identificador-de-imagens.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Copy of minicurso_pytorch-template.ipynb",
      "version": "0.3.2",
      "provenance": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/isacmoura/FLISOL2019-introducao-ao-pytorch/blob/master/Copy_of_minicurso_pytorch_template.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "metadata": {
        "id": "scRyX_BcrWHn",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "# Importando os pacotes necessários"
      ]
    },
    {
      "metadata": {
        "id": "Hr6-ilP_uC8q",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 239
        },
        "outputId": "c8f8bd4c-a2d8-4dc6-c305-e212f1ac6831"
      },
      "cell_type": "code",
      "source": [
        "# Primeiro, vamos baixar um arquivo para simplificar o processo de \"plot\" das imagens\n",
        "!wget https://raw.githubusercontent.com/isacmoura/pytorch_scholarship_challenge/master/lesson4/helper.py"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--2019-04-27 12:59:57--  https://raw.githubusercontent.com/isacmoura/pytorch_scholarship_challenge/master/lesson4/helper.py\n",
            "Resolving raw.githubusercontent.com (raw.githubusercontent.com)... 151.101.0.133, 151.101.64.133, 151.101.128.133, ...\n",
            "Connecting to raw.githubusercontent.com (raw.githubusercontent.com)|151.101.0.133|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 2813 (2.7K) [text/plain]\n",
            "Saving to: ‘helper.py’\n",
            "\n",
            "\rhelper.py             0%[                    ]       0  --.-KB/s               \rhelper.py           100%[===================>]   2.75K  --.-KB/s    in 0s      \n",
            "\n",
            "2019-04-27 12:59:57 (52.4 MB/s) - ‘helper.py’ saved [2813/2813]\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "S9uZQMid4gEw",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        },
        "outputId": "c233d192-2cc0-4f21-90f5-744cb8fa0eff"
      },
      "cell_type": "code",
      "source": [
        "!ls"
      ],
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "helper.py  sample_data\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "NY7zTCXxrN0t",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "import torch\n",
        "from torchvision import datasets, transforms\n",
        "import matplotlib.pyplot as plt\n",
        "# Para renderizar as figuras diretamente no notebook\n",
        "%matplotlib inline\n",
        "# Para renderizar as figuras com a resolução máxima\n",
        "%config InlineBackend.figure_format = 'retina'\n",
        "\n",
        "# Importamos nosso arquivo de ajuda para abstrair conceitos desnecessários neste momento\n",
        "import helper"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "HAu8qBg3u5iU",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "Agora iremos construir nossa rede neural para identificar dígitos em uma imagem.\n",
        "\n",
        "O dataset utilizado será o MNIST, que já é provido pela biblioteca que estamos utilizando. O dataset consiste em dígitos manuscritos em escala de cinza. Cada imagem é 28x28 pixels.\n",
        "\n",
        "Vamos importar nosso dataset e realizar algumas transformações."
      ]
    },
    {
      "metadata": {
        "id": "H7Idx9wHuW92",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 294
        },
        "outputId": "0e15f68b-d187-4ff6-f981-3a50f4a544e5"
      },
      "cell_type": "code",
      "source": [
        "transform = transforms.Compose([transforms.ToTensor(),\n",
        "                              transforms.Normalize([0.5], [0.5])])\n",
        "\n",
        "# Dataset de treino\n",
        "trainset = datasets.FashionMNIST('~/pytorch/FMINST', download=True,\n",
        "                                 transform=transform, train=True)\n",
        "trainloader = torch.utils.data.DataLoader(trainset,  batch_size=64,\n",
        "                                          shuffle=True)\n",
        "\n",
        "# Dataset de teste\n",
        "testset = trainset = datasets.FashionMNIST('~/pytorch/FMINST', download=True,\n",
        "                                           transform=transform, train=False)\n",
        "testloader = torch.utils.data.DataLoader(testset,  batch_size=64,\n",
        "                                         shuffle=True)"
      ],
      "execution_count": 9,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "\r0it [00:00, ?it/s]"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Downloading http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/train-images-idx3-ubyte.gz to /root/pytorch/FMINST/FashionMNIST/raw/train-images-idx3-ubyte.gz\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "26427392it [00:01, 13713682.74it/s]                             \n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Extracting /root/pytorch/FMINST/FashionMNIST/raw/train-images-idx3-ubyte.gz\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "\r0it [00:00, ?it/s]"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Downloading http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/train-labels-idx1-ubyte.gz to /root/pytorch/FMINST/FashionMNIST/raw/train-labels-idx1-ubyte.gz\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "32768it [00:00, 99395.01it/s]                            \n",
