basic_training.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "from zipfile import ZipFile\n",
    "import numpy as np\n",
    "import mxnet as mx\n",
    "\n",
    "display_step=5"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "/home/bt3/15CS10057/dl\n"
     ]
    }
   ],
   "source": [
    "class DataLoader(object):\n",
    "    DIR=None\n",
    "    import os\n",
    "    print(os.getcwd())\n",
    "    \n",
    "    def __init__(self):\n",
    "        self.DIR = './'\n",
    "    \n",
    "    # Returns images and labels corresponding for training and testing. Default mode is train. \n",
    "    # For retrieving test data pass mode as 'test' in function call.\n",
    "    def load_data(self, mode = 'train'):\n",
    "        label_filename = mode + '_labels'\n",
    "        image_filename = mode + '_images'\n",
    "        label_zip = self.DIR + label_filename + '.zip'\n",
    "        image_zip = self.DIR + image_filename + '.zip'\n",
    "        with ZipFile(label_zip, 'r') as lblzip:\n",
    "            labels = np.frombuffer(lblzip.read(label_filename), dtype=np.uint8, offset=8)\n",
    "        with ZipFile(image_zip, 'r') as imgzip:\n",
    "            images = np.frombuffer(imgzip.read(image_filename), dtype=np.uint8, offset=16).reshape(len(labels), 784)\n",
    "        return images, labels\n",
    "    \n",
    "    \n",
    "    def create_batches(self,train_size):\n",
    "        '''\n",
    "        Handles the train-validation split \n",
    "        Params:\n",
    "            train_size: ratio of train_size : float\n",
    "        Returns:\n",
    "            images, labels, validation images, validation labels in that order\n",
    "        '''\n",
    "        img,labels=self.load_data()\n",
    "        n=labels.shape[0]\n",
    "        #split into train and validation data\n",
    "        validation_img=img[int(train_size*n):]\n",
    "        img=img[:int(train_size*n)]\n",
    "        validation_labels=labels[int(train_size*n):]\n",
    "        labels=labels[:int(train_size*n)]\n",
    "        return img, labels, validation_img, validation_labels        "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Network 1:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [],
   "source": [
    "n_epochs=50\n",
    "data = mx.symbol.Variable('data')\n",
    "fc1  = mx.symbol.FullyConnected(data = data, num_hidden=512,name='fc1')\n",
    "act1 = mx.symbol.Activation(data = fc1, act_type=\"relu\",name='act1')\n",
    "fc2  = mx.symbol.FullyConnected(data = act1, num_hidden = 128,name='fc2')\n",
    "act2 = mx.symbol.Activation(data = fc2, act_type=\"relu\",name='act2')\n",
    "fc3  = mx.symbol.FullyConnected(data = act2, num_hidden = 64,name='fc3')\n",
    "act3 = mx.symbol.Activation(data = fc3, act_type=\"relu\",name='act3')\n",
    "fc4  = mx.symbol.FullyConnected(data = act3, num_hidden = 32,name='fc4')\n",
    "act4 = mx.symbol.Activation(data = fc4, act_type=\"relu\",name='act4')\n",
    "fc5  = mx.symbol.FullyConnected(data = act4, num_hidden = 16,name='fc5')\n",
    "act5 = mx.symbol.Activation(data = fc5, act_type=\"relu\",name='act5')\n",
    "#output layer\n",
    "fc6  = mx.symbol.FullyConnected(data = act5, num_hidden=10,name='fc6')\n",
    "mlp  = mx.symbol.SoftmaxOutput(data = fc6,name='softmax')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/svg+xml": [
       "<?xml version=\"1.0\" encoding=\"UTF-8\" standalone=\"no\"?>\n",
       "<!DOCTYPE svg PUBLIC \"-//W3C//DTD SVG 1.1//EN\"\n",
       " \"http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd\">\n",
       "<!-- Generated by graphviz version 2.30.1 (20150306.0020)\n",
       " -->\n",
       "<!-- Title: plot Pages: 1 -->\n",
       "<svg width=\"264pt\" height=\"1194pt\"\n",
       " viewBox=\"0.00 0.00 264.00 1194.00\" xmlns=\"http://www.w3.org/2000/svg\" xmlns:xlink=\"http://www.w3.org/1999/xlink\">\n",
       "<g id=\"graph0\" class=\"graph\" transform=\"scale(1 1) rotate(0) translate(4 1190)\">\n",
       "<title>plot</title>\n",
       "<polygon fill=\"white\" stroke=\"white\" points=\"-4,5 -4,-1190 261,-1190 261,5 -4,5\"/>\n",
       "<!-- data -->\n",
       "<g id=\"node1\" class=\"node\"><title>data</title>\n",
       "<ellipse fill=\"#8dd3c7\" stroke=\"black\" cx=\"68\" cy=\"-29\" rx=\"47\" ry=\"29\"/>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-25.3\" font-family=\"Times,serif\" font-size=\"14.00\">data</text>\n",
       "</g>\n",
       "<!-- fc1 -->\n",
       "<g id=\"node2\" class=\"node\"><title>fc1</title>\n",
       "<ellipse fill=\"#fb8072\" stroke=\"black\" cx=\"68\" cy=\"-123\" rx=\"68.2532\" ry=\"29\"/>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-126.8\" font-family=\"Times,serif\" font-size=\"14.00\">FullyConnected</text>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-111.8\" font-family=\"Times,serif\" font-size=\"14.00\">512</text>\n",
       "</g>\n",
       "<!-- fc1&#45;&gt;data -->\n",
       "<g id=\"edge1\" class=\"edge\"><title>fc1&#45;&gt;data</title>\n",
       "<path fill=\"none\" stroke=\"black\" d=\"M68,-83.7443C68,-75.2043 68,-66.2977 68,-58.2479\"/>\n",
       "<polygon fill=\"black\" stroke=\"black\" points=\"68,-93.8971 63.5001,-83.897 68,-88.8971 68.0001,-83.8971 68.0001,-83.8971 68.0001,-83.8971 68,-88.8971 72.5001,-83.8971 68,-93.8971 68,-93.8971\"/>\n",
       "</g>\n",
       "<!-- act1 -->\n",
       "<g id=\"node3\" class=\"node\"><title>act1</title>\n",
       "<ellipse fill=\"#ffffb3\" stroke=\"black\" cx=\"68\" cy=\"-217\" rx=\"48.4635\" ry=\"29\"/>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-220.8\" font-family=\"Times,serif\" font-size=\"14.00\">Activation</text>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-205.8\" font-family=\"Times,serif\" font-size=\"14.00\">relu</text>\n",
       "</g>\n",
       "<!-- act1&#45;&gt;fc1 -->\n",
       "<g id=\"edge2\" class=\"edge\"><title>act1&#45;&gt;fc1</title>\n",
       "<path fill=\"none\" stroke=\"black\" d=\"M68,-177.744C68,-169.204 68,-160.298 68,-152.248\"/>\n",
       "<polygon fill=\"black\" stroke=\"black\" points=\"68,-187.897 63.5001,-177.897 68,-182.897 68.0001,-177.897 68.0001,-177.897 68.0001,-177.897 68,-182.897 72.5001,-177.897 68,-187.897 68,-187.897\"/>\n",
       "</g>\n",
       "<!-- fc2 -->\n",
       "<g id=\"node4\" class=\"node\"><title>fc2</title>\n",
       "<ellipse fill=\"#fb8072\" stroke=\"black\" cx=\"68\" cy=\"-311\" rx=\"68.2532\" ry=\"29\"/>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-314.8\" font-family=\"Times,serif\" font-size=\"14.00\">FullyConnected</text>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-299.8\" font-family=\"Times,serif\" font-size=\"14.00\">128</text>\n",
       "</g>\n",
       "<!-- fc2&#45;&gt;act1 -->\n",
       "<g id=\"edge3\" class=\"edge\"><title>fc2&#45;&gt;act1</title>\n",
       "<path fill=\"none\" stroke=\"black\" d=\"M68,-271.744C68,-263.204 68,-254.298 68,-246.248\"/>\n",
       "<polygon fill=\"black\" stroke=\"black\" points=\"68,-281.897 63.5001,-271.897 68,-276.897 68.0001,-271.897 68.0001,-271.897 68.0001,-271.897 68,-276.897 72.5001,-271.897 68,-281.897 68,-281.897\"/>\n",
       "</g>\n",
       "<!-- act2 -->\n",
       "<g id=\"node5\" class=\"node\"><title>act2</title>\n",
       "<ellipse fill=\"#ffffb3\" stroke=\"black\" cx=\"68\" cy=\"-405\" rx=\"48.4635\" ry=\"29\"/>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-408.8\" font-family=\"Times,serif\" font-size=\"14.00\">Activation</text>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-393.8\" font-family=\"Times,serif\" font-size=\"14.00\">relu</text>\n",
       "</g>\n",
       "<!-- act2&#45;&gt;fc2 -->\n",
       "<g id=\"edge4\" class=\"edge\"><title>act2&#45;&gt;fc2</title>\n",
       "<path fill=\"none\" stroke=\"black\" d=\"M68,-365.744C68,-357.204 68,-348.298 68,-340.248\"/>\n",
       "<polygon fill=\"black\" stroke=\"black\" points=\"68,-375.897 63.5001,-365.897 68,-370.897 68.0001,-365.897 68.0001,-365.897 68.0001,-365.897 68,-370.897 72.5001,-365.897 68,-375.897 68,-375.897\"/>\n",
       "</g>\n",
       "<!-- fc3 -->\n",
       "<g id=\"node6\" class=\"node\"><title>fc3</title>\n",
       "<ellipse fill=\"#fb8072\" stroke=\"black\" cx=\"68\" cy=\"-499\" rx=\"68.2532\" ry=\"29\"/>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-502.8\" font-family=\"Times,serif\" font-size=\"14.00\">FullyConnected</text>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-487.8\" font-family=\"Times,serif\" font-size=\"14.00\">64</text>\n",
       "</g>\n",
       "<!-- fc3&#45;&gt;act2 -->\n",
       "<g id=\"edge5\" class=\"edge\"><title>fc3&#45;&gt;act2</title>\n",
       "<path fill=\"none\" stroke=\"black\" d=\"M68,-459.744C68,-451.204 68,-442.298 68,-434.248\"/>\n",
       "<polygon fill=\"black\" stroke=\"black\" points=\"68,-469.897 63.5001,-459.897 68,-464.897 68.0001,-459.897 68.0001,-459.897 68.0001,-459.897 68,-464.897 72.5001,-459.897 68,-469.897 68,-469.897\"/>\n",
       "</g>\n",
       "<!-- act3 -->\n",
       "<g id=\"node7\" class=\"node\"><title>act3</title>\n",
       "<ellipse fill=\"#ffffb3\" stroke=\"black\" cx=\"68\" cy=\"-593\" rx=\"48.4635\" ry=\"29\"/>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-596.8\" font-family=\"Times,serif\" font-size=\"14.00\">Activation</text>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-581.8\" font-family=\"Times,serif\" font-size=\"14.00\">relu</text>\n",
       "</g>\n",
       "<!-- act3&#45;&gt;fc3 -->\n",
       "<g id=\"edge6\" class=\"edge\"><title>act3&#45;&gt;fc3</title>\n",
       "<path fill=\"none\" stroke=\"black\" d=\"M68,-553.744C68,-545.204 68,-536.298 68,-528.248\"/>\n",
       "<polygon fill=\"black\" stroke=\"black\" points=\"68,-563.897 63.5001,-553.897 68,-558.897 68.0001,-553.897 68.0001,-553.897 68.0001,-553.897 68,-558.897 72.5001,-553.897 68,-563.897 68,-563.897\"/>\n",
       "</g>\n",
       "<!-- fc4 -->\n",
       "<g id=\"node8\" class=\"node\"><title>fc4</title>\n",
       "<ellipse fill=\"#fb8072\" stroke=\"black\" cx=\"68\" cy=\"-687\" rx=\"68.2532\" ry=\"29\"/>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-690.8\" font-family=\"Times,serif\" font-size=\"14.00\">FullyConnected</text>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-675.8\" font-family=\"Times,serif\" font-size=\"14.00\">32</text>\n",
       "</g>\n",
       "<!-- fc4&#45;&gt;act3 -->\n",
       "<g id=\"edge7\" class=\"edge\"><title>fc4&#45;&gt;act3</title>\n",
       "<path fill=\"none\" stroke=\"black\" d=\"M68,-647.744C68,-639.204 68,-630.298 68,-622.248\"/>\n",
       "<polygon fill=\"black\" stroke=\"black\" points=\"68,-657.897 63.5001,-647.897 68,-652.897 68.0001,-647.897 68.0001,-647.897 68.0001,-647.897 68,-652.897 72.5001,-647.897 68,-657.897 68,-657.897\"/>\n",
       "</g>\n",
       "<!-- act4 -->\n",
       "<g id=\"node9\" class=\"node\"><title>act4</title>\n",
       "<ellipse fill=\"#ffffb3\" stroke=\"black\" cx=\"68\" cy=\"-781\" rx=\"48.4635\" ry=\"29\"/>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-784.8\" font-family=\"Times,serif\" font-size=\"14.00\">Activation</text>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-769.8\" font-family=\"Times,serif\" font-size=\"14.00\">relu</text>\n",
       "</g>\n",
       "<!-- act4&#45;&gt;fc4 -->\n",
       "<g id=\"edge8\" class=\"edge\"><title>act4&#45;&gt;fc4</title>\n",
       "<path fill=\"none\" stroke=\"black\" d=\"M68,-741.744C68,-733.204 68,-724.298 68,-716.248\"/>\n",
       "<polygon fill=\"black\" stroke=\"black\" points=\"68,-751.897 63.5001,-741.897 68,-746.897 68.0001,-741.897 68.0001,-741.897 68.0001,-741.897 68,-746.897 72.5001,-741.897 68,-751.897 68,-751.897\"/>\n",
       "</g>\n",
       "<!-- fc5 -->\n",
       "<g id=\"node10\" class=\"node\"><title>fc5</title>\n",
       "<ellipse fill=\"#fb8072\" stroke=\"black\" cx=\"68\" cy=\"-875\" rx=\"68.2532\" ry=\"29\"/>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-878.8\" font-family=\"Times,serif\" font-size=\"14.00\">FullyConnected</text>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-863.8\" font-family=\"Times,serif\" font-size=\"14.00\">16</text>\n",
       "</g>\n",
       "<!-- fc5&#45;&gt;act4 -->\n",
       "<g id=\"edge9\" class=\"edge\"><title>fc5&#45;&gt;act4</title>\n",
       "<path fill=\"none\" stroke=\"black\" d=\"M68,-835.744C68,-827.204 68,-818.298 68,-810.248\"/>\n",
       "<polygon fill=\"black\" stroke=\"black\" points=\"68,-845.897 63.5001,-835.897 68,-840.897 68.0001,-835.897 68.0001,-835.897 68.0001,-835.897 68,-840.897 72.5001,-835.897 68,-845.897 68,-845.897\"/>\n",
       "</g>\n",
