intro_to_NNs_cpsc553_111021.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "id": "vi-IcCMLiBFs"
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "from torch import nn, optim\n",
    "import torch.nn.functional as F\n",
    "import torch\n",
    "from torch.utils.data import TensorDataset, DataLoader\n",
    "from matplotlib import pyplot as plt\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "id": "s7NcyiIniOau"
   },
   "outputs": [],
   "source": [
    "class ArithmeticOracle(nn.Module):\n",
    "  def __init__(self):\n",
    "    super(ArithmeticOracle, self).__init__()\n",
    "\n",
    "    # Declare first layer\n",
    "    self.fc1 = nn.Linear(2, 5)\n",
    "    # And second layer\n",
    "    self.fc2 = nn.Linear(5, 1)\n",
    "  \n",
    "  def forward(self, x):\n",
    "    # Pass input through the first layer\n",
    "    x = self.fc1(x)\n",
    "    \n",
    "    # Apply activation function\n",
    "    x = F.relu(x)\n",
    "    \n",
    "    # Pass input through the second layer\n",
    "    x = self.fc2(x)\n",
    "    \n",
    "    return x"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "id": "MmuX02aCjxE0"
   },
   "outputs": [],
   "source": [
    "def train(model, train_loader, test_loader, optimizer, epoch):\n",
    "  # Declare the loss function\n",
    "  loss_fn = nn.MSELoss()\n",
    "  for batch_idx, (data, target) in enumerate(train_loader):\n",
    "    # Zero the gradients\n",
    "    optimizer.zero_grad()\n",
    "    \n",
    "    # Run a batch through the model\n",
    "    output = model(data)\n",
    "    \n",
    "    # Compute loss\n",
    "    loss = loss_fn(output, target)\n",
    "                   \n",
    "    #Propagate loss backwards\n",
    "    loss.backward()\n",
    "\n",
    "    # Propel the optimizer to the next step\n",
    "    optimizer.step()\n",
    "    \n",
    "    # Print out our progress every few batches\n",
    "  if epoch % 1 == 0:\n",
    "    print('Train Epoch: {} [{}/{} ({:.0f}%)]\\tLoss: {:.6f}'.format(\n",
    "        epoch, batch_idx * len(data), len(train_loader.dataset),\n",
    "        100. * batch_idx / len(train_loader), loss.item()))\n",
    "  test_loss = test(model, test_loader, epoch)\n",
    "  return test_loss, loss.item()/len(target)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "id": "gUTjdI7muYWO"
   },
   "outputs": [],
   "source": [
    "\n",
    "def test(model, test_loader, epoch):\n",
    "    test_loss = 0\n",
    "    loss_fn = nn.MSELoss()\n",
    "    with torch.no_grad():\n",
    "        for data, target in test_loader:\n",
    "            #Get input data and ground truth from data loader\n",
    "            data, target = data, target\n",
    "\n",
    "            #Run them through the model\n",
    "            output = model(data)\n",
    "\n",
    "            # Add the loss for the current batch to total loss\n",
    "            test_loss += loss_fn(output, target).item()\n",
    "\n",
    "    # Return average loss per point\n",
    "    test_loss /= len(test_loader.dataset)\n",
    "    return test_loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "s-uUsDiekhUB",
    "outputId": "056a743c-2b8e-40f4-cc44-9ce9f186a17d"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Shape of input data torch.Size([1000, 2])\n",
      "Original shape of output data torch.Size([1000])\n",
      "Fixed shape of output data torch.Size([1000, 1])\n"
     ]
    }
   ],
   "source": [
    "# Seed the random generator\n",
    "torch.manual_seed(1)\n",
    "\n",
    "testsize = 1000\n",
    "\n",
    "# Generate pairs of numbers between 0 and 0.5\n",
    "example_input = torch.vstack([torch.rand(testsize, 2) * 0.5 ])\n",
    "print(\"Shape of input data\", example_input.shape)\n",
    "\n",
    "# Add the number pairs to get ground truth\n",
    "example_output = torch.sum(example_input, axis=1)\n",
    "print(\"Original shape of output data\", example_output.shape)\n",
    "\n",
    "# IMPORTANT!\n",
    "# Reshape the output to have the same number of dimensions as input data\n",
    "example_output = example_output.reshape(testsize, 1)\n",
    "print(\"Fixed shape of output data\", example_output.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "id": "HdE7iemtlpiu"
   },
   "outputs": [],
   "source": [
    "\n",
    "# Divide the training data into training and test datasets\n",
    "train_input = example_input[:-50]\n",
    "train_ground_truth = example_output[:-50]\n",
    "test_input = example_input[-50:]\n",
    "test_ground_truth = example_output[-50:]\n",
    "\n",
    "\n",
    "# Turn the data into an object that can be used for training\n",
    "train_dataset = TensorDataset(train_input, train_ground_truth)\n",
    "test_dataset = TensorDataset(test_input, test_ground_truth)\n",
    "\n",
    "#Define hyperparameters\n",
    "batch_size = 32\n",
    "learning_rate = 0.025\n",
    "momentum = 0.8\n",
    "\n",
    "\n",
    "#Create PyTorch's DataLoader objects, designed for gradient descent\n",
    "train_loader = torch.utils.data.DataLoader(train_dataset, \n",
    "                                           batch_size=batch_size, \n",
    "                                           shuffle=True)\n",
    "test_loader = torch.utils.data.DataLoader(test_dataset, \n",
    "                                          batch_size=batch_size, \n",
    "                                          shuffle=True)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "id": "OG_81XgM0SPt"
   },
   "outputs": [],
   "source": [
    "model = ArithmeticOracle()#\n",
    "optimizer = optim.SGD(model.parameters(), lr=learning_rate, momentum=momentum)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 530
    },
    "id": "RhdkFPjrpEYv",
    "outputId": "8a75f088-d51d-4932-a731-b2e4a2a527de"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train Epoch: 1 [638/950 (97%)]\tLoss: 0.036820\n",
      "Train Epoch: 2 [638/950 (97%)]\tLoss: 0.044953\n",
      "Train Epoch: 3 [638/950 (97%)]\tLoss: 0.056406\n",
      "Train Epoch: 4 [638/950 (97%)]\tLoss: 0.054116\n",
      "Train Epoch: 5 [638/950 (97%)]\tLoss: 0.027548\n",
      "Train Epoch: 6 [638/950 (97%)]\tLoss: 0.036967\n",
      "Train Epoch: 7 [638/950 (97%)]\tLoss: 0.024664\n",
      "Train Epoch: 8 [638/950 (97%)]\tLoss: 0.032527\n",
      "Train Epoch: 9 [638/950 (97%)]\tLoss: 0.038842\n",
      "Train Epoch: 10 [638/950 (97%)]\tLoss: 0.029497\n",
      "Train Epoch: 11 [638/950 (97%)]\tLoss: 0.020377\n",
      "Train Epoch: 12 [638/950 (97%)]\tLoss: 0.009334\n",
      "Train Epoch: 13 [638/950 (97%)]\tLoss: 0.009748\n",
      "Train Epoch: 14 [638/950 (97%)]\tLoss: 0.003900\n",
      "Train Epoch: 15 [638/950 (97%)]\tLoss: 0.002707\n",
      "Train Epoch: 16 [638/950 (97%)]\tLoss: 0.001329\n",
      "Train Epoch: 17 [638/950 (97%)]\tLoss: 0.001043\n",
      "Train Epoch: 18 [638/950 (97%)]\tLoss: 0.000650\n",
      "Train Epoch: 19 [638/950 (97%)]\tLoss: 0.000211\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAsMAAAHUCAYAAADftyX8AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAASdAAAEnQB3mYfeAAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+j8jraAAAgAElEQVR4nOzdeXyUVZrw/d+pJftSSQgESFQSFAFFCZugoEAAdURFg7h0z/Q8baM93c909zM20e7p6fTMPM2APd39zmtrB/ud5dPuoiBuaAIqiGwhIEtQIAElrDFJZansVef9oyrVCWSppKpyV1LX9/O5P0PVXefcV2psvDy5znWU1hohhBBCCCHCkcnoAIQQQgghhDCKJMNCCCGEECJsSTIshBBCCCHCliTDQgghhBAibEkyLIQQQgghwpYkw0IIIYQQImxJMiyEEEIIIcKWJMNCCCGEECJsSTIshBBCCCHCliTDQgghhBAibEkyLIQQQgghwpbF6ACGq4SEBJ2ent6vMZWVlaSmpg7oeUaN9Wd8WVkZAFlZWYP+bCPH+jNevjP5zgZjvHxng/+dGRW3kc+W70y+s2COP3r0aL3WOsGXzyqt9YADEz2bNGmSLi0t7dcYpRQD/f+HUWP9GT958mQAjhw5MujPNnKsP+PlO5PvbDDGy3c2+N+ZUXEb+Wz5zuQ7C+Z4pdRRrfUkXz4rZRJCCCGEECJsSTIshBBCCCHCliTDQgghhBAibEkyHEJ++ctfDrmxgRhv1LPlOxs6YwMx3qhny3c2dMYGYrxRz5XvbHDHy3c2tJ7dF9lAFyQD2UAXbgKxSSfcyHfWf/Kd9Z98Z/0n31n/yXfWf/Kd+U420AkhhBBCCOEDSYaFEEIIIUTYkmRYCCGEEEKELakZDhKpGRZCCCGEMIbUDAshhBBCCOEDSYaFEEIIIUTYkmQ4SCorK1FKdbny8/ONDksIIYQQYljJz8+/LOcCUn0dLzXDQSI1w0IIIYQQxpCaYSGEEEIIIXwgybAQQgghhAhbkgyLIcnV1ETdhg20nDhhdChCCCGEGMIsRgcgxEA0fPgh9W+9Rf1bbxExcSIJS5cSMXFiR9G8EEIIIYRPJBkWQ452OnF8/DEmm42oyZNp3LmTb44exZqVRfzSpUTdcIMkxUIIIYTwiSTDYshpPnAAV00N8cuWkXDPPcTfey/1775L46efUv3732O94gp3UjxtGsoklUBCCCGE6JlkCr1QSmUqpXKUUrbOr42OK9w5tm4Fs5nYW28FwJKaStJ3vkPa008Tu2QJ7efPU/2HP3Dx5z+ncccOtNNpcMRCCCFE6CkpKWHRokUkJSUZHYqhQioZVkqtUkrlKqVWKqXWdCSh/ozx4f4az/W6UqrgkvvZQCFQo5TSnj+XB+BHFQPUdu4cLUeOED1tGmZb1388zElJ2B56iFG/+Q1xd92Fs6aGmuef58KTT+L46CN0W5tBUQshhBA9W7t2rSHzZWdns2bNmoA+eygKmTIJpdQqAK31es/rbOB1YNFAx/hwvwDI01rbO73eB2R1esxjQDVQrrUuCcxPKwbK8dFHAMQuWNDjZ8wJCSTm5hJ/xx00FBXR8OGH2P/nf6jbtIn4O+4g5tZbMUVGDlbIQgghRK/27t1r2Hw2W5/rjsNeKK0MPwWs73jhSTynK6Uy/RjT433PCnAO0Hn+NcClpRBFWuv1kggbz9XSQuP27VjGjCFiwoQ+P2+KjSXhnntI+/d/J2HFCnA6qX3pJS488QT177yDq6lpEKIWQggherZu3TpKSgKXYgR6vnAQEivDnhVbG+4V2M6qgVzgsvX+vsYopYr6mHMdkIw7GS7pdA+6Jsg2z7OSgeKOVWQx+Jp27UI3NRF7//396hZhiooi/o47iFu4EMe2bTS8/z5169dT/957xOXkELt4Mea4uCBGLoQQQlxu/fr1FBYWUl1dTV5eHgBPPfWUd7U2Ly+PrKws7HZ36rFq1SrAnfAmJycDUF5e7r3X13z9sXbtWu+4srKyLvP09Py+7oWqkEiGcSeadJNo2oGUAY7p9b7n/UsrxjtWhIs6vbcCKMBdK/y8UqpAa11EH8rKypg8eXJfH+PIkSN9fkaA1hrHli2oqChibr55QHOoiAh38nvbbTR+9hn1775L/aZNNHzwAbHz5xN3++2X1SELIYQYHv71uQ2c+6bW6DAYPSKRf/z+MgByc3Ox2WyUlJRcVru7fPlyFi1axMqVK72v169fj81mw263e9+32+2sXr26z/n6Y/ny5axYsYLc3FzAndROmzaNsrIyioqKenx+b/cCxZfcqr9CpUyitwykp3t9jRnInE8Ba7XW5eCuNdZa52mtyz3JcwHwui8b+0RgtZWV0fb118TMmYMpOtqvuZTFQuy8eYxavZqkxx/HnJpKw+bNnH/iCex//jPtVVUBiloIIYTov5KSEtavX+9NKgFWrFhBQUEBdrudV1991bvqarPZWLFiRcCf3ZEIA2RmZpKZmcnatWt7fX6wYwuWUFkZ7qn0oLeks68x/ZpTKbUGdxlEXi/PLPaMn07X1ePLZGVlyapvADVs3QpA7Pz5AZtTmUzE3HQT0TNn0rx/P/XvvINjyxYcH39MzJw5xP/VX2FJSwvY84QQQhinYzV2KCguLsZms7F+vXfbE3v37qW8vJzc3FwKCgrIysrCZrOxcuXKgHaEKCoqIjPz8u1amZmZ7N27l1WrVvX4/GDHBr7/Rr0/5ZShkgxXAyilbN2UNZQNcIzPcyqlcoGqSxNhpVQNsLyjLEJrbfd8ubIyPIic9fU07dlDxDXXYM3ICPj8ymQieto0orKzaTl8mPq336Zx+3YaP/2U6FmziL/rLqzp6QF/rhBCCHGp8vJy7HY7ycnJXVZnc3NzvYllYWEhJSUlFBUVUVBQQHl5Oa+//nqP83WX3A5ER+1yb8/vT2yhIiTKJDydGux03biG53W3K7B9jfF1Tk/niGSt9dpL3gN3nXB5p/c75pJtmoOocft2aG/vtZ1aICiliLr+elJ/9jNGPPUUkZMn07RrFxf/8R+p+o//oPXkyaA+XwghhCgqKiI7O9tbatCZ3W5n3bp1lJeXk52dzapVqygrK+u1e0RRUZ/bnLro6dkdz+zt+f2NLVSERDLssRr3ZjXA2y2ipKOlmVIq29MH2OcxvswJLAfKPSfN5Xh6E3d0lSjqqB/2yAPWXfKeCCLtcuH46CNMCQlET58+aM+NnDCBEU88Qeo//RNRU6fSXFJC5a9+xTe/+Q0tx44NWhxCCCGGt+nTp3uTT7vdTmZmJjk5OeTk5LBu3boun123bh12u52Cgq7pUOeV3+7m64/unl1eXk55eTlr1qzp9fl9xRaqlNba6Bi8PIloR0lDFrC604EYucAarXWWr2N6u+/ZBHeSbkoetNbqkvHg7lBR1XkFuTeTJk3SpaWlvnxU9KL588+p+t3viF+6lIT77zcsjrbTp6l/5x2a9uwBrYmcMoWUv/97lCVUKo2EEEIMVWvXrqWsrAybzdalxjYvL4+UlBQyMzOprq5m5cqV3oTYZrORnJzsrSPunHT2NN+lSkpKWL16NevXr2fVqlWXtXVLSXE39KqqqvLe6+35vsQ2WJRSR7XWk3z6bCglw8OJJMOB8c1vf0vLoUOM+s1vsKT01GVv8LSdP0/tSy/RcvCgu5TCh8M/hBBCCDG4+pMMh1KZhBBdtF+8SMuhQ0TdeGNIJMIA1rQ04hYvBtyrxUIIIYQY2iQZFiHL8fHHoDWxCxcaHUoXHZ0lJBkWQgghhj5JhkVI0q2tNG7bhnnUKCIn+fRbjkFjSkzEFB9PW0WF0aEIIYQQwk+SDIuQ1LR3L66GBuIWLECZQusfU6UU1vR02isq0C6X0eEIIYQQwg+hlWUI4dGwdSsqIoKYW24xOpRuWTIy0C0tOL/5xuhQhBBCCOEHSYZFyGk9dYq2sjKiZ83CFBtrdDjd6jgJT+qGhRBCiKFNkmERchxbtwKE3Ma5zryb6KRuWAghhBjSJBkWIcXlcNC0axfWzEwirrrK6HB6ZBkzBpSSlWEhhBBiiJNkOEgqKytRSnW58vPzjQ4r5DV++im6tZXYBQuMDqVXpshILKNGSTIshBBCGCw/P/+ynAtI9XW8JMNBkpqaita6yyXJcO+0y+XeOBcbS8zMmUaH0ydrRgbOixdxtbQYHYoQQggRtvLz8y/LuYBKX8dLMixCRsvRozgvXCB23jxURITR4fTJkp4OWtN+5ozRoQghhBBigCQZFiHDsXUrKEXs/PlGh+IT6SghhBBiuCspKWHRokUkJSUZHUrQSDIsQoKzuprmkhIir78ey8iRRofjE28yLB0lhBBC+GHt2rUBnS8vL49p06YFZK7s7GzWrFkTkLlClcXoAIQAcHz8MWgd8hvnOjOnpKCiomRlWAghhF/27t0b0PlmzJgR0PlsNltA5ws1kgwLw+n2dhyffII5JYWoKVOMDsdnymT6y7HMWnfsXhVCCCF8tm7dOkpKSgI6Z25uLrm5uQGdcziTZFgYrqmkBFdtLQm5uSjT0KrcsaSn03riBC67HfMwrqcSQggReOvXr6ewsJDq6mry8vIAeOqppyguLiYvL4/k5GTy8vIoLCykvLyc119/HYDy8nJvAl1YWMi0adNYuXIl4K7xzcvLo7i4mJqaGgCKiorIy8vDbrdTWFhIUVGRd2zHnAOxdu1a76pxWVkZTz31lPf1unXrSE5O9sYLsGrVqj7vGeLSVhRyBeaaOHGiFr65+Otf64rvfle319YaHUq/1RcV6Yq/+Rvd9PnnRocihBBiCCosLNSZmZk9vl9YWKj37dunc3JyvPdycnL0qlWrvK+zs7N1YWGh9/W+ffu0zWa7bD6bzaZff/31LvMUFBT0GWNZWdll8+Xm5naZq6yszPtzFBYW6jVr1njv1dTUeOPt7V4gAaXax5xNVoaFodoqKmj98kuiZ8/GnJBgdDj91nkT3VAq8RBCiHBT9fvf037xotFhYBk5kpQf/7jPzyUnJ1NeXk5OTg7gXsXt0LFq3GH69OkUFhZ6P9tdjW9mZiZ2u937mY73ysrK+v0zlJSUsH79+i6rypmZmWRmZrJ27VoyMzN59dVXyc3NJTMzE5vNxooVKwCw2+093jPK0PqdtBh2HFu3AhC7cKHBkQyMdexYQNqrCSGECLyeNq7l5OSQnZ2N3W6nqKjIW2owkDntdnu/4yoqKiIzM/Oy9zMzM9m7dy+5ubkkJyeTlZVFUlISeXl5ZGdnA/R6zyiyMiwM42pqovGzz7BecQURWVlGhzMgpthYzCkp0l5NCCFCnC+rsaGgvLzcm2h2Xv3trKMuePny5eTk5HSbmBqlI7kuLCykpKSEoqIiCgoKutQ893bPCLIyLAzTuHMnurmZ2AULhnQnBmt6Ou1nz6Lb240ORQghxBDXsbmtJ3a7nWnTprFmzRpWrlzZJRG22+0DWuntr+zs7G5Xo8vLy8nOzmbdunXeP69atYqysjLvhr/e7hlFkmFhCK01jq1bUdHRRM+ebXQ4frFkZIDTSfu5c0aHIoQQYoiZPn26N7G02+3e5La6urrbz3d8tnMS3PFecXFxj+MCmSTn5OSQk5PDunXrusRQXl7OmjVrsNvtFBQUdBnTEW9v94wiyXCQVFZWopTqcuXn5xsdVshoPXaM9ooKYm65BVNkpNHh+MWang7ISXRCCCH6z2azsWbNGh577DFWr15NTk5Ol/KBvLy8LiunHSuqq1evpqioiPXr11NQUEBRURElJSXY7XZvG7WOdm0lJSWsXr0awHtv3bp1FBUVUVRU1OsJeB0lGZ3nBXepQ1lZGWvXrmXt2rUUFBSwb98+78+UkpLCunXrWL9+vfd+X/cGKj8//7KcC0j1dbxyd58QgTZp0iRdWlpqdBghq/q552javZuRv/411jFjjA7HL21nznDx5z8n7s47SXzgAaPDEUIIIcKeUuqo1nqSL5+VlWEx6Jx2O03FxUROmjTkE2EAS1oaWCyyMiyEEEIMQZIMi0Hn2LYNnE5iFywwOpSAUGYz1rFjpb2aEEIIMQRJMiwGlXY6afz4Y0w2G1FTpxodTsBY0tNx1dTgamgwOhQhhBBC9IMkw2JQNR84gLO6mtj581Fms9HhBIxsohNCCCGGJkmGxaBybN0KZjOx8+YZHUpAeY9lllIJIYQQYkiRZFgMmrbz52k5coTo7GzMSUlGhxNQkgwLIYQQQ5Mkw2LQOD76CGDYbJzrzJyYiCk+XsokhBBCiCFGkmExKFwtLTRu345lzBgirr3W6HCCwpqRQXtFBdrlMjoUIYQQQvhIkmExKJp270Y3NhK7YEHHyTDDjiUjA93