Infinite_LR_Schedule.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "OriyDhx9S77u"
      },
      "outputs": [],
      "source": [
        "import math\n",
        "import math\n",
        "import math\n",
        "\n",
        "def get_lr(num_iters, warmup_iter, start_lr, end_iter, min_lr, decay_style, num_repeats=1, constant_lr=0, cooldown_iters_percent=0, cooldown_iter_before_scale=0, timescale=0, constant_iters_percent=0, max_lr=0):\n",
        "    \"\"\"Learning rate decay functions from:\n",
        "    https://openreview.net/pdf?id=BJYwwY9ll pg. 4\"\"\"\n",
        "\n",
        "    constant_iter = constant_iters_percent * end_iter\n",
        "    cooldown_iter = cooldown_iters_percent * end_iter\n",
        "\n",
        "    if decay_style != 'cosine-inf':\n",
        "      num_repeats = 1\n",
        "    repeat_iter_interval = int(end_iter / num_repeats)\n",
        "\n",
        "    if decay_style == \"cosine-inf\":\n",
        "      warmup = warmup_iter / end_iter\n",
        "      warmup_iter = warmup * int(end_iter / num_repeats)\n",
        "\n",
        "    # Warmup.\n",
        "\n",
        "    if warmup_iter > 0 and num_iters <= warmup_iter:\n",
        "        return float(start_lr) * num_iters / warmup_iter\n",
        "    elif decay_style == 'cosine-inf' \\\n",
        "    and warmup_iter > 0 \\\n",
        "    and num_iters % repeat_iter_interval <= warmup_iter:\n",
        "        return float(min_lr) + float(start_lr - min_lr) * (num_iters % repeat_iter_interval) / warmup_iter\n",
        "\n",
        "    num_iters = (num_iters % repeat_iter_interval) - warmup_iter\n",
        "    if decay_style == \"linear\":\n",
        "        lr = start_lr * (end_iter - num_iters) / end_iter\n",
        "    elif decay_style == \"cosine\":\n",
        "        end_iter_ = end_iter - warmup_iter\n",
        "        lr = min_lr + (\n",
        "            (start_lr - min_lr)\n",
        "            / 2.0\n",
        "            * (math.cos(math.pi * num_iters / end_iter_) + 1)\n",
        "        )\n",
        "    elif decay_style == \"cosine-inf\":\n",
        "        end_iter_ = repeat_iter_interval - warmup_iter\n",
        "        lr = min_lr + (\n",
        "            (start_lr-min_lr)\n",
        "            / 2.0\n",
        "            * (math.cos(math.pi * num_iters / end_iter_) + 1)\n",
        "        )\n",
        "    elif decay_style == \"exponential\":\n",
        "        # exp(-0.693) = 1/2\n",
        "        lr = start_lr * math.exp(-0.693 * num_iters / end_iter)\n",
        "    elif \"infinite\" in decay_style:\n",
        "      if num_iters <= constant_iter:\n",
        "          if num_iters <= cooldown_iter:\n",
        "            if decay_style == \"constant_infinite\":\n",
        "              # Do linear decay from start_lr to constant_lr\n",
        "              lr = start_lr - ((start_lr - constant_lr) * num_iters) / cooldown_iter\n",
        "              return lr\n",
        "            elif decay_style == \"inverse_sqrt_infinite\":\n",
        "              def inv_sqrt(x):\n",
        "                return start_lr/math.sqrt((x + timescale)/timescale)\n",
        "\n",
        "              def y_shifted(x, func, A, B_new, x_start, x_end):\n",
        "                return ((B_new - A) / (func(x_end) - func(x_start))) * func(x) + A - ((B_new - A) / (func(x_end) - func(x_start))) * func(x_start)\n",
        "\n",
        "              def x_shifted(x, func, x_start, x_end, x_end_new):\n",
        "                k = (x_end_new - x_start) / (x_end - x_start)\n",
        "                return func(((x - x_start) / k) + x_start)\n",
        "\n",
        "              y_shifted_func = lambda x: y_shifted(x, inv_sqrt, start_lr, constant_lr, 0, cooldown_iter_before_scale)\n",
        "              x_shifted_func = lambda x: x_shifted(x, y_shifted_func, 0, cooldown_iter_before_scale, cooldown_iter)\n",
        "              lr = x_shifted_func(num_iters)\n",
        "              return lr\n",
        "            else:\n",
        "              raise NotImplementedError\n",
        "          else:\n",
        "            # Stay at constant LR\n",
        "            lr = constant_lr\n",
        "            return lr\n",
        "      else:\n",
        "        # Go from constant iters to min LR in remaining iters\n",
        "        end_iter_ = end_iter - warmup_iter - constant_iter\n",
        "        num_iters = num_iters - constant_iter\n",
        "        exp_factor = -math.log(min_lr/constant_lr) / end_iter_\n",
        "        lr = constant_lr * math.exp(-1* exp_factor * num_iters)\n",
        "    else:\n",
        "        lr = start_lr\n",
        "    return max(lr, min_lr)"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import matplotlib.pyplot as plt\n",
        "\n",
        "start_lr = 0.01\n",
        "min_lr = 0.00001\n",
        "constant_lr = 0.005\n",
        "end_iter = 136363\n",
        "warmup_iter_percent = 0.01\n",
        "warmup_iter = int(warmup_iter_percent * end_iter)\n",
        "constant_iters_percent = 0.9\n",
        "cooldown_iters_percent = 0.4\n",
        "cooldown_iter_before_scale = 300\n",
        "timescale = 10\n",
