pfkc/nUS6Xkc1mMZlMMJlMEEURjo6OMBwO0Wg00Gg0kMlkEIYhHMfBgwcP8ODBA8xmM/R6PcxmM6TTaWQyGXQ6Hfzxj39Eu93G6ekpRqPRiqjR21/A620vVnVI8us2CRM4hmEYhnGNpNNpOI4Dz/NQq9VweHgI3/elgsNtpuFwCAAYDAbir3EcB6VSCa7r4sGDB/jqq68wmUzQaDQQxzEymQwymQw8z8PTp08xGo2QyWTkudeZlNd1Ut0FX44JnBsAD069Z8r9VO6dGoZhGLcDVlK0+TedTiMIAnieJ18XxzHy+TziOMbDhw/R6/WQy+VQqVTgui6q1Sq2trYwmUwQBAHG4zFevHiBFy9eoNFo4OjoCM1mE4PBAMvlUozMmUwG5XIZpVIJAOQ1zGYzeev3+5hOpxiPxxiPx/K6NwkTOJ8ZlgzT6bS47ReLBabTKZbLpZjHDMMwjNsDxY02/fq+jzAMxTQ8n89Rr9dXhIfruqjX63BdVwRHHMfwfR9RFOHZs2d4+vQpGo0GXr16hU6nIx1ZmUxGRNTh4SEODw8BANPpVO4r9O4cHx8jiiIxOG+auAHuiMBZlz3wuf+YLBvygMxms3AcB47jYD6fi3kMwBslRcMwDOPmQjERxzFarRZevXqFfr+PbDaLIAiQyWTgOA6Wy6UsZsMwRBiGyOVy0lYOvPbIdLtddLtdtNttdLtdDIdDETZBEMB1XeTzeezt7SEIAuzt7WFvb2/ldbFa0+/3MRwOkU6nEccxhsOhbJFt0n1mowWOFhHaMZ5Mhfwc0BWfz+dx7949+L6PfD4Pz/MwHo9xcnKC0WiEfr8vr9UqOYZhGDef+XyObreLwWCA3//+9zg/P0cQBNjf35ctKnpfHMdBJpPB3//93+Pv/u7v4Hme3LP4NhqN8Pvf/x6vXr3CH/7wB3z33XcYj8eI4xjpdBoPHjzA48ePUavV8Nd//deoVCoIggDFYnFFVFEcnZ2dIZvN4vT0FABkQT0ejzfqPrOxAkcfHDxYdEbA5/a2cEvKdV2EYYhCoYB8Po98Po/RaIRut4v5fC7tgovFYmONYIZhGDedq7RV014wm83QbrexWCzg+z4mkwk8z5PtK3pmPM/DcDiUyk3SIEzB1Gq10Ov1EEURptOpmJmLxSJ2d3dRr9fxxRdfoFqtwvM8eJ4n/5/NZhGGIYIgAACUy2WMRiN4nodsNruR95aNEzg8CLndo01XDFXSW0DaBPYpYDBTrVZDvV5HuVzGL3/5SwmAchwH3W4X4/FY2gn7/f5GHnyGYRg3mXXjDoCr2QVo4I2iSK7rvOfUajV88cUXkpXjOI6InHQ6LR7MTCaD+/fvw/d9ycnJZrPSbfXw4UM8fPgQhUIBh4eHCMNQ7nXcBmNgYKVSwWKxwJdffolyuYwoinByciKVIqvg3GCoeNmix5a7bDYrB8tkMhF/C0XOp3ptzDCo1Wp49OgRtra28F/+y39BtVqV/z87O8Px8TGWyyX6/b4MW7MKjmEYxqclOSPqqtfi0WgkIqfRaAB43TGbTqexvb2NL774AvV6/Q2Bs1gsZBvq/v37qNVqUrUJwxA/+9nPUCwWUa1WUa1W4bquLJZ7vZ4sjilwXNeV7aonT56gVqvh6OgIf/7zn8UIvUn3mY0UONz+YQpkPp8XgTOdTgH8FMQ0n88/afWG22XcmgqCQIxlYRjK9lkul5PqUzLHwDAMw/g0JMXNRZ97130kmTmTz+dRKBREmFQqFaRSKXS7XQDAZDKRigvvXblcDsBPVZ8oiuD7PqrVKsIwRD6fRyqVwnQ6FRHV7XbR6/VkkZ/L5eD7PhzHWbkXZTIZ5HI52fLapITjjRI4/KOxHa9UKsHzPGxvb8N1XYzHY1HT4/FYlCqd6B/7D5pOp5HP5+G6Lmq1Gvb397G9vY2DgwNUKhV5XcPhUPw4nD+yCQebYRjGbUCn/iYrGswq01UWCpLLQFPw119/jcPDQ/zVX/0V9vf30el08Oc//xndbhd//OMfMRgM8NVXX+Grr76C7/vY3t6WjJwvv/xSqjgc/cDuqG+//Ra9Xg/tdhvtdhvFYhG//vWvUS6Xsbu7iyAIVra5PM+D7/siqFjtIbf53rNRAgdY3QZyXRee54lQALBi7tLBep/qdfGgcl0Xvu9L91QQBCK2qKp1BccwDMP4+CSrNckkYF6Xef+Yz+eyM3DR42nBlM1mUSwWsbOzg3q9jmKxCN/30Wg00G63cX5+jh9++AGdTgdhGGJnZwfAT7sOvu/D8zyUSiWZVcXKzXQ6RRRFOD09RavVQrPZRKfTQbVaxf379yVBGVjdItPTzinoNmWbaqMEjj7ocrmcuMi55TOZTN6IsebHwMdXqvp1JSs0wE+thXEci0foU1WWDMMw7jrJ+4EWJbVaDZVKBb7vY2dnB67rotPpoNPpYDwe4+zsDOPxWHyebGgBgK2tLWxtbaFQKODJkyfS8bS7u4tMJoN/+7d/AwBJKO73+/jhhx8QRRGKxSLy+Tzq9Tq2trZkF8DzvJVOradPn+K7775Dq9XCH//4R3Q6HbTbbXQ6HTEw93o9eJ6HarWKOI6lesMBnvSj8n5D7ye5jfehjRI4AKR6Q+8NBU4ul5NJretEzqd6bXxdSYHDVsA4jhHHsZwom+RoNwzDuImsW/BqgbO1tYUnT55I12uhUMAPP/yA58+fo9vtyugDXr9pDl4ul9ja2sJXX32Fvb09/I//8T+kIgMAZ2dn+Id/+Aecnp7i5cuXePnypcyoms/nCIIAjuOg3+/jZz/7mVT88/m85NZwfMO//Mu/oN1u4y9/+Qt6vZ4EA9brdezu7iKKItTrdRweHmKxWEgjTjqdloRjLXDet2vsJrFRAkcfmBQ6yeoMlbWO0f4Ur4tvLAeyJTybff0nmM/nK5HduoJzWw8wwzCM24jeEaAhV+8MbG1tYTaboVgsYjabYTAYoN/vo9/vA4BkmH355Zf48ssvsb29jVqthlKphNFohCiKVu5DyWs/RRL9NcxCI4vFQjyb/X5fQvyiKMJoNBIvDRfO4/FYtqiSOTtJz5H+HWjBc9vuQxslcIDXe4u6TRx4LWyoVDnN9VNuA6XTaTF0sYMqn89LKXAymWAwGGA4HMpBzYP0th1YhmEYt4V1VQvdaeQ4DnK5HFzXRaFQQLlcxtbWFn79618jjmM0Gg2Mx2M8e/YMz549QxAEePDgAQqFAh4/fozHjx/DdV1UKhVks1k8e/YMZ2dniKJIKvv083ChO5/PMRgM0Gg0UCwWZdYURch0OsXZ2Rk6nQ5+/PFH/OUvf8FgMMD5+bkIIpqgu90ustmsbE3x//S9cjqdys+q8+Fuc1fVxgkcrUJp1AXwhojRVR52KelWPv0++fGHvDYeQNrYxcdfLBYrCl63sd/Gg8swDOM2o7eu2OJNCwQT6KfTKYIgkAVzHMcibMrlMr744gs8fPhQ7ktsctEhfvpewOfie21V0BUWdm+NRiMMh0MMBgMJE6TflItnmpBZBVpnzdD3zqTR+DZWb4ANEzi6pMhOJapihhxVq1XUajU8ePAAqVQKk8kE0+l0Jd2Y+5qz2Qyj0WjlcxQhV/lj83Uxj6BaraJcLotpjbkEdMFHUbTixdETaQ3DMIyPC2/ovDccHR0hjmPs7u5iZ2dHPmZKfq1Ww3K5RD6fx4MHD2QiuOd5CMNQtpqeP3+OKIoAAEEQIJ1O4/HjxxgMBpjNZhgOh5hMJpJE/POf/xy/+MUvsL+/j62tLQRBIHk4TCXOZDIol8vSicvcHIqjVCqFYrEo86my2SzG4zHa7bZsbVEcXeTDua33no0SOMBrMUFjMQCZ1sryYhiGEnvNLavJZIJWq4U4jtHpdNDtdhHHMZrNJsbjMXq9HgCI6r7qH5zpyhQ41WpVcnqo3KfTKUajkWxPxXFsW1SGYRifCFZp+DGFwvHxMY6Pj9Fut/H48WOZLbW1tSVCw3EcHB4ergTpAT+F9vH+8vvf/x7NZhOPHj3Cw4cPkc/nkcvlxJ7Ae1AQBJjP5/j666/xd3/3d6hUKtje3kYYhvK4juOgUqlIejGNx/T+cCcglUqhUCjIgpo7Fp1OB61WS4QOF/W6ynPb7zsbJXD0lpMOaNIDNrm/yT1GVlYAiGM9lUrJQed5nrjaOZ6eVR0Kkfl8Lu/XiREeLJlMBr7vy8AzHtzM4lksFiK4tNEMuL0K2jAM4zax7lrLz00mEzSbTWmxzmQykmPjOI7sHujxCOfn55Jx02g00Ov1MJlM5Os47JLenjiOZXZiqVSSVnGOe+D9hPezVCoF3/dl1AOHe/IewtTkQqEgXVPz+VyqN8PhUCpVyfvNbb/vbITAWRfux4ONQob7mNxm6vV6mM/n8DwPruvCcRz4vi+Pue4Py7wAlhGfPn2Kp0+fot/v4/nz53KgTCaTta8vn8/j3r17ODw8xL1797C9vb2SQUAVPxgM1pqMb/vBZhiGcVNJei7XjckZDAb413/9V/i+jz/96U8Sovc//+f/hOu6yOVyyGQyiOMYrVYLo9EI/+f//B/8v//3/2ShDAD3798XgRQEARaLBb744gvx8HQ6Hcznczx58kQ+/+rVK7FUULjs7+/DdV0cHBzgb//2b3F2diY7D9xt2NvbwxdffIHDw0NsbW3JWIYffvgBz549w9HRkXRdJaNJbnsX70YJnHUVHAAiELi1NJlMMB6P5Ws590krcJb5+J5t3bPZDFEUSdmx3+8jk8ng9PQUcRxfmFuTrOAwkVK/rqTJ2Pw3hmEYnxa9TZXsqprP5+h0OuJXieMYvu9jMpm8IZBoAD4/P8fTp0+l2sJuJj1rCgDCMJSRPcBP9y3eK0ajkSx+ee8pFAqo1+vI5XIIggD1eh2z2QxhGK60oIdhiGKxiGKxiFwuJzsFHOfAnQndOaUrOLf53nNrBM5FvfkARHx4nodCoSBbP8DrfVSq5yiKkM1m0el05MBgeZEt26zqUF1zLH2pVEImk0GlUgEAicNOpVJwXVfmeCSh4Tmfz6NYLKJcLiMMQ3ieJ+ZmZhUww8A8OIZhGJ+WdaMZtNUhm82iVqshDEN89dVX+PLLL2XkQj6fl8UybQccnhwEAcbjMbrdLubzOb777jsUi0UUCgUcHh7C933pvOJiF/hJ9DSbTTSbTfzzP/8zGo2G7EjU63WUSiUZ9/Dzn/8c9Xod/X4f5+fnyOVyyGazqNfrePToEba2ttBut/H8+XO8fPkSf/7zn/Hy5Uu0Wi25zyTvNbf9vnOrBM5FIof7odxrZIIx9ylZsel0OjJp9cWLF0ilUqKQc7kcCoWCCJ0wDJHNZlEul+G6ruyV+r6PSqWCXC6HRqMhBxy3xeI4fuN164CopMCh8qcIo8BJdlFd9nekue0Hp2EYxqfiouA7slwu4TgOarUaqtUq/uqv/gp///d/jzAMpWNKeyp5/eVQzMViIdf37777DtPpVITR1tYWwjDE3t6e2CUymQyOjo5wfHyMFy9e4J//+Z9xdHQkr+fw8BBffvmldGzV63U0m00xKzNrjcIpn8/j+fPn+O1vf4ujoyP85S9/wdnZmWxlAe+ein7buBUChx1IyYONWzs6iIligv4bnRDJvUt949dOd5YMmfrI+VWO40iFJZ/PYzQaIZfL4eTkBJ1OB/1+X/Yvk1k73JpyXVeqPMmEZf36uAd62aqNzi3QU275/HwOPpZ+7E07mA3DMK4LXjd5DafwqFQq2NraktZs13VlgaobWhzHWRmqPJ1O5do/Go2k+n96eirbV4VCAQAkG0eLJT4H7xccx1AsFhGGIWazGTKZDKrVqjw336IownQ6RavVwunpKZrNplgqNnmH4FYInGw2K2PhyXw+x2g0wmKxkMqN7/vSLkcX+nQ6lThrbvtQRCyXS9myYqUl6eOhIGFVyHVd1Go15HI56ayKokjaymez2cprp8O9UqnICUFz2Xw+l6oNcwh0Z5beE02i2+F1lHgmk5HH55vO7uFz8Lk3+eA2DMO4KutCYVkN2dvbw69//Wvs7+/jyZMn2N3dxWKxwGAwkKaVfD6PVCqFUqmEQqGA3d1d3Lt3D57nSZ5Oo9HA6ekpSqUS4jhGpVLBf/tv/w2VSmWtnxSAZOlwHMRsNsNf/vIX9Pt9LJdLBEGATCaDX/3qVwAgOxdxHOPo6AjD4RC/+93v8I//+I8YDAbo9XobP9D5Vggctm2v89Uk04H1fCeKhGQ6MEc28HH0CPnkTA6+pxksl8uh2+2KWYuvY51XRs+f4uvnQcvts7cZi/VjJI1vOqlZx4nT78PXy9+Frg6xy2tdmqVhGMZdJDlcUl9vWb0JwxDValWmgzuOI4tobTTm12ezWbE8RFEk12gahufzuTSoULRc1KjCXQt6NaMoQqfTQRAEYj7O5/MolUrI5XLo9/vSpdvv99HtdsU3yg7dTRY3wC0RONz3ZBmOooJChQZeHlDZbBaTyUQUL6s3/J5kZ1LSOb7OZDadTpFOpyUMSY930Nk4+uDk97qui1KpJL6bXC63kmPA16q/l9Ujnlzc5qI4SlZwOCeF1SbdAabb4wGIqOMBblUcwzDuIusWeVqklMtleJ6Hr776Ck+ePMH29jZ+8YtfoFarYT6f45tvvsFgMMC3336L4XAo/pwwDPHgwQOp6LAt/MWLF1JV7/V6GI/HePXqFdrtNh49eoSTkxPZcqJHkwtoLmo5fiGKIvz444/odDqoVqt48OAB5vM5yuUylsslut0ujo6OVgzKP/74o8SZ3AWLwq0QOLlcDvV6HeVyWUYmULgAEPMvA5M450OLG+5BJp3iV1WwqVQKURS9ofYvaudOpVJiLmbQkh6sRrHGvVWiW9hZwSqVSvJvANK+zm0wbk0xzIlfpwUYx06k02nZqtqExErDMIzLklzAAqv3At0xVSwW8atf/Qr//b//d1QqFXz11VcIwxDffPMNvv32WzQaDfz2t79Fp9PB4eEh9vf3UavVEAQBKpUK8vk8Dg8PkcvlsL29DQDodrtYLBYYjUZ49eoVstksXr58iZOTE0wmE2xvbyOdTkuzCXcZtH9zNBrhxx9/xMnJCR4+fIher7cymqjX6+HVq1c4OjrC7373OxwfH+P8/BzD4fCt9odN4lYIHD2Fm74aYHVA5WQykQoLx8jTz0Lx8D4jFpLwOZMCZ92+rd6iYgiUHqrGg4xvHOLGdnddrclms1Kh0ttnVPQMK9TmNG0k1mGHtz28yTAM47rh9TYIAgRBgGKxiK+++gqlUgn7+/syB4pZaL1eT1q4O50O2u22JAa7rovRaATf95FOp+H7vjwmP8/RCvRCahMxAKnMM8qE9w9tPB6Px0ilUhgMBmi32wB+Ej6sEvHref2/a7lqt0LgUEnX63VxoAM/+UriOMZgMBARwAhsGoAZxqeHan4oV1G+2mRcKpWk0sJtL1ZvJpMJcrkcqtWqHMzaBKw7yXR1hgc7P0/Rw8enqKHYo+Dj9pRhGMZdIzlzihYAx3Hwy1/+Er/61a9Qr9fxt3/7t9ja2pJrKwC0223Znvrd736HVquFP//5z+Kh4QL70aNHSKfT0sKdzWbxxRdfIAgCMQuPRiM0Gg2xEfR6PTEp837A5GGOhKCXc7FYoNvtYjAY4IcffsC//uu/Ynd3F5VKBbVaDYvFAoVCQYZ63sWF7a0QONyK4RYU0Ybh8XiMXC4nXVG6esP28E/tNdFeGXZgaQVORc2fgwc1ABSLxZX5IEnjcxJ+Lvn/64zMepvurh3whmEYRI9kYKNGpVLB/v4+dnZ28OjRI9RqNfT7/TcmfrfbbTSbTbTbbfR6PTEOa88nvZvcgSgUChiPx5K/BkAEDH2Z2gvK+562X/A+oK/rnCvleR6iKEIYhvLYrOzrXYXkbsOm3gNutMDRFQkAImbiOJaDhxO4ucVDfwn3Lvk9epDYp4IHExOMC4WCdFPRCzMcDmV6ORU9q0y62yo5NBR47arnx1rMcBXBE40D3obDoTzHdWzZGYZh3FZYxWEjSz6fl+2ldruNP//5zwiCQAZmsuIyHo/x448/4unTpxLr4TiOmH3ZGs60YcdxEAQB7t27h0KhIKMa6PXJZDJ48uQJvvzySwRBgNlshm63i263i16vh5OTE/T7fRFXyfyas7Mz/PGPf0S/38cvfvELsTrs7e1hPp/j/v378jNzzAQfZ5O58QKHbXWsdKwTOEwrpsucrXG8gX/OG7k2GYdhKGMkOKeEAocj67n9pn8W7qHqLTYKGvqRdIWGs0XYSpgcB8F2RhM4hmHcdVgpoaHYdV2Mx2Msl0v8x3/8BzKZDJ49e4ZXr16tDFYejUYYjUby/az+HB4eYnd3F/v7+5JwzLbxe/fuYWtrC/P5XAYwc1RDuVxGsViUOVZM3j8/PxeBo7uB9aL3/PxcWsXPz89RKBQkt2e5XOLw8BAAMBwO0Wg0xOujLRCbeC+48QKHe4esdoxGozdyZ5LGX50x8zlJ5uDoJOPlcilibTgcYjAYSBkUeJ1GrCszOilTCxy+p8DRg+CSVSx9YhiGYdxV9DVwPp9jOByKWAF+qup0u11kMhmcn5+j2WxiNBohiiLJkeHOAa/xQRCgXC6L0TjZFOK6LgCgXC5jZ2cHuVxOfJlMJ57NZuh0OojjGKenpzg9PcXZ2Rm63a5M/U5ew3VobLvdxvn5OcrlsmSesQP51atXMnoo2SK/iSLnRgscdhUtl0ucnp5iMBjg/PwcrVZrJTWYAiDZ2aTff2q02PI8D9VqFeVyGb7vw3VdSb/sdDp4+fIlzs7O0Ol00Ol0RBTxgOPPp7NsKO5Y6mSWDj9OuuaZWHnXXPSGYRgXwevgaDTC8+fPV8bqUJCkUilpWtEJ8LyWptNp8dTcu3cPX3/9NcrlMkqlkmSSUTiVSiVJHT44OJDq+nQ6ldmGvV4P//Ef/4Fut4vnz5/jxYsXGI/HaLfbF4bKsiLTarXwhz/8QWYulkolzGYz/PznP8eDBw/Q7/fx9OlTZDIZGTH0tsDX236fuNECh6qXFRwAcjDoLqDPLWYugq+fc0H0pFlWXZhKSYOa7gjjOAduJyX3TPXnWK3RHhxgtYXdhI1hGMab6ArOcDiU6zQ7V3l9XYc2KNNI7Pu+XO/117E6lMlk4Pu+VIJYeW+1Wmi323j58iXa7TaePn2K58+fvzPig4IrjmO0221ks1nZ0kqn09LBWywWJUCQwutt94TbXtW50QKHOQAM7tODxm66f4QHMBOWuSoAIFUWbk3Rg8ODm9+vlbUWLjqBmZUZ/l6SxuN17w3DMIw3oXjQ1e9UKiVdrsDqEGO+n0wmyGQy6HQ6OD4+RhzHqFarb1gN+PhMmp9MJrL99Oc//1mMwvT5tNvtlXlR77qGj0YjfP/99zg/P8f+/j4ePXoE3/dRq9Xg+z729vbw+PFjtFotEXKMKUluW2lxc1vvHTda4DDJlwIHwK0QODQIM9yJIkePaNDem06nIzNCWq3WyoGmfUjJbiqrzhiGYVwv2ruphQ1ZZ4VgRaTb7cp08IODA/G7sMJCL6Tv+8hkMojjWDq0/vSnP+H//t//i/F4jG63u+Kr1M/1Nphu7DgOnjx5grOzM2l7D4IAu7u7ePz4MQqFAo6Pj2UxTK+O/hn5lhxldJu40QInGaO9rorxsZ+fXPb5knOimGCst6Z40DLoT+fTJGdZXdQerpX1bTzwDMMwbgK6i+hts6mSlZvk9/NjnU4/n8/RbrfFW6m3rZh5xigUzlHUScXJvJrka1r3Wvm8w+EQzWZTwnGZqbO1tYXZbCbdXXq3YNPuJTda4ACv/6hxHEslJ5kBcN3opOCL5pTo9+teczabRbFYXBmymc1mpa17OBxKCiX3YdclLV90UG/agWgYhvG5eFur9NuEDd90nIeONBmPx/jd736H/+//+//g+z4eP36MYrGIBw8ewHVdERphGKJcLmNrawu9Xk/S95OvQQuxty1uF4sFXr16hX/6p3/C/v4+tre3MZ/PUa1W8Zvf/AZPnz7FH//4R0RRJNtryTmNyZ/5NvpxbrzAAV4fSLp68zFawPVBpJOBk2F7yde2jnUVHB4gHLLJCg4NZOtE2207oAzDMG4jyUqN/vxlqic6bR547ZFstVp4+vQpCoUCSqUS5vM56vW63M/YYs4KSxzHK2n0ycV2UuSs+zmWyyUGgwGazSby+bykK7uuizAM0el0ZPAzO3bX/dy3nRstcDgFWwsO5gBcVwWHU7u1Y16XCvnHB14nBzO5UpuzKLh09g2Hq9F/4ziOVKBGoxEGg4H8PJ8jadkwDMO4mHVCImnE5SDkMAxRrVaxvb2NIAikMs+U+vF4LMnInudhd3cXqVQKOzs7qFar+P7772XUgs5L0wJr3TgH/n/ydXU6HXz33XeI4xgvXrzAcrlEqVRCqVTCdDpFoVBArVYTzw99rfo+tG7n4jZxowUOxzAAr3+5ukX8ugQOVSyNXxQjnAKuW/2YMDmdTiUdOGny5UEfhqEMO+PjULUzxTiKIsn0uY0HkGEYxiaTvLbzPT/OZDIIgmBF4GSzWYzHYxm50Ov1sFgs0G634bou7t27h1/84hcIggDb29vIZDKo1WpwXXdl7tS6575oGy1ZgWm329KF9fLlS6TTacxmM2QyGRneubW1hW63C9d15d6qn+O2ezxvnMBhBxKNV7olOtlJ9K7HSSpabQBmtYbt29lsFp7nSecWA544+VsbgLPZrLjrL1L3qVRK5oEk82+4P6vF2qceBGoYhmFcjovMx/q+ou9bHBfEazwT5QHIbMBer4flcolyuSwDOSuVCmazmdyT9H3hfSoptHZ0u120Wi0EQSCvg2KK2TybyI0TOJlMBoVCAY7jrEwBZ3rkZf033GbSQsdxHBl2yTH0vu/D931xuOv5V3S2p1IpRFEkoyKGw6FEdLOFWwsdPlaxWES1WhWT8XK5FFPXcDiUCbSs4HyMwWebuK9qGIbxsVjXTXVR2i+vpwz605WXOI4xGo3Q7/fR7XYxmUwkeO/HH3/Ed999J/OvuF3061//GkdHRzg6OsJ0OpX5hBdtFSU/t+51j0YjfPPNNzg/P0c6nUa9XsdisUClUkEmk8HR0dGNmgJwndw4gUOBkcvlVoLsLitsdNnQcZwVhU3DL4dfep4nGTVU4DqVkp8DILOdkmPn9XPyY77x59CpxKzg6CrOOvf6h7CujEo2pfRoGIbxsUj6Xi76Gk2yVVxX6TlKh57S4XCIfr8Pz/PkHuC6LsrlMqIokspKsmX8XdfspDhjSGGv15PnZWML73NcpG/i/eDGCBwG2tFJzhlULLHp8QM6BA+ATHLNZDKyf5nP5+VjxmYHQSBihpUZHRjI7aLRaCRihAcfqzdxHK+o8WT3kzaF6S4q3ZUFYGVW1HWKG+4H83eQz+dX/j+OY3nd9BCZwdkwDGOVi/JxkoKHjSf9fh/tdhsnJyeYTqc4OTlBFEU4Pz+X+wwFBRfKmUwG+XxeJn9zUby7uyuV/X6/L8/zLvQCm/ciDu3s9XqoVqvI5/MYj8d4/vw5BoMBXrx4gdFo9EaA7iYInhslcKgotcCh1yWO4zeqJhRFHImQy+VQLpdlqivd6rVaTao2hUJB9iQnkwn6/T76/f6KmImiSGZecQQ9Fbie/6QPiOTBoCtRumxJ+LMl04k/lEwmI1tilUoFlUpl5YDv9XryszNRGYAJHOOd3MYcDMO4DtZtWxEOQk6n02g2mzg9PcVoNMKzZ8/Q7/fRaDRkJ4L3Au4MpNNp5PN5BEGAdDqNYrGIdDqNnZ0dRFGEfr8vX3vZhTBb1fn1HAeRTqdl8RtFEX788UcMBgN0u11EUbSRVZwbI3AIKzZ0fOtJ4dx2YkWE+TKe50m1hv4demv4NQBkGjdb9yhgBoPBypTY0WgkLX5RFK1k1ly28sIKCYP8dOYBDc1sJczn84jjWEQP378PXBHk83lUKhXU6/WVdEwaqSni9O+cHxt3G53nsW6L0ybSG3edZAAsF7yDwUAEztnZmUwh15PHk4GA3Lqiz3M4HCKOY0m5v8r2VPI9hRkXsMPhEI1GA+PxeCVkdlPP5RsjcPR2lP4D85fP6k61WsXW1hZyuRxKpZKIHb1FpTuh+Jij0QgnJyfodDoSYz2bzTAej1dybZLeGC2ykl6gi4L5lsufpp+/fPkScRzj4OAAg8FAqlOu66Jer+PBgwcIggCtVgu+76PVaomw4zbZVcnlcqjX66hUKvjqq6/w5MkT5HI55PN5pFIpvHjxAi9evECn08FyuZQWxuSgTuPuQiM+S+l6j17nQPG8MIxNJdkppd/06BymDn/77bfodDpyvxmPxxiNRjKriotNbh1REI3HY3Q6HXQ6Hbx8+RInJydoNBoYDAaXyn3T22ksCCQFz3w+x9HREVqtljTu6Cnlm8iNETjAm/OmdGcRV5WsfLiui1KpBNd1RRWzqqPD+vgYrNa0223J16F6juNYtmv08+qKCl+ffq1vgxUcBjyxKsX2cw7ijOMYhUJBWgmpqBkgqJ/rsuVJ7umWy2Vsb2/Lll06nRZX/2KxgOd5MlH2XemYxt2AlVK9cOCKk4uF68yhMoybjhY5wJuJwskKDu8lrVZL7inJKBHuJFDkcFBnp9NBr9dDFEVy33ifc23dwhv4KSiX2XJ3gRsjcLQaHo1GSKfTK+ZXHmCdTgeLxQK5XA5RFMF1Xdm20gchv5cVIR58vLlzu4kGZl68KbJ0lUZz2QNNh/n1+330ej0RNblcDtvb22Igy+fz6Pf7ouCjKMLx8TGiKJLSJVsGmXq8bmaVDhgMwxBbW1vY29uD67qyt9tqtVY6w5I+Irtp3S24MODWZjabRbVaFTNivV6H67orxzL37lnJMYxNR1dG9D2BIieOY4k1YfOGjjbh99PH+fz5c6RSKYRhiKOjI+TzebkeN5tNnJ+fo9frYTweX/qavG6B+q5F+aYvaG+MwNHCQl80k+qZhijHceQ9O6YAyGPQY6IPOH0zX2fuvU73OI1n9OH0ej0ArzNydJz39va2uPD59s0334hBrdFoSMWJ5c51Aocr7iAIUCqVUKvVsLe3B8/zUC6Xkclk8OzZszeCqD7lhHbjZqHN8KyI3rt3DwcHBygWi/jqq68QhiHa7TaazSYajQY6nY5cjGlSN4xNZt29gYsDWhIAIIqiC9vLubCeTqd48eIFWq0Wcrkc/vKXv8h1O5/PYzAY4Pz8HP1+XypAV32dF/37XV+/adwYgQNcPNFV/5tdVQBkBalNshQ42lujqx66IrSudZvvk36bq/4cfA3ZbBbdbhfn5+eYTqeo1WryczBBslAoSCWK2TxcLTNtcjgcisBhZYdikOa1TCYjHWX5fF7SMNnBxS05fj9XHSZuNod1gV3JiHedusokb9/3sbe3hyAIcP/+fdy/f19i5PP5vIReAkC5XJbjhhdg8+IYdxl9bumP9TaW/n/6P+fzuSwyOH1cZ9V8yHXZruk3TOAAF/9R+HmKFSphXqyTczve5qNJmpEdx1n5Om0y1mXGy75+iptGo4F+v48//elPWC6XqNfrSKVSqNVq0sLt+z6KxaK8ZvpvHj58iPF4jGaziWaziX6/j6dPn0rWQqPRkG6w2WwmnqN6vY7Dw0PU63XZZkin05Kp0Gw28fLlS7RaLRkCt8ku+rsEq4PMd2KVT7eZMoGb3YbFYhHFYhG1Wg1/+7d/i62tLTx58gRffvnlykRjiuKTkxP0ej2cnJzg+++/F5/AVVeahnHb4AL8bf8PvBY27ERcJ1S4AKf3Rsee8P7xvvlkdi1/zY0TOO+CQiSplJNpjLr6su4PrjutdLcV91kBrNwk1vlT1j0uP6f3WzudDs7Pz8VDxHKkbn3XrYMciDaZTCQssN/vS24PKz/MBppMJhITXiqVZMgnKzjA6yGldPVzq4tCzrj96IukbhHVw2Ip8Hnch2GIUqmEarWK3d1d7O7u4vDwEIeHh+Ijm8/n0ok3mUxQrVYxHo/h+z6y2ew7L/yGsSlcdKxfVVQkdxMuejwTKx/GrRM4RAuYZBmQ/3/RwUGTr+/7KJVKKBaLKwKH6pr+HZ2krAMBdUeJ9rHwdbHM2Gg0ZKuK+TT0OXieh2q1Ctd1xUtE89lisYDjOAjDEHEco1wuYzweS1gf293n87lsIYRhiC+++AKFQkEMoszzoR+IOT8XhRQatwuKGp0DxWwoHrP6uA7DEI8fP0ahUMAXX3yBR48ewfd97O7uiuH93//936X9dbFYrBybnJZ8dHQkq1TDuGsk70F8z8rnu+JE1n286abfT82tFTjA+6tcx3EkK6Zer2N7e1u2ppJJxv1+X7bEGLxE0aN9PqzW6MA8Vk2azaZsWS0WCxSLRTQaDbTbbRQKBcnDqdVqEkwYhiHS6TTCMEStVsN8Pse9e/cwm83kteiWem7TcUQDRza4rist671eb2XkRHLMhHE7obhxXReFQkFEc71el789O/YGgwEqlQp+/vOfo16v42/+5m/wN3/zN1gsFlIlPDk5wQ8//LDirdnd3cXOzg5SqRR2dnZkcZDMBrFjydh09H0nuR2lz4O3BWImP5/0yWnsvHp/brXAeV843iEIAhSLRZTLZQCQNnHf90W4FItF8Rjwc8PhENPpVFInGcyX7EpihYT+IN5k5vM5PM+T6OzlcokgCER88Ealh4Um05y5DaF9FjSOMsNEd3KxBZ2jJ8x3s3nwQkuRy2PLdV0sFgvJWyqVStjb20OlUpEWcApxzq05OjqSC3g6nUahUHhjQUHPjzb+X+Y16scwjNtO0hqRTDnmxxeRnHNl58b1cScFDreodnd38eDBAxweHq6U2nXHFc3KOpGy2+0ijmN0Oh20221MJhN0Oh0RQTpEULelx3GM58+fI51O4+nTpyupxvl8HgcHBzg4OEChUMCjR49QKBTET6NTZekd0kNDaTImy+VPc6dYKfrzn/+MVqslKcYUZMZmwAsru/Cq1Sr29vakspPJZMRQ7Ps+9vf34XkexuMxfvjhBwwGAzx9+hSDwQDPnz/HixcvJFiTPrD79++L8X4ymSCbzYqXjLNsNEnRo/1AyTgIw7iNaM9l8nPJf79tEZBsLV93LpnwuTp3UuCwypHP5+H7PsIwlAqI9vPoMiPbs2n8jeNYvp5zpOI4lrZsGoXp3+H212g0Wjngc7kcxuOxrLKXyyXK5bKstheLxYqo4XPyZkGBk+wG442IGTztdhutVkuiv62CszkkszmYRMxZbGEYwnEcbG1tyTZorVaD4zg4OTmRcR0MFzs9PcXp6enKIFtWLXXKuBbWF21R6Qs3j1lmML2tCcAwbhMXeWrWfU3yPFm3ELBz4nq4cwKHF9fz83MAkK0eDujkzCuKCT2TJ5fLYTabwfM8TKdTFItFVKtVTKdT9Hq9lXELzKyhByKZRsxVMAXSdDrFy5cvxYh8dnaGIAiknTyXy60MEs3n82JA5irdcZyVSenHx8cyf+tPf/oTOp0OXr16JYNGN3kGyV1DD+6j4KB4Pzw8RD6fBwDZYm21WpjNZjg6OsLR0REGg4FUcprNJjqdjow6GY/HePbsGf7t3/5NKpEUya7rYjabIZfLibjW2R4cHUJfGEercCSJnm31riBOw7ipsIL6rmP1bZ23ycX1277HuBx3SuDQTzCbzdBoNGRA2mKxgOu6KJfL4n+hE56VEV7AaRJmNYZbPRQzFDn0v8xmM/R6PYnd5gA2Gj51hgg7oxzHwY8//gjP81CpVFCtVuF5Hmq1moxd4GvlipxdM6PRSMY8cKuB8fq9Xk8qOrZFsDnwgkiBwyqJ7/uoVCp4+PAhisUizs/P0Wg0MBwO8ezZMwwGA7x8+RIvX75Ev9/H8+fPpUrJSiVnlj19+hQApCJEX5geIKu7RihwstksKpUKyuWyVJUymQyGw6GInE6nI+eNngeXzLAyjJvMVbLSyLoZgCbqr4+NFzg6tZUrySAIEASBtNEOBgPxy2QyGQwGA3S7XWm75gRYHXwGQEr2PBi59ZXNZmVeFodsZrNZWamOx+OVNt4oimRLiVtbrCZRkE0mE0m45L8dx8FwOEQ2m0Uul4PjOJhMJjg/Pxcx1Ww2EUWRTGjXycV2Em0G/Pv7vo9yuYwwDFEsFmW21Gg0kgwmTig+PT2VUSAcfxLHsRxfFBncfuWg2lwuh+l0KrPifN8HAPi+j1QqJYJfZ/Jks1l4nifVRwqwTCYjVZzpdCpeH13VMUO8cRew4/vjsPECh6mtuVwOh4eHODg4WOlSGg6HePnypVxkdbaBHm/PrituC3Ebi0IkCAJks1nZPuJzAxDjMVerk8lkZdtq3WRzPUV2ufxp1slgMADwOpFWr3C1Z4jfyyoRt8aSwzWN2w8zk8rlMiqVCv7qr/4K5XIZjx49ws7ODpbLJV68eIH5fI5nz57hxYsX6PV6+P7772X4LCuJg8FgJTqAxwvFTK/XE9FPsbK9vS3bYuPxGOfn51JF4nav53nSvs7p9sxjms/nqNVqclyyknN2dobJZIJer4d+v79yXhrGJpD03tg1+frZeIFDAy5FTb1elxVlOp2WsfRs++ZFPdlhVCwWRShxe4hhfLlcTqaZAxBRRC8EP55OpwB+qvyw24VeHXoduILWK1iKHj2mgqMXuA3GGwQAqTTpQD8+hlVuNgd2/rF6UygUUK1WUavVJMk6jmOJH2i1Wjg/P0e325UxIsmZbcluEG7jslLIFG2eQ/TWsNLZ6XTeyMbRYodVGh2cyUgDfm40GomY5zlkGJuKXY8/HhsvcLjdw/I9u0eCIFjpptIGYH3R5+qWIXmZTAbNZlPmWXEcArcD+Dk974pbZDoMSpfv6WmgMGGJnlsE3D7jile/Pr5m5txoPxC3HN4VOmXcHigWXNcVb9be3h52d3dRLpdlCngul8NwOBTzcK/Xw6tXr/Dq1SuMRiP0ej2JPeAx8rZjgwLE8zzs7OygUChIVxZn6rDTcDAYiMdrsViI6Mnlcuj1emKG51YWK508VzKZzMoMNR7PVsUxNgm7Fn98Nl7gMMtDz9xhNYcixPd9GVxJrwuHC9IfQ1Mk8Lq0SJGSfNNjExifH4YhcrkcSqWSzIwqFosrRmHG4SfFDKs1URStCBxm8YzHY2QymZXP83u0R8i4/ejE4gcPHqBYLOLg4ACHh4cIgkDybTg1vt1u4+nTp2g2mzg5OcH5+bnMNGNF8DJ5SNwOzWaz2NnZQa1WE4HDyfb0roVhKBEFs9lMMqJ0CCXFvd7eZVWVfrlsNosoimTcCCughmEYl2HjBQ5Jrvx4IdUt4Wx75U0kn89LN5QWO/qGkEyq1FtcXJ3qmwPNlPQaeJ4HAHJT8n1/ZWwEV67camMViK+Rz0lxw3Rlvia+DhM5t5d0Oi3HaBiGYiLe399HqVTC/v4+dnd3pa17OByKd6XT6Uj1UR+THJJ5mYpIsuqoBT4fi4Z53/fluAReH4MUOEw9Zl4UqzfZbHalVZzbthxJYsevYRhXZeMFDi/i2pgLQC6yLIUvl0uUSqWVVldmhugtqul0KsZdely0f4efA4B+vy/PQ3HC8Qs0I+fzeezs7CCfz6NUKqFcLq+MY9BGZ53JQy8EtwG2t7exv7+PVqsFACtbVnqkg3H7cF0Xe3t7kkDMMQu//vWvZZ7azs4OOp0O/uVf/gWtVgvPnz/Hy5cvMR6P0Wq1RCh4nifHBI3F6yojyfh4LgS0MOK2Uzqdxvb2tqR8N5tNDAYDdDodESu6+pkUTHwM13UljoECn740EzmGYVyVjRc4AFb8AEmvAQUE8FPGB/A6L2c+nyOfz4v3gGF+rMRQeLBdnKF9fE7djcJQP66AmV5MkyYNx4vFQrqxGC7IeVYUR8y94Y0CgPhtuIoej8fy+ngzshvE7YJ/X3buFQoF1Go17O7uolKpyPtarYZKpSIhkxQZJycnK54sPc6DxxIrKOueV3/M80SPWuD/a78Zt1l5vPNrk5UiLXTYqcXEb53GTaFjAscwjKuy8QJHJ/tyFak9NfTa6C0gZtRwtg8AaYudz+eoVCorbd06t4NCQ8+iogDS3VF6hEOj0UAqlUKj0ZBp4hQ42shcKpWki6tSqSCbzaJYLEpliKMn+D16WKdxO6BI4N+bXq2vv/4apVIJjx8/xuPHj8Uon8lk8OLFC/zhD39As9nEH//4RzQaDRwfH+Ps7AzA6+1YTgDnMUfvls530ltYTB7WrymVSmE4HCKVSqFarYqw5zFXLpelFbzVaon5eDQarfyc3HZNpVKYTqdIpVJyrgBYiUHQUQiGYRiX5U4IHHaMUMgwXI8DKXu9nggbx3FQKpUkhZUrU9/3Vy66wGqVRr/nylNPHWeYYLPZlCRjZtM0Go0V744WOFwRM8mYXTOs1NCgSYHDGw1vTNx+M24HmUxGQigPDw9xeHiIarUq21E///nP8fXXX8vU7+FwiBcvXuB3v/sdOp0OvvnmG5k91m63xeROE3u9Xpc2bB5rFDgUHIw9cBwHhUJBKj/82uFwKOfQdDoV8Z9KpWTLLJPJoFQqyfkSRdEbPyuPS55P68S4CRvDMN6XjRc4wGshwkoHADEx8mbAzo1MJiMVFr1VdFElZF0pn6tgZuTw86zs8KZAU2gURStGYW4n6RUt8FNgIBOPue2m28uTgX/G7YFVFtd1USqV4Ps+dnd3cXh4iHK5jHq9jlKpBABotVqIokhGLBwfH8ugzH6/L+MWKLp1JlPyONbHL6uHxWJROgDpj9HbvLoCye0l/gz8Hh73+u1D5vQYhmFclY0XOLqTiKvZ8XiMV69eYTAY4NWrVzg+PkYqlYLrukilUigUCigWi7INwHlUrIrwPYPL1n1uuVxKorG+GdCMnEyQnUwm6HQ6aLfb8jl2n2iTpud5MuDT87yV8DVmmrCl1kr7twfXdeF5HkqlEr7++mtUq1X89V//Nf76r/8avu9jb28Pruvi+fPn+N//+3+j0+ng3//939HpdPDy5Uu8ePFixVBMMcIkYQ5nTcJjw3EcVCoVeJ6Hx48f49GjR5jNZmi324jjWCaOM/xvuVxK5k0+n0cYhlIF5TaYbglnJfFdwtuOVcMwrouNFzjA68oGKzgUArxAdzodAJBJx/TGOI4jMfTcJmJVhxWf5XK51rzJFTH/nxUZVogohmazGRzHEXMlRQ0TZNk2y9fE79c3Cq6gubrWvgXjdqBnNhWLRVQqFWxtbUmHXbFYRDabla2pZrOJFy9eoN1u4/T0dGUECAW9rs7o7cp1xwWrRxzQubu7K2GTmUxGtnR5LLIiyZlorBbp5+JzG4ZhfA42XuCwukJPAXNsOIeH7dTA6xEL/X5/JaGY3UwUOHzPx133Od0SzufW21j8muVyKSMfyuUydnd3JcBPT1jWHqFarYZisShCa7FYII5jDAYDieVn+7p1n9wOtLk3CAKEYYjpdIqzszPM53N0u13EcYzvvvsOP/zwA7rdLn744Qf0+33JPtLCV7dhU2zEcYx+vy8eHAZH0sxer9dRLpdx7949HBwcyLw0ADLtnnlRjuOgXC6j1WrJENhsNouTkxOcnJyg1+vJEM/LJCUbhmFcNxstcHjToPmWU76n0ymiKJJtol6vJx4dYDWhOJnRoVti6TXQ/h1+To9vYDgb2321KVj7I7iNxSGDfJ1cPdPfs7W1JSnIFEla4NDLw9k+dnO5+ehWbN/3ReCcn5+j0+ng97//PVqtFs7OznB6eoooinByciLdScm/r66ksIoyHo9lcCuFsBY429vbqNVq2N/fx71799Dv93F+fi7VQM6HYlWz1WrJ0Np+v490Oo2XL1/i6OgIo9EI3W5XhLYdg4ZhfGo2VuDomT2FQkHMwuym0uMMdAcUu6x4QZ5OpytbQHqLiavWbDaLyWQigshxnJVOKHpiHMeRPBK2cfPmohOKGQTI59ZZJbpll4/NrBOKIbanmwfn9kBDOX00bK+mAD87O0O320Wn05GOvHeN4eCWJbOZuL1E0TGfz6XS6Ps+arUatre3UalUEASBGJV5jPG8oGAZDofodrtSXQQggZh6cKxVEQ3D+BxsrMDxPA9BEKBQKODRo0cIwxCu6+Lo6Ai9Xg+NRgOdTkfm6ACvV8HaN8CxDDqmXr+/6HNaiIRhiCAI4LouKpWKTH/O5/OSa5PL5eTr+Pr5/Mk22iAIEASBDAPlVsarV68kmp83Qeum+vxcpoOIomMymeC3v/2tjDzI5XIYjUY4OzuTrUces2+bzaSDLVut1srAVz3XrFKpoFKp4PDwEH/zN3+D+/fvY2dnB/V6HQDEYKy3mujBOT4+lrRuEkWRbJfxOd51DOocHvp49L9NHBmG8T5srMBhlUQP2kyn0xLuR4MkxYO+sOoLqhY/2t/A9/pj/Z4VnkwmIwZhdmm5rivChZk1/Jiw48VxHHk9FFGs/nA1rWdlcUI0V87Jn8f4PLxL5DDKYDQaodlsot/vS5QAO+x0h9xloCmYHi4er/ShsSKZy+UQBAG2trawvb2NYrEofi+a3pPbnXytjCrg5/V5pY/Bt/1edBv520zJdhwbhnEVNlLgUASUSiXxv3DkQrvdRrfbFZMlV6P0LHCbiFtA9NZcdOHlTYQJxfoCT98M/5/TkWlKplBhEFuhUJBgtWTQXy6Xk04a3qSm0yk6nQ6GwyFarZZ8HMfxSpu48WnRxxKrenpw6tuqEpx/FsfxysgQHl8XzRRbJ85JsqtKb8cyQLJWq2Fvbw97e3uYTqfodrtot9totVpotVoSW8DvA7BSRdLDZi+zNaorSkwQ5zmnf0d6e4wVJMMwjMuwkQIHgLTbFgoFSYZlx1S328VwOFwROPpCy5ZwvX3Er9FVEV7IddcTV67A6xtLHMdyQadngUZjenDYPRMEgTyv67qo1WqoVqvwfR+FQkFunBRT3W5XbkbccktWcIxPC0WyNqvruUpvE5469fey2zNJ8aJ526BVer2q1Sq2trZE4BwfH6PRaIjAaTabksqtH5/nzrtE27rn5fmmj39uyenhuOPxWCpIervWMAzjXWyswNGChVtEuntEVzi0wGH3E/0zvFEBkE4rXfHJZrOYz+coFovibWDiMB9fj3TQq9vka+D/83Vz+8r3fcnSAVZXy5PJRHJz+HjWOfX5SA6m5DHEY4fC9G3iM/l3XvccySoRv17//d/1OnUlMIoiHB8fI51O4+TkBGdnZzg7O8NgMJCOvIte61XFDQUguw15fDOUkD4f7Ynj1hxfs4l3wzDexcYKHJp2HccRUXN2doaTkxMROtwGYOgeW6+LxSLK5TIAiAChSKIAyuVy2NraQq1Wk+0wVnrov6FvgsZf3kRGo5FUkrToocByHAftdluCBinS9vb2AEAu/pwc3el0ViL6rXPl88C/PeMA2J3E7VEawhkBcBFv+7vxOVjxo3jiccQRJBTY6+BgWX2cAsD/+l//C8ViUY7NdruNFy9eSGVwXXXoqscYBSDDCzmgk2nh9J7xZ2KllecBux+Z7G0YhnERGylwdDcTAEkF1i3irHZwFay3jVg50TcOPfuJX++6LsIwhO/72NnZEV9NLpfDZDJBo9FAHMfI5/NwXRf9fl8ydziriu3BXEnzBsmKEkv0OiyNK1hWA/j/l/U/GB8HfWzoikQQBCvmYPqo3udvpLv06OHi8UlD8GUGrOoKzng8lrElrVZL8qEomrlF9KHo85LnGcdIMFiTkQjJY52i7F3dY4ZhGGQjBc5y+dOU8KOjoxWTJyspej+f20MApMrCyoj2L+hRCb1eD9lsFs1mE8+ePZOqD+dQ5XI5eR2s/nA7iZ6DXC4H13XFRMmLOTtJGJEPYMV8ye4VVoE4YJEr93dloxgfDx5rOvk6CAIUi0X5O2ezWYzHY7lZX8YIroMkfd+H7/vI5XIy9Z6PMx6PcXZ2JkGPFz02j8nFYiHzpXq9HkajERzHkeOLlU4K8PeFlRlOSs9msygWi2Ka5+f06+PvkzPe2EbP6AOGFtqxbhjGRWyswOn3+yv/Xi6XcuHWe/wUDXrGDiPp9aTuZAVHd1tpUykrONooTE8PK0ZcuVIIsdz+NoGj51kxJZYDEHu93kp4m130Pz36uGDgI9uvS6WSbKvokRuX9ctQGPCY4pYOjy8+zmAwwGg0EhP627Zx2JHV7/fldTUaDaRSKekETHrG3hemdjMHir4ypnpTvFEE6uqW53nS7cVFwng8lkqoHeuGYVzERgocALJ9A6xWUpIXxWTmjTZNJgVOMgdHix5emLnlpJON6RtghwyTh3ULLNHbX+zmoueHPwO3FOhPSM6dsov+p0f/znl86AA74CehoscnXBQ9oLOUdBo3J3VT4LCrTidlU1jxjcfoRRUYXUWiiTe51XnVY0ovAFh9YtWJxzIrM9wmA16b5pPPxUUEzwH+nDzH7Xg3DGMdGytw1vkGLloxJy/i+sajQ8iS37cu5E+3BjebTfFj0AzK10DvjK4m8fs8z0O1WkWxWMTu7i52dnbkZsUguGfPnkmEP5OLzVx8c9Aihn9bVm2SkQMafQwVCgXxeN2/f18mfddqtZVuPm5TcmAr/SzZbBbT6fStxwaPRVZugKt3RiV/bm3aZ74TTdfcjtLdiRT/DOHUv4swDFEoFEQo5XI56erS8QyGYRhJNlbg6Iv5OsGiSbblrhMyV22DZWurHuOgX5fuctE3Ot4gWLlhF47eoqI3gt1hNE2buPm8XOR3Ibp9XB8nSbQRN5/PIwxDFItFhGGIcrmMUqkkj81uOl3tWSwW8DxPBEtSbK17fdd13GgfEs36PI75+nQVi1tl2jCvfw96y5WiiAZl7Z8zDMNIsrECh9CYmc1mV24OFBU0UzLDhitD3hyAq138190w1gWxrRMjvAHm83ns7u6iVqthd3cX9XpdtrcmkwmGwyHa7bZ0udjE5puBFq4cxLpcLsVAq8dv5HI5CYHUMLYgk8mgVCphf38fpVIJjx8/Xknl5hyy2WwG13WlMjKZTJDP5wG8DuLj130sdNWJW6s6OoEVnGT2Dn1v/X5fzj9WcCgC6WPKZDLyGNqwPRgMRPhbCKBhGJqNFjj0AXDQ5v379/Hw4UMArz01nU4H7XYb4/EYjUZDhI0WOFeFIuOqF1wKHM/zsLOzg+3tbezs7GBrawtxHCOKIkynUxE49OBw+KKFn31ekoZ1+kNoFqfAoREdwEp1j74vVmO4RVmr1fDo0SOUy2UREqzaAZBwPM4nY2QAxVMyEPBjoEebMDmcOTdBEMD3fSwWC9mOoi+NOTwMG6RJmo8XhqF0L9LHQ89RNptFp9NZeVzDMAyy0QIHeJ1o7HkeCoXCyk0CgLTxxnGMcrksFZLBYCA3Kq4Q+TGrJTRj6uGF71tF0Tc2GkTpOeANks/DFeyHPqdx/STzW7TvSm9Xag9NEl3F4dcwV4fbP8xS4raTjiJgqvWn8mTprTdtjmdwJYCVyhY7AVl5ZAWVr5leHXZdVatVabvXGUPc8uJ5+CGLEsMwNo+NFTh63k2pVEKhUMC9e/fw5MmTlfRULVIoXNjSy2GWnPk0GAxWEopp7nzfac/6tbIcz/bZWq2GSqUiWw96EjrfrDX8ZpEMYWSlQhtntYil6NHeGAbaOY4j2zk8NpLjQ1jNmM1mkszNaiTTrfXg14+Frq6Uy2VUq1XxDzG/aTweyzkzmUwkTFAPEuW547ouHj58iEqlgocPH+L+/fuYzWbodDqI4xi+70umVBRF0l2lf8+GYRgbK3CIXv0yfyOfz6NSqcg2QRKuJuM4RqPRkBsGL7BMGOYKmmnDye2Gi9rRtU9Hm03ZfUKPBocQ8kaXrN5YavHNhH8TnYCtjb46Y+miCg4rjHwsVnMokHR1Rydij8fjFeO5HpB53ceJzv3hccvqDV+njk7QOTZ8nayM6sTibDYr1VZWcGisZ4WH5zTbxT8kHdowjM1kYwUOV9OsvkynUzx79kyqJK1WSy7G7OzgCplmyUwmg62tLcxmM5RKJezs7KyYkdcZlBkmeJmvYxorDaCVSkW8N1wRx3GMVquFXq8nE565Mk9OdzZuBvybcisJwIqw0V1AHNbKN+0n4d88lUrh9PQUs9lMjln6V/h90+kUo9EIrVZLOuwoCK6zyqd/DgZY5nI5qdowr4diY7FYIIoimcPV7XYxmUxkBEQ2m0WtVpPZbvV6HUEQ4PDwUM6BTqcj6eKsWnKryvd9ZDIZOe9YETI/mmEYGy9weFEdjUay2vN9H91uV4LTSqXSSsoqU4a5LaAfL/kcenW8XC4lgG9d9YdG5vF4jHa7LWMjmK5cLpext7eHarUK3/fhOA4mk4mMZTg7O5PxDHqopnEz0McCO5hYlQCwkkzNTJxsNrvyd9Rt371eT7qtzs7OMJvNEAQBgiCQaoje2hmPx2g2m+j3+yvzy65zfIf2BxUKBamE8j0N1Fxc0DxMIzEHjdIrlM1mJfPpl7/8JX71q1/J4zPjh+cOJ5vrXCEKHG7t8nHtvDAMY2MFDrAqcgDIhXY8Hkv8+3A4RLfblZUiL5ra0Kgv6vyY2wV6+CFn53BlPp1OpVzPAZzMJymXy7KqZasrV698DKYWx3GMwWCAKIpEPHGLyrh5JNOxdd5RUuysIxkpwOMgiiI51rhtpUUFtzU5r+k6BTCPc8YY8L1eDLBLTPuP2M2lZ64xwC+XyyEMQxwcHMiWFFvrWYXp9Xpot9uI41iEvc7QoT+JmTv83ZovzTCMjRY4wE8X2+FwiHQ6LRO+2e1Bzws9DRwIyAs3DcqO4yAMQ5kHxITW5Oe0B4GpsrzRJI3MuuuFBmW2ztI82u/30el00O120el08OLFC2lrt9bwm0myqkdxQqGsU68v8o1o4zEFAit/nGulh23qXJjt7W24rivt09c1gJWBfa7rolarwXVdSVlmtAGHYlKEcyuVk8n5s+VyOTx58gRPnjxBuVzG119/jWKxKNWX8XiMk5MTjEYjnJyc4OzsTLbgWIHVz5tKpVAsFqU7i+ZlOz8M426z8QJHh4DR36B9BNqsqUclcFXMi3mxWBR/wXw+l8RYAOKNAF7PG+Lbu5hMJmg2m9IB0263kU6nZfU9Go3Q7/cl1I/ehcsOajQ+DzrQUc8pY3VBm4jXoc3H9JWwYsNKCY8xVjAorhl18LZW9KvA84TPwwonFwKs7PDn1VtmFC2stnBBwYyfSqWCvb09FAoFNJtN8dhwq7fb7aLdbstigOccn4+PyeoVvU/pdNrODcO442y8wAFWg/f0x8m4fD0bh+24o9FIVorsamKyqs77YDu37iJhKZ8XX/05ltS5Yo3jGN1uF2dnZyurTrai9/t99Ho9STM2cXNzYdjeYrFAv99fCXBksjA7ndZtM/LvT9/JbDbD8fExPM9Dt9uVsQ1M+QWAYrGIdDqNer2OfD4vlR++XfV4oeCn6T4MQ/GqMciPW2X02dCDxtdMvw1NyL7vY3d3F0EQ4P79+9jZ2RFB1u/3cXp6ilevXqHb7eLHH39Ev9+X6iWFIn8GtslzYcFZVfxdZLPZlWRkwzDuHndK4KwzCq9Dt/C22+03Kj3J6o/24BQKBRQKBbiuKwFlpVJJtrXYMcLPsR2dXSLJcrzOvOn1emIetdL7zYU3/fl8jn6/L5UbbqdQ0K7zUfEYZR4TjbXcguLwyXK5LNsynDaezWZRr9fh+75s1aTTaRmFwMd/F7qamcvlRNSzgklxxS0gmofjOMZoNMJwOJQqzHw+x/b2tuQ6/epXv0KlUpHXz1Tm8XiMs7MzPHv2DJ1OB9999514bjiUVosbChp6cvj7BSC/C54/hmHcTW6swElWV5LDKj929UIbQvVEaOB19UdvAej3i8VCDI80F49GI8nPcV0X3W4XQRDITXA6neL8/BydTkc6qxaLhdwM9Q3RKjc3Gy2kubXC7UpO7aYAokhYl5PEjBg+jv4/nZisP8fvoS/nKkZ0Xclkjg2rlrpaqc8DdkRxa4nHKUdH6IGh9PBQfDDgbzQaYTqdotFoSEAhHy+Z1s2tPnZgcYwDt9AWi4V0qHE7zLw4hnE3uZECh2JBJ7ry4subw8e8aCVvIiQpttZ9PBwO3zAtc7tLb2/xJqJvYszyYItt0qDMz9nF+majQ/5GoxEmk4lU/Pj/vGlTDKyLIOBxro+nMAxXAv606J3NZhgMBuj1ejKnjMfMu0LwtEinz0d70AqFAkql0spz0x82mUzQarXkudj+zgrm/v4+9vf3Zaq44zgYDAY4Pz9HFEV48eIFhsMhzs/PZR5cu92WNned8gxAQjXn87lUceiRy2azK0nkrH5ZNo5h3D1unMBZl+zLzwOvV7Ufc3Cg5qI04ovgqlbfMLSgGQwG4l3gBZfvtRlap7vyZmi+m9uDrsJQYCQneuu/6bq/qw4JpDFZVxL5/TrsT2fMsALyLnQicXImGg33rN7oAD9Wivh8nAXFSgu/n+nh9O0AkO23fr+Ps7MzDAYDqd7wMZPVJ13FZeWGTQPA6zEYjuNIJUebkQ3DuFvcSIHDMjhXfMDragpXtgBu5A1fCxde6LmVxRV4LpeTuHpdqtc3PH3j02/G7SE5kkNDofC2vyuFBxN7Pc9DtVrF9va2+E0YM0BD8osXL0Qs8Li6aOimXkywusjn4nGq4w84u41T7Xu9HrrdLhaLhaR/U1R4nod6vS6eHQYavnz5EvP5fCXu4PT0FKPRSPxmWtjpLjDdPs80Ywb/DYfDlQwqGqPZyfU+RmvDMG43N0rg6Lk27DgKwxAAVsrw4/H4M7/Si0nmoACQvBOaRJmTw4suy+nJxzFuP8lj4SroTibf9xEEAcrlMra3t1c6/k5OTnBycoIoinB6eioDPpmQvc6Ho6s2NBPTHF8ul1EqlaTbT8ceUOCMRiN0u130+32p1lBUsGJDgcOhmzrfptlsotFoYDabiQlaV5y0eZ8/q67g6G1dxjTQEO26roQJ8vVyXMv7/i0Mw7h93BiBoy9o9Ksk8yxu2wpM5+rwwl+pVFaMm2yj1aZQmij1FpVxN6EIocDg+cBjhIb04XC4kjtDb89F50vS58ZtKJ2KzGoJq0Bsf+fjA5DzlFWaMAwRhuHKNHT6kMbjMfr9vnR40R/DY54/b3IgKbd0k9vFuprLtOTlcikCh4sKblvpIbWGYWw+n13gcFXGi5AulXNVpmPfb0vAHUv2TJfd3d1FGIZ4+PAhgiAQv8R0OpUkY2besGWW21hxHAOwleddQwt+Co9UKiX5OTTiHh8f4/j4WPJkaE6+6Eauq6SFQgFhGK7MYmMFRm+f9vt9afum2OHUb8/zsLe3B8/zpPoTxzE6nQ6Gw6FsZXEuHHNzmJ2jTcTcXqJA4iJB+9K0ByiKopVEcnp+KNr4OwiCQLa0RqMRADufDGPT+ewCB1hdpeqOI27tAKsenNvgR9FmTd40CoWCdHtwy40VnMlkIjcd3hS0kfM2/MzG9aP9J6xksHrDifTcMqI5913GYj1bjUZgbiVxm0cb4RkaOBwOVx6HlR56dniMs5uJCcTD4VDiD5iRw2oT8LqaRDGTfLvIIMzfBc+j6XQqs6zoe+LPyesJH8/OJcPYfD6rwGGVg7Oh+DH38/W+/HA4lAvtuuyQm4w2S/JCq8c+1Go1zGYzmTs1GAzg+z6GwyFarZZMYh4OhyudVsZmQyGTTqfR6XRkG4aznljB6Xa7K+M7LkLf7FklDcNQjMQ6o4YpxAzbY5YPj122gVcqFfEIsavp6OgIURTJex7X/Hn4OvU5wffcmtJdk7pbTHeNaU8OBRgNzKze6KrvbbpmGIbx4XxWgcP0UT2kUqcCc4uGpWi9h34bLlT6Qk24FVcsFlGtVgG87rxiKX84HMLzPAwGA7kRcauA21UmcDYf7b/qdruIogiDwQCtVkvMuQzXS2bGJNHzpBzHge/7MjA2DMOV3KbhcIjhcCh5NOxYYjWyVqvB930cHBzg4OBAUpfZJaUFDv1B3I7SHVL0FLFrkuJfV2b0iAadGZQM4ozjWHKltNGY3Vt8vNtw3TAM43r4LAJH+21ouNWTlZOGRl21uE2rMGaXcAzDYrHA+fm5dLckZxTxxgMA5XJZtq7oWeC2RBRFdsG+Q3ALRreba5OxPhbeFeanO7N4DupOJZ5rXEzw6zn/yvd9bG9vIwgC+L4vX8/5UxySyaGwPH/1thGAlTEneiGQTGjWnrt1P6NuQGD+FMUeRdRt8+8ZhnE9pN5xQfwoV4JKpSJtqBQ4vKDq7SgKHF2evi03db1CZcgZ51N5nocHDx7g8PAQxWIRX3zxhSTUptNpTKdTSYhtNBpoNBro9/v4/vvv0e/3cXJygtPT0zcGEG4ay+XyxqWzfaxz4h3PKQsA3UmVrGi8K0+H28Ce58mgS1ZwgNdditpQTL788ks8evQIYRji4OAAnufh/PwcZ2dnGA6HK23q5+fnMjyWhudkJtS67ShtIKYoooB6m/8u6bXhdpkWS5ztxmvMbcXOCcNY5W3nxCet4PBixkoFW6jZJcGLEc2SNA7qFd1tQb9mGkAZF5/L5STEcD6fY39/X24+LNcn23Mdx8H5+TlSqRR6vZ7M27ltVS3j6lDMMNFYCwL+/2X+/rqCQzGgBROPJYoKLUbCMES9XkcQBKjVavA8T7r/2P7NSeKsNrLCpEWMNg8z1Vijt6aSMQm62ypZxdFfQ99SsuPqtiyODMO4Hj6awEmWouk7objJ5/MrpXeWtZnrobelbuuFKRn6x4twNpvFq1evMBqNEIYhOp0OCoWCTFmmkZMepXq9jkKhgHw+j/F4jP39fRwcHGAwGIjfgR01+nmMzUEfR3o7R79/1/cmb/hcSNC7whRhx3EkpG93dxf5fB5bW1vwfR+pVApHR0dYLBZ49uwZnj9/LgnKPId5/gKvz39Wa1iJYkaOfl26WpNM9mYYoc4E0mGZWghxVpWu+OjHMwzjbvBRBE4yAp4ThVkW5747U4knkwmGw6GUxbktdRtFjYavX88K4uBFlvM9z8PR0RHy+Tz29/dx7949hGGI+/fvi88hn88DAA4ODrBYLLC/v4+joyO0220AQLvdFjPobRaExtu5rKC56Hv1lg2rHDwXuS06m82wvb2NWq2GcrmMX/7ylyiVSvIYo9FIRPWLFy9E2Lx69WqtkVhfA7TA0dVJvteBf3yvqzbs8mJ7Orea2GmpfTbrfnbDMO4W1yZwtKjRFzIGlLmuK1/LwC62fesU302/QWvRQ1MkTcPdbhf5fB5xHKNQKKwk0uoVbD6fR7lcBgBsb28jl8tJuZ8VMD2sc1N/l8bV0CnBDMTj1iiHUi6XSxSLRQRBIK3WPEdpdm+1WhgOh9Lxp483/Tx6S4yw+0mf6+uGybJKxa4vx3FQKBRWhu+ympWsaBmGYQDXIHB0xoueWZPP51dm6PAixNZvtn/TUJz022wqFG8Udfy52Z7baDTg+z46nQ7CMMTW1pZUvur1uswKKpfLiKIIpVIJw+EQz58/x/Pnz1dW2Pw927gHg+cnq6l7e3soFAor855o+Pd9X7ajWB3s9Xro9XoYDod4+fKlzHfq9XpSdQVWPTYM1tPVI57jye2opM+G15RqtYp6vY5cLodarQbHcdDpdNDpdOTrLsIEj2HcbT5I4CRXatq4qAffMdCOqzxmd7CCk5xHcxfQ2wUMPaMxcjKZIAxDTCYTSYqdz+crAxD5+51MJgiCAFEUYTgcot/vo91uy+9b/33sgn83SVZT9EgDfd4ywE9vAbE6Q1ERRRHa7faK3ya5LZUclKkriUnz8LoGAm2Edl1XRkmwgjMej2UhBeCNqo9+HDvmDePu8t4CR1/AXNeVLRR2AXFuzmw2kwwYZmLEcSzJq7cpuO9jwQs8qy1cDbuui263i/PzcwRBgF6vJxOly+XySuvvvXv3kMvl0Ov1kEql0O/3cXp6KsKRv2/j7qK79zg2ZGtrC7VaTb5msVig1Wrh7OxM5knxPX1eTNbWeTN6dhTFlG4U4DmebG0HXgsUhn0yTNB1XXm9ANDr9bBcLtHpdDAYDCRnh0Ke15x1uTr6GpPsQtMLgKt2phmGcXP5IIHDag23o1hZ4P8DkGGRenCfnq9kHpHXF1N6HZj+mk6n0Wg0EAQBwjBEFEUoFArY29vDbDZb2bbKZrMolUoyw6rT6UjkPlt2jbuJjmfgsRQEAQqFAu7fv48vv/xSPGCTyQSdTgdnZ2dihB+Px+h0Oivz0bQA0AJHiwVWbNd1RyW3olOpFDzPQ6lUQhiGePz4sVQx+cbRFL1eD4PBQLZ35/M5HMcRgaNDQwHIooH+Hy1m6G/j55LdW4Zh3F7eW+DokjffNHo7ihchfcG7KLTrrsOLK39HbOPlPCI9lJNx+5z8rOP06adgKT9ZvjfuDqyysJKnfXA8NxeLhWxV8Vzm8cevS4ZKJlOI+T0XbUUlz3k9iyoZiElRwtfLFG9Wf/m6gddBiDTb5/N5aSMHsPaao8WMFjjsRFz3PcmfxTCMm821CBydSsp2TXprkmnE6y50xk/o34nuNtFBao7j4NWrV6hUKigWi3jy5AlKpZK8ZbNZFItFZDIZyc4B3vQnGHcLtnjzRn1ycoLJZIJ6vY5ut7tyDrPrUQ/K5EIlmaqsp44nhYEOokxuR1G053I5GUtSLBalO7Df74upmQKHA0C5NUVojC4Wi5Ldk8vl5DXoChOFnRY4ejsrmXisK88UV5ueIG4Ym8K1mYx1twQrNxd1R9mF4d0kV4682PKGQhG5tbUFAGLCXC6X0i3DlfHbOk2MuwGPGV0VoUmYXhp66VjJ4YJl3QRwXblNtnqvMxSvW9DweVi14fv5fC7mZr5Obt3ymkLfDY9xHvOu6yIIAqlqsrLJ/K0oiuA4zooQ48/E80tnVjFrh+cfc6wMw7j5vLfA4U0XAEaj0YrA0abWZPCWiZurkfQDcLAhL7ZPnz5FoVBYMYG2223EcYzz83MZemi/97uL3qKiUODQ12KxKB1Uh4eHMp/q3r17cBwH5XJZFinT6RTA660lLV507lLSx7JuW4qGYj4ft1RZneF27Gg0kqoKqyfaMKyvOQwvbLVaK2KNrzWdTss1S2/D6Y4segj5u+Icq9lshmazuRIqaOeUYdxsPkjg6C2ndDq9Et6lg7/sQvBh6EqO3rZiciznArXbbZnoPJ1O0Wg0ZPyF5eDcbfSWzGg0QrPZxGAwQC6Xw3w+R61WQ6VSESPy/v6+TBHXnp2khyXZJaXjHpLnve6yYvu3Fjj0+4zHY7TbbWlQoEDXokb/XHx+ChLtRXNdV6qeuv1cPxaDBHV2jx4fQYHDbTK9rWUYxs3lg7aoLipJX3QxMj4cfXPhNtVyucRgMJDyup7GrjtYDIPnKT0sURTJ8NZer4dsNrvitSkUChJfoD08STP8Zcy3OkywUCjI3DWmE1O4MwQ0mWy+7rF1ZWYymSCbzWI0GiGTych2Vjqdlq0lbSim6NcCh2nh+jzjFhm9ONZhZRi3gw+q4GhDLPDmcEnj46Avsp1OB5lMBv1+H2dnZ+J/0qZIfTE37i48P6fTKXq9nnhoOIeqXC6j0+kgCAIEQQDP8/DgwQOUy2V8//33aLVab7R8J4dYvk3kOI6Dra0tBEGA3d1dbG9vY7FYiHl4OByi1WrJSAhuBV107OpFFSeY63wuLWjYQk509xe3qLTA4Rsronw/HA7tfDKMW8IHV3AA2FbUZyDpO+AKVm8P3rV0aONyUBSwq2g4HMJxHOnSY1dSJpNBGIYAXntT9PevawFfx7ocnjAMUSqVEMcxBoPBinFeJ5u/65qivWl6G0rn4KTTaZkwzu/RX6cFjZ5YzvNqOByuVHAMw7gdXMuwTRM2nw/dipvcLrBSuvEu4jhGv98HALx8+RL9fh/pdBqVSgWZTAbVahVBEOD8/ByNRgPj8Rjn5+crYX/rxI3ejqpWq6jVaivBlADEZ9NsNjEajdDv92V7iiL9ssev9qglv0cLOk1y1EwqlUIcxyvVHR0SSIOyYRi3g2ubJm58eljFAVZzbkzUGJdhuVyudCYdHR2h1+uhUqlguVwil8uhUqlgsVhge3sbjUYD3W4XzWbzndUVBlG6rosHDx7g0aNHYizOZDI4Pj7GyckJoigS4cTJ5MlOqav8PFf5nnVGYV3lsW0ow7jdmMDZEEzUGO8DxQS3h2jS5XYVU4HDMESlUgEACc0D3jzu2LnEak0+n5dWdHZwAcBwOJQQP3rFdNXmUxzPF5mWbUinYWwGJnAM4w7DLZhUKiVzy46OjpDP51EqlfD48WMEQYCDgwN4nofj42P8+OOPksWkPSmpVAr1eh337t2TlO1isYjRaITxeCzt6dyWYtYNh3dqT8/n4lOJK8MwPj4mcAzjDqNNuhQ6URRJBQd4vd1UKpUkO0fPegJeZ9xwYGa5XMb29jYKhQIajYaE9Q2HQ0RRJG/JxGDDMIzrwgSOYRiS9zKZTNBoNOC6LuI4xu7urmxfcexBuVwWczKzc4rFInK5HGq1msxCOzo6ki0vVnEoZiaTydoBnoZhGNeFCRzDMCRLhqMO2Fp9eHgoAoRVmnK5LH6Z4XAI13VRq9WQz+dRrVZRKBQwm81wfHwsoYGpVGqlWqODKE3gGIbxMTCBYxiGwDwaemZarRYWi4WE4Q0GA8lb0nEE0+l0xaCsW71zuRwcx5EJ95PJBFEUIZv96fJjYw8Mw/gYmMAxDEPQnU7MhPE8D77vI5/PYzQaodPpiH+GgqjT6WA4HCKOY3Q6HWSzWYRhCMdx4Ps+yuWyPD6ndnMQLA3GVskxDOM6MYFjGMYKHFyZSqXQbrfhui7G4zGCIFhp6WYHFSs4rPQsl0u4rosgCGSAZT6fly2u5XKJfD6PbDaL2Wy2EqxnIscwjOvCBI5hGCtwjtl8Ppehm+yq0tO1GRIIQOafMaU4l8uhWq2iWCxiZ2dH5k4Nh0NMJhPk83nk83kZm2Dt2YZhXDcmcAzDWIH5Nuxy0hUWYHWifTJNmx9ns1lsbW2hWq1if38f+/v7GI/HOD09RRRF8DwPnudJZ5ZhGMZ1YwLHMIwLWVdVeddwTaIHWerhlvw/MxcbhvExSVlZ2DAMwzCMTcNqw4ZhGIZhbBwmcAzDMAzD2DhM4BiGYRiGsXGYwDEMwzAMY+MwgWMYhmEYxsZhAscwDMMwjI3DBI5hGIZhGBuHCRzDMAzDMDYOEziGYRiGYWwcJnAMwzAMw9g4TOAYhmEYhrFxmMAxDMMwDGPjMIFjGIZhGMbGYQLHMAzDMIyNwwSOYRiGYRgbhwkcwzAMwzA2DhM4hmEYhmFsHCZwDMMwDMPYOEzgGIZhGIaxcZjAMQzDMAxj4zCBYxiGYRjGxmECxzAMwzCMjcMEjmEYhmEYG4cJHMMwDMMwNg4TOIZhGIZhbBwmcAzDMAzD2DhM4BiGYRiGsXGYwDEMwzAMY+MwgWMYhmEYxsZhAscwDMMwjI3DBI5hGIZhGBtH9m3/mUqllp/qhRhGkuVymfrcryGJnRPG58TOCcNY5W3nhFVwDMMwDMPYOEzgGIZhGIaxcZjAMQzDMAxj4zCBYxiGYRjGxmECxzAMwzCMjcMEjmEYhmEYG4cJHMMwDMMwNg4TOIZhGIZhbBwmcAzDMAzD2DhM4BiGYRiGsXGYwDEMwzAMY+MwgWMYhmEYxsZhAscwDMMwjI3DBI5hGIZhGBuHCRzDMAzDMDYOEziGYRiGYWwc2c/9AgzDMD4mqVQK6XQaqVQKi8UCi8Xic78kwzA+AVbBMQxjo0mlUshms3AcB+m0XfIM465gFRzDMDaKVCoFAMhms8hkMkin08hms0in01gul5jNZp/5FRqG8SkwgWMYxsbAak06nUYQBAjDEKlUCplMBgDQbrcRxzGWy+VnfqWGYXxsTOAYhrFRpNNppNNp5HI5eJ63si2VzdolzzDuCna2G4axMaTTabiuC8dxUC6XsbW1BQCYTqeYz+fo9XpIpVJWwTGMO4AJHMMwPjr0xZCPJTAocFzXRalUQr1ex2KxwHA4xHQ6RS6Xk9djIscwNhsTOIZhfFQymYx4YCgq5vP5R2nXpgdHv9FUPJ/PTdQYxh3CBI5hGB8FVm1c14XneQBei4zJZILJZAIA1yp0WMHxfR+e58F1XaRSKcznc8xmM3kuEzqGsfmYwDEM46PBDqZcLoflcimt2vP5/KNsE/H50um0vNcBfxbyZxh3BxM4hmF8FJggXCgUsL29jeVyKWbfbreLxWKB+XyO6XR6bUInlUrBcRx4nodMJiMiKo5jjMdjy8AxjDuECRzDMK4diptMJoNCoYCdnR0RGrPZDPP5HFEUIZVKYTabXavAyWazyOVyklxMYTUej82HYxh3CBM4hmFcO6ykZLNZ8eDMZjMRGdwuum6xwdTibDa7sjU1m81EWL3rdRMTQsanINlhyM8lj0W+GZfHBI5hGNdONptFsViE67qoVquoVCoYjUZoNpsYDAYYjUaYTCbXKnJYNfJ9H0EQIJ1OYzabYTqdYjQaIYoiTKfTt35/UhjZTcX4mNAzRjGzXC5lcaBFzmKxwHQ6/WgLg03FBI5hGNdOKpWSJOFcLodcLifbU9PpVDqarvtizRRjjmugz4cVnLc9nzYos9KzXC7thmJ8VGiEp7hJGuQBSBWSoseOx8thAscwjGuDF2bf91Gr1VAsFhEEgQiN0WiE4XD4Uao3AOS5fd/HZDLBYDCQitG7TMa5XA77+/vwPE9EURzHaLVa8nqtC8u4DijEOVJE/5tRBxToFDfXaca/K5jAMQzj2uAFO5/PY3t7G6VSCa7ritGXAieO42sXC6zAeJ6HfD6POI4xGAwwHA4xHo8Rx/FbPTiu62Jvbw/FYlG6u4bDIUajkXyfCRzjOmCVkREK6XRaPGs8hrnFOplMRPgA6z07xnpM4BiGcS3oDqZcLicJxvP5HHEcy9bUdXZNEb36JXzeyWQixuZ3PS//n6vo6XQK3/exWCyszdz4YLjFpI9Xfk57bvT/6a0qEzdXwwSOYRgfjN6aKpVKKBQK8H0fuVwOURSh0+mg1+tJ9eZd3UxXhd1aTC5eLBYYjUbodrsyh+oigaNvLJPJBOPxGL7vI5/PI5vNYjweo1Qq4ezsDOPx2LYJjPdCC5tsNisxBjx3AEgYpuM4yGQyWCwWiOPYjrn3xASOYRgfhF59Oo6zkkFDw67eIrruOVTamKlnXrGCc5nqjTZz8rXxRuT7vmy9mcHT+BAuqtgAeONjXY003g8TOIZhfBCs3DiOg1KphFqthlwuJ8bcVquFRqOxYi6+DnTXUxAEKBaLCMMQi8VCZl1Np9NLmTM5H6vRaGAwGCAMQ/mZ+Njdbte6WIxLk9yO0i3hDKBk+7fOjaKpGHgtuPWbHXuXxwSOYRgfRCaTQT6fh+d5KJVKqFarAF5v9zSbTZyenmI6nWIymVx7anEmk0EQBCiXyzLzajKZiP/mXQKH/0eBk8lkEIahCJt6vQ7f93F0dGQeCOPSUNTo4EndEp6MIKAA4ucodNiBaALn6pjAMQzjveAKlV0fzLzRq1CdIHzdF2eueh3HkbRkx3GwXC5Xnvey4xm4rUWBlM1mEUURer2ejHqwm4uxDm0C1pWbpEkYWM1W0kGSWujo9yZq3h8TOIZhvBeZTAaO48D3fVQqFYRhiHK5DN/3MR6PMRwOEUURoiiSTiYA17bFw7Rkz/OwtbUlAz3Z7TQej+Xjd22L8UYzmUzk9U0mE/kZstks2u223WyMN9BbT6zWAFgxDvOYoYBel5TN7Bstkhiv8DE6D+8CJnAMw3gvaOzl9G5WUHhhpw+GF+fkCvVDL9jrKjiz2QxRFL1RObosfJ3c1qI4SqVSVsEx1qKrNDS66yoMRQ2FDYA3BE4yyZis28oyLo8JHMMwrkwqlUI+n5ek4lKpBN/3sVwu0e12EUWRhOwtl0v5PwoH+mM+hHQ6Dc/zEIYhPM+D67oAIJUXiqv3uTlwi41m0FQq9dY5VsbdQXtraHLXwh7ASpVGb5NqUUPhrMURH4OLAj3WxAZuXh0TOIZhvBf5fB6VSgW+76NYLCKfz6Pb7aLX62EwGKDX62E0GsF1XQRBIBdsXqw/VOBkMhkZrOn7PlzXlfEKbEtPVo8uC/04TDQ2DGA1yyaXy8F13ZU0Yi1ItEGYlcSkHyw5ooGVH/0YWiCZuLkaJnAMw7g0ugzveZ4E4ultqdFohOl0KitbvmWzWYRhKOnGes6ONiXrycn6c+zA4gVfB6Yx3I8zpLSQspuC8b6sSxbmIFe+5/bSupbuq7R3r9vGsu6pD8MEjmEYl4Yt2blcDltbW9jb2xPxMp/PMRwOcXp6CgDiy+H2Ua1Ww29+8xvUajXU63Xs7OxgOp2i2+1iMpmg0+mg2+1Ka3kcxyufazQasv00HA7huq5UcDKZjLSlc8AmE2DtxmC8D9o8zJTsZODjOs8WxbUeKPu245D/x7lT2qPGmAMTOO+HCRzDMC5NKpWS0rzneQiCQKoneupxJpORjhK2jnueh+3tbezu7uL+/fs4ODjAbDZDq9VCHMdoNptot9sYjUbwPA/j8Riu68q4hNlshtFoJDcD13UlNTl5w+EqOBnMZzcJ47KsG61AUz0HYSa9NXo76jKVF33M6uOW//cx4hXuEiZwDMN4JzRUckp4Pp9HoVBANpvFdDpFq9XCZDKB4zh48OABXNdFvV5HLpfDy5cv8fLlS8xmM/zTP/0TyuUydnd3sbOzs7Ja1TcEhgXu7u6+0VobRZFsgw0GA8xmM9kqC8MQ4/EY/X5fvEDz+VxW02wd15k3hgG8FsMcN8LjnVUcHqvam5XcUl1nKH4X2pOmO6qSnjU7Vq+OCRzDMC6EF3VWYoIgQL1el+6pTCaDOI7RbrcRRRG2t7exv7+PIAhw//59uK6LXq+HTqeDwWCA8XgMz/NQqVRQLpfhui6q1SpyuRyKxSKKxSJyuRxqtRpc10WxWEShUIDruqhUKjICYjqd4uTkBP/wD/+A8/Nz+brRaCRbZd1uF51OR1rHJ5MJer2e+RqMtdBPQ2+Z4zgi4mlY5zYSt6Cm06mInfc5pnQ3FR+Lnwdw7XPb7homcAzDuBDdMcIqCVe38/lcqilcyXIeVT6fl2ni2lTMyd7ATxfvXC6H2WwmU8eHw6F8nMvlUCgUEIYhcrkcut2ujGIAgE6nIzcbJg1PJhN5vUEQSJs339iezgGgeiVOg/JisUCv10Ov17OV84ajO5hoGuZWFI8dHrv6GNHvr0MsJ7eq+DnjwzCBcwWuK6Bs3eOaT8C4aaTTaWmDrVarqFaryOfzCIIAnueh3+9jMBiIuXexWKBWq+Grr76C67ooFArSdcKp3lEUiY+H4sfzPGSz2RUjp+/7yGQy8rlsNisVIw7VnM/nskU1mUzQbrfFGKpnU+Xzedy7dw++72NrawtbW1uYzWbodruI41gqTHEc4/z8HOPxGL/97W/xL//yL+IpsnNys+A1Vx9fnBqv27VpVGdopRbGfAPwQVWWt1Vr7Lj7MEzgXJKPOWRP7+/qPVjD+JxQLHB7igM1tclyMBisTD+msKEReblcIp1Or5TxAUilRWeIsJ2cFRiuqvmx7/vIZrOoVquoVCorNyNWYeiboGDyfR+FQgG7u7soFos4ODgQczMrQM1mE81mE6PRCEEQYDgc4ttvv5XKk52Pm4UehUDTsDbD64qMrt6wYqO3o657+8iOs+vFBM4loYue5fC3tf5dpdKjL+baSf++AWWGcV1QLNALUywWkU6nMRqNEEUR2u02Wq2WrIRZleHxzK2rOI7feOxk+2wqlcJkMhFDp35P8ROGIbLZLPL5PMrl8sp50m630W63AWBFJLFN/fnz5/A8D7VaDdVq9Q3TKA2iem6W/hrj9sOuPopqinceYzpziccExYw2FL8tdE8nE3OCuK708L3xaTCBc0nY5sqD+6L2veRF8V0CheX6TCbzRidJ0nBmGJ8SJgV7nodCoYBSqSS5NcyqaTQacF0X29vbIm4cx8F8PsdoNJJE4eQxnBTzF8FzyXEcTCYT5HI5VKtV+V4aM8/Pz/Hq1asV4cSVObcfeA7TR8QtLC3estnsiujS8fvG7YWCg9uhFMv6ONSihsImjuOV7qh3waokt1fT6fRKlo0Z2z8tJnDeAS9yjJunyr9IwfOCyqyOd7X46YGFXDXyea00btwUaOKdTqeIoghxHIu4WCwWshrm8UzfAv0K74sWGskVt745JRORuUDgzYavcTweYzQawXEcMTfzc1rg9Pt962C5xehtKFZUkuZhHRVwUXr2u0SJbi1npZFBgBTHfEyrBn56TOC8BT1ErVqtYnt7W8yUo9FIjJP82lQqhSAIUCgUMJ/P0e/3xaS4bpWqDZG6/VW3IprIMT4XWpjTazOZTFYSheM4lot6EARiQI6iCN1uF91uVwZuvg96AcDHZ3dW0h/Bc2c0Gr0R9MfzU4slz/Nkpc1tNd78mKT8MXwWxsdHe7dolNeiJ9kJpbOS+Dng7eZh3WFYqVTgOI7EFdDEzo7B6XQq21VWxfl0mMC5AH0x1Bkg/DdXAnpWCbeb8vk8ZrPZSqDY2x5fV3Co9G3/37gJ6G2kKIqkkqPbrZfL5Yq/QRt/mUD8vn4ybQbl9hcrRbxR6FV28gZ1EfQIaZOpPueiKLIgwFuKvq5qEzvwerufwmVd2/e6yt2667GuKjLZm6b2OI4ldTsZFGh8OkzgrIEnBltTeQHUlRgaK3lhLxaLcBxHTIyTyQSnp6cYj8cSOKbFDh+Tc3ocxxHz8lUGtF0VtkXqlMzL7i8bdwsG5sVxvGLEHA6HK1u1wOo547quGJF7vZ6YjN+nGql9M77vIwxDmdrMLSjHceQGQ+HyruNZ++j4OBoz+d8eeOzpLBseixwjss7fqIez6s8Br0USK/g8FlgRYlcfO/Z47AE/VTt1LtNkMrmUXcG4fkzgrEHvpQZBIEKGF3Vg1Rys4+u3t7dRr9cxHo/hOA6iKAIADIfDlQNcixu21HL762MKHBpG2THCn8fKpkYS5szwAg+8nnCsL9TJdnKumKMoQr/fXxkgeFUcxxGBUygUZMHBxQDNo/w6AJcyBtMjZ9x+dFcURyuQdUF960YraHTVkI/Hr2U3H03q5XJ5xW8Zx7F04tGzxoUxRZbx6TCBswauQilyHMeRFNTZbLZSkszn88jn89ja2pKLsOu6cvFMlrl58vGxs9msfJ0+AT+G2Ein0/A8T7ph9ARcw0jCC7IOoLwocZX+MV2ST2bfvA86H4dbYFrccDss+d68a5vLOvMwjw0d0aGNxMlqzboFpLYb8LijP1I/r+u68vx8bB0CyKqN9lRa5ebzYAJnDb7vo1KpSDthJpOR2TbpdFqG+u3u7mJ7extBEGBnZwee5