ch07-keras-building-models.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "from tensorflow import keras\n",
    "from tensorflow.keras import layers\n",
    "\n",
    "import numpy as np"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**The Sequrntial Model**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-08-08 22:51:07.815530: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:975] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
      "2022-08-08 22:51:07.819907: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:975] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
      "2022-08-08 22:51:07.820135: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:975] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
      "2022-08-08 22:51:07.820510: I tensorflow/core/platform/cpu_feature_guard.cc:193] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations:  AVX2 FMA\n",
      "To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.\n",
      "2022-08-08 22:51:07.821291: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:975] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
      "2022-08-08 22:51:07.821484: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:975] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
      "2022-08-08 22:51:07.821622: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:975] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
      "2022-08-08 22:51:08.229218: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:975] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
      "2022-08-08 22:51:08.229381: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:975] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
      "2022-08-08 22:51:08.229509: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:975] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
      "2022-08-08 22:51:08.229618: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1532] Created device /job:localhost/replica:0/task:0/device:GPU:0 with 4630 MB memory:  -> device: 0, name: NVIDIA GeForce RTX 2060, pci bus id: 0000:01:00.0, compute capability: 7.5\n"
     ]
    }
   ],
   "source": [
    "model = keras.Sequential([\n",
    "    layers.Dense(64, activation=\"relu\"),\n",
    "    layers.Dense(10, activation=\"softmax\")\n",
    "])\n",
    "\n",
    "\n",
    "model = keras.Sequential()\n",
    "model.add(layers.Dense(64, activation=\"relu\"))\n",
    "model.add(layers.Dense(10, activation=\"softmax\"))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# layers are built only when they are called for the first time\n",
    "# because, shape of layer's weight deps on input dimension which is not know during model creation\n",
    "\n",
    "# model.weights"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<tf.Variable 'dense_2/kernel:0' shape=(3, 64) dtype=float32, numpy=\n",
       " array([[-0.01728177,  0.18202105,  0.13344258, -0.06259781,  0.0762296 ,\n",
       "         -0.18444303,  0.11669421,  0.24633586,  0.2709825 , -0.2503717 ,\n",
       "         -0.02201816, -0.17047454,  0.06104934, -0.03936216,  0.25353473,\n",
       "         -0.07832678,  0.28366154, -0.23965922, -0.26351303,  0.20566106,\n",
       "         -0.07706985, -0.01520407, -0.21959433,  0.25699502, -0.00264814,\n",
       "          0.16713771,  0.08944726,  0.20166856,  0.00428656, -0.06611858,\n",
       "         -0.21239217,  0.25794816,  0.02162185, -0.0906793 ,  0.28345   ,\n",
       "         -0.24014968, -0.06032097,  0.12174377, -0.03453949,  0.06549767,\n",
       "          0.08190158, -0.23211288,  0.1715526 ,  0.16433403,  0.0593597 ,\n",
       "          0.02893716,  0.21983176,  0.148424  ,  0.28943014,  0.25018656,\n",
       "          0.23577774, -0.27212653,  0.16660509, -0.18008825,  0.2888024 ,\n",
       "         -0.1104705 , -0.2769277 , -0.02426618,  0.1345532 , -0.02354407,\n",
       "         -0.01406929, -0.0274477 , -0.11243063, -0.081715  ],\n",
       "        [ 0.18554935, -0.10232605,  0.27554333, -0.09230095,  0.14036387,\n",
       "          0.05727893,  0.27932346, -0.19463208, -0.01726222,  0.15928963,\n",
       "          0.08175752, -0.10859178,  0.04167411, -0.08967587,  0.17081758,\n",
       "          0.1488809 ,  0.09570146,  0.08163929,  0.14844438, -0.00192794,\n",
       "          0.24401104,  0.06410831, -0.29324052,  0.0949775 ,  0.19874135,\n",
       "          0.20094627,  0.09880447, -0.09659086, -0.24344748,  0.14400187,\n",
       "          0.2887841 , -0.15637551,  0.13828674, -0.22599635,  0.19904733,\n",
       "         -0.11481278,  0.01062328,  0.00825113, -0.03778994, -0.15379602,\n",
       "         -0.11624187, -0.20866683, -0.16086462,  0.05972636, -0.0004485 ,\n",
       "         -0.04555178,  0.26044565,  0.23296535,  0.0268285 , -0.11006947,\n",
       "         -0.29140508, -0.2273871 , -0.064188  ,  0.23302639,  0.00573304,\n",
       "         -0.17944577, -0.09327592,  0.2865646 ,  0.26122707,  0.12537178,\n",
       "          0.14522535,  0.09372956, -0.09538931,  0.27848494],\n",
       "        [-0.08768186, -0.1836455 ,  0.00138429,  0.03780672, -0.2731816 ,\n",
       "         -0.18813804, -0.20773076,  0.18910551,  0.25391924, -0.05109839,\n",
       "          0.26405948,  0.02011341,  0.00731054,  0.07427403, -0.22963583,\n",
       "         -0.07034707,  0.12339517,  0.14353004,  0.21327561, -0.2609328 ,\n",
       "          0.13040105,  0.16759825,  0.08529949, -0.17013943, -0.28291005,\n",
       "         -0.27000263,  0.2373066 ,  0.00481039, -0.14361082, -0.12778175,\n",
       "          0.12790632,  0.01397225, -0.21260692,  0.25150698,  0.14258897,\n",
       "          0.05552655, -0.12577553,  0.09161773, -0.15425357,  0.2487253 ,\n",
       "          0.24438578,  0.08892649, -0.1387345 , -0.11002666, -0.04427865,\n",
       "          0.23984838,  0.10764372, -0.05750625,  0.14144436,  0.00545281,\n",
       "          0.15628427,  0.27115065, -0.19303581, -0.15417452,  0.12928498,\n",
       "         -0.12847638,  0.23103523, -0.05274637, -0.21798488, -0.09774156,\n",
       "          0.14591855, -0.255366  , -0.09803115, -0.02709177]],\n",
       "       dtype=float32)>,\n",
       " <tf.Variable 'dense_2/bias:0' shape=(64,) dtype=float32, numpy=\n",
       " array([0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
       "        0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
       "        0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
       "        0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.], dtype=float32)>,\n",
       " <tf.Variable 'dense_3/kernel:0' shape=(64, 10) dtype=float32, numpy=\n",
       " array([[ 2.17074484e-01, -2.19056025e-01,  7.96305537e-02,\n",
       "          4.14488614e-02, -1.73471987e-01,  1.25416100e-01,\n",
       "         -1.93215817e-01, -1.97822779e-01, -2.06122145e-01,\n",
       "          1.27461046e-01],\n",
       "        [-1.58436924e-01, -1.60176113e-01, -2.75985092e-01,\n",
       "         -2.23029256e-02,  2.57320493e-01, -1.33338273e-01,\n",
       "          2.63068944e-01,  2.80370742e-01, -1.05902195e-01,\n",
       "          6.43995106e-02],\n",
       "        [ 4.61100638e-02, -7.61910826e-02,  2.43031591e-01,\n",
       "          1.53610140e-01,  1.02945000e-01, -2.01045811e-01,\n",
       "         -2.23917291e-01, -1.96345717e-01, -7.13494420e-03,\n",
       "          1.80781066e-01],\n",
       "        [ 1.18140757e-01,  1.08397901e-02,  1.82681948e-01,\n",
       "         -2.56778181e-01, -1.50109857e-01, -2.82579899e-01,\n",
       "          1.01017654e-02,  2.75529832e-01, -1.04429752e-01,\n",
       "          6.19024634e-02],\n",
       "        [ 9.22510028e-02, -1.24821320e-01, -2.55198866e-01,\n",
       "          7.93420076e-02,  1.82297230e-01, -1.94099665e-01,\n",
       "         -1.19112521e-01,  1.08176261e-01,  2.17510313e-01,\n",
       "          1.21651590e-01],\n",
       "        [-2.13928908e-01,  2.39574313e-02,  2.66536206e-01,\n",
       "          8.58357549e-02, -1.11642182e-02, -2.05067694e-02,\n",
       "          8.47905278e-02,  8.71853232e-02, -1.77813560e-01,\n",
       "          2.31237084e-01],\n",
       "        [-2.43385494e-01,  2.12107092e-01,  8.72815251e-02,\n",
       "          1.20665342e-01, -2.53377199e-01,  1.58619344e-01,\n",
       "          1.88976228e-01, -3.77151668e-02,  2.21209496e-01,\n",
       "          1.51718974e-01],\n",
       "        [-1.47596404e-01,  2.21869648e-02, -1.04874834e-01,\n",
       "          2.82969385e-01,  7.82361627e-02,  1.06002808e-01,\n",
       "         -9.72592831e-03, -2.50848264e-01,  2.36554056e-01,\n",
       "         -1.48026481e-01],\n",
       "        [ 1.35919809e-01, -1.91541612e-02,  1.39136076e-01,\n",
       "          5.94370067e-02, -1.89822733e-01, -1.02386698e-01,\n",
       "          2.05671966e-01, -2.16597497e-01, -7.60444999e-03,\n",
       "         -1.70053363e-01],\n",
       "        [ 1.72791243e-01,  4.28040624e-02, -1.63106889e-01,\n",
       "          2.21172273e-02, -7.23528415e-02, -2.47960478e-01,\n",
       "         -1.86282039e-01, -1.94097430e-01, -1.43139958e-02,\n",
       "          2.11118698e-01],\n",
       "        [ 1.99614763e-02,  2.79899687e-01, -1.65547907e-01,\n",
       "         -6.35915548e-02,  2.41124243e-01,  2.56442577e-01,\n",
       "         -2.44664595e-01, -4.06189114e-02, -1.23906925e-01,\n",
       "          2.55947113e-02],\n",
       "        [-1.16657719e-01,  4.26815152e-02,  1.91374749e-01,\n",
       "         -2.19247043e-02,  2.05174416e-01, -2.06796348e-01,\n",
       "          1.50244743e-01,  2.60449797e-01, -2.32629776e-01,\n",
       "         -1.36347651e-01],\n",
       "        [ 1.76198334e-01, -2.65627652e-01,  1.75982326e-01,\n",
       "          5.29471040e-02,  6.36278093e-02, -1.40531659e-01,\n",
       "         -1.28101110e-01, -5.50817251e-02, -1.10304803e-01,\n",
       "         -6.60612881e-02],\n",
       "        [-1.70593962e-01, -1.11206502e-01, -7.35211521e-02,\n",
       "          2.41804570e-01,  1.34344995e-01, -2.13132650e-01,\n",
       "         -2.06618488e-01,  8.69478285e-02, -1.56354368e-01,\n",
       "         -2.39576176e-01],\n",
       "        [ 6.67087436e-02, -1.85571313e-01, -2.58436978e-01,\n",
       "          1.54694825e-01,  7.87000060e-02,  1.71048611e-01,\n",
       "         -6.93131089e-02, -7.82536268e-02,  7.90117979e-02,\n",
       "          9.80792642e-02],\n",
       "        [ 2.60564983e-02,  2.54491359e-01,  2.40856618e-01,\n",
       "         -6.91896230e-02, -1.61433622e-01,  7.87200928e-02,\n",
       "         -1.72544479e-01, -2.51686841e-01,  2.02265292e-01,\n",
       "          1.58936858e-01],\n",
       "        [ 1.52867854e-01,  2.13724554e-01,  1.77818775e-01,\n",
       "          2.27710336e-01,  2.55345732e-01, -9.20412838e-02,\n",
       "          1.40213192e-01,  8.58690739e-02, -2.30245888e-02,\n",
       "         -8.77007991e-02],\n",
       "        [ 6.42132163e-02, -2.16043442e-01,  8.63406956e-02,\n",
       "          5.03391922e-02,  1.86333299e-01, -7.37154484e-03,\n",
       "          9.95850563e-03,  1.21213764e-01,  2.15497285e-01,\n",
       "         -3.28498781e-02],\n",
       "        [ 8.51219893e-03,  1.31935477e-02, -1.98505938e-01,\n",
