DICE_CIFAR.ipynb

{
 "cells": [
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# DICE\n",
    "[DICE: Leveraging Sparsification for Out-of-Distribution Detection](https://arxiv.org/abs/2111.09805)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from typing import Dict, Tuple, Union\n",
    "\n",
    "import torch\n",
    "from torch import Tensor\n",
    "\n",
    "import torchvision\n",
    "import torchvision.transforms as transforms\n",
    "from torchvision.models.feature_extraction import get_graph_node_names, create_feature_extractor\n",
    "\n",
    "from torchmetrics import Metric, CatMetric, MetricCollection\n",
    "from torchmetrics.classification import BinaryROC, BinaryPrecisionRecallCurve\n",
    "from torchmetrics.utilities.data import dim_zero_cat\n",
    "from torchmetrics.functional.classification.precision_recall_curve import _binary_precision_recall_curve_compute\n",
    "\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "from tqdm.auto import tqdm\n",
    "\n",
    "from models.densenet import DenseNet3\n",
    "import util.svhn_loader as svhn"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Files already downloaded and verified\n",
      "Files already downloaded and verified\n",
      "Files already downloaded and verified\n"
     ]
    }
   ],
   "source": [
    "transform_cifar = transforms.Compose([\n",
    "  transforms.Resize(32),\n",
    "  transforms.CenterCrop(32),\n",
    "  transforms.ToTensor(),\n",
    "  transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2023, 0.1994, 0.2010)),\n",
    "])\n",
    "\n",
    "datasets = {\n",
    "  'CIFAR-10': torchvision.datasets.CIFAR10(root='./datasets/id_datasets/', train=False, download=True, transform=transform_cifar),\n",
    "  'CIFAR-100': torchvision.datasets.CIFAR100(root='./datasets/id_datasets/', train=False, download=True, transform=transform_cifar), \n",
    "  'SVHN': svhn.SVHN('datasets/ood_datasets/svhn/', split='test', transform=transform_cifar, download=False),\n",
    "  'dtd': torchvision.datasets.ImageFolder(root=\"datasets/ood_datasets/dtd/images\", transform=transform_cifar),\n",
    "  'places365': torchvision.datasets.ImageFolder(root=\"datasets/ood_datasets/places365/\", transform=transform_cifar),\n",
    "  'celebA': torchvision.datasets.CelebA(root='datasets/ood_datasets/', split='test', download=True, transform=transform_cifar),\n",
    "  'iSUN': torchvision.datasets.ImageFolder(\"./datasets/ood_datasets/iSUN\", transform=transform_cifar),\n",
    "  'LSUN': torchvision.datasets.ImageFolder(\"./datasets/ood_datasets/LSUN\", transform=transform_cifar),\n",
    "  'LSUN_resize': torchvision.datasets.ImageFolder(\"./datasets/ood_datasets/LSUN_resize\", transform=transform_cifar),\n",
    "}\n",
    "\n",
    "dataloaders = {\n",
    "  k: torch.utils.data.DataLoader(v, batch_size=512, shuffle=False) for k,v in datasets.items()\n",
    "}\n",
    "\n",
    "ood_dls = ['SVHN', 'LSUN', 'LSUN_resize', 'iSUN', 'dtd', 'places365']\n",
    "\n",
    "device = 'cuda:0'\n",
    "result_dfs = {}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "class BinaryPrecisionRecallCurveOut(BinaryPrecisionRecallCurve):\n",
    "  def compute(self) -> Tuple[Tensor, Tensor, Tensor]:\n",
    "    if self.thresholds is None:\n",
    "      state = [1-dim_zero_cat(self.preds), 1-dim_zero_cat(self.target)]\n",
    "    else:\n",
    "      state = self.confmat.flip((1,2))\n",
    "    return _binary_precision_recall_curve_compute(state, self.thresholds)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "curves = MetricCollection({\n",
    "  'roc': BinaryROC(compute_on_cpu=True),\n",
    "  'prc_in': BinaryPrecisionRecallCurve(compute_on_cpu=True),\n",
    "  'prc_out': BinaryPrecisionRecallCurveOut(compute_on_cpu=True)\n",
    "})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def compute_metrics(curves:MetricCollection) -> Dict[str, float]:\n",
    "  res = curves.compute()\n",
    "\n",
    "  fpr, tpr, _ = res['roc']\n",
    "  p_in, r_in, _ = res['prc_in']\n",
    "  p_out, r_out, _ = res['prc_out']\n",
    "\n",
    "  return {\n",
    "    'FPR@95': fpr[tpr <= 0.95].max().item(),\n",
    "    'DTErr': ((1-tpr+fpr)/2).min().item(),\n",
    "    'AUROC': torch.trapz(tpr, fpr).item(),\n",
    "    'AUPR_In': -torch.trapz(p_in, r_in).item(),\n",
    "    'AUPR_Out': -torch.trapz(p_out, r_out).item(),\n",
    "  }\n",
    "\n",
    "def run(\n",
    "  extractor:torch.fx.GraphModule,\n",
    "  dataloader:torch.utils.data.DataLoader,\n",
    "  is_ood:bool,\n",
    "  curves:MetricCollection,\n",
    "  device:str='cuda:0'\n",
    ") -> Dict[str, Union[Tuple[Tensor, Tensor], MetricCollection]]:\n",
    "  features = CatMetric(compute_on_cpu=True)\n",
    "  for x, _ in tqdm(dataloader):\n",
    "    with torch.no_grad():\n",
    "      res = extractor(x.to(device))\n",
    "      energy = torch.logsumexp(res['logit'], -1)/1000.0  # make sure that the value in [0,1] interval.\n",
    "      target = torch.tensor([not is_ood]*energy.shape[0], dtype=torch.int)  # 1 for ID, 0 for OOD.\n",
    "      features.update(res['feature'])\n",
    "      curves.update(energy, target)\n",
    "  return {\n",
    "    'features': (features.compute().mean(dim=0), features.compute().std(dim=0)),\n",
    "    'curves': curves\n",
    "  }"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## CIFAR-100"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* No Sparsity"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['x',\n",
       " 'conv1',\n",
       " 'block1.layer.0.bn1',\n",
       " 'block1.layer.0.relu',\n",
       " 'block1.layer.0.conv1',\n",
       " 'block1.layer.0.bn2',\n",
       " 'block1.layer.0.relu_1',\n",
       " 'block1.layer.0.conv2',\n",
       " 'block1.layer.0.cat',\n",
       " 'block1.layer.1.bn1',\n",
       " 'block1.layer.1.relu',\n",
       " 'block1.layer.1.conv1',\n",
       " 'block1.layer.1.bn2',\n",
       " 'block1.layer.1.relu_1',\n",
       " 'block1.layer.1.conv2',\n",
       " 'block1.layer.1.cat',\n",
       " 'block1.layer.2.bn1',\n",
       " 'block1.layer.2.relu',\n",
       " 'block1.layer.2.conv1',\n",
       " 'block1.layer.2.bn2',\n",
       " 'block1.layer.2.relu_1',\n",
       " 'block1.layer.2.conv2',\n",
       " 'block1.layer.2.cat',\n",
       " 'block1.layer.3.bn1',\n",
       " 'block1.layer.3.relu',\n",
       " 'block1.layer.3.conv1',\n",
       " 'block1.layer.3.bn2',\n",
       " 'block1.layer.3.relu_1',\n",
       " 'block1.layer.3.conv2',\n",
       " 'block1.layer.3.cat',\n",
       " 'block1.layer.4.bn1',\n",
       " 'block1.layer.4.relu',\n",
       " 'block1.layer.4.conv1',\n",
       " 'block1.layer.4.bn2',\n",
       " 'block1.layer.4.relu_1',\n",
       " 'block1.layer.4.conv2',\n",
       " 'block1.layer.4.cat',\n",
       " 'block1.layer.5.bn1',\n",
       " 'block1.layer.5.relu',\n",
       " 'block1.layer.5.conv1',\n",
       " 'block1.layer.5.bn2',\n",
       " 'block1.layer.5.relu_1',\n",
       " 'block1.layer.5.conv2',\n",
       " 'block1.layer.5.cat',\n",
       " 'block1.layer.6.bn1',\n",
       " 'block1.layer.6.relu',\n",
       " 'block1.layer.6.conv1',\n",
       " 'block1.layer.6.bn2',\n",
       " 'block1.layer.6.relu_1',\n",
       " 'block1.layer.6.conv2',\n",
       " 'block1.layer.6.cat',\n",
       " 'block1.layer.7.bn1',\n",
       " 'block1.layer.7.relu',\n",
       " 'block1.layer.7.conv1',\n",
       " 'block1.layer.7.bn2',\n",
       " 'block1.layer.7.relu_1',\n",
       " 'block1.layer.7.conv2',\n",
       " 'block1.layer.7.cat',\n",
       " 'block1.layer.8.bn1',\n",
       " 'block1.layer.8.relu',\n",
       " 'block1.layer.8.conv1',\n",
       " 'block1.layer.8.bn2',\n",
       " 'block1.layer.8.relu_1',\n",
       " 'block1.layer.8.conv2',\n",
       " 'block1.layer.8.cat',\n",
       " 'block1.layer.9.bn1',\n",
       " 'block1.layer.9.relu',\n",
       " 'block1.layer.9.conv1',\n",
       " 'block1.layer.9.bn2',\n",
