ML Pipeline for Airfoil noise prediction using PySpark.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "MnsXVLuPDK0h"
      },
      "source": [
        "<p style=\"text-align:center\">\n",
        "    <a href=\"https://github.com/salimt\">\n",
        "    <img src=\"https://i.imgur.com/7bgtWRh.jpeg\" width=\"200\" alt=\"Skills Network Logo\">\n",
        "    </a>\n",
        "</p>\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "144hyHyQDK0x"
      },
      "source": [
        "## ML Pipeline for Airfoil noise prediction using PySpark\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "nsf2-nQLDK0y"
      },
      "source": [
        "## Scenario\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "dr_2PsEMDK00"
      },
      "source": [
        "As a data engineer at an aeronautics consulting firm renowned for its efficient airfoil design for planes and sports cars, my role involves facilitating data preparation and constructing machine learning pipelines. In this project, we'll be working with a modified version of the NASA Airfoil Self Noise dataset. Our objective is to clean the dataset by removing duplicate rows and eliminating any entries with null values. Subsequently, we'll develop a machine learning pipeline to create a predictive model that estimates sound levels based on all other available columns. Following model evaluation, we'll finalize the process by persisting the model for future use."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "0JvWz6KGDK02"
      },
      "source": [
        "## Objectives\n",
        "\n",
        "### Part 1: Execute ETL Operations\n",
        "- Load a CSV dataset.\n",
        "- Eliminate any duplicate entries.\n",
        "- Remove rows containing null values.\n",
        "- Apply necessary transformations.\n",
        "- Save the cleansed data in the Parquet format.\n",
        "\n",
        "### Part 2: Construct a Machine Learning Pipeline\n",
        "- Develop a machine learning pipeline for prediction.\n",
        "\n",
        "### Part 3: Assess the Model\n",
        "- Assess the model's performance using appropriate metrics.\n",
        "\n",
        "### Part 4: Preserve the Model\n",
        "- Save the model for future deployment purposes.\n",
        "- Load and validate the stored model.\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "c5_9UKC6DK02"
      },
      "source": [
        "## Datasets\n",
        "\n",
        " - The original dataset can be found here NASA airfoil self noise dataset. https://archive.ics.uci.edu/dataset/291/airfoil+self+noise"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "GQ5hBgzhDK03"
      },
      "source": [
        "Diagram of an airfoil. - For informational purpose\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "nq9lkUMxDK04"
      },
      "source": [
        "![Airfoil with flow](https://cf-courses-data.s3.us.cloud-object-storage.appdomain.cloud/IBMSkillsNetwork-BD0231EN-Coursera/images/Airfoil_with_flow.png)\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "24ZdTL8NDK06"
      },
      "source": [
        "Diagram showing the Angle of attack. - For informational purpose\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "fcWNr-RiDK07"
      },
      "source": [
        "![Airfoil angle of attack](https://cf-courses-data.s3.us.cloud-object-storage.appdomain.cloud/IBMSkillsNetwork-BD0231EN-Coursera/images/Airfoil_angle_of_attack.jpg)\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "F_dMyjd3DK09"
      },
      "source": [
        "## Setup\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "NYY5bofuDK0-"
      },
      "source": [
        "*   [`PySpark`](https://spark.apache.org/docs/latest/api/python/index.html?utm_medium=Exinfluencer&utm_source=Exinfluencer&utm_content=000026UJ&utm_term=10006555&utm_id=NA-SkillsNetwork-Channel-SkillsNetworkCoursesIBMSkillsNetworkBD0231ENCoursera2789-2023-01-01) for connecting to the Spark Cluster\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Q-Ah6WcjDK0_"
      },
      "source": [
        "### Installing Required Libraries\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 3,
      "metadata": {
        "tags": [],
        "id": "6LzuyyS9DK0_"
      },
      "outputs": [],
      "source": [
        "!pip install pyspark==3.1.2 -q\n",
        "!pip install findspark -q"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "d-Bacl7NDK1C"
      },
      "source": [
        "### Importing Required Libraries"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 4,
      "metadata": {
        "tags": [],
        "id": "y-9LkzGJDK1C"
      },
      "outputs": [],
      "source": [
        "def warn(*args, **kwargs):\n",
        "    pass\n",
        "import warnings\n",
        "warnings.warn = warn\n",
        "warnings.filterwarnings('ignore')\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "from pyspark.sql import SparkSession\n",
        "import findspark\n",
        "\n",
        "findspark.init()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "inzGkrIwDK1D"
      },
      "source": [
        "## Part 1 - Perform ETL activity\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "tags": [],
        "id": "qV4Z9j90DK1D"
      },
      "source": [
        "### Create a spark session\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 5,
      "metadata": {
        "tags": [],
        "id": "14TasGzJDK1E"
      },
      "outputs": [],
      "source": [
        "spark = SparkSession.builder.appName(\"ETL using Spark\").getOrCreate()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "U7A8lQqtDK1G"
      },
      "source": [
        "### Load the csv file into a dataframe\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "swDaGkHfDK1G"
      },
      "source": [
        "Download the data file."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 6,
      "metadata": {
        "tags": [],
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "rqpZrky2DK1G",
        "outputId": "db024266-826d-4e72-9e35-dc0a0bc84ff5"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "--2024-02-05 16:16:16--  https://cf-courses-data.s3.us.cloud-object-storage.appdomain.cloud/IBMSkillsNetwork-BD0231EN-Coursera/datasets/NASA_airfoil_noise_raw.csv\n",
            "Resolving cf-courses-data.s3.us.cloud-object-storage.appdomain.cloud (cf-courses-data.s3.us.cloud-object-storage.appdomain.cloud)... 169.63.118.104\n",
            "Connecting to cf-courses-data.s3.us.cloud-object-storage.appdomain.cloud (cf-courses-data.s3.us.cloud-object-storage.appdomain.cloud)|169.63.118.104|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 60682 (59K) [text/csv]\n",
            "Saving to: ‘NASA_airfoil_noise_raw.csv’\n",
            "\n",
            "\r          NASA_airf   0%[                    ]       0  --.-KB/s               \rNASA_airfoil_noise_ 100%[===================>]  59.26K  --.-KB/s    in 0.03s   \n",
            "\n",
            "2024-02-05 16:16:16 (2.27 MB/s) - ‘NASA_airfoil_noise_raw.csv’ saved [60682/60682]\n",
            "\n"
          ]
        }
      ],
      "source": [
        "!wget https://cf-courses-data.s3.us.cloud-object-storage.appdomain.cloud/IBMSkillsNetwork-BD0231EN-Coursera/datasets/NASA_airfoil_noise_raw.csv"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xS4jRtoVDK1H"
      },
      "source": [
        "Load the dataset into the spark dataframe\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 7,
      "metadata": {
        "tags": [],
        "id": "z-eUnC4mDK1H"
      },
      "outputs": [],
      "source": [
        "df = spark.read.csv(\"NASA_airfoil_noise_raw.csv\", header=True, inferSchema=True)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 8,
      "metadata": {
        "tags": [],
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "dt28sO0-DK1H",
        "outputId": "c3aaf0b5-8694-474a-911e-d4ede92a9717"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[Row(Frequency=800, AngleOfAttack=0.0, ChordLength=0.3048, FreeStreamVelocity=71.3, SuctionSideDisplacement=0.00266337, SoundLevel=126.201),\n",
              " Row(Frequency=1000, AngleOfAttack=0.0, ChordLength=0.3048, FreeStreamVelocity=71.3, SuctionSideDisplacement=0.00266337, SoundLevel=125.201),\n",
              " Row(Frequency=1250, AngleOfAttack=0.0, ChordLength=0.3048, FreeStreamVelocity=71.3, SuctionSideDisplacement=0.00266337, SoundLevel=125.951),\n",
              " Row(Frequency=1600, AngleOfAttack=0.0, ChordLength=0.3048, FreeStreamVelocity=71.3, SuctionSideDisplacement=0.00266337, SoundLevel=127.591),\n",
              " Row(Frequency=2000, AngleOfAttack=0.0, ChordLength=0.3048, FreeStreamVelocity=71.3, SuctionSideDisplacement=0.00266337, SoundLevel=127.461)]"
            ]
          },
          "metadata": {},
          "execution_count": 8
        }
      ],
      "source": [
        "df.head(5)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "NWUB5mhRDK1I"
      },
      "source": [
        "## Data Preprocessing\n",
        "\n",
        "#### Dropping Duplicates\n",
        "Duplicate rows are removed from the dataset to ensure each observation is unique. This helps prevent biases in analysis caused by redundant data.\n",
        "\n",
        "#### Handling Missing Values\n",
        "Rows containing null values are dropped from the dataset to maintain data integrity. This ensures that analyses are based on complete and reliable information.\n",
        "\n",
        "#### Column Renaming\n",
        "The column named \"SoundLevel\" is renamed to \"SoundLevelDecibels\" for clarity and consistency in column naming conventions. This step enhances the interpretability of the dataset and facilitates downstream analysis."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 9,
      "metadata": {
        "tags": [],
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "QKUodwd1DK1I",
        "outputId": "6c7509e7-b29c-4690-e8a7-d1dcf974385f"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "1522\n"
          ]
        }
      ],
      "source": [
        "rowcount1 = df.count()\n",
        "print(rowcount1)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "fEfm8aLXDK1I"
      },
      "source": [
        "### Drop all the duplicate rows\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 10,
      "metadata": {
        "tags": [],
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "kB6vqHwcDK1J",
        "outputId": "e3ae1103-140d-4d8b-a2fe-f1ee4b62bd5c"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "1503\n"
          ]
        }
      ],
      "source": [
        "df = df.dropDuplicates()\n",
        "rowcount2 = df.count()\n",
        "print(rowcount2)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "8p2FKi29DK1J"
      },
      "source": [
        "### Drop all the rows that contain null values\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 11,
      "metadata": {
        "tags": [],
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "SGSZyJKNDK1J",
        "outputId": "7ce255c3-eea6-4f9e-b485-b831c4fd6d39"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "1499\n"
          ]
        }
      ],
      "source": [
        "df=df.dropna()\n",
        "rowcount3 = df.count()\n",
        "print(rowcount3)\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "6J2d4dQKDK1K"
      },
      "source": [
        "### Rename the column \"SoundLevel\" to \"SoundLevelDecibels\"\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 12,
      "metadata": {
        "tags": [],
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "V-6gnxX2DK1K",
        "outputId": "83d6eba2-79a2-43f1-af8d-484bd9f47fc7"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "['Frequency',\n",
              " 'AngleOfAttack',\n",
              " 'ChordLength',\n",
              " 'FreeStreamVelocity',\n",
              " 'SuctionSideDisplacement',\n",
              " 'SoundLevelDecibels']"
            ]
          },
          "metadata": {},
          "execution_count": 12
        }
      ],
      "source": [
        "df = df.withColumnRenamed(\"SoundLevel\",\"SoundLevelDecibels\")\n",
        "df.columns"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "lv4DjdLTDK1K"
      },
      "source": [
        "### Save the dataframe in parquet formant\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 13,
      "metadata": {
        "tags": [],
        "id": "RyK2CThwDK1K"
      },
      "outputs": [],
      "source": [
        "df.write.mode(\"overwrite\").parquet(\"NASA_airfoil_noise_cleaned.parquet\")"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "lyzvBfxuDK1L"
      },
      "source": [
        "#### Part 1 - Evaluation\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 14,
      "metadata": {
        "tags": [],
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "KZ-oERZnDK1M",
        "outputId": "35629771-c195-44f3-8214-cfc31110792c"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Part 1 - Evaluation\n",
            "Total rows =  1522\n",
            "Total rows after dropping duplicate rows =  1503\n",
            "Total rows after dropping duplicate rows and rows with null values =  1499\n",
            "New column name =  SoundLevelDecibels\n",
            "NASA_airfoil_noise_cleaned.parquet exists : True\n"
          ]
        }
      ],
      "source": [
        "print(\"Part 1 - Evaluation\")\n",
        "\n",
        "print(\"Total rows = \", rowcount1)\n",
        "print(\"Total rows after dropping duplicate rows = \", rowcount2)\n",
        "print(\"Total rows after dropping duplicate rows and rows with null values = \", rowcount3)\n",
        "print(\"New column name = \", df.columns[-1])\n",
        "\n",
        "import os\n",
        "\n",
