visualize-results-GD.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import os\n",
    "import glob\n",
    "import pickle\n",
    "\n",
    "\n",
    "# define constants and other global variables\n",
    "label_size = 28\n",
    "tick_size = 20\n",
    "title_size = 40\n",
    "subtitle_size = 28\n",
    "line_width = 3\n",
    "legend_size = 20"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## CIFAR"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "imp_df = pd.read_csv(\"results/CIFAR10_ResNet20/results/IMP/IMP_cifar10_resnet20_prune_0_2.csv\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "RESULTS_FOLDER = \"results/CIFAR10_ResNet20/results/finetune_results\"\n",
    "sp_list = [1,2,3,4,5,6,8,10,13,16,20,26,32,40,51,63,79] #[1]#\n",
    "vanilla_acc = []\n",
    "mask_shuffle_acc = []\n",
    "weight_shuffle_acc = []\n",
    "#idx = [-10:-1]\n",
    "\n",
    "weight_reinit_df = pd.read_csv(\"{}/finetune_reinit_weight_1.csv\".format(RESULTS_FOLDER))\n",
    "for sp in sp_list:\n",
    "\n",
    "    vanilla_acc_df = pd.read_csv(\"{}/finetune_{}.csv\".format(RESULTS_FOLDER, sp))\n",
    "    mask_shuffle_df = pd.read_csv(\"{}/finetune_shuffle_mask_{}.csv\".format(RESULTS_FOLDER, sp))\n",
    "    weight_shuffle_df = pd.read_csv(\"{}/finetune_shuffle_weight_{}.csv\".format(RESULTS_FOLDER, sp))\n",
    "    \n",
    "    vanilla_acc.append(np.mean(vanilla_acc_df['test_acc'][-10:-1]))\n",
    "    mask_shuffle_acc.append(np.mean(mask_shuffle_df['test_acc'][-10:-1]))\n",
    "    weight_shuffle_acc.append(np.mean(weight_shuffle_df['test_acc'][-10:-1]))\n",
    "\n",
    "    '''\n",
    "    print(np.mean(acc_df['test_acc'][-10:-1]))\n",
    "    print(np.mean(mask_shuffle_df['test_acc'][-10:-1]))\n",
    "    print(np.mean(weight_shuffle_df['test_acc'][-10:-1]))\n",
    "    '''\n",
    "    \n",
    "#     ax1.tick_params(axis='both', labelsize=tick_size)\n",
    "#     ax2.tick_params(axis='both', labelsize=tick_size)\n",
    "\n",
    "#     ax1.set_xlabel(\"Epoch \", fontsize=label_size)\n",
    "#     ax1.set_ylabel(\"Test Accuracy\", fontsize=label_size, color='C0')\n",
    "#     ax2.set_ylabel(\"Model Sparsity\", fontsize=label_size, color='C1')\n",
    "\n",
    "#     #ax2.set_ylim([0, 55])\n",
    "#     ax1.set_ylim([10, 95])\n",
    "#     #ax1.set_xlim([0,152])\n",
    "#     ax1.legend(fontsize=legend_size, loc='center left', bbox_to_anchor=(1.1, 0.8))#, loc='right')\n",
    "#     ax1.grid(True)\n",
    "#     #ax2.grid(True)\n",
    "#     plt.savefig('hc_quantized_resnet20_cifar10.pdf', format='pdf', dpi=600, bbox_inches='tight', pad_inches=0.05)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAtUAAAE+CAYAAABC2Oy3AAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8vihELAAAACXBIWXMAAAsTAAALEwEAmpwYAACmqUlEQVR4nOzdeXgT1foH8O+bpOm+7xste6FggRaEouwIKIuiIiIiqAh6Fa94cUO4P64LyhVQVFAUF0CEiwgo+yKLgkDZoUBZC3Tf9z05vz9m0iZp2qZp2rK8n+fJM5OZMzMnadq+OfOec0gIAcYYY4wxxpjlFM1dAcYYY4wxxm53HFQzxhhjjDHWQBxUM8YYY4wx1kAcVDPGGGOMMdZAHFQzxhhjjDHWQBxUM8YYY4wx1kAcVDPGGGOMMdZAHFTXAxGJWh75RHSNiDYS0WQicmzu+pqDiCYavY6zRFTn54KItukdE9rIdXyYiP5PfriZUZ6IqDcR/Vuu500iKiGiIiK6TkTriehpIrKtZz1CiOgjIjpDRLlEVEBEF4jocyIKt/gF3sGIKJCIpsm/F5eIKJuIyogonYhOENFSInqEiNQ1HB+q9zmLN6OMuY9/1lLn+43K/l89Xm9tddHKn5tLRPQ/Ihpb0+u2FBH5E9EI+XdlExEl1/X+mXFO/twzxpg5hBD8MPMBQNTjcR1Ar+ausxmvaaKJuj9txnHb9MqHNnIdfzD3WgA6AEg082cUByDSzDo8BSC/lnOVAnituX+et8oDgCuAzwGUmPmzSAMwDYCN0XlC9crE13CtUDOvof/4Zy11/86o7DUAZObrrm9dzgBob6X3fEQd1zL5/vHnnh/84Ac/rPNQgVnqEaPnLgC6AngagCeAFgC2EFEXIcT1pq5cA80hojVCiLLmrogFPAEEyOuFAHYBOAggQd52D6QvEr4A2gHYTUS9hRCxNZ2QiB4C8CMAJaRA4hcA2wGUA+gL6WeuBrCAiPKFEN9a+TXdVoioDYDfAYTpbT4CYCeAeAC5kH5OrQEMBdAJgDeAzwCcBrC3AZdPB/CCGeXOmNoo32F63GhzKID+AP5oYF0IgDuAeyEFq46QXvsOIuoshMir5/mNKY2elwM4C+nvUr3x554xxuqpuaP62+kBvRaaWsr4ADivV/ab5q53Ha9pol5dC/XW/1HHcbdqS/V9kFoWpwBwrqGMG6TATXfOfbWczwGGLd/PmCgzGFKgISC16vk298+1GT9PnpDu0ujer1Oo444NgB4Adsjl+xntC0UdLa3mlKlH/fV/H77XW19h5vFm1QVAWwApemVnWuG9jwbwjfzZjwKglrfX+73hzz0/+MEPftT/wTnVViaESAPwL71NI5urLhb4BkCRvP4uETk0Z2UsdBJAOyHE10KIfFMFhBA5AMag6rX2IaKWNZxvMqpavtcKIX40cb6dABbKT51g+PO/2/wI6S4NAPwN4H4hxN+1HSCEOCKEeADAawCa++7IJHlZAeBNVLVoP0pELta6iBDiEoD39DY1+O+EEOKgEGKy/Nk/Khp2p4k/94wxVk8cVDeOP/XWfYjItbbCRNSLiJYQ0TkiypE71d0gojXyLdhaEZFS7nj3u16nvGJ5/bjcGWy0GUFyCoBF8rofgFfrurY5iEhNRM8R0W969cshotNENJ9q6OhIRD8QkQDwjN7mayY6gP2g2ymEKBBClNdVJ/nLz369TZ1rKPqE3vpntZzyc0gtdoAUsDcYEQXLHcSOE1EWEZUSUaL8c55IRMa3+42P36t7j/S2PUFEO4koRT7fdSL6nojCajuXmfXtBUD3ec0H8KSoR0qDEOJTIcTBhtbDUkTUGsD98tPt8mdkhfzcHoafBWvQ/zvRzsrnbqhm+9wzxtjtioPqxlFq9NzOVCEiciSinyHl/E6F1MnOFYAtgGBI/6Q2yb34nWs4hxekFsHlAIYDCJKPt5PXu0JqdVoH4AEz6j4PQI68/gYRuZtxTI2IKArABQDfQupIpaufK6RAdjqAOCKa0pDrWEi/JdveeKfcMtlTfpoL6X02SQhxE8A5+WkLIurYkIrJ70ccpNbSrpBycdWQWg+HQ0pNOFHTFxIT57MjovUAVgMYBCmnXA2pVXkigJNENKwhdQbwT73178Xt15dgIqS8Z6AqmF4JQCOvTzI+oIH0/06Y/BvRHJrzc88YY7cz7qjYOPSHmSqFNLKBAZKGc9uFqn9eNwD8DCBWPqYNgAkA2kNq/dtARIOFEFqjU30DoLu8flk+x0UAxZA6T7YH0AdS56g6CSGyiWgegA8h5R6/AeBtc441Jrdc7oKUnwkAuwFsBXATUhDRC9JrdADwFRGVCiF+0DvFIgAbII0K0V/eNgXV388bltQPhj8nUwFgR1QFWSdNvPfGjuqdsxOqgo16kQPqr/Q2/Q5gM6QvO+0gBXctIX0p+YuIugoh0us47XcAHgZwDFJgfQOAF6QOc9GQvuisJKL2QogMC+pMAAbqbVpRU9lbEUnDSOruiOQB+A0AhBDJRLQb0hfSXkQUJoS4YKXL6n/+LP0MN4Zm+dwzxthtr7mTum+nB8zoqCiXW6tXdm8NZRbqlVkCuVORURkbSDmqunJTjfb7QGpFEwBiADjWUqcQACEmtk/UO/9b8jYHAMmo6rzoZ+K4WjsqAnCGFCgIAAUAhtVQrzao6thWAMDLRJkfaruWhT/L+/TOmQZAaaLMBL0yP5hxzjl65d+1sF6hkHK9BaS83jEmytgD2KR3rbU1nGuv/mcWwPswGhoO0t2q9Xpl3rCw3h30zlEEQGWln1Oo3nnjLS1jxnUG653jW6N94/X2fdTQ+srlbCDdoTL789WA97Be701zfO75wQ9+8ONOeHD6h5UQkTMR9SGi3wA8prfrIxNl/QG8JD/dLYR4UZjoVCSk3ODnAVyVN003KtIKVSk8q4QQhTXVTwhxXZh5O14IUQQpAAOkAHuWOccZmQwphQUAXhRCbK3hWpdRdVvdEeYNh9Yg8oQbX+ptmieE0Jgo6qa3bk7rbWYNx9bHNFSloswXQvzPuIAQohjAOEhffACpE11dObl/CCHeFUII/Y1CaoWcobdpiGXVRqDe+nUhRIWF52moEBM598aP/zNxnH5qh3Er+6+QvvABwIS6ctlrQhJ3IhoKaXi+XvIuDYBPLDlnI3HTW2+qzz1jjN32OP3DQvqdv2oxXQixzcT2MZDyWQFgfm0nEEKUE9EaSCkYbYkoVAgRL+8u0itq7ZnNlgJ4HVKawWQimi+EuFrHMfqelpfJAH6qraAQ4g8iSoKUL/wApNSTxrQY0njVAHACVZ0zjTnprZeYcd5ivXWTOfBmGC0vK1DLZ0MIkUdEiyGNIEGQUjvm1XLeGjubCSEuE9FNSF+CLM2J9dRbz7HwHM2CpFk6dePOX4dhB1YIIYqIaB2k9BB/SF88tphx6hAz/k6UA3hKCHG2XpVuXM3xuW92x44dC1UqlS8oFIphQogG9SVhjN15iChbq9Vu1Wg0SyMjI+NNleGgunGcADChln+U9+ut+xDRw3WcT/8PfAdIE2gAUv61Lhh9Ts5r/QbAEVF3HmSt5GD+35A6QNpAusX7dO1HSeTRTnRBazKAkVLVaqVrCexQ/9qaj4hmAHhOfpoL4AlTdwlMMOdLVIMQkQ+kNB0AOCWk0SdqswNVw7LVlTN/qI79iZCC6ts9mDBn8hfjnOgnUdVRcKVxa75sBapyrp+FeUF1XXYAmKL3JflW1Oif+1vBsWPHQm1sbH719fV1c3Nzy1er1Rlm/M1ijN0lhBAoKyuzycnJGZuamjr02LFjo00F1hxUW05/RkV7SLmUT0FqMe4K4GUieqmG4DZUb/2Hel63MugRQmjkTm3rILV8Pys/cojobwB/QRoa7Fg9r6HzE6SOip0AjCOij81sUQtGVVpKN0g5u+ZqtKCOiF5AVWtuIYAHhTRecE0K9NarjQ5ign4Zk2Nk18Ffb/2iGeX1y/jXWEpS12183UgUtmZc15RbJQWgSAixoZ7H1Jb6obMHUgfbYAAjiMhTCJFZQ1kd4wDfCVIfgomQvjwNgvRF9b1qRwKo48t2kRBiRx3Xt1RTf+6bnVKpfMHX19fN19c3q7nrwhi79RARbG1ty+W/ER7JyckvAHjHuBwH1RYy9Y+biD6CdJv9FUijVGTBxJsOaTg5S6n1nwghNhFRDwD/B2mUEBtIQc0w+fEBEZ0FMKOGVJQaCSG0RPQupBE4FAA+ADDKjEMb8vpsGnBsjYhoAqpG1CgGMELUPSZyjt66Z02FaiiTU1OhWujfOq8xP16PfvBT6233ht65MEOS3noIEamaMa/abEQUjqrRc2KEEHGmysm/Cz8BeAvS7+BTqDltSMdkgE9EHwBYA+mL+X+IKFUIsdTE8bV9Gb0Owy/n1pSjt94Un/tmp1Aohrm5ud2WXwgYY03Lzc0tPzU1dRhMxHfcUdGK5NvGr0EaYgoA3iQiU7fldcFQBQAbIQTV4/GDieueEkI8Aumf2zBILV/7IOVrAlJL8xYiesqC17QRwGH56Ugi6llbeaPXB0ijB9Tn9Vn9nisRPQlpXGeC1CL7iBBijxmH6rcEh5pRPkRv3ZyWZmP6/9QdzSivn/va3AHBeVS1VtsD6NJ8VakX/Vbq7rV1cIQUUJs6rl7kDsgTILV8A8BCIgqp5ZCm1tSf+2YnhHBXq9V1ThrFGGNqtbq8pn4XHFRbmTyKhG6UDgVM9+pPlJcqWHEmNSFEvhBimxBithCiH6SUAN00wgRggYUjF+iPUz3XjPKJeuvW7kBZL0T0GKRb+gpIU2A/JoTYbubh5wDoWni7ymMZ1yZKb92SjmfJeuttzSivXyapxlJNQP5CuVtvk1n5982JiFSQhsuzRBci6mLptYUQBQBmyk8dIN0FMi5T25fPUEuvbYam/tzfEjiHmjFmjtr+VnBQ3QiEEH9CysEEgPuI6EGjIvv01h9BIxFCZAohpqOq5dwH5gVrxufZA2Cn/LQfEdU6M6OQJg/RTQARSUTBtZU3g37qgtn/+eSc1J8BKCHdFXhCCLHJ3OOFNMW2roOfK6om6jF1Lf2RM24IIeo9AYbcMVE37GEXIvKu4xD9n8OR+l6vEeiPMDLpFmt9NeVBSDNLAsApSJ1x63qs1Tu+oTMsroI0YRMg9Vlo1i+gOk39uWeMsTsFB9WNR79F999G+1ZDajUFgNeIyK+R6xKvt25pHr1+7tCHqDu4/VFeKmBe63Zt9NNJzEmLgPxFZg2k16sBMM6CDmyQz6Hzai3lXkHVe1JtbOl6WCcvVTCc9tuAPG29bqxzgfp1Bm0Uco66blQMZwA/y/U0CxG9SkTRjVI50/SD4v8KIf6vrgek91yXJvCUPOa5ReS7WrqOs4TqfyeaU1N/7hlj7LbHQXUjEULshDQlNAD00G+tFkLcBPC5/NQTwHYialPTueRJIwYS0Uyj7UPkQKTGjoHyeQfLTwsAXKn/qwGEEEdRFfBFAuhdxyFfoqrV9SkiWlhbAEJELkQ0jYgGmdh9TW+9W111lc+hGxFFA+BpIcTa2o+q0beoSq0YQ0TPGBeQr/ea/LQADZvI43NUjfv7BhE9auJ6dgBWQhpKEQDW1TGKSVN6BkCCvN4L0jTqtebhE1F3ItoB4FMYdcRtLPJdgIfkp4WQOuPWSb4Lo+vw6wlgRAOr8iOqPl+PEVGnBp7PWpr6c88YY7c9Hv2jcc0F8Iu8/n8wHNv2bUiduQZCGtP5HBFthDTxRAqkUTB8AURACooDIOWs6ude+kMKROYR0R5IHQqvQpoUxgvSqAZjUNW6+6k8G5+l3oU0yYgSdbQYCyEK5fSLfQBcILW6jiGi/wE4DSAPUmtmSwA9APSHNJybqVxc/VzdeXJAFAcppQMAEoUQZwBAznPdiKpxh9cBKDZjLPALQgjj8Yt1E3+8IJ9TCeB7InoIwFb5+n0hdTrT/S69JoRIreNaNRJCxBPRa5BGKlEB+EX+XGyBNLJCW0jDJraSD0kE8A9Lr2dtQogMIhoI4HdI/QXuAfA3ER2GlEIUD+ln7wGgNYChADo3Q1XHo2qkmV9FLbORmrACVcH0JFR92aw3IUQZES2AFJDqWqsft/R8RPQ6ah6W0o2I3jfadk0IscxEvZr0c88YY3cCDqob13pIwV97SCMLPCSE2AxUTq7yIKRZ816E9A/+MRhOcW4s0ei5LtdYDWmWt5qmmBaQhv9q0O1lIcQFIloOM3NJhRAn5eH+foY0dncAaklpgDQyR7XxlIUQp4noZ0iTdPiieovYj5DG/gWkLyoOevvGyI+6zIH0xacaIcRmuaXuK0gjbjyO6oFPOYA3hRDfmnGtWgkhvpYn8lkI6cvBKJgeyvAspKEB65okpkkJIS7Ko958CGmiHTWkyWlqm6AmBdKoNX81fg0BGH6GV9bz2N8hTRzkCmAoEfkLIZLrOKY2X0NKr/KANOV8JzPHgzflFRiOxqHPFVWdI3X2AagWVANN/7lnzY+IIgHAeG4D3XYiwpkzZ86Gh4eXmjr+3nvvbXfkyBFnAPjss8/ip02bVjmW+6OPPhr666+/GgzRaGdnpw0KCiobPHhwzpw5c1K8vb011n5NjDUlTv9oRPLYwB/rbfq30f4yIcQrAMIAfASppTkdUktQEaS0hy2Q/uHeI4QwvgW7AlJL9nRILUqXId3K1kD6p38SwBcAIoUQ/7TSWMX/h6qJQuokj/sbCSko/BHSkFt5ch1zIHUQWw4pKPavZSztpyF9+dgLKfBu0jGQhRA/QRqacB6kmSzzIb3XFyGlunQVQiys+Qz1vt5XkFp6P4b0c8yBlIefDOkzMQlAF3GLzsYnhMgRQrwEqTX6NQCbIN1FyYP0s8uENPPoUkifjWAhxOKmGNtaDhB0reMpMLwTUichRAmqOiwqIbXYWkweCeQLXfVwC+VWN/Xnnt26lEqlEEJgyZIlXqb2nzlzxjYmJsZZqVTWOgvnwIEDc1577bXk1157LfmRRx7JzM3NVS5ZssQvMjKyQ2pqqiWjUzF2yyBhckZexhhj7O5w6tSp+IiIiLpmHb3j1dZS7ePjU+7t7V2emppqk5SUdNrGxnCerhdffDHwq6++8hs0aFDOrl273GpqqTbenpubq4iMjOxw5coVu+nTpyfNnz+/IXd9GGsSp06d8oqIiAg13s4t1Ywxxhir08SJE9MzMjJsVq9e7aa/vbS0lNauXevVtWvXwg4dOtSr346rq6t27NixGQBw7Ngxs0Z3YuxWxUE1Y4wxxur03HPPZdnb22u/++47gxSQn3/+2TUzM1M1ceLEdEvOq5tJlyfgYbc77qjIGGOM1SD0rc2RzV0Hc8V/9NCxuktZzt3dXTtixIisdevWeV25csWmdevW5QCwbNkybycnJ82kSZOyZ86cWa95F3JzcxWrV6/2BICoqKj6jMLD2C2Hg2rGGGOMmWXKlCkZ//vf/7yWLFni9cknnyRfvHhRffDgQZcnn3wy3dnZuc7O8L/99ptbfHy8LQCkpqaqdu7c6Zaenm4THBxc+sYbb9xSIxkxVl8cVNfBy8tLhIaGWnRsYWEhHB05Raw++D2rP37P6offr/qz5D07duxYhhDCu5GqxJrJgAEDCtu2bVv8888/e3388cfJX375pZdWq8WLL75oVurH7t273Xbv3u0GSEPqBQQElI0ePTprzpw5yTykHrvdcVBdh9DQUBw9etSiY/fu3Yt+/fpZt0J3OH7P6o/fs/rh96v+LHnPiOh63aVufY2dUnE7euaZZzLefffd4F9++cV19erVXuHh4UW9e/c2q4Oi8egfjN1JuKMiY4wxxsw2ZcqUTDs7O+2rr74akpaWZmNpB0XG7jQcVDPGGGPMbF5eXpqhQ4dmp6am2tjb22ufe+65rOauE2O3Ak7/YIwxxli9zJs3L2n06NE5vr6+5e7u7taYrZex2x4H1Ywxxhirl7Zt25a1bdu2rLnrwdithNM/GGOMMcYYayBuqWaMMcYYhBAmRzqpabspixYtSlq0aFGS8fZ169bFA4i3uHKM3QY4qGbsLiWEgEYrUKEVKNNoUaERqNBoK9fLNVqUawQqtNrKdQBwVKvgYKusXDrYKKFS8k0vxhhjdzcOqhmrgxBS4KkLQCs02srn5Rqt3nYpAJW268pLAWrdx2qNriGg0WpRrtsmn1tXtlwOeis0AslpJfju6hGUV2jlALhqX7lGi3KtYZBcuU+rhRDWeY/UKgUc1Uo4qFVwtJWWDkbPHdVKONhK243LOtoqYW8jL9VK2CqVsFERbJQKqBQEIrJORRljjLFGwkH1XUS/ZbJCK6CRAyv9AE8XDFYFfHILplHwVxnk6crL56nQCGjl62gFoBUCWq2ARl5qBaR13XatXEY+JjGxFFszTkvPjY7R1V+jlV+LkK+h1V1P94DB88rycj2E/n6t4Xm0ldcQlQGtRmulyLMxpTfvMLFlFVqUVWiRXVRu9XMTATZKBdRKBWyUBLVKUflct2683UZvn1pFAAgKks6VlFSKPblnKwN1BRGIAAKgUBAIAAgg+RilguDlZAs/Vzv4u9rBz9UOXo62UCg40GeMMVaFg+pGcDohB9cyCnE2sRxpMTflgFRr1NpYFaRWBbNyy6TGMEg1DGyrB7mmgmLD61Qdc1tIuNncNbhr6AJWGwXBRqWASiEFqDZKBVRKglpeqhRSsKoVAkVlGhSVVchLDQrLKqzW4m2KEFVBu9XcaNhkfyoFwdelKsiWlvaVzwNc7eHjzIE3Y4zdTTiobgQ/H7mJn4/ckJ6cOd28lWFWoVQQVPJDqZCCzsptcoqC0nhdQZUBqXSMbrsUqJo63uC8xuUUBKUuAJaDXRulAnHnY9El4h4pAJbPp9bbb6OUttkoCTYK/e3StRpKCIGSci0KyypQVKpBUXkFCkulwFu31AXildvLNCgqNQzMi0qlZXGZBmWaqjzuW/FOQYVWIDGnGIk5Nc/MbKtSINTTEaFeDvLSEaGejmjp5QhfF1tOaWGMsTsMB9WNwEZ56/6zVBAqgzj9gM5gXUlQyi2WSoUUiCkrA0T9cvrHVgWPCiIoFdKtdAURlCTdRlco5HV5u+7Wuv76pUuX0CGsPZTyLfnK/QrD81Reg3TrcvnKdWlpfB6Tx1SWI7mc4XukO9+tyiEzDv3a+zTb9YkI9mopFxpO1j+/7k5MmUaL8grdUlQG3mUV2sr90rouf1yLUnmfEFLwLwBcvHgJbdq0gVYAAvJ2AQhIS63euhDS3Z/0glKk5JYgObcEKbnFZqW5lFZoEZeaj7jU/Gr77G2UCPF0QEsvR3QP9cCj3YLg6mBj/TePMcZYk+GguhF0DnTFiIgAZKSnIsjf30RrpV7AatAyaSJgVVa1WiqVVQGu8fG68+oHuAbryqqA9la2tzQe/Xq0aO5qsFuIUkFQKpSws1Fa5Xx7S+PRr3fLBp2jpFxTFWTnFSMpp8TgeWJ27YF3cbkGF1LycSElH1vPpmDe9gsYcU8AxvcMQUSwW4PqxhhjrHlwUN0IHo8KxuNRwdi7dy/69Yto7uowxqzMzkYppXN4OdZYJreoHPGZhYjPLMS1jELEZxTiWmYR4jMKkVtsGHCXlGux9lgC1h5LQOdAVzzdMwQjIgKk1n/GGGO3hXoH1UT0TwArhBCZ1q8OY4zdGVwdbBDh4Gay5Tm7sAzXMgsRm5iL1TE3EZuUV7nvTGIu3lh3Gu9vPodHI4Pw1L0haOPTCHk1jDHGrMqSluoFAD4ioo0AvgewXYjG7PvPGGN3FndHNdwd1ejWwh3je4bg5M0crDh0HZtOJ1eOcpJXUoHvD8Tj+wPx6BzoilFdAjD8ngD4udo1c+0ZY4yZYuk0aGoAjwHYDOA6Ec0hooYlKdaCJM8S0SEiyieiIiI6QUTTiMjk/VEiiiaiLUSUJZc/TUT/rKk8Y4w1ByJC1xbuWDCmCw6/PRAzH+yAUE8HgzJnEnPx/ubz6PXRboxd+jd+PnIDOUVlzVRjxhhjplgSVPcA8DWAXEjzJQQBeBfAJSLaTURPEpGtFesIAD8CWAagJYA1AL6BFNh/BmANGQ3NQESjAOwH0AfAegBfyuUXAlht5boxxphVuDuqMblPK/zxej8sf7YHhoT7Qq03BbwQwKGrWXj71zPo/sEuzN16HtpbcMhBxhi7G9U7/UMIcRTAUTm3+jEAkwD0hxSg95MfuUS0CsB3QojjDakgET0M4GkA1wD0EEJkyNttAPwPwKMAngHwg7zdBVLQrQHQT64viGgWgD8APEZEY4UQHFwzxm5JCgWhTztv9GnnjdyicmyLTcZvp5Jw8Epm5UQ75RqBr/ddRacAabQhxhhjzcvS9A8IIUqFED8JIQYBaAXgPQA3ILVeuwF4EUCMnKbxDyJyt/BSo+XlfF1ALV+/HMAs+ekreuUfA+ANYLUuoJbLl0BqUYdcN8YYu+W5Otjgie4t8NPzPXH47YGYPbwjwvycK/d/siPOurNNMsZYPRFRZI8ePdo3dz2am8VBtT4hxHUhxL+FEC0BPAApRaMUUoAdAWARgCQiWkVEg+p5ej95edXEPt22bkTkJq8PkJfbTJTfD6AIQHQjpKgwxlij8nGxw7P3tcSaF3rB1V6aLOZ6ZhHWxNxo5pqxOwERRRJRZG1lAgMDOxNRZFxcnNrU/suXL9u89NJLgeHh4R1cXFy6qFSqbh4eHhHR0dHt3nvvPZ/MzMwG92sqKiqi2bNn+95zzz1hzs7OXWxsbLp5e3vfEx4e3mHChAktNm/efNcPlzN9+vQAIorctGmTc92lmbVYJajWJ4TYJYR4EoA/gJcBpEAKrm0BPAFgOxFdJaI35VSNuuhap011hGyltx4mL3XflC6aqFsFpDQSldGxjDF223B1sMFL/VpXPv9s92UUllY0Y40YAxYsWOAVHh7eecmSJX4ajYZGjhyZ9eKLL6YOHTo0Oz093Wb27NnBbdq06dSQa+Tm5iq6d+8e9t577wUlJyerhw4dmvPCCy+k9uvXL8/Ozk77888/ey1dutTbWq+JsfpolMlf5FSP8QCeBeALaTZgkh8AEArgQwBvENFrQojltZxuE4AnAUwnotVCiCz5GioAc/TK6dJLXOVlbg3n0213q6X+LwB4AQB8fX2xd+/eWqpXs4KCAouPvVvxe1Z//J7Vz53yfrXUCHjYEbJKBDIKSjFr5R8Y2dpk42GD3SnvGWs8X331lcfrr78e4uLiovnxxx+vjR07ttr/4B07djhOmzYtpCHX+eCDD3zPnTvn0Lt377xdu3ZdtrOzM+ipm56erjx58iSPO8mahVWDaiJ6AFIgPQrSaBu6IDod0ggeKwB0BvAcpM6N7gC+J6JcIcTGGk67GlKAPgzAOSL6DVIKxyAArQFcAtAWUsdEs6opL2vsMi+EWApgKQBERUWJfv36mXlqQ9KMipYde7fi96z++D2rnzvp/cp2uYE3150BAOy4IfDu2Gh4OFo/sL6T3jNmfdnZ2Yq33nqrBQB8//33V0ePHp1nqtwDDzxQeOLEifMNudbhw4cdAWDKlCnpxgE1AHh7e2sGDx5cqL/t0UcfDf311189L1y4cKZ9+/YGY1Fu2rTJecSIEe1ee+215AULFiTp79u3b5/DO++8E3jy5EknIsI999xT+J///Cdxy5YtrgsXLvT//fffLw4fPjxf/5glS5Z4fP75575Xr161d3Bw0PTp0yfv008/TRgzZkyrmJgYJyHEMeM6r1u3zmXRokU+p0+fdiwsLFT6+vqWPfjggzkffPBBspeXl0Fsc/jwYfv333/f7/jx407p6ek2jo6OGn9//7KePXsWLFmyJMHW1lYEBgZ2TkpKUgPAiBEj2ukfr3/9/Px8xYcffuizfv16j+vXr9sSEdq1a1f84osvpk2ZMiXLuJ4lJSU0a9Ysv9WrV3umpqaqvb29yx999NHMjz/+OLn6T+ru1OCgWh6fehKkETiCdJsBaAHshDQSx0Y59QIAzgBYRUTdAayCFBi/CcBkUC2E0BLRSACvQhoF5GkA5QAOytf8AlJQnSYfovt27ArTXIzKMcbYbenRbkFYuv8qrqQXoqC0Al/uuYxZwzs2d7XYXWb58uXuubm5yoiIiMKaAmode3v7Bo0B6e7urgGAuLi4Ru0XtW3bNqeHH364XUVFBQ0ZMiS7ZcuWpefOnbN/6KGH2vfs2dPka5w1a5bv+++/H+Ti4qIZPXp0hqurq2bfvn0uvXv3DnN2djbZ8Pevf/3Lf/78+QGurq6aAQMG5Hh7e1fExsbaL1261Hf37t2uR44cOe/h4aEFpIC6b9++HYhIDBw4MDckJKQ0Ly9Pee3aNdsVK1Z4L1y4MNHW1lZMmTIlddOmTe4xMTFOo0ePzgwJCak2qH1GRoayT58+7c6fP+/QsWPHojFjxmRotVrat2+fy9SpU1vGxsbaLVq0qPJLhlarxfDhw1vt3r3bLTg4uPSZZ55JKysro9WrV3vFxsbaW+t9v91ZFFTLnfweg9Qq3ReGqR2JkGZaXCaEuF7TOYQQMUT0L0jjSNf6X0AOyOfLD/162APoAqAYQKy8OQ5AFIB2AI4ZlVdBys2ugOmOj4wxdttQKRWYMSQMU1dKf+pW/H0dk3qHIsjdoY4jWX28t+lcwLK/rvmbU3bEPf4Zn4/rZvC/75VVx0N+P53sZc7xz93XMnnW8I4GLaZPLj3U5u+rmTU1FJk8piGmT59e4xiNeXl51ToaHjhwwAkA+vTpU2tAbQ1PPPFE1saNGz3mzZsXGB8fbzt8+PDcXr16FYWEhJRb6xoajQZTpkwJLS0tpTVr1lwaM2ZM5euaN2+e95tvvtnC+Jhz586p586dG+jm5lYRExNzrk2bNuUAoNVqE0eNGtVy06ZNHsbH/P77787z588P6NKlS+HOnTsv6bdKL1q0yPPVV18NnTFjRuCyZctuAsC3337rWVpaSitWrLg6fvz4HP1zpaenK52cnLQAMHv27LScnBxVTEyM06RJkzKNW9MBYMqUKcHnz593mDlzZsL777+fqtteVFREQ4YMafPFF1/4jx07Njs6OroYAJYuXeqxe/dut4iIiMKDBw/GOTg4CABITU1NioqK6lDvN/kOVe+OikS0GFLnw+WQxqRWQGqV/g3ACAAhQojZtQXUes7JS0t7pz4NwA7A/+Qh9gBpLGoAGGqifB8ADgAOCiFKLbwmY4zdMoaE+6JrCzcAQJlGiwU7q/XRZqxeFi5c6F/To6CgoFpQnZqaagMAQUFBVgtsa/Lkk0/mvvfeezdtbW21P/30k/eTTz7ZJjQ09B5vb+97Ro4c2XLr1q0NHvlj165dTjdu3LC999578/UDagB4/fXX00NCQqrFDz/88IOnRqOh5557Lk0XUAOAQqHA/PnzE5XK6oOeLFq0yAcAvvnmm3jjNI9p06ZlhoWFFa9fv75aMO7g4FBtDE1vb2+NqWuYkpKSoty4caNneHh4kX5ALZ9bzJs3L0EIgeXLl3vqtuvW33vvvURdQA0Avr6+mhkzZnD6h8ySluqpeutXAXwH4HshhCVvaimksa1rHWSViFyEEHlG27oD+AhAAYD/6O36BcDHAMYS0ed6k7/YAXhfLrPEgroyxtgth4jw5tAwjF16CACw/kQiJkW3ROegGhs2GauVqbxfHf18Xb3yBABE1CTTe7777rtpr776asbGjRtdDhw44Hj69GmHEydOOP3+++8ev//+u8err76a/Omnn1rccn/06FEHAOjZs2eB8T6lUomoqKiC69evG6SfnDp1ygEA+vTpU+2Ydu3alfn6+pYZv28nTpxwUqlUYtWqVR6rVq2qVo/y8nLKzs5WpaSkKP38/DTjxo3L+v77732feuqp1kOHDs0eOHBgfv/+/QvCw8Pr1Uj4119/OWo0GhCRybsS5eXlBAAXL16s7PB57tw5R4VCgQceeKDa6xsyZEi1lvC7lSVBdTmklI1vhBC7G3JxIcQNSCOB1GUnERUDOAsgH0A4gAchBeWjhRCVqRxCiDwimgwpuN5LRKsBZAEYCWm4vV8gjaPNGGN3hJ6tPNG/vTf2xKVDCGDy8qP49aVoBLhxqqM1zBreMakh6RWfj+t2/fNxMOfurUk/v9DzsqXHNgU/P78yAEhISGic4WdMcHZ21o4fPz5HlwZRUlJCCxcu9Jo1a1aLzz77zP/xxx/P7t27d7El587NzVUCgK+vr8mWdx8fn2rb8/PzlQAQEBBg8hgvL69y46A6JydHqdFoaOHChbWmFuXl5Sn9/Pw0/fv3L9q2bduFDz74wH/btm3uGzZs8ASA0NDQkrfeeivZVOdCU9LT01UAcPbsWYezZ8/WmCtWWFhYmc1QUFCgdHFxqbC1ta32xSk4OLjR71DcLiwZpzpACDG2oQF1Pf0CKUVkPIDpkEYQ+RZAuBBiu3FhIcQGSLne+yFNY/4KpC8D0wGMFUI0ybdpxhhrKjMf6ghnW6mdJCWvBJO+j0FuMf+vY42vd+/eBQCwf/9+c+aeaBR2dnbi7bffTh8+fHgWAOzcubOyLgqFFOroWmD1ZWdnV8uZcHFx0QBVaS3G0tLSqm13cnLSAEBSUpLJYzIyMqptd3Z21ri4uGiEEMdqe7Rr166yo+GgQYMK9+zZczkrK+vkjh07Lrz66qvJmZmZNlOnTm25YcMGs1Jp3dzcNADw3HPPpdZ23cOHD1fmkjk5OWny8vJUpaWl1d7DmzdvmnzNd6N6B9VCiMzGqEgd1/yvECJSCOEmhLAVQrQUQkwVQsTXcswBIcSDQgh3IYS9EKKzEGKhEMLcofcYY+y20cbHCV89HQmVQvqfF5eaj6krjvEU5qzRTZgwIdvV1VVz8uRJx7oCu+Li4mpBmTXpglv9tjM3N7cKALh27Vq1lvSYmJhqLbVRUVFFAHDo0KFq+dkajQZHjx6ttj0iIqIIAPbv319t38WLF9WpqanVrt2lS5fCvLw85dGjR+s9rra9vb0YPHhw4aeffpo0d+7cGwCwfv16N91+pVIpdPU11qdPn0KFQoFDhw6Z3Z+tY8eOhVqtFjt27Kj2+rZv386zNsosmlGRiFrIjzqHtCEiO115S67FGGPMPL3beGHeY/dUPv/7aibe+OWUQYDBmLW5u7trdYHdxIkTW69bt85ki/Xu3bsdIyMjw/S36abTrm3EEX3z5s3z3r17t6OpfSdOnLDbvHmzOwD069evMs+3R48ehQCwdOlSgxFYjhw5Yr9s2