Img_Clsf_FL_tf_keras.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "view-in-github"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/paxF3E/Image-Classification-by-Federated-Learning/blob/main/Img_Clsf_FL_tf_keras.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "JlOZ7EcY2DZ_"
      },
      "source": [
        "### Environment Setup"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 2,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "8Vff13xAMtfS",
        "outputId": "5cf6239b-feec-4efd-f8f9-53ebf2083881"
      },
      "outputs": [],
      "source": [
        "# !pip uninstall --yes tensorboard tb-nightly\n",
        "\n",
        "# !pip install --quiet --upgrade tensorflow-federated-nightly\n",
        "# !pip install --quiet --upgrade nest-asyncio\n",
        "# !pip install --quiet --upgrade tb-nightly  # or tensorboard, but not both\n",
        "\n",
        "import nest_asyncio\n",
        "nest_asyncio.apply()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 3,
      "metadata": {
        "id": "W8CuPvLAa-u2"
      },
      "outputs": [
        {
          "ename": "ModuleNotFoundError",
          "evalue": "No module named 'tensorboard'",
          "output_type": "error",
          "traceback": [
            "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[1;31mModuleNotFoundError\u001b[0m                       Traceback (most recent call last)",
            "\u001b[1;32md:\\Open SOSS\\Image-Classification-by-Federated-Learning\\Img_Clsf_FL_tf_keras.ipynb Cell 4\u001b[0m in \u001b[0;36m<cell line: 1>\u001b[1;34m()\u001b[0m\n\u001b[1;32m----> <a href='vscode-notebook-cell:/d%3A/Open%20SOSS/Image-Classification-by-Federated-Learning/Img_Clsf_FL_tf_keras.ipynb#W3sZmlsZQ%3D%3D?line=0'>1</a>\u001b[0m get_ipython()\u001b[39m.\u001b[39;49mrun_line_magic(\u001b[39m'\u001b[39;49m\u001b[39mload_ext\u001b[39;49m\u001b[39m'\u001b[39;49m, \u001b[39m'\u001b[39;49m\u001b[39mtensorboard\u001b[39;49m\u001b[39m'\u001b[39;49m)\n",
            "File \u001b[1;32m~\\AppData\\Roaming\\Python\\Python310\\site-packages\\IPython\\core\\interactiveshell.py:2305\u001b[0m, in \u001b[0;36mInteractiveShell.run_line_magic\u001b[1;34m(self, magic_name, line, _stack_depth)\u001b[0m\n\u001b[0;32m   2303\u001b[0m     kwargs[\u001b[39m'\u001b[39m\u001b[39mlocal_ns\u001b[39m\u001b[39m'\u001b[39m] \u001b[39m=\u001b[39m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mget_local_scope(stack_depth)\n\u001b[0;32m   2304\u001b[0m \u001b[39mwith\u001b[39;00m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mbuiltin_trap:\n\u001b[1;32m-> 2305\u001b[0m     result \u001b[39m=\u001b[39m fn(\u001b[39m*\u001b[39margs, \u001b[39m*\u001b[39m\u001b[39m*\u001b[39mkwargs)\n\u001b[0;32m   2306\u001b[0m \u001b[39mreturn\u001b[39;00m result\n",
            "File \u001b[1;32m~\\AppData\\Roaming\\Python\\Python310\\site-packages\\IPython\\core\\magics\\extension.py:33\u001b[0m, in \u001b[0;36mExtensionMagics.load_ext\u001b[1;34m(self, module_str)\u001b[0m\n\u001b[0;32m     31\u001b[0m \u001b[39mif\u001b[39;00m \u001b[39mnot\u001b[39;00m module_str:\n\u001b[0;32m     32\u001b[0m     \u001b[39mraise\u001b[39;00m UsageError(\u001b[39m'\u001b[39m\u001b[39mMissing module name.\u001b[39m\u001b[39m'\u001b[39m)\n\u001b[1;32m---> 33\u001b[0m res \u001b[39m=\u001b[39m \u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49mshell\u001b[39m.\u001b[39;49mextension_manager\u001b[39m.\u001b[39;49mload_extension(module_str)\n\u001b[0;32m     35\u001b[0m \u001b[39mif\u001b[39;00m res \u001b[39m==\u001b[39m \u001b[39m'\u001b[39m\u001b[39malready loaded\u001b[39m\u001b[39m'\u001b[39m:\n\u001b[0;32m     36\u001b[0m     \u001b[39mprint\u001b[39m(\u001b[39m\"\u001b[39m\u001b[39mThe \u001b[39m\u001b[39m%s\u001b[39;00m\u001b[39m extension is already loaded. To reload it, use:\u001b[39m\u001b[39m\"\u001b[39m \u001b[39m%\u001b[39m module_str)\n",
            "File \u001b[1;32m~\\AppData\\Roaming\\Python\\Python310\\site-packages\\IPython\\core\\extensions.py:76\u001b[0m, in \u001b[0;36mExtensionManager.load_extension\u001b[1;34m(self, module_str)\u001b[0m\n\u001b[0;32m     69\u001b[0m \u001b[39m\"\"\"Load an IPython extension by its module name.\u001b[39;00m\n\u001b[0;32m     70\u001b[0m \n\u001b[0;32m     71\u001b[0m \u001b[39mReturns the string \"already loaded\" if the extension is already loaded,\u001b[39;00m\n\u001b[0;32m     72\u001b[0m \u001b[39m\"no load function\" if the module doesn't have a load_ipython_extension\u001b[39;00m\n\u001b[0;32m     73\u001b[0m \u001b[39mfunction, or None if it succeeded.\u001b[39;00m\n\u001b[0;32m     74\u001b[0m \u001b[39m\"\"\"\u001b[39;00m\n\u001b[0;32m     75\u001b[0m \u001b[39mtry\u001b[39;00m:\n\u001b[1;32m---> 76\u001b[0m     \u001b[39mreturn\u001b[39;00m \u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49m_load_extension(module_str)\n\u001b[0;32m     77\u001b[0m \u001b[39mexcept\u001b[39;00m \u001b[39mModuleNotFoundError\u001b[39;00m:\n\u001b[0;32m     78\u001b[0m     \u001b[39mif\u001b[39;00m module_str \u001b[39min\u001b[39;00m BUILTINS_EXTS:\n",
            "File \u001b[1;32m~\\AppData\\Roaming\\Python\\Python310\\site-packages\\IPython\\core\\extensions.py:92\u001b[0m, in \u001b[0;36mExtensionManager._load_extension\u001b[1;34m(self, module_str)\u001b[0m\n\u001b[0;32m     90\u001b[0m \u001b[39mif\u001b[39;00m module_str \u001b[39mnot\u001b[39;00m \u001b[39min\u001b[39;00m sys\u001b[39m.\u001b[39mmodules:\n\u001b[0;32m     91\u001b[0m     \u001b[39mwith\u001b[39;00m prepended_to_syspath(\u001b[39mself\u001b[39m\u001b[39m.\u001b[39mipython_extension_dir):\n\u001b[1;32m---> 92\u001b[0m         mod \u001b[39m=\u001b[39m import_module(module_str)\n\u001b[0;32m     93\u001b[0m         \u001b[39mif\u001b[39;00m mod\u001b[39m.\u001b[39m\u001b[39m__file__\u001b[39m\u001b[39m.\u001b[39mstartswith(\u001b[39mself\u001b[39m\u001b[39m.\u001b[39mipython_extension_dir):\n\u001b[0;32m     94\u001b[0m             \u001b[39mprint\u001b[39m((\u001b[39m\"\u001b[39m\u001b[39mLoading extensions from \u001b[39m\u001b[39m{dir}\u001b[39;00m\u001b[39m is deprecated. \u001b[39m\u001b[39m\"\u001b[39m\n\u001b[0;32m     95\u001b[0m                    \u001b[39m\"\u001b[39m\u001b[39mWe recommend managing extensions like any \u001b[39m\u001b[39m\"\u001b[39m\n\u001b[0;32m     96\u001b[0m                    \u001b[39m\"\u001b[39m\u001b[39mother Python packages, in site-packages.\u001b[39m\u001b[39m\"\u001b[39m)\u001b[39m.\u001b[39mformat(\n\u001b[0;32m     97\u001b[0m                   \u001b[39mdir\u001b[39m\u001b[39m=\u001b[39mcompress_user(\u001b[39mself\u001b[39m\u001b[39m.\u001b[39mipython_extension_dir)))\n",
            "File \u001b[1;32mc:\\Program Files\\Python310\\lib\\importlib\\__init__.py:126\u001b[0m, in \u001b[0;36mimport_module\u001b[1;34m(name, package)\u001b[0m\n\u001b[0;32m    124\u001b[0m             \u001b[39mbreak\u001b[39;00m\n\u001b[0;32m    125\u001b[0m         level \u001b[39m+\u001b[39m\u001b[39m=\u001b[39m \u001b[39m1\u001b[39m\n\u001b[1;32m--> 126\u001b[0m \u001b[39mreturn\u001b[39;00m _bootstrap\u001b[39m.\u001b[39;49m_gcd_import(name[level:], package, level)\n",
            "File \u001b[1;32m<frozen importlib._bootstrap>:1050\u001b[0m, in \u001b[0;36m_gcd_import\u001b[1;34m(name, package, level)\u001b[0m\n",
            "File \u001b[1;32m<frozen importlib._bootstrap>:1027\u001b[0m, in \u001b[0;36m_find_and_load\u001b[1;34m(name, import_)\u001b[0m\n",
            "File \u001b[1;32m<frozen importlib._bootstrap>:1004\u001b[0m, in \u001b[0;36m_find_and_load_unlocked\u001b[1;34m(name, import_)\u001b[0m\n",
            "\u001b[1;31mModuleNotFoundError\u001b[0m: No module named 'tensorboard'"
          ]
        }
      ],
      "source": [
        "%load_ext tensorboard\n",
        "# tb load"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "mKToZCaRMSYC"
      },
      "outputs": [],
      "source": [
        "import collections\n",
        "import numpy as np\n",
        "import tensorflow as tf\n",
        "import tensorflow_federated as tff\n",
        "\n",
        "np.random.seed(0)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "_hU4hNzP2QUF"
      },
      "source": [
        "### Data prepping\n",
        "- federated learning requires a federated data, i.e. data from multiple users (indentical and independent distributed rvs; i.i.d)\n",
        "- data used here is a modified form of MNIST dataset, keyed over each writer; since writing style is unique, this replicates a federated data\n",
        "- use of `tff.simulation.ClientData` enumerates the set of users, allowing selection of data subsets for simulation[link text](https://)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "SuOPnPpmeo2S",
