From 175bf7b7da010f2d4e900135bfce85fef91dc323 Mon Sep 17 00:00:00 2001
From: kejingfan <jingfan_ke@bjtu.edu.cn>
Date: Tue, 10 Oct 2023 09:42:31 +0800
Subject: [PATCH] first commit

---
 .gitignore                                    |    1 +
 .../Pytorch基本操作实验报告-checkpoint.ipynb  | 1262 +++++++++++++++++
 Lab1/.vscode/settings.json                    |    6 +
 Lab1/Pytorch基本操作实验报告.ipynb            | 1262 +++++++++++++++++
 Lab1/code/1.1.py                              |   39 +
 Lab1/code/1.2.py                              |   23 +
 Lab1/code/1.3.py                              |   12 +
 Lab1/code/2.1.py                              |  143 ++
 Lab1/code/2.2.py                              |   89 ++
 Lab1/code/3.1.py                              |  169 +++
 Lab1/code/3.2.py                              |   96 ++
 images/school_logo.png                        |  Bin 0 -> 116916 bytes
 12 files changed, 3102 insertions(+)
 create mode 100644 .gitignore
 create mode 100644 Lab1/.ipynb_checkpoints/Pytorch基本操作实验报告-checkpoint.ipynb
 create mode 100644 Lab1/.vscode/settings.json
 create mode 100644 Lab1/Pytorch基本操作实验报告.ipynb
 create mode 100644 Lab1/code/1.1.py
 create mode 100644 Lab1/code/1.2.py
 create mode 100644 Lab1/code/1.3.py
 create mode 100644 Lab1/code/2.1.py
 create mode 100644 Lab1/code/2.2.py
 create mode 100644 Lab1/code/3.1.py
 create mode 100644 Lab1/code/3.2.py
 create mode 100644 images/school_logo.png

diff --git a/.gitignore b/.gitignore
new file mode 100644
index 0000000..93db21b
--- /dev/null
+++ b/.gitignore
@@ -0,0 +1 @@
+dataset/
diff --git a/Lab1/.ipynb_checkpoints/Pytorch基本操作实验报告-checkpoint.ipynb b/Lab1/.ipynb_checkpoints/Pytorch基本操作实验报告-checkpoint.ipynb
new file mode 100644
index 0000000..c0645c6
--- /dev/null
+++ b/Lab1/.ipynb_checkpoints/Pytorch基本操作实验报告-checkpoint.ipynb
@@ -0,0 +1,1262 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "id": "3b57686b-7ac8-4897-bf76-3d982b1ff8da",
+   "metadata": {},
+   "source": [
+    "<h1><center>本科生《深度学习》课程<br>实验报告</center></h1>\n",
+    "<div style=\"text-align: center;\">\n",
+    "    <div><span style=\"display: inline-block; width: 65px; text-align: center;\">课程名称</span><span style=\"display: inline-block; width: 25px;\">:</span><span style=\"display: inline-block; width: 210px; font-weight: bold; text-align: left;\">深度学习</span></div>\n",
+    "    <div><span style=\"display: inline-block; width: 65px; text-align: center;\">实验题目</span><span style=\"display: inline-block; width: 25px;\">:</span><span style=\"display: inline-block; width: 210px; font-weight: bold; text-align: left;\">Pytorch基本操作</span></div>\n",
+    "    <div><span style=\"display: inline-block; width: 65px; text-align: center;\">学号</span><span style=\"display: inline-block; width: 25px;\">:</span><span style=\"display: inline-block; width: 210px; font-weight: bold; text-align: left;\">21281280</span></div>\n",
+    "    <div><span style=\"display: inline-block; width: 65px; text-align: center;\">姓名</span><span style=\"display: inline-block; width: 25px;\">:</span><span style=\"display: inline-block; width: 210px; font-weight: bold; text-align: left;\">柯劲帆</span></div>\n",
+    "    <div><span style=\"display: inline-block; width: 65px; text-align: center;\">班级</span><span style=\"display: inline-block; width: 25px;\">:</span><span style=\"display: inline-block; width: 210px; font-weight: bold; text-align: left;\">物联网2101班</span></div>\n",
+    "    <div><span style=\"display: inline-block; width: 65px; text-align: center;\">指导老师</span><span style=\"display: inline-block; width: 25px;\">:</span><span style=\"display: inline-block; width: 210px; font-weight: bold; text-align: left;\">张淳杰</span></div>\n",
+    "    <div><span style=\"display: inline-block; width: 65px; text-align: center;\">报告日期</span><span style=\"display: inline-block; width: 25px;\">:</span><span style=\"display: inline-block; width: 210px; font-weight: bold; text-align: left;\">2023年10月9日</span></div>\n",
+    "</div>"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "e24aa17e-faf9-4d69-9eae-43159116b56f",
+   "metadata": {},
+   "source": [
+    "实验环境：\n",
+    "- OS：Ubuntu 22.04 内核版本 6.2.0-34-generic\n",
+    "- CPU：12th Gen Intel(R) Core(TM) i7-12700H\n",
+    "- GPU：NVIDIA GeForce RTX 3070 Ti Laptop\n",
+    "- conda: miniconda 23.9.0\n",
+    "- python：3.10.13\n",
+    "- pytorch：2.1.0"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "id": "a4e12268-bad4-44c4-92d5-883624d93e25",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import numpy as np\n",
+    "import torch\n",
+    "from torch.autograd import Variable\n",
+    "from torch.utils.data import Dataset, DataLoader\n",
+    "from torch import nn\n",
+    "from torchvision import datasets, transforms"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "cc7f0ce5-d613-425b-807c-78115632cd80",
+   "metadata": {},
+   "source": [
+    "引用相关库。"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "59a43d35-56ac-4ade-995d-1c6fcbcd1262",
+   "metadata": {},
+   "source": [
+    "# 一、Pytorch基本操作考察\n",
+    "## 题目2\n",
+    "**使用 𝐓𝐞𝐧𝐬𝐨𝐫 初始化一个 𝟏×𝟑 的矩阵 𝑴 和一个 𝟐×𝟏 的矩阵 𝑵，对两矩阵进行减法操作（要求实现三种不同的形式），给出结果并分析三种方式的不同（如果出现报错，分析报错的原因），同时需要指出在计算过程中发生了什么。**"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "id": "79ea46db-cf49-436c-9b5b-c6562d0da9e2",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "方法1的结果:\n",
+      "tensor([[-3, -2, -1],\n",
+      "        [-4, -3, -2]])\n",
+      "方法2的结果:\n",
+      "tensor([[-3, -2, -1],\n",
+      "        [-4, -3, -2]])\n",
+      "方法3的结果:\n",
+      "tensor([[-3, -2, -1],\n",
+      "        [-4, -3, -2]])\n"
+     ]
+    }
+   ],
+   "source": [
+    "A = torch.tensor([[1, 2, 3]])\n",
+    "\n",
+    "B = torch.tensor([[4],\n",
+    "                  [5]])\n",
+    "\n",
+    "# 方法1: 使用PyTorch的减法操作符\n",
+    "result1 = A - B\n",
+    "\n",
+    "# 方法2: 使用PyTorch的sub函数\n",
+    "result2 = torch.sub(A, B)\n",
+    "\n",
+    "# 方法3: 手动实现广播机制并作差\n",
+    "def mysub(a:torch.Tensor, b:torch.Tensor):\n",
+    "    if not (\n",
+    "        (a.size(0) == 1 and b.size(1) == 1) \n",
+    "        or \n",
+    "        (a.size(1) == 1 and b.size(0) == 1)\n",
+    "        ):\n",
+    "        raise ValueError(\"输入的张量大小无法满足广播机制的条件。\")\n",
+    "    else:\n",
+    "        target_shape = torch.Size([max(A.size(0), B.size(0)), max(A.size(1), B.size(1))])\n",
+    "        A_broadcasted = A.expand(target_shape)\n",
+    "        B_broadcasted = B.expand(target_shape)\n",
+    "        result = torch.zeros(target_shape, dtype=torch.int64).to(device=A_broadcasted.device)\n",
+    "        for i in range(target_shape[0]):\n",
+    "            for j in range(target_shape[1]):\n",
+    "                result[i, j] = A_broadcasted[i, j] - B_broadcasted[i, j]\n",
+    "        return result\n",
+    "\n",
+    "result3 = mysub(A, B)\n",
+    "\n",
+    "print(\"方法1的结果:\")\n",
+    "print(result1)\n",
+    "print(\"方法2的结果:\")\n",
+    "print(result2)\n",
+    "print(\"方法3的结果:\")\n",
+    "print(result3)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "2489a3ad-f6ff-4561-bb26-e02654090b98",
+   "metadata": {},
+   "source": [
+    "## 题目2\n",
+    "1. **利用Tensor创建两个大小分别3*2和4*2的随机数矩阵P和Q，要求服从均值为0，标准差0.01为的正态分布；**\n",
+    "2. **对第二步得到的矩阵Q进行形状变换得到Q的转置Q^T；**\n",
+    "3. **对上述得到的矩阵P和矩阵Q^T求矩阵相乘。**"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "id": "41e4ee02-1d05-4101-b3f0-477bac0277fb",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "矩阵 P:\n",
+      "tensor([[ 0.0053,  0.0013],\n",
+      "        [-0.0086,  0.0136],\n",
+      "        [-0.0013,  0.0176]])\n",
+      "矩阵 Q:\n",
+      "tensor([[ 0.0044,  0.0014],\n",
+      "        [ 0.0147,  0.0078],\n",
+      "        [-0.0002, -0.0023],\n",
+      "        [ 0.0001, -0.0011]])\n",
+      "矩阵 QT:\n",
+      "tensor([[ 0.0044,  0.0147, -0.0002,  0.0001],\n",
+      "        [ 0.0014,  0.0078, -0.0023, -0.0011]])\n",
+      "矩阵相乘的结果:\n",
+      "tensor([[ 2.4953e-05,  8.7463e-05, -3.8665e-06, -8.9576e-07],\n",
+      "        [-1.9514e-05, -2.0557e-05, -2.9649e-05, -1.5913e-05],\n",
+      "        [ 1.8189e-05,  1.1834e-04, -4.0097e-05, -1.9608e-05]])\n"
+     ]
+    }
+   ],
+   "source": [
+    "mean = 0\n",
+    "stddev = 0.01\n",
+    "\n",
+    "P = torch.normal(mean=mean, std=stddev, size=(3, 2))\n",
+    "Q = torch.normal(mean=mean, std=stddev, size=(4, 2))\n",
+    "\n",
+    "print(\"矩阵 P:\")\n",
+    "print(P)\n",
+    "print(\"矩阵 Q:\")\n",
+    "print(Q)\n",
+    "\n",
+    "# 对矩阵Q进行转置操作，得到矩阵Q的转置Q^T\n",
+    "QT = Q.T\n",
+    "print(\"矩阵 QT:\")\n",
+    "print(QT)\n",
+    "\n",
+    "# 计算矩阵P和矩阵Q^T的矩阵相乘\n",
+    "result = torch.matmul(P, QT)\n",
+    "print(\"矩阵相乘的结果:\")\n",
+    "print(result)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "cea9cb6d-adde-4e08-b9f2-8c417abf4231",
+   "metadata": {},
+   "source": [
+    "## 题目2\n",
+    "**给定公式$ y_3=y_1+y_2=𝑥^2+𝑥^3$，且$x=1$。利用学习所得到的Tensor的相关知识，求$y_3$对$x$的梯度，即$\\frac{dy_3}{dx}$。**"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "id": "951512cd-d915-4d04-959f-eb99d1971e2d",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "梯度(dy_3/dx):  2.0\n"
+     ]
+    }
+   ],
+   "source": [
+    "x = torch.tensor(1.0, requires_grad=True)\n",
+    "y_1 = x ** 2\n",
+    "with torch.no_grad():\n",
+    "    y_2 = x**3\n",
+    "\n",
+    "y3 = y_1 + y_2\n",
+    "\n",
+    "y3.backward()\n",
+    "\n",
+    "print(\"梯度(dy_3/dx): \", x.grad.item())"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "3269dbf6-889a-49eb-8094-1e588e1a6c30",
+   "metadata": {},
+   "source": [
+    "# 二、动手实现logistic回归\n",
+    "## 题目1\n",
+    "**要求动手从0实现 logistic 回归（只借助Tensor和Numpy相关的库）在人工构造的数据集上进行训练和测试，并从loss以及训练集上的准确率等多个角度对结果进行分析（可借助nn.BCELoss或nn.BCEWithLogitsLoss作为损失函数，从零实现二元交叉熵为选作）**"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "bcd12aa9-f187-4d88-8c59-af6d16107edb",
+   "metadata": {},
+   "source": [
+    "给定预测概率$ \\left( \\hat{y} \\right) $和目标标签$ \\left( y \\right)$（通常是0或1），BCELoss的计算公式如下：\n",
+    "$$\n",
+    " \\text{BCELoss}(\\hat{y}, y) = -\\frac{1}{N} \\sum_{i=1}^{N} \\left(y_i \\cdot \\log(\\hat{y}_i) + (1 - y_i) \\cdot \\log(1 - \\hat{y}_i)\\right) \n",
+    "$$\n",
+    "其中，$\\left( N \\right) $是样本数量，$\\left( \\hat{y}_i \\right) $表示模型的预测概率向量中的第$ \\left( i \\right) $个元素，$\\left( y_i \\right) $表示实际的目标标签中的第$ \\left( i \\right) $个元素。在二分类问题中，$\\left( y_i \\right) $通常是0或1。这个公式表示对所有样本的二分类交叉熵损失进行了求和并取平均。\n",
+    "\n",
+    "因此BCELoss的手动实现如下。"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 5,
+   "id": "e31b86ec-4114-48dd-8d73-fe4e0686419a",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "输入：\n",
+      "tensor([0.6900])\n",
+      "标签：\n",
+      "tensor([1.])\n",
+      "My_BCELoss损失值: 0.37110066413879395\n",
+      "nn.BCELoss损失值: 0.37110066413879395\n"
+     ]
+    }
+   ],
+   "source": [
+    "class My_BCELoss:\n",
+    "    def __call__(self, prediction: torch.Tensor, target: torch.Tensor):\n",
+    "        loss = -torch.mean(target * torch.log(prediction) + (1 - target) * torch.log(1 - prediction))\n",
+    "        return loss\n",
+    "\n",
+    "\n",
+    "# 测试\n",
+    "prediction = torch.sigmoid(torch.tensor([0.8]))\n",
+    "target = torch.tensor([1.0])\n",
+    "print(f\"输入：\\n{prediction}\")\n",
+    "print(f\"标签：\\n{target}\")\n",
+    "\n",
+    "my_bce_loss = My_BCELoss()\n",
+    "my_loss = my_bce_loss(prediction, target)\n",
+    "print(\"My_BCELoss损失值:\", my_loss.item())\n",
+    "\n",
+    "nn_bce_loss = nn.BCELoss()\n",
+    "nn_loss = nn_bce_loss(prediction, target)\n",
+    "print(\"nn.BCELoss损失值:\", nn_loss.item())"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "345b0300-8808-4c43-9bf9-05a7e6e1f5af",
+   "metadata": {},
+   "source": [
+    "Optimizer的实现较为简单。\n",
+    "\n",
+    "主要实现：\n",
+    "- 传入参数：`__init__()`\n",
+    "- 对传入的参数进行更新：`step()`\n",
+    "- 清空传入参数存储的梯度：`zero_grad()`"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 6,
+   "id": "0297066c-9fc1-448d-bdcb-29a6f1519117",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "y.backward()之后，x的梯度:  2.0\n",
+      "optimizer_test.step()之后，x的值:  0.800000011920929\n",
+      "optimizer_test.zero_grad()之后，x的梯度:  0.0\n"
+     ]
+    }
+   ],
+   "source": [
+    "class My_optimizer:\n",
+    "    def __init__(self, params: list[torch.Tensor], lr: float):\n",
+    "        self.params = params\n",
+    "        self.lr = lr\n",
+    "\n",
+    "    def step(self):\n",
+    "        for param in self.params:\n",
+    "            param.data = param.data - self.lr * param.grad.data\n",
+    "\n",
+    "    def zero_grad(self):\n",
+    "        for param in self.params:\n",
+    "            if param.grad is not None:\n",
+    "                param.grad.data.zero_()\n",
+    "\n",
+    "\n",
+    "# 测试\n",
+    "x = torch.tensor(1.0, requires_grad=True)\n",
+    "y = x ** 2\n",
+    "optimizer_test = My_optimizer([x], lr=0.1)\n",
+    "\n",
+    "y.backward()\n",
+    "print(\"y.backward()之后，x的梯度: \", x.grad.item())\n",
+    "\n",
+    "optimizer_test.step()\n",
+    "print(\"optimizer_test.step()之后，x的值: \", x.item())\n",
+    "\n",
+    "optimizer_test.zero_grad()\n",
+    "print(\"optimizer_test.zero_grad()之后，x的梯度: \", x.grad.item())"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "6ab83528-a88b-4d66-b0c9-b1315cf75c22",
+   "metadata": {},
+   "source": [
+    "线性层主要有一个权重（weight)和一个偏置（bias）。\n",
+    "线性层的数学公式如下：\n",
+    "$$\n",
+    "x:=x \\times weight^T+bias\n",
+    "$$\n",
+    "因此代码实现如下："
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 7,
+   "id": "8e18695a-d8c5-4f77-8b5c-de40d9240fb9",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "输入：\n",
+      "tensor([[1.],\n",
+      "        [2.]], requires_grad=True)\n",
+      "权重：\n",
+      "tensor([[-1.0980],\n",
+      "        [-0.5413],\n",
+      "        [ 1.5884]], requires_grad=True)\n",
+      "偏置：\n",
+      "tensor([-1.1733], requires_grad=True)\n",
+      "输出：\n",
+      "tensor([[-2.2713, -1.7146,  0.4151],\n",
+      "        [-3.3692, -2.2559,  2.0036]], grad_fn=<AddBackward0>)\n"
+     ]
+    }
+   ],
+   "source": [
+    "class My_Linear:\n",
+    "    def __init__(self, input_feature: int, output_feature: int):\n",
+    "        self.weight = torch.randn((output_feature, input_feature), requires_grad=True, dtype=torch.float32)\n",
+    "        self.bias = torch.randn(1, requires_grad=True, dtype=torch.float32)\n",
+    "        self.params = [self.weight, self.bias]\n",
+    "\n",
+    "    def __call__(self, x: torch.Tensor):\n",
+    "        return self.forward(x)\n",
+    "\n",
+    "    def forward(self, x: torch.Tensor):\n",
+    "        x = torch.matmul(x, self.weight.T) + self.bias\n",
+    "        return x\n",
+    "\n",
+    "    def to(self, device: str):\n",
+    "        for param in self.params:\n",
+    "            param.data = param.data.to(device=device)\n",
+    "        return self\n",
+    "\n",
+    "    def parameters(self):\n",
+    "        return self.params\n",
+    "\n",
+    "        \n",
+    "# 测试\n",
+    "linear_test = My_Linear(1, 3)\n",
+    "x = torch.tensor([[1.], [2.]], requires_grad=True)\n",
+    "print(f\"输入：\\n{x}\")\n",
+    "print(f\"权重：\\n{linear_test.weight}\")\n",
+    "print(f\"偏置：\\n{linear_test.bias}\")\n",
+    "y = linear_test(x)\n",
+    "print(f\"输出：\\n{y}\")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "5ff813cc-c1f0-4c73-a3e8-d6796ef5d366",
+   "metadata": {},
+   "source": [
+    "手动实现logistic回归模型。\n",
+    "\n",
+    "模型很简单，主要由一个线性层和一个sigmoid层组成。\n",
+    "\n",
+    "Sigmoid函数（又称为 Logistic函数）是一种常用的激活函数，通常用于神经网络的输出层或隐藏层，其作用是将输入的实数值压缩到一个范围在0和1之间的数值。"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 8,
+   "id": "e7de7e4b-a084-4793-812e-46e8550ecd8d",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "class Model_2_1():\n",
+    "    def __init__(self):\n",
+    "        self.linear = My_Linear(1, 1)\n",
+    "        self.params = self.linear.params\n",
+    "\n",
+    "    def __call__(self, x):\n",
+    "        return self.forward(x)\n",
+    "\n",
+    "    def forward(self, x):\n",
+    "        x = self.linear(x)\n",
+    "        x = torch.sigmoid(x)\n",
+    "        return x\n",
+    "\n",
+    "    def to(self, device: str):\n",
+    "        for param in self.params:\n",
+    "            param.data = param.data.to(device=device)\n",
+    "        return self\n",
+    "\n",
+    "    def parameters(self):\n",
+    "        return self.params"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "e14acea9-e5ef-4c24-aea9-329647224ce1",
+   "metadata": {},
+   "source": [
+    "人工随机构造数据集。\n",
+    "\n",
+    "这里我遇到了比较大的问题。因为数据构建不合适，会导致后面的训练出现梯度爆炸。\n",
+    "\n",
+    "我采用随机产生数据后归一化的方法，即\n",
+    "$$\n",
+    "\\hat{x} = \\frac{x - \\text{min}_x}{\\text{max}_x - \\text{min}_x} \n",
+    "$$\n",
+    "将数据控制在合适的区间。\n",
+    "\n",
+    "我的y设置为$4-3\\times x + noise$，noise为随机噪声。\n",
+    "\n",
+    "生成完x和y后进行归一化处理，并写好DataLoader访问数据集的接口`__getitem__()`。"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 9,
+   "id": "c39fbafb-62e4-4b8c-9d65-6718d25f2970",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "测试数据集大小：100\n",
+      "测试数据集第0对数据：\n",
+      "x_0 = 0.5531462811708403\n",
+      "y_0 = 0.42036701080526284\n"
+     ]
+    }
+   ],
+   "source": [
+    "class My_Dataset(Dataset):\n",
+    "    def __init__(self, data_size=1000000):\n",
+    "        np.random.seed(0)\n",
+    "        x = 2 * np.random.rand(data_size, 1)\n",
+    "        noise = 0.2 * np.random.randn(data_size, 1)\n",
+    "        y = 4 - 3 * x + noise\n",
+    "        self.min_x, self.max_x = np.min(x), np.max(x)\n",
+    "        min_y, max_y = np.min(y), np.max(y)\n",
+    "        x = (x - self.min_x) / (self.max_x - self.min_x)\n",
+    "        y = (y - min_y) / (max_y - min_y)\n",
+    "        self.data = [[x[i][0], y[i][0]] for i in range(x.shape[0])]\n",
+    "\n",
+    "    def __len__(self):\n",
+    "        return len(self.data)\n",
+    "\n",
+    "    def __getitem__(self, index):\n",
+    "        x, y = self.data[index]\n",
+    "        return x, y\n",
+    "\n",
+    "\n",
+    "# 测试\n",
+    "dataset_test = My_Dataset(data_size=100)\n",
+    "dataset_size = len(dataset_test)\n",
+    "print(f\"测试数据集大小：{dataset_size}\")\n",
+    "x0, y0 = dataset_test[0]\n",
+    "print(f\"测试数据集第0对数据：\")\n",
+    "print(f\"x_0 = {x0}\")\n",
+    "print(f\"y_0 = {y0}\")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "957a76a2-b306-47a8-912e-8fbf00cdfd42",
+   "metadata": {},
+   "source": [
+    "训练Logistic回归模型。\n",
+    "进行如下步骤：\n",
+    "1. 初始化超参数\n",
+    "2. 获取数据集\n",
+    "3. 初始化模型\n",
+    "4. 定义损失函数和优化器\n",
+    "5. 训练\n",
+    "    1. 从训练dataloader中获取批量数据\n",
+    "    2. 传入模型\n",
+    "    3. 使用损失函数计算与ground_truth的损失\n",
+    "    4. 使用优化器进行反向传播\n",
+    "    5. 循环以上步骤\n",
+    "6. 测试\n",
+    "    1. 设置测试数据\n",
+    "    2. 传入模型\n",
+    "    3. 得到预测值"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 10,
+   "id": "5612661e-2809-4d46-96c2-33ee9f44116d",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Epoch 1/10, Loss: 680.9198314547539, Acc: 0.9677169744703272\n",
+      "Epoch 2/10, Loss: 677.2582936882973, Acc: 0.9985965700887113\n",
+      "Epoch 3/10, Loss: 677.1911396384239, Acc: 0.9993738265049104\n",
+      "Epoch 4/10, Loss: 677.1777537465096, Acc: 0.9995470920810262\n",
+      "Epoch 5/10, Loss: 677.1745615005493, Acc: 0.9998228389835642\n",
+      "Epoch 6/10, Loss: 677.1743944883347, Acc: 0.9999690339979311\n",
+      "Epoch 7/10, Loss: 677.1735371947289, Acc: 0.9998205132243208\n",
+      "Epoch 8/10, Loss: 677.1737813353539, Acc: 0.999798559017381\n",
+      "Epoch 9/10, Loss: 677.1740361452103, Acc: 0.9998672931901137\n",
+      "Epoch 10/10, Loss: 677.1736125349998, Acc: 0.9997257713704987\n",
+      "Model weights: -0.0006128809181973338, bias: 0.023128816857933998\n",
+      "Prediction for test data: 0.505628764629364\n"
+     ]
+    }
+   ],
+   "source": [
+    "learning_rate = 5e-2\n",
+    "num_epochs = 10\n",
+    "batch_size = 1024\n",
+    "device = \"cuda:0\" if torch.cuda.is_available() else \"cpu\"\n",
+    "\n",
+    "dataset = My_Dataset()\n",
+    "dataloader = DataLoader(\n",
+    "    dataset=dataset, batch_size=batch_size, shuffle=True, num_workers=5, pin_memory=True\n",
+    ")\n",
+    "\n",
+    "model = Model_2_1().to(device)\n",
+    "criterion = My_BCELoss()\n",
+    "optimizer = My_optimizer(model.parameters(), lr=learning_rate)\n",
+    "\n",
+    "for epoch in range(num_epochs):\n",
+    "    total_epoch_loss = 0\n",
+    "    total_epoch_pred = 0\n",
+    "    total_epoch_target = 0\n",
+    "    for x, targets in dataloader:\n",
+    "        optimizer.zero_grad()\n",
+    "        \n",
+    "        x = x.to(device).to(dtype=torch.float32)\n",
+    "        targets = targets.to(device).to(dtype=torch.float32)\n",
+    "        \n",
+    "        x = x.unsqueeze(1)\n",
+    "        y_pred = model(x)\n",
+    "        loss = criterion(y_pred, targets)\n",
+    "        total_epoch_loss += loss.item()\n",
+    "        total_epoch_target += targets.sum().item()\n",
+    "        total_epoch_pred += y_pred.sum().item()\n",
+    "\n",
+    "        loss.backward()\n",
+    "        optimizer.step()\n",
+    "\n",
+    "    print(\n",
+    "        f\"Epoch {epoch + 1}/{num_epochs}, Loss: {total_epoch_loss}, Acc: {1 - abs(total_epoch_pred - total_epoch_target) / total_epoch_target}\"\n",
+    "    )\n",
+    "\n",
+    "with torch.no_grad():\n",
+    "    test_data = (np.array([[2]]) - dataset.min_x) / (dataset.max_x - dataset.min_x)\n",
+    "    test_data = Variable(\n",
+    "        torch.tensor(test_data, dtype=torch.float32), requires_grad=False\n",
+    "    ).to(device)\n",
+    "    predicted = model(test_data).to(\"cpu\")\n",
+    "    print(\n",
+    "        f\"Model weights: {model.linear.weight.item()}, bias: {model.linear.bias.item()}\"\n",
+    "    )\n",
+    "    print(f\"Prediction for test data: {predicted.item()}\")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "9e416582-a30d-4084-acc6-6e05f80a6aff",
+   "metadata": {},
+   "source": [
+    "## 题目2\n",
+    "**利用 torch.nn 实现 logistic 回归在人工构造的数据集上进行训练和测试，并对结果进行分析，并从loss以及训练集上的准确率等多个角度对结果进行分析**"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "0460d125-7d03-44fe-845c-c4d13792e241",
+   "metadata": {},
+   "source": [
+    "使用torch.nn实现模型。\n",
+    "\n",
+    "将之前的Model_2_1中的手动实现函数改为torch.nn内置函数即可，再加上继承nn.Module以使用torch.nn内置模型模板特性。"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 11,
+   "id": "fa121afd-a1af-4193-9b54-68041e0ed068",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "class Model_2_2(nn.Module):\n",
+    "    def __init__(self):\n",
+    "        super(Model_2_2, self).__init__()\n",
+    "        self.linear = nn.Linear(1, 1, dtype=torch.float64)\n",
+    "\n",
+    "    def forward(self, x):\n",
+    "        x = self.linear(x)\n",
+    "        x = torch.sigmoid(x)\n",
+    "        return x"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "176eee7e-4e3d-470e-8af2-8761bca039f8",
+   "metadata": {},
+   "source": [
+    "训练与测试过程与之前手动实现的几乎一致。仅有少量涉及数据类型（dtype）的代码需要更改以适应torch.nn的内置函数要求。"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 12,
+   "id": "93b0fdb6-be8b-4663-b59e-05ed19a9ea09",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Epoch 1/10, Loss: 600.8090852049173, Acc: 0.9945839732715815\n",
+      "Epoch 2/10, Loss: 565.9542879898308, Acc: 0.9999073566261442\n",
+      "Epoch 3/10, Loss: 565.9275637627202, Acc: 0.9999969933728429\n",
+      "Epoch 4/10, Loss: 565.927609191542, Acc: 0.9999961959888584\n",
+      "Epoch 5/10, Loss: 565.928202885308, Acc: 0.9999953721249991\n",
+      "Epoch 6/10, Loss: 565.9323843971484, Acc: 0.9999969051674709\n",
+      "Epoch 7/10, Loss: 565.9298919086365, Acc: 0.9999935973983517\n",
+      "Epoch 8/10, Loss: 565.9299267993255, Acc: 0.9999985970973472\n",
+      "Epoch 9/10, Loss: 565.9306044380719, Acc: 0.9999947955797296\n",
+      "Epoch 10/10, Loss: 565.9329843268798, Acc: 0.9999973784035556\n",
+      "Model weights: -3.7066140776793373, bias: 1.8709382558479912\n",
+      "Prediction for test data: 0.13756338580653613\n"
+     ]
+    }
+   ],
+   "source": [
+    "learning_rate = 1e-2\n",
+    "num_epochs = 10\n",
+    "batch_size = 1024\n",
+    "device = \"cuda:0\" if torch.cuda.is_available() else \"cpu\"\n",
+    "\n",
+    "dataset = My_Dataset()\n",
+    "dataloader = DataLoader(\n",
+    "    dataset=dataset, batch_size=batch_size, shuffle=True, num_workers=5, pin_memory=True\n",
+    ")\n",
+    "\n",
+    "model = Model_2_2().to(device)\n",
+    "criterion = nn.BCELoss()\n",
+    "optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate)\n",
+    "\n",
+    "for epoch in range(num_epochs):\n",
+    "    total_epoch_loss = 0\n",
+    "    total_epoch_pred = 0\n",
+    "    total_epoch_target = 0\n",
+    "    for x, targets in dataloader:\n",
+    "        optimizer.zero_grad()\n",
+    "\n",
+    "        x = x.to(device)\n",
+    "        targets = targets.to(device)\n",
+    "\n",
+    "        x = x.unsqueeze(1)\n",
+    "        targets = targets.unsqueeze(1)\n",
+    "        y_pred = model(x)\n",
+    "        loss = criterion(y_pred, targets)\n",
+    "        total_epoch_loss += loss.item()\n",
+    "        total_epoch_target += targets.sum().item()\n",
+    "        total_epoch_pred += y_pred.sum().item()\n",
+    "\n",
+    "        loss.backward()\n",
+    "        optimizer.step()\n",
+    "\n",
+    "    print(\n",
+    "        f\"Epoch {epoch + 1}/{num_epochs}, Loss: {total_epoch_loss}, Acc: {1 - abs(total_epoch_pred - total_epoch_target) / total_epoch_target}\"\n",
+    "    )\n",
+    "\n",
+    "with torch.no_grad():\n",
+    "    test_data = (np.array([[2]]) - dataset.min_x) / (dataset.max_x - dataset.min_x)\n",
+    "    test_data = Variable(\n",
+    "        torch.tensor(test_data, dtype=torch.float64), requires_grad=False\n",
+    "    ).to(device)\n",
+    "    predicted = model(test_data).to(\"cpu\")\n",
+    "    print(\n",
+    "        f\"Model weights: {model.linear.weight.item()}, bias: {model.linear.bias.item()}\"\n",
+    "    )\n",
+    "    print(f\"Prediction for test data: {predicted.item()}\")\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "e6bff679-f8d2-46cc-bdcb-82af7dab38b3",
+   "metadata": {},
+   "source": [
+    "对比发现，使用torch.nn的内置损失函数和优化器，正确率提升更快。\n",
+    "\n",
+    "但是为什么相同分布的数据集训练出的权重和偏置，以及预测结果存在较大差别，这个问题的原因还有待我探究。"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "ef41d7fa-c2bf-4024-833b-60af0a87043a",
+   "metadata": {},
+   "source": [
+    "# 三、动手实现softmax回归\n",
+    "\n",
+    "## 问题1\n",
+    "\n",
+    "**要求动手从0实现softmax回归（只借助Tensor和Numpy相关的库）在Fashion-MNIST数据集上进行训练和测试，并从loss、训练集以及测试集上的准确率等多个角度对结果进行分析（要求从零实现交叉熵损失函数）**"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "3c356760-75a8-4814-ba69-73b270396a4e",
+   "metadata": {},
+   "source": [
+    "手动实现nn.one_hot()。\n",
+    "\n",
+    "one-hot向量用于消除线性标签值所映射的类别的非线性。\n",
+    "\n",
+    "one-hot向量是使用一个长度为分类数量的数组表示标签值，其中有且仅有1个值为为1，该值的下标为标签值；其余为0。\n",
+    "\n",
+    "原理很简单，步骤如下：\n",
+    "1. 初始化全零的张量，大小为（标签数量，分类数量）；\n",
+    "2. 将标签值映射到全零张量的\\[下标，标签值\\]中，将该位置为1；\n",
+    "3. 返回修改后的张量，即是ont-hot向量。"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 13,
+   "id": "e605f1b0-1d32-410f-bddf-402a85ccc9ff",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "def my_one_hot(indices: torch.Tensor, num_classes: int):\n",
+    "    one_hot_tensor = torch.zeros(len(indices), num_classes).to(indices.device)\n",
+    "    one_hot_tensor.scatter_(1, indices.view(-1, 1), 1)\n",
+    "    return one_hot_tensor"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "902603a6-bfb9-4ce3-bd0d-b00cebb1d3cb",
+   "metadata": {},
+   "source": [
+    "手动实现CrossEntropyLoss。\n",
+    "\n",
+    "CrossEntropyLoss由一个log_softmax和一个nll_loss组成。\n",
+    "\n",
+    "softmax的数学表达式如下：\n",
+    "$$\n",
+    "\\text{softmax}(y_i) = \\frac{e^{y_i - \\text{max}(y)}}{\\sum_{j=1}^{N} e^{y_j - \\text{max}(y)}} \n",
+    "$$\n",
+    "log_softmax即为$\\log\\left(softmax\\left(y\\right)\\right)$。\n",
+    "\n",
+    "CrossEntropyLoss的数学表达式如下：\n",
+    "$$\n",
+    "\\text{CrossEntropyLoss}(y, \\hat{y}) = -\\frac{1}{N} \\sum_{i=1}^{N} \\hat{y}_i \\cdot \\log(\\text{softmax}(y_i)) \n",
+    "$$\n",
+    "\n",
+    "故代码如下："
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 14,
+   "id": "759a3bb2-b5f4-4ea5-a2d7-15f0c4cdd14b",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "输入：\n",
+      "tensor([[-1.2914,  0.4715, -0.0432,  1.7427, -1.9236],\n",
+      "        [ 0.5361, -0.7551, -0.6810,  1.0945,  0.6135],\n",
+      "        [-1.3398, -0.0026, -1.6066, -0.4659, -1.6076]], requires_grad=True)\n",
+      "标签：\n",
+      "tensor([[1., 0., 0., 0., 0.],\n",
+      "        [0., 0., 0., 0., 1.],\n",
+      "        [0., 0., 0., 0., 1.]])\n",
+      "My_CrossEntropyLoss损失值: 2.4310648441314697\n",
+      "nn.CrossEntropyLoss损失值: 2.4310646057128906\n"
+     ]
+    }
+   ],
+   "source": [
+    "class My_CrossEntropyLoss:\n",
+    "    def __call__(self, predictions: torch.Tensor, targets: torch.Tensor):\n",
+    "        max_values = torch.max(predictions, dim=1, keepdim=True).values\n",
+    "        exp_values = torch.exp(predictions - max_values)\n",
+    "        softmax_output = exp_values / torch.sum(exp_values, dim=1, keepdim=True)\n",
+    "        log_probs = torch.log(softmax_output)\n",
+    "        \n",
+    "        nll_loss = -torch.sum(targets * log_probs, dim=1)\n",
+    "        average_loss = torch.mean(nll_loss)\n",
+    "        return average_loss\n",
+    "\n",
+    "        \n",
+    "# 测试\n",
+    "input = torch.randn(3, 5, requires_grad=True)\n",
+    "target = torch.randn(3, 5).softmax(dim=1).argmax(1)\n",
+    "target = torch.nn.functional.one_hot(target, num_classes=5).to(dtype=torch.float32)\n",
+    "print(f\"输入：\\n{input}\")\n",
+    "print(f\"标签：\\n{target}\")\n",
+    "\n",
+    "my_crossentropyloss = My_CrossEntropyLoss()\n",
+    "my_loss = my_crossentropyloss(input, target)\n",
+    "print(\"My_CrossEntropyLoss损失值:\", my_loss.item())\n",
+    "\n",
+    "nn_crossentropyloss = nn.CrossEntropyLoss()\n",
+    "nn_loss = nn_crossentropyloss(input, target)\n",
+    "print(\"nn.CrossEntropyLoss损失值:\", nn_loss.item())"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "dbf78501-f5be-4008-986c-d331d531491f",
+   "metadata": {},
+   "source": [
+    "手动实现Flatten。\n",
+    "\n",
+    "原理很简单，就是把多维的张量拉直成一个向量。"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 15,
+   "id": "74322629-8325-4823-b80f-f28182d577c1",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "class My_Flatten:\n",
+    "    def __call__(self, x: torch.Tensor):\n",
+    "        return self.forward(x)\n",
+    "\n",
+    "    def forward(self, x: torch.Tensor):\n",
+    "        x = x.view(x.shape[0], -1)\n",
+    "        return x"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "35aee905-ae37-4faa-a7f1-a04cd8579f78",
+   "metadata": {},
+   "source": [
+    "手动实现softmax回归模型。\n",
+    "\n",
+    "模型很简单，主要由一个Flatten层和一个线性层组成。\n",
+    "\n",
+    "Flatten层主要用于将2维的图像展开，直接作为1维的特征量输入网络。"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 16,
+   "id": "bb31a75e-464c-4b94-b927-b219a765e35d",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "class Model_3_1:\n",
+    "    def __init__(self, num_classes):\n",
+    "        self.flatten = My_Flatten()\n",
+    "        self.linear = My_Linear(28 * 28, num_classes)\n",
+    "        self.params = self.linear.params\n",
+    "\n",
+    "    def __call__(self, x: torch.Tensor):\n",
+    "        return self.forward(x)\n",
+    "\n",
+    "    def forward(self, x: torch.Tensor):\n",
+    "        x = self.flatten(x)\n",
+    "        x = self.linear(x)\n",
+    "        return x\n",
+    "\n",
+    "    def to(self, device: str):\n",
+    "        for param in self.params:\n",
+    "            param.data = param.data.to(device=device)\n",
+    "        return self\n",
+    "\n",
+    "    def parameters(self):\n",
+    "        return self.params"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "17e686d1-9c9a-4727-8fdc-9990d348c523",
+   "metadata": {},
+   "source": [
+    "训练与测试过程与之前手动实现的几乎一致。由于数据集的变化，对应超参数也进行了调整。\n",
+    "\n",
+    "数据集也使用了现成的FashionMNIST数据集，且划分了训练集和测试集。\n",
+    "\n",
+    "FashionMNIST数据集直接调用API获取。数据集的image为28*28的单通道灰白图片，label为单个数值标签。\n",
+    "\n",
+    "训练softmax回归模型。\n",
+    "进行如下步骤：\n",
+    "1. 初始化超参数\n",
+    "2. 获取数据集\n",
+    "3. 初始化模型\n",
+    "4. 定义损失函数和优化器\n",
+    "5. 训练\n",
+    "    1. 从训练dataloader中获取批量数据\n",
+    "    2. 传入模型\n",
+    "    3. 使用损失函数计算与ground_truth的损失\n",
+    "    4. 使用优化器进行反向传播\n",
+    "    5. 循环以上步骤\n",
+    "6. 测试\n",
+    "    1. 从测试dataloader中获取批量数据\n",
+    "    2. 传入模型\n",
+    "    3. 将预测值与ground_truth进行比较，得出正确率\n",
+    "    4. 对整个训练集统计正确率，从而分析训练效果"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 17,
+   "id": "d816dae1-5fbe-4c29-9597-19d66b5eb6b4",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Epoch 1/10, Loss: nan, Acc: 0.09999999403953552\n",
+      "Epoch 2/10, Loss: nan, Acc: 0.09999999403953552\n",
+      "Epoch 3/10, Loss: nan, Acc: 0.09999999403953552\n",
+      "Epoch 4/10, Loss: nan, Acc: 0.09999999403953552\n",
+      "Epoch 5/10, Loss: nan, Acc: 0.09999999403953552\n",
+      "Epoch 6/10, Loss: nan, Acc: 0.09999999403953552\n",
+      "Epoch 7/10, Loss: nan, Acc: 0.09999999403953552\n",
+      "Epoch 8/10, Loss: nan, Acc: 0.09999999403953552\n",
+      "Epoch 9/10, Loss: nan, Acc: 0.09999999403953552\n",
+      "Epoch 10/10, Loss: nan, Acc: 0.09999999403953552\n"
+     ]
+    }
+   ],
+   "source": [
+    "learning_rate = 5e-3\n",
+    "num_epochs = 10\n",
+    "batch_size = 4096\n",
+    "num_classes = 10\n",
+    "device = \"cuda:0\" if torch.cuda.is_available() else \"cpu\"\n",
+    "\n",
+    "transform = transforms.Compose(\n",
+    "    [\n",
+    "        transforms.ToTensor(),\n",
+    "        transforms.Normalize((0.5,), (0.5,)),\n",
+    "    ]\n",
+    ")\n",
+    "train_dataset = datasets.FashionMNIST(root=\"./dataset\", train=True, transform=transform, download=True)\n",
+    "test_dataset = datasets.FashionMNIST(root=\"./dataset\", train=False, transform=transform, download=True)\n",
+    "train_loader = DataLoader(\n",
+    "    dataset=train_dataset, batch_size=batch_size,\n",
+    "    shuffle=True, num_workers=4, pin_memory=True,\n",
+    ")\n",
+    "test_loader = DataLoader(\n",
+    "    dataset=test_dataset, batch_size=batch_size,\n",
+    "    shuffle=True, num_workers=4, pin_memory=True,\n",
+    ")\n",
+    "\n",
+    "model = Model_3_1(num_classes).to(device)\n",
+    "criterion = My_CrossEntropyLoss()\n",
+    "optimizer = My_optimizer(model.parameters(), lr=learning_rate)\n",
+    "\n",
+    "for epoch in range(num_epochs):\n",
+    "    total_epoch_loss = 0\n",
+    "    for images, targets in train_loader:\n",
+    "        optimizer.zero_grad()\n",
+    "\n",
+    "        images = images.to(device)\n",
+    "        targets = targets.to(device).to(dtype=torch.long)\n",
+    "\n",
+    "        one_hot_targets = (\n",
+    "            my_one_hot(targets, num_classes=num_classes)\n",
+    "            .to(device)\n",
+    "            .to(dtype=torch.long)\n",
+    "        )\n",
+    "\n",
+    "        outputs = model(images)\n",
+    "        loss = criterion(outputs, one_hot_targets)\n",
+    "        total_epoch_loss += loss\n",
+    "\n",
+    "        loss.backward()\n",
+    "        optimizer.step()\n",
+    "\n",
+    "    total_acc = 0\n",
+    "    with torch.no_grad():\n",
+    "        for image, targets in test_loader:\n",
+    "            image = image.to(device)\n",
+    "            targets = targets.to(device)\n",
+    "            outputs = model(image)\n",
+    "            total_acc += (outputs.argmax(1) == targets).sum()\n",
+    "    print(\n",
+    "        f\"Epoch {epoch + 1}/{num_epochs}, Loss: {total_epoch_loss}, Acc: {total_acc / len(test_dataset)}\"\n",
+    "    )"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "a49d0165-aeb7-48c0-9b67-956bb08cb356",
+   "metadata": {},
+   "source": [
+    "这里发现梯度爆炸。暂时无法解决。"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "3ef5240f-8a11-4678-bfce-f1cbc7e71b77",
+   "metadata": {},
+   "source": [
+    "## 问题2\n",
+    "\n",
+    "**利用torch.nn实现softmax回归在Fashion-MNIST数据集上进行训练和测试，并从loss，训练集以及测试集上的准确率等多个角度对结果进行分析**"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "5c4a88c6-637e-4af5-bed5-f644685dcabc",
+   "metadata": {},
+   "source": [
+    "使用torch.nn实现模型。\n",
+    "\n",
+    "将之前的Model_3_1中的手动实现函数改为torch.nn内置函数即可，再加上继承nn.Module以使用torch.nn内置模型模板特性。"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 18,
+   "id": "0163b9f7-1019-429c-8c29-06436d0a4c98",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "class Model_3_2(nn.Module):\n",
+    "    def __init__(self, num_classes):\n",
+    "        super(Model_3_2, self).__init__()\n",
+    "        self.flatten = nn.Flatten()\n",
+    "        self.linear = nn.Linear(28 * 28, num_classes)\n",
+    "\n",
+    "    def forward(self, x: torch.Tensor):\n",
+    "        x = self.flatten(x)\n",
+    "        x = self.linear(x)\n",
+    "        return x"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "6e765ad7-c1c6-4166-bd7f-361666bd4016",
+   "metadata": {},
+   "source": [
+    "训练与测试过程与之前手动实现的几乎一致。"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 19,
+   "id": "a58a23e1-368c-430a-ad62-0e256dff564d",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Epoch 1/10, Loss: 15.148970603942871, Acc: 0.7520999908447266\n",
+      "Epoch 2/10, Loss: 9.012335777282715, Acc: 0.7996999621391296\n",
+      "Epoch 3/10, Loss: 7.9114227294921875, Acc: 0.8095999956130981\n",
+      "Epoch 4/10, Loss: 7.427404403686523, Acc: 0.8215999603271484\n",
+      "Epoch 5/10, Loss: 7.084254264831543, Acc: 0.8277999758720398\n",
+      "Epoch 6/10, Loss: 6.885956287384033, Acc: 0.8274999856948853\n",
+      "Epoch 7/10, Loss: 6.808426380157471, Acc: 0.8327999711036682\n",
+      "Epoch 8/10, Loss: 6.647855758666992, Acc: 0.8323000073432922\n",
+      "Epoch 9/10, Loss: 6.560361862182617, Acc: 0.8317999839782715\n",
+      "Epoch 10/10, Loss: 6.5211310386657715, Acc: 0.8349999785423279\n"
+     ]
+    }
+   ],
+   "source": [
+    "learning_rate = 5e-3\n",
+    "num_epochs = 10\n",
+    "batch_size = 4096\n",
+    "num_classes = 10\n",
+    "device = \"cuda:0\" if torch.cuda.is_available() else \"cpu\"\n",
+    "\n",
+    "transform = transforms.Compose(\n",
+    "    [\n",
+    "        transforms.ToTensor(),\n",
+    "        transforms.Normalize((0.5,), (0.5,)),\n",
+    "    ]\n",
+    ")\n",
+    "train_dataset = datasets.FashionMNIST(\n",
+    "    root=\"./dataset\", train=True, transform=transform, download=True\n",
+    ")\n",
+    "test_dataset = datasets.FashionMNIST(\n",
+    "    root=\"./dataset\", train=False, transform=transform, download=True\n",
+    ")\n",
+    "train_loader = DataLoader(\n",
+    "    dataset=train_dataset,\n",
+    "    batch_size=batch_size,\n",
+    "    shuffle=True,\n",
+    "    num_workers=4,\n",
+    "    pin_memory=True,\n",
+    ")\n",
+    "test_loader = DataLoader(\n",
+    "    dataset=test_dataset,\n",
+    "    batch_size=batch_size,\n",
+    "    shuffle=True,\n",
+    "    num_workers=4,\n",
+    "    pin_memory=True,\n",
+    ")\n",
+    "\n",
+    "model = Model_3_2(num_classes).to(device)\n",
+    "criterion = nn.CrossEntropyLoss()\n",
+    "optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate)\n",
+    "\n",
+    "for epoch in range(num_epochs):\n",
+    "    total_epoch_loss = 0\n",
+    "    model.train()\n",
+    "    for images, targets in train_loader:\n",
+    "        optimizer.zero_grad()\n",
+    "\n",
+    "        images = images.to(device)\n",
+    "        targets = targets.to(device)\n",
+    "\n",
+    "        one_hot_targets = (\n",
+    "            torch.nn.functional.one_hot(targets, num_classes=num_classes)\n",
+    "            .to(device)\n",
+    "            .to(dtype=torch.float32)\n",
+    "        )\n",
+    "\n",
+    "        outputs = model(images)\n",
+    "        loss = criterion(outputs, one_hot_targets)\n",
+    "        total_epoch_loss += loss\n",
+    "\n",
+    "        loss.backward()\n",
+    "        optimizer.step()\n",
+    "\n",
+    "    model.eval()\n",
+    "    total_acc = 0\n",
+    "    with torch.no_grad():\n",
+    "        for image, targets in test_loader:\n",
+    "            image = image.to(device)\n",
+    "            targets = targets.to(device)\n",
+    "            outputs = model(image)\n",
+    "            total_acc += (outputs.argmax(1) == targets).sum()\n",
+    "    print(\n",
+    "        f\"Epoch {epoch + 1}/{num_epochs}, Loss: {total_epoch_loss}, Acc: {total_acc / len(test_dataset)}\"\n",
+    "    )"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "59555b67-1650-4e1a-a98e-7906878bf3d0",
+   "metadata": {},
+   "source": [
+    "与手动实现的softmax回归相比较，nn.CrossEntropyLoss比手动实现的My_CrossEntropyLoss更加稳定，没有出现梯度爆炸的情况。"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "f40431f2-e77b-4ead-81a3-ff6451a8e452",
+   "metadata": {},
+   "source": [
+    "**实验心得体会**\n",
+    "\n",
+    "通过完成本次Pytorch基本操作实验，让我对Pytorch框架有了更加深入的理解。我接触深度学习主要是在大语言模型领域，比较熟悉微调大模型，但是涉及到底层的深度学习知识，我还有很多短板和不足。这次实验对我这方面的锻炼让我收获良多。\n",
+    "\n",
+    "首先是数据集的设置。如果数据没有进行归一化，很容易出现梯度爆炸。这是在我以前直接使用图片数据集的经历中没有遇到过的问题。\n",
+    "\n",
+    "在实现logistic回归模型时，通过手动实现各个组件如优化器、线性层等，让我对这些模块的工作原理有了更清晰的认识。尤其是在实现广播机制时，需要充分理解张量操作的维度变换规律。而使用Pytorch内置模块进行实现时，通过继承nn.Module可以自动获得許多功能，使代码更加简洁。\n",
+    "\n",
+    "在实现softmax回归时，则遇到了更大的困难。手动实现的模型很容易出现梯度爆炸的问题，而使用Pytorch内置的损失函数和优化器则可以稳定训练。这让我意识到了选择合适的优化方法的重要性。另外，Pytorch强大的自动微分机制也是构建深度神经网络的重要基础。\n",
+    "\n",
+    "通过这个实验，让我对Pytorch框架有了更加直观的感受，也让我看到了仅靠基础模块搭建复杂模型的难点所在。这些经验对我后续使用Pytorch构建数据集模型会很有帮助。"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3 (ipykernel)",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.10.13"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}
diff --git a/Lab1/.vscode/settings.json b/Lab1/.vscode/settings.json
new file mode 100644
index 0000000..d99f2f3
--- /dev/null
+++ b/Lab1/.vscode/settings.json
@@ -0,0 +1,6 @@
+{
+    "[python]": {
+        "editor.defaultFormatter": "ms-python.black-formatter"
+    },
+    "python.formatting.provider": "none"
+}
\ No newline at end of file
diff --git a/Lab1/Pytorch基本操作实验报告.ipynb b/Lab1/Pytorch基本操作实验报告.ipynb
new file mode 100644
index 0000000..c0645c6
--- /dev/null
+++ b/Lab1/Pytorch基本操作实验报告.ipynb
@@ -0,0 +1,1262 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "id": "3b57686b-7ac8-4897-bf76-3d982b1ff8da",
+   "metadata": {},
+   "source": [
+    "<h1><center>本科生《深度学习》课程<br>实验报告</center></h1>\n",
+    "<div style=\"text-align: center;\">\n",
+    "    <div><span style=\"display: inline-block; width: 65px; text-align: center;\">课程名称</span><span style=\"display: inline-block; width: 25px;\">:</span><span style=\"display: inline-block; width: 210px; font-weight: bold; text-align: left;\">深度学习</span></div>\n",
+    "    <div><span style=\"display: inline-block; width: 65px; text-align: center;\">实验题目</span><span style=\"display: inline-block; width: 25px;\">:</span><span style=\"display: inline-block; width: 210px; font-weight: bold; text-align: left;\">Pytorch基本操作</span></div>\n",
+    "    <div><span style=\"display: inline-block; width: 65px; text-align: center;\">学号</span><span style=\"display: inline-block; width: 25px;\">:</span><span style=\"display: inline-block; width: 210px; font-weight: bold; text-align: left;\">21281280</span></div>\n",
+    "    <div><span style=\"display: inline-block; width: 65px; text-align: center;\">姓名</span><span style=\"display: inline-block; width: 25px;\">:</span><span style=\"display: inline-block; width: 210px; font-weight: bold; text-align: left;\">柯劲帆</span></div>\n",
+    "    <div><span style=\"display: inline-block; width: 65px; text-align: center;\">班级</span><span style=\"display: inline-block; width: 25px;\">:</span><span style=\"display: inline-block; width: 210px; font-weight: bold; text-align: left;\">物联网2101班</span></div>\n",
+    "    <div><span style=\"display: inline-block; width: 65px; text-align: center;\">指导老师</span><span style=\"display: inline-block; width: 25px;\">:</span><span style=\"display: inline-block; width: 210px; font-weight: bold; text-align: left;\">张淳杰</span></div>\n",
+    "    <div><span style=\"display: inline-block; width: 65px; text-align: center;\">报告日期</span><span style=\"display: inline-block; width: 25px;\">:</span><span style=\"display: inline-block; width: 210px; font-weight: bold; text-align: left;\">2023年10月9日</span></div>\n",
+    "</div>"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "e24aa17e-faf9-4d69-9eae-43159116b56f",
+   "metadata": {},
+   "source": [
+    "实验环境：\n",
+    "- OS：Ubuntu 22.04 内核版本 6.2.0-34-generic\n",
+    "- CPU：12th Gen Intel(R) Core(TM) i7-12700H\n",
+    "- GPU：NVIDIA GeForce RTX 3070 Ti Laptop\n",
+    "- conda: miniconda 23.9.0\n",
+    "- python：3.10.13\n",
+    "- pytorch：2.1.0"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "id": "a4e12268-bad4-44c4-92d5-883624d93e25",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import numpy as np\n",
+    "import torch\n",
+    "from torch.autograd import Variable\n",
+    "from torch.utils.data import Dataset, DataLoader\n",
+    "from torch import nn\n",
+    "from torchvision import datasets, transforms"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "cc7f0ce5-d613-425b-807c-78115632cd80",
+   "metadata": {},
+   "source": [
+    "引用相关库。"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "59a43d35-56ac-4ade-995d-1c6fcbcd1262",
+   "metadata": {},
+   "source": [
+    "# 一、Pytorch基本操作考察\n",
+    "## 题目2\n",
+    "**使用 𝐓𝐞𝐧𝐬𝐨𝐫 初始化一个 𝟏×𝟑 的矩阵 𝑴 和一个 𝟐×𝟏 的矩阵 𝑵，对两矩阵进行减法操作（要求实现三种不同的形式），给出结果并分析三种方式的不同（如果出现报错，分析报错的原因），同时需要指出在计算过程中发生了什么。**"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "id": "79ea46db-cf49-436c-9b5b-c6562d0da9e2",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "方法1的结果:\n",
+      "tensor([[-3, -2, -1],\n",
+      "        [-4, -3, -2]])\n",
+      "方法2的结果:\n",
+      "tensor([[-3, -2, -1],\n",
+      "        [-4, -3, -2]])\n",
+      "方法3的结果:\n",
+      "tensor([[-3, -2, -1],\n",
+      "        [-4, -3, -2]])\n"
+     ]
+    }
+   ],
+   "source": [
+    "A = torch.tensor([[1, 2, 3]])\n",
+    "\n",
+    "B = torch.tensor([[4],\n",
+    "                  [5]])\n",
+    "\n",
+    "# 方法1: 使用PyTorch的减法操作符\n",
+    "result1 = A - B\n",
+    "\n",
+    "# 方法2: 使用PyTorch的sub函数\n",
+    "result2 = torch.sub(A, B)\n",
+    "\n",
+    "# 方法3: 手动实现广播机制并作差\n",
+    "def mysub(a:torch.Tensor, b:torch.Tensor):\n",
+    "    if not (\n",
+    "        (a.size(0) == 1 and b.size(1) == 1) \n",
+    "        or \n",
+    "        (a.size(1) == 1 and b.size(0) == 1)\n",
+    "        ):\n",
+    "        raise ValueError(\"输入的张量大小无法满足广播机制的条件。\")\n",
+    "    else:\n",
+    "        target_shape = torch.Size([max(A.size(0), B.size(0)), max(A.size(1), B.size(1))])\n",
+    "        A_broadcasted = A.expand(target_shape)\n",
+    "        B_broadcasted = B.expand(target_shape)\n",
+    "        result = torch.zeros(target_shape, dtype=torch.int64).to(device=A_broadcasted.device)\n",
+    "        for i in range(target_shape[0]):\n",
+    "            for j in range(target_shape[1]):\n",
+    "                result[i, j] = A_broadcasted[i, j] - B_broadcasted[i, j]\n",
+    "        return result\n",
+    "\n",
+    "result3 = mysub(A, B)\n",
+    "\n",
+    "print(\"方法1的结果:\")\n",
+    "print(result1)\n",
+    "print(\"方法2的结果:\")\n",
+    "print(result2)\n",
+    "print(\"方法3的结果:\")\n",
+    "print(result3)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "2489a3ad-f6ff-4561-bb26-e02654090b98",
+   "metadata": {},
+   "source": [
+    "## 题目2\n",
+    "1. **利用Tensor创建两个大小分别3*2和4*2的随机数矩阵P和Q，要求服从均值为0，标准差0.01为的正态分布；**\n",
+    "2. **对第二步得到的矩阵Q进行形状变换得到Q的转置Q^T；**\n",
+    "3. **对上述得到的矩阵P和矩阵Q^T求矩阵相乘。**"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "id": "41e4ee02-1d05-4101-b3f0-477bac0277fb",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "矩阵 P:\n",
+      "tensor([[ 0.0053,  0.0013],\n",
+      "        [-0.0086,  0.0136],\n",
+      "        [-0.0013,  0.0176]])\n",
+      "矩阵 Q:\n",
+      "tensor([[ 0.0044,  0.0014],\n",
+      "        [ 0.0147,  0.0078],\n",
+      "        [-0.0002, -0.0023],\n",
+      "        [ 0.0001, -0.0011]])\n",
+      "矩阵 QT:\n",
+      "tensor([[ 0.0044,  0.0147, -0.0002,  0.0001],\n",
+      "        [ 0.0014,  0.0078, -0.0023, -0.0011]])\n",
+      "矩阵相乘的结果:\n",
+      "tensor([[ 2.4953e-05,  8.7463e-05, -3.8665e-06, -8.9576e-07],\n",
+      "        [-1.9514e-05, -2.0557e-05, -2.9649e-05, -1.5913e-05],\n",
+      "        [ 1.8189e-05,  1.1834e-04, -4.0097e-05, -1.9608e-05]])\n"
+     ]
+    }
+   ],
+   "source": [
+    "mean = 0\n",
+    "stddev = 0.01\n",
+    "\n",
+    "P = torch.normal(mean=mean, std=stddev, size=(3, 2))\n",
+    "Q = torch.normal(mean=mean, std=stddev, size=(4, 2))\n",
+    "\n",
+    "print(\"矩阵 P:\")\n",
+    "print(P)\n",
+    "print(\"矩阵 Q:\")\n",
+    "print(Q)\n",
+    "\n",
+    "# 对矩阵Q进行转置操作，得到矩阵Q的转置Q^T\n",
+    "QT = Q.T\n",
+    "print(\"矩阵 QT:\")\n",
+    "print(QT)\n",
+    "\n",
+    "# 计算矩阵P和矩阵Q^T的矩阵相乘\n",
+    "result = torch.matmul(P, QT)\n",
+    "print(\"矩阵相乘的结果:\")\n",
+    "print(result)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "cea9cb6d-adde-4e08-b9f2-8c417abf4231",
+   "metadata": {},
+   "source": [
+    "## 题目2\n",
+    "**给定公式$ y_3=y_1+y_2=𝑥^2+𝑥^3$，且$x=1$。利用学习所得到的Tensor的相关知识，求$y_3$对$x$的梯度，即$\\frac{dy_3}{dx}$。**"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "id": "951512cd-d915-4d04-959f-eb99d1971e2d",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "梯度(dy_3/dx):  2.0\n"
+     ]
+    }
+   ],
+   "source": [
+    "x = torch.tensor(1.0, requires_grad=True)\n",
+    "y_1 = x ** 2\n",
+    "with torch.no_grad():\n",
+    "    y_2 = x**3\n",
+    "\n",
+    "y3 = y_1 + y_2\n",
+    "\n",
+    "y3.backward()\n",
+    "\n",
+    "print(\"梯度(dy_3/dx): \", x.grad.item())"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "3269dbf6-889a-49eb-8094-1e588e1a6c30",
+   "metadata": {},
+   "source": [
+    "# 二、动手实现logistic回归\n",
+    "## 题目1\n",
+    "**要求动手从0实现 logistic 回归（只借助Tensor和Numpy相关的库）在人工构造的数据集上进行训练和测试，并从loss以及训练集上的准确率等多个角度对结果进行分析（可借助nn.BCELoss或nn.BCEWithLogitsLoss作为损失函数，从零实现二元交叉熵为选作）**"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "bcd12aa9-f187-4d88-8c59-af6d16107edb",
+   "metadata": {},
+   "source": [
+    "给定预测概率$ \\left( \\hat{y} \\right) $和目标标签$ \\left( y \\right)$（通常是0或1），BCELoss的计算公式如下：\n",
+    "$$\n",
+    " \\text{BCELoss}(\\hat{y}, y) = -\\frac{1}{N} \\sum_{i=1}^{N} \\left(y_i \\cdot \\log(\\hat{y}_i) + (1 - y_i) \\cdot \\log(1 - \\hat{y}_i)\\right) \n",
+    "$$\n",
+    "其中，$\\left( N \\right) $是样本数量，$\\left( \\hat{y}_i \\right) $表示模型的预测概率向量中的第$ \\left( i \\right) $个元素，$\\left( y_i \\right) $表示实际的目标标签中的第$ \\left( i \\right) $个元素。在二分类问题中，$\\left( y_i \\right) $通常是0或1。这个公式表示对所有样本的二分类交叉熵损失进行了求和并取平均。\n",
+    "\n",
+    "因此BCELoss的手动实现如下。"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 5,