TJf4/MMHjy4IDg4uPTw4cPO//vf/wxey/z5872N86kBYOLEiVlKpVIsW7bM5/Lly5Utt1qtFq+//nqgqeD2n//8ZyoATJ48OTQ+Pr5aa29eXp5C/7Vu27bNydQ077oWdf0OjJ6enhUAEB8fXy2YDwwMrBg5cmRmbGysw4wZM/zLy6vf0YqNjbW9cOFC5bETJkzIBIDZs2cHFhUVVX4xSk1NVX7yySdmjYxzN6h3TjUR9QbwJ6Tc5lBIec21cYA03J09EfUSQsTU95qMMcbMM7pbEJJzS/Df7XEAgA0nkxAe4IrJfVo1c83YnezFF1/MKi4uVrz99tstHnvssbZhYWHFUVFRBe7u7hWZmZmqY8eOOcXFxdnrWo11tFopDlSpVGZ989uxY4fLm2++2SIgIKAsKiqqICgoqKy0tJSuXr1q9+eff7pUVFTQxIkT0/r371+kO+app57K+eCDD0o3bdrkERUVpY6MjCxMSEhQ79q1y23QoEE5W7Zscde/hlKpxOLFi68/+uijbceNG9dm+fLlOS1btiyNjY21P3DggEufPn1y9+/f76prDQaA8PDw0hkzZiR99NFHgZGRkeHDhw/P0o1TnZOTo2rfvn3xxYsXDTo5jBo1Kv+dd95JnDt3bmCHDh069evXLzckJKSsoKBAkZCQoD5y5IhzZGRkwcCBAy8BwCeffOJ74MABlx49euSHhISUOTk5aS5cuGC/f/9+VxcXF83LL7+coTv3Aw88kD9z5ky89957QWfPnrXXje89b968ZABYtmzZjWvXrtl98sknAWvXrvXs3r17gY+PT3lycrLNpUuX7M+ePevw9ddfXw0LCysDgBdeeCFr7dq17n/88YdbWFhY+JAhQ3LKy8tpy5Yt7p07dy68efNmo44bfruwpKPiGHm5UQiRXVdhIUQWEa0DMAHAWAAcVDPGWCN6qV9rJOYUY9XhGwCAH/+Ox/P3twRRo955Z3e56dOnZ4wcOTJ3/vz5Pvv27XPZsGGDR3FxscLZ2VnTrl274jlz5tzUD/wAIDY21l6hUGD8+PFmdbJbsGBBwrp16wr27NnjcuLECccdO3a4aTQa8vT0LO/fv3/uxIkTM8aNG2cwuZuDg4PYtWtX3Kuvvhr8119/uZw9e9axTZs2xUuXLr3q6empMQ6qAWD48OH5W7dujXv33XcD9uzZ47pnzx5EREQUbt68OU43vJyrq6tB8/PcuXNTgoKCyr744gu/X375xUs3o+Jnn32WMHjw4HaOjo7Vmqs/+OCDlD59+hR89tlnPkePHnXatWuXysnJSePr61s+bty49AkTJlS+Ly+99FK6m5ub5sSJE47Hjx93qqioIF9f3/Lx48envfPOO6n6udfdunUr+fzzz68tWrTIb8WKFT66XGhdUO3h4aE9dOhQ3IIFC7zWrl3ruW3bNrfS0lKFp6dneWhoaOmcOXNujhw5snLUNYVCgc2bN1999913/VavXu31448/+nh7e5c/8cQTmfPmzUtycHDoZs7P705H9b0tSERHAXQF8IwQYqWZx4wDsBLAUSFEj3rXshlFRUWJo0ePWnQsT+1bf/ye1R+/Z/Vzt7xfZRVaRL6/E/klUsPg7y/fZ/Ewe5a8Z0R0TAgRZdEFm9ipU6fiIyIiMuouyaxJq9XC09MzolevXvlbtmy5bSZk69atW9jp06cds7KyTri4uNTZaSErK0sREBDQJSwsrOjkyZMXmqKOrHGdOnXKKyIiItR4uyU51bqpyOszw4BuOKBAC67HGGOsntQqBQZ3qEoX3XqW52dgt5ajR4/a5+TkqGbOnHnLfTjz8/MVGRkZ1fKXFy1a5HnixAnH3r175xoH1ElJSdVGxygvL8dLL70UXFpaSsOHD6/z7j67vVmS/uEmL+szVpOurGetpRhjjFnNsM7++PVEIgBg69kUzBjSnlNA2C2jR48exbVNNNOcLl++rO7Vq1fH6OjovFatWpVUVFTQmTNnHI4fP+7k7OysWbhwYYLxMStXrnT/6KOPAnr37p0XFBRUlp2drTp06JDz9evXbcPCworfeuuttOZ4LazpWBJUZwLwg9RJ8YSZx4TIy9xaSzHGGLOa+9t6wVGtRGGZBtcyChGXmo8wv2YbSpix20ZQUFD5qFGjMg8dOuR8+PBh57KyMvLy8ip/7LHHMv/v//4v2dQshr179y6IiooqOHLkiPPOnTtVABAYGFj6yiuvJM+ZMyfFycmJh+G5w1kSVJ+FFFQ/AmlmRXM8Ki/PWXA9xhhjFrCzUWJAB1/8fkqaDHDLmRQOqhkzg7e3t2bNmjX1mgWzd+/exTt27LjSWHVitz5Lcqo3ASAATxLRwLoKE9FgAE8CEPKxjDHGmsiwTn6V69s4r5oxxhqNJUH1twBS5WN/J6J/EVG1GXaIyImI3gCwQS6bAeCrBtSVMcZYPfVr7w07G+lP/cXUAlxOy6/jCMYYY5awZJryYgBPAKgAYAvgYwBpRPQ3Ef1KROuI6G8A6QDmArCH1FFxrBCi0HpVZ4wxVhcHtQr92vlUPt96JqUZa8MYY3cui6YpF0LsBzAAQCKkVBA7AD0AjALwsLxuK+9LADBICLHHCvVljDFWT8M6V6WAbD3LQTVjjDUGi4JqABBCHADQBsBkABshBc+l8iNR3vY8gDZCiD8bXlXGGGOWGBDmA7VS+nN/LjkPCdlFdRzBGGOsviwZ/aOSEKIMwDL5wRhj7BbkbGeDe1t54M9L0qSBB69kYkyUQzPXijHG7iwWt1Qzxhi7fUS39qpc//tKZjPWhDHG7kwcVDPG2F0gunXVhLYHLmdACJ6HgjHGrImDasYYuwt0CnSFs52U8ZeWX4or6TwYE2OMWZPFQTURqYloEhFtJKJ4IiogIk0djwprVp4xxph5lApCz1ZVrdUHr2Q0Y20Yu31Mnz49gIgiN23a5NyQ82zatMmZiCKnT58eYK26WUNgYGDnwMDAzo19naysLMXEiRODAwMDO6tUqkgiijx48KA9AJSWltJrr70WEBIS0kmtVncjosgVK1a4xcXFqYko8tFHHw1tzLoRUWSPHj3aN/Q8FgXVRNQOwElIE8GMANACgAOkIfTqejDGGGsGvfVSQA5e5rxqZoiIIokosrYygYGBnYkoMi4uTm1q/+XLl21eeumlwPDw8A4uLi5dVCpVNw8Pj4jo6Oh27733nk9mZqaycWp/Z7JWQH8rePnll4N+/PFHn/bt2xf/4x//SH7ttdeSg4KCygFgzpw5vp9++qm/j49P+ZQpU1Jfe+215E6dOpU0d53rq96jfxCRI4CtAFoC0EIaOi8d0tB6AsD7ADwARAG4V972N4Cd1qkyY4wxS0S30euseDUTGq2AUsFtHcw6FixY4PX222+3KCsro/bt2xePHDkyy93dXZOZmamMiYlxnj17dvCCBQv8s7OzTzV3XetjxowZaU8//XRWmzZtypq7Lrez3bt3u4WEhJT+8ccfl433bd++3dXBwUH7559/XrSzs6vs8FHTl7dblSVD6k2FFFBrAAwRQvxBROGQgmoIIf6tK0hEXQGsANATwGohxBcNrzJjjDFLtPVxgpeTLTIKSpFbXI7zyXnoFOja3NVid4CvvvrK4/XXXw9xcXHR/Pjjj9fGjh2ba1xmx44djtOmTQtpjvo1hL+/f4W/vz+nrzZQenq6TVRUVIGpfampqWo3N7cK/YD6dmRJ+scISK3P/xNC/FFbQSHECQD9AaQBWFDXbSXGbgXlGi0OXM7A7I1nMXjBPqw8dL1amdkbz2L04gMY89XfGLv0b4z/9jCe+e4Inv0hBpOXH8WLK4/hH6uO49XVJxATn1Xt+CV7r+DDLefx8bYLmL8jDgt3XsTnuy9h8d7L+HrfFXz751X8cOAaVvwdj+Tc4mrHX0rNR3p+KY/gwOqFiKqNAsJYQ2VnZyveeuutFgDw/fffXzUVUAPAAw88UHjixInzll7n1KlTtkQUOXLkyJb62y9cuKDWpa5s27bNSX/f1KlTg4go8rfffjNIn7hy5YrNhAkTWgQFBXVWq9Xd3NzcugwYMKDNvn37qg3gXlsKxpIlSzw6duzYwc7OrpuHh0fEww8/3DI+Pt6mR48e7WuLeQ4ePGjfr1+/Ns7Ozl3s7e27du/evf3OnTsd9csEBgZ2XrhwoT8AjBgxop3uNZoTS2m1Wnz++eeeXbt2DXN3d4+wtbXt5ufnd899993X9ptvvnE3dUx+fr5iypQpQf7+/p3VanW3Fi1adJo5c6afVqs1KFdXbrhxjrbuvRBCICYmxkn3Gnr06NH+0UcfDSWiyMTERHVSUlLlz9GcHO/8/HzF22+/7RcWFtbR3t6+q4ODQ9cuXbqEff311x6mypeUlNCMGTP8g4ODO6nV6m6BgYGdp02bFlBcXGy123WWtFR3lJfrTe0kIhJ6/+mFEOlEtADAPAAvA5hkwTUZa1Ql5Rrsv5iO7bGp2H0hFTlF5ZX7CkurN1BcTM3H8Rs5Zp27X3tvdA81/B1fe/QmrmaYN/pCa28n+LvaG2x7fvlRXM8sgoNaCQ+1Fh1vHEWIpwNaeDighacjQjwcEOhuDxslD/DDDPVu44nfTiUBAA5cycSUvq2buUbsdrd8+XL33NxcZUREROHo0aPzaitrb29vcUtAREREqY+PT/nff/9tENxu2bLFRbe+Y8cOl6FDh1a2hv7111/Otra2YtCgQfrbHEaMGNE2NzdXdd999+U9+OCD2ZmZmaodO3a4DR48OGzFihVXnnjiCZNfDPTNmjXL9/333w9ycXHRjB49OsPV1VWzb98+l969e4c5OztrajruxIkTDkuWLPHt0qVL4ZNPPpmRkJCg3r59u/uIESPaHz58ODYiIqIUAKZMmZK6adMm95iYGKfRo0dnhoSEmJ1+Mm3atMAvv/zSLzAwsGz48OHZrq6umpSUFJtTp045rlu3zn3y5MnZ+uXLy8upX79+bVNTU9X9+/fPUyqVYvv27W4ffvhhYElJCc2fPz/Z3GsbGz9+fMZ9992Xv3DhQv+AgICyJ554IhMAQkNDS93d3TUhISFl33zzjQ8ATJ48OQ0A3Nzcar0zkJGRoezTp0+78+fPO3Ts2LFozJgxGVqtlvbt2+cyderUlrGxsXaLFi1K0pXXarUYPnx4q927d7sFBweXPvPMM2llZWW0evVqr9jYWPvarlUflgTVbvJSv/muVG/dCUC+0TEH5GVfC67HWKPILS7Hngtp2B6bgr1x6SguN/030FTOqUZr/v8FpaJ6YKupRwuzwuj6FRotErOl1uuiMg2KyoCEc6nVjyMgwM0eP0zqgTY+VY03QggUlFbA2c7G7DrcyoQQKNNoUVYhPco1QlrXaODjYgcXo9d5Jr0CBaeT5LJalMnlyzValFdopXNptCivEJjUOxTBHoYNV+tPJMBWpUSPlh7wcrJtypdqFb1aVeVVn7iRDSEEiDivujbzYuYFrDi3wt+cskNDh2b8t+9/DW5vzdg3I2Rb/Davmo7R93THp5Pf6P5Gkv62Z7c/2yYmJabGPB1TxzREbaNT5OXlVetoeODAAScA6NOnT60BtTVER0fnbdiwwfPo0aN2UVFRJQCwZ88eZzc3two/P7/yffv2VQbc6enpyvPnzzv06NEj38HBQQBAeXk5xo0b16qoqEj5+++/xz300EOVwbbcwtzh5ZdfDhk5cuSZ2r4AnDt3Tj137txANze3ipiYmHNt2rQpBwCtVps4atSolps2bTLZWgoAe/fudf3ss8/ip02bVtlb+L///a/XG2+8EfLf//7Xd+XKlTcAYPbs2Wk5OTmqmJgYp0mTJmUOHz7cOLaq0cqVK719fHzKz58/H+vs7GzQ1JycnFwt9ktPT7fp0KFD0b59+846OTkJAEhMTEwKCwvrtHTpUt8PP/wwxdbW1qIvRLrXuXDhQv/AwMCyBQsWGHxWn3766Zw1a9Z4AoDxvppMmTIl+Pz58w4zZ85MeP/99yv/ARYVFdGQIUPafPHFF/5jx47Njo6OLgaApUuXeuzevdstIiKi8ODBg3G6z0NqampSVFRUB0telymWBNVFAJwhpYDo5OittwAQW8OxfhZcjzGrW3HoOv7zeyzKNab/Rvi72uGBjr4Y1NEX9wS5Vdv/wSOdkV9SjgqNgEYroBECFVoBjUZaanXPtVp0Da5+/It9WyOnuFw6Visfo606Rv+5n4udwbG5xeVo7+eMG5lFyDfRiq6jFUBCdjE8HQ37eWQWliHq/V3wcFSjhYdDVQu3hwNCPB0R4ukAbyfbasG8NQkhkF9agayCMmQVlUnLwjJkFpYhu6gMmQVlyCosxRPdgzG0k2EsM3n5URy7nl0ZRJdptDVcBfhyXDc8dI/h8T+eK0PGsRNm1XNYZ79qQfX8HReRIH+paevjhJ6tPHFvKw/c29IT3s63fpAd7GEPdwcbZBeVI7+kAvGZRWjp5Vj3geyuoUs5MFdqaqoNAOhGcmhMAwYMyN+wYYPntm3bXHRB9d9//+0SHR2dHxQUVPbdd9/55ObmKlxdXbVbtmxx1mq16Nu3b2Wwv2bNGrebN2/avvDCC6n6ATUAhIaGlr/yyisps2fPDv7tt99camut/uGHHzw1Gg0999xzabqAGgAUCgXmz5+fuHXrVg+NxnRDTbdu3Qr0A2pACjzffvvtFidPnrTaL6NKpRIqlaraP7ma8sO//PLLm7qAGgACAwMrBg8enLN+/XrP06dP23bv3v2WGI0jJSVFuXHjRs/w8PAi/YAaABwcHMS8efMSoqOjOy5fvtwzOjo6AQCWL1/uCQDvvfdeoi6gBgBfX1/NjBkzkl999dVQa9TNkqD6GoB7AFR+kxVCZBBRFgB3AL1RPajW5f9wz1nW5NLzS6sFO219nKoF1K28HTGwoztaBWYiRxuH42mr8c6xC3A544KWri0R6hIqLV1D0c2nW4Na98b2aGHxsZ5Ottg87X4IIZBTVI71O/+EZ2gYbmYV4XpmEa5nFeFGZhFS8krgbKeCm4NhS+31zCIAQFahFMievJlT7Rq2KgVaeDigra8TFj9Vd1eICo0WOcXlUmAsB8hZRWVo5+OEe/XGRgaA6f87id9PJdX4hUZfr9ae1bblydcxR5mJf2qqemTElFUYBuw3s4oqA2oAuJRWgEtpBVgh59239nbEva080bOVJ3q29ICP0ReiWwERoXOQG/ZfTAcAnE7I4aCaGRBCHKtpX2BgYOekpCS1UXkCACJq9E4eDz74YN60adOwZ88e53fffTctJibGLjMzU9W/f/+8Fi1alC1dutR3+/btTmPGjMnbvXu3CwAMHjy4soX34MGDjgBw8+ZNtakW+cuXL9sCwLlz5+wA1BhUnzp1ygEA+vTpU63jXbt27cp8fX3LjN8nnYiIiCLjbba2tsLT07MiNzfXKkMOPvzww5ny8HXhI0aMyO7fv39+//79Cz09PU1G+k5OTppOnTqVGm8PCgoqA4DMzExL4sVG8ddffzlqNBoQkcm7KuXl5QQAFy9erPwDfO7cOUeFQoEHHnig2s9ryJAhZt8BqIslb9JRSEF1FIDf9LbvBvA4gBlE9IsQIgsAiKgVgLcgtWyfbFBtGTODEAIXUvKxPTYF286m4Ep6AY6+Oxiu9lXBZfdQD3g6qhHgZo+hnfwQ7J+CNVeX4JeMc6hIM/wSn1Oagxv5N7AP+wAAnnae2PvEXoMyyQXJOJ91HqGuoQh2DoaNovFTK4gI7o5qtHJTol+XwGr7S8o1SM8vrRb8ZxaUQq1U1NrCW1qhxaW0ApNpLrvOpWL2xrPwcbFDXokU4OYWl8NURsvE6NBqQbWSyKyAGqge1AKA2kRUrFIQ1CqF9FAqKtftbar/ievspURkK2/YKAk2ctmqJUGtVMJGRVArFQjxNGylVqsUeLl/Gxy+lomTN3OqvY4r6YW4kl6IVYdvAABOzX4Arg63XppNRJBrZVB9JiEXo0x8fliVN7q/kdSQ9Ir/9v3vdeOUkPr4bsh31YYgu5X4+fmVAUBCQkKjD3/WunXr8pCQkNIjR444V1RUYNu2bS4AMGzYsPzg4OBylUoldu3a5TJmzJi8v/76y9nJyUnTp0+fyg4sWVlZKgDYunWr+9atW2u8TkFBQa1fv/Pz85UAEBAQYLJ13svLq7ymoNrNzc1kYKtSqYRWq7XKLcJvv/32ZuvWrUtXrlzptXjxYr/Fixf7KZVK0bdv39zPPvsswTiAdnFxqalOAICKiopbJkcsPT1dBQBnz551OHv2bLWOpTqFhYWVP8OCggKli4tLhakUluDgYKvdYbEkqN4J4DkAIwHM1tu+CFJQ3QrARSLaA8ARwH2Q8qwFgKUNqi1jNdBqBU7czMb22FRsO5uCG1mGDQF7LqTh4a6BSClMwdmMsxgUMgh7ZvSrzLeNzSzF6cOnzbpWS9eW1bYdSDqAOX/PAQCoSIUg5yCEuoaipUvLytbtli4t4Wbn1rAXWg92NspqqQsA8EC4Hy68NxQpeSW4nlkktXBnFeJ6ZhFuZEkPXUfNFp7Vj4/PLERSbgmScuu+E2iqRdlDTkdxUCvh4aiGp6Ma7o5qg3VPRzU8HG0R5ld9voNPn+gCrZACXFs5iK5PqspTHWzRr183s8vr83Wxw7+GSJNuFZdpcOJGNg5dzcSha1k4eSPH4ItKmJ9ztYD65M0c/HDgGnq09ESPlu5o7e3ULPnMnfWG0TudWGd/LMZq1bt374K1a9d67d+/3wWA1XK7axIdHZ33888/e+/bt89xz549Lv7+/mXh4eGlAHDPPfcU7t+/3+X69es2165ds+vfv3+ujU3V76EueFy5cuXlp556yuIPv5OTkwYAkpKSbABU+2OYkZHRrN+mVSoVZs2alTZr1qy0xMRE1a5du5zWrFnjsXXrVvcHH3zQPi4uLtbSTqNKpVIAQEWF6fTD/Px