        "print(\"NASA_airfoil_noise_cleaned.parquet exists :\", os.path.isdir(\"NASA_airfoil_noise_cleaned.parquet\"))"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xkkSK7gxDK1N"
      },
      "source": [
        "## Part - 2 Create a  Machine Learning Pipeline\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "74yKUjtbDK1O"
      },
      "source": [
        "### Load data from parquet into a dataframe\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 15,
      "metadata": {
        "tags": [],
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "30lk6zn7DK1U",
        "outputId": "62cd6f6d-0904-4ef4-b934-f436b259dcf2"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "1499\n"
          ]
        }
      ],
      "source": [
        "df = spark.read.parquet(\"NASA_airfoil_noise_cleaned.parquet\")\n",
        "rowcount4 = df.count()\n",
        "print(rowcount4)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "pZIssobhDK1U"
      },
      "source": [
        "### Define the VectorAssembler pipeline stage\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "tags": [],
        "id": "uiuz1oABDK1V"
      },
      "source": [
        "By converting features into a unified format, the VectorAssembler enhances algorithm compatibility and leverages PySpark's distributed computing for efficient processing of large datasets, ensuring scalability and performance in machine learning workflows."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 16,
      "metadata": {
        "tags": [],
        "id": "--f-NhWUDK1V"
      },
      "outputs": [],
      "source": [
        "\n",
        "from pyspark.ml.feature import VectorAssembler\n",
        "\n",
        "assembler = VectorAssembler(inputCols=['Frequency',\n",
        " 'AngleOfAttack',\n",
        " 'ChordLength',\n",
        " 'FreeStreamVelocity',\n",
        " 'SuctionSideDisplacement'], outputCol=\"features\")"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ctIcw-hZDK1W"
      },
      "source": [
        "### Define the StandardScaler pipeline stage\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "me3aIQwBDK1W"
      },
      "source": [
        "By applying StandardScaler, potential issues related to feature magnitudes and variations are mitigated, leading to more stable and reliable model performance across diverse datasets."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 17,
      "metadata": {
        "tags": [],
        "id": "bAhwv6ZCDK1X"
      },
      "outputs": [],
      "source": [
        "from pyspark.ml.feature import StandardScaler\n",
        "scaler = StandardScaler(inputCol=\"features\", outputCol=\"scaledFeatures\")"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "DH4k0OkwDK1Y"
      },
      "source": [
        "### Define the LinearRegression pipeline stage\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "5LT6NDbRDK1Y"
      },
      "source": [
        "Linear Regression in this context, we aim to capture the linear association between the scaled features and the sound level decibels, enabling us to make predictions based on this relationship."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 18,
      "metadata": {
        "tags": [],
        "id": "-OWonLBXDK1Z"
      },
      "outputs": [],
      "source": [
        "from pyspark.ml.regression import LinearRegression\n",
        "lr = LinearRegression(featuresCol=\"scaledFeatures\", labelCol=\"SoundLevelDecibels\")"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "tags": [],
        "id": "yWkZRL0WDK1Z"
      },
      "source": [
        "### Define the Lasso pipeline stage\n"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Lasso regression, a variant of linear regression, introduces regularization by adding a penalty term that promotes sparsity in the model coefficients. This encourages simpler models with fewer non-zero coefficients, effectively selecting a subset of features and reducing overfitting.\n"
      ],
      "metadata": {
        "id": "l-JddL6zEPXH"
      }
    },
    {
      "cell_type": "code",
      "execution_count": 19,
      "metadata": {
        "tags": [],
        "id": "-E2rxJrYDK1a"
      },
      "outputs": [],
      "source": [
        "lasso = LinearRegression(featuresCol=\"scaledFeatures\", labelCol=\"SoundLevelDecibels\", elasticNetParam=1.0, regParam=0.1)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "tags": [],
        "id": "i0tpUFzbDK1b"
      },
      "source": [
        "### Define the ElasticNet  pipeline stage\n"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "ElasticNet regression is a generalized linear regression model that combines both L1 (Lasso) and L2 (Ridge) regularization penalties. It balances between feature selection (Lasso) and coefficient shrinkage (Ridge), offering a versatile approach to handle multicollinearity and overfitting in regression tasks.\n"
      ],
      "metadata": {
        "id": "HBQ-lcGvEXQI"
      }
    },
    {
      "cell_type": "code",
      "execution_count": 21,
      "metadata": {
        "tags": [],
        "id": "bB55k_EaDK1b"
      },
      "outputs": [],
      "source": [
        "from pyspark.ml.regression import GeneralizedLinearRegression\n",
        "\n",
        "elastic_net = GeneralizedLinearRegression(featuresCol=\"scaledFeatures\", labelCol=\"SoundLevelDecibels\", family=\"gaussian\", link=\"identity\", regParam=0.1)\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "dM9nPgzvDK1c"
      },
      "source": [
        "### Build the pipeline\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "IVh--BJkDK1c"
      },
      "source": [
        "Build a pipeline using the above three stages\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 22,
      "metadata": {
        "tags": [],
        "id": "MvfXvQBxDK1e"
      },
      "outputs": [],
      "source": [
        "from pyspark.ml import Pipeline\n",
        "pipeline = Pipeline(stages=[assembler, scaler, lr])\n",
        "lasso_pipeline = Pipeline(stages=[assembler, scaler, lasso])\n",
        "elastic_pipeline = Pipeline(stages=[assembler, scaler, elastic_net])"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "gU0FGEzTDK1f"
      },
      "source": [
        "### Split the data\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "qeYZsmINDK1g"
      },
      "source": [
        "The larger subset (70%) is designated for training the machine learning model, while the smaller subset (30%) is reserved for evaluating the model's performance. The seed=42 parameter ensures reproducibility by setting the random seed for the split."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 23,
      "metadata": {
        "tags": [],
        "id": "W7FZrS5IDK1g"
      },
      "outputs": [],
      "source": [
        "(trainingData, testData) = df.randomSplit([0.7, 0.3], seed=42)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "mzy5B3FvDK1g"
      },
      "source": [
        "### Fit the pipeline\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "F2ygdAWqDK1g"
      },
      "source": [
        "This process involves applying any data preprocessing steps, such as feature engineering or scaling, followed by training the model using the specified algorithm. Once fitted, the model encapsulates the learned patterns from the training data and is ready for evaluation and prediction tasks."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 24,
      "metadata": {
        "tags": [],
        "id": "UUh7-bFZDK1h"
      },
      "outputs": [],
      "source": [
        "model = pipeline.fit(trainingData)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 25,
      "metadata": {
        "tags": [],
        "id": "GYl0TJb8DK1i"
      },
      "outputs": [],
      "source": [
        "model2 = lasso_pipeline.fit(trainingData)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 26,
      "metadata": {
        "tags": [],
        "id": "Y798yhroDK1i"
      },
      "outputs": [],
      "source": [
        "model3 = elastic_pipeline.fit(trainingData)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "3CncqKrADK1j"
      },
      "source": [
        "#### Part 2 - Evaluation\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 27,
      "metadata": {
        "tags": [],
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "4ErW4MrvDK1j",
        "outputId": "30a2f0cd-061d-4ed6-d860-a7ce9c721e71"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Part 2 - Evaluation\n",
            "Total rows =  1499\n",
            "Linear Regression\n",
            "Pipeline Stage 1 =  VectorAssembler\n",
            "Pipeline Stage 2 =  StandardScaler\n",
            "Pipeline Stage 3 =  LinearRegression\n",
            "Label column =  SoundLevelDecibels\n",
            "\n",
            "\n",
            "Lasso\n",
            "Pipeline Stage 1 =  VectorAssembler\n",
            "Pipeline Stage 2 =  StandardScaler\n",
            "Pipeline Stage 3 =  LinearRegression\n",
            "Label column =  SoundLevelDecibels\n",
            "\n",
            "\n",
            "ElasticNet\n",
            "Pipeline Stage 1 =  VectorAssembler\n",
            "Pipeline Stage 2 =  StandardScaler\n",
            "Pipeline Stage 3 =  GeneralizedLinearRegression\n",
            "Label column =  SoundLevelDecibels\n"
          ]
        }
      ],
      "source": [
        "print(\"Part 2 - Evaluation\")\n",
        "print(\"Total rows = \", rowcount4)\n",
        "ps = [str(x).split(\"_\")[0] for x in pipeline.getStages()]\n",
        "print(\"Linear Regression\")\n",
        "print(\"Pipeline Stage 1 = \", ps[0])\n",
        "print(\"Pipeline Stage 2 = \", ps[1])\n",
        "print(\"Pipeline Stage 3 = \", ps[2])\n",
        "\n",
        "print(\"Label column = \", lr.getLabelCol())\n",
        "\n",
        "print(\"\\n\")\n",
        "\n",
        "ps = [str(x).split(\"_\")[0] for x in lasso_pipeline.getStages()]\n",
        "print(\"Lasso\")\n",
        "print(\"Pipeline Stage 1 = \", ps[0])\n",
        "print(\"Pipeline Stage 2 = \", ps[1])\n",
        "print(\"Pipeline Stage 3 = \", ps[2])\n",
        "\n",
        "print(\"Label column = \", lr.getLabelCol())\n",
        "\n",
        "print(\"\\n\")\n",
        "\n",
        "ps = [str(x).split(\"_\")[0] for x in elastic_pipeline.getStages()]\n",
        "print(\"ElasticNet\")\n",
        "print(\"Pipeline Stage 1 = \", ps[0])\n",
        "print(\"Pipeline Stage 2 = \", ps[1])\n",
        "print(\"Pipeline Stage 3 = \", ps[2])\n",
        "\n",
        "print(\"Label column = \", lr.getLabelCol())"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "KDncLyrlDK1k"
      },
      "source": [
        "## Part 3 - Evaluate the Model\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "r4S1FOiJDK1l"
      },
      "source": [
        "### Predict using the model\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "pudCj4d2DK1l"
      },
      "source": [
        "The resulting predictions are stored in the \"predictions\" DataFrame, which can then be used to evaluate the model's performance against the true labels or values in the test dataset. This step is crucial for assessing how well the trained model generalizes to new, unseen data and for validating its effectiveness in real-world scenarios."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 28,
      "metadata": {
        "tags": [],
        "id": "G3CIYatDDK1m"
      },
      "outputs": [],
      "source": [
        "predictions = model.transform(testData)\n",
        "predictions2 = model2.transform(testData)\n",
        "predictions3 = model3.transform(testData)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "nhyT7BUpDK1m"
      },
      "source": [
        "### MSE\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "CvW4kHGQDK1n"
      },
      "source": [
        "The calculated MSE value quantifies the average squared difference between the predicted and actual values, providing a measure of the model's predictive accuracy. A lower MSE indicates better model performance, with values closer to zero indicating a more precise model."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 29,
      "metadata": {
        "tags": [],
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "PgO9-fcADK1n",
        "outputId": "95e13204-b8f3-44eb-ea40-97030d9ea4c4"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Linear MSE = 25.005633149560463\n"
          ]
        }
      ],
      "source": [
        "from pyspark.ml.evaluation import RegressionEvaluator\n",
        "\n",
        "evaluator = RegressionEvaluator(labelCol=\"SoundLevelDecibels\", predictionCol=\"prediction\", metricName=\"mse\")\n",
        "mse = evaluator.evaluate(predictions)\n",
        "print(\"Linear MSE =\", mse)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 30,
      "metadata": {
        "tags": [],
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "DfVsZUNLDK1n",
        "outputId": "ff2e9d55-a634-4260-d802-2edaf04ae6f9"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Lasso MSE = 25.337662351914624\n"
          ]
        }
      ],
      "source": [
        "evaluator = RegressionEvaluator(labelCol=\"SoundLevelDecibels\", predictionCol=\"prediction\", metricName=\"mse\")\n",
        "mse2 = evaluator.evaluate(predictions2)\n",
        "print(\"Lasso MSE =\", mse2)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 31,
      "metadata": {
        "tags": [],
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "XSfzdkpzDK1n",
        "outputId": "c4d2a2e8-bc6b-4ae4-c1d1-99cbdc4c8779"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "ElasticNet MSE = 25.123461720551557\n"
          ]
        }
      ],
      "source": [
        "evaluator = RegressionEvaluator(labelCol=\"SoundLevelDecibels\", predictionCol=\"prediction\", metricName=\"mse\")\n",
        "mse3 = evaluator.evaluate(predictions3)\n",