       "         -7.38510489e-03,  3.84712517e-02, -2.58263648e-02,\n",
       "          3.61145437e-02,  8.77046585e-02,  2.06857800e-01,\n",
       "         -8.18403363e-02],\n",
       "        [ 2.70011634e-01,  2.54499882e-01,  1.87638730e-01,\n",
       "          2.83424705e-01,  1.73316985e-01,  2.46611923e-01,\n",
       "         -5.02557009e-02, -8.81400406e-02,  4.58987057e-02,\n",
       "         -2.65262336e-01],\n",
       "        [ 2.34016627e-01,  1.12615198e-01, -2.07946926e-01,\n",
       "          1.93745971e-01, -7.48189837e-02, -2.20714837e-01,\n",
       "          1.34008825e-02,  2.20342845e-01,  1.53008997e-01,\n",
       "         -1.84713185e-01],\n",
       "        [ 7.74485171e-02,  2.59026796e-01, -1.77077979e-01,\n",
       "         -6.35819882e-02,  2.43172646e-02, -9.24241841e-02,\n",
       "         -6.35919571e-02,  2.51815408e-01, -1.46374941e-01,\n",
       "         -1.46820500e-01],\n",
       "        [-1.38503909e-02, -2.60964483e-01,  1.55724347e-01,\n",
       "         -8.96395743e-02, -7.73901492e-02,  1.54732555e-01,\n",
       "          1.75412863e-01,  2.56974727e-01,  8.56804848e-02,\n",
       "          1.39071763e-01],\n",
       "        [ 2.47884959e-01,  3.47283185e-02,  1.73696280e-01,\n",
       "          2.75216371e-01,  7.01336861e-02,  1.83018684e-01,\n",
       "         -7.18000531e-03,  5.60210943e-02, -2.78495818e-01,\n",
       "          1.03074193e-01],\n",
       "        [ 1.83330357e-01, -4.99430001e-02, -2.13515058e-01,\n",
       "         -6.03032857e-02,  3.66166234e-03,  1.96499199e-01,\n",
       "          2.22783476e-01, -8.82830769e-02,  2.13314027e-01,\n",
       "          2.06195384e-01],\n",
       "        [ 1.48228109e-01,  7.48572648e-02, -2.19567508e-01,\n",
       "         -2.49093473e-01,  5.54320216e-03,  2.84060687e-01,\n",
       "          9.05822217e-02, -1.83299333e-01,  1.85823143e-02,\n",
       "          5.23413718e-02],\n",
       "        [ 1.41068935e-01,  9.13734734e-02,  7.33092725e-02,\n",
       "          2.27762491e-01, -1.41564906e-02, -1.52656317e-01,\n",
       "         -1.49226487e-01,  4.00558114e-03,  2.12694287e-01,\n",
       "          1.33024901e-01],\n",
       "        [ 1.30200088e-01, -1.48512766e-01, -5.89863062e-02,\n",
       "         -1.52798206e-01, -1.29921690e-01, -8.18261504e-03,\n",
       "         -5.54482639e-02,  1.17237777e-01, -4.88560349e-02,\n",
       "          1.09569013e-01],\n",
       "        [-2.61063278e-01,  2.03845412e-01,  3.58880758e-02,\n",
       "          7.37246871e-03, -1.32807717e-01,  2.65666574e-01,\n",
       "         -1.33398220e-01, -2.30505452e-01, -1.26372382e-01,\n",
       "          2.00670928e-01],\n",
       "        [-2.71452367e-02,  1.36788666e-01, -2.76511639e-01,\n",
       "          1.40151083e-01,  1.39391661e-01, -7.05723166e-02,\n",
       "         -2.17252895e-01, -4.41267341e-02, -3.85114253e-02,\n",
       "         -2.72816777e-01],\n",
       "        [ 1.89958155e-01,  2.68206507e-01,  9.69453156e-02,\n",
       "          3.76256108e-02, -1.97507709e-01,  1.30831271e-01,\n",
       "         -2.21952498e-02, -2.08769411e-01,  6.94386959e-02,\n",
       "         -1.83002248e-01],\n",
       "        [-1.38214529e-02, -1.46303043e-01,  2.07124650e-01,\n",
       "         -2.71091968e-01, -1.08499229e-01, -2.60965705e-01,\n",
       "         -2.63519377e-01,  7.42405653e-02, -1.51957601e-01,\n",
       "          2.39168853e-01],\n",
       "        [-2.25854233e-01,  1.00621909e-01, -1.69933140e-02,\n",
       "          2.07967579e-01, -7.25721270e-02,  4.37198281e-02,\n",
       "         -2.60755539e-01, -2.62857169e-01, -2.76417136e-02,\n",
       "          2.81398982e-01],\n",
       "        [-1.24045208e-01, -1.17256507e-01, -2.19814956e-01,\n",
       "         -1.17856711e-01,  6.85347617e-02, -2.62225926e-01,\n",
       "         -4.63301539e-02, -1.80539906e-01, -1.80377588e-01,\n",
       "         -1.07307836e-01],\n",
       "        [-1.13647118e-01,  2.40069628e-03,  1.20185435e-01,\n",
       "         -5.47287762e-02,  1.23165101e-01,  2.19172239e-03,\n",
       "         -1.27762139e-01,  1.53684229e-01,  1.32840455e-01,\n",
       "          1.22388303e-01],\n",
       "        [ 2.57497102e-01, -2.56958902e-01, -1.40755907e-01,\n",
       "         -2.39729136e-01,  7.39736855e-02,  1.33734256e-01,\n",
       "          1.10358566e-01,  3.67064774e-02,  2.75358826e-01,\n",
       "          1.92936987e-01],\n",
       "        [-2.65690804e-01,  1.93678766e-01, -1.82962328e-01,\n",
       "          1.15052611e-01, -2.28839815e-02,  1.35959804e-01,\n",
       "          5.27476966e-02, -1.84969902e-02, -1.00362316e-01,\n",
       "          1.55186117e-01],\n",
       "        [ 2.70625561e-01,  4.28678691e-02, -1.06020600e-01,\n",
       "          2.69277960e-01,  2.82270938e-01, -1.03347182e-01,\n",
       "         -1.40225545e-01,  1.50209725e-01, -2.94844210e-02,\n",
       "          1.93265170e-01],\n",
       "        [ 2.79125899e-01, -1.17458403e-02,  1.43451720e-01,\n",
       "          1.39283240e-02,  2.42746025e-01,  1.82271838e-01,\n",
       "          8.01951289e-02, -1.77145123e-01,  1.94423497e-02,\n",
       "          1.37212485e-01],\n",
       "        [-1.84941083e-01,  2.79207319e-01, -1.05193570e-01,\n",
       "         -2.33843014e-01,  2.39724487e-01, -9.63825881e-02,\n",
       "         -1.17302537e-01, -1.24360964e-01,  5.09142280e-02,\n",
       "          7.91955292e-02],\n",
       "        [ 2.62835354e-01,  1.18748844e-02, -4.55919206e-02,\n",
       "          1.50599778e-02,  2.71084458e-01, -2.83947945e-01,\n",
       "          1.33505881e-01,  3.30073237e-02, -6.10081106e-02,\n",
       "          2.60763079e-01],\n",
       "        [ 7.76291192e-02,  1.77983344e-01, -5.58765680e-02,\n",
       "         -1.06005251e-01, -8.93856734e-02,  2.07388490e-01,\n",
       "          2.13700116e-01,  1.50020510e-01, -1.02498710e-01,\n",
       "         -5.32885790e-02],\n",
       "        [-6.19277358e-02,  1.90282464e-01, -2.09181085e-01,\n",
       "         -1.72069043e-01,  1.89360529e-01, -2.80439764e-01,\n",
       "          2.94130743e-02,  2.72372961e-02,  2.80460387e-01,\n",
       "         -4.38900143e-02],\n",
       "        [-3.32780778e-02, -1.33803189e-02, -1.94647610e-01,\n",
       "          2.61643827e-02, -2.06757724e-01,  1.06836766e-01,\n",
       "          5.44530749e-02,  1.14609092e-01,  1.86676741e-01,\n",
       "          1.94243312e-01],\n",
       "        [-9.15675610e-02, -2.27465853e-01,  2.64421701e-02,\n",
       "         -1.68085188e-01,  4.43497598e-02,  2.63766140e-01,\n",
       "         -1.05646789e-01,  1.20709419e-01, -1.42370433e-01,\n",
       "          3.65019739e-02],\n",
       "        [ 9.75586176e-02,  1.20581239e-01,  2.20582038e-01,\n",
       "         -2.80744255e-01,  1.30711317e-01,  1.40190184e-01,\n",
       "          1.73987448e-01,  2.60150760e-01, -2.29763493e-01,\n",
       "         -2.19376534e-01],\n",
       "        [ 4.88598347e-02, -1.87220126e-01,  2.82825828e-02,\n",
       "         -2.25723818e-01, -1.63885832e-01, -2.73435175e-01,\n",
       "         -2.67614573e-01,  1.00832790e-01, -6.60801679e-02,\n",
       "         -1.23441473e-01],\n",
       "        [-1.33524761e-01,  1.02311403e-01,  2.21125096e-01,\n",
       "          8.28639567e-02, -2.72041887e-01, -1.20758966e-01,\n",
       "          2.50561327e-01,  1.96752697e-01,  8.90550613e-02,\n",
       "          2.55255371e-01],\n",
       "        [-5.51671386e-02, -2.11536437e-01, -1.91568628e-01,\n",
       "         -1.65533841e-01,  2.08199084e-01, -3.82017940e-02,\n",
       "         -2.04967007e-01, -2.20665336e-01, -3.25707197e-02,\n",
       "          1.05049163e-01],\n",
       "        [-2.37336785e-01,  1.86308235e-01, -9.05234218e-02,\n",
       "         -2.21513405e-01, -8.70283544e-02,  1.97572470e-01,\n",
       "         -2.53556788e-01, -1.00980714e-01, -2.11111248e-01,\n",
       "         -1.72546715e-01],\n",
       "        [ 2.83779293e-01,  1.99901462e-02, -1.54520616e-01,\n",
       "         -9.45167243e-02, -1.36832461e-01,  2.07312316e-01,\n",
       "          3.66804600e-02, -4.94185537e-02,  2.61131734e-01,\n",
       "          1.91707402e-01],\n",
       "        [-4.75141406e-03,  2.49467701e-01,  4.59954143e-03,\n",
       "         -1.24981940e-01, -3.43051553e-02, -1.08729362e-01,\n",
       "         -2.12134331e-01,  2.71622688e-01,  1.62206531e-01,\n",
       "         -2.34395266e-03],\n",
       "        [-1.01741135e-01, -1.12254649e-01,  1.92770064e-01,\n",
       "          2.17611104e-01,  1.24480039e-01,  5.14913499e-02,\n",
       "         -8.55741650e-02,  5.74550033e-02,  9.25431848e-02,\n",
       "         -7.80451447e-02],\n",
       "        [-3.06985676e-02, -9.04743373e-02,  1.06752515e-01,\n",
       "          4.29026186e-02, -1.87769294e-01,  5.93974292e-02,\n",
       "          2.74775296e-01, -1.89676821e-01, -1.11164153e-02,\n",
       "          1.21588856e-01],\n",
       "        [-4.29499149e-03, -2.27952614e-01,  8.16780031e-02,\n",
       "          1.47055864e-01, -2.80024946e-01, -2.59064615e-01,\n",
       "         -2.60767341e-03, -3.31589580e-03,  2.13713646e-01,\n",
       "          7.99350441e-02],\n",
       "        [ 1.34360075e-01, -5.97684532e-02, -2.61035711e-01,\n",
       "          2.14434266e-01, -2.66989589e-01, -9.25461054e-02,\n",
       "          2.11898357e-01, -4.06061560e-02, -2.53517330e-01,\n",
       "         -1.73738107e-01],\n",
       "        [-1.79437876e-01,  3.13853323e-02,  8.62575471e-02,\n",
       "          2.16076672e-02,  2.15762943e-01, -2.18069598e-01,\n",
       "         -1.41472667e-01,  2.69518882e-01, -5.03939390e-03,\n",
       "         -5.93256056e-02],\n",
       "        [ 1.37140125e-01,  2.92468071e-02, -2.51908362e-01,\n",
       "          3.20817828e-02, -7.34728575e-03,  9.56011713e-02,\n",
       "         -8.89499485e-02, -1.64279342e-02, -2.06456691e-01,\n",
       "         -1.70914829e-02],\n",
       "        [ 1.80518925e-01,  1.46829516e-01,  2.10615307e-01,\n",
       "          8.48306417e-02, -1.09908059e-01,  1.14105940e-02,\n",
       "          4.72153425e-02,  1.37407213e-01, -9.81770307e-02,\n",
       "          1.18383795e-01],\n",
       "        [-7.37992227e-02, -2.02234745e-01, -1.06471986e-01,\n",
       "          1.86678588e-01, -4.10178304e-02,  1.32021755e-01,\n",
       "         -2.56512940e-01,  2.06745505e-01, -8.54084492e-02,\n",
       "          2.55252570e-01],\n",
       "        [ 1.85904175e-01, -1.51288137e-01,  7.57500231e-02,\n",
       "         -7.47717917e-02, -1.11712888e-01,  8.05260837e-02,\n",
       "         -7.29172677e-02, -1.73040688e-01,  1.85607553e-01,\n",
       "         -7.20233023e-02],\n",
       "        [-1.62813812e-01,  1.23460680e-01,  6.69803023e-02,\n",
       "         -2.02522606e-01, -6.34837151e-03,  1.59909844e-01,\n",
       "         -2.81164408e-01,  1.75237656e-05, -5.59010804e-02,\n",
       "         -2.58843303e-01],\n",
       "        [-2.05212235e-01, -8.32064748e-02,  2.74249345e-01,\n",
       "          2.16363996e-01, -2.28863284e-01, -2.80138850e-03,\n",
       "          1.40494198e-01, -2.38389075e-02, -1.50325954e-01,\n",