       " 'block1.layer.9.relu_1',\n",
       " 'block1.layer.9.conv2',\n",
       " 'block1.layer.9.cat',\n",
       " 'block1.layer.10.bn1',\n",
       " 'block1.layer.10.relu',\n",
       " 'block1.layer.10.conv1',\n",
       " 'block1.layer.10.bn2',\n",
       " 'block1.layer.10.relu_1',\n",
       " 'block1.layer.10.conv2',\n",
       " 'block1.layer.10.cat',\n",
       " 'block1.layer.11.bn1',\n",
       " 'block1.layer.11.relu',\n",
       " 'block1.layer.11.conv1',\n",
       " 'block1.layer.11.bn2',\n",
       " 'block1.layer.11.relu_1',\n",
       " 'block1.layer.11.conv2',\n",
       " 'block1.layer.11.cat',\n",
       " 'block1.layer.12.bn1',\n",
       " 'block1.layer.12.relu',\n",
       " 'block1.layer.12.conv1',\n",
       " 'block1.layer.12.bn2',\n",
       " 'block1.layer.12.relu_1',\n",
       " 'block1.layer.12.conv2',\n",
       " 'block1.layer.12.cat',\n",
       " 'block1.layer.13.bn1',\n",
       " 'block1.layer.13.relu',\n",
       " 'block1.layer.13.conv1',\n",
       " 'block1.layer.13.bn2',\n",
       " 'block1.layer.13.relu_1',\n",
       " 'block1.layer.13.conv2',\n",
       " 'block1.layer.13.cat',\n",
       " 'block1.layer.14.bn1',\n",
       " 'block1.layer.14.relu',\n",
       " 'block1.layer.14.conv1',\n",
       " 'block1.layer.14.bn2',\n",
       " 'block1.layer.14.relu_1',\n",
       " 'block1.layer.14.conv2',\n",
       " 'block1.layer.14.cat',\n",
       " 'block1.layer.15.bn1',\n",
       " 'block1.layer.15.relu',\n",
       " 'block1.layer.15.conv1',\n",
       " 'block1.layer.15.bn2',\n",
       " 'block1.layer.15.relu_1',\n",
       " 'block1.layer.15.conv2',\n",
       " 'block1.layer.15.cat',\n",
       " 'trans1.bn1',\n",
       " 'trans1.relu',\n",
       " 'trans1.conv1',\n",
       " 'trans1.avg_pool2d',\n",
       " 'block2.layer.0.bn1',\n",
       " 'block2.layer.0.relu',\n",
       " 'block2.layer.0.conv1',\n",
       " 'block2.layer.0.bn2',\n",
       " 'block2.layer.0.relu_1',\n",
       " 'block2.layer.0.conv2',\n",
       " 'block2.layer.0.cat',\n",
       " 'block2.layer.1.bn1',\n",
       " 'block2.layer.1.relu',\n",
       " 'block2.layer.1.conv1',\n",
       " 'block2.layer.1.bn2',\n",
       " 'block2.layer.1.relu_1',\n",
       " 'block2.layer.1.conv2',\n",
       " 'block2.layer.1.cat',\n",
       " 'block2.layer.2.bn1',\n",
       " 'block2.layer.2.relu',\n",
       " 'block2.layer.2.conv1',\n",
       " 'block2.layer.2.bn2',\n",
       " 'block2.layer.2.relu_1',\n",
       " 'block2.layer.2.conv2',\n",
       " 'block2.layer.2.cat',\n",
       " 'block2.layer.3.bn1',\n",
       " 'block2.layer.3.relu',\n",
       " 'block2.layer.3.conv1',\n",
       " 'block2.layer.3.bn2',\n",
       " 'block2.layer.3.relu_1',\n",
       " 'block2.layer.3.conv2',\n",
       " 'block2.layer.3.cat',\n",
       " 'block2.layer.4.bn1',\n",
       " 'block2.layer.4.relu',\n",
       " 'block2.layer.4.conv1',\n",
       " 'block2.layer.4.bn2',\n",
       " 'block2.layer.4.relu_1',\n",
       " 'block2.layer.4.conv2',\n",
       " 'block2.layer.4.cat',\n",
       " 'block2.layer.5.bn1',\n",
       " 'block2.layer.5.relu',\n",
       " 'block2.layer.5.conv1',\n",
       " 'block2.layer.5.bn2',\n",
       " 'block2.layer.5.relu_1',\n",
       " 'block2.layer.5.conv2',\n",
       " 'block2.layer.5.cat',\n",
       " 'block2.layer.6.bn1',\n",
       " 'block2.layer.6.relu',\n",
       " 'block2.layer.6.conv1',\n",
       " 'block2.layer.6.bn2',\n",
       " 'block2.layer.6.relu_1',\n",
       " 'block2.layer.6.conv2',\n",
       " 'block2.layer.6.cat',\n",
       " 'block2.layer.7.bn1',\n",
       " 'block2.layer.7.relu',\n",
       " 'block2.layer.7.conv1',\n",
       " 'block2.layer.7.bn2',\n",
       " 'block2.layer.7.relu_1',\n",
       " 'block2.layer.7.conv2',\n",
       " 'block2.layer.7.cat',\n",
       " 'block2.layer.8.bn1',\n",
       " 'block2.layer.8.relu',\n",
       " 'block2.layer.8.conv1',\n",
       " 'block2.layer.8.bn2',\n",
       " 'block2.layer.8.relu_1',\n",
       " 'block2.layer.8.conv2',\n",
       " 'block2.layer.8.cat',\n",
       " 'block2.layer.9.bn1',\n",
       " 'block2.layer.9.relu',\n",
       " 'block2.layer.9.conv1',\n",
       " 'block2.layer.9.bn2',\n",
       " 'block2.layer.9.relu_1',\n",
       " 'block2.layer.9.conv2',\n",
       " 'block2.layer.9.cat',\n",
       " 'block2.layer.10.bn1',\n",
       " 'block2.layer.10.relu',\n",
       " 'block2.layer.10.conv1',\n",
       " 'block2.layer.10.bn2',\n",
       " 'block2.layer.10.relu_1',\n",
       " 'block2.layer.10.conv2',\n",
       " 'block2.layer.10.cat',\n",
       " 'block2.layer.11.bn1',\n",
       " 'block2.layer.11.relu',\n",
       " 'block2.layer.11.conv1',\n",
       " 'block2.layer.11.bn2',\n",
       " 'block2.layer.11.relu_1',\n",
       " 'block2.layer.11.conv2',\n",
       " 'block2.layer.11.cat',\n",
       " 'block2.layer.12.bn1',\n",
       " 'block2.layer.12.relu',\n",
       " 'block2.layer.12.conv1',\n",
       " 'block2.layer.12.bn2',\n",
       " 'block2.layer.12.relu_1',\n",
       " 'block2.layer.12.conv2',\n",
       " 'block2.layer.12.cat',\n",
       " 'block2.layer.13.bn1',\n",
       " 'block2.layer.13.relu',\n",
       " 'block2.layer.13.conv1',\n",
       " 'block2.layer.13.bn2',\n",
       " 'block2.layer.13.relu_1',\n",
       " 'block2.layer.13.conv2',\n",
       " 'block2.layer.13.cat',\n",
       " 'block2.layer.14.bn1',\n",
       " 'block2.layer.14.relu',\n",
       " 'block2.layer.14.conv1',\n",
       " 'block2.layer.14.bn2',\n",
       " 'block2.layer.14.relu_1',\n",
       " 'block2.layer.14.conv2',\n",
       " 'block2.layer.14.cat',\n",
       " 'block2.layer.15.bn1',\n",
       " 'block2.layer.15.relu',\n",
       " 'block2.layer.15.conv1',\n",
       " 'block2.layer.15.bn2',\n",
       " 'block2.layer.15.relu_1',\n",
       " 'block2.layer.15.conv2',\n",
       " 'block2.layer.15.cat',\n",
       " 'trans2.bn1',\n",
       " 'trans2.relu',\n",
       " 'trans2.conv1',\n",
       " 'trans2.avg_pool2d',\n",
       " 'block3.layer.0.bn1',\n",
       " 'block3.layer.0.relu',\n",
       " 'block3.layer.0.conv1',\n",
       " 'block3.layer.0.bn2',\n",
       " 'block3.layer.0.relu_1',\n",
       " 'block3.layer.0.conv2',\n",
       " 'block3.layer.0.cat',\n",
       " 'block3.layer.1.bn1',\n",
       " 'block3.layer.1.relu',\n",
       " 'block3.layer.1.conv1',\n",
       " 'block3.layer.1.bn2',\n",
       " 'block3.layer.1.relu_1',\n",
       " 'block3.layer.1.conv2',\n",
       " 'block3.layer.1.cat',\n",
       " 'block3.layer.2.bn1',\n",
       " 'block3.layer.2.relu',\n",
       " 'block3.layer.2.conv1',\n",
       " 'block3.layer.2.bn2',\n",
       " 'block3.layer.2.relu_1',\n",
       " 'block3.layer.2.conv2',\n",
       " 'block3.layer.2.cat',\n",
       " 'block3.layer.3.bn1',\n",
       " 'block3.layer.3.relu',\n",
       " 'block3.layer.3.conv1',\n",
       " 'block3.layer.3.bn2',\n",
       " 'block3.layer.3.relu_1',\n",
       " 'block3.layer.3.conv2',\n",
       " 'block3.layer.3.cat',\n",
       " 'block3.layer.4.bn1',\n",
       " 'block3.layer.4.relu',\n",
       " 'block3.layer.4.conv1',\n",
       " 'block3.layer.4.bn2',\n",
       " 'block3.layer.4.relu_1',\n",
       " 'block3.layer.4.conv2',\n",
       " 'block3.layer.4.cat',\n",
       " 'block3.layer.5.bn1',\n",
       " 'block3.layer.5.relu',\n",
       " 'block3.layer.5.conv1',\n",
       " 'block3.layer.5.bn2',\n",
       " 'block3.layer.5.relu_1',\n",
       " 'block3.layer.5.conv2',\n",
       " 'block3.layer.5.cat',\n",
       " 'block3.layer.6.bn1',\n",
       " 'block3.layer.6.relu',\n",
       " 'block3.layer.6.conv1',\n",
       " 'block3.layer.6.bn2',\n",
       " 'block3.layer.6.relu_1',\n",
       " 'block3.layer.6.conv2',\n",
       " 'block3.layer.6.cat',\n",
       " 'block3.layer.7.bn1',\n",
       " 'block3.layer.7.relu',\n",
       " 'block3.layer.7.conv1',\n",
       " 'block3.layer.7.bn2',\n",
       " 'block3.layer.7.relu_1',\n",
       " 'block3.layer.7.conv2',\n",