        "outputId": "a4aac2d7-8af9-46c9-ea0d-2f5f852f3f18"
      },
      "outputs": [
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "Downloading emnist_all.sqlite.lzma: 100%|██████████| 170507172/170507172 [00:48<00:00, 4605093.06it/s]\n"
          ]
        }
      ],
      "source": [
        "emnist_train, emnist_test = tff.simulation.datasets.emnist.load_data()\n",
        "# retrieving an instance of ClientData simulation"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ePFirrxJzwp8",
        "outputId": "8df34004-3801-4262-cb62-0de4cf79bda5"
      },
      "outputs": [
        {
          "data": {
            "text/plain": [
              "3383"
            ]
          },
          "execution_count": 5,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "len(emnist_train.client_ids)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "9JZmQ6ISzz8k",
        "outputId": "5047effe-8b6f-46a4-fd08-e94d5eb7aee9"
      },
      "outputs": [
        {
          "data": {
            "text/plain": [
              "OrderedDict([('label', TensorSpec(shape=(), dtype=tf.int32, name=None)),\n",
              "             ('pixels',\n",
              "              TensorSpec(shape=(28, 28), dtype=tf.float32, name=None))])"
            ]
          },
          "execution_count": 6,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "emnist_train.element_type_structure"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "JiEdcyw_gLl3",
        "outputId": "27a68c02-2881-490c-d87a-58f698295be6"
      },
      "outputs": [
        {
          "data": {
            "text/plain": [
              "1"
            ]
          },
          "execution_count": 7,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "example_dataset = emnist_train.create_tf_dataset_for_client(emnist_train.client_ids[0])\n",
        "example_element = next(iter(example_dataset))\n",
        "example_element['label'].numpy()\n",
        "\n",
        "## allows to iterate over client ids in this case\n",
        "## FL doesnt deal in client ids, just the data subsets"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 54
        },
        "id": "jFnkhJ6egWiL",
        "outputId": "24511751-4ad7-4041-bdaf-66b231c6f65b"
      },
      "outputs": [
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          },
          "output_type": "display_data"
        }
      ],
      "source": [
        "from matplotlib import pyplot as plt\n",
        "plt.imshow(example_element['pixels'].numpy(), cmap='gray', aspect='equal')\n",
        "plt.grid(False)\n",
        "_ = plt.show()\n",
        "\n",
        "# visualisation"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "IvGOTglrOgU0"
      },
      "source": [
        "#### heterogeneity\n",
        "- federated data is usually non i.i.d due to user specific usage patterns\n",
        "- there might be data paucity for some clients while, enough data for some other clients\n",
        "- in real production federated environment, data inspection is not possible for obv reasons"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 54
        },
        "id": "R5EtMhltgf6D",
        "outputId": "e27820da-1276-4574-b045-cf049914301e"
      },
      "outputs": [
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 1440x288 with 40 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          },
          "output_type": "display_data"
        }
      ],
      "source": [
        "## Example MNIST digits for a client\n",
        "figure = plt.figure(figsize=(20, 4))\n",
        "j = 0\n",
        "\n",
        "for example in example_dataset.take(40):\n",
        "  plt.subplot(4, 10, j+1)\n",
        "  plt.imshow(example['pixels'].numpy(), cmap='gray', aspect='equal')\n",
        "  plt.axis('off')\n",
        "  j += 1\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 54
        },
        "id": "5vKBmYxWghuL",
        "outputId": "0fbd1177-ce93-4981-fecb-a3c895c84ec1"
      },
      "outputs": [
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 864x504 with 6 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          },
          "output_type": "display_data"
        }
      ],
      "source": [
        "# #examples for each Client\n",
        "\n",
        "f = plt.figure(figsize=(12, 7)) \n",
        "f.suptitle('Label Counts for a Sample of Clients')\n",
        "for i in range(6):\n",
        "  client_dataset = emnist_train.create_tf_dataset_for_client(emnist_train.client_ids[i])\n",
        "  plot_data = collections.defaultdict(list)\n",
        "  for example in client_dataset:\n",
        "    # Append counts individually per label\n",
        "    label = example['label'].numpy()\n",
        "    plot_data[label].append(label)\n",
        "  plt.subplot(2, 3, i+1)\n",
        "  plt.title('Client {}'.format(i))\n",
        "  for j in range(10):\n",
        "    plt.hist(\n",
        "        plot_data[j],\n",
        "        density=False,\n",
        "        bins=[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10])\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ZnZCRoXDf7yO"
      },
      "source": [
        "- mean image for each label for a Client\n",
        "- mean would result accordingly accounting the noise pixels too"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 124
        },
        "id": "dBnYudTOgo8P",
        "outputId": "76a330be-3165-4351-ff13-a95780b33632"
      },
      "outputs": [
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 864x360 with 10 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          },
          "output_type": "display_data"
        },
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 864x360 with 10 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          },
          "output_type": "display_data"
        },
        {
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAqwAAAE+CAYAAABIhELlAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nO3deXxdVb338d86ycnUpEmazm0KnQcoUKZCKYOMoqCoF1RQnB/Hq/c68Tzq9eKdnRW5iiJXrNcLCIhMYhGkzDK10AIWOtB5SNOkadI008l6/jjhmu/aaYbSJivJ5/168XrxS/bZe5+z19nnl9PvXtt57w0AAACIVWqgdwAAAADoDg0rAAAAokbDCgAAgKjRsAIAACBqNKwAAACIGg0rAAAAokbDCgxCzrmrnXP/3fH/U5xzDc65nIHeL6C3nHM3Ouf+pb8fC2BwomEFIuWcu9w592xHM7rdOXefc25xuJz3fpP3vth7nzkE21zmnPtoL5ctds5t7vj/Dzvnvtfpd7Occ3c653Y552qcc0udc7M7/f5q59zVB7F/3jlX5ZzL7fSzdMfP+n1SaefcB51zj/X3dvuq47g2dYylaufcb51zEw7huns1ZgDgYNGwAhFyzn3ezH5gZv9mZuPMbIqZ/djM3j6Q+xVYYGYrOv7/BDNb3ul3ZWZ2l5nNtuz+P21mdx6i7daa2YWd6gs7fobufcZ7X2xmsyx7fL7f1xXwLT6AgULDCkTGOVdqZv9kZp/23v/We7/Pe9/qvb/be/+lLpY/suObx9zXH++cu6HjW9mtzrl/eb3ReP0bQefcd5xztc6515xzF3b87l/N7HQzu7bjm7hre9jVE83suU7//78Nq/f+ae/9Dd77Gu99q2Wbo9nOuYou9n+0c+4e59yejm9jH3XOdXdu+pWZXdmpvtLMloSvYTevwXTn3J+cc7s7vm38tXOurNNjNzjnvuicW+mcq3PO3eKcK+jhtej82C91PHZfxz6M6/h2vN4594BzrrzT8rc653Z0bOcR59xRnX5X4Zy72zm31zn3TMdzeKzT7+c45/7Y8Zq94py7rDf76L2vMbPbzezontbT8U/vP3HO/d45t8/M3tSbbfTm+XUY3bHteufcw865I97o8wMwNNGwAvE51cwKzOyOg3z8jWbWZmYzLPst6Plm1vmfbBea2StmNtrMvmVmNzjnnPf+q2b2qHV8E+e9/0xXK+9owvaY2TfN7Esd/3+imT3hnHvpAPt0hpnt8N7vNjPz3l/tvb+643dfMLMtZjbGst/GfsXMuvvn/d+Z2RnOubKO5u90S357291r4Mzs381sopnNNbNKM7s6ePxlZvZmM5tqZseY2Qe72Z/Qu8zsPMt+k3mxmd3X8ZzGWPac+9lOy95nZjPNbKxlG/5fd/rdf5rZPjMbb2Yf6Pgv+wScG2FmfzSz/+l47HvM7MfOuXk97ZxzbnTHPq7o5XouN7N/NbMSM+tr/KG752dmdoWZ/bNlx+Lzr//+jTw/AEMTDSsQnwozq/bet/X1gc65cWb2FjP7u45vZqss++3mezotttF7f31H5vWXZjbBso1ir3jvP2LZRm6DZRuNT5nZdd77Mu99+A2aOecmW7b5+vwBVtnasQ9HdHyT/Kj3vruGtcnM7jazd3f8d1fHz17fXrevgfd+rff+j977Zu/9LjP7npmdGWzjGu/9to5vI+82s+O62Z/Qj7z3O733Wy37B8BT3vsV3vsmy/4RsuD1Bb33/+W9r/feN1u2aT6249vhHMs2lf/ovW/03r9s2WP1uovMbIP3/hfe+zbv/QrLfmt6aTf7dU3HHxcvmNl2yx6P3qznTu/949779o7n0GsHen6dFrnXe/9Ix++/amanOucqD/L5ARjCcnteBEA/223ZfyrNPYim9QgzS5vZdufc6z9LmdnmTsvseP1/vPeNHcsV92blzrm3Wfaf39OWPX/sMLMRZtbknLvCzM713j/bafkxZna/mf3Ye3/TAVb7bcs2M/d37MvPvPf/0cOuLLHst6TOzK4Kftfta9DR0P7Qst/MlnT8LszA7uj0/42W/Ta2t3Z2+v/9XdTFHfuRY9lvLi+17Lev7R3LjDazQsu+vp2PW+f/P8LMFnY0oK/LtWxc4kA+673/eecfdPwTfE/r6bzdXuvh+dWF6/beNzjnaiz7Wh/M8wMwhNGwAvF50syazewSM7utj4/d3PHY0QfzDa11/0/x5r2/y8zKnHPXmdnD3vubnHM7LfvtqHz71vHP9feb2V3e+3/tZp31lo0FfME5d7SZ/ck594z3/sFuduVRy34r6y37z9TTO/2up9fg3zoeN997X+Ocu8TMesrrHg6XW/YiunMt+211qWUbZ2dmuywbaZhsZq92LF/Z6bGbLfv6n/cG96E36znY2Re6e36v+9/n5JwrNrNRZratl/sFYBghEgBExntfZ2ZfN7P/dM5d4pwrctmpmy50zn2rh8dut2yT+F3n3EjnXKrjIqPwn7wPZKeZTevFcieY2XLn3FQz295FszrSzJaa2ePe+//b3Yqccxc552a47NehdWaWsb9+G9eljsjAxWb2tjA+0IvXoMTMGsyszjk3ycwSF7L1kxLLNta7zazIso20mZl1xDV+a2ZXdxz/OaYXmt1jZrOcc+/vGBtp59xJzrm5fdyHQ7WeXOdcQaf/0t09v07e4pxb7JzLs2yW9c/e+82HcL8ADBE0rECEvPfftWzG8GuW/bZts5l9xrIXHPXkSjPLM7OXLfuN1m2W/TayN35oZn/jsjMIXNPVAh3NyJGW/ebvePvrTAGdvcPMTjKzD7nsjAOv/zeli2VnmtkDlm0in7RsfOChnnbUe/+S9/5AF3l19xp8o2O/68zsXss2hgNhiZltNLOtlt3PPwe//4xlv5XcYdl/Cr/Jsg3g699