       "<!-- act5 -->\n",
       "<g id=\"node11\" class=\"node\"><title>act5</title>\n",
       "<ellipse fill=\"#ffffb3\" stroke=\"black\" cx=\"68\" cy=\"-969\" rx=\"48.4635\" ry=\"29\"/>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-972.8\" font-family=\"Times,serif\" font-size=\"14.00\">Activation</text>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-957.8\" font-family=\"Times,serif\" font-size=\"14.00\">relu</text>\n",
       "</g>\n",
       "<!-- act5&#45;&gt;fc5 -->\n",
       "<g id=\"edge10\" class=\"edge\"><title>act5&#45;&gt;fc5</title>\n",
       "<path fill=\"none\" stroke=\"black\" d=\"M68,-929.744C68,-921.204 68,-912.298 68,-904.248\"/>\n",
       "<polygon fill=\"black\" stroke=\"black\" points=\"68,-939.897 63.5001,-929.897 68,-934.897 68.0001,-929.897 68.0001,-929.897 68.0001,-929.897 68,-934.897 72.5001,-929.897 68,-939.897 68,-939.897\"/>\n",
       "</g>\n",
       "<!-- fc6 -->\n",
       "<g id=\"node12\" class=\"node\"><title>fc6</title>\n",
       "<ellipse fill=\"#fb8072\" stroke=\"black\" cx=\"68\" cy=\"-1063\" rx=\"68.2532\" ry=\"29\"/>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-1066.8\" font-family=\"Times,serif\" font-size=\"14.00\">FullyConnected</text>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-1051.8\" font-family=\"Times,serif\" font-size=\"14.00\">10</text>\n",
       "</g>\n",
       "<!-- fc6&#45;&gt;act5 -->\n",
       "<g id=\"edge11\" class=\"edge\"><title>fc6&#45;&gt;act5</title>\n",
       "<path fill=\"none\" stroke=\"black\" d=\"M68,-1023.74C68,-1015.2 68,-1006.3 68,-998.248\"/>\n",
       "<polygon fill=\"black\" stroke=\"black\" points=\"68,-1033.9 63.5001,-1023.9 68,-1028.9 68.0001,-1023.9 68.0001,-1023.9 68.0001,-1023.9 68,-1028.9 72.5001,-1023.9 68,-1033.9 68,-1033.9\"/>\n",
       "</g>\n",
       "<!-- softmax_label -->\n",
       "<g id=\"node13\" class=\"node\"><title>softmax_label</title>\n",
       "<ellipse fill=\"#8dd3c7\" stroke=\"black\" cx=\"205\" cy=\"-1063\" rx=\"51.2967\" ry=\"29\"/>\n",
       "<text text-anchor=\"middle\" x=\"205\" y=\"-1059.3\" font-family=\"Times,serif\" font-size=\"14.00\">softmax_label</text>\n",
       "</g>\n",
       "<!-- softmax -->\n",
       "<g id=\"node14\" class=\"node\"><title>softmax</title>\n",
       "<ellipse fill=\"#fccde5\" stroke=\"black\" cx=\"136\" cy=\"-1157\" rx=\"47\" ry=\"29\"/>\n",
       "<text text-anchor=\"middle\" x=\"136\" y=\"-1153.3\" font-family=\"Times,serif\" font-size=\"14.00\">softmax</text>\n",
       "</g>\n",
       "<!-- softmax&#45;&gt;fc6 -->\n",
       "<g id=\"edge12\" class=\"edge\"><title>softmax&#45;&gt;fc6</title>\n",
       "<path fill=\"none\" stroke=\"black\" d=\"M111.08,-1122.28C103.407,-1111.9 95.0676,-1100.62 87.7883,-1090.77\"/>\n",
       "<polygon fill=\"black\" stroke=\"black\" points=\"117.058,-1130.37 107.495,-1125.01 114.086,-1126.35 111.114,-1122.33 111.114,-1122.33 111.114,-1122.33 114.086,-1126.35 114.733,-1119.66 117.058,-1130.37 117.058,-1130.37\"/>\n",
       "</g>\n",
       "<!-- softmax&#45;&gt;softmax_label -->\n",
       "<g id=\"edge13\" class=\"edge\"><title>softmax&#45;&gt;softmax_label</title>\n",
       "<path fill=\"none\" stroke=\"black\" d=\"M161.394,-1122.14C169.321,-1111.57 177.938,-1100.08 185.403,-1090.13\"/>\n",
       "<polygon fill=\"black\" stroke=\"black\" points=\"155.22,-1130.37 157.62,-1119.67 158.22,-1126.37 161.22,-1122.37 161.22,-1122.37 161.22,-1122.37 158.22,-1126.37 164.82,-1125.07 155.22,-1130.37 155.22,-1130.37\"/>\n",
       "</g>\n",
       "</g>\n",
       "</svg>\n"
      ],
      "text/plain": [
       "<graphviz.dot.Digraph at 0x7fa2210eb588>"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mx.viz.plot_network(mlp, node_attrs={\"shape\":\"oval\",\"fixedsize\":\"false\"})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_loader=DataLoader()\n",
    "data,label,val_data,val_labels=data_loader.create_batches(train_size=0.7)\n",
    "batch_size = 100\n",
    "ntrain = int(data.shape[0]*0.8)\n",
    "train_iter = mx.io.NDArrayIter(data, label, batch_size, shuffle=True)\n",
    "val_iter = mx.io.NDArrayIter(val_data, val_labels,batch_size=val_data.shape[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "#starting the training of network 1\n",
    "\n",
    "# data for plotting finally\n",
    "train_1_loss=[]\n",
    "train_1_acc=[]\n",
    "val_1_loss=[]\n",
    "val_1_acc=[]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1, Training ('accuracy', 0.48988095238095236)\n",
      "Epoch 1, Validation ('accuracy', 0.67500000000000004)\n",
      "Epoch 6, Training ('accuracy', 0.81292857142857144)\n",
      "Epoch 6, Validation ('accuracy', 0.80494444444444446)\n",
      "Epoch 11, Training ('accuracy', 0.83907142857142858)\n",
      "Epoch 11, Validation ('accuracy', 0.82627777777777778)\n",
      "Epoch 16, Training ('accuracy', 0.85207142857142859)\n",
      "Epoch 16, Validation ('accuracy', 0.83966666666666667)\n",
      "Epoch 21, Training ('accuracy', 0.86247619047619051)\n",
      "Epoch 21, Validation ('accuracy', 0.84788888888888891)\n",
      "Epoch 26, Training ('accuracy', 0.87009523809523814)\n",
      "Epoch 26, Validation ('accuracy', 0.85211111111111115)\n",
      "Epoch 31, Training ('accuracy', 0.87604761904761907)\n",
      "Epoch 31, Validation ('accuracy', 0.85550000000000004)\n",
      "Epoch 36, Training ('accuracy', 0.88195238095238093)\n",
      "Epoch 36, Validation ('accuracy', 0.85855555555555552)\n",
      "Epoch 41, Training ('accuracy', 0.88661904761904764)\n",
      "Epoch 41, Validation ('accuracy', 0.86083333333333334)\n",
      "Epoch 46, Training ('accuracy', 0.89133333333333331)\n",
      "Epoch 46, Validation ('accuracy', 0.86288888888888893)\n",
      "Epoch 50, Training ('accuracy', 0.89504761904761909)\n",
      "Epoch 50, Validation ('accuracy', 0.86388888888888893)\n"
     ]
    }
   ],
   "source": [
    "mod = mx.mod.Module(symbol=mlp,\n",
    "                    context=mx.cpu(),\n",
    "                    data_names=['data'],\n",
    "                    label_names=['softmax_label'])\n",
    "\n",
    "# allocate memory given the input data and label shapes\n",
    "mod.bind(data_shapes=train_iter.provide_data, label_shapes=train_iter.provide_label)\n",
    "# initialize parameters by uniform random numbers\n",
    "mod.init_params(initializer=mx.init.Uniform(scale=.1))\n",
    "learn_rate=0.001\n",
    "mod.init_optimizer(optimizer='sgd', optimizer_params=(('learning_rate', learn_rate), ))\n",
    "# use accuracy as the metric\n",
    "metric = mx.metric.CrossEntropy()\n",
    "metric2= mx.metric.create('acc')\n",
    "for epoch in range(n_epochs):\n",
    "    train_iter.reset()\n",
    "    metric.reset()\n",
    "    metric2.reset()\n",
    "    for batch in train_iter:\n",
    "        mod.forward(batch, is_train=True)       # compute predictions\n",
    "        mod.update_metric(metric, batch.label)  # accumulate prediction loss\n",
    "        mod.update_metric(metric2, batch.label)  # accumulate prediction accuracy\n",
    "        mod.backward()                          # compute gradients\n",
    "        mod.update()                            # update parameters\n",
    "    loss=metric.get()\n",
    "    acc=metric2.get()\n",
    "    train_1_loss.append(loss)\n",
    "    train_1_acc.append(acc)\n",
    "    if epoch%display_step==0 or epoch==n_epochs-1:\n",
    "        print('Epoch %d, Training %s' % (epoch+1,acc))\n",
    "    #validation error and accuracy\n",
    "    metric.reset()\n",
    "    metric2.reset()\n",
    "    val_iter.reset()\n",
    "    for batch in val_iter:\n",
    "        mod.forward(batch)\n",
    "        mod.update_metric(metric,batch.label)\n",
    "        mod.update_metric(metric2,batch.label)\n",
    "    loss=metric.get()\n",
    "    acc=metric2.get()\n",
    "    val_1_loss.append(loss)\n",
    "    val_1_acc.append(acc)\n",
    "    if epoch%display_step==0 or epoch==n_epochs-1:\n",
    "        print('Epoch %d, Validation %s' % (epoch+1,acc))\n",
    "        \n",
    "mod.save_params('network1')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "***\n",
    "\n",
    "\n",
    "### Network 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "n_epochs=50\n",
    "data_2 = mx.symbol.Variable('data')\n",
    "fc1_2  = mx.symbol.FullyConnected(data = data_2, num_hidden=1024,name='fc1')\n",
    "act1_2 = mx.symbol.Activation(data = fc1_2, act_type=\"relu\",name='act1')\n",
    "fc2_2  = mx.symbol.FullyConnected(data = act1_2, num_hidden = 512,name='fc2')\n",
    "act2_2 = mx.symbol.Activation(data = fc2_2, act_type=\"relu\",name='act2')\n",
    "fc3_2  = mx.symbol.FullyConnected(data = act2_2, num_hidden = 256,name='fc3')\n",
    "act3_2 = mx.symbol.Activation(data = fc3_2, act_type=\"relu\",name='act3')\n",
    "\n",
    "#output layer\n",
    "fc4_2  = mx.symbol.FullyConnected(data = act3_2, num_hidden=10,name='fc4')\n",
    "mlp_2 = mx.symbol.SoftmaxOutput(data = fc4_2,name='softmax')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/svg+xml": [
       "<?xml version=\"1.0\" encoding=\"UTF-8\" standalone=\"no\"?>\n",
       "<!DOCTYPE svg PUBLIC \"-//W3C//DTD SVG 1.1//EN\"\n",
       " \"http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd\">\n",
       "<!-- Generated by graphviz version 2.30.1 (20150306.0020)\n",
       " -->\n",
       "<!-- Title: plot Pages: 1 -->\n",
       "<svg width=\"264pt\" height=\"818pt\"\n",
       " viewBox=\"0.00 0.00 264.00 818.00\" xmlns=\"http://www.w3.org/2000/svg\" xmlns:xlink=\"http://www.w3.org/1999/xlink\">\n",
       "<g id=\"graph0\" class=\"graph\" transform=\"scale(1 1) rotate(0) translate(4 814)\">\n",
       "<title>plot</title>\n",
       "<polygon fill=\"white\" stroke=\"white\" points=\"-4,5 -4,-814 261,-814 261,5 -4,5\"/>\n",
       "<!-- data -->\n",
       "<g id=\"node1\" class=\"node\"><title>data</title>\n",
       "<ellipse fill=\"#8dd3c7\" stroke=\"black\" cx=\"68\" cy=\"-29\" rx=\"47\" ry=\"29\"/>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-25.3\" font-family=\"Times,serif\" font-size=\"14.00\">data</text>\n",
       "</g>\n",
       "<!-- fullyconnected6 -->\n",
       "<g id=\"node2\" class=\"node\"><title>fullyconnected6</title>\n",
       "<ellipse fill=\"#fb8072\" stroke=\"black\" cx=\"68\" cy=\"-123\" rx=\"68.2532\" ry=\"29\"/>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-126.8\" font-family=\"Times,serif\" font-size=\"14.00\">FullyConnected</text>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-111.8\" font-family=\"Times,serif\" font-size=\"14.00\">1024</text>\n",
       "</g>\n",
       "<!-- fullyconnected6&#45;&gt;data -->\n",
       "<g id=\"edge1\" class=\"edge\"><title>fullyconnected6&#45;&gt;data</title>\n",
       "<path fill=\"none\" stroke=\"black\" d=\"M68,-83.7443C68,-75.2043 68,-66.2977 68,-58.2479\"/>\n",
       "<polygon fill=\"black\" stroke=\"black\" points=\"68,-93.8971 63.5001,-83.897 68,-88.8971 68.0001,-83.8971 68.0001,-83.8971 68.0001,-83.8971 68,-88.8971 72.5001,-83.8971 68,-93.8971 68,-93.8971\"/>\n",
       "</g>\n",
       "<!-- activation5 -->\n",
       "<g id=\"node3\" class=\"node\"><title>activation5</title>\n",
       "<ellipse fill=\"#ffffb3\" stroke=\"black\" cx=\"68\" cy=\"-217\" rx=\"48.4635\" ry=\"29\"/>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-220.8\" font-family=\"Times,serif\" font-size=\"14.00\">Activation</text>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-205.8\" font-family=\"Times,serif\" font-size=\"14.00\">relu</text>\n",
       "</g>\n",
       "<!-- activation5&#45;&gt;fullyconnected6 -->\n",
       "<g id=\"edge2\" class=\"edge\"><title>activation5&#45;&gt;fullyconnected6</title>\n",
       "<path fill=\"none\" stroke=\"black\" d=\"M68,-177.744C68,-169.204 68,-160.298 68,-152.248\"/>\n",
       "<polygon fill=\"black\" stroke=\"black\" points=\"68,-187.897 63.5001,-177.897 68,-182.897 68.0001,-177.897 68.0001,-177.897 68.0001,-177.897 68,-182.897 72.5001,-177.897 68,-187.897 68,-187.897\"/>\n",
       "</g>\n",
       "<!-- fullyconnected7 -->\n",
       "<g id=\"node4\" class=\"node\"><title>fullyconnected7</title>\n",
       "<ellipse fill=\"#fb8072\" stroke=\"black\" cx=\"68\" cy=\"-311\" rx=\"68.2532\" ry=\"29\"/>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-314.8\" font-family=\"Times,serif\" font-size=\"14.00\">FullyConnected</text>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-299.8\" font-family=\"Times,serif\" font-size=\"14.00\">512</text>\n",
       "</g>\n",
       "<!-- fullyconnected7&#45;&gt;activation5 -->\n",