       " 'block3.layer.7.cat',\n",
       " 'block3.layer.8.bn1',\n",
       " 'block3.layer.8.relu',\n",
       " 'block3.layer.8.conv1',\n",
       " 'block3.layer.8.bn2',\n",
       " 'block3.layer.8.relu_1',\n",
       " 'block3.layer.8.conv2',\n",
       " 'block3.layer.8.cat',\n",
       " 'block3.layer.9.bn1',\n",
       " 'block3.layer.9.relu',\n",
       " 'block3.layer.9.conv1',\n",
       " 'block3.layer.9.bn2',\n",
       " 'block3.layer.9.relu_1',\n",
       " 'block3.layer.9.conv2',\n",
       " 'block3.layer.9.cat',\n",
       " 'block3.layer.10.bn1',\n",
       " 'block3.layer.10.relu',\n",
       " 'block3.layer.10.conv1',\n",
       " 'block3.layer.10.bn2',\n",
       " 'block3.layer.10.relu_1',\n",
       " 'block3.layer.10.conv2',\n",
       " 'block3.layer.10.cat',\n",
       " 'block3.layer.11.bn1',\n",
       " 'block3.layer.11.relu',\n",
       " 'block3.layer.11.conv1',\n",
       " 'block3.layer.11.bn2',\n",
       " 'block3.layer.11.relu_1',\n",
       " 'block3.layer.11.conv2',\n",
       " 'block3.layer.11.cat',\n",
       " 'block3.layer.12.bn1',\n",
       " 'block3.layer.12.relu',\n",
       " 'block3.layer.12.conv1',\n",
       " 'block3.layer.12.bn2',\n",
       " 'block3.layer.12.relu_1',\n",
       " 'block3.layer.12.conv2',\n",
       " 'block3.layer.12.cat',\n",
       " 'block3.layer.13.bn1',\n",
       " 'block3.layer.13.relu',\n",
       " 'block3.layer.13.conv1',\n",
       " 'block3.layer.13.bn2',\n",
       " 'block3.layer.13.relu_1',\n",
       " 'block3.layer.13.conv2',\n",
       " 'block3.layer.13.cat',\n",
       " 'block3.layer.14.bn1',\n",
       " 'block3.layer.14.relu',\n",
       " 'block3.layer.14.conv1',\n",
       " 'block3.layer.14.bn2',\n",
       " 'block3.layer.14.relu_1',\n",
       " 'block3.layer.14.conv2',\n",
       " 'block3.layer.14.cat',\n",
       " 'block3.layer.15.bn1',\n",
       " 'block3.layer.15.relu',\n",
       " 'block3.layer.15.conv1',\n",
       " 'block3.layer.15.bn2',\n",
       " 'block3.layer.15.relu_1',\n",
       " 'block3.layer.15.conv2',\n",
       " 'block3.layer.15.cat',\n",
       " 'bn1',\n",
       " 'relu',\n",
       " 'avg_pool2d',\n",
       " 'view',\n",
       " 'fc']"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "densenet = DenseNet3(100, 100, 12, reduction=0.5, bottleneck=True, dropRate=0.0, normalizer=None, p=None, info=None)\n",
    "checkpoint = torch.load(\"./checkpoints/CIFAR-100/densenet/checkpoint_100.pth.tar\", \"cpu\")\n",
    "densenet.load_state_dict(checkpoint['state_dict'])\n",
    "densenet.eval()\n",
    "\n",
    "get_graph_node_names(densenet)[1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "return_nodes = {'view': 'feature', 'fc': 'logit'}\n",
    "extractor = create_feature_extractor(densenet, return_nodes).to(device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6740899060dc4510bf23d119ed8c013f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/20 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "curves.reset()\n",
    "\n",
    "res_id = run(extractor, dataloaders['CIFAR-100'], False, curves)\n",
    "\n",
    "id_avg, id_std = res_id['features']\n",
    "id_curves = res_id['curves']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "832e815260ee49598fba699ea8f433e6",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/20 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "4c854b8483404ea98608246e3a6638d3",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/20 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "da4cdc2e53984630933fa2568e650c01",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/20 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "02ea32f806b1472e8de5eae83aa82cd3",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/18 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "9cc7e80487734c27b5b4e83d446dfae4",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/12 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "065f903f659e4767ad248af3e9d91218",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/642 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "ood_features = {}\n",
    "ood_metrics = {}\n",
    "\n",
    "for nm in ood_dls:\n",
    "  curves = id_curves.clone()\n",
    "\n",
    "  res_ood = run(extractor, dataloaders[nm], True, curves)\n",
    "\n",
    "  ood_features[nm] = res_ood['features']\n",
    "  ood_metrics[nm] = compute_metrics(res_ood['curves'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>FPR@95</th>\n",
       "      <th>AUROC</th>\n",
       "      <th>AUPR_In</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>SVHN</th>\n",
       "      <td>0.875400</td>\n",
       "      <td>0.818515</td>\n",
       "      <td>0.863067</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>LSUN</th>\n",
       "      <td>0.147800</td>\n",
       "      <td>0.974348</td>\n",
       "      <td>0.976312</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>LSUN_resize</th>\n",
       "      <td>0.707300</td>\n",
       "      <td>0.801376</td>\n",
       "      <td>0.814240</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>iSUN</th>\n",
       "      <td>0.746106</td>\n",
       "      <td>0.789539</td>\n",
       "      <td>0.817554</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>dtd</th>\n",
       "      <td>0.843617</td>\n",
       "      <td>0.710096</td>\n",
       "      <td>0.765005</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>places365</th>\n",
       "      <td>0.782843</td>\n",
       "      <td>0.781495</td>\n",
       "      <td>0.141730</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Avg.</th>\n",
       "      <td>0.683844</td>\n",
       "      <td>0.812562</td>\n",
       "      <td>0.729651</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "               FPR@95     AUROC   AUPR_In\n",
       "SVHN         0.875400  0.818515  0.863067\n",
       "LSUN         0.147800  0.974348  0.976312\n",
       "LSUN_resize  0.707300  0.801376  0.814240\n",
       "iSUN         0.746106  0.789539  0.817554\n",
       "dtd          0.843617  0.710096  0.765005\n",
       "places365    0.782843  0.781495  0.141730\n",
       "Avg.         0.683844  0.812562  0.729651"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pd.DataFrame(ood_metrics)\n",
    "df['Avg.'] = df.mean(axis=1)\n",
    "result_dfs['CIFAR-100_energy'] = df.T\n",
    "\n",
    "df.T[['FPR@95', 'AUROC', 'AUPR_In']]"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* With Sparsity $p=90$"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "idx = id_avg.argsort()\n",
    "\n",
    "plt.title('Features')\n",
    "plt.plot(range(len(idx)), id_avg[idx], color='red', label='CIFAR-100')\n",
    "plt.fill_between(\n",
    "  range(len(idx)),\n",
    "  id_avg[idx] - 2*id_std[idx],\n",
    "  id_avg[idx] + 2*id_std[idx],\n",
    "  color='red', alpha=.1\n",
    ")\n",
    "\n",
    "for k, (avg, std) in ood_features.items():\n",
    "  plt.scatter(range(len(idx)), avg[idx], s=1, label=k)\n",
    "  # plt.fill_between(range(len(idx)), avg[idx] - 2*std[idx], avg[idx] + 2*std[idx], alpha=.1)\n",
    "\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "weight = extractor.fc.weight.data.cpu()\n",
    "contrib = id_avg * weight\n",
    "thresh = np.percentile(contrib, 90)\n",
    "mask = contrib > thresh\n",
    "extractor.fc.weight.data = (mask * weight).to(device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "id_avg = (id_avg * weight).flatten()\n",
    "id_std = (id_std * weight).flatten()\n",
    "idx = id_avg.argsort()\n",
    "\n",
    "plt.title('Contribution')\n",
    "plt.plot(range(len(idx)), id_avg[idx], color='red', label='CIFAR-100')\n",
    "plt.fill_between(\n",
    "  range(len(idx)),\n",
    "  id_avg[idx] - 2*id_std[idx],\n",
    "  id_avg[idx] + 2*id_std[idx],\n",
    "  color='red', alpha=.1\n",
    ")\n",
    "\n",
    "for k, (avg, std) in ood_features.items():\n",