8ZW0dNRtK96XkueeeS/32228TzDnGyclJk5eXpyotLSXjwPrmzZtW+1lZElRvArAfgJKIWgshrgCAEOIAEf0HUqDtAWC0XF73H+N7IcSqhlaYMZ2yCi0OXc3E9tgU7DiXivR84ztXAgp1OtRO8Vh+eSuWXLmEpELp7/2ux3bB19G3smR79/ZwUDmgqKIIrVxboZtvN3Tz6YYu3l1QVFGEa7nXcC3vGuJz49HWvW21ulzLvVa5XiEqEJ8Xj/i8eOzFXoNybrZueLTto/hn5D+t9TZYRKEgBLjZI8DN3mSKRW5xOW5kFsFUvKdLHzHm5mADD0c1PBzkANlJja4tqo/cNG1gW7w2uB3sbCy7y+l5i3QOtFcrEd3Gq3JClZJyDU7cyMGhq5k4fC3TZC7+nxfTseFkEjaclD6HHo5qdA91ryxbVFaBojINiss0eDwqCJEhhn2dEnOKoSSCn2vD0kr063Y2MZcngWENMmHChOx///vfwSdPnnTcsGGD88MPP1zj7fTi4mJqyAggADBo0KD8n3/+2XvLli0uMTExTsOGDascyaJPnz75n332mf8PP/zgDgD9+vUz6DzZq1evwm+++Qb79+93bkhQHRERUbRz5063/fv3O40cOdLg9V68eFGdmppqlVZ7XQBbU362OQIDAyueeeaZnGeeeSanV69e7Q4dOuR89OhR+/vvv9/0H/M66FJITN2ZOHv2rG1jB9V9+vQpVCgUOHTokNmzTHbs2LHw77//dtmxY4fTiBEjDH5e27dvt9pslfUOqoUQRQD61bDv/4joTwDPAwiXz38JwHIhxLoG1JOxat7+9QzWHdf/kqqBwi4JSod42Dpdh9rxOsrlgWguGf3pOJ52HMNaDqt8rlKo8NXgrxDqEgp3u+qBYHuP9rXWJcQlBL38e+Fa3jWkFKbUWC6nNAcC1f+fLN71Gi6nnUConS+C7D0RaOsBPxsXeKud4SAAlBcB5SXS8t4pgJvlOdnmcLW3Qecg04MNzBreEZN6hyKzsAyu9jZwd1DD3cEGKjOH73O0vWVS86zKzkaJXq09TX5J0TliNGZ5VmEZtsemYnts9dFbugS7VQuq5227gK1nUjCmexBe7NcGgW6WjQTl52oHH2dbpOWXoqhMg6vpBWjre9vPgsyaibu7u3bu3Lk3XnrppZYTJ05svWzZsquPPvpotZFAdu/e7fjKK6+0OHfuXOVY1dOnTw9YuHCh/2uvvZZs7sgPw4YNyyci/PDDDz4FBQXKAQMGVAZJgwYNyvv000/9Fy1a5A9Uz5cdN25czr///e/S5cuXew8YMCDfVGfEXbt2Od57773FxqNm6Js4cWLWwoUL/ZctW+YzderUDL3RP/D6668HNiQI1ufp6VkBAPHx8WYH6cXFxaQLHhV6o0+VlpZSTk6OCgAcHR1rzv+rQ0RERImTk5Nm586dbomJiarAwMAKACgoKKB//OMfwZae11yBgYEVI0eOzNywYYPnjBkz/D/88MNk/bsRABAbG2urVCpFWFhYGQBMmDAh8++//3aZPXt24MCBA/VH/1B+8skn9eqYWxur/3cTQuyGlF/NmFXkFJXhUlpBtbGe+7X3NgiqXVp9BWF7s/K5qSQpe5U97vG6B4421TtmdfXparihohTIvAKU5kvBbIUc1Pp2BrzbGRQdk5WOMSW2AIWgQu2JkuJslJXkoKIsH9qyIigqiqHUlMNWq0VscCaMPRazGj6l5jUapAZHwVcvqNYKLc5tm44SOxdkebVCkb0bNFoNtEILjdBAI+R1edsTYU/AXlUVjOWV5eG7M99Vlq88TqsxeK4V0t/guffPRStvJ7Tylo7PL8vHyvM/w9XWFW62bnCzc5OWtm5wUbtAqbBKv5s7wptDw9C3XSYOX8vC0fgsZBfVnMpXVGb4T/lKegF+P5UErQBWHrqBNTE38Wi3ILzUr43JNJ263BPkil3n0wAApxNyOahmDfLiiy9mFRcXK95+++0Wjz32WNuwsLDiqKioAnd394rMzEzVsWPHnOLi4uyNxx/WTSxiapSKmvj7+1e0a9euOC4uzh6QOi/q9g0cOLDQzs5Om5WVpXJ3d6/o3r27wexZtra2Yu3atVeGDx/eduzYsW0+/vjjwvDw8CIHBwdtYmKi+tSpUw4JCQm2169fP1VbUB0eHl46Y8aMpI8++igwMjIyfPjw4Vm6capzcnJU7du3L7548WKDxz9+4IEH8mfOnIn33nsv6OzZs/bu7u4aAJg3b16N40YXFhYqHn744XYBAQFlXbt2LQwODi4tKSlR7N+/3+Xq1at2AwYMyOnWrZvFI3nY2tqK5557Lu2zzz7z79q1a8ehQ4fmVFRU4M8//3Tx8fEp9/b2bvRRYJYtW3bj2rVrdp988knA2rVrPbt3717g4+NTnpycbHPp0iX7s2fPOnz99ddXdUH1Cy+8kLV27Vr3P/74wy0sLCx8yJAhOeXl5bRlyxb3zp07F968edMqt0DrHVQT0TR59ZQQYp81KsGYsZTcEuw4l4LtsSk4dDULznYq7PpXd5zNPIXjqcfR1r0t+ocNQytvR/Rr54OhnfywNfkYfrl00+A8brZu6OrTFZG+kejm0w1hnmGmOxHmJQEpZ4DUs0BqLJB6Dsi4CAgTrQ2D36sWVCNuKxD/JwDpl8qp+lGV7nFtVW1boanr1CA55yp89Z6T0CD08HdwkHsKpiiVOG2rxmk7W5y2VeO8Wo0SvdaK4a2HGwTVReVFWHZ2mdnXn3v/XIPnKYUpmH9svsmyBIKz2hnudu5wtXWFr4MvFvRbYFAmuyQbl7IvwdXWFe527nCzdYNa2ej9nZpFp0BXdAp0xfP3t4JWK3AlvQCHr2XhanohbG0UcLBRwl6thINahcgQwzsmxWUaRAS74YQ86VC5RmB1zE2sPZaAvu280TXYDRHBbogIcjOrc2TnQLfKoPpMYi4ejQyy+utld5fp06dnjBw5Mnf+/Pk++/btc9mwYYNHcXGxwtnZWdOuXbviOXPm3Hz55ZcNpvGMjY21VygUGD9+fPWpZ2tx//3358XFxdm3bt26pEWLFpWBup2dnYiMjCw4cOCAS69evQxaanXuvffe4pMnT5774IMPfHfu3On2yy+/eBIRvL29y8PDw4vefvvtJHOmJZ87d25KUFBQ2RdffOH3yy+/eDk4OGj69OmT99lnnyUMHjy4naOjY4Obq7t161by+eefX1u0aJHfihUrfEpLSwmoPah2dnbWzpw5M2H//v0ux44dc9y5c6ebg4ODpkWLFqUff/zxjWnTpjV4KtUFCxYkOTg4aFesWOH1888/e3l5eZWPGjUq+7///W9SWFhYeEPPXxcPDw/toUOH4hYsWOC1du1az23btrmVlpYqPD09y0NDQ0vnzJlzc+TIkZVfthQKBTZv3nz13Xff9Vu9erXXjz/+6OPt7V3+xBNPZM6bNy/JwcHBso42Rqi+0xwTkRZSp8PHhRC/WqMSt7KoqChx9OhRi47du3cv+vXrZ90K3cGuphdg8e8HcanYEadu5oBUOVA6xEPpcA1K+3go7apukd8XeB+WDFpicPymq5uw6PiiynzoSN9ItHRtCQXp/VEtzQfKigBnX4Njsfop4MIm8yra7x2g35uG2356HLi0w7zjH1oAdH/OYFPFmvEoz7yEIgAF0CBPW45sTQkyNUUohBalpEAJEUqI8PjDK9AipE/lsUd+/w49jr1W4+XKAVxS2+C0rS1O2drinxP2wdel6g5delE6BqwdYF7dAZyacMrgPY1JicGz258161hfB1/senyXwba9N/filT9eMdhmr7KHm60bvO29EegciGDnYAQ5BaGla0t08elidl1NuZ1/L4UQOHA5E4t2X6qWSqLv4S4B+HRs1xr3A1Ln3Uk/xAAAurVww68v9a6xrCXvGREdE0JE1eugZnLq1Kn4iIgInrO9iWm1Wnh6ekb06tUrf8uWLVebuz7WkpWVpQgICOgSFhZWdPLkyQvNXR9mfadOnfKKiIgINd5uSfpHNqRZFa/VUY6xOp1NzMW2s1KL9OXMFKicY6F0iIdj62tQqHNqPO5k2klotBqD1IIHWz6I4a2GS0+0GiDrGnDuN6nlOe2c1AqdHQ90eQp4eLHhCX3DTQfVbiGAkw9gYw+o7KWlcSs1AHR/Hmj/oLTfxh6wcQBUdtLSxs5wm7p6O7bqiZVQAbAHoJ+RqxVaJBcmIz5X6vh4Lfca/IN6GRyrVTlid4sIhOSlITg/HbYawwYWGwAdy8rRsawco4vKUGJ0fUdB+Ff4ZFTYOkJJSihIAaVCWbWut83gC4rM094T4zuMR25pLnJKc5Bbmovs0mzklOYgv8ywv5KbrVu147NLsqttK64oRnFFMZILk3E6o2pUlnDPcKwevtqg7Jn0MziZfhJBTkEIdg5GoHOgQUv8nYSIcF9bL9zX1guHrkrB9cEr1dOJfM3oyKifMx+blMedFVmTO3r0qH1OTo5q5syZFk+B3ZySkpJUnp6eGv3RJMrLy/HSSy8Fl5aW0vDhw6v/cWN3NEsnf+kKw//9jNVLhbYCKoUK/9l0DkeuSS1uCvtM2Pmvr/EYFanQwbMDuvl0QzffbtBCCyXkoDr7OhRxW6vSN9LOAxXFpk+UamLCz8BIIKS3FFz7dAR8OwE+YYCtmXmm7YaYV66eFKRAoFMgAp0C0TvQdEtiib0vBj67X3qiqQDSLwAJMUDCUWmZEVdZVh18L9RGHTEdru7FM5tmAV7tgaDuQFAkENQV8O4AKOv+E9HKtRXe7PGmyX0V2grkleUhpyQHOaU5JoePc7V1RaRvZGWZ3NJcVAjTd16DnKunKPyZ+CeWnDK8a+Ft740g56DKQDvIOQidvDoh1CW0ztdzu+gpTzJzJb0Ax69n43RCLk4l5OB8ch66mBh5xJiXky3cHGyQU1SO0gotsovKbstp19ntq0ePHsW1TTRzq1u5cqX7Rx99FNC7d++8oKCgsuzsbNWhQ4ecr1+/bhsWFlb81ltvpTV3HVnTsiSoXg+gG6RxqnfVUZYxlFZosCcuAZdyY6FyuI5jacdwOv00tozegqHhfpVBtboiGCRsIEjq46DrVNjNVwqi73EPg0NOApBzA2hhlK6Qdg7YZjqwM6BQAaQAhIDBeHHthjRaYNyklCrAr5P0iJokbSvOAZKOS0G2q4mO2QlSCgAy4qTHyZXScxtHILAbEBQFBEZJAbdx2kwdVAoVPOw84GHnUWOZAS0GYIDez1MIgcLyQmSXZiOlMAUJ+Qm4mX8TCfkJiPStPrtjQn71YUrTi9ORXpyOE2mG05GPbD0SgzG4Xq/hVtfa2wmtvZ3weJT0sy0p15gcCtEUT0d15ZjkmQUcVDNWH7179y6IiooqOHLkiPPOnTtVABAYGFj6yiuvJM+ZMydFf8pvdnewJKheBOBZAC8Q0SYhhJmJpOxu8+mB37H96p+4WRQL2CZASsevcjLtJB4Ij8aZxFwMCfdDn3ZemL15Hzq1Cce9TsFoW1YGVXoccOU4cHCF1HFQWyEFxu8kAyq9zmy+JvpFOPkBvh2lfb6dpKVXO0B1lwUO9m5A6wHSw5SSPOk91Rq1DpcXSp0v5Q6YAID7XwcGzkZjIiI4qZ3gpHZCsHMwuvt1r7X8/UH3w15lj4QCKfhOLkiusaW7g0cHwKjt6IsTXyC/LB8R3hHo4tMF/o7+zTIhi7UYj/+dlleCzWeSMTE6tNrr8nS0xZV0abK5zMJSADwCCGPm6t27d/GOHTuuNHc92K3DknGq84noAQDrAGwmoh8B/ATgFIBsUd+ej+yOtSZuNQqUp0E1pHfGZcdhUMggLHyiC1BRBux4B+/G/w23uLVASU7NJ9ZWSAG2X6eqba7BQOQkKWj2DZcejl7WfDl3rkeWAA/NB5JPyWkjcupIvokhY71MjNd9YJGUM956AODRCmY3k1rJsJbDDMYcr9BWSC3cBVUt3FdyruBU+il08emCjDTD/mibrm5CYkEiVl2Q5qbysfdBhE8Eunh3QRefLujg0QE2ytpH09Bd80b+DdzMu4kb+TdwI/8GEvITYKe0w5j2YzCy9cgmH17wvU3nsOLQdZRVaNHezxnRrQ1/Jzydqr6YZhZUn/2SMcaY+SwZUk9/iBgCMEl+6PbXdrgQQtyZMz+waiK8u+JAltTJTCEEOmi98aBzAHoq1XDv/gK8g/U63CltgLO/wK2oeqcrA24hUquz8QQqRMCIT61a/7uK2gEI6SU9dHITgcSjVUF20gkpFUSfVgP8tQAolvvjuLUAWvWXAuyWfQCHmtM+GotKIU0TH+QchJ7+PSu3CyEgILAf+yu3pRWlIbEg0eD4tOI07Ly+Ezuv7wQAqBVqhHuFo4t3FzwT/gw87au6k2y8vBHfnPkGifmJNbaOA8DZg2ex9uJarBi2okkD6+JyDcoqpDtEi/dcqRZU66aMB0xPKc8YY8x8lgS4xlHz7XuflDWqV52UGHPVCfegFJ4FSaDymwCOSztD+wP6QTWR1Lp8TQ54bF2qWpx9wwGfcMCnA2Dn0uSv467lGig9Oo6SnmvKpTQRfcknqwJqQMp3P/6j9CAFENBVCrBb9ZdyslXNN/40EYGM/ly52rriq0Ff4WT6SZxMO4nT6adRVGE4CU+Ztgwn0k7gRNoJPNfZcChErdDiet51s64f5RvV5C3VU/u0xuojN6AVwF+XM3DiRrbB1PH6U75nFpQ2ad0YY+xOY0lQPcfqtWB3pA45ceiQfc70TlMjcNz3Gs449kHnQWMB16AmTyNgdTCVAuEaDAz7L3B1j/SFqKygap/QAonHpMf+/wL27sDrcbdUTrut0ha9A3tXjqyi0WpwOecyTqWfwsm0kziZfhI386UJhVq7toarreHU7S1cqma29Lb3RrBzMIKdg9HCpUXl2Np/Jf6F9ZfX4/l7nq92/ZKKEtip6h7+zlItPB0wMiIAG05KqTxzt17Amhd6Vt5R9NRrqc7klmrGGGsQS3KqOahm5jHuPOjoU9XybKrTXOsByLypANxMjFDBbk1OPsC9L0gPTbmUKnLlD+DKHmnEEaHXOdWnY/WAOjUWSI8DWvVrllQRY0qFEu092qO9R3uMaT8GAJBRnIHT6adRYdyRE0BHz45YN3IdgpyC4GBjeqrwzt6d8Xzn56vlZScXJOOJTU/gyQ5P4pmOz9R4fENNG9gWm04no0IrcORaFrbHpmJoJz8AnFPNGGPWxPnNrPG0HggMUVSlbzh5N3eNWGNS2gAh0dJjwLtSWsi1/XKQ/QfQun/1Y06uAv7+AgDJqSJyPnZQj2ZNFdHnZe9lMOSfPnuVPdq5m5gMyIipjo6LTixCdmk2Fp9cjF/ifsGc3nNwX+B9Da6vsVbeTni6Vwi+PxAPAJi79TwGhPlArVIY5FRLo38wxhizFAfVrPH4dpQe7O5k7y7lY3ccJY0LbqKlF1f3yitCatlOOg78OV8aIzv0PiC4O+AcII2PHRglDQ94ByipKMHF7IuVz9OK0zDtj2lYMmgJ7vW/1+rXe3VgW/x6PBG5xeW4nlmE5X/H4/n7WxmMS83pH4wx1jDV5xxmjDFrI6qek63VAmHDpQ6MxtOflxcCl7YDf7wPbHwJWPmoNFOksbWTgE2vAXs/Bo79CFzcDiSdBPJTpJFJblF2Kjv8b/j/MCd6DrzspRE5yrXlmPbHNMRmmOhv0EBuDmpMG9i28vmi3ZeQXVhm2FLN6R+MMdYglgyp16CZH4QQ/2nI8YyxO4RCAfR/W3oUZwPX/qxKFckxMaKGk9FsjqUFQOyvNZ+fFFIev7OvNBRjz5cMhwxsZkqFEqPbjkZP/554euvTSCtKQ1FFEV7c9SJ+GPYDWrm2sur1nu4ZgpWHruNaRiHySirw2e5LmDW8I4ikGwm5xeUo12hho+S2FsYYs4Qlfz3/D8C/G/BgjDFD9u5Ax5HSWOP/PA1MOwEMXwhEvwJ0HiONee3sZ3hMQWrt5xRaoCBFmtTm/G9A6tlGq35DBDgFYOngpZUji2SXZmPKzilILki26nXUKgXeHhYGAGjr44QBYT5QKgjuDlWt1dmcAsKY1QQGBnYODAzs3FTXmz59egARRW7atKnRp0b9+uuvPTp06NDR0dGxKxFFPvvss5UjDPz6668uXbt2DXNxcelCRJGDBg1qDQA9evRoT0SRjVmvRx99NJSIIuPi4pqlU46lOdX1GetM1LM8Y+xu59FKetTG0Qt4/AcgP1UKnvPlR0GqtCzOqirb62Wgx+RGrXJDtHZrjcUDF+P5Hc+juKIYKYUpeGHnC/hx2I/wsLPeqCiDO/riq/HdMKiDL1Ryi7Sno7py4pfMwjL4uDTeEH/s1qYLeIQQx2oqExgY2DkpKUl94cKFM+3bt6/2Lezy5cs2CxYs8Nm3b5/LzZs3bYuKihQuLi6asLCw4mHDhuW89NJLmZ6enrdubhar065duxxffPHFlkFBQaVPP/10uoODgzY6OroAAOLi4tTjxo1r4+zsXDFmzJgMFxcXbVhYWHFz17mpWDKkXp2t20TkACAMwHgA/wBwBMBoIUR6vWvIGGOm2LkC4Y/UvL+iFChIA7KuAC2im65eFrrH+x582v9T/GP3P1ChrUB8Xjy+OvUV3rn3Hatdg4gwtJO/wTbOq2bWsmDBAq+33367RVlZGbVv37545MiRWe7u7prMzExlTEyM8+zZs4MXLFjgn52dfaq568os99tvv7kKIbBs2bJrgwcPLtTft3nzZpfS0lL69NNPE6ZOnZpV0znuVI0y+ocQogjS1HnHiWgTgC0AthJRtBCC/2ozxhqfylYa89zUuOdlhcDOfwP937klxsfWiQ6Ixkf3f4QZ+2agX3A/vB71eqNf03AEEB5Wj1nmq6++8nj99ddDXFxcND/++OO1sWPH5hqX2bFjh+O0adNCmqN+zHqSk5PVANCiRYty431JSUk2ABAUFFRt392g0XukCCH+APANgK4AXmns6zHGWK2EBlj3PBDzDbBsMJB1tblrZGBI6BB8+8C3WNBvAWyVNc8+mVOSg9+u/IbZB2bj5ws/QwhR72sJIZBbXPW/j1uqmSWys7MVb731VgsA+P7776+aCqgB4IEHHig8ceLE+YZca9OmTc5EFDl9+vSA/fv3O9x///1tnZ2du7i4uHQZMmRI68uXL9sAwLlz59TDhw9v5e7uHmFnZ9ft3nvvbff333/bG5/v9OnTti+99FJgp06dOri7u0eo1epuAQEBnZ988smQK1euVBtgXqvV4vPPP/fs2rVrmLu7e4StrW03Pz+/e+67776233zzjbs5r+Grr77yUKvV3Vq1ahV+8eLFOnN/Dx8+bD9ixIiWgYGBndVqdTd3d/eIjh07dnj22WeDS0tLTabXfv/99+6dO3fuYG9v39XV1bXL8OHDW127dq3a66kt79s4R3vRokWeRBT5yy+/eAJAWFhYZyKKJKJI3b758+cHAMCIESPa6faZk+O9bt06l759+7bR/QyCg4M7TZkyJSgjI0NpqvyGDRucIyMj2+te36BBg1qfOHGi2XPXmmqc6o2Q0kDGAZjfRNdkjLFq3HJigbgt0pPMy8C3g4An10hjYt8ievj3MLn9Zv5N7LmxB3tu7sHxtOPQyjNWrr+8HtfzruPN7m9WTkFelxuZRfjnmhM4fiOnchu3VDNLLF++3D03N1cZERFROHr06Lzaytrb29f/258Jx48fd1i8eLFf9+7d85988smMc+fO2e/YscNt0KBB9uvXr788YMCA9q1atSp57LHHMm/evKnesWOH+0MPPdTu2rVrZ1xdXSunel29erX7ihUrvHv27JkfFRVVoFarxYULF+zXrFnjtWvXLtcjR46cb9myZeU3z2nTpgV++eWXfoGBgWXDhw/PdnV11aSkpNicOnXKcd26de6TJ0/Orq3e7777ru+HH34Y1LVr14ItW7Zc9vX1rTW//PDhw/Z9+/btQERi4MCBuSEhIaV5eXnKa9eu2a5YscJ74cKFiba2tgbv6eLFi713797tNnDgwJzo6Oj8Y8eOOW7evNn93Llz9rGxsecs/RlERUUVvfbaa8lbtmxxi4uLs580aVKam5ubRn/fX3/95RwTE+M0evTozJCQkDIAaNu2ba1/WP71r3/5z58/P8DV1VUzYMCAHG9v74rY2Fj7pUuX+u7evdv1yJEj5z08PCp/Zt9//737888/38rGxkY89NBDWX5+fuWHDh1y6tu3b1j79u2bNX+7qYLqDHnZuomuxxhjJuW43yN1cPx1CqApBYoygR+HA6OXShPV3KL+te9f2B6/vcb9P53/CQDMDqxLKjQGATWAyg6LzMi2twNwaLF/3QUBdHo0A499Zzgm5C/PhuDsOi+zju/5UjKGzk0y2PbD8DaI/9O1Xsc0wPTp0wNq2peXl1et5fDAgQNOANCnT59aA2pr2rdvn+vixYuvvfjii5V5u2PGjAlZu3atV79+/cKmTp2a8vHHH6fo9s2YMcP/k08+CVi0aJHXrFmz0nTbJ0+enDlr1qxU40Dz119/dXn88cfbvvvuu/4//fTTDd32lStXevv4+JSfP38+1tnZWat/THJyco0xlUajwbPPPhu8fPlynwceeCBn/fr1Vx0cHOoMbr/99lvP0tJSWrFixdXx48fn6O9LT09XOjk5aY2P2b9/v+uff/55vkePHpUB5ogRI1pu2rTJ46effnJ7/vnnaw38axIdHV0cHR1dfP36dXVcXJz9m2++marfWTU6Orp4+vTpATExMU6TJk3KHD58eH5d5/z999+d58+fH9ClS5fCnTt3XvLy8qr8krFo0SLPV199NXTGjBmBy5YtuwkAubm5itdeey2EiMSOHTsu9OnTp0hX/rnnngv+7rvvfCx5bdbSVAOStpSXJpvxGWOsSYU/AjzzO+DgKT2vKAH+9wxw8HNp0OZbiBACC44uqBZQEwgR3hG4169qBsZV51chNtO8yWNaeTnCVmX4LyCD0z8YgIULF/rX9CgoKKj2fzw1NbXJ82i7detWoB9QA8CkSZMyAcDJyUnzwQcfpOjvmzx5ciYAnDp1ykF/e8uWLctNtdyOHj06r3Xr1sX79u2r9mVGpVIJlUpV7Rh/f38T08YCRUVFNGzYsNbLly/3mTBhQtqWLVuumBNQ63NwcKgWPHt7e2uUyuph1bPPPpuqH1ADwAsvvJABAEeOHHGsz3Ub26JFi3wA4JtvvonXD6gBYNq0aZlhYWHF69evr+z4smrVKrfc3FzlyJEjs/QDagCYN29ekpOTU7OOLNPoLdVEpAbwhvz0Ym1lGWOsybS4F3huJ/DT49IIIRDAjneB7Hhg6MeAsqlu5NVOIzRILEgEAKgVavQM6IkBwQPQN7gvvOy9UKGtwJv738TO6zvxwX0foJNXJ7POq1IqEObnjFMJVemvGQWc/sHMG1LPqDwBABE12TfSLl26FBlvCw4OLgeAjh07FqlUhr+/LVu2LAOqOtLpaLVafPXVVx4rV670On/+vH1+fr5Ko6mKy2xsbAxe08MPP5z5448/+rRv3z58xIgR2f3798/v379/YU3DBJaUlCjuu+++9idPnnScOXNmwvvvv1/HAPuGxo0bl/X999/7PvXUU62HDh2aPXDgwPz+/fsXhIeH1/jL2r1792rvje715+Tk3FKNmydOnHBSqVRi1apVHqtWraq2v7y8nLKzs1UpKSlKPz8/zfHjxx0AoE+fPtVawT09PTUdOnQojomJcWqCqptkyYyKLcwopgDgDqA7pM6J4ZDGq/65vtdjjLFG49kaeH4X8POTwM1D0raYb4Gcm8Bj3wG2zfa3uZJKocInfT/Bjfwb8Lb3hoONQ7X9H/f5GGPTxqK7X/3ywjsGuBoE1Tcyq/0vZgAwdG5Sg9IrHvvuerWUkPqYuOmyxcc2AT8/vzIASEhIaLIJN1xdXasFsboA2MXFxdQ+AEBFRYXB7ZnJkycHf/fddz7e3t7lffr0yQsICCjTtVyvWbPG0/gLxLfffnuzdevWpStXrvRavHix3+LFi/2USqXo27dv7meffZbQqVMng2C3qKhIce7cOQdHR0fNQw89VO/0mP79+xdt27btwgcffOC/bds29w0bNngCQGhoaMlbb72VPGXKlGrD1rm7u9f43mg0mltq3pCcnBylRqOhhQsX1ppelZeXp/Tz89Po0o/8/PxM3hXw8fFp1lFHLGmKuWbhtf4CsMjCYxljrHE4eAATNgIbXwLOrpO2XdoutVqP+LRZq6ZDRAhxqXkkMpVCZTKgLiovgr3KvsYc6/AAF4PnmYVlyC0qh6tDtUECGKtR7969C9auXeu1f/9+FwBWy+1ubImJiaoffvjBp23btsWHDx++4O7ubpBi8euvv1Ybb1OlUmHWrFlps2bNSktMTFTt2rXLac2aNR5bt251f/DBB+3j4uJi9dNJPDw8KhYvXhw/duzYNkOHDm3/+++/XzROW6jLoEGDCgcNGnS5uLiY/vrrL4fNmze7fvfddz5Tp05t6evrW/7www/XmbtcEyJCeXm5yT8QTdGq7ezsrNFqtZSbm3vSnPK6L0wpKSkm49e0tLRm/eNlSU411fORCeB9AEN4jGrG2C3Jxg4Y/S1w33TpuUdrYMCs5q1TA+WW5mLitom1dm7saBRUA8C1zEITJRmr2YQJE7JdXV01J0+edNywYUOtw6cVFxffMi2lFy5csNVqtejbt2+ecUB95coVm7pa3gMDAyueeeaZnC1btlzt2bNn/s2bN22PHj1abdi+UaNG5a9fv/6SRqOhhx56qN2uXbssymu2t7cXgwcPLvz000+T5s6dewMA1q9f72bJuXRcXV0rMjMzVaaG5jPOP28MXbp0KczLy1MePXrUrOHwunXrVgQA+/fvr/Y5y8zMVJ4/f77a+9+ULAmqJ5nxeAbAaABdAPgLIWYLIUqsUWHGGGsUCgUw6N/AqC+Bp9YCjp7NXSOLZZdkY/KOyTifdR5LzyytcQzrDn4uUBj9K72WUdAENWR3End3d60uyJs4cWLrdevWVf+2BmD37t2OkZGRYfrbdGMh1zbiSGPRDfV2+PBhp4qKqmyC3NxcxaRJk0KNUyWKi4tp48aNzlqtYZ/B0tJSysnJUQGAo6NjtQ6FADB06NCC33777SIRYdSoUe02b95sVm7Ztm3bnDIzM2vsHGqqA2N9dOnSpVCj0dDnn39u8Adv0aJFnsePH2/0/Ld//vOfqQAwefLk0Pj4+GqtzHl5eYrdu3dXfgkZN25cjouLi+a3337z2L9/v0HQ/8YbbwSY6kjblCyZpvzHxqgIY4zdErqON7097Tzg06Fp62IhAiE+Lx4AcCn7EvYn7Eff4L7VytmrlWjl7YTLaVWB9LV0bqlm9ffiiy9mFRcXK95+++0Wjz32WNuwsLDiqKioAnd394rMzEzVsWPHnOLi4uzd3NwMcmF1Aaqp0TQaW4sWLSqGDx+etWnTJo+OHTt27Nu3b15eXp7yzz//dFGr1SIsLKz4woULlS2fhYWFiocffrhdQEBAWdeuXQuDg4NLS0pKFPv373e5evWq3YABA3K6detWYwPigAEDCrdu3Ro3fPjwdo899ljbVatWXXnkkUdqzbP+5JNPfA8cOODSo0eP/JCQkDInJyfNhQsX7Pfv3+/q4uKiefnllzNqO74u06dPT/vll1+83nzzzZA9e/a4BAYGlp09e9b+xIkTTv3798/ds2dPzUM5WsGoUaPy33nnncS5c+cGdujQoVO/fv1yQ0JCygoKChQJCQnqI0eOOEdGRhYMHDjwEgC4urpqP/300+vPP/98qwceeCBMf5zqS5cu2UdFRRUcPXq02TrDNNWQeowxdvuKXQ8s7gX88f4tN+SeKW52bni83eOVz785802NrdUd/Q0bFa9mcFDNLDN9+vSM2NjYM1OnTk0hIrFhwwaPxYsX+23ZssXd1dW1Ys6cOTevXLlyRv+Y2NhYe4VCgfHjx1frcNcUVq1adf3ll19OKSkpUSxfvtxn3759rgMHDsw9dOjQeWdnZ4MOf87OztqZM2cmtG7duuTYsWOO3333ne+GDRs8nJycNB9//PGNzZs31zk9a+/evYt37NgR5+joqB07dmyb1atX1xq0vvTSS+kPPfRQdkJCgu369es9vv/+e5+rV6/ajR8/Pi0mJuZcbaOAmCMyMrLkt99+u9i1a9eCP/74w3XVqlXearVa7N2794KpEVYawwcffJCydevWuP79++ceO3bMadmyZT6bN292T0lJUY8bNy79gw8+SNQvP2nSpOxffvnlUnh4eOGWLVvcV6xY4e3m5laxb9++Cy1atGjWIYzIkqlt7yZRUVHi6NGjFh27d+9e9OvXz7oVusPxe1Z//J7VT73fr8RjwHfDpIliAKDz48DwhYBtnTPvNqvUwlQM/XUoKrRSw+B3Q74z2Znx631XsHDXRZSUSy2GHf1dsOXV+w3KWPIZI6JjQogoy2rftE6dOhUfERHRoBY/Vn9arRaenp4RvXr1yt+yZUudASljt4pTp055RUREhBpvr3dLNRE5E9Fs+VHnDFNE5K9XvtnnZWeMsXrxage01Asyz6yVWq0v726+OpnB19EXo1pXzRD57ZlvTZZ7JjoUf705oPL5tYzCGlu1GbOmo0eP2ufk5KhmzpyZ3Nx1YcwaLEn/GAng/wCMF0KY84uQAmA8gH8DGGHB9RhjrPnYOgNPrgEiJ1Vty70JrBwNbPwHUJzTbFWry7OdnoWCpD/zB5MOIjaj+myLdjZKeDqq4Wov9REqLtcgNY8ngWGNr0ePHsVCiGO9e/currs0Y7c+S4Lq0ZAmcvmfOYWF1OSxGtLweo/XUZwxxm49SpWU8jH6W8Beb+jaEyuBxT2BuK3NV7datHBpgSEhQyqfzz8232QrNBEhyL1qJKrUPB6siTHG6suSoFrX/f1gPY75W152tOB6jDHW/IiAex4H/nEE6Phw1fb8ZODnscC654Hi7GarXk2mRkyFkqRRpmJSYrDrxi6T5dwdqobkzS7iKQUYY6y+LAmqg+RlfWZN0qWJBFpwPcYYu3U4eQNjfgTGLAccfaq2X/8boGYdItWkVm6t8GTYk5XPP4n5BKUaw/SOtPwSFJZVjXTGQTVjjNVfQ4bUq8+sNboOirXOTsQYY7eNjqOAfxwG7hkrPR/xGWBncs6LZjc1Yircbd0R4BiA16Neh1ph+Kf4xI0cnLiRU/k8u7C8iWvIGGO3P0uC6hR5GVGPY7rIy1QLrleJiB4ioh1ElEBExUR0lYjWElGvGspHE9EWIsoioiIiOk1E/yS6BZuTGGO3HwcPYPTXwNS/gLaDqu+/fvCWGNfa1dYVSwYvwcaHN+KB0AdAZDiNYs+WngYzK97MuvvGquYRTxhj5qjtb4UlQfVBSJ0Op5DxX2YT5DJTIXVu/LuO4rWd52MAmwB0A7ANwGcAjgMYBeAAEY03Kj8KwH4AfQCsB/AlpJbyhZA6TjLGmHX4da6+LW4b8P0wYNUTQPb1pq+TkXDPcNipTI9q6upgg1DPypmAcSap1kne7jhElF1WVlZtimTGGDNWVlZmQ0QmO9BYElSvkJf3AFhSW6svESkAfCWXBYCfLLgeiMgPwL8gtXR3FEI8L4R4SwjxGIAhkIL8/+iVdwHwDQANgH5CiOeEEDMgtZj/DeAxIhprSV0YY6xOxTnApn9K65e2A1/2APZ8CJQ1yQRlFokMca9cv9umKtdqtVtzcnJu7dl8GGO3hJycHGetVmtyyKd6B9VCiN0AtkIKZCcDOE5EzxNReyJykSeHaU9EkwGcAPA8pFbq3UKILRa+hhC5roeFEGlG9dkDIB+At97mx+Tnq4UQR/XKlgB4V376ooV1YYyx2iltgA4jq55XlAD7Pga+6C5Ned7MqQZCCJxMO4mDiVWDON3X1qtyPbOwDNmFd09nRY1GszQ1NTUnNTXVo7S01IZTQRhj+oQQKC0ttUlNTfVITU3N0Wg0S02VU1l4/nEA/gDQFUAnAF/XUpYAnAIwxsJrAcAlAGUAehCRlxCicjpZIuoDwBnABr3yuunBtpk4134ARQCiichWCMGzHDDGrEvtCDw4D+j0KLD1DSD5pLQ9LwFYOxEIuQ8Y9jHg16nJq3Y19yre2v8WzmedR6hLKDY+vBEKUqCll6NBuV3nU/F4VHCT1685REZGxh87dmx0cnLyC6mpqcOEEF51H8UYu5sQUbZWq12t0WiWRkZGxpsqY1FQLYTIJaJoAO9BavF1rKFoIYAlAGbLrcQWEUJkEdGbABYAOEdEGwBkAmgNaYbHnQCm6B3SXl5eNHGuCiK6BiAcQCsA5y2tF2OM1arFvcDkP6RJYnb/ByiS2wOu/wV8fT8Q9SzQf6bU4bGJ+Dn44Wb+TQBAfF48DiUfQnRAtME41QCwPTblrgmqASmwBvCO/GCMsXqjht7mkvOX+0PqQKhLwcgAcAzAHiGE1Xq8ENHDAL4D4K63+TKAfwshVumVuwigLYC2QojLJs5zAEA0gGghRLXOk0T0AoAXAMDX1zdy9WrL+jUWFBTAycnJomPvVvye1R+/Z/XTXO+XqrwAIddXIyhhMwjayu2JAcNwqd3UJq3Luqx12Ju/FwDQyb4TpvhMQXGFwIu7qnK+VQrgiwEOsFORRe9Z//79jwkhoqxYbcYYu6VZmv5RSQ6aN8qPRkNEbwD4EMAiAF9AGtovDMBcAD8RURchxBvmnk5emvxGIYRYCmApAERFRYl+/fpZVOe9e/fC0mPvVvye1R+/Z/XTvO/XcCDtArDtTeDqXsDWBYHjFiHQyafOI60pNDcUezfsBQDEFseiTWQbBDoFwmbPVpRrpD+LFVpA6xuGfp39+TPGGGNmaMjkL02GiPoB+BjAb0KI6UKIq0KIIiHEcQCPAEgE8DoRtZIPyZWXrjWc0sWoHGOMNQ2fMODpDcATPwFDPwKaOKAGgFDXUPQO6A0AEBD49dKvICK46aWAEIC4lPwmrxtjjN2uLAqqiaiF/LA1o6ydrrwl15INl5d7jHcIIYoAHIH0WrrKm+PkZTsT9VEBaAmgAsDVBtSJMcYsQwR0GA50farZqjCi9YjK9UvZlwAAHnpB9fLneuC1wdX+hDLGGKtBvYNqIuoNIB7AGQAOZhziACAWwFUi6l7f68l0wbt3Dft123VjQP0hL4eaKNtHrtNBHvmDMXZLuboXWPc8oNU0+qWCnas6IaYUSRPlujlUzX+iqHtuL8YYY3osaanWDY23UQhhckYZfUKILADr5GtZOuHKn/LyBSIK1N9BRMMA9AZQAmm2RwD4BVJnybFEFKVX1g7A+/LTJRbWhTHGrG/Hu8DyUcCZtcDfXzb65fwc/SrXUwtTAcBgBJDsortnnGrGGLMGSzoq9obUwW9HPY7ZAWACgPstuB4gBcm7AAwCcJ6I1kPqqNgBUmoIAXhLCJEJSJ0n5clnfgGwl4hWA8iCNPxee3n7Ggvrwhhj1mejd+Pvj/eANgMB3/BGu5ynnSdUpEKFqEB2aTZKKkrQvaUHVEqCu4MaQe7m3IhkjDGmY0lQHSQvq40BXQvdsHaBtZaqgRBCS0QPAvgHpNbuRyClcGQB2AJgkRBih9ExG4ioL4CZAB4FYCfXY7pcnqfMYozdOvrMAC7tAJJOAJoy4NcXpDGuVXV2XbGIUqGEt4M3kguTAQCpRal47r6WkLqcMMYYqy9Lgmo3eVlej2N0ZT0tuB4AQAhRDuBT+WHuMQcAPGjpNRljrMkobYBHlkqTwlSUAKlngT0fAoPnNNol+wT1QV5ZHvwc/WCntGu06zDG2N3AkqA6E4AfgFAAJ8w8JkRe8hB2jDFWE+92wOD/SFObA8CBz4B2Q4