610dkynUxkKCKxm3tAwqfMXdFrmdYsclm3r9Tq+/PJLqd5wNs9kMrm25zI2g4vEb/K4ZOAft3G5mh6Px+Lb4Q3hKhd6zgUqFouyHeC6rpiXOceKWwassmpvnN1UNgtuR2mTOKuHhMetHoqZHISpxTnwugpJwURxc//+fblW01vJgMjZbCbddsx80qNLWJG3brzPhwmcBDzQWcGhx4YqnUFjPAHK5TKCIEClUoHrulKapAkyua+rg8s4V0e3Kn6Mk0GveBjaBkD2q838ZlzEZQQCbyjJCo7OqXlfuP1A0cQFQRRFK4sAfW5ZBWdz4d+Z3isteJLZYbpycxmvoW4q4ZZnGIbSGVUoFMRgD0AWpQyz5Hs948yEzefFBM4akvHdyc4pvQLQoU5U+MnWUp44epCbzsvRzv1k+f864NYUn284HK50t5j/xvgQlsslxuOxlPK5OBiNRjKr6iqeMi1WgiBArVYTsycjGrrdrpxrumtGj28wkXO70Z1H/JiWAf23BiDHlx6+qis37/K/6Jyc7e1t7OzsoFAo4MmTJwjDUBakvV4PJycnmE6nGA6HMq2eVRu99W/H3ufHBM4auGqkqEkGjAGvV7Y6C8F1XQyHwze2mJLt4Prx9Umn94avC125yeVyWCx+mujMThMTN8aHwhsLDfW+78NxHIzHYwwGgyvnydA4zC3iarW6MtmZGTv6psVz01pyN4d13ip2zOm/NfDa/M6KIbf7gctVIXnM5XI57Ozs4Msvv0SxWMTjx48RBIE8PplOpxgMBisCxzpRbx4mcNbAUuN8PhcjGfDTyhR4PbtmPB6j3+8jlUrJxZ0eAIY+6cfTF17duqjbGvWwN+DDPAQ6T0f7grhPzBWOCRzjfUlmRQFYWT2/T0WSCwZtxNcmf+2v0FUcz/MsMXZDoJjRE+fXVW2SW1HaZwNc/vqpr5WsyHOOWT6fXwl11cdd8vmMm4UJnATL5RJRFKHZbK7MLmFb4HQ6RafTwXQ6xenpKabTKQqFAtLptJQyKW5+9rOfYTKZ4OnTp2LApHdADyHk93D1wa2jD20t1HHhfKxWqyXl1OFwKKZmw7gqWtzQo7Bc/jT4kvOqeCO4yuMxGTafz6NYLErKd7fbxWg0ktbzyWQiYxey2ezKhHET7rcXVmsKhYJ4HPf29uB5HgaDgWx70uzLzBk9ruMq26HJih87TUulEqrVKsIwRLvdlioNfTb049ixdnMxgbMGOuB1YiqNZzotlRUcAPJeBwMWi0XM53M0Gg14nodUKiUGNAoYllm1uY2BUh/qIdClXb7uyWQiqx49AM4wroo2r3PrgJUWHmdX7ZpKrqB10JrOFtFtvzyPWPWxCs7th36bIAgQhiFqtZpsfaZSKYxGI/R6Pbm2va39+23PoT/Wxx+rR9rgDmDluFvXlWXcLEzgrIHt3ZlMBvP5XFoQmcFBQcAW2CiKpH28VCrJqrNUKolAogGNOQnT6VS2pIIgkI+ZjMkJzFdZAQOr7Y5MLQYgVRo9SE6naxrG+8CthDAMUSqVkEql0Ov1JM31KuV7Vkp935fWcAYGDgYDNJtNDAYDRFEkIr3f72O5XKJYLEobued5IrIY0WDcDrRfsVgsYn9/H9VqFb/61a9QLBYRRZGYe//whz+g0+mg0WjIFtJlj7fkiB0tanRQoBY+/J7pdLqyZfUx0VtyetvNFqWXwwTOGhjOxEh4XWXRKwRmyWSzWURRBNd1UavVUKvVEIYhfN+XbBwtcOI4XjEnU0DxAk9zJveWr1LJ0W2Uvu/D933pPNF5O7o93QSO8T4kt6joOev3+yJwrrKa1mMZmFoM/JQC3u/30W63JbeJ5xTPk3w+L0Z/Chx6zozbBY+rUqmEg4MDbG1t4Ze//CVqtZr4u05OTtDtduE4DuI4RrfbBYArG4uTHa4UOMmRDdrLOJ1OpT38Y/zsGp5f+mey6/blMYGzhqRipxigXyVpcgQgF9ThcAjHcTCbzWRrqtvtrgwnTBqM+RjaZJxMj73sa9ZR4vRFUIRRYFGo6S0q8+EYV2XdllIqlZIFwmVn79A8yqgFTn/mAoMr5mTFMZkdxcUAt3qT86WM2wO3OgeDAcIwlL+967oIggDT6RTb29vydaPRSDyNvM5eZh4Z3/Maz60u3aEFYKXiDeCN+8N1QKGlTdU6SV+Pg7AQwcthAmcNVM1avUdRhPF4DOD1CcGqzGQykS6p4XCIZrOJXC6H4+NjZLNZdDod9Hq9lenKwE8nCT/mTcJ1XbmoX6W8zxJrGIZ48OABwjBEpVJBuVzGZDLB2dmZ5JJwdc2hidozYSeLcVm0/4aGzEwmIxWcyxxLvJjTJMxkYrYCR1GETqcj2U3MGAFe33R4HvJ88X0fAOQYt5Xu7UFfg1qtFn744QcxFAdBgIODAxwcHGB7exvT6RTNZhPffvst6vU6ut0uvvvuO/T7fQyHQ9nKvOh5dGVcz+abz+ciLOj/oudSD469rpwlCiX6fZig7Pu+pCgzNXkymaDRaKDZbGIymWAwGJjV4C2YwFlDsiwJvL6YXtTqPZ/P5fP07XCbi76bdStQ4KcVh26B1LNLLnvQ6pydQqGAUqmEcrmMarWKOI4Rx7G08eqWW7bBv89zGgawOi+KRvbLesd0xSWXy60Yi4HXhn9d/SRJUyn/zyo4txteg+I4FrHMCg29jvP5HFtbW8hms+j1euj1ejIck9c74O0T7Pk8/Jpk7IAe2ppMQr7MQNfLoO81nHGYz+dRLpelAl8sFjGZTGQ7jhYGmq0vWym9i5jAWYPrumJwZKmSq0Tg9cGtTW18zxK9DiDjY9IjwH1cPZCT38/vvWwFh4LLdV0JRXvw4IFUb0qlEhaLBfb29jCfz+ViMB6PcXZ2hvF4jEajgUajgclkgn6/b74F41Lozjz6IbjN2mq1pIz+tmM4k8kgDENpzS0UCgAglc5ut4ter7c2zoDPDUBuavTM6bRbW93eLnhc6erbP/3TP+H58+c4Pz/H+fm5+LX29vaQzWZRr9dxdHSERqMB4KfqHQ3ol3kueha51cprvPbp0NuohdD7wsf0PA/1eh2e52Frawv1eh1BEODevXsyE7FcLsu5EMcxnj17hhcvXqDf7+O7777DYDBYqS4ZrzGBswbOmeKFklWYZHlSx4AzIjwZRMYVRz6fX9nT5b4xJ9DypElWh64icDgb6/DwENvb2ygUCjIslCV/CpzhcIgXL15gOBzi6dOnWC6XknDMao5hvA0dHslBtIRbspfx34RhiCAIZMVKoc2Var/fl3NKo2MPdCYJV8M6edw8ZrcHXveGw6Fck1iZbjabaDab2Nrawm9+8xvUajVsbW1hOp3ixx9/xL/9278hiiL0er1LPxfw03U0n8+LyEmO/KBtQXfFfojA4eP5vo+DgwOUSiU8evQIX3zxBVzXldmGW1tb2Nrawnw+R7fbxXQ6RbVaRbFYxNnZGZrNptyHTOC8iQmcNehkYZq9AKykC7NaQ8c901b1icHqj77YJnNpdGVIC5r33S7SJjTu6XJVzdfC6tTW1hZ830cURdIFxhk/Zl4z3oU+hgeDgfgVgJ86n9YZgwm3kSj+eZ7x8Rjkx0rmRVUYLiL0VhUHL+rZcdfllzDe5LrNtoSPx+NruVzi7OwM+Xwe0+kUvV4PnufJYpLHmxYtV3ldSdMuxTurg9wme98tIe03o5gKwxD1el0Wo6w8cleAPwcrTLPZTLKBhsOhiK7ryE3bREzgJNDGSZYNHcdBv99HLpdbmTuiKzK5XG5twBiVOg9EDu/k4/OE4cVce2EugzbJccry8fExxuOx+G+YwkzjHE2Ye3t7WCwW2NnZwfb2Ns7OzmTlMxqNxFRtGOvQWwlPnz5FLpdbqepwu3Nd9aRYLKJcLsP3fezs7IivgmZiJm4ztXbdOUExT0HE56GYp6/HTJgfFy1sP8bvOI5jnJycIJPJoNFo4E9/+hMePnyI7e1t9Pt9ESbHx8eIokiOFwrmy1xPeQ1m5x6FNtOzG40Gzs7O0Gq1ZFv0Kj8vK0QPHz5EGIbY29vDvXv34HkearWaLEYZZDgcDpFOpyUsNpvNolqtyv+xWsPwQ9uOXY8JnDVQqHBrJ5fLSQeU/n8taJLlzOTj6GnHbP8DsLLS/BAFrju7oigSwxpXOPl8HsDrsRBcTQCvc0smk4msMK7LRGdsLnqFTWHCSgpvGOt8ZBQfzGnyPA+u68pWLcX+2ypAfH692malhsdusmpqfDw+ZvWAOV6pVAqTyUQqN61WS8aDLJdLuYZxkXjVv7k2retKIp+fnbRJ0XTRz63DBBmDUCwWUSqVsLu7i4ODA+maYjwIK0g6h20+n6/MKmTVM7mVxuc0gfMaEzhr4AHGdkHf98WYO5lMMBwOMZ1OV/Jy9LYUS4z6QOPXsYID4I1AqfdpPUy2zI5GIxwfH6PT6aDZbMroiFKpJJNyt7e3pRXR8zy50TBFWXsXDONt8KbAfCW9tZoMJOP5kU6nUSgUUK/XxZhMzwVn/qxrC7/o+flcTAdnHg/9c3Ysf1wYRLfOJ3Vd6AXcYrFAq9XCP/7jP+L777+XY6rVauH58+fodDrSPn2Z6g3/n4vPKIpwenqKbDaLRqOBdDqNk5MTnJycyGKQJviLHjubzco1t1qtolaroVAo4Msvv0ShUMD29jZ2d3elajmbzdBqtdBqtaT6yQolO6tYkdQ5UxQ91jG4HhM4Cm0e5sGby+Wk+sHyJ0+epLChx4UXbH2BZ6gfqyOsrvB5P2SVqb0Qo9EIJycnK65/+m5c18WjR48QxzG2trbw4MEDSY6lwGGSp90UjMvCC/G70H40zhfSlcvRaIRms4k4jmUR8S4fmI5p4KKDxzDfJ7eTjetFexQ/1haJrhYCP2Xk/Pa3v5VYAcdxMB6PcXR0JBaCq3oIed2jwAFe/0zNZlM6A7XAuQjHcVCpVBCGIR4/fozHjx+jUCjg4cOHEuNRKpUwHA7xww8/IIoiNBoN/PDDDxiPxzg/P8dsNpMuK349u3RZ2eE5Zdfr9ZjAUSSju7XY0Z1RvKBqAyO/T5ctKXB0yiq/3nEceV696n3fiwO30/ia+br5f1xl65ENACQHRyctW6aCcZ3wXNIp29wKXS5fZzLRxKlHPFz2OOR2go7dT2ZZGR+Hz7E9Mp/PRciwCviuLc23wYiOdDoto0GA11UpzkC76Dl4nHGBG4ahZNnUajXU63UxFbuuK5Ed/X4frVYLvV4PrVZLugd1h20URcjlcnJtpr3AdV0RN3aMr8cEzn+SzDvgBZLhUuPxWErmOsNG7/nz4qrzQXgycDWhL/Y0++q91/e5QKRSKYRhKOme3W5XJi+Px+OVE69er2M0GslqlyVZthyORqO1mSOGcVV40WVpnduijDPwPE9ymDhMs9PprIwRuQxsE+90OigWixKZwDwTW91+XJKew0/BbDZDp9N5Y04UPTNXhYNboygS8aQfj8nIukVcb72yMlkulyV/7Gc/+xnK5TK+/vpr/OxnP5OKOgCcnp6KleBPf/oTer0eTk9P0Wg0xIeWSqVwdnaGly9fIo5j3L9/X6rt29vbGI/HK55J89+8iQkcRdJPoys4vOEnD6B1lRedZcM3fj8rOcmW8w+tmtB7wIuNbnEEXndz6Y4UXvx5QrPsak584zrgOcQFA4UHRQjPA65YuYDQC4LLwpsCzZk00msDv/Fx+By/W1ZcrgsdrjoajaTqzc/xGF2H9l56nodisShdglrwcHuJz9HpdMR30+l0xDvE6y8X2Lo9HYA0kCQrOFatfBMTOAoeIKyucH83l8uhXq/L+AV6cEajkbj7WfGhw52Jqszz4KwqigqOZNCDAt8Hnlg0bUZRhFartSKygNezV9iWu729LSXU+XyOZrMpBk+r4BjvixY1zLbhRd7zPOkY4YwpHq/JDpWrordZ+d73fezt7Ylh2QIsPw56i+o2VxFYeY/j+A2j/NvEVCaTEdH+4MEDPHr0CGEY4uDgQEZHfP/995hMJmi324jjGC9fvsTR0RH6/T6ePn0qgYa8n+gp4vqeQU8lvaHlclmqS+12+4N2AjYREzgJ9MVZZxPUajVUq1URONPpFK1WS1z2VNJhGCIMw5WWV5Y2tSEslUqJwHnfCk5yz7der6PX6yGXyyGKIgCrE3MXiwWCIHirwNFzfwzjquiKTRAEyOVy2N7exs7Ojhh/U6kUOp0OGo2GZN7o6uH7QA8OQy49z5Mo/06ng5OTk2v+SQ2SFDW3VeTwGsxU7Mv+DNlsVoL6Hjx4gF/+8pdwXRfVahWO46DdbuPk5ATD4RDPnj3DYDDA+fk5Go0GRqMRTk9PV0z6bEJJWh9YjaTA8TwP5XJZziH6Om/r7/9jYAInAU1cFA7szuD/6XRJYDX/hl+jt3mS3hvdocTHe99kTB7srB5pE3SyPVcPM6R44/PTr2MmY+NDoYhhvAI/5jHPUjtXq3EcX9uqU3dTjcdjOZd1Ho6tbj8OemtkE36/b8u24fWW20T5fB47Ozsy+Rt47elJpVIy64+dUsPhEJ1ORwL73jZzkMczfZ8MdNUp33pYsrGKCZz/hGKEB3AQBLIHSt+K9tNo4aLhAUv/C8URV6682OsU1qte4PVWWqFQELXPUqWuDAGvWxbz+Ty2trZQq9UQhqEY6Pr9Pnq93htBWYZxFVKpFDzPQxiGsj3E6AEa9Vkl7HQ6MhaE1c4POeb0PJ52u43JZCKR9qzC0uBp1cnr5XOYjD8XvJa6rov9/X0ZisljPZPJyBR0bkednZ3h/PwcURRJGzs7WXW3K9HRIvTq5PN5DIdD+L4vM7rYhcXOKz6mXbtfYwJnDayM0BQGYMUno83DSWhM023lwOtuK72S1P6b99meopmSpUkdAJWs4HBCum4v1INCuUqwFa7xIbCCo+dMsWTO8j8NxfTcXMfcM31TiOMYmUwG+Xz+jQTxD30e4010/MWmotvAOTutUqlIaGq9Xpc0bh7f9Jg1m03Zjup0OpcagaNzfzgLS/s5ef7wc9Ycsh4TOApeIBmMN5lM0Ol05ABLHpgss6fTaTFU6qhtZhnw4qq3rFhWfB8Pjt5C4+opiiJZJbANnM/lOI4YPQuFgmxPNRoNabfU+SOGcVW0uZjHJsU+t0BZXdF5TNe1HcrtVpbvKWa4lVCpVJDNZtHr9dDr9exGcI2wwr1p1QM9HLNSqcjstP39feTzeQnf42KRkQfNZhODwQDPnj2TyeYMrnxbOCDRwa2dTgcvX76U2YKVSkWqNu12Gy9evECr1ZI0ZBM5q5jAUWgjsO/7WC5/mm/CA5R7qkwtZklQT+qmoma5fLFYSIleCxxOEX+fCg4rN3xepm9S0SeNwtlsFuVyWeLCHccRYzG7WShw7OQwPgTdok3B0e/30e12V7aPktuoHwrPIa54WSVlS22lUpHBm/1+347za4TDJzdt6y+dTiMIAriui4ODA3zxxRcIwxCHh4cIgmClAk9LQrPZxPPnz9Hv9/H9999Lxy2FzWWPO94nut0ugJ/mBbqui0KhgG63K/ekly9fIooieR5jFRM4Cm1I1CXvdVk4OrBPzwPRYU66bVJXbvQWl95OuuzBz/K/nkHC1ROFjRZNbF8PwxCZTEZWud1uF4PBQNT/dd5wjLsJL/bAT1XFpPjmtu3HMrLzXNXGTIZa8nzhosCqldfHJlw3eGzQGO95Hra2tuB5HqrV6kp4JINc6aPUnbWcV0UT8FUXsPreo5tAer0e5vO5CBxuhVnX68WYwPlPWMrmto8ud6/LskmlUtIaSN+BbuHTF1fmKOgwJm0Iu6rBuFgsYmdnB8vlasw9y/40bfICziGb+/v7cF1XTMXffvst2u02jo6OxNFvF33jfeDxyzC0dDqNTqeDVColW7y8KXysY4wVHJouwzBEoVDAYrGQUSY8P9kIsAk35s/NJvwOM5kMfN+H4zgSa+D7Ph48eCCeG4ZTzudz6YhihtPx8bGIkF6vh+l0urINe1W4m8Ak+tlshh9//BG5XE6eYzabSVOJXbfXYwLnP9FjGpiwqis4ySoLv5YdTBRG2oisE1T5ef77Q9rDGUTIbTA+Hm8iF1Vw2BnGFUG320W73Zb9YavgGB+Cjh3QhlNWU64q5t/3NbCCxOcFXp+vPE9pirXj/W7D41THG4RhiGq1KlO/aVcAXlcI2eHErSKaiPmmW7c/5BjjOTWdTmXYMz04XDDYMXwxJnD+Ex23Hccxut0uxuMx0um0zNHRE8PT6bQYd7WQ6ff7K+MQdLlR877pxUwj5iwSndK6roMK+KmCs7W1hd3dXQDAyckJms0mnj17JmGFrATZyWJ8CBQ4yc+9r5i/KtrsysorzaIM4vR9f0XQ2zH/YSQXfclw0Zv2O9Z5ZKVSCcViEfl8Hru7uytJ767rolgsIpPJoN1ui0Ge7d+dTkc8ZZ1OZ6Xa/z5iXosttpz7vi+diEw21jYI7jYY6zGBo+CBwwN2NpvJnizNikxpzWQy4qwHXl/EWUXRJzarKvpg/5ATnoM1+/0+jo6O5KLOVXLysbPZLLa2trCzs4Pz83OcnJzg/PwcL168QKPRkBP3pl2IjNvH2+a2fQp4LmjvAgcUMt/K932Z8aMDOo33g787VsiA1xXrz308rEMnwNdqNRwcHKBQKODJkycoFAqoVCqoVqsAXl/Xz87OcHZ2Jt1Rg8FgpVqjbQEfsl1Ekz5TwJmMr+diaZ/obR+P8bExgaPgwTmZTOTATW7d6Pf06CRXjeuqKcnvvwq6BZcHdzJrZ93jMnOHvgPXdbFcLuXE1BPSP8X2gXF3+JzHES/8TDQGVreIOVPOJjBfD7lcDgDguq5EUPB6qOc63YTrix7Fw61+3/cRhqHEaPi+j0wmg+l0KuGnNPYOBgNEUSS5NEkj8VV/Pl1Nouji9ZqJyXxsbnu9T2PKXcUEzn+ik4UXi4XMcuKFkeJCG4a5HaX3ZZkuyRXk+wb5ES1SqOoBSIgUXfRJoxkrTyyzcqLt06dPcXp6imazKVNsryNJ1jDITTiOeC7O53MUi0VUKhUJxdQl/+vM4rmrsNrBKdra/EphoLvrPsfvmotENpPUajUZsbC3t4dqtYovv/wS5XJZTPHD4RDffvstOp0Onj9/jhcvXoh9gQtMffy8j7jhHCtGkwRBIK+TsEtKN5FcZEcwVjGB859onwyFSXL8PA88hvfpygcPdubJXEe5Mvm8NMEBkAC1dW2IOnVTDwx1HAeLxeKNFEwTN8YmQmNmLpdbORf1wFvjw+Hv0/d9FAoFyWNhJ5H2i/Bv8DmuN3popeM4MpfPdV14nifJ2xQTk8kEvV4PrVZLhBrzbvSInfdBN7XoOYH5fB7Aalq+nkGln/cmVMRuOiZw/hOWVdkGyCoN8NPBOJlMJFhPCwhWdajkk+qaj/2+OI4D13XlJHAcB1EU4ezsDOPxWJz13OemQc1xHNy7d0/mpWQyGURRJHNMuMKyFYCxqbCKOhwOcXp6KrlP3Haw0STXw3Q6laaLJ0+eYD6fi/k2n8/DdV3p/mFn20V+wY8BxSwrJBxtw3EKmUxGQlyDIJBssF6vh++//x69Xk/SiXlveJ/jRifecztPxxfoiiKfQy9CWWm3jtfLYwJH8TY3unauJz+nqz/J8Qsfig71Y6jgeDyW5EqOZdACh4FUDx48wC9/+UuJqY/jGFEUod/vr+Qn2IlibCJc6UZRhNPTU+RyuZXZPfq8Md4fdqqVSiV8+eWXmM1maDQaGI/H0pBBLxTHdPAa+Sl+/6yWBEGAWq0mAoIdUdPpFM1mE51ORxaQo9EIw+EQR0dHiKJIPvch10qdtcMMNY2u0nBOG6vt2vLwoZ7Ou4QJnP9Ej2C4yBycLEeuSyq+zlUJq0RU/TrD46JRD5lMBmEYIggCVKtV1Ot1BEEghj/dymjixth0eG6yGUCb8+3Yvx4oUuhbASDViUqlIsF4jN1gNQLAJzN5L5dL8TICP22rLZdLCWgFII0XrHLrJoyrHi/6fqIzmHSYLICVe43eguJzrhugbFX3y3PnBY4+AOme1+Xrt3UY6QPtup3t3C8OwxD1eh0AVg56VnCSHhrP83D//n3U63X81V/9FX7zm99gPp+j0+mg3++j2WzKDCrt4TGMTWU+n6Pf7690pKxbsBjvB4XL+fk5vv/+exQKBRweHoqJl0Mjc7kc2u02jo+PpYrzKRZZeuCx7/srkR987slkgvPzczERd7vdFaFx1WOFFSOah33fX8lQ43PqTjMOpGXDC5+XAtIWpFfHBI5qweYWkL746RXGRQLnYyhr7fPxPE8UPvB63k8ybwf4qYJTKBRQKpVQqVRQq9UwHo/RarWkckODsp0wxl2ARk2NrYKvDy6wRqMRut2uVCt83xdhkclkUCwWMZlMVtqfPxWMDtAJ9Jx4TxERx7H4GgeDwXsfI7yf6HiOfD6/0hmlt+hYodGNI7YddT3caYGjW6ld10W9XofjOCtdRnTO6y4MHmgUP9ctcNgxpdW/rtpoJ33yOXO5HLa3t3Hv3j3s7OygWq3KALhGo4FutyvlV0tyNe4K6xYnxvXAa1Gj0UAqlcJgMMD9+/eRz+dRq9VQrVbh+z7Ozs4QBIEYv+M4Xrm+fsy/yXK5xGAwwMnJiUzldl0XW1tb2NraEo9NJpPBaDSSeWWXuUbyeq3HPej0bHbxUchoMUOfDT/Hr1mXvWZcnTsrcFghyeVy4q7f2tpCPp8XUxn3X9lZpQ92ffB9jOpNsnWQr4EVGD2oU+M4Dur1Ovb397G9vY1qtYo4jtHv93F+fi75OSyB2slj3AXsOP94cAul2Wyi3+8jjmMMBgOUy2Xk83ncu3cPQRDg7OwMvu+j1+vh/Px8JQn4Uyy0uEDkMGTP87C/v4+DgwP0ej00Gg3M53MJAaTf8V2t7eyOYgJxsVhcyd3Rgays0LBaxI+TvwM7Xq+HOytwiE6QZB4C8Dp2fDwew/O8ldZGvTfKA5Ml8A89MFne5EogmVvzrhk63GcOwxDpdFoi6QeDgUxZNpOlYRjXBbd/WIUYj8fo9/vodruyUFwsFvB9H7PZTMZlLJdLMeJ+ikwiVlFSqRRGo5F02A2HQ0wmE5k5OBgMkM/npZqS7DZl2zm33mgj4H2EvxMA4pPkY9Bjoz9ngX0fjzstcOhx4aTtcrmMIAjk/6fTKXZ2djCZTORE0C18XIHo5E7g/UUOTxjP81AsFmWlcXx8jDiOZcgbT5Lk87AitbOzg4ODA2SzWTQaDZycnODHH3/E06dPcXZ2hiiKzFxsGMa1wUUeDd0//PADut0uXNdFqVRCOp3G7u4uarUaWq0Wzs7O0G63ZdAvHwPAR7su8fXxGprNZlEqlWSwZalUQqlUktEJ4/EYzWZzpfuUSchsAAmCQD5HkaZH/mjzMJ9bh/hd1MBiXA93WuAA6+c1UZ1TsMxms5UZUJlMRgQCv+Y65troVQG3qCaTiaQjJys4ye/VpuR8Po/lcinCTFdwPlX3gmEYdwdWIabTKXq9ngToRVEEx3EQhiE8z5MJ2ePxeCUsVQ+Q/BjXJt09x2rTcDhEv9+H53kol8uy4GXXk+M4K3PLuO3E+wXtA7lcDqlUaiW+413mYf26jI/DnRQ4OoU4n8+jWCxKAFM2mxVnO8utLCcCEINaOp1eEQ2s6PD7rgLFE08snjg6nI8nSXLmFH8eTp2tVqsyaPDk5ASdTgevXr3CyckJ2u02oii6thEShmEYSeI4xtHRETqdDqrVqgyxPDg4gOd5CMMQe3t7yOfz6Pf7KJVK6Pf7MjuM19OPuW1DodNut/Hs2TMUi0URYKVSCalUSpoxmGBPoaNDX3ULN3927bPR9xJdrTFR82m4swJH750GQSCDLClwmIHAmSRcYXALKJfLyUnJ9zx4aQC+ymuhd6ZUKsmclEwmg16vJ9tf66o3XFUUCgUxFafTaUwmExwfH+Obb77B+fm5CBya+gzDMD4Gk8kER0dHyGazqFarqFQqqFarqNVqMsF7d3cXvu8jiiL0ej2cnp6ueFMoCD4WFBmdTgej0QhbW1t49OgRHMeRBS99RFwwjkYjqd5Q4OjoDnps6HNcN8TYhM2n5U4JHJYZGZVN1zsd8ByWyZZBzk+ZTCby/TzAPc/DYrGQx/Q8T05Ofv1lDmbdEs5ZU5lMRlYEyXbwpLjhyeb7vniIeKHgWIbhcLjSgmgYhvGxYKWCCcbNZhPL5VKC9OI4lu5VjpHh+JlUKiUeQeDt43OuA24hsWWdOWjc3k/O69MLWNoV9Od4D0jaAEzYfB7ulMChryUIAty/fx9BEIixbLFYyIC1RqOBVqsl7dWz2Uy+t1qt4mc/+5msStLpNJrNJhqNhuwh6xLru6DRjZUkdhdwYi1XAet8NxzLkMvlcHBwgK+//hqlUgnT6RStVgsnJyd4/vw5ut2ubKVddfvMMAzjqvD6d3JygsVigWKxiH6/j0KhgN3dXezs7CAIAhQKBWnZns1mGAwGYjr+FPk4FCPdbhfffPMNwjCUxabO69GdVBRd9Ecmg1fNPHxzuFMCR5vDmFeQz+elW4k5CcPhEN1ud+VzruvKtpWu4vB7+ThXbXtMp9MrE8M5UfZdgX7A67wc13VXusBYweHAOHp4rDXcMIxPBdONO50OZrMZCoUCRqMRisWi+FoYhcFrsW7o4DX0Y86r0sbowWCA6XQqk74pXJL5Z7weM5mYC1JdHbfr7M3gzggc7YDnCUMxw5yZdruN8XiMTqcjhjd94CYzacrlMra3t8UD0+/34TjOpU5M7QFiuyENdjzZKFIu8vR4noeDgwOUSiV89dVX+MUvfgEAkkMxGAxWPDy2ojAM41PC2U6sanueB9d1USwWpYU8m82iUChga2sLjuPg9PRUhMWnGsY5m81ki4pWAV6P9X1g3VaVBfTdXO6EwNHiRk8M5x4xu6XOzs5kFsloNHpDtc9mM5k4G8cxPM/Dzs4OFouFbC85jiPP+TZ4IlHghGEoXVncGmN7+EX5Oq7rykiGn/3sZ/jFL36BwWCA//iP/0C/3xeRQ9ObCRzDMD4lXBCm02l0u104joNSqYStrS0UCgUUCgXkcjkZKpxKpeD7viwiPxXM79ENKHrr6aLvsYaNm82dEDiEImU6nUq7NBOKaX7T07m1MueBz5EJOrEzjmPZImq1WiudWEkotjzPQ6FQkG4pPqYWNTqMb50wyWazKJfLqNVqKJfLCMNQTM4cx8CtLjMXG4bxuaB3he3X7XYbs9lMOpYGg4GkHuvF2KdckOmgQZvTtxncCYHDA5UhS1Tl2u8yn89lO2hdTozjOMjn83AcR1obj46O5Pvv3buHra0tjMdjnJ6errjpie562t3dxeHhIQBIRWgwGKDT6YgAe9tJlkqlEAQBvvrqKzx58gRfffUV7t27BwDodDo4OjpCo9FAv99/53gHwzCMjwkFznK5xPHxMQBI1Tqfz6PdbqPdbkvVmYuzT3290kGA/Ldxe7kTAodoZR5FkVROKGrWpUwSGovT6fSKibfX60lIH1M6uX+b3KbS22Ss4HCfl5WlZJvhOlgFYrm3Wq2iUChIOKB2/+vBnHayGobxuaB40MZjjp/pdrtSxXnX9e9TvE5jM7hTAkevIuI4FgPbRTkzhFtKlUoFvu+v5MxwSKfrum8MyaTLXj+O4ziSuUPXfqfTEVHyNpGlX0s+n0e5XEa5XEaxWMRsNsPZ2RlOT09xenqKs7MzeVybO2UYxk2AnVWtVgtRFAEAcrkchsOh5HXxvflbjA/lzgkcHUKlP/8uKChyuZxsd0VRhHQ6jfl8jkqlIrOsPM+TNkJdxWGoHys83D5il8FoNJIEz7e9Jr4WZvgUi0VMp9MVgXN6eir+IFuRGIZxU+CQYm73cyQNF3efY2vK2EzulMDRXPUE0iZjdjXFcSxbVu12G47jSG7OutWHNtpxi4smZ37P26pIzIdganEYhiK22IF1dnYmlSC+TsMwjJuGHkj5ruufYbwPd1bgXBVWX4CfTMHL5RLD4VBEx9HREQDg+Ph4ZatJwxwdTirv9/srWQtv20qisHEcBwcHB3jy5Anq9TpmsxlarRa++eYbfPPNN2g2m3j58qXMnbKLhWEYNxE2dgB4q0XAMN4XEziXRPtpuD+c9NgsFouVCs66k1X7ffRME26drYPVG26BFQoFVKtVFItFGfDW6XRkoCYNxraHbRjGTcauUcbHxATOJeHQOB3fva6dnCGBF7UacouKJuSkyTlJpVIRc/O9e/fg+z6++uorfPnll8hkMpLfc3Z2huPjY0kuprnYMAzDMO4iJnAuAdvKKXD0gDX6XGgmftuwtXXTwN9Wkk2lUqhUKnjw4AHK5bIM0/ziiy/w8OFDRFGEp0+fotfr4ezsDEdHRxiNRiJwrNxrGIZh3FVM4FyS2WwmHUkMCqRvRs+eelfLueYyAoThgLlcDqVSCZVKRTJvOM2W87Qmk4lsdxmGYRjGXcYEziXhEMzkYLV1YuI6BQbnooRhiC+++AK7u7uoVqsol8uYTCZotVp49eoVms2mVG5sNINhGIZx1zGBc0nW+W0+JZlMBr7vy/wqTi0fj8dSwdGpxYZhGIZxlzGBc8MZDoc4Pz+H53lot9sIggDdbhepVApHR0f48ccf8fLlS7RaLemcsuqNYRiGcdcxgXPDGQ6HODs7WxE4URQhiiIcHx/jxx9/lPZw5vNYBccwDMO465jAueEw5XM8HqPRaCCbzcrclkaj8cbkXRM3hmEYhmEC58bD7q1Go4Hf/va3KBQKMnV3OByuJCebuDEMwzCMnzCBc8NhOzpFDsVNFEUYj8eIosgm7xqGYRhGAhM4NxgO52T+ztnZGRzHkcm70+nU2sINwzAMYw2pdyTp2p7HZyaVSiGdTkvgH8MEOesqmZa8SSyXy9Tnfg1J7JwwPid2ThjGKm87J6yCcwuggNFzrNbNuTIMwzAM4ydM4NxwdGeU3oYyYWMYhmEYF2MC5xZhosYwDMMwLkf6c78AwzAMwzCM68YEjmEYhmEYG4cJHMMwDMMwNg4TOIZhGIZhbBwmcAzDMAzD2DhM4BiGYRiGsXGYwDEMwzAMY+MwgWMYhmEYxsZhAscwDMMwjI3DBI5hGIZhGBuHCRzDMAzDMDYOEziGYRiGYWwcJnAMwzAMw9g4TOAYhmEYhrFxmMAxDMMwDGPjMIFjGIZhGMbGYQLHMAzDMIyNwwSOYRiGYRgbhwkcwzAMwzA2DhM4hmEYhmFsHCZwDMMwDMPYOEzgGIZhGIaxcZjAMQzDMAxj40gtl8vP/RoMwzAMwzCuFavgGIZhGIaxcZjAMQzDMAxj4zCBYxiGYRjGxmECxzAMwzCMjcMEjmEYhmEYG4cJHMMwDMMwNg4TOIZhGIZhbBwmcAzDMAzD2DhM4BiGYRiGsXGYwDEMwzAMY+MwgWMYhmEYxsZhAscwDMMwjI3DBI5hGIZhGBuHCRzDMAzDMDYOEziGYRiGYWwcJnAMwzAMw9g4TOAYhmEYhrFxmMAxDMMwDGPjMIFjGIZhGMbGYQLHMAzDMIyNwwSOYRiGYRgbhwkcwzAMwzA2DhM4hmEYhmFsHCZwDMMwDMPYOEzgGIZhGIaxcZjAMQzDMAxj4zCBYxiGYRjGxmECxzAMwzCMjcMEjmEYhmEYG4cJHMMwDMMwNg4TOIZhGIZhbBzZt/1nKpVafqoXYhhJlstl6nO/hiR2ThifEzsnDGOVt50TVsExDMMwDGPjMIFjGIZhGMbGYQLHMAzDMIyNwwSOYRiGYRgbhwkcwzAMwzA2DhM4hmEYhmFsHCZwDMMwDMPYOEzgGIZhGIaxcZjAMQzDMAxj4zCBYxiGYRjGxmECxzAMwzCMjcMEjmEYhmEYG4cJHMMwDMMwNg4TOIZhGIZhbBwmcAzDMAzD2DhM4BiG8clIpVJIpVKf+2UYhnEHyH7uF2AYxuaQTqdFxKwTM8vlEgCwWCywWCxWPmcYhnGdmMAxDOODoZBJp9Mrb1rkLJdLETOz2Wzl34ZhGNeNCRzDMK4EBYt+n07/tNvtOA6y2SzS6TQcx1kROovFApPJBIvFAsvlEvP5HIBVcAzD+DiYwDEM451oMZPJZOQ9xUs2m0UqlUI+n0c+n4fjOAiCANlsFo7jwHEcxHGMZrOJOI4xGAxkm8oqOYZhfAxM4BiGcSkoZrTAochh1cbzPOTzeWSzWQRBAMdxkMvl5P1wOAQAjMdjqfqkUikTOIZhXDsmcDaA5JYBzZuGcRW0j4beGYoQiplMJgPP85DJZOA4DlzXRTabRT6fl885jiMf68+xwpNKpTAej5HNZjGfzzGbzUzkGMYngOc2ANki3mRM4NxydLdKOp2Wcr/dLIyrwuOIQkZvQ3GrKZvNIgxDZLNZeJ4H3/eRy+VQKBSQy+UAvOmpyWQyyGb///a+tLmN7Eo2sRVqw86d1NI9bXe3PbbHMeGJ+Ufzayfmgx32656xRImSSOxLoRbs74Mijw5KkERSpAiSNyMYpCgSAMlbt/LmyZPn/VbjOA6y2SyiKBLSM5vNDCk3MPgGYDl5tVpJefghwxCcewxdLqCZU7fhGqJjkEZa7ePHaTKTNgpbliWfLxaLa6SFKoyGXnvZbBa5XA6TyQSz2QzT6RTz+dysUQODWwSVUx5GstksZrMZZrMZFosFFovFgyc5huDcY9DzkM/n5dS9WCyQJIksYN54HvIiNrgcSITTxJjExrIsKUFxXbmui3w+D8uyUCwWsVqtMJvNsFqtMJ/PMZ1OAQCz2QzAhzbx5XKJxWKxdlJcLBYYj8eYz+eIogjz+Vy+xqxPA4ObQyaTQbVaRaPRgO/7ePbsGRzHwevXr/H69WvpZqR6+lAVVENw7jF0OYFMPe1pMN4GA0K3bJPUkODkcrnPEpxisQjLsmQj5Drj5kgiTeJEAkRiw9PidDpd+15DbgwMbg7armDbNsrlMiqVCg4ODuC6LoIgQKvVAvD+Wk0rrw8NhuDcU+jTdz6fR6VSQaVSwWKxQBzHmM/nGA6HGA6Ha6dpc0N52NiUJMzPkQRro7BlWUJsaArWZuPZbIb5fI4kSQBAVJvFYiElJ63SkDiR2PA9iRC/l1K5WY8GBl8PHkocx8H+/j48z8Ph4SEODw/hui729/clqmEymWAwGCAMwzWV9SHCEJx7iHSpIZfLoVQq4fDwEMvlUghOJpNBHMfilte5IwYPD9qLRfKrM2qKxaIoM2zhplFY+224CS6XSyRJgvl8jslkIrI2/286nYrEzY2S6xHAWmmK/0/lxsDA4OZg2zYqlQpqtRr+9Kc/oVar4ejoCIeHhygUCmLuT5IEcRzDcRycn5+LneGhwhCcewqefBeLBXK5nJQJeCMrFAoolUpIkgSz2Uy8Dzw18zEM7h82JQmT3JCkfMoobFmWfI5ERCssADCdTpEkiRAcqjUkOHw/n8/XylBaIdTKjDYzmjVnYHAz0KVmz/NQq9VQq9VQrVZRqVTgOI50NsZxjOVyiTAMEcexXNcP/Xo0BOceQpOb6XSK1WqFMAwxHA7hOA4ajQaKxSJ838fBwQHCMMTZ2RnCMEQURYiiaO2mlJ4XpN8bbA+0QrOpg47KTNpHw/f8HhqFSVKSJFn73HQ6XVNruM7S5ag0kfnUWtIfm3VlYHAzyOVy8H0flmXh2bNn+OGHH1CpVPDDDz+gXC7DcRzYto0oivD69WuMx2Ocnp7Kx1T6H7KiagjOPQZvNvQ0TKdT8VnwpM6WXtu2xT/BLpe0AqAf15iTtwtapUmXnrTRnMF7ruuuGYW1t0YrNiwbkeikS08sR13WM2PWjIHBtwF9dTzM1mo1lMtllEolOegA76/3IAgwHA4xGo0wHo+li/Ghq6qG4Nxj8NTN9ttsNovpdIpyuYzFYiE5CL7v4+joCEmSoNPpoFAoyA2NJa50SCDLWYBJRv5W+FSSsH5fLBbXSCsVGhIbmodJfAEgiiIhNPTDkMzo0hM/pwmQNgo/9M3QwOA+wbIsaQM/ODjA/v4+LMtCFEVIkkQITRiGePXqFaIoQrPZRKfTwXQ6xWQyMQqOwfaCBAcAgiAQUtJoNABAmDxD2ubzudz8qOTwc5rtr1artRqtUXNuH7rrSSszuvzEWrvruuKx4nue2AqFAgAgDEMEQYDZbCZrI0kS6WLS3UzaW/OQDYcGBg8JhUIBOzs7qNfrQnAWiwVGoxEmkwlevnyJ09NTxHGMTqeDJEkQBAGCIFg7tDxkGILzQMAywmQywXA4FOKjSxNsC+Ypn6d5GlLTbbwsZRncLDaVBnX3Ew3A/Lvw78ZsCyo36blP/N5MJiOqDcuSNA6nA76oAD7ksC8Dg/sONgnoPaBcLmN3dxfValUOOIwJiaII4/EYQRBI5xRjHR6T4d8QnAcCxt9PJhOEYYh8Pi9tgrZto16vo1AooFwuA3ifZxKGIebzuZQ4ZrMZer2eGJeZUmtO9TcDXW5KG4X1rCfP89ZmPenPpf1SmUwGtm2vhfRls1n0+33pmOj3+0iSRNq/de7FJrOwgYHBdsGyLOzv78NxHDx58gQnJyfwPE/eO46DQqGAKIpwcXGBwWCA169f4+zsDLPZTPZ6HmofCwzBeSDQeSPT6RSZTAalUgmlUkluaLyROo4jtdfFYiEngslkInH82gdi8PVIB++x/MT3VGOYTcOhljQIl8tlUdm0uXy5XH7UTcVyI83DrLfzzcDA4H4hl8vBcRz4vo9Go4GjoyO4roudnR0Ui0X5usViIeoNu2Z58H1MxIYwBOcBYzgcYrVawXVdzOdzOI4jb6vVSm6Y6Zuj9uM8xoviukiXnDSh0aWntFGYJahisSidEbpElclk1kYc8CTG8hK9V5SwOS2Y5a3HMjnYwOChgXtFuVzGkydPUKlU8PTpU5ycnMgIlWw2i8FggF6vh+FwiHfv3mEwGGA4HMq+8VjLz4bgPFCsViuMRiMEQQDbtjGbzeC6Lk5OTsSEXCwWxYdBszFvxsB6QJvBl6EJTdoorPNofN+XJGGdKGzb9lr7tu5iIomhn0a3dZPksNTIzYzGZACmBGVgcA9RLBZRKpVQr9dxfHyMRqOBJ0+e4Pj4GKvVSjqhWJIaDAZ4+/YtgiDAaDQSu8FjvfYNwXnA4MKm2ZRZJixh6XKJPvHr8gk7qB7rBXIVkMhQpdGlJ5YG+Z4fU8mhoVuPM6BKw8/priemC69WK1F/FovF2mvg+20vM6Zfn1lrBo8d3BNs20apVJJDkeu6AN5HP7BDcjKZoNPpoN/vYzQaIUkSGbHy2PduQ3AeAZh3ArwvW9GUyjZj3X1jWZb4cujVeSwthV8Ly7Lgui6KxSKq1aooNOxwoFGYb9o3xWwaPSyV3hk9MoEkNZ1RA3yIbuffzvM8zOfzrfVTpbN+dNqxNkEbGDwm0G9TKBSwt7eHJ0+eoFqt4ujoCOVyGbPZDKenpwjDEC9fvsR4PEan05FW8G63K56bx24xMATnEUDfSHnj1K3KVHC074Nv/BqDL4Pt2hxuR3nZ9/2NnVAknplMRrw1/PvQGMiTmCY4mnQCHzroGNrIv6NW4bYNadN1OmjysZ88DR4neD1wbpzv+9Is4rouHMfBZDLBeDzGaDRCu93GcDhc89zwgGRgCM6Dhja3WpYl71m6YPlDh8nVajX4vg8AohzoRGNzst4MmoO5CZVKJdi2LZ1QWiUj2WT7JqVmdjxwAjzr6zqYT0/m5vMy44aBf9VqFUEQCKHdRi+VHi9RKpXE8E5PmG5rfawGSYPHg0wmI6Vr13VxfHwM3/fx5MkTPHnyBK7rolqtwrIsnJ+f4+LiAsPhUPw2zL4x18s6DMF5wNBTpRnwp7t1eAOhRySbzaJWqwHA2o1mPB6vRf+b0/XHIMHhBsVhdxx4xzIMAAlknE6nUkPnKYzEJk1O6Mvhm1Y/9LgF3/dRrVbRbDZF7dHlrG2AnnzOjCbOSqNypf0DZsM2eAxwHEemgf/2t79FtVrFyckJjo+PZQ9nPlmz2US/38fFxYUklT/2ctQmGILzCKC7eWhYBSA3xWKxKEFxWs1xHAcA4Ps+crmclExMCeFjaDM3O500qdFfx2TR6XSKMAwxmUwkaZRp0umZULq7io/D9yQMAMSETHVum4iNBtcY/Uqu68rrjqIIQRDIz7utP4OBwU1Al6Vc14XneahUKqhUKtIVOZ1ORaEZDocyDXwb1dltgiE4jwBUF2ggHo/HmM/n8r5YLIq6s7e3J2rPwcEBJpMJHMeRQZ2dTkcIUvqm+9jBAXdJksjvlOoY/TY0E9M3w1JMHMeI4/gjtUUTyfTvmaRKJxgHQYBMJiMR7SQ62+bDYTmvVqvhp59+Qr1elynHbG/l5GOqOUbJMXho4OEzl8uhUqng6OgIjUYD33//Per1upAbDswcj8d4+fIl3r59iziO12YGGnwMQ3AeOPTCZysyFQbePHSEN5UHSqL06hQKBYzHYxnmqG+c5uL6YOTm70+bhzkjhoZhlpS034ak5zo3ct0ark3K20w+dTZQuVxGrVZbm4lm2zaSJNlak7SBwU2ABIdqJrsf2RYeRZF47Hq9HkajEYbDIcIwFF+eUdM/DUNwHjDo1eBpOEkSIS70f/Amy5tNNpuF4zgol8sol8vI5XLwfV88Erx5d7tdIUqcWfXYLzSasUlqaComseRpi6RSe2v4/irgYxeLRem2qFarKJVK6Ha7oh4B2/e34Tqk0uU4jvws2WwWpVJJNnb+/gyZNngoYLBnsVhErVaDbdv47rvv8OzZM4l3CIIArVYLFxcXotwEQYBut7vmuzHXxKdhCM4DhvYvDIdDGQlQKBTWyiO8QGzbluGN2WxWvDee5wGA5ORo3wgAeY7H3mFFUy9PVhx5AWCN4NwU2dC1e01wyuUyPM+T0Q3bRgw4E43khp1nxWJR1lqpVJJ2WHrHDAweCmzbRqVSge/7eP78OXzfF0NxNpsVC8G7d+/w4sULBEGAly9fynypOI7NNXEJGILzgLEpOI0qA0tSuozBDh4AiONYSlI8XfNmyiBA/djsunrsvpz0iAVC/15uktzo8pQu5zA7h/OqtkXBSU9C3zS7Kz0yxMDgIYBeSCqUtVoN5XIZu7u7cjjxPA/L5VKSisMwRBAEGI/HYsI3qs3lYQjOAwdvuAz3o0eE3VSaBE2nUwyHQxQKBaxWK8RxDMdxsLu7K3OU6OxnqzONofxePQCSj/vYoJOj05+/KVAF4VuxWJS/G09/7XYbURSJuXlb/haanKUnq+dyOURRZEiOwYMC1e9qtYpisYjnz5/j+fPnqFar+N3vfodyuSzdrEEQ4O3bt+j3+3j9+jVev34tDQzssDS4HAzBeQQgyeFpmXkJ6QtFqzD05ejZSDTDaVWH/8cbLlWhbSuLfGt8i01ID/bUMQDpdvWbVI5u4jVv+pwmOPxZiG143QYG14XOrKKRuFwuo16vo1arodFooFwuy9eFYYgkSRCGobwZ9eZ6MATnkeAyF8VqtZKLKI5jSd11XRdxHIuzHwA8z5M2aHa8zGYzWJa11vL8mMtVtw0qHJ7nodFooFAooN/vYzAYoN/vS2fWNp74SMTo9apWqzIQll1lzAoy68jgPoOk3XVd7OzsoFQq4fnz5/jpp59QLBYBAGEYotvtotvtYjgc4v/+7/8wGAzQbrfXUr3NNXA1GILziPCli0OHAJLczGYzFItFST9ml4vneeLJKRaLkslAYsPOqm28uT4EaL+K67qo1+tYLBYYDAZIkkTes/V820BFKZPJoFwuy3gJKoia4GyL+mRgcB1wn3QcBzs7O6jVakJw5vM5+v2+5Nz88ssvGI/HODs7QxRF8n9m/V8PhuAYfARdRtA3Fz2ugf/HzixmlwCQ1l4A5tRxS9AD+TgslSVGPXxzm8pTwIfkZbbRW5YF4MPaooJDmAwcg/sIjiLJZrNwXReu66JSqaBarYqZmPtmHMcIw3At44adUtonaXB1GIJjsAbdzQKsT33mDXSxWIhngqnH9ORQueHgSDO/6mbBv0U+n0e5XIbrutL2z8RTbo56FtW2gOuAp9lyuYzVaiWxAxwcyPJVej0aGNwHFAoFac5oNBqo1+uoVCr4+eefUalUcHh4CNd1MZlM0Gw20el08M9//hO//vqr5Jbx+jW4PgzBMQDw8Ul5Uzsvc3VIdgBIFw/wflgcSQ8NygxpM+TmZsHfM3/XwAeTuCY220RuNNj5xXVCBUd34BkY3FdwfArnSjUaDVFwyuWyZI6xJZxjScbjMWazmaSeG3wdDMF5ZEhnj6Q/p4kLkzbZ1cJwwGw2izAMkclkUCqVUCqVsFqtpFxSq9XW4sUnk4nceA3R+TrQsMjBfOVyWaYMLxYLGVTJidzbViKk8sduPBK0fD7/0Tq86ewgA4PbBtd1uVzGkydPUK1W8dNPP+G3v/2t7Km5XE4mgQ+HQ7x58waDwQDj8XjtMPjYO1FvAobgPDKwXTH9ni3GDALkCZukhXOVWBPm1wGQm1ShUJDp0ACk3RFYNzCbi/b6YG1fj2cgwWEw2Hg8ltTkbVLP9GBBrheSG90aTjUn7SMyMNh2sKvU9308efIEe3t7+POf/4z/+I//wGw2k2yqX3/9Fb/++ivCMMSbN2/Eh6PN9yT7Zu1fH4bgPHCkS0+8uehUTb4x64b+B95kdKsxL0B6bUajkRhGPc+TYZyWZUkHFm9SJEfblKx7n5BOL+YEeP7tWKKierNtv2Odd8MbAQMkuSb1Wtk2/5CBwafAQ2CxWITruvA8D47jwHEcKcPOZjMMh0MEQYDBYIAgCJAkyZpq7jjOmvdm2xTY+wZDcK6BTYbHbVuEOlyK/wawdlPxPG8t5p/5NyQ1bNPl8Eh9oma5qdvtwrIseJ6Hp0+fwnVd5HI5lEol5PN5RFEkKbssc3GCNrB9v7dthTYXMyyMHRmFQkFKgkEQyGDVbVM+stmskLJqtYqdnR00Gg2USiU4joNMJoPJZCIt4kzI3jaiZmCgQb9NPp9Ho9GQt729Pezs7CCfz2M8HqPX6+Gvf/0r2u02Li4u0Gw2hdzkcjlUKhXYti0z/pIkwWQykcOkwdVhCM41kW6l3hY5MW0OTvtsWBrgCZoKAAkOO6TYLcVUYl2i0gRnMpkI+WHEPkkOb8bMxeF07clk8ugHc14VmrDq3y3ziai2sU18G1Uylqj42l3XlbBIGqW5xvRa26afwcAgDZaNmfDueZ4oODQTTyYTRFGEXq+HTqcj7eA6W0w3DBQKBcxmMzOq5CthCM4VwJsM1QjecCg/ps1hm24w6c9x0//U9+h/65bttDKzyShMtUZ3NunOm7SCwzRigiqLHu2QNn/SLxGGIc7OzlAsFqUtMpfLYWdnBwDQ7/fR6/WQJImUuGg81q/Z4GPo0hSVm3K5LBvoeDyWHA1tLt4m8GegbF8ul6VNPJPJYDabIQgCdDodMaYbed5gG8G9nx5F27axt7cHx3FkxhSng3ueh9FohFarhW63i19++QXdbhdRFGE8HsP3fRwcHMDzPNnrgyCQwyKAre+I3GYYgnNJaGJRKBTkJMr5SzxFa9KTDmnaRGD4eGwZ1ItYT+rWpEZPjiap0bOIWL/VJlTf94XM8PWT/PB7WeJI13/TQzn1e4IGV2bhsMPq6dOncBxHurFYdqAMa25ilwOVGxKcSqUC13VhWRbm87nU9klwtnEz5OvnaIZGowHXdUUdHI/HaLVa6Pf7QnC28ecweLxIHzZKpRJ838fTp09RKpXw888/4+eff5YmgGw2i3/84x9CbP73f/8Xg8FA9vV8Pi/XAh+XIxqolHNszpcUWdN19TEMwbkCqLaw3kpZcrlcSilGEwCtcuiP+f9MAubNX38vSQUVEq3eaFMmiY7uhOLjuK4rxIYZNSQzOsuGZSlmL+ihbpvadD91EaUHPWo1gaUrvi5gfbin/t0YrCNd2mFpUQ9P5WBNlhm3qTylS2u2ba+VpjjhPpPJyLgPTW625WcwMAAg+y/3s0qlglKphN3dXVSrVfi+D+D9ga/f72O5XIoqORgMEMcxJpOJ7He8bqfTqdwH2EDgeR6iKJJ9/VOvJ21LMPvoBxiCc0noDXpnZ0fKP/RAkGhwsWrwZq8JAz0TlmWhXC5L+YiptOPxWHwI2pBLaZR1W77XxIXkRyfB8nXqSH+afVkqmkwmGAwGkirLsttVbzJRFKHb7WI6naJer2O5XMKyLOzt7UnbeBzH6HQ68vNNJhP5Gc1NbR1UPmjOpRzO7ow4jjEYDKQtn8bwbYEeNri7uytTlB3HAQAMBgNMp1O02210u12MRiMxuG9bqc3gcYL7aT6fF+X06OgIz549Q6PRwF/+8hfUajXxwIVhiNevX2M8HuP09BSvX79GFEXodDqI41j2uXw+L2pNo9GQsvPe3p4kHQ8GA1E5dQu53uvT1QGzj76HIThXABe5Hk9g27Z8LpvNYjKZyM2aizCOY+Tz+TU1hOSF5SM+ZrFYxGw2E1+CDmvjguXgNl1m0u3e+XxeWocJ3YYLQNrASb70e95YrnsSmM/nmEwmKBQKmEwma4ZS4IOCo011Oi3Z4AP4e9EKDkt+Ol2af7ttNObq/BteN8xX4imW3VN828afw+DxQlsUqN4zaLNaraJer6Ner8twTJKZ4XCITqcjw2/pc+S6Zmoxu1f5HMViUe4PPKRuek38P62ApxtgHjMMwbkkeCPhnBCqNHTJc4GxbKXhuq4QDp0pA7y/yddqtbX8hNlsJqfaKIoQRdEaM0/P6NE5KLrDiTc8TY74NVSJqNywnMQZUtpsfNXf03Q6lVROBlvt7OysDZ9jl0Amk0GSJOh2u5J4vE3qw11Cn9Jc14XjOKjVatjd3ZUTHc253EB1eXEbwNfP0lSlUhH/UKFQEN9XGIZiRKd6aAiOwTaBRJ2HtXK5jFqthkKhgLOzMzSbTbRaLbTbbQRBgBcvXmA8HqPf76Pf768FpRKTyQTv3r3DaDSCZVloNBprlgY+J7PJdFer3v91g4muFDx2GIJzSXDBTKdTMXOSuJDgsBRk27YQIgByc2K3Ec1l9N80Gg0Ui0WUy2WUSiUJhKI82e/3AWCN9fNC0RcMVRem2DJLhF1e2v/DWH+qLexuuolxClSB5vM5Wq3WWnmFvw/+PguFgpA4XtjbdIO+S2j1hubcWq2Gvb09TCYTkbtJcCiPb0sNnpuubp8tl8vSAcYuEb7+fr8vxF6vVwODu0b6sOH7vsyWWi6XODs7w3Q6xfn5Oc7Pz9cSiqmubtrTJpMJzs/PYVkWdnd31+I5tNr+qSHIWrknATIDjj/AEJwrQJeXgPdek+FwKB4YtlqzQ4jQacAMYNMdWLZtY7FYiIoDQE64k8lE/Dg0X5JAaLVG54bo5+Lz6fot8CEOP932fZPQ3pogCNDr9cRrxNeRVqJMmeoD+Pei6uV5nhiM+btlV9q2dhyls3uo5NBgTKKtx0sY343BtoF7vx4jwutvPp9LRMNoNFrzkF1GUeX+y4Mmmz6YD8WMsk2vh9Ddu7rZ5LGTHENwrghuyPTbtFqtNS8MoX059MRoKVErOMPhELZtYzKZSPvh3t6elLo4QLHVaiGKIuk20YSJ3SfAh5PzJv8On5egmnMbQw3n8zn6/b78rjqdjrRUuq679rzp34/Be1DlYirq7u4uSqUS5vM5ut2u1Pe3rTQFrHvW2FK7u7uLw8NDlEolCUB78+YN3r17h06nY9rDDbYSWsGnKt5qtaTUzuaMi4sLtFqttcT2z12T3HMXiwWCIMDFxcVawwnVbx520nu0TkJmjAhjQHSTyGO9ngzBuSK4GMni4ziWxcRaKWc6kfToDBsdrKej62ezGaIoQpIka6mYbKdlunCSJEJy5vO5MH6qNbqja9ONguqRfk23xfS5IWQyGfHkUGGiCvHYTxiXAUtU7N6gMZsnyG31q2iVjsbM9HweHhjYAWbIjcG2Ih2DwfIw/YyTyQRhGCIMw7W4hss8LgC5n9DmwGuGh79N35dWcdL7/2M/MBqCcw1owy+RJMmaNMguqLQJLF2SKRQKGI/HsCxLLhjf9xFFETzPAwAcHx+jUqmIFHp2diYXlm4lT48/2EQedJlKJxTfJlhCo0J1cXEB27Zlo0iSRG5w1zU3PxToujpPcaVSSeZOLRYLNJtNdLvdjwb2XYfkpDM0gJup3evN1vM87OzsoF6vw/d9OI4jf/der4fxeCxEzRBeg20GfZRsMqFKw86/MAyvPOiWZS9eE/l8Hr7vw7Ztue5Zxk0r7npQJ/BhmLJ+7MecM2YIzjWQXrhp34C+UfDf6ZuHJjxsmR2NRhgMBiiVSphMJiiVSvjuu+/w/PlzMeFymOLp6ank11BN4gL+1ELWBjXttP8WNxRewMvlEs1mE4VCQZ6XuRFmqNx7sKRJYy43uUqlgjiOJfZ9MBhce7Bm2qyoP3cTc8K098b3fSE4nNFDQzEJDj0Lt+UHMzC4CSyXS0RRJPEfbAAhWGK9LLSvhwSH5Vx9sMnlcuj1emsBsHw9fBxg/X5CjybweGM4DMG5BWxSTdIf6/eUMmlSWy6X6Pf7mM1mKJfL8H0f8/lc8hEajQaePn0qw9tYqtCE5TIL+VvKl+kLWZMw0zHzAXroXrlclnwNnfrL2VOXNTFuAv/2LJcC6+v0a0iOVicLhYJk9+hyK0+97DDZRg+RgUEa3Md0ppjeR69jkOfjaUWbOWKr1WotuT6dXJw+qNLPqDN7+Fo3VR7uAzbl+qR/D5+CITjfCJ9aVGTZmUwG3W5XDMdRFMFxHHQ6Hbx58wau6+Lk5ATVahV//vOf8ac//Qn9fh///d//jX6/j3a7jXa7LReJbhXc9BpYPgC+zTwoyqosp2nVQLc2Plbwgi0UCtJG/Zvf/AY//PADisUiSqUSAGA4HOLVq1fSWk2ic1X1hmUwHWugjZR8zKuuDT4uS2sk5AcHB6jX63AcB4VCAdPpFL1eD/1+f22OllFvDLYdWj1Jq+JXXbu65BSGIXq9npShoyjCZDKB4zji8dz0vcCH/ZXkhg0stEro17yNe21aRdaf12OIuB+R9H0JhuBsAXhh6PwYBrfxxFupVHB4eIhcLodSqSRhUwyGoo9HR3pvUkV0iexTi+q2f04tm+r/M/gwfJW5SAz1Y6soYwPSCs5VoXM9eFLk5sE2VW7g11GHqETpAEumfvNEqQMmjYJjcN9wU+V9Pg6bRthRa1mWNIWkc3D09wLrXhvtf6OZX+fkbAs+VybXn9ODpLkfsYX+SzAEZwvBlsE4jqXmWyqVhNzs7+9jf38f2WwW//Zv/4bpdIoXL17g9PQUQRDg9PRUzG6byI4mOVpy/VY3F3MT+xhaUiYhKJVKqFQqEgYWxzEuLi4k4I/1/usQED06geoQiVLazHiVtZHL5WBZFjzPw97eHkqlEk5OTmSqPFtewzCUIYQ8qW7jydLA4FuBXkR233Jsz3Q6lXIvs9PSzRjab6MPkI7jyCDldMv4t9yH9UGK96RKpYJyubz2WgqFgvz8uguZexD3CH7+i897ez+SwXVBgpPJvB9jMBwOJTeGfhxGhf/xj3+E4zjY2dlBpVJBu93GcDiU4CgAH9VoNesH7o5wGKKzDp1aTIJTLpcxHA7R6/XQ7XaF4HDju67qoWdb+b7/kQmeadk62foyz0OCowdrnpyc4MmTJ7IpkbTTKE2DsRnRYfCYQW+i53kSijmdTiVahCRhU6epJj2caaWHQevW9bswGzMiQkeD1Go1nJycCOlZLpeS5m9ZFiqVipCi9Ou9bCnbEJwtBhkr50WxNZzdJ6zbAu8ZMmPDj4+P4bouBoMBhsPhWl7OpjwFg+2ANmLPZjOZz8R0VLbSp2MBrgp98tGZTKzxMyhMewwuC4aUkYCXy2XJ79FlWBopje/GwOA9eCPn9T+bzZDNZmU+IZsPPpWJo0fdsBzF64oEif+n1ZCbRrFYRLFYXFNhisXiRzMaj4+PcXR0tEa4ONaoWCyiWq1KXhqJmY44uczrNwRny0FvwmQykVBBpmnWajVkMhlUKhXYto0ff/wRSZLg+PgYcRzj73//O/7xj38gjmN0Oh1pw9btwPp0bm4ydwf+LRj4CACvXr2SeWKvXr3CeDwWxeNr26m1B4d5O77vS2gly1RXITiUxDkv67vvvsPe3h4ODg7geR6SJEGz2cR4PEa73Ua325WBr2b9GTx2kKTMZjOMRiMUCgXUajUcHR3B9328fv16LbFeQ2fj8NCgvSvsYAQgasltHCwymQx2dnZweHgI3/fx7NkzGT8EvCc/tVpNBos2Gg1RqzRxs20bu7u7KBaLktzP/Qp4r3Zt+j2kYQjOloNMm6w+k8lIyQIA+v0+Mpn3Qz5p5Mzn85hMJmi323j79i1yuRyCIPiopVA/h7m53D3492HNfTweS/2cYXi6LHUTfzNtNOapS48WuSq4WeksDz3GhEQ9SZK1oZpm/Rk8dmiPDG/gDMqMomhNwdlUZkqrOLp8zetPj3S4adBH6DiO5PccHR2hUqkIASsWi9jf34dt26jVaqhWq+I7zOfzok7bto29vT0Ui0WEYSjZQ2y6oW/vSzAE555htVphOBwCeE9uVquVMOUoiuQk7nkefvzxRzQaDfT7ffz1r3+VklUQBJJHo6U/c5O5eywWC4zHY4mC7/f7mE6nMmWbdfqv+VuRNHNDKhaLMi8ql8tJ5s5loJWgXC4nGU27u7vY2dlBrVZDNpvFeDzGcDhEs9lEv99Hr9f7aDyDWX8Gjx30o4RhiEKhgMViIcZbx3Ek5V6PBUp/P/d1PeqBJuXZbCaPe5VxEl8C4y2KxSK+//57/P73v0e9Xscf//hHSeFnyY0l8Hq9LrYKNk3w50mSBK1WSw54HGC9qUT1L//yL598XYbg3EOQqNi2jTAMRf5frVZi0nIcBz/++CP++Mc/otlsYj6f4+LiAm/evJH6LqXKdJaDwd2BBnOeuGjK1Rf/TUDX6umZ0fNv+NyfIzq6HZVlrkajgSdPnoj8zPESYRhiNBrJmAkSHJqLjYpo8NihPTjj8Ri5XA7z+XwtbkHPo9uk5LLMvVgsZMYhu6/y+bxEjzAfR6v610Umk5GSk+/7+O677/D73/8eu7u7+Mtf/oJ6vb4WCRGGIZbLpYSYRlGEN2/eYDKZyJ4ym83WDuI82HGv0OWqz8EQnHsKXghJkgAAer0ezs/PEUUR6vW6hD7R3FUqlaQNcTweS4lgm3IRDD7IzOnP3fTNn6ZidlLpKfObElM/Ba0CMTSQJkHW/1mWGo1GCIJAIhD0tHsDAwOslZRYJmYjAEmKnnm46ft1s4I2FfOeQGLwNVlXhN4nqChxT+C1zzlaYRhK8wQDaTk4mjEYem4WD+AsaWu/6GUVZkNwviH0H0WXha5LMmazGXq9HrLZLEajEf75z39iZ2cH0+kUOzs7+P7778Vc9pvf/AZHR0fi0xmNRsLgTXvu9iD997huQurnQJnYsiz4vo96vS6lqdVqtdZZ9bmNhJsv04pd18Xe3h52dnbgeZ5soJ1OB2/fvkWv18OLFy8kefuqU5cNDB4y9HXOshQPDHEcw3EcOI6DfD7/2TmCLD/PZjMpAVPFYVcWVRHgg+n4a197evBzkiQ4OztDq9VCs9lEs9lcIzg8VE2nUynFk8wAH3xDw+EQw+EQ2WxWzMiMyVitVviv//qvT74uQ3C+Eb6U2nidDZ71VgDS+bJcLtHtdpHNZrG3tyfZCL7vo1AooFqtolwuy6L/0k3M4NsjLTnfNHjCIqnhcL701+iPP/U6dImKKg49PSyvxXGM4XAoCg4VRMrNBgYGH6CvCa208r3OrPrcY+gyNGcZ6kTjT6UjXxX69WiSM51ORa3tdDq4uLiQLlF97XO+HisMYRjKISyTyaDX62EwGCCXy4kZOUkSIUKfgyE43wA6pdb3fWHgXIBRFMkp9rpMmt8XBAH+/ve/w/d9jEYj9Ho9uK6Lg4MDMZKWy2XM53O5EbH2aW42Dxt6HXKqN09E7NxgZpL2xnwKuq2VbeXj8Vg+Ho1GKBaLaLVaePXqFYbDIdrt9lpKt/HeGBisYzqd4uLiAlEU4fe//z2q1ar4KzltvFgsSpnnU/cMBv5pVZixEMvlUg4hADb6eS4DXU6iMnN6eir3FoYWUsFh5IluU6faw/8jcaHixC6qbDaLIAiEsG0KPEzDEJxvAN5YqKBwcXKB6oV6ldRYDXpygiDA//t//w/ZbBaDwQCdTgf7+/uoVCoyv6pcLmM6nUr5igmz5kbzsKHXoed5cF1Xkk65yTCEjyb0yxCc+XyOOI6F4DDHx3VdTKdTITij0QjtdhtRFJmuKQODT2A2m6HdbmM0GmE6naJSqUi3bBzHYjRmGepT0KSFSr8eUsnHWCwWUqa6Tvif7vyaTCZ49eoVoigSMgVAktj5HNorxI9pkCYxo8JERQiAtLhf9j5pCM43gJYZWRJgqx7ltvTU16/Z/LlooihCv99HNpvF2dmZGDw3zf4xN5uHD8rRehAmy5TceHSXgp7kC2we7aFVHEbC09/FNR8EgZAm47cxMPgyeN2xzMPEen0Yvcx1lM7G4ccMAqTnTjcWXNcuwedgbhfz2bgvULnhfWdT0OymcRK6HZ5702VfoyE43wAMP3NdV6LrSXQovSVJgiAIPvpjXvdmsFqt5BRg2zbOzs4kXKlSqUiLrikRPB6wPFUsFmXQne/7sG1bCEgcx1gsFmuEPJvNCvkBPt6Y5vO5tLWen59L5gaNxq9fv0a/35c2UbPmDAw+DYbl2baN4XCIX375RRpCtOrypcOCJg080OouKsuyZBQCs3Ouq6xq9aXf70tIKVUYmpo/ZcNIEx1iE5kzBGfLkFZw6JAvFosA3sdSL5fLtTkjN2E0ZfYAE3AtyxL3OmueWh40N52HjU0Kjl5zetPURmSuDbZ+pgdzAh86MaIowmg0kjTtTCYjBkI9uNPAwGAztHLKwwOJBz9/lf2aJEcTI50joxWcr4Euh1HRJcHR5EkTnC+RlU+1wl8WhuDcMriQWA7QQUb03nDhFotFlEolkfnSkdvXvTnQQLpcLtHv9wFAaqZmivPjgvbglEolISI8YbFWb1nW2lA+3f5JFSZ94uPw13a7LR1awPs2T+beGO+NgcHnoRtP8vk8Go0GJpOJEB2OLKBq8iXw63S6sY6B4D2Kfpev8YECWAuP5cefuu5vey8wBOcWsSnplQuJBEefanlTYY2SLDhtyLoq0qyatdB0OqTBw0a6m4/DNdm5QMUPwFrwH8tQJNtcN3pD5NpkJwWfC4AQqK/pEjQweCygEkq1vdFoIIoiXFxcyHXHTtzLqC4kOCRG7KLSM+d4X9KhgNfFNlUDDMG5RejFo93rvGHQk0AVRRMYvfjIgPXU5assIP31vEml3esGDxuMU3ccR3JvcrmcrAWqN7pjAfiQdqpTjqneAPhoDZEwa8k7vd4MDAw+D14/k8kEQRBgOp3KQVkP3LwKdNcj8F7FZwlapxo/pNmEhuDcEuhzoMHYtm1hzkxzHI/Ha/kBXLS5XA6lUklOyJwdFQTB2qh74PISHxcvT+C8SZmbzsOGrueXy2Xs7u6iXC6LWsh1GMexdDtpI3B6qjjzMzQpItnRCaYaRrkxMLg89PDaXq+HX375Rbyb9Xodrut+dF1eFmwk4LBO3UmVJjoPAYbg3BJIVPQcEW0g1sSFNwkdka/njujHSxOb69RJb6JLy+D+QOff0NxOSVoPs+N61GtCb6Jp0yAH/vH/qDReJktDb6BmDRoYrIPXE9vEbdtGvV4X4nPdFGLee/R7XSm4CbPxNsEQnFsCSQpvKp7nAYCoJ3Eci3LDjiZNiIAPZSq+ZwLxdDoVH40+bX/uRvHQFq7B58G/dS6Xkynhvu9LTAFPanEco9vtIkkSDIdDTKfTNSJDgqPXpf6YnR5ci5S/N6k2LJPpx9dm5esGjRkYPCSQfPDAwGt5MpmIqZjt3fo6ugz04ZbXrbZO6JEtD8HCYAjOLUA70zXBYZmJQ8VoJGNGgVZwuHhLpRJs2wYAcc4zCl+33X3JgPwQ2bnB58E2b5ZH2TlFkrFavQ+DHAwGSJJEklP1YD6uR2C93KXbWHO5nJRYAaypOmk1SHsI9IgInaYKGFXH4PFCt3UzloGNIvx/Ehxen5dtQOG9ggMr9bUNrAcJap/dfYUhOLcEOtZ1UixPudrMuSnNUcdZs4SliUk+n4fjOGvZBjpa/0sLko/1UIxkBh8j3cHHLj4SDN0RpWdPcU3xMXTuDfBhA0zncuhaPjs8gPUyaC6Xg+M4a6WyxWIhXrQ4jgFgbf0bGDxmcG+n+snrT5uMvyY+RHfp6vwr3VF1n1UcQ3BuCblcDp7nyTBD3kTG47EYvZIkWZPyedPQ0iRNyLlcDsViUSaq1ut1zOdzUYHG4zHCMFxTczYtTH0z0uUAQ3YeFui5KRaLUpaq1WoolUpYLpdCKvr9PgaDAWazmZRKuaFSheRjcYMlSQIghIZhlfpEyTIq8H4TdhwHe3t78DwP+/v72NvbQ5IkePv2LaIoQrfbRa/XE7Kjy10GBo8FWmlPkgSDwQCe56FSqaBYLMoMQV6D14EuDesAUF2uSt9H7iPRMQTnlqBj8SnlU4an2rLJb6AXEfNytBTJUzONoiwNUCniY+j36bKUPqHr/zMk5+FAJxFzsi9Lplx/zL9JK4r6MQB8dKLjuqNRGcCaP4cKkV6PVJM4sqRarWJ3d1e8aIVCAXEcywybyWQiz8fHMDB4bGDXLfd52hj0Hn4dpKMdtIKjCU+6o+q+XYeG4NwwXNeF4zhwHEcUHJ0/EEWR3FQ+B73gtMrDm8Z0OhUCxVO2bduYzWaiEun5QZuITNqXc98Wr8FmsDzqOA5830etVpOZU1QPO50OoihaWyubTmjpiHVdiqInjM/J058Oq9TfzxlY1WoVx8fH+O1vf4vpdIpqtYooivD69Wv4vi+hZpyPxc4uXbYya9XgoUKrJSwhc6yP7obkIeMqHhwNnTKsY0r4Pp/Pr1UU7uP1ZwjODcNxHNRqNRSLRXieB8uyRPpnBHcYhl9cIJsWEp31zNFxHAcHBweSswNAnouTaNM3rTTzN6bjhwUSVhIcz/NQq9VQLpdl7cRxjF6vh+FwiCAIhECku550GZMbHDdTEmotbWsVh5uhDpZkuaxWq+Hk5AQ//PADFosFdnd3kSQJXNdFLpcTs3MQBGu+HL43mToGDx36UKyTwHmt2bYt4X88dFznOfQ1rqsEmuAA69cfv/c+wBCcGwJZr2VZcF0Xruui0WjAdV0EQYAgCLBYLKR2et2WWG1C5s2KNxyeoB3H+Sh3Z5Np05SkHh505k2xWJTyVKFQWMteYsvpZY3pmgBp8qKlc63m6CCytAKpjZLA+/Way+VQrVaxs7MDy7IwHA5hWZZMKV8sFjLpXCuTZv0aPEToda1LvDoVX5eIv/a50s0tm7ontdp/X+4dhuDcAHT7a6VSwf7+PhqNBv793/8dlUoFL1++xOnpKWzbRqfTAQDxP1zHuMUTMTusmJFTKpVQKBSwv78vKZiDwUBMmzo3ATCt4w8NJDa5XA6+72NnZwelUklMxoPBAKPRCKPRCEEQiFJyGXk7bTSkbD2bzYRM6xOmTjxmNyBfH/1iepggAElqDYIAjUYDQRDg4uICzWYTSZKg2+3KdRPH8SeJu4HBfYc+FJBgUJH1PA+u60p37deu/03NLbQ+0HOnr/105+82wxCcGwJPrJz3wxtMvV7HcDhEr9fDYrGAbduSe5MebHbZxaJJESV8Pq+eXM78E5o2te/G4GGC67BYLMJxHNmouEnpqeHae3NZpNUYrieuZ576qNQAm03uJCfsymLm02r1PuOD6a2TyQRxHCOfz2M8HgOAqEbpzdbA4CEh3XCilZu02fhrn0e/1+b+tGpz3+4fhuDcEOgNyOfzUqJisNof/vAH/Pjjj2i326hUKmi32zg9PcXp6ekaM9Ylg6uSHq3S8JQ8n8/hOA6y2axMib5q8qXB9kP7blzXhWVZqFQqqNfr0tq9WCwQhiF6vR7CMJRBr9fxs3Dt6CAwknbLstZeFzdkmiT7/T6WyyUqlQocx4Hrutjb25NuQwYRAu9VzlKphFqthjAM4XkexuMxut0uut3uWkzCfW1jNTDYBKo2LM1yLy+Xy6LgJEki5WftpbkOdEOL7vbV5WYdL3KdeYh3AUNwbgAkJCQ4lBIZjb+/v4+DgwNcXFwgn8/j3bt3yGQy6Ha7a4syiiKR3vm4+v3nQD8Ojcj5fF5UnVwuJ49L1n8f2bjBZmijr+u60q1Uq9XWZk6R4MRxLITkazYnvdElSbLWdaEzdHReB1OTacAvlUooFovwfV828OVyiVKphMViIb6c4XCIbDaL4XCIXC4nPiKaL+/LhmtgcBnwGuIon/l8Dtd1US6X5VphHk6hUBBP5tdAHxB4H6HCqmMn9P9v+7VmCM4NgZtsGIbodDrI5/PodDqSQ+L7PubzOUqlEnZ2dnByciIBazyFhmEo86n4uXRwn4bOK7AsC7Ztr4170POB0h4Kg4cDbUBkjd7zPDiOI2UpbpT8+CbXgTYopo3H6XIWlcTBYIBWq4U4jlEul5EkCarVqnwd17HruqhUKshms2JATpJE4hbYZaLbyW/65zMw+NbQ65fXF1vGddp4+mtv8vnTGTg6C0s3EGzzdWYIzg2Am+piscDp6SmazSb29/dhWRYODg7w/PlzjEYjZLNZnJyc4Pj4GCcnJ/jP//xPjEYj/PrrrwiCAIPBAIPBAOPxGGdnZ0J2SHS4gWvZkIbNarWKWq0G4EOXy2QyEaK0ieSkI78N7he0UkI15NmzZxKit7OzgzAM8eLFCwRBgE6nI23hNzXUUm++2hdDkq3fA8B4PJbU7bdv36JUKqHX66FSqeDJkyd4+vQpHMfB7u6upLbu7+8jiiJUKhWEYYidnR1Uq1WEYYjXr19Lno8+HJBsmeGdBvcRWpXkIaXX6yGbzSKOYzkA3NZYE6qz3Fv0/DmakXXa/rYSHUNwbgj8A1OFyeVyaLVa4iuoVquwbRu7u7tSN61UKkJoBoOBdMAUCgV0Oh2ZFUQywoWklRsudLYF61Ze3ng2jW/YxsVocDVwHXANMKKgVCqJF4flqDAMJTjvuhtimgjrx+D60i3c3AC170urLFEUYTKZoFqtYjqdirqpgyipSubzeSRJAsdxpBOM7eR6o2dnIV+fKVsZ3GdodXQymYh3Lv01t6HgaBN/+r6jW8i3OSjWEJxbwng8xt///necnZ3h/Pwcp6en8DwPx8fHYkKmAfhf//VfsVqt0Ol00G63EUURnj17hiiKcHZ2hnfv3iGOY3S73bVwJ5amaEYbj8ei9NCcFkWR3FAArIU3pS+M+1BTNXgP/u1zuRwqlQrK5bIoN5VKBYvFAq1WC4PBAO12W4g0lcbL/p31uIa02qeznHTtnhsjDcZcozo8jENnF4sFfvnlF9i2jfF4jCAIUKlUMJlMJPqA87M4qPP4+FhKVfv7+4jjGK1WC61WC0mS4Pz8XMpYYRgK8THk3uA+YbVaIQgCadVm3tlisUCSJNIZe5sqvL530O+jE5ABrKk523ZtGYJzSwjDEP/4xz+QzWbx+vVr7O7uolQq4fvvv0epVJJSVa1Wwx/+8AdUKhU0m000m01MJhP0+33EcYz/+Z//AQAEQSBkhSdaTXDm87lcDLPZDMvlEnEci7l4k2GZC1I76A3uBxgqWSgUUK1Wsb+/j2q1ikajgVKphHa7jXa7jeFwiE6ng9FohCiK1m70l30e4MOMKf05LU3rkx43PHZvUV3SoJ8gDEM0m03ZzIfDIRqNBvL5POr1Ovb390WpZHIy83LoW5vNZjg7O8ObN28wGo2Qz+cRBAF6vZ740HT3h1nnBvcFLL3ati1RDCQ4VCpvs1lEe9r0gM+0asqv3bZryxCcWwQ30yRJMB6PsVgsRKEhG57NZmi1WpL3USgUAEC6S+r1OnZ2dpDP59FsNmUx6TRZLiodvqZrs+lSwqZIfoP7A5almFLMSIJisSgEl+SBpIY+rquSG60UstWb389p4WllRBMdPaoBwEfStp61liSJeNVardbaqBHLsuD7/toJkqSnUCigXC5LW/xoNJLRJdq7xt/DNnsGDAw0OCCX61nv9fqAeltr+VNl3nQzgA4J3KbryhCcW8ZqtcJoNBJjWLPZRKFQwNHREY6Pj1GtViUfh6qObduoVqtiLiO56Xa7AD6werb/phc6Cc4mMpMmRPycme9zP8ANxbZt1Go1uK6Lg4MDHB8fi09lMpmg3W7j3bt3CMNQ1t9Vc29yuZyER5bLZVQqFVmTXNfAhzWVnlfDUilLVQwg1O2nLFUB7zNyoiiS8Sb82Q4ODuD7Pp4+fQrf9+WNuT+ZTAaO42Bvbw9RFGFnZwfj8RjNZhMXFxcIggD//Oc/pWTF8SYmBdlg20HivrOzg2w2KwdchrmmD7k3Dd5PtAeU0Q+6TKX3lm1SSQ3B+QbgqTGTySBJEul+4g2J3SHlchmHh4dyOs9kMhJ0FsexnFY1sWFLrp74zBPrJugb0Tabwww+hjb5ceCebdsSmAd8WGtxHIuxmL6bq5JYemZ4gvQ8D8vlUnI3kiT5KK1YqzifqtFroyLfuG6psrTbbSlJZTIZxHEsBIvEJj3kk0Z7Du4kecrlcvA8T8LL2Kq+TRuxgUEaLMfati33A2A9QRy4/dLQpuaUdJMLic+2qTiG4HxDaK/CcDjEYrHAYDDAarWC53kIggDNZlPaZC3LQrvdxmg0khbYdEmA7JqP/6XFrv9vky/HYPugO5Ecx5FRIHt7eyiVSqhUKrAsC9PpVIL0ut0uBoOBhOFd1VjMeTS7u7vwPA+Hh4c4PDzEYrGQVvN8Pi+dflQNgQ/rSZ/u+P869I+G43RCKj1oTOMejUYolUqYzWbwfR+7u7toNBrSQq4DzzzPw8HBAWazGVzXRb1ex2AwQCaTwXA4RKvVQrvdxmw2QxAEQv7M8E6DbcR0OsV4PEaj0UC9Xsfh4SHG4zFarZaQC3rdbhsMky0UCjJ7kSoS0/I/1cByVzAE5xuDm/5gMMBwOESxWEQURXAcB+12G69evUKpVMLz58/heZ7coEhwtIlTP95VsA0Lz+Bq0EZbkpr9/X0ZpFkoFDCZTDAYDCTzpt/vr5GLy4KEigSnWq3iu+++w3fffYfpdIputysmx/F4LGqRJglpHw5VRb0Rk9howyLfSNw5k4rXgud52N/fF3LHwaLValVSnH3fBwAxXw+HQ6xWK/T7ffEyJEkiZEp7iQhzjRhsAxgEu1gs0Gg0cHh4iGazKcSG5d906/htgGGhvDZYieB+wYYWXSG46+vIEJw7BDd8LprRaIRcLid5H47jyA1qMBggiiI5jRs8Hugp4RwD4rquSNcA5EbNSdu6HfyqmwxLPwwP5NiHRqMhag1LZOkW1U0kgeSFpVU+h5a5eS3wsXSIGMtMo9FI5l3lcjmZVUVPDUtqnGtFkkZPWzabFeM153Hx98rXm+4MMzC4S+hRJ4w+0J2QV20c+BqkDyHplHJ9UNkGcgMYgnPnmM/n6Pf7yGazGI1GuLi4gGVZODs7g2VZa51Y7Lb62pkjBvcH3Dg4u2lvb09Mtzs7O3AcR6bVB0GAdruNIAik/HKdzY+dWfV6Hc+ePcPBwQF+/vln/PTTT4iiCLZtYzAY4Pz8fC2lOA0+t27T5qnTsiyZccOJ57pMpcMp6ZuZzWYoFApoNptiNG61WvA8D0+ePMHJyQkcx5EUcRI027ZhWRYmkwkajQZ2d3cRBAFevnwpwzt7vd7awFpNEA3RMbgrsHuQjSZUIrkur6PQXhc6ZoEjI/SE82KxKHuB7vK9SxiCc8fgBg5A0iopOaZPlkyiNRvu40DaVEzDIYftMQeHLdb6Tfu1rgomYxcKBVFwqtWqtGFzrhozmIj0c6W9OOmuP52xo31G2lPGDVK3hrOElSQJbNuWzis9740GTXZ7lMtlzGYzMV5bliUHC30y1kMEzXVmcNfQHUtcp+xiBNZ9nbcNrRrp+xNfJ/+tVd27VnIMwdkicOEsl0uMx+O1m4fOt9kG6c/g9sDNgScjx3FQr9flfblcRqFQEB9Mt9tFq9VCGIYYj8cfbYJXfW62cJfLZTx9+hTPnz/H4eEhSqUSAIgiQsLyJQOzHh9CNYcGZXpwSEgsy1qTv9OGen4PCT9fz2QywWg0gu/7GI/H8DwPjUZDJqqznLe7uyupybZtIwxDnJ+f4927d0iSBL1eD5PJBMPhUEpihuwY3BVo4tfqJ433TOm+i9fEQxQVWABy7bKErb2id3XPMgRni6AXxKaUSrPJPg5ogmNZFhzHQa1Wg+/7qNVq0i5Nr02320Wz2USSJAiCQObVXHVT0cM7i8WiDO/84Ycf0Gg0UC6XsVwuhSwwQVtL15ueUysxJOnskOLPSZ8RN8104jJJDR+DLepxHEtycbvdRrlcljEPNGfSz8a2+t3dXZleHsexkMYoivD27VuEYQgAUhYzwYAGdwGtllD5zOfzYu6/K4LDlHydLs4Ss45ASXt17uL6MQRny2E21ccFnRFDcqNvzpzYrccUxHEsAX83MY5A58ukZ09RSZxOp1e+8ae7qzhbR//sNBynk443PZYew0CiAwCdTmfNQMyS3nQ6lfTnbDYLx3GQy+VQq9WkU3E6ncJ1XURRhNFohEwmI11WfF4Dg28BXZ6mgVeHuN4l0qoswb0rPdblrmAIzj2A2VQfPvRkcJr1arUaqtWqZL94nodCoYDpdLqmNjSbTbRaLczn840tz1cFW60dx1kz+U6nU8RxjMFggG63iyiKrrTZUskhQWMdnxs4x5TQF/Opx9VEiaF90+kUw+FQvDWWZeHg4EBmdIVhiEqlgp2dHTQaDWSzWTQaDWQyGdRqNTx9+hRRFOHo6Ajj8VgmsXPiuTY9Gxh8KxSLRZTLZYkTGQ6HsuaBy2Wf3TR0MwBzcbhvUcGxbXttLMpdXTuG4BgYbAG0oZg3/WKxCM/zRMFhrZsKShiGMoSVhOBrNxK+Bj1Yj4/J0xrJjo5nv46Kozdm/bNTjv/SEMG0CXk6nUppVxuMp9Mpjo6OAACu64qHicZttt87joM4jqU8x/DE25zWbGDwOeiICPpb7lq9AT5cxzrjigSHGT1sDLjLa8cQHAODOwQvfnYtWZaFarUqLeF7e3vyeZrPaTBsNpuIokgGuX7NSU4PzvM8D7VaTYgAvTZUNPRrINm5LLH6lLTN+r2u5V9lc+RzLxYLxHGM6XSKTqeDJEkwHA5lnlav18NoNJLZVSxZcfyELp9tImIGBt8CulRNj4vv+3JN6tyouyQR