        "num_repeats=1\n",
        "decay_style = \"inverse_sqrt_infinite\"\n",
        "\n",
        "iterations = list(range(1, end_iter))\n",
        "lrs = [get_lr(num_iters, warmup_iter, start_lr, end_iter, min_lr, decay_style, num_repeats=num_repeats, constant_lr=constant_lr, cooldown_iters_percent=cooldown_iters_percent, cooldown_iter_before_scale=cooldown_iter_before_scale, timescale=timescale, constant_iters_percent=constant_iters_percent, max_lr=0) for num_iters in iterations]\n",
        "\n",
        "# Plotting\n",
        "plt.plot(iterations, lrs)\n",
        "plt.title('Learning rate per iteration')\n",
        "plt.xlabel('Iteration')\n",
        "plt.ylabel('Learning rate')\n",
        "plt.grid(True)\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 472
        },
        "id": "r56RSQgnS9u-",
        "outputId": "c81fe470-f224-44eb-95d9-af2abb579e3e"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAlgAAAHHCAYAAABjvibXAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjcuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/bCgiHAAAACXBIWXMAAA9hAAAPYQGoP6dpAABoAklEQVR4nO3deVyU5d4G8GsGhhn2VTZFQKVcwA2SMDuWoliWYh2343E7vpqlpVF6skzLLMvU1LTMc3LrraNZvZ4WIxGXUglzzX1FUREQkV1gmLnfP8Z5ZGQdnWEe5Pp+Pn7UZ+555jc/SK7u5577UQghBIiIiIjIYpS2LoCIiIjofsOARURERGRhDFhEREREFsaARURERGRhDFhEREREFsaARURERGRhDFhEREREFsaARURERGRhDFhEREREFsaARUS1CgkJwZgxY2xdBt3hwoULUCgUWLNmja1LqbcdO3ZAoVBgx44dti6FyOoYsIgawJo1a6BQKLBv3z5bl9KklJSU4K233moyP9A3b96Mt956y9Zl4JNPPmlUwY/IGuxtXQARydupU6egVDbO/xcrKSnB22+/DQB47LHHbFuMhQUHB+PmzZtQqVTSsc2bN2P58uU2D1mffPIJfHx8qsx8/uUvf8HNmzfh4OBgm8KIGlDj/FeTiO5KRUUFysvLzXqOWq02+SFuS3dTf2NWXFxc42MKhQIajQZ2dnZWrUEIgZs3b1rkXEqlEhqNptEGdiJz8LucSEauXLmCf/zjH/Dz84NarUaHDh2watUqkzHl5eWYNWsWIiMj4e7uDmdnZzz66KPYvn27yTjjGp0FCxZg8eLFaN26NdRqNY4fP4633noLCoUCZ8+exZgxY+Dh4QF3d3eMHTsWJSUlJue5cw2W8XLn7t27kZCQgGbNmsHZ2RmDBg3CtWvXTJ6r1+vx1ltvITAwEE5OTnj88cdx/Pjxeq3rqq3++vTgwoULaNasGQDg7bffhkKhgEKhMJndOXnyJP7617/Cy8sLGo0GUVFR+P777+v6MpnU9tFHHyE4OBiOjo7o2bMnjh49WmV8fV7H2NedO3fihRdegK+vL1q0aFFnDcZLcWPGjMHy5csBQHqvCoVCGq/X67F48WJ06NABGo0Gfn5+eO6553Djxg2T84aEhOCpp57CL7/8gqioKDg6OuKzzz4DAKxevRq9evWCr68v1Go12rdvj08//bTK848dO4adO3dKNRhnD2tag7Vx40ZERkbC0dERPj4++Pvf/44rV66YjBkzZgxcXFxw5coVxMfHw8XFBc2aNcOrr74KnU5XY5+IbIWXCIlkIisrCw8//DAUCgUmT56MZs2a4eeff8a4ceNQUFCAqVOnAgAKCgrw73//G8OHD8f48eNRWFiIzz//HHFxcdi7dy86d+5sct7Vq1ejtLQUEyZMgFqthpeXl/TYkCFDEBoainnz5uHAgQP497//DV9fX3zwwQd11vviiy/C09MTs2fPxoULF7B48WJMnjwZGzZskMbMmDED8+fPx9NPP424uDgcPnwYcXFxKC0trXdfqqu/Pj1o1qwZPv30Uzz//PMYNGgQnnnmGQBAx44dAQDHjh3DI488gubNm+O1116Ds7Mzvv76a8THx+Pbb7/FoEGD6qxt3bp1KCwsxKRJk1BaWoolS5agV69eOHLkCPz8/O7qdV544QU0a9YMs2bNqnUG607PPfccMjIykJSUhC+++KLax9esWYOxY8fipZdeQlpaGpYtW4aDBw9i9+7dJrOUp06dwvDhw/Hcc89h/PjxePDBBwEAn376KTp06IABAwbA3t4eP/zwA1544QXo9XpMmjQJALB48WK8+OKLcHFxwRtvvAEAUi+qY6zpoYcewrx585CVlYUlS5Zg9+7dOHjwIDw8PKSxOp0OcXFxiI6OxoIFC7B161YsXLgQrVu3xvPPP1/vXhE1CEFEVrd69WoBQPzxxx81jhk3bpwICAgQOTk5JseHDRsm3N3dRUlJiRBCiIqKClFWVmYy5saNG8LPz0/84x//kI6lpaUJAMLNzU1kZ2ebjJ89e7YAYDJeCCEGDRokvL29TY4FBweL0aNHV3kvsbGxQq/XS8dffvllYWdnJ/Ly8oQQQmRmZgp7e3sRHx9vcr633npLADA5Z3Vqq7++Pbh27ZoAIGbPnl3l/L179xYRERGitLRUOqbX60X37t1FWFhYvWpzdHQUly9flo6npqYKAOLll182+3WMfe3Ro4eoqKio9fUr17B69Wrp2KRJk0R1/6z/9ttvAoD48ssvTY4nJiZWOR4cHCwAiMTExCrnMX4PVhYXFydatWplcqxDhw6iZ8+eVcZu375dABDbt28XQghRXl4ufH19RXh4uLh586Y07scffxQAxKxZs6Rjo0ePFgDEnDlzTM7ZpUsXERkZWeW1iGyNlwiJZEAIgW+//RZPP/00hBDIycmRfsXFxSE/Px8HDhwAANjZ2UmLhPV6PXJzc1FRUYGoqChpTGXPPvusdKnsThMnTjT5+6OPPorr16+joKCgzponTJhgcgnq0UcfhU6nw8WLFwEAycnJqKiowAsvvGDyvBdffLHOc9dVv7k9uFNubi62bduGIUOGoLCwUOr19evXERcXhzNnzlS5RFWd+Ph4NG/eXPp7t27dEB0djc2bN9/164wfP97i66o2btwId3d39OnTx+R7KzIyEi4uLlUuL4eGhiIuLq7KeRwdHaU/5+fnIycnBz179sT58+eRn59vdl379u1DdnY2XnjhBWg0Gul4//790bZtW/z0009VnlPd9+z58+fNfm0ia+MlQiIZuHbtGvLy8rBy5UqsXLmy2jHZ2dnSn9euXYuFCxfi5MmT0Gq10vHQ0NAqz6vumFHLli1N/u7p6QkAuHHjBtzc3GqtubbnApCCVps2bUzGeXl5SWPro6b6zenBnc6ePQshBN588028+eab1Y7Jzs42CU/VCQsLq3LsgQcewNdff33Xr1Of+s115swZ5Ofnw9fXt8YaKqupht27d2P27NlISUmpslYvPz8f7u7uZtVl/B4xXoKsrG3btti1a5fJMY1GUyVse3p6VllHRiQHDFhEMqDX6wEAf//73zF69OhqxxjXDv3v//4vxowZg/j4eEybNg2+vr6ws7PDvHnzcO7cuSrPqzzrcKeaZkqEEHXWfC/PNUd19ZvbgzsZ+/3qq69WO1MDVA2Gd+NuXqe2r9e91OHr64svv/yy2sfvDC3V1XDu3Dn07t0bbdu2xaJFixAUFAQHBwds3rwZH330kfRercnan5gksiQGLCIZaNasGVxdXaHT6RAbG1vr2G+++QatWrXCd999Z3KJbvbs2dYu0yzBwcEADLM4lWdErl+/fs8zDvXtQeXHKmvVqhUAQKVS1dnv2pw5c6bKsdOnTyMkJMSir1NfNb3f1q1bY+vWrXjkkUfuOsD98MMPKCsrw/fff28ye3nn5cXa6riT8Xvk1KlT6NWrl8ljp06dkh4naoy4BotIBuzs7PDss8/i22+/rfZj/pW3PzD+X3zlmaLU1FSkpKRYv1Az9O7dG/b29lU+xr9s2bJ7Pnd9e+Dk5AQAyMvLMznu6+uLxx57DJ999hmuXr1a5fx3bjdRk02bNpmsodq7dy9SU1PxxBNPWPR16svZ2RlA1fc7ZMgQ6HQ6vPPOO1WeU1FRUWV8darreX5+PlavXl1tHfU5Z1RUFHx9fbFixQqUlZVJx3/++WecOHEC/fv3r/McRHLFGSyiBrRq1SokJiZWOT5lyhS8//772L59O6KjozF+/Hi0b98eubm5OHDgALZu3Yrc3FwAwFNPPYXvvvsOgwYNQv/+/ZGWloYVK1agffv2KCoqaui3VCM/Pz9MmTIFCxcuxIABA9CvXz8cPnwYP//8M3x8fOo9y1Gd+vbA0dER7du3x4YNG/DAAw/Ay8sL4eHhCA8Px/Lly9GjRw9ERERg/PjxaNWqFbKyspCSkoLLly/j8OHDddbRpk0b9OjRA88//zzKysqwePFieHt7Y/r06dIYS7xOfUVGRgIAXnrpJcTFxcHOzg7Dhg1Dz5498dxzz2HevHk4dOgQ+vbtC5VKhTNnzmDjxo1YsmQJ/vrXv9Z67r59+8LBwQFPP/00nnvuORQVFeFf//oXfH19q4THyMhIfPrpp5g7dy7atGkDX1/fKjNUgGFm74MPPsDYsWPRs2dPDB8+XNqmISQkBC+//LLFekPU4Gz2+UWiJsT4Efyafl26dEkIIURWVpaYNGmSCAoKEiqVSvj7+4vevXuLlStXSufS6/XivffeE8HBwUKtVosuXbqIH3/8UYwePVoEBwdL44wf4//www+r1GPcpuHatWvV1pmWliYdq2mbhju3nLjzI/hCGLZTePPNN4W/v79wdHQUvXr1EidOnBDe3t5i4sSJtfastvrr2wMhhNizZ4+IjIwUDg4OVbZsOHfunBg1apTw9/cXKpVKNG/eXDz11FPim2++qXdtCxcuFEFBQUKtVotHH31UHD58uMr4+rxOfbbyqK6Gyts0VFRUiBdffFE0a9ZMKBSKKls2rFy5UkRGRgpHR0fh6uoqIiIixPTp00VGRoY0Jjg4WPTv37/a1/z+++9Fx44dhUajESEhIeKDDz4Qq1atqvI9k5mZKfr37y9cXV0FAGnLhuq+R4QQYsOGDaJLly5CrVYLLy8vMWLECJPtL4QwbNPg7OxcpSbj9zKR3CiEsPCKVCKiWuTl5cHT0xNz586VNqJsbC5cuIDQ0FB8+OGHePXVV21dDhHJENdgEZHVVHcPu8WLFwO4/26+TERUGddgEZHVbNiwAWvWrMGTTz4JFxcX7Nq1C//5z3/Qt29fPPLII7Yuj4jIahiwiMhqOnbsCHt7e8yfPx8FBQXSwve5c+faujQiIqviGiwiIiIiC+MaLCIiIiILY8AiIiIisjCuwbpLer0eGRkZcHV1vacNE4mIiKjhCCFQWFiIwMBAKJXWm2diwLpLGRkZCAoKsnUZREREdBcuXbqEFi1aWO38DFh3ydXVFYDhC+Tm5max82q1WmzZskW6lUVTxT7cxl4YsA8G7IMB+2DAPhiY04eCggIEBQVJP8ethQHrLhkvC7q5uVk8YDk5OcHNza3J/8fCPhiwFwbsgwH7YMA+GLAPBnfTB2sv7+EidyIiIiILY8AiIiIisjAGLCIiIiILY8AiIiIisjAGLCIiIiILY8AiIiIisjAGLCIiIiILY8AiIiIisjAGLCIiIiILY8AiIiIisjCbB6zly5cjJCQEGo0G0dHR2Lt3b63jN27ciLZt20Kj0SAiIgKbN282efy7775D37594e3tDYVCgUOHDlU5R2lpKSZNmgRvb2+4uLjg2WefRVZWliXfFhERETVhNg1YGzZsQEJCAmbPno0DBw6gU6dOiIuLQ3Z2drXj9+zZg+HDh2PcuHE4ePAg4uPjER8fj6NHj0pjiouL0aNHD3zwwQc1vu7LL7+MH374ARs3bsTOnTuRkZGBZ555xuLvj4iIiJommwasRYsWYfz48Rg7dizat2+PFStWwMnJCatWrap2/JIlS9CvXz9MmzYN7dq1wzvvvIOuXbti2bJl0piRI0di1qxZiI2NrfYc+fn5+Pzzz7Fo0SL06tULkZGRWL16Nfbs2YPff//dKu/zXpVqdRBC2LoMIiIiqid7W71weXk59u/fjxkzZkjHlEolYmNjkZKSUu1zUlJSkJCQYHIsLi4OmzZtqvfr7t+/H1qt1iSAtW3bFi1btkRKSgoefvjhap9XVlaGsrIy6e8FBQUADHfw1mq19X79uhjPZfz9enE5/rLgV8SEeuHfo7pa7HXk7s4+NGXshQH7YMA+GLAPBuyDgTl9aKhe2Sxg5eTkQKfTwc/Pz+S4n58fTp48We1zMjMzqx2fmZlZ79fNzMyEg4MDPDw8zDrPvHnz8Pbbb1c5vmXLFjg5OdX79esrKSkJALArU4HyCjvsPJNTZb1ZU2DsA7EXRuyDAftgwD4YsA8G9elDSUlJA1Riw4DV2MyYMcNk9qygoABBQUHo27cv3NzcLPY6Wq0WSUlJ6NOnD1QqFcoPZWBjmmGN2ZNPPmmx15G7O/vQlLEXBuyDAftgwD4YsA8G5vTBeAXK2mwWsHx8fGBnZ1fl03tZWVnw9/ev9jn+/v5mja/pHOXl5cjLyzOZxarrPGq1Gmq1uspxlUpllW9q43ndnG6/pg5KaFR2Fn8tObNWfxsj9sKAfTBgHwzYBwP2waA+fWioPtlskbuDgwMiIyORnJwsHdPr9UhOTkZMTEy1z4mJiTEZDximA2saX53IyEioVCqT85w6dQrp6elmnaehOFYKVPk3m/Y1diIiosbCppcIExISMHr0aERFRaFbt25YvHgxiouLMXbsWADAqFGj0Lx5c8ybNw8AMGXKFPTs2RMLFy5E//79sX79euzbtw8rV66Uzpmbm4v09HRkZGQAMIQnwDBz5e/vD3d3d4wbNw4JCQnw8vKCm5sbXnzxRcTExNS4wN2W9JU+PZh/Uws/N40NqyEiIqL6sGnAGjp0KK5du4ZZs2YhMzMTnTt3RmJiorSQPT09HUrl7Um27t2746uvvsLMmTPx+uuvIywsDJs2bUJ4eLg05vvvv5cCGgAMGzYMADB79my89dZbAICPPvoISqUSzz77LMrKyhAXF4dPPvmkAd6x+e4MWERERCR/Nl/kPnnyZEyePLnax3bs2FHl2ODBgzF48OAazzdmzBiMGTOm1tfUaDRYvnw5li9fbk6pNqHT3/5zfgkDFhERUWNg81vlUO10+tsJK48zWERERI0CA5bMmcxgMWARERE1CgxYMldRaQaLAYuIiKhxYMCSucqL3AsYsIiIiBoFBiyZq9DxU4RERESNDQOWzFWewcorKbdhJURERFRfDFgyx0XuREREjQ8DlszpuMidiIio0WHAkjmdvvIarAobVkJERET1xYAlcxUmAascotKaLCIiIpInBiyZq7zIXasTKCzjLBYREZHcMWDJXOUZLAC4XsRPEhIREckdA5bM6e8IWLnFZTaqhIiIiOqLAUvmKm/TAAA5nMEiIiKSPQYsmau8TQMA5BYzYBEREckdA5bM6cSda7B4iZCIiEjuGLBkrsoid85gERERyR4DlswZF7nbKxUA+ClCIiKixoABS+aMM1i+rmoAXINFRETUGDBgyZxxBquZmwYAkMM1WERERLLHgCVzxkXuxhksrsEiIiKSPwYsmdPdcYnwRnF5lc1HiYiISF4YsGTOGLCa3QpYFXqB/JtaW5ZEREREdWDAkjnjIndHlR08nVQAgKzCUluWRERERHVgwJI54+VAO6UCfrcWumfmM2ARERHJGQOWzFVUE7CyC/hJQiIiIjljwJI5vbgdsPyNM1gFnMEiIiKSMwYsmdOZzGAZFrpnMWARERHJGgOWzEkBS6GAn7thBosBi4iISN7sbV0A1a7yDJaPk3EGi2uwiIiI5IwzWDJXeZG7vzvXYBERETUGDFgyV3mRu1+l+xFW6PS2LIuIiIhqwYAlcxW62wHL29kB9koFhACu8abPREREssWAJXPSDJZCAaVSId2TkJuNEhERyRcDlsxVXuQOAIEejgCAK3k3bVYTERER1Y4BS+buDFgtPA0B6/INBiwiIiK5YsCSOZ24M2A5AQAu3yixWU1ERERUOwYsmau8yB3gDBYREVFjwIAlc5UXuQNAkJdxBosBi4iISK4YsGSuosY1WCUQt8IXERERyQsDlszp7whYAe6OUCiAUq0e14vLbVkaERER1YABS+buXOTuYK+E/60d3XmZkIiISJ4YsGROd8cid8D0MiERERHJDwOWzN05gwXc3qohPZcBi4iISI4YsGTuzo1GASDE2xkAcCGn2CY1ERERUe0YsGROCliK2wErtJkhYJ2/xoBFREQkRwxYMnfnNg0A0MrHELDSOINFREQkSwxYMnfnNg0AEHorYF0vLkd+idYmdREREVHNGLBkrrpF7s5qe/i5qQEA53OKbFIXERER1YwBS+aqW+QOAK18XABwHRYREZEcMWDJXE0BS1rozhksIiIi2WHAkjEhBG7lK5NPEQJc6E5ERCRnDFgyZpy9AgB7pemXqnUzwyXCs9mcwSIiIpIbBiwZq6gUsO7IV3jA3xWAYQ1WeYW+IcsiIiKiOjBgyZhe3A5Yd67BCnTXwFVjjwq9wLlrnMUiIiKSEwYsGat8ifDOgKVQKNDO3w0AcDKzoEHrIiIiotrZPGAtX74cISEh0Gg0iI6Oxt69e2sdv3HjRrRt2xYajQYRERHYvHmzyeNCCMyaNQsBAQFwdHREbGwszpw5YzLm9OnTGDhwIHx8fODm5oYePXpg+/btFn9v98okYN2xyB0A2gYYLhOevFrYYDURERFR3WwasDZs2ICEhATMnj0bBw4cQKdOnRAXF4fs7Oxqx+/ZswfDhw/HuHHjcPDgQcTHxyM+Ph5Hjx6VxsyfPx9Lly7FihUrkJqaCmdnZ8TFxaG0tFQa89RTT6GiogLbtm3D/v370alTJzz11FPIzMy0+ns2R20zWADQVprBYsAiIiKSE5sGrEWLFmH8+PEYO3Ys2rdvjxUrVsDJyQmrVq2qdvySJUvQr18/TJs2De3atcM777yDrl27YtmyZQAMs1eLFy/GzJkzMXDgQHTs2BHr1q1DRkYGNm3aBADIycnBmTNn8Nprr6Fjx44ICwvD+++/j5KSEpOgJgfGgKVUGC4J3unBWwvdeYmQiIhIXuxt9cLl5eXYv38/ZsyYIR1TKpWIjY1FSkpKtc9JSUlBQkKCybG4uDgpPKWlpSEzMxOxsbHS4+7u7oiOjkZKSgqGDRsGb29vPPjgg1i3bh26du0KtVqNzz77DL6+voiMjKyx3rKyMpSVlUl/LygwhBqtVgut1nL3AzSeS6vVorRcB8Awe1Xda7Ty1gAAsgrKkJVXDC9nB4vVYWuV+9DUsRcG7IMB+2DAPhiwDwbm9KGhemWzgJWTkwOdTgc/Pz+T435+fjh58mS1z8nMzKx2vPHSnvH32sYoFAps3boV8fHxcHV1hVKphK+vLxITE+Hp6VljvfPmzcPbb79d5fiWLVvg5ORUx7s1X1JSEq6XAoA9IPRV1poZ+ajtkFOmwNr/JuNBD1HtmMYsKSnJ1iXIBnthwD4YsA8G7IMB+2BQnz6UlJQ0QCU2DFi2IoTApEmT4Ovri99++w2Ojo7497//jaeffhp//PEHAgICqn3ejBkzTGbPCgoKEBQUhL59+8LNzc1i9Wm1WiQlJaFPnz7IKNQCB3fBwd4eTz4ZV+34LUV/4qcjmXBs/iCe7NnKYnXYWuU+qFQqW5djU+yFAftgwD4YsA8G7IOBOX0wXoGyNpsFLB8fH9jZ2SErK8vkeFZWFvz9/at9jr+/f63jjb9nZWWZBKWsrCx07twZALBt2zb8+OOPuHHjhhSMPvnkEyQlJWHt2rV47bXXqn1ttVoNtVpd5bhKpbLKN7VKpYJSadhA1E6pqPE1urT0xE9HMnEko/C+/I/LWv1tjNgLA/bBgH0wYB8M2AeD+vShofpks0XuDg4OiIyMRHJysnRMr9cjOTkZMTEx1T4nJibGZDxgmA40jg8NDYW/v7/JmIKCAqSmpkpjjFODyju2RlcqldDr5bUjek03eq6sU5AHAODw5bwGqIiIiIjqw6afIkxISMC//vUvrF27FidOnMDzzz+P4uJijB07FgAwatQok0XwU6ZMQWJiIhYuXIiTJ0/irbfewr59+zB58mQAhvVVU6dOxdy5c/H999/jyJEjGDVqFAIDAxEfHw/AENI8PT0xevRoHD58GKdPn8a0adOQlpaG/v37N3gPaqMTxoBV85epQ6Ab7JQKZBWUITO/tMZxRERE1HBsugZr6NChuHbtGmbNmoXMzEx07twZiYmJ0iL19PR0k5mm7t2746uvvsLMmTPx+uuvIywsDJs2bUJ4eLg0Zvr06SguLsaECROQl5eHHj16IDExERqN4RN3Pj4+SExMxBtvvIFevXpBq9WiQ4cO+O9//4tOnTo1bAPqUKEzBqyaxzg52CPM1wUnMwtx+HIe/N2rv7xKREREDcfmi9wnT54szUDdaceOHVWODR48GIMHD67xfAqFAnPmzMGcOXNqHBMVFYVffvnF7FobmvFehPa1zGABQOcgD0PAupSHuA4MWERERLZm81vlUM0qjBuN1vFVMq7D2n/xhpUrIiIiovpgwJIxvXGRezW7uFf2UIgXAODQpTyUVeisXhcRERHVjgFLxurzKUIAaN3MGT4uapRV6HH4Un5DlEZERES1YMCSsfoGLIVCgehQwyxW6vnrVq+LiIiIaseAJWP12abBKLrVrYCVlmvVmoiIiKhuDFgyVqGve5sGo+hQbwCGhe5anbw2TCUiImpqGLBkTFrkXo8ZrDBfF3g4qXBTq8Of3NWdiIjIphiwZEyawap9CRYAQKlU4JHWPgCAnadzrFkWERER1YEBS8b09VzkbtTzwWYAgJ2nsq1WExEREdWNAUvGbi9yr2fAesAQsP68ko/rRWVWq4uIiIhqx4AlY/XdpsHIz02DdgFuEAL47QwvExIREdkKA5aM6cxY5G702K3LhDt4mZCIiMhmGLBkzJxF7kbGy4Q7T19DBbdrICIisgkGLBkzZ5sGo6hgT3g6qXCjRIu93HSUiIjIJhiwZMycjUaN7O2UiOvgDwDYfPSqNcoiIiKiOjBgyZjezE8RGj0REQAASDyaJa3jIiIioobDgCVjd7PIHQBiWnnDTWOPnKIy7L94wxqlERERUS0YsGRMdxeL3AHAwV6JPu0Nlwl/+jPD0mURERFRHRiwZOxuZ7AA4KlOhsuEP/x5FeUV/DQhERFRQ2LAkrG7WeRu9GgbH/i6qpFbXI7t3BOLiIioQTFgydjdbNNgZG+nxKCuzQEA3+y/bNG6iIiIqHYMWDJ2LzNYAPDXri0AANtPZiOH9yYkIiJqMAxYMiZt06Awc5X7LWF+rujUwh0VeoH/O3DFkqURERFRLRiwZKziHi4RGg3r1hIA8MXvF7knFhERUQNhwJIx/T1eIgSA+M7N4e6oQnpuCW8ATURE1EAYsGTsXrZpMHJ0sMPQh4IAAGv2XLBEWURERFQHBiwZu9dF7kYjHw6GUgH8diYHZ7OLLFAZERER1YYBS8Zu34vw3r5MQV5OiG3nBwBY+eu5e66LiIiIaseAJWPSDNZdfoqwsomPtQYAfHfgCi7fKLnn8xEREVHNGLBkzBKL3I26tvTEI228UaEX+Gzn+Xs/IREREdWIAUvGLLFNQ2WTHw8DAGzYdwlZBaUWOScRERFVxYAlY5acwQKAh1t5ISrYE+UVeizbdtYyJyUiIqIqGLBkTGehRe5GCoUCr/R9EADw1d50nL/GTxQSERFZAwOWjN1e5G65c8a09kbvtr7Q6QU+/OWU5U5MREREEgYsGZMuEVrqGuEt/3yiLZQK4Oejmdh/Mdei5yYiIiIGLFmz5DYNlT3g54ohUYbd3Wd/f4z3KCQiIrIwBiwZs/Qi98pejXsQbhp7HL1SgC9SLlj+BYiIiJowBiwZs/Q2DZX5uKgxvV9bAMCCLae5bQMREZEFMWDJ2O1b5Vjn/H/r1hKdgzxQVFaBt384BiF4qZCIiMgSGLBkTGfFGSwAUCoVeHdQOOyVCmw+konvD2dY5XWIiIiaGgYsGbPWIvfKOgS648Vehh3e39x0FFfzb1rttYiIiJoKBiwZu73I3XoBCwAmPd4anYI8UFBagWkb/5Rel4iIiO4OA5aMVTRQwLK3U2LRkE7QqJTYdTYHn+48Z9XXIyIiut8xYMmYtRe5V9a6mQveHtABALBwyynsOpNj/RclIiK6TzFgyViFzrqL3O809KGWGBLVAnoBvLT+INdjERER3SUGLBmTZrCsuMj9TnMGhqN9gBtyi8sxYd1+lJRXNNhrExER3S8YsGRM10BrsCrTqOyw4u+R8HJ2wJEr+XjpPwd5Kx0iIiIzMWDJmC0CFgC09HbCv0ZFwsFeia0nsvHOj8cb9PWJiIgaOwYsGdMJ2wQsAIgM9sJHQzoDANbsuYDl2882eA1ERESN1V0FrIqKCmzduhWfffYZCgsLAQAZGRkoKiqyaHFN3e1F7g0fsACgf8cAzOzfDgDw4S+n8O/fztukDiIiosbG3twnXLx4Ef369UN6ejrKysrQp08fuLq64oMPPkBZWRlWrFhhjTqbJFsscr/T/zzaCkVlFVi89Qzm/nQCanslRsaE2KweIiKixsDsGawpU6YgKioKN27cgKOjo3R80KBBSE5OtmhxTV1DbTRalym9w/D8Y60BAG/+9xhW7UqzaT1ERERyZ/YM1m+//YY9e/bAwcHB5HhISAiuXLliscKo4W6VUxeFQoHpcQ+iQqfHv35Lw5wfjyOvpBwv93kAChvOrhEREcmV2TNYer0eOp2uyvHLly/D1dXVIkWRgS0Xud9JoVDg9Sfb4dW+DwAAlm47i9nfH+MWDkRERNUwO2D17dsXixcvlv6uUChQVFSE2bNn48knn7RkbU2ezsaL3O+kUCgwuVcY3hnYAQoFsC7lIsav24fCUq2tSyMiIpIVswPWwoULsXv3brRv3x6lpaX429/+Jl0e/OCDD6xRY5NlnMGyl0nAMhoZE4KPh3eB2l6JbSez8ddPU3Apt8TWZREREcmG2QGrRYsWOHz4MN544w28/PLL6NKlC95//30cPHgQvr6+ZhewfPlyhISEQKPRIDo6Gnv37q11/MaNG9G2bVtoNBpERERg8+bNJo8LITBr1iwEBATA0dERsbGxOHPmTJXz/PTTT4iOjoajoyM8PT0RHx9vdu3WZlzkrpRZwAKApzoGYsNzMfB1VeNUViEGLt+N3Wd5g2giIiLgLgLWr7/+CgAYMWIE5s+fj08++QT/8z//A5VKJT1WXxs2bEBCQgJmz56NAwcOoFOnToiLi0N2dna14/fs2YPhw4dj3LhxOHjwIOLj4xEfH4+jR49KY+bPn4+lS5dixYoVSE1NhbOzM+Li4lBaWiqN+fbbbzFy5EiMHTsWhw8fxu7du/G3v/3N3FZYnbTIXaYLyTsHeeC/kx9BeHPDvQv//nkqFm89zXVZRETU5JkdsB5//HHk5uZWOZ6fn4/HH3/crHMtWrQI48ePx9ixY9G+fXusWLECTk5OWLVqVbXjlyxZgn79+mHatGlo164d3nnnHXTt2hXLli0DYJi9Wrx4MWbOnImBAweiY8eOWLduHTIyMrBp0yYAhk1Sp0yZgg8//BATJ07EAw88gPbt22PIkCHmNaIByGWbhtoEuDti43PdMTQqCEIAi7eewcjPU5FdWFr3k4mIiO5TZm/TIISo9qP5169fh7Ozc73PU15ejv3792PGjBnSMaVSidjYWKSkpFT7nJSUFCQkJJgci4uLk8JTWloaMjMzERsbKz3u7u6O6OhopKSkYNiwYThw4ACuXLkCpVKJLl26IDMzE507d8aHH36I8PDwGustKytDWVmZ9PeCggIAgFarhVZruUXexnOVlZdLx/S6Cmi18r2rkb0CmDuwHR4KdsesH05gz7nriPvoV7z9dDs8Ee5/V+c09sGSvW2s2AsD9sGAfTBgHwzYBwNz+tBQvap3wHrmmWcAGD5JNmbMGKjVaukxnU6HP//8E927d6/3C+fk5ECn08HPz8/kuJ+fH06ePFntczIzM6sdn5mZKT1uPFbTmPPnDbd7eeutt7Bo0SKEhIRg4cKFeOyxx3D69Gl4eXlV+9rz5s3D22+/XeX4li1b4OTkVNfbNduWLVth/PJsS94KJ7OjcMNTAZjaHlh72g5XSrR4acOf6LL1EP4aqoeL6u7OmZSUZNEaGzP2woB9MGAfDNgHA/bBoD59KClpmA9l1fvHtru7OwDDDJarq6vJLu4ODg54+OGHMX78eMtXaGF6vR4A8MYbb+DZZ58FAKxevRotWrTAxo0b8dxzz1X7vBkzZpjMnhUUFCAoKAh9+/aFm5ubxerTarVISkrCY716AamGNW394vrCRd0IEtYtIyr0+GTneaz4NQ0HryuRXqrBzCcfRP8I/3pvTGrsQ58+faBS3WU6u0+wFwbsgwH7YMA+GLAPBub0wXgFytrq/VN79erVAAw7tr/66qtmXQ6sjo+PD+zs7JCVlWVyPCsrC/7+1V9W8vf3r3W88fesrCwEBASYjOncuTMASMfbt28vPa5Wq9GqVSukp6fXWK9arTaZtTNSqVRW+aZW2t3+0jiqHaBS2Vn8NaxFpQKm9WuHfuGBeGXjIZzOKsLLG4/gm4MZeHtAONr4uphxLuv0tzFiLwzYBwP2wYB9MGAfDOrTh4bqk9kLe2bPnn3P4QowzHpFRkaa3L9Qr9cjOTkZMTEx1T4nJiamyv0Ok5KSpPGhoaHw9/c3GVNQUIDU1FRpTGRkJNRqNU6dOiWN0Wq1uHDhAoKDg+/5fVlK5U/iKWX6KcK6RLRwxw8v9kBCnwegtldi99nreGLJr/gg8SSKyypsXR4REZHV3NV1p2+++QZff/010tPTUV5pMTYAHDhwoN7nSUhIwOjRoxEVFYVu3bph8eLFKC4uxtixYwEAo0aNQvPmzTFv3jwAhhtN9+zZEwsXLkT//v2xfv167Nu3DytXrgRgWB82depUzJ07F2FhYQgNDcWbb76JwMBAaZ8rNzc3TJw4EbNnz0ZQUBCCg4Px4YcfAgAGDx58N+2wCuMmo4C8P0VYF7W9HV7qHYb4zs3x1g/HsO1kNj7dcQ4b913GlNgwDHsoCCo7+S7gJyIiuhtm/2RbunQpxo4dCz8/Pxw8eBDdunWDt7c3zp8/jyeeeMKscw0dOhQLFizArFmz0LlzZxw6dAiJiYnSIvX09HRcvXpVGt+9e3d89dVXWLlyJTp16oRvvvkGmzZtMvn03/Tp0/Hiiy9iwoQJeOihh1BUVITExERoNBppzIcffohhw4Zh5MiReOihh3Dx4kVs27YNnp6e5rbDakxnsGxYiIW09HbC56OjsHJkJIK9nZBTVIY3Nx1F3Ee/IvHoVQjBvbOIiOj+YfYM1ieffIKVK1di+PDhWLNmDaZPn45WrVph1qxZ1e6PVZfJkydj8uTJ1T62Y8eOKscGDx5c60yTQqHAnDlzMGfOnBrHqFQqLFiwAAsWLDC73oaiq7QHVn0XhsudQqFA3w7+eOxBX3yVehFLt53F+ZxiTPzfA+jYwh0v9gpDbDvf++b9EhFR02X2DFZ6erq0HYOjoyMKCwsBACNHjsR//vMfy1bXhBknsBrz5cGaONgrMeaRUOyc9hhe7NUGjio7/Hk5H+PX7cOTS3dh85Gr0i72REREjZHZAcvf31+aqWrZsiV+//13AIZNPnmZx3Iqbm0nIdfb5FiCq0aFV/o+iF3/fBzPP9Yazg52OHG1AC98eQD9l+1BSpYCpVqdrcskIiIym9kBq1evXvj+++8BAGPHjsXLL7+MPn36YOjQoRg0aJDFC2yqbuUr2N+HM1h38nZR45/92mL3a73wUq82cNXY4+y1Yqw/b4e/LPgVi7ac4q13iIioUTF7DdbKlSulzTonTZoEb29v7NmzBwMGDKhxk04yn/E+hMomELCMPJwckND3QfzPX1rhy5QLWLn9FG6UaLF021l8uvMcnu4YiBEPt0TXlp5cp0VERLJmVsCqqKjAe++9h3/84x9o0aIFAGDYsGEYNmyYVYpryvSN4EbP1uKmUeF/eoTAL/847IO7Yu3vl7D/4g18d/AKvjt4BWG+Lhj6UBCe6doCXs4Oti6XiIioCrMuEdrb22P+/PmoqOAmkdZW0YQDlpGdAngi3B/fPt8dmyY9gr9GtoBGpcSZ7CLM/ekEHn4vGZO/OoAdp7JRodPbulwiIiKJ2ZcIe/fujZ07dyIkJMQK5ZCR/tYHBu7nRe7m6Bzkgc5BHpj1dHt8fygD6/9Ix9ErBfjxz6v48c+r8HZ2wFMdAzCgc3N0benBS4hERGRTZgesJ554Aq+99hqOHDmCyMjIKrfNGTBggMWKa8p0nMGqlptGhb8/HIy/PxyMo1fy8fW+S/jxz6u4XlyOtSkXsTblIoK8HDGwU3M8GRGAdgGuDFtERNTgzA5YL7zwAgBg0aJFVR5TKBTQ6fixektgwKpbeHN3hDd3x5tPtceuszn4/lAGfjmWiUu5N7Fs+1ks234WQV6O6NveH3Ed/BEZ7Ml+EhFRgzA7YBk/QUjWZbwXYVPYpuFeqeyUePxBXzz+oC9uluuw9UQWvj+cgV9PX8Ol3Jv4fFcaPt+VBm9nB8S280Of9n6Iae0NZ/Vd3YqTiIioTvwJI1O6JrhNgyU4Otjh6U6BeLpTIErKK/Dr6RxsOZaJ5JPZuF5cjg37LmHDvktwsFPioVBP9HygGXo+4IsH/Fx4KZGIiCyGAUumpEuE/KF/15wc7NEv3B/9wv2h1emxNy0XvxzLxPZT2biUexO7z17H7rPX8d7mk/B306DnA83w6AM+eLiVN3xc1LYun4iIGjEGLJniGizLUtkp8UgbHzzSxgdCCKTlFGPn6WvYefoaUs5dR2ZBqTS7BQBhvi6Iae2Nh1t5IzrUC94MXEREZAYGLJkyrsFiwLI8hUKBVs1c0KqZC8Y+EopSrQ5703Kx49Q17DmXg5OZhTiTXYQz2UVYl3IRAPCAnwsebuWNh0K80DXYE4HuGl5SJCKiGjFgyRRnsBqORmWHvzzQDH95oBkA4EZxOVLTcvH7+ev4/fx1nMwsxOmsIpzOuh24/NzUiAz2RNeWnujS0hPhzd2gtrez5dsgIiIZMTtgFRQUVHtcoVBArVbDwYG3LrEEBizb8XR2kNZuAUBucTn2pl3H7+dzsf/iDRy/WoCsgjJsPpKJzUcyAQAOdkqEN3dDl5aeiLi1fUQrH2d+SIGIqIkyO2B5eNS+S3aLFi0wZswYzJ49G0qlWXfioUoYsOTDy9kB/cID0C88AABQUl6BPy/n40D6DRy4mIcD6TeQW1yOA+l5OJCeJz3P2cEOHQINYSuihRvCA93RqpkLv6ZERE2A2QFrzZo1eOONNzBmzBh069YNALB3716sXbsWM2fOxLVr17BgwQKo1Wq8/vrrFi+4qeCnCOXLycEeD7cyLIAHACEELl4vwf6LN/Dn5TwcuZKP41cLUFyuw94Ludh7IbfSc+3QPsANbQNc8aC/G9r6u+JBf1e4aVS2ejtERGQFZgestWvXYuHChRgyZIh07Omnn0ZERAQ+++wzJCcno2XLlnj33XcZsO6BzpCvONvRCCgUCoT4OCPExxnPRrYAAFTo9Dh3rRhHr+TjyJV8HL2Sj2MZBSgp12HfxRvYd/GGyTmaezjiwVthyxi6Wvm4wMGes8BERI2R2QFrz549WLFiRZXjXbp0QUpKCgCgR48eSE9Pv/fqmjDdrR3zGbAaJ3s7pRSYjKFLpxc4f60IxzIKcDKzEKcyDb9fzS/FlbybuJJ3E9tOZkvnUNkpEOLtjFAfJ4h8JcoOZuCBAHe0aubMGS8iIpkzO2AFBQXh888/x/vvv29y/PPPP0dQUBAA4Pr16/D09LRMhU2U7tYdiRiw7h92SgXC/FwR5udqcjy/RItTWbcD16lbvwrLKqTtIgAlkr47Kj3H11WN1s1c0NrX2fB7MxeE+jgj0MOR3zNERDJgdsBasGABBg8ejJ9//hkPPfQQAGDfvn04efIkvvnmGwDAH3/8gaFDh1q20iZGz32wmgx3JxW6hXqhW6iXdEwIgYz8UpzLLsLpzHxs338COidvnM8pQXZhmfQr5fx1k3Op7BQI8nRCS28nBHs5oaW3M4K9nBDs7YQgLydoVNxKgoioIZgdsAYMGICTJ0/is88+w+nTpwEATzzxBDZt2oSQkBAAwPPPP2/RIpuiCi5yb9IUCgWaeziiuYcjYkI90OzGMTz55ENQqVQoKNXi/LVinMsuwvmcIpzLLsbZa0VIv16Ccp0e53OKcT6nuJpzAv5uGrS8FbiCvZ3RwtPwGs09HeHrqmGgJyKykLvaaDQ0NLTKJUKyLL0xYNnxBx6ZctOo0DnIA52DPEyO6/QCmQWluHi9GOnXS3AxtwQXrxfj4vUSpF8vQWFZBa7ml+JqfilS03KrnNdeqUCAh+ZWsHNCc09HtLgVvpp7OCLAQ8PNVImI6umuAlZeXh727t2L7Oxs6G8txjYaNWqURQpr6jiDReayU96e9ere2vQxIQRulGhxwRi+rpfgYm4xrtwwLK7PzC9FhV7gUu5NXMq9CaBqAFMogGYuajT3dESAuwZ+bhrpd383Dfxv/ZmXIYmI7iJg/fDDDxgxYgSKiorg5uZmsumoQqFgwLIQ4xose16yIQtQKBTwcnaAl7MDuras+gEUnV4gq+DWpxlvha7Lt36/cqMEV/JuolSrl9Z+HazltTydVFL48r8jgPm7a+Dl5AAXjT0cVXa8nyMR3bfMDlivvPIK/vGPf+C9996Dk5OTNWoiABW3NsLirVaoIdgpFQj0cESghyMeCqn6uBAC14vLceXGTWTk3URmQanhV77hV1aB4dJjWYUeN0q0uFGixcnMwnuuq0r+EnaY+vuW24/X+fzaR9T9/Doer+sM9/Zwja+v09nhn/u21vn6dddf1+vfW//u/f3XNkJAq7XD7EPbazyRrb8/6jrDvX99AAigtNQO7x3dWeUJtX1/qOwV+Ge/tniqY2Adr0J3y+yAdeXKFbz00ksMV1YmfYqQ/4dPMqBQKODjooaPixqd7lj7ZSSEQP5NrUnwyiy4Hb6MQSzvpha3vr3rVHWcAqh0rM7T1PeF7pq1z18TBbR3LM9omhQoqdDauggZUCBfW2b2s77Zf5kBy4rMDlhxcXHYt28fWrVqZY166BYdF7lTI6NQKODh5AAPJwe09XercZwQAje1OpSU62oZU/VYRYUWW7cmIza2N+zs6/inq47cU9vDdWUyUcfJa3t+XXFM1PHiQgAVFRXYvn07Hn/8cdjX1Yd61mWozYbv6y5et6KiAr/+uhOPPvoXqFQ196Hu923+a99+7j30zELnrqiowO7du/DIIz3q/f2w62wO3v/5ZK3/DdK9Mztg9e/fH9OmTcPx48cREREBlcp0R+kBAwZYrLimjPcipPuVQqGAk4M9nBzM++dHq1XCzQHwcVFX+XenKdFqtfDWAC08HZt8H045Am18XZp8Hy46Ax0C3erdh+zCUgDATQYsqzI7YI0fPx4AMGfOnCqPKRQK6HT8glmCjhuNEhGRFTjemvErKa+wcSX3N7MD1p3bMpB1SDNYDFhERGRBTg6GrVQ4g2VdSlsXQNUzBixu00BERJZkDFglWgYsa6rXDNbSpUsxYcIEaDQaLF26tNaxL730kkUKa+qMAYvbNBARkSU5GgMWZ7Csql4B66OPPsKIESOg0Wjw0Ucf1ThOoVAwYFnIrW2wuMidiIgsyvgBk/IKPXR6waUoVlKvgJWWllbtn8l69FyDRUREVmC8RAgYFrq7aprupzCtiWuwZKqCAYuIiKxAba+UdpG/yXVYVmP2pwh1Oh3WrFmD5OTkam/2vG3bNosV15TpuU0DERFZgUKhgJPKDsXlOn6S0IrMDlhTpkzBmjVr0L9/f4SHh/NmrVbCGSwiIrIWRwd7FJfXfkcFujdmB6z169fj66+/xpNPPmmNeugWPbdpICIiK3HiJwmtzuw1WA4ODmjTpo01aqFKjDNYSs4QEhGRhXGzUeszO2C98sorWLJkSZ0376R7w08REhGRtdzeC4u3y7EWsy8R7tq1C9u3b8fPP/+MDh06VLm55HfffWex4poy3ouQiIisRZrB4qcIrcbsgOXh4YFBgwZZoxaqhPciJCIia7l9w2cGLGsxK2BVVFTg8ccfR9++feHv72+tmggMWEREZD28XY71mbUGy97eHhMnTkRZWZm16qFbpIDFRe5ERGRhTirjIneuwbIWsxe5d+vWDQcPHrRGLVSJcQ2WvR0DFhERWRZnsKzP7DVYL7zwAl555RVcvnwZkZGRcHZ2Nnm8Y8eOFiuuKdNxmwYiIrIS7oNlfWYHrGHDhgEAXnrpJemYQqGAEAIKhQI6Hb9YlsA1WEREZC3cB8v6zA5YaWlp1qiD7nArXzFgERGRxTk6GH78c5sG6zE7YAUHB1ujDrpDxa2baHOROxERWRovEVqf2QHL6Pjx40hPT0d5ebnJ8QEDBtxzUQTcylew4yJ3IiKysNsbjfJThNZidsA6f/48Bg0ahCNHjkhrrwDDOiwAXINlIRXcpoGIiKzEUcUZLGsze5uGKVOmIDQ0FNnZ2XBycsKxY8fw66+/IioqCjt27LBCiU2T8V6E9lyDRUREFuZkXIPFgGU1Zs9gpaSkYNu2bfDx8YFSqYRSqUSPHj0wb948vPTSS9wjy0KMM1hKBiwiIrIw7oNlfWbPYOl0Ori6ugIAfHx8kJGRAcCw+P3UqVOWra4J0/Nmz0REZCVc5G59Zges8PBwHD58GAAQHR2N+fPnY/fu3ZgzZw5atWp1V0UsX74cISEh0Gg0iI6Oxt69e2sdv3HjRrRt2xYajQYRERHYvHmzyeNCCMyaNQsBAQFwdHREbGwszpw5U+25ysrK0LlzZygUChw6dOiu6rcG7oNFRETW4uxgvNkzF7lbi9kBa+bMmdDf+ojbnDlzkJaWhkcffRSbN2/G0qVLzS5gw4YNSEhIwOzZs3HgwAF06tQJcXFxyM7Ornb8nj17MHz4cIwbNw4HDx5EfHw84uPjcfToUWnM/PnzsXTpUqxYsQKpqalwdnZGXFwcSktLq5xv+vTpCAwMNLtua+O9CImIyFqc1bdnsIxrfsmyzA5YcXFxeOaZZwAAbdq0wcmTJ5GTk4Ps7Gz06tXL7AIWLVqE8ePHY+zYsWjfvj1WrFgBJycnrFq1qtrxS5YsQb9+/TBt2jS0a9cO77zzDrp27Yply5YBMMxeLV68GDNnzsTAgQPRsWNHrFu3DhkZGdi0aZPJuX7++Wds2bIFCxYsMLtua9PxEiEREVmJs/r2EuxizmJZhdkBy+js2bP45ZdfcPPmTXh5ed3VOcrLy7F//37ExsbeLkipRGxsLFJSUqp9TkpKisl4wBD6jOPT0tKQmZlpMsbd3R3R0dEm58zKysL48ePxxRdfwMnJ6a7qtyZeIiQiImtR2yulT6kXl3EdljWY/SnC69evY8iQIdi+fTsUCgXOnDmDVq1aYdy4cfD09MTChQvrfa6cnBzodDr4+fmZHPfz88PJkyerfU5mZma14zMzM6XHjcdqGiOEwJgxYzBx4kRERUXhwoULddZaVlaGsrIy6e8FBQUAAK1WC61WW+fz68t4LmPAEnqdRc/fWBjfc1N873diLwzYBwP2wYB9MLiXPjir7ZB/swJ5xaXwdrKzdGkNypw+NNT3jNkB6+WXX4ZKpUJ6ejratWsnHR86dCgSEhLMCli28vHHH6OwsBAzZsyo93PmzZuHt99+u8rxLVu2WGUGrLSsHIACu377FWccLX76RiMpKcnWJcgGe2HAPhiwDwbsg8Hd9EGpswOgwJZtOxHsavmabKE+fSgpKWmASu4iYG3ZsgW//PILWrRoYXI8LCwMFy9eNOtcPj4+sLOzQ1ZWlsnxrKws+Pv7V/scf3//Wscbf8/KykJAQIDJmM6dOwMAtm3bhpSUFKjVapPzREVFYcSIEVi7dm2V150xYwYSEhKkvxcUFCAoKAh9+/aFm5tbPd9x3bRaLZKSkqC0swd0OvR67DEEe8vvEqa1GfvQp08fqFQqW5djU+yFAftgwD4YsA8G99KH5ef24EZ2ETpFRaN7a28rVdgwzOmD8QqUtZkdsIqLi6udscnNza0SWOri4OCAyMhIJCcnIz4+HgCg1+uRnJyMyZMnV/ucmJgYJCcnY+rUqdKxpKQkxMTEAABCQ0Ph7++P5ORkKVAVFBQgNTUVzz//PABg6dKlmDt3rvT8jIwMxMXFYcOGDYiOjq72ddVqdbXvT6VSWeU/buMid7WDdc7fWFirv40Re2HAPhiwDwbsg8Hd9MFFY4gApTrcNz2sTx8a6r2aHbAeffRRrFu3Du+88w4Awz0I9Xo95s+fj8cff9zsAhISEjB69GhERUWhW7duWLx4MYqLizF27FgAwKhRo9C8eXPMmzcPgOFWPT179sTChQvRv39/rF+/Hvv27cPKlSuleqZOnYq5c+ciLCwMoaGhePPNNxEYGCiFuJYtW5rU4OLiAgBo3bp1lZk5W+EidyIisibjJwmLy/gpQmswO2DNnz8fvXv3xr59+1BeXo7p06fj2LFjyM3Nxe7du80uYOjQobh27RpmzZqFzMxMdO7cGYmJidIi9fT0dCiVtz/s2L17d3z11VeYOXMmXn/9dYSFhWHTpk0IDw+XxkyfPh3FxcWYMGEC8vLy0KNHDyQmJkKj0Zhdn60YtyXhvQiJiMgaXG4FrCIGLKswO2CFh4fj9OnTWLZsGVxdXVFUVIRnnnkGkyZNMlnzZI7JkyfXeEmwuhtIDx48GIMHD67xfAqFAnPmzMGcOXPq9fohISEQQj4brQlxewaL9yIkIiJrcGbAsiqzAxZg2FfqjTfeMDl2+fJlTJgwQbpUR3evctTjDBYREVmDCy8RWtVdbzR6p+vXr+Pzzz+31OmatMp3LeAMFhERWcPtgMWNRq3BYgGLLKdywOK9CImIyBqMlwgLSzmDZQ0MWDJkErA4g0VERFbgcuuGz7xEaB0MWDKkr/RnBiwiIrIGaZsG3uzZKuq9yP2ZZ56p9fG8vLx7rYVu4SVCIiKyNm7TYF31Dlju7u51Pj5q1Kh7LohuByyFgovciYjIOqSAxTVYVlHvgLV69Wpr1kGVcJNRIiKyNu7kbl1cgyVDxjVYSl4eJCIiK+FGo9bFgCVDxhksLnAnIiJrkfbBKtfJ6m4m9wsGLBliwCIiImtz0RgClk4vUKrV1zGazMWAJUMMWEREZG1OKjvpz7xMaHkMWDJk/P8ILnInIiJrUSoVcHbgZqPWwoAlQ8YZLC5yJyIiazJeJuTtciyPAUuGBLdpICKiBuCmUQEACku1Nq7k/sOAJUM64wwWAxYREVmRm6MhYBUwYFkcA5YMGT8sy0XuRERkTW63LhEW3OQlQktjwJIhHT9FSEREDYAzWNbDgCVDQhiCFW/0TERE1mRcg1VwkwHL0hiwZIj7YBERUUNwc7x1iZCfIrQ4BiwZYsAiIqKGwBks62HAkiFuNEpERA2Ba7CshwFLhvTcpoGIiBrA7RksXiK0NAYsGZIuEXKROxERWdHtNVicwbI0BiwZ4hosIiJqCFyDZT0MWDJkXIPFgEVERNZ0ew0WLxFaGgOWDHEGi4iIGoJxJ/eisgpU6PR1jCZzMGDJEAMWERE1BNdblwgBQ8giy2HAkiFjwOI2DUREZE0O9ko4quwA8JOElsaAJUPGSVolP0VIRERWxk8SWgcDlgzxEiERETUU6ZOEDFgWxYAlQwxYRETUUKRPEvISoUUxYMkQAxYRETUU4ycJOYNlWQxYMnQrXzFgERGR1RlnsPJLGLAsiQFLhnS8VQ4RETUQTycHAMCNknIbV3J/YcCSIWHcpsGOAYuIiKzrdsDiDJYlMWDJkHEGi9s0EBGRtXk6Gy4R3ijmDJYlMWDJkBCGYMU1WEREZG28RGgdDFgypLv1OwMWERFZGwOWdTBgyZDgInciImogHk63LhFyDZZFMWDJkLQPFhe5ExGRlXk6G2aw8krKIYz/h0/3jAFLhvScwSIiogbidesSoVYnUFTG3dwthQFLhow3e7bnGiwiIrIyRwc7qO0NcSCPlwkthgFLhowzWEoGLCIiagBc6G55DFgyxEuERETUkIzrsHK5F5bFMGDJEBe5ExFRQ/K89UlCXiK0HAYsGTKuweIMFhERNQReIrQ8BiwZkmawuAaLiIgaAG+XY3kMWDLEgEVERA2JN3y2PAYsGTLu88ZtGoiIqCF43ApYubxEaDEMWDKk4zYNRETUgHxcDAHrelGZjSu5fzBgyZDxRgVc5E5ERA2hmYsaAHCtkAHLUhiwZEjHNVhERNSAfFwNASuniJcILYUBS4YEAxYRETUg4wxW/k0tyip0Nq7m/sCAJUPSPlgMWERE1ADcHVXSB6uucxbLIhiwZEgvDN/kDFhERNQQlEoFfFyMlwm5DssSGLBkSM9tGoiIqIH5uBo+SciAZRmyCFjLly9HSEgINBoNoqOjsXfv3lrHb9y4EW3btoVGo0FERAQ2b95s8rgQArNmzUJAQAAcHR0RGxuLM2fOSI9fuHAB48aNQ2hoKBwdHdG6dWvMnj0b5eXymBY1BiwlP0VIREQNhJ8ktCybB6wNGzYgISEBs2fPxoEDB9CpUyfExcUhOzu72vF79uzB8OHDMW7cOBw8eBDx8fGIj4/H0aNHpTHz58/H0qVLsWLFCqSmpsLZ2RlxcXEoLS0FAJw8eRJ6vR6fffYZjh07ho8++ggrVqzA66+/3iDvuS7cyZ2IiBra7UuE8phsaOxsHrAWLVqE8ePHY+zYsWjfvj1WrFgBJycnrFq1qtrxS5YsQb9+/TBt2jS0a9cO77zzDrp27Yply5YBMMxeLV68GDNnzsTAgQPRsWNHrFu3DhkZGdi0aRMAoF+/fli9ejX69u2LVq1aYcCAAXj11Vfx3XffNdTbrhUXuRMRUUMzbtXAGSzLsLfli5eXl2P//v2YMWOGdEypVCI2NhYpKSnVPiclJQUJCQkmx+Li4qTwlJaWhszMTMTGxkqPu7u7Izo6GikpKRg2bFi1583Pz4eXl1eNtZaVlaGs7PY3XUFBAQBAq9VCq7XcvZu0Wq00gyX0eoueuzExvu+m+v4rYy8M2AcD9sGAfTCwZB+8nAyRILvgZqPrqzl9aKj3ZtOAlZOTA51OBz8/P5Pjfn5+OHnyZLXPyczMrHZ8Zmam9LjxWE1j7nT27Fl8/PHHWLBgQY21zps3D2+//XaV41u2bIGTk1ONz7sbemEHANj/x14UnBZ1jL6/JSUl2boE2WAvDNgHA/bBgH0wsEQfLucoANjhdHomNm++cu9F2UB9+lBSUtIAldg4YMnBlStX0K9fPwwePBjjx4+vcdyMGTNMZs4KCgoQFBSEvn37ws3NzWL1aLVavHtwGwAgJiYa3UJqnlW7n2m1WiQlJaFPnz5QqVS2Lsem2AsD9sGAfTBgHwws2QfvtFysPbMPegcXPPnkIxaqsGGY0wfjFShrs2nA8vHxgZ2dHbKyskyOZ2Vlwd/fv9rn+Pv71zre+HtWVhYCAgJMxnTu3NnkeRkZGXj88cfRvXt3rFy5stZa1Wo11Gp1leMqlcri/3Eb12BpHCx/7sbGGv1trNgLA/bBgH0wYB8MLNEHfw9nAIY1WI21p/XpQ0O9N5sucndwcEBkZCSSk5OlY3q9HsnJyYiJian2OTExMSbjAcOUoHF8aGgo/P39TcYUFBQgNTXV5JxXrlzBY489hsjISKxevRpKpc3X+0u4TQMRETU0f3cNAKCwrAJFZRU2rqbxs/klwoSEBIwePRpRUVHo1q0bFi9ejOLiYowdOxYAMGrUKDRv3hzz5s0DAEyZMgU9e/bEwoUL0b9/f6xfvx779u2TZqAUCgWmTp2KuXPnIiwsDKGhoXjzzTcRGBiI+Ph4ALfDVXBwMBYsWIBr165J9dQ0c9aQuE0DERE1NBe1PVw19igsrUBm/k208XW1dUmNms0D1tChQ3Ht2jXMmjULmZmZ6Ny5MxITE6VF6unp6SazS927d8dXX32FmTNn4vXXX0dYWBg2bdqE8PBwacz06dNRXFyMCRMmIC8vDz169EBiYiI0GkM6T0pKwtmzZ3H27Fm0aNHCpB4hbL+onAGLiIhsIdDdEadKC3E1v5QB6x7ZPGABwOTJkzF58uRqH9uxY0eVY4MHD8bgwYNrPJ9CocCcOXMwZ86cah8fM2YMxowZczelNgjug0VERLbg767BqSxDwKJ7I5+FRyQRvBchERHZQMCtdVhX8xiw7hUDlgxxkTsREdlCgLsjACCz4KaNK2n8GLBkSC/NYPHLQ0REDcc4g5XBGax7xp/gMiTNYPGrQ0REDci4VUMm12DdM/4IlyHjInfOYBERUUMK9Li1BiuflwjvFX+CyxBnsIiIyBb8b63BKiitQDE3G70n/BEuM0IICBgWt9txkTsRETUgF7U9XNWGHZw4i3VvGLBkRqe/vdEpLxESEVFDa+HlBAC4lMuAdS/4E1xmKgcs5isiImpowbcC1sXrxTaupHHjj3CZ0QnOYBERke209DYErHTOYN0T/gSXGc5gERGRLQV5GQNWiY0radz4I1xmdPrbf+YMFhERNbRgaQ0WA9a94E9wmdHpbycs3oqQiIgaWstKM1ii0rIVMg8DlszopPsQAgpu00BERA0s0MMRSgVwU6vDtaIyW5fTaDFgyYxxDZYdp6+IiMgGHOyV0k2feZnw7jFgyQwDFhER2Vqwt3GrBgasu8WAJTPGbRoYsIiIyFZCfJwBAOevcS+su8WAJTO6W4uweJscIiKylTbNXAAAZ7ILbVxJ48WAJTOcwSIiIlsL8zMErLPZRTaupPFiwJIZrsEiIiJba+NrCFgXrpegvEJfx2iqDgOWzEgBi5cIiYjIRvzdNHBR20OnF7wn4V1iwJIZPS8REhGRjSkUCrT2Na7D4mXCu8GAJTMVt2awlAxYRERkQ2G+XId1LxiwZEZ/K2DZM2AREZENGQPWqSx+kvBuMGDJjDSDxTVYRERkQ+0C3AAAxzMKbFxJ48SAJTPGNVicwSIiIlsKb+4OAEjLKUZhqdbG1TQ+DFgywzVYREQkB17ODmjuYbgnIWexzMeAJTN6aR8sGxdCRERNXodAw2XCI1fybVxJ48Mf4zJz60453KaBiIhszniZ8BhnsMzGgCUzvBchERHJRcStgMUZLPMxYMkM70VIRERyEdHCELDOXStCfgkXupuDAUtmeC9CIiKSCx8XNVr5OEMIYN/FXFuX06gwYMkMAxYREcnJQyFeAIC9FxiwzMGAJTO82TMREcnJQ6GGgPVHGgOWORiwZMa4Bov7YBERkRx0uzWDdeRKPkq1OhtX03gwYMkM70VIRERyEuTlCH83DbQ6gVTOYtUbA5bM8F6EREQkJwqFAj0faAYA2HEq28bVNB4MWDLDexESEZHcPN7WELB2nrpm40oaDwYsmeG9CImISG4eaeMDe6UC53OKcSGn2NblNAoMWDLDNVhERCQ3rhqVtF3D1hNZNq6mcWDAkhnOYBERkRz1C/cHAPzw51UbV9I4MGDJjJ77YBERkQw9GREAO6UChy/l8TJhPTBgyUwFd3InIiIZauaqRvfW3gCA/x7KsHE18seAJTN66WbPNi6EiIjoDvGdmwMAvt53SbrzCFWPP8ZlRqc3/G6n5JeGiIjkpX/HAHg4qXAl7yYXu9eBP8Vl5va9CG1cCBER0R00KjsMe6glAGDtngu2LUbmGLBkRie4BouIiORrZEww7JQK7Dl3Hfsv3rB1ObLFgCUzOi5yJyIiGWvu4Yi/dm0BAFjwyykbVyNfDFgyo+M2DUREJHMvxYbBwU6JlPPXse0k12JVhwFLZjiDRUREctfcwxFjHgkBALy56RiKyypsW5AMMWDJjHENFndyJyIiOZsaG4YWno64kncT720+YetyZIcBS2Z4L0IiImoMnBzsMe+ZCCgUwJep6fh2/2VblyQrDFgyI92LkGuwiIhI5h4Na4YpvcMAADO+O4Kdp6/ZuCL5YMCSGWkGixthERFRI/BSrzA8GeGPcp0eE9btw9bjXPQOMGDJDmewiIioMVEqFVg8tAti2/mirEKP8V/sw9LkM6gw3pqkiWLAkhk9NxolIqJGxsFeiU//HomRDwdDCGBR0mkMXL4be9NybV2azcgiYC1fvhwhISHQaDSIjo7G3r17ax2/ceNGtG3bFhqNBhEREdi8ebPJ40IIzJo1CwEBAXB0dERsbCzOnDljMiY3NxcjRoyAm5sbPDw8MG7cOBQVFVn8vZmrgts0EBFRI6SyU+Kd+HAsGtIJ7o4qHMsowJDPUjBsZQp+/DMDpVqdrUtsUDYPWBs2bEBCQgJmz56NAwcOoFOnToiLi0N2dna14/fs2YPhw4dj3LhxOHjwIOLj4xEfH4+jR49KY+bPn4+lS5dixYoVSE1NhbOzM+Li4lBaWiqNGTFiBI4dO4akpCT8+OOP+PXXXzFhwgSrv9+66LnRKBERNWLPdG2BrQk98bfollDZKfD7+VxM/uogHnp3KyZ+sR9f/H4RR6/k3/eBy97WBSxatAjjx4/H2LFjAQArVqzATz/9hFWrVuG1116rMn7JkiXo168fpk2bBgB45513kJSUhGXLlmHFihUQQmDx4sWYOXMmBg4cCABYt24d/Pz8sGnTJgwbNgwnTpxAYmIi/vjjD0RFRQEAPv74Yzz55JNYsGABAgMDG+jdV8V7ERIRUWPXzFWN9wZFYNLjbfCf1HR8d+AyMvJLkXgsE4nHMgEYfs6FeDuhhacTAj008HdzRDNXNdwdVXB3VMHDSYUQH2e4qG0eVe6KTasuLy/H/v37MWPGDOmYUqlEbGwsUlJSqn1OSkoKEhISTI7FxcVh06ZNAIC0tDRkZmYiNjZWetzd3R3R0dFISUnBsGHDkJKSAg8PDylcAUBsbCyUSiVSU1MxaNCgKq9bVlaGsrIy6e8FBQUAAK1WC61Wa/6br4G24taiQKGz6HkbG+N7b8o9MGIvDNgHA/bBgH0wkHsffJ3tMaVXK7z4WCj+vJKPPedykZqWi+NXC5F3U4tz14px7lpxjc//7O9d0OvBZnW+jjl9aKhe2TRg5eTkQKfTwc/Pz+S4n58fTp48We1zMjMzqx2fmZkpPW48VtsYX19fk8ft7e3h5eUljbnTvHnz8Pbbb1c5vmXLFjg5OdX0Fs2Wl6uESqHAiWPHsDn7aN1PuM8lJSXZugTZYC8M2AcD9sGAfTBoLH0IARDiBwhfoEALXC1R4EYZkFdu+L24AiipUKCkArhZAZw89AdKz9X//PXpQ0lJyV3Xb47GOe9mAzNmzDCZOSsoKEBQUBD69u0LNzc3i71Onz5aJCUloU+fPlCpVBY7b2Oj1bIPRuyFAftgwD4YsA8G7IOBOX0wXoGyNpsGLB8fH9jZ2SEry3RTsqysLPj7+1f7HH9//1rHG3/PyspCQECAyZjOnTtLY+5cRF9RUYHc3NwaX1etVkOtVlc5rlKprPJNba3zNjbsw23shQH7YMA+GLAPBuyDQX360FB9sumnCB0cHBAZGYnk5GTpmF6vR3JyMmJiYqp9TkxMjMl4wDAlaBwfGhoKf39/kzEFBQVITU2VxsTExCAvLw/79++Xxmzbtg16vR7R0dEWe39ERETUNNn8EmFCQgJGjx6NqKgodOvWDYsXL0ZxcbH0qcJRo0ahefPmmDdvHgBgypQp6NmzJxYuXIj+/ftj/fr12LdvH1auXAkAUCgUmDp1KubOnYuwsDCEhobizTffRGBgIOLj4wEA7dq1Q79+/TB+/HisWLECWq0WkydPxrBhw2z6CUIiIiK6P9g8YA0dOhTXrl3DrFmzkJmZic6dOyMxMVFapJ6eng6l8vZEW/fu3fHVV19h5syZeP311xEWFoZNmzYhPDxcGjN9+nQUFxdjwoQJyMvLQ48ePZCYmAiNRiON+fLLLzF58mT07t0bSqUSzz77LJYuXdpwb5yIiIjuWzYPWAAwefJkTJ48udrHduzYUeXY4MGDMXjw4BrPp1AoMGfOHMyZM6fGMV5eXvjqq6/MrpWIiIioLjbfyZ2IiIjofsOARURERGRhDFhEREREFsaARURERGRhDFhEREREFsaARURERGRhDFhEREREFsaARURERGRhDFhEREREFiaLndwbIyEEAMONpC1Jq9WipKQEBQUFTfrO6OzDbeyFAftgwD4YsA8G7IOBOX0w/tw2/hy3Fgasu1RYWAgACAoKsnElREREZK7CwkK4u7tb7fwKYe0Id5/S6/XIyMiAq6srFAqFxc5bUFCAoKAgXLp0CW5ubhY7b2PDPtzGXhiwDwbsgwH7YMA+GJjTByEECgsLERgYCKXSeiulOIN1l5RKJVq0aGG187u5uTXp/1iM2Ifb2AsD9sGAfTBgHwzYB4P69sGaM1dGXOROREREZGEMWEREREQWxoAlM2q1GrNnz4ZarbZ1KTbFPtzGXhiwDwbsgwH7YMA+GMixD1zkTkRERGRhnMEiIiIisjAGLCIiIiILY8AiIiIisjAGLCIiIiILY8CSmeXLlyMkJAQajQbR0dHYu3evrUuqt3nz5uGhhx6Cq6srfH19ER8fj1OnTpmMKS0txaRJk+Dt7Q0XFxc8++yzyMrKMhmTnp6O/v37w8nJCb6+vpg2bRoqKipMxuzYsQNdu3aFWq1GmzZtsGbNmir1yKGX77//PhQKBaZOnSoda0o9uHLlCv7+97/D29sbjo6OiIiIwL59+6THhRCYNWsWAgIC4OjoiNjYWJw5c8bkHLm5uRgxYgTc3Nzg4eGBcePGoaioyGTMn3/+iUcffRQajQZBQUGYP39+lVo2btyItm3bQqPRICIiAps3b7bOm76DTqfDm2++idDQUDg6OqJ169Z45513TO6Ddj/24ddff