CQ6Ea53Ls9322U8zLG2N3Iko6KZ+XlI/U45lF5ec6C6zHG2N2j+2SgVT/5iQDWTwFKrDYxLWOMsUZiSVC9CVLHwCeJaGBdhYloMIAnIXVu3GTB9Rhj7O6hUACjFgN28txVOTeA7W83b50YY4zVyZKg+lsAqfKxvxPRv4jIybgQETnJU4tvkMtmAPiqAXVljLG7g2sg8NCCqucnVgIXNjdffRhjjNWp3jnVQohiInoCwE5Ik7J8DOA/RHQKQDKkFukAAF0AqCG1apcDGCuEKLRSvRlj7M7W+TEpkI79VRpur9T6U4ZnlWRhxbkVSClMgVqpxpzoxusUyRhjdzpLOipCCLGfiAYA+BnSEHt2AHoYFdNNx5UA4CkhxJ9gjDFmvofmA9pyYNAcwLO11U+v0Wrw7ZlvAQButm4cVDPGWANYFFQD0nB1RNQGwNOQJmDphqrpwjMAHAPwO4CVQgiemosxxurLwQN4YmWjnd7TvmoCmJzSHBRXFMNeZd9o12OMsTuZxUE1AMjB8jL5wRhj7DaiIAV8HHyQVJgEAEgrSkOIS0gdRzHGGDPFko6KFiGi3kS0tKmuxxhjrG5+jn6V6ymFKc1YE8YYu701alBNRCFENIuILgHYD+C5xrweY4zdkYQAYjcAG18GFnUDyoutdmpfB9/K9dSiVKudlzHG7jYNSv8whYgcATwO4BkA96OqwyJBGhmEMcZYfRABf7wPZF6Snl8/CLSpc5oAs3BLNWOMWYfVWqqJaBARrYA0hvUyAH3k8xOkofa+ANDPWtdjjLG7SptBVeuXd1vttL6OVS3VHFQzxpjlGhRUE1EYEX1IRDcAbAcwDoADpEA6EcCnAO4TQgQJIabxsHqMMWYhg6B6l9VO6+dQ1VLN6R+MMWa5eqd/EJE7pGnHJwDortssL/MAuEBK8/iXEOJ/1qgkY4zd9UJ7Ayo7oKIEyIiTpi93a9Hg03L6B2OMWYdZLdVEpCSiEUT0C4AkAJ9DmuyFAGgAbAYwFoB/Y1WUMcbuajb2QEjvqudWSgHRT//glmrGGLNcrS3VRNQFUofDcQC8dJvlZQyAFQBWCyEy9I6xfi0ZY4xJKSBX5GD68i4galKDT+lh54FJnSbB18EXfo5+EELw33HGGLNAXekfxyGlcuj+wl4D8BOAFUKIS41ZMcYYY0baDAK2vy2tX90HaMoBpU2DTqkgBaZHTrdC5Rhj7O5mbkfFfACThBCthRCzOaBmjLFm4NUWcJXzqMvygYSY5q0PY4yxSuYG1U4AviOiv4hoitxZkTHGWFMiMhyf2oqjgDDGGGuYuoLqEQDWASiDlAISDWAxgGQiWk9Eo4moYfceGWOMmU9/aL3cxOarB2OMMQO15lQLITYD2ExEbqgaRu9eAGoAI+VHLhGtBbBcCHGgcavLGGN3uVZ9gUeWAq0HAE7eVjnlwaSD+PbMt0gpTMH9gffj7Xvftsp5GWPsbmJW+ocQIkcIsUQI0QtAGICPACRAar12A/A8gP1EdLWxKsoYYwyArTMQ8YTVAmoAKKkoQUxKDG7m38TZzLNWOy9jjN1N6j2johDiohDiHQChAAYDWAmgGFKAHQpptBAAmE1EbxJRG+tUlTHGWGPo6tMVSlICAE6nn+ZJYBhjzAIWT1MuJLuFEBMA+AJ4FsA+eTcB6ADgQwBxRHSSiN4lorAG15gxxphVudu5o7tf98rnu65zB0jGGKsvi4NqfUKIQiHED0KI/gBaAfg/AFchBdcEoDOAOQD4viJjjFmDVgPcjAH2zAVSzzX4dENCh1Su77i+o8HnY4yxu41Vgmp9QojrQoj/CCHaArgfwLeQxrnWBdiMMcYaavPrwLJBwL6PgPO/N/h0A1sMrEwBOZF2glNAGGOsnqweVOsTQhwQQrwAwA/AUwC4+YMxxqwh9L6qdSuMV80pIIwx1jCNGlTrCCFKhBA/CyGGNcX1GGPsjtd6ACpv/iUeBeL/avApOQWEMcYs1yRBNWOMMStz8ACCoqR1oQV+eAj47RWgONviUxqngKQWplqjpowxdlfgoJoxxm5XQz8CbF2qnh9fDnzRA4hdDwhRvXx5CfDrFGD1U0B+9YDZOAVk5/WdjVFrxhi7I3FQzRhjt6ugKOAfh4Gw4VXbCtOAtROBn8cCuQmG5fd+CJxeDVzYBBz73uQph4QOgYpUuC/wPoS4hDRe3Rlj7A7DQTVjjN3OXAKAsT8BT6wEnPyqtl/cVr0D44HPqtb3zjV5ugdbPoi9T+zFkkFLcH/Q/Y1QYcYYuzNxUM0YY3eCDiOAl48AUc9Jz1tEA10nVO3XagzL+91j8jQONg5wtXVtpEoyxtidS9XcFWCMMWYldq7A8AXAPWMABy9AoddukpdoWNbJt2nrxhhjdzhuqWaMsTtNi56AVxvDbW4tgGc2VT0vK2jaOjHG2B2Og2rGGLtb2OmNFFLKQTVjjFlTvYNqIpogP1zqLl15jJPuuPpejzHGmIWEANLOAxe3S8/VTlX7SvOap06MMXaHsiSn+gcAAsBRAOfMPMZXPk4LYLkF12SMMVYfuYnAsgeAvATA3gOYcdlwTGtO/2CMMatq6o6K1MTXY4yxu5OzP6Apk9aLs4AtM4CeL0l51bbO0oMxxpjVNFVOtVJeVjTR9Rhj7O6mUABtBlU9P7oM+CISyLwEBHQBPFs3W9UYY+xO1FRBdXt5mdVE12OMMdZmYPVtgVFNXw/GGLsL1Jn+QUR9atjVnYi86jjcFkBrAP+ClId9sl61Y4wxZrnWA6pv8w1v+nowxthdwJyc6r2QAmJ9BOC7elyH5HN8XY9jGGOMNYSDh+FztROgUAJaLVBeCKjsAKVN89SNMcbuMOamf5Dew9S2uh4JAP4hhNhglVozxhgzz6A5euv/BywfBfzHHZgbBCTENFu1GGPsTmNOS3V/vXUC8AekVufnAFyr5TgBoARAshDipsU1ZIwxZrmeL0qjf5AS6PYMELe1ah9PAMMYY1ZTZ1AthNin/5yosrH6iBDC3HGqGWOMNQeVLTD4P1XP9YfS4wlgGGPMaiwZp7qlvEy0ZkUYY4w1AVu9WRV5AhjGGLOaegfVQojrjVERxhhjTUB/VkVO/2CMMatplBkViWg4gDEAvCDlXX8jhDjZGNdijDFWD2q9lurS/OarB2OM3WHqPfkLEfUloiQiukZE7ib2/xvARgBPARgCYCqAw0Q0tsG1ZYwx1jD6OdWc/sEYY1ZjyYyKDwLwA3BKCJGtv4OIwgHM1j0FUCwvbQB8S0RBDagrY4yxhtLPqeaOiowxZjWWBNX3Qxoub5eJfVMhBdF5AKKFEE4A+gDIAWAv72eMMdZcOKeaMcYahSVBtZ+8PGti33BIAfcSIcQhABBC/AVgMaRge7AllWSMMWYlnFPNGGONwpKg2kdeZulvJKIWAELkp78aHaMb67qtBddjjDFmLTykHmOMNQpLRv9QyksXo+33ycsCAMeM9qXLS0cLrscYY8xaAiOBf56VOizqt1ozxhhrEEtaqtPkZWuj7YPk5UEhhDDaZy8vcyy4HmOMMWuxsQfcggF7N0DZKKOqMsbYXcmSoPoopPzo54hIAQBE5AlgNKR86t0mjmkjL1MtqSRjjDHGGGO3MkuC6uXysjeAA0T0CYCDkNJBygH8bOKY3vLyogXXY4wxxhhj7JZmyTTlG4nof5BmTLwXQA9ILdcAMFcIkahfnohUAB5Gza3YjDHGmlJRljRGdWk+4NMRUCjrPoYxxlitLE2oGwfgT0iBtR+AJAA/CCF+NFF2LKpGDNlp4fUYY4xZy+fdgGJ57q4ZVwFHz+atD2OM3QEsCqqFEFoAX8qPusquBLDSkuswxhhrBGrnqqC6NI+DasYYswJLcqoZY4zdzmydq9Z5rGrGGLMKDqoZY+xuY8uzKjLGmLU1KKgmov5EtIKILhNRARFVEFFHozJ9iOglIhrfsKoyxhizCoOpyrmlmjHGrMGinGoisgfwPYDHdZvkpfGkLwCgAfAFAEFEh4QQly25JmOMMSvRT/8ozWu+ejDG2B3E0pbqVZACaoI0Gcz8mgoKIQ4AOCs/HW3h9RhjjFmLfvoH51QzxphV1DuoJqIRAEbJT18SQtwrhJhRx2G/QgrA+9b3eowxxqzM1qVqndM/GGPMKixpqZ4oL1cLIb4y85hj8rJjraUYY4w1PjV3VGSMMWuzJKi+F1Lu9Kp6HJMsL70tuB6IaCIRiToeGhPHRRPRFiLKIqIiIjpNRP8kIp4+jDF29+Ih9RhjzOos6ajoJS8Tay1lSNuA6wHASQBzath3P4ABALbqbySiUQDWASgBsAZAFoARABYC6I2qTpaMMXZ3sXUCSCEF18QjqzLGmDVYEuQWAnCDND25uULlZaYF14MQ4iSkwLoaIvpbXl2qt80FwDeQRh7pJ4Q4Km+fBeAPAI8R0VghxGpL6sMYY7e1bs8AkZMAorrLMsYYM4slTRRX5WV4PY7RdWyMteB6NSKiTgB6Qmo136y36zFIqSardQE1AAghSgC8Kz990Zp1YYyx24ZCyQE1Y4xZWa1BNRG1kB/6OcjbIY3k8QoR1dnSTUS9ATwJKQ97ax3F62uKvFwmhNDPqR4gL7eZOGY/gCIA0URka+X6MMYYY4yxu1BdLdXxkFqm2+tt+wJSUBoM4PvaAlMiehrAJkhpJtmQUjKsQp6AZjykfO1vjXbr6nvR+DghRAWAa3KdWlmrPowxxhhj7O5lTk61wT1CIUQKEf0D0oyK4wAMJqItekWmEJETgEEAguTjtQCeE0JYs5v5GEi53ZuFEDeN9rnKy9wajtVtdzO1k4heAPACAPj6+mLv3r0WVbCgoMDiY+9W/J7VH79n9cPvF0DacjgVxEOpKQIJDbI9utVant8zxhirm0WjcQghfiQpH+9LAD4AnkHVFOUvy0tdMF4M4HkhxMYG1NOUF+Tl1xYcW9u06hBCLIXc8TEqKkr069fPgksAe/fuhaXH3q34Pas/fs/qh98vAAXpwCePSev2HsCb12otzu8ZY4zVzeKxlIQQP0JKs5gP4AqkQFX/cRNSqkiYEOLnhle1ChF1BBANIAHAFhNFdC3Rrib2AYCLUTnGGLt78DjVjDFmdZaOGw0AEEIkApgBYIY8jJ0PACWADCGERcPnmammDoo6cQCiALRD1WyOAAC5c2VLABWoGsmEMcbuHipbQKECtBWApgyoKJW2McYYs5jVRv0XQuQJIS4LIeIaM6AmIjsAT0PK015WQ7E/5OVQE/v6AHAAcFAIUWr9GjLG2C2OyLC1upRbqxljrKHMDapN5h43k8cBuAPYYqKDos4vADIAjCWiKN1GOSB/X366pFFryRhjtzK1flCd13z1YIyxO4S56R87iKjcCtcTQojWDTyHroPi0poKCCHyiGgypOB6LxGthjRN+UhIeeC/QJq6nDHG7k62TlXrnFfNGGMNZm5QHWil6zWoxZuIOgC4DzV3UKy6kBAbiKgvgJkAHgVgB+AygOkAFgkhbqXWd8YYa1oG6R/5zVcPxhi7Q5gbVCcBsEZLdYMIIc7DaNzsOsofAPBg49WIMcZuU2q9lmrOqWaMsQYzN6h+QAhxrlFrwhhjrOkYDKun11KdmwCkXQBa9wcUyqavF2OM3aasNvoHY4yx24h+TrV++sfR74GfHgU+vQc480vT14sxxm5TDRqnmjHG2G3KJRDwaCWlgeharTUVwMmfpPW8BB67mjHG6oGDasYYuxv1f0d66Lu8C8hPltYdvYF2pob6Z4wxZgqnfzDGGJMcX1613mUcoLRpvrowxththoNqxhhjQH4qcHFb1fOuTzdfXRhj7DbEQTVjjN3tkk4A89sBQiM9bxENeLVt3joxxthtpq6c6pbyMrGxK8IYY6wZpF8EVow23NZtQvPUhTHGbmO1BtVCiOtNVRHGGGPN4PpfQPH/t3fn0ZIUddrHvw8CzSaICAKCtCDI0ogsIoKv3AZEUNlmRBkOSIOowzAojI7jvIg2gmd0QGWAGRYRmml0cEABZXFhucCgqA00iOLL1i3I1kAjQ8vWdP/ePyKKyludlbXde+suz+ecPFlVGZEZGRWVFRUVGbGw/nzK6rDlfv1Lj5nZOOXuH2Zmk9kOR8C+Z4Dy18E7Pw4rrtLfNJmZjUMeUs/MbLLb7mOwwY6w8AHY9P39To2Z2bjkSrWZmcE6m6fFzMy64u4fZmZmZmY9cqXazMzMzKxHrlSbmZmZmfXIlWozMzMzsx65Um1mZmZm1iNXqs3MzMzMeuRKtZmZmZlZjxQR/U7DmCbpSaDb6drfADw1jMmZDJxnnXOedcb51blu8myjiFh7JBJjZjYWuVI9giTNiYgd+p2O8cR51jnnWWecX51znpmZtebuH2ZmZmZmPXKl2szMzMysR65Uj6xz+52Acch51jnnWWecX51znpmZteA+1WZmZmZmPXJLtZmZmZlZj1ypNjMzMzPrkSvVZmZmZmY9cqW6CUkflnSGpJsl/a+kkHRRizg7S7pa0kJJz0u6S9Kxkl5TEecwSb+WtEjSs5IGJX1o+M+ofyTNyPlXtSwphJ/aIuzF/Tyf0SBpfsX5P94kTsflb6KQtJakIyVdJul+SS/kz9P/SPq4pOUawk/6MtaKpA0knS/pUUkv5TJ5mqQ1+502M7OxaPl+J2AM+yKwDbAI+BOweVVgSfsBPwBeBL4PLAT2Ab4F7AIcWBLnVOCzef/fBlYEDgJ+LOmYiDhzuE6mz+YCJzbZ9n+A3YBrSrbdCVxe8vrdw5Kqse9Z4LSS1xc1vtBN+ZtgDgTOAh4DbgAeAt4I/BVwHrC3pANj2TuzJ3sZKyVpE+AXwDrAFcAfgB2BzwB7SdolIp7uYxLNzMYcj/7RhKTppMru/cCupC/q70bEISVhV8/h1gB2iYg5+fWVgOuBdwN/ExEXF+LsDNwCPAC8MyKeya9PBW4DVgU2j4j5I3SKY4KkXwI7AftFxI/ya1OBecCFETGjf6nrH0nzASJiahthOy5/E42k3UifmasiYmnh9XWBXwMbAh+OiB/k16cyyctYFUk/BfYEPh0RZxRe/yZwHHBORPxtv9JnZjYWuftHExFxQ0TcV9KyVebDwNrAxbUKTd7Hi6QWb4CjGuLUvpC+WqtQ5zjzgX8HpgCHd5n8cUHSNFKF+hHgqj4nZzzrpvxNKBFxfUT8uFihzq8/Dpydnw6MesLGIUkbkyrU80nXoqIvA38BDpW06ignzcxsTHP3j+GxW17/pGTbTcDzwM6SpkTES23EuQY4IYf58nAmdIz5VF5/JyKWlGxfX9KngLWAp4FfRsRdo5a6/psi6RDgzaSKzF3ATSV51U35m0wW5/UrJdsmexkrUytPPyv5kfKcpFtIle6dgOtGO3FmZmOVK9XD4215fW/jhoh4RdI8YCtgY+Ce3MLzJmBRRDxWsr/78nqzkUjsWCBpZeAQYCmpz2uZ9+WlGG8QOCwiHhrRBI4N6wKzG16bJ+nwiLix8FpH5W9EUjpGSVoe+Fh+WvajY7KXsTJNy1N2H6lSvRmuVJuZvcrdP4bHGnn9bJPttddf12X4iegjpPO7JiIebtj2PHASsD2wZl5q/doHgOsmwV/PFwC7kyrWqwJbA+cAU4FrJG1TCOvy1NzXgGnA1RHx08LrLmPNuTyZmXXBlerRobzu9K7QiXwX6Sfz+pzGDRGxICK+FBG3R8Sf83ITqXXsV8BbgSNHMa2jLiJOzP2En4iI5yPi7nxj2DeBlYGZHeyu2/I3rkn6NGl0nT8Ahxa3uYz1ZFKWJzOzVlypHh61lps1mmxfvSFcq/CtWorGNUlbAjuTRle5ut14EfEK9a4i7x2BpI0HtZvuiuffafmb8CQdDfwb8HtgekQsbCeeyxjg8mRm1hVXqofH/8vrZfpA5z6dbyHdJPUgQET8hTTixWqS1ivZ36Z53axP43jX6gbFKk/m9WT9a35BXhfPv6PyN9FJOhY4kzTW9PQ8AkgnJnsZa1qesol+fTIz64or1cPj+rzeq2Tbe4FVgF80jLxQFWfvhjATRh47+VDSDYrf6WIXO+X1pKgglnh3XhfPv5vyNyFJ+ifShDdzSRXqBdUxSk32MnZDXu9ZMhPla0mTCb0A3DraCTMzG8tcqR4elwJPAQdJ2qH2Yq5AnpyfntUQp/Y3/vHFaX/zpBRHAy+RblabaA4k3RR2dckNigBIepekFUte34008QRA5ZTx45mkrSS9vuT1jUgtsDD0/LspfxOOpBNINybeBuweEU9VhJ3UZaxKRDwA/Ix0U+zRDZtPJLXg/2f+x83MzDLPqNiEpP2B/fPTdYH3k1qubs6vPRURn2sIfylpmuiLSdNE70sanupS4CONE8lI+gbwD6S+xZeSpin/KGnM3Ik0TfmrJN0MvAfYNyJ+3CTMIGkIuEFS3gC8nfr4uSdExMnLxpwYJM0EvkBqMZwHPAdsAnwQWInUD/2AiHi5EGd/Oix/E4mkw4BZwBLgDMr7+86PiFk5/CCTuIy1UjJN+T3Au4DppG4fO3uacjOzoVypbiJXbKomXvlj4xTSknYBjif9Rb8Saero84HTm/UdzpWBvwe2JHWJuB04JSKu7PEUxhxJW5BuHPsTMLUiTz4OHEAaCu0NwArAE8AvgTMj4uayeBOFpF1JM25uS31IvT+TujTMBmaXVZC7KX8TRRufV4AbI2Igh5/UZawdkjYEvkLqVrQW8BhwOXBiuzd+mplNJq5Um5mZmZn1yH2qzczMzMx65Eq1mZmZmVmPXKk2MzMzM+uRK9VmZmZmZj1ypdrMzMzMrEeuVJuZmZmZ9ciVajMzMzOzHrlSPU5ImikpJHlg8QJJK0s6QdIdkhbV8kjSrH6nzcYHSfMncpmRNFD4XAyM4HFqx5g5UscwMxvLxmSluuFLoJ1lZr/TbKNP0grAtaRZ395BmnnQzMzMbNSNyUr