    "  avg, std = (avg * weight).flatten(), (std * weight).flatten()\n",
    "  plt.scatter(range(len(idx)), avg[idx], s=1, label=k)\n",
    "  # plt.fill_between(range(len(idx)), avg[idx] - 2*std[idx], avg[idx] + 2*std[idx], alpha=.1)\n",
    "\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "c504fbfabe71427cb1411b7dfea7d9b1",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/20 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "curves.reset()\n",
    "\n",
    "res_id = run(extractor, dataloaders['CIFAR-100'], False, curves)\n",
    "\n",
    "id_avg, id_std = res_id['features']\n",
    "id_curves = res_id['curves']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "12726d4c9c8e48ad9234967b67f4f069",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/20 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "4be5d71a98364304a6887c7c12784956",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/20 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b7704ab0a2264125876983d9d9f976d2",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/20 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "bdcca6ed084149e1b6afd0fd33b4dc1b",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/18 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "bef991498ef14368ac26fc0750f11d3d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/12 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "02c907f26d474c96a329b22d62ee29af",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/642 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "ood_features = {}\n",
    "ood_metrics = {}\n",
    "\n",
    "for nm in ood_dls:\n",
    "  curves = id_curves.clone()\n",
    "\n",
    "  res_ood = run(extractor, dataloaders[nm], True, curves)\n",
    "\n",
    "  ood_features[nm] = res_ood['features']\n",
    "  ood_metrics[nm] = compute_metrics(res_ood['curves'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>FPR@95</th>\n",
       "      <th>AUROC</th>\n",
       "      <th>AUPR_In</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>SVHN</th>\n",
       "      <td>0.591500</td>\n",
       "      <td>0.886079</td>\n",
       "      <td>0.903695</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>LSUN</th>\n",
       "      <td>0.009200</td>\n",
       "      <td>0.997392</td>\n",
       "      <td>0.997412</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>LSUN_resize</th>\n",
       "      <td>0.516800</td>\n",
       "      <td>0.893147</td>\n",
       "      <td>0.902909</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>iSUN</th>\n",
       "      <td>0.494006</td>\n",
       "      <td>0.894948</td>\n",
       "      <td>0.909927</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>dtd</th>\n",
       "      <td>0.612234</td>\n",
       "      <td>0.771527</td>\n",
       "      <td>0.793554</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>places365</th>\n",
       "      <td>0.802904</td>\n",
       "      <td>0.774476</td>\n",
       "      <td>0.140147</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Avg.</th>\n",
       "      <td>0.504441</td>\n",
       "      <td>0.869595</td>\n",
       "      <td>0.774607</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "               FPR@95     AUROC   AUPR_In\n",
       "SVHN         0.591500  0.886079  0.903695\n",
       "LSUN         0.009200  0.997392  0.997412\n",
       "LSUN_resize  0.516800  0.893147  0.902909\n",
       "iSUN         0.494006  0.894948  0.909927\n",
       "dtd          0.612234  0.771527  0.793554\n",
       "places365    0.802904  0.774476  0.140147\n",
       "Avg.         0.504441  0.869595  0.774607"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pd.DataFrame(ood_metrics)\n",
    "df['Avg.'] = df.mean(axis=1)\n",
    "result_dfs['CIFAR-100_dice'] = df.T\n",
    "\n",
    "df.T[['FPR@95', 'AUROC', 'AUPR_In']]"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## CIFAR-10"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* No Sparsity"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "densenet = DenseNet3(100, 10, 12, reduction=0.5, bottleneck=True, dropRate=0.0, normalizer=None, p=None, info=None)\n",
    "checkpoint = torch.load(\"./checkpoints/CIFAR-10/densenet/checkpoint_100.pth.tar\", \"cpu\")\n",
    "densenet.load_state_dict(checkpoint['state_dict'])\n",
    "densenet.eval()\n",
    "\n",
    "return_nodes = {'view': 'feature', 'fc': 'logit'}\n",
    "extractor = create_feature_extractor(densenet, return_nodes).to(device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "c5d8dbf612554f10815f386763351362",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/20 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "curves.reset()\n",
    "\n",
    "res_id = run(extractor, dataloaders['CIFAR-10'], False, curves)\n",
    "\n",
    "id_avg, id_std = res_id['features']\n",
    "id_curves = res_id['curves']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "265dc2161cd548ce9c4d41443229a41c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/20 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "9cf238832e52458aae437f3ffc6881c2",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/20 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "3e8bae9532e249529af1bb541c6f2165",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/20 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "8ed6be4bb4774057bb2b198be1e53512",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/18 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "acb2a94abda544218c29793e46682d70",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/12 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "40fc59c94ec84f13b9a2373c81701f7d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/642 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "ood_features = {}\n",
    "ood_metrics = {}\n",
    "\n",
    "for nm in ood_dls:\n",
    "  curves = id_curves.clone()\n",
    "\n",
    "  res_ood = run(extractor, dataloaders[nm], True, curves)\n",
    "\n",
    "  ood_features[nm] = res_ood['features']\n",
    "  ood_metrics[nm] = compute_metrics(res_ood['curves'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>FPR@95</th>\n",
       "      <th>AUROC</th>\n",
       "      <th>AUPR_In</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>SVHN</th>\n",
       "      <td>0.406300</td>\n",
       "      <td>0.939926</td>\n",
       "      <td>0.953340</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>LSUN</th>\n",
       "      <td>0.038100</td>\n",
       "      <td>0.991504</td>\n",
       "      <td>0.992562</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>LSUN_resize</th>\n",
       "      <td>0.092800</td>\n",
       "      <td>0.981238</td>\n",
       "      <td>0.984919</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>iSUN</th>\n",
       "      <td>0.100616</td>\n",
       "      <td>0.980683</td>\n",
       "      <td>0.986024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>dtd</th>\n",
       "      <td>0.563121</td>\n",
       "      <td>0.864202</td>\n",
       "      <td>0.898692</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>places365</th>\n",
       "      <td>0.397665</td>\n",
       "      <td>0.918054</td>\n",
       "      <td>0.385097</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Avg.</th>\n",
       "      <td>0.266434</td>\n",
       "      <td>0.945935</td>\n",
       "      <td>0.866772</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "               FPR@95     AUROC   AUPR_In\n",
       "SVHN         0.406300  0.939926  0.953340\n",
       "LSUN         0.038100  0.991504  0.992562\n",
       "LSUN_resize  0.092800  0.981238  0.984919\n",
       "iSUN         0.100616  0.980683  0.986024\n",
       "dtd          0.563121  0.864202  0.898692\n",
       "places365    0.397665  0.918054  0.385097\n",
       "Avg.         0.266434  0.945935  0.866772"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pd.DataFrame(ood_metrics)\n",