7Exk3hTVWB6OqMrSVwd824qEITgGBncMdhBRqanX6zKG4ODgQIbaLRYLjEYjvH37FnEco9PprIX6fe1rYNeW67qoVquymbJbim2qQRBsJDhfgvbf0IBITxHwoX5PgpPNZq/8c7FlHnhfthqNRuj3+zKfKggCxHEM3/exWq3g+z5KpRLK5bJ8f/o1fsswNQMD4EPbtfa0eJ4nyqKOXrhLhVEruZVKRV47fW4kOLo8bQiOgcEjQdp7wzlTHKBZLBZFlWBsexzHa9PBb2LD4Gtg55ZWNqji8MbPxOTrGB51V9SmJFadkfMlo/GXQJJCz9JyucRgMECr1UKSJJKCTJLGYZ/poEFDbgzuClrFYTlIk/5tWJva8E9VNp1urJsJviUMwTEwuGNkMhnJbSmVSnj69Kmc1pjr0ul0MB6PcX5+jmazKcbf67SDb0I2mxWfT7lcXjM4e563lkocBAH6/T7G47G8hqsYjfWGqMP/eBpdLpdCtkhS+L1XAdWXxWKBdruNfD6POI5xcXEhKcgMMqRa5XkestksptPpR+W3u76RGDwOaDWGa7JQKCAMQwn825T4fRfQ+wK7OtnxycMSldi7mDZuCI6BwRaACo5t25LKq6PPkySROUocJ3CVCeGfg1aRaLzVCg7DxbiZ8uavP3dVpFvGqdpo6f0mvAWUxplpQz/RZDKB7/sSljidTmUqOX9e/Ts25MbgLqBLUJzLplVGrYjeJUhyMpmMhIDqxgGdbPwtYQiOgcEdQm8ClmVJQBbbnuM4RhAEODs7E3PvTSZa63A/DvKs1WrY39/Hzs6OlMk482Y6na6Ng7gu0nk4egin/r2kRzdc97n4+yI5nE6nksLqeR48z0OxWMT5+TlyuRzevn271iJ+15HzBo8Hep0lSYJ+v4/lconRaIT5fI7BYIB+vy/jRbYhvoDet9lstmYyJugj+tbpxobgGBjcEfRNnG3hVA+m0ymazSbOz88RhiHevHmzptzc9Gtgq7TruqjVatjb20O9XofruuLBYVlMe4CovFwHWmJPt5Rq78FNQHdvMQU5iiJRzvi7r1aryOfz6PV6az+jgcG3Bgn5YDDAarWS0SE0zl+1PHybIMGhgmPbNoAPJEbPqjIlKgODRwJe6DTvZjIZRFEkNW1m3NxEQvEmaPXI8zz4vg/XddfmTQGQ3B22YGtD4XVek/4e/bGWsm9TztZz37jhFgoFjMdj5PN5UahuqgxoYHAVUOHkWJJ8Pi/dgOPxeK1Uxa/fBuhsnLTyepOq7GVhCI6BwR1jtVohDEN0Oh3k83m5yXY6HXS7XVFObuNmy0GUpVIJx8fHqNfrODo6kuRk4H3WRa/Xw9nZGd69e4fRaCTy+NcQr7S3JX3Ku02Ss1qtpExFWT2Tych4BpbkWN4yMPiW4DUVhqEY+7kv8ONtCv7TYNIyD066owr4cG2bEpWBwSMADXo8qbE8c9vqTbo1nJ1bVHCYYUGTrp5UzgnHX2PA3fR930rB4fNrL5B+/m3wNRg8bnBfACBqIjv80sNotwl6RhzTxIHbP7RsgiE4BgZ3BN5YM5kM4jiWvAi2g5JM3MZwPZ6qHMdBtVpFvV7HwcEB9vb20Gg04Hme+FRWqxU6nQ7Oz89xcXGBfr+P0Wgk4yG+tkSlOy7S0BvibZ36NpXLtvHGYfD4QJKjfWrb3t2nx6cwmZyv+bYOa5+CITgGBncIEhcad3Wd+rY6DnQMvOM4qFQqqNVqMpyyXq/D8zzp3ODYA2bw9Ho9DAaDtSnb1wE36E3mYv4/X+9tYxtvFAYG2hzPUQfbDt04YFkWAKyRm285Dd0QHAODLUBaRdCns9u4+ZLkWJYl6ckcWeA4DgqFgiQWR1EknRtBEEh56ms3qc/J1TeRZGxgYHB30OMZGAXxra9jQ3AMDLYI3BRuE3paeLVaxfHxMY6OjvD999/j4OAA1WoVtm1jMplISeqXX37B3/72N4zHYwyHw0upN5q8pDe2dCeFngHF18hZPHc9kdjAwODq0K3j+tBmZlEZGBjcKkggisWiJCfTZGzbtsyeYltqEAQSPHiZcLG0b2ZTy+iXiIshNQYG9xu3fVj7EgzBMTB4JNCTiR3HgWVZ2N/fxw8//CAG41qthuVyKQnKnU4HzWYTo9HoUrOZtIdIB/XpvB/gfQsp37TvRn+NMfwaGBh8DQzBMTB4JKBqUygUYNs2HMfB3t4evv/+e+zu7mJ3dxfVahXD4RBhGEo2T6vVkrIU21M/9xyaSFEJAtbLb/l8Xt50GUp/7W37kAwMDB42DMExMHgkYPCWbduoVCrwfV/eu64rQVzz+RxxHCOKInlj0ODn2lN1u7cmOMAHwsJ/5/P5jQpOWvHZ1lZYAwOD7YchOAYGjwCZTEYSi6vVKn73u9+hXq/jp59+wrNnz+D7PizLwmq1QhAEuLi4wMXFBc7Pz3F+fo7BYCCzsDaZi/VEcsuykMvlhDRtUmH0rCmSHY1014UhOQYGBleFITgGBo8ENBXbto1arYZGoyEKjm3bUiKazWYfKTiXTVTWKg4Vms+9HpKiTYZio94YGBh8DQzBMTB4wCCByGazKJVK2NnZwf7+Pn7zm99gf38fx8fHqFQqEgE/mUzQ6/Uk1K/dbqPT6SCKos8SnE95Zz7V5q0NyLostVgsJJb+c2ZmAwMDgy/BEBwDgwcMXQLyfV9GMvzwww84OjrC4eEhyuUyFouFTCnu9/toNptCcHq9HsIw/GJQV5rcaAKTz7/fatJeG34fgLVIev12F/kZBgYG9x+G4BgYPFBoT0yhUECpVEK9Xke1WoXv+5JYDLzPqwjDEFEUYTAYoNfrYTgcYjqdCvG4LPQAS87P0e8BbGwL/xS5MTAwMLgODMExMHiAoInXsixUq1U4joPnz5/jD3/4A/b29nBycoKdnR1YliXk5tWrV+j3+/jHP/6Bv/3tbxgOhxgOh0iSRKYDfwk6lp2BgJu6pVii0inGs9kMs9kMi8VCxkHcxEgIAwODxwlDcAwMHij0vCnbtuH7Pmq1GiqVClzXRbFYRDabxXK5xGw2k7TiwWCAfr8vJaurzJDZlHmT9uBoJQf4WLnRQ/lMacrAwOC6MATHwOABgl1MjuPIfKknT57g5OQE5XIZtm0jl8thPB5LoN+LFy/Qbrfx5s0b9Pt9xHF85cGaesDedDpFNpvFfD4XxYYdU1rB4dfPZjMxGPO9ITcGBgbXhSE4BgYPECQ4tm1jf38fu7u7ODk5wcnJyZp6E0WRGIpfvHgh2Tf9fh/T6fRaCo4mOJrMpAkOU45Z/prP5+L3uWxJzMDAwOBTMATHwOCBgmUhKiJsA89ms5hMJgCAKIoQBAFGoxGCIMB4PJaBmtchGXrmlB6/oMtU2nSszciXydkxMDAwuCwMwTEweMBYLBZCYnq9HprNJjzPQz6fh2VZePfuHU5PT3FxcYHT01O0Wi20222Mx+NrdTLRM6N9NprYpGdO8Xv09+rPGRgYGFwXhuAYGDxg0NsymUyQJAmiKAIAxHGM5XKJMAwRBIGoN+Px+EpdU5973k0fa4JjSIyBgcFtwhAcA4MHCLZpR1GEVquFMAxRKpVgWRZ830ccxygWi3jx4gVevnyJTqeDfr+PIAgwnU5v7XUZhcbAwOBbwRAcA4MHCLZ+A0Cr1YJlWbAsC5lMBp7nIY5j2LaNly9f4vT0FMPhUFrDZ7PZrRIQQ24MDAy+BbJf/hIDA4P7ivSYA36ss2ZM5oyBgcFDRMZsaAYGBgYGBgYPDUbBMTAwMDAwMHhwMATHwMDAwMDA4MHBEBwDAwMDAwODBwdDcAwMDAwMDAweHAzBMTAwMDAwMHhwMATHwMDAwMDA4MHh/wPcs9wIJWv0ywAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 720x720 with 9 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(10, 10))\n",
    "\n",
    "t1 = time.perf_counter()\n",
    "augmented_images = img_augmentation(images)\n",
    "t2 = time.perf_counter()\n",
    "print(\"elapsed timer:\", t2 - t1)\n",
    "\n",
    "for i in range(9):\n",
    "    ax = plt.subplot(3, 3, i + 1)\n",
    "    plt.imshow(augmented_images[i].numpy().reshape((64, 64)).astype(np.float64), cmap=\"gray\")\n",
    "    plt.axis(\"off\")\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Make sure the image augmentaiton layers are working."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Create an EfficientNet Lite model and merge preprocessing, augmentation layers with it to one model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "img_augmentation (Sequential (None, 64, 64, 1)         0         \n",
      "_________________________________________________________________\n",
      "efficientnetlite0 (Functiona (None, 6507)              11747915  \n",
      "=================================================================\n",
      "Total params: 11,747,915\n",
      "Trainable params: 11,705,899\n",
      "Non-trainable params: 42,016\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "from efficientnet_lite import EfficientNetLiteB0\n",
    "\n",
    "eff_net_lite = EfficientNetLiteB0(\n",
    "    include_top=True,\n",
    "    weights=None,\n",
    "    input_shape=(64, 64, 1),\n",
    "    classes=len(ls),\n",
    "    pooling=\"avg\",\n",
    "    classifier_activation=\"softmax\",\n",
    ")\n",
    "\n",
    "f16_model = tf.keras.Sequential([\n",
    "    tf.keras.layers.InputLayer(input_shape=(None, None, 1)),\n",
    "    img_augmentation,\n",
    "    eff_net_lite,\n",
    "])\n",
    "f16_model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Set a path to save model realted files and optionally load stored weights to resume training."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "model_dir: E:\\projects\\DaKanjiRecognizerML\\single_kanji_cnn\\model\n"
     ]
    }
   ],
   "source": [
    "#path where the model should be saved\n",
    "model_dir = os.path.join(os.getcwd(), \"model\")\n",
    "print(\"model_dir:\", model_dir)\n",
    "\n",
    "#f16_model.load_weights(os.path.join(model_dir, \"tf\", \"checkpoints\", \"weights-improvement-46-0.94.hdf5\"))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Set the optimizer, loss function and compile the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "l_r = 0.001\n",
    "opt = tf.keras.optimizers.Adam(learning_rate=l_r)\n",
    "\n",
    "METRICS = ['accuracy']\n",
    "\n",
    "f16_model.compile(optimizer=opt, loss=\"categorical_crossentropy\", metrics=METRICS)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Setup callbacks for the model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensorboard log dir: E:\\projects\\DaKanjiRecognizerML\\single_kanji_cnn\\model\\tf\\logs\\20210824-204112\n"
     ]
    }
   ],
   "source": [
    "from tensorflow.keras.callbacks import ModelCheckpoint\n",
    "\n",
    "#CHECKPOINTS\n",
    "filepath = os.path.join(model_dir, \"tf\", \"checkpoints\", \"weights-improvement-{epoch:02d}-{val_accuracy:.2f}.hdf5\")\n",
    "checkpoint = ModelCheckpoint(filepath, monitor='val_accuracy', verbose=1, save_best_only=True, mode='max')\n",
    "\n",
    "callbacks_list = [checkpoint]\n",
    "\n",
    "# TENSORBOARD\n",
    "# load the tensorboard extension\n",
    "%load_ext tensorboard\n",
    "\n",
    "# add the tensorboard callback\n",
    "log_dir = os.path.join(model_dir, \"tf\", \"logs\", datetime.datetime.now().strftime(\"%Y%m%d-%H%M%S\"))\n",
    "tensorboard_callback = tf.keras.callbacks.TensorBoard(log_dir=log_dir, histogram_freq=1)\n",
    "callbacks_list.append(tensorboard_callback)\n",
    "\n",
    "print(\"tensorboard log dir:\", log_dir)\n",
    "\n",
    "# LEARNING RATE SCHEDULER\n",
    "def scheduler(epoch, lr):\n",
    "    return l_r - ((l_r / 100 * 4) * (epoch // 3))\n",
    "\n",
    "scheduler_callback = tf.keras.callbacks.LearningRateScheduler(scheduler, verbose=1)\n",
    "callbacks_list.append(scheduler_callback)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Finally train the model on the data set (in case of an Interrupt creates checkpoints)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "#train the model\n",
    "hist = f16_model.fit(\n",
    "    train_dataset,\n",
    "    validation_data=val_dataset,\n",
    "    epochs=100,\n",
    "    initial_epoch=10,\n",
    "    workers=16,\n",
    "    use_multiprocessing=True,\n",
    "    callbacks=callbacks_list\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Evaluating the model\n",
    "\n",
    "After training plot the loss and accuracy for the test and validation sets."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_2\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "preprocess_tf (PreprocessTFL (None, 64, 64, 1)         0         \n",
      "_________________________________________________________________\n",
      "efficientnetlite0 (Functiona (None, 6507)              11747915  \n",
      "=================================================================\n",
      "Total params: 11,747,915\n",
      "Trainable params: 11,705,899\n",
      "Non-trainable params: 42,016\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "policy = tf.keras.mixed_precision.Policy('float32')\n",
    "tf.keras.mixed_precision.set_global_policy(policy)\n",
    "\n",
    "eff_net_lite = EfficientNetLiteB0(\n",
    "    include_top=True,\n",
    "    weights=None,\n",
    "    input_shape=(64, 64, 1),\n",
    "    classes=len(ls),\n",
    "    pooling=\"avg\",\n",
    "    classifier_activation=\"softmax\",\n",
    ")\n",
    "\n",
    "f16_model = tf.keras.Sequential([\n",
    "    tf.keras.layers.InputLayer(input_shape=(None, None, 1)),\n",
    "    PreprocessTFLayer(),\n",
    "    eff_net_lite,\n",
    "])\n",
    "f16_model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "# reload values from file if necessary\n",
    "f16_model.load_weights(os.path.join(model_dir, \"tf\", \"checkpoints\", \"weights-improvement-47-0.95.hdf5\"))\n",
    "f16_model.compile()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2542/2542 [==============================] - 149s 58ms/step - loss: 0.0000e+004s - loss: 0.0000 - ETA: 4s - loss: 0.000 - ETA - ETA: 2s - ETA: 0s - l\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "0.0"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "f16_model.evaluate(val_dataset, workers=16, use_multiprocessing=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['セ', 'ゼ', '也', 'ヤ', 'た', '巳', 'ピ', '牾', '拒', '七']\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 158.4x158.4 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# load a sample\n",
    "sample = tf.keras.preprocessing.image.load_img(\n",
    "    path= r\"Z:\\data_sets\\dakanji_single_kanji_cnn\\100\\1540.png\",\n",
    "    color_mode=\"grayscale\",\n",
    "    target_size=(64, 64),\n",
    ")\n",
    "sample = tf.keras.preprocessing.image.img_to_array(sample)\n",
    "sample = sample.reshape((1, 64, 64, 1))\n",
    "\n",
    "# run inference\n",
    "prediction = f16_model.predict(sample)[0]\n",
    "\n",
    "# output 10 most liekly predictions\n",
    "ind = np.argpartition(prediction, -10)[-10:]\n",
    "q = ind[np.argsort(prediction[ind])]\n",
    "print([ordered_labels[i] for i in reversed(q)])\n",
    "\n",
    "# show the most confident prediction and the image\n",
    "show_image(sample.reshape((64, 64)), ordered_labels[prediction.argmax()] + \" - %.2f\" % prediction.max())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Saving the model\n",
    "\n",
    "The model is perfoming very well therefore save the trained model as a \"*.pb\" file."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "E:\\projects\\DaKanjiRecognizerML\\.venv\\lib\\site-packages\\tensorflow\\python\\keras\\utils\\generic_utils.py:494: CustomMaskWarning: Custom mask layers require a config and must override get_config. When loading, the custom mask layer must be passed to the custom_objects argument.\n",
      "  warnings.warn('Custom mask layers require a config and must override '\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: E:\\projects\\DaKanjiRecognizerML\\single_kanji_cnn\\model\\tf\\trained_model\\assets\n"
     ]
    }
   ],
   "source": [
    "f16_model.save(\n",
    "    os.path.join(model_dir, \"tf\", \"trained_model\"),\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Finally convert the model to a TF-Light model to be used in applications."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "E:\\projects\\DaKanjiRecognizerML\\.venv\\lib\\site-packages\\tensorflow\\python\\keras\\utils\\generic_utils.py:494: CustomMaskWarning: Custom mask layers require a config and must override get_config. When loading, the custom mask layer must be passed to the custom_objects argument.\n",
