keras_vae_tutorial.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "vae.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "id": "nzdcpXpICR-f",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import keras\n",
        "from keras.layers import Lambda, Input, Dense\n",
        "from keras.models import Model\n",
        "from keras.datasets import mnist\n",
        "from keras.losses import binary_crossentropy\n",
        "from keras import backend as K\n",
        "\n",
        "import numpy as np\n",
        "import matplotlib.pyplot as plt"
      ],
      "execution_count": 6,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VnLCzaVKCbRB",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# MNIST dataset\n",
        "(x_train, _), (x_test, _) = mnist.load_data()\n",
        "\n",
        "x_train = np.reshape(x_train, [60000, 784])\n",
        "x_test = np.reshape(x_test, [10000, 784])\n",
        "x_train = x_train.astype('float32') / 255.\n",
        "x_test = x_test.astype('float32') / 255.\n",
        "\n",
        "def sampling(args):\n",
        "    z_mean, z_log_var = args\n",
        "    batch = K.shape(z_mean)[0]\n",
        "    dim = K.int_shape(z_mean)[1]\n",
        "    epsilon = K.random_normal(shape=(batch, dim))\n",
        "    return z_mean + K.exp(0.5 * z_log_var) * epsilon\n",
        "\n",
        "# VAE model = encoder + decoder\n",
        "# build encoder model\n",
        "inputs = Input(shape=(784, ), name='encoder_input')\n",
        "x = Dense(512, activation='relu')(inputs)\n",
        "z_mean = Dense(2, name='z_mean')(x)\n",
        "z_log_var = Dense(2, name='z_log_var')(x)\n",
        "z = Lambda(sampling, output_shape=(2,), name='z')([z_mean, z_log_var])\n",
        "\n",
        "# instantiate encoder model\n",
        "encoder = Model(inputs, [z_mean, z_log_var, z], name='encoder')\n",
        "# build decoder model\n",
        "latent_inputs = Input(shape=(2,), name='z_sampling')\n",
        "x = Dense(512, activation='relu')(latent_inputs)\n",
        "outputs = Dense(784, activation='sigmoid')(x)\n",
        "# instantiate decoder model\n",
        "decoder = Model(latent_inputs, outputs, name='decoder')\n",
        "\n",
        "# instantiate VAE model\n",
        "outputs = decoder(encoder(inputs)[2])\n",
        "vae = Model(inputs, outputs, name='vae_mlp')"
      ],
      "execution_count": 7,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6WoYnQ2vWw37",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# VAE loss = mse_loss or xent_loss + kl_loss\n",
        "reconstruction_loss = binary_crossentropy(inputs, outputs) * 784\n",
        "\n",
        "kl_loss = 1 + z_log_var - K.square(z_mean) - K.exp(z_log_var)\n",
        "kl_loss = K.sum(kl_loss, axis=-1)*(-0.5)\n",
        "\n",
        "vae_loss = K.mean(reconstruction_loss + kl_loss)\n",
        "vae.add_loss(vae_loss)\n",
        "vae.compile(optimizer='adam')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0JMG5zW-Q-3u",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "cc607945-b0e6-48cd-e318-56509310fcd9"
      },
      "source": [
        "# train the autoencoder\n",
        "vae.fit(x_train,\n",
        "        epochs=50,\n",
        "        batch_size=128,\n",
        "        validation_data=(x_test, None))"
      ],
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/keras/engine/training_utils.py:819: UserWarning: Output decoder missing from loss dictionary. We assume this was done on purpose. The fit and evaluate APIs will not be expecting any data to be passed to decoder.\n",
            "  'be expecting any data to be passed to {0}.'.format(name))\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Train on 60000 samples, validate on 10000 samples\n",
            "Epoch 1/50\n",
            "60000/60000 [==============================] - 14s 232us/step - loss: 198.9698 - val_loss: 174.0658\n",
            "Epoch 2/50\n",
            "60000/60000 [==============================] - 13s 223us/step - loss: 170.7867 - val_loss: 167.3984\n",
            "Epoch 3/50\n",
            "60000/60000 [==============================] - 13s 222us/step - loss: 165.6201 - val_loss: 163.3487\n",
            "Epoch 4/50\n",
            "60000/60000 [==============================] - 13s 223us/step - loss: 162.2058 - val_loss: 160.6898\n",
            "Epoch 5/50\n",
            "60000/60000 [==============================] - 13s 223us/step - loss: 160.0699 - val_loss: 159.2391\n",
            "Epoch 6/50\n",
            "60000/60000 [==============================] - 13s 224us/step - loss: 158.5908 - val_loss: 158.0246\n",
            "Epoch 7/50\n",
            "60000/60000 [==============================] - 13s 222us/step - loss: 157.4262 - val_loss: 157.4132\n",
            "Epoch 8/50\n",
            "60000/60000 [==============================] - 13s 220us/step - loss: 156.5384 - val_loss: 156.3530\n",
            "Epoch 9/50\n",
            "60000/60000 [==============================] - 13s 218us/step - loss: 155.7210 - val_loss: 155.9271\n",
            "Epoch 10/50\n",
            "60000/60000 [==============================] - 13s 220us/step - loss: 155.0308 - val_loss: 155.2988\n",
            "Epoch 11/50\n",