       "<g id=\"edge3\" class=\"edge\"><title>fullyconnected7&#45;&gt;activation5</title>\n",
       "<path fill=\"none\" stroke=\"black\" d=\"M68,-271.744C68,-263.204 68,-254.298 68,-246.248\"/>\n",
       "<polygon fill=\"black\" stroke=\"black\" points=\"68,-281.897 63.5001,-271.897 68,-276.897 68.0001,-271.897 68.0001,-271.897 68.0001,-271.897 68,-276.897 72.5001,-271.897 68,-281.897 68,-281.897\"/>\n",
       "</g>\n",
       "<!-- activation6 -->\n",
       "<g id=\"node5\" class=\"node\"><title>activation6</title>\n",
       "<ellipse fill=\"#ffffb3\" stroke=\"black\" cx=\"68\" cy=\"-405\" rx=\"48.4635\" ry=\"29\"/>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-408.8\" font-family=\"Times,serif\" font-size=\"14.00\">Activation</text>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-393.8\" font-family=\"Times,serif\" font-size=\"14.00\">relu</text>\n",
       "</g>\n",
       "<!-- activation6&#45;&gt;fullyconnected7 -->\n",
       "<g id=\"edge4\" class=\"edge\"><title>activation6&#45;&gt;fullyconnected7</title>\n",
       "<path fill=\"none\" stroke=\"black\" d=\"M68,-365.744C68,-357.204 68,-348.298 68,-340.248\"/>\n",
       "<polygon fill=\"black\" stroke=\"black\" points=\"68,-375.897 63.5001,-365.897 68,-370.897 68.0001,-365.897 68.0001,-365.897 68.0001,-365.897 68,-370.897 72.5001,-365.897 68,-375.897 68,-375.897\"/>\n",
       "</g>\n",
       "<!-- fullyconnected8 -->\n",
       "<g id=\"node6\" class=\"node\"><title>fullyconnected8</title>\n",
       "<ellipse fill=\"#fb8072\" stroke=\"black\" cx=\"68\" cy=\"-499\" rx=\"68.2532\" ry=\"29\"/>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-502.8\" font-family=\"Times,serif\" font-size=\"14.00\">FullyConnected</text>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-487.8\" font-family=\"Times,serif\" font-size=\"14.00\">256</text>\n",
       "</g>\n",
       "<!-- fullyconnected8&#45;&gt;activation6 -->\n",
       "<g id=\"edge5\" class=\"edge\"><title>fullyconnected8&#45;&gt;activation6</title>\n",
       "<path fill=\"none\" stroke=\"black\" d=\"M68,-459.744C68,-451.204 68,-442.298 68,-434.248\"/>\n",
       "<polygon fill=\"black\" stroke=\"black\" points=\"68,-469.897 63.5001,-459.897 68,-464.897 68.0001,-459.897 68.0001,-459.897 68.0001,-459.897 68,-464.897 72.5001,-459.897 68,-469.897 68,-469.897\"/>\n",
       "</g>\n",
       "<!-- activation7 -->\n",
       "<g id=\"node7\" class=\"node\"><title>activation7</title>\n",
       "<ellipse fill=\"#ffffb3\" stroke=\"black\" cx=\"68\" cy=\"-593\" rx=\"48.4635\" ry=\"29\"/>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-596.8\" font-family=\"Times,serif\" font-size=\"14.00\">Activation</text>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-581.8\" font-family=\"Times,serif\" font-size=\"14.00\">relu</text>\n",
       "</g>\n",
       "<!-- activation7&#45;&gt;fullyconnected8 -->\n",
       "<g id=\"edge6\" class=\"edge\"><title>activation7&#45;&gt;fullyconnected8</title>\n",
       "<path fill=\"none\" stroke=\"black\" d=\"M68,-553.744C68,-545.204 68,-536.298 68,-528.248\"/>\n",
       "<polygon fill=\"black\" stroke=\"black\" points=\"68,-563.897 63.5001,-553.897 68,-558.897 68.0001,-553.897 68.0001,-553.897 68.0001,-553.897 68,-558.897 72.5001,-553.897 68,-563.897 68,-563.897\"/>\n",
       "</g>\n",
       "<!-- fullyconnected9 -->\n",
       "<g id=\"node8\" class=\"node\"><title>fullyconnected9</title>\n",
       "<ellipse fill=\"#fb8072\" stroke=\"black\" cx=\"68\" cy=\"-687\" rx=\"68.2532\" ry=\"29\"/>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-690.8\" font-family=\"Times,serif\" font-size=\"14.00\">FullyConnected</text>\n",
       "<text text-anchor=\"middle\" x=\"68\" y=\"-675.8\" font-family=\"Times,serif\" font-size=\"14.00\">10</text>\n",
       "</g>\n",
       "<!-- fullyconnected9&#45;&gt;activation7 -->\n",
       "<g id=\"edge7\" class=\"edge\"><title>fullyconnected9&#45;&gt;activation7</title>\n",
       "<path fill=\"none\" stroke=\"black\" d=\"M68,-647.744C68,-639.204 68,-630.298 68,-622.248\"/>\n",
       "<polygon fill=\"black\" stroke=\"black\" points=\"68,-657.897 63.5001,-647.897 68,-652.897 68.0001,-647.897 68.0001,-647.897 68.0001,-647.897 68,-652.897 72.5001,-647.897 68,-657.897 68,-657.897\"/>\n",
       "</g>\n",
       "<!-- softmax_label -->\n",
       "<g id=\"node9\" class=\"node\"><title>softmax_label</title>\n",
       "<ellipse fill=\"#8dd3c7\" stroke=\"black\" cx=\"205\" cy=\"-687\" rx=\"51.2967\" ry=\"29\"/>\n",
       "<text text-anchor=\"middle\" x=\"205\" y=\"-683.3\" font-family=\"Times,serif\" font-size=\"14.00\">softmax_label</text>\n",
       "</g>\n",
       "<!-- softmax -->\n",
       "<g id=\"node10\" class=\"node\"><title>softmax</title>\n",
       "<ellipse fill=\"#fccde5\" stroke=\"black\" cx=\"136\" cy=\"-781\" rx=\"47\" ry=\"29\"/>\n",
       "<text text-anchor=\"middle\" x=\"136\" y=\"-777.3\" font-family=\"Times,serif\" font-size=\"14.00\">softmax</text>\n",
       "</g>\n",
       "<!-- softmax&#45;&gt;fullyconnected9 -->\n",
       "<g id=\"edge8\" class=\"edge\"><title>softmax&#45;&gt;fullyconnected9</title>\n",
       "<path fill=\"none\" stroke=\"black\" d=\"M111.08,-746.285C103.407,-735.903 95.0676,-724.621 87.7883,-714.772\"/>\n",
       "<polygon fill=\"black\" stroke=\"black\" points=\"117.058,-754.373 107.495,-749.006 114.086,-750.352 111.114,-746.331 111.114,-746.331 111.114,-746.331 114.086,-750.352 114.733,-743.656 117.058,-754.373 117.058,-754.373\"/>\n",
       "</g>\n",
       "<!-- softmax&#45;&gt;softmax_label -->\n",
       "<g id=\"edge9\" class=\"edge\"><title>softmax&#45;&gt;softmax_label</title>\n",
       "<path fill=\"none\" stroke=\"black\" d=\"M161.394,-746.141C169.321,-735.572 177.938,-724.082 185.403,-714.129\"/>\n",
       "<polygon fill=\"black\" stroke=\"black\" points=\"155.22,-754.373 157.62,-743.673 158.22,-750.373 161.22,-746.373 161.22,-746.373 161.22,-746.373 158.22,-750.373 164.82,-749.073 155.22,-754.373 155.22,-754.373\"/>\n",
       "</g>\n",
       "</g>\n",
       "</svg>\n"
      ],
      "text/plain": [
       "<graphviz.dot.Digraph at 0x7fa2211b8fd0>"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mx.viz.plot_network(mlp_2, node_attrs={\"shape\":\"oval\",\"fixedsize\":\"false\"})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1, Training ('accuracy', 0.68414285714285716)\n",
      "Epoch 1, Validation ('accuracy', 0.75461111111111112)\n",
      "Epoch 6, Training ('accuracy', 0.83171428571428574)\n",
      "Epoch 6, Validation ('accuracy', 0.81933333333333336)\n",
      "Epoch 11, Training ('accuracy', 0.85804761904761906)\n",
      "Epoch 11, Validation ('accuracy', 0.8338888888888889)\n",
      "Epoch 16, Training ('accuracy', 0.87480952380952381)\n",
      "Epoch 16, Validation ('accuracy', 0.84183333333333332)\n",
      "Epoch 21, Training ('accuracy', 0.88664285714285718)\n",
      "Epoch 21, Validation ('accuracy', 0.84433333333333338)\n",
      "Epoch 26, Training ('accuracy', 0.89580952380952383)\n",
      "Epoch 26, Validation ('accuracy', 0.8469444444444445)\n",
      "Epoch 31, Training ('accuracy', 0.90364285714285719)\n",
      "Epoch 31, Validation ('accuracy', 0.84783333333333333)\n",
      "Epoch 36, Training ('accuracy', 0.90983333333333338)\n",
      "Epoch 36, Validation ('accuracy', 0.84988888888888892)\n",
      "Epoch 41, Training ('accuracy', 0.91554761904761905)\n",
      "Epoch 41, Validation ('accuracy', 0.85016666666666663)\n",
      "Epoch 46, Training ('accuracy', 0.92047619047619045)\n",
      "Epoch 46, Validation ('accuracy', 0.85099999999999998)\n",
      "Epoch 50, Training ('accuracy', 0.92421428571428577)\n",
      "Epoch 50, Validation ('accuracy', 0.85244444444444445)\n"
     ]
    }
   ],
   "source": [
    "train_2_loss=[]\n",
    "train_2_acc=[]\n",
    "val_2_loss=[]\n",
    "val_2_acc=[]\n",
    "\n",
    "mod_2 = mx.mod.Module(symbol=mlp_2,\n",
    "                    context=mx.cpu(),\n",
    "                    data_names=['data'],\n",
    "                    label_names=['softmax_label'])\n",
    "\n",
    "# allocate memory given the input data and label shapes\n",
    "mod_2.bind(data_shapes=train_iter.provide_data, label_shapes=train_iter.provide_label)\n",
    "# initialize parameters by uniform random numbers\n",
    "mod_2.init_params(initializer=mx.init.Uniform(scale=.1))\n",
    "learn_rate=0.001\n",
    "mod_2.init_optimizer(optimizer='sgd', optimizer_params=(('learning_rate', learn_rate), ))\n",
    "# use accuracy as the metric\n",
    "metric = mx.metric.CrossEntropy()\n",
    "metric2= mx.metric.create('acc')\n",
    "for epoch in range(n_epochs):\n",
    "    train_iter.reset()\n",
    "    metric.reset()\n",
    "    metric2.reset()\n",
    "    for batch in train_iter:\n",
    "        mod_2.forward(batch, is_train=True)       # compute predictions\n",
    "        mod_2.update_metric(metric, batch.label)  # accumulate prediction loss\n",
    "        mod_2.update_metric(metric2, batch.label)  # accumulate prediction accuracy\n",
    "        mod_2.backward()                          # compute gradients\n",
    "        mod_2.update()                            # update parameters\n",
    "    loss=metric.get()\n",
    "    acc=metric2.get()\n",
    "    train_2_loss.append(loss)\n",
    "    train_2_acc.append(acc)\n",
    "    if epoch%display_step==0 or epoch==n_epochs-1:\n",
    "        print('Epoch %d, Training %s' % (epoch+1,acc))\n",
    "    #validation error and accuracy\n",
    "    metric.reset()\n",
    "    metric2.reset()\n",
    "    val_iter.reset()\n",
    "    for batch in val_iter:\n",
    "        mod_2.forward(batch)\n",
    "        mod_2.update_metric(metric,batch.label)\n",
    "        mod_2.update_metric(metric2,batch.label)\n",
    "    loss=metric.get()\n",
    "    acc=metric2.get()\n",
    "    val_2_loss.append(loss)\n",
    "    val_2_acc.append(acc)\n",
    "    if epoch%display_step==0 or epoch==n_epochs-1:\n",
    "        print('Epoch %d, Validation %s' % (epoch+1,acc))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Saving the models and network 2 data:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "mod.save_params('network1')\n",
    "mod_2.save_params('network2')\n",
    "import pickle\n",
    "pickle.dump(train_2_loss, open('train_2_loss','wb'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "***\n",
    "\n",
    "### Comparing the 2 networks on the test data - which has been left untouched so far"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Results on test data: \n",
      "Net 1: Accuracy:  [('accuracy', 0.85919999999999996)]\n",
      "Net 1: Cross-Entropy loss:  [('cross-entropy', 0.39792304687500002)]\n",
      "Net 2: Accuracy:  [('accuracy', 0.84609999999999996)]\n",
      "Net 2: Cross-Entropy loss:  [('cross-entropy', 0.51931982421875)]\n"
     ]
    }
   ],
   "source": [
    "test_img,test_labels=data_loader.load_data(mode='test')\n",
    "metric = mx.metric.CrossEntropy()\n",
    "metric2= mx.metric.create('acc')\n",
    "test_iter = mx.io.NDArrayIter(test_img, test_labels,batch_size=test_img.shape[0])\n",
    "\n",
    "metric.reset()\n",
    "metric2.reset()\n",
    "net1_acc=mod.score(test_iter,metric2)\n",
    "net1_loss=mod.score(test_iter,metric)\n",
    "net2_acc=mod_2.score(test_iter,metric2)\n",
    "net2_loss=mod_2.score(test_iter,metric)\n",
    "\n",
    "print(\"Results on test data: \")\n",
    "print(\"Net 1: Accuracy: \",net1_acc)\n",
    "print(\"Net 1: Cross-Entropy loss: \",net1_loss)\n",
    "print(\"Net 2: Accuracy: \",net2_acc)\n",