airOy0uhQhBBCCOEjSYZF0GmtcWzZgoqMJObmm40OJ2i8HSWkblgIIYQYMiQZHgbOf2Onrd1pdBg9aisvp+2rr4iZMwdTdLTR4QSNdxOd1A0LIYQQQ4bF6ACEf1rb2sl/5k0sZhM/eHgRE7PGGh3SZRq2bgWG58a5zqxjxoBSsjIshBBCDCGyMhwklZWVKKW6XPn5+QF/jlKK+bMmcbbSTt6/v8KzLxfR0Ngc8OcMlLO+nqbdu4m45hrvyulwpSIisKSlycqwEEIIMYjy8/Mvy7mAVF/HK6118KILY5MmTdKlpaWD9rwzF6p59qUiDh0/TXJiLCsfWMDsG682fLNa/XvvUffaayQ9/jgxN91kaCyDofrZZ2nau5fRf/wjpshIo8MRQgghwpJS6qjWepIvn5WV4WFi7Khk/vXHy/n7by+htc3Jvz3/Nv/3jxuprK4zLCbtcuH46CNMCQlET5tmWByDyZKeDlrTfuaM0aEIIYQQwgeSDA8jSilyZl/Hs7/8W26dMZE9h8r5wb/8N5u2luA0oPdty+HDOCsrib31VpTVOujPN4KcRCeEEEIMLZIMD0O2+Bj+4W/vJP+H95MYF8Of1n/ET9e+RNnpC4Mah2PLFlCKmNtuG9TnGknaqwkhhBBDiyTDw1j2pKt45hd/w32LZlBecZF/WPMi//XmJzS3tAX92e2VlTQfPEjUjTdiSUkJ+vNChXnECFRUlCTDQgghxBAhyfAwFxlh5TvL5vHbvG+RlTGKDUXF/PBf/5t9R04G9bmOjz4CrYlduDCozwk1Sims6em0VVQgm1OFEEKI0CfJcJjIzBjJ2p8+xPeWz6e+oYlf/eFNfvOf71JT5wj4s3RrK43bt2MeNYrIST5t5BxWrBkZaIcDl91udChCCCGE6IMkw2HEbDKxdH42z/ziO8ycksW24i/4wT//Nx/uOBTQVcym4mJc9fXEzp+PMoXfP2IW2UQnhBBCDBnhl6kIUpMT+Plj9/Dk9+4mwmrmmRc/5Ge/e42K81UBmd+xdStYrcTecktA5htqpKOEEEIIMXRIMhymlFLMmXo1f/inv+XOeTdQWlbB3//6z7z87me0tbUPeN7Wr76i9cQJYm66CVNcXAAjHjqsY91HYstJdEIIIUTok2Q4zMVGR/L4gzn82/95kDGpNl5+dyc/+vWfOXJiYImcY+tW97xhtnGuM1NMDOaUFFkZFkIIIYYASYYFABOzxvK7p77Nt+6+hQtVtTz121d55sUPaWhs9nkOl8NB086dWDMzibjqquAFOwRYMzJoP3cO3T7wVXYhhBBCBJ/F6ABE6LBazDxw+yxuyb6GZ18u4sMdh9hzqIzv5c7nlmkTUEr1Or7xs8/Qra3ELlgQtBgbm1s5c76aigvVXKiqJToygoS4aBLiokmMjyHR8+fICGNPvLNmZNB84ADt5855a4iFEEIIEXokGQ6SysrKy5LHX/7yl+Tn5xsTUD+MGZnEv/x9Llt3l/Kfb3zM0//5Llt3l/L4gwsZlZLY7RitNY4tW1CxscTMnOnX87XW2OsaOX2+igpP4nv6fBVnzlfzjb3BpzkiIywkxsX8JVGOiyEhPtqbLCfE/SVxToyPITY6ss9kvz8snU6ik2RYCCGECJ78/Hx+9atfXfp2qq/jlRwMEByTJk3SpaWlRofht9r6Rv7zjU/4aE8pkREWHr7rZu6en43Z3LXCprm0lKq1a4m7/XYSH3zQp7mdLhcXvqml4rw72a24UO1Ofs9X42hq6fLZyAgL6aOSSU9LISMtmfS0ZEaNSKSltZ26+kZqG5qobWiivqGJ2gb367r6RuocTdTWN9Hax6ZAs8nkTZzdyXOnpDnenUzHe97vSLIv/Q46azt7los/+xlxd95J4gMP+PR9CCGEECIwlFJHtdY+HXYgK8OiV4nxMfzkO3cw/6ZJPPtSIf/15id8svcoP3xkMeOvGOX9nGPLFoBuSyRaWtu6JLodfz5zsYb2dudlz7tqbKon4U0hPS2ZjLRkUmzxmEwDX7ltbmmjzpMo1zU0UdspUa7rSJ49V/npi5cl45eKirRyx9wbWLZoBrb4mMvuW0aNAotFNtEJIYQQIU5WhoNkuKwMd9bS2sYr7+1iY1ExWmuWzs/m4bvmENHYwPknnsB87bVU3fewN+ntWO2trK6j8z9mSsGolETPSu9fkt70UckkxEUb9wN20u50Ut/Q7F1l9q441zdR52jiyPEKvjr7DRFWC3fMu4H7cmaQlBjbZY6Lv/wlztpaRv/+9wb9FEIIIUR46s/KsCTDQTIck+EOJysq+cNLhRw7dY4RSfEsbbvAzItlPBtxJYcsCd7PWS1mxo5KJn1UUpfyhjEjkwzf4OYvl0uz6/PjvPLeLk6dqSTCamHJLVO4b9EMUmzu/so1zz9P444dpD3zDOYw7bkshBBCGEGS4RAwnJNhcNf7bt7+OS+9tZ1/rDkEFjOFc+8hY/QI70rvyJQEzMP8OGaXS7Pn4AleeX8X5acvYrWYWXLLFO5fPIOIXTuoe+UVRuTlETlxotGhCiGEEGFDkuEQMNyT4Q6O3XuwP/csCbm5xN91l9HhGEZrzd5D5bzy3k5OfH0Bi8XMw9ekMLt4C4mPPELcokVGhyiEEEKEDdlAJwZN40dbwWIhZt48o0MxlFKKmVOymHF9JsWHT/LKezvZcKSC2cDBj3cxPnsWI1MS+pxHCCGEEINLkmExYC3Hj9P6xRdEz56NOUESPXAnxTOuz2T6deMoKT2F47ercZ07y+P5/x8LbppM7pJZpI3ovlezEEIIIQZfSCXDSqlVQDmQDGQBq7XWdn/G+HB/jeePmUA1kNef8eFKu1zUvvgimM3E33OP0eGEHKUU0yaPo3LCeGKOH+eaK0fx4Y5DbNl5hPmzJvHAHbNIG2EzOkwhhBAi7IVMMuxJOtFar/e8zgZeB3ostuxrjA/3C+iU/Hpe78Od9A4opnDRuH07badOEXfHHVjT0owOJ2RFZGTQWlrKvzwyn9LaVl5+bydFOw+zdfcR5s+cxPLbZzFmZJLRYfbK0dTCsZPnOP7Vea7NHMOUCVcYHZIQQggRMCGzgU4pVQNM01qX9/Zef8b0dh/3KvA+YLnWusRzLxMoAxZprYsGElOH4byBzuVwcOHJJ0EpRq1Zgyk6NHoDhyLHp59i/9OfSP7BD4ieMQOAQ8dO88q7Ozl0/DQmpbh1xkQeuGMWY0clGxytuzvG6fNVfHnyLF+Un+PLU+eoOF/l7ROdGB/Dn/7l0SHfGk8IIcTwNuQ20HlWXG24E9TOqoFcYG1/xyilivqYcx3u0odMoKTTPYDMgcQULuo2bcJVX4/tu9+VRLgP1vR0ANoqKrzJ8PXXZHD9NRkcPl7Bq+/t5KM9pXyy9yhzp1/LA7fPImN0yqDFV9fQxLFT5/ji5Dm+PHmW46fO09jc6r0/MjmBW6ZNYMJVY6iy17OhqJiPdpdy+9wbBi1GIYQQIphCIhnGnZTSTS2uHegpM+hrTK/3Pe9f+vvpHM//LcKdJPc3pmGv7exZHEVFWDMzibn5ZqPDCXnWMWNAKdoqKi67d93V6Vz3o+WUlp3h1fd28sneo2wrPsot0yaw4o7ZXBHgpNjpdHHqbCVfnjznvc5erPHej7BauPrKNCaMG+29khP/clhIc0sbRTsPs6GomEU3Xz/se0gLIYQID6GSDPe2k6ine32NGcicTwFrPSUW2QMY71VWVsbkyZP7+hhHjhzp8zOhQmtN7UsvgdOJ7ZFHUJIM9UlFRGAZPZq206d7/MykrLH86n/n8kX5WV55byfbi7/k031fMmfqNay44yauGps6oGfX1Dk6Jb5nOf7VeVpa2733R6fauG3mRCaMG82148Zw5dgRWMzmHueLirRy57wbefX9Xez+/ARzpl4zoLiEEEKIgfIlt+qvUEmGe+rO0FvS2deYfs3p6SpRrLXO8yOmYa15/35aDh8m5pZbiMjKMjqcIcOank7Tnj24mpsxRUX1+LlrM8eQ/8P7OXbqHK+8t4sdJcfYUXKMOVOvZsUdsxmX3nNS3Nbu5FRFJV+cPOtNgC9U1XrvR0damTBuDNd6VnyvuWo0ifEx/f5Z/uq2qWwoKubNwr3MvvFqlFL9nkMIIYQIJaGSDFcDKKVs3ZQllA1wjM9zKqVygapOifBAY/LKysoaUqu+fdGtrdS+/DIqKoqE3FyjwxlSOpLh9jNnfPqPiGuuGs0//d0yjn91nlff38Vn+4/z2f7j3HTDeFbceRNZGaOoste7N7idPMsXJ89R9vUF2tqd3jnSRyWzcPZkJowbw4SrRnPFmJSAlDXY4mNYeNNk3t/+OUdOnOG6q9P9nlMIIYTwla+5VX8Wa/xKhpVSCVrruk6vx+Guuy3TWm/1dR6tdYlSyk7XzWx4XhcNZIyvcyqlcoBkrfXazu95ukn0K6bhrOGDD3BWVpLwwAOYbWG7OD4glowMANpOn+7XivrVV6bxj4/fS9npC7z63i52fX6CXZ+fwJYQg72u0fu52OhIrrsmgwlX/aXWNy6m5xVof92zcBqbP/2cNwv3SjIshBBiyPN3ZfgpzwWA1vok8LxSKlEptVpr/VTPQy+zGliBJ/H01OyWdGp7lg08prV+zNcxPs65HHjdkxQDZPOXZLev+cOCs7qa+rffxpKWRtzixUaHM+RYOyXDA5GVMYqfPXYPJysu8vrm3Zy5WMPM67O8tb5jRyVjMg1eucKYkUnMvvEaPtt/jK/OfsOVY0YM2rOFEEKIQKfseJQAACAASURBVAtKmYTWulYp1a+TBLTWa5VSq5RSKz1vZQELO30kk790e/BpTG/3lVI2YAvuGuCVnaZFa618jCks1L72Grq1lcSHHkJZQqWyZugwp6SgoqO77SjRH+PSR7Lq0aUBiso/9y+awWf7j7GhqJgf//XtRocjhBBCDFi/Mxul1E+BGcBUINlTb3upTCCvm/d71blUoZt764H1/RnT2/0eWqv1K6Zw0HL8OE27dhF5ww1E3SC9ZQdCKYU1PZ22igq01sNi09nVV6Vx/dUZfLLnKN9aejMjkuKNDkkIIYQYkH4nw1rrpwE8q6U5QEE3HyvWWtd2874YQrTLRe0LL4DZjO2hh4wOZ0izZmTQevw4rpoazMnGnzQXCPctnsGhP5xm09Z9/K/7bzM6HCGEEGJABvw7b631OqVUldZ6SyADEqGjcds22r76irg778SSlmZ0OEOa9yS606eHTTKcPekqrhwzgg8+PcgDd9wU1E17QgghRLD41WtJa/1GT/eUUk/4M7cwlsvhoO6NNzAlJhK/NDTqVIcyi5+b6EKRUor7Fs2gqaWNzds/NzocIYQQYkD83g2llHoU98ayS/tt5QC/8Xd+YYy6jRtx1ddje/RRTNHRRocz5HlXhv3cRBdq5k6fwJ83fcrbH+3n7gXTiLDKBkshhBBDi18rw0qp14AHcG9EU5dcYohqO3MGx5YtWDMziZkzx+hwhgVTdDTmESOGXTJsMZu5Z8E0auocfLyn1OhwhBBCiH7zdxmnUGv9fHc3lFLf83NuYQCtNbUvvQQuF7ZvfQsVgFPLhJs1I4PmgwfRbW0oq9XocAJm0c3X88p7O9lQWEzO7OsHteexEEII4a9gZjqvBnFuESTNJSW0HDlCzNy5RGRmGh3OsGJNTwenk/Zz54wOJaBioiK4c96NnLlYw56DJ4wORwghhOgXf5PhQqXUfUqphG7u9ef0ORECdGsrta+8goqKIuH++40OZ9jxbqIbZqUSAHfdNhWrxcwbhXvRWhsdjhBCCOEzf5PhIuBPQI1Sytn5Alb5H97QVVlZiVKqy5Wfn290WL1q2LwZZ2Ul8ffcg9l26X5I4a/O7dWGm6TEWBbcNJkvT57jaNkZo8MRQggRRvLz8y/LuYBUX8f7mwzbgeXA+EuuGUCPbdfCQWpqKlrrLlcoJ8PO6mrq33kHS1oacYsWGR3OsGQZNQqs1mGZDAPcu3AaSsGbhXuNDkUIIUQYyc/PvyznAip9He/vBrq8Hg7dOKmUWu3n3GIQ1b72Grq1lcSHH0ZZpD1WMCizGevYscOyTAJg7KhkbrrhanYeOM7X56q4YnSK0SEJIYQQffL30I0eT5/TWu/3Z24xeFqOHaNp1y6ibriBqClTjA5nWLOmp+Oy23HW1xsdSlDct2gGABuKZHVYCCHE0BC0bhKyMjw0aJeL2hdeALOZxIceMjqcYc/q2UTXPkxXhyeMG83k8el8sucoVfbhmfALIYQYXvqdDCulipVSv/b8uVopVdXNVU2Yb6AbKhq3baPt66+JW7IES1qa0eEMe8PxWOZL3bdoBu1OF5u2lhgdihBCCNGngawMFwHFnj8X4z52eXo3V1hvoBsKXA4HdevXY7LZiF+61OhwwsJw7ijRYdrkcWSMTmHzpwdxNLUYHY4QQgjRq34nw1rrJ7XWb3pe5mmt92utT15ylQNSJhHi6jZuxNXQQOLy5Ziio40OJyyYExIwJSQM2010ACaT4r6c6TQ1t/LBpweNDkcIIYTolb8b6PYDKKWeUEq9qpT6oKOEQjbQhba2M2dwbNmCNSuL6NmzjQ4nrFgzMmg/cwbtchkdStDMmzGR5MQ4Nm0toa2t3ehwhBBCiB75lQwrpRKVUieAxUANcBKYrpQ63sOpdCIEaK2pffFFcLmwPfIIyhTMU7nFpawZGejWVtovXjQ6lKCxWszcsyCb6toGPin+wuhwhBBCiB75mwU9CSzSWi/WWj/uuRYDS5DjmENWc0kJLaWlxMydS0RmptHhhB1vR4lhXDcMsOSWKcRERbChcC8ulxzRLIQQIjT5mwyXa61PXvqmp2a42s+5RRDo1lZqX34ZFR1NQm6u0eGEJUvHJrphXDcMEBMdyR3zbuD0+WqKD5cbHY4QQgjRLX+T4d6We2QpKATVb96M85tviL/7bsyJiUaHE5aso0eDyTSsO0p0uOu2bCwWsxzRLIQQImT5mwwnKaXuu/RNpdSjgJzFGmLaq6poeOcdLGlpxC1aZHQ4YUtFRGBJSwuLZDjFFsf8mRMpLTvD0bIzRocjhBBCXMbfbhJPAw8qpZyeTXPHlVJOYLnWOqxrhisrK1FKdbny8/MNjanutdfQra0kPvwwymIxNJZwZ83IwFlZiaupyehQgm5ZznQANhQV9/FJIYQQov/y8/Mvy7mAVF/H+91GQGv9ADADWOe5Fmutl/g771CXmpqK1rrLZWQy3PLllzTt3k3UjTcSNWWKYXEIt47DN9rPDP/V0vS0FGZNyWL3wRNUnK8yOhwhhBDDTH5+/mU5F1Dp6/iA9NTSWpdorZ/2XFsCMacIHO1yUfvCC2CxkPjQQ0aHI+h0LPMw30TX4b5FM9AaNm7ZZ3QoQgghRBcBSYaVUvcppVZ7rmWBmFMETuMnn9B2+jRxS5ZgGTXK6HAEf2mvFg51wwATs8YyMWsMW3eXUl3bYHQ4QgghhJffybBSqhhYDywClgNvKKU2+zuvCAyXw0HdG29gstmIv+suo8MRHubkZFR0dNgkw+BeHW5vd/L2R3I4pS9av/qKug0b0G1tRocihBDDmr8n0D0HFGitTVrr6Vrr8VprE7BFKbU6MCEKf9Rt2ICroYHE5csxRUcbHY7wUEphzcigraKio7Zp2JtxXRbpo5LZvP1zGptajA4npDlra6n63e+of+stHNu2GR2OEEIMa4E4dOP5S9/0dJmQQzcM1lZRgWPrViLGjyd69myjwxGXsKanoxsbcVaHx/9UTCbFskXTcTS18MGOQ0aHE7K0y0XNunW47HZURAT1b7+Nbm01OiwhhBi2/E2G7b3c63LklFIqwc9niX7QWlP74ougNYmPPIIyBaQ8XARQuBzL3NltMyaSnBjLpq37aGt3Gh1OSKp/+21ajhwh5rbbSLj/flx2O45PPjE6LCGEGLb8zZCqlFI3XvqmUuoqLj+BLqz7Dg+25uJiWo4eJWbuXCLGjTM6HNGNcOsoAWC1Wlg6P5sqewPbir8wOpyQ03L0KPUbN2LNyMD28MPEzp+PKTGR+nfewdUipSVCCBEM/ibDa4F9SqmqzhdQBjx/yXur/I5W+ES3tlL7yiuo6GgScnONDkf0wDp2LBBeyTDA7XNvIDoqgg2Fe3G5wqNe2hfO2lqq//hHVGQkST/4ASoiAhURQfxdd+GqrcXx0UdGhyiEEMNSIMokFgM5l1zTL3m9ApD+w4Ok/v33cVZVkXDvvZgTpDolVJmiozGnpoZVRwmA2OhIbr9lCl+fq2LfkZNGhxMStMtFTUEBrtpabN/5Dta0NO+92FtvxZSURMO778rqsBBCBIG/yXCe1nqL1np/H1cRsCYQAYvetVdV0fDuu1hGjyZ24UKjwxF9sKan037uXNi1z1o6PxuL2cSbhXuNDiUk1G/aREtpKbHz5xNz001d7qmICOKXLsVVX49ji6wpCCFEoPmVDPfntDk5mW5w1L36Krq1lcSHH0ZZLEaHI/pgzcgAl4u2s2eNDmVQjUiK59YZEzlyooIvTobXz36pltJS6t96C+sVV/R4QmTs3LmYU1JoeO89XE1NgxyhEEIMb9JiYBhp+eILmvbsIWrqVKKuv97ocIQPrOnpALSHWd0wwLKc6QBsKCw2OBLjOO12b51w8t/9HSoiotvPKavVvTrc0ICjqGiQoxRCiOFNkuFhQjud2F98ESyWHleXROixhNmxzJ1dMWYEM67LZNfnxzl7scbocAaddrmoLijAVVeH7W//FkunOuHuxNxyC+bUVOo3b8bV2DhIUQohxPAnyXCQVFZWopTqcuXn5wfteY5PPqH99Gnibr8dy8iRQXuOCCzLqFFgtYZdR4kO9y2egdawsSj8Vofr33qL1qNHiV2wgJhZs/r8vLJYiL/7brTDQUNh4SBEKIQQQ0N+fv5lOReQ6ut4SYaDJDU1Fa11lytYybCroYG6N97AZLMRf9ddQXmGCA5lMmFNTx8SK8PNhw7ReupUQOeclDWWCeNGs2XXEWrqHAGdO5Q1HzlC/aZNWK+8ksQHH/R5XMycOZhHjqRh82ZcjvD5voQQojf5+fmX5VxApa/jJRkeBuo2bEA7HCQ+8ACmqCijwxH9ZE1Px1Vbi7OuzuhQetReWUnV739P9TPPoF2ugM2rlOK+RTNoa3fy7sf7AzZvKHPa7dQUFPRZJ9wdZTaTcM896KYmGjZvDmKUQggRPvxKhpVSrymlPghUMKL/XE1NNO3ZQ8T48UTPnm10OGIAhsImuvq33wanE+c339DyRWBPjps5JYuxI5N495MDNDW3BnTuUKOdTqqfew5XXR1J/+t/uctk+in6ppuwpKXRUFiIs6EhCFEKIUR48XdluBx4srsb3R3TLALPFB3NyNWrSVq5sqNGRgwx1hDfRNdeWUnjjh1YxowBoHHbtoDObzaZuDdnOo6mFj787FBA5w419Rs30vrll8QuXEj0zJkDmkOZzcTfey+6uZmG998PcIRCCBF+/E2GC4GkHu6t8HNu4SNzXJxsmhvCLJ6V4VDdRFe/aRM4ndi+/W0irrmGpuLigNerzp81CVtCDG9t2Ue70xnQuUNF8+HD1L/zTr/rhLsTPXMmljFjcBQVhXR5jRBCDAV+n0AHrFNKOZVSVUqp457rBLAqAPEJMeyZExIwJSaGZDLcfvEijTt2EDFhApETJxIzbx60t9O4c2dAnxNhtbD0tmy+qalne/GXAZ07FDhratx1wlFRJP/gByir1a/5lMlEwrJl6JYWGt57L0BRCiFEePI3GU4GHgOmAznAA57rceANP+cWImxYMzJoq6gI6Oa0QKjftAlcLhKWLQMgesYMVFQUjdu3B/xZd8y7gehIK28W7u3YCTwsaKeT6j/+EVd9vbtOOEC/xYmaNg1LRgaOrVtx2u0BmVMIIcKR3yvDWustWuv9l1xFQEEgAhQiHFjT06GtjfYLF4wOxav9wgUaP/uMiIkTibz2WgBMkZFEz5pF21df0frVVwF9XlxMFItvmcJXZ7+hpPRUQOc2krdOOCeH6BkzAjavMplIuPdedGsr9bI6LIQQA+ZXMqy13qKUuk8p9aFS6jiAUipRKfWE1npLYEIUYvjr2ETXHkKb6Oo6VoXvvbfL+7Hz5gGB30gHcPf8bMwmE28W7g343EZoPnTIXSc8bhyJKwK/jSIqOxvrlVe6V4drwu8UPyGECAR/W6v9FHgQ9yrwGwBa61qt9W+UUo8GID4hwkKobaJrO3+eps8+I3LSJCInTOhyz5qZiWXsWBp37UK3BrYVWmpyAvNmXMuhY6c5fup8QOcebM6aGmrWrXPXCX//+37XCXdHKUX8smXQ3k79O+8EfH4hhAgHfh+6obV+QGv9BnDpUo4sUwjhI+uYMWAyhUx7tfq33gKt3YnWJZRSxM6bh3Y4aCopCfizl+VMB+DNoqG7OuztJ1xfT9Kjjwa120vUDTdgHTfOfSR7VVXQniOEEMOVv8lwbwnvOD/nFiJsKKsVy+jRIbEy3HbuHE27dhE5eTKRV1/d7WeiZ88GszkopRJXjU1l+uRx7Nx/nLMXh+Z/U9dt2EDrsWPELlpE9LRpQX2WUsq9wbG9nQZZHRZCiH7zNxke36kcwrv92/Neip9zCxFWrBkZOCsrcTU1GRpHb6vCHcwJCURNnUpLaSntlT4f/+6zZYtm4NKat7bsC/jcwdZ88CANQawT7k7k9dcTMX48jm3bgvL/DyGEGM783UD3JLBYKeUEnvf0GHYCy7XWTwUkwiGqsrISpVSXKz8/3+iwRAjrOJa57cwZw2JoO3uWpt27ibzuOiLHj+/1s96NdJ9+GvA4rrs6nauvTGPLriPY6xsDPn+wOKur3XXCMTEk/93foSyWQXmut3bY6XQfnS2EEGEkPz//spwLSPV1fEBqhoEZuI9lXgcs1lov8XfeoS41NRWtdZdLkmHRm45k2MiOEh2rwgm9rAp3iLzuOszJyTRu3x7w/shKKe5bNIPWtnbe/Xh/QOcOFt3e7q4Tbmgg6bvfxZLq89/DARE5aRIR11xD46ef0n7x4qA+WwghjJSfn39ZzgX4/Gsyv5NhAK11idb6ea3109JSTYiBsXjaqxm1ia7tzBma9uwhcsoUIrKy+vy8MpmIuflmnNXVtBw5EvB4brpxPGmpNt795ABNzYHtWhEMdW++Sevx48QuWRL0OuHueGuHXS73YSlCCCF8EpBk2NNreLXn6ntJSQhxGXNyMiomxrBNdN5V4Uv6CvcmZu5cIDg9h80mE8sWTqehsZminYcDPn8gNR84QMN772HNzCRx+XLD4oicOJGIiRNp3LGD9vNDuzWdEEIMFr+TYaVUMbAeWAQsB95QSm32d14hwo1SCmt6Om2nTw/6ccRtFRU07d1L5A03EJGZ6fM4y8iRREycSFNJCc76+oDHteCmSSTGRbNxSzHtTmfA5w+E9qoqqp9/ftDrhHuScO+9oDV1b71laBxCCDFU+HvoxnNAgdbapLWerrUer7U2AVuUUqsDE6IQ4cOakYFuasJZXT2oz63buLHfq8IdYufOBaeTpp07Ax5XZISVu+ZnU1ldz//7wofsOnCceoex3TY60+3t1Dz3HNrhIOl738MyYoTRIRE5YQKRkyfTtGsXbWfPGh2OEEKEPH9Xhsu11s9f+qbW+mlgcP9tLsQwYDWgbrjt669pLi4m6sYbiRjX//bg0dOno6KjcWzbFpQV7Tvn3UBGWjIf7S7l1+s28chPn+V//+v/UPDqFnaUHDO020TdG2/QeuIEcbffTvTUqYbFcamEZctAa3fpixBCiF75+/s8ey/3yju/UEolaK3r/HyeEMNal44SN944KM/s+HV6/ABWhQFURAQxs2fj2LqVtpMn+1Vm4Yv42Gie+cV3OH2+msPHT3PkeAWHj1fw7icHePeTAwBkpCUzeXw6k69O57qrM0ixxQU0hu40HThAw/vvY83KIiE3N+jP64+I8eOJnDKFpj17aFu61PvPlRBCiMv5mwxXKaVu1Fof6PymUuoqOh3C4fGU5xoylFKZQCZQrLW2d7zWWhcZHJoYpixjxwIM2ia61q++onnfPqKmTiXiqqsGPE/MvHk4tm7FsW1bwJNhcNdTXzE6hStGp3DnvBvRWnOu0s7h46c57EmON396kM2fHgRgdKqN6zyJ8XVXp5OanBDQeNqrqqh5/nlUbCzJ3/++4XXC3UlYtozKgwep27iRlB/+0OhwhBAiZPn7N/haYJxS6tIVYhtgV0o9f8l7vSbDSqlVuFeUk4EsYLXWurfV5z7H+HDfBjwAPKa1vrQfUjbwuudzeOZZ1Fs8QvjDFB2NOTV10Mok6jduBAa+KtzBeuWVWDMyaNq9m8SHHsIUGRmI8HqklGLMyCTGjExi8c1T0FpzoaqOI8dPc/hEBUeOV1D42WEKP3N3oRiZkuBNjCePTydtRGLH/6b7Tbe3U/3ss2iHg+Qf/Sgk6oS7EzFuHFFTp9JcXEzb119jveIKo0MSQoiQFIgyicX0XR+cAqzq7QOepBWt9XrP645EtMfks68xPtzPwb3ya/Nc3XnM8/OVa61L+vg5hfCbNSOD5gMH0K2tqIiIoD2n9dQpmvfvJ2raNCKuvNKvuZRSxMybR+2LL9JcXEzMzTcHKErfn582IpG0EYksnH0dAJXVdRw5ccZbWrF11xG27nL3Q06xxbkTY0+CPHZkks/Jcd369bSVlYVcnXB34u+9l+b9+6nbsIGUH/3I6HCEECIk+ZsM5/l6yIZSqq+dNU8B3pVZrXWJUmq6UipTa10+wDG93u8od1BK9VbwV9TL84UIOGtGBs0lJbSdO+d3ktqbjlXhgXSQ6E7M7NnUvvoqjm3bBj0Z7k5qcgK3zUzgtpkTAaiubfDWGx85UcEne7/gk71fAGBLiOG6TjXHGWkpmEyXJ8dN+/fTsHkzEePHh1ydcHcirrySqGnTaN63j9ZTp/wqhRFCiOHKr2S4P6fN9fZZz4qtjctXmKuBXNzlGP0ao5Qq6u+cPbB5npWMp3bYx3FCDIh3E11FRdCS4daTJ2k+cICo6dO9HSz8ZYqLI3raNJp276b9wgUso0YFZN5ASU6MY+70a5k7/VoAausbOXLijLe0Ysf+Y3xacgyA+NgoJo93J8ZTJmRw1dhU2isrvXXCSSFaJ9ydhHvvpXnfPuo2bGDET35idDhCCBFyQuVv82SAbhJNO+4Si4GMGcic3VkBFOCuF35eKVXgywa6srIyJk+e3OfkR4JwjK0Y2gbjWOa6jlXhe+4J6Lwxc+fStHs3ju3bSQzxldPE+BjmTL2aOVOvBqChsZkjJzwrx8cr2HOwjF2fnwDgrrnXc+/xXejGRlJ+/GMsKf35K8RY1owMomfOpGnPHlrLynw6alsIIUKVL7lVf4VKMtxTvW5v9/oaM5A5u/DUGq/veK2UKgBeV0qNkxViESyWkSNRERFBS4Zby8tp+fxzomfMCNiqcIfISZMwp6TQ+OmnJCxbhjKbAzp/MMXFRDFrynhmTRkPQGNTC0fLz7L+gz1EbvmAtvYqopcsIWqQWt4FUvw999C0d697dfiJJ4wORwghQkqoJMM9JZa9Ja19jRnInH0p9oyfDvS6OpyVlSWrvmJAlMmEZezYoCXDdRs3glJ+d5DojjKZiJk7l/qNG2k5dGhIJo4dYqIjmTZ5HBMav6H24PucMMXwzlft/KzOQVJCrNHh9Yt17FiiZ82iadcuWo4fJ/Lqq40OSQghBsTX3Ko/HYP8PYEuUKrB2+bsUmUDHDOQObtQStV4Ok4AXUou/EmoheiTNT0dV10dzrrAnlPTeuIELQcPEj1zJlZPT+NAi5k7F5TCsX17UOYfTO1VVdT9139hiovj4uK7+fLri6x6+mUqLgy9Azbj77kHlKJ+wwajQxFCiJDiVzKslPqpUmq1UupGz+tXlVInlFIfeA7e8ImnZZkdd5uzzjLpYQW2rzEDmbMb5XQ6Sc9z6AaAtFgTQeU9ljnAh294V4UDXCvcmSUlhcjJk2k+cABnbW3QnjMYal94Ad3YSNKjj7L8wdv5wcOLqKypY9XTL3PkxOAcjBIo1tGjiZ4zh5bSUlq++MLocIQQImQEYmX437TWB5RS/wZka63Ha62X4O7Y0B+rcW9WA7zdIko6evsqpbI9Nbs+j/HhfmfJ3bx3aVu1PGCdtFoTwdaRDLcHsFSi5cQJWg4fJnrWLKxjxgRs3u7EzJ0LTieNn30W1OcEU/OBAzTv30/0zJneco8lt0zhF99fhtPp5Bf/sZ7txUMrqUy4+24wmajbsAGt++p2KYQQ4cHfZLhca92x9LMSd9eFDif7M5HWei3u451XKqVW4k5iF3b6SCaQ058xfd33JNircPcjtiml1nQc1OEZn6eUWuW51gBlWuvH+vNzCTEQFk97tUDWDddv2BD0VeEO0dnZqNhYGrdtG5JJl25txf7ii6jISBIffLDLvWmTx7H6/zxIQmwUT//nu7zx4Z4h8zNaRo0i5uabaf3yS1qPHjU6HCGECAn+bqDTAEqpcbjraDuXHyT1ezJ38trTvS6dHXwZ48OcJbhLHnr7jK/9iIUIGHN8PCabLWBlEi3Hj9Ny5AjRs2djHT06IHP2RlmtxMyZg6OwkNayMiLHjw/6MwOp/r33cFZWkrBiBebky39plJkxkqd/+jD//OwG/mfjdi5U1fHYAwswm0NlG0bP4u++m8bPPnN3lpg4ccDHUgshxHDh79/cKUqpJ3AfcVzkKZcYp5RajWwyE8Iv1vR02s6cQTudfs81mKvCHWLnzQOgcdu2QXtmILRfvEj9O+9gGTOGuEU9ngZPanIC//YPD3LDhCvYvP1z/m/BWzQ1tw5ipANjSU0lZu5cWj3/gSSEEOHOr2RYa/08sB9YrbVe7Hk7B3cnh6G9c0YIg1kzMqCtjfYLF/yap+XLL2kpLSV6zhysaWkBiq5v1owMrOPG0bRnD67m5kF7rj+01thfeAHa27F9+9t9njIXGx3JP/3gPhbcNJniw+X87HevUl3bMEjRDlz80qVgsVD35ptDpsRDCCGCxe/f6Wmtt2it3+j0+nmt9dOeRFkIMUCB6ihRt3EjmEzuzVODLGbuXHRzM0179gz6sweief9+d+u52bOJnDjRpzFWi5kffXsJD945m7LTF/np0y/x9bmqIEfqH0tKCrG33kpbeTktBw8aHY4QQhgqGK3Vjve3tZoQ4nJWzyY6fzpKtBw9SuvRo8TMmYNl1KhAheazmFmzwGodEqUSrpYWal98ERUdTeKKFX0P6EQpxcN3zeFH315Ctd1B3m9e5uCXXwcp0sCIv+su9+qwdJYQQoS5YLRWu3qArdWEEJ1YRo8Gs3nAK8Naa++qcLwBq8IApthYoqdPp/XECdrOnjUkBl81vPMOzqoqEpYtw2wb2JaHhbOv45c/vA+X1uQ/8wYf7S4NcJSBY05KInb+fNpOnaL5wAGjwxFCCMOETGs1IURXymrFMnr0gNurtR49SuuXXxJz881YRo4McHS+826kC+ET6drOn6f+/fexZGQQu3Bh3wN6ceO1V7LmHx7EFh/D7/7nfV57f1fIrrzG/9VfoSIiqH/zTbTLZXQ4QghhCH+T4YC2VhNCdGVNT8f5zTe4mpr6Nc67Kmw2G7Yq3CHi2msxjxxJ444d6PZ2Q2Ppjtaa2s6b5sxmv+e8amwqT696mHHpqbzw9g7+8FIh7QHoChJoZpuN2AULaDt9muYS9mfztwAAIABJREFUOVhTCBGepLVakFRWVqKU6nLl5+cbHZYYYjrqhvtbKtFSWkrrsWPuVeHU1GCE5jOlFLG33IKrro7mzz83NJbuNBcX03L4MDE330zkNdcEbN4UWzyrf7KCqZOu4sMdh/iX5zbSGIKt1+LuvBMVEUHdxo2yOiyEGJLy8/Mvy7kAn//lF8zWanZ/5h7qUlNT0Vp3uSQZFv01kGOZtdbuvsJms7uFVgiImTsXlAq5jXSu5mZqX34ZFRNDQj83zfkiJjqSX3z/XhbNuY79pad46revUGWvD/hz/GFOSCA2J4f2igqaiouNDkcIIfotPz//spwLqPR1fEBaqwHjPJ0kPgAyPa3V/uTv3EKEu4G0V2s5coTWEyeImTvX8FXhDuakJCKvv57mgwdx1tQYHY5X/aZNOKurSbj/fswJCUF5hsVs5oePLOZbS2/mZEUlP137EqfO+Px39KCIu+MOVFQU9bI6LIQIQ/62VktUSp0AFgM1uDfNTfO0VwvOv1mECCOmpCRUbKzPm+i01tSF2Kpwh9h580BrGnfsMDoUANrOnqXhgw+wXnklsfPnB/VZSikeuOMmfvI3d2Cvb+TJf3+FA198FdRn9oc5Pp64RYtoP3uWpt27jQ5HCCEGlb8rw08Ci7TWi7XWj3uuxcAS4Cn/wxMivCml3McyV1T41JGg5dAh2srKiJ03D0tKyiBE6LuoG2/EFB+PY/t2w7sraK2x//nP4HRi++u/RpkC0WWyb/NnTSL/h/ejlOJXz7zJlp2HB+W5vohbsgQVHe2uHQ7BzX5CCBEsgWitdlkLNa11Oe66YSGEn6zp6eimJpxVvZ9q5u0gYbEQd9ddgxSd75TFQsycOTgvXKD12DFDY2navdt9GMm8eURkZQ3qs6dMuII1//AgybZY/p8/f8DL735m+H8cAJji4ohbvBjnhQs07txpdDhCCDFoAtJabQD3hBA+8tYN91Eq0XLwIG3l5SG5KtwhxtNz2GHgRjpXUxO1r7yCio0lYflyQ2K4YswInv7pw2RljOTld3fyH3/+gLZ241dj4xYvRsXEUP/WWyHZBk8IIYLB32Q4SSl136VvKqUeBULz38ZCDDG+HMvceVU4PgRXhTtYx47FmpVF8969/e6dHCh1GzfisttJXL4cc3y8ITEAJCfG8eufrGD65HFs2XWEf/7DmziaWgyLB9wnBsbffjvOysqQqe0WQohg87e12tPAg0opp2fT3HGllBNYrrWWmmEhAsDiQ6/h5s8/p+3kSWJvvRVzcvJghTYgsfPmoVtbDdmo1Xb6NI7CQqyZmd5VaiNFR0Xw88fv5fa5N/D5l1/z5L+/QmV1naExxS5ahIqNpX7TJnRbm6GxCCHEYAhEa7UHgBnAOs+1WGu9xN95hRBupqgozCNH9pgMa62pHwKrwh2iZ85ERUQMeqmEd9Oc1oO6aa4vZrOJ7z+4kL+5dy5fnf2Gnz79MuWnLxoWjyn6/2/vzuOjru79j7/OzGRfSQigLApBwCCLLCqouLBUcUUQbbW1tSq2tb29rYreVh1vF1zbW729LXpv1fpr6wJ1t8qi4r5AQJRF2QRR0UDINlln5vz+mAkGCMlMZk3m/Xw8vo9k5nvO+X7myzfffDg533OyyJs5E9+ePXhWrEhYHCIi8RKV3wbW2vLg3MJ3BucdRlOriURP2sCBeL/4Att88ApmjWvW0PLJJ+ScdhrOXsm/CrojK4us446jZevWsFfWi0TDm2/S/PHH5Jx2GulHHhm344bCGMPsGcdx7eVnUeNp4MbfPcKqdQc9mxw3OdOm4SgooPaZZ/A3JXbohohIrMWya0TDJESiJG3AALCWls8/3+/9fb3CaWnknXVWgqILX7wfpPN7PFQ/+iiOvDzyZ8+OyzG7YsqEEfzqJ3NwOp386k9PsOSNtQmJw5GRQd455+CvrsazfHlCYhARiZewkmFjTKUxZk8IWyVwfYxiFkk5aYcYN9xYXk7L9u2BXuHCwkSE1iXpRx2Fq18/Gt58My6zFtQ88QT+mhry587FkZMT8+NFYuTQAdxx7cWU9Mrnv/+2lJffWZ+QOHJOOQVncTG1zz2XsIcdRUTiIdye4ZXANGBCCNvi6IUpktpap1drO6OE9fupffJJTHo6eTNnJiq0LjHGkD1lCv66OhpXr47psZq3b8ezfDnpQ4eSfeKJMT1WtAzoV8xtP7+Y3oW53Pu3JWzY8lncYzBpaeSddx7W46HuhRfifnwRkXgJNxmeb61dba3d1sm2FVgQi4BFUpGzTx9Mevp+PcON5eW0fPppt+sVbpU9eTI4HDEdKmH9fqoffhiAgm9/O2kemgtFcWEuv/jB+Tgdht8ufIov91THPYbsE0/E1a8fdS++iK+uLu7HFxGJh7B+M1hrQ+7CCadsT1RRUYExZr/N7XYnOizppozDgat//30Lb1i/n5pgr3BuN+sVbuUsLCRzzBiaPvwQbyer63VV/euv07x5MznTppF+xBExOUYslQ7sy8+/O5MaTwO/+p8nqI/zPMTG6SRv1ixsYyN1zz0X12OLiITK7XYflHMBJaHW7z7dJN1MSUkJ1tr9NiXDEom0gQPx19Tgq66mceVKvDt3kjN1Ks6CgkSH1mXZU6aAtdS//nrU2/bX1VHz+OM48vPJnzUr6u3Hywljj+I7553Mji/2cOdfnsPn88f1+FkTJ5I2cCB1y5bh27s3rscWEQmF2+0+KOcCKkKtr2RYpJtouyxzzVNPYTIyyD3zzARHFZnM0aNxFBRQ//rrWH90k7yaxYvx19ZScPHFOLKzo9p2vF0wfSJTJ41k1bpt/OWf8Z371zgc5M2eDS0t1D7zTFyPLSISD0qGRbqJ1hklap98Eu9nnwV6hfO793Texukk+8QT8VVU0LxxY9Tabd66Fc8rr5A+fDhZkyZFrd1EMcbww29OZ+TQATzzcjn/evX9uB4/c8wY0kpL8axYgbci5M4WEZFuQcmwSDfhCvYMN2/ejMnM7Pa9wq2yTz4ZiN6cw9bvD6w0ZwyF3/5269ixbi/N5eTGq86hX0khCx9bzpqN2+N2bGMMBbNng89H7VNPxe24IiLxoGRYpJtw5ubiCM4akTNtGs68vARHFB1phx1G+rBhNKxcid/jibi9+hUraNm2jdwZM/b1pvcU+bnZ3PyD88nKSOf2+59h567YPHjYnoyyMjKOPpr6N944aPEXEZHuTMmwSDeSPnQoJjub3DPOSHQoUZV98sng9VL/1lsRteOrraV60SIchYXknXdelKJLLgP6FXP9FWfT0NTMf/7pSWrq4rcgRv7s2WAttU88EbdjiojEWpeSYWPMkcaY0w9471hjzHXGmGuNMRcYY46MRoAi8rVel19O39/8BmdubqJDiaqs447DZGZS/9prEbVT8/jjWI+Hgm9+E0dWVpSiSz7HHn0kV809nV0VVdx2/9O0eH1xOW760KFkjh1Lw3vv0bw9fsM0RERiKdzlmDcZY94DxgHb2u4LLsZxJ3A/MAPYErUoRQQAR3Y2zl69Eh1G1DkyMsg67jhatm/vcpLVvHkz9a++SkZZGVnHHRflCJPPzCljOefUY/lw007+9I9lrVMJxVz+BRcAgdk6RER6gnB7ho21dqK19p/W2m3tFbDWVltrr+aAZFlEpCM5U6YAUN+FB+n2PTTndFJw6aU95qG5zlw++1TGlR3Jsrc+5MllK+NyzLRBg8g67jia1q6ladOmuBxTRCSWwk2Gl7V+ExwWcaUx5orW7VBlRUQ6k1Zaiuvww6l/+21sc3NYdT0vvUTL9u3knnEGaYcfHqMIk4/T6eC675/NoMOKefDJV3ln7ea4HDfvggvA4aBm8eK49UiLiMRKuMnwvuWHgsstPwZcHXz9vweUrYosNBFJJcYYsqdMwXo8NJSXh1zPV11NzeLFOIuKyDv33BhGmJxysjL45Q9mkZeTxd0PPM/WT7+K+THT+vUj+8QTad64kab162N+PBGRWAp7mETbF9baamBZO4kwQPzm/BGRHiF78mRwOsN6kK76scewDQ0UfOtbODIyYhhd8urXu4D/mHceXp+fX//5SSqr62J+zLzzzgOnk5pFi9Q7LCLdWrjJcHt3vN2HKFsUZtsikuKc+flkjh1L07p1Ia101vTRRzS88QYZxxxD5vjxcYgweZWV9ufHl8xg995afrvwKZqaW2J6PFfv3uSceiot27bRuHp1TI8lIhJL4SbD84wxL7bd2nsv+P5VMYi326ioqMAYs9/mdrsTHZZI0tv3IN3rr3dYzvp8gYfmXC4KU+ihuY6cdnwZF37jeD7+ZBd/+OsL+P2x7bHNO+