Kn2/ZXO62jmW+aWb5fdmBQ7UeM/uJZeMOr//3i148P7PsRVX/aGY1lv0j6H2HeL8ADBGu+2sbAAAxcM5908zGe+8/0OPCADDE8A0rAESoYx7SY1zWyWb2ETv4qc4AYFDjoisAiFOJZWMAEy2bLf6uHbq7hQHAoEIkAAAAAFEjEgAAAICo0bACAAAgajSsAAAAiBoNKwAAAKJGwwoAAICo0bACAAAgajSsAAAAiBoNKwAAAKJGwwoAAICo0bACAAAgajSsAAAAiBoNKwAAAKJGwwoAAICo0bACAAAgajSsAAAAiBoNKwAAAKJGwwoAAICo0bACAAAgajSsAAAAiBoNKwAAAKJGwwoAAICo0bACAAAgajSsAAAAiBoNKwAAAKJGwwoAAICo0bACAAAgarnd/bJ9x0zfXzuCwy81fo07XOtmrAwth2usME6GFs4p6C3OKeiN7sYJ37ACAAAgajSsAAAAiBoNKwAAAKJGwwoAAICo0bACAAAgajSsAAAAiBoNKwAAAKJGwwoAAICo0bACAAAgajSsAAAAiBoNKwAAAKJGwwoAAICo0bACAAAgajSsAAAAiBoNKwAAAKJGwwoAAICo5Q70DgDDTca3S53j+LsRPWPcAOhPrT4jdW17k9Rjc0b05+7wDSsAAADiRsMKAACAqNGwAgAAIGpkWA9CmCXrDfJmeB1jIS4H837uSV+PcZgVMzNLu5xu13nj3rFSt3g9nX9o5OZu14fBodm3Sp3q4num8NiSd0ZvhOOk3bzU711/gdSLR62V+u/KN3S7PrNDO/YYxQAAAIgaDSsAAACiRsMKAACAqJFh7UJPuY6UuR7XQWZoaGoI5qFb26pj4ag8fUttatufWMeDjbOkJmt4aIXHKMz8FaXypD6Y92pfM4LL9uvvn2qcLvVVFWsSj3mttUHqT6+/TOpZI6uk3tJYJvUJBRu0zmdcxSAcO1WZRqmXt4yW+r7ahVKPyatPrHNB0Qapx+fWSb2jrVjq0wuqpR6ZKpCazy81EJngLW36/q9r1/fvEbn62bO5Tfdxbl5RYp1hVj78rLl07XlSr3hxqtTXXXxnsEadh/Vwvy6MSgAAAESNhhUAAABRo2EFAABA1IZlhjXMozzXormOmozmffZkirqtzcy2tIySen8mLfU3xj0hdXGQGcIb11P2+FBkQ3dm2qRe2rBA6kdTLVLfs2N+Yh1b9pRKfenJ10tNnqxv/tyk799rd7xZ6r8Z86zU43M037enXd/PyxuPTGxjct5uqesymt1KOR17rV7H2u+2Hid1fo6Oo5/+6ZzENkvW63HPr9Px/O2rb5f6ppyTpZ6RTs7tiv4XZqqXt+j7+/69p0r96+c0s1r8imauW09MZlhvrF8k9aRJNd3u0/+U7JH6K5N+L/UxecP786k2yBWvb9NW6YR8PSYHI/y82tSm2/zkuncHy+v54COTH5N6YrpW6lldzIlandFrKn5Wq+eMFx+bIfUnLn5A6tE5et7r72wvn4QAAACIGg0rAAAAokbDCgAAgKgNiwzrSy2a2/jjvrlSFzi9V/Pt24+XetsDlVKXv5rMhm1frHOitRdqtuPsc1+W+q1Fmmvi3s9vXOI1OwT3iA+Py+pWnSPxV2s0A+Sf0XxqS5nmDs3Mpp+0SeryHM1QHo572w8lm4L5Cf9po85PesXEp6S+avk7pfbrNIfVNlJf71MWvJrY5mu5etyf2zlZ6rp15VIXTNGcYfP6kfr7aj1fFOspyMzM3vpBzajdvOpEqa+pOlvqyyqelro0VZhcKQ67V1v3SX397sVSP7B5ttTuDzp2pr2iOXifq3XhrXsT29zytklSbw3mci0YrZ+BU0o07wiVcvr+LEmFb9C+ZVi7Oqe3WZC9rz5D6m33HiG1O12P2d52fX+/I1/nZa7KNCe2+fXtF0j9pyf0GouKY3V+3r8tf0nqjNeWsb/7FLoiAAAARI2GFQAAAFGjYQUAAEDUhmSGNcys/qZOs18T0joH3fd/+zapp/9wrdSVR2oebefJJYltpifpMm2bw/nK+Nugvx2KfM3qVs0Bvdw0U+r0g5pZHbVal19/aXLu109VPiQ1mdW++dcder/ruSN3SH1ZsWa5zj71x1KvP0kzwz+vOlPqL46/P7HND710pdT5aZ1HdeHJr0j95Es6n2FRjWbiGudphv3Bs65JbHPJHs1HW53O7Zyf0n04NV/Pe33N2SEpfG++2qrHbem+eYnH3LZZr4HYe/94qSc8oRnsnL06x2/jVD2ntBbpeazp5AmJbZa/dZvUo1Pdn1M+O0Hn1xzu866Gwvx3aR8/SnqaE9zM7OH9eh66955TpJ74gn6WpN+iPcZ5RZq1z3h9vy+p0znCzcweeO4oqcOPyGvn/Y/U+W5gM6shuigAAABEjYYVAAAAUaNhBQAAQNQGXYa1N3m/3zccLfVju6ZL3fJjzQAd+dsnpa65QrMkF375EalvW6/3BTczmzNKc0gv1OgcauNz9f7lZppHG+hsCLKqMzqH4vPNU6R+sGqO1BPv0+xY9Wk6thbO12yjmdnpBTrXXY4rSiwzXPQ0//DN9TpHpZnZhvoKqa+bcbPUKdPXc0JusdQ17Zr13LZPM4Nv+93fJ7bZXqJ50fPn6/yEj9+uebGSYNrGgnN2Sb3yuFuCLSTHwJKHdF7GVDD98wcqHpc67ZJ5abwxzV6Pe3g9xJKVCxOPmfhbzRJOWlMjdfVJOqbrztd8c2GhZlz37tbrISon62eNmdnEEfr50tKuY+HrlfdITWa1fzW0J+dE/fiyj0s95Vkda3XTdRx9pVL7kMm5mrN9vEl7ihtXax9jZpa/S1u+0y5YKfWCvLj7kLj3DgAAAMMeDSsAAACiRsMKAACAqNGwAgAAIGqD7qKr0H7fkvjZT549S+pZP9TJnl1ZcEXEg5OlvGbaj6SuSGlg+nZ3bGKb1fs1GH/6saulbg9uHNDTxSbo2aF4DcN1PLxfL5oqSekFOlW/0YuwJux/TX9/ugbnb628K7HNkanhe5FVqKdjtrhwc/JnM/Uiq8nBRVWhR/Ttb9dsvVjqjbv1Ipixs/UCKTOz8yfq+/m2W/RmA5Mea5R69Lc2Sv2rIx+UOpxI/NK1b0lss2S9vjYLLl8l9Qn53BjgcMtxrueFAlXH63Fr+KD+/uQJz0u9sWGU1Oue1nNMUZ3uw7baccltTtH3wEfmPSH1jFw+fw6n8PUML9Z7z6uXJR4z4QFtvzJ5ek4YfZme+y4aoRfbVWe0L/nO5oukHnF/8ry45016nvrHiX+QOu26P5cONEYpAAAAokbDCgAAgKjRsAIAACBq0WdYW73Olt3sNX96zIOfSjxm6q+03nmqTgx+/JU6We4NUx6TOuN10uXr6uZK3d6e7PNHpDVLu7hsrdQFTjMtOS4/sQ70zaHIXS0LJlsOM6v/vetUqcf9SsfOzvceI/VVizWzOiFHJ3dG3/SUTzXrOY/3SIPe7GFsgU7MftfC66S+s16PqZnZ77Zobr1xqp6HJr1tndQ/r3xY6jbT89gPauZJ/dodenMTM7P6EzR8+/UJ90md8ZqFJod46OU7PT8sHKHHuWhBckL44xdtkPqOGr3ZwNIHj5e6RBe3I17RddYdqZ8VzaOT+9m8T/PMj+6eKfUny/RGF4yVNyY85zR4PWbfrj5Z6p2/01yymVlZnZ5DNl2ox+TLE1ZIvbFNe4zv7jxP6qpfHCl1bkYzsWZm75jzgtTh51Ps2ea49gYAAAAI0LACAAAgajSsAAAAiFp0GdYws5oynYPu3JXvk7rytuRTaByrffiZH3la6i+OWSZ1s9eMUMZr9mP9/jFS76vSOVfNzE6ZrnPrbWjSoNGlxZppRf/7S0tj4md7MuOlLsvZJ/W672l+uXjfU1IXXbpD6veN1IxbypJzZcaWC4pZQ3tT4mf17ZoHL03pa1zktP5E+XNSp4PX/8vbzpW6rjWZO14yd4nUm2eOlPqsQs1+hfGxpY2ao7/hXt2mq0jmzW49Q7O1U3LJrA60NxVq/vmcwlcTy/yoVvOjT/5cM6uTNuv43X205mS3n1ogdX6Njo2CqfWJbZ49Rfdjw76KxDI4fB7er6/3LUsXSz1+q/Y1ZmY7FwbXT0yplfo7K86X+lvVOn/0rP/Wz6qKlzSfuu4fkvPFj07r+P1La3BNUJ6OvdhwxgMAAEDUaFgBAAAQNRpWAAAARG3AM6zhvF9pp3OgXrXzOKnzfqr3Xd70dn28mVn+yP2Jn3U2Jqf7OVC3ZDTX8dBWzSTlVes+mpkdE9zzvDWYy7U8h/vHD7TK3OTfZzPSmht6xxq9H3PxrZpZ3fy1RVJ/b/oNUofzNpIz7Ju/265zVm5pLEssc27FX6S+qPgVqfOdntZG52jm/ONbdG7dJ7ZOlfqfj9a5dM3M1rWWS53j9LyzskWztksbjpL6+hdPk9rpKcZOOkefk5nZcXkDfnoe8ura9bOiOfg8GhuMnVzT83p4z3gzs52tmm+uWaD5xZqTNZM6bkKV1D+Yc4vUlz/2Md3AtuTcxIXBvMDFuTovaPi5ir4J+5R1bTpu7th9ptQlr+nj82uDN7yZTXxUx076Hj1Gqb17pW4v1R7CP7NKf79IM6tnnq1zhpuZzS7YrvvpkuM3ZnyaAgAAIGo0rAAAAIgaDSsAAACi1u8hqTALEnq1VecWW/Z9zZvtOkcf/94TdI5VM7NPVTwhdUlKsyGtXud2LU7p3GO/3qO52fYHdI61ue/SuTbNzDa26LyrnysP513lb4P+Fo61MF9qZnZvMD9mwzcnS11wkh7Xd1+6TOozCnROxNTAx8IHlWav2a5TivW9tWiMZsPNzKbkaoYv47ufn/TbNdOlfuTeBVKXnLxL6q+8cElim/urdRv5O/U4p/Q23xZEXK0wiNWXv2Wb1D+bcn9im+1BXpIc4qG3tFHnYX6iXq9X+Ldx+lkSHoMwL21m9k9jn5H6wvM0S9jk9Tx0fH6N1OXB59G4MXVS1742LrHNV/bqzyry9XM0xefPGxJmlR9t1HPK/owe08Jg7ty8Kj0eZmaZl3Xu3NT0I6WuXqxjM6dZ15k/9iSpR35Zz5VfGb80sc3kXM7JPHTMGMUAAACIGg0rAAAAokbDCgAAgKj1e+CuzXROujBXeNGTn5S6TOOm9p7Tn9S6PJlhHZej9wIPc0dhtvFTW0+R+slf6L2f6xZozu7oUs2fmZnNyNd7yjP/5sBrN838rG1tTizzxbveJ/Ws5eulfvX7E6VeMkrzaWmneTOOe9+E7//3lNQGS/Q9YxW+v3/yp3OlzinScVG3XHPKhVXBScfMRtbqOuumJxYRZWv1PJe3V+tp76mWuqssJGPp0NvepvdSX7LtQqn3Nuv7+aFSnVP1rUU6326rT94jPpyr9bQC/fxImeYh2yxP6vDzqi2jdelaHb9mZmvn6xj+p+PvDNYZ9z3iB5uKXB1HR5doT7DiXXotxK4FOn+8mVnrmBOkPmrGVqn3Vuk2xv9Cj2H6i9pz3DXzD7p+r32Q2eA/pwzuvQcAAMCQR8MKAACAqNGwAgAAIGr9nmHtaT643FWaWWuYor8/o0TvG35MXjKb09iukyI+0qQZoY8+oPdmnnS/7lPd2ZpLeufxz0l98ojkPKxvG9GY+BkGVpgv+38b35FYZtxTWm+5YobU3zvpRqnDORJTlsw74uCF+dODyVxVZfS9OGqlrqPi55qDdyfNl3rnySWJde6b2P1+hPOu7j5ac4dla3ScrN4zVuofFuu4MzP72/I1UjMPa9+F42lbRj8LXlpdKXXFs/oa//P+D0r97++vkvrmeUsS2wymy7Td7flS5wTZ+gKn56nPrn2X1EU/LZO6XuORZmb2mTnLpD4qrc/zULyvhrOilL6eFxftlfrcwuelPvr4LVJvnV+eWGdFjmZUb6vSTGvZnSN0HW/Sc8jjs26WutUP/c8mRi0AAACiRsMKAACAqNGwAgAAIGo0rAAAAIhav1901ZN0cO3SyA0aSL/qxXdK/fm25IUILa/pRRMTH9N1TMzTMHL9+zVA/YVZj0n9zuK/SD0ht++TmePwCy8s+E2DXp3wwqvBFXxmNqZQx0LmtDqpLyjSOhVMCs7FC4fWwbyeq1p0YvZvbbtY6paReoxdvl4EUz9VL25oK0puI6P3N7Cxy3Xy96ZyHReVH9MLptbsmaXbfEUvulqanpfYZk2b7tc3xrwgNWOv70pSOlbSpXozkdLX9CMxP7iRSGONXhx3wUc/nthGUb5uo7FZB8/+Hfr5MWKDjp3yNTq2moOxdcqHVyS2+bHSzVKHN03hgr1DK3zvNbZrj1HVpjecyLHgqkwze2yvnhNWPTJT6sJRet769Ft/L3VFKnljgO72cSgYes8IAAAAQwoNKwAAAKJGwwoAAICoDcCNAzSXUZ3ZJ3Xu2dVSV71QIbVbqZMop1qSk+OmRmh+Z9TnN0r9pcr7pJ6Z3i91ODl82pFZHQzCzM6WllFSl4/TrLKZWfU5mhP86rwHpc41sl+xu2nPQql3N+kxff+Hl0r90xPOkDq9RsdNgc4Nb2bJLH3NXD11Tn+r3kxkTslOqesvCib1btBzSnVjMjj7jB0h9R+LX5X6zUWav2Ry+KTwNTgiVyeAv2LeM1LfeOliqcePmy31yJt1+cn36bgwM8udPEnq9lF6TYVP64Tx+6boWNh2qWZgrzj6cak/Uh7c7cTMzHT8kFk9vML32hNN46Quy9G+Zltr8sYBj26dJnVenfYy0y7VHPz7Rr4UrEEzrMPh/T70nyEAAAAGNRpWAAAARI2GFQAAAFHr9wxrmLNIB/Uv5i+RumaeZnP2tmsWbGSqKbGNY/M0I1SeE+bDdJsZP/yyIMPBGcWrpV49enximffNe1LqMBfI33Tx29em86oeU75V6neP1PlLLzptldT/OetNUr+yV+dINTNbWLFB6n8c87zUYWbw+WYdR0cXbpF6R1up1EvWaQ63K7dWnyT1eZWPSM15q2f5TudE/ULFcqlPPVdzgz876kypn19wstSluriZmTm9hMIaJgfXWcytl/LiGU9LfUqx5qHPLNwudXkqmXfm2PevqoxOGL8no58t5xbp/L0/3qd5dDOzfWv1HHDKJS9L/cnxf5K6PEWfMvyeMQAAAAYVGlYAAABEjYYVAAAAUXPe+wP+sn3HzAP/MmLhHGnhfZXDuWCHSxYkNX5NctLaQ2QwjJWG9mTeudDpvIzDZSz05HCNlb6Ok/C9bJY8Rn9o1AzrtVvOlro1o/lSF4QMJxTp/Lz/NlHnaTYzm5Db/VzMfZ0DNZx/+vNbLkws8/Fxy6QuS2kudlZax+5AzL051M8p4XFtM513tbFd50w1M6sPHtPk9SUaFQyNoiBXm+/00pKhck6K5ZxyKNzeMFLqMbl6DqnJ6PnihxvOSaxj0og6qX9Qea/UwzWz2t04GR6vAAAAAAYtGlYAAABEjYYVAAAAUYs+w9pVhq0nwyXr0VdDPW+GQ2cw5822t+k8zDlOn0omOOf1lE/NPqZvGdW+qmvfn/hZaZBhi9FQP6f0dD1EV8JrJHp6zEBkjwfCYD6nhMJzTEFwPni5VeeL351JnmMuKNIMazhH8HBFhhUAAACDFg0rAAAAokbDCgAAgKjl9rzIwCKPCqAvepNJ7avDfR4aDHnV4Sg87geTNh0eCdXhpadzzGmJg96YWCbjGRl9RTcIAACAqNGwAgAAIGo0rAAAAIha9BlWAACAoYTrc/qOVwwAAABRo2EFAABA1GhYAQAAEDUaVgAAAESNhhUAAABRo2EFAABA1GhYAQAAEDUaVgAAAESNhhUAAABRo2EFAABA1GhYAQAAEDUaVgAAAESNhhUAAABRo2EFAABA1GhYAQAAEDUaVgAAAETNee8Heh8AAACAA+IbVgAAAESNhhUAAABRo2EFAABA1GhYAQAAEDUaVgAAAESNhhUAAABRo2EFAABA1GhYAQAAEDUaVgAAAESNhhUAAABRo2EFAABA1GhYAQAAEDUaVgAAAESNhhUAAABRo2EFAABA1GhYAQAAEDUaVgAAAESNhhUAAABRo2EFAABA1GhYAQAAEDUaVgAAAESNhhUAAABRo2EFAABA1GhYAQAAEDUaVgAAAESNhhUAAABRo2EFAABA1GhYAQAAEDUaVgAAAESNhhUAAABRo2EFAABA1HK7+2X7jpm+v3YEh19q/JXVukIAABUaSURBVBp3uNbNWBlaDtdYYZwMLZxT0FucU9Ab3Y0TvmEFAABA1GhYAQAAEDUaVgAAAESNhhUAAABRo2EFAABA1GhYAQAAEDUaVgAAAESNhhUAAABRo2EFAABA1GhYAQAAEDUaVgAAAEQtd6B3AAAAAAcv49ulznFD7/vIofeMAAAAMKTQsAIAACBqNKwAAACIGg0rAAAAosZFV5EYDoHpoYjjhv4SjrV281KnzEnNWIxTeBx7wnGEmVmrz0gdvt+HA94JAAAAiBoNKwAAAKJGwwoAAICokWHtQk8ZozA71pW+5snIKQ1OHDf0Rl/zp72RdjlvaJ9wePT0+dHTOaOhvUnqfEtLzXEfHpbt13Hyuz0nSt3u9ZwxrXCX1GePWJ1YZ9p0bFbm6jaKUwV93s/+xKctAAAAokbDCgAAgKjRsAIAACBqZFi70GY631m+0wxRbxJEtZlGqTdn9G+D9a2jpa5qGyl1Wc4+qS8o2pHYRmmqsBd7gjcizKNtatPjek/DUVJ/qPQVqTe2JfPOmSCveExe3Lmh2PSUBx2IjF+4T7vb90tdHJxDtmRapW7yyX0Of3Zj9elS78/oOicV7JH6yvI/Sz0rPSKxDfS/55ubpb5i+YelbtpcInV7cZvUb1/wfGKdnxu9TOopuUVSk7WPT3VGP+N/2zBT6m8+e4E+YHe+lHk1ekzDU8iSBQsT28xP61j64vQ/Sv2u4r0H3N8YMIoBAAAQNRpWAAAARI2GFQAAAFEbFhnWMF/2UmuL1A80zJN6e0upLl83Qeq1OzV/amaW+2Kx1KXrdJt7ZunfBmMWbZf6rRNflHp8rubRcobhfYMHQni/5jAP+e2qc6W+79ljpB5ztmaANreOSmxjUdGaN7KLw0pVkPMyM/ufvZobnpiulfpdI7Q+FPm9cFysbNH60cZZUhelgnPM7rlSP795sq5/j+bTzMwKtuvpuWRTkNVt1HrZmzWfdsYZOg/j9FzNTnaFrGP3wmsTVrUWJZZ5qlGziLdvWiB1y91jpB65R49j45n62eGa9Bz0xw2ze9zPq8Ysk3pCbnHXC8LMknPfpoLv8opSeX1aX9hzbAiufTAz+9bO86Re9sfjpK4IPib2j9YeoHGybiO3QX9fu1OvizEzSxXqOeIPo+ZLfcmIRxOP6Wygzw+cnQAAABA1GlYAAABEjYYVAAAAURuSGdYwP/Jokz7Nz7/0XqmbnqqQuuJlzaeVvFQt9fRCnUPRzKxpgv5s17E6R+Lss9dJ/f0jb5d6arqnjBFzdR4O4VgJM6uPaLTJnvmJ5tHcsZo/29ii+eYrSlcktjmZPNkBNXt9Hz3RNC6xTENG3wvTi/Qe2jnujeXNwoy7mdkTjdOl/lPNHKmXb5widcVS3cdRv9FxMKNkm9RtMyYmtlkzT8diUZXmz3yOZtbOmqdzAC8s0Dx1jmPe5r4K58q8aa9e77Bm/9jEY5ZtmSF15ulyqct263jbdqGOeRdcruD26GdJY23yOI6e2SB1aZC5DMf4QGcRY9MavD6twVzsRda3c0pVkHUO86pmZg8+eqzUE57Tfdh+mg6ERYte0rpUe4rvPh9sozHZ3rW36DmltkXHUpi1nd5jX9K/GLUAAACIGg0rAAAAokbDCgAAgKgN+gxrY3syb/bA/jKpb9x+mtT5N2mmqCiYUzGTp9mR1Z/TXGJZpc6RamZ2zuRnpb6k7DmpT8zXbeQ7zYaQMTr8wtfYLHkf+tUteg/4T/zXF6QuSuny04/eqsuXr5S6NBVXBig2YWb1of36epWk9HiYmS0e8arUx+V1fxqra9d1rGgeIfXde46X+vlanSPVzGz9Os3STnhIs2CzH90kddtWzai6uTo3Z9UiPafsPknPD2ZmlqevTeoRzTLWnq/P6+8rNCdb7JJzu6JvljZqNvmh3Trf7tQRuxOPObJc5wF+caLOh1l+hmaur53xe6kfqdd89G0PnyJ14cZknrL+BM1Mr2jR98TstI6V0Tn6HhjuUkFwOL+Hec+7+izp7Jb6o6V+4PFjE8tMXqbr2HSRfrbccf61UqddOLer9jEupb+vvDvZQ2w7XX9W36LjZldGM63T9ZQz4OiKAAAAEDUaVgAAAESNhhUAAABRG3QZ1jA7cv/+5L3ad7SWSv3Ccp1DMW+29ump+XVS37BgidSjcnQyznSQezQzm5Kr95QOs5Fp130YhMzqwGj0moG+aOlnpZ74qo63Xe/Ueep+Ok3n0y1NaQaoq6wTx/qv6oMMelFK73d/fF4wEa6Z1bfrfKQ5rvuc8M6MHoPVzROkfm2fzsO88fnknKizb9H5ONtz9RhWnX+E1PsmHSm1P7Ze6mMn6pypZxYkc/EP/lKzi7Xz9JxyyZwXpD63UOeLznHM3fxGjc/V47K6SrPMq1omJR4zf7Lm2osm6hypf1O5XOpzC3VszM97TOon506Vuv1nyblfn1ysy4xO6zqPLnsx8Rj8VVEwd3OqhwxraG+7nqeuff5NUhdvTJ7zN5+v7+drz9G+Y36e9gyrW/XcuHTPfKknLtHM+oiXtie22f7OMVLPHKl56mnp8HwbV9aZT04AAABEjYYVAAAAUaNhBQAAQNRoWAEAABC1QXfRVVVGL3p5ofHoxDK3/fosqdPFGm4+5c2rpD6zTC+AWFqvYeYwwJ5jyQtprhz5mtT5btC9tENOq+9iMvbAB9ZdIvWk+/VvuB2LdPkvHvOg1OGk9dwAom/CCczPyAmXSF44VNzHl3R6rl4IN27kWqnHp/Wiy2/M1IuwzMzWXK43Izlyvt4Y4KIxz0g9Nb9K6lav46QkRydy/+frr0hsM3+/nrcWn6EXznysQi/OKXT6PNF34ft3cYHevMG54ILbLcnXfEWdXgBVPF4vumrIBJP8N+uAvmb7xVI33jpeH5+cg94+MVlvGvHR0tW6DynGRnfSLnHi6ZMXWvTCz5z1eoxbS5KPWXSC9h2vNOnFnr/coePomZV68fj4R3XclC3Tm9a0Hqs3KzEzmzBOLyJ8xyi94dHYyG8owacpAAAAokbDCgAAgKjRsAIAACBq0Qctk5lAndA3zKuamVWs0tzRnhk6Ae+LP9Xc69MTNLPaOFUf/8Ozfi31rraRiW1WBxOgT8nNSyyDwyscK+Hkz/9QdVziMZtumqbrqNTHvP2sP0v9sdLNUjd7ncS+KMVxj02YIy4Nsp4z0zp59jfm3ZVYR8V8zSHOTmsGtclrtvGJpmBC+SD2/tU7Lpd69JYubjDxAc3Bfmjso1KH2Vzy0odefnDDl2uOu1nqT7a8L/mYNXoTmdxVmn++YecZWrvTpZ51g07e3rJYz0nXXH59YpvnFOqk8tw04vBqM70+YlOr3syh9Qg9HnkvJo/HC3fMk3pVi9alG/SzZWa19hjpbbW6T416fc/Ohck86vFl66U+NV/PY5kgax/bOSWuvQEAAAACNKwAAACIGg0rAAAAohZdhrWnuTM/teHtUuft8Ylltp2uTyu/Rn/fPEozQaHcWn38c406H9qkPM2OmJm1JncDh1mYWQ09uD9f6lseWpRYZmJVkJH+xE6pLy3X+TUvXfdmqetaNEd495zbpCbTGr+j8vQYzkk3HGDJv6oNhl6YWd3WWi71NQ9eIPXov+jjWy4PTlJmdtU0nfP3tHzm+D3cenpNzwrmZb3quKWJZX5QcLbUJct1js7xj+yVuv1FnTO19dwTpP7OpzSzGuZV0f9qMnoMRqQ0X3rp/OVS31mg18mYmbWv13GRX6N9Sf0k7UPSDdobpdu09qfqBL3tZ+qcq2ZmF47SuVrDz6eePlMHGmc8AAAARI2GFQAAAFGjYQUAAEDUBjzD2tPcmdszOrfY85sqpXYzkutsn6Dz2BUv0MzQGePXSf1CrebPqm+aIvXDx+k9ed9fqXNzmtH5x6DBa67oa6+8W+ppdyazX+sv0Zzr2aU6J+d7l35S6tx6ved0eq+O1++NPUb3YbTm08y6mluY0RO7/V4zak81V0i9J6NzHv78ldOkzm3UYzzlI2uk/uaU3yW2eWSuzufJOBl4m9qCuS5bSxPLNGzXbOK4Op1P07VoDtYv0uzhhkv0Y/npRr2H/HH5zyW2Gfs94IeaguC9WBBkWKfk75b61CkbEuvYVqFjZ1ON5t73b9RjmtOqedP8lfp5tu5SzeK/+8gViW1eVLQr+ImuM/ZzTNx7BwAAgGGPhhUAAABRo2EFAABA1AY8w9puOoFpQ7vmMkpSmhm8bdF1Uu9u15yXmdnmVs2XnVmo98+dEmTDnivTuTY/nv85qbctn6Ab0BitmZlNDu7rjUMvzH6GY+er28+SOvdXOg5aRibnmCuo0r/ZVl6rGdRRwWFtLdbMauWtm6S+Z9HRUneVYQ33OyexBAZSVzmuZ5v0nLG1dZTU169bLHXT+hKpK0/YJvU3Ku+WOsyrHmg/0L+qMvukvqH2VKnv2zw38Zi8Wn1HuzbNrK79UHDf+VKdT9MVaeZ1dcN4qfeVdjHpNyeRflWeo+/XxQU6N3tlrs6BOj63LrGOPaW6jsbxej3FdSk9p4z5lX5+1b1Js80LF74i9WVBX2Nmlu/SiZ8NJpwRAQAAEDUaVgAAAESNhhUAAABRG/AM649qdY7Te3doBvCdE3UusZLUfqlH5uicq2Zmx+VvlnpVi2aGrquZJfXtLy+QevQezYrsH6u5xRMKNia2mbLBnQ2JUU/3NV6yV+fPve8xPY4zbtb5clPHzUuso6xAs4a1szUMtn+qzq9nLfo3XtskzTK2tGkGO8zAmZmVpwoSP0M8VrYkzylVmXFSP1Kr55DdG3QOxbHzqqX+2rR7pJ6T1rwaedU43Vo/R+ows1r/kubkzcxG/UUzplVf0PH0jiP0M+03z54kdWq3fpZUN+l8nAX6cYQIlKb0YodjdHpTOyqtmdYs/dnjzXoOKLrnLVK7Jp1DteXKGqmvGPek1HPzkueUwX6eGdx7DwAAgCGPhhUAAABRo2EFAABA1A57hrXZ6xx04TxgP3rsXKnnfk/vwfv9j79VVzheM4LtmWSgxzfq08rbpbnEscs1G1lern17i8YarfykKqnn55FX7Q9h3qY2o/fx/un606WefptmxXJmTpN6/TvKEtsoOUGzhpdPWSn1q/s0/7ziDs1Ybz5H7xt+9gSdC6/FJ+dMTDsmTYzJ9rYGqde0TEwss7JxitRPrJgtdapcs87fmnOb1Kfl6zlnsGfJhoowJ99mOifqTZs0X1q7UbPKec3Jz589b9Px9MDx10v9Qstoqf8wJpjL9WXdRuVizTqOzUnO2Rs+D8ZXXMK5t82SvdEH79X53+f+aYvUf/m8XrPxH7NuknphvvZO+U6zz0MBoxoAAABRo2EFAABA1GhYAQAAEDUaVgAAAERtwG8ckFeuF8pkXl0n9fQvrpc6Z5QG0l1xMljcXqo/8znBRS65QZ/uNMS+5TwNSD981C+DLSRD74TcD70wlH5Lvd5kov7pMVKP3qOh841/M17qlqnJCeFr9uhYWbLyTVIX7tKLKnwwIfSUc/QmEueXrZJ6QhcXSGBgtXq9sGZ5cBHMtlY9x5iZ3fTiifqD4BqKb5x4l9Sn5us2zJjtPUbheXtdq96YZttOvVBzwiN6HLefHh5ns4/OeVrqmnb9mP2Hl9+uD3g4+EwLVvmJsQ8FW0h+bPP5E5fwHJPq4v3/wdcuknrOtfr5tf0tk6W++oJbpT6lYKvUo3P0AuChiFEOAACAqNGwAgAAIGo0rAAAAIjaYc+whjcKCCc4/q+TbpT6yh98Suppd+iNAnKqdfL4TF4XTyHIqLYH9e5jNFc4+rLNUq+arZN+F6eGfjYkBuHYeKpZx8791fOkHvN8m9RtpQVSp3X+biu5PwigmlnjOB0bzWUaTtw7T3O082fpWPnoxEekvrCoXuocbhIQnVdbdZL/zS16U4BbtxyfeIxL6bj49/Nvlvq8wu1Sp0zHIhnDOIXnnB0ZzbTnF+n7v2RdcOOanOAuM2Z208ZztE5pXbZG8431GlW0RVcul/qEfD1vhfuMgRcekzCz+mcdNmZmtuGXek3GyEo9L7330/dLfUy+ZlYnHYLrI8Ksbew3teEsCgAAgKjRsAIAACBqNKwAAACIWr/PwxpmuU7TqJfdcckPpP6v0xdLvby6UurtuwsT2ygeofNtHjdOsx8/nXCf1FPTYUY12Cn0i/ZgcsvNrRVSp1zw+wv18QXbNH/WVKn5s5zxdYltVhRouGj3Ps0FnTRWs4n/b6KOnTnpfN1G5Bmg4SjMaa1p1fl708HEl9NLqxPr+PARj0t9WbGOpYwnszoYhcfpxDzNEX7tmN9r/X/eKXXhpuRxLlurecb6Sl2m9HObpL5wlM41/snyFVIztgafBq+fK5944SOJZUoa9PMs7ys7pF5UtEbquXl63A/FOIg9sxpi5AMAACBqNKwAAACIGg0rAAAAoua89wf8ZfuOmQf+5WESzmfW7HWuzYbg/vKZLva/NKXz1hWlkvNvdrfNoZoRSo1fc9huaH44xkpdu97Xuz041q1h5rVN521t8hrRznTx91mr1wxPRUrn+Z2R1m0Up4ZHvvlwjZWBOKeEajN6jNO9eL+H80mH8ywO1XNGTwbbOeWNeq1VJ3de1TI2sUyT17EyMqXXVIzJ0bmap+XqZ9zI1NDMrA7lc0rYQ1xXd4TU33n8zYnHFI/eJ/VDJ14vdUnQt4TnoKGqu3EyNN4JAAAAGLJoWAEAABA1GlYAAABErd/nYe1JmNcpckGOI8glhnN3dqWnjOpQyQgNNaWp5By7nYXHtaL7qLKZdXUP7nBs5HexDIaS8kNwD24MD+E5ZkpuUVBrprV39DMsx/V44sIg8+YRf5H67PNeSSwzLa2Z1Hw3IrEMFJ0aAAAAokbDCgAAgKjRsAIAACBq0WVYe5LInw7QfmDghWMhzJsBwBvR0/UNB3PO4ZqJoSc8ptPTxQO0J0Mb7xwAAABEjYYVAAAAUaNhBQAAQNQGXYYVOBCyYQD6E+ccoP/wbgMAAEDUaFgBAAAQNRpWAAAARI2GFQAAAFGjYQUAAEDUaFgBAAAQNRpWAAAARI2GFQAAAFGjYQUAAEDUaFgBAAAQNRpWAAAARI2GFQAAAFGjYQUAAEDUaFgBAAAQNRpWAAAARI2GFQAAAFFz3vuB3gcAAADggPiGFQAAAFGjYQUAAEDUaFgBAAAQNRpWAAAARI2GFQAAAFGjYQUAAEDU/j88XmkvSZhhUgAAAABJRU5ErkJggg==",
            "text/plain": [
              "<Figure size 864x360 with 10 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          },
          "output_type": "display_data"
        },
        {
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAqwAAAE+CAYAAABIhELlAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nO3de5RU9ZXo8b3r0S8aaOhuoHkjgiKKEh+oYEwimmBidG6icn0nuZlZK/Hm7WStO7lXM2MemjFjTOI4SYwmxjGP0RjxEaOioPiMQnxFUBR5No+Gbppuuumu+t0/TjHp/TtFV4M0/avq72ct1upddeqcU1W//tXm9P7tUuecAAAAAKFKDPQJAAAAAL0hYQUAAEDQSFgBAAAQNBJWAAAABI2EFQAAAEEjYQUAAEDQSFiBIqSq16jqr3I/T1TVXaqaHOjzAvpKVW9X1WsP9WMBFCcSViBQqnqRqv45l4xuUtWHVHWev51zbq1zrto5lzkIx3xCVf9XH7etVtV1uZ8/rarf73FfnaouU9UmVW1W1WdUdW6P+69R1WsO4Pycqm5R1VSP29K52w55U2lVvUJVnzrUx91fufe1IzeWtqnqParacBD33acxAwAHioQVCJCqfkVEbhSRb4vIaBGZKCI3i8i5A3lentkisjz38/Ei8lKP+3aJyKdFpF5ERojIdSKyqGei+R7sEJEFPeIFudvQuyudc9UiMl1EakTk3/Z3B1zFBzBQSFiBwKjqcBH5ZxH5vHPuHudcm3Ouyzm3yDl3VZ7tJ+euPKb2Pl5Vb81dld2gqtfuTTT2XhFU1X9V1R2q+o6qLsjd9y0ROU1EfpS7EvejAqd6goi82OPn/05YnXMdzrmVzrmsiKiIZCRKXEfmOf86Vb0/dyV2u6o+qaq9zU13iMhlPeLLROSX3j57ew2mquri3NXfbap6p6rW9HjsGlX9mqq+rKotqvobVa0o8Fr0fOxVuce25c5hdO7qeKuqPqqqI3ps/ztVbcwdZ6mqzuxxX62qLlLVnar6Qu45PNXj/iNV9ZHca7ZSVS/oyzk657aLyN0icnSh/eT+9P7vqvqgqraJyAf7coy+PL+cutyxW1V1iapOeq/PD0BpImEFwnOKiFSIyO8P8PG3i0i3iBwu0VXQs0Sk559s54jIShGpE5HrReRWVVXn3D+JyJOSuxLnnLsy385zSVizRFdNr8r9fIKIPK2qr3nbviwiHSJyn4j8zDm3RUTEOXeNc+6a3GZfFZH1El2NHS0i/0dEevvz/r0i8n5Vrcklf6eJyB/24zVQEfmOiIwVkRkiMkFErvEef4GIfEREpojILBG5opfz8X1CRM6U6ErmOSLyUO451Us0536hx7YPicg0ERklUcJ/Z4/7fiwibSIyRkQuz/2LnoDqEBF5RET+M/fYhSJys6oeVejkVLUud47L+7ifi0TkWyIyVET2t/yht+cnInKxiPyLRGNxxd7738vzA1CaSFiB8NSKyDbnXPf+PlBVR4vI2SLypdyV2S0S/el3YY/N3nXO/TRX8/oLEWmQKFHsE+fcZyRK5NZIlGh8TkRucc7VOOdmetvOEpFhEiU9+0p2unLnMCl3JflJ51xvCWuHiCwSkQtz/+7L3SYihV8D59xbzrlHnHOdzrmtIvJ9ETndO8ZNzrmNuauRi0TkuF5fFOuHzrnNzrkNEv0H4Dnn3HLnXIdE/wmZvXdD59zPnXOtzrlOiZLmY3NXh5MSJZVXO+fanXOvS/Re7fUxEVnjnLvNOdftnFsu0VXT83s5r5ty/7n4i4hsEpGv9HE/f3DOLXPOZXPPoc/29fx6bPKAc25p7v5/EpFTVHXCAT4/ACXsYNSTATi4miT6U2nqAJLWSSKSFpFNqrr3toSIrOuxTePeH5xz7bntqvuyc1X9uER/fk9LNH80isgQEelQ1YtFZL5z7s89H5NLcu5S1b+q6grn3F+83X5PomTmT7lz+Ylz7rsFTuWXEl0lVRH5undfr69BLqH9gURXZofm7vNrYBt7/Nwu0dXYvtrc4+fdeeLq3HkkJbpyeb5EV1+zuW3qRKRSote35/vW8+dJIjInl4DulZKoXGJfvuCc+1nPG3J/gi+0n57H7bMCz6/F37dzbpeqbpfotT6Q5weghJGwAuF5RkQ6ReQ8Efmv/Xzsutxj6w7kCq30/qd4cc7dJyI1qnqLiCxxzt2lqpslujpa6OpbWkQOk+gKX899tkpUFvBVVT1aRBar6gvOucd62deTEl2VdRJduZ3a475Cr8G3c487xjm3XVXPE5FC9br94SKJFtHNl+hq9XCJEmcVka0SlTSMF5FVue0n9HjsOole/zPf4zn0ZT8H2n2ht+e3138/J1WtlqjGeWMfzwvAIEJJABAY51yLiPw/Efmxqp6nqlUatW5aoKrXF3jsJhH5k4jcoKrDVDWRW2Tk/8l7XzZLlFQWcryIvKSqU0Rkk5+squrJqjpPVctUtVJVvy5R2cFz/o5U9WOqerhGl0NbJFqglfW36ylXMnCOiHzcLx/ow2swVKIuBi2qOk5EYgvZDpGhEiXWTSJSJVEiLSIiuXKNe0Tkmtz7f6TYhWb3i8h0Vb00NzbSqnqiqs7Yz3M4WPtJqWpFj3/p3p5fD2fvHScS1bI+65xbdxDPC0CJIGEFAuScu0GiGsNvSHS1bZ2IXCnRgqNCLhORMhF5XaIrWv8l0dXIvviBiHxSow4CN+XbIJeMTJboyt/75G+dAnoql2jRUJOIbJCopvSjzrmNebadJiKPSpREPiMiNzvnHi90os6515xzr+3j7t5eg2/mzrtFRB6QKDEcCL8UkXclen1eF5FnvfuvlOiqZKNEfwq/S6IEcO9V6bMkqsvdmNvmOole9z47WPsRkX+XqNxh77/b+vD8RKJFVVeLyHaJ/hN0yUE+LwAlQntf2wAACIGqXiciY5xzlxfcGABKDFdYASBAuT6kszRykoh8Rg681RkAFDUWXQFAmIZKVAYwVqLa4hsk3m8WAAYFSgIAAAAQNEoCAAAAEDQSVgAAAASNhBUAAABBI2EFAABA0EhYAQAAEDQSVgAAAASNhBUAAABBI2EFAABA0EhYAQAAEDQSVgAAAASNhBUAAABBI2EFAABA0EhYAQAAEDQSVgAAAASNhBUAAABBI2EFAABA0EhYAQAAEDQSVgAAAASNhBUAAABBI2EFAABA0EhYAQAAEDQSVgAAAASNhBUAAABBI2EFAABA0EhYAQAAEDQSVgAAAAQt1dud2cZp7lCdCPpfYsyb2l/7ZqyUlv4aK4yT0sKcgr5iTkFf9DZOuMIKAACAoJGwAgAAIGgkrAAAAAgaCSsAAACCRsIKAACAoJGwAgAAIGgkrAAAAAgaCSsAAACCRsIKAACAoJGwAgAAIGgkrAAAAAgaCSsAAACCRsIKAACAoJGwAgAAIGgkrAAAAAgaCSsAAACCRsIKAACAoJGwAgAAIGgkrAAAAAgaCSsAAACCRsIKAACAoJGwAgAAIGgkrAAAAAgaCSsAAACCRsIKAACAoJGwAgAAIGgkrAAAAAgaCSsAAACClhroEwhBxmVNnFTyeAD75s8ZhTCnYC9/7GTFmTghamLGDvIpNAeV4rgpvWcEAACAkkLCCgAAgKCRsAIAACBog7KGdX/rz4C9ulzGxGlNDtCZICSlWC+G/hEbK97nUbdkvPsLPB7By5dzbMm0m7jLu39iqtrEnc5u8VZXt4mr1I6biamq2DGLfewU99kDAACg5JGwAgAAIGgkrAAAAAjaoKxh9es4XuzcY+LVXfUmPrlig4n92hKUrh1endGitokmPq5ivYlnlVX0+zmhf23JtMVue7bDzgk/XvshE7d02vf9pFHvmvi7Y5aZuCpR9l5OEUVkmzeevrdtrokfXHOUiY8dvdHE/zj2jyYem/RqXEWkLjnkvZwi9pO/lqFQ79x722pi+/jd1vkmnlHdaOIvjPyziX/TOs3E33vpLHvMlD2nzx2zNHbMy4e9buIRyXida8i4wgoAAICgkbACAAAgaCSsAAAACFpJ1rD6Pc/8vnZnvnqBiRM31dm4yza++8FXtpt42ax7Ch6z2PudDRb++9aU3W3ii1ctNHHzr8abeNspthfecwtuNHFtojJ2TMZGWFq89/zJ3Q2xbb621M4Zh91l54iaJX8x8RsnzTTxF//N1qz+aPwTJk5JvJ8v4yR8+fprrthj54TzH/iKiY/4SYuJx69628RNRx5m4r+7/Esm/vE5t8WOOa/C7rM6QS19f/prl+2JWp+w7/nIZLmJ72ycE9vHyj/amtQ/10438ZxzV5v4jrV2HyOesO/x7lG2jvZPo21ttEi8hrXYMCMCAAAgaCSsAAAACBoJKwAAAIJGwgoAAICgFf2iK7+Br4hIp7MF0bMe/byJj7zWLqLKvPWCiRNDh5q4cdkx9gCz4ueRFbsII76EAsXgzS67SGrjoxNMPOm+N0y8c+oR/X5O6F9p77e1NRtfsFKx3i6aylTYOUbn2UmhfPUWEz/+pJ1D3vrkoyY+Ms2MUQz8z5Z8i+W6nL0tucteF9pTZ5v8d0+yY2fIWztMPOUP9ott3jgrvijwzEq7cJBFwP2rPZs28Vbv/uasHScr1kwQ39Sl9ktpGk+2TfzX7LGLwTeusXG9PYR0V9oc5KxR8QVW1Yny2G3FhFEMAACAoJGwAgAAIGgkrAAAAAha0dWw+jWrWYk3bj7h6c+aeMY3t5nYVdkatY1XnWricY/vNHHnSHuMfM2iEb5871unsw2f7222zZnHeXVGOnyYiRNH7DKx/0UBfm2zCPXNoalK2PrUaWWNsW1O/cjLJh59rp0jHm+0TcD1W7beLNVum3q3Zu0xGSfFIV/Nqm92uZ1n7lt4g4mvOvkTJm68Y7KJq19qM3HTGfUm/tAQW0cfKctzGw6WXdkO7xZbw5r0fn9/3XKiiatXxOviy1bbL4xInGi/MOK6ZxeYeNhfbbpWud1+du1utfe3Z+L1qn7+lBA7L4Ve6xz22QEAAGDQI2EFAABA0EhYAQAAELTga1j9ukO/5uLqLcfHHjP5eltP0jmp1sRNX7Z1idVlm02c/J1tcDbzfbYGNl+dRzZPP1iEb3vW9jh8dJ3tqzr6uVdNvGeu7ad5wfTn++fEcMj4vTW3ZkbEttncYXszv9Jke2F2PWDrDEc02c6MqaPtMcanbN/MhNgejAhTn2r8vHLkpNo5pvHOySauvfUZE2dmzzTx/E/b+2eVxesh/drEtFIBfTBtznR7t9ga1uasrRd9autUE9ev6Izts3uTrZUft6jSi+32rsoeQzN2oHVX2nnr2R1TYsfsGrnCxOXhp4AGV1gBAAAQNBJWAAAABI2EFQAAAEErrgIGifcrvPvBubFtDt+63sRvf3Wkia+d8ZCJr7vhIhN3jbP1ZefUP2LifP08/dpahCdfr8v13bZuSB62Y8V129qlDafZ7c8a+kqvx2RchKfdq1t+vMP21v3y0oWxx4x5zE6VtS832w3eteOg9cyjTHz+1CdNXJOw+9vt7DmJiFRrvFYRA6slaz8b7m8bH9vmmZ2Hm/iS2qdN3DTb1ptWL7A9O7Nlds7YnbE9VuM9QUUqlT6s/SldYBqvUPs5Ma9+tYl/e1p8nIystT2/Re1Bal7YZO9usr2fd88ca+JNp9vPt6vGPhU7ZpU3TkLvu+orrrMFAADAoEPCCgAAgKCRsAIAACBoRVfDurbb1hCNeD1elyhZe1v6lSEmvv5RW7M66re21+bb/3i0iU+vetM7QLy2rNhqQQYjv1ehiMjdzbaOqGFJk91ghv2O+LrTbF1RfdKORxGvJjYPvwaasdO//D6rf7Wh/HDtGSaecld8Tql89R0TZ7bZceLXOg95Z5eJb3/e1to/NtH2+5023PZtFRH536MfM/F0r5CuKkHd4sFW6Hfz4fYxJv7uTy+M7aNyqx0/S86zPTlTNV69csL29Mym/BpWe39HnnmsklL5fjUxVW3i4Qk77/u1oU1DVpl449k1sX3WnGv7wfs9wFPto02cbrNzzOZ/sOdw++w7TTy3PL7WJlnk/Xn5pAQAAEDQSFgBAAAQNBJWAAAABK3oaljrkrYGo3ViPOce2WG/t3fST1baDbx6s80X2ZrV88+x/cump+mHWIz8erR2r5ZRRGTJJtszsa7R1hI2fdTWFZ3bsNjE45O2vox61PBknK0pbM7aOuP1zba+bOJar8eqSKxHYseZs+0xKuz7Xv3o6yY+4kpbt5g5cYaJF19RGzvk6HLbd/HqUS/Gzwv9yp9DVnU0mDgRn1Kk7vltJtaH7fuYbbX1zVpVZeLVX7Jz0rdH2c+joXlql5l3Dq3hCTuH+HXywxK2V+5n6pfG9jE+ZWtQF2+w6yXUW4vTNMPmIVccscTEp5T7tc2lV9jMKAcAAEDQSFgBAAAQNBJWAAAABK3oalj92pEvXnZvbJvvTDvbxMkd9mmWT2418fXH3mbij1b539VMXl8KRiTiPVLPn/SSiW/7j1NMPL3+bRNfVmPrCKsStj8fwuP3Kz0q3WLib8x80MTXXPux2D6yGfs+f+VY+xi/V/Nn37jExG2LbP/O5B5bn5asjH8//JGVG01crunYNji4CtWCnjLEvs9LPj4tts3aigkmrn1thIm7q+wxNp5l11R8Y+7v7TG92sQ04yA4LVmvt67Xq92vVxURWdk13MTbtwwz8VCvBLX5GDtOTquyvV59pVjXXHrPCAAAACWFhBUAAABBI2EFAABA0EhYAQAAEDR1XlPtnrKN0/Z9ZyC6nN8sV2R9d7zAuaf6pF2EVanxRsw9lUrxcmLMm/3WSbgYxko+fsPnjd32SycqvFesITU4Fln111gJcZz4c0hLNr4Ayl+wt79zwjtdtln8u912gUVVwo47EZHjy+yXpIQ4Dw22OaXdW1yzPhP/5oCtGTtWOpxdJHVsmf0igbrkkIN0dmEr5TnF/4KJHVmbg+Rb8Lsh027i72/9gInfaBlt4kvHPWvihdX2S25CnB8ORG/jpDSeIQAAAEoWCSsAAACCRsIKAACAoBXdFwf40pqM3TY+ZetFsmLrSxIF8vRSqQVBYX4z9omp+HhCafPnkOGJin1s+Td+zZovK7asbmKqyottE/AkzeCLgv8lFNMT8fUPU1N2bCTV1khnXKUX+9vz+VNs/PesL3XJE731EDeMed5uYL9rJM+4GHzjZPA9YwAAABQVElYAAAAEjYQVAAAAQSv6GtZ84nWt1CWib6gfQ766+P1VaA9+3WK+mljGYnEq9L7xviIff1wUqpMfjPjNAQAAQNBIWAEAABA0ElYAAAAErSRrWAEgZNQxAugNc0QcrwgAAACCRsIKAACAoJGwAgAAIGgkrAAAAAgaCSsAAACCRsIKAACAoJGwAgAAIGgkrAAAAAgaCSsAAACCRsIKAACAoJGwAgAAIGgkrAAAAAgaCSsAAACCRsIKAACAoJGwAgAAIGgkrAAAAAgaCSsAAACCRsIKAACAoJGwAgAAIGgkrAAAAAgaCSsAAACCRsIKAACAoJGwAgAAIGgkrAAAAAgaCSsAAACCRsIKAACAoJGwAgAAIGgkrAAAAAiaOucG+hwAAACAfeIKKwAAAIJGwgoAAICgkbACAAAgaCSsAAAACBoJKwAAAIJGwgoAAICgkbACAAAgaCSsAAAACBoJKwAAAIJGwgoAAICgkbACAAAgaCSsAAAACBoJKwAAAIJGwgoAAICgkbACAAAgaCSsAAAACBoJKwAAAIJGwgoAAICgkbACAAAgaCSsAAAACBoJKwAAAIJGwgoAAICgkbACAAAgaCSsAAAACBoJKwAAAIJGwgoAAICgkbACAAAgaCSsAAAACBoJKwAAAIJGwgoAAICgkbACAAAgaKne7sw2TnOH6kTQ/xJj3tT+2jdjpbT011hhnJQW5hT0FXMK+qK3ccIVVgAAAASNhBUAAABBI2EFAABA0EhYAQAAEDQSVgAAAASNhBUAAABBI2EFAABA0EhYAQAAEDQSVgAAAASNhBUAAABBI2EFAABA0EhYAQAAEDQSVgAAAASNhBUAAABBI2EFAABA0EhYAQAAEDQSVgAAAASNhBUAAABBI2EFAABA0EhYAQAAEDQSVgAAAASNhBUAAABBI2EFAABA0EhYAQAAELTUQJ8AAAAA/ibjsibOijNxQtTESS3964+l/wwBAABQ1EhYAQAAEDQSVgAAAARtUNSw+rUg+6tQ7UhfDIb6ksHAH0u8r4NDl8uYOCu9zykJ71pAoTmDcVQ63uvnTSGMldJUaNykNXmIziRcjHwAAAAEjYQVAAAAQSNhBQAAQNAGRQ1roZqf/e13diDHQHHakmkzcZVXR1QuaRNTZ1Sa4u/r/r3Pfg2sj1FTOvzPAv+992N/bPn10ak+jA5q64tPp+sy8fruThMv2nW0iR9qtPERwzeb+Iv1j5t4arr6vZ5icBjVAAAACBoJKwAAAIJGwgoAAICgFX0Na77eZd1ia4Tub6s18bfeWGDi5rU1dgdeyWpqZIfdf1NF7Jjp+t0mPnXSOyaeP+J1E188tCm2Dxx6/vhZ1WXf6zt2nGLiu16YY+LjZ9j3+cZJ95q4IVkVOyb1ZWHble2I3Xb15lNNvHj9dBM3bxpm4vQOW3eYbrWTStmc7Sa+bdYvYsc8pszWRzNuwtOS3R277dbmmSa+a80JJm5+zX4eVW6xY6NzhF1DMXXuuyb+/ITFsWOeXtFs4mqNf0Zh4Gzz1kKIiHxv21wT3/d7O8eMX2LHVnr5ahO/PO84E19/jU3nvjAqPk6OTJebuNjmlOI6WwAAAAw6JKwAAAAIGgkrAAAAglZ0Nax+zeGa7vbYNl9fe56Jt15/mInrttl+Z3UJGyfabX80zXjfI14Rr5vddpztefb8hyfafXq9XT9ZbXumlautV8PBl6/eeVPGjp+r1pxv4ncfmGLiCau67eMn2trFMi3csxcDq9B3dv9q59TYbU/8h61dHvWz522ctXNEsma4id3kcSbevH2EiR+cMit2zOkjXzFxlZbt44xxqKzv3mXiS1deHNtm4/NjTTzMlh7K1OU7TZzYY+eUrjpb975x82QTP3z5MbFjHuX14KxU+rIeSv6cstbLS3663daniogs/qFdHzHx9ufsPt9/rIkbL7K10Q2LbG3zn1bYPq1/f+aS2DH9HvPF1v+ZUQwAAICgkbACAAAgaCSsAAAACBoJKwAAAIIW/KKrQsXMC1/5dOwxtVfbBUwVy180cdf82SZur7cvQ9OsSnsOVfYcXNoWLouIfOC4V038zbEPmXh00jbsZZFV//PHjl9wLiJyy3Zb+L56sV1kNXFJq4nXnzHUxN+Z+kcTD0/YhTEsdgiP/57446Q+Zd9zEZEd8+zCzK5quwhr1xS76Eq77OK7uuU2TnnfTZDU+EKwpLeAzz9Pxtah99TuCSZe8259bJuKTvu+tY+297dcaBdqprwvlWh4xo61Yevsoqy3d9XFjrlupF30OzHV+8JC9K91Gft+/HrZKbFtpt/5kombLj/JxJd97UET3/LGPBO7/7QLAOuesXlM44fsOBMRmSXxLzAoJsx4AAAACBoJKwAAAIJGwgoAAICgBV/D6mvJ2trP9mXxep5R2zeY2NWONPGWw22dYevc3SY+feqbJj6ndoWJJ6eaYsecnrZ1SFWJ6tg26F+FGsI/trsqdttdj8818ZSltriwfbx9zBmfeMHEJ5ZvMXFK4sdA2Pxa0I8Nif9+H3f6TSYe+UH7mFe67Pv+qac/ZeLqDXb79WfYOWhOldddXkRSXltvalYH3tj0DhOXD+vMs5V9b/165W7vo6G82caatbX27fV2HFRm4u3eh+geEyf5kolDKt6Q334WVa2Pv2eJ6iEm3jPM5hA/uvdsE09eZOtPs535xt7f+GNCRCQhxf3FNsyAAAAACBoJKwAAAIJGwgoAAICgBV/D6tdtHe71QD33gqdij7lr0skmnvx7W08y9o8bTbzysDEmrjvK9jebX7nNxPl6qBZ7bUgp2pCxPXv/7xsLY9uMXRrvzdpT00W2bmjVzlEm/p/bLzHxPx9+r4nfX1HwNBGYfL/fk1O2Bm11t617v3H9mSYee7etIcymbZ/WU8+wfZtPKLdjNUIdYmiOTNv54LhxG2LbvPrikSYe/ax9TGqnLWrV3bYWceextnFr6u+2mvjicc/Gjnl0GZ8/A8n//D+p3H6uVM6zOYSISOaJ8SYe+/NX7AZdXSbMdthxk6irNfG2k+wcc1h6Z54zLe41FlxhBQAAQNBIWAEAABA0ElYAAAAELfga1hc7bS+xpoz9Lvfjh6yJPWbu/FUm/tepHzax+2K5ietftI9fOnOqibeOtHWy41PxGjd6JIanLWvfk9YX4z17a9fZOp/GU+33L6eX2SLU7uW2rjDtbK3S5Rf+vYlfOufG2DGHJew+GTthydfPt9PZ73Nf3DbdxG/9YZqJ61ptXeLaBXbO+Eb90/aYEq+lZlwMPH8srOu27+OsofEa1q0vTDGxPvMXu4HXF7zriAkm3vgJ+5n30MxfmHhqqjJ2zGSeumscOv7vqt919e5ZP4895pJvXmribUuPMfGER1rtA563Na7NZ9g555wTbSLTkIyPk2LHjAgAAICgkbACAAAgaCSsAAAACNqA17AW+v73X++YY+Knv2vj3SPj/eeO/9TLJj5z9BsmfqLG9mntqLV5+6UTlpt4ZNJWpKQ1/r3AOPQKjZ2NXr1z5eY8G3k1qDVv2t53qQ7b2y7VYmsTkztsnVFFo+2ttzUbr00cxn8Ti86mjK0rvG6Z/Z7vSa/bcdN0jK2T/4cPP2zi+qTtzYniUKF2Prj1sQ/Gtpn+9AoTd84/3sS7q+3nR9soG8+ZYj+v6hP0WC12+epJHz/6bhP/cPxhJn7gKTu2yhpsv/jNNo2R/zHC1rCWYm94PjoBAAAQNBJWAAAABI2EFQAAAEEb8BrWQvw+q0+mbQ1r/S3PxB6z7kXbz2zV6Jkmrl7r9c6bPcSE84astNurrUdDcehwtjdh62HxmtdRN79m4sqhtu41MbLGxNlt2+0+59uxlTyuZb/PEwPLr4Xulkxsm5u3vd/Eo560U2dbg90+9SH73eEnVr5j4hk88t0AAAPjSURBVCFq+7oOT1T36VwxsNqcfd+Hrc5zzWf6ZBNmKm2NavXKHSZur7X9oWvKdtvHez166c9bGjZk2k180xO2X/yMNRtNvPXDtr/v7BPeNPEJ5XZ/ImXiK/axU9xnDwAAgJJHwgoAAICgkbACAAAgaANew1qopmLhUFvvM/ZfbjHx5R/4bOwx03/aYeLq5etN3HqC7ZU55UJbC3JSua19LNTvEwOj0Ng5tszWEZ469/XYNq987lQTNzzSaOLMMNs/b8sCO3Z2fcD20/zajMdNPCVVETtmsdcRlbqMi/fOfbfdfv97t9dWccdMO0d8+bDnTJz2alYbkvH6MoSvJmH78bbN2xXbpjFrx0rDo7YBdFfDMBM3nWzHxkdqbB/x4Yn4HIKw+TlDvp6oi9tt39Xhr9la50z9cBM3T7ePv7rhKRNXe+OkFPMWPjkBAAAQNBJWAAAABI2EFQAAAEEjYQUAAEDQ1OVZYLBXtnHavu8cIH0pJF6xxxax37HdLqw5sdo28V5YvdXEpbooJjHmzXjl90ES4ljpi11Zu0DvN62TTfxOZ72Jj660C/jeV2Fjf5FVWm0hfbHor7FSDOMk3xyzy3Wa+LkOu3BmWMKOo/qkbf5el7TjYHjCW7VVpAb7nLKiszN22/2tx5p4RMouzByXtguJ55TbhZ7DE3ZBXlWiNBboDeY5JZ9lHXae+dTzV5g4m7V5yPSGLSb+7bR7TFypdpwUax7T2zgpzmcEAACAQYOEFQAAAEEjYQUAAEDQSrKGtZBire14rwZ7vVk+/njyx0ah8VaqY4l6M/QFc0rc/s4ppTqH+JhTrC6XMfGurK2HTnvjwl8PUa72C45KBTWsAAAAKFokrAAAAAgaCSsAAACClhroE9hffan38WuEsuKVuAzSGiLEUbOKAzFY6xCx/w7GuguUHr8mdViiYh9bYi9mWQAAAASNhBUAAABBI2EFAABA0IquhrUv/Hqy4vw2dwwEahHRF4wT7AtjAweCcVMYrxAAAACCRsIKAACAoJGwAgAAIGgkrAAAAAgaCSsAAACCRsIKAACAoJGwAgAAIGgkrAAAAAgaCSsAAACCRsIKAACAoJGwAgAAIGgkrAAAAAgaCSsAAACCRsIKAACAoJGwAgAAIGgkrAAAAAgaCSsAAACCRsIKAACAoJGwAgAAIGgkrAAAAAgaCSsAAACCRsIKAACAoJGwAgAAIGgkrAAAAAgaCSsAAACCRsIKAACAoJGwAgAAIGgkrAAAAAiaOucG+hwAAACAfeIKKwAAAIJGwgoAAICgkbACAAAgaCSsAAAACBoJKwAAAIJGwgoAAICg/X8XyBEZYTBjPAAAAABJRU5ErkJggg==",
            "text/plain": [
              "<Figure size 864x360 with 10 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          },
          "output_type": "display_data"
        },
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 864x360 with 10 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          },
          "output_type": "display_data"
        }
      ],
      "source": [
        "# Each client has different mean images, meaning each client will be nudging\n",
        "# the model in their own directions locally.\n",
        "\n",
        "for i in range(5):\n",
        "  client_dataset = emnist_train.create_tf_dataset_for_client(\n",
        "      emnist_train.client_ids[i])\n",
        "  plot_data = collections.defaultdict(list)\n",
        "  for example in client_dataset:\n",
        "    plot_data[example['label'].numpy()].append(example['pixels'].numpy())\n",
        "  f = plt.figure(i, figsize=(12, 5))\n",
        "  f.suptitle(\"Client #{}'s Mean Image Per Label\".format(i))\n",
        "  for j in range(10):\n",
        "    mean_img = np.mean(plot_data[j], 0)\n",
        "    plt.subplot(2, 5, j+1)\n",
        "    plt.imshow(mean_img.reshape((28, 28)))\n",
        "    plt.axis('off')\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Twdain3NwUx3"
      },
      "source": [
        "### Preprocessing"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "vnCA7c3xguwn"
      },
      "outputs": [],
      "source": [
        "NUM_CLIENTS = 10\n",
        "NUM_EPOCHS = 5\n",
        "BATCH_SIZE = 20\n",
        "SHUFFLE_BUFFER = 100\n",
        "PREFETCH_BUFFER = 10\n",
        "\n",
        "def preprocess(dataset):\n",
        "\n",
        "  def batch_format_fn(element):\n",
        "    \"\"\"Flatten a batch `pixels` and return the features as an `OrderedDict`.\"\"\"\n",
        "    return collections.OrderedDict(\n",
        "        x=tf.reshape(element['pixels'], [-1, 784]),\n",
        "        y=tf.reshape(element['label'], [-1, 1]))\n",
        "\n",
        "  return dataset.repeat(NUM_EPOCHS).shuffle(SHUFFLE_BUFFER, seed=1).batch(BATCH_SIZE).map(batch_format_fn).prefetch(PREFETCH_BUFFER)\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "uNi5SmI-gw1Z",
        "outputId": "a51f7da7-d73f-44a0-ca57-0528bbd5ec6e"
      },
      "outputs": [
        {
          "data": {
            "text/plain": [
              "OrderedDict([('x', array([[1., 1., 1., ..., 1., 1., 1.],\n",
              "                     [1., 1., 1., ..., 1., 1., 1.],\n",
              "                     [1., 1., 1., ..., 1., 1., 1.],\n",
              "                     ...,\n",
              "                     [1., 1., 1., ..., 1., 1., 1.],\n",
              "                     [1., 1., 1., ..., 1., 1., 1.],\n",
              "                     [1., 1., 1., ..., 1., 1., 1.]], dtype=float32)),\n",
              "             ('y', array([[2],\n",
              "                     [1],\n",
              "                     [5],\n",
              "                     [7],\n",
              "                     [1],\n",
              "                     [7],\n",
              "                     [7],\n",
              "                     [1],\n",
              "                     [4],\n",
              "                     [7],\n",
              "                     [4],\n",
              "                     [2],\n",
              "                     [2],\n",
              "                     [5],\n",
              "                     [4],\n",
              "                     [1],\n",
              "                     [1],\n",
              "                     [0],\n",
              "                     [0],\n",
              "                     [9]], dtype=int32))])"
            ]
          },
          "execution_count": 13,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "preprocessed_example_dataset = preprocess(example_dataset)\n",
        "\n",
        "sample_batch = tf.nest.map_structure(lambda x: x.numpy(), next(iter(preprocessed_example_dataset)))\n",
        "sample_batch"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "AWEEiOr1xM5_"
      },
      "source": [
        "constructs a list from users for input to a training round"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "jgcP4jg_g0VF"
      },
      "outputs": [],
      "source": [
        "def make_federated_data(client_data, client_ids):\n",
        "  return [\n",
        "      preprocess(client_data.create_tf_dataset_for_client(x))\n",
        "      for x in client_ids\n",
        "  ]\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "fqFP7NUK-TLX"
      },
      "source": [
        "- typically, among all the users, a small set of them is actually available to provide data for model training\n",
        "- for obvious reasons, it takes enough time to achieve convergence\n",
        "- with this dataset, we take a small set of users for training; also, running the algorithm for that long to converge is unnecessary\n",
        "- instead of randomly shuffling the subset for each round, we take same set(intentionally obtaining high variance i.e. overfitting)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "-DvTCKp5g4Mi",
        "outputId": "7ce5346f-0940-412c-be6f-40d0b4fad977"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Number of client datasets: 10\n",
            "First dataset: <PrefetchDataset element_spec=OrderedDict([('x', TensorSpec(shape=(None, 784), dtype=tf.float32, name=None)), ('y', TensorSpec(shape=(None, 1), dtype=tf.int32, name=None))])>\n"
          ]
        }
      ],
      "source": [
        "sample_clients = emnist_train.client_ids[0:NUM_CLIENTS]\n",
        "\n",
        "federated_train_data = make_federated_data(emnist_train, sample_clients)\n",
        "\n",
        "print('Number of client datasets: {l}'.format(l=len(federated_train_data)))\n",
        "print('First dataset: {d}'.format(d=federated_train_data[0]))\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "jlmlntbOFsk9"
      },
      "source": [
        "### Model training\n",
        "- keras is best high-level model API for TensorFlow, models of which thru `tff.learning.from_keras_model` are used in TFF\n",
        "- though, `tff.learning` provides a low-level interface, minimal functionality\n",
        "- directly implementing from scratch using `tf.keras.layers`, allows max customization"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "PTEWGvJynUoP"
      },
      "source": [
        "1. Model variables, forward pass and metrics\n",
        "    - data structure including variables like *weights*, *bias* to be trained\n",
        "    - variables holding cumulative stats *loss_sum*, *accuracy_sum* and *num_examples*"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "YoUvuCvUnGt6"