+   "id": "e31b86ec-4114-48dd-8d73-fe4e0686419a",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "输入：\n",
+      "tensor([0.6900])\n",
+      "标签：\n",
+      "tensor([1.])\n",
+      "My_BCELoss损失值: 0.37110066413879395\n",
+      "nn.BCELoss损失值: 0.37110066413879395\n"
+     ]
+    }
+   ],
+   "source": [
+    "class My_BCELoss:\n",
+    "    def __call__(self, prediction: torch.Tensor, target: torch.Tensor):\n",
+    "        loss = -torch.mean(target * torch.log(prediction) + (1 - target) * torch.log(1 - prediction))\n",
+    "        return loss\n",
+    "\n",
+    "\n",
+    "# 测试\n",
+    "prediction = torch.sigmoid(torch.tensor([0.8]))\n",
+    "target = torch.tensor([1.0])\n",
+    "print(f\"输入：\\n{prediction}\")\n",
+    "print(f\"标签：\\n{target}\")\n",
+    "\n",
+    "my_bce_loss = My_BCELoss()\n",
+    "my_loss = my_bce_loss(prediction, target)\n",
+    "print(\"My_BCELoss损失值:\", my_loss.item())\n",
+    "\n",
+    "nn_bce_loss = nn.BCELoss()\n",
+    "nn_loss = nn_bce_loss(prediction, target)\n",
+    "print(\"nn.BCELoss损失值:\", nn_loss.item())"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "345b0300-8808-4c43-9bf9-05a7e6e1f5af",
+   "metadata": {},
+   "source": [
+    "Optimizer的实现较为简单。\n",
+    "\n",
+    "主要实现：\n",
+    "- 传入参数：`__init__()`\n",
+    "- 对传入的参数进行更新：`step()`\n",
+    "- 清空传入参数存储的梯度：`zero_grad()`"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 6,
+   "id": "0297066c-9fc1-448d-bdcb-29a6f1519117",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "y.backward()之后，x的梯度:  2.0\n",
+      "optimizer_test.step()之后，x的值:  0.800000011920929\n",
+      "optimizer_test.zero_grad()之后，x的梯度:  0.0\n"
+     ]
+    }
+   ],
+   "source": [
+    "class My_optimizer:\n",
+    "    def __init__(self, params: list[torch.Tensor], lr: float):\n",
+    "        self.params = params\n",
+    "        self.lr = lr\n",
+    "\n",
+    "    def step(self):\n",
+    "        for param in self.params:\n",
+    "            param.data = param.data - self.lr * param.grad.data\n",
+    "\n",
+    "    def zero_grad(self):\n",
+    "        for param in self.params:\n",
+    "            if param.grad is not None:\n",
+    "                param.grad.data.zero_()\n",
+    "\n",
+    "\n",
+    "# 测试\n",
+    "x = torch.tensor(1.0, requires_grad=True)\n",
+    "y = x ** 2\n",
+    "optimizer_test = My_optimizer([x], lr=0.1)\n",
+    "\n",
+    "y.backward()\n",
+    "print(\"y.backward()之后，x的梯度: \", x.grad.item())\n",
+    "\n",
+    "optimizer_test.step()\n",
+    "print(\"optimizer_test.step()之后，x的值: \", x.item())\n",
+    "\n",
+    "optimizer_test.zero_grad()\n",
+    "print(\"optimizer_test.zero_grad()之后，x的梯度: \", x.grad.item())"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "6ab83528-a88b-4d66-b0c9-b1315cf75c22",
+   "metadata": {},
+   "source": [
+    "线性层主要有一个权重（weight)和一个偏置（bias）。\n",
+    "线性层的数学公式如下：\n",
+    "$$\n",
+    "x:=x \\times weight^T+bias\n",
+    "$$\n",
+    "因此代码实现如下："
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 7,
+   "id": "8e18695a-d8c5-4f77-8b5c-de40d9240fb9",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "输入：\n",
+      "tensor([[1.],\n",
+      "        [2.]], requires_grad=True)\n",
+      "权重：\n",
+      "tensor([[-1.0980],\n",
+      "        [-0.5413],\n",
+      "        [ 1.5884]], requires_grad=True)\n",
+      "偏置：\n",
+      "tensor([-1.1733], requires_grad=True)\n",
+      "输出：\n",
+      "tensor([[-2.2713, -1.7146,  0.4151],\n",
+      "        [-3.3692, -2.2559,  2.0036]], grad_fn=<AddBackward0>)\n"
+     ]
+    }
+   ],
+   "source": [
+    "class My_Linear:\n",
+    "    def __init__(self, input_feature: int, output_feature: int):\n",
+    "        self.weight = torch.randn((output_feature, input_feature), requires_grad=True, dtype=torch.float32)\n",
+    "        self.bias = torch.randn(1, requires_grad=True, dtype=torch.float32)\n",
+    "        self.params = [self.weight, self.bias]\n",
+    "\n",
+    "    def __call__(self, x: torch.Tensor):\n",
+    "        return self.forward(x)\n",
+    "\n",
+    "    def forward(self, x: torch.Tensor):\n",
+    "        x = torch.matmul(x, self.weight.T) + self.bias\n",
+    "        return x\n",
+    "\n",
+    "    def to(self, device: str):\n",
+    "        for param in self.params:\n",
+    "            param.data = param.data.to(device=device)\n",
+    "        return self\n",
+    "\n",
+    "    def parameters(self):\n",
+    "        return self.params\n",
+    "\n",
+    "        \n",
+    "# 测试\n",
+    "linear_test = My_Linear(1, 3)\n",
+    "x = torch.tensor([[1.], [2.]], requires_grad=True)\n",
+    "print(f\"输入：\\n{x}\")\n",
+    "print(f\"权重：\\n{linear_test.weight}\")\n",
+    "print(f\"偏置：\\n{linear_test.bias}\")\n",
+    "y = linear_test(x)\n",
+    "print(f\"输出：\\n{y}\")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "5ff813cc-c1f0-4c73-a3e8-d6796ef5d366",
+   "metadata": {},
+   "source": [
+    "手动实现logistic回归模型。\n",
+    "\n",
+    "模型很简单，主要由一个线性层和一个sigmoid层组成。\n",
+    "\n",
+    "Sigmoid函数（又称为 Logistic函数）是一种常用的激活函数，通常用于神经网络的输出层或隐藏层，其作用是将输入的实数值压缩到一个范围在0和1之间的数值。"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 8,
+   "id": "e7de7e4b-a084-4793-812e-46e8550ecd8d",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "class Model_2_1():\n",
+    "    def __init__(self):\n",
+    "        self.linear = My_Linear(1, 1)\n",
+    "        self.params = self.linear.params\n",
+    "\n",
+    "    def __call__(self, x):\n",
+    "        return self.forward(x)\n",
+    "\n",
+    "    def forward(self, x):\n",
+    "        x = self.linear(x)\n",
+    "        x = torch.sigmoid(x)\n",
+    "        return x\n",
+    "\n",
+    "    def to(self, device: str):\n",
+    "        for param in self.params:\n",
+    "            param.data = param.data.to(device=device)\n",
+    "        return self\n",
+    "\n",
+    "    def parameters(self):\n",
+    "        return self.params"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "e14acea9-e5ef-4c24-aea9-329647224ce1",
+   "metadata": {},
+   "source": [
+    "人工随机构造数据集。\n",
+    "\n",
+    "这里我遇到了比较大的问题。因为数据构建不合适，会导致后面的训练出现梯度爆炸。\n",
+    "\n",
+    "我采用随机产生数据后归一化的方法，即\n",
+    "$$\n",
+    "\\hat{x} = \\frac{x - \\text{min}_x}{\\text{max}_x - \\text{min}_x} \n",
+    "$$\n",
+    "将数据控制在合适的区间。\n",
+    "\n",
+    "我的y设置为$4-3\\times x + noise$，noise为随机噪声。\n",
+    "\n",
+    "生成完x和y后进行归一化处理，并写好DataLoader访问数据集的接口`__getitem__()`。"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 9,
+   "id": "c39fbafb-62e4-4b8c-9d65-6718d25f2970",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "测试数据集大小：100\n",
+      "测试数据集第0对数据：\n",
+      "x_0 = 0.5531462811708403\n",
+      "y_0 = 0.42036701080526284\n"
+     ]
+    }
+   ],
+   "source": [
+    "class My_Dataset(Dataset):\n",
+    "    def __init__(self, data_size=1000000):\n",
+    "        np.random.seed(0)\n",
+    "        x = 2 * np.random.rand(data_size, 1)\n",
+    "        noise = 0.2 * np.random.randn(data_size, 1)\n",
+    "        y = 4 - 3 * x + noise\n",
+    "        self.min_x, self.max_x = np.min(x), np.max(x)\n",
+    "        min_y, max_y = np.min(y), np.max(y)\n",
+    "        x = (x - self.min_x) / (self.max_x - self.min_x)\n",
+    "        y = (y - min_y) / (max_y - min_y)\n",
+    "        self.data = [[x[i][0], y[i][0]] for i in range(x.shape[0])]\n",
+    "\n",
+    "    def __len__(self):\n",
+    "        return len(self.data)\n",
+    "\n",
+    "    def __getitem__(self, index):\n",
+    "        x, y = self.data[index]\n",
+    "        return x, y\n",
+    "\n",
+    "\n",
+    "# 测试\n",
+    "dataset_test = My_Dataset(data_size=100)\n",
+    "dataset_size = len(dataset_test)\n",
+    "print(f\"测试数据集大小：{dataset_size}\")\n",
+    "x0, y0 = dataset_test[0]\n",
+    "print(f\"测试数据集第0对数据：\")\n",
+    "print(f\"x_0 = {x0}\")\n",
+    "print(f\"y_0 = {y0}\")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "957a76a2-b306-47a8-912e-8fbf00cdfd42",
+   "metadata": {},
+   "source": [
+    "训练Logistic回归模型。\n",
+    "进行如下步骤：\n",
+    "1. 初始化超参数\n",
+    "2. 获取数据集\n",
+    "3. 初始化模型\n",
+    "4. 定义损失函数和优化器\n",
+    "5. 训练\n",
+    "    1. 从训练dataloader中获取批量数据\n",
+    "    2. 传入模型\n",
+    "    3. 使用损失函数计算与ground_truth的损失\n",
+    "    4. 使用优化器进行反向传播\n",
+    "    5. 循环以上步骤\n",
+    "6. 测试\n",
+    "    1. 设置测试数据\n",
+    "    2. 传入模型\n",
+    "    3. 得到预测值"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 10,
+   "id": "5612661e-2809-4d46-96c2-33ee9f44116d",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Epoch 1/10, Loss: 680.9198314547539, Acc: 0.9677169744703272\n",
+      "Epoch 2/10, Loss: 677.2582936882973, Acc: 0.9985965700887113\n",
+      "Epoch 3/10, Loss: 677.1911396384239, Acc: 0.9993738265049104\n",
+      "Epoch 4/10, Loss: 677.1777537465096, Acc: 0.9995470920810262\n",
+      "Epoch 5/10, Loss: 677.1745615005493, Acc: 0.9998228389835642\n",
+      "Epoch 6/10, Loss: 677.1743944883347, Acc: 0.9999690339979311\n",
+      "Epoch 7/10, Loss: 677.1735371947289, Acc: 0.9998205132243208\n",
+      "Epoch 8/10, Loss: 677.1737813353539, Acc: 0.999798559017381\n",
+      "Epoch 9/10, Loss: 677.1740361452103, Acc: 0.9998672931901137\n",
+      "Epoch 10/10, Loss: 677.1736125349998, Acc: 0.9997257713704987\n",
+      "Model weights: -0.0006128809181973338, bias: 0.023128816857933998\n",
+      "Prediction for test data: 0.505628764629364\n"
+     ]
+    }
+   ],
+   "source": [
+    "learning_rate = 5e-2\n",
+    "num_epochs = 10\n",
+    "batch_size = 1024\n",
+    "device = \"cuda:0\" if torch.cuda.is_available() else \"cpu\"\n",
+    "\n",
+    "dataset = My_Dataset()\n",
+    "dataloader = DataLoader(\n",
+    "    dataset=dataset, batch_size=batch_size, shuffle=True, num_workers=5, pin_memory=True\n",
+    ")\n",
+    "\n",
+    "model = Model_2_1().to(device)\n",
+    "criterion = My_BCELoss()\n",
+    "optimizer = My_optimizer(model.parameters(), lr=learning_rate)\n",
+    "\n",
+    "for epoch in range(num_epochs):\n",
+    "    total_epoch_loss = 0\n",
+    "    total_epoch_pred = 0\n",
+    "    total_epoch_target = 0\n",
+    "    for x, targets in dataloader:\n",
+    "        optimizer.zero_grad()\n",
+    "        \n",
+    "        x = x.to(device).to(dtype=torch.float32)\n",
+    "        targets = targets.to(device).to(dtype=torch.float32)\n",
+    "        \n",
+    "        x = x.unsqueeze(1)\n",
+    "        y_pred = model(x)\n",
+    "        loss = criterion(y_pred, targets)\n",
+    "        total_epoch_loss += loss.item()\n",
+    "        total_epoch_target += targets.sum().item()\n",
+    "        total_epoch_pred += y_pred.sum().item()\n",
+    "\n",
+    "        loss.backward()\n",
+    "        optimizer.step()\n",
+    "\n",
+    "    print(\n",
+    "        f\"Epoch {epoch + 1}/{num_epochs}, Loss: {total_epoch_loss}, Acc: {1 - abs(total_epoch_pred - total_epoch_target) / total_epoch_target}\"\n",
+    "    )\n",
+    "\n",
+    "with torch.no_grad():\n",
+    "    test_data = (np.array([[2]]) - dataset.min_x) / (dataset.max_x - dataset.min_x)\n",
+    "    test_data = Variable(\n",
+    "        torch.tensor(test_data, dtype=torch.float32), requires_grad=False\n",
+    "    ).to(device)\n",
+    "    predicted = model(test_data).to(\"cpu\")\n",
+    "    print(\n",
+    "        f\"Model weights: {model.linear.weight.item()}, bias: {model.linear.bias.item()}\"\n",
+    "    )\n",
+    "    print(f\"Prediction for test data: {predicted.item()}\")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "9e416582-a30d-4084-acc6-6e05f80a6aff",
+   "metadata": {},
+   "source": [
+    "## 题目2\n",
+    "**利用 torch.nn 实现 logistic 回归在人工构造的数据集上进行训练和测试，并对结果进行分析，并从loss以及训练集上的准确率等多个角度对结果进行分析**"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "0460d125-7d03-44fe-845c-c4d13792e241",
+   "metadata": {},
+   "source": [
+    "使用torch.nn实现模型。\n",
+    "\n",
+    "将之前的Model_2_1中的手动实现函数改为torch.nn内置函数即可，再加上继承nn.Module以使用torch.nn内置模型模板特性。"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 11,
+   "id": "fa121afd-a1af-4193-9b54-68041e0ed068",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "class Model_2_2(nn.Module):\n",
+    "    def __init__(self):\n",
+    "        super(Model_2_2, self).__init__()\n",
+    "        self.linear = nn.Linear(1, 1, dtype=torch.float64)\n",
+    "\n",
+    "    def forward(self, x):\n",
+    "        x = self.linear(x)\n",
+    "        x = torch.sigmoid(x)\n",
+    "        return x"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "176eee7e-4e3d-470e-8af2-8761bca039f8",
+   "metadata": {},
+   "source": [
+    "训练与测试过程与之前手动实现的几乎一致。仅有少量涉及数据类型（dtype）的代码需要更改以适应torch.nn的内置函数要求。"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 12,
+   "id": "93b0fdb6-be8b-4663-b59e-05ed19a9ea09",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Epoch 1/10, Loss: 600.8090852049173, Acc: 0.9945839732715815\n",
+      "Epoch 2/10, Loss: 565.9542879898308, Acc: 0.9999073566261442\n",
+      "Epoch 3/10, Loss: 565.9275637627202, Acc: 0.9999969933728429\n",
+      "Epoch 4/10, Loss: 565.927609191542, Acc: 0.9999961959888584\n",
+      "Epoch 5/10, Loss: 565.928202885308, Acc: 0.9999953721249991\n",
+      "Epoch 6/10, Loss: 565.9323843971484, Acc: 0.9999969051674709\n",
+      "Epoch 7/10, Loss: 565.9298919086365, Acc: 0.9999935973983517\n",
+      "Epoch 8/10, Loss: 565.9299267993255, Acc: 0.9999985970973472\n",
+      "Epoch 9/10, Loss: 565.9306044380719, Acc: 0.9999947955797296\n",
+      "Epoch 10/10, Loss: 565.9329843268798, Acc: 0.9999973784035556\n",
+      "Model weights: -3.7066140776793373, bias: 1.8709382558479912\n",
+      "Prediction for test data: 0.13756338580653613\n"
+     ]
+    }
+   ],
+   "source": [
+    "learning_rate = 1e-2\n",
+    "num_epochs = 10\n",
+    "batch_size = 1024\n",
+    "device = \"cuda:0\" if torch.cuda.is_available() else \"cpu\"\n",
+    "\n",
+    "dataset = My_Dataset()\n",
+    "dataloader = DataLoader(\n",
+    "    dataset=dataset, batch_size=batch_size, shuffle=True, num_workers=5, pin_memory=True\n",
+    ")\n",
+    "\n",
+    "model = Model_2_2().to(device)\n",
+    "criterion = nn.BCELoss()\n",
+    "optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate)\n",
+    "\n",
+    "for epoch in range(num_epochs):\n",
+    "    total_epoch_loss = 0\n",
+    "    total_epoch_pred = 0\n",
+    "    total_epoch_target = 0\n",
+    "    for x, targets in dataloader:\n",
+    "        optimizer.zero_grad()\n",
+    "\n",
+    "        x = x.to(device)\n",
+    "        targets = targets.to(device)\n",
+    "\n",
+    "        x = x.unsqueeze(1)\n",
+    "        targets = targets.unsqueeze(1)\n",
+    "        y_pred = model(x)\n",
+    "        loss = criterion(y_pred, targets)\n",
+    "        total_epoch_loss += loss.item()\n",
+    "        total_epoch_target += targets.sum().item()\n",
+    "        total_epoch_pred += y_pred.sum().item()\n",
+    "\n",
+    "        loss.backward()\n",
+    "        optimizer.step()\n",
+    "\n",
+    "    print(\n",
+    "        f\"Epoch {epoch + 1}/{num_epochs}, Loss: {total_epoch_loss}, Acc: {1 - abs(total_epoch_pred - total_epoch_target) / total_epoch_target}\"\n",
+    "    )\n",
+    "\n",
+    "with torch.no_grad():\n",
+    "    test_data = (np.array([[2]]) - dataset.min_x) / (dataset.max_x - dataset.min_x)\n",
+    "    test_data = Variable(\n",
+    "        torch.tensor(test_data, dtype=torch.float64), requires_grad=False\n",
+    "    ).to(device)\n",
+    "    predicted = model(test_data).to(\"cpu\")\n",
+    "    print(\n",
+    "        f\"Model weights: {model.linear.weight.item()}, bias: {model.linear.bias.item()}\"\n",
+    "    )\n",
+    "    print(f\"Prediction for test data: {predicted.item()}\")\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "e6bff679-f8d2-46cc-bdcb-82af7dab38b3",
+   "metadata": {},
+   "source": [
+    "对比发现，使用torch.nn的内置损失函数和优化器，正确率提升更快。\n",
+    "\n",
+    "但是为什么相同分布的数据集训练出的权重和偏置，以及预测结果存在较大差别，这个问题的原因还有待我探究。"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "ef41d7fa-c2bf-4024-833b-60af0a87043a",
+   "metadata": {},
+   "source": [
+    "# 三、动手实现softmax回归\n",
+    "\n",
+    "## 问题1\n",
+    "\n",
+    "**要求动手从0实现softmax回归（只借助Tensor和Numpy相关的库）在Fashion-MNIST数据集上进行训练和测试，并从loss、训练集以及测试集上的准确率等多个角度对结果进行分析（要求从零实现交叉熵损失函数）**"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "3c356760-75a8-4814-ba69-73b270396a4e",
+   "metadata": {},
+   "source": [
+    "手动实现nn.one_hot()。\n",
+    "\n",
+    "one-hot向量用于消除线性标签值所映射的类别的非线性。\n",
+    "\n",
+    "one-hot向量是使用一个长度为分类数量的数组表示标签值，其中有且仅有1个值为为1，该值的下标为标签值；其余为0。\n",
+    "\n",
+    "原理很简单，步骤如下：\n",
+    "1. 初始化全零的张量，大小为（标签数量，分类数量）；\n",
+    "2. 将标签值映射到全零张量的\\[下标，标签值\\]中，将该位置为1；\n",
+    "3. 返回修改后的张量，即是ont-hot向量。"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 13,
+   "id": "e605f1b0-1d32-410f-bddf-402a85ccc9ff",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "def my_one_hot(indices: torch.Tensor, num_classes: int):\n",
+    "    one_hot_tensor = torch.zeros(len(indices), num_classes).to(indices.device)\n",
+    "    one_hot_tensor.scatter_(1, indices.view(-1, 1), 1)\n",
+    "    return one_hot_tensor"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "902603a6-bfb9-4ce3-bd0d-b00cebb1d3cb",
+   "metadata": {},
+   "source": [
+    "手动实现CrossEntropyLoss。\n",
+    "\n",
+    "CrossEntropyLoss由一个log_softmax和一个nll_loss组成。\n",
+    "\n",
+    "softmax的数学表达式如下：\n",
+    "$$\n",
+    "\\text{softmax}(y_i) = \\frac{e^{y_i - \\text{max}(y)}}{\\sum_{j=1}^{N} e^{y_j - \\text{max}(y)}} \n",
+    "$$\n",
+    "log_softmax即为$\\log\\left(softmax\\left(y\\right)\\right)$。\n",
+    "\n",
+    "CrossEntropyLoss的数学表达式如下：\n",
+    "$$\n",
+    "\\text{CrossEntropyLoss}(y, \\hat{y}) = -\\frac{1}{N} \\sum_{i=1}^{N} \\hat{y}_i \\cdot \\log(\\text{softmax}(y_i)) \n",
+    "$$\n",
+    "\n",
+    "故代码如下："
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 14,
+   "id": "759a3bb2-b5f4-4ea5-a2d7-15f0c4cdd14b",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "输入：\n",
+      "tensor([[-1.2914,  0.4715, -0.0432,  1.7427, -1.9236],\n",
+      "        [ 0.5361, -0.7551, -0.6810,  1.0945,  0.6135],\n",
+      "        [-1.3398, -0.0026, -1.6066, -0.4659, -1.6076]], requires_grad=True)\n",
+      "标签：\n",
+      "tensor([[1., 0., 0., 0., 0.],\n",
+      "        [0., 0., 0., 0., 1.],\n",
+      "        [0., 0., 0., 0., 1.]])\n",
+      "My_CrossEntropyLoss损失值: 2.4310648441314697\n",
+      "nn.CrossEntropyLoss损失值: 2.4310646057128906\n"
+     ]
+    }
+   ],
+   "source": [
+    "class My_CrossEntropyLoss:\n",
+    "    def __call__(self, predictions: torch.Tensor, targets: torch.Tensor):\n",
+    "        max_values = torch.max(predictions, dim=1, keepdim=True).values\n",
+    "        exp_values = torch.exp(predictions - max_values)\n",
+    "        softmax_output = exp_values / torch.sum(exp_values, dim=1, keepdim=True)\n",
+    "        log_probs = torch.log(softmax_output)\n",
+    "        \n",
+    "        nll_loss = -torch.sum(targets * log_probs, dim=1)\n",
+    "        average_loss = torch.mean(nll_loss)\n",
+    "        return average_loss\n",
+    "\n",
+    "        \n",
+    "# 测试\n",
+    "input = torch.randn(3, 5, requires_grad=True)\n",
+    "target = torch.randn(3, 5).softmax(dim=1).argmax(1)\n",
+    "target = torch.nn.functional.one_hot(target, num_classes=5).to(dtype=torch.float32)\n",
+    "print(f\"输入：\\n{input}\")\n",
+    "print(f\"标签：\\n{target}\")\n",
+    "\n",
+    "my_crossentropyloss = My_CrossEntropyLoss()\n",
+    "my_loss = my_crossentropyloss(input, target)\n",
+    "print(\"My_CrossEntropyLoss损失值:\", my_loss.item())\n",
+    "\n",
+    "nn_crossentropyloss = nn.CrossEntropyLoss()\n",
+    "nn_loss = nn_crossentropyloss(input, target)\n",
+    "print(\"nn.CrossEntropyLoss损失值:\", nn_loss.item())"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "dbf78501-f5be-4008-986c-d331d531491f",
+   "metadata": {},
+   "source": [
+    "手动实现Flatten。\n",
+    "\n",
+    "原理很简单，就是把多维的张量拉直成一个向量。"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 15,
+   "id": "74322629-8325-4823-b80f-f28182d577c1",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "class My_Flatten:\n",
+    "    def __call__(self, x: torch.Tensor):\n",
+    "        return self.forward(x)\n",
+    "\n",
+    "    def forward(self, x: torch.Tensor):\n",
+    "        x = x.view(x.shape[0], -1)\n",
+    "        return x"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "35aee905-ae37-4faa-a7f1-a04cd8579f78",
+   "metadata": {},
+   "source": [
+    "手动实现softmax回归模型。\n",
+    "\n",
+    "模型很简单，主要由一个Flatten层和一个线性层组成。\n",
+    "\n",
+    "Flatten层主要用于将2维的图像展开，直接作为1维的特征量输入网络。"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 16,
+   "id": "bb31a75e-464c-4b94-b927-b219a765e35d",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "class Model_3_1:\n",
+    "    def __init__(self, num_classes):\n",
+    "        self.flatten = My_Flatten()\n",
+    "        self.linear = My_Linear(28 * 28, num_classes)\n",
+    "        self.params = self.linear.params\n",
+    "\n",
+    "    def __call__(self, x: torch.Tensor):\n",
+    "        return self.forward(x)\n",
+    "\n",
+    "    def forward(self, x: torch.Tensor):\n",
+    "        x = self.flatten(x)\n",
+    "        x = self.linear(x)\n",
+    "        return x\n",
+    "\n",
+    "    def to(self, device: str):\n",
+    "        for param in self.params:\n",
+    "            param.data = param.data.to(device=device)\n",
+    "        return self\n",
+    "\n",
+    "    def parameters(self):\n",
+    "        return self.params"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "17e686d1-9c9a-4727-8fdc-9990d348c523",
+   "metadata": {},
+   "source": [
+    "训练与测试过程与之前手动实现的几乎一致。由于数据集的变化，对应超参数也进行了调整。\n",
+    "\n",
+    "数据集也使用了现成的FashionMNIST数据集，且划分了训练集和测试集。\n",
+    "\n",
+    "FashionMNIST数据集直接调用API获取。数据集的image为28*28的单通道灰白图片，label为单个数值标签。\n",
+    "\n",
+    "训练softmax回归模型。\n",
+    "进行如下步骤：\n",
+    "1. 初始化超参数\n",
+    "2. 获取数据集\n",
+    "3. 初始化模型\n",
+    "4. 定义损失函数和优化器\n",
+    "5. 训练\n",
+    "    1. 从训练dataloader中获取批量数据\n",
+    "    2. 传入模型\n",
+    "    3. 使用损失函数计算与ground_truth的损失\n",
+    "    4. 使用优化器进行反向传播\n",
+    "    5. 循环以上步骤\n",
+    "6. 测试\n",
+    "    1. 从测试dataloader中获取批量数据\n",
+    "    2. 传入模型\n",
+    "    3. 将预测值与ground_truth进行比较，得出正确率\n",
+    "    4. 对整个训练集统计正确率，从而分析训练效果"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 17,
+   "id": "d816dae1-5fbe-4c29-9597-19d66b5eb6b4",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Epoch 1/10, Loss: nan, Acc: 0.09999999403953552\n",
+      "Epoch 2/10, Loss: nan, Acc: 0.09999999403953552\n",
+      "Epoch 3/10, Loss: nan, Acc: 0.09999999403953552\n",
+      "Epoch 4/10, Loss: nan, Acc: 0.09999999403953552\n",
+      "Epoch 5/10, Loss: nan, Acc: 0.09999999403953552\n",
+      "Epoch 6/10, Loss: nan, Acc: 0.09999999403953552\n",
+      "Epoch 7/10, Loss: nan, Acc: 0.09999999403953552\n",
+      "Epoch 8/10, Loss: nan, Acc: 0.09999999403953552\n",
+      "Epoch 9/10, Loss: nan, Acc: 0.09999999403953552\n",
+      "Epoch 10/10, Loss: nan, Acc: 0.09999999403953552\n"
+     ]
+    }
+   ],
+   "source": [
+    "learning_rate = 5e-3\n",
+    "num_epochs = 10\n",
+    "batch_size = 4096\n",
+    "num_classes = 10\n",
+    "device = \"cuda:0\" if torch.cuda.is_available() else \"cpu\"\n",
+    "\n",
+    "transform = transforms.Compose(\n",
+    "    [\n",
+    "        transforms.ToTensor(),\n",
+    "        transforms.Normalize((0.5,), (0.5,)),\n",
+    "    ]\n",
+    ")\n",
+    "train_dataset = datasets.FashionMNIST(root=\"./dataset\", train=True, transform=transform, download=True)\n",
+    "test_dataset = datasets.FashionMNIST(root=\"./dataset\", train=False, transform=transform, download=True)\n",
+    "train_loader = DataLoader(\n",
+    "    dataset=train_dataset, batch_size=batch_size,\n",
+    "    shuffle=True, num_workers=4, pin_memory=True,\n",
+    ")\n",
+    "test_loader = DataLoader(\n",
+    "    dataset=test_dataset, batch_size=batch_size,\n",
+    "    shuffle=True, num_workers=4, pin_memory=True,\n",
+    ")\n",
+    "\n",
+    "model = Model_3_1(num_classes).to(device)\n",
+    "criterion = My_CrossEntropyLoss()\n",
+    "optimizer = My_optimizer(model.parameters(), lr=learning_rate)\n",
+    "\n",
+    "for epoch in range(num_epochs):\n",
+    "    total_epoch_loss = 0\n",
+    "    for images, targets in train_loader:\n",
+    "        optimizer.zero_grad()\n",
+    "\n",
+    "        images = images.to(device)\n",
+    "        targets = targets.to(device).to(dtype=torch.long)\n",
+    "\n",
+    "        one_hot_targets = (\n",
+    "            my_one_hot(targets, num_classes=num_classes)\n",
+    "            .to(device)\n",
+    "            .to(dtype=torch.long)\n",
+    "        )\n",
+    "\n",
+    "        outputs = model(images)\n",
+    "        loss = criterion(outputs, one_hot_targets)\n",
+    "        total_epoch_loss += loss\n",
+    "\n",
+    "        loss.backward()\n",
+    "        optimizer.step()\n",
+    "\n",
+    "    total_acc = 0\n",
+    "    with torch.no_grad():\n",
+    "        for image, targets in test_loader:\n",
+    "            image = image.to(device)\n",
+    "            targets = targets.to(device)\n",
+    "            outputs = model(image)\n",
+    "            total_acc += (outputs.argmax(1) == targets).sum()\n",
+    "    print(\n",
+    "        f\"Epoch {epoch + 1}/{num_epochs}, Loss: {total_epoch_loss}, Acc: {total_acc / len(test_dataset)}\"\n",
+    "    )"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "a49d0165-aeb7-48c0-9b67-956bb08cb356",
+   "metadata": {},
+   "source": [
+    "这里发现梯度爆炸。暂时无法解决。"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "3ef5240f-8a11-4678-bfce-f1cbc7e71b77",
+   "metadata": {},
+   "source": [
+    "## 问题2\n",
+    "\n",
+    "**利用torch.nn实现softmax回归在Fashion-MNIST数据集上进行训练和测试，并从loss，训练集以及测试集上的准确率等多个角度对结果进行分析**"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "5c4a88c6-637e-4af5-bed5-f644685dcabc",
+   "metadata": {},
+   "source": [
+    "使用torch.nn实现模型。\n",
+    "\n",
+    "将之前的Model_3_1中的手动实现函数改为torch.nn内置函数即可，再加上继承nn.Module以使用torch.nn内置模型模板特性。"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 18,
+   "id": "0163b9f7-1019-429c-8c29-06436d0a4c98",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "class Model_3_2(nn.Module):\n",
+    "    def __init__(self, num_classes):\n",
+    "        super(Model_3_2, self).__init__()\n",
+    "        self.flatten = nn.Flatten()\n",
+    "        self.linear = nn.Linear(28 * 28, num_classes)\n",
+    "\n",
+    "    def forward(self, x: torch.Tensor):\n",
+    "        x = self.flatten(x)\n",
+    "        x = self.linear(x)\n",
+    "        return x"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "6e765ad7-c1c6-4166-bd7f-361666bd4016",
+   "metadata": {},
+   "source": [
+    "训练与测试过程与之前手动实现的几乎一致。"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 19,
+   "id": "a58a23e1-368c-430a-ad62-0e256dff564d",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Epoch 1/10, Loss: 15.148970603942871, Acc: 0.7520999908447266\n",
+      "Epoch 2/10, Loss: 9.012335777282715, Acc: 0.7996999621391296\n",
+      "Epoch 3/10, Loss: 7.9114227294921875, Acc: 0.8095999956130981\n",
+      "Epoch 4/10, Loss: 7.427404403686523, Acc: 0.8215999603271484\n",
+      "Epoch 5/10, Loss: 7.084254264831543, Acc: 0.8277999758720398\n",
+      "Epoch 6/10, Loss: 6.885956287384033, Acc: 0.8274999856948853\n",
+      "Epoch 7/10, Loss: 6.808426380157471, Acc: 0.8327999711036682\n",
+      "Epoch 8/10, Loss: 6.647855758666992, Acc: 0.8323000073432922\n",
+      "Epoch 9/10, Loss: 6.560361862182617, Acc: 0.8317999839782715\n",
+      "Epoch 10/10, Loss: 6.5211310386657715, Acc: 0.8349999785423279\n"
+     ]
+    }
+   ],
+   "source": [
+    "learning_rate = 5e-3\n",
+    "num_epochs = 10\n",
+    "batch_size = 4096\n",
+    "num_classes = 10\n",
+    "device = \"cuda:0\" if torch.cuda.is_available() else \"cpu\"\n",
+    "\n",
+    "transform = transforms.Compose(\n",
+    "    [\n",
+    "        transforms.ToTensor(),\n",
+    "        transforms.Normalize((0.5,), (0.5,)),\n",
+    "    ]\n",
+    ")\n",
+    "train_dataset = datasets.FashionMNIST(\n",
+    "    root=\"./dataset\", train=True, transform=transform, download=True\n",
+    ")\n",
+    "test_dataset = datasets.FashionMNIST(\n",
+    "    root=\"./dataset\", train=False, transform=transform, download=True\n",
+    ")\n",
+    "train_loader = DataLoader(\n",
+    "    dataset=train_dataset,\n",
+    "    batch_size=batch_size,\n",
+    "    shuffle=True,\n",
+    "    num_workers=4,\n",
+    "    pin_memory=True,\n",
+    ")\n",
+    "test_loader = DataLoader(\n",
+    "    dataset=test_dataset,\n",
+    "    batch_size=batch_size,\n",
+    "    shuffle=True,\n",
+    "    num_workers=4,\n",
+    "    pin_memory=True,\n",
+    ")\n",
+    "\n",
+    "model = Model_3_2(num_classes).to(device)\n",
+    "criterion = nn.CrossEntropyLoss()\n",
+    "optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate)\n",
+    "\n",
+    "for epoch in range(num_epochs):\n",
+    "    total_epoch_loss = 0\n",
+    "    model.train()\n",
+    "    for images, targets in train_loader:\n",
+    "        optimizer.zero_grad()\n",
+    "\n",
+    "        images = images.to(device)\n",
+    "        targets = targets.to(device)\n",
+    "\n",
+    "        one_hot_targets = (\n",
+    "            torch.nn.functional.one_hot(targets, num_classes=num_classes)\n",
+    "            .to(device)\n",
+    "            .to(dtype=torch.float32)\n",
+    "        )\n",
+    "\n",
+    "        outputs = model(images)\n",
+    "        loss = criterion(outputs, one_hot_targets)\n",
+    "        total_epoch_loss += loss\n",
+    "\n",
+    "        loss.backward()\n",
+    "        optimizer.step()\n",
+    "\n",
+    "    model.eval()\n",
+    "    total_acc = 0\n",
+    "    with torch.no_grad():\n",
+    "        for image, targets in test_loader:\n",
+    "            image = image.to(device)\n",
+    "            targets = targets.to(device)\n",
+    "            outputs = model(image)\n",
+    "            total_acc += (outputs.argmax(1) == targets).sum()\n",
+    "    print(\n",
+    "        f\"Epoch {epoch + 1}/{num_epochs}, Loss: {total_epoch_loss}, Acc: {total_acc / len(test_dataset)}\"\n",
+    "    )"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "59555b67-1650-4e1a-a98e-7906878bf3d0",
+   "metadata": {},
+   "source": [
+    "与手动实现的softmax回归相比较，nn.CrossEntropyLoss比手动实现的My_CrossEntropyLoss更加稳定，没有出现梯度爆炸的情况。"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "f40431f2-e77b-4ead-81a3-ff6451a8e452",
+   "metadata": {},
+   "source": [
+    "**实验心得体会**\n",
+    "\n",
+    "通过完成本次Pytorch基本操作实验，让我对Pytorch框架有了更加深入的理解。我接触深度学习主要是在大语言模型领域，比较熟悉微调大模型，但是涉及到底层的深度学习知识，我还有很多短板和不足。这次实验对我这方面的锻炼让我收获良多。\n",
+    "\n",
+    "首先是数据集的设置。如果数据没有进行归一化，很容易出现梯度爆炸。这是在我以前直接使用图片数据集的经历中没有遇到过的问题。\n",
+    "\n",
+    "在实现logistic回归模型时，通过手动实现各个组件如优化器、线性层等，让我对这些模块的工作原理有了更清晰的认识。尤其是在实现广播机制时，需要充分理解张量操作的维度变换规律。而使用Pytorch内置模块进行实现时，通过继承nn.Module可以自动获得許多功能，使代码更加简洁。\n",
+    "\n",
+    "在实现softmax回归时，则遇到了更大的困难。手动实现的模型很容易出现梯度爆炸的问题，而使用Pytorch内置的损失函数和优化器则可以稳定训练。这让我意识到了选择合适的优化方法的重要性。另外，Pytorch强大的自动微分机制也是构建深度神经网络的重要基础。\n",
+    "\n",
+    "通过这个实验，让我对Pytorch框架有了更加直观的感受，也让我看到了仅靠基础模块搭建复杂模型的难点所在。这些经验对我后续使用Pytorch构建数据集模型会很有帮助。"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3 (ipykernel)",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.10.13"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}
diff --git a/Lab1/code/1.1.py b/Lab1/code/1.1.py
new file mode 100644
index 0000000..5f23159
--- /dev/null
+++ b/Lab1/code/1.1.py
@@ -0,0 +1,39 @@
+import torch
+
+A = torch.tensor([[1, 2, 3]])
+
+B = torch.tensor([[4],
+                  [5]])
+
+# 方法1: 使用PyTorch的减法操作符
+result1 = A - B
+
+# 方法2: 使用PyTorch的sub函数
+result2 = torch.sub(A, B)
+
+# 方法3: 手动实现广播机制并作差
+def mysub(a:torch.Tensor, b:torch.Tensor):
+    if not (
+        (a.size(0) == 1 and b.size(1) == 1) 
+        or 
+        (a.size(1) == 1 and b.size(0) == 1)
+        ):
+        raise ValueError("输入的张量大小无法满足广播机制的条件。")
+    else:
+        target_shape = torch.Size([max(A.size(0), B.size(0)), max(A.size(1), B.size(1))])
+        A_broadcasted = A.expand(target_shape)
+        B_broadcasted = B.expand(target_shape)
+        result = torch.zeros(target_shape, dtype=torch.int64).to(device=A_broadcasted.device)
+        for i in range(target_shape[0]):
+            for j in range(target_shape[1]):
+                result[i, j] = A_broadcasted[i, j] - B_broadcasted[i, j]
+        return result
+
+result3 = mysub(A, B)
+
+print("方法1的结果:")
+print(result1)
+print("方法2的结果:")
+print(result2)
+print("方法3的结果:")
+print(result3)
diff --git a/Lab1/code/1.2.py b/Lab1/code/1.2.py
new file mode 100644
index 0000000..66986c2
--- /dev/null
+++ b/Lab1/code/1.2.py
@@ -0,0 +1,23 @@
+import torch
+
+mean = 0
+stddev = 0.01
+
+P = torch.normal(mean=mean, std=stddev, size=(3, 2))
+Q = torch.normal(mean=mean, std=stddev, size=(4, 2))
+
+print("矩阵 P:")
+print(P)
+print("矩阵 Q:")
+print(Q)
+
+# 对矩阵Q进行转置操作，得到矩阵Q的转置Q^T
+QT = Q.T
+print("矩阵 QT:")
+print(QT)
+
+# 计算矩阵P和矩阵Q^T的矩阵相乘
+result = torch.matmul(P, QT)
+print("矩阵相乘的结果:")
+print(result)
+
diff --git a/Lab1/code/1.3.py b/Lab1/code/1.3.py
new file mode 100644
index 0000000..ef057e1
--- /dev/null
+++ b/Lab1/code/1.3.py
@@ -0,0 +1,12 @@
+import torch
+
+x = torch.tensor(1.0, requires_grad=True)
+y_1 = x**2
+with torch.no_grad():
+    y_2 = x**3
+
+y3 = y_1 + y_2
+
+y3.backward()
+
+print("梯度(dy_3/dx): ", x.grad.item())
diff --git a/Lab1/code/2.1.py b/Lab1/code/2.1.py
new file mode 100644
index 0000000..7b10711
--- /dev/null
+++ b/Lab1/code/2.1.py
@@ -0,0 +1,143 @@
+import numpy as np
+import torch
+from torch.autograd import Variable
+from torch.utils.data import Dataset, DataLoader
+from tqdm import tqdm
+import ipdb
+
+
+class My_BCELoss:
+    def __call__(self, prediction: torch.Tensor, target: torch.Tensor):
+        loss = -torch.mean(
+            target * torch.log(prediction) + (1 - target) * torch.log(1 - prediction)
+        )
+        return loss
+
+
+class My_optimizer:
+    def __init__(self, params: list[torch.Tensor], lr: float):
+        self.params = params
+        self.lr = lr
+
+    def step(self):
+        for param in self.params:
+            param.data = param.data - self.lr * param.grad.data
+
+    def zero_grad(self):
+        for param in self.params:
+            if param.grad is not None:
+                param.grad.data.zero_()
+
+
+class My_Linear:
+    def __init__(self, input_feature: int, output_feature: int):
+        self.weight = torch.randn(
+            (output_feature, input_feature), requires_grad=True, dtype=torch.float32
+        )
+        self.bias = torch.randn(1, requires_grad=True, dtype=torch.float32)
+        self.params = [self.weight, self.bias]
+
+    def __call__(self, x):
+        return self.forward(x)
+
+    def forward(self, x):
+        x = torch.matmul(x, self.weight.T) + self.bias
+        return x
+
+    def to(self, device: str):
+        for param in self.params:
+            param.data = param.data.to(device=device)
+        return self
+
+    def parameters(self):
+        return self.params
+
+
+class Model:
+    def __init__(self):
+        self.linear = My_Linear(1, 1)
+        self.params = self.linear.params
+
+    def __call__(self, x):
+        return self.forward(x)
+
+    def forward(self, x):
+        x = self.linear(x)
+        x = torch.sigmoid(x)
+        return x
+
+    def to(self, device: str):
+        for param in self.params:
+            param.data = param.data.to(device=device)
+        return self
+
+    def parameters(self):
+        return self.params
+
+
+class My_Dataset(Dataset):
+    def __init__(self, data_size=1000000):
+        np.random.seed(0)
+        x = 2 * np.random.rand(data_size, 1)
+        noise = 0.2 * np.random.randn(data_size, 1)
+        y = 4 - 3 * x + noise
+        self.min_x, self.max_x = np.min(x), np.max(x)
+        min_y, max_y = np.min(y), np.max(y)
+        x = (x - self.min_x) / (self.max_x - self.min_x)
+        y = (y - min_y) / (max_y - min_y)
+        self.data = [[x[i][0], y[i][0]] for i in range(x.shape[0])]
+
+    def __len__(self):
+        return len(self.data)
+
+    def __getitem__(self, index):
+        x, y = self.data[index]
+        return x, y
+
+
+learning_rate = 5e-2
+num_epochs = 10
+batch_size = 1024
+device = "cuda:0" if torch.cuda.is_available() else "cpu"
+
+dataset = My_Dataset()
+dataloader = DataLoader(
+    dataset=dataset, batch_size=batch_size, shuffle=True, num_workers=5, pin_memory=True
+)
+
+model = Model().to(device)
+criterion = My_BCELoss()
+optimizer = My_optimizer(model.parameters(), lr=learning_rate)
+
+for epoch in range(num_epochs):
+    total_epoch_loss = 0
+    total_epoch_pred = 0
+    total_epoch_target = 0
+    for index, (x, targets) in tqdm(enumerate(dataloader), total=len(dataloader)):
+        optimizer.zero_grad()
+        x = x.to(device).to(dtype=torch.float32)
+        targets = targets.to(device).to(dtype=torch.float32)
+        x = x.unsqueeze(1)
+        y_pred = model(x)
+        loss = criterion(y_pred, targets)
+        total_epoch_loss += loss.item()
+        total_epoch_target += targets.sum().item()
+        total_epoch_pred += y_pred.sum().item()
+
+        loss.backward()
+        optimizer.step()
+
+    print(
+        f"Epoch {epoch + 1}/{num_epochs}, Loss: {total_epoch_loss}, Acc: {1 - abs(total_epoch_pred - total_epoch_target) / total_epoch_target}"
+    )
+
+with torch.no_grad():
+    test_data = (np.array([[2]]) - dataset.min_x) / (dataset.max_x - dataset.min_x)
+    test_data = Variable(
+        torch.tensor(test_data, dtype=torch.float64), requires_grad=False
+    ).to(device)
+    predicted = model(test_data).to("cpu")
+    print(
+        f"Model weights: {model.linear.weight.item()}, bias: {model.linear.bias.item()}"
+    )
+    print(f"Prediction for test data: {predicted.item()}")
diff --git a/Lab1/code/2.2.py b/Lab1/code/2.2.py
new file mode 100644
index 0000000..1015a09
--- /dev/null
+++ b/Lab1/code/2.2.py
@@ -0,0 +1,89 @@
+import numpy as np
+import torch
+from torch.autograd import Variable
+from torch.utils.data import Dataset, DataLoader
+from torch import nn
+from tqdm import tqdm
+import ipdb
+
+
+class Model(nn.Module):
+    def __init__(self):
+        super(Model, self).__init__()
+        self.linear = nn.Linear(1, 1, dtype=torch.float64)
+
+    def forward(self, x):
+        x = self.linear(x)
+        x = torch.sigmoid(x)
+        return x
+
+
+class My_Dataset(Dataset):
+    def __init__(self, data_size=1000000):
+        np.random.seed(0)
+        x = 2 * np.random.rand(data_size, 1)
+        noise = 0.2 * np.random.randn(data_size, 1)
+        y = 4 - 3 * x + noise
+        self.min_x, self.max_x = np.min(x), np.max(x)
+        min_y, max_y = np.min(y), np.max(y)
+        x = (x - self.min_x) / (self.max_x - self.min_x)
+        y = (y - min_y) / (max_y - min_y)
+        self.data = [[x[i][0], y[i][0]] for i in range(x.shape[0])]
+
+    def __len__(self):
+        return len(self.data)
+
+    def __getitem__(self, index):
+        x, y = self.data[index]
+        return x, y
+
+
+learning_rate = 1e-2
+num_epochs = 10
+batch_size = 1024
+device = "cuda:0" if torch.cuda.is_available() else "cpu"
+
+dataset = My_Dataset()
+dataloader = DataLoader(
+    dataset=dataset, batch_size=batch_size, shuffle=True, num_workers=5, pin_memory=True
+)
+
+model = Model().to(device)
+criterion = nn.BCELoss()
+optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate)
+
+for epoch in range(num_epochs):
+    total_epoch_loss = 0
+    total_epoch_pred = 0
+    total_epoch_target = 0
+    for index, (x, targets) in tqdm(enumerate(dataloader), total=len(dataloader)):
+        optimizer.zero_grad()
+
+        x = x.to(device)
+        targets = targets.to(device)
+
+        x = x.unsqueeze(1)
+        targets = targets.unsqueeze(1)
+        y_pred = model(x)
+        loss = criterion(y_pred, targets)
+        total_epoch_loss += loss.item()
+        total_epoch_target += targets.sum().item()
+        total_epoch_pred += y_pred.sum().item()
+
+        loss.backward()
+        optimizer.step()
+
+    print(
+        f"Epoch {epoch + 1}/{num_epochs}, Loss: {total_epoch_loss}, Acc: {1 - abs(total_epoch_pred - total_epoch_target) / total_epoch_target}"
+    )
+
+with torch.no_grad():
+    test_data = (np.array([[2]]) - dataset.min_x) / (dataset.max_x - dataset.min_x)
+    test_data = Variable(
+        torch.tensor(test_data, dtype=torch.float64), requires_grad=False
+    ).to(device)
+    predicted = model(test_data).to("cpu")
+    print(
+        f"Model weights: {model.linear.weight.item()}, bias: {model.linear.bias.item()}"
+    )
+    print(f"Prediction for test data: {predicted.item()}")
diff --git a/Lab1/code/3.1.py b/Lab1/code/3.1.py
new file mode 100644
index 0000000..6dd8b47
--- /dev/null
+++ b/Lab1/code/3.1.py
@@ -0,0 +1,169 @@
+import numpy as np
+import torch
+from torch.utils.data import Dataset, DataLoader
+from torch import nn
+from tqdm import tqdm
+from torchvision import datasets, transforms
+from torch.utils.data import DataLoader
+
+
+def my_one_hot(indices: torch.Tensor, num_classes: int):
+    one_hot_tensor = torch.zeros(len(indices), num_classes).to(indices.device)
+    one_hot_tensor.scatter_(1, indices.view(-1, 1), 1)
+    return one_hot_tensor
+
+
+class My_CrossEntropyLoss:
+    def __call__(self, predictions: torch.Tensor, targets: torch.Tensor):
+        max_values = torch.max(predictions, dim=1, keepdim=True).values
+        exp_values = torch.exp(predictions - max_values)
+        softmax_output = exp_values / torch.sum(exp_values, dim=1, keepdim=True)
+
+        log_probs = torch.log(softmax_output)
+        nll_loss = -torch.sum(targets * log_probs, dim=1)
+        average_loss = torch.mean(nll_loss)
+        return average_loss
+
+
+class My_optimizer:
+    def __init__(self, params: list[torch.Tensor], lr: float):
+        self.params = params
+        self.lr = lr
+
+    def step(self):
+        for param in self.params:
+            param.data = param.data - self.lr * param.grad.data
+
+    def zero_grad(self):
+        for param in self.params:
+            if param.grad is not None:
+                param.grad.data.zero_()
+
+
+class My_Linear:
+    def __init__(self, input_feature: int, output_feature: int):
+        self.weight = torch.randn(
+            (output_feature, input_feature), requires_grad=True, dtype=torch.float32
+        )
+        self.bias = torch.randn(1, requires_grad=True, dtype=torch.float32)
+        self.params = [self.weight, self.bias]
+
+    def __call__(self, x: torch.Tensor):
+        return self.forward(x)
+
+    def forward(self, x: torch.Tensor):
+        x = torch.matmul(x, self.weight.T) + self.bias
+        return x
+
+    def to(self, device: str):
+        for param in self.params:
+            param.data = param.data.to(device=device)
+        return self
+
+    def parameters(self):
+        return self.params
+
+
+class My_Flatten:
+    def __call__(self, x: torch.Tensor):
+        return self.forward(x)
+
+    def forward(self, x: torch.Tensor):
+        x = x.view(x.shape[0], -1)
+        return x
+
+
+class Model_3_1:
+    def __init__(self, num_classes):
+        self.flatten = My_Flatten()
+        self.linear = My_Linear(28 * 28, num_classes)
+        self.params = self.linear.params
+
+    def __call__(self, x: torch.Tensor):
+        return self.forward(x)
+
+    def forward(self, x: torch.Tensor):
+        x = self.flatten(x)
+        x = self.linear(x)
+        return x
+
+    def to(self, device: str):
+        for param in self.params:
+            param.data = param.data.to(device=device)
+        return self
+
+    def parameters(self):
+        return self.params
+
+
+learning_rate = 5e-3
+num_epochs = 10
+batch_size = 4096
+num_classes = 10
+device = "cuda:0" if torch.cuda.is_available() else "cpu"
+
+transform = transforms.Compose(
+    [
+        transforms.ToTensor(),
+        transforms.Normalize((0.5,), (0.5,)),
+    ]
+)
+train_dataset = datasets.FashionMNIST(
+    root="./dataset", train=True, transform=transform, download=True
+)
+test_dataset = datasets.FashionMNIST(
+    root="./dataset", train=False, transform=transform, download=True
+)
+train_loader = DataLoader(
+    dataset=train_dataset,
+    batch_size=batch_size,
+    shuffle=True,
+    num_workers=4,
+    pin_memory=True,
+)
+test_loader = DataLoader(
+    dataset=test_dataset,
+    batch_size=batch_size,
+    shuffle=True,
+    num_workers=4,
+    pin_memory=True,
+)
+
+model = Model_3_1(num_classes).to(device)
+criterion = My_CrossEntropyLoss()
+optimizer = My_optimizer(model.parameters(), lr=learning_rate)
+
+for epoch in range(num_epochs):
+    total_epoch_loss = 0
+    for index, (images, targets) in tqdm(
+        enumerate(train_loader), total=len(train_loader)
+    ):
+        optimizer.zero_grad()
+
+        images = images.to(device)
+        targets = targets.to(device).to(dtype=torch.long)
+
+        one_hot_targets = (
+            my_one_hot(targets, num_classes=num_classes).to(device).to(dtype=torch.long)
+        )
+
+        outputs = model(images)
+        # ipdb.set_trace()
+        loss = criterion(outputs, one_hot_targets)
+        total_epoch_loss += loss
+
+        loss.backward()
+        optimizer.step()
+
+    total_acc = 0
+    with torch.no_grad():
+        for index, (image, targets) in tqdm(
+            enumerate(test_loader), total=len(test_loader)
+        ):
+            image = image.to(device)
+            targets = targets.to(device)
+            outputs = model(image)
+            total_acc += (outputs.argmax(1) == targets).sum()
+    print(
+        f"Epoch {epoch + 1}/{num_epochs} Train, Loss: {total_epoch_loss}, Acc: {total_acc / len(test_dataset)}"
+    )
diff --git a/Lab1/code/3.2.py b/Lab1/code/3.2.py
new file mode 100644
index 0000000..8624929
--- /dev/null
+++ b/Lab1/code/3.2.py
@@ -0,0 +1,96 @@
+import numpy as np
+import torch
+from torch.utils.data import DataLoader
+from torch import nn
+from tqdm import tqdm
+from torchvision import datasets, transforms
+from torch.utils.data import DataLoader
+import ipdb
+
+
+class Model(nn.Module):
+    def __init__(self, num_classes):
+        super(Model, self).__init__()
+        self.flatten = nn.Flatten()
+        self.linear = nn.Linear(28 * 28, num_classes)
+
+    def forward(self, x: torch.Tensor):
+        x = self.flatten(x)
+        x = self.linear(x)
+        return x
+
+
+learning_rate = 5e-3
+num_epochs = 10
+batch_size = 4096
+num_classes = 10
+device = "cuda:0" if torch.cuda.is_available() else "cpu"
+
+transform = transforms.Compose(
+    [
+        transforms.ToTensor(),
+        transforms.Normalize((0.5,), (0.5,)),
+    ]
+)
+train_dataset = datasets.FashionMNIST(
+    root="./dataset", train=True, transform=transform, download=True
+)
+test_dataset = datasets.FashionMNIST(
+    root="./dataset", train=False, transform=transform, download=True
+)
+train_loader = DataLoader(
+    dataset=train_dataset,
+    batch_size=batch_size,
+    shuffle=True,
+    num_workers=4,
+    pin_memory=True,
+)
+test_loader = DataLoader(
+    dataset=test_dataset,
+    batch_size=batch_size,
+    shuffle=True,
+    num_workers=4,
+    pin_memory=True,
+)
+
+model = Model(num_classes).to(device)
+criterion = nn.CrossEntropyLoss()
+optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate)
+
+for epoch in range(num_epochs):
+    total_epoch_loss = 0
+    model.train()
+    for index, (images, targets) in tqdm(
+        enumerate(train_loader), total=len(train_loader)
+    ):
+        optimizer.zero_grad()
+
+        images = images.to(device)
+        targets = targets.to(device)
+
+        one_hot_targets = (
+            torch.nn.functional.one_hot(targets, num_classes=num_classes)
+            .to(device)
+            .to(dtype=torch.float32)
+        )
+
+        outputs = model(images)
+        loss = criterion(outputs, one_hot_targets)
+        total_epoch_loss += loss
+
+        loss.backward()
+        optimizer.step()
+
+    model.eval()
+    total_acc = 0
+    with torch.no_grad():
+        for index, (image, targets) in tqdm(
+            enumerate(test_loader), total=len(test_loader)
+        ):
+            image = image.to(device)
+            targets = targets.to(device)
+            outputs = model(image)
+            total_acc += (outputs.argmax(1) == targets).sum()
+    print(
+        f"Epoch {epoch + 1}/{num_epochs} Train, Loss: {total_epoch_loss}, Acc: {total_acc / len(test_dataset)}"
+    )
diff --git a/images/school_logo.png b/images/school_logo.png
new file mode 100644
index 0000000000000000000000000000000000000000..777f05e43c5f50e0339071890c449ae5bde88416
GIT binary patch
literal 116916
zcmV)QK(xP!P)<h;3K|Lk000e1NJLTq00N`{00L780{{R3=@0Ze00001b5ch_0Itp)
z=>Px#1ZP1_K>z@;j|==^1poj7{ZLF)Mcv)q@9*#a{{H{}{{;mF1_lNP2L}iU2nh)Z
z3JMAf3kwVk3=Itp4h{|v4-XI!5D^g(5)u*<6B85^6crT}78Vv47Z(^97#SHE8X6iK
z8yg%P9334U9v&VaA0HqfAR!?kA|fIqBO@dvBqb#!CMG5)CnqQ<C@Co^Dk>@~D=RE4
zEG;c9E-o%FFE21KFflPPGBPqVGcz<aG&MCfHa0dlH#ayqI5{~vIyyQ#J3Bl)JUu-<
zK0ZD_KR-Y~KtVx4LPA1ALqkMFL`6kKMn*<QM@L9VNJ&XaN=iygOG`{lOifKqPEJlw
zPft)#P*G7)Qc_Y=Q&Ut_R8>_~R#sM5S65hASXo(FT3T9LTU%UQTwPsVUS3{bUteHg
zU}0flVq#)rV`F4wWMyS#W@ct*XJ=?=XlZF_YHDh0Yin$5Y;A3AZf<UGZ*OpLaB*>Q
za&mHWb8~cbbai!gc6N4mcXxPrczJnwdU|?$dwYC*e0_a=etv#`e}900fPsO5f`WpB
zgM);GgoTBLhK7cRhlhxWh>3}bii(Phi;IkmjE#+rj*gCxkB^X$kdcv*l9G~>larK`
zl$Dj0mX?;6mzS8Bn3<WGnwpxMo12`RoSmJWo}QkcpP!(hprN6mqN1Xsqobsxq@|^$
zrlzK+r>Cf>sHv%`s;a81tE;T6tgWrBuCA`HudlGMu(7eRva+(Xv$M3cw6(Rhwzjsn
zx3{>sxVgExy1Kf%ySu!+yuH1>zP`S{zrVo1z`?=6!otGC!^6bH#KpzM#>U3S$H&OX
z$jQmc%F4>i%gfBn%+1Zs&d$!y&(F}%(9zM+($dn?)6>+{)YaA1*4Eb7*VowC*xA|H
z+S=ON+uPjS+}+*X-rnBd-{0Wi;Njun;^N}t<KyJy<mKh%=H}+-=jZ6?=;`U{>gww2
z>+9_7?CtIC?(XjI@9*&N@bU5S^78WY^Yird^!4@i_V)Jo_xJet`1$$y`uh6&`}_R-
z{Qdp?{{H^||NlD=cr*Y200eYWPE-E={{Fg3O8fu-010qNS#tmYA=Ll?A=Lo{oV1+)
z0Q`(eL_t(|UhMq`v=r6WHjLge?)}I8?|<(<zVCa_d%lwlguOe5s;-VZ6D5csDvIJU
zf|%1W=YTn5!hje-Ma&Ta6QU$Rl$>@B-Ia6dapzpsz1f?*IfCf%*}XefS5;R%^~^cf
znrp8ANAVBr9j;q?vMcud$$n!mYySf!{6GI7>As*Tz$C(@?g#QdK-Ueu^t7)it*fYN
z<OM$c1AD+9B;6-~L7HZ$@JJqjnt10y>+0;>zO8Zn`t{qlukCSLzi#{Xb@1-GwT;^w
z8`rmJz(Q5m^&Q%)>YC;{j_V*<6=dok*c1LB>Av9-G7)aNuFi3q!8UJL^~35VOV)fn
z>)z|GIRE_frcOQc%nN$n&N%C=Gp0_Rdd8XOUw-{HH$L^wm*1^k`OV688#eH|>XwyL
z<9SolRkb9le_#*zgQWWc@KU!7OLJAp(OWjH{&w-VOW%F?+No!qa`MTijvrpbFid!0
zMa94YJ8lC8Rt~7BfLq0Ya43`sM>7q_pEB{(i4)Jf?vXiPE&SrU?|y1ER87^8oGLsk
zZR!u~9)FN@UjVPuu`RoG@%LYR`u?nYE<N-3<4+t>mq;+70Rw_eFhEl@6$emM_bryq
z!Yz((lh^Pc>=#PVG*vVFn32a%oN(Sfv)=po<1bdXu$rupt^6USd%+(h-3MG-+4RLb
zbLKsG?pY_FGQK_?2_h@$Kp+7eq=0yd5~{1I@{ML{su`xb+WUa6uA*UKiYA~aIzcn3
zhLI!2o;~IESLeL*#g>BVsAg$%e_%KGgQR6Z2!po0^oqbS*;QZ7z3=jg!;TpV+6n9e
zRYbDE^KNpM!7w{+-T&oJ;Ldvv@TdywVd&Zs!;kyt4UfI`$&aewDJt0o@Ae1E_=BVb
zzzKr|*U$@&@EynOTJ`?Jw_iA}K9yp?U?a75W#+#FZ;gRavUc1_7fgHX_3v3hhE3}x
zxT!@&FSk+s!Ph@Xih$Tic3s<JgC$B_zGdm__g{R%;8-L{#Z!ZFbSMBidQgCp3PerT
zA2aTfJD>e*lWIENE-8klgQ6CJulUA)GX5YbKv!H1IDw6zdS_?HhIj71@K`EB#bc=~
zc#xq`B$!Bq0y_-$FAd;n9DRwJ;ls|o<GBS}*@6KybsSgGRSh2g)1~wek^)=@yOoND
zq_!=ZcH0$$lSlSXq?iHWXebFmd@vFqwz<Tpp6GTvdp+$lk_;1IvT<KHNe3f!r(AaL
z>&pr}D@v9t=^ASm73>T655^xP1&*qkrMPO>wl}BUG$|F1`QnjOZ7@BuJ{JcDQQ09?