      "  warnings.warn('Custom mask layers require a config and must override '\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: C:\\Users\\dario\\AppData\\Local\\Temp\\tmpj73i01k5\\assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: C:\\Users\\dario\\AppData\\Local\\Temp\\tmpj73i01k5\\assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "E:\\projects\\DaKanjiRecognizerML\\single_kanji_cnn\\model\\tflite\\model.tflite\n"
     ]
    }
   ],
   "source": [
    "# Convert the model\n",
    "converter = tf.lite.TFLiteConverter.from_keras_model(f16_model)\n",
    "\n",
    "# set quantization, additional ops\n",
    "converter.optimizations = [tf.lite.Optimize.DEFAULT]\n",
    "converter.target_spec.supported_types = [tf.float16]\n",
    "\n",
    "tflite_model = converter.convert()\n",
    "\n",
    "# Save the model.\n",
    "print(os.path.join(model_dir, \"tflite\", \"model.tflite\"))\n",
    "with open(os.path.join(model_dir, \"tflite\", \"model.tflite\"), 'wb') as f:\n",
    "  f.write(tflite_model)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Load the tflite model and check that it is working."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAIUAAACcCAYAAABRA2hvAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjQuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8rg+JYAAAACXBIWXMAAAsTAAALEwEAmpwYAAAd8ElEQVR4nO2deWxc173fP+fODIfLcIazUdzFTSIlkRJlS4wSKZIsWYotK3ZkJLHTJk5gBG1f2ry2Lw0K5CGIH9DGLdD3R4tXNEGQGrBjpEiTxnIsGLHlyFZka6Uka6Eoi5IomuuIy3C4DGe9/WN4r2c4Q3JIzkrPByA4y13O3Pu95/zO7/zO7whZlsmRIxwp3QXIkXnkRJEjipwockSRE0WOKHKiyBFFVotCCPEVIcS/FULUL7LNUSHE3wghylNZtmwmq0UBfBt4DBhZZJujwBPAQwAhxNNCiHXzNxJCSEKIR4UQzUkpaRYhss1PIYRoAGyyLJ8XQlwAnpVluU8I8WXABhiA3WG7fBHQAGcIPQTPAb3ALsAP/HPgMPA4YAX+QZbll1ZQLiPwj3PnY+58P5JleXrZPzLNaNNdgBVQB7wlhDhHSAA3hBBXgVpCovgSMAo4ZVn+CEAI8TxQBpwE3gY+kGV5Wghhmtvnj8D/BX6/kgIJIczAKeAG8CgggNeBU0KIA7IsT63sp6YJWZaz6o/QE90z9/p54P251y8B/2fu9fvAvwaeAf4rcB94Efh3wCUgL8ZxawEZeGkFZfp7wAnkh31mBGaBf5Pua7bcv2y3KRbDA1wDthNqNm4BU8CULMveRJ1ECJFHSIBvybI8q3wuy7KLUM30t0KIrLrOWVXYMEqEEC8BXwdq517vn7dN8dz3zwL/DChIUlk2AeXAX2N89yGwAahK0rmTQjbaFBCyF16asxVsc69fAsJ7Dv8KGAR+TKjn8ZDQzRtMcFmUGx6rBzQStk1vgs+bNLJVFAsihCghZDS+IctyvxBiAPieLMsXhRAbgS1CCCEnrttVPfd/NMZ3I/O2yQqyVRRlQoizhJqIKiHEB4Qu/AWgDfjPwDfm2vJy4B+FEEFCvYLdwCtCiH+RIGGY5/7H6noqn1kScJ6UkY2iGAB+APwD8HfAj4D/RUggFwjdJKcsy/uFEHsJObj+CzApy/JDIUQP8IcE1hTDc/9NMb4rmfs/lKBzpYSsMzRlWe6UZfl/E/JR9Id99TeEGXRCiA3A00AnoR7IX+c+g5DTKlEoZYhVG1jmbZMVZJ0oAIQQekJ+gAdhH48ALXxW+x0j5NOoBHzA/yPk8k40fXP/S2N8Z5/7n1WiyMbmA0LdwJ45QxJgGxAg1C3dMbfN/wD+myzLQQAhxJ8Iubv/fYLL0kVInHuBf5r33V7gOqEmL2vIypoCaAfeCXu/m5Cf4L8DOgBZlmfDBFFMyOvYQcggfcAKEEL8RAjxRviAmizLAeB/Ak8IIQrDtjUD+4B/SqD9khrS7VJdyR9wGmiae30OeD7sux3ALxbY7zngb4EXY3y3gZCb+z8tsG8RodpIBv7lvO9KgG7gt4AeyAeOExoLKUr39VruX9bVFEKI3cBVWZZvz33kA9qEEJVzNUIX8B+EEAYhRLEQokQIYZ+LpzhHyMZoF0JohBA7hRAfCiH6AOV4fy+EcAohrgoh/k45rxwa7XwduAP8JbxMsiw7CTUVs8AVQjXSQ+CAnIWjpFk1dD43qvkS8B/lufELIcQLwA/5bJR0Pj5C4yDhf/cI1S6LxWF8bsk2UdQAw7Ise9JdlrVMVokiR2rIOpsiR/LJiSJHFDlR5IhiUY+mECJncKxhZFkWsT7P1RQ5osiJIkcUOVHkiCInihxR5ESRI4qMF4UkZXwR1xwZHWSTn59PY2MjLpcLn8+Hw+EgEAiku1hrnowWRUlJCfn5+eTn5xMMBhkdHc2JIgVkbN0shMBgMKjvZ2Zm8PsTGW+bYyEyVhR6vR6j0QiEosMePnxIMBhMc6k+H2SsKMxmM3NBufh8PiYnJ9Ncos8PGSkKnU5HaWmpKorp6Wm83oRNFM+xBBkpisLCQrTakA2sNB05UkfGiUKSJOx2e0QtkWs6UkvGiaK4uFg1MAGcTie5kMHUklGiEEJgt9tVL+bs7Cyjo7Fm+GcfJSUlFBUVpbsYcZFRorBYLBG1xNjYGD6fL40lSgy1tbX8/Oc/50c/+pHaLGYyGSMKnU5HZWWlWkv4fL41YWBqNBq++c1vUldXx7Zt29i9e/fSO6WZjBCFEIL169eTl5cHQDAYpL+/f03UEkeOHGHfvn3IskxBQQHPPPNMuou0JBkhCrvdHtFsTExMrBlbQpKkiCYjG0Z9015Cq9VKVVWVerGmpqbo7e3NuB6HUostl/k2xJ07dxJRnKSSdlFMT0/j8YRmAfr9foaGhjKu2aipqeGFF16gvn7BvPAx0Wq1PProoxGfXblyJZFFSwppHzqfnZ3l7t271NbWMjw8jNPpTMhx9Xo9QghmZ2eX3niJ47S3t2Oz2Th06BCvvvoqbrc7rn0lSaK6enmJ8UpLYyXE+Qyv16teo+rqakZGRuIuT7ykXRQQEsbt27dX1WTk5eVRUFCA3W7HYrFgsVjQaDT09PTQ1dW14jiMqqoq6urqgNAg3ZYtW7h06VLc+ysju7IsEwwGsdlsPProo7S3t8fcvr29fdFuq9Pp5NatW3z88cds2rSJoaEh/vznPyd0BDkjRAGsSBAGgwG9Xs/69euxWq2YTKEEdUII9cJu3ryZvr4+JiYm4j5uUVERdrudnp4eenp6GBoaorKyEoDW1lY6OjqoqKigoOCzJL4mk4mjR48yNfVZbnZJktDpdExOTqoJQZ5//nn1u5Vgt9uxWq2Ul5czMDCAzWajsLAw4ryrZdFZ55k4Q6yoqAir1Up9fT3FxcXk5+cDIVFNT08zPj7OvXv3kGWZpqYmysvL6ezs5ObNm3GfY+PGjfzkJz+hu7sbk8mE2+1mZmZGPb/X66W6ujptHkqn08mNGzfU2u/06dN0dXUt+zgLzRDLmJoiFsoTb7PZsNlslJeXU1BQQEFBgfrkDQ8PMzs7S09PD5OTk1Htq7LPcrDZbOh0OjZt2gSEboLb7UaWZQwGw5LtfrLx+XxJDUvMOFFotVoqKyvRaDSUl5dTXFxMQUEBWq2WYDBIIBBgeHiY0dFRent7mZmZSfgFmt9jSCTBYDAqNkQIgU6n44MPPiAQCCBJkmojhL8G2Lp1Kz09PUkrH2SgKGpra2lra4uIugoGg9y/f5+hoSFGR0fxer1Je1IMBgMVFRURn4XflIWMQI/Ho95sZRu3283Zs2cjtpuamor6DEI3f2Rk6WxLRqORlpYWWlpaltx2pWScKCRJoq+vj/HxcWRZZmhoiNnZWXw+X0ocWrIsc/r0adrb27FarUAoaFg5t9fr5dSpUwwNfZZZWQjB3bt36e2NTNQfDAZVWyRRuFwu1q2LWgItoWScKLq7u8NTEaac6elp3n77bU6ePIkkSUiSxNGjR2loaADg/v37XLx4EbfbjdvtZmRkBCFEwm9+OkmpKIQQbNy4kWPHjnHq1CnOnz8ftU2mRGwrXlWtVqvWGAAjIyOUlJRQUlKCLMusX78egPHxcUZHR3G5XFkvkJSKQqPR8IMf/ICtW7dy+/btpXfIQMJrMCEEGo0GQO0hBQIB3G43fX19TE5OZqVAUjr24ff7+etfQ6sqHThwQA3OjQez2Ux5+fLWm02UHyHcuFyqadNoNBgMBpqbm2lra2Pr1q3YbLakjY4qAkwkKbcp+vr68Hq9WK1WtFptXLO+rFYrL7zwAgaDgVdffZUHD+JLrV1VVYUsywkbT4GQTXH//n11CoLFYkGSJLXGCEen02E2mykpKWFmZoaxsTEGBgZWPR4Tjt/vT+jxIA2jpNevX2d6eprS0lJ27Nix5PaSJHHs2DEqKysxmUzs27cv5g1YaF8Ah8OxqjKH1wxer5fx8XFu375NV1cXHR0dXL58mcLCQlpbWzGbzVH7CyEoKiqiurqatrY26uvrVU9sJpJyUXi9Xrq6utBoNEt6GrVaLYcPH2bDhg3qZ83NzdTW1sZ1rkQZrYsNUHk8Hvx+Pzt37mTv3r1s376dBw8eMDU1FdOXotfrVXE0NDSsOE4jmaRcFD6fj08++QSA7du3L3rBq6qqOHLkSETNkJeXt2zbwmazUVZWRllZGYWFhUvvMI/wmiKWPdHa2kpzczPBYJATJ07Q09PD1atX+eSTTxaMD9Hr9VRVVbF582ZsNlvctV8qSIufYmJigmAwSFNTE0KIBQ03v9+Px+NZMgK6sLAQu92uvm9ra6Ourg6DwYDBYODYsWPk5eUhyzKvvPIKb7/9dtxlLSsrW7RG0+l0PPXUU+h0Orq6urh8+TIAgUAAh8OBw+HAYDBQVVWl2lHhmEwmjEYjd+7cYXBw6dUxrVZrROhiMkiLKE6fPs2LL75Ifn4+xcXFCw5rK5a+IgrFoGtsbORLX/oSGo0GSZIwm80RUVHhQ+fzUWafxescKykpQa/XA6HmaL5hvG/fPpqamvD7/bz11lsxDeepqSm6uroWFMfMzAzj4+NxlaegoEAtT7JIiyiUG1JRUUFLSwsfffQR8Fm8JoTGIJ566ilKS0vVuARFBDU1NSueP7Fnzx5++9vfrijkz+Vy0d3drb63Wq184xvfIC8vL6KWWAhFHIrRabOFVqJ48OBBwnsQqyFtzce5c+fYv38/3/ve93j22WeB0GBPeDOQDBJpfB48eBCLxYLX6+XEiRNxC216epquri5KS0vRarWLzm+Z799IxWSitIgiEAgwMTGB2+3GbrcnTQh+vx+fz6deyJ6eHq5fv75qYQghOHLkCF/72tcAOHnyZEyXfSy0Wq1aHqXJKCwspL6+Psre0Gg0bNmyJaL7ulRtlAjSNiAWjwEZDz6fTw2s8fl8nDlzBr/fj1arRaPR0NvbS0dHBxBquxMR5PrEE0/w7W9/G61WS2dnJ2+88QayLKPT6SJurN1ujxqG37RpU1QPSPFjxHM95ndh79y5k/C0T2kTxTvvvMORI0eWvBCyLOPz+dSn3u/3EwgEuHbtGv39/QwPD3P9+nV1WyVW0Ww2c/DgQaamphI2sUiZyfaVr3wFIQRTU1Ncv36dtrY2INSFtlg+W7NWq9Wi0+kScu6FSMZ0iLSJwuv1qr0EWZbxeDyq59Hr9XLy5En8fj9Wq5X29nbVOFX+nz17Vq0BUkVJSQnf/e53mZmZ4fbt23i9XhobG2lsbFz1sYPB4II9kIGBAXV8Q5kjk0zSJopAIKDGPP7617/m2rVrEcGnijewoaEhpjs8HfEWsiyrcRTL2cfhcEQMpI2Pj6sOPAWfz0dPT0/M3xW+b2lpaYRhOj2d+MUM0yYKh8PBhQsXOHz4MCMjI8uKtgaorKxc1OhS0i2u1m7p7e1lcHBwQS+qECJiMvSdO3ciwupkWaavry9h4YMOh4M//vGPCTnWQmRE5NXjjz/Ou+++G/O7/v5+hoaGKCsri/h869atnDhxYsGeRGVlJUKIVd8Ml8vF73//e1pbW9WA4tu3b6tPqxCCoaGhNZX0Na2iWKi6DGd2djZmOxoeN7kY8Q6zL8b09DTnzp1b8HshRFTUdTaTVlGcP3+e73//+2i12kUvqsfjUeMhlfD2eGd8JdMwUwJqlFHcy5cvrwlhZETz0dLSQm1tLffu3Yv5/W9+8xskSUpKQMlq2L59O2azGa1Wi9vtzorURfGQ1lQEypM/3+kzn5mZGaampjJKEBByJC0npDBbSKsoXC6XOn9ioWrXaDSyZ8+eZU3VkyRpTd6sVJH2pCUQMtT2798f8ZlWq6WxsZFvfetb7Nq1i927d8cdiKLEV2RaNpxsISNEAUSNERQWFnLo0CGKi4uB0HRCJR3AUii9AafTicvlSnhZ1zppF8VCUVAulyvCja3RaNRZWvGijJUki7ViWM4n7aJQ/P0ajSbqIt+7dy9iwKekpGRZsYxKcI7yl0iMRuOK4j2zgYyxxnbs2KFmZ1EYGxtjcHCQmpoaIDQKqcyhgFAT09jYGCUmvV6PxWKhtLSU5uZmJEnC4/Hw+uuvJyw/1FrteUCaRaGE+fv9fiRJwmg0Mjs7y+bNm9ULXlJSom6v1+v5zne+oxqQSszm/EEhrVaLx+NhfHyc+/fvA6EBtlxTEh9pE0VBQQGHDx+mubmZjo4OZFnmq1/9KsFgEEmSmJiYQJbliCra7/dz5syZiCZleno6ypVtsVgoKCjA7XYzPDyckt+zlpazSpkoamtr1bCykZERPB4PRqMRh8PBBx98oE7tU4Jqent7CQaDPPvss+rkn0AgQGdn55KTdlMlhHCmpqbWTBc4ZaL42c9+RmtrKwC//OUv+dWvfsVrr72WqtPnWAYpEUVzc3NMg3C5ZPuTKISIyOgHod80OjqaUU1PSkRhMBgibIOVXoBMMuYWK4sQQo3/EEKg1WqpqKhAq9WSn58f0T1WZsX39/fz6aefJr3c8ZASUYRHIE9PT/POO++k4rRJZb53Va/XU1NTgyRJrFu3jkceeQSfz0d/f/+ixxFCYDabMRgMjI2NJSW8brmkRBTHjh1TXweDwZT+cI1Gg0ajSdgSlnl5eQgholIJrF+/Xl3GoaamhoaGBgKBQNxLXul0Oqqrq1eUJDXRpEQU4U2H1+tNqW3wxS9+EZ1Ox6lTp5a9r16vR6PRUFxcrMZNrFu3Tm0Swgn3tCq/MS8vjw0bNjA1NRVXsE/49IB0knRRVFRUREz+PXnyZNyTaeezEjGVl5cvmcmmqKiItrY2Ojs7EUJQUVGBJElYrVa16VtOeqKhoSGKi4upra2luLiY6upqdQ6q2+1WnWhFRUURx1WccemO90y6KObPD02lla3X65ecod3c3Ex1dbW6qF34LPfV0NPTg91ux2AwUFNTQ2FhIa+99hqjo6N4PB4kSWLPnj0RebkMBgMlJSVpXxUp6QNi4U+Cz+fj5MmTyT6lSmlpKWazOWqOhUJ5eTm1tbVotVp1iH2lgggEAjx8+JCHDx9y7do1PvzwQ4aHh9Wmpqqqii1btqjNiLLUQziLpVBIJUkXxdGjR9UfGgwGV5V/arkXrKmpiampKcbGxoCQkXjo0CE2btwIwN69e9XUB/ESDAajbqbP5+PixYtcunSJK1euMDMzw969e2lqalIN0kAgwJNPPqnm5FSyCc9n/lSGdJD05iPceEqlgZmfn8+6desYHBxkcnKS8vJyDh8+jM1mw2Kx0N3dzeDgoJrdbqFmze/3o9FoGBwcZHZ2lv7+fkwmE21tbRELvMzOztLS0sJzzz2HzWZTb77H48Hj8RAMBtHpdFRVVanNQyzjMxOG45MqCpPJpGakBThz5syqxiWWIyrFRnj//ffRaDQcPnxYtW0U47e3t5ft27ej0+kIBAJotVq8Xi/Dw8O4XC4CgYDqZ/B4POr5tVotgUBArbmMRiM//OEPaW1tjbJh8vLyVNH5/X6qqqr4+OOPgdB4iTJDXkHJVJOKOaMLkVRRzE87NDo6uirLur+/Py5/gxCCbdu24XA41PiMoaEhVRQajYadO3dy8eJFJEmivLwcnU6HJEm8//77XLt2bUmDODwNYlNTU8zu5OjoKO+99x5er5f9+/cjhKC9vZ1z586pKZtjiaKgoGDtisJoNKpPUyAQ4M0331zV8ZxOZ1wxEevWraO2tpbz58+rT/elS5fYsGGD2sZXVFRQVlZGIBCIcET5/f5FBWGxWDh06BCbN2+mqqoKg8EQZeuMjIzw3nvv8Ze//AWn00lxcTFtbW2YzWaMRiO7du3ixIkT6ohw+PmVtT/SSVJF8dRTT0U4dVKxtKQkSWzfvp2ZmZkI7+D4+Djd3d3qOhlKttyxsbGILnN9fT03b96MaqosFguPPfYYBw8exGq1RgnB7/fz6aefcuHCBU6dOqUatxCKNz179ixPPvkkQggaGxvR6XT4fD4GBwfV4GSFysrKtC7pnTRR6HQ6NdEXwI0bNyJC7ZJFaWmpWkuEx10Eg0GuXLlCU1OT+iRWV1dH3VyLxRIze94jjzzC17/+9agY0WAwyM2bN3nzzTcj1vWaz+XLlzl06BA6nY7169ezYcMGOjs7Y9pJ6V7lOGlnt1gsEcssjo+PLzs+UqfTLStdsUajoa2tjZmZGW7duhX1vdPpjDB0CwoK4q6qz549G5EYPRgMcu3aNV5++WVefvllPv7440XtpYmJCXX1YiGEuiSVkqcjHJPJtOx1zxJJ0moKpYpUWMlSDmazeVlJ0r7whS+wadMmzp8/j9vtRqPRYLVaKS0txWKxUFNTExHzCdFrdC3kQJqenuYXv/gFP/3pT7l79y4nTpzg+vXrcTeJfr+fBw8esGnTJiRJUvNzDgwM0NjYGDGSrIy5pIukiWLHjh0RolByZS4HZVJPPOTl5VFXV0dvby/37t1j27ZtNDU1qdMCdDodN2/e5OrVqxw4cGDBnNgmk4nq6uqYi7f19PTw4x//mImJiRWNup47d46CggIKCwvp6OhQfSCxBGA0GhO61uhySIoohBAR/fWBgYGkG0719fVYLBZ8Ph9PP/00Go0Gp9PJ5cuXuXv3Ll6vV7UxysvL1eBgu90ekbNKCLGgYBRX9kpxuVwcP3486pj9/f3qNAalDDabLSU2WCySIgqj0cjBgwfV9wMDA0kPpg0Gg7hcLgYHBxkdHWV4eJjh4eGY1fupU6fUtrypqSkhicxWQyxbRInNSEcIYlJEYTKZIgzE06dPJ+M0Edy5c4e7d+8SDAaXvJDhvg6Hw4Hb7Y4w7NavX7/gIFoycDqd6tQGBZvNhl6vT0v6haT0Ph577DE1ERnA3bt3k3GaCGRZjmnJL8XExESU9zDVKxQ7nc6ocis1RTpIuCj0ej379u1T3z948IDOzs5En2ZNoMRnCiGivKhKrGc6SHjzodVqI9ILTk1NMTk5mejTJJ1UtOdlZWW0trbicrmYnJyMGD+JFQeaKhJeU3z5y1+OWMfz+PHjGT1fQ/FIhmO1WpMe11BYWEhDQwMajQaz2Zwx8ZmQYFFIkkR7e7va756YmFDzZmcy8405jUaT1LW9lIjv+WMe4ciynLaEKwkVRVFRUUTK5L6+vpQYmdmGyWSK8EvEwu/3RwyqpZKE2hS7du2KmCTzpz/9Ke2RyfEwOzsbFbCbzLGHiYkJrl27Rm1tLSaTKcpr6/f7cTgcETVYS0uLmgIqGAzy0UcfJW115ISJQgjBnj171Pder3fV64Gmip6eHmZmZiIiq1tbW5M2MUeJz3Q4HKxbt46Ghga19zEzM8ODBw+ipiVUV1ezZcsWIOTs6ujoyHxRmEwmtm7dqr6fmJjgwoULqzqmx+OJulnJYP5S1X6/PyXxDMFgkMHBwbhWGUwlCRPFli1bItrJd999d9UhZU6nk4cPH6qiqK+v5+zZswn38vn9fo4fP64ayKkSRaaSEFEIISLmi05OTvK73/0uIRN/wp9gg8GAxWJJ+ECRsiZHjhAJ6X1s2rQpwp745JNPEvakKZHPEAq6qaurS8hxs4m8vLyIOJCenp64E9avhISIYufOnWrshCzLXLp0KWFVvMPhiPCIbt26Ne61ztcKdXV1Ec40r9eb1OmXqxaF3W5X1xWFUJrD1UZth+Nyubhx44b6vqioiCeffJJdu3alPeo5FRQXF7Nz507V3pFlOSFrmCzGqm2Ko0ePqr4JWZZ58803Y06HWw3Xr1+noaFBHb0sKipi9+7dVFdX8+mnn3Lz5s2sHF9ZCrvdzuOPPx5RS4yMjESsopwMVi2KW7du0d3dzYYNGxgdHeUPf/hDIsoVgRKx9Mwzz6jCUJaDrKmpiZj4o6QIWkmgcLoxmUwUFRVhNpvZuHEj1dXVEe52r9eblN7XfMRig1VCiLhGsoqKijhw4AA2m41XXnklYYWbj8lk4umnn6a0tHTJWIPR0dEI544sy1y9enXZsZWBQIDBwcFlJ2YtKCigvLx80bZfkiRaWlpU72lJScmC4yFer5eOjg4+/PDDZZVjMWRZjnkREyKKVFJYWEhrayubN29edq7ulRAIBBgbG1u2YZeXl4fFYln1CLEsywwPD3PhwoWER4OtGVEoFBYWYrVaaW5uxmg0qqsLhq8lnq0EAgFVjL29vVy6dCkpLu01J4pwlFSEWq2W5uZmdcJufX19VHWciqWm40UZiFPo7e1lfHwcl8ulrmCQzInGa1oUC6GkAQintLQ05TGYC9Hd3R1hNPp8vpSOKn8uRZFjcRYSRdoXgcmReeREkSOKnChyRJETRY4ocqLIEUVOFDmiyIkiRxQ5UeSIIieKHFHkRJEjipwockSRE0WOKHKiyBHFoqOkOT6f5GqKHFHkRJEjipwockSRE0WOKHKiyBFFThQ5ovj/oHXtwR+0fJ0AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 158.4x158.4 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "size = 256\n",
    "\n",
    "# create a tf lite interpreter\n",
    "interpreter = tf.lite.Interpreter(os.path.join(model_dir, \"tflite\", \"model.tflite\"))\n",
    "\n",
    "# resize the input tensor\n",
    "interpreter.resize_tensor_input(0, [1, size, size, 1])\n",
    "interpreter.allocate_tensors()\n",
    "\n",
    "# Get input and output tensors.\n",
    "input_details = interpreter.get_input_details()\n",
    "output_details = interpreter.get_output_details()\n",
    "\n",
    "# Test the model on random input data.\n",
    "input_shape = input_details[0]['shape']\n",
    "sample = tf.keras.preprocessing.image.load_img(\n",
    "    path= r\"Z:\\data_sets\\dakanji_single_kanji_cnn\\6340\\300.png\",\n",
    "    color_mode=\"grayscale\",\n",
    "    target_size=(size, size)\n",
    ")\n",
    "sample = tf.keras.preprocessing.image.img_to_array(sample).reshape((1, size, size, 1))\n",
    "interpreter.set_tensor(input_details[0]['index'], sample)\n",
    "\n",
    "# run interpreter\n",
    "interpreter.invoke()\n",
    "output_data = interpreter.get_tensor(interpreter.get_output_details()[0]['index'])\n",
    "\n",
    "# show prediction\n",
    "title = ordered_labels[output_data.argmax()] + str(output_data[0][output_data.argmax()])\n",
    "show_image(sample.reshape(size, size), title)"
   ]
  }
 ],
 "metadata": {
  "interpreter": {
   "hash": "7b66e6ad73e4584ec1c4fbbc9ad8719d99baf4738cd57ba39e580b8e9994d93a"
  },
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.8.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}