            "0it [00:00, ?it/s]"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Extracting /root/pytorch/FMINST/FashionMNIST/raw/train-labels-idx1-ubyte.gz\n",
            "Downloading http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/t10k-images-idx3-ubyte.gz to /root/pytorch/FMINST/FashionMNIST/raw/t10k-images-idx3-ubyte.gz\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "4423680it [00:01, 4065918.66it/s]                             \n",
            "0it [00:00, ?it/s]"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Extracting /root/pytorch/FMINST/FashionMNIST/raw/t10k-images-idx3-ubyte.gz\n",
            "Downloading http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/t10k-labels-idx1-ubyte.gz to /root/pytorch/FMINST/FashionMNIST/raw/t10k-labels-idx1-ubyte.gz\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "8192it [00:00, 33841.82it/s]            "
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Extracting /root/pytorch/FMINST/FashionMNIST/raw/t10k-labels-idx1-ubyte.gz\n",
            "Processing...\n",
            "Done!\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "metadata": {
        "id": "p2sFfOk3xh_3",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 250
        },
        "outputId": "d32a5098-a356-49c4-976f-2ea2df6f2815"
      },
      "cell_type": "code",
      "source": [
        "image, label = next(iter(trainloader))\n",
        "helper.imshow(image[0,:]);"
      ],
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAdMAAAHTCAYAAAB8/vKtAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "image/png": {
              "width": 233,
              "height": 233
            }
          }
        }
      ]
    },
    {
      "metadata": {
        "id": "_gITnRKFM94x",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "# Criando nossa rede neural\n",
        "\n",
        "Agora vamos definir nossa rede.\n",
        "\n",
        "Cada imagem é 28x28, resultando 784 pixels, e há 10 classes. \n",
        "\n",
        "Brinque um pouco e inclua uma camada oculta, pelo menos. Para função de ativação, use as ativações ReLU para as camadas.\n",
        "\n",
        "Cabe a você quantas camadas você adiciona e o tamanho dessas camadas."
      ]
    },
    {
      "metadata": {
        "id": "yKx1kjk-NpsW",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "from torch import nn, optim\n",
        "import torch.nn.functional as F"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "5RkLY31vN3jz",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "class Rede(nn.Module):\n",
        "  # usado para inicializar o objeto quando vai criar uma instância daquela classe (semelhante ao construtor)\n",
        "  def __init__(self):\n",
        "    super().__init__()\n",
        "    \n",
        "    # Adicione mais cadamas ocultas, se quiser tentar melhorar a performance\n",
        "    self.input = nn.Linear(784, 128)\n",
        "    self.hl1 = nn.Linear(128, 64)\n",
        "    self.output = nn.Linear(64, 10)\n",
        "    \n",
        "  def forward(self, x):\n",
        "    x = x.view(x.shape[0], -1)\n",
        "    \n",
        "    x = self.input(x)\n",
        "    x = F.relu(x)\n",
        "    x = self.hl1(x)\n",
        "    x = F.relu(x)\n",
        "    x = F.log_softmax(self.output(x), dim=1)\n",
        "\n",
        "    return x"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "3zxwsMOiPCX6",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "# Hora de treinar\n",
        "\n",
        "Iremos utilizar algo chamado criterion, uma métrica para calcular o quanto nosso algoritmo está errando (Função de perca).\n",
        "\n",
        "Também utilizaremos um otimizador.\n",
        "\n",
        "## Os quatro passos de uma rede neural\n",
        "\n",
        "\n",
        "1. Realizar um passo adiante para obter os logits.\n",
        "2. Utilizar estes logits para calcular a perda.\n",
        "3. Realizar uma passagem para trás (backward pass) através da rede com loss.backward() para calcular os gradientes.\n",
        "4. Atualizar os pesos com o otimizador.\n",
        "\n"
      ]
    },
    {
      "metadata": {
        "id": "scNOdHUMFvl5",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "model = Rede()\n",
        "\n",
        "#Troque a loss function e o otimizador e veja o que acontece com a performance\n",
        "criterion = nn.CrossEntropyLoss()\n",
        "optimizer = optim.SGD(model.parameters(), lr=0.03)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "4UtLgb9IGI6y",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 109
        },
        "outputId": "90fc3885-ae95-4a7c-a9e9-edb81487ab20"
      },
      "cell_type": "code",
      "source": [
        "model"
      ],
      "execution_count": 24,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "Rede(\n",