zi6GE*SI||}W$Mxp)0dtC`ovh8s?O1gaAid}H#i%M`zw6$hDm2#Id8LMAw$tOmH)FX
z{vRX-r>|Ahntyohy2--^hN^OrbPh<y(2<HrG)Sn)Bt!mOPYwb+Vcm_A;$v^d?gLC9
z5(+bks)qV(AestA6PfySW&d#PF{j=5+RB!^fnn5txWWGqBdK?`@s9j^=U&<N?Q=Jr
zH7exGW@A9INQm$a+(ar{Lxa5!Fce*SvkWF2By2=3dQm#~mDoBJ&*iWgwcJ*P@V?;z
zztn@-2*z_Im{d3dk5z+nfoPJ-WoX(LsXz6`XTM#~!XBEqv!$K)cKSaF|6wFG!1Z)=
zxPx%M=o*^Mwg2?(vuB;q5b-DC6hnlwOsg`VXd3)d65Q{gm)Ffi2e7}15D`CN64f<{
zWU{uJ0VfmwaWQZga@jZ%7+FhISJ8A$)ehdmSFx#|0AjQR4PV1Y{4z!P`$Zd0IOCb0
ze%NSgmTD;^+E<MKIGX-vkrZ=|3{G)eT{cuxGxFO$d+uMy*Lx!9QDrwkt;0L2ss_7@
zVBtj2&x8j=QcO*4&9LLf9(VE?Q>UJP{*_l;e*LxA-G0Y)x8HW{b=O^Y+4&cqee&67
zjTt_ouD&*zP6F`=w~26in_~hYpflMB0TT?Do*|e7cB{sQPo8qe$J<!FD4U9*7>*-D
z?EfEy|16S%kQ%b%072EFUg-Gl*{jAiFoZU_+TD}1L@4AYHbMpbK-XYm<glS*PQT>Z
zTkpB&k*8<9{?5mXzFW0!Yg=1eYkND-b1L7_(b2hWYkT|VpI3kR_Ut*cXFl}k^jogI
z^q*sE>*`WThK^Idpr@QWZMElXx+VuiWdi*obnV!yW`DJfH5F4fG!_K)KMVh9O6oYS
z3D(}V#T9Qpcy28+H01B0WxW7~J`7S94u$<lPb!hD9e&)I=Usf)%d=knWWoBb&dzqx
z0@$P}(=besIQTcaV{u(`ZGhY<ng+jZ)q;=bJ^%E>w_G-%rlGc`CJ>Cr(0c{imv&$$
z%EH@JO%AMn7~Ft{aksv`vdvU2w}+DAX8!|>|9_ZIYH=iCSzP(S(-)442Ixe@N2tMs
zcf(W&p%U3_Jf6<RD4^%KiRVt6HE-c}E4OvXvS2%!1|)=)nLthxp188?fXC~)j$-JB
zt`;@b#f(whpk|n+uB)mdOI=N?KL7CTS$ADGWo#;*PSE~99K2TaCzt?4`MtnE5S2(&
zXGtJdJ@$ep7i`uPYX>hCpez3W&ha1TlNK$u<NKLc9#;?HJ4I)!LqS4}o$W4y&SYcZ
z{-IQ+VcdCly)bX#%C*~MRd!vZrlA*A&Cm>1Bf0gi3$HkyH+0?5bOi-l)f`2271>e2
zwct1d9!ii^*OofBZ}{f@SMQoUacp%c8a^rnl3v<uH(;MoXUjtC4>9#8ob~AUZ4gC~
z^}tr~k|582*ZBRDv?M!7B$F3abN$@A&Kwm1<Bg|35-2x-mNbk6_{dn)u0%9A^sGy!
zJ@@X<?HygnHlU|&fMW|Zbae$uVmqp)X&{@}i|K;h2Yx!L0Sh`HwW{i>$WEk^0?Wa^
zxJ4Db0UhN}Wec5KHY}Qc)1_w*VM6`FbajRz`{3;h;}3&Jx<g#SbB4I<n6vKqWE;tn
zMzgP&s_u|$@Ar(~D@nVl3BeNFIj#Zc(~^YmXIyxE2Ai8oC!PXE5KfYqn@MDI6huZo
zDo!yqW6!?k{&ySOJM#*-nWk>@oUQ@kkaEC2Acm{425?nXHP{AN|04WrAf-rDcn;ge
zu6ME{-c_P#nBGL30(he98gSJW+S;~#e&2OdM-8p6#AH<nn7|okV0XN>1hQu(#B;f1
zV9d?WFA*J;*B!V<9EGzq4m+kw+xZ>h_exSjE-@4o%}_m8)ady9whM=o(_T8~1WY7|
z!BI58#51+gfgvgu3})&k+%oeEP8JHH7X`t((6$zVU@kbDz)x7C27*es0oGn6|7d<4
zSJQ#5gvIcjj1eipTMOzdI070jtg)scNP?y*?e9K({Yl4E(^QORk^!dLm#PUym_!nG
ztXoZ)K>zfRR7IrW)Vn|3ZW)3g@FpraZ0h&iF7-PlDN+TA>e>c}<J^EJuJqNk)6)F|
z$zZIm7fI<%6_cbhxlFoWB*9cSWD@_p`I)bpb<<#*I(2j@;p<?~RcsXio2_{K!&Tx#
z_fz}_>LNc0;JJ#=dkta2m%vX#QE(IrhFZYPQKP*eL0Hu?|NhHQtxsg?Q~n63Xf91r
zxrQ|AX?b5kV!;i7#}n@UMg&6-ifoe53AgBN;`fK&DM=00L@6zrECew{L$kQ1&u$x$
zOi-~jHpuq^6Y*!O=twX{Gxbbl$Qh4*y{$tu3|lW+hUJ1KH#inFQippWp^C1k4#bD8
zpp8fN5*nBPy&Jq)dg9%c0sJgMQ5~&-f{RPIlHwTcdGKSE!qU4Z3`yov6dkOnpuz(y
z!nIZ4oqGF9*H-yLHG|S%3~EoienBT|f#)E5Bc)~Rio!K}TlyX0_o}1>5TJ-I#KMZa
zdd^i4EI}v)p;9ukb8}O;vN9M?)(;tc%*8V|3LP47ya-{uDRgkAY=ii9LBtM@C`9&}
z4BjSCmgptAqJ$X8Q}XWtpY4Gw_yr=Qu3Er`0c5Spak`@zu5G)Ts3^Q3Y<%+E<B!Sa
zlIezYJOS2zhkpXTDh(b$q9#KJ`VYTu_C`UlU`v=7O-#OQs=q(G|9kaG0S2}fNf-sM
zdj0C5{r&0cZ~&Z7-%bn#c}msR<#HoWykmZ&B7rIyJgXWIEjczs>jvtlVz4NB3Rrqz
zC{PqAOP-ZBMJU<()gDl$D42U&QGr~luBu@9!RX66TnegfE23hVg)Q&idERlgp%7J*
z_3hkz>b3t<u{cvVdDccA!zSW7!%h)D=l6!+4@t{B>1GIcda9{9jf-EuW{97TW-{Io
z5Sq^P>H?)GCXyI;?!BLH7Bti0G^3!Za#53t2G8q;;((IE8nsTfsOL2tW&$Q&Q#AwR
z6t+>?)_#IewB!+}x~mu@dcs$j-OL%PuJNF%wqk)?>n)#5zw#LD!7nvM0D6UdOoFbA
zB+q<xaid+7byox(HC*X;IPUcKLsE1?P2dnbqD}SqW3?}P^i=Gg$5ZN^Yc%-S5=l-s
z4WHzWzwqu4)<6_u78G4_bOl>LAw0sc+0}`VVpo-pBsKCtGEK%p5g?~MEd2g~P?P|7
z5v{#f3raDtIn~DQDq94)f(w8h28F`<_wPIB*y;qf^y5A=0f_f|0`{M(pYqhQ4zsA`
zO+^Jc{zFL*V|XDq(9}i+h0|_ncE|F^POHVNrEY)}(KMZ{p%N)F296$d@m;f4cL3om
zp2G?{B;G-qVgo&_)doo`G;WnF8K?=NJ<?QB4Uj!e0jpoy_<o{9QFy@WPb#<;1h65v
z2a~TF8qXWDrs%F@Xc~?h?fU8cC(dtR=vY-2E~}nGS^`Za>ZUxsOmSpOW=+u4KYY@|
z7$~Jc0F%QuB26wj((<P#4^35;P9_1$XaZu7R3tJu;bW@CU;FOHtr9q%@Et`$ht2_V
z5;8gDnGP|ywE3mm?w|S2m#dppU~HFWV)v&EPN{N$ss#*1X2Q3TtKKW%rBZ<SNplpG
z#G<S@GMN9O2Gm{p=KUw<(l}TTFS1>IYKk6t(NoK0_>RJ1yJ=}NzZ?9X0w;2I3^2s5
z38d4OKYigSw1+)TDxBp^Ek#EI*{Wc_*a=t8_+~pNC`A)}O)%m`95IRnMd+kc$pPDI
zw!SoWpl`x?7u>sCX5nRCRV7!);2ddu0Pxh4MC1g>QyGL?)O53J`F37c3Wg)<4t6K;
zI1sPU^vz3mj={kK!5B&L+X<ki^;4c+3z5DfnwZA_`@-)ja4LDFCJScUZu#h%!IjyX
zL@YovOv&Ce;aDw%Nr7B07^oh1^XJ<-R7(P-#BNL-b(k=w1bC(c)n&)v`0aDgjUE*V
zW@oazUX&Iu`snQqhKh+K2P#_Na)JYi&rnNoTWVg_1kq}F>il_KJg2!@!Njx_Nd=x_
zvU7gj`%j#iqcVeH17bUADNQr}P))XG$~%oXbng!(J$wNah128OI?M5j{LNjbRKv-m
zGlP@9T)NyaN;9zmK{}m?k2&v!wRx_yRW}Q|$q52al9|YQ%BL7f3UQb!J6gd|mfjeu
zh()j3XlWfEpEWp9{d58NjD1iC2FF1^6t@AZ!QOq>$!}F`e(|kigM(gENKaMK5jBnD
z4c7*{&k3BN{Pg<esYq2_s<)If!6-N|3>~i?d(Fo!_V2K))Nh}p$SuMzu<~-z?ra8s
zZpocvVgvk{A?aXjK)eq0vvfWaiFhg-WsbRXMJF~Xi8j*EwQa|NXba+aBqr8%E<aHf
zSK~OvE*9-=mkl_&V&tqvPhUU$&wu;J=($~>+1R@8?L^;z?N8*6v<_guQ&AT<rQ@|L
z&mJ5faP$m2&tuC1I;)P2`BUhj!j=@H^Q~(q(81mYpRUe$ujz{P#7CBPz~16Ey6=Ab
z`0bO_GxCI@cp`1bT03KMRV<dOjSh&_#Hj>CN|;iFV8WG5a>V^ViAB3W+L=(rOjaub
zL(v2E7z+30Cm0;5ECl_6^M?P3P@pP#)c^W(!?m}4ZmOWc5JvUgDTVbybmTy6?~1CT
zE1D~FR#Dyd^>Y*AXWcROe;@I$4V}Es@tlDa!>nnzW`L=t&O+R7YnvYjPdg6E3BSop
z>htasVDp(k|KnfWLWY)NuSD^8(rEfSF!yjSUDx6iS62+l7Te!EHRF%@874``!6Ri7
zbfz{8F+KWjRIL8Y87(rlLE=a}#n7;26}<%HHzAOB?>&JB>{!xSR&|uSeUa?(CsF;G
zOFwJWZSoO<EY}y1U>@wb<7h4I2Bw(kR{N^Au81Cyy=mQo<KvhA;J^-om+b8ec~lJe
zm5!|xEZug+?`NEurD`H!hOW&*WKS51py}F7zr^tuzq|!J+5%@u*6*a%^!GthxU(En
zHalDLi|-iP|LD*llFIC*iZZbfovFbQ7`dV6zPU*ibr(lABQXKkQojcP1A(E6rYe%5
zeSh*%kt>%hxbUo(n>7_XrVcEy$F-7+0Q`u8Y{`nFacv*pG$Hi2z?9X}!n4OtZvkcB
znfv5Ak_{Ikex<d`G+JIhZ*)4BhIlX;>e;-XWD@>xqVCckq~<P5E|`vtjRU`V{4UHr
zoKsX%Q)pi`YeG0&Lq&q=(rE+GsVeMlqOxOe`lQuXT2+YHkpvn@DLm|1YKT{=8miIO
zqHMXnpYPu7t@9pR)Gl|hIGhQ?9o%x?LCTP;D!36mFN!To?;D%$7d~}vSKF*JCeLmb
zumK7d!6my_ujuF_6lGo3Ek_Xrz3r`QPsq`gq3S^;hm?S#nM7rnqR)I~lc}(l%o&dO
zJ80YgT_`EKr4ajwTVI=6OZ&rvsGi;POtb>?FXN-``)r$B1RIE%cM!Ovq2D{j@MAEy
zilVpvco#GB+3nnluN!qku^j9mB}$5^&wU20670FGDTTbIta#+)tnW`pU%a?kc<=Ok
znpM8F4Fpif`lN)L9#Hj3dw<?k4Ox>dy^vq}z=>(vk0bTUVEhrxlA>!O*?)byP619!
zzl#Q`-#kePli=SLI+xrrE?vQdqd6a1dhfJB$dql!#_CSHdS0gmfx1!PR9(^(Oo%S6
zVQ+ACp657u-Th}=^`T&L%}t<7I;ZPhB+l}-+-G=7szY$pw(P0nga19ipT2*i;I1Bi
z-d0s8C^ke>pqSWuzhm`CS5QSqG8EQ;k6}Nwb?@JMQYzjv)F*>rm;^(`sFUtl(rKZZ
z{-LD5K2Svw?z+b()yA=|4V<*@jL}@CGFwAuCfxmHKCjyaj^kL#F<1`UjtBvWhq!y-
z2N2xLD&My0rP=FrL(>$k(7d*h1u51va0U(tWQ%wQBHi)AIim(=a%ViZN#=^n>j%Br
zB5<6k<aH%)U@L!l-^k*wT0nYgp3cezOUXBWcF%+~wS%S%YAKj}#vi8ZPka1_yoI?@
zzj^$oNs4D1qaS!`t-NLgjoc&yEWAVvhM|}w_@c+%`E{#nDXu0ts^n;jF1f%kFJjUa
z;+XCZ$U@gOxdO|zY-dFwuUTT#{D-c*W8Ru(-Ag~|X3>7ZTbT4TVMeK5)cN(FzxDhR
z&#VBsE@&?%{xWtkILzRXVuw9?rDb!Ek3dXEHn7kRepO{zQ{i~4eeoS*vZPlZfY~z%
z#+S*VqSDleYrf2LIQkhcNW3umxW9)De!C=v<84Zq^<*lx*80r(H8hp3p_!;J2Z9O=
zP12dPKig0>_{K#A4pqsJ@n(Xh2VOgvAwt%&$MZ%$a$UW(pyw6WWt%Lc<-12t9zruW
zek+x})+e9T(?%B=s5q)>gOob9+}YV>Xhvbx=a-H?;hI%A1{5zn{CVUz65u2G_qYZZ
zCi<wd4qk)R@$p?JVvRotnovf8s-<>QnyQ`j<Vx5<%#Sh^3;Xot-(;-++a)P56i8Rp
zrJ~C(xM>g$UBW^0bUYa!5KJ<eno$2}ZMN>BdD}TG>D_&n;rtsQpTNF-+G|Xm(`8%D
zYx0)O&4&2V!|(k(v*u6#@6a4fdd4Mio%;=37(~!jSPsH&l{Ix!Gm1J38&NmD``nvf
z{LsWg2(Rw5))N_fM=ldk<f6L%&1(l|QvN6e_sKwr5Yz)r9dqZ~ojMP$EM}{~H~X}|
zhYfziN@^6a@d<p=P0yS`M@yqZn5dtj(^WV&*`H<tr@p?H<5d&>OJBl4&M1)t-g|oh
zd<=^?m<qt^>POR8uDh*v@+$M)YC3%6qwSp61l4C?-;t)Mj;kmRg!!-$M=7Y<mK7TX
zsd+O5<jB6VmERM{zJu9^XAnjeU9;`oiHU657iDTPa7PXV0IYvy)$w<%bu`Y9K}dg_
z3<dm#eNw9cF_7wXExWNc5~}Ty*vy0ju^<zptB!xYQ$V%VM5Gj)Bf?P9lJ9w(WIcNT
z5n3?kdS2GlHK$fRcHw{5e<=NM?={an{^EjW%!vl{>AaJLF(p-1iQ5Sxsuz`lp|!N@
z^3pkPz4`3-JZX8_ueAb~fu^p^@w_Bx`Pt)ZnW~I0Qrq2b%J`!=w@=lL-)k;-29Ari
zzE3-RxZt<!lVX0eQ+Vy{P^gA3w-!>FdOA8Vkj<sDky|*+76jSWIS4n39z)3&y{~&T
z+#_HCR4mZawl?vZ;9n}@m1it<TEALXknE=IUJF#;eNxaqQpSu0;%Kg|K$N8DD)+-P
z=OoiZ|1$Vn%ql7E;XZ*ar*JRlHML_s%Rkl-_l3$1DS~vj`NPQx^V%JBNpZEmNvHI;
z?UTCN>PJUJDt9ESGvQc6Ff?Lll&-#Nh1k@pISSX=fyqwrjh#>ZKDhua#z~p2rq@Ru
z9XsNv^VV3bW~&O&T?HLP>?fT(cndo~PREf|;H*Yw6w~<jkI$gbd-TPJg4wTi4pxlS
zqV@;gj1P{?>6W2%ZhQBPhLo3hvIBHoIF%gn$R@7Ts^6FWJ#6qBB&m0<A+{H!wvVqJ
zG7w|>?lVItqP|2j5<Ttt^^&gfw$!Xx8dr+s@q7CX*bLl2=Q!@o+KNp2%=fn&MF&}n
z13Mw`>g)L~eF3=1Ha43oNU7?o1=cctxM$48Z*SD?H)_uQ5fvI!v-S&$?#QN!gZ6om
zQyr`2<x}b#!o9}o(RFFY7pk8+yII1d7;+H2U-lcqZ;+(eeD9hfNXW*AM=>;&j793Z
zTkn~ukB&!2Kf4Cvei8fu*)mwoYYg>#(fx+!3^=ac^!n(dCd~X^;CUl&Sb4?PI0$vT
z#{9mVY!VJ(mZ-7q%@@}ThTf$W3oEy3Ur&jj{iSBO!gaaFMO@1365S7A&I$M<a1SAv
zDu}MS?DoM`sUGfmf(gXn3Z|$NU)<(^tq1MJ0usM5tN+_0DV`8hG73U|`K_2PioRD5
zC5@+QYA*k@3zq~VZ|kaoZUNDdzNv?|xUOrTpO(4$t2Rf@n{Audt!)L(0>6?J+5uDD
z2cML<v5MxXe167+>04|yZ<w3j{r0;%#ty!z$uaDeV+OqqF&eIRU*UZj-aeM@Z4h9m
zZSIA&*abxn2;7lrBCxo?sA<azHjvX1v8<nG`+qC=?U9r;OS7W9>Dj5(I2qUu;5ms@
zRGt6UR&2J1Me!cac^@UT7aX&F_61M;yneeaGzs#G^UwHF%qvBeWJ`lD&?n+!pFO6M
zs)g4ZCVkEFhNgUQ(_J^+H2wS=TTEShCU{PZl_$NJeOj#th@wna;cfY+=TAw|32cdC
z{Gg)cmi>73l^+Z6dlZhdusy2uwciAOgGvhVHU!umkBp4?z}S;h0ALvG6^%2aum9E-
z1&%~Yy_BZ!!0&+)Z~AIW<J0%85?IZAv-+r$KPd182+%P}q>n516w*~WW%02U)7l+Z
zZ@w`${KALsK66HgB(_{k&+r7gkIu6P*>GG*b!6677F;(vN#^p4*H%?zakxGl9$&a_
z(o>sVRky)CaO?v-B)<-RgFdOn@?!g<yN{tLiowi#0y>-VC#iVylvmnK+0w)VYijQe
z^g10?bu@9ytZ}EzYb|y@nTu9l$%~CR?EXNq^oYfm75DS8112pr4E5cbzyIsQd#X~m
z{3NNL3^{)t4toL~@2|D*2^UN`GPGdn+}amTZAiwG!Dyl`QVyfQ>{BDAErVdv1kvTZ
zfl$8@{1yWzRu$hpFN;MV=u}PF8D$d5`rM#fzc1)cTeFIV%22w<7Y-EobxeHH47PdZ
zsLYejwmTU5$O+rzPDANY!nrTNVW29jemC}SLtf{c<_rHaa^VkC!*u+Lb<V8=pOZ`n
z`#C)>_X#NLpro3tIg%k4^dD{+6wB6UebEM-5)zRFZ-}nG;xiFLC-)B}J=n0fJT|Tl
zLnbCl#x8lFLy<G)Hu1a!Q5DOQ3=oK!yl)&hTV0)FIo;63?@mq3e0fs*=p!y@6Z2+4
z^J1nxtr(ev4e)un?UMh`z!@FZdsTmWMtwg1zq29V%iI}NUy>rwxXS%)DD8%VF}niZ
zE@&ys@gm<e??PWiLnhV`_DD(-O|8gGozvui^ZG}0e-Hqt7yhsf5vaFf+VHfOc~4?k
z91n^PSi_wwyKEUdOI=OYRa3@PCyy8h96AAIB+-E=-c*0Q+@Fhv24-dzI8Dd*aLrgK
zAD-=<s-=(7BZX&50X)ZdCysupOS<pRM}G9q*fUQI|Gj=e%lHdgB;A$`(!$?YDJ452
zvlb407Ia4{nAXMzhh$Tgk<vVn-m#*n+OdypMF$qX0+s-`(#I|R8u(3+)P-mi)6F;t
z>kakGThpOBJnRTM5%K51=EoD0<~AxtQ*;eh;8g>dN@9DUXSqAH&wy`fn$^~96gA0J
zt;P$FiY4OHn_DzTVR;<$iPdzmLis)hHY1Y(ih>o$9q<(Ln!f(aR(s>M|JSIs(;6O`
za^#>l)lbiQ*>IGC?m7;TxzAFHz?}g>m2l!bUKCs3KCvR?&kV{0Fw!q4p2TbJ-wGE&
zu_&=78S*Syhdob!3naBb6HTFLN{+$F`r^xK;@w^Kkw7#&G~;8U*{i<E+ae~Nps?w(
zhE6Hykte<f1_+$M@x)XmEpPCg^K5J&b^dBuHCkF#r>H@U?`Rs9(%r`vNU{@fNCSUa
zWi{0-w5poK$y+Y?$CS0(pFO7X)Q{BSE0-=-4IDC}dR4`}ul5DrS8%XB{A}E$^y3}1
z>2O83Izx50Bs0EH-EAv)$#z6THgHiOEb#QO#&1SRfpo6iR&0Isl&CKWCLRwu6ZIwN
zbbTaPRd;`*>YBh%qd*d!I8K72zn8T8{Q;b3LP1S6WLH<e7+KN(FQ<OmZt<PAs<<ll
zy(34xCHDevG7{W>2;Q9n>)7Tu56oRF*;o7@H|4dJlXLes>)a#PZ?nkkIrx7-Al*+$
za9Dv1?2|cxf4;_F-9MQ0#|T3aG)>i9wpf!n3p^Ral`Wv?;T}f)R(w)8@~+j@-2V2M
z0aUs!>`z90AP&gXm@ip(`n#-Vv5J{T9kn<d!sHggJ0NzAe4RY>9msc<^wT??SN<h<
z%n6AhGq>_O)<jbwPA|8{?!A~UQpZI~V}Cu+qpd^Q_Q2V%am5#pyvvn-xbE0D3}@>V
z*EWOKnm2HWUg>*%1URK&=t&3#KCNsitf>^1oR`X_d|8O|OHEPe!~0I3^`n7`D_fwr
zP7#=TSmQTBQZ)H$r}N|W$6}$Jm=6y&<~K7DA04T?XN_DmAfPu2SZ5TZ1X+dMr{I<n
z8XZ{Rgew8gfM*M?sW@*)XyTiTZ%fcOEE5DU+7i4A<b?H;iAMGT6`{E*+qP<16DM%5
zj(M<MU3T|-M)8|*$FEVi`6oXDelN%B>i%}z?*`;&!voM1o;$YOq!~XwJTzLJ@dkYn
zFuq)B;BhZCx=O)x9N@B5{6k6qeUL+F%0*?_ZRtuf9+Xaynj|=)iAs%|A=|3OimnO&
zUVo^`vp7eN4mXdU2Xq$TyPn61Q;wKV`Hr}nb$-6&=;*|mTRGieISoF>hNyk93O`3B
zRlzIe^RHg~%(7;E;mC_ND%|#FPJ20c?&c15=I}Rx$RL2Ec=Uns6gNY@i;V*=mNwNC
zj%#^pT&g0nBfB9%V-@~U_pCG&p3DhBVyD9zhc`*dVYk4&#7FJj)9WZ2<P(oH6D563
zF=p6>3$bFZY(Q9RNI0FnE?dA^Y=9C==c|0q`*`kQahxF3Br`Q_y7Ye!nXifbyJP=$
z)ZpvCYBh8TJ_Zpb%a!hTzkUpda0yvc^M?HCotK_+^U|*$dcM^uDu%G;s_KVVEc;;O
ziK~k6DJJT}!UxjjL>iWM<T%N?Xh;Q1Zd9cOXTfpu@38myykhXQFJ#A??vMCHeLFOV
z7Kb%Sfu6`v2gqmY>bz5{N-Bz?0E)^a!~KWfzf@J*9net6fPYhRY)&z|@_8(UCl#@|
z)Ejn(4C<5BCyWF2!_r|u(vRuC+{*zo*I)jRRLy|#PyH;gro{qB9WeW#r<hW|2iFwE
z7*^)sJUEJG73TeO&?&df+sF%+s_R>yym|4myGQ)(pW7Tm(G-oC-aW8=AMk#pcOz+h
zQo)pUMc{wFv$iUQv#6GMipQ`rch%>eHugS&kiznZvzi{(Bn5gDfuVLmGIe3rv87}Z
z^hNOzEPiz4Gn*Y=c3d5TC7k}w1=+J2=fCj8todss;4MVuis3!!YjdfWFX|W^1j$M<
zf5{OSuCSb<!N279XOF0;z3TOE3%X{2+1DIb0goBy8Qs(RRfz93R(Bx!gH^CApE>`e
zGavYDGiT{6yM0|}<IFn$-FXmw&{CWe_ZI<pra)7p=!llqIBNn!)n?*#Ij><KK*ejO
zeAFdyHpEY`2rQ1wE`9ef!{JO)OSZ+L&B=CGUKd|DE?3TYD!W(l>_s0ps_6EcvZH})
z3OK_}`;r^#YwN4V-teu&s&;Fet4jvf8SN880p9YuqUB{(Xn&~xn9n7vpc(476T|*M
z<zFK;lOFwKeIENq!8e5mZ(8Qv1L#XM5)8kpX*}P&^5Z*>A9eM+d4~gztK8=29$u{K
zSSX*&l2_jNzX@0e++-cx3huornRv1?H*81VK2sSv=fh6d6|K_TPoSlTshl3(N@@sh
z(RS>X4Xv}rWp<j%j*g|TSOyvg0TOTkXlk;&(D?Q2Yla`;X9iHd)H5AI`?A;Pn|1;8
z)Enf~_fwCJq^+G@j~>O$>B4$py3;ay!jY9jQ}Lj$>e$onocrZ=Ts3B!I5=1C9)Z10
zxT6l(be8A&&99sh%iPms@NMm;&gJt31ItS1OPx&z=9KnE>BnMksls#1Cet;c9eydz
zFcd{icw5y?hcxMv<@<ho4har(pVU-LFyzkW>0@ekq>dtJDxSV}Jx}`P$#@b?HbqyQ
zf6ef0w11557YzDu>abqO);=o2M+edpMHmgNQhBLqc5VFCPD9f`sU5xXql+`aqkJ)6
zzgT+6F=L<FVqzvUw(RfiBEXDicu;g#V?j^z7W>`(^#3`2VW%O>ps^^ms_saob?j|T
z`wK6L1b%^7FeFY~djIe+<<FJVOUV;z^ekTEELXN<^h*y{9M&X-E88eYwz%!F5j$E?
z;0S{rgaae*Z3RniVxkkAQWGmhScTOez5UFc(_g&yn0RF59RA0D{(obixCJb%Lv(^z
z()|VTNj1&ljQnG@<K7gDjv^N|u8lMKAMdOkc;u0_7d`d-+z(z_Yik^@Y59FLmX<z+
z2gkA<a6WZ@+p{%)I`;VuN@o{%z9#nDV?DiJgrEpm58M>alvsf)Se7`u+E<e;9YX|N
zlZuUee5>OM2KY>9>VKgv|I#=-NeTxGj~Cc_PIT?{w`aq-9mbx<shH^F=jx)!vsl9s
zC$<O2z%@nNQYA|_&5gHH4bKQIS5+K+m0+<T#Xvc?kHbg^SHL`*Td%0P<_pf7P+aYj
zZLMIk-#&ZytOcu^6^$!^EW&&6h&xSqvFok-A5m!UXfqsHQ-!9PssA%P^($Go;c_VG
zs;Ocvy<fuAd%%o5*Xd|CY%y<dYc<|DF<A<n==xANnefLF@sW==s+<Yls15rEM8zE~
zef_Y(VM$Wb^aSb%>SY&=MOWA2JO<?8dI*Q=PWi-8ZAS;66_%4QrNop?j^||8GE`Q*
zH+pj?_xA9=AN#f0f%Afy7~PjyvVVZ3(bw(z`kIqx7A!sw9;bm*jq#jX)KyE&s}Aup
zvDgEvDcJCju@iPH0dEz&$B>G=#5_@~4MDgnDYgjS0c!8o$Ls&^qeox-<a<9Ckm;(d
z;q0Th(7#U<SOq!{pxhG6aUk|mtM<cWxV-5kT<=L=q&^%5gI_)7kq+#{2Z9ole>jFu
zho_Q)UDko9q7CA;<*xE9#h%Hh6S0Bm3%=4?IbJ5l9?iH(8uM*~<%EJFw`{rh@`Y^s
zy#sQO2zH)>zzGe#ciV5^B;Rl$Y<~WVcTA~ORTS8#yw}d87j#oA>O80Ex=<*ps>X7v
zYs-p_lZ<2k6mr#Z9W0!O?F!(sV%HU^VS+5ij1#rM>2~|$nZF<1KQ{8rJLfHLT=VI|
zHL^x@^cPT5Y}99Ml>>2<@`mGpUQ5oVsnx-7Z5j?Go$#)Ern2g;%_56EPh_k?ACJ-@
z!Qn|#OF|`eUAVXxUszWf6Pk=wV&Tzfuzt`*iv&%R3|7IIsYsIe!0m*8h-(Up)B5RC
z3w87B6aRF5htl%pyWbn|32IXx11Al>Ww0Ao8&apu<-tdF9i2?N;L56z*EI#$skk=C
zB=Az_1&G--EJOx0#1pRR25aalIGV^%h4i1ocd=F(coDD;r>oU%eRgzzAMFpuk2`bh
z36pPnmnGNJFQTNdDNG-6A$S+zKuWCJrkKmF9K!hG_0b)1e!Ob*T`O4uf;Wk^J%S!?
zEImw>)WBYN5d!xm|4IWj5p*_J!7!0<ZLHzWjiet5>ymf}73qnOrecYX<M4}?2+GDO
z|8?OewQcU{V_sGioDZkZr=FDAg)<FKr!J#`21+_A$LU4^o?;lKfva1{lVSLd4RB<f
zAlD1_fssI9S0)q%-r^J312u1A_a|xI&uco@_3DYKIOY5Ee+7nL_VQOhb=aPi{sM4;
zbRehL^+}Am=73;2Ez<{uYqC}89qv=oH}tLzPEj(jC^Rwkhnu}WJbhBx5^RnW9bw72
z@idw3#slU?B*#CyombJ&1LeFM+<B4;VNO9)I_#FV;?{?f>36K=r|XYAZoWk3U&aOc
z03;c(+7JX4vs6Ko)B>yP3akk{M-?1h>=DyARx`T5A>}wh<V`k@1Sa+F!4=R!Od+6B
z;0e$ebEbfsT{<|Z*!QFs;B~<kJ~=O)4Aflq)F)dCI;`;*Q&O2VT+vhuq)-|B6a}71
zg48m7Op*?i<|ieGmZGa4`pH1Wh2`+}?{EB<#o?)>u9X*E2vL??l?qmsMlCauNOfhV
zhCboh4H73~Ge4$Ad09{J4RUVbwI*44#o)ySV<Pu#FT7Is=Ogau;vwGYvyfsF6tGNo
zY#^1bX(kJNgqOG;P)d+-Af^tFj@HbZy2<G-$ZeNmBO7r`KuBGc9R(??pbUeM!najm
zw^lIVEi48FyREKz__DiZuP4>0l~T9pFX5Bw;8GYOPOO6tCmMYqqAuDUGf$vm-i43l
zBj7$!IqcpaFbSF{=wTN(9j+v`3!-bY-<}7TI#VuF;^jD>Fsn&&utyz(b%?%6=ma1w
z`i^U1T4`%%aov^S$s3BFjT{iFebLq>5U9Rpl@cL!G}Qzq>1xr@@(NgQ42(4JE)34J
zcqFN2a;&Ou-O}E=ZRIx`1rDPrcHPRAKWtdHcAKu{4OLZj9Rd@L10#>^RFZ-Tc|g#j
zmNzxSV9k!7oAY{6Gc=x6J<0ziloTwyp)cMDKLQwQ$vU_moS_tKwQ25gNK)jf_fOX(
zM@(O1VtI4e?!Uk3UmAy}xtAbj5<0)QYj}*x<a{_&16=RPh(A?(>f8>yfSxDzgu<fT
zXA_^mJGy9bLTg@lV))q~xzb(1VE<9?s^F4#&lHZ#_P((vkcIF8qI(j}qgoariqe5Y
z5cR;d0HSEiUE6=!hQ;(1O<8#3l#9+iZQ{fSn!zbx)g?Eae)3rtoPX|}KPm-X=eVE0
zTHe^WO_m|92Qn%g#8=>&>bfI=yXr_BXXrdoSXOO?wZJDW?ZN&67sj83Z1j9{@}6kG
zvAA`fqmqG4I9&;Z_=aT4X2)0O#N$j7u6ZCS<BP#lYSi?fIRV>C$;u8d9F`<?bqREI
z>DBcN&Jjw}=~}qn)7emU{M_#gvQ>aXjZ5uvEP>h(@Qaqx`ob||-pd!4j2sxPIDJ`>
zQw<hk9?aDPv%jwfQ*3#HCD7D^^qGpQ7j(>`1OD-Xp(9y^mT%|1@Yr?dT>f5{sjF7|
z6E*P!P0^v#*ScWh-MNE_IgN~&4<3N({5190v1i|K%R}#MwX`lnE10Tj7F5S{;36=9
z!z?-y5Z3^EbsIAS@uDj2%>F<D4}l~kvL;TZi|J2j;-vy7uNiV>Bba*PaEh=s2eWP!
zQ(JK6z|@dXGzkPHMD<7Mku#fJ*|ag+<d48<KVV6Qz^=G!M4Vje2__n@OGN$27<JBf
zSmwzA4)08}(i@Iz=2=<zcx?4tS<1iB5UBM}TrPIzWt)_RCQUNVKDwSzlCZsrWKQ9P
zc*xkiVG1(K8#qc-bBvb;Wpn9N(06C62@XO2rG^N$lKaouVq+<@4@QLWWoXh*#R7N=
zt44?EKqeeG;awdQpS6zFU2>-*Z<B3DF%(sT7$3w`W5FN;Pff41?%pq?Uck!_3yw>g
z!om(ph9lsJQ@LPwJuzVPR}O?vK+hsbD~`{TOh@ARpH2@4B5-U=fCC*L{ls=hVQp8y
z0P3*9;YLz8K5$dotu=RckF96CSsSR(>FWf^#=?)KV7>>uHz05ln}tPFMxFm2P;1+D
ze+x%~cWs>Y$V!o9>T_z*z@ia*A^^NkZeH0iFz^P1OFF)Sz&zigetq|hpDaysb@9n$
zT{@D@4Y)=C`){?saEvbqyBIiki>>OM@yW3R;U%4zyw*__t@ue@xH?Y9gJb4tt|J=4
z{S&T#_Twh_UY@r}la_4>=tRH{fc?i^0Q4nvLMg>070JK}#6?TRA@$&3TC>sv2Y-Mc
z<FF^-DzX`#=FMV9$K*tsrfPaLMNy+3Z^Gb8JnX8Q99|@Ky<B|96uRC%sT6_2We$F6
zB3(tDw*lf6-pgIynP%RZbsf3g;hsPH!Od0|=PnxZx5587_O)jN$qy8c(@YGXU^$XG
zvKx>|bI1)Fh+I9d6m1*70X$OmpJrWs{v<m4Bx|ZhEB|&K6%M3Q6}LDR$C=K;(-};B
z51qcgX!2I^ixWb`+@G@nVk6C6JZvD2Y)xMH6)TD@GJk0$Q+vXs3+`EFyLlG0T5D_E
zDuQEB)L2wgAQJw0gxVkAJ}R8)y1XVCDu^jq46yZ<`{U#Re>vrAjKpwVG8%g%t5mS%
zeDRA5GjZR*c&X`xrt6ZiQS%D$A*%4<g~N)ZI93D>sLg5G>qEOskpW5Ts6;q)#l~V+
zyQ}iB;GPbn2_cTv^z9G4&5OpCYmdx@|HqVv#t#_3#KNLZylR4=lK5hGASRwrLnk4-
zrdS>A21su~HTZd>(!u{3%7t!bG5E1xO;skSTyDVC5~1}cr-Z5MYC3rK&q`Y}r@lK9
zEonS=%XYY^B=PO~iclmRWNz7NT9$1sJw6<-q!a(hy=Ay^-q1xscWfDl&7lX&@w|zR
zswk!VsivTmiiVlZ&0V}L!75l=Web+M`pQB6zrF~<hy+EVdWE6E9dO!YZQc6IV&#cR
z>AG+<<%`su_71qroO8Ho>EWfMND=%KUppn#Gsl^ssPrXEB~=rx_BJ5sF4n$GTkyKt
z+HJW_YW`^Y_4WO024#mgL~8D9vzu3cvxWsh1g4f+R(Amsl+4D6G}gLQ6SP-tYSdj>
zG2|!zG_av67pS<UP2{wq^YaCMEZF;xe}31x_KQ{D-gQhRn)~>$N#|UA?!+mRDAKk|
zjk)pZ4_2=I>AqYnUX_g`ZZ8zuI$ZsoW252z(LmsWuUTGm)WW78a8gJEGpZqE!VG?F
zFd!4-?a?vm1y~Z8=Spz#IBu)tih0v<L@WO!^Y`G@n^ew1^(?j2%O)^Y;dI#)oQ;!f
zQah|YHt`KS@oSth_Yacp0pu83vMqz0n)61Jscg$&VgqwG{>))q%1YRyzU$d5w?|1$
zk`3{_@%5zs0|JprAKg-MWb~Rft#3>|>9k9pTMx05?UJSe5AX%DB5!RN_B)zpXb@O|
zU#hF^i{5$iW6Om2+gVf}i&aE}fs1~SHJ5MOc69(qn64dh_LMQXA*1WS0Sy98lfk1R
z;iE!uP2q0SiBvpLKWtcCbt0G!A6fg*I<7-e%qJ`8Y$itqZrEn41<U$jO7`68Uo~NB
zejb$0ftbIDsZ}yIMS+%<_i=adK0qY}*=1RN&336#as@-l+xo`ogO3Vcu*C!?6QU-k
zR7lSiA$Zdr$rM>Z{_e80*9gt{av3_2^dFr*?=zCC1?%;0hXoE3l9ruO9Ea@K^T%bd
z`e-+3X3+Uxh#V^$GG>w>A$QhNa=PG6*9Hp<j_aoTng5PurYkd!^k2MEe!lv@gR#H)
zZ(OCrA|N~-%1$Ug65cyH7SsVJRE8&;znD91VkUI$Mhk*wY4h2kcqW(4G|c2SFZuHI
zJ4VOg)J6tI!v1W=cXSYU5}P6dKAMVCiC~DB0h)?M;y!;MSxrIoTs8Lm`=5UEov-d^
z=ujj{(~n}c9!Gk2P(LO;X!3pUu|-iWsywIRBsw78Xz@Ya_iXlo!=!r(H5a^SncMQf
zjI9EMO_sb}+x}Sgf5h(IEJLs)8s%A>0Fnu^Yf9E8&UTcaE+<_j45?a9*QRR1{VrT?
zD<*0w+zywfsKbM#C1(_bl;?!PM-$_4zVRI6nTE%B>XPLaCu6NASc(v`bjW)HiDt7R
z4!x6CPw^iezIxq)>fozJ#|PJ)F}!j>zvN6(bFTblq$WNGZFY2Qn^bj0Qw`N%IYV7B
zZCENrM`CAuVe1@cwcivDB(YQB!e=j=Fd-8n71L;nrl@c@QOYG`f=E^zH|!-9BH-L(
zRvdZ8gkqWc<4+q-3Ry+QEGURYh3956R4zsZDlXV+bF40_8ZYpeL5-8xfuo6$l-CQj
z3zSlbgH%O#Q)z#2>^&PrES=A7oId2py4%(kB~IW>GIJF93ORh3dMkrevyQ&-!c2^z
z>(bu7GE8kEe)&>x50IpXPqn?nMM=E}L$p|K@l+5VJZ3R6bQuIHJ$0dkeLuRmW6@oK
zXAFc=6N@-$IQ!g~m_Ie^r~A)&v6B<jb$8Xo{#yC0LsF+o?W85H0$ITmlre!+)-_!-
zx*+^=KN(R`nFs~rgP!Mk-6+VjGf~_ru;Bfv>dH6?>j)KTnhANr*@K$|sIBBy0*Wk^
z2oNSGBID+)ZQ=x>@E}u71sVE?t6C&ofv`!J)M8hQ7wuq^DUc8%sq((|g!dkhm|VEB
zrSL1y&EBwZeFw)byKiX!hKGMJHHh_jh|f%g%qfnG7m>2qX$l_{?e$j#GLcfY385(!
zzjC7jpWtc_7aUH36Ba9RxaC)u;(9t=3sxvvnQl05zG4-y8W$$-?IG+zf}-f)w7B5A
z3fzh-lCi&?@%ax|jD4j|migw_FZ|~%KNzkCLb{VMVRIAqPr3$cn2qgbUNt#=*~B0d
zpi%?sA8rzw*Sz=mB{34oF~Q12H~^76J`Z7h)Qg7%_!9Lq7)+5{nF0WKyBttgPyE--
zZ~VA*(WEc~QX3sS7vx)YoGqU%`ccyJz)PZ~uy=69kea1?U=8S~H!+^0a2<1w37vTJ
ztdC}#m-t7`gKJm^q99Cu!daS2i$f3vbjILF<ycD-zP~6P@812oKxB}A$dlV0kXl<J
zzQo~x!$?WdCpE>!hj*00PhtflrgqXofm6`dV<p`^U*M!S9}`T_L$hu9%@O@W!#`R*
z{;yTf6{IFZUb47T!m-UgI-G#jurx?0e8BOBAqn3+_V^}Cfu-bWvAQ6gu1KH$=KR~w
z9A68J^h|s_S4DaBBy40tJxPcFtIsf@a0ti;`;R}N8%&^7!Jg2yrn=#TYi}4DOVBk{
z{Z3lPb9%w<dSt@s*S@@-Gre&n=#UmHAR=5>yY)=*+5(PZ$j%xv+ji^U{~8}NJRFPn
zt9fecwhimoE0PYZ>?Ry6kF=Kz$#fi{V6}HHy)Z$Q_c7^<)~Dm6o@yo?SclE*4~LCU
z3UF;+X?kK*cP&jiQ+0H>hC1~##TIQtAn^fEbdM}W&Jp*`P;Fk2S|6-DV%&n~V?(Nk
zJ+fMdc#0RXK_5KP?uJq__6^hfu<71p*U!D{wCwPgB(VOP{mJMcTrM<SHF9K4yei?N
zv7j`Z$7ugZD&+U}5EzF}Cq>2LR6HJwMx(J<EDrzhG?vc5yI|jesNT|mvW^0i!@(e3
zm1R=+A>$9<DYmz}qPTJ@EHQHG?4NN01{K?(z|n!ws5JIz7l12ZfVTpa(q)e0yWXs?
z@KuB(F<-w(!x`sYHuZuRe$>Fei{%!2^hb%BnvP`2d9XO*d!uvVQlfsmF5{~Tg-@Q{
z3huz+GyB6~qojC1Tt|ItboUI5aPDH2bad#eg5sDw2Pyz^K#ji&WX5VG-dA=D{1B-M
zYZ8lXH$<;qbS8b<Q{nV8g4~V-)*y0ndUV~xqPkSnz?)!=*5@xcrm8AJUAW3I@>;Ru
z^8do2=|HPM03R{_g@;@fjtcas20kWGF&+p-21Mb}9}EUy`CK*^2lulo?t`Ez7+`!p
zY&1vv4|ENdrIfs`sUk5K6Y$sFvQ5{mR`%mDVLDb7Wv<?8Yl?#Hrn**uI|rm~cbtJA
zg>~{VYF9K-Q`+u{BpZVLsIjLW7pn}0kN#8Wi4Kvsv4DZMyoAkME;dQ!El1=*e!2Ol
z#!4ZozdD0Ijy~meRW%NuvZaTMPijDX@4D>!<EvA3KnR@B1st_dZFuO*1#l5Dlpz|4
z<w{Cl*%hz`_z;K+Qpzfpu<fHyZ~q_TK6-D+5u@J0L3$cy*poIVvNWnL8hQxp%`Uw#
zeF%tH*q1o%+oH*u`8S992Kg8!5`y3zPaJ#(@ezY}NI5DVjfX;+C{;aZ+}TsFzW0%*
zpM3t+S6_MU)p>8f`r50nzWUk=Pd@qRHRqpo>Is8re>_RycOYhpg9Au14*|7t;wd2e
zX?o~oZ~auT3)7OZWT=10KXJXIs|Bx$KB+r~ZYlYF-mUZiP5`JP4x_WtC*`HBmqw{0
zDo=X+%Vi5+cx={;>6iWMbq*vEsYwnIiUGEm$`(mGu?3d1^jC+Ja!qQ(k@_^9ps351
z*@}tj0Hv?|x;RWoYJkJZs`k>Wsz8aVE7P^fI1q#e9erKxf^Z6s69_3_uLsN%Sv(z%
zCYl_}+dq%{^JxnnO@)rQNU{o&!D1y7(t8gJnH;A$IItK;b_lJ_kJZFzpkuV|S+05Q
z!ha2`qLVZo&*YeB5aM^Z%VsIyW}FTj8GtB1S2gOwJMMkp)kTY!Z)@WX?D2FEu4BU!
z+p#QD5csxj%NNgo^Qr63I_=bPHQ*Nn`h}@n6^JMhGa0~^EKptwyzfv&?4P%Oy6U{6
zgEg5r)Bn<T9d=9M1xM8t1p*+|Kte*qLvHY#Jb*9ikZMmVFL1i6wmw*0@u%n&UqJwG
zo2Dh?J32avc_+uG1ag3ofUeB5d8=JEpR0|)rNSiO2a^kks=aoji!+je=L#bIYx|~$
zi%;rWV9$S?R+otN3)Ls0;lZ`>Y&M)q&bDOB#8%YORlQeqs|A)MSB!$0e<&Ti_KklI
zOb6>f<}DnMjO#IEFI&%b44&6)PBXz@Z4=g<o{kM1o~%q<GWWLO{r%Nwog)J)Lsf(6
z0G-WdYC`?Pk)wl5DpNQ9oXZ~kaCKhMbi?X)aCd=$bx8r5YAQ1K^P(57Ip>5Ej~h&d
zs2Fe|5MromxDdjXmEr2anQTL-zmFMyOd4DIL!-V^S`Ay~^FMcV$!<IN<Cyefqosx)
zfGzy+Q}6-fDI0tNSU~3a^>;J>2v*(QY8Mqj6kUz5Yq!S^)zk&~mGYLnMz$WzR;Q`D
z9Vt-*20w`zQAJk-rvB=_>0zRz5Z1eH(-R}&i3H9VnxK+k@R?ZcBP~3~i$E#R(fvPE
z5DVF4?A-EW|9i;uTi>4&{om=iyi-s>k1*{S3q@d}QxOEuWtV)HcVy9)+Fu<shzTTV
zDp{3GrTw8~IKa@E9F>d)F<c5Vbh=^4ac5up(A(c^DaeMS+pb{SqCkRfRl_D$?+<P`
zF0|;XhEXtdpsFbXT^v)9#P*MFxZ<J_nVP{l2&#apR8>P=ZJGh0jtAmHz}bMesD?kE
z1MxNA(k{PIfBgGOq0Q1DI)eD!L;=NY5o}u`f(okza+>f`<C<QXcywjus5gXSS0_*E
zOu{}$CU)C<eWE>pR2Ny^as}3IQQX#tYvID#5e!uH8}Ukq0}@I`kp7{h1U4rY+h0GS
z0uE5RE?SAHW3kF)-Bq39Mi4$S2;VdJ`vXuKi^FCl_IbuPX1-cj{=k_JZNQ;QvW_LA
zun`7$tg99;6;|Gn7_*3N!5XBiPs%bCwdwwW7}#G5`wQtrB*f5hsw$hRq6VFD(WQ^C
zYA+a;TeL+B$3CH5w>8yKaQtWY4enels0$t>IFg#CsW!TIx+u#<oVHGG{`~d}&Kgq%
zr0|DBv1mmMCtFBS!TKy62s25#s`fn#LZ-a(?a6c`bKQGgs*1z*G~99ECR`Qp5`QL=
z4OA4T2v^xJZ_fVDez~*WRt(O7iwD>Yo~OccEClpq*`9#UIIdtogs&GJXOq13ie%5$
zGC2Gve^7uQgPYm2tvEzDT$EH-j8D(OP#2C96A3oNe2F-H_V*&nK<NnYr;?Jg`L32%
z472t8egnp@#Yy5;{6r=lGk9$3XK@aE9ZT=#x6QktijJSRPIpw*ZhazkWW0{bGJf#S
zAYjVXghD`{fUmli8FIz_4}8!8A+ljQ0*EJiojNao*9o2=R^A~cb-Wwg;ckGsdZX`E
zjt8N`F?N=#XhlnORGwckd)l>AM(3zF1#U0)o`Z7{3IjW+TFO7<#!s7Z48l7jLRG0F
zsJp&wwp<BElY$S5YbH-*@KCXjUoV0?QCRl$#N>cLV#rGhR@aevjzez3!Z=?O88O-m
zgtu5T!j)~A)n$IoWp(wz?vfF7T`)Xl3H&l}OMm_5(!+zKHZQvyuZ&h;S7@v<n+r9-
z$;wRrl$Wt_zO149$5QNgDq37Yv4wYYNA$l@=XedP@Sw_(K6=T~SXROEOfBzUMh_ew
zKkC@GU7qh;^8A!2HZny>x=5yq4hO5iC@1JaXI%5dds}ov;Z4%d59@*8i`6lrhUXd%
zOF`<CZ{8#9uc|sKn0qX@qo}rQ7%Ubl1L4yINjHSX?`GdK`PgCAzGPLLN(Mu<IbWDz
z0+D1zmLB)yXWKiMT*&xh5nol{(pFb403+caI0+6ZDj_CH$r`vXrqaA*#+1ZSN2MlR
zb?cl)*cfOe4!Xm`fjd(Kn(ic}q^%e}a*8<K1gI`7m{hK22Jpob@tQlff;R+T{7+X>
z6a{jy+@{BBNZJ?)d^1&*bo#6}n-$j(v5*N_cCQVjJAnLfi<)Fft4{v!c=kz6vT@vq
z&S@YisIdwVUsr(3#h)$>r|39%W0!4i`g+z`wREOB5eDCrv@~VI15-6jV(8>)uPtr^
zYifbxDFg3F$|(-Z1ET7N3h%1r2x$kra=KUu3fZ9nv2cJMoN;grRNa+TTjQ;wV+lWh
z_~g|m46E@4YcOI$I|7R8OJ#l3s9QdGYIu~Y&QOB}T-vCx*ditAML7>d1?RMgDfD2>
z!A)&i^5nTS6-NYXFQ2<@<->1xIF4Ab6_Z>`o`kxRRca^gMoC)1AppIF<z@EMbCad!