      },
      "outputs": [],
      "source": [
        "MnistVariables = collections.namedtuple(\n",
        "    'MnistVariables', 'weights bias num_examples loss_sum accuracy_sum')\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "mv8D4_a1n6Bq"
      },
      "outputs": [],
      "source": [
        "def create_mnist_variables():\n",
        "  return MnistVariables(\n",
        "      weights=tf.Variable(\n",
        "          lambda: tf.zeros(dtype=tf.float32, shape=(784, 10)),\n",
        "          name='weights',\n",
        "          trainable=True),\n",
        "      bias=tf.Variable(\n",
        "          lambda: tf.zeros(dtype=tf.float32, shape=(10)),\n",
        "          name='bias',\n",
        "          trainable=True),\n",
        "      num_examples=tf.Variable(0.0, name='num_examples', trainable=False),\n",
        "      loss_sum=tf.Variable(0.0, name='loss_sum', trainable=False),\n",
        "      accuracy_sum=tf.Variable(0.0, name='accuracy_sum', trainable=False))\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "OSUEQjEYoFA5"
      },
      "source": [
        "2.  forward pass method that computes loss, emits predictions, and updates the cumulative statistics for a single batch of input data"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "biQ_Hk4PoHiP"
      },
      "outputs": [],
      "source": [
        "def predict_on_batch(variables, x):\n",
        "  return tf.nn.softmax(tf.matmul(x, variables.weights) + variables.bias)\n",
        "\n",
        "def mnist_forward_pass(variables, batch):\n",
        "  y = predict_on_batch(variables, batch['x'])\n",
        "  predictions = tf.cast(tf.argmax(y, 1), tf.int32)\n",
        "\n",
        "  flat_labels = tf.reshape(batch['y'], [-1])\n",
        "  loss = -tf.reduce_mean(\n",
        "      tf.reduce_sum(tf.one_hot(flat_labels, 10) * tf.math.log(y), axis=[1]))\n",
        "  accuracy = tf.reduce_mean(\n",
        "      tf.cast(tf.equal(predictions, flat_labels), tf.float32))\n",
        "\n",
        "  num_examples = tf.cast(tf.size(batch['y']), tf.float32)\n",
        "\n",
        "  variables.num_examples.assign_add(num_examples)\n",
        "  variables.loss_sum.assign_add(loss * num_examples)\n",
        "  variables.accuracy_sum.assign_add(accuracy * num_examples)\n",
        "\n",
        "  return loss, predictions\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "HvFv1DVPof4m"
      },
      "outputs": [],
      "source": [
        "## local metrics\n",
        "\n",
        "# returns the unfinalized metric values eligible to be aggregated to the server in a federated learning process\n",
        "def get_local_unfinalized_metrics(variables):\n",
        "  return collections.OrderedDict(\n",
        "      num_examples=[variables.num_examples],\n",
        "      loss=[variables.loss_sum, variables.num_examples],\n",
        "      accuracy=[variables.accuracy_sum, variables.num_examples])\n",
        "\n",
        "\n",
        "# returns an OrderedDict of tf.functions with the same keys\n",
        "# takes in the metric's unfinalized values and computes the finalized metric\n",
        "def get_metric_finalizers():\n",
        "  return collections.OrderedDict(\n",
        "      num_examples=tf.function(func=lambda x: x[0]),\n",
        "      loss=tf.function(func=lambda x: x[0] / x[1]),\n",
        "      accuracy=tf.function(func=lambda x: x[0] / x[1]))\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "yF7IacS-pOms"
      },
      "source": [
        "in `tff.learning.build_federated_averaging_process` API, the default value for metrics_aggregator is `tff.learning.metrics.sum_then_finalize`, which first sums the unfinalized metrics from `CLIENTS`, and then applies the metric finalizers at `SERVER`."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "QqzVdXwYpg-S"
      },
      "source": [
        "3. Constructing instance of `tff.learning.Model`\n",
        "    - all states must be captured as TensorFlow variables, as TFF doesnt use python at runtime\n",
        "    - model with declared input format\n",
        "    - wrap all logics as `tf.learning` to ensure serialization"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "MlNAhMaapa79"
      },
      "outputs": [],
      "source": [
        "from typing import Callable, List, OrderedDict\n",
        "\n",
        "class MnistModel(tff.learning.Model):\n",
        "\n",
        "  def __init__(self):\n",
        "    self._variables = create_mnist_variables()\n",
        "\n",
        "  @property\n",
        "  def trainable_variables(self):\n",
        "    return [self._variables.weights, self._variables.bias]\n",
        "\n",
        "  @property\n",
        "  def non_trainable_variables(self):\n",
        "    return []\n",
        "\n",
        "  @property\n",
        "  def local_variables(self):\n",
        "    return [\n",
        "        self._variables.num_examples, self._variables.loss_sum,\n",
        "        self._variables.accuracy_sum\n",
        "    ]\n",
        "\n",
        "  @property\n",
        "  def input_spec(self):\n",
        "    return collections.OrderedDict(\n",
        "        x=tf.TensorSpec([None, 784], tf.float32),\n",
        "        y=tf.TensorSpec([None, 1], tf.int32))\n",
        "\n",
        "  @tf.function\n",
        "  def predict_on_batch(self, x, training=True):\n",
        "    del training\n",
        "    return predict_on_batch(self._variables, x)\n",
        "\n",
        "  @tf.function\n",
        "  def forward_pass(self, batch, training=True):\n",
        "    del training\n",
        "    loss, predictions = mnist_forward_pass(self._variables, batch)\n",
        "    num_exmaples = tf.shape(batch['x'])[0]\n",
        "    return tff.learning.BatchOutput(\n",
        "        loss=loss, predictions=predictions, num_examples=num_exmaples)\n",
        "\n",
        "  @tf.function\n",
        "  def report_local_unfinalized_metrics(\n",
        "      self) -> OrderedDict[str, List[tf.Tensor]]:\n",
        "    \"\"\"Creates an `OrderedDict` of metric names to unfinalized values.\"\"\"\n",
        "    return get_local_unfinalized_metrics(self._variables)\n",
        "\n",
        "  def metric_finalizers(\n",
        "      self) -> OrderedDict[str, Callable[[List[tf.Tensor]], tf.Tensor]]:\n",
        "    \"\"\"Creates an `OrderedDict` of metric names to finalizers.\"\"\"\n",
        "    return get_metric_finalizers()\n",
        "\n",
        "  @tf.function\n",
        "  def reset_metrics(self):\n",
        "    \"\"\"Resets metrics variables to initial value.\"\"\"\n",
        "    for var in self.local_variables:\n",
        "      var.assign(tf.zeros_like(var))\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "H1dztzkbrkhj"
      },
      "source": [
        "4. Simulating federated training\n",
        "    - use constructor of new model\n",
        "    - use two federated computations iteratively"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 319
        },
        "id": "a6LepGxDryeJ",
        "outputId": "07bafd80-c283-487e-8858-f01cf92b9415"
      },
      "outputs": [
        {
          "ename": "TypeError",
          "evalue": "ignored",
          "output_type": "error",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mTypeError\u001b[0m                                 Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-21-c8680a16cc75>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0miterative_process\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtff\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlearning\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbuild_federated_averaging_process\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mMnistModel\u001b[0m\u001b[0;34m,\u001b[0m  \u001b[0mclient_optimizer_fn\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m(\u001b[0m\u001b[0;32mlambda\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mtf\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mkeras\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0moptimizers\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mSGD\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mlearning_rate\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m0.02\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/tensorflow_federated/python/learning/federated_averaging.py\u001b[0m in \u001b[0;36mbuild_federated_averaging_process\u001b[0;34m(model_fn, client_optimizer_fn, server_optimizer_fn, client_weighting, broadcast_process, model_update_aggregation_factory, metrics_aggregator, use_experimental_simulation_loop)\u001b[0m\n\u001b[1;32m    