    "print(\"Net 2: Cross-Entropy loss: \",net2_loss)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Plotting relevant data for both"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAmsAAAJcCAYAAACmM+PxAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvOIA7rQAAIABJREFUeJzs3Xl8leWd///XJ3uAAAmENSAIoixZxFBsRdDKgBvWqqhYodgq0LqA4jY/nKp06NjqVxyKlVqpWrAygwWkdaYodBA30CCbsgmWHSFAgIQkZLt+f9znJCch+3aS+H4+Htfj3Pv9uU80fHLd12LOOURERESkaQoJdgAiIiIiUjElayIiIiJNmJI1ERERkSZMyZqIiIhIE6ZkTURERKQJU7ImIiIi0oQpWRMR+RYyswvNbIOZZZrZA+XsX21md9fTvXqZmTOzsPq4nsi3jZI1kRbEzO4wszQzyzKzw2b2v2Y2LIjx+P+RzipTbqvm+c7M+jZ0nNVhZhPN7MNgx1GPHgVWO+dinHNz6vPCZrbHzEbW5zVFvs30V45IC2FmDwGPA1OAFUAecDXwA+CcJMPMwpxzBY0UXvuGuFcjP0NLcx6wKNhBiEjVVLMm0gKYWTtgJnCvc26Jc+6Mcy7fOfdX59wjvmOeMrO3zGyhmZ0GJppZpJm9YGaHfOUFM4v0Hd/RzP5mZifN7ISZfWBmIb59j5nZQd8rtB1mdlUt437NzF40s3d811pnZn18+9b4Dtvkr40zsyvM7IDv/t8Ar/qOvcfMdvniXG5m3QLu4czsATP72syOmdmzZhbie/YTZpYYcGwnM8sxs/gaPkc3331P+OK4J2Dfd3y1nafN7IiZPe/bHuX7WRz3fcefmVln3752ZjbfVzt60Mz+3cxCffv6mtn7ZnbK9zz/VUlcN5jZl77rrzaz/r7t/wCuBOb6vtt+FVyij5l96rvX22YWV41rLwB6An/1XfvRgOv9yMz2+eKeUZPvWORbzTmnoqLSzAteDVoBEFbJMU8B+cCNeH+oReMleGuBTkA88DHwS9/x/wHMA8J95XLAgAuB/UA333G9gD4V3LMX4CqKC3gNOAF8B6+m/w1gUcB+B/QNWL/C95y/BiJ9z/B94Bgw2Lftt8CaMtf4PyAOL4nYCdzt2/c74NcBx04F/lpBrBOBDyvY977vWlFACpAOXOXb9wkw3rfcBrjUtzwZ+CvQCggFLgHa+vYtA34PtPb9bD4FJvv2vQnM8P0Mo4BhFcTUDzgD/Ivv5/cosAuI8O1f7f8eKjh/NXAQGOSL4y/Awmpeew8wspz/Dv7g+5klA2eB/sH+f0dFpTkU1ayJtAwdgGOu6leCnzjnljnnipxzOcCPgJnOuaPOuXTgaWC879h8oCtwnvNq6T5wzjmgEC8pGmBm4c65Pc653VXc95ivBsZf+gfsW+Kc+9QX+xt4yU5lioAnnXNnA57hj865z51zZ4F/Bb5rZr0Czvm1c+6Ec24f8AIwzrf9deAOf42h79kXVHH/UsysBzAMeMw5l+uc2wi8Qunvsa+ZdXTOZTnn1gZs74CXjBY659Y75077ateuAaY5r4b0KDAbuD3gvPPwkuVc51xF7ehuA95xzr3nnMsHnsNLlL5Xg8db4Jz7wjl3Bvg34FZfDV9tr/20cy7HObcJ2ISXtIlIFZSsibQMx4GOVnVvu/1l1rsBewPW9/q2ATyLV1vyru8V4uMAzrldwDS8mrqjZrbI/9rRSnci6Blw3Y7OufYBZVvAvm8ClrPxap8qk+6cy63oGZxzWXjfR/cKnrv4GZ1z6/BqiEaY2UVAX2B5FfcvqxtwwjmXWeYe/vv/FK8marvvVef1vu0L8NoWLvK9gv6NmYXjJWLhwGF/cotXy9bJd96jeDWcn/peQ/6kkrgCv5ci3/fQvYLjy1P2ewsHOtbh2jX9WYsIStZEWopPgFy8V5yVcWXWD+ElB349fdtwzmU656Y7584HxgAP+dumOef+7Jwb5jvX4b2WxDnXJqDsq+tD1eYZzKw1Xo3VwYBjegQsFz+jz+vAnXg1YW+VSQSr4xAQZ2YxZe5xEMA595VzbhxesvVr4C0za+2rrXzaOTcAr0bqemACXtJzltIJblvn3EDf9b5xzt3jnOuG9yr1d1Z+j9my34v5voeD5RxbkbLfWz7eK+eqrl32ZyQidaBkTaQFcM6dAn4BvGhmN5pZKzMLN7NrzOw3lZz6JvCEmcWbWUffNRYCmNn1vsbsBpzGe/1ZaN74XN83ryNCLpDj29cQjgDnV3HMn4G7zCzFF9OvgHXOuT0BxzxiZrG+V5ZTgcBG+QuAH+IlbH+q4l7m6xhQXJxz+/Ha+v2Hb1sSXm3aG74T7jSzeF/t00nfdQrN7EozS/S9VjyNlwgVOucOA+8C/8/M2prXGaKPmY3wXW+smSX4rpOBlxiV9/3/N3CdmV3lq7GbjpcEflzFMwa608wGmFkrvPaNbznnCqtx7er83ESkmpSsibQQzrnngYeAJ/AauO8H7sNrrF6RfwfSgM3AFuBz3zaAC4CVQBZezd3vnHOr8dqrPYNXw/INXo3R/1dFeCfLvCJ9qJqP9RTwuu914K3lHeCcW4XXnuovwGGgDyXtu/zeBtYDG4F3gPkB5x/Ae24HfFBFPN/DS06Li+/V8zi8RvSHgKV4bere851zNfClmWUB/wnc7qu96wK8hZeobcPrpLDQd84EIALYipeQvYXXfhBgCLDOd73lwFTn3D/L+V524CWgv8X7WY0Bxjjn8qp4xkAL8DqBfIPXmeGBal77P/D+CDhpZg/X4H4iUg7z2guLiLRMZuaAC3xt7So65o/AIefcE40XmYhI9WhQXBH5VvP1Gr0JuDi4kYiIlE+vQUXkW8vMfgl8ATxb3qtEEZGmQK9BRURERJow1ayJiIiINGEtps1ax44dXa9evYIdhoiIiEiV1q9ff8w5V615iIOSrPl6Xl0PHHXODargmCvwpoUJx5tGZ0Rl1+zVqxdpaWn1HaqIiIhIvTOzvVUf5QnWa9DX8MYeKpeZtcebFPkG36jdYxspLhEREZEmJSjJmnNuDXCikkPuwJvceZ/v+KONEpiIiIhIE9NUOxj0A2LNbLWZrTezCeUdZGaTzCzNzNLS09MbOUQRERGRhtdUOxiEAZcAVwHRwCdmttY5tzPwIOfcy8DLAKmpqRqDREREmpX8/HwOHDhAbm5usEORBhIVFUVCQgLh4eG1vkZTTdYO4HUqOAOcMbM1QDKws/LTREREmo8DBw4QExNDr169MLNghyP1zDnH8ePHOXDgAL179671dZrqa9C3gcvNLMzMWgFD8SY6FhERaTFyc3Pp0KGDErUWyszo0KFDnWtOgzV0x5vAFUBHMzsAPIk3RAfOuXnOuW1m9ndgM1AEvOKc+yIYsYqIiDQkJWotW338fIOSrDnnxlXjmGeBZxshHBEREZEmq6m+Bm1y9p7cy0MrHuKZD58JdigiIiLyLaJkrZoycjOYvXY2f97y52CHIiIiUm/MjOnTpxevP/fcczz11FOVnrN69Wo+/vjjSo9Zs2YNgwcPJiwsjLfeeqvC47Zs2UJKSgopKSnExcXRu3dvUlJSGDlyZI2eY/To0WRmZtboHIA777yTZcuW1fi8xqRkrZriouMAOJFT2Vi+IiIizUtkZCRLlizh2LFj1T6nOslaz549ee2117jjjjsqPS4xMZGNGzeyceNGbrjhBp599lk2btzIypUrSx1XUFBQ6XVWrFhBTExM9R6gmVGyVk2xUbGAV8MmIiJS78waplQhLCyMSZMmMXv27HP2paenc/PNNzNkyBCGDBnCRx99xJ49e5g3bx6zZ88mJSWFDz74oNzr9urVi6SkJEJCap9qrFy5kpEjR3L77bdz8cUXAzBmzBguueQSBg4cyCuvvFJ8bEJCAidPnmTXrl0MGjSIn/70pwwcOJBrrrmmxr0xi4qKeOihhxg0aBCJiYnFNYMHDx5k2LBhpKSkMGjQID7++GMKCgoYP348iYmJDBo0iDlz5tT6eSvSVMdZa3LaRLQh1ELJzs8mrzCPiNCIYIckIiJSL+69916SkpJ49NFHS22fOnUqDz74IMOGDWPfvn2MHj2abdu2MWXKFNq0acPDDz/c4LGtXbuWrVu30rNnTwBef/114uLiyM7OJjU1lZtvvpnY2NhS5+zYsYM333yTxMREbrrpJpYtW8btt99e7XsuXryYrVu3smnTJtLT0xkyZAjDhw9n4cKFjBkzhscee4zCwkJycnJYv349x44dY8uWLQCcPHmy/h7eR8laNZkZsdGxHMs+RkZOBp3bdA52SCIi0pK44E3E07ZtWyZMmMCcOXOIjo4u3r5y5Uq2bt1avH769OlatQuri+9+97vFiRrA7NmzWb58OeANKrx7925SU1NLndO3b18SExMBuOSSS9izZ0+N7vnhhx9yxx13EBoaSpcuXRg2bBhpaWkMGTKEyZMnk5uby4033khycjJ9+/Zlx44dTJ06lWuvvZZRo0bV7YHLodegNaBXoSIi0lJNmzaN+fPnc+bMmeJtRUVFfPLJJ8Vtyg4ePNjo7cJat25dvLxy5UrWrFnD2rVr2bRpE0lJSeW+4oyMjCxeDg0NrbK9W1mugsT5+9//PqtXr6Zr16786Ec/4o033qBDhw5s3ryZYcOGMWfOHCZPnlyje1WHkrUaUCcDERFpqeLi4rj11luZP39+8bZRo0Yxd+7c4vWNGzcCEBMT0+g1bACnTp0iLi6O6OhovvzySz777LMGuc/w4cNZtGgRhYWFHDlyhI8++ojU1FT27t1Lly5dmDRpEhMnTmTDhg2kp6fjnGPs2LE8/fTTfP755/Uej5K1GoiN9tWs5ahmTUREWp7p06eX6hU6Z84c0tLSSEpKYsCAAcybNw/wGvkvXbq00g4Gn332GQkJCSxevJjJkyczcODAOsd33XXXkZ2dTXJyMjNnzmTo0KF1vibA3XffTUJCAgkJCVx++eXccsstXHTRRSQnJzNy5Eief/55OnXqxKpVq0hOTubiiy/m7bff5v7772f//v0MHz6clJQU7rnnHn71q1/VS0yBrKKqvuYmNTXVpaWlNeg97vjLHbz5xZss+OEC7ky6s0HvJSIiLd+2bdvo379/sMOQBlbez9nM1jvnUis4pRTVrNVAcZs11ayJiIhII1Fv0BpQmzUREZHSZs2axeLFi0ttGzt2LDNmzDjn2C1btjB+/PhS2yIjI1m3bl2DxggwZcoU1q5dW2rbQw89xIQJExr83nWlZK0GitusqTeoiIgIADNmzCg3MSuPf7aCYPC3t2uO9Bq0BjR0h4iIiDQ2JWs1oN6gIiIi0tiUrNWAatZERESksSlZqwF1MBAREZHGpmStBvQaVEREWhozY/r06cXrzz33HE899VSl56xevZqPP/640mPWrFnD4MGDCQsL46233qr02N69e7Njx45S26ZNm8ZvfvObCs/Zs2cPgwYNqjTG66+/vtL7NhdK1mpAr0FFRKSliYyMZMmSJaVmLqhKdZK1nj178tprr3HHHXdUeb3bb7+dRYsWFa8XFRXx1ltvcdttt1U7ppZMQ3fUQKvwVoSHhJNbkEtOfg7R4dHBDklERFqI1autQa57xRWVz1QUFhbGpEmTmD17NrNmzSq1Lz09nSlTprBv3z4AXnjhBbp37868efMIDQ1l4cKF/Pa3v+Xyyy8/57q9evUCICSk6nqhcePGcdttt/Hkk08CXq1cr169OO+889izZw/jx48vnmB+7ty5fO9736vymhVZtWoVDz/8MAUFBQwZMoSXXnqJyMhIHn/8cZYvX05YWBijRo3iueeeY/HixTz99NOEhobSrl071qxZU+v71oWStRowM2KjYzl65igZuRlK1kREpEW49957SUpK4tFHHy21ferUqTz44IMMGzaMffv2MXr0aLZt28aUKVNo06YNDz/8cL3cPykpiZCQEDZt2kRycjKLFi1i3LhxAHTq1In33nuPqKgovvrqK8aNG0dtp5fMzc1l4sSJrFq1in79+jFhwgReeuklJkyYwNKlS9m+fTtmxsmTJwGYOXMmK1asoHv37sXbgkHJWg3FRcd5yVpOBt1iugU7HBERaSGqqgFrSG3btmXChAnMmTOH6OiSioiVK1eydevW4vXTp0+TmZnZIDGMGzeORYsWMXDgQN5++21mzpwJQH5+Pvfddx8bN24kNDSUnTt31voeO3bsoHfv3vTr1w+AH//4x7z44ovcd999REVFcffdd3PdddcVt3W77LLLmDhxIrfeeis33XRT3R+ylpSs1ZDarYmISEs0bdo0Bg8ezF133VW8raioiE8++aRUAtdQxo0bx6hRoxgxYgRJSUl06tQJgNmzZ9O5c2c2bdpEUVERUVFRtb6Hc+UnxGFhYXz66aesWrWKRYsWMXfuXP7xj38wb9481q1bxzvvvENKSgobN26kQ4cOtb5/bamDQQ2pR6iIiLREcXFx3HrrrcyfP79426hRo5g7d27xun+qqJiYmHqvYevTpw8dOnTg8ccfL34FCnDq1Cm6du1KSEgICxYsoLCwsNb3uOiii9izZw+7du0CYMGCBYwYMYKsrCxOnTrFtddeywsvvFD8nLt372bo0KHMnDmTjh07sn///ro9ZC0FJVkzsz+a2VEz+6KK44aYWaGZ3dJYsVVFNWsiItJSTZ8+vVSv0Dlz5pCWlkZSUhIDBgwonl9zzJgxLF26lJSUFD744INyr/XZZ5+RkJDA4sWLmTx5MgMHDqzy/uPGjWP79u388Ic/LN7285//nNdff51LL72UnTt30rp162o/z6pVq0hISCguGzZs4NVXX2Xs2LEkJiYSEhLClClTyMzM5PrrrycpKYkRI0Ywe/ZsAB555BESExMZNGgQw4cPJzk5udr3rk9WUZVgg97UbDiQBfzJOVfuIClmFgq8B+QCf3TOVTpIS2pqqqttg8OaeOB/H+C3n/6W2aNnM+3SaQ1+PxERabm2bdtG//79gx2GNLDyfs5mtt45l1qd84NSs+acWwNUNQ3A/cBfgKMNH1H1Fdes6TWoiIiINIIm2cHAzLoDPwS+Dwyp5LhJwCTwBt9rDMVt1vQaVEREhFmzZrF48eJS28aOHcuMGTPOOXbLli2MHz++1LbIyEjWrVtX6/uvWLGCxx57rNS23r17s3Tp0lpfs6lpkska8ALwmHOu0KziQQKdcy8DL4P3GrQxAlObNRERkRIzZswoNzErT2JiYnHj/foyevRoRo8eXa/XbGqaarKWCizyJWodgWvNrMA5tyy4YZXUrGkydxEREWkMTTJZc8719i+b2WvA35pCogbeoLigNmsiIiLSOIKSrJnZm8AVQEczOwA8CYQDOOfmBSOm6tJrUBEREWlMQUnWnHPjqj6q+NiJDRhKjWlQXBEREWlMmsGghgJr1oIxRp2IiEh9MjOmT59evP7cc8/x1FNPVXrO6tWr+fjjjys95vnnn2fAgAEkJSVx1VVXsXfv3nKP27JlCykpKaSkpBAXF0fv3r1JSUlh5MiRNXqO0aNH12pWhTvvvJNly5pES6sKKVmroejwaKLCosgrzCM7PzvY4YiIiNRJZGQkS5YsKTVzQVWqk6xdfPHFpKWlsXnzZm655RYeffTRco/z9xDduHEjN9xwA88++ywbN25k5cqVpY4rKCio9H4rVqwgJiam2s/QnDTJDgZNXWxULIezDpORm0HriOpPeyEiIlIRe7rioarqwj1Z+VugsLAwJk2axOzZs5k1a1apfenp6UyZMoV9+/YB8MILL9C9e3fmzZtHaGgoCxcu5Le//S2XX375Ode98sori5cvvfRSFi5cWOPYV65cyTPPPEPHjh358ssv2bJlC2PGjOHQoUPk5uby4IMPcvfddwOQkJDAF198wbFjx7jxxhsZOnQoa9eupWfPnixdurRGE8AXFRXx8MMP8+6772JmPPnkk9xyyy0cPHiQ2267jaysLAoKCnj55Zf5zne+w1133cXGjRtxzjFp0iQeeOCBGj9rZZSs1UJstC9Zy8kgoW1CsMMRERGpk3vvvZekpKRzar+mTp3Kgw8+yLBhw9i3bx+jR49m27ZtTJkyhTZt2vDwww9X6/rz58/nmmuuqVVsa9euZevWrcWD37/++uvExcWRnZ1NamoqN998M7GxsaXO2bFjB2+++SaJiYncdNNNLFu2jNtvv73a91y8eDFbt25l06ZNpKenM2TIEIYPH87ChQsZM2YMjz32GIWFheTk5LB+/XqOHTvGli1bADh58mStnrMyStZqQT1CRUSkvlVVA9aQ2rZty4QJE5gzZw7R0dHF21euXMnWrVuL10+fPl3jdmELFy4kLS2N999/v1axffe73y01S9Hs2bNZvnw5AAcOHGD37t2kppaeYrNv374kJiYCcMkll7Bnz54a3fPDDz/kjjvuIDQ0lC5dujBs2DDS0tIYMmQIkydPJjc3lxtvvJHk5GT69u3Ljh07mDp1Ktdeey2jRo2q1XNWRm3WakED44qISEszbdo05s+fz5kzZ4q3FRUV8cknnxS3KTt48GCN2oWtXLmSWbNmsXz5ciIjI2sVV+vWJc2NVq5cyZo1a1i7di2bNm0iKSmJ3Nzcc84JvFdoaGiV7d3KqqgD4fe//31Wr15N165d+dGPfsQbb7xBhw4d2Lx5M8OGDWPOnDlMnjy5RveqDiVrtaCBcUVEpKWJi4vj1ltvZf78+cXbRo0axdy5c4vX/VNFxcTEVFnDtmHDBiZPnszy5cvp1KlTvcR46tQp4uLiiI6O5ssvv+Szzz6rl+uWNXz4cBYtWkRhYSFHjhzho48+IjU1lb1799KlSxcmTZrExIkT2bBhA+np6TjnGDt2LE8//TSff/55vcejZK0W9BpURERaounTp5fqFTpnzhzS0tJISkpiwIABzJvnjVs/ZswYli5dSkpKCh988EG513rkkUfIyspi7NixpKSkcMMNN9Q5vuuuu47s7GySk5OZOXMmQ4cOrfM1Ae6++24SEhJISEjg8ssv55ZbbuGiiy4iOTmZkSNH8vzzz9OpUydWrVpFcnIyF198MW+//Tb3338/+/fvZ/jw4aSkpHDPPffwq1/9ql5iCmQtZayw1NRUl5aW1ij3enr10zz1/lM8cfkT/PL7v2yUe4qISMuzbds2+vfvH+wwpIGV93M2s/XOudQKTilFNWu1UDyLgWrWREREpIGpN2gt+F+DqoOBiIh8282aNYvFixeX2jZ27FhmzJhxzrFbtmxh/PjxpbZFRkaybt26Bo0RYMqUKaxdu7bUtoceeogJEyY0+L3rSslaLRR3MFDNmoiIfMvNmDGj3MSsPP7ZCoLB396uOdJr0FrQZO4iIiLSWFSzVhOnT8Pp0+oNKiIiIo1GNWvV9eGH0K4d3HabBsUVERGRRqNkrbq6d/c+9+0rqVnLyahwlGMRERGR+qBkrbq6dwczOHSISEJpFd6KQldIVl5WsCMTERGpNTNj+vTpxevPPfccTz31VKXnrF69mo8//rjSY55//nkGDBhAUlISV111FXv37q3w2N69e7Njx45S26ZNm8ZvfvObCs/Zs2cPgwYNqjTG66+/vtIYmwsla9UVEQFdukBRERw6pHZrIiJSr8waplQlMjKSJUuWlJq5oCrVSdYuvvhi0tLS2Lx5M7fccguPPvpohcfefvvtLFq0qHi9qKiIt956i9tuu63aMbVkStZqomdP73P/fvUIFRGRFiEsLIxJkyYxe/bsc/alp6dz8803M2TIEIYMGcJHH33Enj17mDdvHrNnz650uqkrr7ySVq1aAXDppZdy4MCBCmMYN25cqWRtzZo19OrVi/POO489e/Zw+eWXM3jwYAYPHlxlkliVVatWcfHFF5OYmMhPfvITzp49C8Djjz9eXBP48MMPA7B48WIGDRpEcnIyw4cPr9N960K9QWuiZ09Yt65UuzV1MhARkfoQzCbQ9957L0lJSefUfk2dOpUHH3yQYcOGsW/fPkaPHs22bduYMmUKbdq0KU5qqjJ//nyuueaaCvcnJSUREhLCpk2bSE5OZtGiRYwbNw6ATp068d577xEVFcVXX33FuHHjqO30krm5uUycOJFVq1bRr18/JkyYwEsvvcSECRNYunQp27dvx8w4efIkADNnzmTFihV07969eFswqGatJnr08D7379fAuCIi0mK0bduWCRMmMGfOnFLbV65cyX333Vc8Efvp06fJzMys0bUXLlxIWloajzzySKXH+WvXCgoKePvttxk7diwA+fn53HPPPSQmJjJ27Fi2bt1as4cLsGPHDnr37k2/fv0A+PGPf8yaNWto27YtUVFR3H333SxZsqS4RvCyyy5j4sSJ/OEPf6CwsLDW960r1azVhP816L59xF6o16AiItJyTJs2jcGDB3PXXXcVbysqKuKTTz4hOjq6VtdcuXIls2bN4v333ycyMrLSY8eNG8eoUaMYMWIESUlJdOrUCYDZs2fTuXNnNm3aRFFREVFRUbWKBahwBIewsDA+/fRTVq1axaJFi5g7dy7/+Mc/mDdvHuvWreOdd94hJSWFjRs30qFDh1rfv7ZUs1YT/pq1wOE7VLMmIiItQFxcHLfeeivz588v3jZq1Cjmzp1bvO6fKiomJqbKGrYNGzYwefJkli9fXpx4VaZPnz506NCBxx9/vPgVKMCpU6fo2rUrISEhLFiwoE41XBdddBF79uxh165dACxYsIARI0aQlZXFqVOnuPbaa3nhhReKn3P37t0MHTqUmTNn0rFjR/bv31/re9eFkrWaCOxgoDZrIiLSwkyfPr1Ur9A5c+aQlpZGUlISAwYMKJ5fc8yYMSxdurTSDgaPPPIIWVlZjB07tvg1alXGjRvH9u3b+eEPf1i87ec//zmvv/46l156KTt37qR169bVfp5Vq1aRkJBQXDZs2MCrr77K2LFjSUxMJCQkhClTppCZmcn1119PUlISI0aMKO5s8cgjj5CYmMigQYMYPnw4ycnJ1b53fbJgDOpqZn8ErgeOOufOGSTFzH4EPOZbzQJ+5pzbVNk1U1NTXW0bHFbbkSPe8B1xcbz4vzO573/vY8olU3jp+pca9r4zMlLLAAAgAElEQVQiItIibdu2jf79+wc7DGlg5f2czWy9cy61OucHq2btNeDqSvb/ExjhnEsCfgm83BhBVSk+HiIj4cQJYkO8xod6DSoiIiINKSgdDJxza8ysVyX7AwdRWQskNHRM1RISAgkJsHs3sZn5gJI1ERH5dps1axaLFy8utW3s2LHMmDHjnGO3bNnC+PHjS22LjIxk3bp1tb7/ihUreOyxx0pt6927N0uXLq31NZua5tAb9KfA/5a3w8wmAZMAevrbkzW0nj29ZC0jF1BvUBERqRvnHFadqQaaqBkzZpSbmJUnMTGxuPF+fRk9ejSjR4+u12vWp/pobtakOxiY2ZV4ydpj5e13zr3snEt1zqXGx8c3TlC+HqGxR71eMOpgICIitRUVFcXx48fr5R90aXqccxw/frxOw41AE65ZM7Mk4BXgGufc8WDHU8xXgxf3zUkwvQYVEZHaS0hI4MCBA6Snpwc7FGkgUVFRJCTUrTVXk0zWzKwnsAQY75zbGex4SvHVrLXflw7nwcnckxS5IkKsSVdSiohIExQeHk7v3r2DHYY0cUFJ1szsTeAKoKOZHQCeBMIBnHPzgF8AHYDf+d7jF1S3e2uD89Wshe8/SJsL2pCVl0Xm2UzaRbULcmAiIiLSEgWrN+i4KvbfDdzdSOHUTOCUU1GxZOVlkZGboWRNREREGoTe3dVUwJRT/snc1clAREREGoqStZqKiYH27SE3l9iwNoCG7xAREZGGo2StNvzDdxRGAOoRKiIiIg1HyVpt+NqtxeZ5X59q1kRERKShKFmrDX/NWrY3iKHarImIiEhDUbJWG/6BcU9rflARERFpWErWasNfs3Y8G9BrUBEREWk4StZqw99mLd2bH1Q1ayIiItJQlKzVhr9m7aDXVk3JmoiIiDQUJWu10b07mBF70JtfXh0MREREpKEoWauNiAjo0oU4X29QtVkTERGRhqJkrbZ69iQ2x1vUa1ARERFpKErWaqtnT9rneounck9RWFQY3HhERESkRVKyVls9ehDqoC2ROBynzp4KdkQiIiLSAilZqy3/wLj++UHVbk1EREQagJK12vIP33HWALVbExERkYahZK22/APjnikCVLMmIiIiDUPJWm35a9ZOnQVUsyYiIiINQ8labcXHQ2Qksae8ydw1MK6IiIg0BCVrtRUSAgkJxPnHWtNrUBEREWkAStbqomdPYn1jrek1qIiIiDQEJWt10aNHySwGqlkTERGRBqBkrS4CatZO5KrNmoiIiNQ/JWt10aOH2qyJiIhIg1KyVheazF1EREQaWFCSNTP7o5kdNbMvKthvZjbHzHaZ2WYzG9zYMVZLYAcD1ayJiIhIAwhWzdprwNWV7L8GuMBXJgEvNUJMNRfYwUA1ayIiItIAgpKsOefWAJW1yP8B8CfnWQu0N7OujRNdDcTE0C6qHebg9NnTFBQVBDsiERERaWGaapu17sD+gPUDvm2lmNkkM0szs7T09PRGCy5QSI+etPe9Cj2ZezIoMYiIiEjL1VSTNStnmztng3MvO+dSnXOp8fHxjRBWOQI7GajdmoiIiNSzppqsHQB6BKwnAIeCFEvlevTQLAYiIiLSYJpqsrYcmODrFXopcMo5dzjYQZUroGZNk7mLiIhIfQsLxk3N7E3gCqCjmR0AngTCAZxz84D/Aa4FdgHZwF3BiLNaevQgdoO3qNegIiIiUt+Ckqw558ZVsd8B9zZSOHXTs2fJLAZ6DSoiIiL1rE6vQc1sqpm19b2unG9mn5vZqPoKrlnQZO4iIiLSgOraZu0nzrnTwCggHu915TN1jqo56d69pINB9vHgxiIiIiItTl2TNf8QG9cCrzrnNlH+sBstV0QEsRHtADhx4kCQgxEREZGWpq7J2nozexcvWVthZjFAUd3Dal7iYjoBkHHymyBHIiIiIi1NXTsY/BRIAb52zmWbWQeacs/NBhIb1w34ioysY8EORURERFqYutasOWAA8IBvvTUQVcdrNjuxHb3xe9UbVEREROpbXZO13wHfBfxDcWQCL9bxms1ObNfeAJwoyAxyJCIiItLS1DVZG+qcuxfIBXDOZQARdY6qmYntcQEAGeQGORIRERFpaeqarOWbWSi+SdbNLJ5vYQeDtr0uIrQIzoQWkl+YH+xwREREpAWpa7I2B1gKdDKzWcCHwK/qHFUzYz170l6TuYuIiEgDqFNvUOfcG2a2HrgKb3y1G51z2+olsuYkPp7YXON4K0fGiYN0at0p2BGJiIhIC1HX6ab6AP90zr0IfAH8i5m1r5fImpOQEGKLvKZ6J/ZuD3IwIiIi0pLU9TXoX4BCM+sLvAL0Bv5c56iaobiQ1gBkHNwd5EhERESkJalrslbknCsAbgL+0zn3INC17mE1P7ERbQHIOLo3yJGIiIhIS1IfvUHHAROAv/m2hdfxms1SbKsOAGQc0/ygIiIiUn/qmqzdhTco7izn3D/NrDewsO5hNT+x/vlBTx0JciQiIiLSktS1N+hWfFNNmVksEOOce6Y+AmtuYjt0gzNwIlvzg4qIiEj9qWtv0NVm1tbM4oBNwKtm9nz9hNa8xMWfB0BG7skgRyIiIiItSV1fg7Zzzp3G62DwqnPuEmBk3cNqfmK7nQ9ARtEZcC7I0YiIiEhLUddkLczMugK3UtLB4FspNq47ABkRRXBMr0JFRESkftQ1WZsJrAB2O+c+M7Pzga/qHlbzExsVC8CJaGD//uAGIyIiIi1GnZI159xi51ySc+5nvvWvnXM3109ozUtcdBwAGVHAvn3BDUZERERajLp2MEgws6VmdtTMjpjZX8wsob6Ca2qcKyQ//0S5+2KjvZq1DNWsiYiISD2q62vQV4HlQDegO/BX37YW5+TJD1i7tg87d/6s3P2tw1sTRgg54XB239eNHJ2IiIi0VHVN1uKdc6865wp85TUgvqqTzOxqM9thZrvM7PFy9vc0s/8zsw1mttnMrq1jnHUWFdWbs2f3c+zYEvLyzh341syI9c8PekjJmoiIiNSPuiZrx8zsTjML9ZU7geOVnWBmocCLwDXAAGCcmQ0oc9gTwH875y4Gbgd+V8c46ywqKoEOHa7HuQIOHy6/8jA2sj0AJzQ/qIiIiNSTuiZrP8EbtuMb4DBwC94UVJX5DrDL1xkhD1gE/KDMMQ5o61tuBxyqY5z1olu3yQAcPvwHnCs6Z39c644AZJw42KhxiYiISMtV196g+5xzNzjn4p1znZxzN+INkFuZ7kBgC/wDvm2BngLuNLMDwP8A95d3ITObZGZpZpaWnp5eu4eogbi40URG9iQ392syMlaesz+2rW9+0DPHIT+/weMRERGRlq+uNWvleaiK/VbOtrJD/o8DXnPOJQDXAgvM7JxYnXMvO+dSnXOp8fFVNpWrM7NQuna9B4BDh+adsz+2dQcAMiIdHGoSlYEiIiLSzDVEslZeMhboANAjYD2Bc19z/hT4bwDn3CdAFNCxvgKsi65dfwqEcuzYcs6eLR22BsYVERGR+tYQyVpVE2N+BlxgZr3NLAKvA8HyMsfsA64CMLP+eMlaw7/nrIbIyK507PgDoJDDh/9Yap8/WcuIRgPjioiISL2oVbJmZplmdrqckok35lqFnHMFwH1401Rtw+v1+aWZzTSzG3yHTQfuMbNNwJvAROeazuzoJR0NXsa5wuLtpWYxUM2aiIiI1IOw2pzknIupy02dc/+D13EgcNsvApa3ApfV5R4NKTZ2JFFR55Ob+zUnTvydDh2u87ZHq2ZNRERE6ldDvAZt8cxC6NZtEgCHDv2+eHvxa1DVrImIiEg9UbJWS1263IVZOMePv0NurpeY+WvWTqhmTUREROqJkrVaiojoRMeOPwSKOHz4FSCgzZqSNREREaknStbqoFu3KQAcPvwKRUUFpXuDZmRAVlYQoxMREZGWQMlaHbRvfwXR0f3IyzvEiRPvBHQwMG/8ko8+Cmp8IiIi0vwpWasDMwvoaDCP6LBoIkIjOBvqyAkHHn4YCgqCG6SIiIg0a0rW6qhz5x9jFsGJEyvIzd1b8iq0Xw/44guYd+60VCIiIiLVpWStjiIiOhIfPxZwHD78h5JOBjOmewf84hdw7FjwAhQREZFmTclaDWRmeqWskhkN5hMb1R6AjO9eDP/yL15Hg3/7t8YMU0RERFoQJWvVtH49pKTA1Knn7mvXbhitWvUnP/8IrULPApBx9iS88AKEhsLvfw8bNzZyxCIiItISKFmrplat4NAhePVV+MtfSu/zOhp4w3hEFh4C4ETOCRgwAO6/H5yDBx7wPkVERERqQMlaNfXvD8895y3fcw8cPFh6f+fO4wkJiSLSfQNARk6Gt+PJJyE+Hj74AP7rvxoxYhEREWkJlKzVwM9/Dtdc4zVDmzgRiopK9oWHxxIffxsxYd56Rq4vWWvfHn71K2/5kUfgzJlGjVlERESaNyVrNWAGf/wjdOwIK1fCf/5n6f3duk2mjS9ZO5ET0AP0rrvgkkvgwAF45pnGC1hERESaPSVrNdSlC8yf7y0//jhs3lyyr23bS+nYugcAhzMCdoSGwpw53vKzz8LXXzdStCIiItLcKVmrhRtugEmTIC8PfvQjyM31tpsZvTpfB8CR09tKn/S978Gdd8LZs97MBiIiIiLVoGStlp5/Hi64wJuk4F//tWR77643ApCRc4IjRxaVPunXv4bWrWHpUnjvvUaMVkRERJorJWu11Lo1vPEGhIV5w6n5c6/4Nt5r0MwC2L59PMeOLS85qVs3eOIJb3nqVMjPb+SoRUREpLlRslYHQ4bAU095yxMnwvHjFM8Nmu1a4VwBX345lhMnVpac9OCD0KcPbNsGv/tdo8csIiIizYuStTp6/HG47DJvwNxJk6C9L1k7lZdHt2734lweX3zxA06e/NA7ITISZs/2lp98EtLTgxS5iIiINAdK1uooNBQWLICYGFiyBBYtjCIqLIqCogK6nvcrunS5i6KibLZsuZbTp9O8k66/Hq6+Gk6dghkzgvsAIiIi0qQpWasHvXvDiy96yw88AG3PXAxARu5JLrzwD8TH30ZhYSabN48mK2uLN2Db7Nleg7dXXoGPPw5i9CIiItKUKVmrJ3feCbfeCllZkP1fL0NhKM99/BxmofTvv4AOHcZQUHCCTZv+hezsnXDRRTBtmjdf6OjR8Pe/B/sRREREpAlSslZPzGDePEhIgKyvBxHy4S/47ae/5Y3NbxASEs6AAf9NbOxI8vOPsGnTVeTk7IF//3cYN87L8K6/3qtlExEREQmgZK0excbCn/7kJW5u9b/Bqn/n7mU/Y8uRLYSGRjFo0DLatr2Ms2cPsGnTVZzlOCxc6A3UVljozRD/xBNebZuIiIgIQUrWzOxqM9thZrvM7PEKjrnVzLaa2Zdm9ufGjrG2rrzSm1HKDPhgBrm/X8X1cx/iVO4pQkNbk5T0DjExqeTmfs2mTSPJKzjuTfT++997vRVmzYLx472ZDkRERORbr9GTNTMLBV4ErgEGAOPMbECZYy4A/hW4zDk3EJjW2HHWxfTpsHq10fO8Ijg0hH2/XsaI+9+gsKiIsLB2JCX9ndatB5GdvY3Nm0eRl3fMG/fjr38tGW336qshIyPYjyIiIiJBFoyate8Au5xzXzvn8oBFwA/KHHMP8KJzLgPAOXe0kWOss8svh82bQrjx1kzIb82mV35O4ohdpKdDeHgHkpLeIzq6H1lZG/n004s4dOgV3NWjYc0a6NoVVq/2BnDbuzfYjyIiIiJBFIxkrTuwP2D9gG9boH5APzP7yMzWmtnV5V3IzCaZWZqZpaU3wcFl27WDpf8VwyPPfw6RJ9n2YT8uGniWv/8dIiO7kJy8ivbtr6Sg4Dg7d97Dhg2XkXmBwdq1MHCgN8vBpZfC+vXBfhQREREJkmAka1bOtrIt6sOAC4ArgHHAK2bW/pyTnHvZOZfqnEuNj4+v90Dry28eHMzU1/4I563mRHok11wD998PziWQnLyK/v3/TEREV06fXsv69al8dfZZ8le/4zWA++YbGD4c3nkn2I8hIiIiQRCMZO0A0CNgPQE4VM4xbzvn8p1z/wR24CVvzdb/u3Uqo3/5LIx8DAspYO5cSE2FTZuMzp3H8Z3vbCchYRpgHDw4l093fIdv/jQeN/5OyM6GG27wei5o8ncREZFvlWAka58BF5hZbzOLAG4Hlpc5ZhlwJYCZdcR7Lfp1o0ZZz0JDQvnzLQvodf1/4+4eSvvu37B1KwwdCnfdBR9/3JY+fWaTmvo5bdteRn7+Ubbv+gkbH9xH1jOToagIHn3Uez26ZImG9xAREfmWaPRkzTlXANwHrAC2Af/tnPvSzGaa2Q2+w1YAx81sK/B/wCPOueONHWt9i4uO4y+3/oXIHl9y8sfnc+XYbeTlwWuvwYgRcMEF8MILSXTosIYLL3yV8PB4Tp1aQ9rQV9i14kYKEvvCV1/BzTfDsGHwySfBfiQRERFpYOZaSA1NamqqS0tLC3YY1fLHDX/kp8t/SmRoJIuuXM9n/zOQ11+Hgwe9/WZw1VUwfnwWKSlPcOLEHMARGtqWzulJdP31F8R8etI7+Oab4ZlnoG/foD2PiIiI1IyZrXfOpVbrWCVrwTHpr5P4w+d/oFf7Xnx2z2fERnZk5Up49VVYtqxkTNx27eDmm9MZMeIJevR42RtsF2hzujNd/3SCzn/PJ+xsGPzsZ/CLX0DHjsF7KBEREakWJWvNQG5BLpe/ejlph9Lo1b4Xi25exNCEoYA3Fu6iRV7i9tlnJeckJOQxePAmLrpoIYmJ/0P37rsILQwlflUhXf8G7fbFYP/6/8HUqRAdHaQnExERkaooWWsmDp4+yA8W/YD1h9cTFhLGf1z1Hzz03YcIsZKmhF984bVpW7gQjhwpfX7HjsdITFxJcvL7JCe/T3+20e1/oPOWzkTcNgV++lPo0QMRERFpWpSsNSN5hXk8vvJxZq+dDcA1fa/h9RtfJ7516XHjioq8xO39972yZg2UHQe4ffujJCWtIXHQBwyOTGPokc30iE4iZsx07LobICyssR5LREREKqFkrRn6646/MvHtiZzIOUG3mG78+aY/M6LXiAqPdw62by9J3t5/33H4cOnxhs2K6NlzOxf2Wk9K6y1c2i+EYRNuIf7iav23ISIiIg1EyVoztf/Ufu5Ycgcf7vuQEAvhF8N/wRPDnyA0JLTKc52DXbu8xO2jj/JIS8th+/bWFBScW5vWs+tXJF1wmCFXdiYppRcXXhhJ794QFdUQTyUiIiJlKVlrxgqKCnh69dPM+mAWDscVva7gjZveoFtMtxpfKzcXtmyBtWuPsPbd3Wz8ojVfHbiI/ILIc441K6Jbt7NccEEYF1wQTp8+3mggfftCnz7Qpk19PJ2IiIiAkrUWYeXXK7lzyZ0cOXOEjq06suCHC7i6b7nz2ddI7qHDrPv9a3yUdoSNOX35Z+YFHDrUh2++6UVRUcVt2uLj4bzzKi7t21M8rIiIiIhUTslaC3Ek6wjjl47nva/fA+DnqT/nieFP0DWma90v7hx8/jn5f1vE6a1vcaz1AbZ3O4+drfpw4EhfDh3qy8GDfTl0qA+HDvUhP//c2rhAMTFe0tazJ3TpAp07l3wGLiupExERUbLWohS5In794a/5t//7NwpdIVFhUUy5ZAqPXvZo/SRtfl9/DW+/TdFfl5B57CNODXScTILTgyCvtXH8eFeOHDmvuBw7Noj09AEcOdKDQ4diOXOmej1NIyJKErjOnaFTp4pLfDyEh9ffI4qISPOTnw+nTsHJk+WXrCyvEiA0FEJCKv4MCfEGnD9zBrKzS3+Wty0tDdq2bbjnUrLWAm0+spmn33+aJduWADRc0gZw7Bj87W+wbBnuvRWc6ZxL5oWQdQFk9g8hqy8UhRcVH+4cnD4dx4kTQ8jMHEZmZn9OnerNyZPdOHYsjvT0CI4cgW++gczMmoUSG+tNyhAbC3FxFX/6l9u29UqbNt7/oCIiUj0FBSWJS1aW95mb6yVL1S15eV7xL1e2raJP//KZM14ylp0dnO/j0CHoWs//vAZSstaCbfpmEzPXzCyVtP0s9Wc8etmjdGnTpf5vmJ0NK1fCe+/BqlWwbRsuBLJ7QuYFkJUYSebgGLK6ZFIYerbcS4SHd6ZVq4to3bo/ZgM5cyaRU6cu5NSpTqSnh3D0KBw96g36618+etQbR66oqNxLVkubNiXJW9kSE+PtDyzlbWvdGlq18kp4uF7hikjDcA4KC73fef5P/3J+vlcjVJ2SnX1u8dcWld3mT8j85Wz5v8KDLjTUa0LjL+3alV5v08b7/gK/u4o+IyNLfq+3bl358vnnN+zwpErWvgU2frORme/PZOn2pUAjJG1+hw/DP/7hlVWrYO9eAJxBTnfISo0le2hXsi+IILvDGbJDDlBUlFPupUJCWtGq1YW0anWRr/SnVauLiI6+gNDQKIqK4MQJOH7c+8zIKP/Tv5yRAadPe6WmNXjVERrqzeLlT978xb8tMrL8EhVVej06uqT4z69oPTpaNYTy7eacl0Tk5no1P4GlsLDqdf+2ij79NTnllbI1PeVdo+xyYA1TVbVLgYlEU2BW8keqv0RFec1XwsOrLv7jIiJKL9d2W3S098akdeuW+YeykrVvkbJJW3RYNJMumcSkSyYxIH5Aw97cOa+tmz9x+8c/zplWwYWFcHZEf7KH9+ZMYluyuxeSHf4N2dnbyM8/WsGFjaio3gFJXD+io/sSHd2XyMgEzKrOXoqKvL8a/clb2ZKV5ZXMzJLlsuuZmd5fmzk53mewfqH6f2lFR3u/OMsmc1FRXkJXnRIW5l0vLKzy5dBQ7zt0rqSUt+6Pr6JfvoHroaHeL9zqFP+vJf+9ylsO5D8vcLnsemX7ArcFPmN5z+2/t1lJO5iQkHPX/cX/D3h5CUDgekFBxc9T0+fwl6KiyhOQwOJPiHJzS5bL2xYaWr1/cENCqk6oAmuNyt7Pv95Ua3saQtn2Vf7l8PCK/xj0l4gI73dB4NuAikp0dOmkzL8cFdUyk6KmSsnat9CGwxuYuWYmy7YvK942pNsQJqZM5PZBtxMXHdfwQfinVVi3Dtau9cqWLee+y4yLg6FDyR86gOyUOLJ7hZLd+jjZOTvIzt5OTs5uoPzMyCyCqKjevuStT6nPqKhehIRENNjj5ed7rw9yckpeJQQuV+cVhf8fIf+5/vPLLgduE/m2i4goSUj8f1AE/mFR0bbAP1Iq2hb4R0VFxV9zVJtrVpTY+q8XmNjLt4uStW+xjd9s5KXPXmLRl4s4ffY0ABGhEdxw4Q1MTJ7I6L6jCQtpxDlCs7Jg/fqS5G3tWq+nQVkxMZCcDCkpFKUMJCexA9kJhWTn7yYnZ5ev7CYv73AlNzMiIroQGdmTqKgeREZ6JSqqZ/FyRERnzJrPb0XnvFoPf+KWk+Mle2XX/a9gqiqBNRz5+RUvFxaW1Bb5a4wCa2z861D6VVJFtUZ5ed41y9ZQVVSqqhnzL/u/o7K1b5XVzJW3HritvOesqPbP367IX8puKyws/Y93ZbWP/prH8moQa/IcZWv//ElOeQlEYPG/sve/tq/o019bV51XfP4a1aqSq7L3CVz219KJtDRK1oSc/ByWbV/Ga5te473d7+Hwfs5d2nThzsQ7+XHKjxnUaVDjB+Yc7Nvn1b5t3FhSDpeThIWGQv/+kJQEAwbAgAEU9O9FbldHTt7e4gQuJ2cXubm7yc3dB1TeI8EsnMjI7kRGJhAR0Y3IyG7lfoaFxTTM84uIiKBkTco4cPoACzYt4LVNr7Hz+M7i7Zd0vYTbB93OTf1v4vzY84MYIV5X0E2bSidwO3aU3x00MhIuvLA4gWPgQBgwgKLzzyOvKJ3c3H2cPbu/uASu5+cfq1Y4oaFtiIjoRkREVyIjuxIR0dW33K14OSKiK2Fh7TA18hARkRpSsiblcs6x7uA6Xtv4Gou+WMSps6eK96V0SeGmi27ipv43MSB+QNNIQLKzvTZvX3wBW7fCl196n/v3l398eLg3kWm/fl658MKS5c6dwYzCwhzOnj1AXt4hzp49xNmzB4uXAz8r6sFaVkhIVEDy1pmIiM6Eh3f2LXcKWO5MaGhM0/heRUQk6JSsSZVy8nP4286/sWT7Ev62829k5WUV7+vXoV9x4pbaLbXpJRinT8O2bV7i5i9fflk8jEi52rYtncT16eMNonP++d50CQHP6JyjoOAUeXkHOXv2MHl5JSVw/ezZQxQVnal22CEhUYSHdyY8vCPh4XGEh3cgLKxDJcsdfDV3arAjItLSKFmTGsktyGXV16tYsm0Jb+94m+M5x4v39Wjbg5v638QNF97AZT0uIzKs8jlCg+rMGdi1C3bu9F6h7txZsnzyZMXntWpVkriVLb16ef3cK1BQkOlL3r4hL+8IeXlHyM8/Ql7e0YBlrxQV1WYY7hBfAtfRl8B1JDy8g690LP4MC4slLKy97zOW0NDWTS/JFhGRYkrWpNYKigr4YO8HLNm2hCXbl3Ao81DxvlbhrRhx3ghG9RnFqD6j6N+xf/NICJzzptAKTN6+/rqkZGRUfn58vDdLfWDp2bNkOTa2WoMTFRae8SVwx8nPP05BwXHf8omA5eMUFJwoXi4sPF2rRzYLOyeBCw+PJTS0HWFhbQkLa0doaFvCwtpW+BkS0qp5/HxFRJohJWtSL4pcEZ8e/JQl25awYvcKNh/ZXGp/t5huXuJ2/ihGnj+S+NbxQYq0jjIy4J//LEne/Mu7d3s9V/PzKz+/TRsvaUtI8EqPHiXL/tK2ba1GmywqygtI5o4VJ3Ely8d8+zIoKDhJQUEGBQUZ1W5zV7mQMklcTBUJXruAJLCdb3s71fKJiJRDyZo0iMOZh1n59Ure/fpd3tv9HkfOHCm1f3DXwYzsPZLh5w3nez2+R2x0bJAirUeFhd64cHv3emXfvpJlf8nKqvo6bdqUJG7du0O3bueWLl28QaXqQVHR2XMSOBcH72EAACAASURBVK+cprDwNAUFpwKWS7b51wsKTuFcfQ0dX5L0hYa2ITQ0hrCwGF/y5y9tAra19S2fmxB6iZ/a8IlI89fkkzUzuxr4TyAUeMU590wFx90CLAaGOOcqzcSUrDUu5xxbjm7h3d3v8u7ud1mzdw1nC0v/4z6o0yCG9RjGsJ5eOa/9eUGKtgE559XM7dsHBw6UX/bv93q2Vkd8PHTtWjqBCyxdu3qfbdo07HPh1eoVFmZSUJAZkNRllknwyiZ/p4qTPf9y7drqVax0QhdDaGhrXxLYhpCQkmWv+NdbERLSyvcZ7Vv2PkNCon3bo5QIikij+f/Zu+/4yK76/v+vz/TRqJfdlbZ39/Xai41tTDXgAMH0YEhiEjv+JmD/ElrAJN+EEBIDoSbwTUKAQBIChoTigHENBtxwb9vX26RdadX79Dm/P869M3dGo7K70mokfZ6Px3mcc+/cuXNHo9W899xzz63osCb2xo77gFcDHcBjwLXGmF0l29UAPwVCwE0a1ipbPB3ngaMP8L+H/pcH2h/g0WOPksqmirZZXbs6H9xesuYlnNNyzpm9m8J8MQaGhgrB7fjx8qWrq/y8cuXEYsUhbvnyiWXZMlufgWA3lVwu7QS6ESfsjXhCoFtGi4KhbXvrYSf4zfzq21Ph80U8oW66OuIEv6inHXFCYPG6icVdvwR+/5VSZVV6WLsM+Lgx5rXO8i0AxphbS7b7InAv8CHgQxrWFpZEJsETx5/ggaMP8ED7Azxw9AEGE8VXZEYDUS5YfgHbV2znotaL2N66nfOWnUckEJmno55n2Sx0d9vg1tlZCHClpbPT3mNqpqqqikNcS4sNcuVKU5O9/0+FMiZLNjuaD3DZ7JgT9Nx6tOxyLjdOLhcnmx0nlxsnm407tV1vHz+Jn+ms8ReFuELQizo9fm67quSxMCIhfL4wPl/IaYfy60qX3SJity+03fVBHVeo1BlW6WHtbcDVxpgbnOXfAS41xtzk2WY78OfGmLeKyP1MEtZE5EbgRoA1a9ZcfGSqebbUvMqZHLt6dtnwdvQBHmx/kMODhydsF/AFOKflnEKAW7GdbSu2URuuPfMHXamMsXPNeQPciRMTS3e3rU8m2IlAY6MNdN7S3Dx5O7I4wrUxOSfQeYPc5HUul3CCXsJ5TmF5Yl2+GJOZ77edZ08Dh53gGPaEyOJ2IeSFSsJfyBP+3LAYzBefz20HStaFyuyruC3i1zCpFp1KD2tvB15bEtYuMcbc7Cz7gP8F3mOMOTxVWPPSnrWFpz/ez9NdT/Nk55M81fUUT3Y+yd7evfn7mHqtrVvL+cvP57yW8zh/+fmcv+x8tjZvJeSfnQH5i5YxMDJSHODc0tNTvNzdDX19hbuAz1QsZnvkmpuL68nWNTXZ5+iXL7lcBmOSnsAX9/QAll/OZscxJkUul8rXuVxywjpbJ53Hkp7tkp717rpprnied+LpAQyUlHLrAvmeRRsK3V5GbzgMFq0rbBcsCpqTtb1hsziMloZT73N1TKQqqPSwNuVpUBGpA14A3EvsVgD9wBunCmwa1haH0dQoz554lqc6n8oHuJ09OyeMfwPbC7e1aWtRiDun5RzW16/H7/PPw9EvApmMDWzd3XZuup4eW6ZqZ06hdygUsj143gDnLY2NhdLQUGhHoxry5oDtVUx5evySnh7A5IR6Yvgr105hTJpcLo0xhVJYzjjL3oDpBslU0b4gO98/olniK9PLOJPiB/xOD6N/ynXe59nHi/dTaPs8z3fbPs9+ve1yzy9ue4/Fvs/Stvf1CiG7ONQGllQPaqWHtQD2AoNXAcewFxi8yxizc5Lt70d71pa0TC7D/r79PN/9PM91P8dz3c/xfPfzvND/QtleuLA/zNbmrZzdfDbntJyTrzc3bdaeuNnmnpLt67Olt7e4dtu9vdDfX1gXP8V54MLh4vDmtqer6+sreiyempoxWSfAuSGvXCk8VgiE3l5G73Jxu3hduqTtDZ0TA2jhuROXS4Oqmgm/J8D5PaHTGwYnX54YDn1F66d73NveuPHvCATq5uydVnRYAxCR1wFfxE7d8Q1jzN+IyCeAx40xt5dsez8a1lQZY6kxdvXsyoe457ufZ3fvbjqGO8pu7xc/mxo3cXbL2Wxp3MLmps1satzEpsZNtNW04dNTFGdOPF4IbuXKwIAt/f3FJXkac79VV9vQdiqlrk7DnjotxhiMyRaFN7s8Wfj0htBsvoD7nKnWeZ9fvAxZcrk0kHOem/Psx23nnH1mJ+yz8FoT91143lT7dp+T9tRp55gqqwf18stPEAotm7P9V3xYmwsa1pRrODnMnt497OrZxe6e3ezqtfXBgYNle+LAXpm6sXGjDW8Nm/JBbmPDRlbWrlwaU4wsBPF4cXhzA9109dDQzKdFmYw37NXVTV5qa8uvq63VwKfUFGygLQ1ypaF0quXSkFgaHt115ULpxKC5fPnv4PdPfm/o06VhTaky4uk4e/v2srtnNwf6D7C/fz8H+g9woP8APeM9kz7PL35W161mff161tWvY139uqJ2W02bjpGrdLmcvdPE4ODUZWDABrvS9bMR9sBOo+INdOVqbym3rqYG/Pr7ptRCp2FNqZM0lBjKBzc3yO3v38+hgUN0jnZO+dygL8iaujWsrV/L2rq1rKtfx9q6taytt+2VNSsJ+oNn6J2oOVEu7A0NFcrwcPFyuceGh0/+StvJxGI2tLnhbaq6NOh5l/WqXKXmjYY1pWZRIpPg6NBRDg8e5tDAIQ4PHubw0OH8cuk9Ukv5xMfKmpX58La6djWra1ezqnYVq+tsuzHauKSuglqScjkYG5sY4Err6crIyOyFPpGZhb7Sdk2NPS1c2tbTvErNmIY1pc6g8fQ4R4eOcmTwCIcHD3Nk6IgtzvLxkeOTjpVzRQNRVtWuyge4VTWr8ssra1eyqnYVzVXNehGEKoS+kZFCePMGObceGrK19/HS7Wd6z9qZikTKh7iZLFdXF4q7HA5rz59atDSsKVVBUtkUHcMd+fDWPtxOx3BHoR5qZyg5NO1+gr4gbTVthQBXUwhyrdWttNa00lrdSiwUOwPvSi0KmYw9vVsa9Eprb8AbHbXLbu1tz8a4Pi+/f/Ig522XWxeL2eK23bqqSgOgqgga1pRaYEaSIxMCXMdwB8dGjuXr/nj/jPZVG64tCm9tNW1FyyuqV9Ba00pduE5PvarZY4y9WrdciJtu2Vt726mJk2HPCm+Qm654w5+7rlw7FNIQqE6KhjWlFqF4Os6xkWMcGy4EuGPDx+gY6aBzpJPO0U6Ojxwve7eHciKBiA1uboDzBLkV1StYHlvOiuoVLIstIxwIz/G7U6qMVGpigJtJe2TEnip2y+hooX2qEzJPJxAo7tGrqiq0p1rn7fWbbJ2eDl6UNKwptUQZYxhIDBSFN2+7a7SLztFOuka7GE2NTr9DR32kPh/gllcvzwe55bHltMRaaIo20VzVTHNVMw3RBh1bpypXNmvH6nlDXGno8xa3F9Ab/LzPc8up3HZtpny+8mHPuzxZEJwuLFZVQVCvVp8PGtaUUtMaTY3a8DbSWRTi3PrE6Am6RrvoHusma2Y+s7hPfDRGG/PhrbmqmeZoMy2xFpbFlrEstoyWqkK7uapZpzZRC5+3F9ANg2Njk7dLi7f3r3TdXJ0OdgWDxWGuXAisqiqUqZbLtaNRGzhVEQ1rSqlZkzM5+uP9+fB2YuwEJ0ZPcGLMLvfF++gd782XwcTgSb9GY7SxKMg1VzXne+uaqpomLOt4O7WkZDITg95MQ+BUj7vrs2fgNk/R6PShrlw7Gi0su+1y6xZgD6GGNaXUvMnkMvTH+4sCXM9YD73jvXSPddM93m1rp/SO95IzJ3cVYcAXyPfe5UNctKlssHOX6yP1eqcJpUoZA+n05MEuHi+sd0u5Ze86737Gx+dunGCpQKC4x88b5MqtK3289LHLL7fT0cyRkwlrOoOhUmpWBXyBfC/ZTGRzWfrj/fnw1jPeQ994obfO7bnz9uCNpkbz28+UINRH6mmqaqIx2khTtFCXrvOWukidjsFTi5eIvZI1FIKGhrl5jVyuEPpOJgC6Qc+7vtyyu20mU5hiZjZ0dsKKFbOzr9OkYU0pNa/8Pj8tsRZaYi2cy7