        "print(\"ElasticNet MSE =\", mse3)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "KQ8LPVXCDK1o"
      },
      "source": [
        "### MAE\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "AFmMf-2LDK1o"
      },
      "source": [
        "MAE value quantifies the average absolute difference between the predicted and actual values, providing another measure of the model's predictive accuracy. A lower MAE indicates better model performance, with values closer to zero indicating a more accurate model."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 32,
      "metadata": {
        "tags": [],
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "e9yBnrw1DK1o",
        "outputId": "c34508f1-8e95-4671-ddab-45623b2c62f0"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Linear MAE = 3.870800827817513\n"
          ]
        }
      ],
      "source": [
        "evaluator = RegressionEvaluator(labelCol=\"SoundLevelDecibels\", predictionCol=\"prediction\", metricName=\"mae\")\n",
        "mae = evaluator.evaluate(predictions)\n",
        "print(\"Linear MAE =\", mae)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 33,
      "metadata": {
        "tags": [],
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "odC1lO3kDK1p",
        "outputId": "119cb246-284e-4631-9356-cbc22b9f6f6a"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Lasso MAE = 3.922395099708769\n"
          ]
        }
      ],
      "source": [
        "evaluator = RegressionEvaluator(labelCol=\"SoundLevelDecibels\", predictionCol=\"prediction\", metricName=\"mae\")\n",
        "mae2 = evaluator.evaluate(predictions2)\n",
        "print(\"Lasso MAE =\", mae2)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 34,
      "metadata": {
        "tags": [],
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "DL41CCppDK1p",
        "outputId": "ba6fd2e6-734a-4487-ed63-1bf1d28161c0"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "ElasticNet MAE = 3.893721314528024\n"
          ]
        }
      ],
      "source": [
        "evaluator = RegressionEvaluator(labelCol=\"SoundLevelDecibels\", predictionCol=\"prediction\", metricName=\"mae\")\n",
        "mae3 = evaluator.evaluate(predictions3)\n",
        "print(\"ElasticNet MAE =\", mae3)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ZmtlK53PDK1q"
      },
      "source": [
        "### R-Squared(R2)\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 35,
      "metadata": {
        "tags": [],
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "KZD757qyDK1q",
        "outputId": "4181b76b-9ab8-4846-d602-7ec17ed49156"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Linear R2 = 0.5137496926589646\n"
          ]
        }
      ],
      "source": [
        "evaluator = RegressionEvaluator(labelCol=\"SoundLevelDecibels\", predictionCol=\"prediction\", metricName=\"r2\")\n",
        "r2 = evaluator.evaluate(predictions)\n",
        "print(\"Linear R2 =\", r2)\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 36,
      "metadata": {
        "tags": [],
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "19HEptXDDK1q",
        "outputId": "8c046e98-3682-4a66-b183-bed052c48073"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Lasso R2 = 0.5072931754124199\n"
          ]
        }
      ],
      "source": [
        "evaluator = RegressionEvaluator(labelCol=\"SoundLevelDecibels\", predictionCol=\"prediction\", metricName=\"r2\")\n",
        "r22 = evaluator.evaluate(predictions2)\n",
        "print(\"Lasso R2 =\", r22)\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 37,
      "metadata": {
        "tags": [],
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "2E4PszohDK1r",
        "outputId": "6ec2cc6f-a20b-474b-822b-3b34492d2459"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "ElasticNet R2 = 0.5114584417830002\n"
          ]
        }
      ],
      "source": [
        "evaluator = RegressionEvaluator(labelCol=\"SoundLevelDecibels\", predictionCol=\"prediction\", metricName=\"r2\")\n",
        "r23 = evaluator.evaluate(predictions3)\n",
        "print(\"ElasticNet R2 =\", r23)\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "dME87uvPDK1r"
      },
      "source": [
        "#### Part 3 - Evaluation\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 38,
      "metadata": {
        "tags": [],
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "8EMJu0ozDK1r",
        "outputId": "b11fa918-b0c7-4f1c-daf5-3ddd027f5b38"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Part 3 - Evaluation\n",
            "Linear Regression\n",
            "Mean Squared Error =  25.01\n",
            "Mean Absolute Error =  3.87\n",
            "R Squared =  0.51\n",
            "Intercept =  132.43\n",
            "\n",
            "\n",
            "Lasso\n",
            "Mean Squared Error =  25.34\n",
            "Mean Absolute Error =  3.92\n",
            "R Squared =  0.51\n",
            "Intercept =  132.04\n",
            "\n",
            "\n",
            "ElasticNet\n",
            "Mean Squared Error =  25.12\n",
            "Mean Absolute Error =  3.89\n",
            "R Squared =  0.51\n",
            "Intercept =  132.21\n"
          ]
        }
      ],
      "source": [
        "print(\"Part 3 - Evaluation\")\n",
        "print(\"Linear Regression\")\n",
        "print(\"Mean Squared Error = \", round(mse,2))\n",
        "print(\"Mean Absolute Error = \", round(mae,2))\n",
        "print(\"R Squared = \", round(r2,2))\n",
        "\n",
        "lrModel = model.stages[-1]\n",
        "\n",
        "print(\"Intercept = \", round(lrModel.intercept,2))\n",
        "\n",
        "\n",
        "print(\"\\n\")\n",
        "print(\"Lasso\")\n",
        "print(\"Mean Squared Error = \", round(mse2,2))\n",
        "print(\"Mean Absolute Error = \", round(mae2,2))\n",
        "print(\"R Squared = \", round(r22,2))\n",
        "\n",
        "lassoModel = model2.stages[-1]\n",
        "\n",
        "print(\"Intercept = \", round(lassoModel.intercept,2))\n",
        "\n",
        "print(\"\\n\")\n",
        "print(\"ElasticNet\")\n",
        "print(\"Mean Squared Error = \", round(mse3,2))\n",
        "print(\"Mean Absolute Error = \", round(mae3,2))\n",
        "print(\"R Squared = \", round(r23,2))\n",
        "\n",
        "elasticN = model3.stages[-1]\n",
        "\n",
        "print(\"Intercept = \", round(elasticN.intercept,2))"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "aaxPDZTrDK1s"
      },
      "source": [
        "## Part 4 - Persist the Model\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "OiNvDvSnDK1s"
      },
      "source": [
        "### Save the model\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 39,
      "metadata": {
        "tags": [],
        "id": "UrsHyqfZDK1s"
      },
      "outputs": [],
      "source": [
        "model.save(\"Final_Project\")\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "aZKMIGbjDK1s"
      },
      "source": [
        "### Load the model\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 40,
      "metadata": {
        "tags": [],
        "id": "nax0l-pADK1t"
      },
      "outputs": [],
      "source": [
        "loadedPipelineModel = model.load(\"Final_Project\")\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "AK66SASRDK1t"
      },
      "source": [
        "### Make predictions using the loaded model on the testdata\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 41,
      "metadata": {
        "tags": [],
        "id": "hZEoapmHDK19"
      },
      "outputs": [],
      "source": [
        "predictions = loadedPipelineModel.transform(testData)\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "c_j9hJFeDK19"
      },
      "source": [
        "### Task 4 - Show the predictions\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 42,
      "metadata": {
        "tags": [],
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "dDRMiIP8DK1-",
        "outputId": "aeec1de2-c529-4c0e-e4a7-e845be40b400"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "+------------------+------------------+\n",
            "|SoundLevelDecibels|        prediction|\n",
            "+------------------+------------------+\n",
            "|           128.545|121.03522307921861|\n",
            "|           130.898|122.44608224351367|\n",
            "|           109.951|127.63255285986469|\n",
            "|           112.506|129.25004941528528|\n",
            "|           130.089|122.16654986012581|\n",
            "+------------------+------------------+\n",
            "only showing top 5 rows\n",
            "\n"
          ]
        }
      ],
      "source": [
        "predictions.select('SoundLevelDecibels', 'prediction').show(5)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 43,
      "metadata": {
        "tags": [],
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 564
        },
        "id": "QXaM_PMGDK1_",
        "outputId": "7da15280-077c-49d3-9b0c-1843b288e732"
      },
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 800x600 with 1 Axes>"
            ],
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAArcAAAIjCAYAAAAZajMiAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjcuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/bCgiHAAAACXBIWXMAAA9hAAAPYQGoP6dpAADNHUlEQVR4nOzdd3gUxRsH8O8lJCF0EhJaIkhRBBEQlN6kI6GE3qQI2ECKVEFI6EgXRaQJYhApAUT8qSAgVZAmNhQQUAIIIZAQAimX+f0xbnJl9252b6/m/TxPHri9vd25ubvkvdl33jEwxhgIIYQQQgjxAX7ubgAhhBBCCCF6oeCWEEIIIYT4DApuCSGEEEKIz6DglhBCCCGE+AwKbgkhhBBCiM+g4JYQQgghhPgMCm4JIYQQQojPoOCWEEIIIYT4DApuCSGEEEKIz6DglhDiVgaDATExMe5uhkeKiYmBwWAw21a+fHkMHDjQPQ2SIddGX3PlyhUYDAasW7fO3U1xiKe9dwhxFgpuCfEhy5cvh8FgQN26dTUf4/r164iJicHZs2f1a5iXMhgMOT9+fn4oU6YMWrdujQMHDri7aap4ymu6a9cuNG3aFOHh4ShQoAAqVKiAHj164Ouvv3Zru/Ry4MABGAwGbN261d1NISRPo+CWEB8SFxeH8uXL48SJE7h48aKmY1y/fh2xsbFuD4Q8RatWrbBhwwasX78er776Ks6dO4cXXngB//vf/9zSnj/++AOrVq1S9RhPeE0XLFiAjh07wmAwYNKkSVi8eDG6du2KCxcuYNOmTW5rFyHE9+RzdwMIIfq4fPkyjh49ivj4eLzyyiuIi4vDtGnT3N0sr/fEE0+gX79+Obe7dOmCZ555BkuWLEG7du1kH/Po0SMEBgbCz0//8YOgoCDdj+lsWVlZmDFjBlq1aoVvv/3W6v5bt265oVWEEF9FI7eE+Ii4uDgUL14cL774Irp164a4uDjZ/e7du4fRo0ejfPnyCAoKQkREBF566SUkJibiwIEDeO655wAAgwYNyrkkL+UaKuXsNWvWDM2aNcu5nZGRgalTp6J27dooWrQoChYsiMaNG2P//v2qn9e///6LfPnyITY21uq+P/74AwaDAe+//z4AIDMzE7GxsahcuTLy58+P0NBQNGrUCHv27FF9XiXVq1dHiRIlcPnyZQC5l6I3bdqEKVOmoGzZsihQoABSUlIAAMePH0fbtm1RtGhRFChQAE2bNsWRI0esjnv48GE899xzyJ8/PypWrIiPPvpI9vxyr4Ejr6kz2mgpMTERKSkpaNiwoez94eHhZrdv3bqFl19+GSVLlkT+/PlRo0YNrF+/3mwfqd8tU0Tk8mMHDhyIQoUKISEhAZ07d0ahQoUQFhaGsWPHwmg0mj3+3r17GDhwIIoWLYpixYphwIABuHfvntDzFHXv3j2MGjUKkZGRCAoKQqVKlTBv3jxkZ2cD4O/jkJAQDBo0yOqxKSkpyJ8/P8aOHZuzLT09HdOmTUOlSpUQFBSEyMhIjB8/Hunp6bq2mxBvQSO3hPiIuLg4REdHIzAwEL1798aHH36IH3/8MSewAYDU1FQ0btwYv//+OwYPHoxnn30WiYmJ+OKLL3Dt2jU89dRTmD59OqZOnYphw4ahcePGAIAGDRqoaktKSgpWr16N3r17Y+jQobh//z7WrFmDNm3a4MSJE6hZs6bwsUqWLImmTZti8+bNViPRn3/+Ofz9/dG9e3cAfHLTnDlzMGTIEDz//PNISUnByZMncfr0abRq1UrVc1By9+5d3L17F5UqVTLbPmPGDAQGBmLs2LFIT09HYGAg9u3bh3bt2qF27dqYNm0a/Pz88PHHH+OFF17AoUOH8PzzzwMAfv75Z7Ru3RphYWGIiYlBVlYWpk2bhpIlS9ptj6OvqSvaGB4ejuDgYOzatQsjRoxASEiI4r4PHz5Es2bNcPHiRQwfPhyPP/44tmzZgoEDB+LevXsYOXKk3fPJMRqNaNOmDerWrYsFCxZg7969WLhwISpWrIjXXnsNAMAYQ6dOnXD48GG8+uqreOqpp7B9+3YMGDBA0znlpKWloWnTpkhISMArr7yCxx57DEePHsWkSZNw48YNLFmyBAEBAejSpQvi4+Px0UcfITAwMOfxO3bsQHp6Onr16gUAyM7ORseOHXH48GEMGzYMTz31FH7++WcsXrwYf/75J3bs2KFb2wnxGowQ4vVOnjzJALA9e/YwxhjLzs5mERERbOTIkWb7TZ06lQFg8fHxVsfIzs5mjDH2448/MgDs448/ttqnXLlybMCAAVbbmzZtypo2bZpzOysri6Wnp5vtc/fuXVayZEk2ePBgs+0A2LRp02w+v48++ogBYD///LPZ9qpVq7IXXngh53aNGjXYiy++aPNYagBgL7/8Mrt9+za7desWO378OGvRogUDwBYuXMgYY2z//v0MAKtQoQJLS0vLeWx2djarXLkya9OmTU7fMsZYWloae/zxx1mrVq1ytnXu3Jnlz5+fXb16NWfbb7/9xvz9/Znlr2nL18CR19RZbZQjtbNgwYKsXbt2bNasWezUqVNW+y1ZsoQBYJ9++mnOtoyMDFa/fn1WqFAhlpKSwhjL7ff9+/ebPf7y5ctWz3XAgAEMAJs+fbrZvrVq1WK1a9fOub1jxw4GgL377rs527Kysljjxo0VPxOmpDZt2bJFcZ8ZM2awggULsj///NNs+8SJE5m/vz/7+++/GWOMffPNNwwA27Vrl9l+7du3ZxUqVMi5vWHDBubn58cOHTpktt+KFSsYAHbkyJGcbUqfX0J8DaUlEOID4uLiULJkSTRv3hwAn+Xfs2dPbNq0yeyy67Zt21CjRg106dLF6hh6lnPy9/fPGW3Kzs5GUlISsrKyUKdOHZw+fVr18aKjo5EvXz58/vnnOdt++eUX/Pbbb+jZs2fOtmLFiuHXX3/FhQsXHH8S/1mzZg3CwsIQHh6OunXr4siRIxgzZgxGjRpltt+AAQMQHBycc/vs2bO4cOEC+vTpgzt37iAxMRGJiYl48OABWrRogYMHDyI7OxtGoxHffPMNOnfujMceeyzn8U899RTatGljt32OvKauaiMAxMbGYuPGjahVqxa++eYbTJ48GbVr18azzz6L33//PWe/r776CqVKlULv3r1ztgUEBODNN99Eamoqvv/+e6HzyXn11VfNbjdu3Bh//fWX2bnz5cuXM5IL8PfyiBEjNJ/T0pYtW9C4cWMUL148p78TExPRsmVLGI1GHDx4EADwwgsvoESJEmbv+bt372LPnj1m7/ktW7bgqaeeQpUqVcyO98ILLwCAplQgQrwdpSUQ4uWMRiM2bdqE5s2b5+SBAkDdunWxcOFCfPfdd2jdujUA4NKlS+jatatL2rV+/XosXLgQ58+fR2ZmZs72xx9/XPWxSpQogRYtWmDz5s2YMWMGAJ6SkC9fPkRHR+fsN336dHTq1AlPPPEEnn76abRt2xb9+/fHM888o/l5dOrUCcOHD4fBYEDhwoVRrVo1FCxY0Go/y+clBdi2LmknJycjPT0dDx8+ROXKla3uf/LJJ/HVV1/ZbJ8jr6mr2ijp3bs3evfujZSUFBw/fhzr1q3Dxo0bERUVhV9++QX58+fH1atXUblyZavJeE899RQA4OrVq6JPz0z+/PkRFhZmtq148eK4e/duzu2rV6+idOnSKFSokNl+Tz75pKZzyrlw4QLOnTtn1RaJNLkuX7586Nq1KzZu3Ij09HQEBQUhPj4emZmZZsHthQsX8Pvvv9s9HiF5CQW3hHi5ffv24caNG9i0aZNsSaW4uLic4NZRSiOBRqMR/v7+Obc//fRTDBw4EJ07d8a4ceMQHh4Of39/zJkzB5cuXdJ07l69emHQoEE4e/Ysatasic2bN6NFixYoUaJEzj5NmjTBpUuXsHPnTnz77bdYvXo1Fi9ejBUrVmDIkCGazhsREYGWLVva3c901BZAzuSg+fPnK+YYFypUyK2TftzVxiJFiqBVq1Zo1aoVAgICsH79ehw/fhxNmzYVPoat96Ic0/enO2VnZ6NVq1YYP3687P1PPPFEzv979eqFjz76CP/73//QuXNnbN68GVWqVEGNGjXMjle9enUsWrRI9niRkZH6PgFCvAAFt4R4ubi4OISHh+ODDz6wui8+Ph7bt2/HihUrEBwcjIoVK+KXX36xeTxbl7KLFy8uO3P86tWrqFChQs7trVu3okKFCoiPjzc7niOlyTp37oxXXnkl5zLtn3/+iUmTJlntJ80yHzRoEFJTU9GkSRPExMRoDm61qlixIgAeyNkKjsPCwhAcHCybSvHHH38InUfra+qqNtpSp04drF+/Hjdu3AAAlCtXDufOnUN2drbZ6O358+dz7gf4exGA1ftR68iudOzvvvsOqampZqO3jj5HUxUrVkRqaqrQF6YmTZqgdOnS+Pzzz9GoUSPs27cPkydPtjreTz/9hBYtWvj8SnGEiKKcW0K82MOHDxEfH48OHTqgW7duVj/Dhw/H/fv38cUXXwAAunbtip9++gnbt2+3OhZjDAByLrnLBbEVK1bEDz/8gIyMjJxtX375Jf755x+z/aRRMumYAC83dezYMc3PtVixYmjTpg02b96MTZs2ITAwEJ07dzbb586dO2a3CxUqhEqVKpmNPCYnJ+P8+fNITk7W3BYRtWvXRsWKFbFgwQKkpqZa3X/79m0AvK/atGmDHTt24O+//865//fff8c333xj9zyOvKauamNaWpriay8thiFd+m/fvj1u3rxplmualZWFZcuWoVChQjmju+XKlYO/v39Ojqpk+fLldtujpH379sjKysKHH36Ys81oNGLZsmWaj2mpR48eOHbsmGy/3bt3D1lZWTm3/fz80K1bN+zatQsbNmxAVlaWWUqCdLyEhATZhT0ePnyIBw8e6NZ2QrwFjdwS4sW++OIL3L9/Hx07dpS9v169eggLC0NcXBx69uyJcePGYevWrejevTsGDx6M2rVrIykpCV988QVWrFiBGjVqoGLFiihWrBhWrFiBwoULo2DBgqhbty4ef/xxDBkyBFu3bkXbtm3Ro0cPXLp0CZ9++mnOCKCkQ4cOiI+PR5cuXfDiiy/i8uXLWLFiBapWrSobRInq2bMn+vXrh+XLl6NNmzYoVqyY2f1Vq1ZFs2bNULt2bYSEhODkyZPYunUrhg8fnrPP9u3bMWjQIHz88ceyNXv14ufnh9WrV6Ndu3aoVq0aBg0ahLJlyyIhIQH79+9HkSJFsGvXLgB8stXXX3+Nxo0b4/XXX88J5qpVq4Zz587ZPI+jr6kr2piWloYGDRqgXr16aNu2LSIjI3Hv3j3s2LEDhw4dQufOnVGrVi0AwLBhw/DRRx9h4MCBOHXqFMqXL4+tW7fiyJEjWLJkCQoXLgwAKFq0KLp3745ly5bBYDCgYsWK+PLLLx3KMY2KikLDhg0xceJEXLlyBVWrVkV8fLzqL0Lbtm3LGWk2NWDAAIwbNw5ffPEFOnTogIEDB6J27dp48OABfv75Z2zduhVXrlwxS7Xp2bMnli1bhmnTpqF69eo5uceS/v37Y/PmzXj11Vexf/9+NGzYEEajEefPn8fmzZvxzTffoE6dOto6hBBv5eZqDYQQB0RFRbH8+fOzBw8eKO4zcOBAFhAQwBITExljjN25c4cNHz6clS1blgUGBrKIiAg2YMCAnPsZY2znzp2satWqLF++fFYlkBYuXMjKli3LgoKCWMOGDdnJkyetSoFlZ2ez2bNns3LlyrGgoCBWq1Yt9uWXX7IBAwawcuXKmbUPAqXAJCkpKSw4ONiqVJRk5syZ7Pnnn2fFihVjwcHBrEqVKmzWrFksIyMjZ5+PP/5YqKyT1LY33njD5j72yj+dOXOGRUdHs9DQUBYUFMTKlSvHevTowb777juz/b7//ntWu3ZtFhgYyCpUqMBWrFjBpk2bZrcUGGOOv6Z6t9FSZmYmW7VqFevcuXPOe6JAgQKsVq1abP78+VZl4/799182aNAgVqJECRYYGMiqV68u+3rdvn2bde3alRUoUIAVL16cvfLKK+yXX36RLQVWsGBBq8fLtf3OnTusf//+rEiRIqxo0aKsf//+7MyZM6pKgSn9SOW67t+/zyZNmsQqVarEAgMDWYkSJViDBg3YggULzN6rjPHPUmRkJAPAZs6cKXvejIwMNm/ePFatWjUWFBTEihcvzmrXrs1iY2NZcnJyzn5UCozkFQbGTK4bEkIIIYQQ4sUo55YQQgghhPgMCm4JIYQQQojPoOCWEEIIIYT4DApuCSGEEEKIz6DglhBCCCGE+AwKbgkhhBBCiM+gRRzA1+a+fv06ChcuTMsXEkIIIYR4IMYY7t+/jzJlypgtz22JglsA169fR2RkpLubQQghhBBC7Pjnn38QERGheD8Ft0DOco7//PMPihQp4ubW6CMzMxPffvstWrdujYCAAHc3x+NRf6lHfaYe9Zk61F/qUZ+pR32mjjv7KyUlBZGRkTlxmxIKboGcVIQiRYr4VHBboEABFClShD6sAqi/1KM+U4/6TB3qL/Woz9SjPlPHE/rLXgopTSgjhBBCCCE+g4JbQgghhBDiMyi4JYQQQgghPoOCW0IIIYQQ4jMouCWEEEIIIT6DgltCCCGEEOIzKLglhBBCCCE+g4JbQgghhBDiMyi4JYQQQgghPoOCW0IIIYQQ4jMouCWEEEIIIT6DgltCCCGEEOIzKLglhBBCCCE+g4JbQgghhBDiMyi4JYQQQgghPoOCW0IIIYQQ4jPcGtwePHgQUVFRKFOmDAwGA3bs2KG476uvvgqDwYAlS5aYbU9KSkLfvn1RpEgRFCtWDC+//DJSU1Od23BCCCGEkLzMaAQYc3crZLk1uH3w4AFq1KiBDz74wOZ+27dvxw8//IAyZcpY3de3b1/8+uuv2LNnD7788kscPHgQw4YNc1aTCSGEEELytPx37sC/dWtg2TJ3N0VWPneevF27dmjXrp3NfRISEjBixAh88803ePHFF83u+/333/H111/jxx9/RJ06dQAAy5YtQ/v27bFgwQLZYJgQQgghhGhj+OYbNBs9Gn4pKcBvvwGDBgGFC7u7WWbcGtzak52djf79+2PcuHGoVq2a1f3Hjh1DsWLFcgJbAGjZsiX8/Pxw/PhxdOnSRfa46enpSE9Pz7mdkpICAMjMzERmZqbOz8I9pOfhK8/H2ai/1KM+U4/6TB3qL/Woz9SjPhOUmQm/mBjkmz8f+QBkP/MMjBs3AvnzAy7qO9HXyKOD23nz5iFfvnx48803Ze+/efMmwsPDzbbly5cPISEhuHnzpuJx58yZg9jYWKvt3377LQoUKOBYoz3Mnj173N0Er0L9pR71mXrUZ+pQf6lHfaYe9Zmy/Ldvo87ChQg9fx4AcLldO/wyaBCyL14ELl50WTvS0tKE9vPY4PbUqVNYunQpTp8+DYPBoOuxJ02ahDFjxuTcTklJQWRkJFq3bo0iRYroei53yczMxJ49e9CqVSsEBAS4uzkej/pLPeoz9ajP1KH+Uo/6TD3qM/v8hwyB3/nzYEWKIGP5cpwrVMgt/SVdabfHY4PbQ4cO4datW3jsscdythmNRrz11ltYsmQJrly5glKlSuHWrVtmj8vKykJSUhJKlSqleOygoCAEBQVZbQ8ICPC5N7YvPidnov5Sj/pMPeozdai/1KM+U4/6zIYlS4AHD2B49134RUYCX33llv4SPZ/H1rnt378/zp07h7Nnz+b8lClTBuPGjcM333wDAKhfvz7u3buHU6dO5Txu3759yM7ORt26dd3VdEIIIYQQ73XlCjB9em6pr+LFga1bgQoV3NosUW4duU1NTcVFk1yNy5cv4+zZswgJCcFjjz2G0NBQs/0DAgJQqlQpPPnkkwCAp556Cm3btsXQoUOxYsUKZGZmYvjw4ejVqxdVSiCEEEIIUWvHDl4B4d49oHRpYOhQd7dINbeO3J48eRK1atVCrVq1AABjxoxBrVq1MHXqVOFjxMXFoUqVKmjRogXat2+PRo0aYeXKlc5qMiGEEEKI70lPB0aOBLp04YFtvXpA69bubpUmbh25bdasGZiK1S2uXLlitS0kJAQbN27UsVWEEEIIIXnIpUtAz56AlOY5bhwwaxbgpTnIHjuhjBBCCCGEONkXXwD9+wMpKUBoKLB+PWCxaJa3oeCWEEIIISSvKloUSE0FGjUCPvsMiIhwd4scRsEtIYQQQkhe8ugRX1kMAJo2BfbuBRo3BvL5RljosaXACCGEEEKIzjZu5CW9/vgjd1vz5j4T2AIU3BJCCCGE+L60NF7Wq29f4MYNYOlSd7fIaXwnTCeEEEIIIdZ+/x3o0QP45RfAYADeeQdQUXbV21BwSwghhBDiq9avB15/nY/clizJ0xJeeMHdrXIqCm4JIYQQQnzR5s3AwIH8/y1bAp9+ygNcH0fBLSGEEEKIL+rShVdBaN0amDQJ8Pd3d4tcgoJbQgghhBBfwBiwfTvQoQMQGMhXGNu3z6cqIYigagmEEEIIId7u/n2+0ljXrnyUVpLHAluARm4JIYQQQrzbTz/xagh//slTD8LD3d0it6LglhBCCCHEGzEGfPQRMGoUkJ7Ol87dtAlo2NDdLXMrCm4JIYQQQrxNcjIwbBiviADwPNt164DQULc2yxNQzi0hhBBCiLe5fRv43/94Tu3ChcAXX1Bg+x8auSWEEEII8TaVKuXWra1b192t8Sg0cksIIYQQ4unu3gW6dQO++y53W8eOFNjKoJFbQgghhBBPdvw40LMncPUq8OOPwIULvI4tkUUjt4QQQgghnogxnk/bqBEPbCtUAOLjKbC1g0ZuCSGEEEI8zZ07wMCBwJdf8ts9egArVwJFi7q1Wd6AgltCCCGEEE9y6xZQpw7wzz9AUBCwZAnwyiuAweDulnkFCm4JIYQQQjxJWBjQpAlw4gSvY1uzprtb5FUouCWEEEIIcbfbt/nSuSEhfIR2xQqec1u4sLtb5nVoQhkhhBBCiDsdPMhHZwcN4gEtABQqRIGtRhTcEkIIIYS4g9EIzJwJNG8OXL/OS3wlJrq7VV6PgltCCCGEEFe7eRNo0wZ45x0gO5tXRvjxR55vSxxCObeEEEIIIa703XdA377Av/8CBQoAH34IvPSSu1vlMyi4JYQQQghxlYwMYOhQHtg+/TSvhvDUU+5ulU+htARCCCGEEFcJDAQ++4zXrT1xggJbJ6CRW0IIIYQQZ/rmG17qq18/frtuXf5DnIJGbgkhhBBCnCErC5g0CWjblqci/Pqru1uUJ9DILSGEEEKI3v75B+jdGzhyhN8ePBioWNG9bcojKLglhBBCCNHT7t28+kFSElCkCLB6NdC9u7tblWdQWgIhhBBCiF4mTgQ6dOCBbe3awOnTFNi6GAW3hBBCCCF6CQ7m/44cyVMSKBXB5SgtgRBCCCHEEQ8f5ga1U6YATZsCzZq5tUl5GY3cEkIIIYRokZ7OR2gbNAAePeLb/P0psHUzCm4JIYQQQtT66y+gYUPgvfeAs2eBr75yd4vIfyi4JYQQQghRY+tWoFYt4NQpICQE2LULiI52d6vIfyjnlhBCCPEyRiNw6BBw4wZQujTQuDG/Gk6c7NEj4K23gOXL+e2GDflSupGR7m0XMUPBLSGEEOJF4uN5mue1a7nbIiKApUuBqCj3tcsV3B7UjxjBa9YCfOWx2FggIMCFDSAiKLglhBBCvER8PNCtG8CY+faEBL5961bfHcG1FdS7LCNgyhQeXS9dCrRp46KTErUo55YQQgjxAkYjD+4sA1sgd9vEia5tk6tIQb1pYAvkBvXx8U46cVoasG1b7u1y5YBff6XA1sNRcEsIIYR4gUOHrIM7U4zZvt9biQT1o0bx/XT1++9A3bo8et69O3e7rw6N+xAKbgkhhBAvcOOGu1vgHiJB/T//8P1088knQJ06wC+/ACVL5i7QQLwCBbeEEEKIFyhd2t0tcA/RoF6X4P/BA2DQIGDAAJ6S0KIFr2H7wgs6HJy4iluD24MHDyIqKgplypSBwWDAjh07zO6PiYlBlSpVULBgQRQvXhwtW7bE8ePHzfYpX748DAaD2c/cuXNd+CwIIYQQ52vcmE+gMhjk7zcY+P2+RjSodzj4//VX4PnngXXrAD8/YPp04JtvgFKlHDwwcTW3BrcPHjxAjRo18MEHH8je/8QTT+D999/Hzz//jMOHD6N8+fJo3bo1bt++bbbf9OnTcePGjZyfESNGuKL5hBBCiMv4+/NJ+oB1gCvd9sWxHZGgPjKS7+cIw88/A7/9BpQpA+zbB7zzDuXXeim3lgJr164d2rVrp3h/nz59zG4vWrQIa9aswblz59CiRYuc7YULF0Yp+mZFCCHEx0VH83JfciWxlizhdW59bRVYKajv1o0HsqYTy6SAd8kSx+NQ1qsXkJIC9OgBhIU5djDiVl5T5zYjIwMrV65E0aJFUaNGDbP75s6dixkzZuCxxx5Dnz59MHr0aOTLp/zU0tPTkZ6ennM7JSUFAJCZmYnMzEznPAEXk56HrzwfZ6P+Uo/6TD3qM3Wov+RFRQHt2wPHjgE3b/Kr5vXr8+DOV/ssKooH9RMm8PJfkogIPlodFQWofspnz8J/7FhkrlsH4L8+GzYM/93Qpd2+yJ3vMdFzGhiTK67hegaDAdu3b0fnzp3Ntn/55Zfo1asX0tLSULp0aezYsQPPPfdczv2LFi3Cs88+i5CQEBw9ehSTJk3CoEGDsGjRIsVzxcTEIDY21mr7xo0bUaBAAd2eEyGEEEI8DGMo//XXeHrtWvhnZuLv5s1xZuRId7eKCEhLS0OfPn2QnJyMIkWKKO7n8cHtgwcPcOPGDSQmJmLVqlXYt28fjh8/jvDwcNnjrF27Fq+88gpSU1MRFBQku4/cyG1kZCQSExNtdpY3yczMxJ49e9CqVSsE0NKAdlF/qUd9ph71mTrUX+rp2WdGo/zosFdLTob/a6/Bb+tWAEB2+/Z4tGIF9pw+Te8zQe78XKakpKBEiRJ2g1uPT0soWLAgKlWqhEqVKqFevXqoXLky1qxZg0mTJsnuX7duXWRlZeHKlSt48sknZfcJCgqSDXwDAgJ87o3ti8/Jmai/1KM+U4/6TB3qL/Uc7TOPWOpWb6dOAT17ApcuAfnyAXPnwm/MGARkZQGg95la7ugv0fN5XZ3b7Oxss1FXS2fPnoWfn5/iyC4hhBBClLltqVtn+vZboEEDHtiWK8dXfHjrLeUSDMSruXXkNjU1FRcvXsy5ffnyZZw9exYhISEIDQ3FrFmz0LFjR5QuXRqJiYn44IMPkJCQgO7duwMAjh07huPHj6N58+YoXLgwjh07htGjR6Nfv34oXry4u54WIYQQ4pXsLXVrMPClbjt18rIUhfr1eVBbrRqwdi1AMYJPc2twe/LkSTRv3jzn9pgxYwAAAwYMwIoVK3D+/HmsX78eiYmJCA0NxXPPPYdDhw6hWrVqAHh6waZNmxATE4P09HQ8/vjjGD16dM5xCCGEECJOzVK3zZq5rFna/P47UKUKj8gLFwYOH+Ylvmi01ue5Nbht1qwZbM1ni7dz7ePZZ5/FDz/8oHezCCGEEJ9gNPJA9MYNvoJX48a2R1xdutStszAGLF7M64YtWMCHogGA0hXzDI+fUEYIIYQQ9bRMCnPZUrfOcucOMHAg8OWX/PaPP+bmU5A8w+smlBFCCCHEtl27tE0Kc9VSt05x9ChQqxYPbIOCgOXLgQ0bKLDNgyi4JYQQQnzMhAnKk8IAPinMaLS+X1rqFrCOCfVc6lZX2dnAvHlAkyY8IbhyZeCHH4DXXqPANo+i4JYQQgjxMaZL1FoynRQmJzqaL3Vbtqz59ogIvt3j6tz+9hswZQqP1vv04fVsa9Z0d6uIG1HOLSGEEJIH2ZoUFh3Ny33JTUZTO0nN6Z5+Gli4EChQAHj5ZRqtJRTcEkIIIXmRvUlh/v7W5b48YuUyo5GnIURFAdWr821vvumikxNvQGkJhBBCiI8pW1b/SWEesXLZv/8CbdoAkycDPXoAjx654KTE21BwSwghhPiYefP4v3pNCrO3chmgPElNN999B9Sowf8tUACYOBHIn9+JJyTeioJbQgghxMdERek7KUzNymW6MxqBadOAVq34yG21arx+7YABTjgZ8QWUc0sIIYT4IFuTwtQSXZHsu+90nmh27x7QpQtw4AC/PWQIT/AtUMDBAxNfRsEtIYQQ4qNMJ4VprXJgNPIBUxEzZ+b+X5eJZoUL81yKQoWAjz7ipb4IsYOCW0IIIcTHaa1yIPc4UdJEM9VpEFlZPKIOCuLRd1wckJICPPmk+kaQPIlybgkhhBAPYzTyK/Gffcb/dWSiltYqB0qPE6Vpotm1a0Dz5sCYMbnbSpemwJaoQsEtIYQQ4kHi44Hy5XmM16cP/7d8eW2ltrRWObD1ODVUTTTbvZuvLHb4MPDppzD+c123AJ/kLRTcEkIIIR5C71qyWqsc2HucpH9/sXbYnJCWmQmMGwd06ADcuQPUro1v5pxG+QZldAnwSd5DwS0hhBCPJI3Ubd2aN0bunFFLVrTKgeV+oo8rV05sP8XV0K5c4TPbFizgt0eMwI5xR9BueEWHA3w9UzuId6HglhBCPAz9UeYBjLSy6ssv542RO2fUkrW3xK7SfqKPa9aMT0wTWQ3N6n2dngW0aAEcPw4UKwbEx8O4+D2MGBvkcICvZ2oH8T4U3BJCiAehP8q5l+YTEsy3u3SZVwuu+MKhdZQVyG3f1q25twEeVIoGn6ZEH9esGa+4IG2z3Afgq6Ht3Cnzvq6UD8e6LQTq1gXOnAG6dNElwPeIZYKJW1FwSwghHkLpj/K1a0DXrsD06b4/iusRy7xacNUXDq2jrKbte/llvq16db7d318s+LSsd6vmcdHRtldDA3Lf14/jLzTB9wB4sNlwfmfEjz3KnwAcC/ABz3z/ENej4JYQQjyAyOz0adN4jqMvjzy5dZlXGa4cBdQyyqrUvuvXc9tnL/hUqkGr5nHR0Tx9dv9+YONG/u/ly3yFNOl93RVbcQa1sA1dURbXcoPNMX45wabWAF/iae8f4h4U3BJCiAcQnZ3ujkurrswBdnTkTk+uHgVUO8qqpn1Kwae9xRW0Pk5y6BBw+9ojvI83sBXdURQp+AO5NWstg02taRQST3r/EPeh4JYQQjyA2j+2rrq06uocYEdH7vTkjlFAabS0TBnz7WXLWo+Wqm2ftBRv7978X5Gld0Uft2ULUKqU9fvkyLoLOIb6eAPLAQBzMQHNcAAJiDB7vPT+15pGIRF9X/z7b96esOnrKLglhBAPoCZYc9WlVXdMzHF05E5P7hwFVHr+Ws7r7FHK8eOBHj2AxETz7Y2ufYY31z+LWjiL2yiBtvgfJmEushBgdQzp/W80AiEhfEQ6NNR8H3tpFID99w/AA+PRo3ODcKkqB/EdFNwSQogHEPmjbMmZQYu7JuZII3dKuceM2R6505M7RpHVfKHwhFHurVuB+fPl73sB+1AYqTiIJqiFs/gGba32Mf2yYnqVYMkSHiyHhfH3mWg6hK2RX4nle1aqyrFzp+1jE+9BwS0hhHgA0z/KopwZtNDEHMdGkbXkKav9QuHuUW6jEXj9deX7R2IpRmIJXsB3SEBZu6XC5IL6xET+uUhKEv9CozQRzt7jBw3i6RXE+1FwSwghHkL6oxwRYXs/V1yad9clbynAU2IwuC7fWGv+p9Y8ZTVfKIxG/m+3bny7lvxURx06BNy+nXu7Pz7BVnSFAdkAgIcogPcwEkbkw6hRylUXTCsqWFK6SmDvy4PlRLjFi+2/Z4xGnl7hy9VI8goKbgkhxINIf5RjY+Xvd0XQArjvkrdogBcT45rJQGrLaDmSpyz6RcF0QYQlS/g2P4u/5nIT0PQmtbcAHmAtBuETDEBXxKMv4qz27dRJueqC2qsEol8eTCfClSwp/ryoDq73o+CWEEJ05mjpLH9/YOpUYNs261FckUk1enDXJW/RAG/mTNet3iZaDsvRPGXRLwpLllgHg9IxpTSBc+ec/x4pXRqohl/wI57DIKyDEX6YilhsRB+z/cLC+PtEqeqCmqsEWr88qPkS5uvpNnkBBbeEEKIjPUtnOVpj1BH+/vxSrlyg5szRY7Ujwa6q+ytSDsvRPGXTS/y22qHEYAC++ML+frpgDE0urMGPhudRFb/jOkqjBb7DDExFNsxP/sEHttsj+pqHh2v/8iB9WRNFdXC9GwW3hBCiE3ujSrt2qT+m1tqkjoqP5+WS5Dhz9Fht1QhPWlLVkTxloxEYM8b+Y209R8bEFgLRxeTJ8Bs2BMHsIb5Ba9TEWXyPZla7jRsHdO9u+1CiVwkA7V8e1E7YdEUdZeI8FNwSQogORC5JT5zo2jZppRSkSxYudN7osb1SYHI8pXKDI3nKoivUeYzu3YGCBYHZs/Fgy/8QFBFudndYGPD550D79vbTc0Qn7t26JdY0pS8Z0dG8GoK90W9X1VEmzkPBLSGE6EDkkrQ3BC+2gnSA//F/6y33j5LKcfelZEfylN3ddrsYA375Jfd2rVo8R2bSJER387NKn/ngA/4+EU3PEZm4p8ckx27deLAtx1WTNYnz5dPyoMuXL+PQoUO4evUq0tLSEBYWhlq1aqF+/frInz+/3m0khBCP5/HBiSA1eaPNmul/fnulwGxx96VkaQSyWzceKJl+QbAXOIm2vUQJ4M4d5VxoNXmlwlJSgKFDge3bgaNHgTp1+PawsJxdpPQZgAewPXtat1FKz7FMaZHKmqWnA+vW8W23bvE+kSaiAblfHhISbD9/e6Ou3bvzyZojR5q/18uWBebOdU1OO3EuVSO3cXFxeP7551GxYkVMmDABO3bswKFDh7B69Wq0bdsWJUuWxOuvv46rV686q72EEOKR3B1Y6cXdS7pquTzvSZeS1ZYOk4iO+i5fnnvb8n6AB2e6OnUKePZZYPNmHlGeO2dzd7UVIywnYLZsCQwcCAQFWeeYa607LCc6Grh0iU+aHDaMbztzhgJbXyEc3NaqVQvvvfceBg4ciKtXr+LGjRs4deoUDh8+jN9++w0pKSnYuXMnsrOzUadOHWyhZT4IIXmISHDilFE1nbl7SVe1QbMnXkrWUuVCNHDr3t128BwVpdOTYAxYtgxo0IBHgeXK8W8egwfbfJiakX+l3O5r15QrYGj98mApPh6oWJFPmly5km+rWZMWcPAVwmkJc+fORZs2bRTvDwoKQrNmzdCsWTPMmjULV65c0aN9hBDisaTLqTdu8GBv8WK+wpHSJWndR9WcQK9Lv1qpDZojInjQ52kjbqaX6UVJgZvl5XLL5xgdzRdFMH3vSZfvMzPNj2n5HjW9zK/o7l3g5Zd5GgLAT/bxx0Dx4nafg+iXk4QEPsFSKbebMT6i2qmTdXttPX8RUlBtee7r1+XTJoj3EQ5ubQW2lkJDQxEaGqqpQYQQ4g3i4+WDkLFj+YQVueAkKgr46iuXN1UVR/JG9WAaXCsJC+NfJMqWVRfUeAPRwM1e8Gw0AnPm8NcyKSl3e0QE32YzeNu4kQe2AQHA/PnAm28K12YT/XJy+7b99JM7d4BZs/iCJpa0fHkA7KdNMMbTi4sWdW3pPaIvTRPKTp8+jYCAAFSvXh0AsHPnTnz88ceoWrUqYmJiEBgYqGsjCSHEkyiN/CQkAAsW8BJIYWH2R9U8legIojNYBtempNsrVvj2yJrWwM1UpUryXxDkJnVZje4Oew3+v/zCUxCee87qGLZGg0VH/k3motk0fz4webJ+QaZITndSEs/9FfoiQDySplJgr7zyCv78808AwF9//YVevXqhQIEC2LJlC8aPH69rAwkhxJOITJh56y3+R97VCy/oyZ2ro0nBdZky5ttdtfSwN5KWfJ40id82Ha01ZTmpKz4eqPlYEn5qPhJD+jzgJbsq+CG+1Yeyga29FfhEc4ctc2aVpKby5yXK3tLXanK6ExKArl15Xq6WZbSJ+2gKbv/880/UrFkTALBlyxY0adIEGzduxLp167Bt2zY920cIIR7F0SVWvYm7VkcDeAD788/8/2vWuDa49jamAadUTcEW6T06axawsOtR7L5eEyPxHhaDL0mntKSxvRX4pP1FJn01bgwULiz2/ESDW6XAe/r03GA3PNzOQUxIXwSWLHFsGW3ieprSEhhjyM7OBgDs3bsXHTp0AABERkYiMTFRv9YRQoiHcXeprLxECqa7dePpn8SaUoqMPQZkI3vuAnyPt5EPRlxAJXyI1wDwYxkMfHRXmtBl74qF5f72cof9/YE2bXiwK8LexDilfrh2DZg2Lfd22bJAoUJ8RFgtpTq9xPNoCm7r1KmDmTNnomXLlvj+++/x4YcfAuCLO5QsWVLXBhJCiCdxd6kskvcoBXb2VpNTUgK3sR4D0P7h/wAAG9Ebr+AjpCJ3KNVysQ4ti3tY5g5LKQPS8xg2TCy4zZePj5pa5n9L+bBq+sHWREV75IJ44pk0BbdLlixB3759sWPHDkyePBmVKlUCAGzduhUNGjTQtYGEEOJJ3F0qi/g2y0A2MZHnfMoFdiEh6he8eBan8AU6oiyu4yHy4028h9UYAkC+GoJ0BcLRKxZy1UVERlELFwZiYqy3m46iaukHrZy9Qh/Rh6bg9plnnsHPUjKUifnz58OfvsoQQnyYvz8vQ9W9u/V9nrigAPEecgGgHCmw07JM8Q2URgAykVjiSbyQuBk/4xmb+0tXIBy5YmGrrqy90Val4kumo6hz5oi1TU+ennakqb6xD9E0oUxJ/vz5EUCJUYQQHxYfz0fS5NBsfqKV0mQtOVJAGBcnduyCyB0azQgtg5/nf4PiF0/ibsQzdpf7la5AiC4PbHnFQqS6iJzChYHYWF7rVok0inr7tvI+zuKpaUdGI59AFx6uXNEiLxAObosXL46QkBChH0II8UX2ApCFCymwJeppyZ1ljAd1JUrYXl+hOfbhAipjQKFtiI0F/v0XaDG2FvyLFhIq2WU6AUzN/hKRurJy7t8HfvxRbN+wMNuBt56UgnhPEB8PlCzJJ9BZloJTqoDhq4TTEpYsWaL7yQ8ePIj58+fj1KlTuHHjBrZv347OnTvn3B8TE4NNmzbhn3/+QWBgIGrXro1Zs2ahbt26OfskJSVhxIgR2LVrF/z8/NC1a1csXboUhQoV0r29hJC8SyQAeeMNICPDN1fOIs6jNQAEgH79eFBpyY8ZMQ2zMBXT4QeGj59ZDMM70WYRoNrFOpT2L1uWr+qVns4njJm+9x25fC+6ml/Zssqr6qkVFsZHO6VA3pQnpx3Fx/OavEry2mQ44eB2wIABup/8wYMHqFGjBgYPHoxomeGOJ554Au+//z4qVKiAhw8fYvHixWjdujUuXryIsP+WN+nbty9u3LiBPXv2IDMzE4MGDcKwYcOwceNG3dtLCMm7RAKQ27d5sAHQ6kZEnCMBYIcOwIYN5pfv8yclYXdGWzTF9wCAzwq+jB5fvwd/maFN0eV+lfa/cAFYtcq83Jbpe9+Ry/fZ2XwZ3JQU+5M3/f2BzZuB11/XnqZgMOSuftekiXtW6NNC+uJtT16aDKc55/bSpUuYMmUKevfujVu3bgEA/ve//+HXX38VPka7du0wc+ZMdOnSRfb+Pn36oGXLlqhQoQKqVauGRYsWISUlBefOnQMA/P777/j666+xevVq1K1bF40aNcKyZcuwadMmXL9+XetTI4QQK2oDkLx2GVANe6tI5TVaAkCDgY8y7ttnHti2MO5Bs9Gj0TT7e6SiIPriU/R5sBqHThVQPJa/Pw8QS5fm7/NDh2y/JlKJr6AgXsnA1qIO9nJ17ZEKMNlLhZBy4U0D2yJFeCUFU5GRwLhxvE2WNmzIDVzduUKfWmpH/j19MpweNFVL+P7779GuXTs0bNgQBw8exKxZsxAeHo6ffvoJa9aswVbRqswqZGRkYOXKlShatChq1KgBADh27BiKFSuGOnXq5OzXsmVL+Pn54fjx44pBc3p6OtLT03Nup6SkAAAyMzOR6S2Lv9shPQ9feT7ORv2lXl7rs1KlgOBgdY8xGICJE4H27fkf4LzWZ3J27QImTDCvN1q2LDBvHhAVZb5vXumvevWASpXEqgeYSk3llTuk9+WT2b9jZ3oH+GUw/OL/NPrm+wwX/J5EMDJx4wag1I1qXhOJ0cgfkz+//P2m7/2lS3OvaKglTdK0bF9EBDB3Lm9ffDzQvz/vO9PPaFYW/5kxA6hYkX+G69fnn8Xp04Fjx4CbN4GSJTNx/z7Qtm2mVR81bJj7/+xs/uNpbtxQ97upVCnl94IId34uRc9pYEx9dkr9+vXRvXt3jBkzBoULF8ZPP/2EChUq4MSJE4iOjsY1DclDBoPBKucWAL788kv06tULaWlpKF26NHbs2IHn/lvvevbs2Vi/fj3++OMPs8eEh4cjNjYWr732muy5YmJiEBsba7V948aNKFBA+dstIYQQ4smqr1oFg9GIXwYNQnZQkLubQ4iu0tLS0KdPHyQnJ6NIkSKK+2kauf35559lc1rDw8N1X363efPmOHv2LBITE7Fq1Sr06NEDx48fR7iaBaItTJo0CWPGjMm5nZKSgsjISLRu3dpmZ3mTzMxM7NmzB61ataLybAKov9TLi322axcfIQLUjbCtWcMv0+bFPpMYjUD16sorRBkMfLTw3LncfE9f6C+5UVGJ5eio3L4REcCsWbwqwvXrfDTUNA2htfFr/OJXHdcNZQEABfK3wJqPv8Pgwa3w8CHvM39/YO1awHTsaNcuYPx4fkwlBgNQvDgfFbQc1e3UCVi+3P7zX7WKpy44sjLY7t1Ao0by9x0+DLz4omPHkN5nRmMrTJgQoGoE2xNIny17I/8GA0+9cPS5uPNzKV1pt0dTcFusWDHcuHEDjz/+uNn2M2fOoGzZsloOqahgwYKoVKkSKlWqhHr16qFy5cpYs2YNJk2ahFKlSuXk+0qysrKQlJSEUqVKKR4zKCgIQTLfaAMCArz2F6gSX3xOzkT9pV5e6jMp306k0L6p0qUB0y7KS30mOXIEuHjR9j4XLgA//GA92cVb+0tp8QLJxYu5q2xFR9uf3HXgQO77Lh8yMQuTMR7zcRCN8QL2wWjyJ/3hw4Cc4BYAevbMPY+9dplKS7PedukSL3sn4vZt+6+7PTdvmn9+LO97+NCxY0j69w9AWpr5Tpcumb9GjnLG4goBATwA79aN35Z7XUNDgZUr9c0ZdsfnUvR8miaU9erVCxMmTMDNmzdhMBiQnZ2NI0eOYOzYsXjppZe0HFJYdnZ2Tr5s/fr1ce/ePZw6dSrn/n379iE7O9usXBgh3oom3nge04kmn37KJ/WoLWyfFzm6fKu3Ea1dyxgvzyR9tqXJWr17839NAx+pbx7DVRxEE4zHfADAWdSEH+wng44axUvVqa2pK9dmg8F2UCa99/8rbOQQWxPuRCfjXbiQ+3/L36sZGXy7rYUmTF8jU2p+R8fH88UUnLG4glSmzXJ8MSQEOfWNPXEynNMwDdLT09mQIUNYvnz5mMFgYAEBAczPz4/169ePZWVlCR/n/v377MyZM+zMmTMMAFu0aBE7c+YMu3r1KktNTWWTJk1ix44dY1euXGEnT55kgwYNYkFBQeyXX37JOUbbtm1ZrVq12PHjx9nhw4dZ5cqVWe/evVU9n+TkZAaAJScnq3qcJ8vIyGA7duxgGRkZ7m6KV/DE/tq2jbGICMb4r1f+ExHBt3sCT+wzd9i2jTGDgf+YvlbSNtPX6+FD3mcbN2aw/fsZU/Hr0uvt32/eP0o/+/fnPsaT3mNZWbxtGzcyoddO9PnKPW9bx+yIHewOijMGsLsoyrpgm9lxgoN5nwUHZ8ieZ/Fide0S+bH13lfbD5bHiYy03ddZWYyVLWv/WBERfF/536u2+0zpNVLzO1r6PSH3HC1/TzhC7ftUC3d+LkXjNU3BreTq1ats9+7d7PPPP2d//vmn6sfv37+fAbD6GTBgAHv48CHr0qULK1OmDAsMDGSlS5dmHTt2ZCdOnDA7xp07d1jv3r1ZoUKFWJEiRdigQYPY/fv3VbWDglviaf3lql+EjlDbZ674pesucn/kIiPNX6dt2xirVMn8j6gnfVmRo+drlpXFn6/c+1opkPGUz6W9IEaunzZuVBfIbdyofP6sLMYOfJvOfms7KucBx/EcK4+/rI5jL7gdPlzfwHbUKNvvfXuvu73gVuTzERsrdrzYWPl22OszuddIze9oqQ9sPU97Qbwn8fngljHGsrOzWXZ2tqOHcSsKbokn9Ze3/CJU02eePgqtB1uBoPSH0PKPqCd9WbHkjNdMzSg3Y57xubQXxIwbJ99PogGX9KM0ciu9DsF4wH5BVcYAtgBjWADSZY/j6pFb6b1u60uQ0utu66dECR44mx5P6TyiXyQKF5bfLhrcSq9RVhZjoaHiv6O1XLXwZD4d3K5evZpVq1aNBQYGssDAQFatWjW2atUqrYdzKwpuiSf1l7f8IhTtM28YhXYm0y8rcn9EPeXLiilnvmYio9wSRz6Xeow62/uiaSu4AXgAJBLQyb3+WVlSgJyds99T+JVFYafsMSIi+PkKFJAP1KT3WXq69pFUuXOK9uu2bYyFhYkdt0AB6/MofYlwNPVBNLg1fY1Ev7hIv6NFg29bo/eexBuCW00TyqZOnYqRI0ciKioKW7ZswZYtWxAVFYXRo0dj6tSpjicCE5KH+dLEG1uTaqRtShM1fIW91YMYy10S0xM4+zVzxcpPek3cUbvyk4Qx8RW5DIbcVbYk8fHAk+UeIXTaG5iAeTnbf0dV7ELHnNthYXxS4/79vE9XrlQ+B8DPExjIF1Uw3a7V0KHiM/2jo/mCEyIsKzRcuwbMn6+8ElpiovyKY3qSXiOjMbf/7JF+R4tOenNkqWJiTlNw++GHH2LVqlWYM2cOOnbsiI4dO2LOnDlYuXIllosUviOEKPKlX4TeFtg5g7d9WXHFa2arIoCjpDJXtpaEFWE0At99p70djPF6tDExyoFXZKR1ian4eGBC1wvYnNAAb2A5ZuAdROJv2cffvs1nxzduzF+P9HRg0iTr/aRVvkyXlpWbWa/W3bvq9te5UmjOl63Ro/nKY84yalRu3x06BCQliT1O+h1tbwliqqqiP011bjMzM82WvJXUrl0bWVlZDjeKkLxM+kWYkCA/emYw8Pu94RehtwV2zuBtX1Ycfc2cUcdTlL1RZ4OBByqdOtluU3y8+lrGSipX5qOqhw7xz/Tt23zEVQpKTdthNAL/G7AJpzEUhZGK2yiBl/AJ/sFjisffuZMvLCK1NTiYl6V6+22+5KzSayDV1F22jAeHWsTFAQsWiL++9n63acEYf+5btuhzPDmdOuX+X/TzERqa+zva35+P9nbrxt+Dps/ddFTdVZ+TvEDTyG3//v3x4YcfWm1fuXIl+vbt63CjCMnLpF+EgPU3fW/7RehtgZ0zeNuojSOvmTPreIrQY9RZaeRXq9Klc0eq+/blwXXfvjIj1g8f4mzdV7AqtTcKIxUH0Rg1cRZfo53N4y9ZIt/WOXOAoCDbI+P+/kDJkpqeFgAeqIuM4Eu1YDdv5qkMatI23Enusyn6+XjzTfN+VxottxxVJ/oQHrk1Xa7WYDBg9erV+Pbbb1GvXj0AwPHjx/H33387fREHQvIC6Reh5ehRRAT/Y+Ytvwh9aRRaK8tRG1Oe+GVF62umtOqVlA7gij/gWkedpdHmhAQefOoxqqjqvW00gjVpitqnfkQ2DJiFyYjFNLMVx+RIOaBKREapHf1iaa/P5UbBQ0P5v6bLCIeF8WDZUyh9NkVGn0NDgcmTrbfbW4FOb+68iuJuwsHtmTNnzG7Xrl0bAHDp0iUAQIkSJVCiRAn8+uuvOjaPkLzL1b8InYEux3HSl5UJE8y3e+KXFS2vmV7pAI7SMuqsNQUhMhLo1YtflgccfG/7++Nig5dQ5ORV9MOn2ItWQm2wFdiajlJbLmdsavduoVMpstXnSl94pJzV2FietlG6NNCgAf88uCrADQvjE9zKluUT0iwpfTZtfT4kK1faHi239XroRe59HRHB2+5Jv2+cxkXVGzwalQIj1F/qOVrnVqn8ky/zphXK1Lxmzipfp2WhEDULRSiVPLP3M2VK7jE2b+Y1WVW/t1NTGTNZ/GhjXDYLQaLQ+SMieA1YkbJWtspLpacz5u+vrXyWvRJ2IvW6w8IY+/TT3FJtSs9Jzx+5cnbS+2zfvgzh0nGe/DvN2eUXvaEUmKYJZYQQooYvjELrQXq+3boBAQHubYs9oq+ZmqoCzp44qGbU2dZosz0tWvBjxMcDY8aYj/yFhQGLFtkZHfv1V6BHD+DRI+D0aaBoUZQuY0ASQu2ee+BAYPVq/rosWWK/rbZGVpctc6wMn62RaZH859u3gX79+O2ICJ6P62y2rpY0aiT+ufTU32mechXF3YSD2+joaKxbtw5FihRBtJ0x7XhXzR4ghHgNV12OI/qx95qpvaSv58RBpXxC0Xx1LTVsTfNolS65JybyuFU2x5gxYN064I03gIcPgVKleJHfmjWFczlXr+bPUyQ32tZExfh4YNo0dc9fEhlpP51G7ReZhATentBQ81xcOW++CXTsCAwYAFy/Lv4FJTaW58LqFdR54u80NZMqPa3tehKullC0aFEY/vvqW7RoUZs/hBBCfJuaqgJ6V4SwV5VBZKEItcGX6cgvoGGhi9RU4KWXgMGDeWDbqhXw009AzZoAbFdJkbaZ5nLa219qq2Vu9IEDvPRX167Agwc2n7KZsDD+nEQX3VD7RUZNBYXt23lg9t57/LbI4wwG/sXA11H5RU545Pbjjz+W/T8hhBDv4ugsajWX9PWeOChalUFuVM30ef/7r7rzmo78HjigcnTs3DmwHj1g+OMPZBv8cGXwDJT7cCL8A8zHl5RGnZVGSpX2B4ANG6wXh9Aycc5gAL75BnjhBXWvn5aatozZH7UFcvvW1vOXO3ZeGLGk8oucpjq3ly9fxoULF6y2X7hwAVeuXHG0TYQQQpxEj1q0ai7pm9bxlEYOP/uM/6s239ORpYEtn/fo0baDNYPBfHlb09FK0VGvnTv5vwnDYmD44w9cQ1k0ZQdQcc3bKF/Bz6zPpb5JT+eZC3v3ii1PbDlKLVU/iIoyf+5aa/eOGcMHmdV+MREZWXaE9BpIz3/KFHWP81XeVlfbWTQFtwMHDsTRo0etth8/fhwDBw50tE2EEEKcQK+laUUDhClTcgMze0G1SOCrdZEGpeetFFxLgcGKFfILLoiOesXF8cC+1vGP8DEGoibO4jB4VGHa55Z907IlnzgWFJS7tK6tfjFdzrhRI+vnqHXiHAB06KDtcQB/3T//PLeurZ5MXwN/fz7JT+3jfJEvLQLkCE3VEs6cOYOGDRtaba9Xrx6GDx/ucKMIIYToS89Z1KIBgmlVAVupBGPH8uBNrianaVu05BOKBHeWiyHYqz/cuLHtRQdq4TQ6YSdibsfi9deB2wjDYJin80l9PmyY/KX4hASeF2s5wUptrVItE+dMqRnptEx3SUy0ribh5wdkZ8s/3mDIXcFL7SIitGBMLl9ZBMgRmoJbg8GA+/fvW21PTk6G0ZG6IoS4gdEIHD7M/3/4MNCkie9/qyV5j56zqKVAwl7QlJgolkowf771fdeu8TJRn32W+7nUkk8oEtwZjbygvxSwhoUBISF8u9zvAn9/PqJrXYqL4Q18gIV4C0HIwG+ois23eyqe11aOqdQ3lveLrvhmNAJHjgDbtinvI0K0z0Vzem0FtkDuqKPahV9owRhznlqqzFU0pSU0adIEc+bMMQtkjUYj5syZg0aW10QI8WDS5cAXX+S3X3xRff6hVo7mHxKihl6zqKXROZHRnzFj7E++EiF9LhMT1ecTij7v774DJk7kubj9+innIkufW8vnVBT3sAXd8T5GIAgZ2ImO2CO40pga9nKLJdWr8+fw/vvazxUWJjbSqSWn1zLIknKzO3XiXyxGjgRKlJDfx1b+8datuaO/oo/zVabpKpapNb5O08jtvHnz0KRJEzz55JNo/N87/9ChQ0hJScG+fft0bSAhzmJ6qTQ4OHe76MiIo+fO00sjEpfTYxa12hn3//zDA0E9JCTw+rFjx/Ilb0VH50Sf95dfyp/T9HeB0vOvgx/xOXqiAi4jAwEYj3exFCNRpIgBSFH1NIXYGmXfsQMIDORtd1TfvvYDIq05vdJoecmSuaOKO3fyLxSm/VuiBP+y0amT2MhjXh+xJJymkduqVavi3Llz6NGjB27duoX79+/jpZdewvnz5/H000/r3UZCdOfIrGtH6TWphxA1HJ1F7ciMez1In8tNm/gkJdHROel5az0nYzwvNiZG/vkPwSocQUNUwGX8hcfREEewFKMAGJAvn+0+d5TlqPTWrbyMrl46dbK/jyM5vSVL5o4q7twp37937vAv/UlJ4gFqXh6xJJzm5XfLlCmD2bNn69kWQlzGXau40NKIxF0cyUl0ZMZ9s2a8tJWaeqdKpM9lWBgv/yQyOufvD7z8Ml+dSqs7d5Qf/w8iEYhMbEVXDMFqJKNYzn1JSbzfYmKU+7xQIUBmCosQ01Hp+Hige3fzq1COMp0IJrGcNObICLHUfvq9SPSmaeQW4GkI/fr1Q4MGDZDw37t7w4YNOCzNzCHEg7lrFRetpYwI0YPWnEStS9VGRvLg1l5pIrVu3FA3Oqf3FZhCyI1Gv0Fb1MdRdMcWs8BWUrmyfJ+HhPAfLYGt5Si7FByKGj5cbOndMWPM+06unNvrr6tqOgDr9tPvRaI3TcHttm3b0KZNGwQHB+P06dNIT08HwKsl0Ggu8QbuWsWFlkYk7iayNK0lR5aq9fdXDqpLlOA5tAaDukDXXbVKDcjGWMzHX6iA8rics/0H1Acg/wRKl7bu89hYPhpsazUuqTasSK1StV8+unYVuyJlGlAqpaWkqMwplms//V4ketMU3M6cORMrVqzAqlWrEBAQkLO9YcOGOH36tG6NI8RZ3LWKCy2NSDyB2pxEte9HuZHg6Ghg0SLzGfC3b/Mc2rFjrQNfOVo/l3qkFoUiEbsQhfkYjzAkoj822H2MaVulPu/RA1i1yvbjwsKA69d5GS+RUXY1QZ/UJjUBpaMLQZiSa7/MgqeyRN+HVImGaMq5/eOPP9CkSROr7UWLFsW9e/ccbRMhTmeZf2jKmTURqdA48UYNGvCgVC4HE+Dv2xIl+Oz3smXl81/j44GePeUXcliwgE8SCwvjE4uWLNH3c9msmfViCGo0wiF8ht6IQAIeIj9GYilWYajdx8m1VWSU9fZt4OhR8Zn/ar58SG1S80XbkUlj06bx/ldqv9FoP9gHzH8vWub9mh6TKtEQQOPIbalSpXDx4kWr7YcPH0aFChUcbhQhruCOmoi0NCLxNvHxQMWKtgNbQHmpWkCsOslbb/EgZfFi8RFLy3Mojdb5+wMrV9p+np06WVdVMCAbb2MWDqAZIpCA83gSdXEcqzAMSmkIklGj5Nuq9hK8yCi7vStR0nE2b85tk5qrVzt3irVZzpNP2m6/aOA8dGjuandKyzhTJRoi0RTcDh06FCNHjsTx48dhMBhw/fp1xMXFYezYsXjttdf0biMhTiPlwu3ezW/v3m0//1CPc1KhceINRMp/iQSdy5apmzCk5nNpNALTpwPh4fIBjyQ6Wj5oLlGCB307dvDA2jRt4jV8iFmYAn9k4xP0Rx2cxM94RvmJmFAqo+WM1CR/f952W2kDmzbxagqmjxH5oi2NpGtl73mIBvuVK9sPXocN07e8I6U3eDGmQXZ2Nps5cyYrWLAgMxgMzGAwsPz587MpU6ZoOZzbJScnMwAsOTnZ3U3RTUZGBtuxYwfLyMhwd1O8gjv6KyuLsf37Gdu4kf+bleWyU+uC3mPqeVOfZWUxFhEhVXqV/wkLYyw9XfkY27bZP4bpz8aN5o+311/btjEWGmr7mG++af75UvrcbdvGmMFg/thAPGIH0YgNxFoGZOfcX6SI7XP6+zO2ZYvtfrU8l/RjMDAWGanu94FpPwcH8z4LDs5gAD/Wtm3qXiPpMSLvAaUf0eexf7/Y8fbu1d4W05/9+63bIPc+k+uXiAjbfZlXuPP3mGi8pinn1mAwYPLkyRg3bhwuXryI1NRUVK1aFYUKFdI38ibEh0mXGwnxRGpyQ+Xex6YrAIpSM1opevz33uM/pnmXlu2V0iYMzIiBWI9P8BKMyIcMBKEJDkJKQShThh8LsH1uo5FPHJMb0Xak3rAce/2wcKHtq0G28nq1Lp2s5nmIzkMA9FlARGSkWKlPXbF6pTPYylH2VZrr3DLGkJKSgpIlS+L555+nwJYQQnyII+WZtMyuV1MFQcvxr11Tzrs8cADIunYDe9AKa/EypmK6yb251+z/q3qJ6Gg+Ac5egKB0GVyv1CR7/WAw8Fxme5fTlfJ6RRdosPzzr+Z5iKZH3Lol1hZ77H2Bcufqlc5gK0dZLSlNY+vW3NueSnVwe/PmTbz00ksoXrw4SpYsifDwcBQvXhyDBw/Gv//+64w2EkIIcTE1uaGWuYlaRvx69RIfTdI6e58x68AkPh5Y3mUPzqImXsB+pKIg/sQTso9PTMwNkMPCbP9xt7fwgJZ6w5b0WvxALrc0Pp73lYgdOxx7HiLBvqPlEUXLyPnSghJ6TbCzzGt/+WW+vXp1z52kpyotISUlBQ0aNEBqaioGDRqEKlWqgDGG3377DZ999hkOHz6M06dP0yguIUQ3el5Sy4uX5+xR6hPRy8WJiXwkyPQPaEiI+nZs2gTMmSP2ejhSzN90We3tW7Lwe48YbMFs+IHhHKqjBzbjD1SxeYxRo3hbRdhqq6OpSXosfiBXOktN2TTpvRIYyG9r/YzZK3sm8n4MCeHtdiTdw1cWlNBrSeP4eD5RT+79cP2656ZpqApuly5dCn9/f/z6668ICwszu2/KlClo2LAh3nvvPbz99tu6NpIQkjfpWbOS6l9as9cn9nJDe/XiuaWWf0CTktS3xTTotGQ0AkeO5AY94eHqj2/qxg0g43ICSvXpjS7gQ3Ar8ApGYzEeIdjmY6WRu9u3xc7lzAVZHK28oJRbqqYesNGYm3ft6GfMVrAvkqsslXuTa8OSJWJt8JWFdtSMQCv1ub18btM0DXtBsqupSkvYvXs33n77bavAFgDCw8MxadIk7Nq1S7fGEULyLj1rVlL9S2sifWLrcvHnn/PL2HqsWiVRGg2rXt08Z3DgQD66qGbJXlMXLgAtat9DjayTSEFh9MJneA0r7Aa2pkJDeWqCEmetcmhKJMBWaoOeq47duOGaz5hI+oKj6R7uWr1Sb46OQIu+Pzw1TUPVyO2ff/6JBg0aKN7foEEDjB071uFGEUJ8i9pLlXpdUtP7WL5CTZ8oXS52ZNUqJZajYbt28dfEcmKT6aVpy1E82xgKFTIgJgZgrBp64zP8imq4hEqq2zpiBJCcLH+f6WVwgOex3riRO+J865Y+KTZjxtjfb9Ei+XPo+fqFh/MvHK74jIms2uZIuofe1SzcxdERaLXvD09L01A1cpuSkoJixYop3l+sWDGkpKQ42iZCiA/RMltXz0kdvjRBRC9q+0RuNr3oHzOR/Fu50TCjEZgwQbl9AB89LVNGrB2R+BsH0Az1so/mPP4LdNIU2ALKgS2QO5IImL/3W7bkP47OWgfEgw/TRSlM6RWMSKPXen3GRBZOEFm1zRG+sNCOoyPQat8fnpamoSq4ZYzBz0/5IQaDAUzPa1SEEK+m9VKlnpM6fGWCiJ609olp4CFaHGfzZiA21rpclERpNOzQIfulqO7cAUaPBqZNs71fFL7AWdREUxzEwrRXYUC2WOM1KFECkFant7e6myOX7B19X+sVjHzwgXiZLntt1rNslaP0qGbhTo4u9a7m/eGJaRqq0hIYY3jiiSdgUPgqQIEtIURKQUhI4IGHlkuVek7q8JUJInrS0idyk4X8/ZXLYUnVFJKS8F8agPx+ISF8IpBl0CBaY3XsWOU/0AHIwFxMxBgsBgBcLfkcOv+7CUxwXEdN1QBJYiIwezawZo1YvqLp5wAQT99x9H1tr/qAiHHj+JK+Bw6I7W9rIqAnLpzg7QvtSCPQWibYqXl/eGKahqrg9uOPP3ZWOwghPkAuAFJia7auaBkqkdECPY/lK9T2iVLgYSuwBXi+p9IXHElwcG5gZ0q0GoFSO8rjMj5HTzyPHwEAXz0xCvmXzsPldoF2j1eiBPDRR7xdy5bx56BGbKz4vtLnYNYsYNUq25UGpJHzAweA7GzezsRE+ePau+xsL7eUMeXgPiwMWL6cPxYQD4QGDOArvFkGVd6WF+9NJQVFcpTl2Hp/SEJC5F9PT6AquB0wYICz2kEI8XJallsF5C9V6jmpw1cmiOhJTZ+IzJq2HMGVRoZCQux/0bl2Tf4Ljq1KBPZUxEWcRB0UQzKSUBwDsQ5H73TEZ4Kv8YQJuYFUyZLa26GGXGqF6agloFxvVA5jwJAhyvcbjfz1GTkSiIsz/zIhvX5SUJSQwO8PC+N5qHITuOwFQoByXVQ9yla5ijeWFNQ6Aq008ivl0V+8COTPr0sTdSecc0spB4QQJY6UFVK6bKrnpA5fmCCiN9E+EZm4ZDQCixdb5yaK5oXKpSBYtkuNS6iIfXgBx1APtXAGu9ARd+4ABw+KPX7cuNxcT3emq0ifp2HDgK5d1adITJsmn7Nqmtu6ZAkPXEuU4KOjpq+fFBT17cvv69tXeQKX0vtJ7vlYrhLnLXnxebGkoFzusZRT7skDAsIjt9WqVcPUqVMRHR2NwEDlyzoXLlzAokWLUK5cOUycOFGXRhJCPJuWskIi6QBaL6k5+1i+QqRPRAOKkiX57HVTooHha68Bly4Bkyebr0ilJsCtgEu4jTDcRxEABgzAejxCfmQhQPwgJqSAZfNmx3NTHcGYWFBbpAggV6zIMmfV1sINS5c69pmIjgaKFuUVIZTIjcJ6Q168t6VO6Mly5Dcz021NESYc3C5btgwTJkzA66+/jlatWqFOnTooU6YM8ufPj7t37+K3337D4cOH8euvv2L48OF47bXXnNluQogHUTuioiYdQM9JHd4+QcQZ7PWJI4FHgwa2J51J7t/no4zvvZc7uczfH5g3T+zcPfA5VmEovkJ79MZnAAxIRWGr/Zo1A9atEwtUpYBlzBieO9yjh1hb1FBXo9c2pSqcpoFXhw7Am286N0DTUjnBG/LivSl1gqhIS2jRogVOnjyJL774AuHh4YiLi8Pw4cPRt29fxMTE4MKFC3jppZdw7do1zJs3D0WLFnVmuwkhHkTtiEpeTgfwNo7Uyzx61H5ga+rOHX75Xbq8GxXF/1Uawc2Ph/gQr+Jz9EIR3EdZJKAgHii2sVkzPjopGlBKActXX4nV61WrRAl1k8+0kp7Hq6/arkKhR81nLV+GHC1b5QrekjpBOFUTygCgUaNGaNSokTPaQgjxUiKzpcPCeF6m3IQU4h4is74dmZCn9Q+9aWksAPj5Z+Ddd4H584HUVL7tCfyBzeiBGjiHbBgwB5MwDbEwWvxZs2xjp07qS3ytW6ftedizeDHQqxfw/vvqqkNoJVrwyJEATesorCNlq1zBG1InSC5VizgQQogceyMvBgOwYoXtCSkiRFYv8kbueF5qCuZrnZB34YK2tlmOHn71Fa+VKwW2ffEpTqE2auAcbiEMPQp/jae2zcLmbfkQEWG7jYcOqZ+Y5SylSvHPQt++7m6JOUcCNEdGYT154QRHV/wirkXBLSFEF86uSODM1YvcGTS7Y1UmLbO+1QYeRiOv26qV6ejhhAm5o4CFcB/vYjwK4QH2oTlq4Cfsy9caRYvyUVnTNu7dy0cr09NzX1dPvGwsV+dXrdBQ5cBLDT0CNF+sTuINqROu5tGDDcyNvv/+e9ahQwdWunRpBoBt3749576MjAw2fvx49vTTT7MCBQqw0qVLs/79+7OEhASzY5QrV44BMPuZM2eOqnYkJyczACw5OVmPp+URMjIy2I4dO1hGRoa7m+IVqL/UU+qzrCzG9u9nbONG/m9WluPn2raNMYOBMR7i5P4YDPxn2zZ1xzNtY2wsY2XLmh83IkL+mI4+N8s+0/t5icjK4s/P8pym546MdPx1279f+RwiP/v35/ZXcHCG2X1NcIBNwzTmhyyz7WXL8tfT1usaG+tYu/T82bhR/DUpVMj2scaN4/sVKCDfZ6I/er7n1H5etm2z7gelz6KepPfZw4cZdtsr18bISOe30ZNI/VWpUobLXyvReM2twe1XX33FJk+ezOLj462C23v37rGWLVuyzz//nJ0/f54dO3aMPf/886x27dpmxyhXrhybPn06u3HjRs5PamqqqnZQcEuov9RzVZ/pHYzJ/XESCS71+MNr2meuCjItiQad+/c7dp6NG7UFfKbPOyMjg+3Yvp0NC1jFeuIzh4NJ6XUNDXV/YGvZx9IXHcsvO9Jte22OiGBsyxbGKlXSHtzGxtp/XZ3x5dX0+Yt8FtWy12alYC0khPeJ5f7O6gNvsW2b/HvMmV/KJV4R3JqyDG7lnDhxggFgV69ezdlWrlw5tnjxYofOTcEtof5Sz1V9pmcwpvQH1F6QpdcfXtM+c1WQaUk06JRGFbXSMnJr2Z8ZSUns72bNGAPYfRRk5XDZ4YDSVcGtrfeZ0hcXpVFB0dHm2FjGHj7k77EyZdQFtxER5u2RC+CcNbLqzC96Im1WCtakn9DQvDUya0tWlu0vUM76Ui4RjdeEqyWkKBXRk1GkSBHhfdVITk6GwWBAsWLFzLbPnTsXM2bMwGOPPYY+ffpg9OjRyJdP+amlp6cjPT0957b03DIzM5HpDdWJBUjPw1eej7NRf6nnqj67cQMIDhbbz1ZTjEaeu6lmucjERJ6/aetxBgMwcSLQvr39fDvTPtPrealVqpTYeUuVUn9eoxE4dgy4eZNXx6hUiS+5ypj8/pbVFyIigLlzeQmwzFPn4N+3LyL/+ANG+GFBwHjc9i+NYIPjnZGWBsyYwSc5JiWZ3+fnB2RnO3wKdOkCbN9uvd00PzM72/xcUVH8fST1YalSQP36wJQpYq/Z3LnAgweZaNAAePQoU+gxUnuWLs1tz65d/D1vWjYsJCS3r0yPm5QE9O+f234tDh/mk/xstTcxka8wp6ZY065dvG2MKbe5fXtg6tRMzJoFBAfLv7fS0oB+/YBPP819jqbvdel1Esm31fo4T3H4MJCUxPtJqb+0vFaiRP/eGBhT+rVjzs/PDwY72eqMMRgMBhg1ZBUbDAZs374dnTt3lr3/0aNHaNiwIapUqYK4uLic7YsWLcKzzz6LkJAQHD16FJMmTcKgQYOwaNEixXPFxMQgVqa44MaNG1GgQAHVbSeEEKITxlDu229Rfc0a+Gdk4GFICE699RbuVKvm7pYRQtwsLS0Nffr0QXJyss2BVOHg9vvvvxc+edOmTYX3zWmIjeA2MzMTXbt2xbVr13DgwAGbT2jt2rV45ZVXkJqaiqCgINl95EZuIyMjkZiY6LRRZ1fLzMzEnj170KpVKwQEaFt+Mi+h/lLPVX1mNALVqyuPABoMfFb2uXO2R0C2bgVefln9+ceN4/VV7VmzhlcasMW0z/z8AjQ9L7lRtbJl+WpeoiNn0ogWYH5uafxiwwaxY0mjULt3A8uXyz8HxsxH/QDzEVoz2dnwHzgQfps28eO3bo1v+/ZF027d8PXXAVbP2xG7dyuPLMn1caFCfFQzLU38HBERwJkzwIkT2kbqpPe+muccHJyJtWv3YPDgVnj40Pbncs4c4JVXctuj5Xymdu8G7t5V//48fBh48UWx44uOBooec9w44P33xfvs7bd5v1l+Zu19dkxHkdU8ztMcPgx062a/v9S8VmqkpKSgRIkSdoNbOCcrQj1APuc2IyODde7cmT3zzDMsMTHR7nF++eUXBoCdP39e+NyUc0uov9RzZZ/ZmmwjmvO6d6/6nMnISPHHieTGKlVLEH1eek66cXTWt8jEPKltERG8H4Um4Iwfz5i/P2Pz5rGMR4/M+isrix8nJMSxXFiRnMCsLJ7D6si5RN8XSrTkLQcH259QZtkHUn7tlCmOPdeuXZXPZ+v9KeXcKuUpa8njFM0tnzJFrM+kH1vvB6V2umvyqDN4S86t5jq3hw4dQr9+/dCgQQMk/PcVbcOGDTh8+LDWQ1rJzMxEjx49cOHCBezduxehoaF2H3P27Fn4+fkhPDxct3YQQtzL1XUzTfMimzWzXbxdaoeW2qBqnpfRyFdvYsz6ONK2UaPEa006UjBfqU6uHMb4fv7+QO/eMot4MAaYzumYOZMPdY4fzxNgTfj7Ay1a8Pq50uIgapi+rgCvzRkXx2/HxZnX6ty5ky8cYZmTq9bOndof64yavJY1WU3rLM+c6dixt22T327v/emMGrKiC1E0a6a8vLMcW+8HxuSXLz50yPZnRelxnsjfn4/CA55d71dTcLtt2za0adMGwcHBOH36dM4l/uTkZMyePVv4OKmpqTh79izOnj0LALh8+TLOnj2Lv//+G5mZmejWrRtOnjyJuLg4GI1G3Lx5Ezdv3kRGRgYA4NixY1iyZAl++ukn/PXXX4iLi8Po0aPRr18/FC9eXMtTI4R4KEdXL7p1S/xcpsGlrT+8krQ07UGM6PNyxh9If3/+x1026FRgK8i2RTZQS04GevQAWrcG/vu9joAA4NlnbR5L6UuBPdLrCuQGdP36AaNH83+lBTS2bNH2HOUsWaJ9QQ5nLOVq+t5W8yXFUfben3p/gRVdUaxZs9xgzRaDgS+UIcLyvS76JcUTFxiRI6VPlCljvt2jFunQMixcs2ZNtn79esYYY4UKFWKXLl1ijDF2+vRpVrJkSeHj7N+/nwGw+hkwYAC7fPmy7H0A2P7/rvOcOnWK1a1blxUtWpTlz5+fPfXUU2z27Nns0aNHqp4PpSUQ6i/1vK3PRC/xLl6sXLzdXgkpe5f0HekzV5Xwskfr4gxWl+dPnGDs8cf5nQEBjH3/vdW57PWX5WIcciWfpEUdTMtZiZaD0+PHkcu09i7Xy/3IXWJfvNg6JcTepXJn/dh7f+pZQ1Y07Ud6n5UtK5+WIO0vWpLN8r3urrJ/zqJm0Qu9ObXObXBwMLt8+TJjzDy4vXTpEgsKCtJySLei4JZQf6nnbX3maF5fVpb94DY01PYveEf6TDT3d+9e1YdWRe3iDFb9mp3No62AAL5D+fKMHT8uey61/WUvMHJXQOdI4KIUoCn9mAa3tt7Tar+kiC4m4WkBnEhuuWmwJpdnLe2v9XeIM3KK3cmdv/udmnNbqlQpXLx40Wr74cOHUaFCBS2HJIQQp3I0r+/AAV6L05Y7d/h+3s7WmvFqLpVb9WtSEtC5M88DyMzk1y/PnAGef16XdgK20yzspXY4k9ZLzrYu18fG8jxWOfbe02rTaCIieE7tv//mptBMmSL+eCkNQEtuuiPUpDP5+wNTp/IUJrn9tf4OcUZOMbFNU3A7dOhQjBw5EsePH4fBYMD169cRFxeHsWPH4rXXXtO7jYQQogtH8vpEg1ZnBbeiOcNqcovlmE4w6tOH/1u6dG6uqr1cRlNW/Tp4MPDFF0BgILBsGb+zWDGbwbSadpYvbzu/1Z05jY7kz8oFaFeu8EBs8WJg82agRAnzx5Qtq/yeNhr5BDoRU6ZYB3jSF4gWLdQ9D3cFcGpzy23tr/V3iKsnxeZ1wiuUmZo4cSKys7PRokULpKWloUmTJggKCsLYsWMxYsQIvdtICCG6iY4GOnXio3g3bvCgo3Fj81qfSve5k2hw5EgQJU0wYsx8++3bQPfuvCbou+/yUahu3axXGJNujxrF+9iq7+bPB/7+G1i9Gnj2WWRkAK++yidwpabm7hYRwc+hVPdTqZ0JCXy7UrCg9wStyEhg0SI+EJ2QYN0egPeJ1moapqSAy1J8PDBmDF8VytTs2coB06FD/DW1JyyMV4xQev9LX3SUnrtEej19JYCz9ztE78cRDRzJfUhPT2e//vorO378OLt//74jh3Iryrkl1F/q+WKf2VqHXo+cV0f6TCRfVE3enmV+anq6WD7q5s3KfWVVJ/f2bcY++8z8xNnZjDHGxo1jzM9POQeRT/ix7i+RfoiIkO+HrCzGwsIcyxktUoSxUaPMc3ptTVJTW39YDbnzSjm3BQpkKJ5XNG961CjxNig9/9hYz88l9cXfZc7kszm3n376KdLS0hAYGIiqVavi+eefR6FChfSNugkhxIWUyiJJo4FJSfZLAYWGyo+uKcnI4JdqR4zg/0rVsORIdWJt6dVLbBRI7pJ+2bJi+ahvvMFHt+3mMh4+DNSsyU+wb1/uAQwGjB/PB3Gzs+XPIY0CTpxofZ9I3uy1a8CsWdbb/f2Bvn3tPEE74uN5KoDp5eroaGDsWPlcy7FjnTNiKVKSTamurOgIdqdO9vdRutweGclzdKdOpZFJ4nqagtvRo0cjPDwcffr0wVdffQWjaNVwQgjxQCILJLz1FrBihe3jrFwp/od8/HigQAF+Sfv99/m/BQrw7Upt/Owz28fctMl+vqpSEG95WVvJ7du8r+LieKDZuLFFbmJ2Nl+ftFkz/s2gcmWzbwUZGfxSvj2MyQexonmz06bJ59+KBGxyTOuiWoqPBxYssO777Gy+XWudW1scqXssWgNWNJXC0RrUhOhNU3B748YNbNq0CQaDAT169EDp0qXxxhtv4OjRo3q3jxBCnE40UChRgo9GRUSY3y/NJBf9Yz51Kh+5tAyGjEa+XS7AFRmxtLeIg9YFGCx98IH5ogc5wdutW0C7dsDbb/OT9esHnDoF1KiR89jly8VXUZOjJm9WbuRSCuzUkILAIUP45C3TSW96rxwnypGFAZwxe1/LgiCEOIum4DZfvnzo0KED4uLicOvWLSxevBhXrlxB8+bNUbFiRb3bSAghTqUmUFCaua5mlOr9923fv2iRdYqCHqscOaMU1rVrfCT44PQDPA3h22+B4GBgzRrgk08Ai5S1S5ccO5+a4FQu2JcCOzXL94aE8J9p06wrM7hraVVHJxjS7H3iyzRVSzBVoEABtGnTBnfv3sXVq1fx+++/69EuQggRokd1A7WBgtLMdVH2RvGMRj7CaVrDVI9qCc4qhcUYsGvJJTS5ewN46ile/qBaNdl91Yx/yAWxUnDatavYMeSesxTYjRwpFuzL1TeWcrFHjtTeDomW97C9SgUiqQXOnL3vyOfSUyuWEO+haeQWANLS0hAXF4f27dujbNmyWLJkCbp06YJff/1Vz/YRQogiLbVO5eidg6gHyxFOPdqodyksIDeqWnB3MP54ayXw44+KgS0AvP66WKBiMABz58rfFx3NFzAQYWvk8soVYO9ePiqrlhRQitaLLV1afnEMre9hW6kFEpHUAumLWo8e/LZl2oUWjnwu9fpMk7xNU3Dbq1cvhIeHY/To0ahQoQIOHDiAixcvYsaMGahSpYrebSSEECv2qhuo+WPoiSsIWY5wqmmj0qIIt2/bfg5SgLxpk/WiAJZaYC+Ooy6KI0l6NE7XHgoULGjzcYGBQIcOto9duDAfWVWqcwsAkyfbTk8QCfb9/flPUpLyPrYwxvs0LMz+l47bt62DtpIl+Qi00nt4yxbbi1sopRYAwIYN4qkFegaUjnwu9fxMu5OWRUmIzrTUGevTpw/bvXs3y/L04nWCqM4tof5Sz519Zq/Wqda12oVqtzpA6rNChTJs1hf182MsLU1bG5Vq9Y4bp1yLVK4ma3o6r+lquY8/Mtl0TGFG8IMtxOic+/bvt98HtmrCAow1bJj7utl7jynVWM2tk2u/PaI1X+3Vg7XVDpG+l/vx97d+HeWek2nN4n371H0ulV4PNX1o2g6tn0tnfaZF6Pm7zFatbF/hDXVuNQW3ph4+fOjoIdyOgltC/aWeO/ts/36x4EAk2LJkubiBnn9MpT6bONF2cGvvD2JWFl8sYsoU/rN3L99mK1ARCaS2bMk9h1wfl8E1dgBNcjaswDCWH2kMYCw01H5fqQ1gRN5jjn4hEX0v2XufKbVjyxaxxTFEfkQCTjWfSz0DyqwsxhYv1v65dOZn2h69fpfp+UXBk3lDcKspLSE7OxszZsxA2bJlUahQIfz1118AgHfeeQdr1qzRbVSZEELk6FE5QIkrShpNn86XsrV1bFuXYnfuBAYOBGbO5D8tWwLlygHDhvE/p5bktlkyGs1TESz7rg2+xlnURFMcxH0UQi98hlfxER4h2P7B/+OMygKO1ljVUhpMYpr6oNSOEiX0q1AhvY56lRbT6/WQ0hpGjxY7r9zn0pmfaVdwV0k4Ik9TcDtz5kysW7cO7777LgIDA3O2P/3001i9erVujSOEEDl6VA5wt3ffBe7dU05RVfqDaCsvUW5WvxqmgYNp3/VBHL5GO4QhEWdQE8/iND5HL7PH3rkDLFtm+4+3swIYR76QmOYyqyGXiy3XDr2DMS1fAJTo8XoovR9tkftcevtn2l0l4Yg8TcHtJ598gpUrV6Jv377wN/ktUqNGDZw/f163xhFCiBxPrG6gVnw8UKEC8OCB8j6WfxD1WoRBiWngYNrH/0M7XEE5vI83UB/HcBGVZR8/erTtiUieGsBER5uXXRMhWg/WWc9Fj6DZ0ddD7fvR1ufS2z/T3j7y7Gs0BbcJCQmoVKmS1fbs7GxkZmY63ChCCLHFE6sbqLFrFx/tun1bbH/pD6IzFmEA5AMH/zMnsXQJj1ruIgQ1cRYj8D7Skd/msWylU3hyACO6LO+UKepSHxxJe7BFj6DZ0ddDzfvR3ufS2z/TnvrFLa/SFNxWrVoVh2TG1rdu3YpatWo53ChCCLHHm1dYmjBB3eir9AfRGaM+VoFDRgbw1lvAc88hOmk1tm7lfZqMYkLHs5Vf6MkBjGigFxOjLvVBS9qDSLk2Pb4A2GqbyOuh5v0o8rn05s+0J39xy4s0BbdTp07F8OHDMW/ePGRnZyM+Ph5Dhw7FrFmzMHXqVL3bSAghshydTOQuCQli+1n+QXR01Cd/fiA01HybWeBw+TI/2aJF/M6//srpY9FFEwD5/EKp9md6Og8QPS2AcWbgHR3Na9b62fmLGxLCF5XYtEl+eWBnfQGQW8QiJMT+6yH6fly8WPxz6a2faU/+4pYXaVp+t1OnTti1axemT5+OggULYurUqXj22Wexa9cutGrVSu82EkKIIkeXwvV0pn8Q7S25as+jR/wnNhaoXNliadPt24FBg4DkZKBYMWDdupxr9f7+wNSpwNNPiy9ZC+SO7MXHWz+ubFmFdgiyt0SrliVclZbljYjgr4PWAMtoBH77DQgOls+xloKfVauAFi34/53RDkvSZDC595LI5ESRJYAjIoARI9RP8vPGz7Sz3j9EAz3rj929e5fFxcXpeUiXoDq3hPpLPeoz9aQ+Cw62X+c2LEy+LqbSwgVqfsxqlz56xIzDR+Tcea9qPZZ16Yric1Bbz9SR2p9y77GsLMZiYxkLCTE/nmldYEcL6etZ63jbNl4D2FY/hYbaX5xBtB0PH/I+27gxw+Zj7NW4lfrM3jn1WEjD3fT+XebMWtmewBvq3Ooa3J49e5b5+fnpeUiXoOCWUH+p5yl95qo/JGrPI7e/1GeVKmXYDE7DwvgKYUrkgjd7AZRSMfz9s4+yLPgxBrB5GMfyIcNuICgFRkrPQSr+n57u2CIBlu8xW4GivdXA3BFsbdsm9lqIBJGi56tUyfwLVIkSfAU1y/es6KIJsbFi53Xmyn7O5im/y7yFNwS3mnJuCSHEE0jF45s3B/r04f/aKkXlqvMo7b9rF79/9mzYTCsYOBAwKSFuRS4v8fPP1T2nnTt5O1+YXB/jMB8v4ktMwLvIQoDNigeAWH7hokXA8uX61f6Mjwe6dlW+XC7156JF8n0rbXNVIX2pTJaIa9ccr38qpRhY5nMnJvJL4pbvWdHJYNOm2f88eWueLPFdFNwSQrySrcUMbAVmzj6Prf379+f/f/tt2+dcsMB++y0XDGjWTKzkVH48xGKMwu4lf+asaLYYY/AVXszZRyQQtDWzfexYXvPWkRWrTIkGiozZDlzVBNOOUlu2zZFKGKL1Zk3fs2omJ4p8IXDFyn6EiKLglhDidVy11KXa89jbX9ouUi1BbftFSk49gT/wA+phFJZiE3oh6U624r4igaDciN3ChTw4d3TFKlPHjulb39cVhfTVnuPff4HPPuMVJdS+b0UDadP3bIMG4vV3aWUt4m1UVUt47733bN6fIFrfhhBCHKBmqUtHZl2rPY9eiyxIx122DChZUt1s/23bgAEDgNRU8/v6IA4f4RUUwgPcQhgmYi6YwPiGZZAmVSFISOCLUISF8dHbHj34/eXL2x9BlEiz6e3V/rx5U+x4osLD9T2eHDUjo/7+5qPcERH8i4roZX01gbT03jp6lJ+ja1f9z0GIu6kKbhcvXmx3n8cee0xzYwghRISrlrrcuVPdefQOALQEPJ068UpeUnAbjDS8hzcxBGsAAPvRDH2wETchFn2ZBmlyJb1M2zd0qLrgnjGx2p+lSokfU8SAAcB77zk3J1RN2TbLkVopfUC07q+W+sc3bvAUgthYnlfrjHMQ4i6q0hIuX74s9EMIIc7kiqUujUbg00/VnceZAYBoLrHp6HEZJOAEnscQrEE2DIjBNLTEXqHA1nIBCaVcYsm1a2JBkhb169te/Umt69f1zcuWY2vSnT1qU2vsrY4lR3qvTp5sOz2BVtYi3ohybgkhXscVS10eOsRnmtsTFpZ7Hi1BhijRgMd09PgWwnEfhXETJdESexGLGGTD/kwfyxWVRCcsaTF0KPDdd7afkyOBohwp/3nkSOdWTlCadFeokP3Hqpn8pmaJX8vPhvRYV66IRoizUXBLCPE6zlrqUloi9rPPeMAl4uHD3PQFvYMwSyIBT9liD5APmQCALASgBzajJs5iP15QfIxlPxUvzpfI/W+BMt1yieUkJQEtW9ov4aYUKBYurP3c164Bs2bx/5u+9lomdSmxnHS3dy9PGxElmuqi1D+mlD4btipfuHNJZEK0ouCWEC/nrD/Knk7vP8iWtWlnzhR7XGqq+SVukSDDUYoBz88/o/GYOlhW6O2cQOYaIvEvbCetGo28tm5ICL+dlMRTDKSAUzT32JTa4F4k7UKuOkNiomOj5dOmAePHO7desmmZLH9/fStJmIqOBn7+mf//9df5VQVTtj4bVKuW+BIKbgnxYq5axMARzgy+9fqDbC+fVIRpuoDUrr17cwNGW0JD+eMF5uwCkAl4GANWrwaefx6G8+fxUtAmFGHJqgK+det4UGsqIYHPpl+7Vvw4AJ+kpDa4F027MA0UGzfms/67dXMsZWL+fOfXS5aomXSoJbVGGpGdM4efS81ng2rVEl9BwS0hXspVixg4whXBt6N/kPXIJ5VLF/D35z+WAaMcg4EHMSNGaMglvn8f6NuXJ68+egS0bYsCv5/G2m1FHR49lvokJUVsf6l9kyerC+5Nz6dmxTLpvbVkCd/mp+NfNGetaKZmJNbRXFcKVkleJfyrICUlRfiHEOJcrlrEwBHeEHwD+uaTWo7KiY7SJSbyPtm503bOLmN8edmcIOXsWaB2bT4s7u8PzJsH7N4NhIWpHj3WixSQ+fsDLVoAq1bJT1ayxV6/Kb23spXXpNDEGSuaSZMObfH3BzZvppQAQrQSDm6LFSuG4sWLC/0QQpxLzeIC7uANwbdEz9q0lqNyakuDjRrFJ3HZytkdPfq/LwZpaUCrVsCFC3y49OBBnjxqMnzpSICpVliYfD6nlhxkW/3mzMoNSvR8j+zcySch2rJpE9C9u+198mquPSEihIPb/fv3Y9++fdi3bx/Wrl2L8PBwjB8/Htu3b8f27dsxfvx4lCxZEmvVJmcRQlRz1SIGWnl68G1KNAAtUUJ96TE1pcGkPomJ4SOt774rv1/OyPfXBfhKBFFRwJkzfD1VBc6e5BYWxl9vpZFG0VFkkRJuzqzcoESv+sXSiPOdO/L3h4byFea6dbO+zzSYnT4dKFfOPN2ndGn+GhNCVKxQ1rRp05z/T58+HYsWLULv3r1ztnXs2BHVq1fHypUrMWDAAH1bSQgx44pFDBzhjOBbWvb1xg3x5WhF2FtJSloiduFCoGdPftt0P1ulx6TSYHLBipKZM5UrNdTGSeRnj3DU0IiP8l7uDf9evYSi5+hoPiq8bJn5ymd6WL4cCAy0vY/pKLLUH6L9aErPL2yWr6Xc/SLLA9tiulzxqFG2zxccnFt+zZStleEkt2/z0d6JE4F69bS3lxBfoCn9/tixY6hTp47V9jp16uDEiRMON4oQYpvIiGBYmM3BPKfSO/h25sQ00Zq53btrKz0mjZqWKOFIKxnexFIcRQNsQXeEsX9zR75V5Br4+wMlSzrSDnlqnpujJdxE3zOWZbBCQ/mP5TnHjXPeAgam79t+/ewvCnLtmvXVDLWVPOwt5kDpDCQv0BTcRkZGYtWqVVbbV69ejcjISIcbRQixTWSxgNu3gYoV3TNxS88VxJT+uF+7xstUSc/PkT/aogGX1tJj0dF85E5LgFsMdxGPaCzFKAQiE0fRAOkIAqBtFFPNaL5lMKhEbTscKeEm+t66ds38+P/+y38sz/nuu85ZwEBreTnTvnQkv1ju/e8NpQMJ0YNwWoKpxYsXo2vXrvjf//6HunXrAgBOnDiBCxcuYNu2bbo2kBAiTwrIbF2ulPIzXb3KkOnleDWX8S1lZACvvGL7j/uwYXyW/OjR5v0QEcHbIPq8pcv29lIfpPJKagUG5pasEh1sfR7H8Tl6ojyuIh2BeAsL8QHeAMAPIBKoWqZzNGhgOw0D4Hmxmzfz/7dsaf8cWtJftPaj6HsrMFD++HLbRF97UY4EpaZ96Uh+8bFjPHiVSMG2ZZvc9TuCEGfSNHLbvn17XLhwAVFRUUhKSkJSUhKioqLw559/on379nq3kRAiw2jkQcjs2UDRovL7uLMygaOXn+Pj+b72LuXeucNTBvQoOebsuqBRUfzfMmXs7ckwBgtxGI1QHldxERXRAEfxAYZDCmxFRr7lRuoqVuTPD5C/FG8w8LzYFi14H+g1Aq8nZywXq+drrzUoDQsz70tH8otv3sz9vzdVLyFED5pGbgEgIiICs2fP1rMthBBBIhNMJKaVCbSMlDlCbkSsQQO+qtRnnymPkG3dar8Ukj2M8eBLKq/ligL2opPefv6ZXw5fsICvwSCnNk4hAFn4HD0wDCuRAvNvMPZGvm2N1C1YAIwdy18Dy9HuJUtyg0O9RuCdQe/RVj1pDUr79jVvvyMTQkuZrLispnqJq39HEOIMmoPbe/fu4cSJE7h16xayLSpnv/TSSw43jBAiTylosUePWeamwZvpH09bTC8/x8fzkUNb6QNbtuSOLDrKGX+0lQJYuS8cSqkRX33FS35Zv4YMfGTWgFfwEb5GW2xAf0ijtQA/12ef2R6dtDdSZzDwWqqXLvEvGraCQ6X0F8tA2B20pjY4m9ag1LJSgr1KHrbUr5/7f08vHUiI3jQFt7t27ULfvn2RmpqKIkWKwGByzcpgMFBwS4iT6JXLp4Vl8BYczIOsXbvEAhyRnD8A6NHDsXbK0euPtlIA27s3Hw0VzWecMMFiFBTZGI93URun0AObARiQisLYAOvfpZs22S8tJjpSd/SoWHDoyaOknkhtUKpUcszWyLmtY0mPlXh66UBC9KYp5/att97C4MGDkZqainv37uHu3bs5P0kiC6kTQjTRksunR16krZnf/fvbz2sVyfkbORJ4803tbbRFjz/atpYTnj9fXT5jQkLu/8NwC1+hPeZiErpjK9riawBmC40B4K+hUoF/S84YqXN2PrIvEalmIrGX3qGUXxwaChQubL4tMhLYsMH6GHpWLyHEG2gKbhMSEvDmm2+iQIECereHEGKD2hFIPfIiRUaL7U1GERlJvHbNPOjTg15/tEWCcyW2VmNrgu9xFjXRFt8gDcEYjDX4Gm0BAOvWaSuVBdBInSdQCkotP4cik+DkSqf9+y9w9671e0SatGh5TpFazvSFhfgKTcFtmzZtcPLkSYdPfvDgQURFRaFMmTIwGAzYsWNHzn2ZmZmYMGECqlevjoIFC6JMmTJ46aWXcP36dbNjJCUloW/fvihSpAiKFSuGl19+GampqQ63jRBPpDYYcbRWJyB+iTsmRrm+rCty+Zz5R1uPJV9N+8CPGfEOpmMfXkAZ3MBveArP4Ud8jMEwrYagdaSURuo8g1xQmpam7UuL3Mi5mtF0Z1SYIMRTacq5ffHFFzFu3Dj89ttvqF69OgICAszu79ixo9BxHjx4gBo1amDw4MGItvhkpaWl4fTp03jnnXdQo0YN3L17FyNHjkTHjh3NAuu+ffvixo0b2LNnDzIzMzFo0CAMGzYMGzdu1PLUCPFoIkvFligBLF7M/4jpkRcpGphKy8ZGRABDhwKVK+fmZrpihNDPzzyw1nPCkx7BuWkfrA94BV0ffQIA+BgDMRzvIw0FAeiz5Kuzqhw4awlkXyY36c1dk+Aod5rkGUwDg8Gg+OPn56flkAwA2759u819Tpw4wQCwq1evMsYY++233xgA9uOPP+bs87///Y8ZDAaWkJAgfO7k5GQGgCUnJ2tquyfKyMhgO3bsYBkZGe5uilfwpv7ato0xg4H/8LCF/0jbtm3T93z795ufR/oJDuZ9FhycIXu/9BMRwdjmzfxfyzabtj0igrGyZZX3ARjz82OsQQPl+wHGRo3ibc7Kcn4fiPwYDIxFRvL2SO+z/bMPsrsoyl7Ceqe+htu28X41PUdkpLbjyx0rIkL/95spb/pcegrqM/Woz9RxZ3+JxmuaRm4tS3+5SnJyMgwGA4oVKwYAOHbsGIoVK4Y6derk7NOyZUv4+fnh+PHj6NKli+xx0tPTkZ6ennM7JSUFAE+FyMzMdN4TcCHpefjK83E2b+qvqCh+GXHCBPMc1YgIYO5cfr+eT6NePaBSJeD6dfMRwODgTLN/lSQlAQMH8sli773Ht8mNJEo5gf37W+8jeeMN4IMPeKUGJbt3A3Pm8FXL9PpVVa8eL2FmkRVlV85zW5gF9uNPyHzmGQBAdt3aWDD8IvbvLIrgxNz+0/s1jIoC2rfnq1XdvMnLt9Wvz0fq1Bx/1y7+ujBm3vdJSbmvl1yup6O86XPpKajP1KM+U8ed/SV6TgNjWooK6c9gMGD79u3o3Lmz7P2PHj1Cw4YNUaVKFcTFxQEAZs+ejfXr1+OPP/4w2zc8PByxsbF47bXXZI8VExOD2NhYq+0bN26kSXKEEF3lT0pC7YULUfzCBXz/7ru4X768u5tECCFeKS0tDX369EFycjKKFCmiuJ+mkdvp06fbvH/q1KlaDqsoMzMTPXr0AGMMH374ocPHmzRpEsaMGZNzOyUlBZGRkWjdurXNzvImmZmZ2LNnD1q1amWVE02s+WJ/GY3yI3Za7dplPlocHJyJtWv3YPDgVnj4UKzPdu/m7bDXLrm2HzsGvPiiWFvXrFEumaW2X0xHLeWEhPAR6ago62M3vP8NAsaPhyExEZn5C2H1OxHouQFmfSaN7m7Y4JzRT0cdPizW77t3A40a6XtuX/xcOhv1mXrUZ+q4s7+kK+32aAput2/fbnY7MzMTly9fRr58+VCxYkVdg1spsL169Sr27dtnFnyWKlUKt27dMts/KysLSUlJKGVj+aSgoCAEBQVZbQ8ICPC5N7YvPidn8pX+UrNalijTySjffQcsXMi3P3wYIBzc3rwJ5M8PNG9ue7+AAOt9bt4EHj4Ua2vp0vwYltT2i1QCLC1N+VwGQ+7yvjntzsoC3nmH5xgAYDVr4oUbm3EqpTx64iurPnP1MsFqiPb7zZvyfa4HX/lcuhL1mXrUZ+q4o79Ez6cpuD1z5ozVtpSUFAwcOFAxz1ULKbC9cOEC9u/fj9DQULP769evj3v37uHUqVOoXbs2AGDfvn3Izs5G3bp1dWsHIc6k9wx0kZXAtAa40szvxo35SllqOVI1QfSxYWHylQa09ItICbBr1yyW9/3nH16b6cgRfvuNNzC7+AIcnplfMT/ZGcsE64Vq5hJCvI2mOrdyihQpgtjYWLzzzjvCj0lNTcXZs2dx9uxZAMDly5dx9uxZ/P3338jMzES3bt1w8uRJxMXFwWg04ubNm7h58yYyMjIAAE899RTatm2LoUOH4sSJEzhy5AiGDx+OXr16oUyZMno9NUKcJj4eKF+ej/b16cP/LV/e/opfSkQWG7C34IIIf39g3jz+f3srMEn7OFpXVSqDZs/y5fJpDmr7xWjkI9QizEqFbdjAA9siRYAtWxD/wvuYMjO/0HG++44vaaxUL9gdqGYuIcTb6BbcAryaQXJysvD+J0+eRK1atVCrVi0AwJgxY1CrVi1MnToVCQkJ+OKLL3Dt2jXUrFkTpUuXzvk5evRozjHi4uJQpUoVtGjRAu3bt0ejRo2wcuVKPZ8WIU5haznXbt2sA1yjkQc9csGPdF9MjNiCC3KrZakl5Yfa+x6p12IKUu1WW8H0uHHyubaiC1FI/SJ96Zg5U6xtZqOWEyYAI0YAZ87A2KUbRo4UOwbAz6fHlxw90epWhBBvoykt4T2pns9/GGO4ceMGNmzYgHbt2gkfp1mzZrBVrEGkkENISAgt2EC8jr2RRMscTFu5ooD1ffbouWLYzz8D777L25KUZH2/nospSKssWT7fsDBeIqx7d/nHiT7fGzeU0xfkGAxAvVJX0GTDTKDe+zyh2N8/p+bZoQPaVzbTI41EL0r9rudrSwghetEU3C5evNjstp+fH8LCwjBgwABMmjRJl4YR4svUjCQmJSnninbtqu38euZHfvUVHzFWCgYXLdI3+NGyypLo8w0P5zV5RQPbzmw7NqUMht/ae0DxYsCCBWb7OPIlQu5LjjvR6laEEG+hKbi9fPmy3u0gJE8RDXoSEoCJE23niqqhx9KuliZMUG6LwQCMGQN06aJvECS3pKkSo5H/hITIjywDuf0CiI20BiIdKwqOw6DUZcAD8FUeRoyw2s/RLxGeNtFMTb8TQoi7aApuTV377y9BhMhMD0IIAPGg5/Zt7Ze1LTkrP9J0lTRLrg7OLCtPJCYCo0fb7kPTfrGoLCirAi7hYOmeKHvjFN8wbhwwa5ZsHSxpMlZCgrYvIxI900gIIcTXaZpQlp2djenTp6No0aIoV64cypUrh2LFimHGjBluW5qXEG8iOgM9LEy/c0ZEuC9/0xXBmVzlie7d7X85MO0Xe186WmIPTuNZHtiGhgJffskTjhVqL4pMxhJBZbYIIUScpuB28uTJeP/99zF37lycOXMGZ86cwezZs7Fs2TJVpcAIyatEZ6CXLev4uaZMAfbvBy5fdt/EH2cHZ0qVJ2wJCQH27jXvF3tfOi6iMmDwA2vYCDh7VmjpLmkyluVrGREBbNlCZbYIIURvmoLb9evXY/Xq1XjttdfwzDPP4JlnnsHrr7+OVatWYd26dTo3kRDfpBT0lCjBZ6WHhAANGtgOfmyRAqOYGJ4S4KyJP2XLujc4s1V5wpakJN4npv0i96WjOJJybl81lMePiw7BcGC/WNHd/0RHA1eu8CVqAf7v5cs8IKcyW4QQoi9NwW1SUhKqVKlitb1KlSpIUpqxQQixIgU9+/cDb77J6/7fvs0DmubNgYoV+WJXgO3L2u4MjAYMkA8sXdUGkVXElMilS5h+6eiNjbiC8miHr3LSF1qOehrIp366gr8/0KgR/3+jRrl9Ymtk1xPKgBFCiLfRFNzWqFED77//vtX2999/HzVq1HC4UYTkJf7+vJzWe+8BKSnm9127xqtLjR0rH/xs28Z/3BEY7drF/509W/5+Z7dBWrhi2zbtx1BKl4hum4arbYZiI/qiCO5jfZO1Tk3rMP2Ss3Gj+9NICCHEm2mqlvDuu+/ixRdfxN69e1G/fn0AwLFjx/DPP//gq6++0rWBhPi6rVuB+fOV72cM2LQJuHQJOHqUjzaGh/P7bt3iAZrpfa6oPxofD/TvzwMxObGxwOTJzmuD3KIWatgsifb770CPHvD75Re+4zvvIGzqVMDJI+BUZosQQvShaeS2adOm+PPPP9GlSxfcu3cP9+7dQ3R0NP744w80ppkPhAgzGoHXX7e/3z//8OC1WTMgKIgvNNCyZW5VgIoVeQ5p797Oza+V2mwvx3XVKnXHU1pWWI6WyWOmbKZLrF8P1KkD/PILULIkn3EWG0tJr4QQ4kU017ktU6YMZs2apWdbCMlzDh3iObYibC0N68qlWqUc1+Bg5X2uXeOlX6dOtX0sW8sKyz0PrZPHTCkuGXv0KP/WAPBvDp9+ygNcQgghXkVVcJuYmIgHDx6gXLlyOdt+/fVXLFiwAA8ePEDnzp3Rp08f3RtJiKtZLgbgrMv8auq/2loa1pVLtYq2edo04KmneK1euX7UEqirnTwWGQksXKjcBjMNGgCvvMIfNHEijdYSQoiXUhXcjhgxAmXKlMHChQsBALdu3ULjxo1RpkwZVKxYEQMHDoTRaET//v2d0lhCXEFuNDEkhG/TO49UtP6rtJiDrcBOWg3swAGgRQuHm6ZITc3a3r3N0wykUdlOnZRHYG0F6qKB9fDhQNeudr6UMAZs2AC0bZubxPzhh9rqrhFCCPEYqnJuf/jhB3Ts2DHn9ieffIKQkBCcPXsWO3fuxOzZs/HBBx/o3khCXGXXLvl8zqQkPhJZsiQPfvUiLRpgzwcfiC0NCwA9eujbRkuibQas82elUdlZs8QC9UOHzLeLBtZdu9rJPb5/H+jXj9cx698fkFZWpMCWEEK8nqrg9ubNmyhfvnzO7X379iE6Ohr5/qv52LFjR1y4cEHXBhLiShMm2M7nvHOHB056BY/SogG2Yqpx4/gysqKBXVISDyCdFeCaLnSgltS3oo+3HKkVXbbY5rzWs2f5pLGNG/mTeeEFscYQp1E7qZAQQmxRFdwWKVIE9+7dy7l94sQJ1K1bN+e2wWBAenq6bo0jxNUSEsT2GzVKvz/AUhF/y9HQsDC+POu77/Lb9gI7Z7bRUnQ08Pbb2h7LGA/ARVgG9KLLFsuO2DLG0w7q1QP+/JN35vff8280fpoKx/gkVwea8fFA+fK86odU/aN8eedefSCE+DZVv9Hr1auH9957D9nZ2di6dSvu37+PF0xGPf78809ERkbq3khCPI3cJXNHyBXxv3ED6NIlN9A4dAhYvFjseEqX9fU0dqxjjw8J0TYCq2lFr5QUoFcvXnctPR3o0IGP4DZs6NiT0ImnjFy6OtBUKusmpa9IC4UQQogaqiaUzZgxAy1atMCnn36KrKwsvP322yhevHjO/Zs2bULTpk11byQhnkhNpQMRlkX8lcpkjR0LrFkjNvqpdxtNSaOjBgP/UVuea+RIICbG+rEiy/ZGR/PJZqoqWpw+zZfNnTcPGD3aY/Jr1ZZDc2Y7XFlmzlZZN2lS4cSJthc4IYQQOaqC22eeeQa///47jhw5glKlSpmlJABAr169ULVqVV0bSIgrlS0LXLwotq9IDqytkmK27rMVaCxYwOvHxsbq00ZHbdigfrUwg4GXCdu6VT6wk61Da8Huil5S5xkMQJEiPMcjPR2w+L3lTp5QtxgQCzT1LjNnr6wbY9oX6iCE5G2qE81KlCiBTp06WQW2APDiiy/i8ccf16VhhLjDvHn2B/SEJi3B9iVeW/fZCzQAPnLr8MQqnURF5aZUTJki9hjGgJ49+f8t0zEuX9YhoLt7l8/8e//93G01a3pUYCvyOjszb9qUSKCpd5qLM68qEELyNppFQYiJqCg+WhYaKn+/yCVzwHYuYdeu/Ecpz1CkTNa1a8DQoeZtUttGPUmjqDEx6ie9Afyxui0dfPw48OyzwPbtvDCx6Ow1F3NHQKlENNDUMyB1xVUFQkjeRMEtIRaio4F//+WX/UNCzO+zOWnpPyIjcnKk+957T6ydlStrmFjlZGrKhOkevDHGlyNr1IgPB1esyIeCLV9EjfSe9OWOgFKJaKCpZ0AqUtZNtJ4yIYSYUpVzS0he4e/P81onT1a/DK/aJWJNMcZr6YooXZqPdHboACxfDly6xOO5118HAgO1nd+SlmWIpWoGQ4e6cNLbnTt8beIvv+S3e/QAVq4EihbV4eDOmfTljoBSiRRoJiTIfwGTAk0901ykL0LduilPKpw7V7/zEULyDhq5JcQG6XK7mkvmegRromWy4uN5QDt6NE8vHT2a39ajdJMjZaGio4HNm8XO43Dw9ugR8NxzPLANCuK1bDdt0jWwtVWuSmtf67IghU4cqh/sAHtl3aKi9D0fISRvEA5uU1JShH8Iycv0GGkbOZL/ayvQ2LnTOUEXoE9A16yZi4K3/PmBV14BnniC59u++qpuZb6cOenLXQGlEk31g3U6r1MmFRJC8izh4LZYsWIoXry40A8heZk0IqeFFPBNnmw70OjUyXlBl14BnVODt9u3AdOlvseN43Vsa9TQcDBlzp705a6A0lZ73BFoarlCQgghSoRzbvfv35/z/ytXrmDixIkYOHAg6tevDwA4duwY1q9fjzlz5ujfSkK8iBTUde2q7nGWAZ+thQoOHBAPumzWgpUhGtAdO2b/WFLwprWWrayDB3kUVLQo8OOPQMGCfPncggU1HMw2V0z60rQghRPZrR9MCCEeTji4NV15bPr06Vi0aBF69+6ds61jx46oXr06Vq5ciQEDBujbSkK8THQ0H91cskT8MXIBn1KgoTboUjMxTPTYN28CBQrY30+34M1oBObMAaZNA7KzeXB76xbgxNrarpr0RQElIYToR1O1hGPHjmHFihVW2+vUqYMhQ4Y43ChC3E0kGLS3T6dOYsHtlClAixbqAj41QZfamf6ixy5VChBNsXc4ePv3X6BvX+C77/jtgQP5DDonjNaackcVAUIIIY7RVC0hMjISq1atstq+evVqREZGOtwoQtxJpEqAyD6is+FjYtTnGYoeOzFR/cQw0WP/l5HkfN99x3Npv/uODxWvXw98/LHTA1vA8yZ9EUIIsU9TcLt48WIsW7YM1atXx5AhQzBkyBA888wzWLZsGRYvXqx3GwlxmV277AeDopUEnBkYiRx74UJeGkztxDCPCugYA2bP5iO3Tz8NnDoFvPSSC06cy9MmfRFCCLFNU3Dbvn17/Pnnn4iKikJSUhKSkpIQFRWFP//8E+3bt9e7jYS4zIQJtoPBkSPVVRJwZmAkHbtMGfPtZcvy7WFh2mf6e0xAZzAAGzbwKP3ECaBKFRed2ByVqyKEEO+heYWyyMhIzJ49W8+2EOJ2CQnK9zFmf+UxuSoFzp4Nr5Q+4OhMf9N2JyTw6lthYXyBCUeXnrXp22+Bw4eB6dP57TJlgEWLnHhCMTTpixBCvIPm4PbQoUP46KOP8Ndff2HLli0oW7YsNmzYgMcffxyNGjXSs42EeB3LgNEZgZGUHmE5iiylR8TEiB3H1gQyf3++hO7EifIT0tQG6DYn4WVl8TWPpXKC9esD7dqpOwEhhJA8T1NawrZt29CmTRsEBwfj9OnTSE9PBwAkJyfTaC4h0GeVMltEFlpYtcrxFcJs5Rf376+uzXKT8MLD+QCt8co/PPqXAtvXX+c7EEIIISppCm5nzpyJFStWYNWqVQgICMjZ3rBhQ5w+fVq3xhHiamXL2g4GIyJctKSsHSILLVy7BgwdmtsuU9LtIUOAzZv5ohCWqQYiAbS0nz1KQXJSEnB82m4kV6gJHDkCFCnCG/TBB3xZXUIIIUQlTcHtH3/8gSZNmlhtL1q0KO7du+domwhxm3nz+L9KweDSpbmVBCy5spKAaD5t5cryE8NCQvjPtGnKpcxEAmjA/kpltoLkdzAdu9EBISwJJ1Eb38w5DXTvLvTcCCGEEDmagttSpUrh4sWLVtsPHz6MChUqONwoQtwlKkqsSkBIiPVjQ0JcV0lAzSIOljP9Y2OBO3f4jynLUmZqViqzxVaQ/DOqAwCWYCQa4giGzq3o3MlqhBBCfJ6mCWVDhw7FyJEjsXbtWhgMBly/fh3Hjh3D2LFj8c477+jdRkJcylZ1A6VJXIB1sOhMalfOkia0GY18hFYOY/xxo0bx569mpTJbLIPkENxBEkIBADvQBdVxDr/8F+RKlSYaN3ZedQlCCCG+TVNwO3HiRGRnZ6NFixZIS0tDkyZNEBQUhLFjx2LEiBF6t5EQl5OrbmDr8jpgHhg6OxCTFlro1o2f17RNttIjRFINTANMewE0YH+lMilIDkAG3sV49MFG1MIZXAcfHpcCW8nOnXyymuhywYQQQogpTWkJBoMBkydPRlJSEn755Rf88MMPuH37NmbMmKF3+wjxGGoCQ1fQstCCmtq3IiuVAfYD+caNgQal/sIRNMQoLEU4bqMDvlTcf8kSdcsFE0IIIaY0BbeDBw/G/fv3ERgYiKpVq+L5559HoUKF8ODBAwwePFjvNhLiERxdFMEZ1K6cpSZXVzq+UgC9YYPYsfy3b8WBlFp4DidxByHogF1YiVfk91UIlG0tF0wIIYSY0hTcrl+/Hg8fPrTa/vDhQ3zyyScON4oQJUYjL1v12Wfy5aucSW1g6CpSCkXv3vxfWyOpUqqBmlJmSgF0VJSdhj16BLzxBtC9OwLSUpBYpSGaFzuL3egge17A9uvp6pFx4h3c+TuBEOKZVAW3KSkpSE5OBmMM9+/fR0pKSs7P3bt38dVXXyE8PNxZbSV5nNwiAJblq5xJS2DoaURSDeRyddUE0DnmzQOWL+f/nzQJJc7tx5nESMTGWlebiIjgo7IiXDkyTjybu38nEEI8k6rgtlixYggJCYHBYMATTzyB4sWL5/yUKFECgwcPxhtvvOGstpI8zNZKWa7KxdQaGHoaLbm6mowdyyPhr78GZs8GAgLg789X2L11y3okuFMnscO6emSceCZP+J1ACPFMqqol7N+/H4wxvPDCC9i2bRtCTIZfAgMDUa5cOZQpU0b3RpK8zd5KWXpXKTAalctQSYHhyJHWs/mXLPGe2fy2yp1p9vAhX/N3+HDAzw8oWBDYt092qFuuGoXa8mYk73L17wRCiHdRNXLbtGlTNGvWDJcvX0bnzp3RtGnTnJ/69eurDmwPHjyIqKgolClTBgaDATt27DC7Pz4+Hq1bt0ZoaCgMBgPOnj1rdYxmzZrBYDCY/bz66quq2kE8myurFOzaZf8yp9pJXJ5KU6qBkvPngeef5xHH/Pm525VyOBTa4wsj48T5PK1yCSHEs2iaULZv3z5s3brVavuWLVuwfv164eM8ePAANWrUwAcffKB4f6NGjTBPWhNVwdChQ3Hjxo2cn3fffVe4DcTzubJKgWV9VUD+MqeugaGXM2zYANSuDfzyC1CyJP+/Ri5LmSBezRMrlxBCPIemRRzmzJmDjz76yGp7eHg4hg0bhgEDBggdp127dmjXrp3i/f379wcAXLlyxeZxChQogFL2lkkiXssVVQqkGdZ0mVOFBw9Q6733kG/fPn67RQvg00/tL1lmh1NSJohP8dTKJYQQz6ApuP3777/x+OOPW20vV64c/v77b4cbpVZcXBw+/fRTlCpVClFRUXjnnXdQoEABxf3T09ORnp6eczslJQUAkJmZiczMTKe31xWk5+ELz6dePaBSJeD6deVczLJl+X5an+7Ro/yBwcHKB0hMBA4eBBo10nYOn/Lbb/Dv3RuP/f47mJ8fst95B9kTJ/IIVKf3XMOGuf/PzuY/3s6XPpeuoNRfrvid4K3oPaYe9Zk67uwv0XMaGFNaTFTZY489hvfffx8dO3Y0275z50688cYbuGYrGUqpIQYDtm/fjs6dO1vdd+XKFTz++OM4c+YMatasaXbfypUrcyaynTt3DhMmTMDzzz+PeBtTZWNiYhAbG2u1fePGjTaDYkIIV/TiRTSeOBEZRYrg1JgxuPP00+5uEiGEEB+XlpaGPn36IDk5GUWKFFHcT9PIbe/evfHmm2+icOHCaNKkCQDg+++/x8iRI9GrVy9tLdZo2LBhOf+vXr06SpcujRYtWuDSpUuoWLGi7GMmTZqEMWPG5NxOSUlBZGQkWrdubbOzvElmZib27NmDVq1aISAgwN3N0cWuXcCECTwHVhIRAcydK7CggB2HDmXi/v09GDy4FR4+VO6v3bvz8MhtdjavgvCfrHLlcCAtDU27dfOZ95iz+eLn0pns9Zczfyd4K3qPqUd9po47+0u60m6PpuB2xowZuHLlClq0aIF8+fghsrOz8dJLL2H27NlaDqmbunXrAgAuXryoGNwGBQUhKCjIantAQIDPvbF96Tk5Mxezbl1g717g4cMAxeDW359fKveR7lTnp5/4bLuPP86ZMJbZqRMyvvrKp95jrkJ9po5Sf1F+tjJ6j6lHfaaOO/pL9HyagtvAwEB8/vnnmDFjBn766ScEBwejevXqKFeunJbD6UoqF1aaZhL4JLn6qHo4ccL+PkYjcPSoc87vsRgDPvqIz6ZLTwfGjeO1awnxEM76nUAI8V6aglvJE088gSeeeELz41NTU3Hx4sWc25cvX8bZs2cREhKCxx57DElJSfj7779x/fp1AMAff/wBAChVqhRKlSqFS5cuYePGjWjfvj1CQ0Nx7tw5jB49Gk2aNMEzzzzjyFMjXs7WQgxybt4ERNKt81RpoeRkYNgwYPNmfrtDB2DdOrc2iRBCCLFHOLgdM2YMZsyYgYIFC5rlq8pZtGiR0DFPnjyJ5s2bm50DAAYMGIB169bhiy++wKBBg3Lul/J5p02bhpiYGAQGBmLv3r1YsmQJHjx4gMjISHTt2hVTpkwRfVrEB8XHy68gtnSpcp3UUqUAkVSePHNB4NQpoGdP4NIlIF8+YN48YPRoVYsyEEIIIe4gHNyeOXMmpwTDmTNnFPczqPjj16xZM9gq1jBw4EAMHDhQ8f7IyEh8//33wucjvk9ab97ybSUtxKC0EED9+sA33yjHbnlh6VdptPvRkVNoHdsAfpkZQLlywOef86RkQgghxAsIB7f79++X/T8hnsKR9eZNbxsM5sfIC0u/mo52G1ALX6E5kD8YmdPXIqpucXc3jxBCCBGmafldQjyRHuvNb9iQ95Z+jY8HZnc9hTvX0gAADH7oim1o/ygenQYWh42S0YQQQojHER65jVbxl93WAgqEOIse681HReWt0kLGLIZfBi/GMUzAJ3gJQ7AGAJCGggAAA2jZYUIIId5FOLgtWrRozv8ZY9i+fTuKFi2KOnXqAABOnTqFe/fuqQqCCTGltsKBJb3Wm88zpYWSknA3aiCmJu8CABRBCvIhE1nIrSNoOtqdJ/qEEEKI1xMObj/++OOc/0+YMAE9evTAihUr4P9f9GE0GvH666/7zApfxLW0VDiw1Lgxf0xCgvJ6874+KUzY0aNAr14o8c8/eIQgjMZirMCr4GO11vJUCTRCCCFeTVPO7dq1azF27NicwBYA/P39MWbMGKxdu1a3xpG8QapwYJkvK1U4EM1y8ffnwTBgXfUgL0wKE5Kdzct6NWkC/PMP0iIqox5+wAq8BqXAFshDJdAIIYR4PU3BbVZWFs6fP2+1/fz588jOzna4USTvsFfhAOA5n0aj2PGio/nkr7w2KUxYYiKwYAHv0N69EfTzKdyJqGmzBFpkJI12E0II8R6aVigbNGgQXn75ZVy6dAnPP/88AOD48eOYO3eu2aILhNijpsKBaM4nrTdvQ3g4Lwnxzz/AkCHwNxiwdCkfIc+LJdAIIYT4Hk3B7YIFC1CqVCksXLgQN/5LxitdujTGjRuHt956S9cGEt+mR4UDOXpMCnN0gptHMBqBOXOAqlVzh63btjXbRRrtlst5XrKERrsJIYR4F03BrZ+fH8aPH4/x48cj5b81S2kiGdFCrwoHetNjgpvb/fsv0K8fsHcvULQoj87DwmR3pdFuQgghvkJTcAvwvNsDBw7g0qVL6NOnDwDg+vXrKFKkCAoVKqRbA4lv88QKB1qX8PUo+/YBffrwALdAAR6VKwS2kjxTAo0QQohP0zSh7OrVq6hevTo6deqEN954A7dv3wYAzJs3D2PHjtW1gcS3eVqFA70nuLmc0QhMmwa0bMkD26efBn78ERgwwN0tI4QQQlxCU3A7cuRI1KlTB3fv3kVwcHDO9i5duuC7777TrXEkb/CkCgfHjjm+hK/bZGTwoHb6dN7QIUOA48d5vi0hhBCSR2hKSzh06BCOHj2KwMBAs+3ly5dHQkKCLg0jeYun5HzevCm2n7sWNbA5yS0wEKhWDTh5EvjoI56WQAghhOQxmoLb7OxsGGWuy167dg2FCxd2uFHEd6ipOOAJOZ+lSont545FDeQmuZUrm4Vls1IQNSCEb1i4kOdNVKrk+gYSj+MTFT8IIUQlTWkJrVu3xpIlS3JuGwwGpKamYtq0aWjfvr1ebSNeLj4eKF8eaN6cDyI2b85vi6445g716/N0CE9b1EBuFbcI/INPE5qh8MBobN+SxTcGBVFgSwB45+ePEEL0oCm4XbBgAY4cOYKqVavi0aNH6NOnT05Kwrx58/RuI3EjoxE4cAD47DP+r+hEKr2W1HU1T5vgBshPcmuP3TiLmmiEI3gWp7F8xO+eO8mNuJy3fv4IIUQPmoLbyMhI/PTTT5g8eTJGjx6NWrVqYe7cuThz5gzCw8P1biNxE60jP95eccCTJrgB5qu45UMm3sU47EYHhCIJJ1EbtXAGe/+t7pmT3IjLefvnjxBCHKU65zYzMxNVqlTBl19+ib59+6Jv377OaBdxM0dqvTpjSV1X85QJbkDu5LXHcBWfoyfq4TgA4D2MwDjMRwaCzPYjeZsvfP4IIcQRqoPbgIAAPHr0yBltIR7C3siPwcBHfjp1kg/2nLWkrqt5wgQ3IHfy2qfoh3o4jrsohsFYix3oIrsfydt85fNHCCFaaUpLeOONNzBv3jxkZWXp3R7iAdSM/Mjx1CV1vZW0ittrWIF9aI5aOGMW2LprkhvxTPT5I4TkdZpKgf3444/47rvv8O2336J69eooWLCg2f3xNFvBqzk68uOOJXV9suTRX38BR4/Cv18/LF0KdOtWDS2xz6xP3TXJjXguT1zSmhBCXEnTyG2xYsXQtWtXtGnTBmXKlEHRokXNfoh3c3Tkx9UVB3yy5NHWrUCtWsDAgcDRox43yY14Lk+s+EEIIa6kaeT2448/1rsdxIPoMfIjBWOWiw5ERPA/rHoFY45MfPNIjx4Bb70FLF/ObzdowHMO4FmT3Ihnc9XnjxBCPJGq4DY7Oxvz58/HF198gYyMDLRo0QLTpk1DcHCws9pH3EAa+enWjQeyWi+DOzsYc3Tim8e5cAHo0QM4e5bfnjgRmD4dCAjI2cVTJrkRz0dfhggheZWq4HbWrFmIiYlBy5YtERwcjKVLl+LWrVtYu3ats9pH3ESvkR9nBmM+VfLo88+BIUOA1FSgRAlgwwagbVt3t4p4OfoyRAjJi1QFt5988gmWL1+OV155BQCwd+9evPjii1i9ejX8/DSl7xIP5ukjPz5V8ujff3lg26QJsHGjdXItIYQQQoSoCm7//vtvtG/fPud2y5YtYTAYcP36dUREROjeOOJ+njzy4/Ulj7KzAelL4YgRQGgo0LMnkE9TKjwhhBBCoLJaQlZWFvLnz2+2LSAgAJmZmbo2ihAR0sQ3yxnhEo+u//rJJ8CzzwIpKfy2wQD07UuBLSGEEOIgVX9JGWMYOHAggoKCcrY9evQIr776qlmtW6pzS1xBr4lvehCus/vgATB8OLBuHb/9/vvA2287v4GEEEJIHqEquB0wYIDVtn79+unWGELU8oSSR/Hx8udfutTi/L/+yqsh/PYbT0eYNg2YMMH5DSSEEELyEFXBLdW3JZ7InRPfhOrsdmHA2rU8r/bhQ97AjRs9N5mZEEII8WKU4Ed8gjsmvonW2e18eQn8xo7hd7Ruzct8hYe7tK2EEEJIXkH1uwjRSLTO7rEKffnMtjlzgP/9jwJbQgghxIlo5JYQjZTr5zI0xff4Hs0AAH8/CkfD8+eBAgVc1TRCCCEkz6KRW0I0kqufWxgp2IReOIDm6IO43P0osCU6MhqBAweAzz7j/xqN7m4RIYR4DgpuCdHIss7usziF03gWPbEZmciHUCR5bp1d4rXi44Hy5YHmzYE+ffi/5cvz7YQQQii4JUQzqc4uGMMILMNRNEAlXMJVPIYmOIT3DSNcVmeX5A1SdQ7LXG+pOgcFuIQQQsEtIQ6JfuEe/qnbDe/hTQQhAzvQCbVwBgmR9XgZMBfU2SV5g73qHACvzkEpCoSQvI4mlBHiiB9/RNnj8WABAbg4bD4eNngT8WUMLquzS/IO0eochw5RCWVCSN5GwS0hjmjVCli4EIbGjVH5uedQ2d3tIT5LuTqHtv0IIcRXUVoCIWokJQH9+gFXruRuGzMGeO45tzWJ5A1y1Tkc2Y8QQnwVBbeEiDp6FKhZE4iLAwYOlE9+JMRJLKtzWDIYQNU5CCEEFNwSYl92NvDuu0CTJjypsXJlYMkS5SiDECfIqc4B67eedJuqcxBCCAW3hNh2+zbQoQMwYQKfht67N3DqFB/BJcTFoqOBrVuBsmXNt0dEgKpzEELIf2hCGSFKzp8HWrQArl8H8ucH3nsPGDKERmyJW0VHA5068aoIN27wHFuqzkEIIbncOnJ78OBBREVFoUyZMjAYDNixY4fZ/fHx8WjdujVCQ0NhMBhw9uxZq2M8evQIb7zxBkJDQ1GoUCF07doV//77r2ueAPFt5csD4eFAlSrAiRPA0KEU2BKP4O/Py3317s3/pcCWEEJyuTW4ffDgAWrUqIEPPvhA8f5GjRph3rx5iscYPXo0du3ahS1btuD777/H9evXEU3X5ohWt2/nVsHPnx/44gvgxx+B6tXd2y5CCCGECHFrWkK7du3Qrl07xfv79+8PALhiWnbJRHJyMtasWYONGzfihRdeAAB8/PHHeOqpp/DDDz+gXr16ureZeDejUflybolz55Dv1VeBV14BYmL4xshIt7VVjq32E0IIIcTLc25PnTqFzMxMtGzZMmdblSpV8Nhjj/2/vTuPi6rc/wD+GbYZZBlZlGEUhHK3UHBBvKWYuC8obpgmGmlmucC96vV3c8Pdukn9Ms2itMKykiw3Cr3eNEPMBb2kF9EQXEATREQUWZ7fH/NjdARkRmHOzPB5v17zkrPMOd/z7WH6zsNznoPk5OQai9uSkhKUlJRolwsLCwEApaWlKC0trd+gjaTyOizleurCjh2a+8IuX76/rlkzYPWKcgw9EYMeq1ZBJgTEtm0o+9vfALlcumCrUWP8q4GhQ40fD9uY4ZgzwzBfhmPODMecGUbKfOl7TrMubnNzc2FnZ4fGjRvrrPfw8EBubm6N71u5ciWWLFlSZf1PP/2ERo0a1XWYkkpKSpI6BJNhbQ28/bbuOkV+PgJi3oFtWhoAICskBP+ZMgXl+/ZJEOGjVRd/pd27jRvLg9jGDMecGYb5MhxzZjjmzDBS5Ku4uFiv/cy6uH1c8+fPR3R0tHa5sLAQXl5e6NevH5ydnSWMrO6UlpYiKSkJffv2ha2trdThSKq8XDNk9sEeTwDoU56EuHvz0AR/otjKAWdmTUW7ZcvQ38TyVVP8lWQyTQ/uqVPGHaLANmY45swwzJfhmDPDMWeGkTJflX9pr41ZF7cqlQr37t1DQUGBTu/t1atXoVKpanyfXC6HvJo/Odva2lpcw7bEazLUoUPAuXO669xwHV9hNBxQjJPww0TbeMT0Og8/E8xXdfE/LCMDOHxYc+e8sbGNGY45MwzzZTjmzHDMmWGkyJe+5zPrhzh07twZtra22PfAn5DT09ORnZ2NoKAgCSMjU5KTU3VdHtwxC+9iPaahOw4jw6qN8QPTU3XxP8l+RERElkzSntuioiKce6BLKjMzE6mpqXB1dYW3tzfy8/ORnZ2NK1euANAUroCmx1alUkGpVCIyMhLR0dFwdXWFs7MzZsyYgaCgIM6UQFqenpp/B2I38uGKFGjaRhxeQRxeAQDYw3RvJKiMv672IyIismSS9twePXoU/v7+8Pf3BwBER0fD398fCxcuBAD88MMP8Pf3x+DBgwEA4eHh8Pf3x4YNG7THWLt2LYYMGYKRI0eiZ8+eUKlUSEhIMP7FWJjycuDf/wa+/FLzb+XUr+bo+e6l+MBxLnZjMLZiLFyQr7NdJtM8vtRUPf+8Jr6anh8hk2lmLHv+eePGRUREZIok7bkNDg6GEKLG7ZMmTcKkSZMeeQyFQoF169bV+CAIMlxCAjBrFnDp0v11zZsD775rhs+uz8qCdXg4Xis6DAD4AaG4DQft5sqCcdUqKYLTj7W1JvejRmniffBXpjL+2FjOd0tERASY+ZhbqnsJCZoi6sHCFtDcqT9qlGa72fj+e8DfX3OnlVKJw3O2YU3z93AP928mbN4c+PZbaeaJNURYmCbOZs1011fGb3ZfOoiIiOqJWc+WQHWrvFzTY1tdZ7oQml7C2bOB0FAT7yUsKwPmzNF0ZwJA167A1q3o7uuLCyurf8KXOczdHRamyT2fUEZERFQzFrdGZsqPTz14sGqP7YOEAC5e1OwnxZRTerO2BjIzNT9HRwMrVwJ2dtpNJh17Lcw9fiIiovrG4taITH0sqzlNOVXtlwSUa6o/mQz49FMgORkYNEjqUImIiMiIOObWSMxhLKu5TDmVkAD4+AC9ewMvvggM6H0XnyvfwMXgl+6PqXBxYWFLRETUALG4NYLaxrICmrGsUk+3ZQ5TTj38JaElMvAremDS7XXw+uVL7F/zm3TBERERkeRY3BqBIWNZpVQ55RRQtcA1hSmnHv6SMBZf4Rg6IwAn8CfcMQi7EbGum+RfEoiIiEg6LG6NwJzGsprylFOVXxIUuIMNeBVfYRyccQsH8Dw6IRV7MNAkviQQERGRdHhDmRGYy1jWSqY65VRl8Z+AMAxEIiogw3L8A0uwCOUPNGVT+JJARERE0mBxawSVY1kvX65+3G3l419N6fGppjjlVGXxvwZz0REnEYHN2Iu+Ne5HREREDQ+HJRiBqY9lNXm3bwPJydovCT/LeuNpnK9S2JrCDW9EREQkLRa3RmLKY1lN2u+/A926Af36wfr8We2XhBKZvc5u/JJAREREAItbowoLAy5cAPbvB7Zs0fybmcnCtlpCAJ98onl07unTgKMjkJfHLwlERET0SBxza2SmOJbV5BQVAa+9BnzxhWa5b1/Nz02bAjDdG96IiIhIeixuybScOgWMHg2cPQtYWQFLlwJ//7vm5wfwSwIRERFVh8UtmZYtWzSFbbNmwJdf8u4wIiIiMgiLWzItS5dqxtvOmQO4u0sdDREREZkZ3lBG0jp+HJgwASgt1Szb2gKrV7OwJSIiosfC4pakIQTw/vtAUBAQH68paImIiIieEIclkPEVFACvvAJs26ZZHjYMmD5d0pCIiIjIMrDnlozryBHA319T2Nraap66sH074OoqdWRERERkAdhzS8azZQswaZJmfK2vL7B1q+YhDURERER1hD23ZDxdugByOTBypOZGMha2REREVMfYc0v1q/IRYgDQujVw7BjQqhUgk0kbFxEREVkk9txS/aioANasAXx8gP37769v3ZqFLREREdUbFrdU965fB4YMAebNA+7d09wwRkRERGQEHJZAdevgQWDcOODyZc342vfeA6ZMkToqIiIiaiDYc0t1o6ICWL4cCA7WFLZt2mim/Zo6lcMQiIiIyGhY3FLd2LULePNNTZE7YQJw9Cjg5yd1VERERNTAcFgC1Y0hQzRPHQsKAiZPZm8tERERSYLFLT2e8nLN08UmT9Y8XUwmAz76SOqoiIiIqIHjsAQyXE4O0Lcv8Le/AZGRgBBSR0REREQEgMUtGSopCejUSTN3rYOD5mljHIJAREREJoLFLemnrExzw1j//sC1a5qbxY4e1dw8RkRERGQiOOaWapeTA4wdq5nDFgBefRVYuxawt5c2LiIiIqKHsLil2snlQFYW4OQEbNwIhIdLHRERERFRtVjcUvXKygBra814WldXICEBUCqBli2ljoyIiIioRhxzS1VlZwM9ewKffHJ/XefOLGyJiIjI5LG4JV0//KCZDSE5WXMD2Z07UkdEREREpDcWt6Rx7x4QHQ2EhgI3bgBduwKHDvGmMSIiIjIrHHNLQGamZjaE337TLM+eDaxeDdjZSRoWERERkaFY3DZ0N24AXboA+fmAiwuwaRMwbJjUURERERE9Fg5LaOhcXIA33gCCgoATJ1jYEhERkVljcdsQnTsHnD9/f3nhQuDnn4EWLaSLiYiIiKgOsLhtaLZuBQICgNGjgbt3NeusrQFbW2njIiIiIqoDLG4bijt3gGnTNE8Xu3ULcHDQ/EtERERkQSQtbg8cOIChQ4dCrVZDJpNh+/btOtuFEFi4cCE8PT1hb2+PkJAQZGRk6Ozj4+MDmUym81q1apURr8IMpKcD3bsDH36oeeLY//wPsH8/0KSJ1JERERER1SlJi9vbt2+jY8eOWLduXbXb16xZg/feew8bNmxASkoKHBwc0L9/f9yt/HP6/4uJiUFOTo72NWPGDGOEbxZk8fGap4udOqUpZhMTgeXLARtOlEFERESWR9IKZ+DAgRg4cGC124QQiI2NxZtvvonQ0FAAwGeffQYPDw9s374d4eHh2n2dnJygUqmMErNZKS+H1fr1wO3bQO/eQHw84OkpdVRERERE9cZku+8yMzORm5uLkJAQ7TqlUonAwEAkJyfrFLerVq3C0qVL4e3tjRdffBFRUVGweUTPZElJCUpKSrTLhYWFAIDS0lKUlpbWw9UYX2lpKWBtjZJNm2D3zTeomDtXc+OYhVxfXav8724p//2NgTkzHHNmGObLcMyZ4Zgzw0iZL33PabLFbW5uLgDAw8NDZ72Hh4d2GwDMnDkTAQEBcHV1xa+//or58+cjJycH77zzTo3HXrlyJZYsWVJl/U8//YRGjRrV0RVIQAh4/etfaHTtGtLHjQMA/HT2LNCxI/DjjxIHZx6SkpKkDsHsMGeGY84Mw3wZjjkzHHNmGCnyVVxcrNd+Jlvc6is6Olr7s5+fH+zs7PDqq69i5cqVkMvl1b5n/vz5Ou8rLCyEl5cX+vXrB2dn53qPuV4UFcF6xgxYxccDAHymTsWPt26hb9++sOU0X7UqLS1FUlIS82UA5sxwzJlhmC/DMWeGY84MI2W+Kv/SXhuTLW4rx9BevXoVng+ME7169So6depU4/sCAwNRVlaGCxcuoE2bNtXuI5fLqy18bW1tzbNhnzoFjB0L/Pe/gJUVEBMD67/8BUhMNN9rkgjzZTjmzHDMmWGYL8MxZ4ZjzgwjRb70PZ/JznPr6+sLlUqFffv2adcVFhYiJSUFQUFBNb4vNTUVVlZWaNq0qTHClJYQwMaNQGCgprBVqzVTfP3jH5oil4iIiKiBkbTntqioCOfOndMuZ2ZmIjU1Fa6urvD29sbs2bOxbNkytGrVCr6+vliwYAHUajWGDx8OAEhOTkZKSgp69+4NJycnJCcnIyoqChMmTICLi4tEV2VEU6cCH3+s+XngQGDzZs5dS0RERA2apMXt0aNH0bt3b+1y5TjYiIgIbNq0CXPnzsXt27cxdepUFBQU4LnnnkNiYiIUCgUAzfCCr776CosXL0ZJSQl8fX0RFRWlM57WovXqBWzaBKxYAfz1r+ytJSIiogZP0uI2ODgYQogat8tkMsTExCAmJqba7QEBATh8+HB9hWd6hACuXAGaNdMsT5igefJYy5bSxkVERERkItjVZy4KCoAxY4CuXYFr1+6vZ2FLREREpMXi1hz89hsQEAB8+y1w/Tpw6JDUERERERGZJBa3pkwIIDYW+MtfgMxMwMcH+OUXYMQIqSMjIiIiMkkmO89tg5efD7z8MvD995rlsDAgLg5o3FjSsIiIiIhMGXtuTdWSJZrC1s4O+N//1QxJYGFLRERE9EjsuTVVS5cCZ88Cy5YBnTtLHQ0RERGRWWDPram4fh1YvVozzhYAnJ2BPXtY2BIREREZgD23puCXX4DwcODyZcDBAXjjDakjIiIiIjJL7LmVUkUFsHIlEBysKWxbtwZ69pQ6KiIiIiKzxZ5bqVy7Brz0EvDTT5rlCROA9esBR0dp4yIiIiIyYyxupXDwIDB2LJCTA9jbA++/D0yeDMhkUkdGREREZNZY3ErB2lrTc9u+PfD110CHDlJHRERERGQRWNxKoUcPYMcOzfhaBwepoyEiIiKyGCxupTJwoNQREBEREVkczpZARERERBaDxS0RERERWQwWt0RERERkMVjcEhEREZHFYHFLRERERBaDxS0RERERWQwWt0RERERkMVjcEhEREZHFYHFLRERERBaDxS0RERERWQwWt0RERERkMVjcEhEREZHFYHFLRERERBaDxS0RERERWQwWt0RERERkMVjcEhEREZHFYHFLRERERBaDxS0RERERWQwbqQMwBUIIAEBhYaHEkdSd0tJSFBcXo7CwELa2tlKHY/KYL8MxZ4ZjzgzDfBmOOTMcc2YYKfNVWadV1m01YXEL4NatWwAALy8viSMhIiIioke5desWlEpljdtlorbytwGoqKjAlStX4OTkBJlMJnU4daKwsBBeXl64ePEinJ2dpQ7H5DFfhmPODMecGYb5MhxzZjjmzDBS5ksIgVu3bkGtVsPKquaRtey5BWBlZYXmzZtLHUa9cHZ25i+rAZgvwzFnhmPODMN8GY45MxxzZhip8vWoHttKvKGMiIiIiCwGi1siIiIishgsbi2UXC7HokWLIJfLpQ7FLDBfhmPODMecGYb5MhxzZjjmzDDmkC/eUEZEREREFoM9t0RERERkMVjcEhEREZHFYHFLRERERBaDxS0RERERWQwWt2bkwIEDGDp0KNRqNWQyGbZv366zXQiBhQsXwtPTE/b29ggJCUFGRobOPj4+PpDJZDqvVatWGfEqjKu2nCUkJKBfv35wc3ODTCZDampqlWPcvXsXr7/+Otzc3ODo6IiRI0fi6tWrxrkAI6uLfAUHB1dpY9OmTTPOBUjgUTkrLS3FvHnz8Oyzz8LBwQFqtRoTJ07ElStXdI6Rn5+P8ePHw9nZGY0bN0ZkZCSKioqMfCXGUxc5a0ifZbX9Xi5evBht27aFg4MDXFxcEBISgpSUFJ192Ma262zXJ2cNqY0BtefsQdOmTYNMJkNsbKzOelNpZyxuzcjt27fRsWNHrFu3rtrta9aswXvvvYcNGzYgJSUFDg4O6N+/P+7evauzX0xMDHJycrSvGTNmGCN8SdSWs9u3b+O5557D6tWrazxGVFQUduzYgW+++QY///wzrly5grCwsPoKWVJ1kS8AmDJlik4bW7NmTX2EaxIelbPi4mIcP34cCxYswPHjx5GQkID09HQMGzZMZ7/x48fj999/R1JSEnbu3IkDBw5g6tSpxroEo6uLnAEN57Ostt/L1q1b4/3338d//vMf/PLLL/Dx8UG/fv3w559/avdhG9OlT86AhtPGgNpzVum7777D4cOHoVarq2wzmXYmyCwBEN999512uaKiQqhUKvHWW29p1xUUFAi5XC6+/PJL7boWLVqItWvXGjFS0/Fwzh6UmZkpAIgTJ07orC8oKBC2trbim2++0a47c+aMACCSk5PrMVrpPU6+hBCiV69eYtasWfUam6l6VM4qHTlyRAAQWVlZQgghTp8+LQCI3377TbvPnj17hEwmE5cvX67PcE3C4+RMiIb7WaZPvm7evCkAiL179woh2MYeJ2dCNNw2JkTNObt06ZJo1qyZSEtLq5IfU2pn7Lm1EJmZmcjNzUVISIh2nVKpRGBgIJKTk3X2XbVqFdzc3ODv74+33noLZWVlxg7XbBw7dgylpaU6eW3bti28vb2r5JXui4+Ph7u7O5555hnMnz8fxcXFUodkMm7evAmZTIbGjRsDAJKTk9G4cWN06dJFu09ISAisrKyq/Jm0oXo4Z5X4WVbVvXv3sHHjRiiVSnTs2BEA21htqstZJbax+yoqKvDSSy9hzpw56NChQ5XtptTObIx6Nqo3ubm5AAAPDw+d9R4eHtptADBz5kwEBATA1dUVv/76K+bPn4+cnBy88847Ro3XXOTm5sLOzq7K/1Qfzivd9+KLL6JFixZQq9U4deoU5s2bh/T0dCQkJEgdmuTu3r2LefPmYdy4cXB2dgagaWNNmzbV2c/Gxgaurq5sY6g+ZwA/yx62c+dOhIeHo7i4GJ6enkhKSoK7uzsAtrGaPCpnANvYw1avXg0bGxvMnDmz2u2m1M5Y3DYw0dHR2p/9/PxgZ2eHV199FStXrjTpR+mR+XhwfNWzzz4LT09P9OnTB+fPn8fTTz8tYWTSKi0txZgxYyCEwPr166UOxyw8Kmf8LNPVu3dvpKam4vr16/joo48wZswYpKSkVCk26L7acsY2dt+xY8fw7rvv4vjx45DJZFKHUysOS7AQKpUKAKrcxX/16lXttuoEBgairKwMFy5cqM/wzJZKpcK9e/dQUFCgs762vNJ9gYGBAIBz585JHIl0Kou0rKwsJCUl6fRAqlQqXLt2TWf/srIy5OfnN+g29qicVaehf5Y5ODigZcuW6N69O+Li4mBjY4O4uDgAbGM1eVTOqtOQ29jBgwdx7do1eHt7w8bGBjY2NsjKysJf//pX+Pj4ADCtdsbi1kL4+vpCpVJh37592nWFhYVISUlBUFBQje9LTU2FlZUVv93XoHPnzrC1tdXJa3p6OrKzsx+ZV7qvcrowT09PaQORSGWRlpGRgb1798LNzU1ne1BQEAoKCnDs2DHtun/961+oqKjQfjFoaGrLWXX4WaaroqICJSUlANjG9PVgzqrTkNvYSy+9hFOnTiE1NVX7UqvVmDNnDn788UcAptXOOCzBjBQVFen0fmVmZiI1NRWurq7w9vbG7NmzsWzZMrRq1Qq+vr5YsGAB1Go1hg8fDkAz2DslJQW9e/eGk5MTkpOTERUVhQkTJsDFxUWiq6pfteUsPz8f2dnZ2jk009PTAWi+gapUKiiVSkRGRiI6Ohqurq5wdnbGjBkzEBQUhO7du0tyTfXpSfN1/vx5bNmyBYMGDYKbmxtOnTqFqKgo9OzZE35+fpJcU317VM48PT0xatQoHD9+HDt37kR5ebl27Jmrqyvs7OzQrl07DBgwAFOmTMGGDRtQWlqKN954A+Hh4dVOtWMJnjRnDe2z7FH5cnNzw/LlyzFs2DB4enri+vXrWLduHS5fvozRo0cDANsYDM9ZQ2tjQO2f/w9/ybS1tYVKpUKbNm0AmFg7M+rcDPRE9u/fLwBUeUVERAghNNOBLViwQHh4eAi5XC769Okj0tPTte8/duyYCAwMFEqlUigUCtGuXTuxYsUKcffuXYmuqP7VlrNPP/202u2LFi3SHuPOnTti+vTpwsXFRTRq1EiMGDFC5OTkSHNB9exJ85WdnS169uwpXF1dhVwuFy1bthRz5swRN2/elO6i6tmjclY5ZVp1r/3792uPkZeXJ8aNGyccHR2Fs7OzmDx5srh165Z0F1XPnjRnDe2z7FH5unPnjhgxYoRQq9XCzs5OeHp6imHDhokjR47oHINtzLCcNbQ2JkTtn/8Pq26qNFNpZzIhhKiLIpmIiIiISGocc0tEREREFoPFLRERERFZDBa3RERERGQxWNwSERERkcVgcUtEREREFoPFLRERERFZDBa3RERERGQxWNwSERERkcVgcUtEFk0mk2H79u1Sh/HYzCF+fWLMy8tD06ZNceHCBaPE9KBNmzahcePG2uXFixejU6dOT3TMCxcuQCaTITU1tdZ9r1+/jqZNm+LSpUtPdE4i0g+LWyKqE8nJybC2tsbgwYMNfq+Pjw9iY2PrPig9/Pnnn3jttdfg7e0NuVwOlUqF/v3749ChQ5LEY6hJkyZh+PDhUodRq+XLlyM0NBQ+Pj4A7heHlS8nJyd06NABr7/+OjIyMur03GPHjsXZs2fr9JiGcHd3x8SJE7Fo0SLJYiBqSFjcElGdiIuLw4wZM3DgwAFcuXJF6nD0NnLkSJw4cQKbN2/G2bNn8cMPPyA4OBh5eXlSh2YxiouLERcXh8jIyCrb9u7di5ycHJw8eRIrVqzAmTNn0LFjR+zbt6/Ozm9vb4+mTZvW2fEex+TJkxEfH4/8/HxJ4yBqCFjcEtETKyoqwtatW/Haa69h8ODB2LRpU5V9duzYga5du0KhUMDd3R0jRowAAAQHByMrKwtRUVHaXjyg+j8dx8bGanv+AOC3335D37594e7uDqVSiV69euH48eN6x11QUICDBw9i9erV6N27N1q0aIFu3bph/vz5GDZsmHa/7OxshIaGwtHREc7OzhgzZgyuXr2q3V5d7+ns2bMRHBysXQ4ODsbMmTMxd+5cuLq6QqVSYfHixTrvycjIQM+ePaFQKNC+fXskJSXpfS01SUtLw8CBA+Ho6AgPDw+89NJLuH79OgBg48aNUKvVqKio0HlPaGgoXn75Ze3y999/j4CAACgUCjz11FNYsmQJysrK9I5h9+7dkMvl6N69e5Vtbm5uUKlUeOqppxAaGoq9e/ciMDAQkZGRKC8v1zuGgoICvPrqq/Dw8IBCocAzzzyDnTt3Aqg6LKHShx9+CC8vLzRq1AhjxozBzZs3dbZ//PHHaNeuHRQKBdq2bYsPPvigxmu8ceMGxo8fjyZNmsDe3h6tWrXCp59+qt3eoUMHqNVqfPfdd3rnjYgeD4tbInpiX3/9Ndq2bYs2bdpgwoQJ+OSTTyCE0G7ftWsXRowYgUGDBuHEiRPYt28funXrBgBISEhA8+bNERMTg5ycHOTk5Oh93lu3biEiIgK//PILDh8+jFatWmHQoEG4deuWXu93dHSEo6Mjtm/fjpKSkmr3qaioQGhoKPLz8/Hzzz8jKSkJf/zxB8aOHat3nJU2b94MBwcHpKSkYM2aNYiJidEWsBUVFQgLC4OdnR1SUlKwYcMGzJs3z+BzPKigoAAvvPAC/P39cfToUSQmJuLq1asYM2YMAGD06NHIy8vD/v37te/Jz89HYmIixo8fDwA4ePAgJk6ciFmzZuH06dP48MMPsWnTJixfvlzvOA4ePIjOnTvrta+VlRVmzZqFrKwsHDt2TK8YKioqMHDgQBw6dAhffPEFTp8+jVWrVsHa2rrG85w7dw5ff/01duzYgcTERJw4cQLTp0/Xbo+Pj8fChQuxfPlynDlzBitWrMCCBQuwefPmao+3YMECnD59Gnv27MGZM2ewfv16uLu76+zTrVs3HDx4UK88ENETEERET6hHjx4iNjZWCCFEaWmpcHd3F/v379duDwoKEuPHj6/x/S1atBBr167VWbdo0SLRsWNHnXVr164VLVq0qPE45eXlwsnJSezYsUO7DoD47rvvanzPt99+K1xcXIRCoRA9evQQ8+fPFydPntRu/+mnn4S1tbXIzs7Wrvv9998FAHHkyBEhhBAREREiNDRU57izZs0SvXr10i736tVLPPfcczr7dO3aVcybN08IIcSPP/4obGxsxOXLl7Xb9+zZU2v81Z270tKlS0W/fv101l28eFEAEOnp6UIIIUJDQ8XLL7+s3f7hhx8KtVotysvLhRBC9OnTR6xYsULnGJ9//rnw9PTULtcW48PnEEKIzMxMAUCcOHGiyv5nzpwRAMTWrVv1iuHHH38UVlZW2mt62KeffiqUSqV2edGiRcLa2lpcunRJu27Pnj3CyspK5OTkCCGEePrpp8WWLVt0jrN06VIRFBRUbfxDhw4VkydPrjEHQggRFRUlgoODH7kPET059twS0RNJT0/HkSNHMG7cOACAjY0Nxo4di7i4OO0+qamp6NOnT52f++rVq5gyZQpatWoFpVIJZ2dnFBUVITs7W+9jjBw5EleuXMEPP/yAAQMG4N///jcCAgK0QyvOnDkDLy8veHl5ad/Tvn17NG7cGGfOnDEoXj8/P51lT09PXLt2Tec8arVauz0oKMig4z/s5MmT2L9/v7aH2tHREW3btgUAnD9/HgAwfvx4bNu2TdtzHR8fj/DwcFhZWWmPERMTo3OMKVOmICcnB8XFxXrFcefOHSgUCr3jFv/f6185RKW2GFJTU9G8eXO0bt1a73N4e3ujWbNm2uWgoCBUVFQgPT0dt2/fxvnz5xEZGalzzmXLlmnz9rDXXnsNX331FTp16oS5c+fi119/rbKPvb293jkjosdnI3UARGTe4uLiUFZWplOUCSEgl8vx/vvvQ6lUwt7e3uDjWllZ6QxtAIDS0lKd5YiICOTl5eHdd99FixYtIJfLERQUhHv37hl0LoVCgb59+6Jv375YsGABXnnlFSxatAiTJk2qs1gBwNbWVmdZJpNVGe9al4qKijB06FCsXr26yjZPT08AwNChQyGEwK5du9C1a1ccPHgQa9eu1TnGkiVLEBYWVuUY+has7u7uuHHjht5xV35p8PX11SuGx2lfj1JUVAQA+OijjxAYGKizraahDgMHDkRWVhZ2796NpKQk9OnTB6+//jrefvtt7T75+flo0qRJncZKRFWxuCWix1ZWVobPPvsM//znP9GvXz+dbcOHD8eXX36JadOmwc/PD/v27cPkyZOrPY6dnZ3OzUMA0KRJE+Tm5kIIoe3Be3hO0UOHDuGDDz7AoEGDAAAXL17U3iz1JNq3b6+dt7Vdu3a4ePEiLl68qO29PX36NAoKCtC+fXttrGlpaTrHSE1NrVLMPkrleXJycrSF5+HDh5/oOgICArBt2zb4+PjAxqb6j3uFQoGwsDDEx8fj3LlzaNOmDQICAnSOkZ6ejpYtWz52HP7+/vjiiy/02reiogLvvfcefH194e/vr1cMfn5+uHTpEs6ePat37212djauXLmi/VJ2+PBhWFlZoU2bNvDw8IBarcYff/yhHXusjyZNmiAiIgIRERF4/vnnMWfOHJ3iNi0tTecmQyKqHyxuieix7dy5Ezdu3EBkZCSUSqXOtpEjRyIuLg7Tpk3DokWL0KdPHzz99NMIDw9HWVkZdu/erb1hysfHBwcOHEB4eDjkcjnc3d0RHByMP//8E2vWrMGoUaOQmJiIPXv2wNnZWXuOVq1a4fPPP0eXLl1QWFiIOXPmGNSLl5eXh9GjR+Pll1+Gn58fnJyccPToUaxZswahoaEAgJCQEDz77LMYP348YmNjUVZWhunTp6NXr17o0qULAOCFF17AW2+9hc8++wxBQUH44osvkJaWpi3O9BESEoLWrVsjIiICb731FgoLC/GPf/xDr/fevHmzSuHv5uaG119/HR999BHGjRunnaXh3Llz+Oqrr/Dxxx9reyHHjx+PIUOG4Pfff8eECRN0jrNw4UIMGTIE3t7eGDVqFKysrHDy5EmkpaVh2bJlesXXv39/zJ8/Hzdu3ICLi4vOtry8POTm5qK4uBhpaWmIjY3FkSNHsGvXLm18tcXQq1cv9OzZEyNHjsQ777yDli1b4r///S9kMhkGDBhQbUwKhQIRERF4++23UVhYiJkzZ2LMmDFQqVQAgCVLlmDmzJlQKpUYMGAASkpKcPToUdy4cQPR0dFVjrdw4UJ07twZHTp0QElJCXbu3Il27dpptxcXF+PYsWNYsWKFXjkjoicg5YBfIjJvQ4YMEYMGDap2W0pKigCgvTlr27ZtolOnTsLOzk64u7uLsLAw7b7JycnCz89PyOVy8eDH0vr164WXl5dwcHAQEydOFMuXL9e5oez48eOiS5cuQqFQiFatWolvvvmmys1peMTNTnfv3hV///vfRUBAgFAqlaJRo0aiTZs24s033xTFxcXa/bKyssSwYcOEg4ODcHJyEqNHjxa5ubk6x1q4cKHw8PAQSqVSREVFiTfeeKPKDWWzZs3SeU9oaKiIiIjQLqenp4vnnntO2NnZidatW4vExES9bigDUOUVGRkphBDi7NmzYsSIEaJx48bC3t5etG3bVsyePVtUVFRoj1FeXi48PT0FAHH+/Pkq50hMTBQ9evQQ9vb2wtnZWXTr1k1s3LhRrxxX6tatm9iwYYN2ufKGrMpXo0aNRLt27cT06dNFRkaGwTHk5eWJyZMnCzc3N6FQKMQzzzwjdu7cKYSo/oayjh07ig8++ECo1WqhUCjEqFGjRH5+vs454+PjtW3WxcVF9OzZUyQkJOjEX3lD2dKlS0W7du2Evb29cHV1FaGhoeKPP/7QHmvLli2iTZs2j8wREdUNmRAPDRQjIiKqY7t27cKcOXOQlpamvVmtIenevTtmzpyJF198UepQiCwehyUQEVG9Gzx4MDIyMnD58mWdmScaguvXryMsLEw7owgR1S/23BIRERGRxWh4fxsiIiIiIovF4paIiIiILAaLWyIiIiKyGCxuiYiIiMhisLglIiIiIovB4paIiIiILAaLWyIiIiKyGCxuiYiIiMhisLglIiIiIovxf7W5wAp5nh4KAAAAAElFTkSuQmCC\n"
          },
          "metadata": {}
        }
      ],
      "source": [
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "\n",
        "# Assuming 'predictions' is a DataFrame containing 'SoundLevelDecibels' and 'prediction' columns\n",
        "\n",
        "# Extract actual and predicted values from the DataFrame\n",
        "actual_values = np.array(predictions.select('SoundLevelDecibels').collect())\n",
        "predicted_values = np.array(predictions.select('prediction').collect())\n",
        "\n",
        "# Plot\n",
        "plt.figure(figsize=(8, 6))\n",
        "plt.scatter(actual_values, predicted_values, color='blue')\n",
        "plt.plot([actual_values.min(), actual_values.max()], [actual_values.min(), actual_values.max()], linestyle='--', color='red')\n",
        "plt.xlabel('Actual Sound Level (Decibels)')\n",
        "plt.ylabel('Predicted Sound Level (Decibels)')\n",
        "plt.title('Actual vs. Predicted Sound Level')\n",
        "plt.grid(True)\n",
        "plt.show()\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "sxUpx2g8DK2A"
      },
      "source": [
        "#### Part 4 - Evaluation\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 44,
      "metadata": {
        "tags": [],
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "0kxkCAkCDK2B",
        "outputId": "9eb6293c-2cbf-415c-9665-252e643b5e00"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Part 4 - Evaluation\n",
            "Number of stages in the pipeline =  3\n",
            "Coefficient for Frequency is -3.7978\n",
            "Coefficient for AngleOfAttack is -2.1493\n",
            "Coefficient for ChordLength is -3.3588\n",
            "Coefficient for FreeStreamVelocity is 1.5243\n",
            "Coefficient for SuctionSideDisplacement is -2.2251\n"
          ]
        }
      ],
      "source": [
        "print(\"Part 4 - Evaluation\")\n",
        "\n",
        "loadedmodel = loadedPipelineModel.stages[-1]\n",
        "totalstages = len(loadedPipelineModel.stages)\n",
        "inputcolumns = loadedPipelineModel.stages[0].getInputCols()\n",
        "\n",
        "print(\"Number of stages in the pipeline = \", totalstages)\n",
        "for i,j in zip(inputcolumns, loadedmodel.coefficients):\n",
        "    print(f\"Coefficient for {i} is {round(j,4)}\")"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "vcc0YLLDDK2B"
      },
      "source": [
        "### Stop Spark Session\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 45,
      "metadata": {
        "tags": [],
        "id": "kO7tGzFdDK2B"
      },
      "outputs": [],
      "source": [
        "spark.stop()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "fLLFouN3DK2C"
      },
      "source": [
        "## Authors\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "iw0C81BGDK2C"
      },
      "source": [
        "[salimt](https://github.com/salimt)\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ghoY2mH9DK2C"
      },
      "source": [
        "|Date (YYYY-MM-DD)|Version|Changed By|Change Description|\n",
        "|-|-|-|-|\n",
        "|2024-02-05|1|salimt|Published|\n"
      ]
    }
  ],
  "metadata": {
    "kernelspec": {
      "display_name": "Python",
      "language": "python",
      "name": "conda-env-python-py"
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.7.12"
    },
    "colab": {
      "provenance": []
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}