289\u001b[0m       \u001b[0mbroadcast_process\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mbroadcast_process\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    290\u001b[0m       \u001b[0mmodel_update_aggregation_factory\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mmodel_update_aggregation_factory\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 291\u001b[0;31m       metrics_aggregator=metrics_aggregator)\n\u001b[0m\u001b[1;32m    292\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    293\u001b[0m   \u001b[0mserver_state_type\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0miter_proc\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mstate_type\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmember\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/tensorflow_federated/python/learning/framework/optimizer_utils.py\u001b[0m in \u001b[0;36mbuild_model_delta_optimizer_process\u001b[0;34m(model_fn, model_to_client_delta_fn, server_optimizer_fn, broadcast_process, model_update_aggregation_factory, metrics_aggregator)\u001b[0m\n\u001b[1;32m    567\u001b[0m   \u001b[0;34m@\u001b[0m\u001b[0mcomputations\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtf_computation\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    568\u001b[0m   def model_and_optimizer_init_fn(\n\u001b[0;32m--> 569\u001b[0;31m   ) -> Tuple[model_utils.ModelWeights, List[tf.Variable]]:\n\u001b[0m\u001b[1;32m    570\u001b[0m     \u001b[0;34m\"\"\"Returns initial model weights and state of the global optimizer.\"\"\"\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    571\u001b[0m     \u001b[0mmodel_variables\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mmodel_utils\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mModelWeights\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfrom_model\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodel_fn\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/tensorflow_federated/python/core/impl/wrappers/computation_wrapper.py\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, tff_internal_types, *args)\u001b[0m\n\u001b[1;32m    493\u001b[0m       \u001b[0mparameter_type\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0m_parameter_type\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mparameters\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mparameter_types\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    494\u001b[0m       wrapped_func = self._strategy(\n\u001b[0;32m--> 495\u001b[0;31m           fn_to_wrap, fn_name, parameter_type, unpack=None)\n\u001b[0m\u001b[1;32m    496\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    497\u001b[0m     \u001b[0;31m# Copy the __doc__ attribute with the documentation in triple-quotes from\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/tensorflow_federated/python/core/impl/wrappers/computation_wrapper.py\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, fn_to_wrap, fn_name, parameter_type, unpack)\u001b[0m\n\u001b[1;32m    220\u001b[0m     \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    221\u001b[0m       \u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mkwargs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0munpack_arguments_fn\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mpacked_args\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 222\u001b[0;31m       \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mfn_to_wrap\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    223\u001b[0m       \u001b[0;32mif\u001b[0m \u001b[0mresult\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    224\u001b[0m         \u001b[0;32mraise\u001b[0m \u001b[0mComputationReturnedNoneError\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mfn_to_wrap\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/tensorflow_federated/python/learning/framework/optimizer_utils.py\u001b[0m in \u001b[0;36mmodel_and_optimizer_init_fn\u001b[0;34m()\u001b[0m\n\u001b[1;32m    569\u001b[0m   ) -> Tuple[model_utils.ModelWeights, List[tf.Variable]]:\n\u001b[1;32m    570\u001b[0m     \u001b[0;34m\"\"\"Returns initial model weights and state of the global optimizer.\"\"\"\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 571\u001b[0;31m     \u001b[0mmodel_variables\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mmodel_utils\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mModelWeights\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfrom_model\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodel_fn\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    572\u001b[0m     optimizer = keras_optimizer.build_or_verify_tff_optimizer(\n\u001b[1;32m    573\u001b[0m         \u001b[0mserver_optimizer_fn\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;31mTypeError\u001b[0m: Can't instantiate abstract class MnistModel with abstract methods federated_output_computation, report_local_outputs"
          ]
        }
      ],
      "source": [
        "iterative_process = tff.learning.build_federated_averaging_process(MnistModel,  client_optimizer_fn = (lambda: tf.keras.optimizers.SGD(learning_rate=0.02)))\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "so6aTUCMr1Aq"
      },
      "outputs": [],
      "source": [
        "state = iterative_process.initialize()\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "zmTKrzxEr3Ja"
      },
      "outputs": [],
      "source": [
        "state, metrics = iterative_process.next(state, federated_train_data)\n",
        "print('round  1, metrics={}'.format(metrics))\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "87Jn03QBsD5Y"
      },
      "outputs": [],
      "source": [
        "for round_num in range(2, 11):\n",
        "  state, metrics = iterative_process.next(state, federated_train_data)\n",
        "  print('round {:2d}, metrics={}'.format(round_num, metrics))\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "2tazlrobgM-n"
      },
      "source": [
        "### Evaluation\n",
        "- average metrics over all batches of data across all clients concerns about overfitting the user data\n",
        "- as training is done locally over single and similar batch of data, thus local accuracy will approach to 1\n",
        "- this implies that the training is just 'progressing'\n",
        "\n",
        "\n",
        "---\n",
        "\n",
        "- to perform proper eval on federated data, another federated computation is created using `tff.learning.build_federated_evaluation`"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "ujlLjtCJwYjv"
      },
      "outputs": [],
      "source": [
        "evaluation = tff.learning.build_federated_evaluation(MnistModel)\n",
        "\n",
        "# signature type\n",
        "print(evaluation.type_signature.formatted_representation())\n",
        "    "
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "eF7k4eEyxJ3R"
      },
      "source": [
        "- not returning server state, its stateless as eval doesnt modify model\n",
        "- eval needs only model, not anything else from server\n",
        "`SERVER_MODEL, FEDERATED_DATA -> TRAINING_METRICS`"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "JayMNyD5xfgD"
      },
      "outputs": [],
      "source": [
        "# invoking evaluations\n",
        "train_metrics = evaluation(state.model, federated_train_data)\n",
        "str(train_metrics)\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "fs04XkOLxrHg"
      },
      "outputs": [],
      "source": [
        "# re-running on compiled sample of federated data from same sample of users, but distinct held-out dataset\n",
        "federated_test_data = make_federated_data(emnist_test, sample_clients)\n",
        "len(federated_test_data), federated_test_data[0]\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "KYECDo_Rx5oX"
      },
      "outputs": [],
      "source": [
        "test_metrics = evaluation(state.model, federated_test_data)\n",
        "str(test_metrics)\n"
      ]
    }
  ],
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "authorship_tag": "ABX9TyM7hvTOMmHZl7UkcfQf/D3t",
      "collapsed_sections": [
        "JlOZ7EcY2DZ_",
        "_hU4hNzP2QUF",
        "Twdain3NwUx3"
      ],
      "include_colab_link": true,
      "provenance": []
    },
    "kernelspec": {
      "display_name": "Python 3.10.4 64-bit",
      "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": "3.10.4"
    },
    "vscode": {
      "interpreter": {
        "hash": "26de051ba29f2982a8de78e945f0abaf191376122a1563185a90213a26c5da77"
      }
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}