ccTHo6Nf/8J9bvj+mxREQOxe12H5RzASWh1u/KPMPFBIZLtG7bDnhtgmUKu9B2j1FSUoK1dr9NybBI5zJGjcLRqxf1r73WYYLlWbYM786d5M2ciatfvzhGmNwuOedEJh97FK+Xf8wjz78Z02M5CwvJmTYN786dNLzzTkyPJSJyKG63+6CcC+j8z4tBrjCPd5+19oZQChpjbguzbRERjMNBzkknUfvMMzStW0fmqFEHlfHt3UvNE0/g7N2b3LPOSkCUycvhMPz7ZWfy1Z4aHnn+bfr3LeKUiUfH7Hh5M2fiefllap58kqyJEzGucH+tiIgkVrg9w4/GqKyIyD7ZJ58McMgH6aoffRTb2EjBJZek7ENzHclIT+MXV59PcWEu9zz8Ihu3fh6zYzlyc8k94wx8X35J/RtvxOw4IiKxElYyHJxOrV3GmHxjzFhjTH5nZUVEOuLq04f0ESNoKC/HV7f/zAhNGzbQ8PbbZI4ZQ9axxyYowuRXXJjLL68+H4fD8JuFT/HVnpqYHSt3xgwceXnUPvVU2HNEi4gkWrjLMd/WuiSzMebaNu8vIDAH8UvAKmOMeoVFJCI5U6aA10vDm1+Pe7Veb+ChubQ0Ci69NIHRdQ+lg/rys+/OpLq2nl/96QnqG2OTqDqyssidORNfZSWeV16JyTFERGIl3GESC4HlwSWZ7wIwxlwJXA/cYK0tstYeBdwXTJBFRLoka8IETFYWnldf3TePbd2SJXg//5y8s8/GVRLyg8IpbdLYo/jOeSex/fPd3PWX5/DFaNaH3KlTcRQWUvvss/ibmmJyDBGRWAg3Gb7KWnv1Ae/NB7Zaa+9sfcNau5wDFugQEQmHSU8ne9IkvDt30rJtG77KSmqfegpnnz7knXlmosPrVmbPOI7TTxjJyg+38uA/X43JMUx6Ovnnnou/pgbP0qUxOYaISCxEtAKdMaYAGAIsa6fsoRbjEBEJSXbrnMOvvUb1P/6BbWoKzCmcnp7gyLoXYww/+uY0yob256mXVvHC62tjcpzsKVNwlpRQ+/zz+D2emBxDRCTawk2GD0xw5xJYla69boDeXYpIRCQo7YgjcA0ciOe112h47z0yx48nc/ToRIfVLaWlufiPq86lX+8CFj6ynPc/2hH1YxiXi/zzz8fW11P3wgtRb19EJBbC7hk2xlwQ/OZ04Hag3Fr7zwMKXQA8Ep0QRSRVGWP2PUhn0tMp+Na3Eh1St5afm80vfzCLjHQXt933NJ99WRn1Y2RNmoTr8MOpW7IEX03sZrAQEYmWcKdWuxM4zhhTSWBoxFZgWut+Y8yVxpg/A4uAi6IZqIikpuxJk3Addhj5F12Eq7g40eF0e4MOK+b6K86hoamZX/3pSWo9DVFt3zgc5M+ahW1qova556LatohILJjWp7Sj1mBgHDHW2uqoNtzNlJWV2fXr1yc6DBGRdj23YjULH32JUcMGcuuPZ+NyOqPWtrWWCrebls8+o98dd+AsKopa2yIioTDGbLDWloVSNtx5hk/vrIy1ttpaWx1KWRERSYyzTjmWs04Zywcff8qfH1lONDtGjDHkz54NXi81Tz8dtXZFRGIh3DHDF4ZRdnqYbfcoFRUVGGP229xud6LDEhHZ54o5p3Fs2ZEseeMDnnppVVTbzhg1ivSjjqL+tdfwfvVVVNsWEWnL7XYflHMBIU9GH9YwCWOMH9gSYvEh1tro/d2tm9EwCRHpDjwNTVx/1z/YuWsPv5h3PseNLo1a200ffcTuBQvImjSJonnzotauiEhnwhkm4Qqz7dUEkuH3OinXG7ggzLZFRCTOcrIyuOkH53PtHX/n7gee47aff5PBA6Kzul/G8OFkHHMMDW+/TctZZ5E2YEBU2hURiaZwZ5MYDywgMIvEVmChtfbOdrb5wH0xiFdERKKsX+9C/uOqc2nx+fn1n59gb3X0FszIv+ACsJaaJ56IWpsiItEU7phhrLWrrbWLrbWLgYnGmAsO8bCckmERkW6ibOgArvnWdCoqa/nNwqdoam6JSrvpQ4aQOX48jatW0bx1a1TaFBGJprCT4bastcuDC26sMsbMDm5jg/tSemo1EZHu5vQTRjLnG8fx8Sdf8N9/WxK1GSbyL7gAjKHmn//svLCISJxFlAy3Ck6n1tpbPN0Ys8kYc2002hYRkfi59JyTOGHMUFa8t5FX3t0QlTbT+vcn64QTaPrwQ5o++igqbYqIREtUkmFjzFhjzJ+MMXsILNG8PLiJiEg34nAYrrlkBkUFOSx8dDkVldFZUjn//PPB4aBm8eKozmksIhKpLifDxpgjjTHXGmM2AeVAKXCVtdZhrb3aWrs6alGKiEjc5Odm8eNLv0F9YzN/ePhF/P7Ik1dX375kT5lC88cf0/TBB1GIUkQkOsJOho0xVxhj3iMwm8TFBB6U62WtnREcJtFaLuwV6Iwx1xtj5hhjrjLG3G6MKYy0Tgj7C4P72p1xvisxiYh0d+NHDuaMk8ew9qMdPLciOn0b+eeeCy6XeodFJKmENc+wMcYHVBFIgOdaa7d1UHwe8FIYbV8PYK1dFHw9DnicDlay66xOCPunAUOAwuAWcUwiIj3F92ZNYc3G7Tz05GuMHXEEAw8rjqg9Z1EROaefjmfJEhpXriRr4sQoRSoi0nXhrkBXSWCeYYC2FU2b9wwwEZgdzgp0xpi9wHhr7daO3gunTqhtGmPmALdba/dbeqkrMbXSCnQi0hNs2PIZN/7uUYYM7MMd130TlzOyhUV9NTV8ed11OIuL6fPrX2McUXl0RURkP7Fcge4+a+2dIQbxWKiNBntcC4HKA3ZVAnOAO8KtY4xZFm6bkcYkItLTHF3anwtmTGTRi+/y2L/e4VtnT46oPWd+PrkzZlD7zDM0vPUW2SeeGKVIRUS6Jtxk+NEwyi7ovMg+RQDW2qoD3q8CDvV3uc7qdKXNSGPaZ8uWLYwcObLTg6xbty6EUEREEuebZ01m1bptPPbC20w4ZjDDjjwsovZyzziDuuXLqXnySbKOPx7jCvdXkYikqlByq3CFuxxzyE9RhDmbREcPpR1qX2d1utJmOO2LiKSENJeTn313Jg6Hg98/9K+IV6dz5OSQd+aZ+CoqqH/11ShFKSLSNcny3/EDe19bdZR0dlanK22G036HSktL1esrIj3GEYf35tvnnMgDT7zKQ0++xlVzw54waD8506dTt3QpNU8/TfZJJ2HS06MUqYj0ZKHmVsaYzgsFJcuTC5UQmOasnX1bulinK21GGpOISI917tTxlA3tz7OvrGbNxu0RteXIzCTv7LPxV1VRt1xrNIlI4iRFMmytLSfQEzvkgF1DgGVdqdOVNiONSUSkJ3M6HPz0O2eSlZHGPX99gbr6xojayzn1VJxFRdQ9/zz+hoYoRSkiEp6kSIaDFgAXtb4IzuZQHkxKMcaMM8YsDKdOCPvbKgo3JhGRVNOvdwHfn3Mau6vquO+xkKeSb5dJTyfv3HPx19ZSt2RJlCIUEQlP0iTD1to7gD3Bld6uIpCETm1TZAgwLZw6ne0PJtjXAzcChcEV5q4PIyYRkZQzffIxHDdqCK+8u4E3yj+OqK3sk07C2bcvdS+8gL+uLkoRioiELqxFN9ptwJgLgKuBwdbao4wxBcCV1tq7ohFgd6VFN0SkJ9tb4+HHv34Iay33/vIyigpyu9xW/VtvsXfhQnJnzqRg7twoRikiqSqcRTci6hk2xlwHXAwsBBYDWGurrbV3GWOuiKRtERFJXr3yc/jRt6ZT62nkv//fEiLpWMk6/nhcAwbgWbYMX9WhJvIREYmNiIdJWGvnWmsXA+8dsGtvpG2LiEjymjT2KE47voyV67ax5I0PutyOcTjInzUL29xM7bPPRjFCEZHORZoMd5TwDo6wbRERSXJXzT2d3r3y+L/Fr/BFRdd7dTPHjSNt8GA8L7+Md/fuKEYoItKxSJPhoW2GQ+z7G1nwvVCWPBYRkW4sJyuDf/vOGTQ2tfBff/0XPr+/S+0YY8ifPRt8PmqffjrKUYqIHFpEybC19gZghjHGB9xvjNkU/P5Ca+2NUYlQRESS2pjhgzj39HFs2PI5Ty5b2eV2MkaOJH3oUOrffBNfTU0UIxQRObSojBkGJgI3APcBM6y134i0XRER6T6+fe5JDOxXxN+eeYNtOyu61IYxhpwZM8Drpf6VV6IboIjIIUQ6m8R7xph8a225tfZ+a+2d1trlbfYvMMa8aIy5NvJQRUQkWWWkp/Hv352JtfC7B5+npcXbpXayxo3D0asXdS+9hPV2rQ0RkXBE2jN8G4Ax5nRjzJFtdxhjFhBYKONqoFpTrYmI9GxDB/XlopknsP3z3fz9uTe71IZxucg9/XT8VVU0rFoV5QhFRA4WaTI8BPgEWARsMcb8qc2+OcB8a+02a+39QK8Ij9WtVFRUYIzZb3O73YkOS0Qkpi78xvEMO7If/1z6Hus37+xSG9mnnAIuF56lS6McnYj0RG63+6CcCygJtX6kyXApgZXniqy1TqDcGDM2uK8YqGxTNqVmUi8pKcFau9+mZFhEejqn08FPLzuTNJeL3z/0AvWNzeG3kZ9P9qRJNG/eTPO2bTGIUkR6ErfbfVDOBYT88EKkyfBma211m9ePEegtBiiw1rZ9HDiydZ9FRKRbGNC3iO/OmsKXe6r5y+JXutRGzrRpANQtWxbFyEREDhaNeYavDY4ZPp1AMrzMGFMAGGNMXtuyER5LRES6iZlTxjJmxCCWvPEB732wNez66UccQfqwYTS88w6+6urOK4iIdFGkyfB84DhgGbCQwNjhiwhMsTYd+D9jzBXGmEeBJREeS0REugmHw/Bv3z6DnKwM7v3bi9TU1YfdRu706eD14tE0ayISQ5EuulFtrZ1rrXVYa48KPij3GHBDcIq1Kwk8OHebtfalKMQrIiLdRO9eecy7aCpVNfX8zz+WtY7jC1nmuHE4i4rwvPyyplkTkZiJeNGNdlhgsDHm2mCyfKe1dnUMjiMiIknulIkjOHHcMN5cvYlX3t0QVl3jdJIzdWpgmrWVXV/ZTkSkIxEnw8HxwgtaN+BGAkMkLoo4OhER6daMMfzg4mn0ys/hvsdeoqIyvGWWs085BdLSqNM0ayISI5GuQHclcAeBoRDjg19bv78w4uhERKTby8/N4seXzsDT0MQ9D7+I3x/6cAlnbi7ZkybRsmULzVvDfxBPRKQzkfYMj7fWTrDWXg3cTmBs8NXW2hnAuMjDExGRnmDCMUP4xkmjef+jHTy/IryRc7nTpwOod1hEYiLSZLjtnWklMK3N65RaZENERDp2+QWn0K93AQ8++Ro7d+0JuV7awIGkjxhBw7vv4qvSrxYRia5ojBleYIwZG1x8Y64xZowxJp/AuGEREREAsjLT+ellZ+L1+vjdg//C6/OFXDd3+nTw+fC8/HIMIxSRVBRpMlxOYHxwUfD1DcBqYG+E7YqISA9UVtqfWdMnsnnHlzz+wjsh18scOxZncXFgmrWWlhhGKCKpJtJ5hrdZa2e0ziFsrS0n8ADdBGvtjdEIUEREepZvnTWJI/uX8Oi/3mbTJ7tCqrNvmrWaGhreey/GEYpIKon6PMPBuYVTfl7hiooKjDH7bW63O9FhiYgkXFqai3+/7EwcDge/e+h5mppD6+nNmTIFk55O3dKlYS/gISI9l9vtPijnAkpCrR/p1GqPGWNejKSNnqqkpARr7X6bkmERkYDBA0q45OzJfPblXv765Gsh1XHk5pI1eTIt27bRsmVLjCMUke7C7XYflHMBFaHWj7RneCuBccIHMcaMjbBtERHpwc6fNoGy0v4888pq3t+4PaQ6udMCkxbVLVsWy9BEJIVEY2q1XofYpxXoRETkkJwOBz+97AyyMtL4w19foK6+sdM6aQMGkHH00TS89x6+vXpWW0QiF2kyPB+4zxjjM8bsMcZsCm6bgeujEJ+IiPRg/XoX8v05p7K7qo77Hwtt2rQcTbMmIlHkirB+ETAPqDzg/WLgqgjbFhGRFDB98ijeeX8LL7+7nuPHlDL52GEdls8cOxZn7954Xn6ZvHPOwaSlxSlSEemJIu4ZttYut9auPmBbBiyMRoAiItKzGWO45pIZ5OVk8se/L2Vvtafj8g5HYJq12loa3n03TlGKSE8V6TzDy40xFxhjlhhjNgEYYwqMMddaa5dHJ0QREenpehXk8KNvTafW08hdDzxHS4u3w/KaZk1EoiXSqdWuAy4m0Au8GPbNM3yXMeaKKMQnIiIpYvKxw5h5ylg++PhT7nrgeXw+/yHLOnJyyD7xRFo++YTmzZvjGKWI9DQRL7phrZ1rrV0MHLgkkB7zFRGRsFx54WmcPGE4b63ZxH//fQl+/6F7fXOC06x5li6NV3gi0gNF+gBdRwnv4AjbFhGRFON0OPj3y86koaGZ5W+tIzcrk8tnn9K6otR+0vr3J2PkSBpWrsRXWYmzqCgBEYtIdxdpz/DQNsMh9v33PfhecYRti4hICnI5ncy/8hzKhvbnqZdW8ei/3j5k2Zzp08Hv1zRrItJlkT5AdwMwwxjjA+4PzjHsAy601t4YlQhFRCTlZKSncdMPZjFkYB/+/uybPPtyebvlMkePxtmnD55XXsE2N8c5ShHpCSJ9gG6stXYuMJHAssz3ATNzrNWOAAAgAElEQVSstd+IRnAiIpK6crIycF8zm/59enHf4y/z8jvrDypjHA5yg9Os1b/zTgKiFJHuLtJhEvcbY/KsteXW2vuttXdqSrWAiooKjDH7bW63O9FhiYh0K4V52fznT+bQu1cef3j4Bd5+/+CZI7JPPhmTkYFn2TJNsyaSgtxu90E5F1ASan0TyY0juOzyn4Mvl1lr13S5sR6mrKzMrl9/cC+GiIiE77MvK7nh7keob2zmlh9dwOjhg/bbX/Xww3iWL6f3jTeSMXx4gqIUkWRhjNlgrS0LpWykPcPTrbV3WWvvChzXXGeMucIYc2SE7YqIiOzTv28R7h/PJs3l5Dd/fpJNn+zab3/O1KkA1GmaNREJU6QP0G1r8/1qAotvGKDcGPNihLGJiIjsUzqwLzf/cBY+v8X9x8Xs+GLPvn1phx9OxjHH0FhejnfPng5aERHZX6QP0F0R/Hq6MeZRAvMOXwXMB+ZGHp6IiMjXyoYO4MarzqW+oZmb71nErt3V+/bltk6z9tJLCYxQRLqbSIdJ3GGM2QM8DmwDhlprJwYfpqvupK6IiEjYxo8czM++eyZ7a+q4+d5FVFbXAZAxahTOvn01zZqIhCXSZHgrMNdaW2ytvaHtsAkREZFYOXnCCH74zensqqjilnsXU+tpCEyzNm0a1uOh/q23Eh2iiHQTkSbDCw6cSs0YU6AH6EREJNa+cdJoLjv/ZLZ/vpv//J8naGhsJvukkzCZmdRpmjURCVGkD9Atbu9toNQYc20kbYuIiHRm9ozjmD1jIh9t+4IF9z2Nz5VG9kkn4f30U5o/+ijR4YlIN+CKtAFjzOnA9HZ2TQPuirR9ERGRjnznvJPx1DfxwutrueuB5/nZWafjWbaMuqVLyRgxItHhiUiSiygZNsZcCcwDVgJDCIwhJvj9hZGFJiIi0jljDPMunoqnoYnXVn3En7PSuWTUqMA0a7t34+rdO9EhikgSi7RneLy1dgKAMWYqsMVa+0nw9QXAJxG2LyIi0imnw8FPLzuT+sZmlr+1jiOP6csE+wGel16iYK5m+hSRQ4v0Abq2S/2sJDA0olVVhG2LiIiELM3l5IYrz6FsaH/+8sEu6vMK8KxYgb+pKdGhiUgSizQZxhizwBgzNjiv8FxjzBhjTD7tjyMWERGJmYz0NG76wSwGD+rL043ZWI+HBk2zJiIdiDQZLgfGA0XB1zcAqwmsRCciIhJ3OVkZuK+ZzY7DBtOAgy+fflbTrInIIUU6tdo2a+0Ma+1LwdflQC9ggrX2xmgE2F1VVFRgjNlvc7vdiQ5LRCQlFOZl88ufXszqnL5kVO7m/eeWd15JRLolt9t9UM4FlIRa3+h/y7FRVlZm169fn+gwRERS2s51H2HvXMAHrgIGXPdzRg8flOiQRCQOjDEbrLVloZSNeMywiIhIshowcjh2+NGM8lbzpz8+yqZPdiU6JBFJMkqGRUSkR+tzzlk4gEmNFbj/uJgdX+xJdEgikkSUDIuISI+WMXIkrsMP5zRTg7e+gZvvWcSu3dWJDktEkoSSYRER6dGMMeRMm4ajqZEbJ/Znb00dN9+7iMrqukSHJiJJQMmwiIj0eNmTJ2Oysjjs4w/4wcXT2FVRxS33LqbW05Do0EQkwZQMi4hIj+fIzCRnyhS8n3/OqUUuLjv/ZLZ/vpv//J8naGhsTnR4IpJASoY7YIwZYoyZZowpbPs60XGJiEj4cqZOBWPwLF3K7BnHMXvGRD7a9gUL7nualhZvosMTkQRxJTqAtowx1wNbCaxoVwossNZWRVInwv3jgMeD5QiW0zLTIiLdkKtPHzLHjqVxzRq8X33Fd847GU99Ey+8vpa7H3ye6y4/G6dTfUQiqSZpfuqDSSnW2kXW2vuARwkmol2tE+n+oHnAhcB4a22ptXZrRB9UREQSJmf6dLCWuuXLMcYw7+KpnDx+OG+u3sQf/75UyzaLpKBk6hm+ERjf+sJaW26MmWCMGdJBAtpZnUj3AyxTAiwi0jNkHH00rsMPp/7VV8mfNQtnZiY/vexM6hubWfbWh7icDuZdPBWnI2n6ikQkxpLip90YMw4oBCoP2FUJzOlKnUj3t3ldaIwZ13bssIiIdE/GGHKnT8c2NFD/xhsApLmc3HDlOYwZPigwZOIvz9Hi9SU4UhGJl2TpGS4CaGd8cBVQ3MU6ke5vdRGwkMB44fuNMQuttcs6+Txs2bKFkSNHdlaMdevWdVpGRESiJ2vyZKoffxzPsmXknHYaxuEgIz2Nm384i7sffJ7Xyz/G09DEDVeeS1ZmeqLDFZE2QsmtwpUUPcMEemjD3ddZnUj3t44lnm+t3RpMmhcCj6uHWESk+3JkZASmWfviC5rWr9/3flqai+u+fzbfOGk0qzds56Z7FlFTp3mIRXq6ZOkZPtSMER0lnZ3ViXR/e1YG908AOuwdLi0tVa+viEiSypk6lboXX6Ru6VIyjzlm3/tOh4MffnMa+TlZPP7iO9z4u0e49cdz6N0rL4HRikirUHOr4CxgIUmWnuFKgEP0uG7pYp1I92OM2dt2XuE2QyrUMywi0o25SkrIPPZYmt5/H++uXfvtM8bw7fNO4vLZp/Dprkrm3/0PPvvywMdLRKSnSIpk2FpbTqCndsgBu4ZwiB7YzupEuj/4/dbgBgQW3Qh+W97JRxIRkSSXOz0wbXzd8uXt7j9/6gR++p0z2FNVxw13P8LmHV/GMzwRiZOkSIaDFhB4WA3YN1tEeTBpJTijw8Jw6kRh/4HTqs0H7tNUayIi3V/6iBG4Bgyg/rXX8De0Pzb49BNG8h9XnUdDUwu/+K/HWPvRjjhHKSKxZpJpgvHgIhitQxH2Ww3OGDMHuN1aWxpqnSjuh8AME3ustXeE8lnKysrs+jYPZoiISPLxrFhB1QMPUHDppeROm3bIch9u2smv//QEzV4f111+FpPGHhXHKEUkXMaYDdbaspDKJlMy3JMoGRYRSX7+piZ2/exnOPPy6PPb32I6WGxj66dfcct/L6a2roEfXTKd6ZNHxTFSEQlHOMlwsswmISIiEneOjAxyTjmFuuefp+nDD8kcPfqQZYcM7MPtP7+Ym+9dxL3/bwk1dQ3MnnHcIctbvx9/fT3+mhr81dX4amsP/r6mBl9NDVhL4fe+R2ZZSL+7RSSK1DMcI+oZFhHpHrx79vDltdeSccwx9P75zzssa1ta2LNzF3/+v6fwVFRw6tEDmXTUYdhgUuuvrf062a2tBV8HK9k5HDjy8nDk5+OrqACnkz4334yrb98of0KR1KOeYRERkRC5iovJHD+expUr8axYgfV69+u1bfvVBh+0+15r5TXbqV3zdVsmIwNHfj7O4mLSBw8OfJ+fjyO4tf3ekZOzb1hG49q17Pn979nzhz9QctNNOLKy4nsSRFKYeoZjRD3DIiLdR9NHH7F7wYKDdxiDIzcXR0EBzrw8HAUFOPLycObn48vO4Ym3N7J6xx5Ky0q5+vvnkZGb0+UYap9/nprHHiPz2GMp+vGPOxy/LCId0wN0SUDJsIhI99JQXo5tbg704LYmvrm5HSalXp+PP/z1BVa8t5FRRw3kF1efR3ZWRpeOb61l78KFNLz9NnnnnUf+rFld/SgiKU/DJERERMKUNW5c2HVcTif/ftlMcrMzeW7FGn7xX49xyzWzKczLDrstYwyF3/se3i++oPapp0gbOJCsCRPCbkdEwqO/wYiIiETA4TBcNfd0vnX2ZLZ8+hU33P0IX+2p6VpbGRkU/eQnOPLz2Xv//bR8+mmUoxWRAykZjpGKigqMMfttbrc70WGJiEgMGGO4eOYkrr5oKl9U7GX+3f9gx+e7u9SWq7iYomuuwXq97LnnHnx1dVGOVqRncbvdB+VcQEmo9TVmOEY0ZlhEJDW9unIjv3/wX2RlpnPzj2YxYvDhXWrH8/LLVD30EBllZRT//OcYpzPKkYr0XOGMGVbPsIiISBRNmTCCX/7gfFq8Xm76w+OUr/+kS+3knHYaOaedRtP69VQ/+mh0gxSRfZQMi4iIRNn4kYP51U8uJM3l4td/eoLXVm7sUjsFl1xC+rBheJYswfPaa1GOUkRAybCIiEhMjBhyOLf97CLyc7O564HneP7VNZ1XOoBxuSi65hqcRUVUPfQQzVu2xCBSkdSmZFhERCRGBh3em9uvvZh+vQv58yPLeeT5twj3WR1nfj5FP/kJxuFgz7334tu7N0bRiqQmJcMiIiIx1Le4gNt+fjFDBvbh78++yf2Pv4zfH15CnH7kkRRefjn+qir23Hsvtrk5RtGKpB4lwyIiIjHWKz+H3/x0LsccNYBnX1nN7x96Hq/PF1Yb2SecQO7MmbRs3UrVX/8adg+ziLRPybCIiEgc5GRl4L5mNsePLmXFexv5zZ+foqm5Jaw28ufMIWP0aOpffx3PsmUxilQktSgZFhERiZP0NBc3XHkuU08Yyap127jpnkXU1TeGXN84HBTNm4erXz+q//EPGjWfvUjElAyLiIjEkdPp4Cff/gazpk1g49bPufF3j7KnKvRV5hw5OYEH6jIyqPzjH/F+9VUMoxXp+ZQMi4iIxJkxhu9dcAqXnX8y2z/fzQ13P8KGLZ+FXD/t8MPpNW8etr6ePffcg78x9N5lEdmfkmEREZEEmT3jOK65ZAa799Yy/+5HuPMvz/LVnpqQ6maNHUv+7Nl4d+5k7//+L9bvj3G0Ij2TK9EBiIiIpLIZJ46ibGh/Hli8gtdWfsQ772/h/GkTmD19IlmZ6R3WzT3rLFp27KDh3XepfeYZ8s87L05Ri/Qc6hmOkYqKCowx+21utzvRYYmISBIa0LeIm344i1uvmU2/3gU89q+3+cGtf2H5Wx92OCexMYbC73+ftEGDqH3iCRrKy+MYtUhycLvdB+VcQEmo9Y3mKYyNsrIyu15P+YqISJh8Pj9L3viAvz37BjV1DQwd1JfvzzmVkUMHHLKOd/duKm69FdvSQslNN5HWv38cIxZJPsaYDdbaspDKKhmODSXDIiISibr6Rh7719s8+8pqvD4/J44bxmXnT6Ff74J2yzdt3MjuO+/EWVxMn5tvxpGbG+eIRZKHkuEkoGRYRESi4fOv9vLAP1fwztotpLmcnHv6eC4843iy2xlPXLd8OdUPP0zGyJEU/+xnGKczARGLJJ6S4SSgZFhERKLp/Y3b+d9Fr7D9890U5mdz6TknMXXSSJyOrx//sdZS9eCD1K9YQe4ZZ1Bw8cUJjFgkcZQMJwElwyIiEm0+v5+lb37I355+neq6BgYPKOGKOacxatjAfWWs18vu226jefNmel15JdknnpjAiEUSQ8lwElAyLCIiseJpaOLxF97h6ZfL8Xp9TBp7FN+dNYXDSgoB8FVV8dWtt+KvraXkF78gffDgBEcsEl9KhpOAkmEREYm1LyqqePCJV3lrzSZcLifnnjaOuWccT3ZWBs3btlHx29/iyM2lzy234CwsTHS4InGjZDgJKBkWEZF4+fDjT7l/0cts21lBQW4Wl5x7EtMnH0PT22+z9777SB86lN7z52PS0hIdqkhcKBlOAkqGRUQknnx+Py+9vY6Hn36dqpp6juxfwvfnnMqRq9+k7oUXyJ4yhcLvfa91QQKRHk3JcBJQMiwiIolQ39jMohff4anlq2jx+jhh1BAu27sRPv6IgksvJXfatESHKBJzSoaTgJJhERFJpC/3VPPQE6/yevnH5Dn83GJ3kNPgofd115Fx9NGJDk8kpsJJhh2dFxEREZHupm9xAddfcQ4LfnYRfQccxl2+fjRZ2PWHe2j68qtEhyeSNNQzHCPqGRYRkWTh91tefnc97z36FN/e+zFfpudgfvhvHDt2WKJDE4kJDZNIAkqGRUQk2TQ0NrPynoUMWb+KVc4C1ow7hctnn8KAfsWJDk0kqjRMIglUVFRgjNlvc7vdiQ5LRERSWFZmOidd+yPM6DGM91VTvOYdfvifD3LtHX9j0YvvsHPXnkSHKBI2t9t9UM4FlIRaXz3DMaKeYRERSVb+xkYqfv1rWnZ+xoqySTy9q4mGxmYA+vftxQljjuKEMUM56oh+OByaik26Hw2TSAJKhkVEJJl5Kyr4yu3GNjWRNmwYlcX9WN3kYsn2KvbUNgBQVJDL8aNLOWHMUI4ZNpA0lzPBUYuERslwElAyLCIiya5582aqFy+medMm8HoBMBkZtAwYxLb0fF7b62NVtRe/MeRkZTDhmMEcP2Yo48oGk52ZnuDoRQ5NyXASUDIsIiLdhW1upnnbNpo2bKBp40aaN2/elxzb9AyqivvwoTeDN2osnzqycKa5GDPiCE4YU8rEUaX0ys9J8CcQ2Z+S4SSgZFhERLor29xM85YtNG3cGEiOt2zZlxz7XGl8ll1IeaOLjSabnc4shpcO4PgxQzlhzFAOKylMcPQiSoaTgpJhERHpKTpKjpudLjaZbDaabD525OAYMIDjxw7jhDFDGTKwT+uT/SJxpWQ4CSgZFhGRnsrf1ETLgcmxzwdAo3GyyWTzkTOHil59OGL8aE44djhlpf1xOjWjq8SHkuEkoGRYRERShb+pieYtW2jesIHGDRto3roN4w8kx/U42OzMYXtmIZllRzNi8gSOHTmYjPS0BEctPZmS4SSgZFhERFKVv6mJ5s2badqwgZq1H8CnO3AE8416HGxx5dHYfxBFE45lxOTx9CkuSHDE0tMoGU4CSoZFREQC/E1NNG/aRGX5Gmo/+JCsii9xEsg/9hoXW3N64x06jL7HT+CYY4ZSkJed4Iilu1MynASUDIuIiLTP39jI3rUfsuvNd3B+/BF59TUAeDFscuTwRXE/0kaOonTcMYwcNlBzGkvYlAwnASXDIiIiofF+9RVfvfkONStXkf3ZDpzWD8BXJp31rnyqDh9Er2PHMKpsCCMGH0ZamivBEUuyUzKcBJQMi4iIhM/f1ETThg3sfutdWj78gAxPLQDNGDY6ctmYXkBL6TCGjB7OmOFHMGRQH5wOzVIh+1MynARKSkrs7t2793vvlltuwe12JyYgERGRbsZai/fzz2lYs4a9763Csf0THMFe489NBh8689iSXUTO0UczasQRjB5xBAP7FWlu4xTjdru59dZbD3x7t7W2JJT6SoZjRD3DIiIi0eWvr6dp3Trq16yhfs37ODx1ADTgYL0zl3XOPD4r7Mvgo4cyevhAxgw/gj7F+QmOWhJBPcNJQMmwiIhI7Fi/n5YdO2h8/33q16zBu20brf3B2x1ZfOjIY50zj8a+hzF6xBGMHjaQ0cMHaaaKFKFkOAkoGRYREYkfX20tTR98QOPatTSsXQv19QB4HC4+NLl86MxjvTOXPgMOY8zwQQwffBgD+hXRv08vPZDXAykZTgJKhkVERBLD+nw0b91K4/vv07R2LS07dgDgB3ak5fI+Oex0ZNKAkyaHk7ziXvQ+rA/9+vdlwGHFDDqsmAH9irRKXjemZDgJKBkWERFJDr69e2lcu5bGtWtpWrcO29jYbjk/0IiDBuOk3jjxpaVjsrJJy80hszCf3F6FFJQUkVGQjyMrC5OdjSM7O/A1KyvwXrrmRE4GSoaTgJJhERGR5GNbWmjevBnv7t3Y+nr89fX4Gxrw1XloqKqmsaYWr8cDDQ04m5vJ8LfgDOcALheO1iQ5K2u/r/sS59xcnHl5OPLycOTnB77m5GA0RVzUhJMMa5CMiIiIpAyTlkbG0UeTEWJ5r9fHl7sq+Gz7F3y1cxd7dlVQVVGJp7IKV0sz2fjJsj6y8JHvMvRKc5DngOzGFtI99Tibv4Cm9nui9w/M4MjN3Zcg70uWg5uzNWlu/Zqbq+Q5SpQMi4iIiByCy+Wk/4B+9B/Qb7/3/X5Lxd4aPv1iT2DbVcnqXYHv6xubvy7ohNyidEr7FjC4OJcBBVkUOiHH10KWr5mM5ibSmhpwNNRj6+rw19TQ8umnNHs8HQfWNnluTZjbJsutr9tsxhlWH3fKSKphEsaY64GtQBFQCiyw1lZFUifW+w9FwyRERERSj7WWyuo6dnyxh527Kvn0iz3sCCbJtZ6Oe4hzsjLIyc4gNzuTgqx0eqU5KHZZChx+8q2PHOsl29dMRksz6U2NuBrrMfUeqKvD7/FAJzmdycnBWVCAo6Bg/6/5+TjbfO/Iz+/2vc7dcphEMOnEWrso+Hoc8Dgwvat1Yr1fREREpC1jDMWFeRQX5nHs0Ufue99aS3VdAzt3VVJTV09dfSO1nkbq6puoq28MbJ7A19r6Rr7cXY2noamTo2UD2TgdfcnvnU7vTCe90wzFaYZCh5984yfPesn2t5DZ0kRGcyPOhnqc27ZhDvEQYfBDBHqWgwlzu0lz65aT0+1X/EuanmFjzF5gvLV2a0fvhVMn1vs7+jzqGRYREZFI+Hx+PA1tkuX6Jmo9DXjqm6htJ4H21DcFE+xGmlu8Hbbtsn7yrZc866XAeinASy+npZfxUYCXPL+XXH8zOd5m0vy+Q7ZjHQ582TnYnFwI9io7CwtJKywkvagXmcXFOAsDCbTJzIxb4tzteoaDPa6FQOUBuyqBOcAd4dYxxiyL5f72YhIRERGJFqfTQX5uFvm5WWHXbW7x7kug6zwN+3qgG5taaGxuobGpOfD9vtct7Ghq4eOmZhqbvYH9wfdtYyN51kt+MHFuTaLzbQsFeMmv95Ln2U3+l7twEuhkbQlubUc+p914E32Gl0bl3ERTUiTDBMbj0s5Y3CqguIt1Yr2/Q1u2bGHkyJGdFWPdunWdlhEREREJR3qai6KCXIoKciNuy++3NLd490+ig4ly6/Z5cwtbG5vw1nnw19Rg6mpxeOpw1ntIa6gnvamB43oXRRxLKLlVuJIlGS7swr7O6sR6v4iIiEiP53AYMjPSyMxIg7zsRIcTdcmSDB9qdoaOks7O6sR6f4dKS0vV6ysiIiISRaHmVuGMTU6WeTMqAYwx7SWaW7pYJ9b7RURERKSbS4pk2FpbTqAndsgBu4YAy7pSJ9b7O/tMIiIiIpL8kiIZDloAXNT6IjhbRHkwKcUYM84YszCcOnHYLyIiIiLdWNLMMwz7FrloHau732pvxpg5wO3W2tJQ68Rj/6FonmERERGRxAhnnuGkSoZ7EiXDIiIiIokRTjKcTMMkJMWMHDkyJvMF9mQ6Z+HTOQufzln4dM7Cp3MWPp2z2FAyLCIiIiIpS8mwiIiIiKQsJcMiIiIikrKUDIuIiIhIylIynETcbne3qxuN+ok6ts5Z96kbjfqJOrbOWfepG436iTquzll86+ucda9jd0ZTq8VIV6ZWM8bQ1X+PRNWNpH7rE7GhrjMezWMnsm4k9XXOdM7iUV/nLP7nLFFxJ/LYOmc6Z7Gsr6nVkkBFRQXGmP22RPbSiIiIiPREbrf7oJwLKAm1vpLhGCkpKcFau9+mZFhEREQkutxu90E5F1ARan0lwyIiIiKSsjRmOEaMMTXAzjCrlRDG/2SSpG406kdC5yy+x9Y56z51o1E/Ejpn8T2uzll86+ucJf+xB1hr80MpqGRYRERERFKWhkmIiIiISMpSMiwiIiIiKUvJsIiIiIikLCXDIiIiIpKylAxLUjHGDDHGTDPGFLZ9nei44iXVP39XtD1fYdTReZa4SfXrLdU/f1fovhZfrkQHID2bMeb24LdDgEpgvrW2qoMq44DHg3UBtgLTYxljkunS5zfGXB8sWwSUAgs6Oc89yXxgWvB8tbXVWlt6iDopc50FfzHOBeZZa8e3sz/sa6enX28hnDPd1w7QyTnTfa0dnZwz3dfiSMmwxIwxZiFtfkkEX68icFPryDwCv2C2WmvLYxtlUgrr8wd/YWCtXRR83XpDTJWbYHs3/AuBhZ3U6/HXWbBXaAhQGNwO3B/2tdPTr7cQzpnuawfo7JwF6b7WRgjnTPe1OFIyLDER/B9v6w976w/k7cBVxphp1tplHVRfZq3dGusYk1i4n/9GYF+vgrW23BgzwRgzJEXO49K215MxZgiwJYRfBD3+Oms9L8aYOYco0pVrp0dfbx2dM93X2hfCdQa6r+0nhHOm+1ocacywxFIRgV8arSqDX4e0U7atQmPMuK6MmeohQv78wd6SQr4+t60qgY5+MfUYrT1Hbcyz1t4RQtWUvs66cu3oegN0X+sq3dfCoPtafCkZlpiw1lZZa3sd8APdOpC/o94TgIuAKmAlcH8KPgAQzucvgsD5PuD9KqA4NuElL2PMVcDSEIun+nXWlWsnpa833dciovtaF+m+FnsaJiHxdCNwR0d/wgn+ktn3iyY4Hu9xY8zgnvTgxKF04fN39D//VOwVaPeBpwOl+nUW1JVrR9fbwXRf64TuaxHTfS3G1DMscRF8+nqltXZ+mFVXErj5TYh+VN1CZ5//UDe4lPuFERx719Wxcql4nXXl2tH11obua12m+1qIdF+LDyXDEnPBH+Y91tp5IZTd2/bPOm3+N5sSN8EufP7KYL329m+JcnjJbh4h/tJI9essqCvXjq63IN3XQqf7WkR0X4sDJcMSU8EfzKK2A/87GcO0lTY/+MEnaOHrJ7d7urA+f/DJ4ioOfnhnCJ2PYexpJgB7Qiyb6tdZl64dXW8Buq+FTfe1rtN9LQ6UDEvMBJ8IvhDYGnyydVpw7sjW//WPC45pauvAaWHmA/el0FQxHX7+Q5yzBQQemqC1DFCegnNMttv7oetsn6J23uvw2tH1dvA5032tU+1dZ7qvday9c9ZK97U4MNbaRMcgPVDwz1vbaOcH2VprgmXmALcfuJpO62TrBJ4a3hPidDI9Rkefv5Nz1vonsR63UlMojDFbCDxosuyA91P2OgsmENMIJBXjgDs4+Jo65LWTitdbR+dM97X2hXGdge5rQGjnLFhO97U4UDIsIiIiIilLwyREREREJGUpGRYRERGRlKVkWERERERSlpJhEREREUlZSoZFREREJGUpGRYRERGRlKVkWERERERSlpJhEREREUlZSoZFRCRiwWWJlwZXzBIR6Ta0Ap2IiPF/i8YAAAKdSURBVERFcInZ5dbaXomORUQkVOoZFhGRaKlKdAAiIuFSMiwiIiIiKcuV6ABERCQ6jDG3A1uAQgBr7R3GmGnA7cH35gNFwe9LrbXz2mnjer7u4S0FFlhrqw4oczuwB9gabO+xtmWMMUOAacGX0621F0btQ4qIRJmSYRGRHsAY8ziw1Fp7X+trY8wca+0iY8x8YCmw1Vq7KLj/KmPMKmvt+APaeLRNmSHAKgJJcWuZhUCVtfaONnXGA20T63Ft4rjQGHNV62sRkWSjB+hERLq54INrq6y1ps17c4B51trph3qwzRhjgQuDCfNBbQTLLCWQZN8RTI63AL1ae4KDPc+V1tryQ+xvTZ7nx+rzi4hEQj3DIiLd3wSgKpgAt5oIDOmkXnmw3CICwxq2tlNma7AMwDiAtkMirLXLDqxw4LAKgsM2RESSkZJhEZHur5BA7+yiNu8tIjBGuLN6obYvItIjaTYJEZHur5x2eoGNMZ0lsUOA9zpqI/heeZsyrWOJ2x6nsx5oEZGkpWRYRKSbCw5VWGaMueqAXW1fF7ZNWoOzRixr7U1ur41g+SGt432ttVuB+zi4x7l1eIZ6kEWk29EDdCIiPcSBU561mdFhHLAcuDJYdAhQ3N5DbW3aACim/anVrg/u2wJgrb0veIwbCSTGdwALgLl8nTgvbJ2BQkQkmSgZFhHp4bRMsojIoWmYhIiIiIikLCXDIiI9X1GiAxARSVZKhkVEerDgohjzCTxAd3sIM0yIiKQUjRkWERERkZSlnmERERERSVlKhkVEREQkZSkZFhEREZGUpWRYRERERFKWkmERERERSVlKhkVEREQkZSkZFhEREZGUpWRYRERERFKWkmERERERSVlKhkVEREQkZSkZFhEREZGU9f8BLfvc1royT9cAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 720x480 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "epochs = []\n",
    "testlosses =  []\n",
    "trainlosses = []\n",
    "for epoch in range(1, 20):\n",
    "    test_loss, train_loss = train(model, train_loader,test_loader, optimizer, epoch)\n",
    "    testlosses.append(test_loss)\n",
    "    trainlosses.append(train_loss)\n",
    "    epochs.append(epoch)\n",
    "fig, axes = plt.subplots(figsize=(6,4), dpi=120)\n",
    "axes.plot(epochs, testlosses, label='test loss')\n",
    "axes.plot(epochs, trainlosses, label=\"train loss\")\n",
    "axes.set(xlabel='epoch', ylabel='average MSE loss per point')\n",
    "axes.legend();\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "vxB2jd1SpeUO",
    "outputId": "5acb7fc4-6fab-4366-f1bf-63a1926466bd"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The oracle says 0.05 plus 0.75 equals 8e-01.\n"
     ]
    }
   ],
   "source": [
    "# Have your new arithmetic oracle do some additions!\n",
    "\n",
    "a = 0.05\n",
    "b = 0.75\n",
    "\n",
    "\n",
    "please_add = model(torch.Tensor([a,b]))\n",
    "print(f\"The oracle says {a} plus {b} equals {please_add[0]:0.0e}.\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "QpozNjlNvm64"
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "colab": {
   "collapsed_sections": [],
   "name": "intro_to_NNs_cpsc553_111021.ipynb",
   "provenance": []
  },
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.7.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}