1ZCFpRuGHwdR+p2ccOhDYOT+eDUwHts7L8f1KlJl1R9LUwjVxRr/TY2bWieX7nYA2nAX8R4swC0YjIf0UETOBmX1OxlizR14/ARwREa/0MzFmY1FEDALqdzrMzCa68VCpXhARd/c7ETYmvSmvH3CF2szMzPrJ3T9sPJuS14v7mgozMzOb9CZUpbqxj7KkFSV9WtIvJD0paamk0wrhV5S0j6QzJf1G0jOSFkt6WtKv8ogbb2jz2CtIOlzSFZIelvSipKckzZV0nqT3S1o+hx3Io3hcUNjFvJIbMAcK+29r9A9JG0o6VdJvJT0r6QVJ8yRdKGnnFnGHjIIgaTNJZ+f4L+Y8vErS7u3kSTsk7SXp+4U8e0bSbZK+Upb3xT6XwK755V0b867DNMzK8ea3CFc5ioKkwbxtMD9fP78X9+b34RlJ10s6sMVxhtyAK2k3SZdJeiTn0R8lfVvSJu2mVckMSddJelzSEkmXF8K3NQJGq3JYkvbtJX1X0kOSXpL0mKRLJG1fdZzC/taTdLKkOZIW5n08IumHkvZtcx97S7o6l9/n8/vxTUlvah275b6vzOd7R5PtWxTyZImk15eEUU5bSDqr4ljT8+f4gXwez0n6vaTTJW1cEa+t0T8kvVnSWYXP+6OSLpc0PW/vaASiTt/7vN95hZcuaPxcq+GmdEmrS/pnSbcoXbdr1+97lK5Vx0raqJ30mpn1LCLG3AIMAJGXmR3Em1GItz1wW+F5bTmtEH5WyfbG5SlglxbH3Qq4t419DZScX8vwOc7M2usV6TgYeKHFPk8HlmsSf34OMwvYH1hUsZ9jenyPpwDfb5HWZ4A9GuJNbSfvOkxLrRzM76BcDpRsH8zbBkk3UC6oSOOpFcd5tewDJwBLm+zjeWDfNtK6F/DTkviXl733LfKgshw2pP0o0r8IZWl/GTigxbEOA/7S4r2+HFi1Yh/frIi7ANih3XNvsv/P5bhLgdeXbD+q4Zj7l4TZurD9oyXbVwb+u0U+vES6r6DjcpvD7AY812TfS4H/O9LvfYvze3XfhfCbAw+3EedrvVyrvHjx4qXdZTz0qe7W+aQvq+8CFwOPkvrgvqYQZnngQeAy4NfAQ8ArwEakm+COANYCLpM0LSKWuSFS0qbALcAa+aWrgO8B95FuDtqYNCpFsXXyNzlt+wEn59fen9NYNK/dk5W0F3BRPuYLwGnANaQv2x2ALwAbAscALwKfr9jd1sBHgCdJo2j8GlhCahk+Pp/rNyT9PCL+0G4aG1yQjwHwe+BU4LfA6qQK/VHA64CrJO0UEXfksI/k9NX2sQMwBzi8y3SMhPWAK/Lj44GbSO/JO4Ev5e2flXRNRFxXsZ8P5Dj3A18H5gKrAfsCf0+qbF0iaceIuLNiP18H3k4qmxeQKpFrAW/s4tzatSfwLuAeUlm8i/R525tU9lYEzpd0U0Q83RhZ0mGkHzuQKk5nAL8DHid9jg8GDiJ9hoplqbiPY4Hj8tPHgX8BbiX9oPsgcCxwCbBKD+c5WDsc6fNxWcP2gZLnl1eEubG4QdJywI+o35T7M9I1bR7pc7wd6Tw2B86TtCAiruzkBHIr949IQ1IuBc4BfgA8C0wD/hH4Kuk60I5u3/utgfVJPwABvkj9c1RTvAbPBjYgXbPPI13vHsvb1id9dvZrM81mZr3rd62+bGFoy8p/kC7szZY3FeLNYGgLxadaHGcTQBXbt6beenNSkzC/ot6aM6NiX6sBaza8Vkzv1BZpnVkLW7JtBeotNs8DO5WEWYv0JRekCvI2JWHmF9JzR2N6c5hdC2G+1eX7u3dhH7cAK5eE+VBOZwC3NdnPYN4+2GN5m8XwtlQH6QfahiVhNiNVhgK4rMlximV4LrB6SZi9CvlzS4u0BvAvLc6t9t7P6rYclqT9J8CUkjAfK4T5TMn2Dai3UH8PWLHJsYqtwLs3bFunsI9HKFwnCmGmM7Q1tfLcm6ThNcCfc/zTS7Y/nrddkdd3loT5Qd52T8m24/K2Vyhp5c5hVi6UvXnA8h2W28sK2w8q2b4K9evcSL/3UwvbZ1Tk+8aFcJX/mlHyD4IXL168jMQyHvpUH0VqwWy2fLVJvMGIOKdqxxHxQERExfbfklpAILWeDiFpD2DH/PSciJhVsa9FEfFMVXp6sD+pIgJwSkTcWnL8p4FP5qfLkVo6qxxelt6IuJH0BQv1Ps2dqh17KXBYRLxQcpwrqbdUbidply6P1S/HRMTDjS9GxL3UWyrf28Z+PhER/1uyn58AF+anO0vatmIf95O6kYymF0nv7Usl2y6i3qJYVoY+Q6rIPQkcGREvlx0gIs4i/esD6V+losOot0B/PiIeKYl/A/DtqpNoJSKWADfnpwPFbZK2IP0bsAQ4Kb+8taS1CmFEvRwMNsRfgdS9BNL15fImaXgB+Lv8dCrpx0JbJK0P7JOfXhkRF5fs/3nq14529PLet2vdwuMbm4YCImJhD8cxM2vbeKhUd+uiTiNIWlPSJpK2kjRN0jRSKxTAlvlLruhDhcff6jKdw+F9hcfnNQsUETcDte4a72sWDrg7IuZWbJ+T101vjmpG6WbNgfx0MCLurwh+buFxVXrHmmeBH1dsr+Xf6yW9riLc3RHxm4rt5xceV+XPxTH6Qw5eGxFPlG2IiKXA7flpWRnaP6+vzhW6Kjfl9bsbXq91l1gEXFoR//yKbe0azOtpGnpz7UBe3xERc4A/Ue8mUjMNqMUZZKgdSd0YIHVTaSoifg/UulI05kWV6dS7xM2u2P+dQFUXo6Je3vt2FbvKzcg/TszM+mo8VKpPjAhVLDOaxGvrC0DS1pLOl/QYsJDUqnc39ZbwmTnocsCaDdG3y+sFuQWyX6bl9aNlraMNaq3YG0l6bZMw97TYR63lp1n8KhtTb0FcpkW9we3Uh8vbuirgGHNvrjw0U2w5q8rDVn1Y55Ba+yH1mW6m3crQcOqqDElaA3hrfnpYyegPjaO8fDaHLbZcQr283NWkxbRmLunGuV4M1pLP0NbqgYbtNza83vi4scV1h8LjG9rIi1oLeGNeVJlWeHxbi7BzWmyvGcnrBwARMZ96fh0H/E7SSZL2kLRat/s1M+vFeKhUd6tlVwtJHydV3A6nvS+ilRuer53XjzUGHGW1YbramVny8ZJ4jVq1DtYqct2Un+IxK9MbEYupt741S+tY1G7+wdAbZxu1yp8XgVrXkLUqgo5Ut6Mq7eZB4/mv0+XxGj+bbX0mcgt+r90D7qD+j1ax60WtRXqwYT1QCFN7/IeIKH42ofu86OTGy2JDwZMtwrbaXtPte9+pv6He9WYL0o2NPweekfRLScdVNByYmQ27iTz6x5KqjZI2B84m5cEC4BTgetINW8/lCh2SjgC+U4vWZHdN+2WPsnbSMZb+Jh1v6R1tw5U/lZ+FMaZY0TqTNBJFL0a8jEXEUkk3k/omD8Ay/alrFb8b8rrWTeRpmvSnzop58T6G/iCu0u2PqLFyHWtLRDwGvDePvX0AKe+nka7pO+Xl85L2j4hfNdmNmdmwmciV6lZmkM5/CbBrNB8arqqF9Km8Xr8izGiotbS1M0RaMUw/buApHrMyvbkPey3/RzKt7ba8rzqCaSjTKn9Wov4X+jLD0nVorOTBU4XHq0XE3V3u5xnSv0+t8nB5lu3W1Y1BUqV6S0lvpN4CfXvtRtOIeEDSw6ShLXcljW3frD81DM2LxT3kRZViBXwd0mhHzaxdsa1vImKQnH/5HoXppH8f9yGVgR9K2iT/s2NmNmImcvePVrbK6zsrKtQwtF9jo1ofxHUkva2LNAxXy1Dty3Z9SRtUhkzjxwL8MSKqvkBHyoPU/x5+V1VAYFvScIGQ+rePlFo+vK5FuM1HMA1ldmyxfXvqn+Fe86eWB60qmCOaBxHxFGl4SIA98jjN3ajlx9slrVgRbhvSuMm9Giw8HqBeqb6hSbjpVPenhtStpGavbhPWwu8Kj1vNcll1LRwuPV0TI+LPEXFZROxL/Ubn9YH39JwyM7MWJnOlutZK37TlTdJ6pIk2mimO8HBc01DNFVtOpnQRv+bnhceNQ4u9Kg9Lt0VJnFGT+7AO5qcDkt5SEfwThccjmd4H8/q1uVvQMvLoAgePYBrKTGs2pXNWfK97zZ9aHmzfbCQFSetQH1VjJNUm/NgAOLTLfVyb16sBf10RrunnpUNzGdqvurE/NQ3PB6juTw3wP9Rbqz+hkinOh8EN1P+laJrXkrYh/QAZacN1TQQoTqz0hqahzMyGyWSuVN+X15tK2rlxo6RVSDOXNd4A9apIs+HV7oj/ZJ4FrpSkVSU1tgIWb3DcpK1Ul7ucNFwXpD6E2zUGyMeutdwEqb9qv9SO/RrgAknLfHlK+gD1Cs/tEXHLCKan2ErYbKbJL1If7WU0fbtsNANJe5K6MAHcGhG3N4bpUC0P1qOkcpXfo1nASj0epx2nkGagBDhdUmUro6T3SGoc8/jCwj7+NY/H3BhvVzobf7mpPNpLbXi/g0jdTl4hVYyLBvN6S+o/UAYpkUct+df8dC3g0jw6SilJUyQdnbsFtZvuR0gzbQJ8SNJHS/a7MkOHtxxJT1MfjaXpNVHSO1qMzQ5Dh5mc12vCzMxamcx9qmeTpuxejjQV9imkL8AXSX+DHgfUpiCvmnjkENIEFK8FZkn6MEOnKZ9KapE6iNRiNliIe0c+3krASZIWA3+k3nL0SNnEKI0iYrGkTwBXk1reb5L0LdJ0v7Vpyv+JNP06wKlRPa31iIqIayT9F+nu/V2BOZJOJXVjWZ00tfDRpPfmZeDIEU7PXEm19/nw3F3gAlJ/07eQJhLZh9ZlYbjNIb13t0v6OmlovFVJ/57Uyu7L1Cf+6MVFwJdJXWDOlbQJadrnJaQWyk+TukzdSroBbMRExEOSjsxpWp00nNx/kVqw55POez3S5/QA0vB5x1D4cRQRT0g6ATiV1OI9R1JxmvIPkD7jj5BGyxiO/sKDpPemVvG9rbGLVUQ8KOkh4M353GrxmvkGqeV777y+R9LZpGvVQlJ52JTUveGvSN13LizfVVP/AOxOyofvSnovy05TvhXpOvfODvfdkYh4RdJvSJ+zIyTdQfoXoDa05sI8mcs7SD/IbyP9Y3g7qZFiOVLeHkwqG5C66bU7xbqZWfdGa+rGThaGTqs7s4N4MwrxprYR/kuF8GXLqe3sk3SBn9diX82mCP56O+FpMT10DnMwqXWuKg1nAMs1iT+fYZiqus33agrw/RZpfQbYo2Ifgznc4DCUuU1JX8rN0nIRsFuL97Kt9LQqU8WyT6roNkvTCzSfunqgKq1N4hzA0Gm7i8ti0g+dyve+mPYWx5qVw81vkZ6FLcpIbflYk338W0WcJ0mVxLbKfRv5t23D/r/WJNyFDeHWbeOzci7px3arfFgErNxpWSC16i6q2O9M4Cu1cjeS7z3wwYpznVnyGapafgu8udfrgxcvXry0s0zm7h9ExFdIF/CfkSpwL5O6UfwQ2DMiPlcRvbifuaQbuI7K+3qCVAl5ktTKci6pJeimkuhfIPUdvplUgeh6+LOI+B6wGal1627SzWcvkVq/ZwO7RMQxUT0xyaiIiJci4qOkFrhLSPn+Mql17A7gZGDTiLi2+V6GNT33kbp3nEHqX/wy6a/o64GDIuIQho4vPSoi4kRgT1Jr3OM5XQ+Thnl8ezSZurrLY11GaoW+hHoZfpT042eXiPj34TpWB+mZSpqq+zrS+S8m/Zh4CPgJ8M/A2yLiP5vs4zOkz/hPSZ+vF0kTPJ0ObBvVM1Z26k6GjqYx2CTcDYXHzfpTvyp/Vj5J+vF+BnAXqf/2EtI45XeTPt+HAutFG/9ulRzj56RW6XNI14uXSWXgKmCviJhJvWX92U7332FariJdL68glb/FJcG+R2q5/yrpuvog8Jec7sdI/7IcCWwXEQ+NZHrNzGoUEf1Og5kV5NnxIM0mOrOfaTGrkXQtqbJ7S0R4NA0zswaTuqXazMxayzd61iaqubWfaTEzG6tcqTYzm+QkvbVi28qkftC1MeNnj0aazMzGm8k8+oeZmSVn59kI/5s0WsbTpBGNdiDdK7JpDjcr+jhykJnZWOZKtZmZQRqisGqyoR+RRoAxM7MSrlSbmdk/koYw3J00rvfapHH2F5D6UM+OiCv7lzwzs7HPo3+YmZmZmfXINyqamZmZmfXIlWozMzMzsx65Um1mZmZm1iNXqs3MzMzMeuRKtZmZmZlZj/4/gC9dZPBuf5cAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax1 = plt.subplots(figsize=(6, 4))\n",
    "ax1.plot(imp_df['weight_remain'], imp_df['test_acc'], linewidth=line_width, c='C0', label='IMP')\n",
    "ax1.plot(sp_list, vanilla_acc, linewidth=line_width, c='C0', label='HC, Suggested', linestyle='dashed')\n",
    "ax1.plot(sp_list, weight_shuffle_acc, linewidth=line_width, c='C2', label='HC, weight shuffled', linestyle='dashed')\n",
    "ax1.plot(sp_list, mask_shuffle_acc, linewidth=line_width, c='C1', label='HC, mask shuffled', linestyle='dashed')\n",
    "\n",
    "plt.title(\"ResNet20 on CIFAR-10\", fontsize=title_size-10)\n",
    "#ax1.set_xlim([0,10])\n",
    "ax1.invert_xaxis()\n",
    "ax1.tick_params(axis='both', labelsize=tick_size)\n",
    "ax1.set_xlabel(\"Fraction of unpruned weights\", fontsize=label_size)\n",
    "ax1.set_ylabel(\"Test Accuracy\", fontsize=label_size)\n",
    "ax1.legend(fontsize=legend_size, loc='center left', bbox_to_anchor=(1.1, 0.8))#, loc='right')\n",
    "ax1.grid(True)\n",
    "plt.savefig('hc_resnet20_cifar10.pdf', format='pdf', dpi=600, bbox_inches='tight', pad_inches=0.05)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    " \n",
    "# creating an x sequence\n",
    "x = np.linspace(5, 15, 35)\n",
    " \n",
    "# equation of a straight line\n",
    "y = 3*x+4\n",
    " \n",
    "# creating graph space for two graphs\n",
    "graph, (plot1, plot2) = plt.subplots(1, 2)\n",
    " \n",
    "# plot1 graph for normal axes\n",
    "plot1.plot(x, y)\n",
    "plot1.set_title(\"Normal Plot\")\n",
    " \n",
    "# plot2 graph for inverted axes\n",
    "plot2.plot(x, y)\n",
    "plot2.set_title(\"Inverted Plot\")\n",
    "plot2.invert_xaxis()\n",
    "plot2.invert_yaxis()\n",
    " \n",
    "# display the graph\n",
    "graph.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#weight_reinit_df = pd.read_csv(\"{}/acc_and_sparsity_weight_reinit.csv\".format(RESULTS_FOLDER))\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "fig, ax1 = plt.subplots(figsize=(15, 10))\n",
    "ax2 = ax1.twinx()\n",
    "\n",
    "ax1.plot(acc_df['epoch'], acc_df['test_acc'], linewidth=line_width,\n",
    "         c='C0', label='HC Quantized with fine-tuning (round in forward pass)',\n",
    "         marker='^', markersize=12, markevery=15)\n",
    "ax2.plot(acc_df['epoch'], acc_df['model_sparsity'], linewidth=line_width,\n",
    "         c='C0', label='HC Quantized (round in forward pass)', linestyle='dotted')#, marker='^', markersize=12)\n",
    "\n",
    "ax1.plot(mask_shuffle_df['epoch'], mask_shuffle_df['test_acc'], linewidth=line_width,\n",
    "         c='C1', label='HC Quantized with mask shuffle', linestyle='dotted')#, marker='^', markersize=12)\n",