    "df['Avg.'] = df.mean(axis=1)\n",
    "result_dfs['CIFAR-10_energy'] = df.T\n",
    "\n",
    "df.T[['FPR@95', 'AUROC', 'AUPR_In']]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* With Sparsity $p=90$"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAiQAAAGzCAYAAAAfeAwWAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjYuMywgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/P9b71AAAACXBIWXMAAA9hAAAPYQGoP6dpAACIyElEQVR4nO3dd3gU1frA8e+WFBJSgNBLKKGXUEMXAZVwr4oIKFdEQEBRforYUbioqIhdAZUmSlEUBAtXglKUKgjSS2ihBkJJ78lmfn/MzmZ3swlJyGY3yft5nnmSnZ2dOTsu7ptz3vMeHaAghBBCCOFCelc3QAghhBBCAhIhhBBCuJwEJEIIIYRwOQlIhBBCCOFyEpAIIYQQwuUkIBFCCCGEy0lAIoQQQgiXk4BECCGEEC4nAYkQQgghXE4CEiGEEEK4nAQkQlRAo0aNQlEUh9vMmTNL/Hrdu3dn+vTpBAQElPi5hRDlg9HVDRBCuM60adOIioqy2Xf48OESv06PHj147bXX+Oqrr0hISCjx8wshyj4JSISowNatW8fevXtd3Yxi8/HxITU11dXNEEKUABmyEUI4FB4ezpYtW0hOTiYxMZG1a9fSqlUrm2Patm3L4sWLOX36NGlpaVy+fJlFixZRtWpVyzHTp0/n/fffB+Ds2bOWoaHg4GCCg4NRFIVRo0blub6iKEyfPt3mPIqi0LJlS5YvX05sbCzbtm2zPD9ixAj27NlDamoqN27c4Ntvv6VevXo25wwJCWHVqlVcvnyZtLQ0Lly4wLfffou/v3+J3DMhRPFJD4kQFVhAQADVqlWz2Xfjxg0efvhhvv76a9avX89LL72Ej48PTzzxBNu2baNDhw6cO3cOgDvvvJPGjRuzePFirly5QuvWrXnsscdo3bo13bp1A2D16tU0a9aMhx56iGeeeYbr168DcO3aNapXr17kNq9cuZKTJ0/yyiuvoNPpAHjllVeYMWMG33//PQsXLqR69eo89dRTbNmyhQ4dOpCQkICHhwfr16/Hy8uL2bNnc+XKFerWrcvdd99NYGAgiYmJt3IrhRAlQJFNNtkq1jZq1CglP76+vkpsbKwyb948m9fUqFFDiYuLs9nv7e2d59wPPvigoiiK0qtXL8u+5557TlEURQkODrY5Njg4WFEURRk1alSe8yiKokyfPt3yePr06YqiKMry5cttjmvQoIGSlZWlTJkyxWZ/69atlczMTMv+0NBQRVEUZciQIS6//7LJJlveTXpIhKjAnnzySU6cOGGz784776RKlSp8++23Nr0nJpOJXbt20bdvX8u+9PR0y+9eXl5UrlyZv/76C4COHTvaDKmUlC+++MLm8f33349er+f777+3ae+VK1c4efIkffv2ZebMmZZk2gEDBvDrr7+SlpZW4m0TQhSfBCRCVGC7d+/Ok9T6wgsvALB582aHr7GeJVOlShWmT5/O8OHDqVmzps1xzpriaz8rqGnTpuj1ek6dOuXw+KysLEDNX/nggw947rnnGDFiBFu3buXnn39m2bJlMlwjhBuQgEQIYUOvV3PdH374Ya5cuZLn+ezsbMvv33//PT169OC9995j//79JCcno9frWb9+veU8BVEUpcA2OGLfs6HX68nJyWHgwIGYTKY8xycnJ1t+f/755/nqq68YNGgQd911F59++ilTpkyhW7duXLp06abtFUI4jwQkQggbp0+fBuDq1ats3Lgx3+MCAwO54447+O9//8uMGTMs+0NCQvIcm1/gERcXZzmXteDg4CK1V6/XExUVxcmTJ296/OHDhzl8+DBvvfUW3bt3Z8eOHUyYMIFp06YV+ppCiJIn036FEDbWr19PQkICr7zyCkZj3r9ZgoKCACy9EdpMF80zzzyT5zUpKSlA3sAjKSmJa9eucdttt9nsf/LJJwvd3tWrV5OdnW0zRdiaNgXZz88Pg8Fg89yhQ4cwmUx4eXkV+npCCOeQHhIhhI2kpCSeeOIJli5dyj///MOKFSu4du0aDRo04N///jfbt2/nqaeeIikpiT///JMXX3wRDw8PLl26xF133UWjRo3ynFPLU3nrrbdYsWIFWVlZ/PLLL6SmprJw4UKmTJnCggUL2LNnD7fddhvNmjUrdHvPnDnD1KlTeeedd2jYsCE//vgjSUlJNGrUiMGDBzN//nw++OAD+vXrx5w5c1i5ciUnTpzAaDQycuRITCYTP/zwQ4ndPyFE8bl8qo9ssslWups27bdTp075HtOnTx9l3bp1SlxcnJKamqqcPHlS+fLLL5WOHTtajqlTp47yww8/KLGxsUpcXJzy3XffKbVq1cozZRdQXn31VeXChQtKdna2zRRgb29vZcGCBUpcXJySkJCgrFixQgkKCsp32m+1atUctnfw4MHKli1blKSkJCUpKUk5evSoMnv2bKVp06YKoDRs2FBZuHChcvLkSSU1NVW5fv26snHjRqVfv34u/+8hm2yyoejMvwghhBBCuIzkkAghhBDC5SQgEUIIIYTLSUAihBBCCJeTgEQIIYQQLicBiRBCCCFcTgISIYQQQric2xdGq1OnDklJSa5uhhBCCCGKwM/Pj+jo6EIf79YBSZ06dWTBKyGEEKKMqlu3bqGDErcOSLSekbp160oviRBCCFFG+Pn5cenSpSJ9d7t1QKJJSkqSgEQIIYQoxySpVQghhBAuJwGJEEIIIVxOAhIhhBBCuFyZyCEpiE6nIzAwED8/P3Q6naubI1xIURSSkpKIj49HUWQRayGEKEvKdEBSvXp1xo8fT4sWLVzdFOFGjh8/zoIFC7h27ZqrmyKEEKKQdIDb/inp5+dHYmIi/v7+eWbZGI1GPvvsM5KTk/n++++5evUqJpPJRS0V7sBgMFCjRg0eeOABKleuzJNPPkl2drarmyWEEBVOQd/f+SmzPSS1a9fG29ub999/nxMnTri6OcJNnDlzhtjYWKZOnUqtWrW4ePGiq5skhBCiEMpsUqterzY9IyPDxS0R7kb7TBgMBhe3RAghRGGV2YBECCGEEOWHBCRCCCGEcDkJSIQQQgjhchKQuEjNmjX59NNPOX36NOnp6Zw/f56ff/6Zfv36ARAVFcWkSZMsx0dFRaEois124cIFm3NGRESQnZ1N586d81xv8eLFltdlZmZy5swZZs2ahZeXV4Ht7N27Nz///DOXLl1CURQGDRrk8LjXX3+d6OhoUlNT+f333wkJCSnqLRFCCFGBSUDiAsHBwezdu5d+/frxwgsv0LZtW8LDw9m8eTNz587N93XTpk2jVq1alq1Dhw6W5+rXr0+PHj2YM2cOjz76qMPXr1u3jlq1atG4cWMmT57M448/zuuvv15gW319fTlw4AATJ07M95gXX3yRp59+mgkTJtC1a1dSUlJYv379TYMdIYQQQlNmp/3mx8dF100twrGfffYZiqIQFhZGamruK48ePcqXX36Z7+uSkpKIiYlx+NyYMWNYu3Ytn3/+OX/99RfPPvss6enpNsdkZGRYXn/x4kU2bNjAnXfeycsvv5zvNSMiIoiIiCjw/TzzzDO8+eab/PzzzwA88sgjxMTEcN999/Hdd98V+FohhBCu4wF4AimubgjlrIfEB/WmumIrbCBUpUoVwsPDmTt3rk0woklISCjiu1aNGTOGZcuWERkZyalTpxg6dGiBx7du3ZoePXqQmZlZrOtpGjVqRO3atdmwYYNlX2JiIrt27aJ79+63dG4hhBDOFQQEuLoRZuUqICkLQkJC0Ov1HD9+vMivnTVrFklJSZbtqaeeAuCOO+7Ax8eH9evXA7Bs2TLGjh2b5/V33303SUlJpKWlcfjwYWrUqMF77713S++nVq1aAHl6bmJiYizPCSGEcF/uEgiUqyGbVMDXhdcujFtZAPC9997jq6++sjy+fv06AI8++ijfffedpXT+t99+y3vvvUfjxo05c+aM5fjNmzfzxBNP4Ovry+TJk8nOzmb16tUA9OrVi3Xr1lmOffzxx/nmm2+K3VYhhBDuT4+6how7KFcBCRQtl8MVTp48SU5OTrEWBLx+/TqnT5+22VelShUGDx6Mh4cHTzzxhGW/0Wjk0UcfZerUqZZ9KSkpltc/+uijHDhwgEcffZQvv/ySPXv20L59e8ux+eWq2Lty5QqgzhrSftce79+/v6hvUQghRCky4D4L2rlLT02FERcXx/r165k4cSI+PnkzTwICijaaN2LECC5evEhoaCjt27e3bM8++yyjR4+2lNi3pygKb7/9Nm+++Sbe3t6kp6dz+vRpy5acnFyo60dFRXH58mX69+9v2efn50fXrl3ZuXNnkd6LEEKI0uVOC2xIQOICEydOxGAwsHv3bu6//35CQkJo0aIFTz31VJG/xMeOHcuqVas4cuSIzbZo0SKCgoIIDw/P97UrV67EZDIVOKXX19eX0NBQQkNDATWJNTQ0lPr161uO+fjjj5k6dSr33HMPbdq0YcmSJURHR/Pjjz8W6b0IIYQoPQbcKwhwp7ZUGFFRUXTs2JHNmzfzwQcfcPjwYX7//Xf69+9vM+xyMx07dqR9+/b88MMPeZ5LTExk48aNDpNbNSaTiTlz5vDiiy867K0B6Ny5M/v377cMv3z00Ufs37+fN954w3LMu+++y+zZs5k/fz5///03lStXJjw8XBY+FEIIN2bAvXpIQB0+csvNz89PURRF8fPzy/NccHCwsmTJEiU4ONjl7ZTNvTb5bMgmm2yy3XyrBEpLUOo44dwFfX/nt0kPiRBCCFEByZCNEEIIIVzO3YZsJCARQgghKiDpIRFCCCGEy3m4ugF2JCARQgghKiB3Gq4BCUiEEEKICsmdqrSCBCRCCCFEhSQBiRBCCCFcTocEJEIIIYRwIW2V3xxXN8SKBCRCCCFEBaMzb+5EAhIXCAoK4rPPPuPcuXOkp6dz+fJlIiIiuO2227h27RovvfSSw9dNnTqVK1euYDQaGTVqFHFxcQ6PUxSFQYMG2TxOS0ujQYMGNsetWbOGxYsXl9wbE0IIUSZoAYn0kFRwP/zwAx06dGDUqFE0a9aMe++9lz/++IOAgACWLVvGmDFjHL5u9OjRLFmyhOzs7CJfU1EUmwXxhBBCVFx68+ZOOSRGVzegogkICOC2226jT58+bNmyBYDz58/z999/A+pKwM888ww9e/Zk+/btltf16dOHJk2asGjRomJdd86cOTz77LO89957HDly5NbfiBBCiDJL6yFxp4BEekhKWXJyMklJSdx33314enrmef7w4cPs3r2bRx991Gb/mDFj2L59O5GRkcW67vbt21m7di3vvPNOsV4vhBCi/JAhGzfWfdhgXl2/mu7DBjv1OiaTidGjRzNq1Cji4+PZtm0bb731Fm3btrUcs2jRIoYNG4avry8AlStXZujQoXz55Zc25woMDCQpKSnPlp8pU6YQHh5Or169nPPmhBBClAnakI07cbf2uEy/cSOpWqc2/caNdPq1Vq9eTZ06dbj33nuJiIjg9ttv559//mHUqFEAfPvttxgMBh544AEAHnzwQXJycvjuu+9szpOYmEj79u3zbPk5duwYS5YskV4SIYSo4NxtuAYkILHYtHApsdGX2bRwaalcLyMjgw0bNvDmm2/Ss2dPvvrqK15//XUAkpKSWLVqlSW5dcyYMXz//fekpKTYnCMnJ4fTp0/n2Qoyffp0OnbsaDMLRwghRMXijl/+7tgml9i5cg1vDbifnSvXuOT6R48etQzRgDps07t3b/7973/Ts2fPYiez2rt48SJz5szh7bffxmBwt6WVhBBClAapQyKoWrUqGzduZMSIEbRt25aGDRsydOhQXnzxRX766SfLcVu2bOHkyZMsWbKEY8eOsXPnzhJrw8yZM6lTpw533HFHiZ1TCCFE2SFDNoLk5GR27drF5MmT2bJlC4cPH2bGjBksWLCA//u//7M59ssvv6Rq1ap5kllvVVxcHLNmzaJSpUolel4hhBBlg7t++Svuuvn5+SmKoih+fn55ngsODlaWLFmiBAcHu7ydsrnXJp8N2WSTTbaCt6qgtAClKSh1nHD+gr6/89vcNUgSQgghhJO4W5VWcHKvzYQJEzhw4AAJCQkkJCSwY8cOwsPDnXlJIYQQQtyEgQoWkFy8eJGXX36ZTp060blzZzZt2sRPP/1Eq1atnHlZIYQQQhTAgHtVaQUnr2Wzdu1am8dTp07liSeeoFu3bhw9etSZlxZCCCFEPrQhG3ea+ltqi+vp9XpLOfT8prB6enri5eVleezn51dazRNCCCEqDCMVbMgGoE2bNiQlJZGRkcEXX3zB4MGDOXbsmMNjp0yZQmJiomW7dOmSs5snhBBCVDgVsg5JZGQk7du3p2vXrnz++ed8/fXXtGzZ0uGxM2fOxN/f37LVrVvX2c0TQgghKhStSqu7BSROH7LJysqyrK/yzz//0KVLFyZNmsSECRPyHJuZmUlmZqazmySEEEJUWHrcMyAp9Toker3eJk9ECCGEEKXHA7U3IsvVDbHj1B6St99+m3Xr1nH+/Hn8/Px46KGHuP322xkwYIAzLyuEEEKIfHhRAeuQ1KhRgyVLlhAZGcnGjRvp0qULAwYMYMOGDc68rNtbvHgxa9ascfhcu3bt+Omnn4iJiSEtLY2oqChWrFhB9erVAejTpw+KohAQEJDntVFRUUyaNMnyWFEU0tLSaNCggc1xa9asYfHixSX4joQQQpQVPkC2qxvhgFN7SMaNG+fM05c7QUFBbNy4kbVr1zJgwADi4+Np2LAh9957L76+vly7dq3I51QUhTfeeIPRo0eXfIOFEEKUKQagEuCO2ZqlVodE3FzPnj0JCAhg3LhxmEwmAM6ePcsff/xR7HPOmTOHZ599lvfee48jR46UUEuFEEKURV7mLd7F7XBEFtdzI1euXMHDw4PBgweX2Dm3b9/O2rVreeedd0rsnEIIIcomA+5VndWaBCRmjz8ezpmohTz+uOsW/9u1axdvvfUW33zzDdevX+fXX3/l+eefp0aNGrd03ilTphAeHk6vXr1KqKVCCCHKIndc5VcjAYnZSy8PpWHDmrz08lCXtmPq1KnUqlWLCRMmcOTIESZMmMDx48dp06ZNsc957NgxlixZIr0kQghRwbnzl747t61UzXpnFWfPxjDrnVWubgqxsbGsWrWKF154gZYtWxIdHc3zzz8PQGJiIoDDWTaBgYEkJCQ4POf06dPp2LEjgwYNcl7DhRBCuDV3XOVXIwGJ2bx5ETRuNI558yJc3RQbWqVbX19fAE6ePInJZKJTp042xzVq1IjAwEBOnDjh8DwXL15kzpw5vP322xgMBqe3WwghhPsx4r4BicyycZGAgABCQ0Nt9rVt25YBAwawYsUKTpw4gU6n45577uFf//oXY8aMASA5OZmFCxfywQcfkJ2dzaFDh6hfvz6zZs1i586d7NixI99rzpw5k/Hjx9OoUSO+++47p74/IYQQ7kcCEpFH37592b9/v82+TZs2cerUKT744APq169PRkYGJ0+eZNy4cSxbtsxy3KRJk3j55ZeZNWsWwcHBXLlyhd9//51XX321wGvGxcUxa9YsZs6c6Yy3JIQQws0Zcd+kVlDb5pabn5+foiiK4ufnl+e54OBgZcmSJUpwcLDL2ymbe23y2ZBNNtlky7vpQGkMSkNQapu3pqDUccK1Cvr+zm+THBIhhBCiAtBW+XXXIRsJSIQQQogKQG/eJCARQgghhMtIQCKEEEIIl9MCEsXVDcmHBCRCCCFEBeDuX/ju3j4hhBBClAB3/8J39/YJIYQQ4hbo7X66K3dvnxBCCCGKyQeobv7d3b/w3b19QgghhCgmPeBl/unOC+uBBCTCBRYvXsyaNWtc3QwhhCj39Kjl4o2AN2BybXMKJAGJCxT0hdyuXTt++uknYmJiSEtLIyoqihUrVlC9utrp1qdPHxRFISAgIM9ro6KimDRpkuWxoiikpaXRoEEDm+PWrFnD4sWLS/AdFc2kSZMYPXq0y64vhBAVhQ7wtNqyXducAklA4kaCgoLYuHEjsbGxDBgwgJYtWzJmzBiio6Px9fUt1jkVReGNN94osTZ6eHjc8jkSExNJSEgogdYIIYQoiNZDUsn8M8u1zSmQBCRupGfPngQEBDBu3Dj279/P2bNn+eOPP3j22Wc5e/Zssc45Z84cHn74YVq3bl2s12/evJnZs2fz0Ucfce3aNdavXw9A69at+fXXX0lKSuLKlSssWbKEatWqWV43ZMgQDh48SGpqKtevX+f333/Hx8cHsO0hCg4ORlGUPNvmzZtt7suWLVtITU3l/PnzfPLJJ5ZzCSGEyJ+O3CEbySERhXblyhU8PDwYPHhwiZ1z+/btrF27lnfeeafY5xg1ahSZmZn07NmTCRMmEBAQwKZNm9i3bx+dO3cmPDycmjVr8v333wNQq1Ytvv32W7788ktatmzJ7bffzurVq9HpdHnOfeHCBWrVqmXZ2rdvz/Xr19myZQsAjRs3JiIigh9++IF27drx4IMP0qtXL+bMmVPs9yOEEBWF0fzTgBqcuDPjzQ+pGO65ezgPDX+Mb1bM55e1K1zShl27dvHWW2/xzTff8MUXX7B79242bdrEkiVLuHr1arHPO2XKFA4ePEivXr3Ytm1bkV9/8uRJXnrpJcvjV199lX379vHqq69a9j366KNcvHiRpk2bUrlyZTw8PFi9ejXnz58H4PDhww7PnZOTQ0xMDABeXl78+OOP7Ny5k9dee83S9uXLl/PJJ58AcOrUKZ5++mn+/PNPnnjiCTIyMor8foQQoqLQekW8cO/8EZAeEouHhj9GrZp1eWj4Yy5tx9SpU6lVqxYTJkzgyJEjTJgwgePHj9OmTZtin/PYsWMsWbKk2L0ke/futXkcGhpK3759SUpKsmzHjx8HoEmTJhw4cIANGzZw6NAhvv/+e8aNG0dgYOBNr/Pll1/i5+fHQw89hKIolmuNHj3a5lrr16/HYDDQqFGjYr0fIYSoKAyogYg3EpCUGd+smM+VmEt8s2K+q5tCbGwsq1at4oUXXqBly5ZER0fz/PPPA2pCKOBwlk1gYGC+yaLTp0+nY8eODBo0qMjtSUlJsXlcuXJlfvnlF9q3b2+zhYSEsGXLFnJycrjzzjsZOHAgR48e5amnniIyMpKGDRvme41XX32VAQMGcO+995KcnGxzrXnz5tlcJzQ0lJCQEE6fPl3k9yKEEBWJHshA7SVx9/5kGbIx+2XtCpcN1RQkKyuL06dPW2bZnDx5EpPJRKdOnSzDIQCNGjUiMDCQEydOODzPxYsXmTNnDm+//fYtf5H/888/DBkyhLNnz2Iy5T+rfceOHezYsYM33niDc+fOMXjwYD766KM8x91///3897//ZeDAgZw5cybPtVq1aiXBhxBCFIM2ZBPv4nYUhgQkLhIQEEBoaKjNvrZt2zJgwABWrFjBiRMn0Ol03HPPPfzrX/9izJgxACQnJ7Nw4UI++OADsrOzOXToEPXr12fWrFns3LmTHTt25HvNmTNnMn78eBo1asR3331X7LbPnTuX8ePH8+233/Luu+8SGxtLSEgIw4cPZ9y4cXTu3Jn+/fvz22+/cfXqVbp27Ur16tU5duxYnnO1bt2aJUuWMGvWLI4cOULNmjUByMzMJC4ujlmzZvHXX38xe/ZsFi5cSEpKCq1ateLOO+/kqaeeKvZ7EEKI8k5n3tx5Zo01CUhcpG/fvuzfv99m36ZNmzh16hQffPAB9evXJyMjg5MnTzJu3DiWLVtmOW7SpEm8/PLLzJo1i+DgYK5cucLvv/9uk2TqiPYFP3PmzFtq++XLl+nZsyezZs3it99+w8vLi3PnzhEREUFOTg6JiYncdtttPPPMM/j7+3Pu3Dmee+45IiIi8pyrc+fO+Pr6Mm3aNKZNm2bZ/8cff9C3b18OHTpEnz59eOutt9i6dSs6nY7Tp0/fUkAlhBDlnS/qEE1ZCkgAFHfd/Pz8FEVRFD8/vzzPBQcHK0uWLFGCg4Nd3k7Z3GuTz4ZssslWkTdPUOqB4gdKCCj1Qamdz9YUlDpOaENB39/5bZLUKoQQQpQj3qg9JB6oSa1lpYdEhmwqsPr163P06NF8n2/VqhUXLlwoxRYJIYQoDh1qNwOowYgnakBivd/dSUBSgUVHR9O+ffsCnxdCCOH+agKxqKv5+qAGInrzJgGJcHsmk0mm0wohRBmnR+0N8UD9UvcE0lGn/JYlEpAIIYQQZZgWkGgBiA61KqsEJEIIIYQoNXrU4EP7QldQE1nLWkAis2yEEEKIMkzLF/E0b9lID4kQQgghSpmWvOpp/mkyb2UtIJEeEiGEEKIM00rEG8kNSEANSMrKDBuQgMQlNm/e7HCROSGEEKKo9OQO2xhRh2tADUbyX/7U/UhA4gL333+/Zd2Whg0bsnz5ci5dukRaWhoXLlzgxx9/pHnz5gAEBwejKEqehfggb2ATFRWFoih07drV5riPPvqIzZs3O/EdCSGEcBWt+JmR3NV9AeKAZFc1qhgkIHGBuLg4kpOTMRqN/P777wQEBHD//ffTvHlzHnzwQQ4dOkRgYGCxzp2WlsasWbNKtsFCCCHcllb8TOfqhtwiSWp1gc2bN7N//36++uorQkJC6N+/P+fPnwfg/Pnz7Nixo9jnnj9/PhMmTGDgwIGsW7eupJoshBDCzejtfpalMvGOSA+JC127dg2TycTQoUPR60vmP0VUVBRffPEFM2fORKcr6/GyEEKI/FQHAsgNRFKBFJe26NZIQGI2vnMoJyePZ3znvLkazhIdHc3TTz/NG2+8QVxcHBs3bmTq1Kk0atTols775ptv0qhRI0aMGFFCLRVCCOFKlbAdkvFGDUY8UYc6coAMIKv0m1ZiJCAxe7F3GMGBAbzYO6xUr/vZZ59Rq1YtRowYwc6dOxk2bBhHjhzhjjvuKPY5r1+/zvvvv88bb7yBh4dHCbZWCCGEKwSiBiUaP/NjT8re9N78SEBi9u7W3ZyLT+DdrbtL/drJycmsXbuWqVOnEhoaytatW5k6dSoAiYmJAAQEBOR5XWBgIAkJCQ7P+eGHH1KpUiWefPJJ5zVcCCGE0+lQgw6d1ePKQCa5C+rlOH5pmSIBidmCPQdo+tECFuw54OqmcPz4cXx9fQF1Rs61a9fo1KmTzTF+fn6EhIRw4sQJh+dISUlhxowZvPrqq/j5+Tm9zUIIIZxDW6tGZ/VYjxqQaL9LD4m4JaGhofz4448MGTKEli1b0qRJEx599FEeffRRfvrpJ8txH374Ia+88goPPfQQjRs3pkuXLixfvpxr166xevXqfM8/f/58EhISeOihh0rj7QghhCiGmw2sa0GHfUCShRqolJeARKb9utDFixc5e/Ys06dPp2HDhiiKYnlsXfDs3XffJTk5mZdeeokmTZoQGxvL9u3b6du3L+np6fmePzs7m2nTpvHtt9+WxtsRQghRRB5AEBCDOuyiR01WjTM/Zx2E2Ack2Vb7y8OQjQQkLtC3b1/L788888xNj8/JyWHOnDnMmTOnwOMczc5ZsWIFK1asKHIbhRBCOJ/eastBTVKtBCQCvqg9ICnk9oRA3l6R8hKQyJCNEEII4SLaGjTWxc0MVvs9yZvUat1boj2WgEQIIYRwke7DBvPq+tV0HzbY1U0pNkfBhtYDoi2W5yiHROsdMZmPKQ85JBKQCCGEKJP6jRtJ1Tq16TdupKubUmxaT4jOwWMjuQGJth9sv7i1gER6SIQQQggX2bRwKbHRl9m0cKmrm1Js+Q3ZWPeKGMkbkGg9IpmU7eqs1iSpVQghRJm0c+Uadq5c4+pmFJqOvAmo9j0k1jVHtN4R62EbUGffaOfIBhyXxyx7pIdECCGEKAWVgWp2+6x7ROwfa5t1kivYBiTFETpsMOPXrybUzXJvJCARQgghSoEe2/VowHEPiaOgBPM+uPWckbBxIwmoU5swN8u9kYBECCGEKAXWwy8a+xwS64AEq+cUbHtKbiUg2b1wKQnRl9ntZrk3kkMihBBClALrmTOZVvuse0i0gMVgtV+HOpvGeorwrQQkB1au4YA596byLZynpEkPiZvYvHmzTbn4whg1ahRxcXFOapEQQoiS4GP+aUQNKAxWz+msNrDtIdH2az0k1omu5WGarz0JSNxQVFQUkyZNcnUzhBBC3CJPIBA1EVULRqyHJrQvYaPdT+uhGq2HBKtzSEAihBBCiELzALzIHYbRtgDUYMW+2Jl1EKINzRjIHbLRapKUh8qs9iQgcQEfHx++/vprkpKSiI6O5tlnn7U8t3nzZho2bMjHH3+MoigoSu7HbtSoUZw7d46UlBRWr15NtWr2E8iEEEK4Ew/UwEPrGdEB3qi9JlqgArbFz0zk9pCYsA1IrId7yhunBiQvv/wyu3fvJjExkZiYGNasWUOzZs2cecky4b333qNPnz4MGjSIu+66i9tvv52OHTsCcP/993PhwgWmTZtGrVq1qFWrFgBhYWEsWrSIOXPm0L59ezZv3szUqVNd+TaEEEKYeaMGH/a8yM0d0aFWVfVEXclX6zVRsE1szSG3F8S6NLw2S6e8cmpA0qdPH+bOnUu3bt2488478fDw4LfffsPHx+fmLy5lncOHMXlBBJ3Dhzn1Or6+vowdO5bnn3+eTZs2cfjwYUaNGoXRqH7M4uLiMJlMJCUlERMTQ0xMDACTJk0iIiKC9957j5MnTzJ79mzWr1/v1LYKIYQoHD9yk1etVcJ2qCUbNUjxwLYXRO/gsRaUaDkjWg9JeRyuAScHJAMHDuTrr7/m6NGjHDx4kNGjRxMcHEynTp2cedli6T1kLIE16tB7yFinXqdJkyZ4eXmxa9cuy764uDgiIyMLfF3Lli1tXgOwc+dOp7RRCCFE0WgBhjXrxfG0n5moQUo2tjNorCuyasMzivk4rI7Xhm/Ko1LNIQkICAAgNjbW4fOenp74+fnZbKVl6w+LiL8azdYfFpXaNYUQQpQPWkBhzcO8ZWM7ZHPN/FMLUqxn0GhTenVWx18nN9G1vE75hVIMSHQ6HR9//DHbtm3jyJEjDo+ZMmUKiYmJlu3SpUul1Tz2RKzko/Hh7IlY6dTrnD59mszMTLp27WrZFxgYaJNbk5mZicFg+9E+duyYzWsAunXr5tS2CiGEKBytN8R+nzZMYx1sYP7pQe5Mmvx6SOyV1ym/UIoBydy5c2nTpg3Dhw/P95iZM2fi7+9v2erWrVtazSs1KSkpLFq0iPfee4++ffvSunVrvvrqK3Jycj9iZ8+e5bbbbqNOnTqWmTSffvop4eHhPPfcc4SEhDBx4kTCw8Nd9TaEEEJY0fJErBlRgwot4NCGZzD/1AIW6xk0WtCSX66I9JDcotmzZ3P33XfTt2/fAns9MjMzSUpKstnKoxdeeIGtW7fyyy+/sGHDBrZt28bevXstz//3v/+lYcOGnD59muvXrwOwa9cuxo8fz6RJkzhw4AB33XUXb775pqveghBCCDP79Wc0HqjBhhaQWNOGYKx/14IQLSCxDzyspwWXR06fQTR79mwGDx7M7bffztmzZ519uTIhJSWFRx55hEceecSy7/3337f8vmvXLtq3b5/ndYsXL2bx4sU2+z788EOntVMIIcTNWZd4t+aNGjwoDp7TghRtSEerVaKzOj6bvMpzUqtTA5K5c+fy0EMPMWjQIJKSkqhZsyYACQkJpKenO/PSQgghRKmwXgTPOinVSG5AogUbmhyrfTnYzrixnuprzVFgU5449b09+eSTBAYG8ueff3LlyhXL9uCDDzrzskIIIYTT1MT2r3ktmNA2yC2GZj1kY50TYt1Dog3ZVCF3FeBsHPeQlNcaJODkHhKdTnfzg4QQQogyQpsNYyQ3YLDuHdECEm3Nmmyr11jTeju0HpEEq98BKuI67uW590cIIYQoUVrgYf3nth7HPSTa745mzVjXGtESWW/W+5Fs3gqjVfgwRiyIoJWTq4+XJAlIhBBCiCKwD0h0Vvv1QGXUJFXF7hhHs2aKsnJvfsM4jnQYMha/GnXoYFd9vP/dw/lk2Sb6351/CQ5XkYBECCGEKCTroRlvoAa2uSDeqDkmfuTtEXEUTDgrsWHfD4tIuhrNPrvq4/cOf4zqNety7/DHnHTl4ivPCwcKIYQQJcp6yMaI2hOSQW7w4UnuCr/Wc0njS6+JAByNWMlRB5XHf14xn3uHP8bPK+aXcotuTgISIYQQopB0Vhuoxc/sZ9wAJBbyXKU9a2bj2hVsXLuilK9aOBKQCCGEEEVgHZRoM240BtQF8QrDulqrkBwStxQVFcWkSZNc3QwhhBB2rHNItGEb69k0XhQ+8VSbaSNUEpCIm/rpp584d+4caWlpREdHs2TJEmrXrp3nuOeee47IyEjS09O5ePEir7zyiuW5Pn36oChKnk2r3iuEEGWBdQ6J9Qq92tRdLwrfQ5IOpDmhjWWVDNmIm9q8eTNvv/02ly9fpm7durz//vusWrWKnj17Wo755JNPuOuuu3j++ec5dOgQVatWpWrVqnnO1axZMxITc0dXr169WirvQQghSoL9ujXW5d9NqAFJYdeaybz5IRWK9JC4wObNm5k9ezazZ88mPj6ea9eu8cYbb+R7/OTJkzl48CDJycmcP3+euXPn4uvra3NMjx492Lx5MykpKcTGxhIREUFgYCCgVsx9+eWXOXPmDKmpqezfv58hQ4ZYXhsYGMiyZcu4evUqqampnDhxgtGjR1ue//jjj9m1axfnz59n586dvPPOO3Tr1g2jUY1nW7RowRNPPMGgQYP45ZdfOHv2LP/88w8bNmzI816uXr1KTEyMZVOU8lwIWQhRXmm9I9pPBbW3I9aVjSrjJCBxkVGjRpGdnU1YWBiTJk3i2WefZdy4cQ6PzcnJ4emnn6Z169aMGjWKfv368e6771qeDw0NZePGjRw9epTu3bvTq1cvfvnlFwwGtVjxlClTeOSRR5gwYQKtW7fmo48+YtmyZdx2220AzJgxg1atWjFw4EBatmzJE088wfXr1x22pUqVKowYMYIdO3aQna2OlN5zzz2cOXOGu+++mzNnzhAVFcWCBQuoUqVKntfv37+f6OhofvvtN3r06HFL91AIIUqbdQ+JVg7eSG7RM/kT69Yo7rr5+fkpiqIofn5+eZ4LDg5WlixZogQHB5fItR4f2Fw5vWiY8vjA5k5/X5s3b1aOHDlis2/mzJmWfVFRUcqkSZPyff2QIUOUa9euWR4vX75c2bp1q8NjPT09leTkZKVbt242+xcsWKAsX75cAZSffvpJWbRoUYFtfuedd5Tk5GRFURRlx44dStWqVS3Pff7550paWpqyc+dOpVevXkqfPn2Uf/75R9m4caPlmGbNmimPPfaY0rFjR6V79+7KokWLlMzMTKVDhw4lfn9L+rMhm2yyyaZtvqC0BqWeeWsLSlNQGoNS+xa38GGDlf+uX62EDxt8y+cq7NYUlDpOuE8FfX/nt0kPidmLQ9vRsKYfLw5tVyrX++uvv2we79y5k6ZNm6LX5/1P0r9/fzZs2MDFixdJTExk6dKlBAUFUalSJQDat2/Pxo0bHV4nJCQEX19ffv/9d5KSkizbI488QpMmTQD4/PPPGT58OPv27WPWrFl07949z3nee+89OnTowJ133onJZGLJkiWW5/R6Pd7e3jzyyCNs27aNP//8k7Fjx9KvXz+aNWsGwIkTJ5g/fz7//PMPO3fuZOzYsezYsYPJkycX7wYKIYQL6VF7RkyotUiUEjhn2LiRBNSpTdi4kSVwtrJHAhKzd1cd5GxMEu+uOujqptgIDg5m7dq1HDx4kCFDhtCpUycmTpwIgKenJwBpafnnaVeuXBmAf//737Rv396ytWrViqFDhwIQERFBcHAwH330EXXq1GHjxo289957Nue5ceMGJ0+eZMOGDQwfPpx///vfdOvWDYDLly+TlZXFyZMnLccfO3YMgAYNGuTbtt27dxMSElLUWyKEEC5jvW6NDnVGjf3CecW1e+FSEqIvs3vh0hI4W9kjAYnZvHWRNBm7knnrIkvlel27drV53K1bN06ePElOju3yS506dUKv1/Pcc8+xa9cuTp48SZ06dWyOOXjwIP3793d4naNHj5Kenk6DBg04ffq0zXbx4kXLcdevX2fJkiWMHDmSZ555hscey3+dA60Xx8vLC4Dt27fj4eFB48aNLcdoPSPnzp3L9zzt27fn8uXL+T4vhBDuRme3aQveFXaqb0EOrFzDggH3c2DlmhI4m2MjOoeyffJ4RnQOddo1ikum/bpIgwYN+OCDD5g3bx4dO3bkqaee4rnnnstz3KlTp/D09OSpp57il19+oWfPnkyYMMHmmJkzZ3Lo0CHmzp3LF198QWZmJn379mXlypXcuHGD999/n48++gi9Xs+2bdsICAigZ8+eJCYmsmTJEl5//XX27t3LkSNH8PLy4u6777b0cISFhdGlSxe2bdtGXFwcTZo0YcaMGZw6dYqdO3cCsGHDBvbu3cuXX37JM888g16vZ+7cufz222+WXpNJkyYRFRXFkSNH8Pb2Zty4cfTr14+77rrLyXdaCCFKjqOAJN6VDSqiJ3uHUS8wgCd7h7F8zwFXN8eG9JC4yJIlS6hUqRK7d+9m7ty5fPLJJ8yfn3exo4MHDzJ58mReeuklDh8+zIgRI5gyZYrNMSdPnuSuu+4iNDSU3bt3s3PnTgYNGmSZBTNt2jRmzJjBlClTOHbsGBEREfz73/8mKioKgMzMTGbOnMnBgwfZsmULJpOJ4cPVpalTU1O5//772bhxI5GRkSxatIiDBw/Sp08fMjPVWfSKonDPPfdw/fp1tmzZwv/+9z+OHTtmOQeow0sffPABhw4d4s8//yQ0NJQ77riDTZs2OeX+CiGEM2kBSUkM1ZSmz7bu5mJ8Ap9t3e3qpuTh1vfTz8+PxMRE/P39SUpKsnkuODiYGTNmMG3atAKHBdzR5s2b2b9/vyR0OklZ/mwIIdxbANAAteaIHkghd8pvWVQZ9T1El/B5C/r+zo/0kAghhBCFpP0V7y49JKHDBjN+/WpChw12cUtunQQkQgghRCFZByR6XB+QlKepwpLU6gJ9+/Z1dROEEELcAnf5a373wqWEjRtZLqYKS0AihBBC3ISR3J6RHHKHbFztwMo1Tp0mXJrcJcgrMm1RNm2BNyE02mdCFu4TQhRGAGrA4QnUMu+rAgRaHVMN8MN2mEb+D1OyymxAcuPGDUBdaVYIa9pnIr8FAoUQwpovajBiMP/0ACqRO4RgMB+jfWFqgYg79JA4UlYTXcts90JKSgp//PEHDzzwAADHjx+31N0QFZPRaKRFixY88MAD/PHHH6Smprq6SUIIN6clp2o9Hx6AF+BNbuDhbd6sj3PWDJv+dw/n3uGP8fOK+Wxcu6JY57BOdC1LwzllNiABWLx4MQAPPvigi1si3Mkff/xh+WwIIURBdKg9IFpvhydq74gBNTgBNUDxIPcLU1ue1hnuHf4Y1WvW5d7hjxU7INm9cCkT7rmTzpGRpHYOdbuKrPkp0wGJoih8+eWXrFixgqCgIHQ6d+1AE6VBURSuX78uPSNCiELTeki0JFUtEPFALQuvA/xRAxDr/BEd6kq/Je3nFfMtPSTFdWDlGtrVC6KGm5aIz0+ZDkg0qampnD9/3tXNEEIIUcbosR2K0ZPbG6Iz/25AXTzPiBqkKEACzukl2bh2RbF7Rqx9tnU3T/YOc8sS8fkpFwGJEEIIURz2QzagDtuYUIMTLdk1C9vAxd3LxS/fc6DM9IxoyuwsGyGEEOJWWQ/V6FEDEW8gg9zhG4N5v3acTPd1DglIhBBCVFhaQKL1fmQDSUAqucM3kFsMTTiPBCRCCCEqLPseEgW1d0TrBTGYf1ovqFeWtQofxogFEbQKHwbAgwObs3PRMB4f2NzFLZOARAghRAWm9YwYUJMqHeWGaDkj2mycsjBk0//u4XyybBP97x5us7/DkLH41ahDhyFjAXhsaDvq1fTjxaHtXNFMGxKQCCGEqLC0Hg/rhFXr57T8EesekrIQkFjXM7G274dFJF2NZt8PiwCYv+ogF2OSeHfVQVc004Zb31s/Pz8SExPx9/cnKSnJ1c0RQghRzvgBwUAKai+IAUgzPxcApKMGKmnmx5nm4zKc1J4RnUMt03VvZZZMYSu+VkZ979HFvpJjxfn+loBECCFEheVPbkCifRmmm3/6mfdloQYggeafJtTAxBm2Tx5PvcAALsYn0POjBU66Si53CkhkyEYIIUSFpQ3VQN6/0HNQZ9mYrJ539l/xn23dzcX4hDJV0KykSGE0IYQQFZYWYGj5I/YBiZHcgETLI3GmsljQrKRIQCKEEKLC0tak0QIN61k26dgGKdoxbpvnUMZJQCKEEKLC0npGHAUbCrn5JFgdJwGJc0hAIoQQosLSpvUWdo0aCUicR5JahRBCVFgGcsvC3yzYKI0ckopMAhIhhBAVlp7cXpGbBRvWya+i5ElAIoQQosIqSoChlY+XgMQ5JCARQghRYVnXILkZrYekLHjYvGjew26waF5hlZV7K4QQQpQoHbb1Rwo7ZFMWTBzajvo1/ZjoBovmFVZZubdCCCFEiTEAnubftRySmw3FpAPJTmtRyZq76iAXYpKYe5NF8x7oHMq2yeMZ3zm0lFqWPwlIhBBClEk6oCrFm/niDwSRO7PmZtN9MR+TWoxrOUPosMGMX7+a0GGDHT6/bF0k3ceuZNm6yALPM653GPUCA3ixd5gzmlkkEpAIIYQokwyAF0UrqOVl/ulj3rSAJBFIKNHWOVfYuJEE1KlN2LiRxT5HXeDQD79y7egJ3nWDtXMkIBFCCFEmaTkghe0h8UTtFfFGDUy8sZ01U5Zmz+xeuJSE6Muc+/MYIxZE0Cp8WJHPEQbcc/4SF777mQVusH6OBCRCCCHKJC0gKewXmR41CPFDXcU3AzVIKcxwjbs5sHINCwbcT/VGnfCrUYcOQ8YW+Rza/Jugfvdwz93DS7aBxSABiRBCiDKpqD0ketQAxMP8exJqTkhZ6hmxt++HRSRdjWbfD4uK/FotIMlpGcpDwx8r2YYVg6xlI4QQokwqakCiHe9ttS+tpBtVyo5GrORoxMp8nw8dNpiwcSPZvXApB1ausXmumfnnJf9Avlkx34mtLBzpIRFCCFEm6ay2wtC+8HyATKe0yP3kl/xaCQg2/z5p5vP8snZFqbfNngQkQgghyqTiDNnkoM6mcZfpu86mJb/uXrjUZn8I6v2IA667omEOyJCNEEKIMkmHOvW3KEM2ANnOaY5beHhgcyYObcfcVQdZti6SAyvX5BmqgdzhmpOl27wCSQ+JEEKIMquoPSTlXWFLxktAIoQQQpSQ4kz7LYtTfIuisCXjtRk2p5zfpEKTIRshhBBlUlGTWo2U7Sm+hbFsXeRNy8WD9JAIIYQQJaY4s2xc2UPSKnxYsauqlqRKQAPz7+7UQyIBiRBCiDJJC0YK+0VmxLUBSYchY4tdVbUkaTNsbgCxLm2JLQlIhBBCuKXxnUM5OXk84zuHOnw+v4DE6GAf+ewrTbdSVdXerfS2aPkjNx/YKV2u/u8jhBBCOPRi7zCCAwN4sXeYw+e1oRq93b7aqCvZVrLb7+ohm6MRK1k+PrzAyqqFdSu9LVpAYrwnnFbDBt9yW0qKBCRCCCHcUkR8DilGbyLi8w8jtFokmsqoi+f5ouZJ1MI2z6Q8JLU+PLA5z/quoFXylmL1tmgJrUnt2tDBroKrK0lAIoQQwi0F9R/B/rCRBPUf4fB5LcgwWD0OBLLIrcYaAHiROz24PEz7nTi0HTX99XRJ/9Nhb8vDA5uzc9EwHh7Y3MGrc3tIzgUGsM+ugqsrOTUg6d27Nz///DOXLl1CURQGDRrkzMsJIYQoR75ZMZ8rMZfyXfhNb/fTC3WdGq0sfDZqsKIFJDrKRw/JzWqNFFQcTes5AvjwhWkcdVDF1VWcGpD4+vpy4MABJk6c6MzLCCGEKId+WbuC/zzcz+HCb92HDebJ9asJGTbY0lPihRqAWPeCZKPmkrhDDklJWbYuku5jV+Zbb6SggKW1+ecl3GuGDTi5MFpERAQRERHOvIQQQogKqJ95Fdu240byz8o16FB7R+zXqck0709GDUjKQw/JzRRUHK2t+eeh0mtOoblVDomnpyd+fn42mxBCCGFvk3kV2wPmHAgP1J6QTLvjMgFPbGfcVGTuHJC4Ven4KVOm8Nprr7m6GUIIIdzczpVruLByDX6owzHeqEM2CXbHaT0inhS+omt55s4BiVv1kMycORN/f3/LVrduXVc3SQghhJvSoeaE6FGDkfySVlNQpwNXhOGaglQid8qvBCQ3kZmZSVJSks0mhBBCOKIFJDrUIRtTPsdlowYj5SkgudnUXkdaoyb9XgVinNWwW+BWAYkQQghRWNqsGS0gsU9otZZE3uEcZwkdNpjx61cTWgJVUPMrEV/Q1F5HHh7YnBXt6wCw65Zb5RxOn/YbGhpKaKi6DkGjRo0IDQ2lfv36zrysEEKICkAbotGqtRYUkJSmMPMMoLASqIKaX4n4m9UisfdC/UB89kejGHS8f8utcg6nBiSdO3dm//797N+/H4CPPvqI/fv388YbbzjzskIIIcqxSuTmjBS2h6Q07TbPANpdAlVQ004cwJhjIu3EAZv9N6tFYm0GUH2+2i9yqFNLPG5hYT5ncuosmz///BOdTvKahRBClAwPoCaQaH6s5YW4U9GzAyvXcOAWKqCO6BzKk73D+GzrbsJCu1C9sg/XQruwuhjneg7Q+lYu3jeSE/c9TIcsxdLr4gW0GjKWP35YxJ4SWPTvVrjVtF8hhBCiIIGAP2p5eO3P3fLyZ2+r8GF0GDKWO+IjqZd6gSd7h7Fp43LuCqnNqlOXi3QuT+BVYLz58cJ2XQgYOoafV8zncraJDkPGsu+HRXQcMpbKNerQe8hYlwckktQqhBCiTPBEDUhMqD0l1jkk7jJccyu0fJE99TqRavRmU3wO/QL1+GSn0y9Q/bouTMJsCLCW3GDkfSDSM5POe7+l1pVjHI1YyfLx4RyNWMmRHxaRdDWarcVYNbikSUAihBCiTPBCDUTSyV3hV0GdPZNcQtewntXS/+7hfLJsE/3vHl5CZy/YPnNwUF2vZ1/YSPz7j+Czrbu5GJ/AZ1t3AzdPmG0H/Ay0AW4AI4EPgSd7h1EvMIAne4fZHH8yYiXfjA93ee8ISEAihBCijPBB7R3JQQ1ItB4SEyVXY8R6Vsu9wx+jes263Dv8sRI6e8G0noutSz/lWswlfl4xn+V7DtDzowUs36MmteaXMNsNWA/8hNqLtAfoD2w0P28f2LgjySERQgjh9vSAL5CBGpBYD9mUpH0/LLLkV1wxGrh3+GP8vGJ+CV+lYLVN+wjL+Jy/TXmn9DpKmNUDs4Cm5sfbgDGoFWo1y/ccsAQ17soZ/z1LjJ+fH4mJifj7+0vVViGEqMB8gQZAPLnBiQ614JnbfokV085Fw6hf048LMUl0H2s7lPLwwOZMHNqOuasOWqb83gd8BsSZfz9ZhGtVRg1cokug3daK8/0tQzZCCCHcXiC503rth2zKk1bhw/jbuw8xiTkOi55ZV2j1BF4D3jM/N4+iBSPuRgISIYQQbq0y6lRf+8TVokz3LU459xGdQ9k+eTwjOocW6vjirC9jr8OQsRytfBvzvJ/mHyVvWXitQutnKw/wKfAYam/RP8CXTmxXaZCARAghhNvSAVVRp/XaFz4rSkBSnHLu+c1MyU9R15dxZN8Pi8gxZaM3GPOUiwe1QuudY1fSO+IE9wKZqIXP7ib/mUYl0a7SIAGJEEIIt+WH2gOQcrMDb6I45dyLOjOlqOvLONJRd5DH0z+lVfIW9jmoDaIHvkUNQDKBp4FqDnpARj4ezl9RCxn5eHiJtKs0uPUQnCS1CiFExaVDTWT1IG9A4mfeH1vajXIC60RVrTfDUUIrwH+AD4A0g564Yffw0amz/N8TzfK85q+ohdRvWJMLZ2Po1mhcvteWpFYhhBDCzA/1i1HjjVqV1Rt1Ib1UB6/JQq0/Uh5YD6lY92ZUBVqiDsl8AMwH/mt+TVbfXlRv2ZRn7+rrsAdk7juruHA2hrnvrCr191Nc0kMihBDCpWqhFsW6aH5c2/wzA3UhvXgXtMmZ7Kfu2j+uixp43JPP648AaZ8soHnaJY77BTNo4phit0V6SIQQQggzb9QqrF6ofyVXQp1V448alJSmos6sKQ77JNNl6yLpPnYly9ZF0hT4H7nBSKpfZS507sCmkcOZBowC7gU++X0jEfVu49N164vVBq1EftPwYbf+hkqIBCRCCCFcxohaU8QLNSjxNO8zAm3Ch3G/eV2Z0lLUmTXFYT3E4gmEAs8Cv5u3Gqi9IP2AIeF3svCtabyXnsEi8/NpwMa1K5j0cD82rl1RrDZoJfJbO5jJ4yoSkAghhHAZI2pyahpqj4i3eV8C0NJqXZnS4sw1X7SZL7oGjeg+diU710WyHVgHPA+0Rg3IzgV68wBwHLVU/IIB9+cpF19cWk2SOseXkHQ1miNusMqvRtayEUII4TJG1L+MU1CrsWai1hsxAf9YrStTWpyx5ouWI1K5fShVfbx4vmMTagLDgTqoJd/3AB3GhlG1RzAoCnHjnJOMqg0X3U8S3ceG2yQTu5r0kAghhHAZI7kzK7JQh22yzY+11W+PRuSd/lpW+ALPdAum/oV4fJ9bDPVGU/3xuTyLGoxEAr1Rc0PejU7gAjD3h0OW1xenyqqWH+JoqMuda5LILBshhBAuUxMIQF0kTwdUARLJDUpKWuiwwYSNG8nuhUtLbBjEXl0gDAgGnkCd1mztGHAQOAT8gDo8lZ+CFtrLz4gFEfjVqEPS1WiWjw8v8FiZZSOEEEKgJrNqwYeCWujMWcEIFK+EfH60GTmvNWvMj8AmYDuwC5gLvIgajMQBZ4BlqDNk+gOTUdeesQ5GtByTOcues6myGpOYw9/efRz2eDjqQdn3wyKSrkaX6lBXSZAeEiGEEC5hABqiBiCZpXTNkuwh2T55PPVyFAyfL0GfafsO9gI3gAjgOwr3RatVV83ONmE0Goj95wDJ+w/wt3cfjla+zWGPx6HvH6GKj5G41GzaPrCkyO/BnXpIJKlVCCFEqfJA/WtYh/ollFaK1z6wck2hA5FW4cMsSbXWeSy9gWGA9xdLMWSb0Gdlcal9W2KfGcXx2CrUaePBsLseyXOeOseXcH+rLEsBNHtz31nFxJeHsmf7MTr3bInv6RPUr+mHZ+Iedl0Ncdzj0aQZxF+G2rXzPpcP60JsPzpoh6tID4kQQohSVQ01XyANqI97VWLtf/dw7h3+GD+vmE+tQaPz5GJ8Btxn95rrlbxZMu9jMoPCyMwy4mHMYva/O1me13I6JirzqWJMteSD2FdotZff8zPGjGVoSG1WnbrMCc9LTHx5KHPfWcXSeRGFeo/WeSl3jl3pNj0kkkMihBCiVHmhzj7xRP2r2J3cO/wxqtesy73DH8uTi/FWp7rcB+QY9WTc046/XhvH7UBYWjrvPzKB49vWoZBF5HbbwEA7zy/nfIinEj+dUb967Su22rOu4Aq5+SL/6eqPT3Y6Q0Nqs3ReBN0ajbtpMGI988ZdZ9pID4kQQohS1RB12CYVNTApzf+75zcMo7HuIbGugqoHztT1x/NSIqb726B7NIyYZBOdhn9d6GuPX7+agDq1SYi+zIIB91t6QH46o+fGbQ9a8lryy3PRejaSU0A5