       "          4.56034541e-02],\n",
       "        [-1.29941851e-01, -7.01679736e-02, -2.35103920e-01,\n",
       "         -2.10668743e-01, -1.45267606e-01, -4.85550761e-02,\n",
       "         -2.04793543e-01, -1.08572409e-01,  6.01774454e-03,\n",
       "         -9.24997330e-02]], dtype=float32)>,\n",
       " <tf.Variable 'dense_3/bias:0' shape=(10,) dtype=float32, numpy=array([0., 0., 0., 0., 0., 0., 0., 0., 0., 0.], dtype=float32)>]"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.build(input_shape=(None,3))\n",
    "model.weights"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_1\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " dense_2 (Dense)             (None, 64)                256       \n",
      "                                                                 \n",
      " dense_3 (Dense)             (None, 10)                650       \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 906\n",
      "Trainable params: 906\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"test_model\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " test_layer_1 (Dense)        (None, 64)                256       \n",
      "                                                                 \n",
      " test_layer_2 (Dense)        (None, 10)                650       \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 906\n",
      "Trainable params: 906\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# naming a model\n",
    "\n",
    "model = keras.Sequential(name=\"test_model\")\n",
    "model.add(layers.Dense(64, activation=\"relu\", name=\"test_layer_1\"))\n",
    "model.add(layers.Dense(10, activation=\"softmax\", name=\"test_layer_2\"))\n",
    "\n",
    "model.build(input_shape=(None,3))\n",
    "model.summary()\n",
    "\n",
    "# params calc\n",
    "# \n",
    "# test_layer_1: \n",
    "# weights = 3 x 64 = 192\n",
    "# bias = 1 x 64 = 64\n",
    "# total = 256\n",
    "# \n",
    "# test_layer_2:\n",
    "# weights = 64 x 10 = 60\n",
    "# bias = 1 x 10 = 10\n",
    "# total = 650"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_2\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " dense_4 (Dense)             (None, 64)                256       \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 256\n",
      "Trainable params: 256\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# we cant viwe the summary of the model until its built\n",
    "# model can be built on the go by using keras Input class\n",
    "\n",
    "model1 = keras.Sequential()\n",
    "model1.add(keras.Input(shape=(3,)))\n",
    "model1.add(layers.Dense(64, activation=\"relu\"))\n",
    "model1.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_2\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " dense_4 (Dense)             (None, 64)                256       \n",
      "                                                                 \n",
      " dense_5 (Dense)             (None, 10)                650       \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 906\n",
      "Trainable params: 906\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model1.add(layers.Dense(10, activation=\"softmax\"))\n",
    "model1.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**The Functional API**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " input_layer (InputLayer)    [(None, 3)]               0         \n",
      "                                                                 \n",
      " feature_layer_1 (Dense)     (None, 64)                256       \n",
      "                                                                 \n",
      " feature_layer_2 (Dense)     (None, 64)                4160      \n",
      "                                                                 \n",
      " output_layer (Dense)        (None, 10)                650       \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 5,066\n",
      "Trainable params: 5,066\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "inputs = keras.Input(shape=(3,), name=\"input_layer\")\n",
    "feature_1 = layers.Dense(64, activation=\"relu\", name=\"feature_layer_1\")(inputs)\n",
    "feature_2 = layers.Dense(64, activation=\"relu\", name=\"feature_layer_2\")(feature_1)\n",
    "outputs = layers.Dense(10, activation=\"softmax\", name=\"output_layer\")(feature_2)\n",
    "\n",
    "model = keras.Model(inputs=inputs, outputs=outputs)\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(TensorShape([None, 3]), tf.float32)"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# the \"input_layer\" object is call \"Symbolic Tensor\" as it doen't contain actual data but encodes the specs of the actual tensor \n",
    "\n",
    "inputs.shape, inputs.dtype"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**The Functional API**\n",
    "\n",
    "*Multi-Inputs Multi-oututs model*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model_1\"\n",
      "__________________________________________________________________________________________________\n",
      " Layer (type)                   Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      " title (InputLayer)             [(None, 1000)]       0           []                               \n",
      "                                                                                                  \n",
      " text_body (InputLayer)         [(None, 1000)]       0           []                               \n",
      "                                                                                                  \n",
      " tags (InputLayer)              [(None, 100)]        0           []                               \n",
      "                                                                                                  \n",
      " concatenate (Concatenate)      (None, 2100)         0           ['title[0][0]',                  \n",
      "                                                                  'text_body[0][0]',              \n",
      "                                                                  'tags[0][0]']                   \n",
      "                                                                                                  \n",
      " feature_layer_2 (Dense)        (None, 64)           134464      ['concatenate[0][0]']            \n",
      "                                                                                                  \n",
      " priority (Dense)               (None, 1)            65          ['feature_layer_2[0][0]']        \n",
      "                                                                                                  \n",
      " department (Dense)             (None, 4)            260         ['feature_layer_2[0][0]']        \n",
      "                                                                                                  \n",
      "==================================================================================================\n",
      "Total params: 134,789\n",
      "Trainable params: 134,789\n",
      "Non-trainable params: 0\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "from ast import mod\n",
    "from unicodedata import name\n",
    "\n",
    "\n",
    "voc_size = 1000\n",
    "num_tags = 100\n",
    "num_dept = 4\n",
    "\n",
    "title = keras.Input(shape=(voc_size,), name=\"title\")\n",
    "text_body = keras.Input(shape=(voc_size,), name=\"text_body\")\n",
    "tags = keras.Input(shape=(num_tags,), name=\"tags\")\n",
    "\n",
    "feature_layer_1 = layers.Concatenate()([title, text_body, tags])\n",
    "feature_layer_2 = layers.Dense(64, activation=\"relu\", name=\"feature_layer_2\")(feature_layer_1)\n",
    "\n",
    "priority = layers.Dense(1, activation=\"sigmoid\", name=\"priority\")(feature_layer_2)\n",
    "dept = layers.Dense(num_dept, activation=\"softmax\", name=\"department\")(feature_layer_2)\n",
    "\n",
    "mimo_model = keras.Model(inputs=[title, text_body, tags], outputs=[priority, dept])\n",
    "mimo_model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "num_samples = 1280\n",
    "\n",
    "# dummy inputs\n",
    "title_data = np.random.randint(0, 2, size=(num_samples, voc_size))\n",
    "text_body_data = np.random.randint(0, 2, size=(num_samples, voc_size))\n",
    "tags_data = np.random.randint(0, 2, size=(num_samples, num_tags))\n",
    "\n",
    "# dummy targets\n",
    "priority_data = np.random.random(size=(num_samples,1))\n",
    "dept_data = np.random.randint(0, 2, size=(num_samples, num_dept))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "40/40 [==============================] - 1s 2ms/step - loss: 4.4124 - priority_loss: 0.1431 - department_loss: 4.2693 - priority_mean_absolute_error: 0.3107 - department_accuracy: 0.2211\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x7feb40e9b820>"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mimo_model.compile(optimizer=\"rmsprop\",\n",
    "                    loss=[\"mean_squared_error\", \"categorical_crossentropy\"],\n",
    "                    metrics=[[\"mean_absolute_error\"], [\"accuracy\"]])\n",
    "\n",
    "mimo_model.fit([title_data, text_body_data, tags_data],\n",
    "                    [priority_data, dept_data],\n",
    "                    epochs=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "40/40 [==============================] - 0s 2ms/step - loss: 3.5415 - priority_loss: 0.1392 - department_loss: 3.4023 - priority_mean_absolute_error: 0.3077 - department_accuracy: 0.4461\n",
      "40/40 [==============================] - 0s 1ms/step\n"
     ]
    }
   ],
   "source": [
    "mimo_model.evaluate([title_data, text_body_data, tags_data],\n",
    "                [priority_data, dept_data])\n",
    "\n",
    "priority_preds, dept_preds = mimo_model.predict([title_data, text_body_data, tags_data])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "40/40 [==============================] - 0s 2ms/step - loss: 4.5784 - priority_loss: 0.1682 - department_loss: 4.4102 - priority_mean_absolute_error: 0.3349 - department_accuracy: 0.2492\n",
      "40/40 [==============================] - 0s 2ms/step - loss: 7.3814 - priority_loss: 0.1079 - department_loss: 7.2735 - priority_mean_absolute_error: 0.2758 - department_accuracy: 0.1172\n",
      "40/40 [==============================] - 0s 1ms/step\n"
     ]
    }
   ],
   "source": [
    "# providing inputs & targets as dicts using layer names\n",
    "\n",
    "mimo_model.compile(optimizer=\"rmsprop\",\n",
    "                    loss={\"priority\": \"mean_squared_error\", \"department\": \"categorical_crossentropy\"},\n",
    "                    metrics={\"priority\": [\"mean_absolute_error\"], \"department\": [\"accuracy\"]})\n",
    "\n",
    "mimo_model.fit({\"title\": title_data, \"text_body\": text_body_data, \"tags\": tags_data},\n",
    "                    {\"priority\": priority_data, \"department\": dept_data},\n",
    "                    epochs=1)\n",
    "\n",
    "mimo_model.evaluate({\"title\": title_data, \"text_body\": text_body_data, \"tags\": tags_data},\n",
    "                    {\"priority\": priority_data, \"department\": dept_data})\n",
    "\n",