    "ax2.plot(mask_shuffle_df['epoch'], mask_shuffle_df['model_sparsity'], linewidth=line_width,\n",
    "         c='C1', label='HC Quantized mask shuffle', linestyle='dotted')#, marker='^', markersize=12)\n",
    "\n",
    "ax1.plot(weight_reinit_df['epoch'], weight_reinit_df['test_acc'], linewidth=line_width,\n",
    "         c='C2', label='HC Quantized with weight reinit)', linestyle='dotted')#, marker='^', markersize=12)\n",
    "ax2.plot(weight_reinit_df['epoch'], weight_reinit_df['model_sparsity'], linewidth=line_width,\n",
    "         c='C2', label='HC Quantized with weight reinit', linestyle='dotted')#, marker='^', markersize=12)\n",
    "\n",
    "ax1.plot(weight_shuffle_df['epoch'], weight_shuffle_df['test_acc'], linewidth=line_width,\n",
    "         c='C3', label='HC Quantized with weight shuffle', linestyle='dotted')#, marker='^', markersize=12)\n",
    "ax2.plot(weight_shuffle_df['epoch'], weight_shuffle_df['model_sparsity'], linewidth=line_width,\n",
    "         c='C3', label='HC Quantized with weight shuffle', linestyle='dotted')#, marker='^', markersize=12)\n",
    "\n",
    "plt.title(\"ResNet20 on CIFAR-10 with Finetuning after 150 epochs\", fontsize=title_size)\n",
    "ax1.tick_params(axis='both', labelsize=tick_size)\n",
    "ax2.tick_params(axis='both', labelsize=tick_size)\n",
    "\n",
    "ax1.set_xlabel(\"Epoch \", fontsize=label_size)\n",
    "ax1.set_ylabel(\"Test Accuracy\", fontsize=label_size, color='C0')\n",
    "ax2.set_ylabel(\"Model Sparsity\", fontsize=label_size, color='C1')\n",
    "\n",
    "#ax2.set_ylim([0, 55])\n",
    "ax1.set_ylim([10, 95])\n",
    "#ax1.set_xlim([0,152])\n",
    "ax1.legend(fontsize=legend_size, loc='center left', bbox_to_anchor=(1.1, 0.8))#, loc='right')\n",
    "ax1.grid(True)\n",
    "#ax2.grid(True)\n",
    "plt.savefig('hc_quantized_resnet20_cifar10.pdf', format='pdf', dpi=600, bbox_inches='tight', pad_inches=0.05)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "acc_df.tail()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## MNIST"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# get all file names in directory\n",
    "path = 'results/MNIST'\n",
    "extension = 'csv'\n",
    "os.chdir(path)\n",
    "filenames = glob.glob('*.{}'.format(extension))\n",
    "\n",
    "# read in each csv\n",
    "csv_dict = {}\n",
    "for file in filenames:\n",
    "    print(file)\n",
    "    csv_dict[file] = pd.read_csv(file)\n",
    "    ignore_rows = (csv_dict[file].epoch < 1) & (csv_dict[file].epoch > 0)\n",
    "    ignore_rows = ignore_rows.index[ignore_rows == True]\n",
    "    csv_dict[file].drop(index=ignore_rows, inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "files_FC = [    'hc_iter_period_1_adam_lr_1e-2_100epoch.csv',\n",
    "                'hc_vanilla.csv',\n",
    "                'hc_rewind_iter_period_1.csv',\n",
    "                'vanilla_ep_sparsity_0.9.csv',\n",
    "                'vanilla_ep_sparsity_0.5.csv',\n",
    "                'lth_iter5.csv',\n",
    "                'hc_iter_period_5_adam_lr_1e-2_100epoch.csv',\n",
    "                'hc_iter_period_1_switch_to_wt_at_50_epoch_to_100.csv',\n",
    "                'hc_iter_period_5_switch_to_wt_at_50_epoch_to_100.csv'\n",
    "           ]\n",
    "\n",
    "df0 = csv_dict[files_FC[0]]\n",
    "df1 = csv_dict[files_FC[1]]\n",
    "df2 = csv_dict[files_FC[2]]\n",
    "df3 = csv_dict[files_FC[3]]\n",
    "df4 = csv_dict[files_FC[4]]\n",
    "df5 = csv_dict[files_FC[5]]\n",
    "df6 = csv_dict[files_FC[6]]\n",
    "df7 = csv_dict[files_FC[7]]\n",
    "df8 = csv_dict[files_FC[8]]\n",
    "# df9 = csv_dict[files_FC[9]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df7['test_acc'][80:100]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "fig, ax1 = plt.subplots(figsize=(15, 10))\n",
    "ax2 = ax1.twinx()\n",
    "\n",
    "ax1.plot(df5['epoch'], df5['test_acc'], linewidth=line_width-2, c='C0', label='IMP (50 epochs, Prune every T=5 epochs)', linestyle='dotted')#, marker='^', markersize=12)\n",
    "ax2.plot(df5['epoch'], df5['model_sparsity'], linewidth=line_width-2, c='C1', label='Weak LTH (w/ rewinding, Prune every 5 epochs)', linestyle='dotted')#, marker='^', markersize=12)\n",
    "\n",
    "ax1.plot(df7['epoch'], df7['test_acc'], linewidth=line_width-2, c='C0', label='HyperCube (50 epochs, T=1) + Weight Training (50 epochs)', linestyle='solid')\n",
    "ax2.plot(df7['epoch'], df7['model_sparsity'], linewidth=line_width-2, c='C1', label='HyperCube (50 epochs, T=1) + Weight Training (50 epochs)', linestyle='solid')\n",
    "\n",
    "#ax1.plot(df8['epoch'], df8['test_acc'], linewidth=line_width-2, c='C0', label='HyperCube (50 epochs, T=5) + Weight Training (50 epochs)', linestyle='dashed')\n",
    "#ax2.plot(df8['epoch'], df8['model_sparsity'], linewidth=line_width-2, c='C1', label='HyperCube (50 epochs, T=5) + Weight Training (50 epochs)', linestyle='dashed')\n",
    "\n",
    "\n",
    "\n",
    "#ax1.plot(df1['epoch'], df1['test_acc'], linewidth=line_width-2, c='C0', label='HyperCube (vanilla)', linestyle='dashed')#(0, (1, 10)))\n",
    "#ax2.plot(df1['epoch'], df1['model_sparsity'], linewidth=line_width-2, c='C1', label='HyperCube (vanilla)', linestyle='dashed')#(0, (1, 10)))\n",
    "\n",
    "#ax1.plot(df2['epoch'], df2['test_acc'], linewidth=line_width-2, c='C0', label='HyperCube (iterative, w/ rewinding)', linestyle='dotted')\n",
    "#ax2.plot(df2['epoch'], df2['model_sparsity'], linewidth=line_width-2, c='C1', label='HyperCube (iterative, w/ rewinding)', linestyle='dotted')\n",
    "\n",
    "#ax1.plot(df6['epoch'], df6['test_acc'], linewidth=line_width-2, c='C0', label='HyperCube (Prune every 5 epoch)', linestyle='dashed')#', w/o rewinding)')\n",
    "#ax2.plot(df6['epoch'], df6['model_sparsity'], linewidth=line_width-2, c='C1', label='HyperCube (Prune every 5 epoch)', linestyle='dashed') #, w/o rewinding)')\n",
    "\n",
    "#ax1.plot(df0['epoch'], df0['test_acc'], linewidth=line_width-2, c='C0', label='HyperCube (Prune every 1 epoch)')#', w/o rewinding)')\n",
    "#ax2.plot(df0['epoch'], df0['model_sparsity'], linewidth=line_width-2, c='C1', label='HyperCube (Prune every 1 epoch)') #, w/o rewinding)')\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "#ax1.plot(df4['epoch'], df4['test_acc'], linewidth=line_width-2, c='C0', label='Edge-Popup (50% sparsity)', linestyle='dashdot', marker='^', markersize=12)\n",
    "#ax1.plot(df3['epoch'], df3['test_acc'], linewidth=line_width-2, c='C0', label='Edge-Popup (10% sparsity)', linestyle='dashdot', marker='o', markersize=12)\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "#ax1.plot(df2['epoch'], df2['test_acc'], linewidth=line_width-2, c='C2', label='DG (50% init sparsity, 25% importance)')\n",
    "# ax1.plot(df3['epoch'], df3['test_acc'], linewidth=line_width-2, c='C3', label='DG (0.1% init sparsity, 25% importance)')\n",
    "# ax1.plot(df4['epoch'], df4['test_acc'], linewidth=line_width-2, c='C4', label='DG (50% init sparsity, 25% importance, combination test)')\n",
    "\n",
    "# ax1.plot(df9['epoch'], df9['test_acc'], linewidth=line_width-2, c='C9', label='DG (0.1% init sparsity, 25% importance)')\n",
    "# ax1.plot(df2['epoch'], df2['test_acc'], linewidth=line_width-2, c='C2', label='DG')\n",
    "# ax1.plot(df5['epoch'], df5['test_acc'], linewidth=line_width-2, c='C5', label='DG (10 epoch) + DG-low (q=0.25)')\n",
    "# ax1.plot(df6['epoch'], df6['test_acc'], linewidth=line_width-2, c='C6', label='DG-low (q=0.5)')\n",
    "# ax1.plot(df7['epoch'], df7['test_acc'], linewidth=line_width-2, c='C7', label='DG-low (q=0.25)')\n",
    "# ax1.plot(df8['epoch'], df8['test_acc'], linewidth=line_width-2, c='C8', label='DG-low (q=0.1)')\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "####ax1.plot(df4['epoch'], df4['test_acc'], linewidth=line_width-2, c='C4', label='DG (50% init sparsity, 10% importance)')\n",
    "\n",
    "\n",
    "\n",
    "plt.title(\"RamanujanNet on MNIST\", fontsize=title_size)\n",
    "ax1.tick_params(axis='both', labelsize=tick_size)\n",
    "ax2.tick_params(axis='both', labelsize=tick_size)\n",
    "\n",
    "ax1.set_xlabel(\"Epoch \", fontsize=label_size)\n",
    "ax1.set_ylabel(\"Test Accuracy\", fontsize=label_size, color='C0')\n",
    "ax2.set_ylabel(\"Model Sparsity\", fontsize=label_size, color='C1')\n",
    "\n",
    "ax2.set_ylim([0, 50])\n",
    "ax1.set_ylim([93, 100])\n",
    "#ax1.set_xlim([0,52])\n",
    "ax1.legend(fontsize=legend_size, loc='center left', bbox_to_anchor=(1.1, 0.8))#, loc='right')\n",
    "ax1.grid(True)\n",
    "ax2.grid(True)\n",
    "plt.savefig('HC variants.jpg', dpi=400)\n",
    "#plt.savefig('DG (follow low positive) .jpg', dpi=400)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "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": "3.8.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}