zQrrpnuX}vGBXf@XAMTUdk_$IWEEdJ5j~SS$Lt9U#4~tbKj+@yeIY1@+lz^zyegU{>
zC{Ejq@SnrO-vMJR+u(zObw`ro6E&~6n&>*4Zf2N>KUp0gJahWV^)yAJ2b!(HQ$(jT
z@knU&CHH-_vfTngVPXmy>IC=L{ZIC#`<-3(%zFo|J=%VFRaywkMz~lPN(Bc~GYy{8
zyVfsx?y5<nYkYnjfs#lh83-8ziH0<F?6?e-Ncj>CL*4`4@m#^tyTGs)EbM-gIi9yv
zOD!}le)8PPzYUDmp8L`&S?|2(hBe>|pf`#9gA2vm)6yR8K<PDnfc3}=mdLfv9o2It
zokSvj(rgJl3a?oCuZdq-Ny(TxaMFaYC&Vcxj+2N*d<lxC62tFolx<D|Q?t|2-4~D%
zXoNNM%PIb&&e(u8%t$2(@}Vy2swoJn$#URNzB3{U!E|-XH@M1=Dn`XY$FKpPjz<PG
zTzmf;8@g0oGfBk;te)d}$34&Y&gZ@t@Wb8<c(33dKe1hv<*cH@<+m+)=H~GMW>6ZO
zi)vDW8Ms=BeFq6TG_dlKPD?LnN?zt0HLbv6D=Ek77T?v;zUI|iPK^Ea?}ILQ@TKp%
zc!8HcIR9x;C({LE8&$Vr?gcnF&}r@DIPQgpTqx?vC_%-U3k~%!=p6mEeNs$pa&3Oa
zzbFbjUPw<U6Z27-+qUy!(FPMwPG}#1UI98;uGKRCwo5*eS=G!#sE;3kfd;<nirukM
z*F{^C*ItyYiv+L>iOSSaICeS`fW1g%LxJOOe|=+@XlgnvUvB;<kG~9HZNNKp_`r&r
zJm0kG+gC2AO;S`Mh*K{l8EoyR>DpLeMDp~_a+~gScK&qx&C3jpGjOPgp)`H_>3dII
zb=L6=HTuH)UtYLfhd?TCs9VlIX}y7S+v*TqnVx`qm&WcuK7cRUt&&vW*WQwg(!T0)
zPktg!-Mp>O$NwPU*C#2)MO*KzMvmfvA$=b4<kSTk=H8O(kNe&BM1bc71k*7%w$Rxo
z0z<*81v`(_wOTt3Rw^ty`x?b&1x{1m8lOg2n9hPT8c0&<9G%38AgeE)v1C(cQDln-
z$DxIGaPr2zp6)(E(Tm{k@rG8=CDnmw2vl|3r?;P26_3VfI$o9Y)GS%a)c@B*or+el
z3+o=J&7Au1R##_Da25*lpMU<j$8Ngy(dRx|)zqcQoM0;jP($^u^b0==js#+g%~nK5
z;eHQ0PP*sufH@M&Hy75P0|8S#I2k2SwU3I1E&qqNliHHWiN%iBhWt5Q@68yBhbEmo
z@#j1+t56`t?)yp%WTb1zSVrH%AtxlVhxc8Bm32XCeC7E5HJ=tmTjh!?#-&m|2ACO+
z`f|0Afx&D;GB*0QPq*d4xEF03Tl-=2VCrR33G$aADXaw?2UW9ewV-P(&#Sh|8bz+H
zYuz)a*9F4C0rApY=}aJ+h-S`PRgiS){Xy|qh(7OoERm`TZHpIeThk`A$`U6EKt-TB
z2&ruH?^TUnj@JPEW;FR^TYI$n-p?V<0k1o*GXIpCK(t}UEF}YnzXw|*3;l}twUyM8
z!8Cs~DTB@UplI+9#y*&y<M6W0vY-cL*MGl3l+?wH6|n8P*luAG9uJ{4aM#f}{)Y?c
zL@II2I;Fj%<GWj{!UMuVsy3EEDNko0a$s)S$ab_D5P9jm*wL!vnZz?q2uezp{@d^h
zJ_AiLt65QOr_j!8is2}tAaS2uJ3htKrNP$2!Jw*hflw~X+_hETaB(h@%0yx#?%M!9
zq-{Z{1J0>x;}0h7`>cUoQu^1U{$G!am^4MIi(}YBtbS?JJA=IM7A4k_IZNW44~F-P
zVv#5hG)+zXjO*L+KM44>N$LU_R$Ud2j*3;L>xh=piCFr6t<$k_t_C=MHrF?EPkd4t
zd%4Ym<mjxx;+%M}9bI%yq3y-`M0I#*@^90$4VxdNd_$<3!TlP-!Dx_4ga-_u&Y0aS
zf(qH1qO&SeK~+gNf9XW+INA7&yy||kJHQ7!KI>i*-vx@QsM5f&f&r4PU<rO(?P!#g
z#n)9;MiTxo#7+JfQx%USB2{NEeq#vu0igkr=$Lt(5cBaST)kZ#*iLV_4*4$_m}>ID
zd4E0XF+t}!GAJ5T^2tci9lOyBrF1+@gLc|FtfF&k6^3gi_^OhzTegs0CbwS+zbZ+A
zfvn`Vy*w&hnM%+x5@1IvllA{<0!zr1i#i^Vl43a>xB@sJ8g_xg>C#!85$c=cf{|dV
zGF*TBNAH~CPci-JY><R(6dlhrOnPaZ?r1LNLhFj-V*UxPY<CKZ?~r$Ko@TFEbQh2o
zsvV!9fZ{)S4p##FV>SU!0s#RU$ST-;+bZgsxb&fubMZKcDOD3cdT=_B^hM}e91xI9
zRZ-V&5?LJ!lxeJLVLm=)NMVP*%$hRmt{dMkJ#t2~qwqX$De#U>MvCs$-UC-%$+B%%
za@r;Z>Qdm8B2mGt_`;)~X@R9g8%O5-O8E6i3JjHiq^kPnM2uuXLgCOP<Gy(GpDU0T
zUTcrHY~Mh9Qo=1z2$KiFGFc#<F6*4K`Px)<oQYG(?C7DfD8s}Vh7LzD1wUSY!OLsm
zn|hv9$i=T?cZ;`5va((vpX0mTPj-chHKlfZhWk{0Rr*%x4Np>u^N4oaDk{q#otUGs
z1B$6gqpng>i0`OlLKW1>t1Uw==&mbZF20VvP`H<5-C*zat<(M-iB&(nrmavgBo)Wm
zlTDDjJpxw%Tjp7=xainiIGRXcN>p76uan82>#8k+)A}DrQWqGC$HlQ$ToD5)^FXKS
zs6f^GE;c+NykRAM136peGXf%N>n!jS3+&*`qvHA-(l}KaoN+1?gzzL+LnXpNIy>Z_
zFK)FAUN360rs|r3{Yrg6iDkRo$iMeFpsb_403Tuba_@Q3l{JX<i&|m%qo)kY;21eS
zIEmijVM5{P=+8Brn2hIz4fwmj#Dh~$z6)9kbanN&Bf-d@f)nq1Yh@?btdY_sxZ=H`
zw20#ftl%sdod&zcz_DR`5hg*^O#2aMFkyeaKKx&aq(D(}U~FmoJyjLuNtz&LVxkQ*
zc^PX(0KDetKEm@!OD}ZU;uI`SiUMgCG(*>39$ig_DA5Tj3ksTy1QV(HkyBs%8B>&O
zbVgxm_#>n8_ZyyF-|1HR0D+xP@Pl?-K05%+Ex!*AfX1<;8idNT%O5>C8;>(|)zDaZ
z%yez+J{E|n==x6|{@BuG%PeQeSUwuyx{{%>MzQJoSa{T^6HlLV|FVJ(z9z;^d$RT=
zkSY)m<6GdF%KVE%qV=IDrs&g^ps5wZpK6Ac3%{C>{(4I4-Q2c0C&yA%<;EvWE4j5p
zZ$p-NRZYrl+kfB*@1}w-xhjaM#!2fIby*^~aze)wNvf2UM!F&6$z1k~m)CI|1oQ^S
zYY@pHM`1mCYq|F**Xw`R8@^2*NSs8}vRgBEfGmOWIV|hAj*Y`Rb+uq>e9Qd%#>A>K
z5IuQUF<qSv-muZtSY7X$HsqX-RD;u8C0`y1ij_%4)>79#dF#wYKYstoM?b=9EL`0l
z<E9e4eFV8QM7y<%pO&u8czv{VWsIRJPWseRt-Y_|pyJo_NkL34#aVhDc0`tx6wH0~
zr0*nI$D9(NC>)jpIWlFR6oAIiWX1gU1vmZcb(!aQL2P~iHpgH;lefvBf5MFAoMz>*
zdWmGa5M*O+G_nj(TIDVv-SXgadfoG|9Nx%@?sZ@4xi4y1FbCl|ySoD4Mg25%&_YFZ
z6;;hcWYqfk%_qeCB=ZKqg!?5<x#I_(71eh}B$;u~cR;8m=Yij3+puy7?`4kP-l%}a
z7Jyh1kaS-R-2vB;x2sY?{QiPOO~{*b953d0^&MLs@mFlq|Mf@;N+}jwNiu#n=*WmU
z9gfAxC6ASy_&x%ufVFb*PsRzu^n%gx;;A)twVxDaS6Ol2aO_Yaks5|e)Uavawg4^7
z0;giRGKjGtT*4W|!A|d4OC2&a8tCb0uslg4)sb4rUa;Tr2H1Jbfx@h3TpZ^j(GBm@
zJpegOlK*f3v&yk1EUGqse#_8-;6lUhCg_@Qus(bG(`%iLlTz`cLnE&KsI{dX%k8_E
zA7!dIJl}OhQxQZSB-NG;44g_U_a1i(?=ws_5el}%^7HEgq|Leny8fwlaWCU!4mN&O
zlFG$Jqq~m<nwMD3gY7m>y5upm)Fny|;eN9feNsGCjti=2u(HVu+n*jDJemrhDHdBk
zyEX?L1$si<9QM<(+N(cr(uyFQI712Pp@#=_Bk{>z&MZ)qB$DDl5eVWPpdpb~q$9E%
ztC(Yfyi!t%1`4O)kRqJ0th5Tenh>Rz2*uDqO|f+fyY<27qmSD9#<&>na56z-KU6$E
z^npdUgi}!*K7H;-Z5;ZYWavFnQdk|P+mng#ia_0R4%1$pQm--@sURUhAg{JO=;ck8
zK*!Gc7KIcw<JZV8{q=lOoZfc*)KaZ4K&mbg&s_hL0wx-DltlJ;RJ;&>|D99RQHX&!
z)dCxiP4f2odqy*rbh;`%ZSC{NW-H-(h~QAB1QkvHb51i0Y6(Xk$x<Rr>0z(QB(ACy
zRp1nOyzmHABm{+TB4f3JVdT{!%MxMKG~g^Il4`sFKBQb$QqWjzCG9D!gtKH%h>uE!
zGeI=8a@yE*w0cMc!l#-@BtbQdu8K2$u*X^YK2a}NiYcnFD2{(ZqLHKKHET&&fxNV8
z?{T;AJ|p)owuj1VuIo{JCy}7&tG7yaQ39sQ+@WXbuS8NT`Xm(F9;@*<N>kY?KSNU|
zFA!O&h&8}rneL-nI(YjJENuoi$FUUIF}vOwHIPo%hr$h`PoSdJC=qm~Zs3s&NYZ*v
z6j&K_(Myx+g}t8EqAH=1YCO+kCNxk}MRQJqdQFpo!3DTPc!B412y{RoRb40Qi1Xrk
z@MLrkTPeNm*u}1P-st%J>Ke))0uvkxRcB~$Qs6g$_5a6&Z#V(x(d%fpFyQLB(<Ggl
zbT;)7cx^O)HYNfw6FE76%Fw7+Iv|Yj{E^afhYr6MNd*v0cYcYZOf*!Vh||%!S&Chd
zY)&CkiigVsdBih7j$rp`*geSSKvGRCs#{yI0{PEZMFX)|wqLv^Szql{aY{rX8kzbH
zC-S1jcbR#t2DOj%_J*q!Fl+)FD7INez2x|nVSqEBuspB^1efeL0ZukvKq?`;?d@Z$
z5Xd`D*Oo4;YvyQ>wqPtyhocOtDC%5g&3srNYzR!g3|B+xOCIn@?K7<36}Y%$2*yiw
z;T=+18KXyk1@GIKrGMxX^?xmMZ-HWNyXe2Wuhm4XG7+89QM7p160o}&mmyR|*Oa_K
z9~N%tbM6h;9SY6_3ufH0ZKEK&=7-0oQUeDa6TlMUeoP4rhho{|uV3pHZKJ(K(F$ns
z`<5ySJ}J(i!!^sgE=aOqh@xo<f+>iSB=Ne;VsEDmOMrqxxP<l_LZ<?0E%kuu^%x?R
zUhW<cZ|klj7&>pPxnqpKX7CXewOP<m8cUy14PV+Ui-SSI(2IJhO}{7l%1{8fr?63>
zZTr7KJWA&V$K@S)Un_io#UlE?&VLub9!V8-R<_cc28Yh1kE#k~MlJ#k!s?Es>N5bu
z0H1-Cgcm?Qc!X#BI7o0}z$Zm=>YAxw=_jG{#lfHosc<0PKTuT@3`WCsRmZ-j8k)_@
zme!uffVz(-6aynrNqgDjcRl*ZL(e{X_hXOUd*8iyyXUd#H{P+>0uu7`-~8l@CEqUo
zcEQ3$AAYgmz3<v}GH;^?%w2+)d*3STmXbHH?NqZHpB@>btHCGr6wa5be%$8TxxDU*
z9V}l`Xt+B7X7&JlS+X1uibWHudQ{Y>sk)aS3bTHdLHNHCNo~=wJ{%J-U#m<anyIRK
z(t(i4CCN`ErxYth*%nrYz++EplM<fp9pH&!3aPGG?VXrcQZVy3CTb&rM5a2H&Grvu
z8WPESIy4^JjyPVioua|9ee4D`3J~?Ex;vc?$Ej#0<sq4byL2X)c&ybl*!idAMvNRa
zYS7?OqehMx-B5M*SMZ&}4(Z&F3j>=@^PJc!7G-cv=AV;_)5$=&1vC9XqhKhk(DBlZ
z?`+7UbKzo5)4qG5y#Z0A)Tk=lOQ}*n7oalU;5%a(Ud!N@mX<qoIOHUSpNPNG)bm$H
zyNA#+QLx09Y<4Bru{j82O!N><8H+%{KMABPtcM3Y^%%Q5z*oEj2tw-GCMR?){cNis
za4ekO*XrnWbs`#yg(Bf<I)3_B1rx3llf_UR0vT7!lZ?Fm2ey(a5ZPb*zo=9u?)NAB
zeww23jsnugZduo9+6#sc2m}K`@+M#)^CFH$Atd$o6_16t%X<R;a#^!Y*OgRof;y&W
z5`#+mhSg85Un7}{`1$aEjD57#;(1#(9FdeY+51}e2A9ZXf#-Pss&YMe0-C;^_u~8C
zy-x}u5!~SXupbPspi14LRDw=rM=o+?2aUVyTA~GZSk?u2?HZ)&BZwoGF9efMM0R(8
z4^e8#nYG{`bMq%vU$as(S#kSQBeKarijHS$C<u)kZvTnoL}N=PtjtsGeXwO`0JZf3
zs}~;*Ci{~yme@Ps-6fcy|4!CX^-o5Hah#I(W+D+DHIFlhXF(K_3~t`332Ikh^Jh`z
zELco7HAns8{BQ`igPUT4N7nsoAuF%GIuK819x1@jmUJ>Ee_vhodxJz$r5)Rm+%Haq
z6XS_#UAnT08S$mywUr)f9C9TEN}$u}{yddR){z!tChn&bvHDp&&TC7KfFW6eDH=Sh
zSfVLo{ak~UkX9hD9uId1I5*hR3=d%Jb&j1sDNTp2{ZYxUcwlHe98A#B0K?E!>cl4+
zizdWMxG>6R-z2mQRntLz#e4k>wWFp&2^4ixm#w+qo(6o{f!MIwnhTCFvJwd8Dyk!^
zMY!z+upH17t|i!42^?5WT=~$5Xab_9U=*%xf9Rj@es^bGEEuay-m@8eN$hySk<8>u
zI3RdmlsVf~_{X!kV2)HVgJ_dU47_snfv@-9!67Fp=qQ$UYPzo`o{E!#$)KV>I&u-m
zlgUK@@VP13c^Ndp#8UYd%PD3ar^dnC9+eb;490)Z;*K|tPgT*;+Iu#Bb7?gluFPgC
zvuS@KUVZsjvdQ75fSkCpy{>R~z!L>}2p)ndHD4X8-nAe(6~BWw_2L&Ng-dS`#0I}6
zn7}u(IZ#Gd<qE*5y(%dvE(TcOYQi=#$%*6mc@yEX4%g;<;f8P`d+h1;RD!C>#Itv9
z25Ck|q_l~?0Dls>mEe-vq-_&>UuDO<tg3JUl$JSkIP`%NmV+qjpG}U0ssf}MMBK+@
z=wWL_mK93J325V(lBr_(9aGXJh^3IAB5-dnm7d|$5Ypk%cGVBZrBePxj>%qqepMWT
zQwHY<i6zHB`~$kFtVd96cIsQx6RaKc@$~`-sjevYy_IDnzXMDlaC3+2I!ks^(xIrN
zr5t3(!H)U70Y=|j@$NyK1NfJK(-;(qEHCokoX12-9v73SOl5IU0SH~a@5CRlhG|h|
zt-f1(8T?Rk^8g#PMay*;jL3#}H0dWQPWjFO*X#G+PKpQJQ8ru(*MBrvi=_@BOif3h
zEOL?wH`4TE$rh_fTUypP*(T4SrMDHMz=}mA=^n4-vW$WOSMBdFsZIOwL{h0pB||5Y
ziD0xY;Ga0V%i>uUtMckN5|w~_dp`jT#V<H2%UT_aHy3AC?5LjJ4Jzg->509RG;-|Q
zEcQ~8&<UJT2-+N=skh!e0{ce^Pq9C~*umKnFRi#UOUL?Ua&!_~^%;LG83<R`%mA5#
zYXB_%9uLtT;T_iQuc2cqS_<#HoJghN6qP_#-`$Ai_e)D1GW-gG6Ub!i8x!J5tXD#%
z8zRB_=zz&sQU_QH#6f_|sDFOTb$31Z#=<QH+2Itob-{a$INEcsvyXU;BlnFrRL9bF
zksvss;XpDRs|$saF=paB`8HjXfbhUIGD4|uI$Bz}>*^{i6kdGj2knVy&j`2-`lc>R
zDSkV?x49qoI%i_LKK=sGah2t5%=YtEyeD`|!wu_KX@#av-cY#a*9KE5Buxg3WYRU^
zR4Q_RGn!~wx6v>6zPsPR_4oJ|dcg3?wPki{grd8P>(&Lve`YC%zVrUjD=F50`tqV&
zu$D^IRwOc26$}#|zR+rvut&7ipJ54({=tyHr|DqBHS-m_YulnHPMWw-g_wUYfm2yW
z$-T4T=2$X9k4Qv(*=mS}qv4UMzhAwPRWz1|^<Z6c?^N{pd}3U$I-I~xPoKY<f3Vm5
zk|m&$8rrT(I!`MPAOU~GFgUYq%f6y}57-!ho&hF_Ix@>k9q&(wCj8NGq&5q#X~n?A
z;Aw3ZM>>mmzVGJVGxgo`=%A3|pVY+9uMYaVNB2OCc>DI^A?N6!CMi(DF%|voVS%bZ
zG(G~(Fcqo`#UJb}N-Z{5;8a@#o;pRVBVV|WIm#D3y7J`Lm%V>|t?w@bo>e)aM6Dhm
z!sgw;4aAHL*oY=u(#Cs7Co7ZSu0ddr!D&VK-i@Zk@iI1XVkc<dMHC_uFw!nW6^7Y*
zTYaG66KDE>(h2MV^hvuct+-?o`lKYo9T6KkN6<@^mC%UG9c!+;cqKTR@D}Vjxm?KI
zuD~s0)f}9K#3X5a_R@1AvAPrmK|Y<L{rz9oM8k0<ULe)OOG>zxHI{(e1up_)Ex37U
zec*+Z+CwG7X}_s~!F;Gv8R+=v*GR7ieh)W+w}X#EOHxa~0eI%Zf6_EV(R4Bq3k}Ko
zVyCTj93FxrR<v!o&DGl>NV52qmxU5TvVZ?a?YR%%I(|^)k@p)04#U|=NkKR9rq~>(
zn3CPuj9o4qM8Q_mb=mL$e>NKk2jCco2R{Rm4jOB($*FV>_ZzOFnk<%#P&6J4>#D0$
ziT=laXYZ_}<XrajN$Kk9(g}_YdDZ0GEFgmo!MfZ$>*8_#0XJ@QAOgd|%ov%MJkVW1
znoE&LumHB33WELh=-PT;1iWs#I#U%mWsM~pYTj(x+{K%&j3t~6&{L9QvKPplt#GU@
za|K&rEl040RvjmFfQ=P_?nRmZ?zn6?&M>}e(uY9PQ&*x+OPmAo6UjC?1T;OQBz0|8
zfCJq%tCUg;2RL2hi&j0@rho!kf^Bs*$%bI-5F?qQ{rT{t{VDJXDjsdxI_I?Lm75eS
zt-rI9!byii3NeTXR$Q|5{MJnc0}DcM57V_7>WHcuIuVKHYF{wfZOBhN)ur$54+uHn
zW)xJ{wrz)R;aFb&@}wgZwf#qZ={y@LhxUYtNJ)IS38MRwPs%WKbrrrD8$MHX45z@0
z+dBAFkN*>=<H^KcEoudh)v!2(x3%2@+o9kitdb88uBfTaf^=gb2AfJIgVA6*Tm7lQ
z7sM@hkGgY<%2{1)mV{9nPSRXj#cn|)RD@`tRdsl{vfcM@Tnv{HYyypbt0uAH!@1m`
zRBe*>#qkHmM$Xn0&hj$$$a06EGyl+$l#H{-!p*|+3&P#{!^8$sla@Ke+r->X(X}}q
zBoXYg)IMuapf*-Fq~g3!o4+~r=rdOe_(3VQ>W1ftT0E!V^tpL%-tgJ&UBG7U?P1AU
zCV>O%>2zlF`z9;%oe+E|AiLy?`v(ZGyc;+Xd`BCv08P7I$W^4P62XaINDp_9AuP=|
zy0elZJSB}B`<7;J>a@80=5HSw#ZcL7w5s+Wm$SO6L5z<@%k~IlgD0X$R8V!+kl13=
zjaAjrP;Hiii=7Hj{+e&gPp=MUAK`S}7TG)wDA}iyf@y+P15L5m8E1E0dgBkA?KMcb
zV%`=D?uyf}--+=>F?}pi8M$he?T9@R_I5fQ2Lp$Uq`*-8J%;Eu&lpzGEkE(f=&(0<
z8O%G-*C=%5SqT4Ku>68q5F4kD^#3(CDB>G_<-dmf<AgP~Xh_BG;mRI7T?8d{^nwfu
z`sHak=8<+$mtGo1V;#pN6Y<ljGd?ubCZ1(EEc<XkN{S?v9b3`u_JtSFF*@2mTsdly
z^GtWvrN=}B_@ofsmwi%zw+KY{f+=@)ZJK#nT`C@@nPjMcCNlg(fydN+qLjFu-GO)k
zWFN?bqRXtQG%ER5Mu!riC_~p*3>X<db8+({b@7T|;{KLSRbjV##S8ZSTR{j-SG0hl
zKQ`Kqt<E1ZljU(jXZ#vg*JT~~jiFJ0oaC>9n9>#1FLxCMuTN|5>;C0%D3uicEy*c-
zc1mUW!~!DK(>6M+1X?PzFMjLEXXdWX%MLGTjV#M6Tc+juLFf~wn3z8_dT|km?479{
z=#J<?Ddkpv@wxq~Vzsl|Tb{2E)l!LYw2q?U*^9rF1lbnau|7Dkye|hAH;CuMz+!eq
zL$}sHeq0=<Kc(s5*hTI`WhLziDt4#PGrBJ;={(I)KDqMjp%p%G$l)EDqO0jknyrGV
zV@VR&+8%sT@)@9%GKxN@u6D65_egD38r$+~q631}|Gcq|%K5@IN8QY6EL*_oM|M71
zyTU2T1zT<Ta?Tc45nHaT{>T;aa>1`TMFD5gZ@MW}g@t$UVxtpOeCkq1W*r=5Lez9$
z8~6oqNJ$EIL@aK;zSkhH=!p58<eI$Xw!C&yFjrrH@nc`M%WZc1HqmkHrf0GhI08IU
z$@tQ9Abtfl?sQa2`eL#Y_yXTgP8}6s!XqA=F)o?%W95%zBsBc;g#}jy%ciNMs(7Em
zdd#L*F?G~Yb(7<^d~`)NhAu=^O=954Ma5@(<gf2Qe3}UEZ@Hbc?2|%tudcd;@dYqv
zIYhdQ{9$VNTV&EZ9DldZGhQOA>&S%yg!-D;I^*~xU73xA>qB9j(5^CEll7(i_jK?D
zQ^pddq$K&CfvG59xSaLZKU&ZYW!|8x*EsDOMn<5fqTQ;A1xxvCL<~g~4T3M8pz7{x
zc0o#kq-1Y?Pd=&Fif_vH$HxpLrw&KkGc+^!<|a;rC`s&iq3Y<P1`O!ut2*n&wLDkA
zZfZepd12^)Otz9vRUMIh!-aJ?#N2xt;r(G|9!RMUEL?p_4jT<Zq2YrQOeB)6tV)M-
zwYRTO4J=khf_oA>dArmNcpCsuN+AQT72hV$abOS9S~u+W<u6}x3{8asH8mKqhsG^(
z9z-SWw)CYZyI0cKkk@!?PJPUWUHyco$)rC}U3FW*K$3c6hsv6{SAb*9Vo`E13@x)K
z5|u-9OjC?M8i)@Z96qawl~l~vho8{fxq<bUe<0rBWX-0lZo1`^cM49|H&aJ_YP-A*
z48^oja0B>HR#<A|E$JAp9<6=C7o#VCkJF5kb$Dw!*f@kr3N*Dvm%BYeNJ^)wD>HRe
z{In$k3jvAK^wRJn2gD=Mcp~_}vd=YlwK-;9v`t-_If4%RVu566&^tyeXfZZ==%^Ah
z1f^K}0YYX7s0_u>#HTWF9Mg611)RbguYp8#(Cw>aktY=0Q~Zn>b3k542a^t(so@MG
z8V=?J$G|AyfO?3`*v`#!&#y^%bAuDsg~u<l9)swft|Cf_lIaCc;FEUt^hrHnbYFZX
z87Gz=KvOkUV5(!|zQ9~)uL)|8ZBV^Ht|=KO7gc!vtBVE)lf$EA_Gh{_TbT}Jr_ASB
zOVYhe6O_swzZyl-bm736E(T2wW=?V3-1Mj)RPpCa<9Bax)Gksr)luQD@YPKpm<s>x
zM4U^&7bSb0pzH4CRh;7wHS`eZ=piI2X)(4W_x*a3*@pRfsoF$5_lSzso;z1Qdd{em
zCkzQxR;IE?_8a=LYT$f>*vH=T%E`ctRPKlqSL9nwv1o8u!`Xxw%F0UykPmaC&Ae<B
zgs%4|C2O$KC{0Iw2~sRMT65P%%p1i<Yp;QHcfhmgXa&`AjJ$&N_iUXj$hO9Ex}jQz
z#dFP@KDzwGL3DY}Z30x%2T)0qrMk>JlyrAJPT^zUGK$Zod-yuu;>?|0W<f8d?CtIv
z+C7kZKtxrVe+EU>r;$F0x^Vpi>numWx$p#@wPjM1oG{h<mJ3G?4z0psf;9O^d%xzt
zu4<Ck+!PwT$aab=K5T4iH@NoB&1;_-TRRu1-1*>0rqXLDMZ`~8A^^KfN_q%zs7UJd
zU=)k_ODpThHJ{FgY6gUAPF;qk+FbY2^!eX@``s5WT%GIZ%VhhVwO(W;&`KOI+_vb}
z+HfrO*Aw28;Uyf`0w+TpqZG{o1Y9^XgsxZ|-_fX;&7WKp%?=uvibU#yky?<cNd1jV
zU5*9QhXo+^5Y~H2gQKD}1xuGGAgQWa5G4cWuyrkQ`}c3$F{wVAjrB^(V+M}=vbU0U
zV|RV`OhxS*R`KOp47z%Ns%B1G#2H$ld$R97K*E9h*0Ph6S>NDfm|>EY+4Rt78WjO&
z8xy+>vCEV>Z1?QZ6eXJq_^TSuxnsdIAg%kzk%_nMA1^vG{!F_p6+gT9!=D$u`M|YT
zUwnMkUzoRfu<Soy4cAt(`?phfH60pumT(A_6#FoocR=MJL?kK_^H<l+>}mrSNRroX
zF>#*e{L1@A$0<5MKhgpEh=i1F#awY)Z9J2yJZm|qCy1!zSej&;t}dB*u=cJYIYmn=
zfFSeYmP`LDUYSezQi)I?5=#W9wSgGsv6T~Udyvvn7pbaPxF9%sdEj$b-qbZ+*uLts
zXRbf@#JWT_9*72V#GIEAJZ?$x{z~{EL=j6Wis;==TK*tBdY)xJk?k2Z!iRw{UHznp
z1tiMtOnm?jwRcr#;ql=bI6zcgHs0{JtD7WCM2B7!c%I`d+v8{ViopjES$mtJ8lq-P
zqTsYy&)oKnqJJ?YcHUBnGuEFu_*3!y^Xo$YJLC(8j{gYEiqiUQLv=9QooC`dbKxOP
zg*t>J#qr2n{zc~o`Y{zB&JRRGXRc+zgxjoS7g{9S5(HIw?}Q`&6t120L($}9L%?Ap
z4*%t;m7!c^sNwS>m@3gxx3=5bCW({nF0qss-P!5D6A1pdT-}d}!upfZ$|@#uV`I^=
zHPO?`Ju!fE<`-QJ=WZ}qp65kTR=DqH-go_#7fd-eM}=yVM+TxHI+qBhywgMgO8P}{
zdjBe~+q2}4k`H?2la^w8jP7UeMoGbQKI+0B3%cSd>3;d7Wq`Sd??}A->DY9d_T!xO
z3F?^VCB0yZEF2aA+)@mTdi-pZR97$*uKl}h(uSX#gpT$uYi*O#&bPgP_Wf+fiq}r@
z1#elkY2owt4GTy5{jU?h6Cji_*I$sJ{G`hXz{KfkEeeVA4h;^OxtFlxUVCj&=*T)M
zO=>;HtLwrap}jT*lanmHvr)7yx48X<bNvHrN55W#un7z`2&rn$WsXevGO;5bb76T<
z2}zS=-PJ^kV?q98S7>z&%@Q4*ley_2)DeFon@i>Bn+i713!G=|OM9{_NRyO{S%9uq
zU^PS1+g>|o%<xz=K5#&gilws{>*R2*(Q?e=f%v4Lq#i#@f=P%{Qc^E7zpSJX-CMnV
zQi7UT#jw{+u9N_`hwEtfkTtkreI5J5QBhx*@nvJdQL*Eebt-ymVZqHWwMeYff%CQP
z0QQqAth43r%P#*wl@`46^ds-LiH6Myvf6s{`EN{{Jox_#S5A21{jDnd?OBn(XJ#r2
z=rUh?JC*U;N&&R*=#zgw6n*w(91@b^?;+zpf4DN44b<cB#976%nUh)t#gg^yveDJT
z7Ycm7h35qJyVL*cFPUi)&tvH_kU6{U*%PKcIXdL8zD59_z+}~q`$kWh)go{bs3dSx
z+1RE2xKXh?nic)+{~1X1ucw&Uz)<|um7q)luESH(oq_rZZ{q`S*fk2e*jY4M7u_+b
zj*7*jscZ~f4>}p^AE>HkA_1%rT_PMoIlAv@=?P(J@JUUG?sv+}N0L61+KHr0D4aa{
zqD>|l`a!%B+|k|b?JjV^X_i^jv1QI)I4G7)&=4{vnZHh2<tY06VSk&jh~q#46mO!O
z(svx7sonA5+3z)PZ*1)7T)(Epu-m_$H~qF7r=49L`O}eAzCR!J<cD|8`eM28(}(YU
z^t0)REZDlXi)*_BCx)hLpWHxLiu>!W_+a8tkQ4(ZxN}`)Mm?3O@-yBnBdKKeGYF?e
z#}d>wTQ5ki&RJmKi|Uux`Hvj4P&4@=cI4aQmZu-z&^34P(MO#lD!N#7Hs2BmB$<~T
zy}&~Zr7KdK`{6NXeJQf4{75`AsFH$cn@ZGOv&rNX3!)|t)5}07z1`V?k~3Pw4o~2*
zj<f1?etN~oupe6>7%ULRgn~pwN!vTH&pS;Xx}*D^K*G{gPbDp54@yeaPy@%k+3tlm
zSo0Cd>a|Dh1Lzt+FsL;towD=l&=_zOwx5U%xa_NK&kqkr<EMY&s0OUnn`P7kSRh4S
zJhbA0C0je%f#g<E5!bzT|4nyJW}<atZ@=}731{5BYR#MPfB(fho8STvIG#0StIc{h
z9V}gobbYEeF!CeP7APH(syhdtq@|0^F{R@6+u~G$0juGi>(~@<-Lb`!j9Zpmn=2?q
zi<64>(lh_+zpcYz(E&9I(v~+?D0%+ThChw?QG<7yUO)M0Cg6Wq=VgP(+!RY$eO_hd
z=^y1=KRxE?WGdu?qe&$uFX8jRI|mtvi6y(qE5VL@*gKtLIi6G8qP6Ygn?{7NkrhEG
z33pg@u@^f%bVv8QfTA!9s(Xx6>EsfU?&6ce$21kXVVleunCXOrL;!nehv<#cGOjJ@
z3NO7m5_g5JiyfVbOu4HDOPUV2VLOf=B|o#gLadbIY`JCpE!SQ>Ywj1{ZNu)<wiTbR
z{9(Zz(b4a1?f7Bw&ztj0KHSc4|M0o@mUKv#P~c66ciJwdDKACeR}+p?@w-|iOSZkd
z(u0kIPEsPH@F3f={>E|9vhaW~g$KP|#NUX2MO0ACKEXRUk_#ICH28&1n-v|qv(42y
zySB6!*Pj2UVaqjJY<qcPx-$Ov>>S-Lf=(KuX*NzesxFq8@Y=g4(!kJYC9eMLuT)1Z
z;Iz!Hp+fXUHq#3@3dB@(Sy%KzK{Yrr|K;stX$o_E_X}QhUwVNku@NqyJ9|X;<;V$0
zYIjHXJv;;^7>*8~YpR;6y5M7C$HIOI=?=&SIPis62uqnnMJ5xd4A(@$iCn{7Kz>Xw
z#YX(nvbrmoc45Vu51)DN(Oa*3?ENKc6|u``Rn!F|j{U+?ERgZWSLU{gTw&gW*MFeM
z1=f;yu!HXmijpi7a4Kk;j!#*JC3=4+jrlJ56)yHTHr;{UK;(Bq=oIr`+U|5u6JErz
z_~Z|6h9<u<w*SNhs>up0XeEaECSRDF%YLZ2&X$|P{r%zKH5+*a6auGW+;DGyieVy^
zL&qga2os9d#ZUQI;yE5jiXJD3C257~4SdhlSO}G{YKp;ad~#yi7blf!_Xl2d-y<`>
zJHW?r_n_P5cG9wvlIXsMC@EdjqiGdnI(Wq<F!AIM$A#{)t9u9d?T*5CJXjq=XFuVO
zQL#iInn)z6(5Hq0;XD@9Ex`c+v^3t9RlTtJ%el`#J7?}E->%Cy7v$yVkDl@Ux{j@#
z{2N!j-UfDF+xGOG^VY)&kPKIJ+Ru(sUQQ|5=lZAGG4J`1#7~DtNpW10Y<)h#=M^Y}
z^A-<Ok9k9}d2b=CPe#To!xNXxK-M*HSo+#!qF~8wj%>giW<hGciF!ri#r&K>(NO<?
zuCgTqVtdIDJ0467%2tBaq!JN-RUKUC+1!w4+j+Jik+P$N_@vXmH_!&cYJsSVsya>Y
zU0jtO%p6VC_Uu*K8@%Yg^a4?q$FfhlQx?>ic}8(&W)~$zM8>_3(=kD;?p;&6`p<iV
zf>Yot;t#jh#49to1O?F(cHbtbc*AlDbH4CaS|g|`SbxcscmZNri`%q(`3Ey!d4Dr2
zEWQ1f&p&=+?jvu#bi(8%9O$Pg73SXjVH;;TIIE-kQa$OxC;Q7#^wgz-jpIzb)gDwF
ze4mtfL*RIdEsxiAs}!Bi`Yvx$%(9Y_g$%5*L#pV3$!~jcc+~@K`3_Ra)WD`xQ~z?r
z{j%s<i$+)a{(OFO(NGM@WLb05grhK91y@T$+!U&;$_}2<$`>GL0*+$R6B2Yc=N+X~
zkuej1mwtNS<Xk+FOvHE0`LG9c^NU3H(H)A|4T$cEPg?4`@6LkSp`<&5W^QP8aN0%H
z&`KrA`v9CYhcgxahkvE<0!z|08Jh7&67)$gtyh4RNYb*shr^&i(BIz4=54jH!;!o6
zmKD$5{nWw@uT5VkZ2$VD2X8%Q)GHkl58;-?=fAk^zU91Vik9HCUWY9co|x7JvxA>%
zD-x&l;N#$v)N@J=$y#t)vW^6x0IbL~55jsldgMkF6w^(?J%Br<*>zi9XsCXxAc7xY
z!a*Y0Pit;@QBa(&c_$q^CNO>lSbm_HS>RUPlJS+O!i1v@!Ekl*wsst^?`R;C;FIn#
z4#6_{ummS`eST3D$teg2Nf^Hu&^R82U>7AVqufrqQx?=<@JXNCsmBR{vDtv*mlS1|
z({Yp?oOk@+e@L35fn8CBZ_dPW<#mxLhyhKFxaJE@X_OW8NXwTfc)gaqd08uGw6@#w
znzxn~c!A}%eKG6dTTXpWGi|43<A?XY-^_KjcWhj{;DgUz9G`epbVQsRSDiNmTnPB>
zcoOLZ^{>@n>#?ftLC3)-DY@3bt<+wuq2occ_cX)A2U3@9>#%{Oz+t2@@)H-v8auYd
zcC<TE%k0wz&1_Lk$;y*jJBr%4eR;bfZM^ZKg>Q}<_`GR>(Xv5G+08GV(X$>Cn4%wL
z@MAe%H6dnqc4O%YFmDfnC)nV&r%xE1z%~ISOgDBi$z_z^2A_0MFQ2pwFEf89pOj%x
zN!^!r6Ws?wV4VZ^bqSiIYB;w8mfkJx<9<P5O_eht0F>sA$YiJ_9052A!ynot3Z_sn
zMDQdeY>EQzONLX-&$@NBB5YSxZo~c8f6Y0qnk#Sr>bgM}eX9wb;^tL9E`R&|Sr6QN
z(Q!kw4e9uVH@b?VDHx8-p5vJfUo`}-3F^4lyMUu8(0h=8UknGIq?QQk$O-PM3n}mx
z8JdZO>Jy1T>TPgHk(oqE0p(xnf@3e495`oPZx0{#YO^TWd2EOxou^$L+Z*R!{z9SU
z_TaS~@YdoLqaYNTzr6x0_5jCnSsz86yugNl9p?tYwmfX7FRi;bFnJFr9M#%qR0n#|
zlYq_<$?Qqh-n0#l1ts<7Ku46g2zR~nNr|<GNGW_Q>AS~K>^l}kkS&K!Tc~TggY}=m
z*zd+S*$c2C$N{QyeA|p6F@L5Cr(29o_)HX0Cvljw1XsCQusb^7;JTd;9Q*VJN7Y4E
zeDAUweh_%GnYGk~r-x6u?Vbm(z4YdXAG!7Ni!Qied@fY^xBouz4FT3J6*%739?gPR
znk0P&u-COWY$>I!9Bef`1m<3Zut}Hf4@SlqoR~2|ry3Ge;*v%IZ2PW(EWCy{Eil$a
zTloIiz;sS`SXK~N(Q;h1NjAQ@{rP6aeJ4C|i)uTLWpV|<w(OrSCE1RNL@wQL@GFJ9
z#o>gaO34}B{VB&LD5Jos{DT$DF@at)7ZDH^?Xd2>pToz!mDKBN@&f$$Gzm!BtDO{-
zbl4p5NvqRUG>Pqd1^58p;}CVTUdcUg|M~`qh6<d}d3z$4#mZL<Q#pC9%$k;HnvSHq
zw&Q}=q=A@fo3A@EdUvzjTCi+$?Tnk}aE)I-{jnf@`^>+pqqohR`^|!nzWx^Q%_p;;
zdUWRUE@Yr97R;iw^`x*bURTOrpy^XTFZbn=`-6u=5;(aIFK@rPf2aX?;GLN0pw~t0
zYQjwDULaN<n*s_2#TJww|9RBiT@b?Cwj-=nA(U6lWp8b?SaV5w@Xrp*cNH8(X=XPx
zuNYf0qf9jYtkQ1C4g~J#jp9o8VD2&T$&}o6=TJ10q<8A`>_$)3>lQx(;*;({NqJ+J
z9;e8$^Yr2i(E#|1J=a+RbM&N5I93_XI2mN$2Y5@%rkaN@uALG~_``MS>P$TH&_>lQ
zbT-YH&p9&BnnlbcRlzlQZ_3~PNA#XnLut_+?xRbt{yIPZxraU$+gjFL`&3uEXmZ#B
zAQlXphc!w(HZ!?6mpK1yb)*zI0qD9XT0Ko6J3;Ob4h}j=2@74<`eYP6s6LJb${4th
zpSm0(B5Z4d>$KPJgBO+raf2!9lELXePWao|OJqmjWvf$y5Yu3JLvB)>rfX`Kx@<v|
z*dq8#eDi(&QsFSle~p>%AT>NRsT9`l;hbaMQ&Dj~{<D8<g0%mW73>bZCA5t2_yxQ7
zNrQn~wpryKC*Yz-$Jo$0qWx-B1fTB?zUko9VA&lqdbtlE!}GxPb|lvBGQJ#4W$9!x
z5UovS-)hh6-&_(od6Qj$_|uhy)jz5VXPM2jPYPCF^xOthYhC@t$p7Pn@9gzYUwq$k
z+x+SN*Trp}aPTBYF7R;JRKc=c(HlN0HLV&SNmP`}?9tTeOI?WXJ!5|`P5vPWoKP5x
z9RP$td%urU=~_RSdGJc9WL3=!u;#pLlQAVbiK$yOvFx42cQr~yXY1*IIbmj_U|`Ks
zv$Ih#S%HIVUH$fs4VuL*z5BlPd_lKo{1qrhZvJ!I49mv$QP?mE>j_ay>+S_oSFbbw
zWeqbhv<De2qnGNHU+ko$I}ne0N{T)yH~{fd6V%|>4EKfTZmRGh=-R*u%Up=lk)H>k
zq_U}Trc^K#`<-l}3P+;`qS4b<Y<p~ckg1;D${U!Zr@#F47S~pI+gbC&9izwJ@M&|)
zyN_K~{a+87YT<?BE_=JRxb*qAHepgfXUkyV$@MK-B36F}n<z9tRfBD%CD5bhbiv1@
zqu{_?l>NlPC#e@Xt-2~sgR7K^1?%IL3HpRZt^|H3q0&waLW56|?_D#m=uIu3o%#>@
z%GcI#yr#67T@n@>!)a~#RkEQKrUl}czPH|89El{c@P5=68TNjWBm%g|Qyim*$9ebm
z9`st1b?4JK9j8e^ybF3;<PyrVy65OV=)VgblkN&8J(SdXrKaRN6ZBYr(qB7hhAgms
zH1*wZ9mkfqf+=(Qb=eG-a-b-t^6LAmgFx70-YAF`&|jPP;7<x?h@7hy*ROj1(n%M+
z*R=Y(_bz|2U2-=*k!_gSDy+TZ#$`OuT5xpSB1kim6nj5muSCoKZd4BshE80&3JXhM
z!P0{oI~@Wg1(Fu?vm5A`k51MlsBCp*Jbh!U%St9G!QRbL?<wg(#dLeI=ya@E`tz3j
zhvWWo^vKC?G;&-)G7I1eV7fjlo1Bn;XDnUqA2xUCZT@U65GUe`RXwKQ7$~s_(hDmq
z@pJblURU8VC>G}&4dTW`dY5Y^)WY~4c|^EWdg9#(x}*E@8)fV~TGRvhq=xlGsmF<c
z#7^Fq>AGk#dG02M%qP0rz1<7Q&jTsV8=!$~aqW3=+z8{#3>pxw9u(=Ht-WZqEkPiu
zu6=BwDX}=nAHLJF{Q40W&0ANnjY5~HaLu<xV>f@$xb2FYmxJtr6Dpe`Ys(OExkX~s
zMN3_KZO{Fuj(tPN4i(&m-?Wlqik@3+yaUrVVx%b&L~Gm|O3@HN=yq6pY$z??EYWdI
zw(#?ui%vfOu9f;b<743gL+_ivQQxkRnq?wzFR#jNKb;?AlJV%s(~u|FYaOe(<>#W*
zj^1dQm*vOZz;!gyHO&ulaNOx^so&ir4bU@A7>cq<q_gxSPGpoUlN>nm3m28t3+a0@
z6~{R3I$CrX_@vL4^GV6#aJM7KXW})pLGLuD*kiV*{Q{0%M%P9%9A0%LXW<xfb<p4(
zS4U%sYOw$JHHsVr`uPW5Xp=DIi{l)v>8podd;6j1R_3wxH3R}19y`uI{O+$_N}s>n
z!ComFyZli_aeNxE5KEnWG`@$)q3JtY%3WE%ZJ!k4BgOf2Ea?!(a~bzh>8saa2R#Js
zCHW*q-g^p4YKS&Y9<ubB!J+>hpr>x_eC332ML0D6fwjbp1A}3$AQ@}!&5?|yNdHO(
zJ2PoIKKUo7XxpHocy#x|PY=2bf9_||7)4X$hI83utMEBoN^k>!BS8MkPssCJRh*)#
zgQLE)?=LH9IlBa(Q@iPLVwmvgdAj>t%4>z_!A(ML#+Ry!B`@A=Yek&!p>J7G@Ix`t
zNq1a@V@;V?UKr##6;vWg+NP`H=dXozwrcG2_bf3aLn=Uw-#q8y$G5jUIN^*3pIlsE
zZOKx8dgY|SRKvObBGU?*4$ecVJR12!778H67vq5`fsUW{UAGu`4+jMYpQM&hbfqWI
z8hbws7*O}T44x_;tkN=o?xR7J63qJ6JJJKFWcDw)`R=-ZW#i$BMDh+CVJ~A56H{`O
zt_4{tO0*f615yZ%*zngS&E`}zY_QZ$Rik!+t14*kKa2DaG0A8+>aT7H1(Vps2*e?`
z(z6{Xzet5cOl^G7XU@Idfs;oz<Z5Y*Q?CjpbWP2`n%8af`IKi(N{jCR$X=MPc^Bd>
zB@c3nS6N@qJ1VKnTCObuO|`cBv!xa}V3a>XrD~sS6M%<?y7YqknnZZwN<Y1F%4JJi
zmpw4{ui5wsUkR2ga{TsXZ{2*{F;&Ui8jA{t8G59Wi3gBU0m7*{4LzJ{y6)+Y(#HoA
z2cM+2;1pL(D!1=2RAs8+obNykvA9F8(_nd{o*|skB}d@i8Fo}fEK~EBngvSXwRFD<
zrurY_3g7`bSO>+`6}xNRFkmN;K2wKlXA;q>yEcQNbZ{n}J&hRY2?(uVOT|^wE}k;=
zlB=(|Z1SboUw-KoQ_s8f@=LCmaw*&{zXYFMh7Xrta@nPq!F!jHhsl>+dg<kpFPOg0
zAWGU3NWR@GX>{<bmNh-oBRt*-5LjN<VrrnIlCA++`zEJYT^H_kM>G`i$5XvLdd45~
zWplMR&D)C6x-0XWFZ<^TxX>+q;T6Z+uvXx<uX=Dmw4&mEL9-;iDBG<Ymc8-zd#hB}
z<xNZ=z(Yv%1EiF6LoYj}N4_l`pSp=`{NUmc1WtOZyQnlBa{|kY#H#Oa7J#4l*8|=g
zuFYXzr!B5N+uuJ{;b#J=*Hrf1Q;tdhBmAceG>#}KW}CQ{u>HjolGR}5K?bp1ho-0e
z<Q0sE^^)2<`vCFdz~DEpU$<`64;wdZShajx<GQu$)&t;<Ja#{L*zw+m)!+P_=bY(Y
z%^(kDb;L1FyRBd%6`!t;@1e1auA_qSV?HxDQ{y>I+1uy-_W%kn`N8<#E~~=p!!jg|
z9;|)!<X6{p+N$Y@z|!_>vQKqLcH37MA940lUI4ST>XOKx{(Rm>N7O~r!SW^oTYxYM
z%)jilJfWdSkJA<4Gok73su*^QdTD0^Ml9*eZXO^GI!VFt6gnU6uJs6l7a8)IYvVK{
z@bh;3LTq^=FX1D{Y?^b9@4rH)O&i}oTK%k}etP#)(@vgx*UyINx)`>j2D>`n_Q8~N
z1_K#XQkpvHLy03r_3;e5`!e<bi{o^&fSo*s0-JVpUbgN1)Gt%FWKnnTAK3lNy4_L`
zm|hi3O56<}`<7lT`=kI*U(3(%$k6echqjpxm;s#Ad7qY`2e3UKmnQ#$uC_Kk5F?db
zko)(xN*t_J&O^}7Uw-r{-w12(Z|HaKPYTXcu5fE^oN#pZbCDN#9@JCjE%Gm5BZFt`
zVWr-Bfw;CjcWh-YQyH%h;tv^E^-9<N@8yBw;FDA=y5FDH!^Tm`3)ee1<1(0bZ$=ys
z!~?(?@QRM|-IK$QN{_v3N!J_MBO6{(<e%=mfAzX89ow7bl0pJU<xZu;;$9kFS~DF>
z->{kYWEQ7)#P#h1uA`9gsJaS1hOWu2nynj-D%*}@lYd0{ac9U~k|u+a{=G-jRWBv&
z&VusNNtq<6VDd%v0527Wgic=gMDD$;jo1kv3A_o>y#fZe58)GDl`^*1L!2*}ouZ{J
zJ|P6gJ_>sjKVy|EVL%T-ysbZx3*GeA<Rky@Q&vlc+?3}{iT`ow&BNYgE&Q#>W*H}!
zEQ+j&ZYVj3-jS4mxu$G4-;f$mlZvNE0W&H-X(RdYpyD8sl-M7*bM+_69w{O^_<bI%
z2T&Af+Ji_arLL6+85Qy$Uwg;Q7k}U_k$v%q&?k=g_4u>5;-`+%jw=M)gY!6wx}<zR
zISQNrc0?*Bf8UCQ9f78ReMU_HlT}=k)!_>U&uhA5s$buE`>Z)LAD=nv$tU6esi&TP
z`pFmI`BU(WydXDt1$gr5yRVzqp+307M(n^&J}Gh+h!`3@&or#djvyq3_wa3x*7(9#
zp9MdFO5R=2b+yX?hT^;=`!A**MPYf_Pz*&7q}PW<&|zo%sgcjL3x>q9nj={3%R@sA
z@goyAG#0nC*zh)l20*-Tx8bxUHfip39mxhukN2g(35`9TN?6#C#5oN|S7m*<G&!ji
zvG4JdgW61b5J~DqL2%PMr}PJ<z_S4-E0sNV1xX&o!{6%^;0u#?ZC27ZY%v_TkXf;~
zWkUZ&nzQ=kkzYsxuM{2I1)XWsq%O7H>FBaMUK>s`Q6JbJX7nt>#tKoT^SU3W7duU{
zvLha2xHkm#*^0_wI1~s5gXBLDO9X<UKqwHP=|CWk)j8o7z^RwvuCn5kuiPCq^LL=T
zf=N%Gl;o3|TJb%Sn!lq3stjzC#lv&f_o}YsbvSZ;>%_<3OMWw6#V#DD=4+E=q*HAq
zeeyS&DDsP6TgE!-qDlRa`Y-ynjfw-N8*`yOdO=jAcM}a(QNAZfk=y}n*el*$IDzt2
zCrf!IRP08sh&3V)CQA>pl7i^D*8GtJ6F9RNK|J%2h0U0|2Ra8DPIzC00LsowuJv5q
zyP~#kO8u-3$1w`Bh2~sTn!7p+im8i>PE6sDq$qIbPEin9^3^@9e0R`Q5K?#q7uIpy
zR}+~KlTM~f|CvY{9+>)6G8;+33-5*#+h?M&Sj>0kDr=APIPE4gf5>d!DDqQ}z*H9R
ztzE$QBGWr1R#!A^(C=$HR2j0RaGbN?WM8B<Lj(7t*LLdp_a-GTY&7|ecT`qRczlz@
znhJJzdJ{NVMX$o%K7jGP?Y2Bo69GrU+hLkI<$G7O;aV>(b1-m_eNtO8HSr$gQAs$M
zVDcTTDqi}0XMnchg!eYb^Q>%vb<&nhd0V%ZO|5ypO|#fM2r~{**4rPOwzRXcW9x&1
z>6njUBEIOcpUI*vv|Hu*#P<`x0!;y~Dvrh)#TRHM94fVx0+PTbe1Lx(d|$eexA3=w
zPW!(2IQHG|!nj^OX~`)?l5(cvJ{L{|L%XLatjRy%jAf2wXa(RXNss9}uwxngO}^v)
zA!NsCI(phipWZmADv^GT6U5J-c=WS2ER&DjNg~)Q{05G+)4ech&u|?9XVY=#XM#Ig
z^l9cP9LM8%1P2KRJ8*(Hym@SQcQ07|*s1Gmp?d({P9XhGWLgJee<g>N*FUqu<(E(K
z-v=DUlKEBvd`D+)r24E`o3_mzPGjc~ovlhgBC{N)=puf<w6^^Np#tWkz$Rb7!2{)8
z_5QmbN-ug4GIZ*9#V5PFMZJ&6$Ec(|`|iE(#D=_P6dUtf&Y*MMrx?*=jj*ohuvr~~
z?%Jda>41b3>y^M!kc3s|VRT;`i6m-I8s?8AeZkW=x~|aGMN<2KpQSEM*a(g;@unKv
z3nY*n_@1-klvt?+3U@emhgg)%->6SYfG*EhRiy@Rl1Q3;2unEafuv*<9vm}60NS8e
z;5f=wRyLQO+y9Ct8*I5D1637g{p8AYMa`9OUK2?~qIDFV$(+2-k$|D)8T0!9uh0%s
zTen{CZm8}iTp6VobXB>J-y>-cB}J|hzM@asW$E4Rq@}NdPpTJ{uHrMjGfIhjx?@w!
zz*CkvilM0>Z`h>Yci?4;2SE}(m=K4+DVN1gr&x8aBAa+rfjH624u!=^#|jh_pTL*)
ze(bPBIV>hx&&4z8%2Mt#T`_SP5d2^Y=|Lqaj<U0L`5v?lrC*c?5B<y)NU*XyAVo2#
zGO;k3>sSy@x<!){zZ!e=Ia|PAwXtB7WEbX)PWuuRQ<Vm#pksm9fZCT`tWsi#4rwMW
zt!}^2y|k_sXJH2`+0k4LT^i%$fh3tAmB9FuH{f~b)bH&FDthn>KgL&}q#XSEl9HCJ
zeFA4`Hi}Yl$+6+GEEB+2dc;w5q-wUrX=DgK_P!k$Kq6f*__z22NvbZ3<{|0Nrt6Z&
zEp=3^?L=xrAzKlIdhjPhQ1897dz6?c6&DSS?Dz##%?yoHqB(dtcqGMkP>h<J{#ip)
zDX%I76AnyS;j+A34)1q{cf4>;G6x6bfi%i+NZD`x={xpoQ7T$p_)`Aqvifuo=f0sZ
zi4=J1o7$=Avc+>)v1ET*pLzp%iXm}vPPb#yO}H|UN+cV~qb@3HaN_q~eNu0U9U$$b
zYVXmaprmA$=Df{cPE)<}0()$V353J{+AJ4zSJH7RoYEfl74U;(LzX@neRL)Z`vbSS
zU?fS`#$Vtp8*DpzQ@u_AX?5!L&aS{u$&qTgXketf&I?^}WkDlODSqZH`3vKqkrXLv
ziMDFK-oQX?Lq=~ULw_Cd495z#2|o!I*fWT{0c}j#EJ!y0?b#Jo@A5!WESB%IJU2Y*
zPxVM~uNbnV$YG&KBab0jw>IuC(BkWwp@Wjf%UZfS5Gr|mLRsy9+>gckdMIf(u*b>V
zO+M*c*hpV2zE&SC8~9zn43=T|r#i=4stOAqn356#Ix@@3k7X*rf5bNZq%WH2pE>T$
zg2b7w)3eSV!M+Gu269p6XU2OpuEoZD<~6XwXC<-yg>ev-6i5nQq;Pi)#l$@&WugPd
zFEU$f9%r`OkE22+7!gg`S$k11b$7dpvTCyG#)}e(_;AnBrL(nk#D8m-;$Za%T%h#b
zKER8jL0qiass00Z7P;Mm35OYGz=*GUWI^>nxq`_~^+|`l&YQTY#;Jjlf!}R&X(D*l
zCRbKnh?@>jNx|473*~~-d|Wh(E)lGms?DX&{h-yf1w-Y)8|dqm>;cFqn##80LOset
zC7w_TJ7J9nw$gp!Ao`>rl@R+Z`)4dzgX8b;=OlAyfi4yea4lhheg7h?nlBV>b>pS}
z{%3D-S;58W7lh}JjR(vAA`^`yPh11i0742Xdmx7ig%n~k)sp7qn5rI0J_yvh-RNC5
zLowlr-&;GX>h)NND5+5zE!zD)juyq%d5xQ%j+frvdCiPJRZS0Uc-tmZ<N?tS=&Zw+
zKq6d6EI1F;rabqNrjNbutM#JF^E|5)(|llnpx1QdbD}#MDCp|zw~}^JT;~C=cW*ez
zffJv9eKbUv<?|k|n(lIv#bVwv`mFl^2KZ`g(PqCqBlM5V938B^B{(+QwsgjEN7v&@
zaRTOK;%N?>kU%>Mm?zR_AQf~4yUS<#{MlHj+uZ^Rl@juM6Bj;-R)lM)39H>l%eA1q
zxZdL&QBucK(jD((kCU!Ct|~8?*vk*-35alf$;9QYMJaD-1xq`S@E(2<T5pBtxK{3r
zqwAxvHO7~iw#n(*Dm$VsbDXX8s-(ZaAWU>6PJFIzM>YmcANQFcfRdt`9u#|jFi8p|
zQ^fmXJ^ExqW3~x`#RAc=&vai$31`1?O<7tz`<w(54BoQUH3WEVbhIk!+RN*(%>W$J
z6mwdWjZ>0fDm1C0*JohiGMs6w_}G6BtgWZBu^1J{TRIbsLwFX$|9CINlGSyI|JktE
zB}z&-h>R{Zn3k0E%d(Ps@qKLQ91cs7X=dS`bayTkqQq?vOl2yUU05t=MX9@wx355Z
zZ`ihAO2t>HRJIa_<weh5Uy$V{MP?0G(r`|azONf7h&&anmecxmoBPUG;w6v>Lg9U$
z(*wf6^huG2Eq9PXVIHXHWEFxb0XskU<L3ZLv1;kc69)Jhrt(i`t<W7-cH}1b{tulS
z#tnd@M5pWfN9P!1iac1OLAp5m4wO>a*0fKL&qaLwl1#ih0XHfc3c^k^6#j!X=<&yd
zBB82=i#HITw9L(Jws}5j@rQ|JpOm~G8S@ScF^lV}>brFnu$D4*_Dt#ezut+_cOM6l
z^Z+WU%<~1qQQI$$pij#9qOmvN0=L?`1m3dSx@@Pr32OfVN0*Enm7S?a=Q5?89+*qs
z)p`&qdJsv98BV~{1(Uk3I8ejXhq5D?f+-M=?(-+%^fwOw?Wq4>Ak**ff0?=!+$@7t
zG?tf4{pD~4u9#u`r*779P!s4Vd|vuyp8&psYA6QJi7oRUe|5&g)2Bc7INsoq{CoG)
zAM5cy{n1Bfyfo|m_55x}i_-Bs1ye66X*a;!Lv*hdHCA1HF{x5k27bT14w|ZqqyyuZ
z6}3JOA?XSDs*NM=Oy@<K$pJ|#)AR+)ctK!0+KskO7OrYMSbbl;OQwC4p6xDSnXHV{
z=dJKMMh}XyKbT5NnnKlQs{G}`xl}rF3@5ub&&gJgVWN8hmLpd>rJ2XXE5f5Mc}Kuh
zG11~>T`aCYbx54SF=mm5r+6%z4?=1vaDM<G6XU9?CR<HU-9GQ}Cm%0k`m-~qd(WPD
zqWAyfk3To-q1$F`+e38UyB1UlI7_poYObs}(ksW~)B!!{P1gXE08Aj9_NA%=xALxJ
zVCCo1KK2!0@Npm$m|^yeB+<M;B=@AEZ~NiF8<#Z7uyWET+IKO9lPZA}|2(mO)Qg5|
zlQcE>b(UOC2YN~E0|(J3bzDbozKg;ta&QuGlqh|vYKukNaf<ud^6OX}1niA(-g)Ok
zZ?Be>HoL%gY|C>3XMPfjhU>$T`cUA?tzgAOqM<~e`wl$5y6k9b@uj+kT>ao|u8gWG
z?=d@QNVexc`6wGdeW^EEbZ5{w?q1)u7SvlRY)jQNPW^T=2_}0O`fepfcY=u=x6;+X
zbv$q-1;sOpj%+IG%=#!cY|)8+Q#P!eer_&0tr47616z|0$ns#bfgoN%QNC&%q2dRz
zsC{yug7fd-Ad(aezWK!kq>T<33dcTH_nZPQsbWj^ek&e`rl}b1tlEXm_!i49y5of}
zx6949N5j4(tUFRSql;It|Fral1B9b$pncClxD@nLrRb;!Aav^WPu>e2zmSZ(-xdCw
zI>DOmtfa@hAQlx3!u!0?a*sDvPPdZo1lsqsWZ`6D-X(NE!0PxwbZ+z1SY<W~JRLA*
z+UYreI5z%^)-GGIT!~EM&}T@NRIuIWYrH9Wu+N*KFJ5(^d$~V2m`ZAD>dfOxN`cdk
z#7|xW#C2>-boSSGu_OoG21|v5S^&~3Yfi6hIOD$M@7330G!U(%&RSy$E>QX3JV;m%
z+I45*@YAxk?GvF>me@UOL3J~Zir&#(3kpzrAC(k*QeD>ytp46Hdn{ZAU;+cqUjzFi
z>3uC;)&sJElZeW!GP52hbcA1)%vPq-6rH<o8yIXd)Z{>v)RqdIt1KS5;|EkmPkrlv
zgS`E~!6PZg_N`Br2K0b7fTwoG&$8?Az%_pVtOF1af_BGbc_gT4asvBmGLegCCY+ii
z&7_HV)%_@>q|5$*fm($FsrBOXJGJER9pMue^;FXCAm7V^+G&gv@kx0Tm?vvm-fp_G
zyZ^PD+xVpPf{xY9mU&B2RsDdIOdLlp=ta>`H(tfWN#AI^IuuPNsMv{L2srmL{PKOJ
zknRqWA%erZ?ZVP*R{%QE@ZcWu+I|rnL?v}y<D2W?2b2-3eWFpomU9~?RoV}b(96bJ
zCuRX>A5=_f<9UJV0e;kFJegFg=G1TDVgQ%4&q*acK)UV~<mprX{nk-faCb^dNJi~C
z#t9G`{F)BtzNmnn^4#nZB_-<iOEGrpj-L{dsas4<DQGS!?06thI8!o<qA5#n=KM*F
zGW~Jb?PM@AWJU`QQIktzr#=I#?U+)_{jrGmy8yMxJDN(L9t<2zpH%FcciN7VPE*H!