    "priority_preds, dept_preds = mimo_model.predict({\"title\": title_data, \"text_body\": text_body_data, \"tags\": tags_data})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABJsAAAGVCAIAAADWp2edAAAABmJLR0QA/wD/AP+gvaeTAAAgAElEQVR4nOzde1xTV7o//hUIlxBCQFBAxCq02lPESNFRVIY7SEGtFESrvcwUy9iplFo8glZ75lhttbZqZ2y91fZM1Qq1R7+Dt4qodQS0qAXrlQpWQS4Cyv0igf37Y/26z24CIcBOdhI+7z98mZXF3s++5EmeZO21RQzDEAAAAAAAADBCZkIHAAAAAAAAAP2Eig4AAAAAAMBYoaIDAAAAAAAwVqjoAAAAAAAAjJWY+yAvL++TTz4RKhQAGFSWLl3q5+c3wIXExcXxEgwAgGZ+fn5Lly4d+HKQtQBg4FQy0u9+oystLT1w4IDeQwJCCDl//vz58+eFjsKAHDhwoKysTOgoQFcOHDhQWlrKy3JwngiirKwM7xdcyOGm7fz583l5ebwsCllLKMhaKpC1jJd6RhKrd/r222/1FQ/8H/qlHXY+SyQSvf3223PnzhU6ENAJkUjE16JwnggiIyMjPj4eKYuFHG7a+P1hDVlLEMhaKpC1jJd6RsJ1dAAAAAAAAMYKFR0AAAAAAICxQkUHAAAAAABgrFDRAQAAAAAAGCvTrOg2btwoEolEItGIESOEjgV0aM+ePaLf2Nraqjx79+7dWbNmNTQ01NTUsN18fHza2tq43bjPikSiiRMn6nEL+uDo0aNjxowRi7uZzYgqKCiIioqyt7eXyWShoaE5OTm66JOampqenq7yJ6mpqewOnDJlSr+2b5DSUbLav38/Xay1tTWPi4WBQ9biQtYCMHzIWlwGm5GMrKJramp66qmnoqOjNXdLSUlhGEahUOgnKmFpuU9M2Oeff84wTFNTE7exoKBg4sSJ4eHhdnZ2Tk5ODMPk5+fT9uTkZG5P+mxeXp6joyPDMBcvXtRr9FooLi6eNWtWWlpaVVVVT30uXLgwdepUmUx248aNO3fueHh4BAYGnjhxgvc+ixYtSktLW7VqFfevPvzwQ4ZhGIYxNzfnaaMHCx0lq3nz5jEMExISwu9ieYGURZC1CCHIWmA8kLUIshYhxLAzkuFWdLa2ttOnT1dpZBimq6urq6ur156DR7f7RJ8McP83NDTMnDnzhRdeePPNN7ntVlZWjo6O27dv/+abb4SKrR9WrVo1derUS5cuyWSybjt0dXW99tpr9vb2X375paurq5OT0+eff+7p6ZmQkNDe3s5vH09Pz4MHD65duzYjI0MP2647ujtvDfAVYVCQsrqFrIWspWeG+UIwTMha3ULWMqiMZLgVXbdkMllxcfHRo0eFDsSAYJ+o27BhQ2Vl5erVq1Xara2t9+7da2ZmlpiYWFRUJEhs/fDFF1+kpqZqGANw9uzZa9euxcbGSiQS2mJubj5//vzS0tLDhw/z24cQolAoYmNj33nnHaVSyfvGgslDyuoWshayFhgsZK1uIWsZVEYysooOoFcMw+zatWvy5MnDhw9XfzYiIuLdd99tbGyMi4tTGeRtsNi80JNTp04RQlRGpdOH2dnZ/Pah5syZU1ZWduTIkb5uCwCoQ9aikLUAjAWyFmU4GckQKzo6VUBzc3NOTg69RpBWzIcOHWKvGqTnR089e1JdXZ2UlDRq1ChLS8uhQ4fGxMQUFBToaat0Q32fcFt+/fXX+Ph4e3t7R0fH6Ojo4uJi+lfcyRjy8/NDQkJkMpmNjU1QUBB7+eb7779P+7A/9B8/fpy2ODk5cZej/f7Xj8LCwqqqKg0XJr333nvh4eFXrlxZsmSJhuXU1tYuXbrU09PT0tLSwcEhMjLy9OnT9CltdjKln1Pu5s2bhBCVqTXc3NwIIezXY3z1oSZMmEAI+f7773ncCr3RfN5qOGTTp09nj/vChQsJIaGhoWxLXV3dAF8RN2/ejIqKksvlKi9GSsMJyf75888/L5fLpVKpv7//uXPn2Kfq6uq4l6S///77hBClUsm2xMbG9n1H9gdSVreQtShkLf3Q8EJQKpXp6elhYWEuLi4SicTb23vLli0qow3ZVGNjY/OHP/zh8OHDbCZMSEgghLS3t69evfrpp5+2sbEZMmTIzJkz//Wvf3V2dgqwqXxA1uoWshZlQBmJ4aBTrzCGQSqVTps2Tb199uzZhJDW1tZeeyoUCjc3N/ZheXn5E0884ezsfOTIkcbGxqtXrwYEBFhbW+fm5uoi/r6KjY2NjY3t39+q7xPaMnv27Nzc3KampqysLIlEMmnSJO5fKRQKqVTq5+dH++Tn548fP97S0vLMmTNsH/V96+vrS69q1dCHCgoKGjJkSF5eXv82ihCSnp6uuc/XX39NfrtaV6Vx3bp1Kp3z8/Plcjn9f3V1tbu7OyFkz549tIW9WpeqqKgYPXq0s7NzZmZmfX39rVu3YmJiRCLRzp072T697mTeTzk3Nzdzc3P19rCwMELI+fPnuY2//PILIeTZZ5/ltw9VX19PCPH391eJxNzcfPLkydpsizbHV6fL6fa87fWQFRQUSKVShULR1NTEMExbW9vkyZO/+eabXpesmUKhkMvlQUFB586da2xsVH8x9npC/vLLL/b29m5ubidOnGhsbLxy5Up4ePioUaOsrKzYtURERJiZmd2+fZu7aj8/v7179/YpWmog7xcmmbK0zOHIWpTRZa2BvEer4Cv7DVC3L4TMzEx6Kj58+LC6uvrTTz81MzOjEzhRKqnm6tWroaGhQ4cO5aaahIQEuVx+4sSJlpaWysrKlJQUQsjp06f1s10aIGupQNbqlq6zFsV7RjLE3+h0JC0t7e7du5988slzzz1na2vr5eW1f/9+hmE0f3lg1BISEvz8/KRSaWhoaFRUVH5+fk1NDbdDc3PzZ599RvtMnDhxz549jx8/fuutt3hZe1dXFz3JeFma9ioqKgghcrlcQx8nJ6eMjAwLC4vExET6hYqKtLS0O3fubN68OTo62s7ObsyYMfv27XN1dU1KSlKZBEnDThb2lKN7XiQS6aKPnZ2dSCSiu9qU9HrIFArFl19+WVhY+PLLLzMMk5iYGBISMm/evIGvur6+ft26ddOmTbO1tVV/MfZ6Qq5YsaKurm7Lli1hYWG2trbe3t5ffvmlygFaunRpV1fXJ598wrbk5OTcu3cvLi5u4PHzYnCmLIKs9RtkLUMQGBiYlpbm4ODg5OS0ZMmSF198ccuWLQ0NDfRZlVTj5eW1b9++5uZm7hKys7O9vLzCwsIkEomzs/NHH300ZswYITZFH5C1NPRB1tLQh/eMNIgqukOHDpmZmXEnn3VxcfHy8rp06VJZWZmAgenOpEmT2P/Tr0nKy8u5HaRSKf3Zl/L29h4+fHhhYSEvZ9iZM2cePnzo5+c38EX1CR0UYWFhobnblClTNm7c2NzcHBcX19raqvLswYMHCSFRUVFsi5WVVUhISGtrq8pP5Bp2st5OOXt7e0KIynsqfUif4rEPSywWq+83Y6fNIYuLi1u5cuX//u//Tp8+vba2ds2aNbys2traevLkyexDlRdjryfk8ePHCSERERFsh+HDh6t8igoPD/f29v7qq69qa2tpy0cffbRkyZJeXyx6MzhTFkHW+g2yluCio6NVhnMrFIqOjo5r167Rh+qpZujQoU8//TT3T2bMmJGbm/v666+fP3+eDra8detWYGCgroMXBLKW5m7IWup9WPxmpMFS0bW3t9fX13d1dcnlcu71JJcvXyaE0N9DTQ/3uxNLS0tCiMpoePXTa9iwYYSQBw8e6D46XaG3VO7o6Oi1Z1JSUnx8/NWrV1Um3qVni7W1tcoMts7OzoSQyspKbmNPO1mfpxx9N1XJXPfv3yeEsJ/p+erDUiqVvV5GbFy0P2Rr1qyZPHlybm5uXFycmRk/WdTR0VHlCzz2xdjrCdne3t7Y2Ghtba1y71e6BK7k5OSWlpbPPvuMEFJUVHTq1KnXX3+dl/h5MThTFkHW+g2yluDq6+tXr17t7e3t4OBAj/6yZcsIIS0tLYSQnlKNg4MD9+HWrVv/+c9/lpSUhISE2NnZzZgxg35wN0nIWr32RNbST0Yy3IpO8y+Yfe1pZWVlb28vFos7OjrUR6MGBQUNLFhjVVtbq/JjPU0x7AdBMzOzx48fczvU1dWpLET7I6Ufrq6uhBA6QLlXu3btGjt27O7du+mIcMrKykoul7e1tTU2NnI70zEALi4u2ixZn6ccXdqlS5e4jfQhe49pvvpQDQ0NDMPQXW2k1M9b7Q/ZmTNn6uvrvb2933jjjcLCwl6XrA31M5Z9MfZ6QlpZWclksra2NpV7vz58+FBlmQsWLHB2dv7HP/7R3t7+8ccfv/LKKyofxQycSaYsgqz1G2Qtfer2hTBz5sw1a9YsWrSoqKiIjujbtGkT+W3YWE+pRqU4EYlEL7300smTJ+vq6g4dOsQwTExMDHe896CCrEWQtfSSkQy3orOxsWFP8bFjx+7YsWOAPWNiYpRKpcr0cevXrx85cuSgvUFNW1tbfn4++/Dnn38uLy9XKBTsGebq6kq/WqAqKyvv3bunshDtj5R+jBs3jqh9QdITW1vb7777TiqV0l8tWHPmzCGEcGeVbW9vz87Olkgk3NEmmuntlAsICHjmmWcOHDjAzhHc2dm5f/9+d3d3djADX30oelbQXW2kuj1vtTlkd+7cee2117777rt//etfEolk9uzZ1dXVvS65V01NTdziUOXF2OsJGRkZSX4bEEXV1NTcunVLZS1WVlZvvPHGgwcPPv7447179/J1LYfemGTKIshahBBkLb1TfyF0dnbm5OS4uLgkJSUNHTqUVhEqo8LUU01lZaXKJH729vb0oikLC4uwsDA6Y+GgvW8EshZB1tJLRjLciu7ZZ58tKioqLS3Ny8srKSnx9/cfYM8PPvjA09Pzz3/+87Fjx+rr6x8+fLh9+/b//u//3rhxo4FMBat/crl8xYoVeXl5zc3NFy9eXLhwoaWl5ZYtW9gO4eHh5eXl//jHP5qamoqLi9966y31cVw97f/g4GBHR8fz58/rb3sIIYQoFIphw4ap/3LSEy8vr+3bt6s0fvDBB6NHj05OTj58+HBjY2NRUdGLL75YUVGxZcsWOh5AG9qccgsXLhSJRHfu3NFymd0yMzP74osvHj58+Kc//amysrK2tvavf/3rL7/8snPnTjougsc+FJ0XODw8fCBhC6vb87bXQ9bU1PT8889v3rz5mWeeGTVq1IEDB8rLy2NjY7kjT7TPXVxSqfTNN9+8cOFCty/GXk/IdevWDRkyJDk5OSsrq6mp6fr16wsXLlQZGUW98cYbEonk3XffDQ0NffLJJwe+J/XJJFMWQdZC1hKC+gvB3Nw8MDCwsrLyo48+qqmpaW1tPX369LZt27h/pZJqrl69+qc//Un955S//OUvV65caW9vf/DgwYYNGxiGCQ4O1uPGGRBkLQpZS+cZifvrpEHdveDmzZv+/v5SqdTd3X3r1q0Mw6iMw16wYEFPPT/66CNuz5UrV9Ke9K4XHh4eFhYWQ4cODQ8Pz8rKEmoDVfRvZmT1fZKXl6e+7dyWqKgo+rf07g7Xr1+PiIiQyWQSiSQgIODcuXPc5dfV1SUkJLi6ukokkunTp+fn5/v6+tLlLF++nPZR3/+Uv7+/g4NDv2ePJf29ewHDMCtWrBCLxffv36cPVX5C8fX1VV/U4sWLVSYLrqmpSU5OHj16tIWFhVwuj4iIyM7Opk9pv5N7PeWCg4NtbW2VSqWGzaTTSavgzu1LXb58OTIy0s7OztbWNjg4WOVQ8tsnLi7Ozc3t8ePHKu1GdPeCns5bDYfsr3/9K7v/f/75Z5Xzas2aNZqX3C02Wbm5uf34449BQUG2trbdvhg1nJDUrVu3nn/+eTs7Ozqz8+HDh9kxHq+99hq356JFiwghP/zwQz/2G6t/7xcmnLIGMg84g6xl8FnL9O5e0O0Lobq6OjEx0d3d3cLCwtnZ+dVXX01NTVU5CdlUY2NjM3Xq1B9++CEwMNDGxoZdckFBQWJi4n/8x3/Q+9FNmTJl586d7KyMAkLWUoGsxaXnrMV7RjLcim6w4fHdQksq9+szNNq85/WUZerq6tzc3BITE3UWHT8ePXokkUgSEhKEDqRvCgoKRCKRyk3YKCOq6Aaz3bt3d/te2yf6f78w8JQ1wM9GyFo6NfCsZXoVHY/Gjh07cuRIoaPoHbKWCmQtoegiIxnuqEuAfpPL5ZmZmQcOHNi6davQsfSIYZikpCQ7Ozu+pr/Xj5KSkpiYmLS0NF5uwgaC2LZt29KlS4WOAn4HWUt3kLV4VFlZOWTIEO44819//bW4uHjQDqoczJC1+kdHGQkVHRi9xYsXi0QilUuGfHx8Ll68eOzYMfa+qIamqqqqpKQkOztbywmdDMT27dvXrl27du1abmNqaiqdL5jeeggM0K5du+bMmdPU1LRt27ZHjx7NnTtX6IgGNWQtfULW4tejR48SExNLS0tbWlp+/PHH+Ph4Ozu7VatWCR0X6BayFl90lZG4P9hh1KWA9DnqsqfrDA0KMblxKcDF1/E1ivNEQwZ+77339BPDzp07CSFisXj8+PGXLl0a+AL1+X5hFClL/yPnQZ8w6pJ18uTJOXPmjBo1ytLS0tnZecGCBbdv3xY6KK0ga6lA1jJe6sdukM7xOMilpKSkpKQIHQXAYMFoLOr0IyEhISEhQego+gkpC8BwhISEqNxZC9Qha4GeYdQlAAAAAACAsUJFBwAAAAAAYKxQ0QEAAAAAABgrVHQAAAAAAADGChUdAAAAAACA0eJOfEnndQUA0AO+7l4AAKAHPN69AABggHq/ewHqOkFs2rSJEPL2228LHYihiI+PT05O9vPzEzoQ0In4+Hi+FoXzRBB5eXmbN2/G+wULOdy00ePLF2QtQSBrqUDWMl7qGambim7u3Ll6CQZ+59tvvyXY+Rzx8fF+fn7YIaaKx4oO54lQNm/ejD3PQg43bfT48gVZSyjIWlzIWsZLPSPhOjoAAAAAAABjhYoOAAAAAADAWKGiAwAAAAAAMFao6AAAAAAAAIxVfyo6W1tbEcfGjRt5D6t/DDYw0JE9e/awh9vW1lbl2bt3786aNauhoaGmpobt5uPj09bWxu3GfVYkEk2cOFGPW9AHR48eHTNmjFjczWxGVEFBQVRUlL29vUwmCw0NzcnJ0UWf1NRU9YnCUlNT2R04ZcqUfm2fDhlsZjDYwEB3kLW4kLV6YrDJwWADA91B1uIy2IzUn4quqanpp59+IoTMnj2bYZiUlJR+LEQXDDYw0KnPP/+cYZimpiZuY0FBwcSJE8PDw+3s7JycnBiGyc/Pp+3JycncnvTZvLw8R0dHhmEuXryo1+i1UFxcPGvWrLS0tKqqqp76XLhwYerUqTKZ7MaNG3fu3PHw8AgMDDxx4gTvfRYtWpSWlrZq1SruX3344Yf0dijm5uY8bTSfDDYzGGxgoGvIWgRZSyODTQ4GGxjoGrIWMeyMZKyjLm1tbadPny50FEZM1ztQ2APU0NAwc+bMF1544c033+S2W1lZOTo6bt++/ZtvvhEqtn5YtWrV1KlTL126JJPJuu3Q1dX12muv2dvbf/nll66urk5OTp9//rmnp2dCQkJ7ezu/fTw9PQ8ePLh27dqMjAw9bLvJQMoaINNOWQRZC1nL8Aj+ojB2yFrIWvrMSMZa0QFosGHDhsrKytWrV6u0W1tb792718zMLDExsaioSJDY+uGLL75ITU3VMAbg7Nmz165di42NlUgktMXc3Hz+/PmlpaWHDx/mtw8hRKFQxMbGvvPOO0qlkveNBRickLWQtQCMC7KWQWUkVHRgahiG2bVr1+TJk4cPH67+bERExLvvvtvY2BgXF6cyyNtgsXmhJ6dOnSKEqIxKpw+zs7P57UPNmTOnrKzsyJEjfd0WAFCHrEUhawEYC2QtynAyEj8V3aFDh9jr+X799df4+Hh7e3tHR8fo6Oji4mLaZ+PGjbTDiBEj8vPzQ0JCZDKZjY1NUFAQe8ng+++/T/uwvyMfP36ctjg5OXGX09zcnJOTQ5/SUE+rUyqV6enpYWFhLi4uEonE29t7y5YtXV1dhJC6ujruVZvvv/8+7c+2xMbG0oVUV1cnJSWNGjXK0tJy6NChMTExBQUF6rvi1q1bc+fOdXR0pA9ramoGuqMJIYTU1tYuXbrU09PT0tLSwcEhMjLy9OnT9KmB7EADOUADV1hYWFVVpVAoeurw3nvvhYeHX7lyZcmSJRqWo2E/a3PCUxpOFR7dvHmTEDJixAhuo5ubGyGE/XqMrz7UhAkTCCHff/89j1uhT0hZSFmGk7IIstZvkLU0QNZC1kLWMo2sRfGfkRgOOvUKowXudbGs2bNn08bc3NympqasrCyJRDJp0iRuH4VCIZVK/fz8aJ/8/Pzx48dbWlqeOXOG7SOVSqdNm8b9K19fX3olpYY+GgLjyszMJISsW7fu4cOH1dXVn376qZmZWUpKCtshIiLCzMzs9u3b3L/y8/Pbu3cv/X95efkTTzzh7Ox85MiRxsbGq1evBgQEWFtb5+bmquyKgICA06dPNzc3nz9/3tzcvLq6uqeoGIaJjY2NjY3V0IGqqKgYPXq0s7NzZmZmfX39rVu3YmJiRCLRzp072T4D2YG6PkBBQUFDhgzJy8vrdUsJIenp6Zr7fP311+S3q3VVGtetW6fSOT8/Xy6X0/9XV1e7u7sTQvbs2UNb2Kt1KW32c68nvDanSp+4ubmZm5urt4eFhRFCzp8/z2385ZdfCCHPPvssv32o+vp6Qoi/v79KJObm5pMnT9ZmW7Q5vjwuBymL95Sl5fvF4ElZWuZwZC3K6LKWlsdXG8hayFrIWtTgzFoU7xmJ/1GXCQkJfn5+Uqk0NDQ0KioqPz9f5SuT5ubmzz77jPaZOHHinj17Hj9+/NZbb/EeSU8CAwPT0tIcHBycnJyWLFny4osvbtmypaGhgT67dOnSrq6uTz75hO2fk5Nz7969uLg4+jAtLe3u3buffPLJc889Z2tr6+XltX//foZh1L+EWL58eWBgoI2NzeTJk5VKJfvlykCkpaXduXNn8+bN0dHRdnZ2Y8aM2bdvn6ura1JSkobJefpEpweoq6uLnnm8LK1bFRUVhBC5XK6hj5OTU0ZGhoWFRWJiIv1CRYX2+1nDCa/9qaILdCeLRCJd9LGzsxOJRHRXGzukLBZSljo9pCyCrPUbZC0tIWuxkLXUIWsZRdbiPSPxX9FNmjSJ/T8tzcvLy7kdpFIp/amR8vb2Hj58eGFhoX7ybHR0NPt7LqVQKDo6Oq5du0YfhoeHe3t7f/XVV7W1tbTlo48+WrJkiYWFBX146NAhMzOz6OhodgkuLi5eXl6XLl0qKyvjLvkPf/gD7/EfPHiQEBIVFcW2WFlZhYSEtLa28vXTrU4P0JkzZx4+fOjn5zfwRfWEjthmj1dPpkyZsnHjxubm5ri4uNbWVpVntd/PGk547U+VAbK3tyeENDc3cxvpQ/oUj31YYrFYfb8ZI6QsFlKWOj2kLIKs9RtkLS0ha7GQtdQhaxl41mLxm5H4r+i49bqlpSUhhA6eZqlv0rBhwwghDx484D0YdfX19atXr/b29nZwcKBjc5ctW0YIaWlpYfskJye3tLR89tlnhJCioqJTp069/vrr9Kn29vb6+vquri65XM4dC3758mVCCP1dlSWVSvkNnq7d2tpaZWZVZ2dnQkhlZSUvaxH2AA2ctbU1IaSjo6PXnklJSfHx8VevXlWZeLdP+7mnE75Pp8oAPf3004QQlcx1//59QsiYMWP47cNSKpW9XkZsFJCyWEhZQkHWopC1tISsxULWEgqyFmU4GUmAuS5ra2tVfgumZzA9mwkhZmZmjx8/5naoq6tTWYjm3zc1mDlz5po1axYtWlRUVER/mN60aRP57SdRasGCBc7Ozv/4xz/a29s//vjjV155xcHBgT5lZWVlb28vFos7OjrUR7UGBQX1LyotWVlZyeXytra2xsZGbjv9bdrFxYU+HOAOFPYADZyrqyshhA5Q7tWuXbvGjh27e/duOiKc0nI/a6bPU4Uu7dKlS9xG+jAkJITfPlRDQwPDMHRXmzykrH5DytISshaFrMUXZK1+Q9bSErIWZTgZSYCKrq2tjd5Rnvr555/Ly8sVCgW7Va6urrScpSorK+/du6eyEBsbG/ZcHzt27I4dO3pdr1gsvnbtWk5OjouLS1JS0tChQ+mLQf0XTysrqzfeeOPBgwcff/zx3r17VUY2x8TEKJVKdlYiav369SNHjtTDjW7mzJlDCOHOdtre3p6dnS2RSCIiImjLAHegUAeIL+PGjSNqX5D0xNbW9rvvvpNKpfSbQpY2+7lXejtVAgICnnnmmQMHDrBzBHd2du7fv9/d3Z0dzMBXH4qeAHRXmzykrIFAytIGshZB1uIVstZAIGtpA1mLGFpG4tayvMx12drayrYsX76cEPLTTz+xLQqFQi6Xh4SEaJjhh/4s+/e//72xsfH27dtz5851c3NTmeFnxowZcrn83r17ubm5YrH4+vXrGgKjzM3Nb9y4ERwcTAjZsGFDdXV1S0vLqVOnRo4cSQjJysridq6urpZIJCKRSH1RVVVVnp6eHh4eR48eraurq62t3bZtm42NDXfqKvVd0at+zHXZ0NDAzgu0Y8cOts9AdqCuD5Ae5rrs6uoaNmyY+gRQ3PmXVOzZs4cQ0tP8Sz3t515PeG1OlQULFhBCSkpKNG8p1dP8SwzD5OXlWVtbz5s3r6KioqamJjExUSwWHz9+XBd9GIbZt28fIeTgwYMq7cY41yVSVre7olf9mDXOtFPWQGaNQ9Yy/KxlUHNdImt1uyt6haylAlmrW3rIWowOMlJ/KjqVUcsfffRRXl4et2XlypXM739NjoqKon+rUCjc3NyuX78eEREhk8kkEklAQMC5c+e4y6+rq0tISHB1dZVIJNOnT8/Pz/f19aXLWb58Oe1z8+ZNf39/qVTq7u6+devWbgNTd+PGjerq6sTERHd3dwsLC2dn51dffTU1NZU+6+vryw1j0U41xlAAACAASURBVKJFhJAffvhBfQ/Qu2d4eHhYWFgMHTo0PDyczVMqu0Kb/Ulp/25RU1OTnJw8evRoCwsLuVweERGRnZ3Nyw5kdHyAGIbx9/d3cHDQZkpZbd7zus0yDMOsWLFCLBbfv3+fPqyuruYeEZUDTS1evFglV2rYz9qf8BpOFSo4ONjW1lapVGrYTDoTtAru3L7U5cuXIyMj7ezsbG1tg4ODVY4av33i4uLc3NweP36s0m6YFR1Sli5SlvbfAA6SlDWQz0YMspbBZy09V3TIWshayFqmnbV4z0j9/I2u3+h5rNNV8GX37t3dno46wuO7xUAYzgEiA6jo6urq3NzcEhMTdRYdPx49eiSRSBISEoQOpG8KCgpEItE333yj/pRhVnQDYTiviF7pOWXp4f1CG4ZzgAb42QhZS6cGnrX0/xtdvxnOi6JXyFrCQtYSii4ykgDX0RmLbdu2LV26VOgooD/kcnlmZuaBAwe2bt0qdCw9YhgmKSnJzs5uzZo1QsfSByUlJTExMWlpafPmzRM6FvgdpCyjhqylO8haBgtZy6gha/WPjjISKrrf2bVr15w5c5qamrZt2/bo0aO5c+cKHRH0bvHixSKRyNbWltvo4+Nz8eLFY8eOsbc0NTRVVVUlJSXZ2dlaTuhkILZv37527dq1a9dyG1NTU+l8wZ2dnUIFNjghZRkpZC19QtYyKMhaRgpZiy+6ykjcH+x0+nv0Rx99xF0vHQ5raHbu3EkIEYvF48ePv3Tpkj5XLfioS0M7QETH41JAWHwdX92dJ4b2iuiWgClL8PFLhnaABM/hoFNGMerS0F4U3ULWMpwDhKxlvNSPnbifhWDfpaSkpKSk6G11/ZOQkJCQkCB0FMIwigMEoDdG8YpAyhI6CgADYhQvCmQtoaMA04RRlwAAAAAAAMYKFR0AAAAAAICxQkUHAAAAAABgrFDRAQAAAAAAGKtuZkbJyMjQfxxQVlZGsPN/Ly8vT+gQ9KqqqsrJycnc3FzoQIzMYDtPDATd7UhZLORw01ZWVjZixAi+lmbCWau9vb2pqcnR0VHoQLqBrKUCWct4dZORuBNf0nldAQD0gK+7FwAA6AGPdy8AABgglYwkQnIBMBAlJSUnT548efJkVlZWXV2di4uLv79/aGhodHT08OHDhY4OwERkZGTEx8fjvQ9ggB48ePDDDz+cPHnyyJEj9+/fd3JyCgoKCg0NjYyMdHd3Fzo6gMEFFR2Awens7CwoKKDV3dmzZx8/fuzh4UFLu7CwMGtra6EDBDBiqOgA+q21tTUnJ4e+PV2+fNnc3FyhUERHR8+cOdPHx8fMDLMzAAgDFR2AQWtubs7Ly2PfPq2tradNmxYaGhoaGoq3T4B+QEUH0FclJSWZmZmHDx8+d+5cW1sb/ZIxNDQ0IiLCzs5O6OgAABUdgPGorKz897//rT7EJSIi4oknnhA6OgDjgIoOQBsYVAlgRFDRARgl9qK748ePNzY2st+YhoWF2dvbCx0dgOFCRQfQEwyqBDBSqOgAjBv3Dfinn34SiUQTJkyg1d0f//hHS0tLoQMEMCyo6ABUYFAlgLFDRQdgOmpqak6fPn3y5MkTJ078+uuvUqnUz8+PvjE/++yzIpFI6AABhIeKDoBgUCWAaUFFB2Ca2GGZJ0+efPToEXsvhKioKDc3N6GjAxAMKjoYtDCoEsBUoaIDMHE93QshNDR0xowZMplM6AAB9AoVHQw2GFQJYPJQ0QEMIrgXAgAqOhgMMKgSYFBBRQcwSPV0L4Tw8PBRo0YJHR2ArqCiA1OFQZUAgxYqOgDAvRBgEEFFByYGgyoBABUdAPwfpVJ5/vz5w4cP414IYKpQ0YEJwKBKAOBCRQcA3cO9EMAkoaIDI4VBlQDQE1R0ANA73AsBTAYqOjAuGFQJAL1CRQcAfYB7IYCxQ0UHhg+DKgGgT1DRAUA/qdwLwdzcfPLkyTNnzsS9EMCQoaIDw4RBlQDQb6joAIAHVVVVZ8+exb0QwPChogODgkGVADBwqOgAgGc93QshNDTUwcFB6OhgsENFB4LraVDljBkzRo4cKXR0AGB8UNEBgK4olcrCwsKTJ09mZmbm5eXhXghgCFDRgSAwqBIAdAcVHQDoQ21t7alTp3AvBBAcKjrQJwyqBAA9QEUHAPqmci8EZ2fnP/7xj7gXAugHKjrQNQyqBAA9Q0UHAILh3gvh3//+d3t7O+6FALqGig50AYMqAUBAqOgAwCC0tLTk5ubi8xDoGio64BEGVQKAIUBFBwAGh70XwtGjR8vKynAvBOARKjoYIAyqBABDg4oOAAwae9Hd999/39DQgHshwAChooN+wKBKADBkqOgAwDiw90I4efLkmTNnGIZh74Xg7+9vZWUldIBgHFDRgfYwqBIAjAIqOgAwPuy9ELKysu7cuWNjYzN16lTcCwG0gYoONMOgSgAwOqjoAMC49XQvhOeee27EiBFCRwcGBxUdqMOgSgAwaqjoAMBEaLgXAoZIAQsVHbAwqBIATAMqOgAwQd3eCyE0NDQ6Onrq1Kn4xn0wQ0U3yGFQJQCYHlR0AGDiVO6F4OjoGBwcHBoaGhYWNnr0aKGjA31DRTcIYVAlAJg2VHQAMIjgXgiAim7wwKBKABgkUNEBwGCEeyEMWqjoTBsGVQLAIISKDgAGu8bGxjNnzhw+fBj3QjBJDx48+PLLL9mHV65c2bdv34cffsi2DBkyZNGiRUKEBvzAoEoAGORQ0QEA/B92WGZ2dvbDhw9xLwQToFQqXVxcHj16ZGFhof5se3t7YmLitm3b9B8YDBAGVQIAUKjoAAC6gXshmJIlS5Zs3769o6Oj22d/+OGHP/7xj3oOCfoHgyoBANShogMA6EVP90IIDQ0NDAwUi8VCBwi9yM3NnTZtWrdPubi43L9/HwPzDBkGVQIAaIaKDgCgD9ifCI4dO1ZaWiqTyQICAmbOnIl7IRgyhmFGjhxZVlam0m5pafnWW29t2LBBkKhAMwyqBADQEio6AIB+6uleCCEhIUOGDOnr0oqKisaMGaOLOIEQkpaW9vHHH6sPvLx8+bKPj48gIZm227dvu7q6SqXSPv0VBlUCAPQDKjoAgIHi3gvhhx9+6Orq6uu9ECorK0eOHLly5cqVK1diGKcuXLlyRaFQqDR6eHgUFxcLEo8JYxhm+/bt77zzTkZGRlRUVK/9MagSAGCAUNEBAPCpqanp/PnzmZmZmZmZ2t8LYc+ePa+88opIJBo3bty+ffueeeYZPYc9GDz99NO3bt1iH1paWq5cuXL16tUChmR6ysvL//SnP2VlZYlEoiVLlmzevLmnnhhUCQDAF1R0AAC6onIvhGHDhgUEBHR7L4SXX355//79HR0ddIb9NWvWpKSkmJubCxS4aVq7du3f/vY37sDLW7duYaQrj7799ttFixa1tLTQnfzkk0/+8ssv3A4YVAkAoAuo6AAAdK7XeyE4Ozs/ePCA7W9mZjZx4sQ9e/Y89dRTAoZtYu7evTt69Gj6ricSicaPH19QUCB0UCaivr4+JSVl165dItHvPlfcu3fPyckJgyoBAHQKFR0AgF41Nzf/8MMPWVlZJ0+evHr1qpWVlUKh+PHHH1W6WVhYmJmZrV+/PikpqaexmtBXEydOvHz5MsMwYrF4w4YNb7/9ttARmYKsrKyXXnrp4cOHKhPPmJubjx07tqSkpK2tzcvLKzw8PCwsLCAgwMbGRqhQAQBMEio6AADBVFRUZGVl7dix48KFC0qlUr2DSCQKCQn56quv3Nzc9B+e6fn000/feecdpVIpEolKS0uxVweotbX1b3/724YNG0QiUVdXl8qzYrF49OjRK1asCA8PHz58uCARAgAMBqjoAAAEFhkZeeLECfUPxJSFhYWVldXHH3/8+uuv6zkw01NZWenm5sYwzLRp0/79738LHY5xu3Dhwvz580tLS7v9MoKSy+UPHz7E0EoAAJ1CkgUAEJJSqTx79mxP5RwhpKOjo7m5+S9/+UtsbGxNTY0+YzM9Li4uAQEBDMO88sorQsdixJRK5fr166dNm6a5nCOE1NfX//TTT3oLDABgcMJvdAC9yMjIiI+PFzoKAAAA6B4+zcIgh/vYAmglPT1d6BCgF/Hx8cnJyX5+fkIH0jcZGRnfffcdt8XS0tLGxkYmk8nlcrlcbmtrK5PJ2H9tbW09PDx6HcaWl5e3efNmnLfqWltbd+zY8dZbbwkdiLEqLS199OhR02+am5ubmpoaGhoaGhoaGxubm5tbW1s7OzvZ/l5eXrjpH+gOzXVCRwEgMFR0AFqZO3eu0CFAL+Lj4/38/IzuSDk4OCxYsGDIkCGOjo6Ojo5DhgyxsrLiZcmbN282ur2hH2FhYZioQ6eampoe/qa5uXnmzJlCRwSmDBUdACo6AAAhhYWFCR3CoINyTtfoj8m4aTgAgH5gZhQAAAAAAABjhYoOAAAAAADAWKGiAwAAAAAAMFao6AAABq89e/aIfmNra6vy7N27d2fNmtXQ0FBTU8N28/HxaWtr43bjPisSiSZOnKjHLeiDo0ePjhkzRizu8QLygoKCqKgoe3t7mUwWGhqak5Ojiz6pqakDnIMUx0UXfXBcVOjhuHS7z1NTU9mdM2XKFF62BcDkoaIDgEGtqanpqaeeio6OFjoQIX3++ecMwzQ1NXEbCwoKJk6cGB4ebmdn5+TkxDBMfn4+bU9OTub2pM/m5eU5OjoyDHPx4kW9Rq+F4uLiWbNmpaWlVVVV9dTnwoULU6dOlclkN27cuHPnjoeHR2Bg4IkTJ3jvs2jRorS0tFWrVvVvW3BccFx0TW/Hpdt9/uGHHzIMwzCMubk575sGYLIYANCIfoModBR8kkql06ZNM721E0LS09P7+lcNDQ0eHh6RkZG6CEkbOtohWp63X3/9NfmtouOqr68fMWJEYmIitzE/P9/KysrR0ZEQsm/fPpU/YT+hGqD58+d/8MEHHR0dbm5u5ubm6h06Ozu9vLxcXV1bWlpoi1KpHDt2rLu7e1tbG799GIYpKCgQiUT9OF1xXHBc9EBvx4XRuM/Nzc0nT57ca7Sm9x4N0A/4jQ4ABjWZTFZcXHz06FGhAzEsGzZsqKysVL8xtLW19d69e83MzBITE4uKigSJrR+++OKL1NRUDePHzp49e+3atdjYWIlEQlvMzc3nz59fWlp6+PBhfvsQQhQKRWxs7DvvvKNUKvu0ITguOC56oLfjQgawzwGACxUdAAD8DsMwu3btmjx5crf3bYuIiHj33XcbGxvj4uJULhAyWOxnyp6cOnWKEKJyRRN9mJ2dzW8fas6cOWVlZUeOHNF+K3BcKBwXXdPbcaH6sc8BQAUqOgCe1dbWLl261NPT08rKasSIEaGhoV999VVra6vKs5aWlg4ODpGRkadPn6ZPHTp0iL0c/Ndff42Pj7e3t3d0dIyOji4uLtZyFUqlMj09PSwszMXFRSKReHt7b9mypauri/7hxo0bRSJRc3NzTk4OXRH3W9jq6uqkpKRRo0ZZWloOHTo0JiamoKCgT7HpaO26w90u+mFLmy2lGyISiUaMGJGfnx8SEiKTyWxsbIKCgtjr/t9//33aZ/r06bTl+PHjtMXJyUmbHSKgwsLCqqoqhULRU4f33nsvPDz8ypUrS5Ys0bAcXs52/ZwYN2/eJISMGDGC2+jm5kYIYX9a4asPNWHCBELI999/r32QOC4UjovJHBeqH/scAFQJPOoTwOD1aYx+RUXF6NGjXVxcMjMzGxoaKisr16xZQwjZtGkT+6yzs3NmZmZ9ff2tW7diYmJEItHOnTvZJcyePZsQMnv27Nzc3KampqysLIlEMmnSJC1XkZmZSQhZt27dw4cPq6urP/30UzMzs5SUFG6Q3V64VV5e/sQTTzg7Ox85cqSxsfHq1asBAQHW1ta5ubnax6bTtfeK9Os6Ona7Wltbtd9ShmEUCoVUKvXz86N98vPzx48fb2lpeebMGQ0b6+vrq3LxTE/X0QUFBQ0ZMiQvL68fW8QM7Do62rhu3TqVzvn5+XK5nP6/urra3d2dELJnzx7aonJdEC9nOy8nBldP1wWFhYURQs6fP89t/OWXXwghzz77LL99qPr6ekKIv7+/9sHjuFA4LiZzXKie9jmuowPQHl4DAL3o07vFq6++ql5XzJgxg5Zb9NlvvvmGfaqtrW348OESiaSyspK20PfszMxMtk9sbCwhpLq6WptVZGZmBgYGcp9auHChhYVFfX0929JtCfHKK68QQvbu3cu2VFRUWFlZ+fr6si29xqbTtfeK94pOw5YyDEO/kv/pp5/YlitXrhBCFAoF2zKQii4gIMDBwaHfH8UGUtFt2LCBELJ161aVztxPqAzD5OXlWVhYSKXSGzduMGqfUHk523k5Mbj69AmV/pLArouvPiyRSPTkk09qHzyOC4XjwvYx9uPC6nafo6ID0B5GXQLw6eDBg4SQyMhIbuOxY8fo7NX02aioKPYpKyurkJCQ1tZWlQEnkyZNYv9Pv9ktLy/XZhXR0dHsQB1KoVB0dHRcu3ZNc+SHDh0yMzPjTuLv4uLi5eV16dKlsrIyLWPTw9r1ScOWUlKplI4Xory9vYcPH15YWFhRUTHwtZ85c+bhw4d+fn4DX1Rf0QGoFhYWmrtNmTJl48aNzc3NcXFx7LhiFi9nu95ODHt7e0JIc3Mzt5E+pE/x2IclFovV95sGOC4Ujgv7f2M/Lqy+7nMAUIGKDoA37e3t9fX11tbWMplM+2ednZ0JIZWVldxGuVzO/t/S0pIQQq9G07wKQkh9ff3q1au9vb0dHBzoRRfLli0jhLS0tPQaeVdXl1wu59769vLly4QQOk6m19j0s3Z90rCllPrnkmHDhhFCHjx4oPvodMja2poQ0tHR0WvPpKSk+Pj4q1evvvnmm9x2Hs92/ZwYTz/9NCFE5VPv/fv3CSFjxozhtw9LqVT2OgUFF44LhePC/t/Yjwurr/scAFSgogPgjZWVlVwub2tra2xs1P5ZegtXFxeXga+CEDJz5sw1a9YsWrSoqKioq6uLYZhNmzYRQhiGYfuIRCL1xdrb24vF4o6ODvWf8oOCgrSJTfC1619tbS1308hvtRyt6wghZmZmjx8/5naoq6tTWYj6DhGcq6srIYRe3NKrXbt2jR07dvfu3XQAJ8XX2a63E4Mu7dKlS9xG+jAkJITfPlRDQwPDMHRXawnHhcJx6ZYxHheqH/scAFSgogPg05w5cwghKjc38/Hxefvtt9lnuXM0t7e3Z2dnSySSiIiIga+is7MzJyfHxcUlKSlp6NChtFRQH8piY2PDlhljx47dsWMHISQmJkapVLJTNVLr168fOXKklrcJEnbtgmhra8vPz2cf/vzzz+Xl5QqFgv1o4urqSr+TpiorK+/du6eykG53iLDGjRtH1L5c74mtre13330nlUo/++wzbjsvZ7veToyAgIBnnnnmwIED7PzynZ2d+/fvd3d3ZwfC8dWHoicG3dVawnEhOC4aGd1xofqxzwFAlfaX3AEMTv2Y69LV1fXw4cMNDQ2lpaWLFy92dna+e/cu8/vZzBoaGtjZzHbs2MEuQX2WjuXLlxPODByaVxEcHEwI2bBhQ3V1dUtLy6lTp0aOHEkIycrKYhc4Y8YMuVx+79693NxcsVh8/fp1hmGqqqo8PT09PDyOHj1aV1dXW1u7bds2Gxsb7lwjvcam07X3ivA9M4qGLWUYRqFQyOXykJAQDXNd0rFVf//73xsbG2/fvj137lw3NzeVmVG63SGMoHNddnV1DRs2TH2+FpWZHrj27NlDCOlp7r5+n+3anBgLFiwghJSUlPS6sUzPMz0wDJOXl2dtbT1v3ryKioqamprExESxWHz8+HFd9GEYZt++fYSQgwcPar8hOC44LqZ3XJju9jmFmVEAtIfXAEAv+vpuUVNTk5ycPHr0aAsLC1dX13nz5hUVFXX7rFwuj4iIyM7Opk/l5eVxv21ZuXIl8/sRfVFRUb2uorq6OjEx0d3d3cLCwtnZ+dVXX01NTaV/zk4vdvPmTX9/f6lU6u7uzp2fjd4NycPDw8LCYujQoeHh4WwlpmVsOlq7lvpR0dHZCFgLFizQ/igoFAo3N7fr169HRETIZDKJRBIQEHDu3Dnu8uvq6hISElxdXSUSyfTp0/Pz8319felyli9frnmH+Pv7CzXXJcMwK1asEIvF9+/fpw+rq6u5O6HbqfMWL16sUqzycrb3emIEBwfb2toqlUoNm0nvq6GCOy88dfny5cjISDs7O1tb2+DgYJWjyW+fuLg4Nze3x48f92lDcFxwXEzvuKjvcwoVHYD28BoA6AXeLYwF6e9vdP1DKzq9ra6vBljR1dXVubm5JSYm6iY63jx69EgikSQkJAgdSN8UFBSIRCLuVPVabgiOi07huOif+j5noaID0B6uowMAAFVyuTwzM/PAgQNbt24VOpYeMQyTlJRkZ2e3Zs0aoWPpg5KSkpiYmLS0tHnz5tEW7TcEx0V3cFz0T32fA0D/oKIDABjsFi9eLBKJbG1tuY0+Pj4XL148duxYQ0ODUIFpVlVVVVJSkp2dreVkgAZi+/bta9euXbt2LdvSpw3BcdERHBf9U9/nhJDU1FR634XOzk6hAgMwOiLm90OxAUBFRkZGfHw8XimGTyQSpaenz507V9cr2rhxI73PHrVy5cr3339f1yvtK5y3ADAYINcBEELEQgcAAGBkUlJSUlJShI4CAAAAgBCMugQAAAAAADBeqOgAAAAAAACMFSo6AAAAAAAAY4WKDgAAAAAAwFhhZhQArWRkZAgdAvQuLy9P6BAMBd0VOG8BwLQh7QMQ3L0AoFd0ZmShowAAAIDu4dMsDHL4jQ5AK3i3MHx6ux+dUcA9mgBgMMC3rgAE19EBAAAAAAAYL1R0AAAAAAAAxgoVHQAAAAAAgLFCRQcAAAAAAGCsUNEBAAAAAAAYK1R0ADzLz89/9dVXR48eLZFIhgwZMm7cuBdeeOHzzz8vLi4WOjQAMD53796dNWtWQ0NDTU2N6Dc+Pj5tbW3cbtxnRSLRxIkThQq4W48ePdq2bVtwcPCQIUMkEslTTz21YMGCwsLCbjsfPXp0zJgxYnGP03EXFBRERUXZ29vLZLLQ0NCcnJy+9klNTU1PTx/gRgEAGAhUdAC86erqWrZs2dSpU4cNG3bs2LG6urobN25s2rSpoaHhjTfeePLJJ5VKpdAxAoAxKSgomDhxYnh4uJ2dnZOTE8Mw+fn5tD05OZnbkz6bl5fn6OjIMMzFixcFCrl7y5YtW7JkyezZs69fv15bW7t79+6CggJfX99Dhw5xuxUXF8+aNSstLa2qqqqnRV24cGHq1KkymezGjRt37tzx8PAIDAw8ceJEn/osWrQoLS1t1apVvG8pAIAAGADQiH6Pq03PFStWEEJ27Nih0q5UKiMjIwkhHR0dOgiwn6RS6bRp00xp7YSQ9PR0fpfZE13vvYEvX/vzFgxWfX39iBEjEhMTuY35+flWVlaOjo6EkH379qn8CVvRGZrXXnvt9ddf57YUFBQQQp566ilu4/z58z/44IOOjg43Nzdzc3P15XR2dnp5ebm6ura0tNAWpVI5duxYd3f3trY27fvQAOhNLHncTNA/5DoAhmHwGx0AP27evPnhhx/6+vouWrRI5Slzc3N8EwwAfbVhw4bKysrVq1ertFtbW+/du9fMzCwxMbGoqEiQ2Ppq165d27dv57YoFAqJRFJcXMwwDNv4xRdfpKamahhvefbs2WvXrsXGxkokEtpibm4+f/780tLSw4cPa9+HBhAbG/vOO+9g9AQAGDtUdAD82LFjR1dXV1xcXLfP+vn5MQyj4WMKAAAXwzC7du2aPHny8OHD1Z+NiIh49913Gxsb4+LiVC6oMxbNzc2tra3jxo0TiURsI1uD9eTUqVOEEJWrBOnD7Oxs7ftQc+bMKSsrO3LkSL+3AgDAEKCiA+DH2bNnCSHjx4/XpnNtbe3SpUs9PT0tLS0dHBwiIyNPnz5Nnzp06BA7t8Gvv/4aHx9vb2/v6OgYHR2tMrcKuxArK6sRI0aEhoZ+9dVXra2thBClUpmenh4WFubi4iKRSLy9vbds2dLV1UX/cOPGjSKRqLm5OScnh66IW2pWV1cnJSWNGjXK0tJy6NChMTExdHCU9rHpaO180bDz33//fRrS9OnTacvx48dpi5OTk+b4abtIJBoxYkR+fn5ISIhMJrOxsQkKCmKnZBjI8mGwKSwsrKqqUigUPXV47733wsPDr1y5smTJEg3L4SXb6OKF+e233xJCVq5c2ae/unnzJiFkxIgR3EY3NzdCCPtzpTZ9qAkTJhBCvv/++z7GDgBgYAQe9Qlg8LQco+/q6koIuXDhQq89KyoqRo8e7ezsnJmZWV9ff+vWrZiYGJFItHPnTrbP7NmzCSGzZ8/Ozc1tamrKysqSSCSTJk1SWYiLi0tmZmZDQ0NlZeWaNWsIIZs2bWIYJjMzkxCybt26hw8fVldXf/rpp2ZmZikpKdwwur1Sq7y8/IknnnB2dj5y5EhjY+PVq1cDAgKsra1zc3O1j02na9eAaHEdnTY7Xz02X19flWuTerrOTaFQSKVSPz8/unPy8/PHjx9vaWl55swZXpYfFBQ0ZMiQvLw8zZvJ4NoS4/f111/T15FKe35+vlwup/+vrq52d3cnhOzZs4e2qFxHx0u2GeALs1uVlZXOzs4JCQk9dejpOrqwsDBCyPnz57mNv/zyCyHk2Wef1b4PVV9fTwjx9/fv94aA4JDrABiGwWsAoBd9quh+/PHHXnu++uqrhJBvvvmGbWlraxs+fLhEIqmsrKQt9DNWZmYm2yc2NpYQUl1dzV2ISgEzY8YMtqILDAzkPrVw4UILC4v6+nq2pdua4ZVXXiGE7N27l22pqKiwsrLy9fVlW3qNTadr10Cbik6bnT/Aio4Q8tNPP7EtV65cIYQoFAoNf6v98gMCmlVkPAAAIABJREFUAhwcHLT5JI1POcZuw4YNhJCtW7eqtHMrOoZh8vLyLCwspFLpjRs3GLWKjpdsM8AXprqampoJEybEx8crlcqe+vSpoqO/vLHxaNOHJRKJnnzyyf5tCBgC5DoABjOjAPCFXutSU1PTa8+DBw8SQqKiotgWKyurkJCQ1tZWlcE/kyZNYv9Pv4kvLy/nLoROock6duwYndA8OjqaHVhFKRSKjo6Oa9euaY7t0KFDZmZm0dHRbIuLi4uXl9elS5fKysq0jE0Pa+837Xd+v0mlUjqUi/L29h4+fHhhYWFFRcXAF37mzJmHDx/6+fkNfFFg4OjVcRYWFpq7TZkyZePGjc3NzXFxcXTQNRcv2YbfF2Zzc3NERMQzzzyzd+9ec3Pzvv65vb09XYjKMtmntOzDEovF6vsNAMC4oKID4EdAQAAhhP4go0F7e3t9fb21tbVMJuO2Ozs7E0IqKyu5jXK5nP2/paUlIYRejdbTQlj19fWrV6/29vZ2cHCgF8ksW7aMENLS0tJrbF1dXXK5nHur4suXLxNC6JilXmPTz9r7p087v9/UPzIOGzaMEPLgwQNelg+DhLW1NSGko6Oj155JSUnx8fFXr1598803ue08Zhu+XphKpTIuLs7Nze1//ud/+lHOEUKefvppQohKJXn//n1CyJgxY7Tvww2p1+lYAAAMHCo6AH4kJiaKxeIDBw50++x//ud/mpmZ3bx508rKSi6Xt7W1NTY2cjvQ2+m6uLhos66eFsKaOXPmmjVrFi1aVFRU1NXVxTDMpk2bCCEMZ5Zw7vxy7GLt7e3FYnG3980LCgrSJjbB166BljvfzMzs8ePH3A51dXUqi1KPn1VbW8vdUvJbLUfruoEvHwYJOpCbXujVq127do0dO3b37t306juKr2zD4wszMTGxvb09IyODne/nySefPH/+vPZLoGu8dOkSt5E+DAkJ0b4P1dDQwPw2Zh4AwHihogPgx5gxY957772LFy/u3r1b5albt25t37597ty59JvjOXPmEEK482W3t7dnZ2dLJJKIiAgtV0cXcvToUW6jj4/P22+/3dnZmZOT4+LikpSUNHToUFobqA8rsrGxYeuKsWPH7tixgxASExOjVCrZuRmp9evXjxw5UstbNgm79l5ps/NdXV3p1/lUZWXlvXv3tImfamtry8/PZx/+/PPP5eXlCoWC/dQ4wOXDIDFu3Dii9kNTT2xtbb/77jupVPrZZ59x23nJNny9MP/rv/7r2rVr/+///T8rKyvt/0pFQEDAM888c+DAAfaeDZ2dnfv373d3d2cHl2rTh6KvRLqrAQCMGJ8X5QGYoj5ddZ2ammphYbF8+fJbt261t7eXlZXt2rXL1dV1+vTpTU1NtA939rmGhgZ29rkdO3awy6FzFbS2trIty5cvJ5wpN+hCXF1dDx8+3NDQUFpaunjxYmdn57t37zIMExwcTAjZsGFDdXV1S0vLqVOnRo4cSQjJyspiFzhjxgy5XH7v3r3c3FyxWHz9+nWGYaqqqjw9PT08PI4ePVpXV1dbW7tt2zYbGxvujCO9xqbTtWtA+jjXZU87nw5d+/vf/97Y2Hj79u25c+e6ubmpzFzSbfwMwygUCrlcHhISomGuy4EsH3NdDh5dXV3Dhg1TnyBHZWYUrj179hBCeprrst/ZRpsX5oIFCwghJSUlPW3Ol19+2dPnkG7P555mRmEYJi8vz9raet68eRUVFTU1NXR8xPHjx/vah2GYffv2EUIOHjzYU9hg+JDrABjMdQnQq76+W/z4448vvfSSu7u7hYWFTCabMmXKli1b2tvbuX1qamqSk5NHjx5tYWEhl8sjIiKys7PpU3l5edzPOitXrmR+P4QvKipKfSGurq7z5s0rKiqiT1VXVycmJtIYnJ2dX3311dTUVPrn7FRvN2/e9Pf3l0ql7u7u3Pn06N2rPDw8LCwshg4dGh4ezlZiWsamo7X3SpuKTvPOp+rq6hISElxdXSUSyfTp0/Pz8319fWn8y5cv1xy/QqFwc3O7fv16RESETCaTSCQBAQHnzp3ja/n+/v6Y63LwWLFihVgsvn//Pn1YXV3Nfbl1O9Xk4sWLVb4d4CXb9PrCDA4OtrW11TB3pcqPYz1VdPTeJyq491qgLl++HBkZaWdnZ2trGxwcrPIS074Pvajv8ePHPYUNhg+5DoBhGBHz+/QNACoyMjLi4+PxSjF8IpEoPT197ty5AsYwYcKEmpoavmbmHAictyagvr7ey8srOjp627ZtQseiSV1d3fDhwxcsWLBz506hY+mDwsJCHx+fffv2zZs3T+hYoP+Q6wAIrqMDAAAwTHK5PDMz88CBA1u3bhU6lh4xDJOUlGRnZ7dmzRqhY+mDkpKSmJiYtLQ0lHMAYAJQ0QEAABgoHx+fixcvHjt2rKGhQehYuldVVVVSUpKdna3l5JkGYvv27WvXrl27dq3QgQAA8AAVHQAADzZu3CgSiQoLC+/fvy8Sid59912hIwITMWrUqMOHD9vZ2QkdSPdcXFzOnTvn5eUldCB9s379evw6BwAmQyx0AAAApiAlJSUlJUXoKAAAAGDQwW90AAAAAAAAxgoVHQAAAAAAgLFCRQcAAAAAAGCsUNEBAAAAAAAYK8yMAqCVuLg4oUOA3m3atOnbb78VOgqDQO9yjvMWAEwbzXUAg5yIYRihYwAwaHl5eZ988onQUQAAP6qqqq5evRoSEiJ0IADAG3yXB4McKjoAABhEMjIy4uPj8d4HAAAmA9fRAQAAAAAAGCtUdAAAAAAAAMYKFR0AAAAAAICxQkUHAAAAAABgrFDRAQAAAAAAGCtUdAAAAAAAAMYKFR0AAAAAAICxQkUHAAAAAABgrFDRAQAAAAAAGCtUdAAAAAAAAMYKFR0AAAAAAICxQkUHAAAAAABgrFDRAQAAAAAAGCtUdAAAAAAAAMYKFR0AAAAAAICxQkUHAAAAAABgrFDRAQAAAAAAGCtUdAAAAAAAAMYKFR0AAAAAAICxQkUHAAAAAABgrFDRAQAAAAAAGCtUdAAAAAAAAMYKFR0AAAAAAICxQkUHAAAAAABgrFDRAQAAAAAAGCtUdAAAAAAAAMYKFR0AAAAAAICxQkUHAAAAAABgrFDRAQAAAAAAGCtUdAAAAAAAAMYKFR0AAAAAAICxQkUHAAAAAABgrFDRAQAAAAAAGCsRwzBCxwAAAKAr5eXl0dHRHR0d9GFLS0ttba27uzvbwcfH55///KdA0QEAAAyUWOgAAAAAdGj48OGPHz++du0at7G+vp79/7x58/QeFAAAAG8w6hIAAEzcyy+/LBZ3/w2mSCR68cUX9RwPAAAAjzDqEgAATFxpaekTTzyh/n4nEol8fX3z8/MFiQoAAIAX+I0OAABMnLu7+5QpU8zMVN/yzM3NX375ZUFCAgAA4AsqOgAAMH0vvfSSSCRSaezq6po7d64g8QAAAPAFFR0AAJi+uLg4lRZzc/PAwEBnZ2dB4gEAAOALKjoAADB9Tk5OISEh5ubm3MaXXnpJqHgAAAD4gooOAAAGhYULF3InRzEzM5szZ46A8QAAAPACFR0AAAwKzz//vIWFBf2/WCyOioqSy+XChgQAADBwqOgAAGBQkMlkM2fOpEVdZ2fnwoULhY4IAACAB6joAABgsFiwYIFSqSSESCSS5557TuhwAAAAeICKDgAABovIyEipVEoIiY2NlUgkQocDAADAA7HQAQAA/E5eXl5paanQUYDJmjRp0unTp93d3TMyMoSOBUzW1KlTR4wYIXQUADBYiLgTfwEACC4uLu7AgQNCRwEA0H/p6em4eT0A6A1+owMAgxMbG/vtt98KHQXoW0ZGRnx8vK6/Z+zq6lq/fn1aWppO18ILeld0vBaMjkgkEjoEABhccB0dAAAMImZmZsuWLRM6CgAAAN6gogMAgMFFLMb4FAAAMB2o6AAAAAAAAIwVKjoAAAAAAABjhYoOAAAAAADAWKGiAwAjlp6ePmHCBIlEIhKJRCLR1atXhY6IZxs3bqSbZhT3tnr06NG2bduCg4OHDBkikUieeuqpBQsWFBYWCh3XYHH37t1Zs2Y1NDTU1NSIfuPj49PW1sbtxn1WJBJNnDhRqIC71aez6OjRo2PGjNFwYWRBQUFUVJS9vb1MJgsNDc3Jyelrn9TU1PT09AFuFACATqGiAwBjlZOTM3/+/PDw8Orq6tu3bxtFzdNXKSkpDMMoFAqhA9HKsmXLlixZMnv27OvXr9fW1u7evbugoMDX1/fQoUM6XW9TU9NTTz0VHR2t07UYuIKCgokTJ4aHh9vZ2Tk5OTEMk5+fT9uTk5O5PemzeXl5jo6ODMNcvHhRoJC7p+VZVFxcPGvWrLS0tKqqqp4WdeHChalTp8pkshs3bty5c8fDwyMwMPDEiRN96rNo0aK0tLRVq1bxvqUAAHxBRQcAxurbb79lGOatt96ytbX19PQsLS0dN27cwBdra2s7ffr0gS9ncPrzn//81ltvubi42NjY+Pv779u3r7Oz8z//8z91ulKGYbq6urq6unS6Fg0EP2caGhpmzpz5wgsvvPnmm9x2KysrR0fH7du3f/PNN0LF1g/anEWrVq2aOnXqpUuXZDJZtwvp6up67bXX7O3tv/zyS1dXVycnp88//9zT0zMhIaG9vV37Pp6engcPHly7dm1GRoZOtxoAoN9Q0QGAsSotLSWEODo6Ch0I/P927dq1fft2botCoZBIJMXFxTq9b7hMJisuLj569KjuVmHgNmzYUFlZuXr1apV2a2vrvXv3mpmZJSYmFhUVCRJbX2l5Fn3xxRepqakaxluePXv22rVrsbGxEomEtpibm8+fP7+0tPTw4cPa9/n/2rvTsCiufH/gp4GmaRpoEJUtGlFGSJS0BI2gMihEWqNIZECcwaiTQflLEmBcxiVqeOIWDIlhEhNFwjiocYl51OAWDaL3sjgBFRAVXEhikEUW2deW+r8499btNNAU0HQ18P28ok+fPvWrw+lD/aiqUzSAgICANWvWKBQKTe0mAIAGIaMDgIHq+fPnfIcA3WhoaGhqapo4caJAIOA7lkGLYZj4+PipU6fa2tp2fFcul2/evLmuri4wMFDlhrqBotNRxOZgXbly5QohROUuQfoyOTmZex1q4cKFRUVF586d6/VeAAD0H2R0ADDwnD59WiAQnDlzhhBCl0Vxc3Ojb5WXl4eHh48ZM8bQ0HDEiBH+/v7Z2dnsBxUKxfHjx2fPnm1tbS0Wi52dnWNjY9mr9egyJA0NDWlpaXTRCHoGYPv27fQle2XdxYsXacnw4cOVQ6IKCgoWLVpkaWlJX1ZUVHQbWI+o2Yvq6mrlRS+2b99O67MlAQEB3XZUt/vC3bfffksIef/993u3p1woR0szFuWSX375JSgoyNzc3NLScv78+Y8ePaKfUl5yJjMz09vb29TU1NjYeNasWezCGFx+712NGW3KyckpKytTc7PlBx984OPjk5ub+95776lpp7KycvXq1ePGjTM0NLSwsJg7d25KSgp9i0uXUhoc56zejaL8/HxCiMrttXZ2doQQ9nQllzrUpEmTCCE//PBDD2MHANAKBgBAlwQEBAQEBHCp6efnRwhpampiS4qLi1988UUrK6tz587V1dXl5eV5enoaGRmlp6fTCklJSYSQnTt3VlVVlZeX//Of/9TT06Orj7AkEsn06dM7bq5juaurK11bQiUkT0/PlJSUhoaG69ev6+vrl5eXdxuYejKZzM7Ojn3Z7V7I5XI9Pb2HDx8qN+Lu7n7kyBGOHaVmX7gETJWWllpZWYWEhHCsT1cU5N6+so6DgZb4+fmlp6fX19dfvnxZLBZPmTJF+VMymUwikbi7u9M6mZmZr7zyiqGh4dWrV9k6XH7vXY2ZWbNmDRs2LCMjo3c7xfG7cOjQIToeVMozMzOlUin9uby8fNSoUYSQw4cP0xJ2ZRSqpKTE3t7eysoqKSmppqamoKDA399fIBAcOHCArdNtl/ZxnHeq21FkZ2enr6/fsXz27NmEkOvXrysXPnjwgBDy6quvcq9D1dTUEEI8PDy4xEwIOX78OJeaAAAagYwOAHRLXzK6ZcuWEULYvIVhmJKSEpFI5OrqSl8mJSXNnDlTuZElS5YIhcKamhq2pO8Z3fnz51U+221g6nXM6NTvBT2TEBYWxlZITU21s7NrbW3lHk9X+8JRRUXFpEmTgoKCFAoFx4/0R0aXlJTEltDzk8pJKT2vdevWLbYkNzeXECKTydiSvmR0np6eFhYWvc5nOH4Xdu/eTQjZu3evSrlyRscwTEZGhlAolEgk9+7dYzpkdMuXLyeEHD16lC1pbm62tbUVi8WlpaW0pNsu7eM474jLKOpRRkfPvLHxcKnDEggEDg4OXMJGRgcAWoarLgFg8Dh9+rSenp7yKvbW1tYTJky4ceNGUVERIWT+/PnshWSUTCZra2u7c+eOBsN47bXXehpYj3S7Fz4+Ps7OzgcPHqysrKQlH3/88XvvvScUCnsaT8d94aKhoUEul7/88stHjhzR19fvRQuaMmXKFPZnepKquLhYuYJEIqEX1FHOzs62trY5OTklJSV93/rVq1erqqrc3d373pQa9FpT9pfbFTc3t5iYmIaGhsDAwKamJpV3T506RQiZN28eWyISiby9vZuamlQuNVTTpZod530cRebm5rQRlTbZtzjWYRkYGHTsNwAAXYCMDgAGiZaWlpqamvb2dqlUqnwv2c2bNwkh9EqqmpqarVu3Ojs7W1hY0HfXrVtHCGlsbNRgJBKJpKeB9QiXvYiMjGxsbPzyyy8JIffv379y5crKlSt7EY/KvnChUCgCAwPt7Oz+/e9/85vOEUKkUin7s6GhISFE5SEHHQ/cR44cSQh5+vRp/0enGUZGRoSQtra2bmuGh4cHBQXl5eWpPOSADgkjIyOVJwFYWVkRQkpLS5ULu+pSzY7zvo8iJycnQohKJvnkyRNCyPjx47nXUQ6p2+VYAAB4gYwOAAYJkUhkbm5uYGDQ1tbW8YKEWbNmEUJ8fX23bdu2YsWK+/fvt7e3MwyzZ88eQgijtCq6oItVGfX09FpbW5VLqqurNRVYj3DZi+DgYCsrqy+++KKlpeWTTz5ZtmyZhYVFP8WjIjQ0tKWl5cSJE+waIQ4ODtevX+9js/2ksrKS+f2TFWguR/M6wu333tWY0Q4bGxtCCL3Rq1vx8fGOjo4JCQn07jtKJBJJpdLm5ua6ujrlyvTh3dbW1lxa1uy46vsoolu8ceOGciF96e3tzb0OVVtbyzAM7WoAAF2DjA4ABg9/f3+FQsGuVUhFR0ePHj1aoVA8f/48LS3N2to6PDx8xIgR9Ci842VUxsbG7BG8o6NjXFwc/dnGxob+854qLS19/PixRgLryS4SjnshEonCwsKePn36ySefHDlyJCIiop/iUREVFXXnzp0zZ86IRKK+tKM1zc3NmZmZ7Mvbt28XFxfLZDL22J3L772rMaMdEydOJB1ONHXFxMTku+++k0gk9Pwta+HChYQQ5dX5W1pakpOTxWKxXC7nGImmxpVGRpGnp+fLL7988uRJ9pkNz58/P3bs2KhRo9iLS7nUoegYoF0NAKBz+nITHgCAxvVlZZSysrJx48aNHTv2/Pnz1dXVlZWV+/btMzY2Zlcp8PLyIoTs3r27vLy8sbHxypUro0ePJoRcvnyZbWTOnDlSqfTx48fp6ekGBgZ3796l5fRCtc8//7yuru7hw4eLFi2ys7PrdGUU5ZA4BqaeysooXPaCYZjy8nL6aAc/P79exNPVvqjxr3/9q6u/NVzWe+yPlVGUS9avX09+vw6KTCaTSqXe3t5q1rrk8nvvasxoZ63L9vb2kSNHdlyaRWVlFGWHDx8mhHS11mVtbS271mVcXBxbp9su5TKugoODCSGFhYVd7U5PR1FXK6MwDJORkWFkZLR48eKSkpKKiorQ0FADA4OLFy/2tA7DMN988w0h5NSpU12FrYxgZRQA0C5kdACgW7gcxdJVHDo91KPP1Bo7dqxQKBwxYoSPj49ynlNeXh4aGjpq1CihUGhlZbV8+fINGzbQFtil7fLz8z08PCQSyahRo5TXD6yurg4JCbGxsRGLxTNmzMjMzHR1daWfXb9+fUZGhvr/l6kPrCsff/yxcpvvv/8+x72gVqxYQQi5du1ax5bVxNPtvnRF5bSGdjI6lcEQHBysEj/tNOWSefPm0c/SVPnu3btyudzU1FQsFnt6eqampiq3r/73Tut0NWY8PDy0sNYlwzCbNm0yMDB48uQJfVleXq68v50uNblq1SqVvLSioiIyMtLe3l4oFEqlUrlcnpycTN/i3qXdjnMvLy8TExM1a1dyHEX0GR4qlJ+1QN28eXPu3LlmZmYmJiZeXl4qv1zudehNfexqseoRZHQAoF0C5veTMgAAvwIDA8n/PlMY+uhf//rX3r17s7Ky+A6EkxMnTgQFBWnzr9KkSZMqKip6sQyjdnD/LtTU1EyYMGH+/Pn79u3r/7h6r7q62tbWNjg4+MCBA3zH0gM5OTkuLi7ffPPN4sWLudQXCATHjx9ftGhRfwcGAEDhPjoAgEFr3759q1ev5jsK6HdSqTQpKenkyZN79+7lO5YuMQwTHh5uZma2bds2vmPpgcLCQn9//40bN3JM5wAAtA8ZHQDAoBIfH79w4cL6+vp9+/Y9e/YMJwqGCBcXl6ysrAsXLtTW1vIdS+fKysoKCwuTk5M5Lp6pI/bv379jx44dO3bwHQgAQJeQ0QEA8EzQtaioqF40ePr0aQsLi6+++urYsWPs4u+6GSpfYmJiBAJBTk7OkydPBALB5s2b+Y5IA8aMGXP27FkzMzO+A+mctbV1amrqhAkT+A6kZ6Kjo3F2DgB0nCb/0gMAQC9o9s6xkJCQkJAQDTaobNDcer127dq1a9fyHQUAAIAG4BwdAAAAAADAQIWMDgAAAAAAYKBCRgcAAAAAADBQIaMDAAAAAAAYqLAyCgDonOvXr9NnK8OQQp/0jV896/r16wQdAgAA3cE5OgAAAAAAgIEK5+gAQOe4ubl9++23fEcB2nbixImgoCD86ln07Bw6ZMARCAR8hwAAQwvO0QEAAAAAAAxUyOgAAAAAAAAGKmR0AAAAAAAAAxUyOgAAAAAAgIEKGR0AQC+ZmJgIlOjp6VlYWMhksrCwsBs3bvAdHUAnfv311wULFtTW1lZUVLBD18XFpbm5Wbma8rsCgWDy5Ml8BaxGW1vbnj17XF1dTU1NR44cOXfu3KSkJIZhOq28YMECgUCwfft25cINGzYcP35cK8ECAPQjZHQAAL1UX19/69YtQoifnx/DMG1tbfn5+R9++GF+fv7kyZP/+te/NjY28h0jwP/Jzs6ePHmyj4+PmZnZ8OHDGYbJzMyk5ZGRkco16bsZGRmWlpYMw2RlZfEUcpcaGhq8vLwOHjy4Z8+ep0+fZmVlmZiYLFiw4M6dOx0rJyYmJiUldSxfsWLFxo0bt2zZ0v/xAgD0I2R0AACaoa+vb2Vl5efnd+XKlX/84x8HDx7885//3NUZA+CdiYnJjBkzBm77PVVbW+vr6/unP/3p3XffVS4XiUSWlpb79+8/evQoX7H1wrp163Jzcy9duvTHP/5RLBaPHj364MGDIpGoY83i4uLIyMi33nqr41vjxo07derUjh07Tpw40f8hAwD0F2R0AACa99FHH02dOvX7778/duwY37EAEELI7t27S0tLt27dqlJuZGR05MgRPT290NDQ+/fv8xJbT5WVlcXFxQUHB1tZWbGFEomkubl54sSJKpVXrFgRGBjo4+PTaVMymSwgIGDNmjUKhaIfIwYA6E/I6AAANE8gENAzIV9++SXfsQAQhmHi4+OnTp1qa2vb8V25XL558+a6urrAwECVG+p00/fff//8+XMup0ATEhLu3LkTExOjps7ChQuLiorOnTunuQABALQKGR0AQL+gh5vXr19va2ujJeXl5eHh4WPGjDE0NBwxYoS/v392djZ96/Tp0+wqFL/88ktQUJC5ubmlpeX8+fMfPXrEttnS0rJ161YnJydjY+Nhw4b5+vrSQ1u2gppNDCaVlZWrV68eN26coaGhhYXF3LlzU1JS6Fvbt2+n3cge7l+8eJGWDB8+nJbExMQIBIKGhoa0tDT6loGBAVsuEAheeOGFzMxMb29vU1NTY2PjWbNmpaWl9b19HuXk5JSVlclksq4qfPDBBz4+Prm5ue+9956adtT0PMcxTDQxSm/evEkIsbCwWLNmzahRowwNDV988cXw8PCqqirlakVFRWvWrElISDA1NVXT2qRJkwghP/zwQ49iAADQIQwAgC4JCAgICAjgOwqulFdGUdHU1ESn2eLiYoZhiouLX3zxRSsrq3PnztXV1eXl5Xl6ehoZGaWnp7Mf8fPzo62lp6fX19dfvnxZLBZPmTKFrRASEiKVSi9dutTY2FhaWrp27VpCSEpKCn2XyyZ0GV11sNtqJSUl9vb2VlZWSUlJNTU1BQUF/v7+AoHgwIEDbB2JRDJ9+nTlT7m6utJFPtTUoWQymUQicXd3p7+FzMzMV155xdDQ8OrVqxppf9asWcOGDcvIyOh2TzX4XTh06BAhZOfOnSrlmZmZUqmU/lxeXj5q1ChCyOHDh2kJuzIKxaXnux3DGhmldCvW1tbBwcGPHj169uzZv//9b4lEMn78+OrqaraaXC4PCwtT7oFt27Z1bK2mpoYQ4uHhwT0A9Qghx48f11RrAADdQkYHALpl0GR07EKXNKNbtmwZIeTIkSNshZKSEpFI5OrqypbQ41S6AjsVEBBACCkvL6cv7e3tp02bpryV8ePHsxkdl03oMo4Z3fLlywkhR48eZUuam5ttbW3FYnFpaSkt6WNGRwi5desWW5Kbm0sIkclkaj7LvX1PT08LCwsuCYwGvwu7d+8mhOzdu1elXDmjYxgmIyNDKBRKJJJ79+4xHTIGnAh7AAAgAElEQVQ6Lj3f7RjWyCiVy+WEEHt7+7a2NraQPplgy5Yt9GVcXNzYsWPr6+vpSzUZHcMwAoHAwcGBewDqIaMDAC3DVZcAAP2ipKSEECIUCunFeKdPn9bT05s/fz5bwdraesKECTdu3CgqKlL+4JQpU9if6TmT4uJi+nLOnDnp6ekrV668fv06vdiyoKBg5syZ9F3umxjQTp06RQiZN28eWyISiby9vZuamjR14ZxEIqFX4lHOzs62trY5OTn0d9pHV69eraqqcnd373tT3NG744RCofpqbm5uMTExDQ0NgYGB7ElmFveeVzOGNTJKJRIJIeT1119XvpzV19eX/O/Fk48fP163bl1CQgKt2S0DA4OO+wsAMFAgowMA6BepqamEEHd3d6FQ2NLSUlNT097eLpVKlR/cTG8HevDggfIHpVIp+7OhoSEhpL29nb7cu3dvYmJiYWGht7e3mZnZnDlz6EE2IaRHmxi46G4aGRmp3BlF1zwsLS3VyFbMzc1VSkaOHEkIefr0qUba1z4jIyNCCHtLpxrh4eFBQUF5eXkqDznoUc93NYY1NUrHjBlDCLG0tFQupL+j8vJyQgi9LnTmzJnsJujTC7Zs2UJfPnz4UPmzCoVCLBZz3DoAgK5BRgcAoHnt7e179+4lhLzzzjuEEJFIZG5ubmBgoHyRGGvWrFkcm6UHpj/++GN1dfXp06cZhvH39//00081uAkdJxKJpFJpc3NzXV2dcnlZWRkhxNramr7U09NrbW1VrlBdXa3SlEAg6GorlZWVzO8fJEhzOZoz9L197bOxsSGE0BvGuhUfH+/o6JiQkECvVKQ49rx6mhqldFkalVOm9HdEM8x33nlHpXGVqy4dHBzYD9bW1jIMQ7sIAGAgQkYHAKB5Gzdu/OmnnxYuXBgYGEhL/P39FQoFu2QiFR0dPXr0aO4PwjI3N8/PzyeECIXC2bNn09UF2VXXNbIJ3bdw4UJCiPJa8y0tLcnJyWKxmN5eRQixsbF58uQJW6G0tPTx48cq7RgbG7NZmaOjY1xcHPtWc3NzZmYm+/L27dvFxcUymYw96O9j+9pHn9LG8bJGExOT7777TiKRqDx7g0vPd0sjo/SNN96ws7O7ePGi8rMWkpKSCCFvvvkmx0ZY9FfZ8UF2AAADBTI6AADNaG9vf/r06ZkzZ7y9vXfv3v32228fOXKEPVGza9eucePGvf322