              "  (input): Linear(in_features=784, out_features=128, bias=True)\n",
              "  (hl1): Linear(in_features=128, out_features=64, bias=True)\n",
              "  (output): Linear(in_features=64, out_features=10, bias=True)\n",
              ")"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 24
        }
      ]
    },
    {
      "metadata": {
        "id": "ViPDTKrkPCHs",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 109
        },
        "outputId": "6287a347-469e-441d-8459-245c3c564562"
      },
      "cell_type": "code",
      "source": [
        "# Baixe ou suba o número de épocas\n",
        "epochs = 5\n",
        "\n",
        "for e in range(epochs):\n",
        "  running_loss = 0\n",
        "  \n",
        "  for images, labels in trainloader:\n",
        "    # Limpar os gradientes\n",
        "    optimizer.zero_grad()\n",
        "\n",
        "    # Forward\n",
        "    saida = model(images)\n",
        "    # Perda\n",
        "    perca = criterion(saida, labels)\n",
        "    # passo pra tras\n",
        "    perca.backward()\n",
        "    # otimizar os pesos\n",
        "    optimizer.step()\n",
        "    running_loss += perca.item()\n",
        "  else:\n",
        "    print(f\"Training loss: {running_loss/len(trainloader)}\")"
      ],
      "execution_count": 25,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Training loss: 0.7183913413141327\n",
            "Training loss: 0.4596957409623335\n",
            "Training loss: 0.4099561448639898\n",
            "Training loss: 0.38224119743876367\n",
            "Training loss: 0.36055089609582286\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "DuhCSLteHDA2",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "Sua vez! Agora iremos, desenhar as predições realizadas do algoritmo construído"
      ]
    },
    {
      "metadata": {
        "id": "3WUayHz6HALR",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 223
        },
        "outputId": "a0ff922e-dd48-4a00-ff57-5ae8c6d3c172"
      },
      "cell_type": "code",
      "source": [
        "# Testar\n",
        "\n",
        "dataiter = iter(testloader)\n",
        "images, labels = dataiter.next()\n",
        "img = images[0]\n",
        "\n",
        "# Converter 2D para um vetor 1D\n",
        "img = img.resize_(1, 784)\n",
        "\n",
        "# Passa a imagem através da rede\n",
        "ps = torch.exp(model(img))\n",
        "\n",
        "# Plot the image and probabilities\n",
        "helper.view_classify(img.resize_(1, 28, 28), ps, version='Fashion')"
      ],
      "execution_count": 27,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAA04AAAGdCAYAAAAllSMpAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAAWJQAAFiUBSVIk8AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4zLCBo\ndHRwOi8vbWF0cGxvdGxpYi5vcmcvnQurowAAIABJREFUeJzs3XecZXV9//HXe3thASkCYnQBCyAq\n4s+GRsTYu6ixoBE1MYoxlqiJJcaWX4wx/mwhdtFgjIoldgIqiUpUwApSRFikL1KXZevM5/fHOeNe\nrjN7z9nd2dldXs/H4z7O3HM+53O+987s7Hzut5xUFZIkSZKkqc2a6QZIkiRJ0rbOwkmSJEmSRrBw\nkiRJkqQRLJwkSZIkaQQLJ0mSJEkawcJJkiRJkkawcJIkSZKkESycJEmSJGkECydJkiRJGsHCSZIk\nSZJGsHCSJEmSpBEsnCRJkiRpBAsnSZIkSRrBwkmSJGkKSap9LJ3pttxazNR7vjnXTXJ8e+6buuZN\ncky7/9RNa7G2NgsnSZK0w0uyKMmLk3wlyW+S3JxkZZKLkpyY5NlJFs50O7eWJMsG/qCfeIwluSbJ\nd5O8IsmimW7nrVVbVL0pyaEz3RZtMGemGyBJkjSdkjwe+BCw98DulcA4sLR9PAX4xyTPqapvb+02\nzqCVwE3t1/OA3YAHtY8/TXJkVS2fqcZtR64AzgN+2+OcG9pzfjPJsWOAI4BlwE83s23aQuxxkiRJ\nO6wkxwBfoimazgOeA+xRVTtV1c7ArsBTgVOB2wEPnpmWzph3VtXe7WM3YA/g74ECDqYpODVCVb22\nqg6sqvf3OOeL7Tl/Mp1t05Zj4SRJknZISe4JfIDm752vA/eqqhOq6pqJmKq6oao+X1VHAs8AVsxM\na7cNVXVNVb0B+Hi764lJbjeTbZK2FRZOkiRpR/U2YD5wGfCsqlq1seCq+gzwri6Jk8xO8ugkH0xy\nZpKrkqxNcnmSLyZ56EbOndXOYflOO6doXZKrk5yd5GNJHjXJOfsl+dck5ydZ1c7RujjJqUlem2SP\nLu3u4dMDXx820I7fLYKQZH6S1yf5eZIV7f5dh9p9ZJIvJLmyfX+uHPX+DJ1/SJL/aM9bneTcJH+b\nZP4U8Uva9/azSc5Kcn37fl2Q5ENJ7jxN151ycYiNXOP3FoeY2EczTA/g40Pz0Ja1cR9rn5844hpv\nbuNO69ouTc05TpIkaYeTZF/gse3T91bVDV3Oq6rqeImDaHqxJtwIrAX2AZ4EPCnJ66rqHyY599+A\nZw08vwHYmWaY3MHt45sTB5McRjOUcEm7ax3N3KQ7tI8jgJ8MnrMFXDbw9c6THF8A/A9w37Y9Nw8H\nJHkb8Pr2adG8ztuy4f15e1W9diNtOJxmqOBimvc3wF2BtwCPSfLwqrpp6JznAu9rvx5rrzkLOKB9\nPCvJk6rqlC183S1lFXAVzVyzue31Bwv+q9vtR4DnAY9PsvtgL+qEJLNo3g+Aj01Te29V7HGSJEk7\noofQ/MEL8OVpyL+W5o/RRwK7VNUuVbUTsBfwtzR/tP99kvsNnpTkwTRF0xjwCmDnqtqVphC5Hc2i\nAN8butY7aYqmHwKHVdW8qroNzR/29wHeTVMgbEl3GPj6+kmOvwS4C83wxp3a17CUpqAjyTPYUDS9\nH7ht2+Y92VDY/E2SZ2+kDccBvwTuUVW70LwHz6MpJO7P5L2Dv6WZo3VfYFFV7U7z3h4EfIrmPfv3\nJIu38HW3iKr6TFXtDUz0EL1sYA7a3lV1nzbutLaN84Cjp0j3UOCONN+Tz0xXm29NLJwkSdKO6KB2\nu4ZmUYgtqqrOr6oXVNV/VdWNA/uXV9XbgDfTFG4vGjr1/u325Kp6d1WtaM+rqrqiqj5RVa+a4pyX\nVdVPBq51c1WdUVWvqKr/3aIvEP6s3Y4Dp09yfCfg6e0f+mvb9lxcVeuSBHhrG/cfVfXSqvptG3NN\nVf0lG4YCvrXtGZnMGuBRVfWL9ty1VXU8cGx7/AVJBgs8quo/quoNVXX6QLuqqs6lWRjkFJri7akb\nee29rztDPtJunzfF8ee32xMnfs60eSycJEnSjmj3dntdj+F3W9JX2u0Dh/ZPFFm33UjBMGzinH02\nu1UbkWRekoOTfIRmeXaAz1TV1ZOE/7yq/muKVIcCd2q/ftsUMW9ut0tpeocm84GqunaS/Z8ELqX5\nO/aoKc79Pe3Pwdfap8Pfl2m77jT6JE3P56FJ7jV4oJ1r9uT2qcP0thALJ0mSpE2QZGF7o9hTkyxv\nF3modnL/RM/Q8Ip036L5Y/cw4NQ0N94dtWrdxFyqTyZ5e5L7J5m7hV7G3w20eQ1wNvCC9tgP2NDL\nMmxjPVwTi0lcXVVnTxZQVeexYR7VYZPF0MzrmuzcceC7U52b5PZJ/rFdtOP6NDf2nXiN/68N29h7\nvknX3draeU1fap8O9zo9k2aI4q+q6n+2asN2YBZOkiRpRzQxWf427dCxLSrJPjQ3Jn0XzeIMe9IU\nHlfTTO6fuBHqLebSVNWvgBfTzJf5Q5qFIi5LclG7at4teg5ar6aZ87IE+GuaouXGJN9O8uIkCzfj\npaxs23sVcDlwDvAFmmFtf1hVk81vgg2LFExmz3Z72UZioOm9GYwftrHzJ47d4twkR9C8htfQFDe7\n0CwxP/EaJ3rvNjbHqfd1Z9DEcL1nJZk3sH9imN7H0RZj4SRJknZE57Tb+TQrom1p76ZZHOFCmmFt\nu7U31b1tO7n//lOdWFUfA/YDXg78J02Rt5RmPtSZSV43FH8N8CDg4cB7aXqz5gFH0ixkcFaS22/i\n6xi8Ae6+VXVwVT2lvd/V+o2cN9Yh94JNbNMmaXvhTqCZf3UKzc2MF1bVrhOvEXjlRPjWbNs0OgW4\niGZo6hOgWUod+D8036NPzFzTdjwWTpIkaUf03zRLYEP7B+WW0n6y/8T26dFV9YWqum4obK+N5aiq\nq6rqPVX1JJrei/sCX6T5g/6tSe4xFF9VdUpVvayqDqNZuvzPgWuB/dkwBG1bMNEb9Qcj4iaKval6\nrzY2nG7i2OC5D2hzXgs8saq+W1Wrh87b6PdlE687Y9p5WxNzmCaG6030Np1UVZdv/VbtuCycJEnS\nDqeqLmXD3KCXJpnsXkS/p+Owvj1oerJgw1ymYQ/rcj34XVF0OvA0Niw+8KAR51xXVR8CJnqnjthY\n/Fb243a7OMmkCz8kuQuw71D8sElfU/s9evAk504UYudX1e/dV6rV5fvS97rTYXzish1iP07Tu/TI\nJHcEJpZ4d1GILczCSZIk7ajeQDPv6PY09+7Z6NCxJH/MhqFcG7OCDb1Zd58kzz7AS6e4xrzJ9gNU\n1RjNzWShLcySzEoyZyNtWTUYv434KXBB+/Xrpoh5U7tdBvxoipgXt6vDDXs2zfd0nGY+1oSJe1nd\nebLvdZJH0AxvHKXvdafDxFysydpxC1V1GfANYDbNvar2pOkRm477l92qWThJkqQdUlX9lOZGrQU8\nFvhJu4rdbhMxSXZJclSS79DcJHRJh7wraFacA/hYkkPbXLOS/BHNMMGpegr+b5ITkzxpqB17JXkv\nzdynAk5uD+0MXJDk9UnunmT20LX+vo07afQ7snW0w8fe0D59YpL3JdkdIMnu7et8Znv8De1qdZNZ\nAHyznbNDkrlJngt8oD3+0ar6zUD894Gbaeb7fLItYCdWP3w+8Hk2LBqyMX2vOx0mViM8KskuHeIn\nFomYWGb9hKpaN1WwNs3GPsGQJEnarlXVR5NcA3wQOJBmFTuS3ERToAwWShcD3+6Y+hXAd2h6nH6S\nZCXNB9ILaebYPJ8NS0UPmkOzmMRT2nbcSFNkDbbjDVV11sDzO9LcD+ltwLokK2hWi5vdHr+Qbj1l\nW01VfSbJ3YHXA38BHJvkBpp2T3xw//aq+tRG0hwLfBj4RXvuQppFMaApXG/xmqvq+iSvBd5DM+zx\nae15i2ne95/SDF9774jm97ruNPk34FU0QzZ/m2Q5TW/kpVU12TDOrwFXsOFeXw7Tmwb2OEmSpB1a\nVX2JZgGFl9DMe7qU5g/pOTRDxU4EngXctes9b6rqhzSLEXwJuA6YCyynKdAOBX42xan/D/hLmtX0\nzqcpmuYDl9D0eD24qv7vQPyNwONoVvH7Ec0QrCU0y4ifTlOYHNrO6dqmVNUbgD+iea2/pVnt7hqa\nIWQPq6rXjkhxGnA/4LM0Qy4LOA94I/CQqrppkmu+l+bmtBO9T3OAc4G/Aw6nGWY5Su/rbmlVdS7N\nKorfpBmCuDdNAT3p6ontCogTN10+fajw1haSmbmZtiRJkqQtJcn5wJ2BF1fVB0bFqz8LJ0mSJGk7\n1s53O4WmJ/J2VXXjiFO0CRyqJ0mSJG2nkuwB/FP79GMWTdPHHidJkiRpO5PkncAf08x/mkszj+xu\nVbV8Rhu2A7PHSZIkSdr+7AH8Ac29vP4LeKhF0/Syx0mSJEmSRrDHSZIkSZJGsHCSJEmSpBEsnCRJ\nkiRphDkz3YDp8vBZT9s+J28l05d7uuezzZrdPXZ8bNqaccnrD+8V/4dP/Enn2LOu3adX7t0X3twr\nfhbdv0cH73xFr9yf/Xb39+WAv/pBr9y99Pk5gWn9Wbm1OHn8c9P4i0WSpFsHe5wkSZIkaYQdtsdJ\nkqRbsyQXATsDy2a4KZI0k5YCN1bVfpubyMJJkqQd084LFy7c7aCDDtptphsiSTPlnHPOYdWqVVsk\nl4WTJEk7pmUHHXTQbmeeeeZMt0OSZsy9731vfvzjHy/bErmc4yRJkiRJI1g4SZIkSdIIFk6SJEmS\nNIKFkyRJkiSNYOEkSZIkSSNYOEmSJEnSCBZOkiRJkjSC93Ha1lRNW+rMn98rvtas6XeB8bF+8T2s\necx9Ose+7/kf7JV791k3d45ducfcXrn3nN3vhmvXji3oHLtk1tpeuZ/xlB91jn3pd/6yV+4FX+2e\nezp/TgAyp/uvtVq/fhpbIkmSdiT2OEmSJEnSCBZOkiRJkjSChZMkSZIkjWDhJEmSJEkjWDhJkiRJ\n0ggWTpIkSZI0goWTJEmSJI1g4SRJkiRJI1g4SZIkSdIIFk6SJEmSNMKcmW6Abilz5/WKr7Gx7rFr\n1vRtTi/1wEM7x+72jt/0yv3K2x3XOfbqsSW9cv9m/W16xffxxguf1Cv+EXud0zn2kIWX9Mo9l+4/\nK29974d65X7Znz2jc+zt/vy6XrnXX3lVr/hav75zbOb0+xXYJ7c2X5LjgadW1U4j4k4FqKqHTH+r\nJEm3VvY4SZK2mCTHJqkkP5zptmyqJMe3r2HisT7JJUn+I8nB03ztRUnelOQh03kdSVJ/9jhJkrak\no4FlwH2T3KmqLpjh9myqNcCftl/PAQ4AXgQ8KsnBVXX5NF13EfB37denTtM1JEmbwMJJkrRFJNkP\nOBw4CvggTRH15hlt1KZbX1UnDO5I8gPgq8BjgQ/PSKskSTPGoXqSpC3laOA64GvAie3zW0iytB3+\n9qokL0zy6yRrkpye5D6jLpDk0CRXJzk1yZRzn5LMT/LmJBe0+S9J8o4k8zfj9V3Zbm8x2S3J/kk+\nl+TaJDcn+UGSx07Sptsm+WiSq5KsTvKzJM8dOL4UuLp9+ncDQwXftBltliRtIfY4SZK2lKOBL1TV\n2iSfBl6c5D5Vdfoksc8CltD0TBXwGuALSfavqnWTJW8Lq5OAM4AnVtWqKeJmAV8GHgR8CDgHuDvw\nCuAuQKdVW5Ls0X45G9gf+EfgGppep4mYvYDTaIbYvbc9/lzgy0meWlVfbOMW0gy9uxPwfuAi4GnA\n8Ul2rar30BRNLwb+Ffgi8IX2Mj8f0c4zpzh0YJfXKUnqxsJJkrTZktyb5g/1l7a7vgdcSlNMTVY4\n3QG4c1Vd155/HvCfwCMZKEwG8j8Q+DrwXeApVbWxZUKfBTwMOKKqvjeQ4yzgA0kOr6rTRrykxWzo\n/ZlwGfCIqhrc/zfAXsAfTlwryYdpip13JfnPqhoHXggcBDy7qj7Vxn0A+G/gbUk+VlUrkpxIUzj9\nfHiooCRpZjlUT5K0JRwNXAV8B6CqCvgM8IwksyeJ/8xE0dT6brvdfzgwyZE0PU3fAo4aUTRB05Nz\nDnBukj0mHsC32+NHdng9q4GHt49HAn8O3AR8PcldBuIeA/xosECrqptoerqWAgcPxF0JfHogbh1N\nL9VOwBEd2jSpqrr3ZA/g3E3NKUn6ffY4SZI2S1sYPYOmaNovycShHwJ/BfwR8F9Dp93iZm5VdV17\n3vCN1RbQzJk6E/jjqupyM6070/TuDPcYTbhthxxjVXXK4I4kXwd+BfwD8JR29x1pXuewcwaOn9Vu\nf9X2Pk0VJ0nahlk4SZI210OBfWiKp8nuhnw0v184TXVH5gw9X0MzRO+JwKOYZBjfJGYBvwBeOcXx\nfnePblXVpe2QwgdvyvmSpO2bhZMkaXMdDSwHXjLJsaOAJyd50VSLOYxQbf7/BD6X5NFVdeqIc34N\n3BP4VjtkcEuaQzO0bsLFwF0niTtw4PjE9h5JZg31Og3Hben2SpK2EOc4SZI2Wbta3FHAV6vqxOEH\nzQpyS4AnbOo1qmpte43Tga8kue+IUz4L7Av82WTtTbJ4U9rRzm26K/Czgd1fp7nZ7wMG4hbTLAax\nDPjlQNzewNMH4ubQLKZxE80iEQA3t9tdN6WNkqTpY4/TNqbGphq9MoXx7vGz77Rfr9Q3H9fvg8/3\n3uUD3dvS80PVq8e6/50zd8oRQJNbMnt159idM2pO+lDuuf3i77VwWefY68cX9co9Lz1+Vn5vGsbG\nfeVeH+kc+8vvDU9h2bhjf/jsXvEHHP2TzrG1vst0mQ0yp/uvzL65t2NPoCmMvjzF8R/QzDU6mmax\niE1SVauSPI5mgYdvJDmiqs6aIvzfgD+mWUHvSOD7NEuKH9jufyTNkuYbMyfJxA/fLJqFHl7Ufj14\nU9+3A89s2/Re4Fqa5cj3o1n9b+If04doFpg4vl2BcBnwVOCBwMurasXA6/wl8PQk57f5ztrIa5Uk\nbSUWTpKkzXE0zQp0J092sKrGk3wNODrJ7ptzoaq6Mckjgf8BTk7yh1V1wRTXfBLNfZv+BHgyTU/O\nhcB7gPM7XG4+TQE24UaaHq/nVNW3Bq51VZLDae7x9FKaxSx+Djy+qr42ELcqyUNoCq3nAjsD5wHP\nq6rjh679p8D7gP8HzKMp1CycJGmGWThJkjZZVY0cgldVzwOe1z69ht9fAGIiLkPPjwGOGdp3DXC3\noX0PmSTXOuAd7aOXya47Iv5CmiXQR8UtB57fIe5/gf/T9fqSpK3DOU6SJEmSNIKFkyRJkiSNYOEk\nSZIkSSNYOEmSJEnSCBZOkiRJkjSChZMkSZIkjWDhJEmSJEkjWDhJkiRJ0gjeAHcrmLVgQefY8dWr\ne+WeveeenWNf9I2TeuVeOvfaXvGXr9+lc+wsxnvlXjBrXffgmt0r93h1//zgmlrUK/fBO1/RK76P\nm8fn94q/7dzfdo69cbz7zyzA91b9QefY2829rlfu0x/8L73iX/L9x3aOveaB/dpS69d3D57V7+eQ\n8bF+8ZIkaauyx0mSJEmSRrBwkiRJkqQRLJwkSZIkaQQLJ0mSJEkawcJJkiRJkkawcJIkSZKkESyc\nJEnqIMkxSWrosTzJd5I8eqbbJ0maXt7HSZKkft4IXAQE2As4Bvh6ksdX1VdnsmGSpOlj4SRJUj/f\nqKozJp4k+ShwFfBMwMJJknZQDtWTJGnzXA+sAtZP7EjyqiSnJbkmyaokZyZ56vCJSRYmeW+S3yZZ\nkeTLSfZthwG+aSu+BknSCPY4SZLUzy5J9qAZqndb4KXATsAJAzEvA74MfAqYBzwD+FySx1XV1wbi\njgf+GPg34AfAEcDgcUnSNsLCaSsYX7tu2nKf967bd449cN7VvXL/eHX33AB7zrmxc+zqmtsr9+wa\n7x6b7rEAs3rE9+2ivc3clb3if7b6Dp1jD15wWa/ca2t259i+72Gf7/2V63fplXvZ2j16xf/97b/S\nOfYZf/LqXrl3/eT/do7N3H6/XmvNWK94zahThp6vAZ5fVScP7LtLVa2aeJLk/cCPgVfSFkZJDqMp\nmt5dVa9oQ49L8nHgnl0bk+TMKQ4d2DWHJGk0CydJkvp5CXB++/VewLOBjyRZUVVfABgqmm4DzAa+\nSzMPasKj2u1xQ/nfR7PghCRpG2LhJElSPz8aWhzi08BPgPcn+WpVrU3yOOANwKHA/IFza+DrOwLj\nNCv0DbqgT2Oq6t6T7W97og7rk0uSNDUXh5AkaTNU1TjwHWAf4M5J/pBmftNq4FjgMcDDgX+nmRcl\nSdoO2eMkSdLmm/j/dCfgKTRF0yOras1EQJLnDZ1zMc0HmPsBvxrYf6dpbKckaRPZ4yRJ0mZIMhd4\nBLAWOAcYoxmSN3sgZinwpKFTT2q3xw7tf+l0tFOStHnscZIkqZ9HJ5lYse62wLOAOwNvr6obk3yN\nZvW8byb59zbmJTRzl+4xkaSqzkzyeeDlSXZnw3Lkd5kI2SqvRpLUiYWTJEn9vGXg69XAucCLgQ8C\nVNW3k7wA+Bvg3TSLP/w1sJSBwqn1J8CVNKvtPZlmqfOnA+e1uSVJ2wgLJ0mSOqiq42luWNsl9mPA\nxyY59KahuJuBv2gfACQ5tP3y0k1opiRpmjjHSZKkGZJk4SS7X06zTPn/bOXmSJI2wh4nSZJmzmuS\n3JtmOfP1wKPbx4eq6pIZbZkk6RYsnLaG8bFpS/3Mu58xOqi1bN2uvXIvmLWuV/zK8fmjg1qLZ60Z\nHTRgxfhkH8pObhbjvXKvq+7/DFaOz+uV+5c33a5X/CNvc1bn2GvW79Qr986zu0+XmJv1vXL/78o7\nd47dY86KXrn3n7e8V/wN43M7x/724f2mkOz6ye6xtXZtr9y61TqN5h5Pf0uzlPlvaIbz/f0MtkmS\nNAkLJ0mSZkhVnQycPNPtkCSN5hwnSZIkSRrBwkmSJEmSRrBwkiRJkqQRLJwkSZIkaQQLJ0mSJEka\nwcJJkiRJkkawcJIkSZKkESycJEmSJGkECydJkiRJGsHCSZIkSZJGmDPTDdAtzdl7r17x+83/WefY\ndczulXv1+Nxe8XvPuaFz7LJ1e/TKPVbda/zFs9b0yn3O6tt1jt133nW9cs+iesV/6/qDO8fee8my\nXrl/tOKA7u1YdpdeuR98xws6x95pl6t65Z6d8V7xl43t0jl26T7X9MrdS/X73mv7kmQpcBHw6qp6\n58y2RpK0NdjjJEnaJiW5e5ITk1ycZHWSy5KcnOSlM902SdKtj4WTJGmbk+Rw4AzgnsCHgb8APgKM\nAy+bwaZJkm6lHKonSdoWvR64AbhPVV0/eCDJbWemSVtXkjnArKpaO9NtkSTZ4yRJ2jYdAJw9XDQB\nVNXyia+TVJL3J3lSkrOSrElydpJHDZ+XZN8kH0ty1UDc84di5iV5S5Izk9yQZGWS7yY5clSD0/hQ\nkrVJjhrYv2uSdye5pL3uBUn+OsmsgZil7Wt5VZKXJ/k1sAboPvFRkjSt7HGSJG2LLgYekOSQqjpr\nROyDgKOA44AVwF8Cn09yh6q6BiDJXsAPgALeD1wNPBr4aJKdq+rdba6dgT8FPk0zRHAJ8ALgpCT3\nraqfTtaAJLOBjwFPB55cVV9r9y8C/hvYF/gg8BvgcOAfgH2Alw+leh6wAPgQTeF07YjXLknaSiyc\nJEnboncC3wB+muRHwHeBbwHfqap1Q7EHAQdX1a8BknwH+BnwTJoiCeDvgdnA3SeKKeADST4NvCnJ\nB6tqFXAdsHRweFySDwPnAi+lKaJuoR1SdwLwBOAJVfVfA4dfSdN7dq+q+lW774NJLgdeneSfq+qS\ngfjbA3eqqqu7vU2Q5MwpDh3YNYckaTSH6kmStjlVdTLwAODLNAtEvAY4CbgsyROGwk+ZKJrac38O\n3AjsD80QOuApwFfap3tMPNqcuwCHteeOTRRNSWYl2Y3mQ8YzJmKGzAM+BzwOeMxQ0QTwNJqi77qh\n655CU8g9eCj+832KJknS1mOPkyRpm1RVpwNHJZlHUzw9GXgFcGKSQ6vql23obyY5/TrgNu3XewK7\nAi9sH5P53YITSZ4L/BVNj83gDe0umuS81wI7AY+uqlMnOX5n4B40QwM3et2NXGOjqurek+1ve6Im\nK/YkSZvAwkmStE1re4BOB05Pcj7wcZqenDe3IWNTnJp2OzG64gTgE1PE/hwgybOB44EvAf8ELG/z\nv5ZmyN2wk4BHAa9JcmpVrR46Pgs4GXjHFNc9f+j5qiniJEkzzMJJkrQ9OaPd7tPjnKtpFo2YXVWn\njIh9KnAhcFRV1cTOJG+eIv4HwAeArwKfS/Lkqlo/cPzXwE4dritJ2sY5x0mStM1JcmQ7N2nYY9rt\neV1zVdUY8HngKUkOmeRaew48nei9ysDx+9HMt5oq/ynAM2h6nv5tcJlx4LM0qwM+cpLr7touLCFJ\n2g74C3sbc/0fLu0VP3fKxZR+3zXrd+qVe/c5N/WK72Nt9fvR23vO793KZUqzM94r969u6n4vzevm\nL+qVe8X6+b3i95i/snPsDWP92vKtZXfpHHuf21/cK/c9drq0c+zSub/tlfua8cW94s9YuX/n2Afs\n0W86yenM7hWvzfI+YFGSL9KsaDePZhnvpwPLaIbr9fE3wJHAD9tV8n4J7EYzB+hh7dfQ9BwdBXwx\nydeA/YAXtfFT/hKtqi8leR7wSZqFKf68PfRPNKvtfTXJ8cCZwGLg7jS9W0uBfv8oJEkzwsJJkrQt\nehXNPKbH0CzoMI9mEYjjgLdNdmPcjamqq5LcF3gjTWF0LHANcDbw1wOhxwN70xQ+j6QpmJ7dtuUh\nI65xQpIlwHFJbqyqV1fVzUmOAF7X5vgTmsLqfODvgBv6vA5J0syxcJIkbXOq6pvANzvETTacj6pa\nOsm+5cBftI+p8hXNzWn/YejQ14biljEwnG9g/78C/zq07yaawul1G7nupPkkSdsO5zhJkiRJ0ggW\nTpIkSZI0goWTJEmSJI1g4SRJkiRJI1g4SZIkSdIIFk6SJEmSNIKFkyRJkiSNYOEkSZIkSSN4A9xt\nzI13nN0r/oB5yzvHnr3m9r1yz2WsV/yCrOsce+iCi3vlvmTd7j2i1/fKvf/i33aOvWL1Lr1y37Ru\nfq/4x+3xs86xv1m7R6/cT76xw7HxAAAgAElEQVRT99z33enCXrmX9WjLlWP93sOx6vf5zm5zVnaO\n3WV291iA0+n3b0iSJO047HGSJEmSpBHscZIkaQd11mU3sPRvvrbRmGVvf+xWao0kbd/scZIkSZKk\nESycJEmSJGkECydJkiRJGsHCSZIkSZJGsHCSJKmnJJXk/R3ijmljl05/qyRJ08nCSZKkAUnunuTE\nJBcnWZ3ksiQnJ3npVrj265I8abqvI0nqz8JJkqRWksOBM4B7Ah8G/gL4CDAOvGwTUv4bsBDoetfv\n1wEWTpK0DfI+TpIkbfB64AbgPlV1/eCBJLftm6yqxoCxjcUkCbCgqlb1zS9J2nrscZIkaYMDgLOH\niyaAqlo+vC/Jk5KclWRNkrOTPGro+O/NcUqyLMlXkzwyyRnAKuDPkxSwGHhue04lOX7LvjxJ0qay\nx2kbc9NBa3vFL5nVPf4Oc6/plXvxrDW94lfWvM6x31tx116577f4193bMT6/V+5DFl7aOfYJu/y4\nV+53XProXvHjPT7LuGH9wl65L1l1m86x+86/rlfuC1ft2Tn2Zyv+oFfu5+75/V7x14zt1Dl27zk3\n9MrN/R/TPfYHP++XW9uKi4EHJDmkqs4aEfsg4CjgOGAF8JfA55PcoapG/cK9K/Bp4IM0QwLPA55D\nMyzwR8CH2rjuv/wkSdPKwkmSpA3eCXwD+GmSHwHfBb4FfKeq1g3FHgQcXFW/BkjyHeBnwDOBUSvu\n3Ql4VFWdNLgzyQeAC6vqhK4NTnLmFIcO7JpDkjSaQ/UkSWpV1cnAA4Av0ywQ8RrgJOCyJE8YCj9l\nomhqz/05cCOwf4dLXTRcNEmStm32OEmSNKCqTgeOSjKPpnh6MvAK4MQkh1bVL9vQ30xy+nVAl3Gx\nF22RxgJVde/J9rc9UYdtqetI0q2dPU6SJE2iqtZW1elV9TrgxcBc4GkDIVOtlpcO6V1BT5K2MxZO\nkiSNdka73Wear1PTnF+StIksnCRJaiU5sr2v0rCJJRXPm+YmrAR2neZrSJI2gXOcJEna4H3AoiRf\nBM4F5gGHA08HlgEfn+brnwk8LMkrgctpFpH44TRfU5LUgYWTJEkbvIpmHtNjgBfSFE6/oblX09sm\nuzHuFvZKmns4vQ1YCHwCsHCSpG2AhZMkSa2q+ibwzQ5xky4AUVVLh54fDxy/sZihY+cBR4xsqCRp\nq3OOkyRJkiSNYI/TNubed1nWK37F+LzOsYtmremVe0HW94pfXXM7x64Z7/ej980b7t459oAFV/fK\nfcd53eOvXL9Lr9yrx7q/JwC/WHn7zrG7z13ZK/cZN9+hc+xli7rchmaDI3buPl9+Xc3ulXvJrNW9\n4scn7wiY1Irxhb1yX3O3xZ1jd/9Br9SSJGkbZ4+TJEmSJI1gj5MkSTuoQ/bdhTPf/tiZboYk7RDs\ncZIkSZKkESycJEmSJGkECydJkiRJGsHCSZIkSZJGsHCSJEmSpBEsnCRJkiRpBAsnSZIkSRrBwkmS\nJEmSRrBwkiRJkqQR5sx0A3RLh+1ySa/4c9fcrnPszePze+U+fNGvesWvGF/YOfYRO5/VK/e5a/bp\nHLv/vOW9cl89tnP32PVLeuXedd6qXvE7zVnTOfbnK/btlXvPhTd1jr3vThf2yr10zjWdY8dIr9wr\nxhf0ir92/U6dY5fO+22v3NcdUp1jd++VWZIkbevscZIkaQtLckySSrJ0E849PsmyLd4oSdJmsXCS\nJO0Qktw9yYlJLk6yOsllSU5O8tKZbpskaftn4SRJ2u4lORw4A7gn8GHgL4CPAOPAy2awaZKkHYRz\nnCRJO4LXAzcA96mq6wcPJLntzDRJkrQjscdJkrQjOAA4e7hoAqiq360Yk+R5Sb6dZHmSNUl+meTF\nw+ckWZbkq0kelORH7dC/C5P8ySSxd2tzrkpyaZI3MMn/r0memORrSS5vr/3rJH+bZPbmv3xJ0nSz\nx0mStCO4GHhAkkOqamPLdr4YOBv4MrAeeDxwXJJZVfUvQ7F3Ak4EPgp8Ang+cHySM6vqbIAkewPf\nofn/9O3ASuCFwGRLah4D3AS8q90+FHgLsDPw6r4vWJK0dVk4SZJ2BO8EvgH8NMmPgO8C3wK+U1Xr\nBuKOqKrBoub9Sb4JvBIYLpzuCjy4qr4LkOSzwCXA84BXtTF/DewJ3K+qftTGfQKY7H4Ozxq69geS\nfAA4Nskbqqr7/QgGJDlzikMHbko+SdLkHKonSdruVdXJwANoepLuCbwGOAm4LMkTBuJ+V7gk2SXJ\nHsB/A/sn2WUo7S8niqb23KuB84D9B2IeA/xgomgaiPvUJG0cvPaS9trfBRZhkSNJ2zx7nCRJO4Sq\nOh04Ksk8muLpycArgBOTHFpVv0zyQODNNEXWoqEUu9AsMDHhN5Nc5jrgNgPP7wj8cJK484Z3JLkb\n8DaaIXrDd94eLto6q6p7T7a/7Yk6bFPzSpJuycJJkrRDqaq1wOnA6UnOBz4OPC3JCTTD986lGZp3\nCbCWptfoFfz+KIyxKS6Rvm1KsitNz9aNwBuBXwOraQqbf5zk2pKkbYyF0zbmrguu6BU/N+s7x16y\nbvdeuZfMWjc6aMD1Y8Mf3k5tQfrlPnB+9/dldc3tlfuAuctHB7V2n31Tr9ynjR/QK/42c1Z2jn3I\nbX7vA+2NOmzhss6xs6leuZeP7dQ5dvfZ3V8jwLJ1e/SKv3j1bp1j77f4gl65D7rXxZ1j+/2Eaxqd\n0W73oVkIYj7whKr6XW9SkiM3I//FwJ0n2X/XoecPAXYHjqqq/xm49n6bcW1J0lbkJ1ySpO1ekiOT\nTNYT9Jh2ex4bepB+F9fOa3reZlz668D9k9x3IOeewNFDcZNdex5w7GZcW5K0FdnjJEnaEbwPWJTk\nizRD8eYBhwNPB5bRDNfbi2Zo3leSfBDYCfgzYDlNj9SmeAfwHOCbSd7DhuXILwbuMRB3Gs38qE8k\neS9Q7Xm9h/1JkmaGPU6SpB3Bq2jup/QYmvskvQu4L3AczVLh11fVecBTaYqWdwIvAj4EvGdTL1pV\nVwBHAj8H/gZ4OfDJ4ZxVdQ3wOOAKmgUiXgWcTLP6nyRpO2CPkyRpu1dV3wS+2SHuK8BXJjn08aG4\npVOc/5BJ9v2CZg7TsI8NxZ1Gs5rfsAzFHTPZtSVJM8seJ0mSJEkawcJJkiRJkkawcJIkSZKkESyc\nJEmSJGkECydJkiRJGsHCSZIkSZJGsHCSJEmSpBG8j9M25syVS3vFP2LnszrH3m3+Zb1y79azrN51\n9s2dY2cx3iv3yprXOXbJrNW9ci8fW9I5dl7GeuVePGdtr/ibxhZ0jj3l2oN65T53yT6dY5cu+G2v\n3Af3+Nn69bo9e+W+34JlveLPvnnfzrFn3Lx/r9wP2/OczrHfYNdeuSVJ0rbNHidJkiRJGsHCSZIk\nSZJGcKieJEk7qLMuu4Glf/O1mW6GpB3Ysrc/dqabsNXY4yRJkiRJI1g4SZIkSdIIFk6SJEmSNIKF\nkyRJkiSNYOEkSZIkSSNYOEmSNIkkByT5YJILk6xOcmOS7yd5WZKF03TNZyV5+XTkliRtHpcjlyRp\nSJLHAp8D1gCfBM4C5gEPAv4JuBvwwmm49LOAQ4B3T0NuSdJmsHCSJGlAkv2A/wAuBh5aVVcMHP6X\nJHcCbj03LpEkARZO25wX7HZar/j/XHGPzrF3mn9lr9yr68Ze8WPVfeTnaub2yr377Jt6xfdx/dii\nzrEHzFs+be0AOGThpZ1jL1+8a6/csxnvHLuuZvfKvY7u8T9decdeue+/4LJe8ffZ6cLOsees2rdX\n7ics+Vnn2G9weK/c2qa8BtgJeMFQ0QRAVV0AvAcgyRzgtcAxwO2BK4B/B95cVWsmzknyRJoeqnsB\nuwOXAscD/7eqxtqYU4Ej2q+rPfXiqlq6hV+fJGkTWDhJknRLjwcurKoun2R9BHgucCLwz8D9aAqp\ng4AnD8QdA9wEvKvdPhR4C7Az8Oo25u+BXWgKsFe0+0Z+apTkzCkOHdih/ZKkjiycJElqJdkZ2Bf4\nzw6x96Qpmj5SVX/W7j4uyXLgVUmOrKrvtPufVVWrBk7/QJIPAMcmeUNVramqk5NcBtymqk7Ycq9K\nkrQluKqeJEkb7NxuV3SIfUy7fdfQ/n9ut7+bBzVYNCVZkmQP4LvAIjazZ6iq7j3ZAzh3c/JKkm7J\nHidJkjaYmNy5pEPsHYFx4ILBnVV1ZZLr2+MAJLkb8DaaIXo7c0u7bHJrJUlbjYWTJEmtqroxyeU0\nS4J3Pm1jB5PsCvw3TVH2RuDXwGrgMOAfcfSHJG0XLJwkSbqlrwIvTPKAqvrfjcRdTFP03Bk4Z2Jn\nkr2AXdvjAA+hWUnvqKr6n4G4/SbJudEiTJI0c/yUS5KkW3oHsBL4SFsE3UKSA5K8DPh6u+vlQyGv\nbLdfa7djE6cO5JgHHDvJtVfi0D1J2ibZ4yRJ0oCq+nWSZwGfAc5J8kngLGAecDjwNOD4qnpPkk/Q\n9E5NDMe7L81Ke18aWFHvNOA64BNJ3kvTq/QcBgqpAWcCT0/yLuB04Kaq+sp0vVZJUncWTpIkDamq\nLye5B809lp4IvBhYA/wc+Cvgw23onwIX0tyn6cnAlcA/AG8eyHVNksfRrLb3Npoi6gTgW8BJQ5c+\nDjgUeB7NvZwuBiycJGkbYOEkSdIkqupXwAtHxKynuZHtW0bEnQY8YJJDGYpbCRzdr6WSpK3Bwmkb\ns8/seb3ir12/uHPsggXreuW+uSYbRTK1sUlHnUxuSfq1ZazHdLzV43N75V4677edY68fX9Qr94p1\n83vFX7mu+9SGgxZd0Sv3LrNXdo69w9xre+Xu875cv77fe7hyvN9UzH3nXNc5dvbCfvPwv3jjvXrF\nS5KkHYeLQ0iSJEnSCBZOkiRJkjSCQ/UkSdpBHbLvLpz59sfOdDMkaYdgj5MkSZIkjWDhJEmSJEkj\nWDhJkiRJ0ggWTpIkSZI0goWTJEmSJI1g4SRJkiRJI1g4SZIkSdIIFk6SJEmSNII3wN0KZu++W+fY\nb9y8R7/cGe8ce8m63XvlPnjedf3aQnWP7dFugHU1u3PsvIz1yr04azvHXj52m16558zq9zqvWrdL\n59j95i/vlXt2un9/dp21qlfun66+Q+fY282/vlfucdIrfqzH50FjPXPvMXdF9+BZO/XKzXi/n1tJ\nkrR12eMkSZIkSSNYOEmSJEnSCBZOkqRbjSTHJKmBx+oklyc5KclfJlky022UJG2bnOMkSbo1eiNw\nETAX2Bt4CPBu4JVJnlBVP5/BtkmStkEWTpKkW6NvVNUZA8//IclDga8CX05yUFVNukpKksVVtXKr\ntFKStM1wqJ4kSUBVfRt4K3BH4NkASY5PclOSA5J8PckK4FMT5yS5X5JvJrkhyc1J/jvJAwfzJlmS\n5N1JliVZk2R5kpOTHDYQc+ckn09yZTt88NIk/5Gk+1KbkqRpZeEkSdIG/9ZuHzGwbw5wErAceBXw\neYC2h+p/gJ2BNwOvA3YFvp3kvgPnfwB4cXvescA7gVXAQW2eeW3++wPvA14CfAjYv80nSdoGOFRP\nkqRWVV2a5AbggIHd84HPVdVrJ3YkCU1B9B3g0VVV7f4PAmcDb2ND8fVY4MNV9VcDOd8x8PXBwH7A\n06rqxIH9b+nS5iRnTnHowC7nS5K6scdJkqRbugkYXl3vX4eeHwrcGfh3YPckeyTZA1gMfAt4cJKJ\n/2OvB+6X5HZTXO+GdvvIJIs2u/WSpGlhj5MkSbe0E82wvAnrgUuHYu7cbj+xkTy7ANcBr2njLml7\nh74OfLKqLgSoqouSvAt4JXB0ku8CXwZOqKobJk+9QVXde7L97bUOm+yYJKk/C6etYHz/fTvH7j77\n1F65F81a2zn2inX9hsrPS3rFz85491iqV+4+FmRdr/iVNa977ln9cq8fn75O3S9ffWiv+Mft2X11\n5evHF/bKvdvsmzrH7j1n5N+BtzC3x88V9Pv+X71+5165D5i3fHRQa/b+9+uVe+yCi3rFa3okuT1N\nwXPBwO41VTX8gzjxj/vVwE+nSHcTQFV9ti2GnkwzfO/VwF8nOaqqvtHG/FWS44EntjHvBV6b5P5V\nNVy0SZJmgIWTJEkbPKfdnjQi7tft9saqOmVU0qq6AjgOOC7JbYEfA68HvjEQ8wvgF8DbkhwOfB94\nEfCGXq9AkjQtnOMkSRK/WyXvb2lujPupEeFn0hRPr0qy0yS59my3s4eXFK+q5cDlNItOkGTnJMMf\nZP4CGJ+IkSTNPHucJEm3Ro9OciDN/4N7AQ8FHg5cDDyhqlZv7OSqGk/ypzQ9Rmcn+ThwGbAvcCRw\nI/B4mkUmLk1yIvAzmuF7DwPuA0yssvdQ4P1JPgec37bpOcAY7dLnkqSZZ+EkSbo1mljqey1wLU0P\nz8uBj1fVii4JqurUJA+g6aX6C5pFJa4Efgh8sA27mWaI3iOAo2hGelwAHFtVEyv1/YxmaODjaQqv\nm9t9j66qH2zGa5QkbUEWTpKkW42qOh44vkf8McAxGzn+U+ApGzm+lmZVvddsJOYi4AVd2yRJmhnO\ncZIkSZKkESycJEmSJGkECydJkiRJGsHCSZIkSZJGsHCSJEmSpBEsnCRJkiRpBJcj3wrW7zS3c+yi\nWWumrR3n3rR3vxNu0y98rLrX4Quyrl9u0jl2LbN75b56/c6dYxf3/P7MmTXeK/6ei37TOfbS1f2+\nQT9deYfOsUuWbPTen7/nwPlXdI49d80+vXL3Nd7j5/C69Yt75X7Urt2///+87669cs+6oFe4JEna\nyuxxkiRJkqQRLJwkSZIkaQQLJ0mSJEkawcJJkqQd1FmX3TDTTZCkHYaFkyRJkiSNYOEkSZIkSSNY\nOEmSJEnSCBZOkiRJkjSChZMk6VYjyalJTh14vjRJJTlm5lolSdoeWDhJkrZZSY5pC5uJx+ok5yd5\nf5K9Zrp9kqRbjzkz3YBbg1mrx6Yt9x5zVnSOPXCnK3vl3ilze8UvmLW2c+z144t65V40a03n2HU1\nu1fuPtb2zL33ght7xS9fv3Pn2Jfu9a1euU+7+YDOsZevu02v3Etmreocu/Ps1b1y7zkrveKvHuse\nf8P6hb1y9zE2v9/Pip9ijfRG4CJgAfAg4MXAY5IcUlU3z2jLJEm3ChZOkqTtwTeq6oz2648kuQZ4\nJfBE4NMz16zplWRxVa2c6XZIkvyQU5K0ffp2u90vyZuS1HDAwDC/pX2TJ3loku8mWZnk+iT/meSg\ngeNPbXMfMcm5f94eO2Rg34FJTkxybTvc8IwkT5iivUckOS7JcuDSvm2XJE0Pe5wkSdujibGn1wD7\nbMnESR4GfAO4EHgTsBB4KfD9JIdV1TLga8BNwB8D/z2U4unA2VV1VpvvbsD3gcuAtwMr2/O+lOQp\nVfXFofOPA64G3gIs7tDeM6c4dOCocyVJ3Vk4SZK2B7sk2YNmjtMDaeY8rQK+CvzZFr7WPwHXAg+o\nqmsBknwJ+AnwZuC5VbUqyVeApyb5y6oaa+P2Bo6gKbgmvAf4DXCfqlrTxh0HfA/4R2C4cLoW+KOJ\nnJKkbYOFkyRpe3DK0POLgaOr6rKk3wIiG5NkH+BQ4B0TRRNAVf08ycnAYwbCPwM8E3gIMLFay1Np\nhsF/ps23G/BQmkJvSZIlA+efBLw5yb5VddnA/g/3KZqq6t5TvJYzgcO65pEkbZyFkyRpe/AS4Hxg\nPXAVcF5VjU/Dde7Ybs+b5Ng5wCMHFmz4JnADzdC8icLp6cBPq+r89vmdgABvbR+TuS3NML4JF216\n8yVJ08XCSZK0PfjRwKp6w35vYYjW9N2bAKiqNe0QvicnORbYi2YY4esGwiYWYXonTQ/TZC4Yet59\nfX9J0lZj4SRJ2t5dB5Bk16q6fmD/HaeI35iL2+1dJzl2IPDboeXBPwM8F/gj4CCa3qXPDBy/sN2u\nq6rh4YaSpO2Iy5FLkrZ3v263D57YkWQxTUHTS1VdAfwUeG6SXQfyHQI8Avj60Cmn0Czm8PT28aOq\n+t1Qu6paDpwK/Hk7f+oWkuzZt42SpJlhj5MkaXv3XzSr1n00yT8BY8DzaZb0vsMm5Hs1zXLk/5vk\no2xYjvwGbrlaHlW1LskXgGfQLB3+qknyvYRmBb1fJPkwTS/UXsADgNsD99yENkqStjJ7nCRJ27Wq\nWgc8mabn6a3AXwIfAd6/iflOAR5Fc4+ot9AUQz8AHjjYmzTgM8BO7defnSTfL4H/Q3Pvp2OAfwFe\nBIy3+SVJ2wF7nLaCmte9Pr3v/Lm9cl++/qbOsb9efdteua8aW9srvo8x+i0fPI8etzPpuTLx7HRf\nmGvxrDW9cu8yp98c7+vWj7zX5e+88eIn9sp97L7f7hy7Ynxhr9yrq/vP7aL0ew/XTTnvf4r46v5r\nbf6s9b1y97F6t36/Xvv9y7/1qKrjgeM7xP0YuP8kh44finvI0PNlTPJbo6q+xYaV8kZd+5TJcgzF\nXMiIoYNdX6skaWbY4yRJkiRJI1g4SZIkSdIIFk6SJEmSNIKFkyRJO6hD9t1lppsgSTsMCydJkiRJ\nGsHCSZIkSZJGsHCSJEmSpBEsnCRJkiRpBAsnSZIkSRrBwkmSJEmSRrBwkiRJkqQR5sx0A24NVu4z\nf9pyL1u7R+fYuyy8slfuBenbmu5mU73iV9fczrFra3av3GPV/fODFeMLe+Wem7Fe8fdc+JvOsfP3\nWNcr9wlXP6Bz7CN3O7tX7gWz1naOHe/5ec2i9Pt+9rFk9uppy71u0TT+A5IkSVudPU6SJEmSNIKF\nkyRJkiSNYOEkSZIkSSNYOEmStnlJquPjITPdVknSjsnFISRJ24PnDD3/E+Dhk+w/Z+s0R5J0a2Ph\nJEna5lXVCYPPk9wfePjw/qkkWQisrqp+S3puA5IsqqqbZ7odknRr51A9SdIOJcmj2mF7RyX5xySX\nAyuB+e3xOyf5QpLrk9yc5LQkjxjK8aI2x95T5L7/wL6DknwpyVVJVie5JMmnkiweOvf5SX6SZFWS\na5KckGSfoZgfJDkjyf2TfC/JKuCNW/gtkiRtAnucJEk7qrcCNwPvABYDY0luD5xG8//fe/9/e3ce\nLUdVJ3D8+3shbGEJi8gSIIJAcA6iRBYFWWUXzLCMc0AF95kBOSrqeBhnWMRR5wjiuAEiIIvKoo7I\nICAiiiIzkhAUDYQdIjAYIJBANvJ+80fV8zTNe6mqpPu9fsn3c06doqt+de/t+ypN//pW3QLmAO8D\nrouIwzPzuiYVlCNZN5YvzwGeAjYHDgfWokjYiIjPAqcA3wPOAzYGTgJ2jYg3Zua8lmJfDVwLXApc\nAvy52duWJHWDiZMkaUXVB+yRmQsHNkTEvwAbALtm5u/Kbd8G/gicDTRKnIAdgQnAYZl5bcv201rq\n3JYiafpkZp7dsv0a4HfAh8q6B0wAjs/M79RpQERMHWLXpDrHS5Lq8VI9SdKK6qLWpKl0CHDrQNIE\nkJnPARcA20XEaxvWMadcHxQRqw8RcySQwA8iYsOBBXgUeBjYpy1+LlDr3i1J0vBxxGkYzN+we/np\ntOe3qB2767oPNSt70YZNm1Pbmn3t32WWblWWdKklsDjH1I5dQjQqe60xCxrFP/XS2rVjn1g0vlHZ\nW685u3bsBmPmVQe1GBsv1Y6dNn9io7Kb9vn4vvr30C/oH9uo7CbmTWjW7vW71I6V3Ms+9CKij+Iy\nuhsGiR2YjW9L4P66FWTmPRHxDeAE4L0R8SvgGuCyzJxbhm0DjKFIkgbT/o/zscys/aGXmZMH216O\nRO1UtxxJ0tKZOEmSVlTzl+PYoWbfe8UvLZl5QkR8i+K+pgOArwP/HBG7ZeaTFFd3LKYY7RrM822v\nl6fdkqQuMXGSJK0UMrM/Ih4Dthtk98D9QI+U62fL9XjgyZa4LYcoezowHTgjIvYFfg58ADgTeAAY\nC8zMzEeX601IkkaM9zhJklYm1wFvjYi/XsIWEetQJDn3ZubAZXoPlOs9W+LGAh9sLSwi1o2I9lGo\nu8r1auX6aooRrFPbGxMRfRHhlZqSNAo44iRJWpl8DjgKuCki/pPiMrn3ApsChw0EZebUiLgT+FJE\nvLqMOxZeccPlwcB/RMRVwH0UydJ7gIXAD8uyZkTEGcCp5eQTP6GYpnwr4Ajgy8DXuvN2JUmdYuIk\nSVppZOasiNgd+CLwMWBVikvsDsnM9kkj/h44F/gX4BngfIrpw1unHZ9KcVne3wKbUCREdwIHZuad\nLfWeFhEzKJ7ddBrFCNRjZVk/7ey7lCR1g4mTJGnUycwTgROH2Hc9DD0dY2bOpEh0quqYCew7yK5o\nibmPYsSqUmZeAVxREbNbnbIkScPPe5wkSZIkqYKJkyRJkiRVMHGSJEmSpAomTpIkSZJUwcRJkiRJ\nkio4q94wWLR2/djfL1rQqOyt15xdO/ZvVvtzo7LPf3KvRvGf2vT62rF9kY3KntO/eqP4JlbvW1w7\nto/+ZmVH/bIBxvUtqh27/RqPNyp741Weqx27/ph5jcoeQ/2/55ar1j9nAf60YLNG8Qeu9cfasYuz\n/bmlSzd7yQu1Yxdu0OxckSRJvc0RJ0mSJEmqYOIkSZIkSRVMnCRJkiSpgomTJEmSJFUwcZIkSZKk\nCiZOkiRJklTBxEmSJEmSKpg4SZIkSVIFEydJ0qgSEZdFxJwacatEREbEZ4ajXZKkFZuJkySpI8ok\npc6y90i3dTAR8faI+LeKmK9ExO/L/94jIk6LiHWGp4WSpJG0ykg3YGUwf5MltWPn9q/aqOz1V3mh\nduyrxtSPBZh+3faN4t+z88a1Yy/Z8eJGZY+LRbVjV++r398A4/vm144dG/2Nyu5v+NvExmOeqx27\nIMc2KruvQdvHkI3KXkLUjv3M1CmNyu4b06zPj9j1rtqxcxav2ajsG17con7w+Prn7Ark3W2v3wPs\nP8j2GcPRmMx8KSLWABbXPOTtwAeAM5YScwhwdfnfewCnAhcAzy9rOyVJo4OJkySpIzLzstbXEbEb\nsH/79uGUmQuqYiJiXGXTuKEAAA5ISURBVGZW/rIUEdsCrwX+uxNtkySNLl6qJ0nqGRExNiJOj4j7\nI2JBRMyOiFsjYr9BYjePiGsiYl5E/CUivhgRfS37X3GPU0ScWW7bLiKuKO+VuiUiLgM+DIxpuaTw\npbYqDwWeBX4bEWcCny+3P9ZyzISW93FqRDwYEQsj4qGI+GxEvOyygoiYFRH/FREHR8Rd5Xv+Y0Q0\nG5qVJHWdI06SpF5yJvBJ4HzgDmBdYGfgjcDPW+