zW=mMT47;03%j^|6;0}oD+L#6{LA>P%3&+M1shUhp$Pa5|VinbQY*9fHz~29xN(%p|
zq*QPpQ;(Fvgn~?Xd@m(Ml+?8M>?0+3`J|fbD7vEBD=sV_ni7Y-C#4rOoy*OVg@U0R
zu#&p)ML5!MbqlM`iq=4c&roqD>B}_K`%eE3f+knC4$R-hX&yNH)8oJ?E$=#g#=-+}
zZ}$@igQR%Qfr_G|&mPpx(F7em`Fl)^LO)>NtxpJ4QaH@u-pfVX<axQs-kXKPLer^4
z3?4GI!I2+m7N-<VEF$T#?SSCX)Nx;+_m%|J3$SG9)bFgFXCEP_Z6_ryAu{G|9R9B8
zu97#cmV46_gZ7|&o_m4`M1t|lHd|dlQVoZI9so!S6WC?lk=yUZO8s<Qq+c#e4UA^0
z@7oH&f58ALJrHWSg5gN+@>2#x8_JX7(x<#d3P&C23D`Rh0!d3GbWBBgs=EXf$e#a_
zM$F_W-}VDI5+5EM%q=ZCwqa~LtujN?1O3_LKOnSE(#fkfsYM$rLxHbH9M}Vai&H)6
zs-ZitRPDJ4PA`D%3<pCKzc+U`x9HBZkCZ?f^Hm*!6&zIWymL&dimHt<Brl*_l1iXy
zDmVQ3E|CS{1?4U6XrJI+ctuy{EaSt`=vy!}UKW9q9(sABz}X55Zt4D=hn*pEpwcbZ
z`y+!wSoWDo)Fm5csH!P}BUf7F7sWwSQi$V)RTpBTp9eZI;%SjksZYS=k*Ua2ll|$L
z6OM=uuS@%xI1`D~Mh4Hrx!uT8rQ!Am0_?iS6Y6SDBtwB6nfH4J6ACbNV&JGRod<WG
zr5UmN`lR5K3RqtWEPhed?dF@}G*#!3bk~$U5Wdu9tBXpPsp-J91N&hFIAC}C?Op*h
zpy{Y*Hmjz{8d(40z<g3zrGwQ==hR2Nvs;&<>Dn8&Via~j%?10!K_F@AfI6=E>9NV`
z@=2uQCw;6q+<w{loq;xAbmcaytH6CZt{;S{3<FYDNu9kGds$(NNU{UeQF6?Y2D;`v
zo2iO=g-Uyk9oZG&Pm;jNeKy=ZbPrHwc6^$MJx;_Y?PhyPNsE$-{-Rr)TU!}QQeI=@
zE=pP(55-5$vvn+u18Vs1H1J5K?98tt#r!ZAYP;IbwMeSa+y(4Ckmw!{w#Yj2S0}{>
zl7Y<(RwFoVA*meo8_9w~Q;(c+#KP=`?oowIH2ANzj*W8#^c|M$VwPT)V7G5J_=OjQ
zDs$ONCfdJthJ>}rO>Z>H0V^q1g9bI#RQ(CeX2z19NIdee2f!~1(BW}kNe}xlgz|o4
zcOz6!B_;0>4w5lWa6w>nuS1l!;%smlLgAkD+`&~)QaX9#25@8GhoSZ!I7CwtS6`eY
z4MueQ%pVoaw%TP|Fb*tT4_OcEEXqqJGZh2yIN?mA)S&k{{OsS^(LLgzD5>X_+D-TN
zm`gU6y{}2`o|U<8!20*PR1|)eP0MgPuFJ*~$*?b*8TF}*IeG>r`lF5>sFs2a2J+~t
zJ3AIhHRMnxaj0}jBJml%=!H0Zc=XusoM$8Of-HtWk}*tgpA_CJrRoiP9p|*yfTF7A
zI6C`iGMlUNnw&h88=ySUvC*$7MXXQ$@0@uO-xuU(YrSd?bnZ1)<ZvQV)3qh;fOOBr
zOeoiBnwG6f)&^s~SbZ=?)y`-;I1B|G44;%-Pp&I1J0rd$_Kh~o=0p^R0{|!~aXIz&
zPMa&ttxZI5d=tf7mxmP^IL|F%*#Wz5SXUn%Q|rBB>&A{5f9k0d$B#Q<>}li2PDG3!
zHwkXXAAiF5@#9Y@-Jdve{G{;>m#^6=x+jFJrgw?%OG^3{n0pn+fMO56>wYt=0;^5*
zzz!va=YFPoTHY*x`*^@V3?Lhl0ot$tLMObM=<MZ<CdXU6rgEHhfX!Rj>n{nybG6?1
zdml(qW$e0brF7F@9tY1Sb*)d2jmOJ@6HSf(RMp$i7w9{{y%k-XRc+bex;8wKj`$LE
zrrw`<x2Pa_ag@n{%P1am^c0+;!EJl@{gq$L|9sJhAAa!RhachXqmMpZu;B9r^S}7u
z{m&P^x8t_>^Y=e^_nQv+o}LLN-KKp{W1IvaToLo0F>^%crbR20J4Z}AK;z7d99J+^
z<AA$H;m3hX3SP#xo3oTZ%7B3w|Gpu0E}8L(gw^>EC<_V=zGS-2;**F^T8Ewl6+eAB
z`Q*3blj4JF%xQ@EN;ROsUY_6RTKs-@+V_N|;COqp(A;A4JpavEXlfIQT;g;bnn!Z+
z@zOqUwZR4`rD&*T`}#K?eeJ0iX3u`@wbxz+%zGE_;W@m1Z8p43ZUl2?J^uX5pLepi
z@2sSVvXbsHBc6tn+v2X`zADAfhK6^DXHOL+U=s1#2e*j@!;$|TpA^%fR8D!JAqGxT
zWi&DD;U%x%dPe-XFI|<juy3@l0Aml~YSC5JT$RKV;q5pz=2NUO@f(Zo-L|K&OgOfV
zGjU%$eM1}2&1-b;D_m1BIA#3<Pc2y`uycn{a9k)lp14ylsASlPYaU=GsRz_z2d*kz
zoHsT$xOUXwh9M0@z*}q>HEKw%x;j@?T?H>j4;nOh5TJfE-ZFy*XR0pv-r4OuPT-Sv
zVg-|3^GOW{Ylh<{0eb7Xfq24;wYvS$|37zs9VW%ut&QUUpYQz6`M$m1y<dsINV}A^
zj^i!?LINS~?(T2g-QAri5QUHg5)zUSLXhAvFxFk&<zClW>#6RUo?#%&z~tKb-P6-u
zUDhRc-D|B!)>GmtC^Ew_kyBR5T_nx;0f79LyZU&>pUKu{()6@rTT=d1_0z7b4azA6
z)+4t?YuD}B`f_PL%QU~9b0kanYba?M&aYPhM=`Yy3TeFMWx;VQ?5{O&I3^GN?1znG
zC!Vul+08L<tg=->|A^%_IJuyt#N;1<)&Mls%*&4aG*^uSPaxtH&4dC>ESZj`GuXq)
zn@!U&fiEhP^&h>WqCIJa65`mRa7ZPew7eib@>S0#1z>ou<*oavK&0%HdaISV%K20E
zzEpJ7OCrA$8{8ep*@*`BvFER90^MWSV5A`x9T&^~>n2Gx1#}}|-vit|DGhZT;kA}{
z_gOQ{Ll`;z#YpOYKZhikg)5VYg&L3exL6QK4xIr5(>)qO$Afe>$~H|N0Ui*?(ro<9
z%_=IWca0C^E+WO@1&-p=7`S^Eoz3D<If;067*o>$V1c&`-n@0fb5=XuE%VF8PXt&_
zTFzinZUjY?l=QnNKWW>N?EZFapr`!9gZE>cKh1>VlU{DM<W5sMAfMC`bqCYsKi@<J
z;wgVRS{283%Jt13gVjct6y63{T<WEM68UeAqUa`)njcP+KIfMM){4Dg=tydRI@v4g
zrGm{Y*D!aLY~}{n>=SF79B0*OnZplFfTH0(BWw%%nutS{;9YkPAZh}DLy9+3QM8A%
z4W&{~ran~_Nm9v?nG%b*a4;FJXng|TBXcWT=67ofQBo`?T{5$ml18V@$M&S;C*jvL
zRnz!~qoomZ`U0Ma?6_C>Vj-uJR;6UC!GLrMmAg#+=ab^`hEYj4I(YcF+AQ;?Tp*cs
z$ORk_y!?AYVa+AU8n5F8l63C(t~3}Q_m44jB-LMQsP#%cm=r`%XRLKBAn70g>T7F?
z@y_T(9oGONN-jw@mi*qqR?y&#0z(hT9%AX0#uxIc^BP5^OS;D;VszH`&uFSlqHZNE
zfnh>Nf7`be6d@!RLZ4K1zwk&}MyZbz{vei?0#xzmS;I?9`+`Z&d^=yzq%Kp1D17ih
z(+rkbbF?REQsv8V!8ptMqoXg}1}bY}D&2vS6m=B*cxU^S$({zdbjzDyC;lSQJvn7~
z?YfPkINHa>!wtca;rOH5C9H2644Kd+)e3hfd?`ARZp6IZbRF}rOSj`dOnWOHMa$Fw
zl08A%mq@DUI4?#HkC3cYaLqII{@UhDbx-wZnXP4TH4~QFy>sT5Y(7;|(mr+fqrfL6
zheRlf%}lh`J9<fJj~X-S3Gb`LJPwv)?RVq+eotD$KKQ^!{hN{ggg=%DMjKMTa6QY8
zUS>(ymO#`8YVJ`<MN4!x-d10cKY)2eQA*y-LB<e})Dm1<bllBn)Ue?FgGbsBXo`;i
z))Y)(KeCqe2VK<k*0Un<bR#(GOr|LsXb25kWaD5tdu!}W5<~5EsfY3n)kNE|HCff{
zcN>Ny%|I~|(Zg=cKR2!}<S#p-*rTdf(K{2G^R4@2oD{>9QBu+XDydVB?n~5*VFMFv
z^-O+MQI(vjN}rCZ#u`v>m7WNn&P@2k1ZpX^tK|2y_L7V}f-ho>yMbqkgM><LC(tP{
z^^IXFNhiikx_EuZX4%Cyr6MW49RN@-Ax1K-uV&ShN55wq{g>|`5!J!R5Rnv2^iN0o
zn#f=UTq5pcPgvmy*kfrhPns#2y!_kA9Fxj|?7$sO*JmQ}FEQ0_pF#C10Q2_{l|Jyw
zN73x6SbGnC??=(<7C7K&n9U@c7P$G}-#N1(iB23FW*S3bHd~RSnG1$aS?=D+jjDk|
z>J8d|%`QVz7flRf=4d*b!EsC3+Qv5xMN<@yH@2+Wwk)kaA%J@-wSvN;HP~ByGBSIG
zlh<4kBouAMeu0cV;=TF;=SrP-Rr^zcc#_Gq%y{slor-K&qHY0!ZDF8A`s5cJ!C8EK
zPhN7i=G0%@-Gl8$uulwyxwkEBSh#F5#E`5fJrr~F_qK?6a7wET2w3;2=0AM;>~Uk8
z88#k38daT>b2Qi^T>a8oy+Cf%UT<#cPGU`yO~NMpjv0MkB<WU|OdS)gp7XP#xa!7F
z9y+l;5%y;rLUcV7t*J=qz(i&(aUVXkI*2W$J$AxO;3FGL-L?i|#E&Lc2l1H-hdCfP
z9h&i_%Ik`P9;qv9wu6cBmHSdLjAY=SAP`B`Cg^|7UG69y=D-R_$vN*8K;h*LR65?6
zW)F?O-KD!q&T@-<!E$gs`@z5wtlv(@Z15#e)4z2KgQ&s!V+@6oLZIV1osV-&+?Q-7
z$DbH^`F7XAKKG?{1_ErE4>T?6t=~LyWir%6bFn~(dd|s{Sh9bMj-5YY$zJXo=q4Ek
zUz70z{X<M*qN&caTm%F7L<}VGVbDBbm!a=o_QX}Qv*1!P$uQNc5;JtvFYarTC(M~P
zZO&01X3swA=xNjDjvo7Jm+D%}Pai*J+SDmW9XWa0?Ag=Cj-2-)j-%|=i9&d<D>6{^
zt5cGV<Hj{6S{hrLnj2fjPMuQMJR1B6p|yzp?!5y(2!Nm5In@`8B{;f{zT@{ofzKIY
z7y6{|8-w~8(U7-a9qR7r#4uCd(&eFun}#NE!j$;AD<JR#FAk4CS>N<%o9i0FV3ib8
z<;$jGbuMVA!6IoUI{srxARR~g&;EoI1)<!VPm1@2Xes$1X5UeO(@&+c6v`Tvp|bC(
zk|T?zP*BuO3;uoF_^>ZqPjj%e40iYt-)ox|FIc){@s|she!1k!B}*4ATKeVUFF#)R
z*(P4o3tJYy^ZtA9zxToW?|u5-`=8E#dBG+FB6aVXKpB@cB`<U=e)7?mo_pch7oL6o
zjc1>K{q^S`f8ojZ*R&b7X^ZNh5}4qL?Fq+iE8NIHoWljE`0=acye{&E940b_)d$zC
zR5ZlLdLk!Zyyfl#>1nnP3;{_U^h(`L$CkPaG07%2I_eV(f+lmX9^^iO`CttJbLnij
zJI=-e9FtDW*=9*z6T8w%eF3)$<+Svz4`4Qg^n@+IP#Z_{1WG#UV+o3lh9dZP>`zjm
zW9_mvxq=|_a^b5Nt{z?AKr<mU;KZym7&ys_jU5DK%tZw|!S+p#ZCY|JugQumStfx6
zpSB=MIbK)0aLiB$*FjNb8#cFXLxWWe__yI-5Vc~vYPT0vzL3`k+DZw}%jv(lJ?3`*
zw2llE%+^d;Y}r^!D&$NXOY{fFEnBjmpH!;YAdb$JI|jL@Kfu8!DOh{t#_#h=BxSi&
ztY-GI0?Zy{^8;``v8kTU%WKbyK;#4$gu9{u^aNqvODR5K=iVsl8VWJ@@BuIQ)J)9;
zQ7*`<E*yFEIVa6Idd`u@Ke<gaavz*<{L`zq=LEsfcWhjC$8mG!%{}^<xpR*>>6E#1
z=FU0h=2ebf5RhUdUhv*cQIZ5+*Yk#E*f1@Lwp9e;nxe?7uA0+a)r&4Pp4Z{cBK6(6
z$1L5{MZq*$iyDryIItG^aJD2lbpK}!Pmj84LJgLbay5ykyF|^{zMIdP62F%X{t)QQ
zE;`O@OpUjhsXb%MU!@fkh#MO9-;XIJ4`32aG;{KDz6fr>K$R3$hnJNTz|<FHUHz`b
zN1ohR(~EW<9LxSXO%Z+EyFW@mA|~v^K}uPP1lp>s^X<QU_2Vx;{_^d&%U5pCDfW)*
z0(AD+TR!Yk6rC?<{K_BJ{q)UGKmD-o$De*)wtV^5Yr8alhiqAvZApqM<C~&Lpqrv-
zOR_B4sGz`4Fx>#l5*4*ml@$xv2QqHSovI_s5^N_smIT2Rju$P<w!(MibP0YFgDi4@
zeNKhU*<&^^)D}^E$->gi9*CyYc9k!cf$K+SuKE49_nx`-u~nK|RHZ>GshG2s<)?Os
z_Y6bL-7vU4{Sk(Oq}X5H`FezxJ(r^b35Gj!Ek^u~KCmh$6BC+)Lgo3~_6_ZMNt+*S
zh(cr&sX6ucBKkl5`JAAw=#LsX)ynT&`}^u2y#ea+Z~eM;>(;Mbzj5`CIjgXB@#ia-
zE?)87(l5SWy5y@RUw!f2iZ8$T?z<IBKmTRbS1WfGc*Xs7Mpf9yWM^Ky07Ovdi>p3g
zx@5`Hua+$R60rEIMIV3s@xm{be*DQNpM3i9!Y`M?w=ch32wy+__~TDL{&*2C_+;tQ
zPd>qKFbn4`{AAH0SWLbzT?{imS+wvgaB0_w3V5je+OL-{Ui#g)Uwo0%bGoc52B@xV
z^mR>TL&n*C(ndIP9^xP=Z^__k?rHLiV2oa%L!`KcnRvt56SM62=Ov&Tj&eT;pVYBI
zNq1fm!hxU>{x~!8o6@p_jG-|1SYqG)IP2AzF^NWo%G}Zldga*p0oIFxBv`;xOaAzb
z8Ml12X~%8<Nk^Nwm@j%x0oaR*QChJAL`QTS8_#Ne^5`?pI_bDG&pPS&lTSG1xU)|>
z;gl0jnSIJ7M_;?e+5O<O38x-)^xU~eA3f*jGLD)v2W<VkqneKWqF_7L%ULQM1gQw0
z^o!x>#dl_on>lm3*Hr%qY^*;W|1h$jJ$ueHm^gac5l2j$J9R4T=%`Z655fM?M^BwT
zjr_sd)8|fw7nwJA+7b9=+SnP#Oq~1rZcEn8mA8zYcnloSk>{?qvG5i>UetNi{R*HV
z@ZNgC77T^Y;rKUr5&8m)>^~h>Zal~|q$oOe)n+i-;P4N|4HGR}wB@_J#A~=mmTr7~
zXu|u0Pg3kc>$=;n%6dW?4>kE>4Nn&^BPk}&9SC4;O!ce_?N4Q?^u(#NCWIT~=|(!9
zyqX99c5kgpU^82j98FOlKIG5@8wo~(Q9l<6L<8U^$A;G&{{K7T2a~@wlVuqi`-YN;
zo#w(`#1v$yG(F=hQE>~G9G-xPp3X8St}-;myt{_VWa)5_p;&CKM}j8?x8*`1Y~kmX
zmiD}1fCW28OGrQpLa`i`L|(8V<;`Wdz~Q&cuBchRofn0B9tcH`_{4N9M}wfPSg<t<
zJXd(@{!>j?wKROI{L^^Tm2?H%(*7bRpsi^CHY<cT1#Zi~u)sHd;xeG8@P~tn_6uxG
zE!od>C(}im-XZ=)NNNLh9H;dNa*L3pjkU2+A6q4tbRaB6YY#EEDcfz2)hDZy*;w7U
zcugo7_04);Lx%RhGqm?21kL$k`#r;b(Eu9`Fkyc*5D7%80xS~_#{PEHitU}ZhBzi1
zNF*3G90C@DVgg-|ut3(@3)UN|z4h2YJ(CCrsrnaktwMgqX(1po3o8T@kr2lb;=#s|
zFw6^u6X9@~n!zA`fe(+4C9uJ;2Qr6+EkdwTD9i!h1Gg%APIK1I3&J(?M`|aobk!Y#
zRe+dWaX_;0t>=&EC@N`h0NYei>%asJdnzd~7xs5{DVGIPKN5aCKK)(IaRl%ZN-GTr
zjw2e%Ym@O5Juvj;gX<0aM;H<%wetGzDJ6%RN!AA%XMFB}3jk(lV0HIs?F|tmW%t$5
z+CV&zjQk@qs)1pGN3OL=%+!xn^KOvNOAvVH9}9t8(djym7G5`jXm)aZ)-RU%C@8(B
zq*S`WYy5;&O`v`CY~+4fmE6x~Ww30Y^EE9fbU5m3qf0$Q;XRYd)R*EyI+KN6*VWNk
zui_o&VJfF?NiEGJvkAN}*$iEmg%wi`@nGGE$GQrdBK&?#HFgiKXO8;KmGisbee%^;
z<}X<A=KPO7TJYw(Z<YV%FMM-8cEb8-VfW1UaQ1>XUU~K9`M(#<{)(aC;tPgzODURV
z6Ujgg<&WM|)R3Zsl2XyMMeEa<<w6|16eo0Oe+?xC9$H1ab!ypHNl^ag)7F4>CPF%(
zb8v9{M#mDmK0CF(sga>m^|5pl$1<_Ad7G3z?Zw~&O>I@MZRP1$fT|8t-5ukBaq+_r
z4gKfnHE#QDAcj~?pCJD+G=DgS^vRyILfKsuF0D#{Y0lQgE?lSE){fhvRi#Q%E*yzQ
zVx_%ePknfYa4u46NdidV7Y^KjWVoUtwGv!7imj<(#Z)L73O!*ex>{U+UYf3oB@(q0
ze|9u;>D20II1-73y|-{Q0u$sNrlT+qXO?Hddo+}y{i#O^IIu`RO6mxv^#(j>__W}o
z*vv+npWFqG?A{BT_X!9-1bf|SJ+=1?eMaYBL`glt{BB}R>Cm}&4SUXRP*NO1dH+*9
z1Hjya7Yl?_-kUN!%w_7Tg4wzN%Tk%!^CVK}S4m+Wmf{<HL2AFJCe##9H*{Y{K)Shw
z{l~GtS@OeFssXeWmV%lxUfvQq1GBhL@UjiMWv`hYNK%b##GiREryIi7OJbE$NoUJd
zq;!^S3*7c5aJoK|&DOzg8IaiWMtwlz5~d+TGj&-yaA%k4*xLFt4o{>&gd@{_a+Tc6
zBk3kC8Y?53PK47OykJ(8ynA!G2)r`nm4#>U%ANgP6)YIk9cRm_)$nWtVjSm>XTr?L
z2e)H~QLi|Ep!||_r)Z1kmuv6gH0kN<O3Mx|hJvJE(&eWb{N({1Sc<y=Bhi5?sqH}I
zDe0>EOu}DBG1zGY$3uut_@KD=@fl$OsT2j*t?)qLkW{&)903G5iDdl0rvB>YAFK*u
zaSX~StWa8sd_+gR0A{N=e+>K<{ATdz&pK3H|Fro~FS%F=J-hARwsKbaThF|{AWPs4
zbeXMtq}@^r((kAHL17bI)r_?eE`2z;CPqq(N=Sh4l9xAKLigMR#|5L!@mmZ1(#-*m
z%6un@%t+8T>AahseeK5_R+xdT08a;M^RXDmlCMpcNb08&bmK>Z@f-dSLqJk=B@0(l
z<$eV~()9hV3|_#1eNvE8T{QBF_THS9G{>ZyVgb0R3^nf?5!`^%N_|0e#35br^+$p|
zEyodD6hfMK`fpQKyUIQOy+|Aznf95iSe>`V(e32I{!2F$6~}%)MzIhqmDV|!z@mlw
zWnHzLE$7D3XXUCU{0yh?(d6&|e~xCj$PGK(K7-p5NV+Q;ZTt<aFUoR(c<q!|zsn1X
zWebG@Z()1W0V^qN<cQY2qe_{S7`i#(yRE&n@StKSNa~te?#!AV1&KGm3AUMx%M3gn
z03bw{u)DGT+x*AxyZE$mez+cR6XU0DvekW;W|8J4uKs9<v})=BFo5A=ht2-Q6>i2r
zx<4?~d0R}`-Ev%29CPQz61VTvZ0Gk2;7di`r-7oEqV|zj&d)vy9*W|uKi(_Hj86N>
zG3A%XdZ}>$eKshgpV{H-JK;Iq)8fM%!4RdV8>4>A5m9~gMoTUTMLwq$bhiM0X=%Ov
z2X+mWoM)TL2kD0(^_D_u;X%a^kQD8D?$p}G(y=qi%(!o4qM@*)w9bCPl)y80T+H^=
z)wz4)$_s0ofnfobI`vnvz++pxy#j6tn0@Uj98aPz7^Vu8RJ{kh+kS<SO5L(kckCsj
z!(2QSOa@X<E4t!-F(#S>n)bx9dqe+x>XqDak0J0V=6*%jZFloIanMnYjZgm(PXB|6
zJw@-`?r<f*qupb6=#DlBgZe2dl}yo)tmRW9<fLPx-Y*opbRo|dcmso{!36jP({?|Z
zQd1gE4n6RzTT6=$Dh@tLfryUVHLtqIn~F@{0naIUR0sCFg5O|DhHT_zD=)iSFAOx)
zf(TIQn|F%NA2kp{B^B=u_c$FTfW?m;4oa%s8-T#Px6XQqj(r(xt8NH!kpLEkp752c
zY4UwQPjHL9E%qX8h+bWrE6{6`-rIA*F!Pw@*owX3gfJmwblQ*LM7}qQ4Eje_>Hi1&
znY}k91tTWtj=JeIQuxk={`HzA<@ua$LqvsbOz>}D@RW4j<nogQ-UfX+Nv(IVF%*Fl
zoNvb|zOs_Sb(^-kv}k~U9XXm{Il71)O@Evi8;5r-*7A6pgjG@d0BK_4s!xCuTRMnt
zpv#9!3MB1cNyEpjG<9q9DIrXi%Tk$JIvm+-JvYcjs=}UA+Rw1fJ=h0<_qMFBl5#Z@
zes*;0qlu;Zp}yGrPjFo1#{PX$P*O>A?XD{s()pmF>KUPEDn+qi2!<^Vrco)5V&FK%
z<6_-H;ENsCQCfIVF%%>fox&H>dxjLH>2twT8R&Bg0CNvIYVxvyqw2^DnQ$}2E>Yi{
zZ!7^Ks?u8h0$V}BWxP973Oy^p)TIxbz1q#+4<zjmRPyE>pe$cbixCdT#=e>_^6vMO
zBM=&tjeVIV`yymNAZEUwZ8$&os0cnOQBo|+iA+T$eYZzRd%?5v{rxu6m}KtEk%zl<
zA$kC&Mbwq{JG>I%VBnz+p;PK;?b_a9g06U_Py+&3s=O^1&u3~IO38K@ddiy8!h?z-
z@kw3v6&&c>0~ZTk-bTu$1`HcC)Uhp1u`x6LM=2%_!bnrceCG&%V4dJ5!E)`9ntpOG
za70NJ^hx&s!*KDjFN=;V-A{s6hK(HkjVRgLiwPeV7WE0k>iReBdj$|{&&=Zhig`fM
zv`QuY37*l9CiPfSkCfg&hviRYf0V|3!V8|C1j{4>hn|7$<s=;57B52gRl-3W#Js7!
z(n9)K)3M{1E_>kG6VF-3V=><#x;eoXPSLQO^QjnzJ!bLB&@FG15NzOk+b4!XNhP;<
zO@$$)sf)mW987f?xcY+07YZG_a`G<agA^9efNebSM->Z0ODpXK*xy-rFb>q}3oeR2
zX}{4uc%cp4oF6R1S$j$X2pb79>1(z_sK4qm(vYIBd$5O^?J>hc<?Qlw#XN}apR*KA
zwK2Ny@k!@T@%rtPHE}jmpXS)cR5pdo1g<hnA{k6&nA${|@ki)}XoP8;^8vOnaa@`2
z-DD4GD`>91Xwq<u?zsRpengn^CEw6hLl{gSPwdw06pQW|w4Z7Y#knSsQ@ZJ89T$*{
zW(N;Lqojq)E27|d_CC^@Y2YwKt4lPL+yh5l`rN0Bukza4(1}ZaR~&n<h4};!Ibe3^
zhk|>Wdq7P&y59o~olTtnyJ0&^rc>b{upr*_1oqbZWqMT~`c?9-`&%h$Dt+r-r8^x|
z1$rqe%|57N|2!-w#m(Z=e?p(sQ_@ms53{*&v?0xf{k731kWsom9?CQ|1w#ogFz&cB
zCirS8Hg@g0qLA0H@ri}e6NZnaD?*A&9edMBQHsS(r8E_f)A3QywZp?am?C)CA(rO5
zPtJr}0tW~u3(xS~L_NS0IGA<#L*kRVxpTS;(K)Jm`ZuJuX%KLXoaj2sC;c}Ypc_h=
zvEoe+b|OXhxkL0xZCAXf#$T#c>H)AIaHJRAccW*8gg^VNU>3WcYGOk1Muv$@{RH!J
ze8T;+l-Z<&ir5J@?hfh6j!NGu;(Ri@*HQHTN@t(5RLwPhKf{xEPf5!@Y4@5;T_#)|
zt#2lkJY1+c75n?}RCM%(*IfVZmco6h<cMVBBO5h{*7-uQfMecOltmHhxT5gu>_`Ad
z4e<_*uDg4CzhwdY1v13W$76<<TkHAb%tP282QbJz_K%?;spIC(s43qcuw~7wrGZt4
zmf@I1+ji~6<7*pY@Ys9%i;a3sl+oiVt<*nAZWHy%NTdnw2iC}!!G)>JKc<$V`x1JF
z8e!^VBR`ZhqxDKmSdxeY*^4)rda?6f(jmVR70dxH_9kHITkn|_(%E|ffZ^^}Q1>>Y
zlEU(mPue%S=R%Q&6rBjilA#z&Wt!`nkG}HRmp|DeYJw=;mI_iFb=B`u4rr=bMV;`o
z`zpaV_yy72dU71C9M<2%yFYaP?{ePqGLQ`r;N{|7{cPmn<sblvKzF~HVdy)+z#UYA
zA`U)DvA{le1`bQ%0Z6({8dThbrO&49bUahfvdQj|bYtT`-jA%gXz4&ny>ilCq@Wqj
zcVwUFzGnq4bn+HE=lpa8sYhWFW1g{H+1zqmUu$21Jt)^3-HW>?_Kcn4b&txr=31*T
zx(D=D(jL$Z#!NI-*VvS5nRM)J^S|D#i<%<Xs$t*Cq*GMWgX{9G9i}Z7$Ur9*gn=(|
z9(>j9m!NG%LXkbjkNri?!-X!bH4umvY2LnOLbxdfp&M#5Q*#sCmV@3IZzz0H*Ug_*
zLDIlz2+;@ibm^ERS1sJ#&=4hN8^KV=EY(O^Y-y!_0Wyk8`ox|!PW>rq1>%kOcDh>O
zoiW6OG4b*5xq8lB)jP3x8Gk@ef<A@&WRt$zalYge5I2>S6qfe#NfD(ju!#+}%T8ve
zpZolezvV1Pwp=A|cF0?no=%10EOp#Vt0YIt=L|WAfm4sUhY7o=J5I*~O|jAe(G=Cd
zj{QI<V~`FUpvw?6xI3oewSln7PjO83<pL3i(kcfDLqJkE#KI-r=AKEVrwQ0lue5ML
z&_y$+3b%&>=tOzuJ{&r4r3Ny(kNu&igrS2g+!OY!G~DBEV!(m<z$e`cNh6aM*qYgT
z4LZPF#Lrx?&F<8d`}!?w>RYDo{Y3eP`hF7hTL6Vh`n;v6o=@5n-M=%1v`~eAoQEGx
z^kK;?(>(90`J3B1O;<7TTqRxIw(Oos!DxL7<gxj}MXd$wvM=DsIu(JFjV+<&V%tlj
zt4e+6f^>3Z=J-`$w1r-t!F~d`D4K%3>zKHYBz%SwW2!GDZLt35NB3awHyqa;I5DZr
zwB4>fxZEW`(k`v*%3zG5=s1>9a?zT^d8-|7ypjIlRW@m|uGpPVk+F1GCS6ZjihAIp
zf1moRo4c=@q?I%Z&0Fip_O~Y|vD`5b9dVClc9}cR3s$bvy+!Yto@SPnv$ukFues+0
zf?*#}E0ompN$odAN3h%pgL{sn;e)PEQO(C*{pydcx~m!_#RLxHto-oAyiB|<9f)Ij
zDLwm|C5p`R1_bjxLW+|lmxl7iRIkE-VNSmN_J<#Px3yrpMO<y*HBNA?i?(=NtQ0N4
zPODDe>f$H{|Kp>3h{@+q@eUh7$4<ZoP=k52Owj?7=B`WDW!Vs685gZhUa~<ZL*ABF
z>I;M9*O4{Bw)qDSB@YW11(G2th$F^D4?S{~TkI|;tr($^i6`!q9kcz#3Bgd9PWY44
z-ceK|zvS3zx@Ra~SZ$ALQL=ZBcV+hBXL)+hGN|5?%21EmprqYCX#|zjd2>vxdna5l
z%1}&V(q*syuoKgmkul4FbCM>Dt1layjM1r35X<*DUz9m+VY`Wyokpdb<$xDpx}y8}
zk>J?|;)&YEH|2tCSQf}AkaXZS8rMf+A6;@vI3LyYkJ)hG*s5YMw|lP`5|Zk#wv>Zm
zE?yfu83)rWEgKNv5?eV*|LKGGUwGuI3EmFlncLcVu=aZ`Cj|rO^}CmC`%P+?zm(Vy
zfn>$cnz7%y>fL)PDHo|8{edI9JFdpyA`+-e9{-D|TKP9emz$%NR_)%Rgg%aE89lT6
zZqR2AQBuuuNR1QACTc-RRrw=q1yx#)iPM49-1|P;QYhGpq3B+w64`)H{oQe?cvY+^
z<)_H0GBwPjyM>%7>o}548L)N{*j==LIlcxX9hS~4a8+A@)kIwwc#~%W@(`Ryc}m*E
zF<jNePa$6Y3j|J<tUg`eGX@vMUF-F48Za<1grf_NndfuGRmTHSKnbbF$20>IL+opZ
zj;`pU!Arm0FfJ6Sr`c>t2T@5GHg*^&sc>^26An=c!vv50SyY_Wr-!o*Sg@bELDOV=
z>n$NB>G=))v@9We*^`6>eSe|!weKez&pu~5;AW7zdr(<+>W{81U@Iu55ez;s$`_q<
z_QT({@uq9p<wT@#o~VC?SI5TK#x&zkr7^5(iiX%HG&KkDP-V&;F#M#Ivh&@swGk{M
zU{fE8c~jyoa1Bh{&j9Zbt}a@V?Os%h7VvCp8s5-x$pH8D2N)71l^^jsCBjiNsk*>*
zAbR_557Ivz3pazuyy)7>hO;P^iv_q?<0E+-XmqdF0q-6}O{(B(`9kYQmyH}&O4jTF
zqWkoJP5aeV?(tO?pCSR10XBL=UNiXxM?~Q}aD{u>)n#Y-G2wFIcLnQ8ll|4-SFBZu
zve7S_`fih-2t*$0qoiIdDC{iFC4(WFop9Zv%{jiHIm)i~O6qCwEWbFSF4Y(x7G-O~
zT*4O*)P=9(Z5vowd5Pe%VCZDKzntx*Z=xeFVN+7eF-=$iIyyitMZH8Nb*{%QJta`=
z%enHxgNLCYsoQpIEe*PqY=Rq+%7pH~9M6MzwoD0Ao+!K3-Ec8h&V@V5FkCSGs_N>+
zy_<Vj0FjfKmmC*}YHfaIMx?4LnDCrZ0K-KOf#^Q}!0-@OBJ~i8;~0v48Xk>pw<SZN
zXppT-9P=YQC&CM(nBEOb5Z$8Mo00gfukr6*qdd{)lW9n%p2O(AS2?MR<)q2T@Z<=F
znsD3de6CnDWxk-Brr1-}jsr1SZ%u2d@+GMTib=+)KqfJI!n0ihFt4;uHzXa{=&U&}
zg{0zYnO96rkR-9S%Z5D;N=cE}@?FuMk)~<tr7m1Fh`YTH3;{_Yw7WeVz;1YW?1?D#
zglrY}ykz?U2!Mc%1xIi1aF-qDRlY^Wexx}@7uGrT4>W<8Izba0J*T)<$2T{OkH=E=
z=|*Dh;i#hj9RHoG+*gaeMJuBwh6x|J!qLTFP7jl&Eb;WsosQ#l-bDKAccVnDZt9gl
zzI*)ay{NyM``#c*s+OaBKx{h1AE>1DBNKEad&LiVQ^WSjl4dz7meb?a#5c73)=ML4
zU$`;Vc<qAQvxm1d2dhI5KwMc0?8&MG4!Wqez4?YL1`J$n@*dvQRn-I~Ee^&fCC4Dz
z_LG^O+BmxQVQ_p8wo`iWNot9%d~46}KFREeg^o$`Lk$?PSx$mzUC2xB(rI3UrpVM~
zaQ%h7D=9!%4AIeY-pHc{-}%XL(P2S!MzMzy@XtT%7dHn=+N)ll!$c+z7IJ3cV=_os
zed?d}FDq6-U433O2omnC1>5)ImjX)iS;5bqJ>x`g;`Rafq?QWL-tT7+$>xxxiu~3X
zEOsAz%TKCeXgQMA(Sc|Po_CU}4Bu=A;`4XLMwq7XlxI4%+~-$jBaMN`Gn!UvtJhr!
zg$X0zyIgF0sF74-()Fi4`}`9(-tcLMfJwx=xxC*1DuN5)%1<$wp1~WbM{wYZLF{c`
z7!oCwZ>}aPN)R9Q85o&CDJcX$MU(GrYwO^}PsWqBKr}t~TbPdyR%xZaz&qmrH8Dr7
zs|gS#+v3_g#xRlL!Au>)Bx1qXKfovLy1%=(sK*~J6i?T`>?&IEg=Xxf!g1B3zjAHM
z`fOU1s}6Km%l1RJepyIZ+dZ@Ff0k$TUK^u(AfUPRycozX-58v)(p5X>PYN>QuU!I<
zw}=_cWgJymCEUbQBSZksI_CTd!%}r4KgKjn%Fc(IX+Ql4TyR%1vC4_83Lq4Pua^2m
zV-~9_7&e(nr-RW;el)GYl)b?291NVCg)QX<CUARd9s&pCAQzJkjZZ3EQI+tL<pZK)
zm*Y?mr9}e)II>Esqb|Jkn)}{bbPvYYT(mZI&M#!xk-Z8jC@<U%uMZR%3}3Wed;14h
zOrYp!I2Nd@XTc|3RnZEn+v0P>kL@rN{rj^cEEfrJOzg5Qlb2d==NJy+7qUuU5U%z0
zK}&#T6-zu;?h!Kqa6RhLvXZh*R5)FQ8BC1DCp6vgEoN6za^N$<S;+>Fw9Gm}6HiHJ
z`9q021x;6ShWq8D2$gtR#j^Wet((Y20ba@Ct0Uunk|Bb@Dw6d1YaDqn<KTjSfbN6q
zem1JRQ!C3<BS}T;e|(J-JhlI>X1iOqMaQkM2iXeBcFitTc_f}-<MkP=w&tQWsq@xg
zaKHD0S5o5y1|N$QP*60!pcuR)e12{{6$7#*>QjFQpR}SqX-U->)_;juP@J#E2b<6*
zrIL>WN9}Ew!8GuQtW!pL6ECzced`B%EGf_ES%7>+B`y1;9Qd%qX8z#H;xDg%*JZeB
z4y%e>9B>RchSf-AV3JaZ^v!o>ae;r{rsO0g&$~+|NBzvBZQvAC)Ham>b0kB1ZB*13
z#qu`%ah9I_jWDPY#X&~lumsCpHa6;=37(7ZL2{hZUT@D}Ff>YfWfj+*za%{2H`f|a
zdYx{VMaQu_AB|8{ERN1OGAD8277MfJ>}!WaH1N#^n!>v(YrG(=y?<6LK&Km+=wY*0
zxox-hX_?Q3!{Z;39fP0W=p%*Vk;z}Wu<3WlVgYTB`-0d@XzzW+w?6at30Rk^!|2{A
zMfV&R9kblY8KSHjzy*aAG!a$Alj-3T%z|%@`R@2g!|?lf1IHjW-yIjFxo2RT(pbA>
zyAsIe+J_TLEll8~0x^2}_qJDeKTwdy(m>J`<D%t9*iYAjlEN?k&sS2XQc1aBORFOe
zPDw4yQ=-2ZB5IW&#iR^<GH{+q<lX^Sh+HI)R2)rAlMOzfFKD^N_aDKq(c!`HoF8nZ
zr^cxSi0zj^xUumP1{uh)zVlmkquBoZv^0|lQyweJ<m;b=#Jb~KF9%?+gzZ^CYMgY{
z!g5l49;(KDqbZm%-;s0hCX~l>BHCG+CIBUs+y!HYQ-?mK6!S>6FD68o@RNMEl2+;|
z_H&Y*k7r<E1@<6`bMVw>KZdAw5F~{w3?zl<Z2Ocj>-U<JdP*u<2YHCmAt0%x3RiV2
zDU)JHf*uaaCk0nlm+iune}fCaV0jT-S(-b4i_POe52bbb0h)WV7=191M%i(-E=kX+
z*3OTw9~bg5!J~e5#hU}Y)e#)S3)YE6?S~m)?SX1c;8{l%?HyNvtk!cu^3d%ivhN0=
z9lli%v!Vp5JlFfI@Jku$F$;`_x$49)8%$-X=y7YUoCz+bnp1#TVDvqZ6Dmy;fKRGh
zprq^}_n8XNLwA>s@CRbgsTj)>DXm;z(rw$;7EA#1jE$a2PTyK)LDAZ46uJ+14yj<t
z+oy)gO3Je6lVaQI(n<#he}a;ZDM)*d#rFk^#|=9k2|}dJ#e*!9$@*F9%<Wi6-}_!w
z0t|i6t<*tC!SrB>snI1v#I@!56Gq4W&k@UA?Y?2LMncmPbvR%4l}_2#UTqA9@N0C^
z5?9O18!i}5H?V9eVO|CKy0;@Al?(cSEW)?GHBLR-Fzo%Rt}E8ob7E9owx;Ql_k?^N
z9&JgrKu#r$?fW@nnB)1RGdS+hn`JofqV6u92uk`y9(lKiQ%VM9))tcnpxDYQkxV9<
zS{4x{9q<jp&>j=>I;~R<?=C2%NZ>T6UYGm8(EFsJBl3eQB?X<Z9aB;6KRnLW#WB9d
zI-&HxJKTNN-S^U1xQ{p(jxK?MNIK6e_KqbFPx;%F6*iDGkifXUl=YG2s;92DRekk!
z2?jGhG4#b7OifWgoDoGTSCY`1E~%;)JMBHd_u%of{Izmt2qs?t2%M5^wH{Xk4$b5j
zwt*6M0y`BlUM>zx=0#4u0hS*hesli#NK?%{hGwXF-TCMUmPtHN1l9DEp_1E}8roGq
zngs;q{BdU71-IUL`)yBcaxfPZ$qz9gz_SmA-EntL36;$~ME79s4|d!%G$id6-G`^`
z7-Vpdg-~Xp@G{Lr*)W$PC6#pIvh5&@`>^+Y6$ihA(a%X5$Yee*>7AdSaqbGO?Y1!Z
z4p>_nEfw+8XZ&P1?&`S_68OgJo`a~#-g#?}DJ!9D^Gj6oKKk;G%-N#^s(*075cgoG
z1<v!<i?~$g(yuy-MMLQVl4_(xUWR~Z&+{U4NZtMXal<my?MN6~mOhyhWh3{1&km}I
zN?IwWSaRy9i>HB-g53x<FUl#owk}{NCej_ajV5uMXt#n-T0YH2->ov1K$X@wNcc09
zv}g~yvuM#ZikqH)^uc@YJ<4mvKy#OM!rmZq8rUASJTKd>o^Rg?0m-Vbwz};Pq>@3F
z$~3Yx7%MIkq?#V>P+jx&gfD@~bZYBAcELf|amL~O(eIBz2)bWu!n!o|nCXbRu;P^P
zl;?69D83_sGfE^BDZ1B_B_w@ke3U!v-%4J>RX-XNWw@tR6<pH3N(zj}%KOhyN#SYt
zjrhV<3-A<(rj2V2?1BQo-dnKGoj}quN=o`y^GVszjDk#B-3=H9=%{NJY>n6Tjyo`g
z2wvMuw_{i|uq*1|IN=cJ>mux|fbOZi`=-A~>!Z~{mL)mJ7$!XPOVe>ST@q%pjiGSV
zcjj7GRkbhX^xT0z0#cIDvgd6Nl;`$dO4IkL)^<VL_1Z0;$!1=2WYq@#fye2&*L#}!
z5|D@13zJyxu$v?so8vnlj*GG3rwW$6r%MVmuo=`hbI~6SM(SpLDWRR`u~Z(LK@AjG
zVIl%aA#f^#s(u&`^)D9Pqml|<E%Jb2h*nNC^RjCT>h>G!h?G+4n>xrtKcGEgaVc)$
zs8~UZ!{_AHWk-bqnfh>5xq(yQly!#g{yd$lV>leF;US1U%+3c(DPqg$pLzt{WlwuJ
zs689?sc|YT-~yrhB}dfcPKlRYMTP?<Cce9zZLcSbG`3zCQ<I85oVP5~66H^4Q*rhU
zi$wCcRmG-o>JBLBav&+wnDRx(F1Nr%C~8>XiPxvJ+WrGP4ET**^Qz<J6C_HCzvQiO
zknkrc>CB=sxWEbGaw9K+18S>k$Mtn&Z*&rTQp7$>^ZNpCIF4&?@SvkuSK1|6+M3%Y
z*T*x-+H515KPq+ec5nfgjt*6WkIGREFPpGx$CbVFLGRNa#ZNL_o`HbrQdHk-5&7&c
zgre#DJAo&;-9oXTTCy%fJdf<!tAdsQ>&S-n#=JV;-ybT<yE=;nXX(g*pPCPLxQ|bY
zyTfV6S%yl=F-@VCC5l)q+H%e$q#CG=#)k+i?~BKg!6L|s5T6uw^%n@7z$fjlyJw;&
z!qYY=^|yG@wJkLIyxw(v9a0IE6!UL-7kwb`#1tNV6^r*JYx_oB`uyC6Y(p@ZM1vol
z{8quXcKthqU4|jtJMstH*3_?O_`I;nJDD=c_9wFU(Bqm`6pMDFOj(AxSJpMhvgA%Z
zx7%>!obJ_w5;5F6U>&%&FHIXcEb#C3dBqj=t<Q|fh9jS77~A(!QXnW&(Q(zK$AFS@
zzPjYp&kU=e*mhBekP0a0t+?L+lP+S7?e1xqHQkHvxtja26%=VWaquuCBn2hyZjix*
zW|9#?2YLw*8oH#s00EQ%B*mVe9K&6eADD`kKyH=}ZYhuzv{@IGoNjJ;_uNK04uK1>
z@Pw}v-rhXV*Th9aY&>;WzMz_g7wWKqXWwjk-R$x{%2ewYJqx?1dTmU=(Dx|di@?Ju
zYc}!)_(cx*52Tb-N*Rl$#Nv&0C*8ZeYt4t(Oi49|BcB<(C3{L*2J8(RLP@6m^<0Vz
z)J0-k!(}fodh68>wp&QifyX+)dF;>R6ptN_P29<ecuGogL>)v7k&>24%2EFC)K<GV
zxGE>is%IIwd@e6+x(X`{nRGpSIcfSmU?nAhjc_|0>)uvHT{VlpKR*}DeK^D<Z)`VY
z>;2}M1YGY()s*?7rntXf?8O8nCEXVr`&qV^)k~7vt*o#bDk<uZgq~>SaJTzDOE?=*
zSLFA`Mg7?{!=8M@sUwms!_`JXNx^9J+89q8Y3k|=X8ST}a4{fy8a1&YG2t^|FvB{)
z2AJL){Efnihu4+zbd`KkQ1=1vZeJK8C9N>`;VHXp5W~SdT8?dYDf-H1o_hU_XHOuF
z7C9zy$#!x<2cxDvu=N5jJBqBU;_^pNVWRc1#^=Bf;IA!>v=};S8I}R9m|Gd!@2gU@
zugyp1?adAF$XlW?LpNmOndxuyhJq#<80xvB`vV#5wC))1jsvHVqvKq{ml&Dy#r*Mi
zB~yl=9nifi&mYBW`t{s6R*Yc1C_~d#+}nI{FgHiEEX>B|IKpwLq#nE)r-NOZKg3E3
z2u&}7Ss0X(8loZEU5|vvk7^u6Iyl1f&t2Yyt)TuuWA+U)&Xa8$owCXEs{GY8lN+dc
zUn+U`mkHtK5O@_->c)0mGPH%GQNO!U;#KLR`;&lt@#esLk1JY2uXPbj1C8}sY&gO(
z*<)YZS>!dO=zhgZWnj2&yXME3$CL|1S$~*I1|o2c7Fd$*iCH;rA`dpU^OMD;$M~XT
z7#E_Z)X0wq)#jNfDN4DsYkIKz8vV7}llI)(@YLP*AS>?;95uJ|SwBrtRI(9d6fSGx
z;%)99Bq#kt=-5L_C&i^|mum6b-aCKtRa?Mu{vcUL{Lk2!m-9v5?79U=)Yl#D7F6=t
zi+rAT?m<vMWq@A-GL#F38#2)m*DUDLRF9%wFtooHi?oU|=a0kYOz?-vA@FCCFYplE
zV?$}eO#+<2RwtJI$&oaTI*P=OC)kF!`N7;90MHbx?_1|qch}uF{*~I3Rw!wB(hk=e
zTp5&6aP0iHhpS?KFi&hI>*p9MdG4>aNQz7b54d@{ges)v6i1bdIcdd*OS=rc_-|}u
zN<*M@<WdtN)HNqmc_TUWMwx2x>BaZm-18<XsR$rn$y?t5A6Uv#@e%iKgow#gQJ~8o
zMf8<`Rc%Rme+-i#8No}|CeA#qxl+!{^IRr8kyP7}3-SlEU{&w57^X2XZ845?GdO@b
z66~E*gP5$0{4`Ng_{G11lHSzglX8K^b&fQ+GAILwztvs{lPe0er?|K;g(N*dCB<6@
z9;m87T;}RLpXU{A_fEy#cwW?x9rNqs7p^yS)%ko%n6#6sKpz5<?|XGV$>NHpdcC&B
zTvN%hkwd4yzFjPIX)?IL5HkU%Dg&oKfUG&PX0JQ9=3mW?er($WyNxx(N4=q{T{=)6
zNL;q}$Qq)g>T9E8C5loM&DKo&+8R{%5wb@z2JXaR6e%7A5TEp~(4N$Bi<eg@X=K70
z*Blh9ObG4s#o{Xw8ZIcqaWQ|oA(lMn7gv)=+lIjdTSj?RrFjJi%eR|~E$58I6RB_r
z8v-{xDjSlWzp=Wf>4b;w*r^v;`!cs1{aJnh&fr4Hy4d*GAs4R}cumO{_&g>H$IAM$
z_U#88>`Jc`)_it(D&82wfuU(OGxxzYJ39(_FJ*qWF5?I0BbUY3N5zP-XSsAbK_7nf
zYDKqk@jx#qwv>i@ZT~j0+~^s&UqzJke|+6NSa#vkp7f+_bnFVNRTz~1>yni(2=i<G
z3=;`qzauX0XXbtD2oNpodqliH0O0|=X=wCaSL;%BzE~)}JqDaRaO<Me7TJy_wq97>
zSTRl`AzO)Y{pBs`5B+kP46Db~WqnC1bZ@JzYKrPOqB2m1B1N!?6gC}g{o&S0Kv1fo
z?(8M4@Z1TSMZ&6{j!rNE`-bo8d*h-+_^4TrKmYHOCttC_0#13a0RbE8foo@fI~LQM
zBYf544!aLKvmR8BbO@D{95sLoW|zCBRTz|0suyj$^9iPoNhA_j&k7cwIeHnkPQs?l
z2M)Mxs-jD-B59)SY&|PN*Czu}f9A@~mf|>Tju_5$4;bDfq6B-&V|i9TOFVRYq~YpF
zg5x4Jk^02V1s53#ha%}JiJJxj>`DS&W`Va9D}Q}3lsY7GXS-^thOEncZ<iD&Ku9Ic
z%{|f_2j`VxVwZ2SR8f@kd0;18&H-OgjPG61{%mBqw<uHFSo;7bhxjXyRJy5>q?sQa
zh^7a_DaXo*j`8gkC!Bx%-PeqxStgx{W+!|GPwxR1llB0}iCLq%6i0KHHTlC##?Qr?
zp2(@TroYOBf~8mlu~+taD0!yc7QS+IZ@Tmz<g!$KhGvrEr@bn-@78Qt1*)loxJ0O=
zaFN8EDd+V^qQOQkecfgyj{`GlIJ|vnwGtd0w^1;2x1{5ZX*Lk0uIh3X(bacgy8gi!
zXz;lN3&P6=lFW^u@n83M9qh1ah|IlsMX!|kn<ZmV)uM*r>Umz&yF^7bewazK(I7{$
zGvC*+2h+j24iG(cDZ1#Wdf|@x20zu1ZU|3afT_~9ob7L9`ZR*-gI;7TO~3~dwKAHj
zjGf>cn_$B=k%s7$FW}6&_yQ*URR$6-bpv}i1HUa>-Y(q5(!t26J2yKbx|r}*kur(^
z?gEh1+I~q8v+6Ny=9*4DrwWGc2zjK{fUhXTuc)MJ>phZ^N3`l$cs7V1a1b#hN-9<=
zX(+jfmj;#T*$_?4=OkIWtGZiGs*eWPn4g{gnkW%R|A2uqgjP<G99!?Sx18IQ@F##|
z;S+u~VZkp)huOM@zTtfZD~V@Hywjdp9`JkRygtjhV6ZMj9lO+V+6#&TEGr$#fB~Vs
z0}SUuJk{H9RLRI)f~pZSFB%xjSMC!WRW$coFJiH&0@IkdI_C&GIxI)Z8)%;gET))#
z7k+3#((~Ifc%o|_)fJ%VU!d;3?Uov^kORO)B6o^|8^_79O#vR(Jh+Rxvi2m3p*fC;
zk9t|vA@)89G}UB$0US*=^6EFU=wPbF&&BH>mdw0vz0?#4Cibift(;RLW2I{LMIU~`
zj6i|~i_m=i&z4q-=?7g>ffRKdTkO&u`;JhUV}f_Liv_F*B1Nb@VG{xN?-Z@=r{Q2@
zNq;Q;vZ?WgEyDkxT0y~c0B^2zSH#Pz2bTbzlb0}^-$9O?hJ>WTZFJADA2fY29QFP_
zS>#HUE|)3VW?sPj_@-pHT^^75(~WFAeP^el3;Rf#zuzdy7k<EFu=~YvF$SIjHa6yM
zRZ+}M=dnR18StFay~twslh~Ju-KgX*S&Lx<5suC@U%6H;@(S`(*41*q`hfydLAi3y
zkaR_TI2Ggk)B`(hxYQ0l++g2(u^-PB!m|yh#(_&s;Yjp_EwW%>t3a&ZBsz*$qvEx1
zpgVzC^)AF@wBAYjW87<?q<@9>q>g-Nxo{UibJsi4{$?;CZTh<Vn&=_~HiitYf~D@f
zCWd9DTsU&4V(S7p0;QD>7}!u>G3C{l(tfzs9K#&{wZa$VH75?w*24@u6Hreu`|#@o
zco_JtH#fUy;oQ=aif^zPdZuvv98JfYAK7lHiVgh4Qm0;4c_6@2D@W?mC0$hSf<05I
zJGNoWFL#-wK~*m$#V@wte0K~DN}6C<>XciaU$E%QodvL0CGGPHqPk!EJ}d~rkf_oH
znCTg8JzN)tZvKdeJ^j!3NiA7_k><K<oTC04tbHdO=>cdOFqAz1@DBzHe0@omtWI9r
zak*E}L35XGRzW9<-WCT8V2G-y@HBjJN|K3$ScXd9(g7}BVb`mtj*G?N1kg?8m}~=8
zSC_5JWMLwUZ*}<Uc@{x3sq(8FHFbZ~F<AbD(F&c-(7|ByBdvr!aE1q@IbiH1-4XL$
z7B5_tOaoQ!+X;f8TY_$p<{v#|!TKlIpv(DY4!ES`xVdC{%(x>yFQRYh^=Kv+-Fsof
zW#H-F;b5P8P*U+$cslS=ZVJ|?sFp<q_|b#wk`9fM8gDh!ltqS%Ri7t;b6HxpH;@a1
z66#n)NZ}1$R-Ts}b;F7mAHCy}$zFF+CVuK>xVM8zZ{9D^D|Iy46kczl;ZjFJ@d*nQ
z1r$y#u6=QQh^DhyDqF`zLWHXc_~h_Qw`O+3`(Bz`T2k?W3&N3PFh&iJPk(Ahu|S4d
za+JYZdIBu22fgH5)#B2LI38R4*q1k2FvE~}9IUyt8UoF~Zrh@`aC~=uF`8m)8s_Vm
zE~jMg%RrvN>ZNb}0y{W~Cb+W~ReR$?p_0~4T7jFwsU1uV2}#`_=2TQU1&^{6vrt;L
z7a$>lcA%kVDQ^@^+Ky4Qfh>17j2qhm^u(QUQQy4pTth}y9~2O=aWrB3ZR7{&Omfm!