            "60000/60000 [==============================] - 13s 217us/step - loss: 154.3679 - val_loss: 154.7108\n",
            "Epoch 12/50\n",
            "60000/60000 [==============================] - 13s 216us/step - loss: 153.7677 - val_loss: 154.3229\n",
            "Epoch 13/50\n",
            "60000/60000 [==============================] - 13s 218us/step - loss: 153.2029 - val_loss: 154.0794\n",
            "Epoch 14/50\n",
            "60000/60000 [==============================] - 13s 218us/step - loss: 152.7498 - val_loss: 153.4220\n",
            "Epoch 15/50\n",
            "60000/60000 [==============================] - 13s 220us/step - loss: 152.2137 - val_loss: 153.2595\n",
            "Epoch 16/50\n",
            "60000/60000 [==============================] - 13s 218us/step - loss: 151.7724 - val_loss: 152.6796\n",
            "Epoch 17/50\n",
            "60000/60000 [==============================] - 13s 218us/step - loss: 151.3753 - val_loss: 152.2634\n",
            "Epoch 18/50\n",
            "60000/60000 [==============================] - 13s 219us/step - loss: 151.0102 - val_loss: 152.3400\n",
            "Epoch 19/50\n",
            "60000/60000 [==============================] - 13s 219us/step - loss: 150.6425 - val_loss: 151.9980\n",
            "Epoch 20/50\n",
            "60000/60000 [==============================] - 13s 220us/step - loss: 150.3087 - val_loss: 151.6340\n",
            "Epoch 21/50\n",
            "60000/60000 [==============================] - 13s 223us/step - loss: 149.9717 - val_loss: 151.4053\n",
            "Epoch 22/50\n",
            "60000/60000 [==============================] - 13s 217us/step - loss: 149.6346 - val_loss: 151.3647\n",
            "Epoch 23/50\n",
            "60000/60000 [==============================] - 13s 215us/step - loss: 149.3938 - val_loss: 150.9589\n",
            "Epoch 24/50\n",
            "60000/60000 [==============================] - 13s 216us/step - loss: 149.1045 - val_loss: 150.8904\n",
            "Epoch 25/50\n",
            "60000/60000 [==============================] - 13s 217us/step - loss: 148.8915 - val_loss: 150.9730\n",
            "Epoch 26/50\n",
            "60000/60000 [==============================] - 13s 216us/step - loss: 148.6069 - val_loss: 150.5543\n",
            "Epoch 27/50\n",
            "60000/60000 [==============================] - 13s 216us/step - loss: 148.3617 - val_loss: 150.6350\n",
            "Epoch 28/50\n",
            "60000/60000 [==============================] - 13s 218us/step - loss: 148.1753 - val_loss: 150.3051\n",
            "Epoch 29/50\n",
            "60000/60000 [==============================] - 13s 216us/step - loss: 147.9698 - val_loss: 150.1609\n",
            "Epoch 30/50\n",
            "60000/60000 [==============================] - 13s 216us/step - loss: 147.7724 - val_loss: 149.8966\n",
            "Epoch 31/50\n",
            "60000/60000 [==============================] - 13s 216us/step - loss: 147.5611 - val_loss: 149.8795\n",
            "Epoch 32/50\n",
            "60000/60000 [==============================] - 13s 214us/step - loss: 147.4307 - val_loss: 149.8245\n",
            "Epoch 33/50\n",
            "60000/60000 [==============================] - 13s 218us/step - loss: 147.2158 - val_loss: 149.7996\n",
            "Epoch 34/50\n",
            "60000/60000 [==============================] - 13s 215us/step - loss: 147.0176 - val_loss: 149.9469\n",
            "Epoch 35/50\n",
            "60000/60000 [==============================] - 13s 216us/step - loss: 146.8577 - val_loss: 149.6944\n",
            "Epoch 36/50\n",
            "60000/60000 [==============================] - 13s 215us/step - loss: 146.7518 - val_loss: 149.4186\n",
            "Epoch 37/50\n",
            "60000/60000 [==============================] - 13s 214us/step - loss: 146.5554 - val_loss: 149.4122\n",
            "Epoch 38/50\n",
            "60000/60000 [==============================] - 13s 214us/step - loss: 146.4773 - val_loss: 149.7762\n",
            "Epoch 39/50\n",
            "60000/60000 [==============================] - 13s 214us/step - loss: 146.3052 - val_loss: 149.3730\n",
            "Epoch 40/50\n",
            "60000/60000 [==============================] - 13s 216us/step - loss: 146.1531 - val_loss: 149.4910\n",
            "Epoch 41/50\n",
            "60000/60000 [==============================] - 13s 215us/step - loss: 146.0227 - val_loss: 149.3480\n",
            "Epoch 42/50\n",
            "60000/60000 [==============================] - 13s 215us/step - loss: 145.8989 - val_loss: 149.3663\n",
            "Epoch 43/50\n",
            "60000/60000 [==============================] - 13s 215us/step - loss: 145.8087 - val_loss: 149.1855\n",
            "Epoch 44/50\n",
            "60000/60000 [==============================] - 13s 217us/step - loss: 145.6913 - val_loss: 149.3420\n",
            "Epoch 45/50\n",
            "60000/60000 [==============================] - 13s 220us/step - loss: 145.5636 - val_loss: 149.1410\n",
            "Epoch 46/50\n",
            "60000/60000 [==============================] - 13s 212us/step - loss: 145.4245 - val_loss: 149.1533\n",
            "Epoch 47/50\n",
            "60000/60000 [==============================] - 13s 213us/step - loss: 145.3657 - val_loss: 149.0805\n",
            "Epoch 48/50\n",
            "60000/60000 [==============================] - 13s 212us/step - loss: 145.2897 - val_loss: 149.3635\n",
            "Epoch 49/50\n",
            "60000/60000 [==============================] - 13s 210us/step - loss: 145.1547 - val_loss: 148.8374\n",
            "Epoch 50/50\n",
            "60000/60000 [==============================] - 13s 212us/step - loss: 145.0585 - val_loss: 149.0456\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<keras.callbacks.callbacks.History at 0x7f127346eeb8>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 8
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "a13fEZIlS-gD",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 293
        },
        "outputId": "343478b5-9ca2-4c0b-a4a7-be9436845c6c"
      },
      "source": [
        "x_test_predicted = vae.predict(x_test)\n",
        "\n",
        "n=10\n",
        "plt.figure(figsize=(20,5))\n",
        "for i in range(n):\n",
        "  #original\n",
        "  plt.subplot(2,n,i+1)\n",
        "  plt.imshow(x_test[i].reshape(28,28))\n",
        "  plt.gray()\n",
        "\n",
        "  #reconstructed\n",
        "  plt.subplot(2,n,i+1+n)\n",
        "  plt.imshow(x_test_predicted[i].reshape(28,28))\n",
        "  plt.gray()"
      ],
      "execution_count": 13,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAABH4AAAEUCAYAAABd4vGCAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nO3deZgdVbX38bVFQBCNBEgYEhISxjCHEBCZQRlkVFCU6+U6oVe84gwO13m4LyrOE/eKICKCgAKCIERGBSRhCCSQkACBYEiIIJOgIPX+kWbz2ytdle7TZ6ja/f08Tx5WdVWfs/uss+vUKfbaOxRFYQAAAAAAAMjPS3rdAAAAAAAAAHQGN34AAAAAAAAyxY0fAAAAAACATHHjBwAAAAAAIFPc+AEAAAAAAMgUN34AAAAAAAAyNaQbPyGE/UMIc0II80IIJ7arUegu8th85DAP5LH5yGEeyGPzkcM8kMfmI4d5II/NF4qiaO0XQ1jJzOaa2WvNbKGZ3WRmbymKYnb7modOI4/NRw7zQB6bjxzmgTw2HznMA3lsPnKYB/KYh5cO4Xenmtm8oijuMTMLIfzSzA41s9I3QAihtbtMGLKiKELJrkHlkRz21NKiKNbp5+f0xQahL2aBvpgB+mIW6IsZoC9mgb6YAfpiFsr64pBKvTYwswdke2Hfz9As5LE5FpT8nBzmgTw2B30xb+SxOeiLeSOPzUFfzBt5bI6yvjikET8DEkI41syO7fTzoHPIYR7IY/ORwzyQx+Yjh3kgj81HDvNAHpuPHNbfUG78PGhmY2V7TN/PEkVRnGJmp5gx7KumVphHclh79MU80Bebj76YB/pi89EX80BfbD76Yh7oixkYSqnXTWa2SQhhoxDCKmZ2lJld2J5moYvIY/ORwzyQx+Yjh3kgj81HDvNAHpuPHOaBPGag5RE/RVE8F0J4v5ldZmYrmdmpRVHMalvL0BXksfnIYR7IY/ORwzyQx+Yjh3kgj81HDvNAHvPQ8nLuLT0Zw756pmKW9kEhhz01oyiKKe14IPLYO/TFLNAXM0BfzAJ9MQP0xSzQFzNAX8xCaV8cSqkXAAAAAAAAaowbPwAAAAAAAJnixg8AAAAAAECmhrKcO9AzH/3oR2O82mqrJfu22WabGB9xxBGlj/HDH/4wxtdff32y74wzzhhqEwEAAAAA6DlG/AAAAAAAAGSKGz8AAAAAAACZ4sYPAAAAAABAppjjB41x9tlnx7hq7h71/PPPl+57z3veE+N999032Xf11VfH+P777x9oE9Fjm266abJ91113xfj444+P8Xe/+92utWk4e/nLXx7jr33tazHWvmdmNmPGjBgfeeSRyb4FCxZ0qHUAAAC9seaaa8Z4ww03HNDv+GuiD33oQzG+4447Yjx37tzkuNtuu62VJiIzjPgBAAAAAADIFDd+AAAAAAAAMkWpF2pLS7vMBl7epeU9l112WYwnTJiQHHfwwQfHeOLEicm+o48+OsZf/epXB/S86L3tt98+2dZSv4ULF3a7OcPeeuutF+N3v/vdMfYlmDvssEOMDzrooGTf97///Q61Dmry5MkxPv/885N948eP79jzvu51r0u277zzzhg/8MADHXterJh+RpqZXXjhhTF+//vfH+Mf/ehHyXH/+te/OtuwDI0aNSrG55xzToz/9Kc/JcedcsopMb7vvvs63q4XjBgxItnefffdY3zppZfG+Nlnn+1am4AmeP3rXx/jQw45JNm35557xnjjjTce0OP5Eq5x48bFeNVVVy39vZVWWmlAj4+8MeIHAAAAAAAgU9z4AQAAAAAAyBSlXqiVKVOmxPjwww8vPW7WrFkx9kMnly5dGuMnn3wyxqusskpy3A033BDjbbfdNtm31lprDbDFqJPtttsu2X7qqadi/Otf/7rbzRl21llnnWT79NNP71FLMFj77bdfjKuGi7ebLyd6xzveEeOjjjqqa+3AMvrZ94Mf/KD0uO9973sxPvXUU5N9Tz/9dPsblhldzccsvabRsqrFixcnx/WqvEtXXjRLz/Vaqjtv3rzON6xhXvnKVybbOn3AVlttFWO/uixlc/WmU0Qcd9xxMdaydjOz1VZbLcYhhCE/r1+9FhgMRvwAAAAAAABkihs/AAAAAAAAmeLGDwAAAAAAQKYaPcePX95b6yr/8pe/JPueeeaZGJ955pkxfuihh5LjqE/uLV3+2dfCag28zkexaNGiAT32Rz7ykWR70qRJpcdefPHFA3pM9J7WyOsSw2ZmZ5xxRrebM+x84AMfiPFhhx2W7Js6deqgH0+XCTYze8lLXvz/E7fddluMr7nmmkE/NlIvfemLlwAHHnhgT9rg5w758Ic/HOOXv/zlyT6dswudof1vzJgxpcedddZZMdbrK5Rbe+21Y3z22Wcn+0aOHBljnVvpv/7rvzrfsBKf/vSnY7zRRhsl+97znvfEmOvm5R199NEx/vKXv5zsGzt2bL+/4+cC+utf/9r+hqFt9Px4/PHHd/S57rrrrhjrdyG0z8YbbxxjPVebpXPO7rnnnsm+559/PsY/+tGPYvzHP/4xOa4u50lG/AAAAAAAAGSKGz8AAAAAAACZanSp10knnZRsjx8/fkC/p0NUn3jiiWRfN4fQLVy4MMb+b5k+fXrX2lEnF110UYx12J1ZmqtHHnlk0I/tlwZeeeWVB/0YqJ/NN988xr40xA+nR/t985vfjLEOeW3VG97whtLtBQsWxPjNb35zcpwvGcKK7bXXXjF+9atfHWP/edRJfllrLcFdffXVk32UerXfqquummx/6lOfGtDvaRltURRtbVOuJk+eHGNfLqC+8IUvdKE1y9tyyy2TbS2P//Wvf53s47N1eVr6861vfSvGa621VnJcWX/57ne/m2xr6Xor17wYGF/Wo2VbWq5z6aWXJsf94x//iPFjjz0WY/85pdelv//975N9d9xxR4xvvPHGGN9yyy3JcU8//XTp42PgdGoIs7SP6bWmf08M1E477RTj5557Ltk3Z86cGF933XXJPn3P/fOf/2zpuQeKET8AAAAAAACZ4sYPAAAAAABAprjxAwAAAAAAkKlGz/Gjy7ebmW2zzTYxvvPOO5N9W2yxRYyr6qx33nnnGD/wwAMxLlt+sT9a1/fwww/HWJcq9+6///5ke7jO8aN0Po9WfexjH4vxpptuWnqc1tb2t436+vjHPx5j/56hH3XGJZdcEmNdbr1Vumztk08+mewbN25cjHVJ4T//+c/JcSuttNKQ25E7X9+uS3LPnz8/xl/5yle61qZDDz20a8+F5W299dbJ9g477FB6rF7b/O53v+tYm3IxatSoZPuNb3xj6bHvfOc7Y6zXjZ2m8/pcccUVpcf5OX78/Jgw++hHPxrjkSNHDvr3/bx1+++/f4z9kvA6H1Cn5wTJUdW8O9tuu22MdRlv74Ybboixfq+87777kuM23HDDGOvcrmbtmRcRy9P7Accdd1yMfR975Stf2e/vP/jgg8n2tddeG+N777032affQXSuyalTpybH6TnhwAMPTPbddtttMdYl4TthhVfsIYRTQwhLQgh3yM9GhhAuDyHc3fffNaseA71HHrMwnhw2H30xC/TFDNAXs0BfzAB9MQv0xQzQF/M2kP9Ve5qZ7e9+dqKZTSuKYhMzm9a3jXo7zchj0y01cpiD04w8Nh19MQ+nGXlsOvpiHk4z8th09MU8nGbkMVsrLPUqiuKaEMJ49+NDzWzPvvh0M7vKzE5oY7sGZNq0aZXbyi/D9wK/lOx2220XYx2yteOOOw64Xc8880yM586dG2NffqbDvnSYfSfUOY/tdtBBB8VYl0VdZZVVkuOWLFkS40984hPJvr///e8dat2QPGlmfk3PLHNYZfz48cn2lClTYqz9zayey142sS/uscceyfZmm20WYx2qPNBhy34oqw611mVRzcz23nvvGFctNf2f//mfMf7hD384oHYMQSP74qc//elkW4e7a1mBL7drN/3s8++tbg59b2JfbLeq8iPPl0TURG374je+8Y1k+9/+7d9irNeXZma/+tWvutImb7fddovx6NGjk32nnXZajH/+8593tB1N7Itahmxm9va3v73f42bOnJlsL168OMb77rtv6eOPGDEixlpGZmZ25plnxvihhx5acWO7o7Z90V///+IXv4ixlnaZpaXOVeWPypd3KT+VR901sS/++Mc/Tra1RK9qaXa9b3D77bfH+JOf/GRynH6v93bZZZcY63Xoqaeemhyn9xf0HGBm9v3vfz/G5513Xow7Ufbb6hw/o4uiWNQXP2Rmo8sODCEca2bHtvg86KwB5ZEc1hp9MQ/0xeajL+aBvth89MU80Bebj76YB/piJoY8uXNRFEUIoajYf4qZnWJmVnUceqsqj+SwGeiLeaAvNh99MQ/0xeajL+aBvth89MU80BebrdUbP4tDCOsVRbEohLCemS1Z4W/U1KOPPppsX3nllf0eV1VGVkWHUfuyMh1WdvbZZ7f0+EOUTR6Vlv744Z1KX/Orr766o23qoCxzWMWXhqhurobSZrXLo5bU/fKXv0z2VQ2dVbrKmg5f/fznP58cV1VaqY9x7LEv/o+kddZZJznupJNOivHLXvayZN/3vve9GD/77LMranarapdDM7Mjjjgixn4liXnz5sW4myvgacmeL+266qqrYvy3v/2tW01Stcxjp+y+++6l+/xqQVWlljVTixwWRfq9R9/rf/nLX5J9nVyZabXVVku2tYzhfe97X4x9e9/xjnd0rE0DVIs8ltHSDTOzV7ziFTHWVYD8NYt+Pr3lLW+JsS8vmThxYozXXXfdZN8FF1wQ4wMOOCDGjzziK616rmc5XGONNWLsp3PQKSGWLl2a7Pv6178e45pO+9ALPe+L/rpOV9N617velewLIcRYvxf4aQC+9rWvxbjVqSHWWmutGOvqsp/73OeS43TKGV8m2k2trsN7oZkd0xcfY2YXVByL+iKPzUcO80Aem48c5oE8Nh85zAN5bD5ymAfymImBLOd+lpldb2abhRAWhhDeaWb/Y2avDSHcbWb79m2jxshjFjYycth49MUs0BczQF/MAn0xA/TFLNAXM0BfzNtAVvV6S8mufdrcFnQQeczCvUVRTOnn5+SwQeiLWaAvZoC+mAX6Ygboi1mgL2aAvpi3IU/ujOWNGjUqxj/4wQ9i/JKXpAOsdKnxGtblNsZvfvObZPt1r3tdv8f97Gc/S7b90sZohq233rp0n87zgqF56Utf/HgY6Jw+fq6so446Ksa+jn6gdI6fr371qzE++eSTk+NWX331GPv3wYUXXhjj+fPnt9SOpjryyCNjrK+RWfr51Gk6Z9TRRx8d43/961/JcV/60pdi3MH5mIY1XX5WY8/PeXDrrbd2rE3Dzetf//pk+/e//32MdW4rPyfFQOm8MnvuuWeyb+edd+73d84999yWnmu4WnXVVZNtnSPpm9/8Zunv6dLQP/3pT2Os52ozswkTJpQ+hs4908n5oZrssMMOi/GJJ56Y7NMl1nfbbbdk32OPPdbZhqEl/jz2sY99LMY6p4+Z2YMPPhhjnWv3z3/+c0vPrXP3jB07Ntmn3y0vueSSGPt5fZVv7xlnnBHjTs9t2OocPwAAAAAAAKg5bvwAAAAAAABkilKvDjjuuONirEsO+6Xj58yZ07U25Wa99daLsR+qrsNvtbxESwjMzJ588skOtQ7tpkPT3/72tyf7brnllhhffvnlXWsTltFlwP3yv62Wd5XRki0tFzIz23HHHdv6XE01YsSIZLusrMOs9TKSVhx77LEx1tLBO++8Mznuyiuv7FqbhquB9pVuvj9y9O1vfzvZ3muvvWK8/vrrJ/t23333GGsZwCGHHNLSc+tj+GXa1T333BNjv5w4qulS7J6W8vnpCMpMmdLf9Dj9u+GGG2LMtWz/qspY9bpx4cKF3WgOhkjLrcyWLxNXzz33XIx32mmnGB9xxBHJcZtvvnm/v//0008n21tssUW/sVl6nTt69OjSNqnFixcn290scWfEDwAAAAAAQKa48QMAAAAAAJApSr3a4DWveU2y7WePf4HOMG9mdscdd3SsTbk777zzYrzWWmuVHvfzn/88xsNtNZ+c7LvvvjEeOXJksu/SSy+Nsa6WgfbxKxIqHUbbaVq+4NtU1cbPfe5zMX7b297W9nbViV9pZoMNNojxWWed1e3mRBMnTuz353wOdl9VSUk7VpTCMjNmzEi2t9lmmxhvt912yb79998/xrpazcMPP5wcd/rppw/ouXWVmNtuu630uD/96U8x5hppcPz5VMvytJzSl5PoyqSHH354jP0qQNoX/b53v/vdMdZcz549e0BtHw58WY/S/vbZz3422XfBBRfEmJUM6+MPf/hDsq1l4fodwcxsww03jPF3vvOdGFeVvWrpmC8rq1JW3vX8888n27/+9a9j/IEPfCDZt2jRogE/31Ax4gcAAAAAACBT3PgBAAAAAADIFDd+AAAAAAAAMsUcP21w4IEHJtsrr7xyjKdNmxbj66+/vmttypHWT0+ePLn0uKuuuirGvnYXzbTtttvG2Nfonnvuud1uzrDw3ve+N8a+VrlXDj744Bhvv/32yT5to2+vzvGTuyeeeCLZ1jkKdI4Rs3S+rEceeaSt7Rg1alSyXTbfwnXXXdfW50X/dt111xi/9a1vLT3uscceizHLHLfXo48+GmOdn8Jvn3DCCUN+rgkTJsRY50YzS88JH/3oR4f8XMPVFVdckWxr39F5fPy8O2XzjPjHO+6442L829/+Ntm3ySabxFjnC9HP7eFunXXWibG/JtC58D7zmc8k+z796U/H+Ec/+lGMb7jhhuQ4nUdm3rx5MZ41a1Zpm7bccstkW78Xcr6t5pdY1/mxXvWqVyX7dK5dnYf3r3/9a3Lc/fffH2N9T+h3DjOzqVOnDrq9p5xySrL9yU9+MsY6f1e3MeIHAAAAAAAgU9z4AQAAAAAAyBSlXi1abbXVYqzLApqZ/fOf/4yxlho9++yznW9YRvwy7TpMTsvpPB3G/OSTT7a/YeiKddddN8a77bZbjOfMmZMcp0skon20rKqbdHi2mdmkSZNirOeAKn4J5OF07vXDoXWJ5je+8Y3JvosvvjjGJ5988qCfa6uttkq2tbxk/Pjxyb6y8oa6lBHmTj9PX/KS8v/nd/nll3ejOegwLV/xfU9Lyfy5EgPny2Pf9KY3xVhL0EeMGFH6GN/97ndj7Ev8nnnmmRiff/75yT4tZdlvv/1iPHHixOQ4Pf8PN1//+tdj/OEPf3jAv6fnx/e97339xu2i/U+nqTjqqKPa/lw586VT2j9a8bOf/SzZrir10vJ6fZ+ddtppyXG6XHwvMeIHAAAAAAAgU9z4AQAAAAAAyBQ3fgAAAAAAADLFHD8t+tjHPhZjv6zwpZdeGuM//elPXWtTbj7ykY8k2zvuuGO/x/3mN79JtlnCPQ//8R//EWNdGvp3v/tdD1qDbvnUpz6VbOuStlXuu+++GB9zzDHJPl2yc7jR86Ff1vn1r399jM8666xBP/bSpUuTbZ1LZO211x7QY/g6eHTGEUcc0e/P/dwIP/7xj7vRHLTZkUcemWz/+7//e4x1Dgqz5Zc0Rnvocuza39761rcmx2mf07mYdE4f74tf/GKyvcUWW8T4kEMO6ffxzJb/LBxOdJ6Xs88+O9n3i1/8IsYvfWn6VXjs2LExrpoPrR10TkN9z+iS8mZmX/rSlzraDph9/OMfj/Fg5lh673vfG+NWrqO6jRE/AAAAAAAAmeLGDwAAAAAAQKYo9RogHRJvZvbf//3fMX788ceTfV/4whe60qbcDXT5xfe///3JNku452HcuHH9/vzRRx/tckvQaZdcckmMN9tss5YeY/bs2TG+7rrrhtymXNx1110x1uWGzcy22267GG+88caDfmxdstg7/fTTk+2jjz663+P88vNojzFjxiTbvtzkBQsXLky2p0+f3rE2oXMOOOCA0n2//e1vk+2bb765080Z9rTsS+NW+fOkli5pqddee+2VHDdy5MgY++Xnc6fLZ/vz2qabblr6e/vss0+MV1555Rh/7nOfS44rm36iVVqKvcMOO7T1sdG/d73rXTHW8jpf/qdmzZqVbJ9//vntb1gHMeIHAAAAAAAgU9z4AQAAAAAAyBSlXhXWWmutGH/nO99J9q200kox1jIFM7Mbbrihsw1DQoeympk9++yzg36Mxx57rPQxdKjniBEjSh/jVa96VbI90FI1HY56wgknJPv+/ve/D+gxcnTQQQf1+/OLLrqoyy0ZnnTYcdXKFlUlBqecckqM119//dLj9PGff/75gTYxcfDBB7f0e8PZrbfe2m/cDvfcc8+Ajttqq62S7TvuuKOt7Riudtlll2S7rA/7VTHRTP48/NRTT8X4G9/4Rrebgw4755xzYqylXm9+85uT43QqBKahGJhp06b1+3MtjTZLS72ee+65GP/0pz9Njvvf//3fGH/wgx9M9pWV4KIzpk6dmmzruXGNNdYo/T2dQkRX8TIz+8c//tGm1nUHI34AAAAAAAAytcIbPyGEsSGEK0MIs0MIs0IIx/f9fGQI4fIQwt19/12z881Fq8hhFlYmj81HDrNAX8wAOcwCfTED5DAL9MUMkMO8DWTEz3Nm9pGiKCaZ2c5mdlwIYZKZnWhm04qi2MTMpvVto77IYR7IY/ORwzyQx+Yjh3kgj81HDvNAHpuPHGZshXP8FEWxyMwW9cVPhBDuNLMNzOxQM9uz77DTzewqMzuhn4doFJ2759JLL43xRhttlBw3f/78GOvS7nVVFMXNff/NLoczZ84c8mP86le/SrYXLVoU49GjR8fY10+320MPPZRsf/nLX9bNZ3PO46677ppsr7vuuj1qSWc1JYc//OEPY3zSSSeVHqdLBVfNzzPQuXsGetyPfvSjAR3XIVn3xXbQOaL6235BL+f0yTmHOkeht3Tp0hh/+9vf7kZzOmnY9kWda0KvU8zMlixZEuMmLN8+XHPYKv2c1M/nQw89NDnus5/9bIx/+ctfJvvmzp3b7mZl3Rd///vfJ9t6fa7Lf7/73e9Ojtt4441jvOeeew7ouRYuXNhCC9sj5xz6uSBf8YpX9HuczpFmls6j9cc//rH9DeuiQU3uHEIYb2bbm9mNZja676aQmdlDZja65HeONbNjW28i2okc5oE8Nh85zAN5bD5ymAfy2HzkMA/ksfnIYZ4GPLlzCGENMzvPzD5YFMXjuq8oisLMiv5+ryiKU4qimFIUxZQhtRRDRg7zQB6bjxzmgTw2HznMA3lsPnKYB/LYfOQwXwMa8RNCWNmWvQHOLIri/L4fLw4hrFcUxaIQwnpmtqT8EZpj4sSJMd5hhx1Kj9OlurXsq66amMNLLrkk2fZDWNvpyCOPbOn3dAnHqhKVCy+8MMbTp08vPe7aa6+tfL4m5nGgDj/88GRbyy5vueWWGF9zzTVda1MnNCWH559/fow/9rGPJfvWWWedjj3vww8/nGzfeeedMT722Bf/R5KWY/ZCU/LYK8uuDcu36yDnHO63336l++6///4YP/bYY91oTkflnMcqWurl+9fFF19c+nta3rDmmi/O0arvi24brjlsh1tvvTXGn/nMZ5J9X/va12L8la98Jdn3tre9LcZPP/10W9qScx71WsTM7Jxzzonxm970ptLf22uvvUr3/etf/4qx9tkTT+zdFDq55VDPdx//+McH9Dtnnnlmsn3VVVe1s0k9NZBVvYKZ/cTM7iyK4mTZdaGZHdMXH2NmF7S/eWgjcpgH8th85DAP5LH5yGEeyGPzkcM8kMfmI4cZG8iIn9eY2dvM7PYQwgu3lT9pZv9jZueEEN5pZgvMrPx2J+qAHDbfGkYec0AOm4++mAdy2Hz0xTyQw+ajL+aBHGZsIKt6XWdm/S/HYbZPe5uDTimKghw235PksfnIYRboixkgh1mgL2aAHGaBvpgBcpi3Qa3qlaNx48Yl2365vhf4OS50CWN0xhve8IZkW2szV1555QE9xpZbbhnjwSzFfuqpp8b4vvvuKz3uvPPOi/Fdd9014MfHMquvvnqMDzzwwNLjzj333BhrTTQ6Z8GCBTE+6qijkn2HHXZYjI8//vi2Pq8ukWpm9v3vf7+tj4/ueNnLXla6r13zSSCln4s6X6H3zDPPxPjZZ5/taJvQG/o5efTRRyf7PvShD8V41qxZMT7mmGMMzfazn/0s2X7Pe94TY39N/YUvfCHGM2fO7GzDMuA/tz74wQ/GeI011ojxlCnpvMajRo2Ksf8+ccYZZ8T4c5/7XBtaCbM0H7Nnz45x1XdH7QOa29wMeFUvAAAAAAAANAs3fgAAAAAAADI17Eu9dHlgM7MNN9yw3+OuvvrqZLuOS9Pm7qSTThrS77/1rW9tU0vQLlpm8Oijjyb7Lrzwwhh/+9vf7lqbsLxrrrmmdFvLY/359OCDD46x5vOUU05Jjlu2eOQyOiwXzfX2t7892f7b3/4W4y9+8Yvdbs6w8Pzzz8d4+vTpyb6tttoqxvPmzetam9Ab73rXu2L8zne+M9n3k5/8JMb0xbw8/PDDyfa+++4bY19mdMIJJ8TYlwNixRYvXhxjvdZ529velhy38847x/jzn/98sm/Jksasit4oe++9d4zHjBkT46rv7loCq+XQuWHEDwAAAAAAQKa48QMAAAAAAJCp0M2SpRBCLeqjdt111xhfcsklyT6dCVxNnTo12fbDqOuuYnm+QalLDoepGUVRTFnxYStGHnuHvpgF+uIKXHTRRcn2ySefHOMrr7yy283pV859cf3110+2v/SlL8V4xowZMc5g1bxh2xf1WlZXaDJLy3F/+MMfJvu0rPqf//xnh1o3ODn3xbrwqxa/+tWvjvFOO+0U4yGUWw/bvpiTHPribbfdFuOtt9669Livfe1rMdbSxwyU9kVG/AAAAAAAAGSKGz8AAAAAAACZ4sYPAAAAAABApoblcu677bZbjMvm9DEzmz9/foyffPLJjrYJAIBc6PK26L6//OUvyfY73vGOHrUEnXLdddfFWJcvBvpzxBFHJNs6D8rGG28c4yHM8QPUwsiRI2McwotTFi1ZsiQ57lvf+lbX2lQXjPgBAAAAAADIFDd+ALYNnDEAACAASURBVAAAAAAAMjUsS72q6NDHffbZJ8aPPPJIL5oDAAAAAC17/PHHk+2NNtqoRy0BOuvkk0/uN/7iF7+YHLdo0aKutakuGPEDAAAAAACQKW78AAAAAAAAZIobPwAAAAAAAJkKRVF078lC6N6TIVEURVjxUStGDntqRlEUU9rxQOSxd+iLWaAvZoC+mAX6Ygboi1mgL2aAvpiF0r7IiB8AAAAAAIBMceMHAAAAAAAgU91ezn2pmS0ws7X74l6qQxvMutOOcW18rDrl0Gx4taPdeXzKhs9rNxBNzCF9cXlNzCN9MdXEHNIXl9fEPNIXU03MIX1xeU3MI30x1cQc0hd704bSPHZ1jp/4pCFMb1cdaJPbUKd2DFZd2k07WleXNtOOoalLu2lH6+rSZtoxNHVpN+1oXV3aTDuGpi7tph2tq0ubacfQ1KXddWhHHdpAqRcAAAAAAECmuPEDAAAAAACQqV7d+DmlR8+r6tAGs/q0Y7Dq0m7a0bq6tJl2DE1d2k07WleXNtOOoalLu2lH6+rSZtoxNHVpN+1oXV3aTDuGpi7trkM7et6GnszxAwAAAAAAgM6j1AsAAAAAACBTXb3xE0LYP4QwJ4QwL4RwYhef99QQwpIQwh3ys5EhhMtDCHf3/XfNLrRjbAjhyhDC7BDCrBDC8b1qy1AM5zySwyE/Lzlsk17lsO+5yWOb0BfJ4RCfmzy2CX2RHA7xucljm9AXyeEQn5s8limKoiv/zGwlM5tvZhPMbBUzu83MJnXpuXc3s8lmdof87CQzO7EvPtHM/l8X2rGemU3ui19hZnPNbFIv2kIeySE5JIfkcfjmkRw2P4fkMY88ksPm55A85pFHctj8HJLHFbSri0l4tZldJtufMLNPdPH5x7s3wBwzW0+SM6ebL3zf815gZq+tQ1vIIzkkh+SQPA6vPJLD5ueQPOaRR3LY/BySxzzySA6bn0PyWP6vm6VeG5jZA7K9sO9nvTK6KIpFffFDZja6m08eQhhvZtub2Y29bssgkcc+5LBtyOHg1S2HZuSxFXXLIzkcvLrl0Iw8tqJueSSHg1e3HJqRx1bULY/kcPDqlkMz8mhmTO5sZmbFsttuRbeeL4SwhpmdZ2YfLIri8V62JSfdfO3IYWeQwzyQx+Yjh3kgj81HDvNAHpuPHOZhOOexmzd+HjSzsbI9pu9nvbI4hLCemVnff5d040lDCCvbsjfAmUVRnN/LtrRo2OeRHLYdORy8uuXQjDy2om55JIeDV7ccmpHHVtQtj+Rw8OqWQzPy2Iq65ZEcDl7dcmhGHs2suzd+bjKzTUIIG4UQVjGzo8zswi4+v3ehmR3TFx9jy2rvOiqEEMzsJ2Z2Z1EUJ/eyLUMwrPNIDjuCHA5e3XJoRh5bUbc8ksPBq1sOzchjK+qWR3I4eHXLoRl5bEXd8kgOB69uOTQjj8t0c0IhMzvQls1qPd/MPtXF5z3LzBaZ2bO2rM7wnWa2lplNM7O7zewKMxvZhXbsasuGdM00s1v7/h3Yi7aQR3JIDskheez9P/oiOSSP9fhHXySH5LEe/+iL5JA8duZf6GscAAAAAAAAMsPkzgAAAAAAAJnixg8AAAAAAECmuPEDAAAAAACQKW78AAAAAAAAZIobPwAAAAAAAJnixg8AAAAAAECmuPEDAAAAAACQKW78AAAAAAAAZIobPwAAAAAAAJnixg8AAAAAAECmuPEDAAAAAACQKW78AAAAAAAAZIobPwAAAAAAAJnixg8AAAAAAECmuPEDAAAAAACQKW78AAAAAAAAZIobPwAAAAAAAJnixg8AAAAAAECmuPEDAAAAAACQKW78AAAAAAAAZIobPwAAAAAAAJnixg8AAAAAAECmuPEDAAAAAACQKW78AAAAAAAAZIobPwAAAAAAAJnixg8AAAAAAECmuPEDAAAAAACQKW78AAAAAAAAZIobPwAAAAAAAJnixg8AAAAAAECmuPEDAAAAAACQKW78AAAAAAAAZIobPwAAAAAAAJnixg8AAAAAAECmuPEDAAAAAACQqSHd+Akh7B9CmBNCmBdCOLFdjUJ3kcfmI4d5II/NRw7zQB6bjxzmgTw2HznMA3lsvlAURWu/GMJKZjbXzF5rZgvN7CYze0tRFLPb1zx0GnlsPnKYB/LYfOQwD+Sx+chhHshj85HDPJDHPLx0CL871czmFUVxj5lZCOGXZnaomZW+AUIIrd1lwpAVRRFKdg0qj+Swp5YWRbFOPz+nLzYIfTEL9MUM0BezQF/MAH0xC/TFDNAXs1DWF4dU6rWBmT0g2wv7foZmIY/NsaDk5+QwD+SxOeiLeSOPzUFfzBt5bA76Yt7IY3OU9cUhjfgZkBDCsWZ2bKefB51DDvNAHpuPHOaBPDYfOcwDeWw+cpgH8th85LD+hnLj50EzGyvbY/p+liiK4hQzO8WMYV81tcI8ksPaoy/mgb7YfPTFPNAXm4++mAf6YvPRF/NAX8zAUEq9bjKzTUIIG4UQVjGzo8zswvY0C11EHpuPHOaBPDYfOcwDeWw+cpgH8th85DAP5DEDLY/4KYriuRDC+83sMjNbycxOLYpiVttahq4gj81HDvNAHpuPHOaBPDYfOcwDeWw+cpgH8piHlpdzb+nJGPbVMxWztA8KOeypGUVRTGnHA5HH3qEvZoG+mAH6YhboixmgL2aBvpgB+mIWSvtixyd3BgAAAIDhIITy787d/B/uAKCGMscPAAAAAAAAaowbPwAAAAAAAJmi1Au18pKXvHgv8qUvTd+eK620Ur9x1XG6T39uZvavf/0rxs8880yy7+mnn+73OD9EV7cZvgukyvqi77Mrr7xyjP0Q+WeffTbGzz33XIx9f9Pjnn/++RZbDADA4Olnl17LamxW/TkGAJ3EiB8AAAAAAIBMceMHAAAAAAAgU9z4AQAAAAAAyBRz/KDrdN6PVVZZJdn3qle9KsZrrrlmsm/06NExHj9+fIzHjRuXHLfhhhvGeKONNoqxnzvkgQceiPHMmTOTfTfffHOMFyxYEONHHnkkOe7vf/97jHWOEbN0nhHquNvH18uvscYaMT7ggANK9910000xnjVrVnKczuOEwdF86Fw9ZmaveMUrYrz55pvHeMqUKclxkyZNirGf/+f++++PseZt9uzZyXF/+ctfYvzUU08l+7Rv0hdfpOfEstistXOZ76dlc174+Ziq5lRD+/m579Zaa60Yv/GNb0z2rbfeejGeO3dujC+++OLkuL/97W8xJof1pn3Rvxf0fO77s+ZVz6/+Oijn/PvzZNm1rf9M++c//xljne+nam46/xg+Hy/w1zLMJ9QcVfNCaf7Lcm+W5l/fZ2bkH8sw4gcAAAAAACBT3PgBAAAAAADI1LAp9fJDMl9QNfStauh7K49RdVzuQ/D0dXjZy14W47XXXjs5buzYsTHebLPNkn1bb711jCdMmBDjTTfdNDlugw02iLGW+vjX+IknnojxlltumezbYostYnzZZZfF+Pbbb0+OW7x4cYyffPLJZF/ZEN7cc91pfjj0qquuGuODDjoo2adlfzps/c4770yOo9RrcMrKu0aMGJEcN3ny5Bhr2chOO+2UHDdmzJh+H88s7afXXnttjK+44orkuOuvvz7GWh5mluY3977oP6uqho9r39HYDxH322WPp+UNPo+rrbZav7/36KOPJsdp+ayeQ9EZvg+88pWvjLEvnZ04cWKMb7311hhPmzatQ63Ll++nuq0lQ1UlkwM9f/njtGxE+6Xm3sxs5MiRMdbrNrP0vPzQQw/F2JfD+9Kvpqk6n/ryK/38W2eddWLsS4/1tdPPI/9c+pr7Mryy8/Vf//rX5Dh9fK5z+lfVF6vyrTmp+o6oqvqi5lT7pZnZy1/+8tJ9zzzzTIy1//m+qO+THK99ygw0N1Xf+fX1qirJHOhzed3MByN+AAAAAAAAMsWNHwAAAAAAgEw1rtSrathd1RDYsqF7fji6bletPqVD8vw+HU739NNPx/jxxx9PjvvHP/4R49xXQtDXX187XcXLLF2hy5dwadmOri6y+uqrJ8fpiiIPPvhgjP2QaR06qY/t2/vwww/H2Jcl6JBdHW5pxrDaTvFDKXVI9S677JLs06HSWkJSNVQTy/N9R8+Tuvrea17zmuS4Qw45JMZaTqk5M6suidX+ve2228bY9y/t9xqbpX0zx9xXrcxT9VmlpR2jRo2KsS/rKCuP85+fVUPVNcean7vvvjs5rmzFG3SG79tadrnddtsl+/T9oitfDvfVYwa6Ol5ZOZdZ2uf0nKql62ZpDvS85l/zsjJOf6zu86W6ek7w7r333hjruVivicyauapU2fcFs/Qc6s9xOnWBXr/qSnlmaX/Rc2FVWZw/1+pjLFmyJMZ+ygH9njHcDLQv+u+SmlfNna4SbJbmW3Pn+7Z+1/Cfrdp3/O8pXSHV/y2af53GwL+fHnvssX6f16wZ10VVORxoibs/F2qudZ//Xlm14rSe4/Q7v640a5b2dT3OP0anz5mM+AEAAAAAAMgUN34AAAAAAAAyxY0fAAAAAACATNVyjh9fq6f1l1pbt+666ybHldU+m6W1e1Xzw2gdpS4F7usytS7Q195qe3WZy7vuuis5Tmuk582bl+zLbT4SfU309dLaV7Pl5/xRWh+p+dB5fMzMFi1aFOOlS5fG2NfP6vtAl6k1S98v66+/fox9zbsuG+3nFUFn+Jr7SZMmxdjX3mpOZs6cGeMc+lSnVc1zoOc/nXfnoIMOSo6bMmVKjLXP+trnhQsXxtgvQarn5NGjR8fY91mdX2j27NnJPp2bq4lzTvSnLD9+3gldktnPNbHNNtv0G/t5IfTzSOeW8HMI6HP7c7vmf86cOTHWeQf8tv8cR/v565cdd9wxxn55b823nk/1/TEclc3d4+cO0T7gX1u9Htlpp51iXHVtoucy3490Dg8/n4fOU6PXuf76S+cw8ddZek7Vx/OfFVXztzVB1RLrem41M5s6dWqM995779LHXLx4cYz1s9DPj6R9089NovR87Y/z+cjdQJdi1zz69/3YsWNjvOeee8ZYrzXN0r5TtYy3XsNUzYun16X+s0/br3P6mJnNnTs3xnou1u+f/jH949f1mljPp/610xxWzW2o39v8NdCECRNirHO9+r6t52t/baOvnX7n0O+iZun30TvuuCPZd99998VYzw9PPfVUclw7zqHD64wAAAAAAAAwjHDjBwAAAAAAIFO1LPXyQxO1HEtjv9y3Drvzw2h1Gb7x48fHWIeymqXLDOswMj+8SofJ+SVnddjrX//61xj7YYKzZs2K8bnnnpvs09IHHVpf1+F4nh/2qMPwqpYs1CGM/n3w+OOPx1iHxelrbJYOnS0rDzNL3we+tGGzzTaLsb6vNPaP6f+usqUGmzjcuU7867zvvvvG2L/vbr/99hhrOSU5WDHtf36IrS4xfMABB8R4r732So7TIdQ67FjzYmY2ffr0GOtwWLO0fG+PPfaIsS+B2GKLLWKsQ3bN0iG3ej7174Omvi+03VWfn1tttVWyT8sRNKe+rENLWvVc6T+P9Ll8KbYOndbf03O5mdkDDzwQY780cVPzUzd6nvTD1rXMyL+XtA9feeWVMfbLueeuallhfc38NYeey8aNG5fs22WXXfqN/XLuer2pfcX3RS3F8tc3ej7XNmoprVla5u5LDvQ6V1WVevnXra79uapd+jf43GiZ5CabbBJjX9qs17KaJ1/qpWUp/tpTz7VayqJTSJhR6vUC/37Va5ONN9442aflXXrNoSVg/vGrSpSrvv9o/9OSTF+yrfn2n4ta1qSP598zTSmjLiub82WMVX+rnru23HLLGO+8887JcZMnT46x9jdfHqt91pfVagm0lvX561A9R/tzrd4PmDFjRoz1XoBZeh5u9fw5vM4IAAAAAAAAwwg3fgAAAAAAADLFjR8AAAAAAIBM1XKOH0/r56qWntTlMP2y21rDqbW3VcuHa12ur4fUmmk/x4/OGzRixIgY67wxZmk9r19iTmvp9e9vSo20p+3Uv6dqCdKqZZ31dfDLqD/88MMxfvrpp2NcNT/P5ptvnuzTY7VN2gaztGazKblouqo5KfxcBr/+9a9jPNyXHF4Rf27R+nWtLzcz22GHHWK8++67x1jnSDNLz2NXXXVVjP/whz+UHufnC9Hzqx6nddu+jX5+Ga2r17+rqedTs/K2+r9JPwu32267ZJ/Ok6Svs55DzdJ5IxYsWBBjv0Sufk76+fM0J1ovP2bMmOQ4v8wpOsvPNaDXKf7aRnOjczM1Ze7BbtBzjb/m0DlC9HPLLJ1XRK9X/bwW2hfvueeefn9ulvYxveYyS/umzvfj56TQ+ST89ZjOraifrVXn0CadX8vodaMuBW2WfrfQ87DOxWRmNmfOnBjPnz+/398xS689/XcafS/pnGx+GWo/Z95wop/1fs4c/b7o533ReX00p36eIP2OeNddd8XYf6/R86O/vtF26fnW51E/a/28ePoe0qXA/VxA+vh+Dpu60r/bz5mm51d/zbfNNtvE+OCDD46xvwbS+YD1+tIvxa592PdnPYfqd35975il70edY9YsPb/q3+WvsdphhSN+QginhhCWhBDukJ+NDCFcHkK4u++/a1Y9BnqPPGZhPDlsPvpiFuiLGaAvZoG+mAH6YhboixmgL+ZtIKVep5nZ/u5nJ5rZtKIoNjGzaX3bqLfTjDw23VIjhzk4zchj09EX83Cakcemoy/m4TQjj01HX8zDaUYes7XCMURFUVwTQhjvfnyome3ZF59uZleZ2QntapQfNqxD43QYm19yVoeu+SF5uky4Djf1ZSM6XPbOO+/stw1m6ZAzPxRXh89vu+22MfZlZVqa4IfA6pA83dfqUNle5FFpbrQcwA8f1uHJvrxEy6x0OLVfZlSHWep7yQ/11Bz6sgTdp8tt+vdmValXB4Y1P2lmj7ifdS2HdeFLJrVURMtQzMyuuOKKGNelHKHXfbGMX/ZV+5gv4dLlhvX19yWxf/zjH2N83nnnxdjnSXPjz906VF2XtvRlKNpGX77il6Nvg1r1RX39fLmALiGtn03+WP28u/3225PjZs6cGWM9H/qh1zps2p//9Ln0fOuHaOswZ/+ebPfw9Lr2xU7Tvq192SzNhy+jvvTSS2Osn7M9LuHpeV/U97YOzffnza233jrGWi5rlpbu6Ovpz5XXX399jP/85z/HWK9xzaqvfZQvIVLa1/31trZL3wv+vDzQ90ZT+qJeh06aNCnZp98n9LNq3rx5yXFa6qWfb76sQz/H/PcH3dayEZ8nbYe/Vs79GlX7pX/9tFRcv6eZpX1R38/+tb3xxhtjrJ+ZOo2H55dzL1vC3ZdAV50T9P2l5wH9PmWWXp+toCSzln3Rv3Z6/aHLt5ulfVM/0/w1hfbFa6+9Nsa6vLpZWv7u+5FeY+29994x1ushszTX/jqtrETPT1HRjj7b6uTOo4uieKEA7iEzG111MGqLPDYfOcwDeWw+cpgH8th85DAP5LH5yGEeyGMmhjxrUFEURQih9BZUCOFYMzt2qM+DzqrKIzlsBvpiHuiLzUdfzAN9sfnoi3mgLzYffTEP9MVma/XGz+IQwnpFUSwKIaxnZkvKDiyK4hQzO8XMrKrDu99JtnXIqg6B0qHpZulQLF9yoPu0lMAPyZs7d26/v+PLRPQxNtpoo2SfrtClZWD+MXTYoJ9BvEurRQ0oj+3IoQ5x07I5//rrUD4/ZLhsOLU/Tofy6evvh3rq8Fg/rFJff22jb6/+Lb4MoR0legPQ0b5YF5rvQw45JNmnpSG+RMWvTlRjHeuLVbRP+aGn2nd8ed3UqVNjrP1NV7YwM/v9738f47LSALO0FMufJzWHupqFL7/V8iFdscNs+fKxF/i/eYhq0Rf9UGYtN9EVJ8zSVQ/1tdXhz2blq675XFWV/5Sdi325dQ30pC92kw5B32effZJ9WmKt10NmZrfddluM/eduzXS1L5Zdc/jrCi0/8NeNSqccmD59erJPS730OH/Nq+dUfz7Uz1N9L/hSCl25S1cQM0vLAKvOCUO89ul5X/TnUz13bbDBBsk+zb2WyfkV1/R11ZVI/XPptpaTmKXndT2P+zxp+ZD/vOtSiWZX+2LZ94SqlZ41NkuvKXXlrhkzZiTHaV/Uz0+/ip72K/8ZrP1Fc6zvEbO0j/lSL32v6Xdkv8ptVUn4AN4LXeuLZddl/jNHr+t8X9TSL82Hlq2bmV122WUx1qkJ9JrHLP1+56cG2XTTTWOsq4b5VaA1bzfffHOyT1cK03Orfy+1Q6ulXhea2TF98TFmdkF7moMuI4/NRw7zQB6bjxzmgTw2HznMA3lsPnKYB/KYiYEs536WmV1vZpuFEBaGEN5pZv9jZq8NIdxtZvv2baPGyGMWNjJy2Hj0xSzQFzNAX8wCfTED9MUs0BczQF/M20BW9XpLya59Sn6OGiKPWbi3KIop/fycHDYIfTEL9MUM0BezQF/MAH0xC/TFDNAX8zbkyZ27QWsPtc5O61rN0tpGX5+nx2qdo6+H1fkktJ7Q1z6PHDkyxn5J23HjxvW7T2tFzdJaQ1+frTWOPV4qtS30byhbqt4srcn1c+ZoraOvhVZa06y1mFtssUVynC7hqDW9ZulyfZo3v5yj5q1qqeEccthLWs9++OGHJ/u0r1x99dXJPj8PDFJl9fBmab35XnvtlezTecx0OdorrrgiOU7np9C6ZT8PhPYdv3yl9kWdY8vXwOucGXp+Nlv+HJ0zfy6bMGFCjP2Sp/oa6rnML+Ot/UjPZf49o/OF+Ndc58bQx/D51nb4z90uzX2XJe3r22+/fb+xWfWcXffff3+Mh/Pr7+eg0H6g50a/TLQu5+7PUXr9qq+zn5NC5xLRc6M/p+pcIn4eLe2Lvh1K5+zS9pml12N6TVc112ET6bw9ZulS3/6aUj8z9T3h57TTXOlxfk4QnZfGz1Gp8zbp9xt/zqz6jNccNjVPvi/qts5XpnkzM9tkk01irH3WLH2ddJ4c/x1Ol04v6w9maV59fnQuLv3u4h9D+7r/7qvb+nnqr3+b+Pmpr5e/phg9+sWFxfwcWPp7ep3jr1GnTZsW47J8mqXvEX3vmJkdeOCBMda5Y3XuJTOz2bNnx/jWW29N9ul7S6+/OpGnVuf4AQAAAAAAQM1x4wcAAAAAACBTjS718kvVaRmALwUqK/XyJVY6PEyH3fkhZjpsUJfoNDPbaqutYqzD+HSYl1laNlTzpVGHTHPohyQrff39kGHdLhseaZaWG2huJk+enBynQ/L8ENiFCxfGWMsBtazFbPn3INpHh+zut99+MfbDLHUZxBtuuCHZ15ThrN2kr6v2N+1TZmYTJ06M8ZZbbpns0/6ipQd++OojjzwS46qyu7JSoir+/KBLe1aVelUtYd9U+nf4Ui8tk/TL2uu5U4ej6+/4x9ff8cPnN9988xhvuOGGyT4dlq3D0X0JifLD4jXnVaW1WJ6WPbzxjW+MsX4OmqV99pprrkn2Pf74451pXMP468uyZYV1aV+z9LX2j6HXg3pdUdWftbzBX6Pqcb6URfumlgz5/qbXx768pOzaJ7fPXO03Zml+/dLceu2p51M995ml5zzNr37mmqXfJbSU2bdDr0v9eVE/q/35X6/Fc/kOUvZZ5XOg1wj+OkDf6/oYOo2HWVrK40uWlZYL+usgzY++f3z/0hxXlXppHv13rSb2zarPeT0/6fd/s7RsS3Pjy/W0T6y77rox9mWXeu5+7Wtfm+x79atf3W/75s+fn2zr95M5c+Yk+wZ6rdwOjPgBAAAAAADIFDd+AAAAAAAAMtWIUi+lQ9V8mZbu88NodcibDt30qx3ocFkdAuaHSOrwWL9yg64kpcN3r7vuuuQ4XXnM/y06XC+3Fb5U1d/jc6hDIjVvuvKBmdnGG28cYy3v8sNodfizXyFIV9XQHOrM+l7V6gI557BTNMcHH3xwjP1w9D/84Q8xXrBgQecb1nD6XtQ+5s9x22yzTYx92Y4OVdchq/fee29yXFlpjh9WXjWct2zotj93699SVU5atSJgU1WVrOmQcX+cvp5azle1YqaWN/hShzFjxsTYryDmV8d5gV9BTPPjyw/1fTLQ8uHhyudaS2R33XXXGPt+r6XN/pqF8rr+6TWllpBsuummyXFaSuCv+fR8pmU8vgxd86h9yn8uaqmCLwPTkhJ9Lr8KjfYrX36QS2lQf6pWwqrap/nQ0qwddtghOU4/TzUX+t3BLL1m9eW3mm/NhT93a5t8Xy9bYbfJ/Vw/PzRX+jqbLZ87pSs6ab/UVfnMli8fe4Ev09LX1n/e6Wervu6LFi1KjqvKsZ5LcvgsLFsF2peF6/c2X8Kl/UPfB/5cqH1R+4q/ttHVL/fYY49kn7635s2bF+Pf/va3yXE33XRTv2036+60IfldAQMAAAAAAMDMuPEDAAAAAACQLW78AAAAAAAAZKpxc/xo/aLWYfp9fh4VPVaP8zWvWh+qNc2+HlSX9fNLduqxd999d4x1iXCztJ7a/y05zwOjf5ufh6BqeWld9lJra3VOH7N0bhKtkfbLcuoSf76etmz+Jf8+0HrRqrreXOqnu0mXhtblEn1980UXXRTjbtbJNpWe4zT2ywbrXBK+7yxevDjGOh+Wr8Eum9enan4IP1eFtkvrsXW+DLO0jw20L+ZQD+/5uUPuu+++GPs6eK1j1+VK/XlZ81U2z46Z2ate9aoY61wnXtXcUiqH5Wh7xb+uBx10UIz1+sWfT6+99toYaz834/V/QdV8Ojp3oM6hZZZeP/hrCZ0nQucV8fMElc0d6Pu9P2crvbbSJeF9v6+6bimbK87PodbE652qORl1jhZdMtosfR/oPKBTpkxJjtO5IvW5qt5X/vNZz8n6e37uO/+Yquy9VPU+qBvfVn1v6/W5/yzR817VeU4/I/2cPjo/lr7v/Vygei72+/S7xmOPPRZj32/0c9z/LWV9yRnwFQAAF8JJREFUsWruvzoru0bzr51+p/ZzkOk+Pcf5z8Wy184v56592M9fqO8lXbL9lltuSY7T66+qa9Sy9lUdNxiM+AEAAAAAAMgUN34AAAAAAAAy1bhSL+WHwlUN9S8b/uaHUelxOhRXh/SZme22224x9sP/dLjezTffHGNd9tjM7PHHH++37bkrG1pnlg5L9UNbdeislgFNmjQpOU5LVHSotZZ2mZnNnz8/xroEn1k6/FL5pQB1+b+qodVlpSZmeZabtMIPST700ENjrMuc+lzpcEpeyxUrO//5soSqssuykltfwlW2XGzVEGTfxzT3Wsbpl77Vob4PPvhgsq9sydRczrv6dzzyyCPJvhkzZsTYl8fp8rR6Lvalx7rth1QrfW4/LFuXI9bPVn+cfi7691PVeRQpvxztvvvuG2P9bPV95eqrr46x/8xE//Scoq/Z9OnTk+OeeOKJGPvlpZWeA/35UPuO9kt/TtWSBn/+1nK0qr6opQlV5SW5fe7qucWXQt57770x1nOrWVoeoq+/Ly/R7yf63vHXwzodhF/+WfOmn93+nKmlaX6f5q2pn4v+fV92beGvGzUnvnRHS5a1r2hslpYQVZVY6fcEX4qnj6E5rlrOXR/PLH0v6OdzU0u9lL4X/bWHvmf9NYueh7WUz/cx/d620UYbxdh/r9QpRXz/uOOOO2J8wQUXxFinejFLr0P9Y5Rd23SiLzLiBwAAAAAAIFPc+AEAAAAAAMgUN34AAAAAAAAy1eg5fryqpSe1llVrdn39nNZEag22zoVgZjZ58uR+f8csreu76qqrYvzwww8nx1XV1DapxnYgyuYV8Uuaau3qqFGjkn3bbrttjHfYYYcYjxkzpvQxtE7W1/jec889MfbzYmgtvtaO+nmHlJ8jxT/mC/z7RetWc6uVHwy//PP+++8fY31dzjvvvOQ4v0Q1qpX1RT+XRFVttfK11cr37/4e2yytt/dzpm233XYx1qWNfX/TOm5d+tQsnTdG30u5nGf1b/LzsixYsCDGV155ZbJP5/TQ2velS5cmx+lnpi4b7ee/0NxNmDAh2bfhhhvGWGvd/VwG+ph+ToqmzkPRLZrDPfbYI9mnc9/p6+iXnNV5S4bz51EVf87T84u+n//4xz8mx+l8Sv7zTuf60GsYPyeF9j+dv8XPa6nzeflrJJ3XQueknD17dnKcnlP1bzRL+7C2yffZJqqa40dfE59fnQNGX2M9B5ul14b62ernWNK5mXxf1Dnu9DX3527Nk/8cL5uDr0n861L2+e6XbNdrnyVLliT7dD467Zd+jh/NcVUf0DlifV/Ux9Q+7PubnnP891vNa9VxTeffo/r3+fOfvn7aB/xcovr66zyyu+66a3KcnpP9d8kzzjgjxn/6059irOdW316vm/2PET8AAAAAAACZ4sYPAAAAAABAprIt9fLDpnQYmA6d9cfpcC5d4u8Nb3hDcpwOD/NDK88999wY6zBqPxRtuC5Nq0Pt/LKEWl6nwyPN0iXcJ06cGGM/1FPzoUtv3n///clxVSUqOsRWl3b0dNimL+3SpT21TbpEp9+uKqnJkQ631VI+M7Px48fHWF+jSy+9NDkuh6HlvaKlWH6ZUe2nvnxIh79Xlc7qUGhfsqC0LGGrrbZK9u2+++4xHjduXIz9UOibb745xn4Ivg6hz718xX/OPPTQQzHW4ehmaemJ5kqHRvvf09jnW5c81eXDzdL3UNkwbLM0P7nnqt00h/vtt1+yT0sWtM9ef/31yXF+eDqW59/3+h6eP39+jH15/+233x5jf77VMnItffXPpecy3TdixIjkOJ2ewC8dr+cEnZrAl3rpcb4kU9uh11K5Xcv6azItLb/hhhuSfVp6p9evfil2PU5fR1/qpeVcWiprVr4kfFV5bNVy7k1VVf5TtqS3WXoO9O9tvXbXfun7rH5P0NfWl6FrXv0+bb9+R/FTGOhj+O8a+rdoX8whv1X0tfPvbc295sm//vpd8ogjjoixln2ZpdcsZ555ZrLvnHPOibF+1/Ovf13OjYz4AQAAAAAAyBQ3fgAAAAAAADKVVamXqlrVS8tL/AzfOsT2sMMOi/H222+fHKdDtnTlLjOzs88+O8Y6PKwuw7x6oew198PuBroqjA6/9EP8dEikDnHWIX3+93wJhD6GlsP4IdPKDyXV4aO6qoMvMdNh4n5f7u8ZLf957Wtfm+zTVaa0ZHLWrFnJcbm/RkPlV5FT+vr7Fbj0ddVV7szSVTB0yLlf9a5s+LNfuUvLu3baaadkn65GpO2dM2dOcpwOt9VyC9/G3N8vVWUofoUaLUHQ17ZqpRQ9TodQ+8fz51TNgbbJlzcMp9Uu20H7t5aG6MqXZmmu9LPKl84Ot3LjVvj3ob7XtUTDXxPo9aUvcy9bAdGXyOp2VWm8nrN9v9dSEb020ZVOzdJS+apVhnIuKfF/m76u/hynZR76eeffL/raaf/1udbf8+XW2i7/PabsMYbD+VNfWy2p86+flrT6FU11u+r7ip5TtcxWy2rN0qlB/DlBr5G01MuXn+m2vx7Tx8i5L1apulbQfuVXVTvyyCNjrN/z/es4bdq0GP/kJz9J9mlJbxNef0b8AAAAAAAAZGqFN35CCGNDCFeGEGaHEGaFEI7v+/nIEMLlIYS7+/675ooeC71DDrOwMnlsPnKYBfpiBshhFuiLGSCHWaAvZoAc5m0gI36eM7OPFEUxycx2NrPjQgiTzOxEM5tWFMUmZjatbxv1RQ7zQB6bjxzmgTw2HznMA3lsPnKYB/LYfOQwYyuc46coikVmtqgvfiKEcKeZbWBmh5rZnn2HnW5mV5nZCR1pZRuU1bb6+S90HpjXv/71Mfb12AsXLozxd7/73WSfLoFZl5raoihu7vtvV3JYNa+I1sz62tqRI0fGWJduNkuX1xs1alSMq5aQrlqSWvlaba2Z1eXc/ZKpWmurz2uW1gPr4+t7xyytCV3B++XZbuex03TZzB133DHZp3MK