kzek4yk6Qv3pcPdflAN943Idy56wYTgwwkBhhIDJzU8QlCQ7RhQohriDQU19GJy5HA3P2vXKkFw71AIjaHc0C6PYSThbqpwt5k7erquTvek6RhTSm14IQDYdpq2miraZvxc7K5LAOJAfrG++iP99MXd2rPcl+8j4H4AP3x/nwZSg7l2ycrGojSEG0oCnANkYnLbrsuXEddpI76SD3RQFTH5Sk1U94ewvr6+T6aWadhTSm1JPh9/vzVqScjk8swmBgsCnB9430MJGyoG4gP0J/oz4c87/p4Jk58JM7xkeMnfbwBX6AovHnb9eF6W5cpDdEG6iP1VIeq9fStUouEhjWllJpCwBc4pZBnjGEsPcZAfKAowA0kBoqCnfexoeQQQ4khhpJDJDKJfG/fqfCJj7pw3YQQ5wa9/LJTasO11IXrbB2xdcCnXxFKVQL9l6iUUnNARKgOVVMdqmZ13eqTfn4ykywKb4OJQYYSTu0sl5aBxEC+PZoaPaUxel5VwaqyIS6/7NTlHnNLdahar8JV6jRpWFNKqQoUDoRZFpj5VbWl0tk0w8nhCSFuMDHIQNyznLTLw8lhhpPDDCWH8u3x9Djj6XG6RrtO673EgjFqw7XUhGtsHaopCnT5wDdZKNSePrXE6W++UkotQkF/kKYqOy3JqXBP4w4lhvIhbigxxEhqpGhdacjzPjaSHGE0NcpYeoyx9Bido52n9Z4igQg1oZp8j2VN2LaL1jltbzCsCdfka+86DX9qodDfVKWUUhN4T+OuZOUp7ydncoylxhhJjeSD3Uiy0PaWfOArE/yGk8MkMgkSmQQ94z2z8h4jgciEQOf2ANaEyixPUevVu2ouaVhTSik1Z3zis6EmXHNSU62UMsYQz8QZTY3me+xGU6OMpEbKrhtJjjCcssHQDYqlbTf8nczkypPxi7+oty8WjNk6ZOvqYKHtPlaud9C7XBWs0gCoAA1rSimlFgARoSpYRVWw6pTH8Xm5p3ndAFdae8Ndvi6znVsnMgl7qjg5NAvv1hJk6lAXnLg+FooRC8aKgqF3XSwYI+hfWPfQVBrWlFJKLUHe07yttJ72/tLZdL5Xbyw1lh+r5/b2lVtXrmfQDYCjqVHimXg+JJ7ueD+voC+YD25VwSpiIaf2Lgeq8tt4A+FkPYHVoWrC/rD2BM4RDWtKKaXUaQr6g/aOFNHZu79mJpcpDnZTnP51l8fSY4ylxvJ1uXXpXDp/NfBs8omvKPR5ixsIi0LhVGGxZNltL9VpYOYlrInI1cCXAD/wNWPMp0oe/wBwA5ABeoDfN8YcOeMHqpRSSs2TgC+Qn7R4thhjSGVTjKXHGE+P50Oc2x5Pj08Id/lgOElAdJfTuXQ+TM6VsD88ZZjz9gqWC4DlSjQQLbSD0Yq888cZD2si4ge+Arwa6AAeE5HbjTG7PJs9BewwxoyLyB8BnwF+60wfq1JKKbWYiAjhQJhwIExjtHFW953Opoln4vnQ5xY3DHoD4YT1ZQKjN0y665PZJMl48pTu1TtTkUCEaCDKzvfupLXm9E+Rz4b56Fm7BDhgjDkIICLfBa4B8mHNGPNzz/aPAL99Ro9QKaWUUicl6A8S9AepDdfOyf6NMSQyiUmD3FSBz12OZ+JFQXI8PU48XVgXz8TzVwlHApE5eR+nYj7C2kqg3bPcAVw6xfbXAz8r94CI3AjcCLBmzZrZOj6llFJKVRgRIRqMEg1G5+w1ciZHIpMgno5TF6mbs9c5WfNxYrbcpSKm7IYivw3sAP6u3OPGmK8aY3YYY3a0tLTM4iEqpZRSaqlxL5JoqmqqqLFr89Gz1gF472q8CjheupGIXAX8GfAyY0zyDB2bUkoppVRFmY/Y+BiwWUTWi0gIeCdwu3cDEdkO/DPwRmPM6U8trZRSSim1QJ3xsGaMyQA3AXcBu4HvGWN2isgnROSNzmZ/B1QD3xeRp0Xk9kl2p5RSSim1qM3LPGvGmDuAO0rW/YWnfdUZPyillFJKqQpUOaPnlFJKKaXUBBrWlFJKKaUqmIY1pZRSSqkKpmFNKaWUUqqCiTFl56NdcESkBzjVm703A72zeDhqdunnU9n086lc+tlUNv18KteZ+GzWGmNmNKP/oglrp0NEHjfG7Jjv41Dl6edT2fTzqVz62VQ2/XwqV6V9NnoaVCmllFKqgmlYU0oppZSqYBrWrK/O9wGoKennU9n086lc+tlUNv18KldFfTY6Zk0ppZRSqoJpz5pSSimlVAXTsKaUUkopVcGWfFgTkatFZK+IHBCRj8738Sx1IvINEekWkec96xpF5B4R2e/UDfN5jEuViKwWkZ+LyG4R2Skif+ys18+nAohIREQeFZFnnM/nr5z160Xk187nc5uIhOb7WJcqEfGLyFMi8hNnWT+bCiEih0XkORF5WkQed9ZVzN+2JR3WRMQPfAX4DeAc4FoROWd+j2rJ+yZwdcm6jwL3GWM2A/c5y+rMywAfNMacDbwYeJ/z70U/n8qQBF5pjNkGXAhcLSIvBj4NfMH5fAaA6+fxGJe6PwZ2e5b1s6ksrzDGXOiZX61i/rYt6bAGXAIcMMYcNMakgO8C18zzMS1pxphfAv0lq68BvuW0vwW86YwelALAGNNpjHnSaY9gv3RWop9PRTDWqLMYdIoBXgn8l7NeP595IiKrgNcDX3OWBf1sKl3F/G1b6mFtJdDuWe5w1qnKstwY0wk2MADL5vl4ljwRWQdsB36Nfj4VwznN9jTQDdwDvAAMGmMyzib6N27+fBH4UyDnLDehn00lMcDdIvKEiNzorKuYv22B+XrhCiFl1ulcJkpNQUSqgf8G/sQYM2w7CFQlMMZkgQtFpB74IXB2uc3O7FEpEXkD0G2MeUJEXu6uLrOpfjbz5wpjzHERWQbcIyJ75vuAvJZ6z1oHsNqzvAo4Pk/HoiZ3QkRaAZy6e56PZ8kSkSA2qH3bGPMDZ7V+PhXGGDMI3I8dW1gvIu5/zPVv3Py4AnijiBzGDrd5JbanTT+bCmGMOe7U3dj/6FxCBf1tW+ph7TFgs3NFTgh4J3D7PB+Tmuh24DqnfR3w43k8liXLGWPzdWC3Mebznof086kAItLi9KghIlHgKuy4wp8Db3M2089nHhhjbjHGrDLGrMN+z/yvMebd6GdTEUQkJiI1bht4DfA8FfS3bcnfwUBEXof9H44f+IYx5m/m+ZCWNBH5DvByoBk4Afwl8CPge8Aa4CjwdmNM6UUIao6JyEuAXwHPURh38zHsuDX9fOaZiFyAHQTtx/5H/HvGmE+IyAZsb04j8BTw28aY5Pwd6dLmnAb9kDHmDfrZVAbnc/ihsxgA/tMY8zci0kSF/G1b8mFNKaWUUqqSLfXToEoppZRSFU3DmlJKKaVUBdOwppRSSilVwTSsKaWUUkpVMA1rSimllFIVTMOaUmpREpFbReTlIvImETkjN2AWkcMi0nwmXksptXRoWFNKLVaXYueAexl2fjillFqQNKwppRYVEfk7EXkWeBHwMHAD8I8i8hcislFE7nRu1vwrETnLec43ReSfnHX7nHs5IiIREflXEXlORJ4SkVc46/0i8lln/bMicrPnEG4WkSedx9z9v0xEnnbKU+5s6UopNRNL/UbuSqlFxhjzYRH5PvA7wAeA+40xVwCIyH3AHxpj9ovIpcD/w96nEWAdthduI/BzEdkEvM/Z5/lO8LpbRLYAvwesB7YbYzIi0ug5hF5jzEUi8l7gQ9iw+CHgfcaYB0WkGkjM5c9AKbW4aFhTSi1G24GngbOAXQBOSLoc+L69zSkAYc9zvmeMyQH7ReSg89yXAP8AYIzZIyJHgC3Y+27+kzEm4zzmvQWNe4P7J4C3OO0Hgc+LyLeBHxhjOmbxvSqlFjkNa0qpRUNELgS+CawCeoEqu1qexvaaDRpjLpzk6aX33jOAlNvQWT/Zvfrceztmcf7GGmM+JSI/BV4HPCIiVxlj9kz/jpRSSsesKaUWEWPM004Y2wecA/wv8FpjzIXGmCHgkIi8HWyCE5Ftnqe/XUR8IrIR2ADsBX4JvNvZfgv2hs57gbuBPxSRgPOY9zToBCKy0RjznDHm08Dj2F47pZSaEQ1rSqlFRURagAHnlOZZxphdnoffDVwvIs8AO4FrPI/tBX4B/Aw7ri2BHdPmF5HngNuA9xhjksDXgKPAs86+3jXNYf2JiDzvbBt3XkMppWZEjJmsJ18ppZYGEfkm8BNjzH/N97EopVQp7VlTSimllKpg2rOmlFJKKVXBtGdNKaWUUqqCaVhTSimllKpgGtaUUgoQkVER2XAGX2+5iPxSREZE5HNlHv+miHxyFl/POHdlUEotMDoprlJqVonI/cA2YIUzzcWCYIypPsMveSN24t5aM8uDh53P4D+MMV+bzf0qpeaH9qwppWaNiKwDrsTO7v/GM/zaC+0/n2uBXbMd1JRSi4+GNaXUbPpd4BHsLZ+u8z4gIlER+ZyIHBGRIRF5QESizmMvEZGHRGRQRNpF5D3O+vtF5AbPPt4jIg94lo2IvE9E9gP7nXVfcvYxLCJPiMiVnu39IvIxEXnBOf34hIis9uxrk9MOi8hnReSoiJwQkX/yHGuziPzEOdZ+EfmViJT9Wyoil4vIY877fUxELnfWuz+fP3VOv141yc+zWUTucY71FyKydgb7/htsYP6ys+8ve/Z3lYjsF5EBEfmKeG6SqpSqXBrWlFKz6XeBbzvltSKy3PPYZ4GLsTdTbwT+FMiJyBrsjP7/ALQAF2Jvwj5TbwIuxd5eCuAxZx+NwH9ib9wecR77AHAt9h6dtcDvA+Nl9vlp7A3bLwQ2ASuBv3Ae+yDQ4RzrcuBjlLlPqHMLqp8Cfw80AZ8HfioiTcaY92B/Rp8xxlQbY+6d5L29G/hroBn7M/n2DPb9Z8CvgJucfd/k2d8bgBdhT1O/A3jtJK+rlKogGtaUUrNCRF6CPbX3PWPME8ALOLdhcnqefh/4Y2PMMWNM1hjzkDOm7d3AvcaY7xhj0saYPmPMyYS1W40x/caYOIAx5j+cfWSMMZ8DwsBWZ9sbgD83xuw11jPGmL6S9yHAHwDvd/Y7Avwt8E5nkzTQCqx1jvdXk5zKfD2w3xjz786xfAfYA/zmSby3nxpjfun8nP4MuMzpCTzVfX/KGDNojDkK/BwbRpVSFU7DmlJqtlwH3G2M6XWW/5PCqdBmIIINcKVWT7J+ptq9CyLyQRHZ7ZweHATqnNef6Wu1AFXAE86pzkHgTmc9wN8BB4C7ReSgiHx0kv20AUdK1h3B9tLNVP69GWNGgX5nv6e67y5Pexw40xdVKKVOgYY1pdRpc8ZzvQN4mYh0iUgX8H5gm4hsw171mAA2lnl6+yTrAcawwcm1osw2+V4tZ3zaR5xjaTDG1ANDgDs2a6rXcvVib7Z+rjGm3il17tWixpgRY8wHjTEbsD1ZHxCRV5XZz3FsT6PXGuDYNK/vtdrz3qqxp3aPz2DfetGCUouIhjWl1Gx4E5DFjhu70ClnY8dO/a4xJgd8A/i8iLQ5A/0vE5EwdhzWVSLyDhEJiEiTiLin554G3iIiVc7g/+unOY4aIAP0AAER+Qvs2DTX14C/FpHNYl0gIk3eHTjH+i/AF0RkGYCIrBSR1zrtN4jIJud06bDzvrNljuUOYIuIvMt5X7/l/Hx+Ms178Hqdc/FFCDt27dfGmPYZ7PsEcMbmjFNKzS0Na0qp2XAd8K/GmKPGmC63AF8G3i12Wo0PAc9hLwDoxw7i9znjp16HHbjfjw1o25z9fgFIYcPHt3AG2E/hLuzFCvuwpwUTFJ8m/TzwPeBubND6OhAts5+PYE91PiIiw8C9FMa9bXaWR4GHgf9njLm/dAfOWLg3OO+rD3tBxRs8p4ln4j+Bv8T+XC7Gju+byb6/BLzNuerz70/i9ZRSFUhv5K6UUkopVcG0Z00ppZRSqoJpWFNKKaWUqmAa1pRSSimlKpiGNaWUUkqpCrbQbnw8qebmZrNu3br5PgyllFJKqWk98cQTvcaYlum3XERhbd26dTz++OPzfRhKKaWUUtMSkdK7kExKT4MqpZRSSlUwDWtKKaWUUhVMw5pSSimlVAXTsKaUUkopVcE0rCmllFJKVTANa0oppZRSFUzDmlJKKaVUBdOwppRSSinlSGVTdI50YoyZ70PJWzST4iqllFJKZXIZxtPjRWUsNcZYeoyB+AC94730xfvoHe8tKu664eQwAEMfHaI2XDvP78bSsKaUUkqpeWGMIZFJMJYeYyw1xmhqNN8eSY0wnByeUIaSQ0XLI8kRxtJj+WCWyqZO65j84qepqomhhIY1pZRSSi0AxhjG0+P0x/snLUPJIVLZFOlcmlQ2ZdtZT9uz3u3lcuucyc3q8frER1WwiqpgFbFgLN+uClbREG2gOdpMc1WhNFU1FS3XhmvxSWWNEtOwppRSSi1SbtAqPeU3khphJDnCaGqU0dQoI6mJ7ZHkCEPJIfrj/afdWzWVkD9ELBijOlRNLBTLt6tD1dRF6qgN1VIbLpS6SF3Rck2oJv+8qmAVIX8IEZmz450PGtaUUkqpCmeMIZ6JMxAfYCAxwEB8gMHEYL49kBigP94/IZT1jPeQyCRO+/WjgSiN0cZJS124jnAgTNAXJOQPEfTbOuQP5de5690eLzecBXwaRaajPyGllFJqDrnjstzxWG4P1lBiiMHE4MSSLF52w9ip9m6F/WFaYi00VzXTUtVCU1UTtaFaasI1VIeqqQnV5Huy3HXu+tpwLY3RRqLB6Cz/VNTJ0LCmlFJKTcE7Zqsv3kffeF++zq9zlgcTg0WBbDbHZYX9YRqiDdRH6mmINNAQbbC1p+2G5+CRugAAIABJREFUMre0VLVQFaxadKcFlxoNa0oppZakZCbJibETdI12TVvimfhpvVbYH873WMVC9hRgfaSe+kg9deG6fLu01IXr8kFMe7eWLg1rSimlFpx0Ns1QcqjoVKH31KF3iod8nShePpmxXGF/mKaqJpqiTTRVNdEYbbTtqNN2HquP1FMTrikaJK/jstTp0t8epZRSFWE0NcrxkeNFpXOkk+Ojtt0f788Hs7H02Gm/XsAXYHlsOSuqV0xZlseWUxOumYV3qNSp0bCmlFJqVhhjGE4O0zvem+/dyk9cOskEp4OJQTpHOzk+cpzR1OiMX8snvgmnDBsiDflTh+70DnXhuqLpHrzL0UBUx3KpBUHDmlJKqSLGGEZSI0XTQ7g9Wu5Aeu/tebyD7jO5zCm/biQQoa2mjbaaNlqrW/Ntd7m5qjk/wL46VF1xE5cqNVfmNKyJyNXAlwA/8DVjzKdKHl8LfANoAfqB3zbGdDiPXQf8ubPpJ40x35rLY1VKqcUqm8vSO95bNJj+xKjTHuuie6x7QjA71asXq0PVNEWbaIg25HuxasI1EyY2rQnX5Hu6Wmtaaa1upT5Srz1dSpUxZ2FNRPzAV4BXAx3AYyJyuzFml2ezzwL/Zoz5loi8ErgV+B0RaQT+EtgBGOAJ57kDc3W8Sim1UBhjGE2N0jPeQ/dYNz1jTu0su+0Toyc4MXaC7rHukw5fsWBswjQR9ZH6/KB69xY9TdGm/C17mqJNhAPhOXrXSi1dc9mzdglwwBhzEEBEvgtcA3jD2jnA+532z4EfOe3XAvcYY/qd594DXA18Zw6PVyml5lUqm8r3fHWOdNp6tLjuGrU9YSc7K31TtKkwYL56OStihfay2DIao435MV/1kXqC/uAcvUul1Mmay7C2Emj3LHcAl5Zs8wzwVuyp0jcDNSLSNMlzV5a+gIjcCNwIsGbNmlk7cKWUmg05k6NvvK+4x6ukF8z7WH+8f8b7jgaiLIstY1lsGS2xFltXFdfLYstYUb2CZbFlGr6UWsDmMqyVG3hgSpY/BHxZRN4D/BI4BmRm+FyMMV8FvgqwY8eOCY8rpdRcSGVTnBg9QedoJ50jncW1p31i9ARZk53xfr1TSbTWtOZ7v1prWidMJRELxebwHSqlKslchrUOYLVneRVw3LuBMeY48BYAEakG3mqMGRKRDuDlJc+9fw6PVSm1xI2mRukc6cz3crljvSa0R08wkJj58NmGSEO+l6uo9yvWMmFdU1WTXuGo1OkyBhIJGBmB0VFbe8vwcKF2S7nlZ5+FmsqYX28uw9pjwGYRWY/tMXsn8C7vBiLSDPQbY3LALdgrQwHuAv5WRBqc5dc4jyul1ElJZ9P5ebyODR+z9cix/KSrbns4OTzjffrEx/LY8vxVjK3VrYW2p14eW64D7pWajDGQTk8sqZStx8dhaAgGB8vX3rYbxNxwlp15j/akhocXf1gzxmRE5CZs8PID3zDG7BSRTwCPG2Nux/ae3SoiBnsa9H3Oc/tF5K+xgQ/gE+7FBkop5TLG0Dvey9Gho7QPt3N06OiEdudIJ2biKIoJ3Dm+lseWF/WELYstm7CuMdqI3+c/A+9QqQplDMTjMDBQKP39E0vp+oEB2+uVTs9OoJpMOAzV1TZsuaW6GmprC6WmZurl5cvn7vhOkhizOIZ67dixwzz++OPzfRhKqVmQyCTy0064c4Ll22Mn8nOEtQ+3T3tVpNsLtrJ2JStrVtJW01aoawvLOseXWnJSqULv1GS9V94wVlpSqdM/hmCwUAKB4uVoFOrrC6WurnzthitvOAtW/gU1IvKEMWbHTLbVOxgopc64ZCbJ4cHDHBw4mC8vDLzAwYGDHB06ylByaMb7qo/Us7p2NWvq1uSLd7mtpk2vhFSLh9ujVXo60C2l47MmK0NDdj+nIxyGhoZCaWycvjQ02BAWDILfD/ofpBnRsKaUmlVur1h+pnynR+zI4JF8IOsY7pjy1KR7VeTy6uX5qyPdZW97Ve0qasO1Z/DdKTWFTMaGoEQCksmJdem6eNyOyyotpeuHh4sDWebUb+lVJBCYvLfKrevriwOZt0Sjs3Mcaloa1pRS08rkMnSPdecDWNEti8aKl2fSK+YXP2vq1rChYQMbGzayoWFDvqytX0tjtFGvilSVwRgYG4PubujshK4uW3uLu667224/18LhQqByi3s60DtGq3TZW+rqoKpKe7YWCA1rSilGU6Ps69vHnt497Ondw+HBw0WhrHe8d0aD9AGCvmC+ByzfCxZbzpq6NWxstMFsde1qPTWp5l4uV+iVcsdlecvQUOFx77QNpSV3Erfqqq+HSMSWcHjyOhy2YalciUaL2zU1xeEsrFcYLzUa1pRaIowxdAx3sKd3D3v79uaD2d6+vXQMd0z5XEHys+HnS8zeqqi1urUonDVEGnSgvjo52aw99TfZaUG3jI3ZqRmmK8PDNowND89OT1c0Ci0t0Npqy4oV5dvLli2Ige1q4dGwptQiMxAfYF/fvkLpL7TH0+NlnxPyh9jUuImzms9ia9NWNjZspK2mLR/AWmItBHz650JNwhg7DssNS2NjNiz19U1f3Kkc5kptbfH4q9KrC+vqiqdrcLf3TucQ0N99Nb/0N1CpBap7rJsnO5/k6a6ni8JZz3jPpM9ZFlvG1qat+VB2VvNZbG3eyrr6dRrGlqpUCnp7p56+wVt7e7C8PV0nc6qwnJmcDqyuLkzP4LYnKw0NNmz5dT48tfDpX2elKpwxhuMjx3mi8wme7HwyX46NHCu7fVWwii1NW9jatJUtTVvyZXPjZhqiDWWfoxYZN4B1d8OJE3YAfFdX+Xb/LM03HgrZkBSLFSYfbWoqX5qbC+2GBh3ortQ0NKwpVSHc2fjdOcd2du/kyS4bzLrHuidsXxOqYXvrdrav2M7ZzWfnQ1lbTZuOGVsM3Pm0yt3fsK8PenomL0Mzn6cOn8+Ox5pq8lHvKUNvr5YbzGIxHaul1BzSsKbUGZTKpormGyudEHY0NVr2eQ2RBi5qvShfLm69mI2NG3V6i0qXy9ng1N9fPE6r3PLAQHEgO51Tiz6f7bVatswOgF+xwt46p1y7qUlPFSpV4TSsKTVHjDEcGjzEIx2P5MvTXU+TzqUnfU5duC4/vcWWxi02mLVdzNq6tdpbdiYZU+jBKi29vTZcuWO2Jivu5KanIxIpvoWO225qsr1hzc22Li0NDTawKaUWBQ1rSs2S4eQwjx17zAazYzac9Y73Fm0jCOvq15WdDHZjw0YdUzbbjLHTN7gBq/TG0qVtb09XevJQfVLcsVuNjcXjtrzL7q14vJOW6qlFpZRDw5pSp6B3vJdnup7hmRPP8HTX0zzZ+SS7enZNmDi2paqFF696cb7saNuht0eaLbmcDWEdHbYcO1a+Hi1/anlaVVWTD5BvaioEqqlKNKo9XErNI2Oy5HJpjHFLDsgBxmkXasgVrYtG1yNSGUMENKwpNYVsLsuB/gP5UObWx0eOT9g26AuyvXU7L15ZCGfr6tfp6cuTlc0Wbu1TWo4fL77Fz0x6v6qq7Ngt98pD92bS3htLe2s3jEUic/9elVoAbOBJksulMMbWuVwSY7x1CmOy2MBTqG34sbVdn3XWZz3LGc+6TH69fZ1xstnxMnU8v5zLxZ3Xt4HMG86Y4Z1Xyrn88m5CoZbZ+BGeNg1rSjmMMRwcOMijxx7l0WOP8utjv+aZE8+UnUg2FoxxwfIL2LZ8G9tWbOPCFRdy4YoLiQT0C35KuZwNWe3tttervX1i6eyc+cD6xkZYtQpWriyuve26Op0WQs0KYwy5XIJsdpRsdsSpS9tjnlAzecCZrF1uHYBIEJEgPl8w3y63bP9z6EPEBxS3C+vECTUJ57UKtT3WRP4Y7Otn5+tHPguk6GdV+HkU/0zKrbOlMmhYU0tW73gvjx17jF8f+3U+oPXF+yZst7p2tQ1kyy9k24ptbFu+Ta/ELGdkxPZ8TVba2+2pyUxm+n15b+1TWtraCrf5iUbn/n2pimCMe5qqXG9NrmRdxgk6aU+vT8rT+5LKhycbtIbJZIanqW0gW9jB5VQJPl8YkVBRXbwuiIjfOW3oc9o+wNaF9T5EAs56PyIBz/OKl0VC+P0x/P4qfL4oPl+V0/bWUXy+qBPGQmWCbGWcxjxdGtbUkpDOpnnmxDM83P4wD3c8zKPHHuWFgRcmbLcstoxLV17KJSsv4ZKVl3Bx68U0VTXNwxFXGGNsj9jBg3DoUKEcPlwIYyMjM9tXczOsXj15aWuzE6yqBc2YnBN0hshkBp0yVFIPks0OO71R8UlOdxVOe53OKa3ZYgNEDX5/NYGArW2xbZ+vKh9kygecECK2tusnaxdqwDPmynuaLzNhnf0ZueOwcp6Q6x2XZfKv4/NFnNe1bW/trrcBqnJ6mZYiDWtqUeoa7coHs4c7Hubx44+TyBTff7AqWMXFrRfng9mlKy9lTd2apflHKZezM9ofPWp7wI4csWHMDWeHD09//8ZIxJ56bGubWFpbC6cntTfsjLBf0u4Xt9umZJ27nHNOfSWc03zxomU7JijhBKoRJ2zZYsPYsKftrh9h9sOVb9LemuJ1ficUhfK9LcXLhXV+fy2BQO00dQ1+fy1+fwyfT/8joc48DWtqwUtn0zx74lkean8oH84ODx6esN3Wpq28eNWLuWzVZbx41Ys5d9m5S+d+mMPDNoAdOVIYG+YGs6NH7enJ6QbrNzXB+vW2bNhg63XrbABra7Oz3C/FoDtLjDFks6NkMgNOr5Ot0+niZdsbNeYEqLgToOJll+1Vb/PH768hEKh3Sl2+7ffXeda7IajcKS7v6a/oojmlpdTJWiLfVGox6Y/380jHIzx49EEe6niIR489OuEigOpQNZeuvJTLVl3GZasv49KVly7u05kjI4XesMOHJ5aZ3P+xpaVwKnLNmkIgc0tNzZy+hYXCDjJP5q9Cs8FofIrQNObpeZp6bNTcjIeSfCn0GgveAdQiPufUV9SpS9sRZ2xQxAlgdU6PU53Trsu3C4/VarhSapZoWFMVzRjDvr59PNT+EA+2P8hD7Q+xu3f3hO22NG3hslWXcfnqy22vWcu5+H2L6ItidHTqMNY38cKIItEorF1rixvGvPUSOD1pQ1Z8ih6rASdQea/uK1zl5w4wn8tB5j5fLN/jFAw2OO3Sus4ZG2UHVrsDrL2DrQtF/8QrtRjM6b9kEbka+BLgB75mjPlUyeNrgG8B9c42HzXG3CEi64DdwF5n00eMMX84l8eqKkfveC/3vHAPd75wJ3cduIsTYyeKHg/7w7xo5Yu4YvUVXL76ci5bdRktscqYC+e0DA3Bnj2FcuBAIYz19k793HDYnpJct65wetJbli1bsKco3Z6sidMkjHgGsHvHSg2XjKdyy4AzAPv0iQSdU3fRMqGpeNlezVbnGf9UN8nYqBodD6XULDLGllzOTt/ozgjk/ikUKZTSZe+6SjBnYU1s//dXgFcDHcBjInK7MWaXZ7M/B75njPlHETkHuANY5zz2gjHmwrk6PlU5srksjx57lDsP3MmdL9zJY8ceK7oTwIrqFflgdsXqK9jeup2Qf4F+qRlj5xfbswd27y4OZ52dkz8vHLa9YuUC2dq19sbcFThTfi6XJJ3uI53uLQlTU0+TUNqrZcwMpvuYAZ8vUqa3qiHfk2VD1cQr/IqXdZC5WtyMsdcTjY4WhrJOFWzc5VyuEI68be+6XM7eOndkxJbh4UK7tIyPQzJpSypVXJeu8wYyt57pdI2T6eqyf1orwVz2rF0CHDDGHAQQke8C1wDesGYA9947dcDEaeHVotQ50pkPZ/e8cA8DiYH8YyF/iJetfRlXb7qaqzddzdnNZy+8KzQTCdi/vziM7dkDe/fav1TlRCKwdSucdRacfTZs3lwIZfMcxozJkskMkU73O6cLB/LtdLo3H8hsKCu07SnD0ycScIJScXgqN2bKO5aq8Jg7sF0nLa40mYwNBdN9cY+M2C9fEftPwa3Ltf1+O/tLKGRvr1quHQpBwPMN6PbCuMW7Duy+AwFbgsFCu1zx+wt1ads9RoB4HAYH7W1pp6qDQTvjTUuLLW7brUtvtuG9JW5fn629pa+v8DMdHS38/N326KgNPIuF+3P3/gmd7LP2rqukr525DGsrgXbPcgdwack2HwfuFpGbgRhwleex9SLyFDAM/Lkx5lelLyAiNwI3AqxZs2b2jlzNie6xbv5r13/xnee/wwNHHyh6bHPj5nw4e9nalxELxebpKE+CO/fYvn02hHkD2aFDhX/xpZYts4HMW84+244fm8NAZq82HHbGag2QyfTPuJ3NDp3Sa4oECAabCQSanJ6r6adJsAPYi3u3lmJPVjJpv6gHB+0ZcrceGrJfIuXCR2kw8fsn9nKUa7uBqfSLu9yXuTHlg1LpulzO/p8lHre1t3jXzeSOYYuN329/RjOZH3omqqttcKuqstcS9fae/r7DYbvfUGjqYONddj//qX4vfD5729yamuJSWztxuarKHkcoVFyXrguFCoGstK6kwHU65jKslfsRlX57XQt80xjzORG5DPh3ETkP6ATWGGP6RORi4Ecicq4xZrhoZ8Z8FfgqwI4dO+Z/tkQ1wWBikB/u/iHf3fld7jt4H1lj/7sWCUS4asNVXL3RBrSNjRvn+UinMDxsA5m37N1r68luEu73w8aNhTDm9pht3WqnwJgjmcwIicQh4vGDJBIHS+rDGJM85X37/XUEg435U4d2AHwDwWALwWATwWBzSWnC76+d915RY2xnZmmPRbleDDeMTPelZIwNGalUoXZL6TKU730p7Z1xe0PcgJY89Y9qQRGxoaD0y7r0S7y6uvBzmi6AZrNTfy7e9kzGLLnhM5OZWNLpieuy2UJd2naPD2zIaGiws95MVtfX29fo7YWeHlvctlu7gdrLDXBuaWoqbnt/rtXVxe3qavv7qSrHXIa1DmC1Z3kVE09zXg9cDWCMeVhEIkCzMaYbSDrrnxCRF4AtwONzeLxqloynx/mfvf/Dd3d+lzv230Eqa7+xAr4Ar9/0eq4971reuPWN1IQrbCqIVMqOI3vmGXj6aVt277a9Z5NpaLABbPNm2zvmhrONG2d1Fn5jDJnMAKlUJ6lUF8mkrVOpTpLJYyQSh0gkDpJOT30hgs8Xyweu4uBVvl1YrpswDUM6XThd4n5JplK2t6TceJV0emZjT+JxG65GRyfW3vb4uN1n6ZdhuS/HhSgYLHxZ19UV6tpaGx6mCyGplH3/bi/OZKcM3V4I7xe1+8Vd+mUeixV668qNR/K2fT57gXEkUiily5GI/WeyWHo/ZsKYwu9lMHj6790Y2+vZ02P/TTQ22jBWempULWxzGdYeAzaLyHrgGPBO4F0l2xwFXgV8U0TOBiJAj4i0AP3GmKyIbAA2Awfn8FjVaRqID3DvwXv54Z4fcvve2xlL23FZgvCKda/g2vOu5a3nvJXGaOM8H6mjv9+GMm8w27Wr/DmZSMSGsS1bJpamptP6a2tD2CDJ5DFSqeOe+rhTd+YDmjGpaffn80WIRNYTiWwgGt1AJLLeqW07EKjGGPtH3T211t9faHtPtQ0PT12mu6FBJYlGp+/BaGiwocQNMdNdMVZ6CnKy05Lu6a7pemSMKYSy+nr7a7eUQsxSIVI8Vm429ldba4tavOYsrBljMiJyE3AXdlqObxhjdorIJ4DHjTG3Ax8E/kVE3o89RfoeY4wRkZcCnxCRDHZCoz80xsxgVk91puRMjqc6n+JnB37GnQfu5JGOR/KnOAEuXXkp1553Le849x201rTO44Hm7C2T3FDmBrSjR8tvv2kTXHihLdu2wfnn23nITnEsWTY7TiJx2Dk1eYhE4jDJZEdRMMvlEhgDAwPLOXr0LI4e3Up7+1kMDFyOMT6yWT/G+DAmDFRhTBUiEYyJAhGMse1cLooxQbJZmbSnaXTUBrLZGCvj9sYEAtOPUXHDTel4k3LtSMTuNxabuq6qKozLmmwwt1s09CilFjIp3B9uYduxY4d5/HE9SzqXesd7ufuFu7nzwJ3c9cJddI915x8L+AJcsfoKfmPTb/COc9/B+ob1Z/4Ax8fhueeKg9mzz5YfVxaJwAUXFELZhRfaYHaSs/TnckkSiXaSySNOGCuUePwQ6XTxHHGpVJhjxzbS3r6Vo0fPor19K+3t53L06FbGxs7caeFIpLgXp7TtFvd/7OVKNKohSCmlTpWIPGGM2TGTbXV6azWl9qF2/u2Zf+P2fbdPmP9sde1qfmPTb3D1pqt51YZXURs+w/3wJ07Ar35lyy9/aYNZuQFKra2FQLZtmy2bN095LmJoCPbtM3R3jzI+3kMy2UM83ksi0Uci0U8y2U8yOUgyOUou5yedDjE2VsfYWB2joy9jdPQap93A+HgzY2P1jIzUkExOPo6tvr74eoSVKwu9VpNd6eTWpb1Jky1XV9sQFg7PxgeglFLqTNCwpiZIZVPcvvd2vv7U17nrwF35gDav858ZY2fyd4PZr35lr8b08vvhvPOKQ9m2bXaqjBK5XIaRoWPs2dPH7t3j7NuXZf/+EAcP1nLkSAt9fY3YC5prnLJhVt5GIGCnTfNeHOrWLS3aU6WUUmoiDWsqb2f3Tr7+1Nf592f/nd5xe1VhyB/izWe9mXef/25euf6VZ27+M2PsnGW/+EUhnHV0FG8Ti8Fll8FLXwpXXgmXXIKJVjE8DN3dhuPHBzl+Xw/Hjx+iq2uM7u40PT0+enoidHSs4sSJtRizsuzLh8PjrFy5n6amHoLBEMFgmGAwTCgUIRiMEgpVEQxWEQpFCQR8BIPFV+yVO7VYX6+nDpVSSp08DWtL3EhyhNt23sbXn/o6j3Q8kl9//rLzueGiG3j3+e+mqWru5gXLy+Vg504bztyA1t1dvE1jow1lV17JwLaX87z/fJ7ZOcazz47x/Md8HDqUoq8vSDodxPaKNTilPL8/Q1tbB+vWnWDDhmE2bUqyZYufs8+Osm5dM9HoOvz+C+Z9njCllFJLm4a1Jeq5E8/xxUe+yG07b8tPs1EbruXa867l+u3Xs6Ntx9yGlFzOjjHzhrO+vuJtVqxg6PLX8tyWV/BMeBPPddaya3eYvZ9porvbDZAhSgNZNDpCfX0P9fX9NDWN09ycYdkyYfnyCCtW1NLW1sjWrU1s3BgiFFoD6N0vlFJKVS4Na0vMg0cf5NYHbuWn+3+aX3flmiu54aIbeNs5b6MqWDV3L97eDnfdZct998HAABn8dLCKF+QC9q7Zzr6V2zko6zg8uJz2480M/qB8z1g4PM7atbvYuPEAW7b0cvbZSc45J8zq1ctpbFxLNLqRQGC99ooppZRa8DSsLQHGGO7YfwefevBT+XtyRgNRrt9+PTdfejNbmrbMzQvH4/CLX5D92d20//RZ9r7gZy9b2R1+Gbujf8Ch2s0cH1tFJhu0s+wddYpHKBRn1aqDbN7cwVlnDXHOOTm2batiy5ZWqqs3EwzO6KpnpZRSasHSsLaIZXIZbnv+Nj794Kd5rvs5ABoiDdx0yU3cfMnNtMRaZu210mnoPmFo/+Uh9v7PPvb9up+9RyLszW1mP39LEs+9T5JOcTQ3d9DWdpCVK7tYty7Ohg3Cpk01nHXWCtav30godA4i587asSqllFILiYa1RSiejvONp77BZx/+LIcHDwPQVtPGBy/7IH9w0R+c9D05jx6F/fvtLTLLluM5evt92EH9Gyg3zUVz8zFWrdrH6tV7WbPmEFu3BjnrrDo2bVpBU9MWqqq2EQi89LTfu1JKKbXYaFhbRBKZBF9+9Mt85sHP0DPeA8CWpi185IqP8O7z3004MLOZUDMZeOQR+MlPbNm5c7pn+PCRpTHWTeOK46xauy8fzFav3sv69b20tm6nvv6l1NW9lOrqG/D59FdPKaWUmgn9xlwEcibHbc/fxi333cKRoSMAXNx6Mbe85BbedNab8Pv80+6jv9+O+//JT+BnP4OBgcJjtbVw0UX2RgDLQwOsOPwILTvvINa2j8BlXfhf2kXN8l78fnv3gFColfr6l1FXdyV1dR8jFjsHkVO7t6ZSSim11GlYW+B+cfgXfOieD/H4cXtf1POWncdnrvoMV2+6esorIY2B3bttOPvpT+HBB+2Nvl2bN8Mb3mDLS7aPEfrx90l/+5/pk0foeTX0fwyM585JNTU7aG5+Ky0tbyYa3aJXYSqllFKzRMPaArW3dy8fufcj/HjvjwForW7lk6/8JNdtu27SnrShITtjxp132l60o54rLwMBeOUrbTh7/ethy2YDDzxA8rav0PuVH9F7SZKBPwUTLDyntvYKWlreSkvLW4hE1s7l21VKKaWWLA1rC0z3WDd/df9f8c9P/DNZkyUWjPGRKz7CBy77wIRbQeVy8NRThXD20EPFvWctLfC619lw9prX2Fsi5Y4cZPjHt3LwX/6Lvq2DjL3Ns0Mj1NdcScuKd9Dc/GbC4bYz86aVUkqpJUzD2gIxnh7ni498kU898ClGUiP4xMeNF93IX73ir1hRvSK/XXc33HNPIaD19BT24ffbuzVdfbUtF14IPh+kho7Qf9+nONb5Q/pXnyBzAXCBfY4vG6QhdiVNa95Jc/ObCIVmb7oPpZRSSk1Pw9oC8HTX07z5tjfnp+F4/ebX8+mrPs25y85lbMwGs3vvtSHt2WeLn7t6dSGcvepVtvcMID5+gKMPfpq+Ez9mpLEHGrEFiA7W0FRzFY3bbqSu4eX4/RGUUkopNT80rFW42/fezrv++12MpcfYtnwbf3fV56nrfyU/+he46V57ajOVKmwficBLX1oIaGedBe5Y/2x2nK6uH9B14EsMZuwFCTSDpKD+UB1N9a+h8eUfoWr5xWf+jSqllFKqLA1rFcoYw+ce/hx/es+fYrJ+Lh/4Z1qevp533OJncLCwnQjs2AGvfjVcdRVcfrkNbN79DA8/QVfX1znR9Z9kc8MA+BLQ8nD84vL4AAAgAElEQVSQltrfpOHqW/C/Rm/bpJRSSlUiDWsVKJVN8d6fvpevP/EN2PlbND7yFR461ph/fONGG8xe/Wp4xSugsXHiPtLpPk6c+DadnV9nbKxwbrRmN7Te7WPZ5j8k8NG/Lv9kpZRSSlUMDWsVpj/ez1tueyu/uC+E3PckpvNC+oEtW+D974fXvhbWr5/8+cPDj9Le/nl6e3+IMfb8aGBYWHGXYcXPoHrH2+Gzt9rEp5RSSqmKp2Gtguzr28dVt36M9v/+v3D4lRigrQ0+/nH4vd+zc6FNJpXq5uDBj9LV9a92hREadsVo/f4YzQ8ZfDsug//4rD1PqpRSSqkFY07DmohcDXwJ8ANfM8Z8quTxNcC3gHpnm48aY+5wHrsFuB7IAv+fMeauuTzW+fbNex/hxvd3k37+vwCorcvxsVt83HwzVFVN/rxcLsPx41/h0KG/JJsdQgiy6sFWVv79USLdY7BhA/znp+Gtby1caaCUUkqpBWPOwpqI+IGvAK8GOoDHROR2Y8wuz2Z/DnzPGPOPInIOcAewzmm/EzgXaAPuFZEtxpgsi8yxY/DOm/bwwO07IBfAF0zyx38s/N+PhWhomPq5AwP3c+DAzYyNPQ9AY2Ibm27aQ9ULR6GhAb7wF/BHfwThmd3AXSmllFKVZy571i4BDhhjDgKIyHeBawBvWDNArdOuA4477WuA7xpjksAhETng7O/hOTzeM8oY+Icv5/jAh7JkU2eBZLjgN37NT776Ilavmvqm54lEBy+88CF6em4DIBLZwKZfXkDTR36EAFx3HXzhC0yb9pRSSilV8eYyrK0E2j3LHcClJdt8HLhbRG4GYsBVnuc+UvLclaUvICI3AjcCrFmzZlYO+kwYG4P/83/g29/2AT7knP/mk580fOzNb5vyeblckvb2z3PkyCfJ5cbx+aKsafkTVv/Jw/jv/pEd1PbFL8J736unPJVSSqlFYi7DWrm0YEqWrwW+aYz5nIhcBvy7iJw3w+dijPkq8FWAHTt2THi8Eu3fD295Czz/PBAcQ675A+7+9O9z1Yarpnxef/9d7N9/E/H4AQCam9/KptTvE3nd++DwYVi2DL7/fTsjrlJKKaUWjbkMax3Aas/yKgqnOV3XA1cDGGMeFpEI0DzD5y44P/4x/O7vwvAw+FsOkH37NXzit66dNqh1dn6DvXuvB6Cq6iw2bfoHGu/qgevfBvE4vOhF8N//be8tpZRSSqlFZerBUafnMWCziKwXkRD2goHbS7Y5CrwKQETOBiJAj7PdO0UkLCLrgc3Ao3N4rHMqk4FbboE3vckGtRUveojs9Rdx5Y4mbnnJLVM+t6vrP9i79wYA1q79C3Zc+ASNf3snvOtdNqi95z3wy19qUFNKKaUWqTnrWTPGZETkJuAu7LQc3zDG7BSRTwCPG2NuBz4I/IuIvB97mvM9xhgD7BSR72EvRsgA71uoV4L29MC118J994HPB29870P8qOkK6qP1/Mdb/gO/zz/pc7u7b2PPnusAw/r1f8va6hvhdW+0O9PxaUoppdSSIDYbLXw7duwwjz/++HwfRpFf/xre9jbo6LBDym79x0O8d9fZJLNJvv/27/O2cya/oKCn5wfs3PkOIMu6dR9n3eA18OY36/g0pZRSahEQkSeMMTO6MfdcngZdsoyBf/onuPJKG9Quuwwe/HWcz3f9Jslskhu23zBlUOvtvZ1du34LyLJmzZ+xdvAaeMlLbFB70Yvg8cc1qCmllFJLhIa1OfDhD9u5aNNpuPlmuP9++OKuD7OzZydbmrbwxau/OOlz+/ruYOfOt2FMhtWrP8z62j9B3vxmO9/H29+u49OUUkqpJUbvDTrLnn0WPv95CAbhm9+01wH8z97/4SuPfYWgL8h33vodYqFY2ef299/D88+/BWPSrFz5x2xY8zfI615ne9R27IB/+zeIRM7o+1FKKaXU/NKwNsv+7M/sadA/+iMb1I6PHOf3fvx7ANz6qlu5qPWiss8bGPg5zz//RoxJ0tb2R2za9AXkwx+Ge++1Y9R+8AMNakoppdQSpKdBZ9GDD8JPfgKxmA1tOZPjuh9dR1+8j9dsfA3vv+z9ZZ83OPgrnnvuDeRyCVpbb/j/27vz8Kqqe//j728SSMIohymWiKSAA0MSYiiitNWrBC0qVkSJLVTqT8CiVxC01qFSvPR6Ldf0UlT0gspPe4OCYrH1gqYF4qPCj1SDVCZRIk0pEiCMESHJ+v2xT8LJRE6GkxySz+t5znP2XnvtvdfOfki+rJH+/RdgmZnwn//pjfpcvlxNnyIiIq2UgrVG4pw3lxrAjBleZdh/fvCfZH2RRfd23Vly4xIirOqP+/DhD9m8+QeUlhbRs+dPuOCC57DcTXCHNwku//Vf3kgFERERaZUUrDWSVavgvffA54NZsyBnTw4P/eUhAF4c8yJxHeKqnHPs2Cd88sk1lJQco0eP27joosXY/gPe7LknTsBPf+q1p4qIiEirpWCtEZSWwkNeXMYvfgGRsce47fXbKC4t5p7v3MPoC0ZXc84ptm37CSUlR+je/WYuumgJVuLg1lth924YNgyefloT3oqIiLRyGmDQCF57DXJzoVcvmDYNZr77AJ8d/IzBPQbz5Mgnqz0nPz+DY8dyiYnpw0UXvURERBTMnAFr1kBcnLfWpwYUiIiItHqqWWugU6fg0Ue97cceg5gYx2ufvgbAkhuXEBNVNeD6+uvPyct7DIALLlhIZGR7ePllb/moNm28AQW9ejXZM4iIiEj4UrDWQC+8ADt3wgUXwKRJkH8knwNfH8AX6yM5LrlKfucc27dPobT0BD16/Aifb5S3IsGdd3oZfvc7uPzyJn4KERERCVcK1hqgqAjmzPG2H3/cm2Ujd28uAEPihmDV9Df76quXOXToz0RFdaVfvwzYt89b8/Obb2DyZJgypSkfQURERMJcrcGamd1tZl2aojBnmwULYM8eSEnxFmyH08FadbVqJ08WsHPnfQD06/cUbe0cbwmpsgVE589vsrKLiIjI2SGYmrU4YKOZvWZm11h11UWt0KFD8MQT3vavfw0R/p/kx3s/BqoP1nbunEFx8QG6dBlJz54TvDnUsrPh3HO9AQXR0U1VfBERETlL1BqsOeceAfoDi4Hbgc/M7Ndm1jfEZQtrv/kNFBbCFVdAWtrp9MBm0EAHD65m377fExERywUXLMRKS73+aQDPPecFbCIiIiKVBNVnzTnngL3+TzHQBVhuZtXPS9HC7d3rDdwE+Pd/Pz0V2qETh9h1aBfRkdFc2O3C8vwlJcfZsWMqAH36/IrY2G9761Lt3g19+8LoqvOwiYiIiEAQ86yZ2b8CPwH2A4uA+51zp8wsAvgMeCC0RQw/c+d6gwtuuAEuvfR0+qa9mwAY3HMwURGnf7S7dv2SEyfy6NBhCPHx/vVBFyzwvqdNO92GKiIiIlJJMJPidgNucs59GZjonCs1s+tCU6zwtWuX12pp5gVtgaprAj1yJIf8/N8CEVx44X97k99u2wZZWdCuHdx+e9MVXkRERM46wVTpvA0cLNsxs45mNgzAObc1VAULV4895k2E++Mfw6BBFY9VHlxQWnqKHTvuBEqJj59Bx46XeBmfecb7/vGPoYsG2oqIiEjNggnWngWOBewf96e1Ops3wyuveIsM/OpXVY9XnrYjcEmphAT/CUePwksvedvTpjVBqUVERORsFkywZv4BBoDX/EkrXVP0kUfAOW/e2oSEisdOlpxkS8EWDCOxZ2L1S0qBt6zU0aPw3e9CYmITP4GIiIicbYIJ1r4ws381szb+z73AF8Fc3D8v23Yz22lmD1ZzPMPMcv2fHWZ2KOBYScCxlcE/Umh88AGsXAnt23tBW2Wf7vuUU6Wn6N+1P+3btK+6pBR4kV7ZwIK77266wouIiMhZK5gasqnAfOARwAF/BibXdpKZRQJPAyOBfLyJdVc657aU5XHOzQjIfw8QODnZ1865qjPLNpO4OBg/Hvr3h549qx4PbAL96qv/W3FJqTJr18LWrd6caj/8YdMUXERERM5qtQZrzrl9wPh6XPs7wE7n3BcAZrYUGANsqSF/OvBYPe7TJL79bcjM9CrHqlMerPVMJi/P65/Wr99TtG3b/XSmslq1qVO9jm8iIiIitQhmnrUY4A5gIBBTlu6c+2ktp/YC/h6wnw8Mq+Ee5wMJwF8CkmPMLAdvEt4nnHNvVnPeZPy1fL17967tURpFTYttlY0EvbiLjxP7d9GmTXd69vzx6Qy7d8Obb3qrvd95ZxOUVERERFqCYPqsvYy3PugoYB0QDxwN4rzqwpoa6qUYDyx3zpUEpPV2zqUCtwG/rW55K+fc8865VOdcavfu3SsfbjKlrrS8Zi0+yotPfb4f4M0b7Pfcc1Ba6q34rqWlREREJEjBBGv9nHOPAsedc0uA0cDgIM7LB84L2I8H9tSQdzyQGZjgnNvj//4CWEvF/mxhJe9QHkdPHiWuQxyRX2cD0LVrwHzB33wD//3f3rYGFoiIiEgdBBOsnfJ/HzKzQUBnoE8Q520E+ptZgpm1xQvIqozqNLML8dYa/TAgrYuZRfu3uwGXU3Nft2b38T+9JtCkHgM5fPh9zKLw+UaezrBsGRQUQHIyXHZZM5VSREREzkbBjAZ93sy64I0GXQl0AB6t7STnXLGZ3Q2sBiKBF5xzn5rZHCDHOVcWuKUDSwPncgMuBp4zs1K8gPKJwFGk4aasCfSCzu2BUjp3voKoqM6nMwRO11FTpzcRERGRapwxWPMv1n7EOVcIZAPfrsvFnXNv4y1XFZj2y0r7s6s57wOCa2oNC7lfecFan+gDAHTtOvr0wY0bYcMGb1mp9PTmKJ6IiIicxc7YDOpfrUCdrGpR1gwaZ38DKvVXe/pp7/unP/UWbhcRERGpg2D6rL1rZrPM7Dwz85V9Ql6ys0TB8QL+cfQftG8TS1ybw8TG9qNduwu8g/v3w9KlXtPnXXc1b0FFRETkrBRMn7Wy+dQCVx131LFJtKXa9NUmAC48x0eE/aNirdrixd5I0NGjoW+VmUdEREREahXMCgYJteVpzcqaQL8d+zUAPp+/v1pJCTzzjLc9bVp1p4qIiIjUKpgVDCZWl+6c+7+NX5yzT9nggvNjDhIZ2YFzzvmed+CPf/RWLejbF0aNasYSioiIyNksmGbQoQHbMcBVwEeAgjVOT9vRvwN06ZJGRERb70DZdB3TpkFEMF0DRURERKoKphn0nsB9M+uMtwRVq1d0qoht+7cRaUafdu50f7Vt2yAryxv9efvtzVpGERERObvVp8qnCOjf2AU5G/1t398odaX0bgfRkeDzXesdKOur9uMfe/OriYiIiNRTMH3W3uL0AuwRwADgtVAW6mxR1gTar72jY8ehREfHeQf+93+97zvuaKaSiYiISEsRTJ+1eQHbxcCXzrn8EJXnrFI2ErRfh4BVC5yDPf716i+8sJlKJiIiIi1FMMHabuCfzrkTAGYWa2Z9nHN5IS3ZWaCsZq1vh4BVC44ehaIiiI2FTp2asXQiIiLSEgTTZ20ZUBqwX+JPa9VKSkv4xD9tx4AuPenQYYh34J//9L7PPVeLtouIiEiDBROsRTnnTpbt+Lfbhq5IZ4fPDn5GUfEJekRDQtx1eGveUzFYExEREWmgYIK1AjO7oWzHzMYA+0NXpLND+eCCDpUWblewJiIiIo0omD5rU4Hfm5l/llfygWpXNWhN/vqPDwDo3yGCLl2uPn1g717vW8GaiIiINIJgJsX9HLjUzDoA5pw7Gvpihb+N+WsASOo5mKioDqcPqGZNREREGlGtzaBm9mszO8c5d8w5d9TMupjZvzVF4cKVc47NBZ8BcGmfMRUPKlgTERGRRhRMn7VrnXOHynacc4XAD0JXpPC358jfOfjNN7SPhKTzJ1Q8qGBNREREGlEwwVqkmUWX7ZhZLBB9hvwt3vtf/A8AF3aOpV27fhUPKlgTERGRRhTMAINXgD+b2Yv+/UnAktAVKfxt2P0nABJ7DKh6UMGaiIiINKJgBhg8aWafAFcDBqwCzg91wcLZpr2bABh63tUVD5w4AYWFEBUFXbs2Q8lERESkpQmmGRRgL94qBmOBq4CtwZxkZteY2XYz22lmD1ZzPMPMcv2fHWZ2KODYT8zsM//nJ0GWM+S+/vpzth32BsQO73NzxYNl03bExUFEsD9aERERkZrVWLNmZhcA44F04ADwKt7UHVcGc2EziwSeBkbizc220cxWOue2lOVxzs0IyH8PMMS/7QMeA1IBB/zVf25h3R6v8X35z9f5x9fQJiKCgT0TKx4sawKNi2v6gomIiEiLdKbqn214tWjXO+dGOOd+h7cuaLC+A+x0zn3hX6JqKTDmDPnTgUz/9ijgXefcQX+A9i5wTR3uHTLr85YDcKHvPNpGVlp1S/3VREREpJGdKVgbi9f8ucbM/tvMrsLrsxasXsDfA/bz/WlVmNn5QALwl7qca2aTzSzHzHIKCgrqULT6KS4+xsf//AiAS751WdUMCtZERESkkdUYrDnnVjjnbgUuAtYCM4CeZvasmaUFce3qAjtXQ97xwHLnXFnNXVDnOueed86lOudSu3fvHkSRGqawMIudx7wiXvKt4VUzKFgTERGRRlZrL3jn3HHn3O+dc9cB8UAuUGWwQDXygfMC9uOBPTXkHc/pJtC6nttkDhz4IzuPedvJcclVMyhYExERkUZWpyGL/j5kzznn/iWI7BuB/maWYGZt8QKylZUzmdmFQBfgw4Dk1UCaf2mrLkCaP63ZOFfKvv1/Ytdxbz8pLqlqJgVrIiIi0siCmRS3XpxzxWZ2N16QFQm84Jz71MzmADnOubLALR1Y6pxzAeceNLPH8QI+gDnOuYOhKmswjh37mJ2H9nLKQd8ufekU3alqJgVrIiIi0shCFqwBOOfeBt6ulPbLSvuzazj3BeCFkBWujiIjO3Og7Ujg3eqbQEHBmoiIiDQ6zdwapHbt+vFPNwioob9acTHs2wdm0LNnE5dOREREWioFa3WQuzcXgCFxQ6oeLCgA56BbN2jTpolLJiIiIi2VgrUgOefKgzWNBBUREZGmomAtSLsP76bwRCHd2nXjWx2/VTWDgjUREREJAQVrQco/ko8v1seQuCGYVTNnr4I1ERERCYGQjgZtSS7vfTn779/P8VPHq8+gYE1ERERCQDVrdWBmdGjbofqDCtZEREQkBBSsNRYFayIiIhICCtYai4I1ERERCQEFa41FwZqIiIiEgIK1xuAc7N3rbStYExERkUakYK0xHDwIJ09C584QG9vcpREREZEWRMFaY1ATqIiIiISIgrXGUBasxcU1bzlERESkxVGw1hhUsyYiIiIhomCtMShYExERkRBRsNYYFKyJiIhIiChYawwK1kRERCREFKw1BgVrIiIiEiIK1hqDJsQVERGREFGw1hhUsyYiIiIhEtJgzcyuMbPtZrbTzB6sIc8tZrbFzD41s/8JSC8xs1z/Z2Uoy9kgx455n5gYbwUDERERkUYUFaoLm1kk8DQwEsgHNprZSufcloA8/YFfAJc75wrNrEfAJb52ziWHqnyNJrBWzax5yyIiIiItTihr1r4D7HTOfeGcOwksBcZUynMn8LRzrhDAObcvhOUJDTWBioiISAiFrGYN6AX8PWA/HxhWKc8FAGb2PhAJzHbOrfIfizGzHKAYeMI592blG5jZZGAyQO/evRu39MFSsCYiIo3k1KlT5Ofnc+LEieYuijSSmJgY4uPjadOmTb2vEcpgrbo2QVfN/fsDVwDxwHtmNsg5dwjo7ZzbY2bfBv5iZpudc59XuJhzzwPPA6Smpla+dtNQsCYiIo0kPz+fjh070qdPH0xda856zjkOHDhAfn4+CQkJ9b5OKJtB84HzAvbjgT3V5PmDc+6Uc24XsB0veMM5t8f//QWwFhgSwrLWn4I1ERFpJCdOnKBr164K1FoIM6Nr164NrikNZbC2EehvZglm1hYYD1Qe1fkmcCWAmXXDaxb9wsy6mFl0QPrlwBbCkYI1ERFpRArUWpbGeJ8hawZ1zhWb2d3Aarz+aC845z41szlAjnNupf9YmpltAUqA+51zB8zsMuA5MyvFCyifCBxFGlYUrImIiEgIhbLPGs65t4G3K6X9MmDbAff5P4F5PgAGh7JsjUbBmoiIiISQVjBoKAVrIiLSgpgZM2fOLN+fN28es2fPPuM5a9eu5YMPPjhjnuzsbFJSUoiKimL58uU15tu8eTPJyckkJyfj8/lISEggOTmZq6++uk7PMWrUKI4ePVqnc8qcPHkSn8/Ho48+Wq/zG5uCtYb45hs4eBCioqBbt+YujYiISINFR0fzxhtvsH///qDPCSZY6927Ny+99BK33XbbGfMNHjyY3NxccnNzueGGG/jNb35Dbm4uWVlZFfIVFxef8TqrV6+mY8eOwT1AJatWrWLAgAG8+uqr9Tq/sSlYa4iyBdx79oQI/ShFRKQRmYXmU4uoqCgmT55MRkZGlWMFBQWMHTuWoUOHMnToUN5//33y8vJYuHAhGRkZJCcn895771V73T59+pCYmEhEA/5eZmVlcfXVVzN+/HiGDPEmibj++uu55JJLGDhwIIsWLSrPGx8fz6FDh9i5cyeDBg3ijjvuYODAgVx77bW1js7MzMzkvvvuo2fPnmzcuLE8fcOGDQwfPpykpCSGDRtGUVERxcXFzJgxg0GDBpGYmMgzzzxT7+erSUj7rLV4agIVEZEWaNq0aSQmJvLAAw9USL/33nuZMWMGI0aMYPfu3YwaNYqtW7cydepUOnTowKxZs0JetvXr17Nly5byyfCXLFmCz+ejqKiI1NRUxo4dS5cuXSqcs337djIzMxk8eDA33XQTb775JuPHj6/2+sePH2fdunW8+OKL7N27l8zMTIYOHcqJEycYP348r7/+OikpKRw+fJjo6GieeeYZ9uzZw6ZNm4iMjOTgwYON/swK1hqiLFiLi2vecoiISMvjmmeud4BOnToxceJE5s+fT2xsbHl6VlYWW7acnpzhyJEj9e4XVl/Dhw+vsGpRRkYGK1d6M4Pl5+fz+eefk5qaWuGcfv36MXiwN27xkksuIS8vr8brr1y5kpEjRxITE8O4ceNITU1l3rx5bN26ld69e5OSkgJA586dAe9nMn36dCIjIwHw+XyN9qxlFKw1hGrWRESkhZo+fTopKSlMmjSpPK20tJQPP/ywQgDX1Nq3b1++nZWVRXZ2NuvXryc2NpYRI0ZU28QZHR1dvh0ZGXnG/m6ZmZls2LCBPn36ALBv3z6ys7Pp1KlTtXOmOedCPjeeOlo1hII1ERFpoXw+H7fccguLFy8uT0tLS2PBggXl+7m5uQB07NixyWvYAA4fPozP5yM2NpZPP/20Qv+y+igsLGTDhg3k5+eTl5dHXl4e8+fPJzMzk4EDB/Lll1/y0UcfAV6tYklJCWlpaTz77LOUlJQAhKQZVMFaQ5QNMFCwJiIiLdDMmTMrjAqdP38+OTk5JCYmMmDAABYuXAh4nfxXrFhxxgEGGzduJD4+nmXLljFlyhQGDhzY4PKNHj2aoqIikpKSmDNnDsOGDWvQ9V5//XVGjhxZYdH1G2+8kRUrVhAREUFmZiZ33XUXSUlJpKWl8c033zBlyhTi4uJITEwkKSmJ1157raGPVYW5ZmwTb0ypqakuJyenaW963XXwpz/Bm2/CmDFNe28REWlxtm7dysUXX9zcxZBGVt17NbO/OudSazilAtWsNYSaQUVERCTENMCgIRSsiYiIVDB37lyWLVtWIW3cuHE8/PDDVfJu3ryZCRMmVEiLjo5mw4YNIS0jwNSpU1m/fn2FtPvuu4+JEyeG/N51pWbQ+iopgbZtobTUW8mgbdumu7eIiLRIagZtmdQM2lwKCrxArVs3BWoiIiISMgrW6ktNoCIiItIEFKzVl4I1ERERaQIK1upLwZqIiIg0AQVr9aVgTUREWiAzY+bMmeX78+bNY/bs2Wc8Z+3atXzwwQdnzJOdnU1KSgpRUVEsX778jHkTEhLYvn17hbTp06fz5JNP1nhOXl4egwYNOuN1wVuMvlevXpSWltaaN1woWKsvBWsiItICRUdH88Ybb1RYuaA2wQRrvXv35qWXXuK2226r9Xrjx49n6dKl5fulpaUsX76cW2+9NegyVae0tJQVK1Zw3nnnkZ2d3aBrNSXNs1ZfCtZERCSE1q4NzeLgV1xx5im7oqKimDx5MhkZGcydO7fCsYKCAqZOncru3bsB+O1vf0uvXr1YuHAhkZGRvPLKK/zud7/ju9/9bpXrli2MHhFRez1Reno6t956K4899hjg1cr16dOH888/n7y8PCZMmMDx48cBWLBgAZdddlmt1wRYs2YNgwYN4tZbbyUzM5MrrrgCgGPHjnHPPfeQk5ODmfHYY48xduxYVq1axUMPPURJSQndunXjz3/+c1D3aWwK1upLwZqIiLRQ06ZNIzExkQceeKBC+r333suMGTMYMWIEu3fvZtSoUWzdupWpU6fSoUMHZs2a1Sj3T0xMJCIigk2bNpGUlMTSpUtJT08HoEePHrz77rvExMTw2WefkZ6eTrDzrGZmZpKens6YMWN46KGHOHXqFG3atOHxxx+nc+fObN68GfAWdC8oKODOO+8kOzubhISEkCzQHiwFa/WlYE1EREKothqwUOrUqRMTJ05k/vz5xMbGlqdnZWWxZcuW8v0jR45w9OjRkJQhPT2dpUuXMnDgQP7whz8wZ84cAE6dOsXdd99Nbm4ukZGR7NixI6jrnTx5krfffpuMjAw6duzIsGHDeOeddxg9ejRZWVkVml27dOnCW2+9xfe+9z0SEhIA8Pl8jf+QQQppsGZm1wD/BUQCi5xzT1ST5xZgNuCATc652/zpPwEe8Wf7N+fcklCWtU6cU7AmIiIt2vTp00lJSWHSpEnlaaWlpXz44YcVArhQSU9PJy0tje9///skJl5wdesAABXLSURBVCbSo0cPADIyMujZsyebNm2itLSUmJiYoK63atUqDh8+zODBgwEoKiqiXbt2jB49GuccZhWbnatLay4hG2BgZpHA08C1wAAg3cwGVMrTH/gFcLlzbiAw3Z/uAx4DhgHfAR4zsy6hKmudFRbCyZPQqRO0a9fcpREREWl0Pp+PW265hcWLF5enpaWlsWDBgvL93NxcADp27NjoNWx9+/ala9euPPjgg+VNoACHDx/m3HPPJSIigpdffpmSkpKgrpeZmcmiRYvIy8sjLy+PXbt28c4771BUVFTluQoLCxk+fDjr1q1j165dAM3aDBrK0aDfAXY6575wzp0ElgJjKuW5E3jaOVcI4Jzb508fBbzrnDvoP/YucE0Iy1o3qlUTEZFWYObMmRVGhc6fP5+cnBwSExMZMGAACxcuBOD6669nxYoVJCcn895771V7rY0bNxIfH8+yZcuYMmUKAwcOrPX+6enpbNu2jR/+8IflaT/72c9YsmQJl156KTt27KB9+/a1XqeoqIjVq1czevTo8rT27dszYsQI3nrrLR555BEKCwsZNGgQSUlJrFmzhu7du/P8889z0003kZSU1OCRqA0RsoXczexm4Brn3P/x708Ahjnn7g7I8yawA7gcr6l0tnNulZnNAmKcc//mz/co8LVzbl6le0wGJgP07t37ki+//DIkz1JFVhaMHAlXXAFr1jTNPUVEpMXTQu4tUzgv5F5dQ2/lyDAK6A9cAaQDi8zsnCDPxTn3vHMu1TmX2r179wYWtw727vW+4+Ka7p4iIiLSKoVygEE+cF7Afjywp5o8651zp4BdZrYdL3jLxwvgAs9dG7KS1pWaQUVERKo1d+5cli1bViFt3LhxPPzww1Xybt68mQkTJlRIi46OZsOGDfW+/+rVq/n5z39eIS0hIYEVK1bU+5rNLZTB2kagv5klAP8AxgOVpy1+E69G7SUz6wZcAHwBfA78OmBQQRreQITwoGBNRESkWg8//HC1gVl1Bg8eXD5IobGMGjWKUaNGNeo1m1vIgjXnXLGZ3Q2sxuuP9oJz7lMzmwPkOOdW+o+lmdkWoAS43zl3AMDMHscL+ADmOOeabxhGZQrWREREpImEdJ4159zbwNuV0n4ZsO2A+/yfyue+ALwQyvLVm4I1ERERaSJayL0+FKyJiIhIE1GwVh8K1kRERKSJKFirq+PH4ehRiI6Gc85p7tKIiIg0KjNj5syZ5fvz5s1j9uzZZzxn7dq1fPDBB2fM89RTTzFgwAASExO56qqrqGlu1M2bN5OcnExycjI+n4+EhASSk5O5+uqr6/Qco0aNqveqCidPnsTn8/Hoo4/W6/zGpoXc6yqwVi1M1gwTEZGWx34Vmr8x7rEzT4YfHR3NG2+8wS9+8Qu6desW1DXXrl1Lhw4duOyyy2rMM2TIEHJycmjXrh3PPvssDzzwAK+++mqVfIEjRG+//Xauu+46br755ir5iouLiYqqOYxZvXp1UGWvzqpVqxgwYACvvvoqjz/+eL2v01hUs1ZXagIVEZEWLCoqismTJ5ORkVHlWEFBAWPHjmXo0KEMHTqU999/n7y8PBYuXEhGRsYZl5u68soraedfT/vSSy8lPz+/zmXLysri6quvZvz48QwZMgTwlrq65JJLGDhwIIsWLSrPGx8fz6FDh9i5cyeDBg3ijjvuYODAgVx77bWcOHHijPfJzMzkvvvuo2fPnmzcuLE8fcOGDQwfPpykpCSGDRtGUVERxcXFzJgxg0GDBpGYmMgzzzxT5+eqjWrW6krBmoiINIHaasBCadq0aSQmJvLAAw9USL/33nuZMWMGI0aMYPfu3YwaNYqtW7cydepUOnTowKxZs4K6/uLFi7n22mvrVbb169ezZcsWevfuDcCSJUvw+XwUFRWRmprK2LFj6dKlS4Vztm/fTmZmJoMHD+amm27izTffZPz48dVe//jx46xbt44XX3yRvXv3kpmZydChQzlx4gTjx4/n9ddfJyUlhcOHDxMdHc0zzzzDnj172LRpE5GRkSFZ8F3BWl0pWBMRkRauU6dOTJw4kfnz5xMbG1uenpWVxZYtW8r3jxw5Uud+Ya+88go5OTmsW7euXmUbPnx4eaAGkJGRwcqVKwHIz8/n888/JzW14pKb/fr1Y/DgwQBccskl5OXl1Xj9lStXMnLkSGJiYhg3bhypqanMmzePrVu30rt3b1JSUgDo3Lkz4P1Mpk+fTmRkJAA+n69ez3UmCtbqSsGaiIi0AtOnTyclJYVJkyaVp5WWlvLhhx9WCODqIisri7lz57Ju3Tqio6PrdY327dtXuF52djbr168nNjaWESNGVNvEGXivyMhIiouLa7x+ZmYmGzZsoE+fPgDs27eP7OxsOnXqhFXTV905V216Y1KftbpSsCYiIq2Az+fjlltuYfHixeVpaWlpLFiwoHy/bCBAx44da61h+/jjj5kyZQorV66kR48ejVLGw4cP4/P5iI2N5dNPP63Qv6w+CgsL2bBhA/n5+eTl5ZGXl8f8+fPJzMxk4MCBfPnll3z00UeAV6tYUlJCWloazz77LCUlJQAhaQZVsFZXCtZERKSVmDlzJvv37y/fnz9/Pjk5OSQmJjJgwAAWLlwIeJ38V6xYccYBBvfffz/Hjh1j3LhxJCcnc8MNNzS4fKNHj6aoqIikpCTmzJnDsGHDGnS9119/nZEjR9KmTZvytBtvvJEVK1YQERFBZmYmd911F0lJSaSlpfHNN98wZcoU4uLiSExMJCkpiddee62hj1WFeSs+nf1SU1NdTk5O6G80eDD87W/w8ceQnBz6+4mISKuxdetWLr744uYuhjSy6t6rmf3VOZdawykVqGatrlSzJiIiIk1IAwzq4uRJOHAAIiOhe/fmLo2IiEjYmTt3LsuWLauQNm7cOB5++OEqeTdv3syECRMqpEVHR7Nhw4aQlhFg6tSprF+/vkLafffdx8SJE0N+77pSM2hd/P3v0Ls3fOtb8I9/hPZeIiLS6qgZtGVSM2hTUhOoiIiINDEFa3WhYE1ERESamIK1uigL1uLimrccIiIi0mooWKsL1ayJiIhIE1OwVhcK1kREpIUzM2bOnFm+P2/ePGbPnn3Gc9auXcsHH3xwxjxPPfUUAwYMIDExkauuuoovv/yyxrwJCQls3769Qtr06dN58sknazwnLy+PQYMGnbEM4C1G36tXL0pLS2vNGy4UrNWFgjUREWkiZqH51CY6Opo33nijwsoFtQkmWBsyZAg5OTl88skn3HzzzTzwwAM15h0/fjxLly4t3y8tLWX58uXceuutQZepOqWlpaxYsYLzzjuP7OzsBl2rKSlYqwsFayIi0sJFRUUxefJkMjIyqhwrKChg7NixDB06lKFDh/L++++Tl5fHwoULycjIOONyU1deeSXt2rUD4NJLLyU/P7/GMqSnp1cI1rKzs+nTpw/nn38+eXl5fPe73yUlJYWUlJRag8RAa9asYdCgQdx1111kZmaWpx87doxJkyYxePBgEhMTef311wFYtWoVKSkpJCUlcdVVVwV9n8YW0klxzewa4L+ASGCRc+6JSsdvB34DlE1atsA5t8h/rATY7E/f7Zxr+CJiDaVgTUREmkhzToM6bdo0EhMTq9R+3XvvvcyYMYMRI0awe/duRo0axdatW5k6dSodOnRg1qxZQV1/8eLFXHvttTUeT0xMJCIigk2bNpGUlMTSpUtJT08HoEePHrz77rvExMTw2WefkZ6eTrDzrGZmZpKens6YMWN46KGHOHXqFG3atOHxxx+nc+fObN7shR2FhYUUFBRw5513kp2dTUJCQkgWaA9WyII1M4sEngZGAvnARjNb6ZzbUinrq865u6u5xNfOufBZfLOkBL76ytvWaFAREWnBOnXqxMSJE5k/fz6xsbHl6VlZWWzZcvrP+JEjRzh69Gidrv3KK6+Qk5PDunXrzpivrHZt4MCB/OEPf2DOnDkAnDp1irvvvpvc3FwiIyPZsWNHUPc9efIkb7/9NhkZGXTs2JFhw4bxzjvvMHr0aLKysirU5HXp0oW33nqL733veyQkJADg8/nq9JyNKZQ1a98BdjrnvgAws6XAGKBysHZ22L/fC9i6doW2bZu7NCIiIiE1ffp0UlJSmDRpUnlaaWkpH374YYUAri6ysrKYO3cu69atIzo6+ox509PTSUtL4/vf/z6JiYn06NEDgIyMDHr27MmmTZsoLS0lJiYmqHuvWrWKw4cPM3jwYACKiopo164do0ePxjmHVerQV11acwlln7VewN8D9vP9aZWNNbNPzGy5mZ0XkB5jZjlmtt7MbqzuBmY22Z8np6CgoBGLXg01gYqISCvi8/m45ZZbWLx4cXlaWloaCxYsKN/Pzc0FoGPHjrXWsH388cdMmTKFlStXlgdeZ9K3b1+6du3Kgw8+WN4ECnD48GHOPfdcIiIiePnllykpKQnqeTIzM1m0aBF5eXnk5eWxa9cu3nnnHYqKiqo8V2FhIcOHD2fdunXs2rULoFmbQUMZrFUXjlZugX8L6OOcSwSygCUBx3r718y6DfitmfWtcjHnnnfOpTrnUruHemH1c86BWbPgRz8K7X1ERETCxMyZMyuMCp0/fz45OTkkJiYyYMAAFi5cCMD111/PihUrzjjA4P777+fYsWOMGzeO5ORkbrih9q7o6enpbNu2jR/+8IflaT/72c9YsmQJl156KTt27KB9+/a1XqeoqIjVq1czevTo8rT27dszYsQI3nrrLR555BEKCwsZNGgQSUlJrFmzhu7du/P8889z0003kZSU1OCRqA0RsoXczWw4MNs5N8q//wsA59y/15A/EjjonOtczbGXgD8655bXdL8mWchdREQkhLSQe8sUzgu5bwT6m1mCmbUFxgMrAzOYWWCb4g3AVn96FzOL9m93Ay7nbO3rJiIiItIAIRtg4JwrNrO7gdV4U3e84Jz71MzmADnOuZXAv5rZDUAxcBC43X/6xcBzZlaKF1A+Uc0oUhEREQkzc+fOZdmyZRXSxo0bx8MPP1wl7+bNm5kwYUKFtOjoaDZs2FDv+69evZqf//znFdISEhJYsWJFva/Z3ELWDNrU1AwqIiJnu61bt3LRRReFzShEaTjnHNu2bQvbZlARERGpg5iYGA4cOEBLqUhp7ZxzHDhwIOjpRWoS0hUMREREJHjx8fHk5+cT8umopMnExMQQHx/foGsoWBMREQkTbdq0KZ8xX6SMmkFFREREwpiCNREREZEwpmBNREREJIy1mKk7zKwA+LKep3cD9teaS5qL3k940/sJX3o34U3vJ3w1xbs53zkX1FqZLSZYawgzywl2rhNpeno/4U3vJ3zp3YQ3vZ/wFW7vRs2gIiIiImFMwZqIiIhIGFOw5nm+uQsgZ6T3E970fsKX3k140/sJX2H1btRnTURERCSMqWZNREREJIwpWBMREREJY60+WDOza8xsu5ntNLMHm7s8rZ2ZvWBm+8zsbwFpPjN718w+8393ac4ytlZmdp6ZrTGzrWb2qZnd60/X+wkDZhZjZv/PzDb538+v/OkJZrbB/35eNbO2zV3W1srMIs3sYzP7o39f7yZMmFmemW02s1wzy/Gnhc3vtlYdrJlZJPA0cC0wAEg3swHNW6pW7yXgmkppDwJ/ds71B/7s35emVwzMdM5dDFwKTPP/e9H7CQ/fAP/inEsCkoFrzOxS4D+ADP/7KQTuaMYytnb3AlsD9vVuwsuVzrnkgPnVwuZ3W6sO1oDvADudc184504CS4ExzVymVs05lw0crJQ8Blji314C3NikhRIAnHP/dM595N8+ivdHpxd6P2HBeY75d9v4Pw74F2C5P13vp5mYWTwwGljk3zf0bsJd2Pxua+3BWi/g7wH7+f40CS89nXP/BC9gAHo0c3laPTPrAwwBNqD3Ezb8zWy5wD7gXeBz4JBzrtifRb/jms9vgQeAUv9+V/RuwokD3jGzv5rZZH9a2Pxui2quG4cJqyZNc5mInIGZdQBeB6Y75454FQQSDpxzJUCymZ0DrAAuri5b05ZKzOw6YJ9z7q9mdkVZcjVZ9W6az+XOuT1m1gN418y2NXeBArX2mrV84LyA/XhgTzOVRWr2lZmdC+D/3tfM5Wm1zKwNXqD2e+fcG/5kvZ8w45w7BKzF61t4jpmV/cdcv+Oax+XADWaWh9fd5l/watr0bsKEc26P/3sf3n90vkMY/W5r7cHaRqC/f0ROW2A8sLKZyyRVrQR+4t/+CfCHZixLq+XvY7MY2OqceyrgkN5PGDCz7v4aNcwsFrgar1/hGuBmfza9n2bgnPuFcy7eOdcH7+/MX5xzP0LvJiyYWXsz61i2DaQBfyOMfre1+hUMzOwHeP/DiQRecM7NbeYitWpmlglcAXQDvgIeA94EXgN6A7uBcc65yoMQJMTMbATwHrCZ0/1uHsLrt6b308zMLBGvE3Qk3n/EX3POzTGzb+PV5viAj4EfO+e+ab6Stm7+ZtBZzrnr9G7Cg/89rPDvRgH/45yba2ZdCZPfba0+WBMREREJZ629GVREREQkrClYExEREQljCtZEREREwpiCNREREZEwpmBNREREJIwpWBORFsnM/t3MrjCzG82sSRZgNrM8M+vWFPcSkdZDwZqItFTD8OaA+z7e/HAiImclBWsi0qKY2W/M7BNgKPAh8H+AZ83sl2bW18xW+Rdrfs/MLvKf85KZLfSn7fCv5YiZxZjZi2a22cw+NrMr/emRZjbPn/6Jmd0TUIR7zOwj/7Gy63/fzHL9n4/LZksXEQlGa1/IXURaGOfc/Wa2DJgA3Aesdc5dDmBmfwamOuc+M7NhwDN46zQC9MGrhesLrDGzfsA0/zUH+wOvd8zsAmASkAAMcc4Vm5kvoAj7nXMpZvYzYBZesDgLmOace9/MOgAnQvkzEJGWRcGaiLREQ4Bc4CJgC4A/SLoMWOYtcwpAdMA5rznnSoHPzOwL/7kjgN8BOOe2mdmXwAV4624udM4V+48FLkFTtsD9X4Gb/NvvA0+Z2e+BN5xz+Y34rCLSwilYE5EWw8ySgZeAeGA/0M5Ltly8WrNDzrnkGk6vvPaeA6y6jP70mtbqK1vbsQT/71jn3BNm9ifgB8B6M7vaObet9icSEVGfNRFpQZxzuf5gbAcwAPgLMMo5l+ycOwzsMrNx4EVwZpYUcPo4M4sws77At4HtQDbwI3/+C/AWdN4OvANMNbMo/7HAZtAqzKyvc26zc+4/gBy8WjsRkaAoWBORFsXMugOF/ibNi5xzWwIO/wi4w8w2AZ8CYwKObQfWAf+L16/tBF6ftkgz2wy8CtzunPsGWATsBj7xX+u2Woo13cz+5s/7tf8eIiJBMedqqskXEWkdzOwl4I/OueXNXRYRkcpUsyYiIiISxlSzJiIiIhLGVLMmIiIiEsYUrImIiIiEMQVrIiIiImFMwZqIiIhIGFOwJiIiIhLG/j9HPFOwmBC7qQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "plt.figure(figsize=(10, 10))\n",
    "\n",
    "\n",
    "plt.subplot(2, 1, 1)\n",
    "l1,l2,l3,l4=plt.plot(range(1,51),[v[1] for v in train_1_loss],range(1,51),[v[1] for v in val_1_loss],\\\n",
    "              range(1,51),[v[1] for v in train_2_loss],range(1,51),[v[1] for v in val_2_loss])\n",
    "plt.setp(l1,linewidth=2,color='r')\n",
    "plt.setp(l2,linewidth=2,color='y')\n",
    "plt.setp(l3,linewidth=2,color='g')\n",
    "plt.setp(l4,linewidth=2,color='b')\n",
    "plt.legend([l1,l2,l3,l4],['Net_1_Train_Loss','Net_1_Val_Loss','Net_2_Train_Loss','Net_2_Val_Loss'])\n",
    "plt.xlabel('#epochs')\n",
    "plt.ylabel('Losses')\n",
    "plt.title('Cross-Entropy Losses of both')\n",
    "\n",
    "plt.subplot(2, 1, 2)\n",
    "l1,l2,l3,l4=plt.plot(range(1,51),[v[1] for v in train_1_acc],range(1,51),[v[1] for v in val_1_acc],\\\n",
    "              range(1,51),[v[1] for v in train_2_acc],range(1,51),[v[1] for v in val_2_acc])\n",
    "plt.setp(l1,linewidth=2,color='r')\n",
    "plt.setp(l2,linewidth=2,color='y')\n",
    "plt.setp(l3,linewidth=2,color='g')\n",
    "plt.setp(l4,linewidth=2,color='b')\n",
    "plt.legend([l1,l2,l3,l4],['Net_1_Train_Acc','Net_1_Val_Acc','Net_2_Train_Acc','Net_2_Val_Acc'])\n",
    "plt.xlabel('#epochs')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.title('Accuracies of both')\n",
    "\n",
    "\n",
    "plt.show()\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python [conda env:anaconda2]",
   "language": "python",
   "name": "conda-env-anaconda2-py"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.15"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}