zE|%*$j#gNE*BCZZ9gT<Ty>5VT5V%rzR5&mI-)Xw8g=>7Yt$nwK^eh$R4;nHS+Z&20
ztvGjdhGyx;G}C;+M}h)QD#rG`loT$h#G8)xLAh9yZcNdQY-W*Y5jlqy$fY3vD6Gyt
zg%np5b>O3OUWud#`mD{w*yCOf9)`l)i|&q!dQm1c3$%5=loV|}ypd8+BBiEi<%M1M
zByjLs%Dc8SHRl^rA$RbugQ~!t{`5$6C*UWTbGNvT-lbT(b8fM9#RVf%i8_i+0)?^Q
zW^nW*;DV84AldxS%$=)6UG7w|8P)-w1ndI4m9{<Ek_;y6*g!lGW`ZN{U#Wmc3iPS;
zN#U0y4XCF-)#&+6*iMg2_^13NTT;%8nvk01e^~zyjt<rl!2NZO4weLjWgFQmc=(-z
zJxb2dC~5JsD(|RSjtYd25FPOk6oXt86j4|Y-f$cU2g^xCvs2PvPge!#L_<l-q7yzc
z%si+^Y10D%v6f)$^@8|%1LlaJ*?7ZailzWbUCR=77Nu>k9mOVsH5tao27FA=Psb@L
z4u6<{7G5+>(cV8!!YBU2WE|#^Deo)Hh0Wr0EES1fv%a0TJ9nCzMxv(Dp$rNTBf;ak
z?)+d~qd$gi0)jENE}WVBsZit#mRwm*3VcN+bvqwuDz*M(7`lN=U${;vkrb$Di-zsM
zKYla*C?O;eQQ*Pm&kS?r1}5A;E{0g)ub`xG?~51zqp@THXf8c@XFm-};WWG#zL^p%
z2Vt=adBL%G(ZMviu)q+EyrO?Njm=comKq;(@sTg<MiKq`1Ba#>q_#=4yKd*05S<>z
zQe#&*T?(3eO|fLj(%Y9kctR7+WSU2&GnwWQ>1+mqr@BnKWkd>J&CM-M@YaOiGU*ZI
z9~X>hh9xcVLEbW%5iRM=$mUGQ|Fr7Z%Fb35P5?MWTIdZHaFkNsa-JN+9PV$-_+!EF
z_(mp}Oa>m;Y@y^ylz9rWC~0ub)|{WBK=&}M5M5hc9lrV(-twB}LtrJE0t8%==@d;{
zFtK+CxkGp>ys0v9nmEKor)@+6hh0ZC_1D*40we|YR<3*C7z&bFYT?pq!YEYIaH4Uo
z11@K2VF^G=??on&pm1#j$LPvSu3EH2&+=n_2vhiKS}mKbC(l1ar|UAe<t@p`A4n@G
z4{(ib+tzcco>zZ5J{k#d9F^ft-6+bAs_HofW*cp-nw?+%;TH?ve(#-y3*VXl-di8P
z^Y+4z7QFM$`>)K0*LyF&``*j%z4z*?@4frVd-LCUcRu{T3zP4V*{{4i|6O={7d~Hk
z|DAU}efg#3i`TfCu61gfH--pSHw_xt$4Pa}H^)UAqEQZb$0QRB8;pjY;2}f;pA=BG
z_@eIO=p)YR<Kw;<2ev+$$|Ptia`k4*Y4ffngh6%@dytz_-jqzNUIoV#bP-tRy}E&O
z1y~12x}zx-CdncE6x-zUMrrP~?tx=yl(g$?d|FEgXO;s6_o}4CHh>3Sw9Om@*|udu
zM30vT-BL0@!@6_gXq&*?b1{E{&Rn<6)<JPgI~qJ7um_=bZ3yQ1Pbb%)W0mx0o)>IK
zmuzfp3U|UX^<q&5qh?yB4N?!P120fm{De2V^zuH+Ad}@;uqNDnvbJfeDh?Zq^n|@2
zKg&ut7{I2sqsZ^h^kwR5q6t#n>CdDo?im$p^$9=A`xhaU(i~U)`N+dDx~J<V{QI?s
zE<9%Db8?3Qs)=g>8}aeqy?Ap0^wj3_Joc<7gv6gJ?T7%0D_D-m+s=2Hq&M0a-tn6I
zyl1X*-2=yvkksuu+Zz%Oz@(DQYl@B4qU96qbq1c8@}h;*)QUDt!cFxkjr;{Owd1b(
zI<9ea!i!uO>Wqy=`Ct@5VccS4m8)K8f4VUa$4b@vQlEjH2j0SSqA5sZV3Z*c=$fYM
z2L36ouEUeHzfD4PWQ%HOhM~KP*X+Y7^=LXU;E*|v@y>)OlklZd4Nc8y8mo@IOwYZ;
zGN6?XtW{RdAZgeWfy4q!%l-7#cBd7qn7v3|(4AuOyNhpKnJ-#xn>)Zcg<C)b17D?&
z@(T@-<4CX4(O@~v4n^H{FhYgC7#bxlTotb^9XFeZhwdRMOnRC3igU#+z$>zK(Jtmi
z5LK<Me0fO#W#{HkJ@S@!-@UjY=1XV&Z2b6Ni96)2av;DK{@B?`6W5%B9WG;4q3}st
zOpsC80*2yGfKb63=%Fmw8qb3#i9e0xRu%XM-5_7dOHo`^ap4^&y&_QZ1@5J)sfq&5
zC|GCj+`OiF2YM1fQMh(5PD#{eGj!7>?>=;7ie@-Jg_)jv=Z1je6tJ1T`Z^gaglS}^
zZ2(&hQIgn&L+qhiV&%2O+i$SRbKlx7<iXXF%nEy7`Y63%eSx=K$-2p(E#GRkHvVFt
zrdJ0AL*kRl@)KEKsmh5-WvefCEqq$xyvR+lF;ztO3+EXHaDl)jbqvYjc@c*W1;7@X
zY)X!~{oD0zqP=i@V>AFhS9sc2I=Djz;gNL|I4nh06=Ts%uz_?oS=IRD?z|2@X<jD5
z1n^UlL0Um=^D3BZl2Hnw0m(F!TUUc9-37L)5UG@ihm$a;^aTVMnF_RY6-~jl0kA;P
zNQ&G80X`qD`PR&2bxZRRPqnGJC0Dh?SvvV#zFSGVp937KOyb*b^Tw}B4!@yQ(4~SU
z=W><+R-TmS^QL8MFV<GGlkZ%W!^0<T0`L<hwZiP<e2DEa(X)FE*Ay%#hoG9@+;@q<
z;4magY8O9l@^{yZQZ*OY;KOz6Coq$UkN4dpEO>j8YZdrh*Y@RKuL5y{)ksMxTxxIx
ztb!oB-;PRVu_7xv?gPyb$es=q8boQbku&o5w@?fhrkjW1cs{BGN(!2Y_ZOv62BNAm
zD7{$JA?g8hkK8iA*n|1*1#|*>(DO|*a9V?ZwDyF%@PVT2EgvLcYlxl9cgLmUhaSJQ
zt7xm{)>|4WI{7T-%=diKa^M6jOL;?XKQ~5vrpEt#LE$Y?wVh(l0wslANxGnmIa|3e
z9108{cia0rfQN99gqvit=jFY@(L-c#YF(m+Cd0MBe&>E)`n7|KAs{Ip{)RafqohSo
zz!Ijie!_AH<WVc}WJ}(sB{_!LzUre_@4xi82{(1vL?-c4IH0d8mW&w6npuZ7V2)A>
z!h7@JkQ8(j9|QN-Guc#tW7tI1k*h5u&#Nkqr-!O(TO~8DcNLXlTSsRHL1$-&Aa-;J
z0#1n{yaPIfVxc?({@ch>ZwW#Y+B&*A+B$R<j6dAr1FRAS07Kz%Ha;2YYp%Xg)^oZe
zyPwTseDn)FK56+mfDZu_ecRjs!v$iA;J@B>6;m!iBu`2ia1C%;^Sr(Fw3-k_v%&F?
zZbh}kmbK+S6jPXo$#Y?S!Gf=jeNxO<?&HKzC$y7Y9b_DQlER^jZpSe_1Ku<7c{@Fd
zk}wE)Y6uoK6~%H6*uoH$4QsxB`R*G}nLUjO#D*Qe5l3FZ>uMXaE?9g{GI`PIy!F4D
zY0QEUKWzs*{0GC_W0rD9Q_WAORO5AKDCS;O#VpbwvEKdFNq17X-}W!}U3cdlxA)+V
zJK-N^+;PJV-uydm`1ikY5@(SA%GGg$8?S%(j!Xajs|DVyq7nDNqsL$YTN5Q6A4vor
z<%<S3vweGHES`8qfv~){PYTgKkW~HZ*jg;CV(AgD?8@a$#};5d5mFRQ@LI*4ubmtp
z7G~+1?9D6l4v@5HV1X$z6D|Xm^aKBK1HpvQU*2&{3}R$#1O?Zx2FvmfK88X`!T-*k
zTH9FSDBSv4D_l$m1$0EB;sr4zyCB19^F_UF^#`v%ciz!Q9Z?sgYlBf={O}21J1D&d
zES9jWNzVz60jl=$KN_NSprzp>*FboMyE#Cpl4~oHqYJM!27(DL9H5iWlUvZOKWLNT
z9gM#5<~Tap4IH?m8{J<L*k6!*_F2P2qOl>FVCa`zHKzgV!Jj=~pA?VJd}jvB9`+z_
zl0*M|A{h}6WYIhHGJsxE&g^`9HYs;u8Z)ELyMO-IYdQ+N4Q_y!R}Z$AcvD>V!1%xW
z8v<3n*&l7nIa0oW^-A!vi|80&nlhrP0MWA6Ov3*D@Ps!d8K&l(eCg*7Du#ljw&->q
zQ$q+!5F547v5IiT!1qKQ#fyuER9W}yH}5@g>4~GWoUg{`udZS<(O9C{pIqPqnZd-n
z5J+hlrt$?*>*{!l9TftSGSu8JNmO#cP(_eZMK{_n4Y3dvM|{B(w^*gfr~*Xs%;&Q)
z9QcTU3Hq7RWN%<`Za=_kQvk#I!~W=N;0HT0381>4g+W4w*vqiykB_l_%)ta|=`I)>
zrISx9D386$NgZqmU1+<#H10P?$C{bM*mJ%C-eJK%7L@{9dBXxPwd?E4=)VOUxPOkj
zdU+mfpk<4NBK-3<hE}DDC?c(5aZ!0>QB+du=09qqF(?d$xra08Jgd7jA4rwCmt@T+
zB8u0{5FKUh(zl*Ie#(rbzcxUJ0(DG7WAgACf50C|`h2e-D=^^`Rylz%o4Z%O|I8iN
zj|<R#V(!Pjn8R`a*dKWf8jg%fsx;@93DwC6X3wv_1II1x3rxG~xNnkS;!D^|LzBJr
zBPli61OL1s=6ebECMgHwDZ|#JjyZo^lso(ubf96EIF3{5VX17xdup6aH0<pcRC#?p
z5}~SWb3Ad<GGwM{V}%r~3acnPyNZso<;ml+HQ{XHU$frWDT<<C*fux@mWAc_rGkGD
zY6zU{XX<;R4~Dt+uVU`;QtiII>`Nyb{8Zw4tf%%w6wc2Q?XEA+ADyQDUQ<&`#pzfr
zOV<t`mTfua-1~0M)JFU_cG+O>y_Tim#0vR^$AGg>2RP6oE*79t_lOeKqhiNqcnuoh
zk)oP9Z_yO6E*x(e^-6acQw88=Xo~xGT?I+^hHf_YLPc-UL?HgUtEyOPTH5;o0k1WX
zG{*TKLY8R`j?&B|p3=HWT7HO&hNRnVJI)<ONXj&fNV1`ts>?gHTpoQ+6cAFifob%C
z(%SZWM*Q<I2%uW7T%lPI*UCjpDz<i7;G%lw9%zqF*-c(`69LoosN((F!NpLJ6sTo<
z)Qq<juLlQa?m`*$m0T|pKsm*)#vNW&!!%{+I<$hIj@0z$zyEDx?%P@aO8wVKo4|#&
zP14y8uAF6lF+Iv;B6Xv&h5*>)zq$j<Nxk^BbQ1;vbPM%--fFu#iCLt9r)MlHmTzh=
z;17K`qGZ<wfo|?H3>AODRk|R;AN+AwfDDl{e{3py*aM)L;CG^uvXOfwn8a^A!1hgT
zonK8mJW6g#kV>%e?3kz8#5@QrHa*3vQK<moJ}9J^ceQOxucRW4BMu84`_#haJGwe`
z(FCUrVl7fRg%=(+6Yc9ONjiN#JSqntLqJj_kNDG^T82rIAxJoiIX5rC(}==IoB>#6
z!->8IhV})5vHHm;pFbk*<4#%W7Dai*v_tEM&-%)dz|Lcc4-sTR5PrBg8~4|Q!P?{g
zB2}ktz~qfM;21vtg9V7Dqd@?+VoZ{>1Egcu$^9uQ37p(_STaUW|5~}9;DQmNq?nGO
z-=|@q@JL!#Qmk75pA?l;)zB4yHOfDOnaD-?!5FC&Wm)i&*+86mtla{x*|08F((4X*
zq|%Nqp-`~IuG~*I#{V<g6sQW0Jn^!dAAEn=n)PjFk;kk)usirKn4;BoF7|Npz|d!G
zAis66F%(J)WVAcZM5h#PD3fUPQOB$Ua(V%I>6-1jhoby+-T6<v^w#&kuRJeChqDXa
zqHS-woT)p6d)+njf`#&nm8G`1>)}war6HbdBEBFK8~cU<;S_kKhCH~MCr{KJ%mwAz
z?*s!0<bG|_v#zW3w?9<UH#6vb?b*e7{{aUNp_3ncQd?FJsQFU~#FRJ&_@wONx4>}&
z%iJ#}MVa)&BFY?Dr|grGa9_P6S@Kyp&L5@Xqu<t4kO7G8NslNn;IMYFtE+R<%Fmy@
z>$>A2IAjAIrGmBoz=+v%PkyUYFa%Oc3g#YuLbTVO%2tr{xXs>=9b^na;KU1#t#=&n
zbr|L1p=LU9?57S;7gevsQ7892L&v$(mhriQX?9(BcqGZZZx@Qv_8S@MkjN9ZOzM=d
zl|GiNw!b(!5==#djX0P%!z4!DX*fEzgTkBS?QvkhfpKwDEvLGhukbPz)A3VQm}ahj
zpH#D7^nB96D=8NYb1dhJz2WLPP3=06hNLCPhN3&N{4NAefroRZBfx2WG@Ydr&uHLP
zk^?9I2y7K4{CY|$o~7rWFn)9_bo^=s)0%(++i1lhHj?sNe)#;Y`>#6g$ff`jiZ+Lb
zWs|kBcsw4hsg6<4X|_Pd)<w-Gl>T&PcgJ9cq0j2nu$cc~V`u^=$L>0-eDK)Ob^7bF
zT?E@LfHr!tmW*d&{<Ai^g@Wj`J(z&N{h1D1m!+qgGlz$--2wtf*27B-k>wXN<FRC>
zItqIOp0d<=9jXp29(0L5@(!g~gIHt5IBWm~M;knEr%oD_^nn-Mzfl?84+{7jiGcq_
z@Y*%C6G(KR0H@lKG%)v(#_9))MIK@!_k*!9I`pb+mFh*ip9DAq(Oy;>VUf<h^2@7t
zUwzC?9gcwc%CK4m0w%$b`Ohvo>Zqn<sJbRr6^_?L877_P68<<-8;Qs2zQDnIu+j;O
z2eDMuetR-8ZU{fcURhB7A|wS^YT?T6gJ<Ku@sCN^U(^!HNg{0V`?(Agne>AtfOZ)l
zjriNJ*xhZmWb+@43TK9$xY6mU__TP%T6u0fk)3wc)lqV*>DckV*m*4OhX-P?fH+Q-
zY`36(7H7cL2e?@MBN7gox5q9^<)oPy13?~^cTyCVhvH<pSd@I-OCf1DU^!JEi9TcL
z3h3Y9kGu!64kq3LbD#Lv<F2a9dE0q+Y|I~h3$6=lMP)B135?IvjWN7_T=mi4+D2~k
z_p239T+-_vyCuna!`XD<VTaZZkH+IMswNho)6*|`@bT*|J#Ioak?<dW<XXiPF)KCp
zFU2-emh=QRaV&vuyi52CkrdpUyPCU6%2M1#IQ%SJv>qjGyF3+2C*Bn#Gbg&=O!uYz
zXZ-A1it%1cF!R4>ZBX=Lx3_5<LhI9WZ+P;f*7rb4!EVN?Tb?y>(0iMdmJR~26}~_=
zb>Z<^@D*t`QGLV@hMrTgZmTaqOU)M=$e_+xsT2(-yh-vEf$1W8zU>9LJn83NGQl6U
z_jmAyeFBK+k(3*D_HTJp(}mn?V`5COq@(~+073xgyX~R`R@`%m=oKA?P_!V5M{g7s
zVk0PwcinF$`eSj5_Qiwrs3T9l`=u{;<Z_#T{O+B{AH8ng*ju+3E!!#<Kts`5+Cs5#
zeX`pnZ4NfRF(kqLkdPEFi~jc19+J|`nY*x;9q~#DEUERDaID$)x@Z-0qP_BjVEwSs
zZ$kXJ^@fO#@y*<9<7jxLmBAKuuK9LTLDPSnRfj$y6HnbK!3rdF8kEHc=S`DQHZ^<W
zoPULqkw7*Zy|h*AA`P_m0M7gF=5x-w`l9o%xcu_VFT4DL^RK+%{43!9O89#D<yV}4
z!G-5vc|KXVS6p%4`Bz?f{)zwIXsV8qQ#^O(z<~WnbQ5cwn8sso|61De)I~E>5VpU9
z@j7`9djcod)xH_cra?%V`uax|Ma5J+sG_(K-$bm7|8V5t!-rGJmRaZD^!C?l+IbVB
zB-<%;bZuU-=+`a?=y7aa5<H1k(P}-bC%EV0lfNt9--C-GAgNc%wC<P+r!;=dYA_U)
zeMlVj#dIij=(X!)lg~@ruNVgL&`~cg`R3sXEPd!d&)<v>HfenVfJ>Rn7x;o~=WeB2
zFuG=_lh<0Pq_`(<rGWwO3W@C<=fehUyv9arn;uup)*KMOv`!CTZAZ=RShsdd`_}av
z5ZiXF+qPrv_HA2tknihZYWuqF+r5Q*!}@hQx36EfmUl$emAVdw=-}~7(mOL*swqP?
z%)IpCF;rtD!l9Dl6OIHepAERH&Z0~rj-%uF({GwGU*PlT#Nrx6O|fsT@-RzfMxS)!
z^I!h9o9As8YwimM#8luI2|1XMz$eAFCFBVZobPAXba%REYLD57QUAfmP$(%G-hSh;
z-Qj&ptoDdcK=Vo!qXdre+05Fg@6_ekj8s`S?|(xe8JseEEE68tJk0-e2LwgkPN~Tk
z@{*A+8eROC4Cd)z*~r*WT@kBBd+x#j0oT~JK~USS!*M$}jt<WG#gbJC<NdzS6j#ry
z!0MvlQLExN2l(@j(^b%O*nYp;4j%w8yqD#driNl1!}(GfrXj`Q01t11k{UqZipv9L
z-M*<dm8Q8sJT~`7O|YDzE@9~t);D1RJ$66Xc+KQ{-}qsDYmUc(@EwWgO`&UNCmMLc
zl!^uz`#kT}Phn|*`^j`1*3tt_UAFCDXMyrC6iO;U81HUAoAL|+lW6eMlb**BNP2uy
zyC|(aEmTW2J-L%K!b=4ksbRt9_~D_fzxI%`RykII7s|KMaWp}f^}K8uKTVH=GP7Jb
z`Y`sEQ_<vmD-9GH7<|dLoS)_~evYY2`k4zhb#=Ik<aISBKn2A+4F}>_j?QZkO(SD9
z4Yme9RZ}%hFI5C$dH>#^=o)6ulQoSm7F64{B^<R$E&t4*0S>1PKIz-z!xMrvBT|f?
z3}42w;n%R^r&r-bB7AZll5*O=VhM1+o5E8cU8ffEJmE3+uE)U175SpsdSTs-i+2@x
z3kz37V5s2=c^&ABee~e&6^fRSFMwK-mZh@uR!cc<;#ePb?asf9l3E4VD!RLFi+fE<
zIVzm>jeDf%*gTG93s(g1zQM1*{?JhB?=zqIc=4CFCx(x!&5WR_Sd6Cs{=bh|YGNI-
z!Glo3%HWa~cuZ<<OFK56$)+Ps43kKo-LB|lhlBL(umnXf2*P{QSu#pgtnpz<m-H^x
z01}pBC*Y&17^;?&Y=hU}SajV`uw#e=<OJRUOJOD5&~;d->ZM+K<&T~})MiCPu97KP
zQI#AJ9tR93zcW~Xv6eN*cwsV^pr|G)%KGD>P*XY<f2&PX^Q6KFFQ8<~rotQccT++T
z*#{u_Or)k?cl!sem_LHlOJNa|D`>X%;jI5BHU5_0@)p+p!wW7hIYw-m2k!>9r7k;J
z@bC(Eml{=JOFgRb6>YG0W8WAWCAEr<{A5$<Ap>u7Sn%|fjw$E`I5iMci|6yA@m9E+
zt{Gl6c}CJ-6;2*@)zcGwwV_1aQ8%wObM1DaAX=K&lSvnBU~Im9XZzOAkM<>Kj*Er-
zEkC$<K@$gAR0@|I`+?d{{`Mqhlw{b*<OOhnH3`ce5o9f#G$o=U{zCu&$e2k)K~%oD
zWtSnVnl8h4>{_QOL~ikRVP7ZQLd9ec<PO4H<v%WV9YsSm##`&E#f|NHE{7GD2U_bS
zYVeb4$7`d*Y#<&^W>~5r<Bul~{g0O{i=<&jpdrA+kd=E=gi*m7ilzP4bn}~DSA4MB
zsI;ylSoUKLhe!Ueqgr(nloMM!p&Q`sr2JkXm5%i5CEdY2$I>Gf*`=Og2N^@+le$(;
zvOb$wZk(TB=-AP}6a)v$NiiRp#pm;)^Yi(&Y)krIe?R=c**YpbZfWuTdoH`?)_WH0
zvSpCbqHW|oC3V4}2T}d~jW;jOhNGi0ez+h^b?7b2$jev;<^AZu;0@#?>(+)7q6`bh
zJ`tI>68$nqQ*jwq>%)#gO0BKGtE#f|%Vo=#E&t++Z<oUdyp}9pvE+*tOBR0%lS{r_
z{4LH}vSRU)C5yjY0{9Z9O8+=x#p1=^eX#_#UB3L=FFsqjJzvOcDmbNp((w!qu(hN)
z`r<iMLo%DfP9YRmm&i69u}p96EYS;95?d;}cHyKDaR%x~Bx$D6pNhYQ1z2TM1e`9~
zThHfetAck3NLr@|<lU2V-A9atrQaRZ-69hJmOyF0+=w|p4BD>k8$+R_qHE_1?uL0@
z<b=&kxafpsJj6}ds1!Sn>ZpO8UtS!q_QhlIc&xg1)KjgB(cUGs<xMp2g}mSz*dht9
z38*C~+-G9~v24JX<QOKC2{5r!*1I<BWpG!8@Wat~;oa#>fJ;RDnGsiQ!OS||5@aaa
zd@N<O<h;1;w)zp1CQh6<0saY&m^pF61bBxT<Bymz9%s#%0kbBIpD|;6<v*D*amEq2
zx`*ab_iiz53DZj+n2sV(kfh+XJblirC|Gg2z9pGxnSJ#FK~}pX80;b96b-rf<IH3P
zljm}&KpY<P`U&4^@c5U+(kU2@=zKjP!ca3m!o0;0P?fWm_7Bl27QdO*T`39%^Z4I;
z3GKndP$((feu$iQT;0Twnm|xeCf4#Q*ldy9e^63c-Gy%HcV~~F16(}7HJ<m%E<?17
zVo`BKfiLV_{>?g&T}+`1`vWPE?EHIU!ZhZGfIUJi&iO|F>KHk3aP3O~$U#+ayFP_G
zWSHp04=nf<FIMwtifqNT^kVzHb?JDi+Q>^j#xN92z&qwAV|r$JA6T`7%diT{n<%%<
z2T;}H-{&DlCS9EN&Vl^_>4>NBMYkw^a_Q708=wLsuY78gT*Q%`awT(*JeKo@+;M$^
z^JgHshcLV`<Esx|*r{P76z>qEoF$w3i)@gMpZc9G!p0)dAG1S|#r*(eIRDy+Y_iN!
zrs<BI<-Hwb3;{_=Upw%(o@iz`nxxLBeRS%Y?UI)Ok`(EiMH39Z@#}kco_5OIbMAio
z7XhB#ykJO{Xtb?b_}FnrKO)GcfQ?en1{;DQ{C*kwxTNYHlsO)w7fOX3h&K7nfdbg!
zU>B!f&MXaRADQ(Xc#f(Us{=JbQ}G{Us!LNp7+>x6*+oF?QF{2iC#LlBKDxh_U%fsF
zm@IqDr}=`4{^$X=e1=C##?)kTp<}mlCsxWvuGz#Fr2=@sLYbtFEkV2^nYrWr4JkGl
zV;O&hqp2qPAqWb~oO0|U1gmZT4oCcl-PR@-Ae4vbopi?U0Zil9(k!@wtygtZ6mD|M
z8(rNC4=xTqNs*zJ3oi7+7CPBTL3AIgucf)uf8|Ap`{C3`h8znR2_9<a?`zkt+0+SB
z;Ii@s*=Sw<!QH1!<AztCv;|hdY)qI3(gfit+>^#uVdWYQin%%uk<}p6?da(4f?xRB
zq!{dy;o?m<uQACW_~a^M87guBSK^BcXGOcKRbcI2NyzK!DR_Aq$f^3eY$+jm`IW_m
zb@<ucHlGWJQfIFLw;Tux`!Bx+15i>&GUPm;Q)K&wC>scmd_*kl)*R7P@>*%Pm`qpH
zE$!n8(LjA9mG#A^k6``L>d8wSSqB;tNTy)O!k4E8qqTv@L@eT$F_K5lkvB2{rW~y~
z{Np^)cGsWL(@u}2CohNJ`>U9Huh+8UE}sl$4wV!~?1~??ya;NEp*^Z2iJ|fZ$u(q~
z$SW3tbhZ8T!p&!mWUKu&9h|UO0Bz-EOSE-xk&J@)L3295#EBDxUa5b=Dp5ATcO5vu
z%wiiT+iZWT2|WKW)fAfYP61e{lA3$Gw;C2TVcrtiUVC<;f}`kclqL}H#d~=l-CxVE
ziqA5@_NOoSQPdSxQ5^}3PY2m+l=$N~`iO=poNr;k@5o-W0XxfK=Q>~-7U-kgx&kkB
z+?I_~&EZToI_>3GFPYc!pI7h1*xuU(*dsA#D)YxixZ23+55Kz{;(N(eEDWAR1N-P<
zDF7CzdaGk6Jj-_0oW}ID45O!i4_qIdVfzCNg-`08i@WkT%IBFyI$d8o{S(_1AnwP7
z5aWYVnMDgs5rq1Zqe7(By7Yw`PizkQd`vu=91)2>EIC#V;!My=Jab*%a9VABG&K6C
zW|TgTj;HU@4GCM;4;Uo2$Kv;2FHW&xKU<fmJ@q>SRMT@xy^x9TR3%MOh4okCsHP<Y
ze=xv>!*vapuXWTspXYh>{08Zp;uubFq;L^6u+F!0G|QxJ+6;_?m>u{DOvP3v*rr)h
zo!^d3$D#@FRp=9!^IO)we)=ax3my-YQtYRUW94<--%wQ-iKJ4QV{UrvgDtu#3I(`h
zLe7#35MotmB>bEs7K$Iws4b0i57&pfe5;E?6AuA~hNJ{LZfl^iw1}hp;7T?<y4!`z
zMNGZu>Ur=!O)#&PCEKE-Z&@_|+L<l2{zN7cgJ9pEP5UqU9TVEYZ6NijhU7S}RcB}4
z`|+12)<^x>jGqf1wGL=Cm~v921-x6%k_mAx;;&2jxU*K8Sa3>wQv6jM1;4e^QN?Zd
zG^Xp}Ji?_Uz6XGkn$mSyhH84Wz1XF7=50|Ov}cMr^+?qyNU`ZWn2CdhxwQiq;4-WT
zMUMsS7ubL2-4R)c5jl<;dH*)i>1<oY%Q$A9jhZ1EVjeVb(|L*NI9-!Y#r@2vnHN6z
z;jfbENP;bxHnzhD!g@bT48LI5t&cQCn@d@sz~z3d({-^j>fmDtNvb>sW}XFxV$49N
zF8$5LQYBnso8W;O1sl6BNtV8I@nctv8%z7ZUQxsTKCF7y$=O<#J^edgbj`d31QpOd
zb)8=ydg0Ty-P^8;<48&nInVq^#I8`}#|8vlF%%qSM|f{soT^I&qTwSR*+m+sq8kP;
zP0>V)-=^w(;jOXhh7b_+z?HN#N;4OXrUG>Hls9&@6;w;<BH7FaaS{mJjQG*;cPA(*
z-STKFc>oDuLNc+*Y2K7K-c(2XvXOMQdgP~K!4PsL=}+nvL1E_+iFdx50M@>`dRRIT
zj5ntI)YJ<edS&T4#St6@2wKDrCb(O$^}6F0EqC2zsp^I@N$H7;U<Ki?MpARp1O_PT
zodnIDyaF_oc#z<bVx=dJFmKDN-@E_9Bj_4G9gim)X*M<af){>Tab{Jly7{5q61oK7
ztdjKPX4jT&f>tO9?={pD>q{rEX#;O{5ce09)K<G>d)IYM$-^@vYcdT-E|3*WR7zrc
z64k4GPId|%-_7M3LIHFu$!Q!AN=gcc&V{0}W0p&r#^=bmqJyn@BIk~+OyJ}3G{rLM
zhju|Qk8{0+gubFKi%aKGOl^HE%EhnPWLdjAfOR7F|17PBtt!MvW1$iM8gW*5_&*OH
z9-*n)>RK*4^{N-X2A+a+7ESa-VO{V5z*Pm3etQyELt&XKf}ZwSv54X35Mby?YKtqT
z<H$%J80yHcfKPZ5VC~IfZoA;v%DTl5p2<}o8Xp$S9Ok2fvHxzoY4J9*xbg0IY?$wo
zcDTS`>Om=R07rO%qbQs5hEr=;7E7W+Q&+fyGWXz<lIpB)XU9o_0T-FO8m_(P8#<~C
z(gfb9Wb4W=XG9Xw5R+}_DYOPuL_uZrGiz_)7^*HyQ`}{1u@?yhPK2j}0`W@GsfEM7
zlLeBd9w=(w08tp$<0_!0pw-&b%~esRDNR#T-{U)SJUFr7uy&Qo>aobPV0B&YizJ3$
zvGv&-&zThd@5AHRl%iqSf7Q;t<lZlSYs=>``!lQze3f|MtR=lZl>zbgPKu$<{K>Y!
zu_Ze^$QVK;b+%kmVI$*BkMEF6N(u}%c*hjlo~-|SZEY+@$LQ!ko2Dij{p{1bi;AVb
z*WyoC9k)`$Hlrl*o-V+%4qlXCX|nu8CK^EY2id1|R8n%I$WQJk&?kkfY0MuR$BrWj
zDs@w*iHXX*rP!?*cBthI;l5C;AvrP;N=C{qDX_7h9;4l<Oa7Rw3>&Vl&M={H^=+Fa
z2SW3K9eXd3OwWX?uKjQ<#4)J{cM~x~f|A`5OKzweFG<v(mjn*+!&_9#RSG;N%rRt$
zo3PwS>@?-y&SRPC#5?w`-OJy+_ne6-mi7&g*Hh8JKmX3oxa+HS+s5)GEK;!GvuL$F
z+z?NI;|1pg*Mz-erwjM@uSQaL$MsD;H<M0Xw$`z_#~g#8scX0KcI|&8n&LjHikf%L
zv$xa*WB$u`O3F@k-JHKuwNqb|$>5BHPdSLLL|svK3ZnS=bP6nHEI=pD_zn}H0Z|8N
z^>rDCE8BE!6rLi6VZujXiZok=sGfk(#!)2uwWb)Gh=dy$sysjsA!u(VdbbJQ;NK%_
z>5I3dgz_{O46`)Ng+ffE@kODaN&9Hxw;uqSYVgcK;Ph}h21<HAw&8(?78l_Z7J1sD
zuyA6OZi-da#bR7)-V<ML*DXO4v0*4s4k9b?QSFZBk6_rp9kt!!VOe4GNB3SZvN>H7
z_s6M7r0V~SeXGsu$PYw?z}|3**3PSg(dJM+I;BlXX3|0)pZx*dhW*9RkrX_w*QZpF
zlsn}+9DmhYXbCnDPy1<XjNzj3W6pi>^9|w~vtmqa_F^Glv<tUIA~ki_Z7Ra$!wP>>
zGIG4EOKO*3i$6Xv78wJfD%J9=B0<~_CVya7UlVLq)xW(Ui7`7Dil6(nO<FaV76Tlu
z(AKr;lo0+9jeH{~0jMM%0U7X~OdvsVF;0~K$r3;*YQib_MCA!4x!KD*gnhL`{R4Q6
zz~X!ytodHBymac;yhd_9!ZNR*6xNV<%u~j3bX|sSOvjnjh)X~1w6WVAmf~an5-DfO
zt@mVyMGwEFjg0mrnU<bg^7IWSOia>&>YB)~<JRpao++|g)U8%cDtvo>FbYbF=agU@
zXMAcpB6jD)V>-wfGLl-N!+&vXw@(VUdgek<ymHw)RwIed?`OoShea=1+NFW*`T4w>
z_%QB;f@l>xU(D3dv6D8MB6ib5Wx(NcWM#|N4PV@Se2cfkOk*s4OBeX6c;yE62t(AS
zn8mk7;BXx*=cgVnVr*Zz3-BuRUFTHw9RD8ZmdbA0_L}HkTtekE0o}kJa>7dc0BT0D
z019A3WTcqx?VtmG`u>83dWZJj5y+HaNX2C{BG~0P9K%WlhOSB7-04_klzZ>W*@7wU
zy85tx4G(@G^16-baTOq3SKF?ow;p%ogj99#Mqc4@xF=%qg+iC*ii^g^;I!aOz=@!l
zGk?OO0_5<)$51M%^~+Ti_CDV9*mkMg-aA-#?<mhTQ1SF--^+ZbCCN_&f{D=e>t!o1
ztv#bEK5Xnq0vP$C*Z3Z;rI}N9UE7i*Bf$W;MASd$V;x*cl09|+0ILLH(ayU(3UUGZ
zbr;-)vb85<b+o+u)7VfF^FQ|swENv6Iv6BET-tvpCfpe;N}IyTV{6rWFp5dtwharU
z@P(D3_`XY0+8%9+vw>)n*Paxj&S{_6Dg-}{jg6g15tPX9dh3FtvopRbVu~H~aRuIV
z6oFSZfBx3Lr;UEG!vPBqzlr7s2YYgaSD6%c>tmV5q>o~5-wAFiYUvPQC<7<A_3oCQ
z3PqYdcbnuzPN1`rRkSSGU49fz(?@<RTBcbnd^mb|b>Q^x3WBX~fcX9}>i*WeWJw0N
z2Eb4WB64lzs49qyNgfm2XRPVLHdPZHFyG#O_YauMS(nI&BY0r+yl-8J3`IhKh5GPe
zyI)R;g_x8#r0JgME0qHT(?Er|*kgNdWwgHlPmn0-9hf;ReIF>P7uTbymkW7Q$`^k+
zg=YQfG?Va?Y-MchicPMKkv*)4PAi1@Vs6!{^-G_9X?2%j6|rmz3s(yus;-jDZTjx(
z^_{j`EJax`;RwaGSKt_9ev0!k3A$m#%U!^D9JP4}FmxofL|1)dd{0(#HZo<6)ZJ)_
z1nsuF@#2Ut9KUZ@kr!<9#|xrVF!?IK+b(v#*Bne8K5u0~bZ`tZ>}G1~k|A!qBpMB3
z!rNGY4C@r0xx&c{ps{6!?H3#^r<#QiC*T-n;F3n}>LR1jmX|sj>YL_%Q(w!~CrOgD
zUf7eGdm-ySIh;rYLiFRM7A=DUd6qy(u~nUGKahzNpH#;{3KpXKVnYD$h;2_aVV99$
zRYN4qP)veJy_^Tn6Lb+)B*F_}{`Z$(_0r1L4!&5#u6DN2Ww;iE^P&Ng3hD~liFNlt
zQP?+#uV8*RyDov<Y*}Ajish!hqmedKe=VOBu7F)wd>q8#Wkg4PBzhw@AmH+Vjo$r8
zx;9#S{`LZ&m)zESBY|-A{yYxNyzGdo_@Sc~%a}rnjP5|Z*h2m_Hq;aY-cYF~HWrLm
zH9eBkagBinPL9>_D93P-Ajd?S-*wfzQZ9i4*fuuSbC+|0@jWHI-Pp4Psv`d=pqrYN
zffGR`X<-|h@ju|?_|lIK9G*{#N{Y2scd``mNlAr05lPWhz`>M@n=TDSAU>=~j!aD(
z>5DWZW-PX39X(9o6-G|FBky`P85wcnE%P@Un8?(Xa(PokC)5@TpvJ&Z9rY7$8}THZ
zV)2a;rF#pkXPVCW5yC21_*dyIS^~Br*>SUHmJtlaUbx%!nt2hOh2@UDWCF#8Cw#7e
z@9E^;8xvzgla?qD((kw<I{Lb2zAwtIS_IYrpB>u}4dweIs03JOGL8V~AWa|tjc#^<
z6S4nx22~QGv0u*);~u!ku$e1eMFHI}mq1|?D7+KjO|7bEFA0Dj#ow*qT6qphVS-T!
z$00S#v&$=b-;v8?R(U=XdqU28z46G~08!co2fcSB_lanTi>DrH15=H2K}ZG7wuQW*
zel!uBM+lr~|LIGfKc%6Xxnn!22epWd+AiiG+j!F9VSjAIxZ|JrW`iPQhFm#kfB}~b
z+bWoX0Uj!R!Mlf@NL{hG{hAa-#%5;NreLh)u2xtCGWu7G?p;$6?kPu392JQ9PF-h%
zj|ocvXlTGGU3nTEt!jEkvOqrzD^K=C{}y~lC`uN;{LT-yv=<!H=-7?^C@86{0T*}N
znW4C*=84BA1K{OyY;@!+n$^}ekRm7@J2RWYe9{{isN=f=yin|c=iPI-^?3I|OTTHV
zA>1Nun*mI=j>>{&_7IoJW~poz=G4_?GTFLJma5BQyMof}GVps{2Fs@ySPpB$8+aKB
zw)QDQ@kYuyfUG-MItex(mDE@NSgsVX<4^;;Wp}rzZ8;nJ#&iB`vT=bUZhrpAIUlH!
zA!D$G<2k^k&>d}Y!~Y$^99MzblmGqfPaQb!L;*9(g11oMEexVC4D*Uf3pgg(x3ekl
zviRvFLpQ&ICjra7Yj}_`gi7jKvi8PgyguHYVj^P}fk`D#BLKPTI6EKA4y)xZ*iaPB
zyd*r`<j)*4zs=C|g4$^ocDClWw%bYpc>|t4FFBUEXarY3=A})~u?_L+SQ8tmId6w-
zTIxUw^RZWmQSh?lF>HL~8>DR>z9GC%$5!31M#kA}9T$zn!=RLGA{mP$qtQ~*C-|O#
zIZ>E{-;<GO5+>mTmT=K%ip(w{8I2{wF<8K2FC&h;v32QeeeHt^1Se<&kN^XPw+)t)
z5}z~;K56>)ZLUlPY=lcDcA0`E2@4wn0SK6wrob`V3{#fcc6R15Poyamih=+dqW$UI
zQR$lDHB^JY_AqMRouB`(Nf0dC!2z4G-G~7{Mnn>S5=V$AK3&dK#x#O!I`y60AWmR^
zF@!!TTqbkLv7{p&f=fiBcZ0klH_Lx^vFno=fz0ruez0vbFSb2({7qkrYRBewqnK}3
z^_{Q1xx=zLw}UrIGCy1DmaYvK&Ha#9+|Bdi=|&vYDRA_<jxKFGQRn>tvE!QP$lGo|
z3@I9-qBAy=nI$1D-FmS2?$@J|jp1MxhqTY6>(V1&qAocin;Mx-k#X%?yd@*jsG;zY
zZh>zj)0u1r<~O&bvzfYVT{fG^j)2uNBT_I8Qz`ORSI33Ijj2b02!k#u^?Xun6sp6u
zyfYN$nACkdQC2t!AfM{&mej6`sxpx%M>j>QJ~Vc!+HT%ec65MYMk&RW!SBq2!TRc&
z>8UJR6RsPd`saT$$IRc-mdEy_HV-6)$A|>(<kH|4Lh**~&dmO5y78*b`|oZI979J^
z(gL6N+LQvAR1*_Fz8%w)TA1$&SOses(#8v@@c&M}Bo_tS(b@%t*Id1*%Vwdt?2aQ2
z3%n&swgy2Jc=L*~+fjH!lyp(CjhCrF3x}!h!zXP5$8~>6sw)OqdvN2vn_QLPBH;)X
zJ1&o$1Sg=xQUZvbobN{tV<;*ViiV=WaDWSkA<PPexDdSKWH=lh9>LEb!v=8)uE27^
zAj5=U$Kh}&h^_)IiiAU4Fa+O%WD4G3ngZOT=m+4Fl6S&M5X1*TNtyI*proLqShi$1
zZKAzPT{|Y8fu}JLr^c?-727hvlv~&>2RlRHIJr7-yl88@-1YoirlzK<nxdjL)uHS;
zZ*1-Y*L7zOZWkyi3Mmm&coQt~yi$T1ObgtU9GU?g${0c=#kR`oQ%&84P)ss4={M{w
zLXtS+W9(Rxz5U<OnmBu%fQ9>JQ3I=<m+WHO>Nn3$)%dfSYj#6uj|LI0s$Oul&D-*@
zV|j<Ub=LohVRrIFRrF1D05wi<iyQ}(;4Oc;fo8G|k?fPODI86=y(hcURb5|F|F>8w
ziB=mkJtpx3zIiV&@bC{_CVYBBi+PzKalo-8EW%B23TMC~Z?*CiS+A)v8IC@nz^~#D
z4#p?-P8s7SS+(su64)~T{+*;y3w8;o2@z4p&KrZ#KsFX-nb2dZQWSP|S#sVYRr=s&
zVvlGrQsyosnW6|^-G4!Hc$D#FQ^NyHYWf{Ze%NWi)px{FvL5_FaC(c*%8A2DZS<NT
zxNQ1hD1~$=N$OgbY%Dpgl+lWd`>E=NXJpJvj%P%|Cz$SdWu!mzuja*?S>OdrFQ|6D
zbMtq%&T8~gVSo64v7ZAcF#`l#A0sbn1=G~vyU=O9(NKf^I+$eUl%IAhq{;su!*O)k
zP|UoZza2csKwTqu$mrkTS%R<&O(X#!E8J>(=QIEI##<z>J+Zg>9(pb6XFZs8%bnla
zN|%B5zVzDzf@ki5jFKi7G*3xy-9nTUAc66*MEi?ICK^s)uot`Pqt%)yW2kOmDV1bc
zA~u7v!QAI89C<{L3rcRu%|~Pyx{mWz4gc@z5ywBdY+WbTNkJ%4c1rPnJX2pDi-iN5
zFmLHl1ohn*LXrXvZD-q+WSBq%1FX1Wm&*$fiNmP^?_dqfUN$G{PoKQlaV(uL%C@od
z`%kWDuB{msXksG6Yv=uF!}G3V|9T*6j#n(6mkrU9t)(;lBu{T5kbDQOc994Y?3BFr
z0M3FXx}sw~3Dx~+B9`PQqW;*GqN6D)sF8P2L{)*Qs*zV^UMm*Gf?mXzQ0()^>qU4~
z&MEzow}M`=MX~Zr=}+irZ<pje7D17N!{@O7W8WLRRY8MoNdqPI2U#k4^CmPEWGQed
zZ|^)SjzKuS8tWRbTl~X@uDpTz3Fj&D20n~Z-jws3*Z=U#y7jwwQQ*6Ne&Ldl5r&G#
zGfYiY)2s_V>GX=8N)=9yEn4cSQBQjSff)M$N%A#h7&?;TCASpm@j4v9kzA;B{jjm0
zyH-xLiHstN;hEfW_etkIvv`-y7eHm@H6Px6{K(qjf%<R_$JCwr<hz^g9Ja)Ro#_%9
zejS2NBcI#($)%}qGZ?r`#$P?Jt;<CJ76&+i_dP4}NefYuB;VE`7<P5-lLU@R&N9)_
zFLvS3^`eb(ibbWWj8t*3RWKPYeCVL+t_nYnSBf-E#?kPVf&8M!ZtQpJNa`ceP{PmM
zwiR4WFSsx2QkVM*orpy_Oz#YmS4)k!@S8l(gOExlMoz9Hz&A11`RT={o-yyF8(!QB
zs%p7>+mcr=ZAk=b;#BqUs`|%wdm4&F1!op+fV5@mVI^zNvW@=nZ{tY0<dB90Lq<~1
zoICvDqiPtIqy0qm!r{A#xglx=|5%0J`P;Hpogk@sN!{}EOV>`1{VVLRkNOko@uxle
zZH~9F`yQ#i#{xhUR9@)%@#zbuvNRiqxCtU!HaXwM7Wy_yF06>J-q0cXQb0hkrv(tz
zwRhYVBcn30p*brB2pzC3sVW;?_^v77ks5gjrGVF<s4DoUBrd=`!$R*>n#SCEDvlGX
zl23q6Mh(TaaTX@hgEx{=k1Q)s=z3lO$I+Hmn1gvCRojtOc|c>3dG;R2CYign1!tdj
z=BZaKfS3jwKS`#N;|+E5aY53|!7I|Fsp{l2vaF+R!tG$Sydm(#FRqvrjt8r%vX`xQ
zA!scKBG3Qw_Pr;LP5FjLVn==jfjvG-X!5~R6HKXaf3~W-pD0&-%8w9B{S`@yw-my6
zyYo`8H_4V5cEIQ2RVTC-1+$10Mix7kt%ECy(r0(BUwqf`BRM7>_C<Y6(|PyJ|G8+}
z1sQIn*@eS3!3&A4i}vbkW;1jokU&p@iv_au)L%r&0==~jl54AU1U*0jb!;580eAs)
zb(b}j8U*>z=!9Pf^@RK2D1xJPmEu$_Pxwb#Dd7$Y9dVp`+&|9rrg5S7R{AWxkey`3
ziZ@t*Z)6RGs=4q8lnbELyyhr4ZVp*zzu>J2&mNXFwXI#ZcHKHb$&()PaGrvuiqFz4
zsYGD0kd{p(BlEtr(OD&jA({>ZMPmNbi<$onuvOvV$(y#gyHwzt$U91|>*sfFnl?HW
zzhQ@j?L^@M!&cZF&yjW<7pI%bUGOOSif!bih73bWQp`Wzc00?)LrpAp5@Oi!q~A;=
zDTbi%XBP{SWm}4q-?aYo%SVBG<BQb*^_nk!W?5&KYy(3blb7_ek`h~Ps6Wj}q-tSz
zB;^i6v#IpuI;@8L)%vscuAY}21O9=k4oInHyjhzeZSAS#ojF1#jFwPH!QM)0R}S;@
zs;I9TpI7yqs&^@htmPDqFC$N;^D4aOwbE=&%aK<e=F5ue&Fn^ASHaZ6dyY(id#&Vv
zUrJ5~y$0V@>e4Liv@{@#k4%*y7{mdVNysDs=>;_8;MO>ckM_k99Nn0r>NuLC>!t-C
zg~(nbiF4t4f&xp1^3&|9bfP*Qi2wJTPgRKg3mez}u%^S5U8%EU=^K|G`MwVFNxUB{
z)5kt|_KFE%Qm{zGw08W{B*+*#45hh;7r1qcCezq$j`Cywdn)sQfN7t~<)q+W7Jymp
z-`;WBwA$KuoT8%s?2H>f*plaqV81Ox)&+1gNu?)xl&DCCQoJvgj777gqZ2`_`88K9
z0uS?Em#4naM9?kRQ7#Wm#lcMOU>MSCDLj3NC?i3Us6b5H(SCjBHNd+cFIu=@;lhOr
z7A#z}2tEm5#)58?rpbG0Hm+TsgRA#}HwV`s|8Kv$@S{aIy$FBet#^KGFBCvbAqFZP
z^&i1IQRFoy!p1C)Fb7ELh<w44R9k7kHW6p&#%MfD)4oVERh6BzT+ibuf>_0gMWo=9
ziZ3#LmZ_a^`f)R#YZG{D!y{Lned<N`zPHtp3k9Cv_-!k8Tp~S11s=zAaCVyyj`R_0
zkHt?6ea!dpvqNa@hgeC$=(wBDqDY@YR8q9}+Y5pLb{sFTV@WV+<-WOgWWX0^Gu2hY
zLNl&^`twZ!dCsjuK{9#Z7ZyR`9RQ8Qd?s?|+S3x_L(wLhV$kE_qT#wTHi`MXWOkJb
z4*J5;^Ld3N-zR0)-%kn+3n06P&E96`N<%7v%xVR7?d2oJjT<{|+=Pjvve^-%GcB03
zaYXav5hF%c{*4?3@1sVJN{@t(^eCK$|I8?uhF8yed;X0~kA{yClbT12$Yw{47}YXj
z0{lYrh?ZNoo3;o+oV>qX^Ux(FegpOdf%r)_RLrIa+_No+*?H%KF%f?}M6q;w#zS|H
z%`|bLCv{kk*IqKvH3#?7SUu~o`r(12U-@m_nvI%VSbKG*Ia1{dv|ROBo3l&CA^<G;
zM^on6KhV;)S;MGiY<q)5Wg4%-hL3+ulEPV8qQxuH!!eBSONh42cSM-)6>HjN0n6~e
znOfBl_^+y(cxv(`3pzV?6eK$@TOgd^23ZEI0oqdbNHI}_<-Br8G!ks&GMPrwu#}_I
z4~cwE(olB$xA%F`lCg)C0oK`hIT7;XW?|n|JH6(Y00%r%%Wl8Q&x8Y*p^l2j<LL~D
zJ4{mWssvdOPiIm4;Tya%nRq;thMDB+zThp!WRS3R7B<0zkn#H3(|0-wp9d$^+0T-F
z!cQQojN%K{Muq1ZJSgn&BpJ3~?YJ~i8)NFyEa#uTwD8f)+SH_Pjl2$y0J09eOEG8I
zx#!qGA~NH{4qGaKf!lCR?LX_nVJgcpEzikvp|#T%v5!6$Lg9@wK~4(`M}?cPT@|rp
zO!MoY&3{oPC4mt<e8o@4)$}Azq^ZlWg^foHbP05xSJxi#cQ!IC7M=ars@>bVY#ZYF
zA~*_?AzFDX6(!NV1Hn76O>p#_?5;X7845I9@xgh?x@1#0GKT)=(O)`>ybadCTL*9W
zBLDcN=S0g`cK_2$w*Z9&r|Wu#$~1<9T=2+m1(BqNB3sy6LEds{5Z&AoIF1XGagEE4
zMH#)C+#ztf^7|j5w8QRA5{I6?31}u^u_-x){Q;VKyvjJ?iX<@$xC;`Ysi{BH0^v=x
zI-6-2`%+h%xbx-{-jbZ64r-0HrX;?X3a#h-Eg3!RmLdeOhFn~CNtjC{8WJ%-<vV=*
zD~e?mbnGexKS2_gV4Hkv<M|2NyQ6UQ3_WiL97O3s9~um)k`mbMXR_U`gt=JFQ9roF
zLP<&;D=&h<mtUy+Zzg%o7aIz2fyJE?wv+-I8@y!Y1c<(@jy7N_7=G9U;vOC1nnkhk
zH@vZ3T{*ipNd>qh8_V92xAiXY3yA*;Z-mlh8Y}C?oc3__q3o$IeV=cWw$J;ohBSJ#
zp<~y(In0D_gN*8`E?Umci?D}i1-#OmAxH9WC+-GIZ;LufroX>hS}}poC9)ghC|(D3
zSTcvExTG%~O*UV<(YD$(Z8z3gK*(m|1Bq)(rn2PdVRgZ#&m>@@oZoh1IvSuF!faIt
z%*XICKX+>K4oxQ2NwAGzLEx{uzn;QU-bHCjFbw@b=TP+$En^5t>cLWs*BWbl<eUHu
zJNiM#0zLFj9Bem|=-attU-(%86O>=^*rL3w>_+zyxC->qAyk(19JmG8=o>Dq3M>`e
zT`zov{U6#MO;f4*bR)w>rhK?tc68AsREB#+?i;y@IBo08PK>1>f|z;j*V`XU2S-G(
zd4J+oaEVnQG(_oGeyXbOrc3&e^v?m)RCPhYP-S#RT;)I#Mu1h;>^4!~bbm|SpGi|J
z=bQMe$y-H3D}tvADht;Ggce)^xGu_v6Nfka^P~+H1Y7pT8>vVj#sntZ`N}!<T<~xI
zylYnxJW^8jgqM-@&@TuNG+=XQLQ{X7X?mTcOxfRPZ!j2glES&|Ji2>CDkhcnpVx|G
zJfM=o8#tX1<L7_-+Oh;e-0|jFndv{<ykH1637lkuFBH30|15&h!I+3drr>yj|0#Du
zz$vXcJ<cY;>Tu!6#XqT-GYW43UNv}?|G^VgomXT2L>(7S`6gd5%TG0<7Q{xrSm?-M
z9$vh0nuJD+DCvQy=e}(IK9;5ac5cB@aw;ap0Up6I4+^-Ul(WQ~uy796`v&5H40Xe9
z2p98~s5x*INpE|6b0BU~iYtzZHT|!%)(X5S7hY_RU}uMu7Va)=dHINVyn4*f3J;-^
zjX9t|ENu~3>a-po!|Q|1^8GQEI&&qt-a{#+Lr+pu&7YfSEZ=pGjUV+5kksQ5{@IQx
zSdy*pEarZ=b3!l_xu?xBazz&$Q7oX6jP|cDJm*ItFOrc);LprSuCUuRx<FwS`I#|Z
zQt))T>9O`AcJw5f_B~^d?MX<LsBLY%8pqxMl2GAbg04$r5o=_|V$D*ph#!u|mBCKB
z4k{^jeJUMqKj_P}-rx7=*i-?MAp%$LhpgPcF5nHtR#Z%1w!h|nPk^gQ`qdRzU&MKh
zx*0BZ*P8Ys51xc6cbTFgLtu}_9>{Ds&hlew>A#Qv+>-g?doybz98-P#r=4Qa-F#sv
zN=M&-Pzg&%an}~6IMYRE@t7E1aP-?@@63yxgrQ`WA;O<PQptM1Iqp?L;&~-T-bhMG
ziI`gOY&vGq*7>}sel(Y3s9??9uN^)Ib8`j7mW{3NTsyI*@qt}=Fyx?hcr74?#~>d#
znpfQKPpFFqu<={GYWgA-J3|@ZvU(2z7TiO)WgB@x{Afa=DN1^NGNfJs{;BlUzqDz3
z(S=`^A-Dp6!~J6=?b{EpZ|D2|)UrGeK9Q*9B#@PBrvB@sc1)QA^u&+B-gDRv#Lkh0
zr&EV{qkeL9L;cCG{i<S(7HBNiHjzf8*tkS6oK4q;xm4|}ueSaC!7;vQAX+<SaREmn
zDZbGhs}0>RD@alT(Euft3dJWI>VmO^4_ll=AUx$)a9PRa9y<I9=H7NU%nJv+=J^1c
znztUFN~|)4rRe3jX0ZT_`*`-S8ZL0?<XgTMa`60O@6h(u?_F76?V}@;z9`COPPDNg
z3Rnsezg~27zS#NGBS&B#B=AaUuKA*01n}YYBE(wWU50pF#*j#GIxnqaF<CB<DhnPg
zww7k&_4l>sHED<Hs9g#c0%E=CAN!<zS=b*dD}cnT>`jWRlTCUu?*N#_s`Gg8NmaZk
z`(2#h7jV`NW_mF6k4Ms^kO=!O#2I?x&EIKq-Xwkm`IB_p1=lHG5Iz|HpFxH?=CZ5D
z)n?MM>d|j&iV2$A`amsv2>qO*m_?q{Phkm^4TrYwScdb(n}cvQ!7-&T=anJr%3Q|K
zlN8vfJVf#-Avns%OnDuGBJ8J(iyh1t0nWL-?v!}AhHg6Nv0sZ=c?7!F`SVj3OpOhz
zp_yRpj!s}63L}Ule3<#XY~%~;o;a&6P!=VQYI&@^Ad^ALu$T&=W5#_@4Rz^sxM$=T
z$cHDDv1zZY%9)aieWJ02i2{^v?Hk?q;1AeXfie-{pP2vtx9l}HjY=wM3Kl92uwt|f
zhvdnJ9kgldPm6mc?M+djC`(cG7q5f3N5+Pv*htF7f>lzeD(}2Ca%eo^uc3YEH2e1v
zFKx(KPBCx)e1?zmrREo~6CN@U{=p|TCE=|R*(43*$6*9ny6IIN`<7x*J9HR&l7d*e
z-%Ki#6qPi2?t0r49J7E=K0c!m0SSfo*uPV?jd!ozVR!8iEzs7Dk00NXh}AOD!)seE
zU9<}mKkAZ|$DFzl<p^e5`>NwJjWoEJ;2Pncnefc7Gy|Mh4TCFBOL4ex2;aBdoZ`4}
zv<dh^M6?9D`TVzj+$9%5OGTois@ZmNB}prJ)KA98y#R?>{_ojE|4c2-W*V!{&%0P2
zkM_QF@B@UxV~0H36?FctbQXn_VZhi&qlrK?cG>T!p4dW)>{|lc%^PU``E}RV$6{4c
ze}t`#Haxj|S35@f;ya_FjojET9XSVHE^Gk*VqV9t%!$&yg#B~1NB)LO26MCnz@I=;
zQ*<|;nJAlkj;iI3U8LAKuu4`AV|;)v*vj+te<WJY`ucap<n^ME|NV&*M?`8FD$d2?