LGAjcCvwY+ARwAfAq4H/hWjXp+CNwLfLp8/Xtg\nE2Bv4LhyW/t1oocAN2Tmkoi4imL06Z3ASRQJFcAz5foi4FjgSuAsYDfgM8Ak4Oi2cicB3wW+CVwM\nvB+4OiIOyMyba7wXSdIwMHGSJPWSQ4FrMvMfKuLWBC7JzIFRn3Mj4i6KpKNO4jQtM19271VE3Afs\nNdilhRGxFrBnWT6ZeVdETKdInH6UmbNaYidTJE3nZuY/lpu/HhGzgY9GxFsz89aW4rcD3pGZ15TH\nX0SR1H0B2KXqjUTE1CF2Tao6VpJUn5fqSZJ6yRxgh4h4bY3Y89pe/xrYqmY932zUKngbxY+N19eI\nPaRcn922/axyfWjb9kcHkiaAzJwDXArsHBEbNmynJKlLHHGSJA27iGifhnNOOZHDvwI/Au6LiD8A\nPwUuzcy72+LnZeYzbdueBdar2YSHGjb5UOB/MnN2jdgtgZeAB1o3ZuasiJhb7m91/yBlzCzXE4Gl\n1pmZkwfbXo5E7VSjvZKkGhxxkiQNq4hYBXiibTkKIDN/AWxNcUncn4APAdMj4vi2YoZ67kDdufHr\nP4egcDBwXcNjJEkrEEecJEnDbQnF851a/XVEKTOfBi4ELoyItSkuwTuNYuKEbhr0AWYR8QZgM145\nDflQDzx7hOL/r1sD97WUsxmwdrm/1WCXJW5brh9eaoslScPGESdJ0rDKwk1ty5MAEbFBW+xcikve\nVhuGpr1AMR35Wm3bDwEez8w7B4kHGN+2fWBk6qNt2z9ertsTsC0i4vCBFxExnuKhwXfUvDRQkjQM\nHHGSJPWSmRHxM2AqxT1LuwBTgHOGoe6B2em+GhE3AYsz80qK+5sGu0xvIP7fy+nJFwM/zsypEXE5\n8E8RsT5wK8V05O8Grm6bUQ+KGfS+ExHfoLif6f3AhhQz80mSeoSJkySpl5wDHAYcSDHK9DBwCvCl\nYaj7SmAP4O8onuXUXyZQuw5Wf2b+NiJOo7gP61CKqzg2B2YB76WY9OE44EiK+7g+B5wxSL33AB+j\nmH58O4qJK47OzJs6+N4kScspMoe6RHt027/v6J55Y9tPrZ+fnrrRrxqVfcOLm9WOfduas6qDWuz/\nhU82it/o67c1im+ib9y42rGxVv1YgBi3Zu3Y/nXqxwLEwsWN4nPMmAZlL2pUNs8+Vzu0/7m5jYrO\nxQ3b0sDMb1Q+xuZlHppyfu3Y78+tOwFb4XWrPVE79trnd2xU9i9fv0aj+CZ+1n9V3QkT1CYijqF4\nmO0GmTmvC+XPorgkb0oXyp6600477TR16lCPeZKkFd/kyZOZNm3atKFmIG3Ce5wkSRraM8BJ3Uia\nJEmji5fqSZI0hMys88BbSdJKwBEnSZIkSargiJMkSSMkMyeMdBskSfU44iRJkiRJFUycJEmSJKmC\niZMkSZIkVTBxkiRJkqQKJk6SJEmSVMHESZIkSZIqOB35MJh5zMTasW8+ZpdGZS/Z5sXasVdvPqtR\n2Zte+1ij+JcaRTfT/8IL9YObxKrnbXP5okbxe2x1RO3YJ59Zp1HZS14YWzv21beMaVT2utzeKF6S\nJA0vR5wkSZIkqYKJkyRJkiRVMHGSJEmSpAomTpIkSZJUwcRJkiRJkiqYOEmSJElSBacjlyRpxTRx\nxowZTJ48eaTbIUkjZsaMGQATO1GWiZMkSSumtebPn79k2rRpd410Q3rYpHJ9z4i2ovfZT9Xso3pG\nop8mAs93oiATJ0mSVkx3A2SmQ05DiIipYB9VsZ+q2Uf1jPZ+8h4nSZIkSapg4iRJkiRJFVbYS/V+\n1n9VjHQbRr2HRroB0ij2/obxl3alFZIkqUMccZIkSZKkCiZOkiRJklQhMnOk2yBJkiRJPc0RJ0mS\nJEmqYOIkSZIkSRVMnCRJkiSpgomTJEmSJFUwcZIkSZKkCiZOkiRJklTBxEmSJEmSKpg4SZIkSVIF\nEydJknpIREyIiAsj4vGIWBgRD0fEORGxXsNy1i+Pe7gs5/Gy3AndrrvblredETEuIo6NiO9GxD0R\n8UJEzI2IOyLi5IhYdYjjcinL7Z19l8uvE3/PiLil4n2vPsRxr4uIKyPiqYhYEBH3RsTpEbFG597h\n8uvAubR3Rf8MLJu3HTdqzqWIOCoivhoRt0bE82UbL1vGshr3dy+dS5GZw12nJEkaRERsDdwGbAT8\nGLgH2AXYB7gX2D0zn65RzgZlOdsCNwO/AyYB7wCeAt6cmQ92o+5u60Q7I+Ig4KfAM8AvgPuB9YDD\ngY3L8vfLzAVtxyXwCHDxIMXOyswLlvmNdVgHz6VbgL2A04cIOTMzX2o7ZleK824scDXwGLAv8Cbg\nNxR9u7D5u+qsDp1LE4Hjh9i9A3AEcHdm7tB23Gg6l6YDOwLzgFkUnyWXZ+a7GpbTuL977lzKTBcX\nFxcXF5ceWIAbgAQ+0rb97HL7uTXLOa+MP6tt+0nl9uu7Vfdo6CPgDcCxwKpt29cGppblnDzIcQnc\nMtJ9MMzn0i3F18Xa9Y4B/lTWcXjL9j6KL74JfHqk+6eTfbSU8r9XlnPSKD+X9gG2AQLYu2z7Zd3u\n7148lxxxkiSpB5S/xt4PPAxsnZn9LfvWBp6g+OKyUWa+sJRy1qIYVeoHNsnMuS37+oAHgS3LOh7s\nZN3dNhztjIhjgMuBazPzsLZ9CfwyM/depjcwTDrZTwMjTpkZNeveF/g58KvM3Ktt31bAAxQjLa/J\nEfwS2u1zKSI2pBid6Qc2zcw5bftHxbnULiL2philbTTitCz93Yvnkvc4SZLUG/Yp1ze2fqkAKJOf\n3wBrArtVlLMbsAbwm9akqSynn+JX39b6Oll3tw1HOxeX65eG2D8+It4XEadExAkRMdJ9MpiO91NE\nvDMiPh0RH4+IgyNitSFC9y3X17fvKBP1mRSJ+1Z16+6Sbp9LxwGrAVe1J00tRsO51CnL0t89dy6Z\nOEmS1Bu2K9czh9h/X7netgvldKrubhuOdr6vXL/iy1ppR+DbwOeArwG/jYjpEbHDEPEjoRv99H3g\n88BZwHXAoxFx1DDV3Q3dbucHy/V5S4kZDedSp6wQn0smTpIk9YZ1y/VzQ+wf2D6+C+V0qu5u62o7\nI+JE4CBgOnDhICFnA7sDr6K4H2pninstdgRujojNlqXeLuhkP/0YOAyYQDGSOYkigRoPXFFOtNGt\nurupa+2MiL0ovvTfnZm3DRE2Ws6lTlkhPpdMnCRJ0kovIo4AzgGeBI7MzMXtMZl5cmbelpmzM3Ne\nZt6RmUcDPwA2BD4xvK3uvsz8cmZem5l/zswFmXlvZp4CnEzxPfLzI9zEXvShcn3+UAEr47m0IjBx\nkiSpNwz8erruEPsHtg91v8TylNOpurutK+2MiCkUl6I9BeydbVO113Buud6z4XHdMhx/zwso7gN7\nQ3lz/3DW3QndOpfWB44E5gOXLkO7eu1c6pQV4nPJxEmSpN5wb7ke6nr9bcr1UNf7L085naq72zre\nzog4GrgK+D+K2ePurThkMH8p1+OW4dhu6PrfM4tnXA1MPtL6vlfac6k0MCnElUuZFGJpeu1c6pQV\n4nPJxEmSpN7wi3J9QDlt+F+Vv+jvDrwI3F5Rzu0Uv3bv3jYSMDAd+QFt9XWy7m7raDsj4liKZ+08\nTpE03VdxyFAGZgJrOlLVLV3/e0bEdhQPDZ4LzG7ZdXO5br/3aWAK6W0pppAe6b7qVh8NTAox5GV6\nFXrtXOqUZenvnjuXTJwkSeoBmfkAcCMwETihbffpFL9AX9r6TJmImBQRk9rKmUdxidA44LS2ck4s\ny7+h9XK0Zal7JHSqj8rtxwGXAI8Ce1ZdnhcRr4+IsYNtp5gVDeCy+u+mezrVTxHxmvLSM9q2vwq4\nqHz5/cxsnbr9l8AMYM+IOLzlmD7gi+XLc0fyGU7Q2XOpZf9bge1Z+qQQo+pcaioixpb9tHXr9mX8\njOm5c8kH4EqS1CPKLxu3ARtRzGY2A9iV4hkoM4G3ZObTLfEJ0P5w0ojYoCxnW4pfbf+X4gvdOyju\n43lL+UVmmeseKZ3oo4jYB7iJ4gfkC4HHBqlqTmae03LMxRSzy91axi+kmGHuIGAM8C3gwyOdEAzo\nUD8dT3HPza8pftV/BtgCOITi/pI7gP0HebjrrhTn3ViKmeIeBfYD3kTxvJ79MnNhp99zU53699ay\n/1LgXcBJmfnVpdR7MaPrXJoCTClfbgwcSHE+3Fpum52ZnyhjJwIPAY9k5sS2chp/xvTcuZSZLi4u\nLi4uLj2yAJtT/Jr/BLCI4lKUc4D1BonN4n/lg5azPvCV8vhFZXkXAhM6Ufdo7iPg+IHtS1kebjtm\nCvBD4H7g+ZY+/Qlw+Ej3SZf6aQfgYuAPwNMUDwd+huIL80eAVZdS9+so7h2bTZEYzKQYWVhjpPul\nk33Usm89iktkXwTGV9Q5qs4lipHrWv9WKEaUXvHvZ1n6uxfPJUecJEmSJKmC9zhJkiRJUgUTJ0mS\nJEmqYOIkSZIkSRVMnCRJkiSpgomTJEmSJFUwcZIkSZKkCiZOkiRJklTBxEmSJEmSKpg4SZIkSVIF\nEydJkiRJqmDiJEmSJEkVTJwkSZIkqYKJkyRJkiRVMHGSJEmSpAomTpIkSZJUwcRJkiRJkiqYOEmS\nJElSBRMnSZIkSapg4iRJkiRJFUycJEmSJKnC/wNcG/9To8IOYQAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x648 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "image/png": {
              "width": 423,
              "height": 206
            }
          }
        }
      ]
    },
    {
      "metadata": {
        "id": "jnLG6oNBHrOx",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}