xcuXKipqamqqtq/f/+HH34YExPTo6Xt/9//+3+5ubktLS1Pnz7dvXs3wzBeXl6a3YSO27Vrl729fWRk5NmzZ+vq6u7fv/+Xv/ylpKQkNjaWfd60j49PcXHxF198UV9f/+jRo4iICPb0GuvVV1+9f//+b7/9lpGRUVhY6OHhwb4llUo3bdqUkZHR0NCQlZW1ZMkSQ0PD2NhYtkJf2vfy8rK0tLx+/brmu6ZrMpls5MiROTk5HOtPmDBh//79KoVcer5bXEbpkiVLBALBzz//3FUjIpEoPj6+srJy8eLFDx48qK6uPnTo0K5du6ZOnRoeHs4xEhZ9dkJXjyAHABgA+rSuCgCApg2gtS5VFl0QCARSqdTZ2XnVqlU3btzoWJ8+y2vs2LFCoXDEiBE+Pj6XL1+mb2VkZCg39f777zO/v+pv3rx5DMNkZ2eHhoa+9NJL9Hl0bm5uBw4caG9v57IJ3cdxrUuGYSoqKiIjI+3t7YVCoVQqlcvlycnJyhWqq6tDQkJsbGzEYvGMGTMyMzNdXV1pT65fv57Wyc/P9/DwkEgko0aNUl4EUiaT2dnZ3b17Vy6Xm5qaisViT0/P1NRUTbXv4eGh/bUuGYbZtGmTgYHBkydP6Et6vxmr06UmV61apbKAp5qe5ziGGQ6j1MvLy8TERKFQqN+j9PR0uVwulUoNDQ2dnJyioqIaGxs7VgsNDVU58pHL5coVAgMD7ezsWltb1W+OO4K1LgFAuwTM7ydcAAB+0csUv/32W74DAW07ceJEUFAQ73+VJk2aVFFRoQurg2r2u1BTUzNhwoT58+fv27dPIw32k+rqaltb2+Dg4AMHDmhhczk5OS4uLt98883ixYs11aZAIDh+/PiiRYs01SAAgHq46hIAAGDwk0qlSUlJJ0+epGv26CaGYcLDw83MzLZt26aFzRUWFvr7+2/cuFGD6RwAgPYhowMAABgSXFxcsrKyLly4UFtby3csnSsrKyssLExOTua4eGYf7d+/f8eOHTt27NDCtgAA+g8yOgAAAEIIiYmJEQgEOTk5T548EQgEmzdv5jsizRszZszZs2fNzMz4DqRz1tbWqampEyZM0M7moqOjcXYOAAaBwbP6GQAAQF+sXbt27dq1fEcBAADQMzhHBwAAAAAAMFAhowMAAAAAABiokNEBAAAAAAAMVMjoAAAAAAAABipkdAAAAAAAAAMV1roEAJ1z8uRJgUDAdxTAD/zqVaBDAABAPQHDMHzHAADwfzIyMn777Te+owANuHPnztdff11RUREZGfnqq6/yHY7u+vrrry9duuTu7r5s2TILCwu+wwENmDZt2gsvvMB3FAAwVCCjAwAADSsrK1u3bt3hw4fnzZv3+eefjxkzhu+IdF1KSkpYWNhvv/22ZcuWNWvWGBjgChoAAOAKGR0AAGgMwzCHDh1avXq1kZHRnj17AgMD+Y5owGhubv7oo4+io6MdHR2/+uord3d3viMCAICBASujAACAZmRnZ0+bNi0kJCQ4ODg/Px/pXI8YGRlFRUXdvn3b2tp6+vTpS5curaio4DsoAAAYAJDRAQBAXzU0NGzYsGHy5MlCofDmzZuxsbEmJiZ8BzUgOTg4XLx48cyZMykpKY6OjnFxcbiUBgAA1ENGBwAAfZKUlPTSSy8dOHDgyy+/vHbt2sSJE/mOaMDz9fXNy8tbsmRJWFiYp6fnnTt3+I4IAAB0FzI6AADopUePHs2ZM8fPz2/mzJkFBQUrV67EUvuaIpVKY2Njs7Ky2traXFxcIiIi6uvr+Q4KAAB0ETI6AADosdbW1ujo6IkTJ5aWlqampiYmJg4fPpzvoAahSZMmpaWlxcfHHzlyxMnJ6eTJk3xHBAAAOgcZHQAA9ExKSopMJtu2bdv69eszMzOnTZvGd0SDmZ6e3tKlS/Py8ry8vBYtWuTr6/vrr7/yHRQAAOgQZHQAAMBVSUnJ0qVLvby8HBwc7t69GxUVJRQK+Q5qSLC2tk5MTExJSSksLHz55ZejoqJaW1v5DgoAAHQCMjoAAOhee3t7XFyck5NTWlra+fPnk5KSRo8ezXdQQ46np2d2dvbOnTtjYmKcnZ2Tk5P5jggAAPiHjA4AALpx8+ZNNze3d999d/ny5bm5uXPnzuU7oqFLKBRGRETk5uY6ODjMnj176dKlT58+5TsoAADgEzI6AADoUnV1dURExGuvvWZsbJydnR0bGyuRSPgOCsjYsWPPnTt35syZa9euOTk5xcbGtre38x0UAADwQ4BHlwIAQKe+/fbbd999V09PLzo6+q233sKTCXRQY2Pj7t27d+3a5ezsvG/fvsmTJ/MdEQAAaBvO0QEAgKoHDx74+PgsXrxYLpfn5eUtXboU6ZxuMjY2joqKyszMFIlE7u7uERERdXV1fAcFAABahYwOAAD+T1NTU1RUlLOzc0VFRXp6emJioqWlJd9BQTdeeeWV1NTUr7/++ujRo05OTomJiXxHBAAA2oOMDgAA/sfZs2dffvnlzz77LDo6OjMzc+rUqXxHBFwJBIKlS5fm5+cHBAT89a9/9fb2Ligo4DsoAADQBmR0AABAiouLly5d6uvrO3HixLy8vIiICH19fb6Dgh4bNmxYbGzstWvXnj596uLiEhUV1dLSwndQAADQv5DRAQAMaQqFIjY21snJKSMj44cffkhKSnrhhRf4Dgr6ZMaMGbdu3dq1a9enn346ceLES5cu8R0RAAD0I2R0AABDV2pqqouLy8aNG1evXp2Xl+fj48N3RKAZBgYGERER9+7dc3d3l8vlvr6+RUVFfAcFAAD9AhkdAMBQ9OzZs4iICE9Pz5EjR966dSsqKkokEvEdFGiYnZ1dYmJiUlJSXl6es7NzbGzs8+fP+Q4KAAA0DBkdAMDQwjBMYmKio6PjyZMn//WvfyUnJzs6OvIdFPSj+fPn3717NyIiYv369VOmTPnPf/7Dd0QAAKBJyOgAAIaQ3NzcGTNm/O1vf/vzn/987969pUuX8h0RaINYLI6KisrNzbW0tJw2bdrSpUurqqr4DgoAADQDGR0AwJDQ2NgYFRU1ZcqUlpaWjIyM2NhYMzMzvoMCrRo/fvzly5ePHTv2ww8/TJgwITExkWEYvoMCAIC+QkYHADD4JSUlvfTSS//85z937979008/TZ48me+IgDeBgYEFBQWLFi16++23vby87t27x3dEAADQJ8joAAAGs8LCwnnz5vn5+Xl6eubn50dEROjpYeYf6szNzWNjY3/66aeGhgaZTLZhw4bm5ma+gwIAgF7C33UAgMGpra0tNjZWJpM9fPjw0qVLiYmJI0eO5Dso0CGvvvpqenr6xx9//NVXX02cOPHChQt8RwQAAL2BjA4AYBD6r//6r0mTJm3atGnNmjW3b99+/fXX+Y4IdBF9bF1+fv60adPeeOMNX1/fx48f8x0UAAD0DDI6AIBBpaysbOnSpTNnzhw7duydO3eioqIMDQ35Dgp0mo2NTWJi4pUrVx4+fPjyyy9HR0crFAq+gwIAAK4EWOcKAGBwaG9vP3z48OrVq42MjD777LOAgAC+I4IBprm5+aOPPoqOjnZ0dPzqq6/c3d35jggAALqHc3QAAINBdnb29OnTQ0JCgoOD8/Pzkc5BLxgZGUVFRd2+fdva2nr69OlLly6tqKjgOygAAOgGMjoAgIGtpqYmIiJi8uTJQqHw5s2bsbGxJiYmfAcFA5iDg8PFixfPnDmTkpLi6OgYFxeHy3kAAHQZMjoAgAEsKSnJ2dn58OHDX3755bVr1yZOnMh3RDBI+Pr65uXlLVmyJCwszNPT886dO3xHBAAAnUNGBwAwID18+HDOnDl+fn4zZ84sKChYuXKlQCDgOygYVKRSaWxsbFZWVltbm4uLS0RERH19Pd9BAQCAKmR0AAADTHNzc1RUlLOzc2lpaVpaWmJi4vDhw/kOCgatSZMmpaWlxcfHHzlyxMnJ6eTJk3xHBAAAv4OMDgBgIElJSXFxcYmJiYmKisrKysJqhKAFenp6S5cuvXPnjpeX16JFi3x9fX/99Ve+gwIAgP+BjA4AQIc0NDR09VZJScnSpUu9vLwcHBzu3r27fv16AwMDbcYGQ5yVlVViYmJKSkphYeHLL78cFRXV2traVWU1IxkAADQLGR0AgK5ISUmZNWvW8+fPVcrb29vj4uKcnJzS0tLOnz+flJQ0evRoXiIE8PT0zM7O3rlzZ0xMjLOzc3Jycsc6eXl5bm5uNTU12g8PAGAIQkYHAKATfv75Z39//8zMzC+++EK5/ObNm25ubu++++7y5ctzc3Pnzp3LV4QAlFAojIiIyM3NdXBwmD179tKlS58+fcq+yzDMihUr8vLygoKCOv57AgAANA4ZHQAA/+rr6+fNm0cvVNu0aVNxcTEhpLq6OiIi4rXXXjM2Ns7JyYmNjZVIJHxHCvA/xo4de+7cuTNnzly7ds3JySk2Nra9vZ0QcvDgwZ9++okQ8uOPP77//vt8hwkAMPjpR0VF8R0DAMCQxjBMcHBwenq6QqGgJYWFhQKB4I033rh79+4XX3zx6aefjhgxgt8gATrl6OgYGhqqUCg++OCD77//3sHBISQkpLm5mRDCMEx6evof/vAHZ2dnvsMEABjMBAzD8B0DAMCQFhUVtW3bNnp+g6WnpxcWFrZ9+3apVMpXYADc5eTkrFq16qeffhIIBOz/JgghQqEwLS1typQpPMYGADC4IaMDAODT6dOn/f39VaZiPT09KyurwsJCIyMjvgID6Knr169PmzZNZTDr6+tbWlreunXL1taWr8AAAAY33EcHAMCbe/fuBQcHdyxvb28vLy//6KOPtB8SQO8oFIqQkBB9fX2V8ufPnz979uzNN99saWnhJTAAgEEPGR0AAD+qqqrmzJnT2tra6bUSCoVi586dBQUF2g8MoBc+++yze/fuKV9vyWpra7t169bKlSu1HxUAwFCAqy4BAHigUChef/319PT0trY2NdW8vb1//PFHrUUF0Du//fabo6NjU1OTmjoCgeDzzz9/5513tBYVAMAQgXN0AAA8WL16dWpqasd0TigUCgQCQoiFhcW8efO8vLzq6+v5CBCgB549exYWFubm5iYSiQghBgYGQqFQpQ7DMBERESkpKXwECAAwmOEcHQCAtiUkJPztb3+jP+vr69O1AYVCobOz8x//+MepU6e6ubmNGTOG1xgBekOhUOTk5PznP//5z3/+k5qa+vPPPzMMY2Rk1Nra2t7eLhAIzMzMbt26ZW9vz3ekAACDBzI6gE5kZGR8+umnfEcBg1NlZeW1a9foswqMjY2HDx8+bNiwYcOGmZub6+nhuglOvv322z62gO+4drS2tlZVVVVVVVVWVlZWVtK77MzMzLy8vAwMDPiODkBV3+cWAF5gPgXoxG+//Xby5MmAgAC+A9FF169fJ4S4ubnxHYiuOHnypJub2wsvvMClcltbW2FhoaOjI83i6CVqwF1RUREdgX2E77h2GBoaWltbW1tb05d1dXU0wXv48KGTk5NKZcwtKno0t0AfaWpuAeAFMjqALuF/dZ0KDAwk6BwlAoHg73//+6JFi/gOZEg4ceJEUFCQplrDMNYpmFtUYG7RJs3OLQBahit8AAAAAAAABipkdAAAAAAAAAMVMjoAAAAAAICBChkdAAAAAADAQIWMDgBAe3799dcFCxbU1tZWVFQI/peLi0tzc7NyNeV3BQLB5MmT+QpYjba2tj179ri6upqamo4cOXLu3LlJSUldPRFnwYIFAoFg+/btyoUbNmw4fvy4VoLVjGPHjtHfiJGREd+xAPwO5hblwgE3twD0ETI6AN7U19f/4Q9/mD9/vg621h90P8L+lp2dPXnyZB8fHzMzs+HDhzMMk5mZScsjIyOVa9J3MzIyLC0tGYbJysriKeQuNTQ0eHl5HTx4cM+ePU+fPs3KyjIxMVmwYMGdO3c6Vk5MTExKSupYvmLFio0bN27ZsqX/49WMxYsXMwzj7e3NdyDwO5hbMLeoGHBzC0AfIaMD4A3DMO3t7fRJ0/3UmomJyYwZMzTSft9pdn97gd/eqK2t9fX1/dOf/vTuu+8ql4tEIktLy/379x89epSv2Hph3bp1ubm5ly5d+uMf/ygWi0ePHn3w4MFOH69XXFwcGRn51ltvdXxr3Lhxp06d2rFjx4kTJ/o/5MFMp77plDZDwtyCuUUF5hYYapDRAfDG1NT00aNH58+f18HW+oPuR9ivdu/eXVpaunXrVpVyIyOjI0eO6OnphYaG3r9/n5fYeqqsrCwuLi44ONjKyootlEgkzc3NEydOVKm8YsWKwMBAHx+fTpuSyWQBAQFr1qxRKBT9GDEMaphbMLd0hLkFhhRkdAAA/Y5hmPj4+KlTp9ra2nZ8Vy6Xb968ua6uLjAwUOWmF930/fffP3/+nMtJiYSEhDt37sTExKips3DhwqKionPnzmkuQIChAnOLmjqYW2DoQEYH0EsxMTH0zvIXXnghMzPT29vb1NTU2Nh41qxZaWlptM7p06fZG9ALCgoWLVpkaWlJX8bHx7NvKf+hraysXL169bhx4wwNDS0sLObOnZuSktKL1mh4DQ0NaWlptNzAwKC6ulr5nnh6K7lCoWBLAgIC+qm7lIOnESqX/PLLL0FBQebm5paWlvPnz3/06BH3Tt6+fTutwx4EXLx4kZYMHz5cuR2V3uinPe1UTk5OWVmZTCbrqsIHH3zg4+OTm5v73nvvqWmH4/Doqj+p8vLy8PDwMWPGGBoajhgxwt/fPzs7u0e7c/PmTUKIhYXFmjVrRo0aZWho+OKLL4aHh1dVVSlXKyoqWrNmTUJCgqmpqZrWJk2aRAj54YcfehSD1uTn57/55ptSqVQikXh4eKSmpnaso6ZLuYxhQohCoTh+/Pjs2bOtra3FYrGzs3NsbCx7GaGa7/6GDRs6HdvKH/n111+DgoJMTU0tLS3feuutZ8+e/fLLL76+vqampjY2NitWrKirq+O4O9y/tlr7umFuwdzSFR2fWwA0iQGADugaWVxqymQyiUTi7u6enp5eX1+fmZn5yiuvGBoaXr16la3j5+dHCPH09ExJSWloaLh+/bq+vn55eTn7VlNTE61ZUlJib29vZWWVlJRUU1NTUFDg7+8vEAgOHDjQi9YYhpFIJNOnT1eJWS6X6+npPXz4ULnQ3d39yJEjXHY5ICAgICCAS82OOkZIS/z8/GgHXr58WSwWT5kyRflTXDq54566urrSW//V1KFmzZo1bNiwjIyM3u0UIeT48ePq6xw6dIgQsnPnTpXyzMxMqVRKfy4vLx81ahQh5PDhw7SEXb2A4j481PRncXHxiy++aGVlde7cubq6ury8PE9PTyMjo/T0dO67TLdibW0dHBz86NGjZ8+e/fvf/5ZIJOPHj6+urmaryeXysLAw5R7Ytm1bx9ZqamoIIR4eHlw2zf27qZF2Hjx4YG5ubmdnd+nSpbq6utzcXB8fnzFjxohEIrYOly7tdgzT1R127txZVVVVXl7+z3/+U09Pb+3atcrBqPnudzW26Uf8/f2zsrLq6+sTExMJIXPnzvXz87t161ZdXd2+ffsIIX//+997tDtcvrZdhaQG5hYVmFsG4twCwAuMXYBO9CijI4TcunWLLcnNzSWEyGQytoT+iTp//nzHj6schSxfvpwQcvToUbZCc3Ozra2tWCwuLS3taWtMF8cZ9B+W7J9DhmFSU1Pt7OxaW1u57HJ/HHXRlanZ9gkh9DiV4tLJfTnq8vT0tLCw6NFhhzIuR127d+8mhOzdu1elXPmoi2GYjIwMoVAokUju3bvHdDjq4j481PTnsmXLCCHK2XtJSYlIJHJ1deW+y3K5nBBib2/f1tbGFtJTvlu2bKEv4+Lixo4dW19fT1+qOepiGEYgEDg4OHDZtJYzusDAQELIyZMn2ZInT56IRCLljI5Ll3Y7hpOSkmbOnKm86SVLlgiFwpqaGrZEzXdffUZ37tw5tmTChAmEkGvXrrEl9vb2jo6OPdodLl9bHcnoMLdQmFu6hYwOBjRcdQnQVxKJhF7aQTk7O9va2ubk5JSUlChXe+2117pt6tSpU4SQefPmsSUikcjb27upqUnluhEurXXFx8fH2dn54MGDlZWVtOTjjz9+7733hEJhr9vsoylTprA/0/8lFxcXK1fg2Mm9c/Xq1aqqKnd397431RV6MVi3Pezm5hYTE9PQ0BAYGNjU1KTyLvfhoaY/T58+raenp7zOu7W19YQJE27cuFFUVMRxdyQSCSHk9ddfV77AzNfXl/zvBU6PHz9et25dQkICrdktAwODjvurCy5evEgIoUeZlK2t7fjx45XrcOxS9WN4/vz57EVulEwma2tr67hiey+++8rPHKN3WymX2NnZKX/duI+Qbr+2ugBzC4W5BWBwQ0YH0Ffm5uYqJSNHjiSEPH36VLmw2z8/LS0tNTU1RkZGKjcG0CW/SktLe9SaepGRkY2NjV9++SUh5P79+1euXFm5cmVfGuwjqVTK/mxoaEgIUVmInGMn6yz6QOq2trZua4aHhwcFBeXl5aksRN6j4dFVf9JG2tvbpVKpQAm9d+XBgwccd2fMmDGEEEtLS+VC+hspLy8nhNBrt2bOnMlugq4wvmXLFvry4cOHyp9VKBRisZjj1rWmpaWlrq7OyMjIxMREuZzuKVuHY5eqH8M1NTVbt251dna2sLCgLaxbt44Q0tjYqPKpXnz3zczM2J/19PT09fWNjY3ZEn19ffbr1qMR0u3XVhdgbmFhbgEYxJDRAfRVZWUlwzDKJfRQQPmwjwuRSCSVSpubm1VWKSgrKyOEWFtb9yI2gUDQaTldG/qLL75oaWn55JNPli1bZmFh0Yv2tabbTtbT02ttbVWuUF1drdJIV72hBTY2NoQQelNHt+Lj4x0dHRMSEujVRJRGhodIJDI3NzcwMFC+ook1a9YsjrtDF4pQOYlBfyP0KPCdd95RaVzlyigHBwf2g7W1tQzD0C7SKSKRyNTUtLm5ub6+XrlceZEG7l2qfgz7+vpu27ZtxYoV9+/fb29vZxhmz549hBCVj3RFU2NbUyNEgyH1N8wtmFsABgFkdAB91dzcnJmZyb68fft2cXGxTCbrxV+RhQsXEkKUl1puaWlJTk4Wi8XK131xZ2xszB6LODo6xsXF0Z9FIlFYWNjTp08/+eSTI0eORERE9KJxbeq2k21sbJ48ecJWKC0tffz4sUojXfWGFtAnKXG89MjExOS7776TSCT0JCpLI8PD399foVAor7JICImOjh49ejT3pza98cYbdnZ2Fy9eVF6mla7t8eabb3JshEV/cR0fNqUL5s6dS/732kuqoqKioKBAuQ7HLlUzhp8/f56WlmZtbR0eHj5ixAiaHvToUjENjm2NjBDNhtSvMLcQzC0AAx8yOoC+kkqlmzZtysjIaGhoyMrKWrJkiaGhYWxsbC+a2rVrl729fWRk5NmzZ+vq6u7fv/+Xv/ylpKQkNjZW+XGr3L366qv379//7bffMjIyCgsLPTw82LfCwsLEYvHmzZtff/115X9q6qZuO9nHx6e4uPiLL76or69/9OhRREREx3OkXfWGl5eXpaXl9evX+y9+mUw2cuTInJwcjvUnTJiwf/9+lUKNDI9du3aNGzfu7bffvnDhQk1NTVVV1f79+z/88MOYmBj2xpUlS5YIBIKff/65q0ZEIlF8fHxlZeXixYsfPHhQXV196NChXbt2TZ06NTw8nGMkLLq+eVePCebXzp07h/b2Y+oAABCISURBVA0bFhkZefny5fr6+rt37y5ZskTlIkwuXUrUjmF9ff2ZM2eWlpZ+/PHHFRUVTU1NKSkpdBVKjtR803uK4+5oM6R+hbmFYG4BGAS6WjIFYCjr0VqXdnZ2d+/elcvlpqamYrHY09MzNTWVvpuRkdHVN47eic4KDg6m5RUVFZGRkfb29kKhUCqVyuXy5OTkXreWn5/v4eEhkUhGjRrVcTG0FStWkN+vesdF79aj6xihyu68//77zO+vfZo3bx79rPpOpqqrq0NCQmxsbMRi8YwZMzIzM11dXWk769evV98bHh4e/b0eHcMwmzZtMjAwePLkCX1J7wlhdboc3KpVq1SW1OM+PNT0J33w1NixY4VC4YgRI3x8fC5fvqy8FS8vLxMTE4VCoX6P0tPT5XK5VCo1NDR0cnKKiopqbGzsWC00NFRl3MrlcuUKgYGB3Jda1fJalwzDFBQUvPnmm2ZmZnSt9rNnz3p7e9Md+dvf/kbrdNul3Y7h8vLy0NDQUaNGCYVCKyur5cuXb9iwgR0bar77VMex3XE8KJ+JIoTs2rXrv//7v5VLPvjgg253h/swUz/5dApziwqCuWUAzi0AvMDYBehETzO6/o6nnyQkJPRoXWmqLyuM946OdzLHo67q6mo7O7vQ0FAthNQXz549E4vFISEh2tlcdna2QCBQXjZdPe1ndBqh42NYR2BuUYG5pS/4mlsAeIGrLgGGrn379q1evZrvKIYKqVSalJR08uTJvXv38h1LlxiGCQ8PNzMz27ZtmxY2V1hY6O/vv3HjxsWLF2thcwCDEuaWjjC3wFCDjA5gaImPj1+4cGF9ff2+ffuePXu2aNEiviMaQlxcXLKysi5cuFBbW8t3LJ0rKysrLCxMTk7u3dqqPbV///4dO3bs2LFDC9sCGMQwt6jA3AJDDTI6gF6KiYkRCAQ5OTlPnjwRCASbN2/mOyKuTp8+bWFh8dVXXx07dqxHSx1o38Dt5K6MGTPm7Nmzys8H0ynW1tapqakTJkzQzuaio6MH/X/QB98YHhwG3+8Fc4uyoTC3ACjT6YM5AF22du3atWvX8h1Fj4WEhISEhPAdBVcDtJMBWBjDugm/FwAYTHCODgAAAAAAYKBCRgcAAAAAADBQIaMDAAAAAAAYqJDRAQAAAAAADFTI6AAAAAAAAAYqAcMwfMcAoHNOnDgRFBTEdxQA0Lm+/+XCdxwAOsJRMQxQeHoBQJeOHz/Odwi6aM+ePYSQv//973wHoiuCgoIiIyPd3d35DmRIyMjI+OyzzzTVGr7jOgVziwrMLdqk2bkFQMuQ0QF0adGiRXyHoIu+/fZbgs5REhQU5O7ujg7RGg0edeG3plMwt6jA3KJlyOhg4MJ9dAAAAAAAAAMVMjoAAAAAAICBChkdAAAAAADAQIWMDgAAAAAAYKBCRgegYSYmJgIlenp6FhYWMpksLCzsxo0bfEcHOufXX39dsGBBbW1tRUUFO2xcXFyam5uVqym/KxAIJk+ezFfA6p0/f378+PEGBp0su7VhwwasLYn5AbRmkM0t1IIFCwQCwfbt25ULMbcAIKMD0LD6+vpbt24RQvz8/BiGaWtry8/P//DDD/Pz8ydPnvzXv/61sbGR7xhBV2RnZ0+ePNnHx8fMzGz48OEMw2RmZtLyyMhI5Zr03YyMDEtLS4ZhsrKyeAq5S48ePVqwYMHGjRvLyso6rbBixYqNGzdu2bJFy4HpFMwPoB2DaW5hJSYmJiUldSzH3AKAjA6gf+nr61tZWfn5+V25cuUf//jHwYMH//znP+MZptyZmJjMmDFj4LavRm1tra+v75/+9Kd3331XuVwkEllaWu7fv//o0aO8BNY7W7ZsmTZt2o0bN0xNTTutMG7cuFOnTu3YsePEiRNajk1nYX7gEeaWgaW4uDgyMvKtt97q+BbmFgBkdADa89FHH02dOvX7778/duwY37EA/3bv3l1aWrp161aVciMjoyNHjujp6YWGht6/f5+X2Hrh66+/3rBhQ6fXW7JkMllAQMCaNWsUCoXWAhsoMD+ApgyyuYVasWJFYGCgj49Pp+9iboEhDhkdgPYIBAL6H9Mvv/yS71iAZwzDxMfHT5061dbWtuO7crl88+bNdXV1gYGBKje96CyxWMyl2sKFC4uKis6dO9ff8Qw4mB9AIwbf3EIISUhIuHPnTkxMjJo6mFtgKENGB6BV9CKc69evt7W10ZLy8vLw8PAxY8YYGhqOGDHC398/OzubvnX69Gn2bvVffvklKCjI3Nzc0tJy/vz5jx49YttsaWnZunWrk5OTsbHxsGHDfH19v//+++fPn7MV1GyiX1VWVq5evXrcuHGGhoYWFhZz585NSUmhb23fvp3uF3tV0sWLF2nJ8OHDaUlMTIxAIGhoaEhLS6Nv0fM/tFwgELzwwguZmZne3t6mpqbGxsazZs1KS0vre/vakZOTU1ZWJpPJuqrwwQcf+Pj45Obmvvfee2raUdPJHMcP0e4ImTRpEiHkhx9+6Kf2B7QhNT/0BeYWNQbf3FJUVLRmzZqEhISuruimMLfAkMYAQAd01axef1x55QMVTU1N9KtXXFzMMExxcfGLL75oZWV17ty5urq6vLw8T09PIyOj9PR09iN+fn60tfT09Pr6+suXL4vF4ilTprAVQkJCpFLppUuXGhsbS0tL165dSwhJSUmh73LZRI8EBAQEBAR0W62kpMTe3t7KyiopKammpqagoMDf318gEBw4cICtI5FIpk+frvwpV1dXene+mjqUTCaTSCTu7u60WzIzM1955RVDQ8OrV69qpP1Zs2YNGzYsIyOj2z0lhBw/frzbaioOHTpECNm5c6dKeWZmplQqpT+Xl5ePGjWKEHL48GFawq5eQHHp5G7Hj8ZHiJ2dnb6+flfv1tTUEEI8PDx613gfv5sab6cXBvf80BeYW1RgbqHkcnlYWJjy3m3btq1jNR2ZWwB4gbEL0In+y+jYhezoEduyZcsIIUeOHGErlJSUiEQiV1dXtoT+1UxKSmJLAgICCCHl5eX0pb29/bRp05S3Mn78ePaIjcsmeoTjUdfy5csJIUePHmVLmpubbW1txWJxaWkpLenjURch5NatW2xJbm4uIUQmk6n5LPf2PT09LSwsuBx59O6oa/fu3YSQvXv3qpQrH3UxDJORkSEUCiUSyb1795gOR11cOrnb8aPxEaI+o2MYRiAQODg49K7xwZ3RDYL5oS8wt6jA3MIwTFxc3NixY+vr6+lLNRkdoxtzCwAvcNUlgFaVlJQQQoRCIb0+5/Tp03p6evPnz2crWFtbT5gw4caNG0VFRcofnDJlCvsz/d9qcXExfTlnzpz09PSVK1dev36dXkxVUFAwc+ZM+i73TWjWqVOnCCHz5s1jS0Qikbe3d1NTk6auipFIJPQyG8rZ2dnW1jYnJ4d2ch9dvXq1qqrK3d297011it7BIhQK1Vdzc3OLiYlpaGgIDAxkT+CwuHeymvGj/RFiYGDQcV+ADKX5oS8wt6g3mOaWx48fr1u3LiEhQSKRcKmPuQWGLGR0AFqVmppKCHF3dxcKhS0tLTU1Ne3t7VKpVPkBrzdv3iSEPHjwQPmDUqmU/dnQ0JAQ0t7eTl/u3bs3MTGxsLDQ29vbzMxszpw59I8xIaRHm9Agul0jIyOV2x6srKwIIaWlpRrZirm5uUrJyJEjCSFPnz7VSPv9ysjIiBDC3i6lRnh4eFBQUF5enspC5D3q5K7GDy8jRKFQcFxGZagZIvNDX2Bu6dZgmlvoNZ8zZ85kP06fXrBlyxb68uHDh8r1MbfAkIWMDkB72tvb9+7dSwh55513CCEikcjc3NzAwKCtra3jCfRZs2ZxbJb+kfvxxx+rq6tPnz7NMIy/v/+nn36qwU30lEgkkkqlzc3NdXV1yuX02dPW1tb0pZ6eXmtrq3KF6urqjnvX1VYqKyuZ3z+5ix5v0WOvvrffr2xsbAgh9MaPbsXHxzs6OiYkJNArjiiOnaye9kdIbW0twzB090HZ0Jkf+gJzS7cG09zyzjvvqHxQ5apLBwcHtjLmFhjKkNEBaM/GjRt/+umnhQsXBgYG0hJ/f3+FQsGuokZFR0ePHj2a+0N1zM3N8/PzCSFCoXD27Nl0FTJ2BWeNbKIXFi5cSAhRXki6paUlOTlZLBbL5XJaYmNj8+TJE7ZCaWnp48ePVdoxNjZmj5wcHR3j4uLYt5qbmzMzM9mXt2/fLi4ulslk7F/0PrbfryZOnEgI4XjpkYmJyXfffSeRSFTWtefSyd3S8gihvxG6+6BsSM0PfYG5RT3MLRpvGWAA6PiPEwDQ4Mooz58/LysrO336tJeXFyHk7bffbmxsZGuWlZWNGzdu7Nix58+fr66urqys3Ldvn7GxsfLd8PTu86amJrZk/fr1ROnGfalU6unpmZOT09zcXFZWFhUVRQjZvn079030SC/Wo6utrWWXSouLi2Pr0Et9Pv/887q6uocPHy5atMjOzk5ldYE5c+ZIpdLHjx+np6cbGBjcvXuXlstkMqlU6u3trWY9ur6039/r0bW3t48cObLjwgkqqxcoO3z4MCGkq/XouurkbscPlxESHBxMCCksLOSya+pXRvnmm28IIadOneLSVEeDbGWUwTc/9AXmFhWYWzpSszKKjswtALzA2AXoRF9mdpUbuAUCgVQqdXZ2XrVq1Y0bNzrWp8/8GTt2rFAoHDFihI+Pz+XLl+lbGRkZyk29//77zO8vBJo3bx7DMNnZ2aGhoS+99BJ93pSbm9uBAwfa29u5bKIXOB51MQxTUVERGRlpb28vFAqlUqlcLk9OTlauUF1dHRISYmNjIxaLZ8yYkZmZ6erqSndt/fr1tE5+fr6Hh4dEIhk1apTy6m0ymczOzu7u3btyudzU1FQsFnt6eqampmqqfQ8Pj35dj45hmE2bNhkYGDx58oS+LC8vV/7ldroc3KpVq1SOGtV0Msfxw3AYIV5eXiYmJgqFQs3uJCUldfynofJa51RgYKCdnV1ra2sPekrJQM/oBv380BeYW1RgblEWGhqqMr3I5XLlCjoytwDwAmMXoBOY2dXgftTVr+hRF99RMEwfjrqqq6vt7OxCQ0M1HpJmPXv2TCwWh4SE9L2p7OxsgUCgvCR6Tw30jA7UwNyiAnMLd7oztwDwAvfRAQDwQyqVJiUlnTx5kq6HoZsYhgkPDzczM9u2bVsfmyosLPT399+4cePixYs1EhsAdApzC8BQg4wOAIA3Li4uWVlZFy5cqK2t5TuWzpWVlRUWFiYnJ3Nc4E6N/fv379ixY8eOHRoJDADUwNwCMKQgowOAASYmJkYgEOTk5Dx58kQgEGzevJnviPpkzJgxZ8+eNTMz4zuQzllbW6empk6YMKHvTUVHR+M/6KDLMLdoE+YWAA0y4DsAAICeWbt27dq1a/mOAgAGG8wtADBA4RwdAAAAAADAQIWMDgAAAAAAYKBCRgcAAAAAADBQIaMDAAAAAAAYqLAyCkCXTpw4wXcIuqioqIigc34vIyOD7xCGCs12NYaxTsHc0hHmFq1BV8OAJmAYhu8YAHTOiRMngoKC+I4CADrX979c+I4DQEc4KoYBChkdAAAAAADAQIX76AAAAAAAAAYqZHQAAAAAAAADFTI6AAAAAACAgQoZHQAAAAAAwED1/wELiGDvF+4IUgAAAABJRU5ErkJggg==",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "keras.utils.plot_model(mimo_model, show_shapes=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<keras.engine.input_layer.InputLayer at 0x7feb5fc88850>,\n",
       " <keras.engine.input_layer.InputLayer at 0x7feb5fc888b0>,\n",
       " <keras.engine.input_layer.InputLayer at 0x7feb5fc88c10>,\n",
       " <keras.layers.merging.concatenate.Concatenate at 0x7feb5fd33910>,\n",
       " <keras.layers.core.dense.Dense at 0x7feb5fc88430>,\n",
       " <keras.layers.core.dense.Dense at 0x7feb5fc837f0>,\n",
       " <keras.layers.core.dense.Dense at 0x7feb5fc9dca0>]"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mimo_model.layers"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(<KerasTensor: shape=(None, 1000) dtype=float32 (created by layer 'title')>,\n",
       " <KerasTensor: shape=(None, 1000) dtype=float32 (created by layer 'title')>)"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mimo_model.layers[0].input, mimo_model.layers[0].output\n",
    "\n",
    "# this enables to do feature extraction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "features = mimo_model.layers[4].output\n",
    "difficulty = layers.Dense(3, activation=\"softmax\", name=\"difficulty\")(features)\n",
    "\n",
    "new_model = keras.Model(inputs=[title, text_body, tags],\n",
    "                        outputs=[priority, dept, difficulty])\n",
    "\n",
    "keras.utils.plot_model(new_model, show_shapes=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Subclassing Model Class**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "class CustomerTicketModel(keras.Model):\n",
    "\n",
    "    def __init__(self, num_depts):\n",
    "        super().__init__()\n",
    "        self.concat_layer = layers.Concatenate()\n",
    "        self.mixing_layer = layers.Dense(64, activation=\"relu\")\n",
    "        self.priority_scorer = layers.Dense(1, activation=\"sigmoid\", name=\"priority\")\n",
    "        self.dept_classifier= layers.Dense(num_dept, activation=\"softmax\", name=\"department\")\n",
    "\n",
    "    def call(self, inputs):\n",
    "        title = inputs[\"title\"]\n",
    "        text_body = inputs[\"text_body\"]\n",
    "        tags = inputs[\"tags\"]\n",
    "\n",
    "        features = self.concat_layer([title, text_body, tags])\n",
    "        features = self.mixing_layer(features)\n",
    "\n",
    "        priority = self.priority_scorer(features)\n",
    "        dept = self.dept_classifier(features)\n",
    "\n",
    "        return priority, dept"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"customer_ticket_model\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " concatenate_1 (Concatenate)  multiple                 0         \n",
      "                                                                 \n",
      " dense_6 (Dense)             multiple                  134464    \n",
      "                                                                 \n",
      " priority (Dense)            multiple                  65        \n",
      "                                                                 \n",
      " department (Dense)          multiple                  260       \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 134,789\n",
      "Trainable params: 134,789\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "ct_model = CustomerTicketModel(num_depts=4)\n",
    "\n",
    "priority, dept = ct_model({\"title\": title_data, \"text_body\": text_body_data, \"tags\": tags_data})\n",
    "\n",
    "ct_model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "40/40 [==============================] - 1s 2ms/step - loss: 4.4532 - output_1_loss: 0.1317 - output_2_loss: 4.3215 - output_1_mean_absolute_error: 0.2984 - output_2_accuracy: 0.3602\n",
      "40/40 [==============================] - 0s 2ms/step - loss: 4.4251 - output_1_loss: 0.1448 - output_2_loss: 4.2804 - output_1_mean_absolute_error: 0.3081 - output_2_accuracy: 0.5719\n",
      "40/40 [==============================] - 0s 1ms/step\n"
     ]
    }
   ],
   "source": [
    "ct_model.compile(optimizer=\"rmsprop\",\n",
    "                    loss=[\"mean_squared_error\", \"categorical_crossentropy\"],\n",
    "                    metrics=[[\"mean_absolute_error\"], [\"accuracy\"]])\n",
    "\n",
    "ct_model.fit({\"title\": title_data, \"text_body\": text_body_data, \"tags\": tags_data},\n",
    "                    [priority_data, dept_data],\n",
    "                    epochs=1)\n",
    "\n",
    "ct_model.evaluate({\"title\": title_data, \"text_body\": text_body_data, \"tags\": tags_data},\n",
    "                    [priority_data, dept_data],)\n",
    "\n",
    "priority_preds, dept_preds = ct_model.predict({\"title\": title_data, \"text_body\": text_body_data, \"tags\": tags_data})"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Mixing & Matching Different Components**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "*Creating a functional model that includes a sub-classed model*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "class Classifier(keras.Model):\n",
    "\n",
    "    def __init__(self, num_classes=2):\n",
    "        super().__init__()\n",
    "\n",
    "        if num_classes == 2:\n",
    "            num_units = 1\n",
    "            activation = \"sigmoid\"\n",
    "        else:\n",
    "            num_units = num_classes\n",
    "            activation = \"softmax\"\n",
    "\n",
    "        self.dense = layers.Dense(num_units, activation=activation)\n",
    "\n",
    "    def call (self, inputs):\n",
    "        return self.dense(inputs)\n",
    "\n",
    "inputs = keras.Input(shape=(3,))\n",
    "features = layers.Dense(64, activation=\"relu\")(inputs)\n",
    "outputs = Classifier(num_classes=10)(features)\n",
    "model2 = keras.Model(inputs=inputs, outputs=outputs)\n",
    "\n",
    "keras.utils.plot_model(model2, show_shapes=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "*Creating a sub-classed model that includes a functional model*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "inputs = keras.Input(shape=(64,))\n",
    "outputs = layers.Dense(1, activation=\"sigmoid\")(inputs)\n",
    "binary_classifier = keras.Model(inputs=inputs, outputs=outputs)\n",
    "\n",
    "class MyModel(keras.Model):\n",
    "\n",
    "    def __init__(self, num_classes=2):\n",
    "        super().__init__()\n",
    "        self.dense = layers.Dense(64, activation=\"relu\")\n",
    "        self.classifier = binary_classifier\n",
    "\n",
    "    def call(self, inputs):\n",
    "        features = self.dense(inputs)\n",
    "        return self.classifier(features)\n",
    "\n",
    "model3 = MyModel()\n",
    "inputs = keras.Input(shape=(3,))\n",
    "output = model3(inputs)\n",
    "\n",
    "keras.utils.plot_model(model3, show_shapes=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "from numba import cuda\n",
    "cuda.select_device(0)\n",
    "cuda.close()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3.9.12 ('tflab')",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.12"
  },
  "orig_nbformat": 4,
  "vscode": {
   "interpreter": {
    "hash": "03093e31b4ca43731012cb7d90aef1852be82b91fabe0f61f11e024946a263ab"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}