/N///V+MfX1z0/S6L5bV/Pt5IMqWCDUzW2eddWI8efLkGPs+oPOF6HwUvm/vvPPOMfbz/2j7dc6Jiy++ODlu+vTpMa5aIrxN5+RG9kU/953P+QsGev729LX1y8rrtj6Gf7yq/Gi72pHHJubQ09dvu+22i/G6666bHKe5/t3vfhfjuXPnJsfV5ZplEHreF/X84t/3SnPg3/e6rXOH+HlfdJ4gnRfIH6fnW78MtZ7Ptb1+Hp9HH300xv7c0W69zuFAaf/w82Hp66Xzy1TN06S59nOmaZ789avu0+8ner3qH7PqvN4mteqL+ppVzePpv9/pdYz2Hd8H9Pc0p37e0arXvWz+Qf/e0veTf/x2n7N7ncNWVL0GOsfS4Ycfnuzbb7/9Yqzn0xtvvDE57qtf/WqMdR43s2bM66MGNblzCGG8mW1vZjea2ei+m0JmZg+Z2eiS3znWzI5tvYloJ3KYB/LYfOQwD+Sx+chhHshj85HDPJDH5iOHeRrw5M4hhDXM7Dwz+2BRFMn/FiiW3Wrr93ZbURSnFEUxpSiKKUNqKYaMHOaBPDYfOcwDeWw+cpgH8th85DAP5LH5yGG+BjTiJ4Swsi17A5xZFMX5fT9eHEJYryiKRSGE9cxsSfkj1IsOyfPDIt/+9rfHWJcP96UD5513XoxvvvnmZF8dh311O4d+2F3ZEH2/hKmWR1UtbanDL/2Qaf29siWEzdKhnn4oprZDl2b0pRG6JKQf4q1DS/W5fTv8sM0qOfRFfa11+UQtHzJLlwzXPtbpIeed1uu+qNv63vNDyR944IEYb7PNNsk+LYmdMGFCjHU5b7N0OLr2KS3xM0uHTPtSPl3iVJfU9MtQa4mtHybdifKVHPpimYGev72yJanN0hIVfS/4cl99Ln9ebvdnaw451NdSh63r622Wflb94Q9/iHFVaVJT1CmPVf1joPv0fe/LfzTfWha78cYbJ8dtuummMfbnWy0b0feFnmvN0mukTpcA1imHrSoro676DK6in2OaJ7P0Okj7ur+O8ueBVtoxGL3Oo/5N+hpV5cB/zuh3CL028aXsWkKk3yX1u6NZer3k86PnX23vgw8+mByn10X+c7Ddeex1Dlvhvwdqqetee+0V4ze84Q3JcZpDnUrg5JNPTo6bNWtWjOv4HX8wBrKqVzCzn5jZnUVR6CtxoZkd0xcfY2YXtL95aCNymAfy2HzkMA/ksfnIYR7IY/ORwzyQx+YjhxkbyIif15jZ28zs9hDCrX0/+6SZ/Y+ZnRNCeKeZLTCzN3WmiWgTcth8axh5zAE5bD76Yh7IYfPRF/NADpuPvpgHcpixgazqdZ2ZlU1Jvk97m4NOKYqCHDbfk+Sx+chhFuiLGSCHWaAvZoAcZoG+mAFymLdBrerVZFqzqTXSU6dOTY7TJaX1d+67777kOF0O1S8TiGpax+rnFdFl8rSm0izNm9Y+6xw8Zmld+uLFi2Ps59bROnc/N4XOfaI1uEuXLk2O07prXcrTzGzRokUxvuuuu2L8yCOPJMc1fc6awdJ+NWXKi/O/+SU1ddl7zV3V8uRYXtXro+9z7Q9mZjNnzozx+uuvn+zTPrf55pvHeOzYsclxWgut73PfF7Wefc6cOcm+yy+/PMZXX311jHUOIrPll5xHZ+j7ydfV6zwHo0aNSvbpXD4654h/z2j+/XxPWJ7O4aFzSfg58rR/63xY6I2qOUc09nO0jBw5MsaTJk2K8d57750ct9VWW5U+14IFC2J82223xdgvU6yfD1WfIwOdA2y48q+Jfi7qPt9n9TOtaglvnQdKzwFmZiNGjIixzps43FT1N38Nrq+1xn6On7XXXjvGOqfWrrvumhyn+/ycXfPnz4/x3LlzY1x1fdPqHHy50e9t+j43M9tll11i/Na3vjXG48ePT47Tz8Xf/OY3Mb7pppuS43K6vhzwql4AAAAAAABoFm78AAAAAAAAZCrbUi9fDqJD0tddd90Y77NPWrKoS+1pydCMGTOS4+bNmxfj4Vaq0wod2qpDJ315lL6uft8tt9wSYy0V0BIwszT3+lw+Tzok0i9pq6VkuoS7H26rv+eHAmo5iw4n9EvCN31pwKHQIed+mPk999wT424uK5s77QfaV3x/02HHVe9tLRvZYIMNkuN0KO7jjz8eY82tmdmNN94Y4yuvvDLZp8dq3+n0kqZ4Udlr68+Hyi/TrvR9p8Pl/e/5kkAsT/uBlsmNGzcuOW727Nkx1jLa4fz5U1faP3yply4/PGHChBj7ksk11lgjxvo5a5ae27XUxE9bUHVtq2VJnHsHp6ysz1+H6rQA/vro4YcfjrGWIPlpC/R86r8XoX/6Oul3R98XtZx5++23j/HkyZOT47Q8008bot959BztP/uqyi61vTmXffnS8pe//OUx3mKLLZJ9BxxwQIx1OgL//ev666+P8a9+9asYa/8yy+tzkhE/AAAAAAAAmeLGDwAAAAAAQKaGTamXDnfUWe91tnWzdHisruCkpQhm6RDM3IbTdZoOH/bD7nSosR9qp0OLy2KzdKirvg/8cWUrDvltPc4P9ytbnaG/bSyjr9lFF10UY19OqfyKU2idvi+1hMuXc2nf1NIQM7M///nPMdZh5r7sUp9L+7ZfzU+HuFeVZKJe/PlQV1GcPn16sk9LrLXsz78XKEcYHC0J+PznPx/j008/PTlO+7OuOEn/qoey8hJf3qDlJnpNoysj+m3fF2+++eYY60pPg1m5JqfSh7rwfVE/MzVnZmZXXXVVjDfbbLMY33333clxen4Yzn19oKvomaX9qqyMymz5Fbpe4PuiluBed911yT4tNdKydl/2V5W7qr+l6fT8568vX/nKV8Z4k002SfZtueWWMdZrSl9qp6+/lsTmPIULI34AAAAAAAAyxY0fAAAAAACATHHjBwAAAAAAIFPZzvHj6xx13hetC/R1lDrfwMyZM2OsdYBm6fLSaB/NW9W8O2guzbHOO6HLyqL3qua20qVFORcOL9p//XLuWiOv83d5OmeXLi1tZvbEE0/EmHP+ium8LDovocaov7K51/yyzjp/yA033BDjWbNmJcfp3D2PPvposk/n4tL+5s/zuc0X0jR6fvXz7J1xxhkx1vlN/Pwyugw859MX6Xvbv+/1e6Fe3+jcaGbpfEraT6s+F7XvmaXnaZ071s+3NVz7os6r5OdUWnXVVWPs39v6Wur3+t/+9rfJcdOmTYuxXpfk/Hoz4gcAAAAAACBT3PgBAAAAAADIVOjmcKYQQs/GTq255poxPuyww2K89dZbJ8fpsD5dLtEvh/nYY4+1uYWdVRRFW9bI7WUOYTOKopjSjgcij71DX8wCfbEfOiz7ZS97WbJvrbXW6vc4LTUxS5d67/SS0fTFLGTdF/0S0gNdalr7TtVS1gP5eTfQFwdHp69YZZVVYly1JHwX8ptFX/R9qYz2xVaXVK9jSVGd+qLmQku7zNLv9RMnTkz2afmjloFpaZeZ2UMPPdTvcRko7YuM+AEAAAAAAMgUN34AAAAAAAAyxY0fAAAAAACATA2bOX5WWmmlGK+++uox9svD6TJ8Guuy02b1rMusUqeaTbQsi/rp4Y6+mAX6Ygboi1mgL2aAvpgF+mIGmtIXdf4fnefKLJ3jTOPM5vGpwhw/AAAAAAAAww03fgAAAAAAADL10hUf0lZLzWyBma3dF3eNDu/qWz62620o0Y12jGvjY/UshyWGUzvancenbPi8dgPRxBzSF5fXxDzSF1NNzCF9cXlNzCN9MdXEHNIXl9fEPNIXU03MYcf6ok658o9//GMgv1KHPHarDaV57OocP/FJQ5jerjrQJrehTu0YrLq0m3a0ri5tph1DU5d2047W1aXNtGNo6tJu2tG6urSZdgxNXdpNO1pXlzbTjqGpS7vr0I46tIFSLwAAAAAAgExx4wcAAAAAACBTvbrxc0qPnlfVoQ1m9WnHYNWl3bSjdXVpM+0Ymrq0m3a0ri5tph1DU5d2047W1aXNtGNo6tJu2tG6urSZdgxNXdpdh3b0vA09meMHAAAAAAAAnUepFwAAAAAAQKa6euMnhLB/CGFOCGFeCOHELj7vqSGEJSGEO+RnI0MIl4cQ7u7775pdaMfYEMKVIYTZIYRZIYTje9WWoRjOeSSHQ35ectgmvcph33OTxzahL5LDIT43eWwT+iI5HOJzk8c2oS+SwyE+N3ksUxRFV/6Z2UpmNt/MJpjZKmZ2m5lN6tJz725mk83sDvnZSWZ2Yl98opn9vy60Yz0zm9wXv8LM5prZpF60hTySQ3JIDsnj8M0jOWx+DsljHnkkh83PIXnMI4/ksPk5JI8raFcXk/BqM7tMtj9hZp/o4vOPd2+AOWa2niRnTjdf+L7nvcDMXluHtpBHckgOySF5HF55JIfNzyF5zCOP5LD5OSSPeeSRHDY/h+Sx/F83S702MLMHZHth3896ZXRRFIv64ofMbHQ3nzyEMN7MtjezG3vdlkEij33IYduQw8GrWw7NyGMr6pZHcjh4dcuhGXlsRd3ySA4Hr245NCOPrahbHsnh4NUth2bk0cyY3NnMzIplt92Kbj1fCGENMzvPzD5YFMXjvWxLTrr52pHDziCHeSCPzUcO80Aem48c5oE8Nh85zMNwzmM3b/w8aGZjZXtM3896ZXEIYT0zs77/LunGk4YQVrZlb4Azi6I4v5dtadGwzyM5bDtyOHh1y6EZeWxF3fJIDgevbjk0I4+tqFseyeHg1S2HZuSxFXXLIzkcvLrl0Iw8mll3b/zcZGabhBA2CiGsYmZHmdmFXXx+70IzO6YvPsaW1d51VAghmNlPzOzOoihO7mVbhmBY55EcdgQ5HLy65dCMPLaibnkkh4NXtxyakcdW1C2P5HDw6pZDM/LYirrlkRwOXt1yaEYel+nmhEJmdqAtm9V6vpl9qovPe5aZLTKzZ21ZneE7zWwtM5tmZneb2RVmNrIL7djVlg3pmmlmt/b9O7AXbSGP5JAckkPy2Pt/9EVySB7r8Y++SA7JYz3+0RfJIXnszL/Q1zgAAAAAAABkhsmdAQAAAAAAMsWNHwAAAAAAgExx4wcAAAAAACBT3PgBAAAAAADIFDd+AAAAAAAAMsWNHwAAAAAAgExx4wcAAAAAACBT3PgBAAAAAADI1P8Hn4Q24K8p8vMAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 1440x360 with 20 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    }
  ]
}