zr@y=!yO+qjL3|;8w@Gb2(Dj{gIKVNfWUSst?g&Gjy;%pB6hbSo`ouGEz<A{qXW5Z)
zij9$QsK?)=Gxf3h5tCkS<86V*Hhd6eIk{^{eW{1uyzF7wUij8MA3w|gW%RP{Jt)sk
z3_EMLE0GaL;jjl2IAMbk&6atqqwr2$0GmB=0UZA&#xYEM+B=;EowUyL0KpZk4vvAG
zw>CU6`OrhsZ1sNxCOy11Z^{C1No&pz!~&tyH(&?mvL$zl1;hG$Mr~<?dw-m3OkA;}
zv~Vc#Cj?GKus@GAdQ}u4EDTN8-y>jKdRyI9egq9odXW3in)>@i!{}^P^>t62+(Jiv
zY;;(7@&&JM$rTHtu#*Q*6-xsk%E!RnF+ZOkr`bgC&}>6A$Z;`WT`c%`r^9!UAe7u}
z9NXEot=8M<>_4l=Bq~G{Zeb*tWD?DpRP)vAcPm;E9WKpm>*bU7&&CSwmR9gS_7&Op
z-B<Uw@`V4qR#4KChz?9iVQqMU83zqC9Krc*s#h+}P}#b%kz}1O&0f9<`zd)O#hn95
zB@+ZSZ-~WW?x*MGQlVt(v}ZTl5>~|Lm*3hLW2zzxzzYR7lOMtD9C_zmb>+ds8Mu`b
zKO{|BhZui`PYPb?H%HXt&4~L_bv}-wj#&ZTWgaYk`Cd3q@xgj361{e#uv<1ZJ~};J
zAMr7<NOkiiul_1ns<ZXYSKADM<jciUcP!PkySBWT!9ox=;p31sV92TLv-)N|kLkzA
zUiCbW0=eOmNRkbtJ+Gh#Og33fMMBLSci!U7t(pr;YHh!^LP__8M=A1-Z>7+_%+505
z_6@ROUtQ&fowW-|Iv9skMpqP+6e;RBihLf^oQ7R8fq3m*%PvnxV$96XOu}T6&k0LO
zfzd$IE#EErXz7|RNyu;fWd7o>e{Ij(&idz{d*hi?Mh9cQ!_V#PYKQI0L6K-W){0Xx
zFPewuQ2e-+14|V*I1D{WG2^}4dJD@n$NgZ{s8EBCp7J^`TDo8rNbH1etCRb9y1#**
z@pk*}t<O(M#1A`+p#q0BpY{%~=|xTb>HaZe-<Nrjlzn1(DMUxECTzSck&ZSGW6mJW
z8@V7m=0A_-m39cONL@0P|7+MfPhNgWqOLK6v2G8TOhd%qG$t5~*Cx3U@4y3t#fO_N
z+cR)NR(brZ&`@u(M9WHcR!*1EJr5>4ev&zrs}d#E6qOwL0F@Ls0D5<UuORaFy=+4@
zn`O8_y6NnVX8R33|FkzdikO7NAZ;dqqFB2x?^t;G$V_(B885Hu6kKUrSJ9I4rt-yz
z|EQ~nz`gdc=?l%>?FDSt=lzT=6k4B{RPm(K6Q2+;q4`i_$Vp0Sl^p%E88sY7r<xc#
z+04+1bJpdyOE_Yvw-QXL?dKjEh{cXQ_v{(N|8HNiF%$NUxoK&qY3|r1=UzTC>i>U_
z-Gb!-*ux&XiP)B7<dv0k*eo|^f$(H?1H*xoCg|V=vaCTs4|^p=0oaxlB(m`3wT;0{
z!jC~|=_+&45X;2EaM45Y?2(HV8GTZ>tI{X+fOIUgvw~KY468tYHov<JiV(DW!?K*>
zjIxq?T|x%rlX|fb5Kl0aJnt-FQ;BdQ5MvwIX)Ct>ZWNdP``TZXoaXfxA$^lcYv+z<
z#?%ZC)4qTF*XS2pf#DEV<qc`)UC~4Sc38ZD8=ifnl{C`F6i@^py=%^@7#+|TJ$8k+
zWQZ~JN(w@1@$$xVB7tVOEq-!GqGJ~pa|H`R@X|^~URIw^SJ9Z%v6@ZOp}&VlT)AYs
zAPJJO^Ye>YuHOIOqaT+<+u*VC5?oFbqkVVBopp5=tnS*f{)lj-ITBC%>SKrAncEE!
zKK!t*KyU?j1a6B_RFpN>O=?KhW*fSP4P&q*g~sEFRGqfP>QcZbH7b45GMfmAO4EM~
zvS!aF-Cqg1CkVW69#=h|6b=`!$$*p;d*0&{2^1G|_WC2LSpP^C&Vs9X*pMZuDEzvq
z3j%oqNQ#~kW{I*LY1jKL|8KA{9HHVh|EKvO-of;FhFn-V_P<7k*~H(6HQt|7uzv}-
z>ZOAbEcxScQF7n}6m#<qV7#|*C^7UTh3MWC?9SK521dq1rRq_dy{KK^g`Gf4FuU^d
z*6YGG(ONnbO#aQ68h`ncEf^fB>z3R;C3bkVpAH8uYU8oD3D6Vnp<_YxWOuE6^x@xB
zr_lM$)OZ3I&ctFg^B6WQ#a>O0tw^PuC6Xzq>d%j!l4UuXbn-=j>1Eh#V=$g=dA_TF
z6{wE2qsJ#@vfZ|QA1KqV7yY^mp#C73@_?H{bU#?1l(gcK3<V5`Xze^d8yTLaBk;?y
z>$l2Q=Wf1WDmx_1ii`9_+JRGGs^6dEV@8I<{;KHY=>O*4&Y4b8gPr_v_Tf}I5}+=g
zzXKzy-EcHm3W`*;#T|E&1X^WGdZP=I(RfEOg!mIkiuT2IzMsoA!L9Jlj-h5RlZ?Dj
zenwrB7i{ahQ{(>V@T#hs@Qi<dv_*97f{_2=w#hKl7d||C#4Sr(Wy}VH9gwl93F?i(
z@7jg=U2XZ{EED&GM-h!@kN$8M4*`=ZTX}4Thu#JdR?CZ;^wq5=Oic!{gaND;{D*;&
zTrk>j%4!^Q0$1JLM@f5e&qJ@-<yujXjNP+)*6E&AF|{X{#+nOv+VY@$QqWOsSn3G6
z!YhvfjcC$ep?1uoor<hse`HA&b!^N8vM7UKT5^|ZULS}AtK*|)O%7+H;p$_*D>_mE
z>-oNU^5N^QyMO-Hf{7*cr0E&ZRDubqxO|eA9Gie?IA<k9{{uPFLE_IKDZ$P=>V4pn
zmhKakzOKW`iRJ9%V1_^+)g^ZxJ0d%7?1?Xaze%uo)o%ao<zwrsYAH4vntJ1cZSX`Z
zZMG^|=yIarBA%zg=dmmOZ^s2Qwb_g>oA4**ek#d21WZ~H+)QuxIOsS~TGm!Qa@E<h
zCvqGIc0T|-Wx(C6Yq_@_dyjwvYvg<Tq!p1;nPi@e+O3ae>fz@;WwSD%-;=M&!_V3c
zuINC0Qe-pG8kvGk^xEH@TN+V-W#bK}zq}?_)MeiAN<^`;33w{z42j?MZT;a1U+9AO
zK702{Ce+M@Uo8~D+UG5|pcXc5Xvdn-LU#cv8hydE#rvY+o&#rE9&aPbXoehr8cB^$
z=f+toR4&sGPg!dfY_Db%fP;&OR_BJLue`D7qi^zsVnK4%HS;fT7#^w**Cr=k_|{g*
z$*=oivxK=N&`rXYBqH1%gBNwdk=j<D9EA{wVX5kPGIYgOO95Kuz%TH`8kr37Q>S9{
zyt4kim+rd!;*;haHGBHh>62!ka`h$m{=zGc0(`cevQJvcJrCV{#aU$bo>bd&P6aA<
zRS`dVAE2Zc<k4^s&?iMsOTcK@KvdO^Hz&s+swbuGTp%9KT)n1fLhNeTu9PE{O%O3b
zfTh{1j|%xBbtf-1q^=zgG_iq0Zdto@>53gWTh0q0$*#onBo{OR5z>OCd^@MxQ-ULA
zPW!<sJs(4iKaHf;nwvnh%1X*o*?U_Hyf+>$K+j2_b3i|>6X*(lfLYl6=`~a0HFT`H
zHhba|>-1u=_}jI!o-jcYcp1Af;SI%99I61p)V4Rz;bQ(oEWpv3WTZZQZL10ENgBAM
zUM0SxN;;@(PBsb<K^Ns68&)m*>f?_;c;%fhep<g#6t;Wm$Tg!><J8+9^@!A$ft4z|
zPgHK$y{nS2mi3j?IAQMqbwEBTUIiqn1SX#^{CIJyCd!c^Cb0myE*LpynTg9J9lJ5Z
z2|9KWVklwv3qd-5*yxwqr9#1yu51c~=3F`Zh?DRBV0)1_oL!wo-W2kDxvmKt_7(Gv
zd=I!1{!)Dt9Qd6(MVs^@9cuhpB(;j2uTL&zfke>s>=lkuu2Zvk(aZ~$kyo(&9!!e4
z?{A-(tz{#%3^(Q0Wx1kI-1O{8hyDGWtsUTiVu>b7DFpD6Q2?xaWO{9?fo5qY932rD
zJ~lb#)$O}vOGVy!&w&Hx90H3j$x->DYCsgNgL7-kMN8)kMS;)p66%KIY%Z0PRuXG(
zZ0uR5a_wH<aI&%yRE5nityNx`ME44I03*jij{1**IayG9k*w-B=Z#DPJBgA8qAVAV
z)-liG@VE|E+Cy-KO`mes?%mF(vxjF=kt^HmLYLVlyxJ0B5=o{uHe&j%E4zxtoQb24
z=W?YsdM@xA!Y=3c>6pNs<RAx7Qy1onR<XN2Y6$UXk<=8dRhM*+lEB5o>Dc|6)?M*y
z+ZOncrT~`PEEHvP>&wSCN2_W%-yug{_r`Xx#k-f>I=Yrk)%|-D(pZ;G2!)V;q`>sI
z3r99kv0yS5hzBCEMk>L@CcZ5cAYRA2jW-<iRKtj0hgiyjz{^l@Yj}5EMYPo}May9-
z5y$P^vz)XS>-t4+U$i2I>Jv@@_O>f-*qOUrInR4ax}QENdIvUXAuSk6apSG*uvk+x
z0?#VL(Ed2rkdDp$K>@0PFA6>;xTd^`O3Jt08u^b*c*;^q*xqLE(q2y-Mln=vm|=a?
z>Cdd{g0&!sBDGTl1>P>`?(UbGNo!~Ccxn2^wRy|IuJ}WaKh4~ms_^*8I!v}hz_77-
zzdHqB1#xRkQsj?Ah=NdB+uuBA1Y9(ViA2v^y0z%oLhGw1#%pOh9H?KYSs<dMm>%LL
zTL%d?cbpZ0TMr>S+@7c}k$_Kn?jnejNscmXHbUC2VvrBvlMa?0eoa#hEXr~*-xGKO
zvRCqNw)aH$z^6ZAXBoZM!*58>es@s8<{rRmfeTt4NtLj}(0(f^#`T5_XCUPbebeoY
znIsd8U~xsugbdw~WaH`iMXWCZokf!^Y8Hk~j{3^zVaZYKgNjq^%I6EepB|5MzTq)n
zgoUe4U-Is<?{=AvfK6NR-Z-wH+pU`W!?Y;rGXkf;aWp&WO$cG7p&7WT5B@Zgf@-Zj
zo$fAW<Jef^i`dfJ=Dl|MWF1p5M0L|=S2P4-p@v{+{E1I(l`N%G`RVTDAxt>dO#A9T
zbFdH!Z9TT>=Vg=6^X3nu!c00-9S5tNs0ohoCm1Gj)N&KE>|<k5EGos@4*#VW{1z~m
z{3tS*uo%_Wu{*CWKS)fbjwW^<iRSUG&#PjpXOes)Gb;48Vo~X{^iO7iKaphu*R%>c
z4t(z&)_#E;xdcJ|Ze10g9TA1Sfwf}l>aLrA(a4C8s<~Cv(P(1<DXb0=KF^yn|KrgS
znx&?H;i@?>{<+8Ns6aM6{>+&fD(<hL0_^mQRun98M<+HE#gRB1k<Xdx)Ag~^6Xvf^
z`bJ;7i4geDz@Ha5NlwQ@-K|QQbVE)2*fm995GWXG7y;P&w(st%571mwcv$TCr?<jW
zzrB6avq#m`Br~Cr&7tTG8x$Km;NzWyYh-~U3Sz7M+CQV!v2pd;lSd_e44Y2*I4XA8
zwoVhIG>2WB&;>31#-0Er1uZ2t_eO5*1;emoAApO2D8ZCh<kUdbH2`Pfy?X)+qUgGM
z3b_b`gir6)t(D$>0N|BU^>sA*rr?xP;FqFrTAGBlxJWc{$VI%S$>4?TpQ-nrHVp0^
zIj@OD<<+eBba6nqNxv9Bo?SPFId45!ChYY@gcL3qXsILpeq%Gk1rHzbz(yVdB<06h
zIJW(Xi`M`C?6`mW#*7M9g|FQz<#h{8JZ=M5x@ZW(&LhDi^}ukxY-;|{2TW!BX(YwV
zw0vIq1~RF7e<XAVutFfsq`Y-Z$rN_o6Zl6uTO0E=U9~(XN?mRGrkjtC#kfX>NyZMD
zwzNpnJCKGZ$UIZD+S-f)|7x16qetEU+f7lL9v-HAI4J8C8+L%=k_ewrdT)Sx2ynq0
z1plG^ifCQBo|K=aGFTqb4bu0l*F?n>qO`PUMlV#X+xs({q3KLI`Clit8LE;uWNfZi
zI<kF$l;9)tNK_=7D~>I7$z|%B$A6N~6?fhgU_X?0;;BmpI;9xh+pfa%?Ux+#uW)0m
znx6MEuL#m3*(hB*`aMlHi;Hd;_rL#F^OR}t>b8)##d37-6v1k}K!?0%j^S9wS3hH`
zx9rcrpQogP>$W{qw)b2-)I>**+wJm2<Z0=qfGXPJ9lpkT|KA%g{%|WM7qfS~b#}JK
zAN8ledWQe;v><7c0ozDkL8)VO$zrRU+mXAuC41yDk#9XG&=`z10Uy|)?~1KC)EiwS
zDapLw*gL$Mto5f>!)*u$fIZO=#2-MdaFAm{L52<CfRQ+i15R)-7Z!zJDiq?pH<}HF
zfQzsd+C}7f86gNtaG-kzBrFUoGE5L$U)TV?g(5im!)WH5O){@q&VKe4h0`e+AX#@5
zVfozQsnW<%O#R1sQRMB`1*>wrUew?sf|`;JcsS^|m@oW#ebYaE^~sv>#9P15Egg9X
z{g3+R1dZR3GYiX(PXvcQ+#&EL5VPD?RLB*C*6~a!*Q2i?6l-jGzMQi1Pr#o>QUmnT
zUOlcf%mW8LXCzHMAcI?3>NE$yfpF~KudVrC)bRjiK$^dg{fYs+wrl4zqpSQ(v}V+4
zBSX!>d0losFImW5lxDb7Mouz1C0pG2{jGQXqS~UnW-bnb#j%NKj0@f&s9152DarO>
z?n}@lM^W7TE2kVa2XM^XxyPS8cP@OKc;fLVAAQnEN1X(pbLXDmO_bm8>j|Z|xyQ^o
z@q`mknhQATMDMeV6Hh$q<l|2~aqiqXrH{EMpKua>o^-;|C!a8P-kiBdj=8hbvXvb6
z7u^pf#fPtyWXE=t!q+F$lvjq&ve6f9r(?I-(V4f~MP)|^W-K8_+Xb>p`FyTW;OAc$
z`hU1oZFTig51j>}*03`-%9hZfD+M#R^s?iAazxP4Jc;gcJ0WLDPc!ApCL}2{Z=Lf0
zf94)LhniaPj_v|ShNaS}sq0;{U|Zc~`nHkh^`(iqkN$#UX$l>mTpT%siYJ?AKK4My
z7f62sf>1<jh%$;v?!kZ=Jl`pso%xOqh^Ta4{r*@w7QoT)>H2W=*<Hv_>=Z?=_CBG?
znDIH^y6(5NYk%9kWz*I*TefV1kG1R8t=+tB?Qi50zTw0kZ}7i%%NF=pyLQW_&0E$2
z;IsU<8MeYTwr(zc{JM23tbt$GZiU}z+qik>)&eqAQhPk~eZeaqMViW{Dt<mIh5-|~
zu<;|7cPe15b9!fIK9|>}99BwUHV9NwzIEvbn~QGyhG%9*hew&3SSAo{4vv_QSvT?p
zQz*D%TdR(}qH+8OxM(7#JSsXrP8l8XhIqyn`Ap60`*oBD34fkX3N*Fk)pJR82!M<G
z(rN0c4m&4W`I1lSnxc`{I$rs@fPt3Px$ga0KCUiO-F)fWi!aWG4n2Xl1#quqtPq7A
z>ga39h9GT|%p9gewat7^ao%InDL=-|V5G-9@;f%Yz``dY3V3CS{Pc5)F@ul6=S!KM
z|7;i{#P&HP-9BwSY9hYChGSwflAKKa>C$8z2fYITQv>tf6<nzcoLfuYjxEVxD{Li2
zszBu5E^fJFwI&zxtM3{C@e+Vxt1klw6M`ev1*t9AH~=}iumZeeD-#j(k7wvWDN8<F
zT58g^KYr*B9{x0vA_HBg^{#ABS|~2&KXrwZFE|E}(gS*#f~m`v<XY_u&mKu7>xTuV
z-o1YN{h8|FlfO{S9CpFm{{0pOGdIH%59{Y;O(c0sHKe%y{pB%u)*)VQVj_O3`JUZ^
zV#wHL55gk!O^M;}-}tN2{>j)-00VI1W%l>{^uUXk@MS2N>;e4I;;J*7vbE@ug2xM7
zrKs^Qw-s`_4ZMU{9*i~#ns8h(-q~_fU3KKjrEQj6*zwlflz(_Uk*H<wSfye6=DY<C
zfra^^fUU3@ygQhMT(ZABlCG^Q`J@cn$lb4?#Qs_M6G@6!-(EeZl#mU<#4|S#DJ7v3
zxj{DAZ%|6X(N{lyG(%I#ROIZ1JMs%=#-pjb<w7BsFLteZ;DjdyP*5AY+mSnqFO%o<
zqHTRQFM~P?;Y%XwXVVP#;4WUqJ|ub<c)UmeLPR1keX(zT{Eq+}A{;-S0LY>QHeD%v
zb5S&rt;uF-U}ypiRf35%-myi*+<D#xBxrI))h-%ckB%6gI@EXf4ygcM=g*Iykd4Kg
z!o!-*dFwY_{r}qg3n)3Rq+Jw!_pP_?TIard&N}b^=O2@xsTeY|vKYF>Y)K|bwk$I<
zGn4Ew^O(joMwyvewrt6g#Voa&yQ>l^8Ed_WUDYkwV|#3$bFG$LUso4qW@Tl5`9;Lu
zv15~InX$3IjO8A1T}Lu)P*bP(#tcoe!x7$S?69xlGhE{L!S6@XYLk86eVG<A^%Vg_
z9rR(PRK^L&@N=s=hqDWgYTx@8Wm_;!fAYiI1YzZ6bfEF*Rp2KTyZYWbDcLgr!>W{G
zLlgjvzU1m<Q|N#7$S4Jy$I<l+ox#F7)2>+|^FUCJQ_1K&lu=lVi0YmQ(l`n~KrywF
zp@`ido|TTW{?M36D9%I+kx;CV@kNhWv1?^M@&uxKL3wso%+nN@{Yt;(iiXCQzPf30
zODN_I$EIKWWk2W0*kunN)jC8a+bIi<D~az-54Q$~=EPz0kHd1mpOF5ZN{W@qD$d$N
zJgqn)uL_zyw+r}##KapoTG%^ji`(uzFy8<qop}69oKori$Bfip!%yorr_0|xcZ_%6
z;C{#TtELD}DK^!kj|NXH*|68niUxi07#*XjcsSUK$-i{^ip^ynyab0-D#1O$mi?Wl
za1R3nOBgZ14{*zxD!y|B3$}@AO;dCsnTW&X@5zjbJy5i<v%a<p0M-$8chy0oLan|r
z|L6j*v$x-pMPb_~m&SbQWUM7J`)|uRM=csIXz<`l;7{QA(+wBR3`T|qF}B7JTZa>z
z{C@cT0w)NCtg@{scY(SMS$rmX@H?V7SR-OkN^SmmhM}VjHTI4UQxud9CpHEBS9hqU
zCN4T_k|!1%J?e<9n#IZBz*f_tDis~9KWA-T;PYf!L;iTklTI~Z9X&X0<1X*8I0Zh&
zetbOic*@B4?Rg-V8SoUI%7)f8@ct1D9f+`nv?t2a)9*j6kn%N-iQL|2Dn+1IbsdDl
z+p^Ps;Yg1^l(}-V%=17~Nw>>_u>INbe~l-bDLQw}tJ|!eeh1>GK|q9$0sm8NOw5xT
zlv1|vxDNdP9gpDeMbfIV$G=tc*w#UkveCe}(>FR`Qt(sqvE$m7?7TlMIU+FW*k{|z
zs^j!Mlo-`8{pCJUYX9f4>4s)H5WC>5-s%7`oPG?I6ihy^fFyr&NYj{bD96s6n_#B;
z{gLpbq%WGk?mH8t*1;l>)u0Ld-aQjl2_=V0E;+@g=fu6Ce6m0_hV#tKPvqq%N5d2R
zC-n$@_F%P=>QP5Ww*Ou_mWnnXzCs3mRyx*f5z6Q=2rF)w=<!X8HZbEaScc8n+>j*~
z!Sr)>&n0o}&8nR=rf|eM4E)94VJveEzoU}Y9(o>_8X;xq5llRtf6@^IG7mnsXPT&_
zrnKYs*gi8CE;D&a=hS7hT71;aTiw#P7mXan(B8df+`m*HbA|v-vBw>oO>v?{v&7r)
z)2!d`%fGhq+(cU-lus2>g<wnil1{B3C&?kZfjL1n!@tJ?b{lyLcZ#klPs|<_%`$;V
zq9r~l(Ds4ZTl#9Ir{&giQ8K}mtXgDDE|IaRRC{>;ec9-Uxu2E=Ll!D`jeSusbAsts
z)Q;CL@cmC;!RIf$XxZSb4jmdGVwa{rF*X|W=0df;g6PzzR2dUY|CrG|{9{<O<}hCl
ze=QEmW9cT}Ny}YB0=tcA_I8y=mazEr)1K+HMM>S>v*XH0aMbM2rQSsg*nLu?TQVno
zfD0SIYF@R`7sbhQ;Q?4~^$V+J?UT*jqslvO3e)jeJnT;}j6ZtuGR^HCz;MY0n@<w>
zi2nXkz_R!UR|OVQN1EePyhv~>r1J}w0pR<%X!Wh3b@93&a#nlqw}EUGHl#q{qU+Wh
zt=?QDO$B4DH<KLqxntO6d)0AgcS~irzt6_K!CiK2blz2A{R}GW^=BT0u(4OsKHUGm
z&%9%)iL>PBj@16(EPtdqc*X{`Q-{-lB8lsYrYqPRE=WKu$`%4q(vI&vX^YBZ3x9Q&
zeh>V9Bt?UdKi%P8orKfcBnm)MCXo-fJ}k;6FWRDmMe@nAhNP`pQQ{$_l1=fGX-)p*
zGXo!-6Ku#dvypMPZMFq)o8VB&jzx0qWpq!G6X5uom5;_A@sMV7X2+TNF`jBE2!^8O
zyjzwnoT>?ed<n^l*{0;nei1ZP0~jUE6+3l<GfaWkbzac*lA)_+Nv)`grW?E<bocVA
zYT!pkvDPo3Y88#+1q<FZRWc9cZw1cDsF;d|DJ4SrfrOWVE0Bh`DVqwsgT2Zias6bz
zY6o==UV)Bs<sFJts;oJ@AuuiMy}qQCtg`s+^nD`D@mo2gWVk&8uG{t5Wn&s=y`eiC
zL{2zIAj#6h7J=^Ds#9EobMK|w0^vEIR4=vvg`e?j@jIG(LZS+2<)%|w^D&mHX27#-
zXx0)DX9q7Swoq*&!s#ABRZFB~bMf5&Xi1*C;@JayK3}>qbHYnlsT{->R;j36=qQvh
z<z(=f8LF~!Rj;E;oyz7*nlRA>!6d11uM~So@lq@m1tOy>WT3_`gsWoDzNP@jwUTCI
zBZOL1;7)b560jE7s$!0Osn6oJNJdEqY8E+I4}Ueh&vBrCitPehe^Dmh_W@fWH>95J
zDiHMu<&D?itir9)2-soPA2?vK=HUE|8Z0t+EMOhSl?QZ@>+Dyrq4J|gHJ-a!?d%t&
zjVF#8F}1nrnl)ybvmos6>$mtfkG-Qq!hp){))Z_h1tl$6%9n=@&UnDkttqN->6+?^
z`~x=T6CK4npjP+8Q?U&SGuPN?uq_z5r5BFBRW?L4iMX^O*<jJZvyxr@`Aj4=@7h}?
zwb0Q9&zQ@<7RzOU@7cI%<>I9qw_}%+j<;$FXn<MN)P7R|=BoU=$-y#uY{VCy{79!D
z$c~|CSn>mm-mhvZ1*HUzn!<*q9pEbnpylvd)pT>i^5x4wom2<@U03DHLjU$}zuf_>
zRp25}K~3RJp?%%@bsYlGK}KKt*Fw?B%~U}eIqVGO#P7~v(kb|1Om0k)fnyn%@u^dt
z6ykerPqOKB_lqU7@9~UpY``=BTkuV7SG}$A|Ms^sC$BQG=BTamMwyp8)^+nZu{f}n
z^z_52=u#^^f6EV+(er})L?8M#+6U~>$3=c0{GNdmG7!jTmp5I6qtRHLY!rAJ3r%@X
z5+xV=-?6wEF6^3cNlQ2v&Z<Mbt#s=l1y9&Z)AQfjsyJNFy7zCn@XXoM7o2&|XFGV{
z9ng8`n3+Yz1Q)<4DsKBlg`rVzU=_pTZdf5=oe~*`;p6`Nn&S)-sOp@%?S@0|>IRxP
zDzIGRg}x6iJ>jThZu_E_H}oRtDG2M9M~^-7r1KwLE23tiW2o`c$5)?u<otzSh#<W3
zZ^&^Y2M>>*D;Eu=Wa(WWp2S3|!y4H@G!D<G$bt`f%o<g(2=Cx}09eD3DxZ#Nf+HP1
zX^CK0l=3qZLUe5J>7SK>zU7kL(ItwU1s6JJ0zt8D6xXdtrf8`zOvvH{jR>aIA1W;D
zz-gU+&5`?G$L|<8Rh?3>>jL-23~+Pcr;}gLhP|h6bU28eAb=u2-;%*W#nSgBy?kvG
z&T7wKD#X}x-gL_4(uxPq+&{sL{;NO6qz=4!DfqKI8gSe=96gA^cv7moxc=^R^)wQM
z{4A9{`wM~73~=Zm0)Z8N&7sy5jTiaOe@y+4qxuA3pDM%6>AY(i(BGFi?bUumD*{I~
z`RmiuSkfwW!D?)91Z`C-<yYs@EbV0%JkuqZ+Hc8VsoDFAQiPC+7rURE6~^XK_+6~u
zpG#A1(NKXseuV<TI+=B;MpNv#7xhZnG`>B~8)hKBI{%#>NqT)om>%QLKd&nsZ}n(f
zZ@hcC?uw>TwB(KfP}LIls7jb*qOU(12j2~%C34ui$Y9)h^&5T{{Ju&GH&X<Uv*Ri-
z^h22@u}JQ24U0ZhCklh59IFV*D)6%NQFip?#&oXH6N;Yj>^4q!+g~_uOvFn^!?A3@
z6KXD8_?cWPW116L6I@vX^G<lK{N0t4A|M_BILd6;lP=KF_%SbTCn5`m6K8y=uJen4
z^L}o5<JuY2en)IE90;Qnqi70GPil(!15_$`@FxlfPN`O2bABw!`2BPwe?up5MS~#3
z{%Cq66ZO*R(4k+*IHUcq2ZZxr%N4V%Rus-zciTa6PpFWj*%(Xbvc8BfouS#t*oOt2
z_|^t%kL(1Xl1ir1v1#MhRaZ~-V}Z=6_p~oRCY$i2<2Q7It+!=s;E_b&oDcgtG+4D;
zayqx@5Hk^<)RpWfTBDw!{!mY3`u&|)kQDvVKdMiP`FC8|;uP!M1#OWQX0Q{ViOyRJ
zt`v?=CKo8EEyqi)<5;UN*e9H8iKP<_<8D|YfCg>5V=5-LherFOO%2Vl@aXVG8*!%l
za_!uMdkO*!K~?)Lx!~x?82ADj1AK;I<6$2iKk(koBA9ykCM?5?mkpr0>dynhtrZte
zgMhDL?rQLBUA+kWTQN7pP&3Y+<Yl7QZxPIrVU%8)5u~$+9Ui6AV_$MK4XdrJIh)GT
zu|p;?k<8=Z1N<764yq!=WEyPA=`X{@vhw0=G@qiAfe3^$!ReRIV`-LQ>FCLe1<{sN
zT%mR&KB-dPc=N^I=^GxM8m980k+ZKqB~8K77j6dYuJD%r>a@`fk%M0!z+sUCw$k1U
zVIdL_=&gM-GnpI=K4UQ|cgoT-Upq;^=4bpb;`dZiyeCuNK-hTSR5}tE>cppG7l}YF
zPEe4{IH1I_4cStR=co~xmPEKYdC))Fu~`0wE7C1=j1KxTGY+0L)d$BqoV{TahDl`e
z0Bo7U)IIL~d0{_Qpkmbcxdn2Z8Lae3<t|>%n<hk1pw}p+Iz&;mReu()rY<?KF`dc<
z57|rz0O5Y+wNx}S=J6eO<*DEiU*pUWM%NwLme46*EIufdjoj34IGUkWzMK%s#;^F|
zhD<1RXOH=-DwJSyDUNbhAZV8jjnia7eD|mXg=JRr$wE_8ZrUx|zCI+L_QsRvu0~5g
zH0A@_pG^3XlHT)ZCi3@QxBT*n(UGa74-SJb#2oxFHtE{7vh;`umGzCfUv~wZ-B!}U
zjTLyaBJpOgexpA^PEgHn8grlImO(;b^S>iUe@~yZit1%(y+5ChQ9C>F!{c7ydM%M$
zq?mUKZYdC2?7hv~5+0HFGnakI3xdRbbc!by@`f9d7e4&Tw`<<JX>KSG9X<V}5>Deh
z2;78g@tXX?RC?UL9ya;tvP;sDSR7p6Y@Vj+<ouTe2yX^7QYs2{^cNm)d-LDVIFQNv
z5AU%QjM_EMUX(}&j@co7eUdM5_}AdJR;;a8_&mAW%gXh%&wDL6pb%jhPe#0v!&mg(
zIX0TUVW2WpVd<9u_fFD*qH+mbX;~{3H{CY_^OONB%hJ^3dwVPP6XU{Z|26I3_p@<T
zxGBQc8`+T~5B*%)+_~|>(C9D|^JLS_vtNY}&XNRa<JAqG_`WR{Z*p<Gh=bh)l4+MA
zjxyx(OR0FuLx$S7C9yU3d{_#cDf|V$qqYD2Nm_&QSU#Ti@9f0)Pg*TX5X>X+IJ-6o
zs$?vlvrpc$*UabIu{5#Teqswn$AXdL7VQ|2EMDw;VQwZCY&x~W#rn?#a9;4f;^><8
z-rSaWI&pcsy6H+f0ZtpkfW^;d<1=sCQIUGA5=c6>Lj59+lJh;#`^J%}QAhS6Nx`Kp
zR^Cr~{3mx;UYqS{IDAz_QCy?@Qcq;uL)+yS+9JWD*Ma_OZt12(G`s&%a}UbT`1=~u
z>=tXA{3U>|RdiWbIaB6&ODhY@&WNXz{#p+uoj>N?j&@Zo-=B|Mu^H#I9hA`_$hy7u
z_{IR8zI-!;@tYo)(im(_6%xrCMH^_(S$&U=9~JfZ<}7zeV-%KqfG?FSoV`hCpN_pQ
z<eTuG8S}KEl?00r^cNkq|4rCkB&}?{2(zJT!4s3rh0invuIgFh9p*z-ta52l`2S9y
z{w2;MUa`J8FB8v?^3A;VLmWb)fQ0lvKZ%a)HU0~!I?XtM^x{Jl1=qVMe>FcspS8m^
zxzY<`o4nbSFAlCLlZpq<+g2_Dr%d!pNo&-<1@<`UdS%5itZzX#c7tNeXywgJb3A_S
z+H2Dp-wE4nRoCpTmopyvu0H9hvElfFPr#XUoHZu~^I30fOuBLGhkUnT{u&7abMIo+
zJW*G1;G6cum>@mE+e#+8i^ZaI-!t7Z*P$AJzi*&aCOxLRf?Qc*&%LdUQE%((PXB-?
zyf)7tDnuGmXKmAXp=1v%n&SzL*=NefZeK5USGlAhDVTiGEQ_5NG|grsq@=8`nF9af
zP63Oc!q?Qc|9fFKk<=<zHcbhqvO|MM>9On+BE(Mk8Qzg*eb*@W+?lvhvitkNP3^z6
zITCD(=5FlQw)Mk@76dB2=lgt(-UmyLQIk?QFD7O}o19__){6NjtcSp%UDMy5>kY(`
zIsB@oNVX}xh)i#fLP|W-e+yTEt!mnm!y*lHw>l7<sEVpP>rSJ_(hZGG5qiP{yanEa
z(Q_^B^<Tby%QXd0bnd4WP17ph&0z{njj>>JOUotO1cMtIEAtD0G{auVe$keJHGlK?
zC$==VMXLihnauu6wO)?X6q}c{lJ<Qeg|#ro2Y+Mv3C$t4asFDxkZtY5Yf_^^H0wY3
zNkQkNkLNd~QqjnLN~K)nOg!dfq>03FPOtWKTO%FIlUCBg*aQ<jbZtepNgX}B7JkX4
z{l5vjjihe*n!z!7h%hx_=2DWpf&3&y#nydGY+u~r@MX*7rH`gJXGaB({QDZ+5-m91
z0?$i*_tD-!_$IDm>|#X8;nRw;-dF5zw}k%g{tcb$PiB)|DgXgMDidj9#w5>KuGu){
zt%hx;zW_jGRdUvz93OE+H>`o(Pp;;^Oz-WV)S69uPv2y@Ahw2lg^x}iaN|{Da#Rr9
zR7G_w-%M$VjT=K3+9nt1ha?^Z_?H8ireVDZNA2ks-<utb(DBTej2BB^0odk?wr%OK
z6kY4wu0pI1KO9!5LB^>@P=?)s7pDc8=G+^iBkC1<(=B5gLoBsd{J{=+(-q-Sh2Y2w
z`%F?)zii1EJIPkLShCzt=H%lo)!}yxJ0?Lh4;!jckt{etf6ykVAHX#Qb=ACJpbZW+
zQk4w{3e2TtK>=!#27L^jtRg2{oTBe2TLNcvp3$;zChEJto9ARh0Q;_sj?;U8il$<h
z3)tIQ-7ryL*gh}xcgi|uMq9S7_FPV7k}a8dq`*XinM{aDhYxzbQ>|d|go8?O9lVi~
z3@?UG^-j1N&uc4=jf^_7*G479N%zFN!--tP(;N)V|60*pwP@-uOb@YvM1hVmk?CJl
z6j!M%nE}F=KL3$}8^Y0>`Y|`=-(zrB6y0@f(k0)&{<d^vY$Oy9wnX8WRj==K?D7qk
zTH@_3{Z5a7`we1B4gqcg<9oyDQPfS>QKQ1tkt?;%0YMZyADa}7`GPca&a$U6zSd;J
ziR)FXyWfJ~%)(9%7s4i<6U7VJ7|XPgE(32g?#Z0L*>Nxh5VW+q{of6{iKMpVINB2h
zY>mdgQDmuyp8b;SK<s2!35WAPky<5YiI)sPcp>IXH)pTd=5Vsyi!<DslBJbyOvS?9
zYb<wA4o$_XCha+nt;jaVD+aIG1CPX$baTp=fQ7U9U?7|DrqcP#zQgVU5J<4fz*IE*
zt`1SV@}mQwKsjCa%X#s@f*z14u;aR^emJ>_&Fz2KM4AnpvqdzpxMbH|tu#%0*+4Kc
zds#)(OO?;!bavD!+xdqlh61Opb+9-8=Kv?)1yQd0qgYcEyQtJne?K(p^G6~0qq8Zj
zfC9&iX=`rEUcX)DbXipdu0+a|5N;Bija+rgmT0Kr=L@2qQ1~C)Wk)U=mEPyhDMXl9
zOJ?p2mSLM_eyduFDRLYZI+ZL}hE180_~h6$Sb9thz`s~@-n(W$4sDW$Mwk3<*j<4W
z&UW8>0hYt59`9JRIYu9~uFPYMgjXss6^<FiLAvB(QPK7ryk6FS&}v)3Ng1u4?TT5p
zOG{4ihuP3ws#+aGg69`b4g9${>xcwV5fI1jduB#B5sail1>jFEjgvXS<q=KJ{re^X
zoCZ~s!0PJ~P){cIA6Ne&uvxU?etR^KwATVoV}fYkm8>s#)QhX1ng=TV6sNnIAe1-U
zG9i;=AfOK&ybcqeD&HLv^hX~ZD8JQ4g(rPTQcivv$WK?%9a+OVZi)f^f}>ca53U>^
zp@5*R(O`^C_(L;iVeAxdiO1$HfyFo;2v)9j+Qan<kq_3-lTEE;f|yF;GZRDJzyYro
zK?^xc<=0&>(LXYs-W!7V=EQTRYQicU&pSA*v}`yc=;-PrvOy4093qPK9;vpw#af{7
zdxuTCkEC!s9qxk@D6EJBM<No+HiHL$+YaC;7<UZpZ4y9%oVu3mIK~Hu`!b_~x4D93
zmNaAeLr)K=J#NoGCPu>zQ{OTmSgLMz2;_KzGL}@s+<vkrn)D=Ch8}+iR=`8)2GEfS
z7r(K+Q#UzG9x3uT@JGesW=O}{W`+PHh3J0yapBQN^jM0cU>&}7C*@P&yBxdcN^h#|
zK2A3jTif>jt%uJ)?Si8rRynTS)^yI@zMz>&KT;N6m=X@mS&YRUe+I};F!A7KmcRo9
z;j?ss-~7OwB&a9ZY1S9WOgZcA=jLG+8O=<*WmVCJNC$_>5AHM2QsxY0$FkJ}mQ=E2
zffu)2OZj7exu{bTED3y9W&2}OM~B%KOzFR+WWYs_4aN?d^GeyOly$r3!ojLN047E^
zU9$T3C+8q`7fB(!w`BFPR-7a+*~TzTA{=N9O<vTcpuvaZ0n&&O6k(eKtoUF?Q#?HG
zrHa9GYI*U^sp)S(45Gd=y(QRi+7<;Hg}w)@4hatsZ@J|G<*Dh>P$Wl3#y|4<F`2>4
zGZgX20S_(Z!Er6|eVrxnPtm5MvHw?l58=G+INu%<8Z~zd_@@wVDbBic<1MLsC2sBI
zRCL@EUA&`-r8kZmJMr3$%TJG_W9Ro*6iqgT8(Ne8(>C?pJI=e;2^%HtU--W(h>}*|
z9}7BSX{d_2Zqdav6BOfvXo>YEv0~&UUlo-n$3~fGZtjD9IM+AI+t6mgG35+dD8GLC
zHSex3307H?OYT>PH%D6NSIZ{WgW_$K|6l>QrO|xs^o@NAj$}p~E?M9nU`MXf{oHJd
z8f@um4Mt)!p3~Id-OLYS_mLC@a`p8Eie*wcFSe96f*C&fD?!5S5?I2nwngExim9qk
zjA@E_=3_N?$^Po>LZEs6mo_JDzhlhandc?7)(x-|q_Ph3TUU&d)Op{eWW3pvxN=9|
zdza+tK_?@|2E5rL9$0OgTuBjjV0#ptYU0iP2$M8H)Rr6`X_&JKK6WY)l=a@?^QMkk
z^vyj}sPNPmK#?@F?{8FuJ>%6IQ^Dwz*Ky`BO>-ZoM+aK(SafuNiCoohpu72Vz>pkx
zrn|OM)GJ!&$LCMY&`jJ<s)K^2O+MoOEoDL4dc_!D;;c8jBp!YtswdtPfXxQARNtEK
z%S^gtl^}=$I7ZwPDK`7x4qq<|hEvvnlk$>FTRngA9`YfmE|xSSVgR<jqHwliEj^s6
zRdoiiOk~OvJviWLH(|H=q&P%LvX>s!jBLT2s8q%qi)JogCn|b7P!6LfJaPaWYbQUH
z@P_}-84y76s`b*?#(2x9(?1mzd+QD3ui2<teZ6F>yQ12rub2bimfD?f932`HJaDyP
z@EuPcfN@%N?tAc21MyQ9b(ST8A5bb<Q3VnGF@_St+xF^nV*YunAY|872cn|q67eiE
z<%n4sDlq3098{#qFQ%Gfxr4K*G2S!Qk&!@ZW#z%qana=dF&~wAtStPzl7e5M<8YIr
zp>w<_@Soo_FBzr%!9u#4C&NY$d7&Ge)`9-^LlgI{lvU70Oj*H6oN60(%BEYEzBr;O
zz_di>-M19%1H?qz&x<il>@7Q7mA90VsR;IlzaP{*?(yETS=L1ew6OgNM}gX+t*pB&
z9UW{zKzNS*wj%$IBkz8I-9=KYE+%kV`GKi5Un}BC6`~%h^&Y<L>LR+LaO8=W+B}%<
z(xM5m)`s!RRYxz|8!iHq9i@+1<%r_OH`j<<*(R+by8uZAe;~WP9pGo$1K(aVX7&?~
zRnpZRpBxpcI;KEX@aw{@)6QP2aKL_2u8fR*`WK#JkCQ2AUmY6oELaa8J1X|yuQ<NN
z7fEI!l$W``B<PwUwy!%sK>K_YyYK!>$~xL^*X6HI_<a#78Vw(}%vJtf*MSS6y6#At
zGf-BS-*`ZVp?%>(Bpi?XLoDk{q}vviErr+R%JVN3MG>MRjg)}IQo7X*$VF4*<)^ay
zg?&NBn>qQxjg};c(&v*}SZds7aLK#nqG{VoXa6gwJlx@yIYB5{n3w{;*>Q31hu+%@
zJAEPmJ?A;)_Z-6fL+rl532Dm9TdzyCfIdWmtxVJ(qm!u_?{~{iiA>W3L?xpXaM~F4
z&6zFnK=?LUvj@864VP!Acqn+z8k-YKWmj(3hK7&SfFnX-U85-5ifmwQ{*I?V6f|Q%
zwd8fjG~i4!pcI*e52ChUq>ww~!w#L)>I3qDB%z$*nTH6wN?7m}v6|VJhkKfjTm_$N
znu;@#esxMF6bSF#Fy+>j0|ur%%i_v&lZ@9NoBi}w348d#;9I?SPx6iq2dFbX9e{ZI
z2U$?$i{JshYJBc0B@NvK!R+|^(QQ$R@w4OUFhhfu(k$zZCyraGp=YTUOQ!3{JkM39
z(ZhHPk5;8zEORT?Y<=OxY}6ah1GUe1c{Lb#>FH!5vd{V3WeZp=^P*MexbM~uKy0Ri
zCrZ8^He$oE9(ZS7HHD=HI)8U3I0(4y-GkjnQaqrp&Aodl9RZh=!R&iDAXNI|W&K9k
zc5P5pa9auaEJ09L9TlNG^xQ9O$+9eU<8gmYjE#kE=mDm4s)_B`)y)n9&o-KSju!+)
zmt8{?%XUTQImzw4o{mlR2NEHiW`x1QG)yeYOuFt{PBdMe(}<9|Ad9M|nw(k<rcg_D
z&EE0+#aFx`nzboQ!6{kw%oPicI`g`hx9T|3Rp$l6UU%=AM;(9bQUlbT@Emre_uaD>
zoO1fD%PrEhRNcctDCnhvqX6-^0pH0Bx}zv=Mce+`nFR>x!P3(aDjN;5bSx0h`I?fM
zyF~{Kwrr}Hx()Or?f3AH`OetBS{5HVWZ_nM{oMzR^@RKphM98pl9K4|I6oe3E<9T-
z3ng39VOMtaTT&5XB_4L21Wsg-snosY9QZB*;*?VK@n5@&`Mc&leIL8e+*cpn@)Of`
zUbbvFlxaDu*l**2IUMqYIgb!PIgZ_TcY2IJ>pAi3?d`m=qx->Vb2z%!w67!-zQOLM
zAAp=XU8)Xfa$F09v~2YCESivsM?Kj@C>scdB83oDfY@zpys_=Zm0BNX+nS;wO(8TY
z!N!3wB!wTi<|rVQQg2biAt6=hn%YysUi=EUkjOP$%+%zf2#n=wnyHfB*HV|{Xq*NC
zfA!r#;3^Q#I}T{94YuBqb!&6K{Psy~W03WQ;0nrRMvbB=I@LP1iONh&j85Fu1py9L
zD8W_XuDT(W122IyU8{UUcx2-5y{@tG#!M(xh|@l5+Jo(aymIPZ+0j#%IgV(^c3;0O
zLFkWDj$+|FOhttUr9&C$yJ=_!(N?VX*nH4)z1Vv4LyzkB!|qj5FmmGZiwZ-X_$<pX
z!M1-$727Cbx!f`qAg(xAE!SDSVBb+~;ly!Ie!XVNz4K$?vFXsH2ewpo>8B1KehGY5
z{Vwpv3|=g~b6|5eOX?V=(rxhuZz@jZ*@!=rj#BIe%Le*;45I|1iS4B-c0=L%@Q?ja
zMI8H~13}%v6IOjNA?|>f%f-1vbU`pHrU4YyhR&*nmAwr^2e%jW?nfH)!J$!fFc6%n
z=z6(Z=zD%<Q=&CasyDOV5Sz(QVGn-c;$WOg(!osX-1Tq?Nk2^ScWs$79a-qSXzvCl
zeAugfvZ-%hnDaITv&p^L3!WcvZVxBf#NV*&qwa{XE11y)e`v{Z#E4>*ExyxuC|65#
zCN23fCVy8a4kY;#Drr?q(Jw82a6)(}W3MghNz>6cD}pev1FJPc_~gRDD}oEAKR<zr
zHjdnfof`ATM|+r1^VH`=QfvoBphlPP0~}MLfrHJShN1}@F3YDQO;ny{7d&u5EE{iZ
zp%cDr-tSM)@nj%&_J`X`8U*PMI;R>s3bie(u)8Y6P7s}gF93>Ig`??)=Gq{Up!tew
zkbzDdxU8VTgKTuc@RRZRuqw8Eeh=X2!kb|7_3l0c$JFx!-`sIx+7phEx;EH03IS7c
z>@Az@4QI!*^r%E`(qoFwSvW2me>u4kD7u8@9Ho20`$Xa`(_b3sHiS)U@0=BCjL^*u
zlP-I2<+W5pXv%v$xTJ!m^3{wKyulwfp}3OStGzyDbb7FX)E{P<){B<G6=Cku5c|7h
zcagMuc);6~9QmnfJLez+$CUORvJM<btt5!pW(dd6Dq+3x?k5jVkMyQ|BS$rbLaDt+
z&wJA{Wvnw^>#zS8VAKOQjG=0RtM_f~|NN*zBH!rAj(J+$^vJPo6La1K*4|^{k)|*m
zrYD^G_y&RJ4HJhksRpJu6BeLn3UF*B+Ym2w@L63)FKG_Azd&5qM%!*-VXB?|pL~yr
zMjluU>;7;c9;heyzC~S8K}p4O_m+<@o{;nha6o+;2Z=B-I+q;3@Vf!q&>y7(RBGlG
zpY~dc4!_LBvQX6<fm1fs64%jRUO2KT?%nUWce}yAEp2<}x&z|0XH=M-`oyB?ftI#w
zm-8mJ*b7CQq@ci>;E|$JDw{pV@`-Jc#=#a+Z<LK5yF|drb9bMP?lzK=p-gyl3{l#4
zUB2e10}_;niJT<}0xy{yW<G;U3T(5%^OxSacm@-VMO&$4W^ez(FS-Q9kaWy)h7JAT
z+{3mpaH^Jkf~Q1EVS|dIes}l&iGAZ^PV7{K^0!w`2r?N`8akE1bW@rRQpf${!*vk5
zSBwD#JVgv*(89YwMP1W%6HAp0lGL$50yPj`bSB_RP{Cj2MZqux9)j)KLbe5-El9Hg
z0u1K<`#vcSK?cD#G(nQM(z-XUol2z>kz59+8fWrZAIPXbx^TemmX*qPM@Oh5Ug{9R
zT)Tu__{)cWItESypS-eJmN#GM_cIiI#HXSt=x%x7qcaLjgrb`#op%s8x!GrmB1WW~
z1>|!bWM@rDZAs`_5Km#@9s*A$nIC_T1bzVo{nanoE!cfZ3OvPI*%FE$oiIpJCNaJ>
z6wDUx>JT7?Hp=AKU|>YzD0SV(kKTLV+4)FAbMCTlHM!R&)7aNG_yb^%(_kfNJe7Fw
zn7nH0;tPi-^9Q~SvL*_p$H%h%I7oP;DHvx-8$&b}pLW?(U+z#1u<jfYP;!Q_9t7&C
z;{pY1VyZ*rL}(0D#XH3TC3PWMM#hp6_OKYZsXVL!)CMN#L`!j_2za7`D3XLHtM|S8
zj?KquIuQ(_Hwa7tp5j4?-PkSb%79sZbjD??Esf`Ro-@FR40;ZDIlu+NnOomHF8!dU
zmlq%F8y%s;=dP1XiImEH^U`F*8}Lwx80+y}w9LRE_PT>YYT>0sd{R^4Me(LUYOs7B
zf-c;?12dh`C#@dI-GSZblj4tt$h{0~(%(FUV#qLsSS-%Q=u~Fh6TAoq*s111K@3^~
z(S+rV;=mWDQlSy4JJt)lY`Gdl(d6I_W<l))lBmkfT>$*iq>8!mff-jLbxOSaI2(&*
z^3iNQic^f`u`myaP&_yP`WMz?6PAi?Pe43k?_q<QYN!r2H<6IPrm66V1{Cc%I?3RX
zk}t!L!001mK|nF$!YCdhCtxMGve;#+;Gkz%zPNeGGglm(g|Ch=xoTE3FqER1R?1%p
zkAKzD`Hmi|cX1!5sJf)dB@k02Dftcs0yFG!>b(1a<X_XzYPzuW?BFQcKkA%KDrfhW
z!P9Mj_S`X^P;+QZnx=BM?7+(SB<BvpCemq-6PZZ-6Js;M5IGO%Xteo*Cwv3H3n&U-
z^Jfg4Y5*JKR>Zqg{sPes@(=g3o^^N}m?-1?Ij)7HpInO<%IfyZsetEzf9z09&|ra+
ztD{B!m5x6K*M-1`=Tyra_;|BDARFDKC3918ZzAl=#XUIsGMlSvI!!UTnaAF}WQU9j
zsX*KVK^Kq=gcG<4bVI(@*a%OsBa__73#9ZAhMt2}*MP1HxC~lJN3{hgs;dgN*8bfK
z*U!%u($&T|;)w!RQ%nL2==dYC(CKR=oF$AmDiR1WkW$uF1NQ-gC$|h<MA=a6uFHY}
zreWH<g3EnyR5Y7xWd62Qv@i(|BD=K@oH5JqE2LPD_q3(9!dt{0sb1qiRgO2+&P7ur
zZT{-0rdCYUqo%!Ls<q)FWWn8o-Az(*oGO(K7o|Mx&QkhdbdUVb<~U4<0;ZOUUYU|n
zcATx}jcA<wB)Fp}vO133tJNd_1;Fw;n&grsQ_{>Y_TAZ5@Fn2u;^}xakz+!GErc;T
znV~X=od3v|+d(=j2KFyu{HubS3c?7l-~|C%b-0ITgR+XRYM+uP_!IGt=qG9^c&HjU
zt+1?WT73WZkFPywQjVb*<mMn~Dpol0rc<)fST03nCf(Ih28$zM{)b7vsTMdnE=2hy
zOXGC-M1JS^K*;A^u*}uEUO5ax^OmvKY*1`c+)3t3?Ju4{HH4X_#MPTH{oVpwk2kW5
zBFBmDr-xFl;n6vg=L{DL(>(e9&g#DJUhHO)R@?MlTUmP|4vMIqAUH*_@t3X>i)FX3
zw*tXE;i7>hBgFnUkC^!sC@3-)2?2)?72y{F*T6$=06lpbxMICJJ=_$Z5M?8?9=UaD
zBoM5YqeQUbAhuE#W}p7Zmuvehh|Uen#@7s?uiL2DV06I)tuYif30^Bs3QG7|K|_w8
z!-wzyZUWqE)yN(eE(xwu)Er5#=%&)UW${f%%+Aq1Kky=0n{y7~PaONmhW;leMq|Dx
zOC`?RVS<m2W%F$8{lS>tL@mYg`Is?PHe3kqwHI4GF%NamMqTUv=RqvkgUAE@Hmo8U
zI4G~Y<lZ?6kN@IjCBX(}<Dq~Xc&(TkE!kTy&(N*@$XIYny>K=>;r!KG!6E(Alaw3)
zWUBM&JZiAv0~}Mz7l~fiUF@?4%AlBLNv>ck6Fk+G^82rKlzA-p1Ooa~txrD%{5oKY
z?SPt=_%ioEHrvLI2&Tqf)6uv5wi&HBS#5O}m~<wcrK!Z2OnmIoH$D7*zvY^y0n8Rn
z%hfS*LPaDd6AZjw(t)xNS%Kxp!lxKNk#HYK2?kzM0ggiaNE`;h3%1Rv<*lDRb;0CZ
zlBS|G8;r!6Os#zq5Iz3E<zjbf{b{+_h+HeZ;3ZkZz9`svoO%?m3lf~wo~gVJ|E5yz
zc{1A)Ynt*{SH&H8{2*^b&Rh6Lw<L%hk8|w_M)}KyC!X<{D3}#TtTkK#Nx_w{WWM{3
zTsGl}w5A}q$Ngwdo%#yq+w9iXCxYGXlOiBA68P7~?lVYICecbIQez)e%%WtYj6!gM
zO+1E(l}IInlN}w*J@5n05gOrNdaZ-K$JQtl=Pl{!&`fdNHCcM>xO8y-H<E63efhuv
zo(Rd|1F-RMARZ65G)7z7lEM8iy7#><MG_Qcpijj4KQYNf2Zbd20RAe8<ncRDH+5J-
zHQ>YQMFBJhqG~F5g1l)6uBBOG@6xBQIy6NU;)%4ssU@6iYfHy=Drt1wJEeXJqN9h#
zjHjqWp4bY}cS*o{I}npsQ%4n31E1W0m|hcXh?{Ij88GziXE%ra|1t9eTeH{PK0eUe
zwD<mxb%{85o9A#gXJKGdyH2|3t4d1t4?C2Ujz=aoQh{jO;7-u=+&5IW%+-~25Nek_
za3h$uDxyA?VdMV7t9*xSm%*^xvg6vE43QJW^cAaYl!z6=qet}gmue|e6zGMPR~I34
zF}Tvk%NzG<Y`dA~G^Nw*zc8E~YO-Y+I-R4bbgO4?dMp#7qtj2idl@IImTp=QB~{zL
zfT2}>0&R(~;pK#(zJ{&**lh)02XEkoq>HFt4)AK}Ggr)=*2Yq)1QUm7Iy!!g-|GpJ
zmiy}0Gwfrw?Ld&UboPJEexp+ySle0VFuU264HDH8ZrSA$Z_1X^+uh5RMMr5@6no8a
z6x}@Xu+M}t$K6fuml`)B^QM5)f@48P$96y$6+`JzaU->H`nZRV^752aoX(F;dEhKy
ze{Ays4|jr}%afuyyBEWQq?W91@7s7jmJ~yFq%w`z5)|D3V;h16vm`kNj}>*{wA;3>
zSVDiF+*ji!s^~9P(%PcHRD;tj@DVjlZeM#tY{AM(QFeIYi}9m_@i=^SGzx(}1AZyX
z#=#@ZC1R;`I2wzl=Nx_c+wGm*1H4$BHco?32)C+Z7$yjYLApf21FbruD4~YXTNW)7
z`Duw;zkU0`bEb?P6ZB)TdluWfz_ccU%(w&%R|O4UM@X7Fad}DBc-~s^%qHD*VNJzR
zaL7JhW7QuGmv%`9{@ZJBnRWUnaMeOQSaQVIW{nEcsWVnOYFFucPm&rnYUax-YzYiL
z2)QWoHpE`w!y+>!4DsEvsfc~=%%?q3P=-ir+(Y@J-bnsBtiS;x%kOS$KTJsK>Sfy$
zzBz+RfLEAojj^Q9g9#k+ozWriWs3uvf@>;QY@9z7EWd_+(y!p=_kk{g?P`*ySlyj^
zrTqEFoU0C~mT*@j9nFkOr`te5b6A8AyfZo$kNDGB67EtSf5<;(=ClR3zqM@3*46DL
zEK3fg2gYIqg~TDk3IUaApwX{XbVCPkw7<V^>$-10diKhBZDU(w-jM;C&f<)<5G-Mf
z4+G@%$J5l<11`C2JdRstvZ-e^Ls3ENWWmM+6&w^_l~4dSJi)(>^E>o@a0Bz-=-F4j
z*9EaUI2fge{0;F~>Yg5_?CiKOFe*wn&-;uwC0#ab30O^hS#U(rlEWjgmj{n@)!Y#5
z1RJB7gg?UMgMs{o-&7^E+LN(cF+50W^AZHi%5!sQqNL;&V<Um=@oV|A0U}D~d4j`*
zEk`QJVXm$DD?ua(yYmgzkd2aJ4S<IS8wMfddH2&+Iz*=${n^WwJa<$Zlg<H&LV;K|
z7ibD4gF#>uoz8l^3>^rKpEzO0+>7pe^Ydlv)@@k7t+&6ZY6BSW<18h5Nt0EzUr{?Y
zZP>nT?YCdOeeV@#AA8`~Y$m~Ye4wCk4YI)yi$a<nmx8N-WxcT}lTLa0J8|{l@ti*#
z4>X;)%_;GQVq#B%E%TTjFmzSAIQZC6AXMVT?@k@r91Hw)<oNrRtWz{o<$KSgno|BL
zFAO-cz2;Ch;Y-C&UM#r1WpH0XJ7r6TP#^TOh&fP_F5w{ijfVw67D#3ulTUgmhCXhw
zfGMWK0>i{7g~M-_D?Lw4{ob&J7#%sIQ^azkaC2RgWZMIwI52F#MMkTA`1r45ce$#;
zR&kDdu+nA6(aVp9DT<A>#Ah$6NZ-A6>8v);x=a?+Jg9UIZg2|YVL!Y@d%aYYqT^{c
zHetqr$DDNb#S0&M?wO}vdi(9SKV0(hTW`Ph(krh#^T^{*-g^0&#~gmp^l@3LIUM);
zqN!B1^#HFyc$UjB@lZ4vWOCV9xb23;J2+A8zC9nwrXuO&j8DuGL}tK8oYg_$1cA)M
zNdDoC3U*d0ye+&i%^xi!BaxPt*$dz81gE3C`SiVfff4gQEeoQ!Y(YaJLs6%#a(nu$
zvI*Og3{sE>B6{1EEvIZalC$Y=O$=rW;4x>t5!kgk&-P&(>hQoYAt@YpAZexRRyaO*
z77<J;+u)mU{SNH^<cP5cDq+(-1UwL(|B8mEL6E2k&odZipt*wuc67;ACG73A%f&~>
zg00I(=(E;YhHCe|b$ozSM}S>oL&0POULm*OIXRnPq^U?G8f5~(IGad~9n&_pZR(V%
z)8T*W<SAoP$uVQ2En!@Wih_@eE-0ucHooFk;EoN1=u9$7#RBlQH~X5|Q<fFy-SI8(
z(d5ft>vt7QU{%0rAZ=NW3)CF6^w>hG3GLu;Y`$UoC~pv`9`#UDZ~L^}R4T8{V(8}B
zr5hxp&scnNIOGpxuUu9{15`3Au4I6jcgqIKs4H>2Aa>nA`EmR**)!Hd#i^<Hwqti_
z^|0-B3=@*#h&wzH8_p|4yj2RonF808iC@yGxL9tdjHys?oQaS0D_Zb(K~;rFY82d8
zhl!9O^qPisdCc|CUpIwLH6?E2dClxmw;wzT%bA0~K#&tnXNIz1z-7$WE+sY?!1Cxc
z4Z=z@KCtC9+%bNC%nNP-1>CG&bTwLoS_OdOte@~4mJO%IwE{W(fwqyy4_I7xZ)MAc
z<CBNo|ML1WZ>nJLafeYj$^Ii*N&7tn`HsvlyXX9u`rLI_M_OX72`ot7l0E$T4>~Jd
zcTZ>tF{!(IdL>o-^azUbg>q-E)^I#HZlziz3Vu2kAvMZE$91_#%ooM(ST^p7w2oo(
zH>?scQ;K}+?#J*^QWreavatB%{7`Z!p=qRWpP-U5CRW)sOEn4oO+pGlexehKUx+-%
zY-sRvRS3l(eo)J6KD=elm_y%E6rC%YZ!qBk$V=j&V-G!`HPkTFV+sI$9eRc1vw(X6
zmdO$xBP-z_ABYU&TV%m1h*c%;()lSfPyE}nH;lvK!)^ZY-?&B9GPb{P(>=?1oMl7@
zT_x8Q0q!!kL7{7oA(Xn|`P1p>)teOg{ewm)v%b`%DPcD0ZJdAG;^O8TV*$3M^|}5s
zFLb|uaESJd9QUpxn>gJlZU{Fk^E{~(AqY1|<H;BeTAuP`GtJSl!PJH8Z0x|Q9kkt!