8XTTz+NwMBAKBQKbNq0yeRxOb3n+tRijT5otVr885//REREBJydnREYGIhRo0YhIyPjvutDXb2408SJE6FQKLB48WKT442qF4JkY/369cLBwUGsWrVKHDt2TIwfP154eHiIrKwsW5dWL3FxcWL16tXi6NGj4tChQ+LJJ58ULVu2FEVFRdKYiRMniqCgIJGcnCz27dsnHn74YdG9e3fp8YqKChEeHi5iY2PFwYMHxebNm4WPj4+YMWOGNOb8+fPCyclJJCQkiOPHj4uPP/5Y2NnZicTERGmMHHq5d+9eERISIjp27CimTJkiHW8qPcjNzRXBwcFizJgxIjU1VZw/f1788ssv4uzZs9KY999/X7i7u4tNmzaJw4cPiwEDBojQ0FBx8+ZNaUy/fv1Ep06dxO+//y5+++030aZNGzF8+HDp8fz8fOHn5ydGjBghjh49Kv7zn/8IR0dH8dlnn0ljdu/eLezs7MT8+fPF8ePHxcyZM4VKpRJHjhyxeh/effdd4e3tLX788UeRlpYmNm7cKFxcXMSSJUvu6z5s3rxZvPHGG+K7774TAMT//d//mTwup/dcn1qs0Ye8vDwRGxsrNmzYIE6ePClSUlJEt27dRGRkpMk57oc+1NWLyr777jvRqVMnERgYKD766KNG2wsGLBnp1q2bmDRpkvR3nU4nAgMDxbx582xY1d3Lzs4WAMTOnTuFEIZ/TFQqldi4caM05sSJEwKASElJEUIY/gNUKpUiMzNTGvPpp58KNzc3UVZWJoQQYvr06aJDhw4mrzV06FARFxcn/d3WvSwsLBRhYWEiKSlJ9OzZUwpYTakH//znP0WPHj1qfFyv1wt/f3/x4YcfSsfy8vKEWq0W//nPf4QQQhw/flwAEH/88Yc05ueffxYKhUJcuXJFCCHEJ598Ijw9PaXeGF/7wQcflP4+ZMgQ0b9/f5PXj46OFs8999y9vcl66N+/v/jHP/5hcuyZZ54RI0aMEEI0jT7c+cNUTu+5PrVYSm2hwmjv3r0CgLh48aIQ4v7sgxA19+Ly5cuiefPm4ujRoyI4ONgkYDW2XvASoUyUl5dj//79iI2NlY4plUrExsYiJSXFhpXdvfz8fACAl5cXAGD//v3QarUm77Ft27Zo2bKl9B5TUlIQEREBPz8/aUxcXBwKCgpw7NgxaUzlcxjHGM8hh15OmjQJ/fv3r1JnU+rB999/j6ioKAwePBi+vr7o0qUL/vWvf0mPp6WlITMz06RGd3d3REdHm/TCw8MDUVFR0pjY2FgolUqkpqZKY/7yl7/AwcFBGhMXF4dTp07hxo0b0pja+mVN3bt3R3JyMk6fPg0AOHz4MHbt2oUnnngCQNPpQ2Vyes/1qaUh5efnQ6FQwMPDA0DT6oNer8fIkSMxbdo0dOjQocrjja0XDFgykZOTA51OZ/JDFQD8/PyQmZlpo6runl6vx9SpU/HII48gPDwcAJCZmQkHBwfpHw6jyu8xMzOz2h4YH6ttTEFBAW7evGnzXq5fvx4HDhzAvHnzqjzWVHoAAOfPn8enn36KsLAw/PLLL3j++efx0ksvYe3atSbvpbYaMzMz4evra/K4vb09vLy8LNKvhujFa6+9hmHDhqFt27ZQqVTo0qULpk6dihEjRpjUeL/3oTI5vef61NJQSktL8c9//hPDhw+XbljclPrwwQcfwN7eHi+99FK1jze2XtjXeySRGSZNmoSjR49i165dti6lQV26dAlTpkxBUlISNBqNrcuxKb1ej6ioKLz33nsAgC5duuDo0aNYsWIFRo8ebePqGs7XX3+NL7/8El999RU6dOiAQ4cOYerUqQgMDGxSfaDaabVaDBkyBEIIfPrpp7Yup8Ht378fS5YswYEDB6BQKGxdjkVwBksmfHx8YGdnV+XTZFlZWfD397dRVXdn8uTJ+PHHH7F9+3a0aNFCOu7v74/y8nLk5eWZjK/8Hv39/avtgfGx2sa4ubnB0dHRpr3cv38/srOz0bVrV9jb28Pe3h47d+7E0qVLYW9vDz8/v/u+B0YBAQFo3769ybF27dohPT0dwO33UluN/v7+yM7ONnm8oqICubm5FulXQ/Ri2rRp0ixWREQERo4ciZdfflma4WwqfahMTu+5PrVYmzFcXbx4EUlJSdLslbG+ptCH3377DdnZ2WjZsqX0b+fFixfxyiuvICQkRKqxMfWCAUsmHBwcEBkZieTkZOmYXq9HcnIyYmJibFhZ/QkhMHnyZPzf//0ftm3bhtDQUJPHIyMjoVKpTN7jqVOnkJ6eLr3HmJgYHDlyxOQ/IuM/OMYf1jExMSbnMI4xnsOWvezduzeOHDmCQ4cOSb+ioqIwYsQI6c/3ew+MHnnkkSrbdJw+fRrBwcEAgNDQUPj7+5vUWFBQgNTUVJNe5OXlYf/+/dKYbdu2Qa/XIzo6Whrz66+/QqvVSmOSkpLw4IMPwtPTUxpTW7+sqaSkBEql6T+1dnZ20Ov1AJpOHyqT03uuTy3WZAxXZ86cwdatW+Ht7W3yeFPpw8iRI/Hnn3+a/NsZGBiIadOm4ZdffpHeQ6PqRb2Xw5PVrV+/XqjVarFmzRpx/PhxMWHCBOHh4WHyaTI5e/7554W7u7vYsWOHuHr1qvSrpKREGjNx4kTRsmVLsW3bNrFv3z4RExMjYmJipMeNWxT07dtXHDp0SCQmJopmzZpVu0XBtGnTxIkTJ8Ty5cur3aJALr2s/ClCIZpOD/bu3Svs7e3Fu+++K86cOSO+/PJL4eTkJP73f/9XGvP+++8LDw8P8d///lf8+eefYuDAgdV+VL9Lly4iNTVV7Nq1S4SFhZl8LDsvL0/4+fmJkSNHiqNHj4r169cLJyenKh/Ltre3FwsWLBAnTpwQs2fPbrBtGkaPHi2aN28ubdPw3XffCR8fHzF9+vT7ug+FhYXi4MGD4uDBgwKAWLRokTh48KD06Tg5vef61GKNPpSXl4sBAwaIFi1aiEOHDpn8u1n5U3D3Qx/q6kV17vwUYWPrBQOWzHz88ceiZcuWwsHBQXTr1k38/vvvti6p3gBU+2v16tXSmJs3b4oXXnhBeHp6CicnJzFo0CBx9epVk/NcuHBBPPHEE8LR0VH4+PiIV155RWi1WpMx27dvF507dxYODg6iVatWJq9hJJde3hmwmlIPfvjhBxEeHi7UarVo27atWLlypcnjer1evPnmm8LPz0+o1WrRu3dvcerUKZMx169fF8OHDxcuLi7Czc1NjB07VhQWFpqMOXz4sOjRo4dQq9WiefPm4v33369Sy9dffy0eeOAB4eDgIDp06CB++ukny7/hahQUFIgpU6aIli1bCo1GI1q1aiXeeOMNkx+g92Mftm/fXu2/B6NHj5bde65PLdboQ1paWo3/bm7fvv2+6kNdvahOdQGrMfVCIUSl7YSJiIiI6J5xDRYRERGRhTFgEREREVkYAxYRERGRhTFgEREREVkYAxYRERGRhTFgEREREVkYAxYRERGRhTFgERHdpZCQECxevNjWZRCRDDFgEVGjMGbMGMTHxwMAHnvsMUydOrXBXnvNmjXw8PCocvyPP/7AhAkTGqwOImo87G1dABGRrZSXl8PBweGun9+sWTMLVkNE9xPOYBFRozJmzBjs3LkTS5YsgUKhgEKhwIULFwAAR48exRNPPAEXFxf4+flh5MiRyMnJkZ772GOPYfLkyZg6dSp8fHwQFxcHAFi0aBEiIiLg7OyMoKAgvPDCCygqKgIA7NixA2PHjkV+fr70em+99RaAqpcI09PTMXDgQLi4uMDNzQ1DhgxBVlaW9Phbb72Fzp0744svvkBISAjc3d0xbNgwFBYWWrdpRNTgGLCIqFFZsmQJYmJiMH78eFy9ehVXr15FUFAQ8vLy0KtXL3Tp0gX79u1DYmIisrKyMGTIEJPnr127Fg4ODti9ezdWrFgBAFAqlVi6dCmOHTuGtWvXYtu2bZg+fToAoHv37li8eDHc3Nyk13v11Ver1KXX6zFw4EDk5uZi586dSEpKwvnz5zF06FCTcefOncOmTZvw448/4scff8TOnTvx/vvvW6lbRGQrvERIRI2Ku7s7HBwc4OTkBH9/f+n4smXL0KVLF7z33nvSsVWrViEoKAinT5/GAw88AAAICwvD/PnzTc5ZeT1XSEgI5s6di4kTJ+KTTz6Bg4MD3N3doVAoTF7vTsnJyThy5AjS0tIQFBQEAFi3bh06dOiAP/74Aw899BAAQxBbs2YNXF1dAQAjR45EcnIy3n333XtrDBHJCmewiOi+cPjwYWzfvh0uLi7Sr7Zt2wIwzBoZRUZGVnnu1q1b0bt3bzRv3hyurq4YOXIkrl+/jpKSknq//okTJxAUFCSFKwBo3749PDw8cOLECelYSEiIFK4AICAgANnZ2Wa9VyKSP85gEdF9oaioCE8//TQ++OCDKo8FBARIf3Z2djZ57MKFC3jqqafw/PPP491334WXlxd27dqFcePGoby8HE5OThatU6VSmfxdoVBAr9db9DWIyPYYsIio0XFwcIBOpzM51rVrV3z77bcICQmBvX39/2nbv38/9Ho9Fi5cCKXSMKn/9ddf1/l6d2rXrh0uXbqES5cuSbNYx48fR15eHtq3b1/veojo/sBLhETU6ISEhCA1NRUXLlxATk4O9Ho9Jk2ahNzcXAwfPhx//PEHzp07h19++QVjx46tNRy1adMGWq0WH3/8Mc6fP48vvvhCWvxe+fWKioqQnJyMnJycai8dxsbGIiIiAiNGjMCBAwewd+9ejBo1Cj179kRUVJTFe0BE8saARUSNzquvvgo7Ozu0b98ezZo1Q3p6OgIDA7F7927odDr07dsXERERmDp1Kjw8PKSZqep06tQJixYtwgcffIDw8HB8+eWXmDdvnsmY7t27Y+LEiRg6dCiaNWtWZZE8YLjU99///heenp74y1/+gtjYWLRq1QobNmyw+PsnIvlTCCGErYsgIiIiup9wBouIiIjIwhiwiIiIiCyMAYuIiIjIwhiwiIiIiCyMAYuIiIjIwhiwiIiIiCyMAYuIiIjIwhiwiIiIiCyMAYuIiIjIwhiwiIiIiCyMAYuIiIjIwhiwiIiIiCzs/wElYOC2/TSL2gAAAABJRU5ErkJggg==\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "HZByKcHfU0xE"
      },
      "execution_count": null,
      "outputs": []
    }
  ]
}