z;o_5`mV%ZE#dnUS@GP>J<-nveo^(9(SbxQk9m^zxo=8h%HL2v6qk59Tw#zP=%nnPC
zRHYOwA8Z?2qzsPJO~u;s)!VOZGK>;e?!Vld^rGBd_|>wvZais!BAzIWNHVbEgn6K;
z0UU=E$4}xr=SZ*85Z(f%g&#A-XrL|jE8%M<eeR2&tm-U(dt?Bj6*^AcMEc;>_KqTt
z`Lw3SaR!PgvGVwjgZS{Z0WN^cy**{0V07Hm15(c`vl<#wiTO93G9%=nypcIKEnRk5
zo@N^6e8yWc*ZtDWh7pOIR&y2^a)SM%5I9v_(ZIeYVaH<=lHqE7M^I9xC6@}Hy-X|v
ze}B~>{(lm~gQUdT!<|)j`ktIZ+Mz(h@RY(~qYi$lq+0zQn5+j%h`pe{j-ud<s@2&W
zbQ#BgSh8F0wUMS369?L({{a6$WEGdw`eaux>DEi*+2rV8+?RNS)AY{OpI!pq2Y8ji
zXeI>VUAQTTEmUCdshmFm!Bi~hFMxg_6OqWlFbh!;&ap((nX#B;j~`H>WX2!H4Y9G%
zoOi%QEDmhGA_doFB-?ZrXX^3*a6{)xI?f^@Rnw~Q*OGs^I;|+!YSmEX&+`8>M>kx&
zvgmf*pJ`_4@%Ml8($T?`-|r6``RHv2#9|)r1>c$qC-yu(DRa|SY#PN{Pq-xqHdQG@
zm|qsU@7})<Okr4$V7=j1_TX>9v#J)07#<ilN?N@G9=6{zm4<WXt>u>jsPXp<IAu+6
zT?@>8ZCP^vO9A$cG~l?J<bagoP-R|5a)Pj;$;Xf2m0p4{8tye+wq5s;us;&d(#%nx
zs{$}azlTEqGo5Xl8jKdgi8KxVC|*aQFv~`os?4j(XAt<C9?$slX(k+vHbtW?5eVQx
zedAO^Xx4FO%<u&A8O9TT31T2l;2tc5Xl7)5;!V7vYefShdI{^5;Q{&n&xFvaWGOf>
z0Eb3vj}%hreIj>txJuVUfu>;apcneqKRiD}MMnH*^5jDx8uR(D>6COwS@*&_9gYfD
zgiHqSxj<2{{SeoedH$7yX>Vkw(!)+pe^S+KnH1g`CKxV0X$_WWOWUrbsaPb~T03lD
zaH)e|;Z<!1DlIv0aP)qiPfC;&d{OXeExugVz~pO+#VOcgiW}55i|18cHA<Sn^ROt#
zsqT{7&(2b@VB(4mrVEC5?P()3xmY|p?ZNL>Jb&Kwd<Oi@SRj^($788fGL(z?aDW!5
zYcQD0#1f$pJPR}Ne4fgsSnvtpqKh7M*E^qnvt!%!$tbu10Es|$zch6_*Wax<_PZ0B
z$4x!y#g+Z0gsG^qrd0<-!H(A=CW31Fy;CXm*&1i4y!y@Y`?j<(b3ZDJ(vIU=s95xX
zS4!gQr{{VKWA_~uV?C*fBLXi-W$+w|INwkaR1_OX!PtY6YT%O!qWI4P6P|GGIO7#Z
zN2WX`sl@fgi)46Un32@vIN4r$cBm<}lcT^=_P~FZfkBwDXO<K!!Sm~ad*aE$X*Y_h
zA*q~J?(0KwRj}EWd<?90G~hkFs=BNkuc^Ga{^|Mb%%>&r0|qL~4rvN7sqxHVpK+F>
z3X4yJfGJKFqW)|!Mzw@mW8OrlF;Vb(Q>}?`VqDf6jMMpqHx!`)qXUH$gjCVoP2Dcn
ztIGF+55V|Cfzvv$0*Sow`m+{o=us=ewtk4_ftc{yRD;a-M3`EA#9n(x5k=j!!E759
z$rK**`z8mXH}t!*v#f1Y+~YlGqh@d2`pC4#ef?1;mSCB^bDv6>b1L%KB8nCT>^vsn
zyS<8Cww0Hshx2h7>$xD9*3@3U+q<v^@6h19;eg>qQgDNcCB<HPDk)=D0~__lLNni&
zD}<$(=ZxcfYHR#b7!YlMp31JteQ^1MJLC?5)2t#mpRge`@#wM=;=?y!mNU@WEbb6(
zZsn34h5?jy@A5LimX@&poB`X_P3OHC9yXhe#t&Vz`GbdVxaKsH;FcPnD@>m`scj4!
ziPP|Ddj6eHz4Fb*9XI+y;9Yw%Pm7{vS?)^{BY_Y@H%|NpEN@Y>O1#>m8is4xm}8H`
zB=Pqyz;H=*Dxij?MSovI2A#mgB%Tro{#WC?4|q#e7UlL0QHh6&y+Gc@cOK9F#Z#Ct
z+S}SP=M7E8344K<7%(B-70!?h8|#$6IxUs(Qr-#lVEPr6yQ;U!(YliU6kOXZabm?@
zdRE9Y<R%eEDZKxS-4KAw)wo{6G`(L3T{f`S-;g()`k&s5SMXd3L{(RDuqjF>o)Tm#
z{F8_Zo=KMFm}OH}$W(XI->GP_Eg1@5-QF(>z1*h<GMQMUvEl6YiY^-Zms6T?9A9+q
z^P9J<>*t@F7>>n5qvDrtwThNi`Cwdv4#t_lWu2ncz0Le+Vk3t5X%OYIu4<xG2F~Cv
zsKwfZd1&T$r78H&t_;C{`xkc<TCUnI3qqGVAUp5QZ}zdCqn8MrBVO(E2O1KO^^3Zz
zwtsi$tfVK*(D4@ELIFG&oGlzLJhDwk;yHmAgzm=<WMWt=mrULOU_IfRJEUJV8`|%J
z;YCtg;3e!`a#o&L8*GHYfP@Lo{6JJ36Kg*qQ~&*C4(EeVuO<mNE52g#Qs3jL=9d3)
z*d}lv6jOzF2{&O8GSH5!g^CJt$_qNqDWmGxT4~5jFPJ$llkz1Vl59n?^+$8TW`8!*
zaHFZ2yyLvSe;}Lp#iHl)6;o8S7jkfsF;wi_?XFr;DxXaAK%AeYPuOm0vSP0|I+34z
z+&Oo@+o#B&inass5Xd){Kp|gFcH?{gC}xmCyl=ia<uCg-AN;Q2NU9_1dhb6%jckN{
zaDcPS?@sah@}9}>i`{*e!t)>BIF)J(1PW(<4mw)afq5<}BsM<5<;t1es{Zn{2os}Q
zQ;erp4T8>Ixfz1~s*(;P3^Sh;9YRC06z7|>u^kfW8IO9>8E-t?a^y#X>S6``BGT#S
z4Ssb21I#&?d1B8okwh=4)`tf+rYAH^>2OM2+wg7Dg$2xn4cFEsw!cPJjSLpWnmd}X
zdGR}s{q2;~-^alrcHiG1zNcdGeXdqHSJw3xXL>*?;|-@kIISw~m$QPf_jLTU<uc|&
zFQ4n9W4?HJ;zu}TgR%X^d+&Xz8%v7{9Iy`uHM@$gmJAo44+4og2;3J*bNSDv{r{4l
zk(Y0i422U)Cb#q?%GW&Npl|zSNBivPmWbDL-cn5}Szyk!r8gZNop$dAO9g)vM~0G4
zCQQsD{q-gO!`Vrkqlsm`nFtxh%(Pyx)KSdJ@M*=bVVIGW0DD30#YfXD&@&Zkg#ZPc
zb}0JrkA*VMWnXqo#T0+mC}CX)&_4$pQQ5|96kP&~X_Ta{n`2ZgJoP18U-QJ@KIq~=
zCqYl)55guihRBSa>>%5~H$i6|a55}q{c0KGEvxt5R+4zS&jq%ls;2Pdv;@V_+2|Qt
zEe-sI&!&W8v4D>`bCrT+BECJSG0x^EjXip?rn?}emSH*-9TS{n@HENy)f9YJX6_1z
zrX*7}`_B*2t-f*CmBg8SlGgY9lpxa*y}i54i@ncGjd%i?8&<<dYR7;q_Aj~Pi2+m4
z9pEP%E!Z;lhZarnY!#vF@u`touNgxcd}B=Rs&7pUB!6NTb{~c#Ng;sZH07zo@-!0(
z<4jKkY%qBI%RC1fVwK8Z^sB9#KXE|ueJrm7^uno2!1FY~2bC@1&56zNVDk+cU;Jdg
zf9{Q6l#zVER2`%Ui|SNu{~&O}oaz8&S1TE237clX!_z8G`Llm4JmaLJ_aFDDi9_x5
zhcgu8^RvFQdMnuD{!SsBi{=W^`CqzV%r$-C^ph^W@$b*P)eCgzILi=BtSXH^0HT=c
zcl|tYhzeY0&S#UGTBCcPv`*3;Ntbn}^DK&CnzA1iWmVYp$fOY4-1?xXm#i`nu}cvK
z1f1HTcBD{8D>z3J#|!+Evq`%3PSE)i-_eInK>s}qPm*E+Y1dP8DV9l)6b}N1icWpB
z107S2S8#-eIXLM4M*$}VDFqG~a9+lecp$HvFA2uz$VqF2u5T|N;~5#fe!zk)VCxf(
z<*c5|A&`@bP5<~p1)sJEp_{J43M$y?(vEd2KY8QbPEy9S_|_?N4wyK5_JrF+i{n)1
z>jj4#chEUEU;FS5yJYAkXZ^b?I(M{dScF*x$5SE2d2o!Bs$xpgPyL1=zyjD>WtTc$
zO#Bz!oW8$P;!2im=zXsr&iWc!&Rz?Fm9Xl<6y@<A`mxBHz+dn?1Y71f^de*n%s6(4
zx{`*8PaTg>j<kkq@e%<;AG4^-*^Tube+I*nq?Uvw^SdA2KS#&G7}X9V8xCZ~-`!_(
zW!upetP%wW^rtdSP(1PMllj8I(<OuVbnVgBM8T7OLekBz#y2$Y$J|?XRoO&-YGph>
zxXRFJ0#(HvKJ1yQ;Fu;bg_?rnnDwG!I~p(Ys;U?z9KOtTuV4P<M{m9T>UWk3n#}WS
zmVLWy!-gFL9L7U(5gbla(+m+_7D?bv+z9Yjfv0c*V0!=0Lc!_wbj+@oE)M3S`^^5Z
z+*x)d@Ev*|9vh@%;ag-K+|*5HGza}7j$Y}4mxJR&MF^2N@Qp|-F==9x3<%cv&b#*q
zuZJ$|9G%M0bDr*W$t2Cg55rSQ9T$hy@3?1TpfwCm>EMB6(uuKmZYU3QRFxDdMVR`p
zuy_glQ4^NKAyFKhGsh^&?#jcPB3|mOE?eDrRcn}voU&R}O+^<Sy`<Z`iZyg_ex@3L
zu0Tn6lRQV^tAGRP;Gks#8S1LKp;c6zzZq*lRTM=l>PE>xv8?EvY4T+<uv7yl5u9F#
zpd42%5kdkjF__ngiV|n^=K--9_@uD;3jgW60PAhOq*pg&j0(loCwl^IzFA8ZOXZYz
zXNDN>h)cJDg5pd#5aokMN?P#Yh9#jTxJ=6HZyQg=e9_u8;Rq&r@KaqPiaLH-^*e_T
zh9^mFPDX;Y-!+9MiEvdgnfSi(DOYWAdOBURsNkS`l*=E3i_sDUOR8dZ^jbU*KBv{)
z*M098Dl~f9^F_^Bc5uK)7w#<yBn3(kN;qf=>zjSA6&+tyr4<{ffS<r)0jIFlWKd@?
z*f{VWyjDjOu!b|~18^MBJRI=^XWYg@kp@;hk#U|oO{-w`9WDji#-t`(h~zT=JkaX^
zuDSAnRqVXU&t{u*|I~EsjOT>DXD2sAV<WHcah$Ty`*hx0hzD-&E9*GQA9j-P6TSw%
zwp_HZT!GtjLoS<SKu5iUH69se`n?@Eii&*Ou=7iYCrM4wl5A6U+V9AdQBPp&NyZV_
zC@O#1cWQ?wciV(e)s`s2qZ-J0H-J6xVDj>sh4*ccHM!hx6}3;NMcHWbqK%HT^|Bb-
zLLc|FY3n7q0v3AXKcCr-LV~hcr4HT<oN2WB6bnCKyFONfL<Z_u4FZ%Gt2yeNhEwE~
z)QZ6wCb+F2k~%y!1i>IKtif}z5z@|Na2z&=BAelTP*hC2`~ENZS16cug2V9oM2Gw2
z&_FCtAGs35)s;#`ar@<LBsgNi=b+N2zU9(HOYlFB*kN#{3y!FP<iv;_nJSeGThjFn
z*A+r|<+CK$fN+#L<e#06TQ-rNuzdAPhYyA+Ng+xBzm(TYH;>D~)y||^s|ow?l#U(w
zm0glefmbXnyobkhSH~2dgCc^l$8k|s@zLgePJe!r$VrCObLC!sHgM=?J$mK!$^L+!
zy{~L>I^SE-#T_^1C!P4v3Q;V<H`k(kAfbbEGXN)nnq&w)p=*^;NHe@cQ(e8od-!*3
zWCi&Im8`lf@oUL^IS>G8-~jO0wNz75-0PoLQ&<j!7r&@%YJHDpyzyZ3T_uR$AOtEK
zUrygUmHnT$2)1On+KL6E|1$aAu0Bvt_$~<prfS55?NGA7cW=3v%{E}QOFW!RA&F!C
zr##w2`s%TvVHn2u!;_@0L*{XUGrRrniNOL6G9pcR1kw0eUv{FPNhUg*B)=3{QUf{f
zH6PTl*BsXP-~VI!M<!<9zaD9cwa|BoQpvgQe`K1eQ`eLgM=Qz|Y3l>j-v6;LJ^y91
z+<|3B2<KocWW+Ks4QFJilBtSapZ^eK*nRE!_fcC5cMv<%f%oW>lK05N&utZ#0HRu0
zWM|{~LGQGt>@tYoCDrQdD_zS*#_o06cAz1+;jbKX$~O*Jb{WS(V_#@(!>*0Z`TZ-7
zYHDgu59W`8oPv<f`TKTo25Qsf{OnHb9t=;CVwyf4TsXYj?mL)?Mg6s7O29-8{%8Qs
z0>~)V)x$&ky%-<Y#95aO$r0|1C)42*SDK*B8%`RLrT_ET6+OE4#s2@9-8=h4kyo(t
zgx&qZtbIL+v4OvwxY@4NK?B<ka*TyYRSkkE`TGSDhFVjrr+?>pg{Rfq5RmM(r$fMM
z`dB&w9f85h`moe4y-hy^*@jFey^}5RxhdhexB1+Sim7ru$1Cs7`U@MF{2rDj;VotH
zH@a3b)glgD-bpzZJ`yc$$??tX*f{GSDwIo8xf7n@aa^9W+iO1k6ow~B;n-rHzU**v
zW!nRDsKns7w<_q+%vU+AMQ-!F4E#i1)K&yw+bKv&2_k!S%?aTkeZ-R-Hbp(=3;LN^
zFUg#;@eDd29C6HA(<sU=KVW@wSR=#46D{mN6&`K9rkN$|yVoi@@DmfJ(cY7d6V6qI
z8;Ju!Q`L-<58)|sX=^VCutBu;bPyQg<5qA(K-lV29K2ln{6_%Z063r!&0{vGx%%P>
zqnU8xMR0_b5`^!c&kC~56BZ2s$!r%63&ep2*4cv<K|HHlkaSz>EWL3+V={ta{g9G+
z!*ibQt|%O5!!=M{bl72-k`zT03=}r%yPZ#gmezXkYM`Uz?;Egr*(lqF1o6@LlA_Sq
zyN)iphHAY(*V~dlcS%`QJC`5*|D}^HXRg&<^Sz0Cr97#}%dRmXSIWJcE{|db-N=aS
zB5*)S?^8uV9fhru%sq&mNV$91G--&c_Cn!3cEtnLYI0arO;Q_b5X3FTw|CyK%t$R-
z0zVVziEUBfRH!&m`6mFMfT0I-UV?=jXW22KJoVR;HxBfhw#-#NnU>5n=AOdJl0Zs@
z*BzM`Oj1<efoPtrE|tnXn_ie5h+?kxPB8hyz}CyAEn?~CVJfA=l%zQ5iF6d<dE55D
z!Bln3Qx$AD+}bkc*|nq=2k}3_G*{QF0^Y-Un>9_ki}8dfEt0vCuHDr<Dw>+|d>>~G
zTom;C{6}ta`4afFm98tJk$9v)rNWt~uzVx16rv{85=E{=rWaR1hH=&r72ZW}lJr6W
z@kp4D>O=Gt0E~tG{T|gc1C&&U1+Y23dc&>*N8x#Rsw2BW!eA9l`uy4G3AbF4G}%x&
zTe>?INbDW|TbHAVf>6AR@yDAdzU62TE#geen89y@i6N~|=$K+7RIk#1X(|^A$98GL
zk5T!Pp4(x=5;*v&+G{ZkF?^L2?6&AsN9S$5AN%o%mZG1J<1*>!*o!{Wk);sgb0m_-
zlL)I3j)1eOSS7P~Nq`BD`KKn9`H!aV*X9o`>=$_Lv+>PwFLi%O!>&?8emLFCGP!Y~
zmi;Dwg(+5=jPr<Av>o5P{K&(vyz^xzr|Y~3oWjUmBP~x5_mdWU?Dq$~g-4>H;7FsR
zNp^uez(1M73CSsISaA%=iq*q7T^RswDmWXnjsgbGFDX0<Pk$8E4}f_kLo$t(=Ldk7
z`M<C2x75y84r9FG@Cjd;SeO^N=}MN$<6IG>0m>!0`HuYMEvbE@AfsfQH5nHOFV23o
zvn-I{e|UD=6AV|95+Fs%fcqWKA0G-8=y)jD%ETf*JirVUJK`e<d?YYyWvf&ks;7%F
zl%awLU=*#Mb3<&9xoWwpzdzsWi-(W=T9(}{SF?1y;h=4HNd~IfPb8>VG<Wt{394cC
zw~kp7IB-WrX~&BfA3cRF%sS-An?LQ+uqv)D0I<_Y0j5<ONi{5=i;9S~N`YvkPf4rb
zOw90rDhd<?&k^~8LzhrffzWb^(<*$AZCJW4E4IQJ5Jr)x=%=-``U&Ww4m4NFA5Ls+
zi}(v?ytV9;+opPgtqJxa1uH*SuNI&tX5y2EQpvz{X;*oD7L&(PDR`y(B5fGo)A=iw
z>Nv+eLG>~iju_S?tsXd#Q@-@oIk`}%E$oeleS<y~6Pf;Grw%OTB&&q&P~@*BW1x0c
zQPZAC`wNk=mp=9Msa{VZ;JLSK^5W|=d~}R@7-&({Ozq?89wy;A<DKgxp1^|T*ii)r
zU$&Q@JwEOadD&2NL-Le=td+2uID{!^^A#B!36j3=Km=5s+EG?Pe{s+}>MPMs)Lq9`
zR2!E8c7l?EoJuON7XoiZ*JPo$SH}6p1Xuu=M_BkXN?Q8_HVo!N7`UDG0olT{4{i&F
z*p_(W&}Hx))kUgsB@^6X$pD5*?fSNdr-k$3P*p%tPJI!E%3ry>Dx$+2!<?iDQij}I
zaz=zsFig~23!#`)fnsLe*Uk$nCXbLdeZbHW*9mi1AGxnT9GN(2VxCIJ_d9;8Zb|De
zNYQ~2$8XocZ{?KD=lhbmhG`$IxiUJk<-&GDN6uNr#plxgpqGwBqY*05GX3t=CWu{0
z1qIb?oRI-L@PU;e#YId&(TloX;#KSrMNP&2{ffp5qJR}p6&)uf1r{ofIsg%p-YHt0
zo0dF%!v$C0_2{R(t#S~Q$(VQY?*UX*Aj07c<EwKU8t7aqN_+ft>*&88`kJf`<@~`?
zaC2~;CS<A5HL&5jwrC;KQWa1nsW;5Fq^|n5_NBv|r^DOaR}UYK84<-Vj!6_kp;kIw
z3!$)AFFx_)*TCWfA#0WM2ir?A%ei8dH0AlJ`vh`Kq|uxA{<niafD@^GG?i}ICvhjX
z0GNE=ZK0-yX8+^bXGcc@O_y{Sn6sfvD^Ct3(-adPv7e{G8w-xi{%wb?xZ>BZf3<Nt
z3KUMjZ14suD(M)NWz71=bU?xnt)u}>6&drPL70IpVCHo}6nO~Xbl284zj(<(IX1>7
z*;&`VQ8G$o_At~+{Qnt1@DAK_%bLQAE01pa3!U57!+J(Fh3CHuyFIiniI^%1=0Opb
zmcBkGl}&)-+M4!8;A<Hl1@2Prs^z9ytw}z7-BE(!Oj6QRicd^ZH-FXn$!1_?&0;}}
z?@vaW*@K_Ji3{LlR72C+iNef$Y&R0Qx5odq#gmGAeg9Yf0}%+_y)YOV*>uWgS20zm
z{oeejfS0+s*L`<Jhz(q~!yrleYwk#RqK&@n!Dn82=Gdk{KE8MUC07@B+%jXq3FqGa
z=*yohSzj8EEcix@@-Z{78s3910A4Bhqe=xsFSVq*ih`Y&7}MjBP4G)WTcvNWnHp$}
z1RF*L{Y_2dAMY;nMes<`-v4ZTU;S7WAnqy|s<`6nF`foL<8NtdyZPIqh%-UgK&})F
z=dg;NT<(1N0Geil&EeLJH;h>j-psf-UASc%ELJ7x5WR;d_5dYyD`awqio5-cL~CSd
z5*wiFxTGiMPdz9por;5P`a6#{c!Xw2g+Ny*ess=cFx(V#^yg(xag7gVHN^e&L)dz0
z^}amQmy3FjS+9uiOl$Eo_k%SCrkO8}_0ZAC#A`p^QRY6sX-X&;{Et(6)SiyZ_uY5Y
zC_j~)aQFo`-U3|xuA`(v7zH|Nib82%02~0FWSlraPy=V;G4-MW24B@XHf&n_Rj&aK
zrCI*&BEK)1Y>T&y4A7ZWQ~I_6vxx2Vxc`Gc;KzaF!(nv_c(sPR{rS_TPMS3H{JR!y
z#%$BUbq61C*4m;a+kN*;2v4E|*)fS&oS~VvNJC~~Ffiwd4*b5M#fLM7zfX!0z6}&K
z1+(|g*_duZz@#%x&CvpcPucUoGnGEq!Tx&Cuj)ZJNtcu3iUaPZXU{)y%DjbNxl*4g
zb=?|m$xsW{l_6#_KR$&T6>ZGEQg-CW(jIT_!5&y$m$#hji?!@E`-PHY38u6Cnvo5O
zrp#M*dF{F3K)f~O8@=Cuj`UIK!psHdKCo_}OXMM*U-$Yw*WL2So8J~O^A5A99jUim
z-c-bfTu0_g?_YD=0jItVpJ<ihLV83f*fe5fevCiBvflm1yeOFcJh24-mZYHnreZh}
zHX0aq*V0%2`TiQwlntQdpie4cZV7&=g@JqDwyVH#4MqbkOdwoHd&6TAk+x7^?#l{R
zI9IB6ewY(=_>&ZlAf}C4l4I)1O9v;g-ZFy4x}a>Tz);NLZ$Wr$Nd^xNC1$3;pQA%C
zg)LFGp=iBbo7z>5mvz;8e-6aLd)?+16|=nQN@lc&o$&;CyaSIUS|Ssl95Bq1ZoD*h
zzW|lEzUXR>q8Rd;a~Re?;;cURt#N;82nSf2NhKqZ2pw!{*{A6#dqCCuzPxF6l%l-8
z$c%+w^zpD-$y8px@S01GcmT9i7rO;}5&d5y3y+o!rQ&=PACZgs=p%1>`H2gsQi0K|
zf5Fyrr$q+b|4jEi0_7F~G*xik5s2WxL)LJvU=-QvLZXS;Q34L-2VecevtXCK$y7c;
z<x{>$jEVcAZ1l91SdX}jy-|p|Px`+Bcr0BUQo=ubEViH2av?!f**MFD{D(ZaPPIyk
zqlr$1@DzPg42{4o<&;Xr1~ZNWN$efBME!C9oRw}-t#sZ>wXng!k#F|#a>vafPh|S*
zHUvG=)+<6@U-Z0hEOaW3l3BXFm1)@fq5<>s3#U&i<lE?cJeUkM(5WP~?~GR!SK9RC
zfj%FQH0o{Gr{RzX*2{wCnC++kcc1rv&ivE`FBAf(Hzzi>WS-{058iwx<q1b7|80e4
z%iLQF;?2Rv%*$Oo>?{06y!#mpS#+SJ4N;V8Ry14Fu`i+e0X~ror(Du3MJ$)a^|v1o
zk1{wk91c_*t}i%ZbZ+dKtEvl=d%(|s>u$%MK~e(Ov3Q~M#o75}fljrCtCNB;>9i-+
z+PLq9i#ue;?%zR@K6Qzlail6vrf2es2~>47U43>^Lxzq&r3v6ZzC3<kADc-Rj=kd5
z&aY1k#X|ePLptieIjAA&$-SiLm=9gi1o5?5RB+VcpUQgsigzD<;PyGRKbp(rLe1gu
z-iLf68^s4xe+fdgOfyt6)$D1#zRS=Iq5rJ^Pk<fKbmm&itpJU_oCho2*JtbMYg0S{
zCUkM1Ef^Ke`R35ZXfX9~pM;a!{=2woClmu!m5_CapZE&<Kph$4s{US3QfbL~bVIFz
z4pT_dI0ZJ7k59cGF2G^V%kN<iA}RW%oWL!;ek#POu~aQF3diJUa>4lAC%1K~`ha4g
zTZ*;E0C)~DLQ)m0h<r8ruZ@mqY&dMGVY}*clSlYsG!qU6_HMg$;k<akcj#&t8{;2|
zHWk8WuCokS2birmU(JgZ8m4{g=q4zn%-u`-N7HjIIOl}9W5=DpN!$Jq+ni2QfpB6>
zs<}B7-Y@%bm!;^^rTh8wnc(<`L<3v#mmajYFMEgJR=Ar(6dgSEGtQLNO24QqI`ygt
z9{ECcB%DuYC*l4LzR%7YfS@SFiYfDgQ5JhvE;=|oacJ_iIM{k1Djl18=hiBl_ZXh;
zQA!G)DKCgy@0^O0B<!qTO8nEvq+7q}D&rIla>W?RigI-vXa~WIhRvp8|1+~f`y?M3
z5CnD2vHN)0SSS!{VR9pbt*zj99>2i^B6eRIV@3xa!ro4$qH~&|t~z{lqIvw671+2T
zmYv6={+9HO?VFdp|JW^0b*ih5`U{h2^dEBl!iR4;EE&mp_n!H=%2~oK5pOue22NQk
z>8e`Uc+9AVrc1gcyZf}iG_%ZYePt{f(cdkWwzR93%;A)bznnGI1+n(du7G!edN?ZC
z23Lh!dhVonB);=7F-$Ds3o*w$+m1aSn6y8<J<*-mlav%x5`rPV-0|#D)mq>PIP@$I
zEn%qK$uI06Wt>S`MOBp`;>Q$y2rSS~HH6KJuA6=A62p+TTpFe*DzV_4%wIi$Xc+uj
zI&tw<OK~f!P6|c#8UMD81DbW&R#oS_1^Wbp)4!;2oUSR>!1dmkC-G>x44V~ut<uE-
zI_S-v{Vpe|y{}!E86EOQ@95zz@lM|;5c$Z22TQJ^Icrbw)AZ$irt-}JqnXgmXS+-p
zr-`-|UbGAwyB$DDe~q;V{7_2aB`R#6<o5U5y)T`Tp>VGJT|ra%v%k<)-d3<z{|BoN
zU+hsz>X=yNp<F4ydv=bhS$mRhg3lS69(VOO7$#x85;D*Utv%*2g98fo*Kv&!-@E3+
z&w6-c;Kn!=X$YS3#fqCxJ3Qw3YakZ$MJ``wX^Q*x!3_*`(qaiVQBf5`F}M#8izW7&
zxx&>3uq}Sw8UC0j^|)s6qG;*v@~Mq+ialk?K(ARbdzN2MdE$*TK37cr@kA4PgW-9L
zHC?kh&!?iX^Ve70_hyCSjYoW060ws>vv34-X~()w!GOU2-^EQs7^Elu!kw#-AxbzO
z1kaT|x#7S#%`o|0((>RCc_!Sjj4zvH<Y{%CJpy|WNzo}aa6TtXGnZXiU}7|ye3D~c
zJrgt)Kk~Vaj!h<ZM1BIJanvXY^i3SgSyC-562Zw&?6+^EDS7&rvQS+A;tg}Nu(mgT
z!zS0@rPs&p>toK@f;p;sQK_hUdC}CS#E28N@)a_~{=@m9(CE2~u?~p@QhS@1VS@7)
zE1br68dmAuc}+oI^g-S<UP|tpWEnb=xuer?&8?UCMthE1>Qr8v9A><yt+N$X9e_;~
zcYOEJD>t3~U^mI;#B9CVkNtNb{3P*H?Kvi!!h@_!wz1)nWAn|J@r04mU>6-+#RuQJ
z3G584jYISH9PCLX#S=(waG*EeGe7RB22Uh;uR0!rN=!ZZEzMFPn8(?~Y>>}ttp`{Q
zRSJmUR4iUBJ(A4EN6vad#F>+eJxlIACg!7C?`YS7*^637j7}|FZ{n!Y0U085@zKOS
zN$<HmKmkoNl*Ka}*wKe=wLzg(%^Wz#8}~8S^*Tje?^j*D>$Zr8^_|yVc0Y;l)tV{9
zgPzIH2^GETYL6#=^!t3}{k$i<_Ytd9L(%oFm9O7<#c|W7#rHYp8$-u7(0|uZx)X$-
zHFm-i$y793GE`7!(b#ltD;>jLdI0OK)pdl6`sg=%#cJ|B9?CrjdlE?p0pXM8@NXV6
zHH2w<VDLdxt4XGqWjg1H4pESCbg0geksrI966~%4Kfq9gFCI8-!Xq3gE$NindPh8z
z9`~r|>ZWjWxH&QPjy|jrg;^tRW#u{HXokHV11Si=^tbc-hJ0spc&$KWwR}#H2~B-o
z#j#4VsvE}B+2)x4gf+79PFr&<aoRbBX5VFfHYeR2^|Z|U%&B}dsfismZ@I03vUgqP
z85tb8Pa?hF>@Pq`e~rP%-&9)`v+=>3l{nGD0t1%(SfMe>(1Vtqijz@y3^nD#FZLim
zbXV+wB*j6ZaICFW*H58BIORA=gRY(tHtY#77kwj?1&BvD;2Ak7RW$_w2S~O$P}tb;
z)eha1Kpdq`ed|Td0e}7}34~0!Bj}|jJt12~ECJyFF~6J_YuhjLUd1Sab7`21C=X3t
z*h5;U+{!y+J^t{5uhF~EOB(Fhk_C-{@SIOw_rn8W=k9!ZVslI8bxv>|ii~QW`H@!n
z?$FWUmML!uIGj_vaztaY5TpGgQxB_#0XqMSLnlPl(qn!p{7hHUELRpqZSkdJ8`$xo
zP;Ew$s*-}KpK|ABEIzs?Voy|3Y+<r```z}(51}DAAk|H(9EEcd^~7^CpWMz3bgMS$
zc!Cv%l+;j7@Jf}62@J$x&xVY1>{-sYM~4{y0Uz4HX6}vx<zXKH>#br(yCL*FQD{ky
zp0lZ<VyP3?xveGUi{7QW8fbN;@)*w2ctMv9&Z-XHF0ZUR#m|He{J2tCdT;}ky><Dy
z{=nYz)>nkv($S^~A1F@m(W7F~anBgw-iiYkjTqI`+~R5W)3?Buang@pp>#xb6;m}F
zEN&v}$_}|=Z@y`~H<U{bUbFzL?n=j)Lmn+*Qq-^cf?<O_kEC#zu_uY^>8nVu9bX7}
z;55X;Rcnt^E5yBNHhj!WJy^n&1oyiDb*F-JGus+p7G>8cs+w$x+N$FNp{CjIVEbYD
ziE*Pwx16^T)Jc`V2D>XxYV!O1_Y>Vz6{mFZe#xf7^WdC1nr@m;B`BJ?q}Nqc$HB37
z;<iiaSkt)o+{)JZ{}qT`)%VV{h-cqNEBaGy9?!(*Ma#OhB}T_?8xS?ecAjJXW2PK_
zU~6RbIXyNR)BH<01V;gd6doN2pA^n;6infJ^N6v5STfME%VgIK!+JtfE_%BM(*?nI
zt**aEVb3Eej$d{S$t;zWivIb<Q<<<gllIik2zaQ0c*Gl_#$UaZ2b&Hw!*zD*2RNx%
z(OlpqTN?nE)fkXm+xc?x|M1breQjZh+;^rmhDIH+Ou}3yt6PvCZ*5BKm-(_ns_d5R
z&f`ZV{yJfajdXHd>t33nQaAD7!5ZklYUOQL`UA~VKULlK1*83e%i8;Ir2~!ouc#~<
z(-a-|l;#@uq*?;(1#4`b=PO^|e)p?uzP->NY&d)+=wh`H<Szm+l<WXLbFrds$#Toz
zT%Yz8QoiQI*i<cSf&=0U)AS*KTkX{5Wv;EiM_~_CQf$D)L82g&s<Z9s<8#qn2IYsn
zsR9%7r{gWt?pe<ZngvP<)FbzBN+I%5!N$vuZ3qI-$rY=+Xz~1K$9TMf`D-1yqDyOz
z8XXI_-ML-I%u<ef|M;jk<~h5^)O8$2s(v{)6#4J7*H}2RsbY!`r@h|%^*t8&qmoil
zRcrGVK5uy1yPDm7k=K(rW^v`?`3y7a(*Ad*wD@B8_uKCKllKWmXTNT`uza~!v8_#)
z2bkcLcX{k}|MB`fKLpZ~{Gpfvxh1XA-|MWo?clL2gh|0dG#anAb|AnZI^~@%5p}du
zC+Y77ltny6q$Y&)&N*%L;HZ2y6mAXsW9dTJM@93eKeGvY?scs5fX$wun=0@V9aUg0
ze1vuKZQJ0=o$p*XH+#nVvf_eR-VpRf8V`783wWhg-xnu^*+6sQMHeD_Rn;}?`AJl4
z#F1OE*&nOoJ{B9%GWMPh1Kd##<9m1Y`GHa18E?7v*2}%V$dOAb{2h!pG~=C*4+?r>
z_wcT@>R3+<o|R29KChxHi|03nJ-J0V*)18j|L=h#MO@ibb(~lSUMRdRx4&{~t_7P$
znb<B2B@xrrAIkV^(bU(?{qK$H2~}4A<In<x234?e%;AI6#L`r(aPFHM$|BEUdZnUc
z>3qk<=6zJu>K1^9Kv&1H#m^sjZi`{^B|~{}dSg5gns@8F->h8q$dR-+?)6@{0eeY-
zRGRhlxClG?m|jZ(2UgXbrz73~bA7j^I<l&BobIkZd9*)t@Fx|k>pG7wH1|ENvf;SV
zzWAlD9Np;6-Mrl~xW6?A0?mhipc^<%3B+zsP7KAojaPFZj9A2!?AXr(>y*gg!YVj`
z-%xo|Zh!ak@i^gr?I;WZ=z7R~-+?}3>plMHUvv=+H|)tIB`{^X=k1H8VZ{hIDe0l2
zQEbFZV@K&1?E=sg97+WX;}f%Aow1n!^aTs)m}*7u>&F1Qq&n*trb4ZLUv|!^Cm%Ov
zv_Bpgee^O>!|FRgQ|W%z8;f4X<FI8-cC5##aE7|L+bVGoNDXL)`uX%mHn`v$*Xp~$
z<7?jk6W8Qk7_+Z`+$BeZe8H=?I=cGt#KsuSUbxn-XeG;#R-EP!#C@YKDVaDD3Ui_U
zJzNLldBZJpT*<1)9UojV5hSzr%Xb20^x6+PVTr-)<~<X8K1n&XVr~8B2@vM9OeRID
znBw_Kw+4MN${#)Onxz9^)lJ2~a?jXm^DhnN8n%|&oMv*o3h_*N<2eC_4vY>YBaMEV
zp@Rp!S`@Gd9*|V*znP+<$-50xg*Zx8E1!-V5oFF>t%9cUhN@`X)9L0|;FNW;W!z7V
zY?=A0r5jxrw|JPz6Dh`b#s<f6wS~?8geQ6J*RozTjMaZjw=glb;lOn`hcZqc{tG|~
z>S8fxLpCaXU)*-!m_#j?VOPLqd+=jBY{kMZ*F91~YuMvSid~%*d*JJ9Cu5UoWY^gq
zW8j`9vZ;BuFXOd}QMUAwf<vCa-zo*bHYvwZ6|<=5B~{a{H;zb*2B#M27^RuWfiL&!
zSjAZ-!Tnjm(7vs&z*;(v!K*Bu6AUyR^r2(hGB~MS1MLfAqeoG<agJ*|nQDqmf78|l
z^^>-J5?DOMe`>n};mdmmHU<*j+}Tfk@%5+ApW8}%6O$%2GM}PL`U@NLVKG<c1m4gE
zap|pda1bc7b2{)U7@C@L?Rz3-?@3tQWzTHKC)o2zYLFrK*4C$vD}Z(mS^F?(DwS-7
zlQ(9}VUMns?SXDhm$7Elkf^{Td5SR=&ts9ticu=bw%NbrmKm`~(`auz>~A~s?S4U5
z%PyvuDDJw`g2BBee}oOL$kEF7lYOj@zN3GjOLuV?)mQTa@qOv%1XEYv9J8Nq;v>DL
ztMfNEdvVVKM=uAb)7tTmLPIV^v8@L!I3N$cbMnk9W;J{6QHY=Uiv$mL98}b;XvI~J
z9*|E)vT1K<s16xnE4rtLJig7UU>&})fpJr9-8~R{zPXoN*LF<7D1UIx%q%t38U?b-
zM*M6n;)hTwGw-&gWl&s$BeOTwq!f$i;cycd9o8}g`&fsJ@9R=sz5R`aXCKy@Kje%{
z9{A4CWuv>Ks}POr`up=E@zL{EIyzP`seq???Y>mOH~&kC1GUvEn=fo06%Wi`?SOas
z&e*1w+(RX|;*^&i_*abO8fJfmc}~WPb3&sVL##Kz`bGtP&GSE5KL3B@|JiGUl>VZI
zQs5i#xuSCG@7<rxB)wsrdpA&P8D;WVrKBZ3^XkQ_YAdG1^Cn4#-gB|XlN1!f5_ohr
z&Cb`3j!{vX1o)t(u}q!`MGEkoq3G1y+m~V45iG5XsUz4X1OB1w08b4~R}B>_Nb;~D
zM^U9NT_5OJvG~LHziUU8<rGW9>6$c6cxHbl=Q(XFxOuKBL!4FlBD){!4V|(?5sZqp
z;rj6@|A^R)eYR{n-yFF2$lSx-w!!JnvoTL7m~5Q67~3|vp7qzn_Vs{!;`7Hm`=syg
z<e$FonT?`}874mvJiLTJH>*$Z8NLM54L5Dc-;YRT65a@0j||hA#8wR)Z|A3C6jeC<
z(IQss!}fdZ$l4>Do(8e!n|lDBJ@huUrMFFP@+R_RkRF|GW#QyihfgrTi0Suk!kKJw
zx@H+PRW~812VI4;k1RkTtvHyQRMC{dI=I+f1rn;@3{I}9=tA$K$-hKAm+Clelg{y)
zV{N^Q{l7ecz)7!uyS?j!Gozr}jnlt!T&}CKcFJFA`lcSNTw?Z}HHxP5f${IVSh%cc
z_1!bIp(#c+d)Pq!0p3;nB@6fCUrjPxK{$_y=E$Zl2v|Euu)u?^RK)F1&4R0st&RZo
zgP6u+xv}x=*c6oyg{R%RZjhDLZT`rn_Z<8oNQ!_H2b#(+zivi@KbeiOm>)$C=KBH|
ziptGA?=wym6pK>~1sg|yzzci|mY+PuN{&^KCBU+42FIJdH)Z^Z$dyW&BRNtSbbWW|
zzM1%5{}G%#cWztE7YMdYynUMl?6sFq-7Dn3YP%#@T+vxNbDv0LMB&A<0FDCiPFeck
z(fM3HJ?pM5B&|gM@$p8Wn{bn14ITE^#xhW_<63vmvfF0`!`~Z3!{pNePcsuprP<si
z?|0z!^+&-UhNNWXF*vK+o;rDQ$eT%du%$G!vpzJOMu^uZF8sV*P+UW{4PF$$al}C;
zLp-f2Dxu}hH*4=e*d#gDw)^IeVaGn@ItqBC2vygg&U<52IGKw2M+F(m8$1s@)}mob
z+m0W_P?v5t9arIt>la2E=~1n3+cFm8Rrs<Xtb6mb)vH$VW?AHQ7w3%p5mcl7GOPm<
zDvG?on|jaoEni$WF;0gv>7ic-05vTnqxm>pIPU%pAiO`m!95ax93>^fR<RUg{XK_f
zF?TOr*hNX%updjC(#)94K3-m|m?Dog?-VSeH`KPTtKbK~mIec>8bDj&M*2kgQpb01
zKJmmVN5v=!Q}b0#=zMSz1zvD)v@a43j!vKdA@E*?3t{US%~ar+Wgx<;ZRsnH7!i+T
zo{<Ecg-peX9W}nE*Q#K~A5N(I<B<)~daMGWyAp`Mq$@(_x_7Ra;`3+I{+3-Qu*HBr
z&VpZh$W8C_6@|0Nh^IdW{xBrPq!YMvO;J-8Up*@irmz<6*5n3wOJ|ZTbU1nX^UM2n
z$<$0Tu?Wg%6{IsC0x9GLif0WqCB=FuIQ_RQnvPn;$(1YEo6qy&!1KpW3N`rI(7u12
zc<fDUO{0W$@}=z;G=)>=ueK|W0wOr@;Xht_{?*L}D5<4ril%Wm8Lp|R@TwXU{W#z%
z(nt!4gJZi&TUI=OTrN&UV}S^h{oW`$08O*gu3FrSHC!~ePSRlqG6e~6$-q&bcE=lM
z=Sfo`k<c1h$sH4;{r<wdJ3n19fcYjUo!_SiM$e$BL+w<9eAP4)7>a4AqGe)^KKZ1g
zN{%K7tDn7M{+v07EI9e^U#=2?SPGVW6yIEU%U$pFL*&D&vg(v&bwJm69$xEO3GO+}
zns!xAhucrJ;^R+}bs(sr@s?Uz_rZfl<X9@ofWOUD8$@>ke%q9*-|gTntpuS~-P{j9
z;N(rU;iT}yF0a4;IGjkYdhP&(t9S>ms3@I1@S-Odcj~C7;FZGqNJYXr5;5)4jaW7I
z02jyxbiu?UoG%n)6&v$4p2J)wxnsr0FTS;M{Q$OL;nX*pTH!iMJK8}pg_6#dTphx4
z@PQRaDdBX@it4Bkh`F#qMYCm)OY&`lJ3{saHcu+!x-w^C0f}$_dB*fSl^&P&M|O5v
z3=);0smT{SyTyh}1}!~q;17d8j*`~y;D&bJ<vzLVusrj<_!sVCKvSS<EE-E>j(F&c
zZL-B1usFu_*j|c7&ycAGGPQb^Yiz+$cMw;*2B@B@U^yPF;S7%&sW8s#x@<X`Zs_nn
zQWU5x<HQa?OIOiBLmi2Hpc_T_R0UG%Dr^qhXfX$n>^Wg3&XJ3A@L}66_PJxaf63Gb
zw!e4JDPw3XBp8q7GCoq_39iCiiqI6EwI2G9Rf=Yj3k-LB&l<J-v-kt~q(gvnZ9(MT
zzj)#N)}b7}8j!2tntIYXn$GTj%0r)ZTM*Z9KtM3u4z`ox+zmi6Y=k20tO0bBOw4TK
zI1=<*k<IW_lO3!|&Y1@B0fBGu4;#YS#3A;_d-Pp#QWE4Mau>)-ZouKHGg_r65*9cK
z;)&Bm0;z44GhtocR=2IV|Cs%A;B!{5w0Ly{H@1!j!i5yre!6hzUEe4Q)*&10T-tT#
z_i+3XNs2&iv<E)A>(D}!&=fmm@R0ztCmbn+Mn?+q@#j6exLq*}hto`6;1wH_^?+F<
z?UZzy);0)Fh}7Y<4fr?|U}-o34n8GIV!8`Z6k?$&@HMPlT_2w!kl(e<0e7naZsNkg
zUI$`*QQ%FD(`;9@wax23x^Vv-!%+W{i^ni5#RP-7Bz@40i+jPFzztRR;}3;DDoG8=
zwFKSX@yYGeKs1?PAQ2Aij5C;Aa&#b_p_yblG5zpM7H!6XpROX|q+*!5Umf?1M!4#a
z?gDtP8dH%js_GjRV0CQ<QG6g1jsdXW3SSS>1|F+7wYEU@z1kD_$BCF+xnzLDiVaS#
zA(#8Ne{|9KapU7~srf>?TxSGvCKiN?Ei~<_cT0}?6RlW(0Q^x&3TK~)n5(b4d3GV#
zlFX)ZwNneg6lpdbj5BN^<YQ73_doZwZnZ3zRok|~iPThL>yfxaB-#ao@4;j3DHwBN
z{;O~7@;Us2%UM%Pc*JZxY!?`clXe2u#mODC5^M^Bc~O!qv-I`C11F7Rs9-o4hLA5%
zt2J5!Q=kHg-1x&UeXYaP1t2)tOzq46Nch8&)Z{_O;DlM4{P|Tg3t=Ce)WJ|V9&HR{
z^U+kEVbWPT+|&@7Fm~MaYx@mLDwjpgRKU@Mm<N2%UH#G_{_Vnps>=y6I!@$_MV$vf
z+PTbNR74uL@L_Nhga0)^#AK?vqBs>*5F!3CxNq(`D4p~Kf{n@X$qcrM{;*+`$wl@{
z-MC)0ab{BRB5*ULQp6t&e_V5q=2EeF)zqAUcWyj1$A(&Jg?H&xE|CufSl}!RP5_h2
zg575VxoLN-?druz-*KKCpcHTpCl7_Kk@Jqvh?lz4%vbNgtfONJ{DKGxKdQc0LzOeA
zqNExb-l|MRHLV(cvIu&kvW~Hms@eRO_ikR$7LP`-`US<rAwYti{h^XF@q_MLt6{%C
zS`CHA{(p$f{y1TM4S!UU!g(jPAJMlIbVvN`mQyDOu$7dwmd0_YUhPCf92IASewK<*
zG&AGcS3X+LfuL1gQWC|(wunSxcm$gR#Uv!d&#L^YzDAaj$H6zM&k0D31a@C1!df-^
z#FMBNfWlWOgxK)p$i?>WUOs0M)e>YpWGYZL&<M63cXwBxl#b@dpK$+@0sLa{a(SF(
zdGO1skAEEeVa@$6C(ChdQ5C=b`zbRER4Rwv`dE18dmyKuj;907@#!Z&^6_^YJFvwG
zrvNpu+a4!n)-(`U5KE*RG8L5;WBZ+)7<yKXmWT!;&?`kQA&p!I1QQ={h&|HQu|Z<7
z?@Q-QSL)xf^3A)?J2)TkM<G-iypXD0ygL&S;GIIY^@Mx24y8`v)$_;2ADE=O!Zk%j
z-|+HP`?s+)L#GPCK>T~UgkVGdd@dSj2&N{SaK+tkt=wMZcnzqA1$9)%R#g>Cn_@dC
zmWe=y;;hW@8R}$>BjhOw+0+`J0YGH%6{?M;z)BEL(CDI$yoXKjrebJ@&hzaXR=se`
zRVPdcd7Av0R4wy>0G>jmSIAWNm<Gkq9eTr}Z+og`9RK+Evyc=55w&7%df}Gi3Ycn2
zB@3}oty6ypgaZ&s(f(jEJLkl6?|bvR_3MkeYT$rWP6Jjc4pt_UZB-_!i1X!ZAdQ$J
zN@xm(8=mVrsw8q!0jY$QkgkgAU>OG-Sr0S?_Y*d(DW;)23dfaptY7igz2_gfe=ZRS
zMAJzIg8Aw`)e!fi<*zy^zVMjCE_-$h&cpjB!JmVqI6#zimKyCJ-+S_`JOe)HzbGlu
zQZ|;$GjuXTu|cM7!m%e^{LClIH@24)oOc=QI1eJ|svL@-0#+SED;>K)vHc!Dht~u^
zSw~iZnebXM3_$>XV$!HX-UXWKB5xaprdo2xiY0IT<A&o8ot1+<!hTO|g!sX}G5Itd
z_Xl(1!#=7o^W29&+-4}kV58F?AAcGpMH`H<x!$|{#T({L%!AFWF<=)EZeVB?WN}_5
zZ?wQnI^f_7Z+q&Uk2VwsWb`sk!%)G@<D|^Qlw&q1cBPX3O>BwM4FjYU0)B7|(0kP|
zDa8g))YSx06gl`5<8j-z#lH5%FF$t2c}LDFBtRfT{vhrHOGFc&?DD{}$WwnP7N2(Y
z%ir}F6^#Qi{=>H3{{Vj)lA1h(LbAoHuFh@${MqvlXa(o)M|esAYKBczK_ANmd`vvi
zI_rotuDs(PZ>?zW7bM9w4U_N#LMv5;kQ^u~!vjW4RV4-XB4LS33fBgQ)y@ob0sx9)
zm&NkBkKcXv_Nz~vH*;bhbdxkR<<iw?uBM_)J_Apf>H|E&+l7h8Ji4q4XBEYAq{bgQ
zSlShT8gmbJ8SXG)3FOc%Z~W?>Q>T5e`fjc4J^{W@7y|zgXi*f9HrP68)}bd|a@GAW
zeYw4dSIXc~;+Cswh77F=CniD>HlkpHs17cHZK6)OTa=f*@#<p>&pP?YnUf0;yv2Ne
zq#vTHoa6;Ojbk%zZTu(6EMW`B-}=_-UI^$_UB>=XmlP=dGvLob(rWOHEHWXCA#a(*
zZ6Dk|Z$h4mQow~6MEsa>3KRiu5a0BB0B;e@KYp5TEgX2*DHoi1=WB1iyKeiIZcY_?
zB^)BB85QGSo6((e8H{{)Pj~m0Z$5bH`ZG^GWA4PMZ8(G#9Z=Zf54;Xy?3r9Dl1<}o
z<;JC{LMrMFPC5GGr@j@c2HwFz_kilJ`V-(!V(x1&s<ofVFcEmgue#^-8PNbkMPtz>
zCPyPX8LW9$)3zUia0;A+fG?JrG&YkNf5_p-pMKieSKRjKD{sI4#oASyH*MOyv8%JQ
zYxA1bYnFYn_=_(-eEHrRuetQxW9A>U|D*!zr{nQZL!ipeAO8yS)oiH6<B1f~*%TWL
zL<>}S&bjx0wB4xaJh=*fVJ)6N4*pyub*qI*aaP<a)Xv2ZTreZr6d9BB1<>Hf2S=Oy
z3>YThj|V(76^W+7j|_QzbT*p-iES;wZStf^lO|5g7xKA$A{u6+;W#Lw&rhWjp~RSY
zDi!?+cJ2a(%Cxm6vEhl$_#@L#ePkJDqW+^1AYSgD41X??!a2qs{c8UV$!i9la!YPI
zZwliHg5T-OjLZ4|JxgPmEF1LZ3;sZ)F%Y9ti7*w7MkCRv*NgX2iab)$c#e&F{eb{T
zGwC#nhg({H&;a_cfS}DEN`&ZOfX<FfA9(p4ul3nD3#h7-!p@5Mr*TDx@Mj|_TKt+T
zM;fF^@WX9<<-w~?o-r<yN;CwALaCny5H6E}P@JJN7?H=r)kY(D36VIijy9Xg<w8J7
zhRJ5=Y$oOp!ZLKW7Qz2C&?#jYFFhv76b`@OZ*L3`4F&zQNg@J1`?I*BL-^B?w04T&
z0reCNC>A)a*p~R!!<Qd(KpSaJ|FI_2UBCwY@o*#=3P6a3>|~Ie_;v``T8+^WY;2H{
z@JX%h03!9B3;Z}pgqF{@&A;@)FE}jc1X^#ZrpB2#ZuC!yKNU%L(ID)D0vSa$wN(uV
z-npi{=9PypIkdGfCg8;wsm7)sAr*X>9-B^QYg}eSL`j)!7U-4B)!xO>2)i-aK#(xB
zwizZjxEkX78z((Zv3T3WxtHJd?9wir&BWAj$djyLr`Ax3-9MX_{;5cc%&Y;=GpZ>(
z7&wVDsVq^kwJo2#`p_kFCrrx4s6+-8mB|iOINufVaA4=K_o_-(88=8fHV_I1;sH=w
z@+mx2-{0jKtXOTSM*uz1Z*7}A@6x*;e{WkI<tsO_VQz$>cx2>Rch^5R_gu1HJ?S0O
zq0n&cP{}t|k52i=?f<@J@$zMpZit$^aIW+J#9*5A&%{GJxKy(uEP?H@;k1h7Tm3gI
zmD|_4NJ9R|KX$p=lg;aEHbre{dsKes@xi_Kc7C`2XZ>%-kIO%o*fTn>-8NUrBvk&!
zmygli>+F)gO)+~Suxq9L&DLMW5<;O1p85qnHs0nH+IDKmBH+p0e&=RXUO9Wd{!eUr
z=i#D7r8WGelbANV6@0Sg8`oSW;cU}V{mS3HyZ!Ae>$FeAt}t}i)sf}$J5@qO%_i(=
z@dw$qL%EG!zm*pj+<vGY_v66t_`1`pT@4;z{qyJ6x+{(W>Y5Jg#V)Vidx&Xg-?BuX
zrxtg9h0eFDe;ny9R~(@x$NwWBRPm3R?UD}p&Q)biyetR)s~nQ#nQ#EOR{BtO8>h*H
zmgU!&`z$|xX+5uZcSCZ6fbbswUSpXZN&j-#{yq;>*JSc^oaH09>ygwPwv8fYUz00k
za}$owTGzjB=d(G<eAX*;&mD7K7@+VX{lLT}K`J|baUM~8$FJ5TVI62P)z8f+b4>`N
z_LH=8*59s7*8BCldbN4pi$^_Pp;n&8D}ohM%U(t8;(nzf+vD7JgnOO^-^@1`uZB)u
zZ}vMPCO*Er{P*?qcPkzqoj#}dNe`!`O2P_5pBr7We*e2tn!dOO@;v5NYu43$n6jtf
zK;iPaRi`4&b$_L~`<K6WSnJ8T;L#IRwm2OZ_v}c6Qh_U;UbnpW^S{wMI@5XM1in+!
z9NK%bxzd&<N!#d`>ePN~oZYzQ`MjDZipJ{OT9pI)zsEn6Hoat@wZwc@M()}3*8ksZ
z*4y`Emh}7IIqfToVpC<zeJ8nw$huoUo(jAucE68lfrsY4*y94PRPMSuvm0uxb-vZ*
z)5`v->44EerPnq8Pi>E@{?lkJc5lZA_rjl3x$`3`9`}~peY-SCHdW?t%9|k5Ex&k{
z7Qc|JOZ^(%zHb+w!`vnRvp3}UTsHUlA<wi&!Tj=t)$iZud|&f;`vSj9_qmkH+7hDn
zulUV=eAcyj*FI(ks=v%<4Kpg-_`&!Wla{8GYtGj0t!1mWFElriVTj%SV83YTq<>}#
ze@!O1J1d(mYG^#+&FiE$b;^^KM$<$t9<B0^IJD&4rTbl0p7qkJ&bWkddRAEd5Ni1D
zas8I`47H%`-ysHs%nYw~^!fcg(Dd8Y=PfH*)|dWKyW;ul_QuUzU-o~>zg`})_Lf-G
z@qCxLOZLm|50UDebZy@CkJ16^nX-rFN+(@2d=ogz95mtZ@&CHXz|*Ptp8sC~q#yE5
d|6l+2KcnPY=Y)B?x%?S`z|+;wWt~$(69D<mBi#T1

literal 0
HcmV?d00001