VYdVpy4zbfGiQr2v/GbeuFMOifSQCCGEKDUG8+aJGpTc6oya0GGDGb9+NaHDBhfq+A4FVH99eGBz3hxs4u/lT+QpyT4E1GDEx4P/1fHnQkwSH339V5HaunvhUhKiL7N74VJA7QGZu+ogD3cPopdfnGXmT34zgbQZN79ld2TI1khLMGL/vhytx2Pf26P1tjxYhPomziYBiRBCiFKjrV2j/W66xfMVdRpvQVNi8xtC8UKdwgvwdmoWE+bssNQE2btiFF/8+pHDgMg+WNLKwAOW/ROHtiOQNDrF/mMJVOwDFy14APgwZThHK99Gr1H/Z5kabP++HK3HY19kTnuvj+UzXOQKMmQjhBCi1PigFgzLNP9+g1v7EirJabz2SaQhqEXOHgb+jTqs0QtINx9/cNVjVPXOJgE/PjU9wtbP37EZLrEfonG0P+3jKZZrKte8ebJ3GJ9t3W1Tvn7/t6MJ8tNzPSmHh75OocOQsdzZIYM7O+m5cDaGbo3G2byPEZ1DeWbA7exp3py4uLPc2yiNw1ntWbD4d0vPj/ZeF646yJfrIt1iyEam/QohhCg1RtS/hNNQ16y51b+IC5rG+8JzD/FI76os2RrL/45kFZhjAeoQiq5BI56eMYFB5z+i55HzludygNdRgxEtX2OXxzW6K3vYoXRBhwcdhoy1OffuhUsJGzeS6H2HGL9+tSVo0vbvXriUA+siLUmr2yePt/RsWAckKZnNqJoZRUpmCEcjXuFoxEr0j4fTopo6u8belVotOdRnHAkBWdx7NZJAXQrtPA7w9MABXK+UYbl297Er3WotG+khEUIIUWqqopaLjy+Fax368f+oYkwlLtuHd67dfdNy6mOB/7atj/HIRXQ56lfjMeAC8AFqqXfITRA1pafh4elFtimbzPRUdi+b7TDYya+nxJ71dF7r/JC9rzxDkJeRLQFNiPBtVGBQBfDJsk1Ur1mXdCWTK8eWcl+jdGLjqlMp8gZf9LnNpi2S1CqEEKJC8uTW80YKa8nWWOKyfViyNfam5dTDgRmAx6EL6HIUshtV5T+oZd5HkxuMAHhe/QtfjwyyT60j4foV9B6eZKWl5hsk2OeE5Me//wj2hY3Ev/8Im/0f/raZi/EJbK3SLN+E3E8n9OfsykeZ+9aTRHr7EJOTzql9P/LKpHd5640/0e0/z4e/bS50W1xBekiEEEKUmgaUfnXWwlgDdAX+rutP7ad78ekfpx0WK4PcEu8Xzsbw6BPrbzoUVFj5TTnWFDS19+zKRzFWgsQcXz5VxhNQKZvBTf/Jk19iz516SCQgEUIIUSpKe+2a0GGD6fX04wBs+3Sew1yTMKAvMMncpq5AzE3OO/LxcJvqqCM6hzpMRoX8q62WtE8n9Ofe/sH874Q3Bxo/SI+m8Wxc8PlNC6a5U0AiSa1CCCFKhZGS6x25WSEwUKcEVwoMsPxuHZAEoi5619bq+B+xDUbyCyZy9l7G9OkucvZeBrCZZqurnm55DcCbE7pjNOiZOLRdiQUkjt77019s5OkvtCM+Y36JXKl0SQ6JEEKIUuGB2ktSEt3yBRU40+xeuJS0+ATS4hMIOrfbpm7HG6jBSDJqIPIW8Jrd6/OrS/LsXX2pFxjAs3f1BWBv2hmUVrHsTTtj85qJQ9thNOjJNuXYlGm3LuPu6LG1kY+HW9qt1TUJG/mUzXu3fv0rs2bw/eYDvDJrRr73xfqcwx8PZ2fUQh5/3HGib2mSIRshhBB5dB82mH7jRrJp4VJ23mJ9D00VoBYlM8OmqCXgtbyPi8cvsKrlk0xGTa4dBPyTzzXy6yH5ae5iWiSd47hfMIMmjrGpE/L+kp02PSSOXq/N0tFlplGZHJLRo3hWcjgDyDpfZd2JJjQ01uRoejzR0YmW966dz5hjYkjKaepcPMzl+q25fUTeAMf+nHqgbsOanD0bQ+Ob5JsUhcyyEUIIUSL6jRtJ1Tq16VfICqiF4UXJzbD5T21/pl7dxn9q+1v2aRVNHx7Y3GaRPIC576zi0r7T6Hq/zGTz8XPJPxiBvIvbaT5dt56Ierfx6br1AJw3dCMdf84butm8Jr/Xe/y9Ht/0RLz0epI8ffHU6/OdATT3nVVcOBvD3HdWUSulEl45Brr4Vka/52M1GOkcyh3xkXhnZ9DQy4s6Fw7inZFM0Knd+b4v63PON/8+y0E9k9ImPSRCCCHycEYPSUPUv4KLm0NiXZV1RfvW+GSnk2r0puWrbwK5i89diEli6hoDD42bRMMqZ9n75iyift/PQ0AT1OqwrwI/3+Qahan8+vDA5jz/SHfOG7rxyZI/2Lh2hU3Pyj9Kuzw9OVpdkV1Vm/F3rbYYju7inenP3PRaM8aM5cHm9cjo1ISUkEy6NRpnKaZ2PSObba3uJXHjcvoF6h0m2DoiSa1CCCHc2s6Va0osEIHchNZbmV1jvW7NqhX/sxQR08xddZCJQ9vx2coDREacoGeoB36frqZlUorlmKRKnmT++CoBq3eCgxko1tcoTEAycWg7gvz0pMVsZONa23ViJg5tx5yc3FwXLSC5WDeUyknn8NJ78N6w7oV+//0C9XhlpGA4cJR3V+0H4LOtu9Uy8S2a8+VPszmwcg3THLy2JEvsO4sM2QghhHA6H9SiaFnFfH3osMF4+viQFp/A7oVLmbZ4ES1ffZNpixcRBnwB3LcukuyxK5kRcYLdgP9bS9ElpWAK9Gcdaq9I8qrXCcyO5emR/7IZ4gE1LyXrehCJ51NsCofZH2dt7qqDXIhJsklatd7nqCCb/ZDPp5P/w+n3X+bTyf8p8B58tnU3F+MT+O+KdZbpvMv3HGD+bX043aFDgQsMWgdajlYDdgfSQyKEEMKpPIFqQMYtnEObwpsQfdnmL/xFjZozMCrvdFrFqIdafuS0q8u0ywa+2qfWWp2WfQFPQzrVbhy39GS89PidlqEV78qBJEVH21zDusfDPh9kmdVaNI73ReZJuq1TN56uPc+wZ2s8AP+q3RhjSgb/qt2Ypwu4B8v3HHA4DBN0bjcjb+vJDm8vdg8bnKcHxBJoZaew+6ulfHlXX4K8jDx7V19+KsSwTmmRHhIhhBBOUxuoD1QCUm9yrDatNXTY4DzP7V64lJALfzEh7iceHtgcA/AUMDAqEkWn43Tfu3kcGAZ0AX549HayPhrCmga1+WrfIR4e2JyD3zyEkXSUbB3pZ2FjpBcJih879d0seR6Okksd9YIUR/+7h/Pz91/w+n1NqKdLZeLLQwE4VbUV6V6VOVW1VbHO2aHpf0i5HkDPBtDr6cfz3EMt0EqJzubAyjVcrBtKltEL/0qVeMCNekkkIBFCCOEUPkAAamXWuEIcbz2sYO/AyjWEXtpOTbJ5vlFV9vl4MMX8XHp4f/5oWJNfgO3AJaCTbxN0R6tyZ/3u9L97OBOHtqOqvzd6A5iyFDYci2RYryqc0HfloKm1Jel0+fjwPD0a+c2WsTdjzFhOzniZI9MmOxwOuXf4Y7TyOY6nKY3syGOWlXrf+/EXIurdxns//lKIu5T3nNnZ3kSdDmTLKXVAzP4eWgdaIx8Pp1LP6mTrDHjmZDGud1iRr+ksEpAIIYRwCj/zz8Imslov/KblbYy/I4Q+wBjA7+kf0Q9fTo3PdhKUmoXi74VpVBjGrqH0C7T9Ovts625Sjd5cbtCJe4c/xtxVB4lNTCcuMZ2pX26nS5eq+OpTaGXaybFFs255HRqAoSG18czJprLRwJO9w/IUPPt5xXz2ZLQlPseXub8lWPJANq5dwaSH+zlcv+Zmfl4xn2sxl5j3/ts83vo+tn06L8/iedaB1sSXh9I6VCGhZV1Sjd78GZ9zy++7pMi0XyGEECXKH3U6aWXUvJHizKzZOX8IDVYdgo0n0Zlsv6ZSAcbejvfgZmTofDCkJvLzxnM8/cVGm+MKWqyusGvMPDywOZNHdeNv//Z8vngLbZOP5vu6GWPG0rJzd3ZVb8W1DSvJ6jIAvxp1bAqeaUXMHBVBs2df/M1+DZ3CsH+N9vjIgdZU8q3KjZhLPP1wP7eY9is9JEIIIUpMNaAO4Is6oya/YMS6fLk9HRDzxgZ0v51AZ1LIqlmZtBY1yBkbxvm5g2kKvG705mLrMFJTYjFWgi5dquY5T0E9D8vWRTJvVz1qjl7E6Mcm2zxnPatm4tB21Kxs4DZdFGHjRuZbTh5g2uJFrPGsR5JXZbK6DMiTk6IVMcuJv+awCJo9+/L4E18eSv2GNS25J4W5n/avWTovgm6NxvHNwk+4FnOJdSvcZ9UbCUiEEEKUiCpADdQejGQgvYBjHX259r97OHMW/o8fatcn7GICOcDVJ7vzase6fHvvXaQM6sqGc74o5H6xzlq8iwsxSew5dtXh1NyCpuxW6z+EFG9/qvUfYts2q6Bj7qqDxGUYuUgYmdtP3DTBNe3EAYw5JtJOHLAMlXS4foLtk8fzQv9e9Eu9wEtn/keH6yccvt46sLAPaLQKq3u2H8sTfOQXrFhXZbWmBWvXK2Uwbv1qujtIJC5tEpAIIYS4ZV5AEGowUphaI9oX5fwZ3/EQ8ALwQUY0g58fRrfLF8jW64j9cCwfmAJYti6StsH94Wg19acVLeG0c8saNj0XWv7G84/2oX5NPyaPVJNbP1m2if53DwfgxsYf8E1P5MbGH2zbZhV0LFsXyZkr3Qk+cIr/NKjP/332Ap/tTmLZukiHvRJ9QrvQo7IPfUK7WPZpqwEDmBQFo15vWZjPnnVgYZ9kqwVhnXu2zBN85Bd4aK8BHPagdDAnEpfkEgHFJQGJEEKIW+aNuppvYfNFls6LoHujcQz48nfeByYDNX7fgi4+FVP1ytxY/jzVJt9n+dKtd+kA3hnJ1LvkuG7GqSu+mDL1nLriC+QOd+zy6EW84sd2pbPN+jatwoeR1WUANzb+wHjfZEZ0DmXGmLEce2sqzWv1splVo107JPaoTSCgBQ8vvjXS8mWvJZn+bB4KGdE5lCp+vuR46tmSGcuh6m1I96rMxbqhDqc55xdYWHN0jBZ45Jdbkl8Pyj5zIvEmqyRYV5GkViGEEEVmv9bNi4+H89TLQ5ldhITLr0OqceepG2QadHxjUgjr1JBmvYL5KVXPbr2XTTLmiM6hPNk7LN81Wg7PnMZJv/rsrNkG/dHd/LxzOx2GjOXKsX3UatmBfT8s4j+1/S3l5o/3GIZfjTr4pify9PEfuRifQNWgmnnWx5kxZizDG1cnMzOD05mJtK5enf9dOsfTH33LyMfDeWFwf/yOxmLs3ZBLQQa6NRpnk4y6tGNdgryMZHj5cja8C7OmLrMk2jZ+6gEC6tQmIfoyCwbcX+z/FgUl72ryS4h1p7VspIdECCFEkVmvBmwA/u/lodTNJ+FS0yp8GOM+Xsm05qFsaVKPO0/dACDxkU4cHdickJf7QbO2dKwcwt5zfszeUI9o3z6MWBDBvqBmjD8WD4/PskyjtZ5W+2e9HvxWL4xEz8qYWnW1DHds/HCKZdijX6Aen+x0+gXqLfkZjU99g9Iqlv1BHsxpHM5f1VvbrI8zNKQ2h6s05qt2D3AjqB3GlGw6URlQeyWS1p9An5RF5p9Rlh4LrXem+8NPE9e8t7lHpD0XzlazSbTdvXApHfdF8uze/cwYMzbP/SpsifenBw4g/OIWnh44wOHzjoIRbbhpuIOkYleRgEQIIUSR/WHu6t+xcCm+wKJ3VvHnr8ms+rulZdaKFjCMfmwy2yeP518DhvDU+y/yROQBQk5fRNHrMI3uwruXE5k4tB3GSpBTI5XPtu62fKk36TXAMtPEftZJr1H/h1+NOvR+/BX2e1ZF0enRKTkYju6ytNM6z0NbC+azrbstAUvHWtcwVoI+bQ2YfPz5zT+EaYsXWV6b2Ko+O2q2JcnTl3+Cw0jz8uVi5z6WYZa9JJPta+R/l85Zvuy1YKd9djz1Lx/iQp12nPSpwzcLP7G5h+2qptE5OgqfrFSGhtTOc4+fvasv9QID8s030RQ0nNX/7uHcM3wWBmMzm2BRG8J5rIAAsrRJQCKEEKLQPAEDcGTlGn4YcD8xK9dQF/huXgT/pDQg2cuPoDvULzktgAi6Yyj1AvwZ98ELVL8YhamqD8oD7TC9dw+/Dn2UZRG5s1emfrmd5XsOWL7UG5z5jonKfFombMbD24f0pHjLrJMeTePRoaDTG9Dpc0i6Gs2fX7zFO9OfsbTXOndC36k2hqe7ou+U++W/MdKLlBxfjuV0wksH9T0MNq+td18dWje6StLVaPwq+fJ5qyFEVWtvqYQa9nR/vJ/rSdjTarJt6LDB9Jz8Hw7/No8eJzbinZFMwyt72fbb9DzDKRNfHsredmHMaTWEr7OD8tzri3VDLfkmBfnwt81cjE/gw98253nu6YEDCN21hpQttjknlqTiAnJVSpsEJEIIIW6q+7DBvLp+NXcOG0wN1JLw6agl4eOBewY25zGvb+hh+pvg1Fggt6cgODWWrLOX8b94gSRgVrtaXLm/A7+27MtnX/8G5C3PrvVg9K9yhirGVAaFpODtH0h1Lw/+U9sfgKTfvuZ5r4X09NxPj0ZX0e/5mC8/H2Azk8Q6AdRRYqdfu0fZW+n/8L3mwYRD32OIWJTntRsXfM7y8eFcT0slQ4ELmdmWSqj2CabW5e8//G0z2b5GvPo2cDiUNfedVeyhBkmevlyo0yLP8/arAmvsk2GX7zlAz48WOMyt0XpPgs4f5GBsJct+LQl2RSHzfUqDBCRCCCFuSssZ6TJuJFVRC5+lWT0/cWg7fAxpdDXux89whKNrn+b/7mvA8vHh/LhyGSk71NV2lwJz/jhDp+FfM+Ffk/OsTGufN6H1nGy94oVfZgo9Yg5Zhjc61zTglZ1Cb/1ONi743GHAYT37xDp4CB02mHlHfqRNx2zSUmJpmLqdyi2u8HDPypaAxn7myrYvPyA9MZ6UxESykowARF8KZNf2xkRfCgRsy98v33OA107u51KQweGsmaXzIlj36bsOF/SD/Au7FbTmj70Pf9tMvKcXe1q1KNTxriSzbIQQQuTLgFpfpMuwwXQcN5IDC5eyd+UaDNhO8bUuxf7KkwPw1yWTaPJlybVmBL/zEYNOnSFLr+fjZfPZtOZ/eQKRGWPG0qxLLzJq5dBT/zcZ5/S0mvSZ5fm/ohbie8oTr72n2WUyMfLVN22uqWvQiBffGolXzEUyjhxh1uJdBZaEH79+Nc/e7oW/Zw4Xzsbgu30zVQK9UPSebKwxkN/3eVlKtluzL/3+ybJNVK9Zl2sxl5j0cL9bv+GFEDpsMGHjRrJ74dI897Gox7vTLBsJSIQQooLyRS1kZv8l4GfekoBA1LVpMlHzRxIA003O+8fXM6hpOE+lN/6H54ncr7r/PTmOyCfG0nFfJB0unGLVqcuWBNJjb/+XxU3DecT7OwJ1SaSafNiqjKB+wxss/XwOAG/OmYDh0nmyI48x9eNNNgHHX1ELqd+wJlm//YoxK4MLMUl0H7sy3+nCocMG8+Rro+hSPZWPpi1BOR/FS2O6QpNmzD7RG71PFYfrzbz8+seYWnXFcHQX70x/Js+UW+sZLUCR154pbRKQFJIEJEII4RzeqGvOxKLmgOhQAxRQy79XAnLMW2Ihz/lyj2BGexsxXtPjHRWNPjkFxcNATm1/Yv/dgpRunajnl0F2jC+Gq0abeh/vr1jLBY+atDYep4d+D3+bOtDj+FliGrah2r2ePDRqMe2qpjHj/kY2AYdGCwT+/vZXOtc0WKqs7n3lGYK8jFzPyGbkP5cs9UGsa5JoQZHGuo5IvY7tadQ1nKhdEfz29qsce2tqnlol1rTA6MLZGADqN6xJfJrCQ6MWc2DlmiL3btgrTM2RonCngMRYwm0QQghRBviSG4BkmX+vhhqA6IF/F7CyrP2X6nO9GjK2VW0CFu9Gl5Xbf5Ie4Iv+tQFk1K1MNR8jVZVUdDow1kwmNb4uG6wWuzls8se7khdHDK3Zk9SA3jFH8c5Ipu6ZvWR/6slL9wzioUdGo5zPHaaZMWZsblAxb5HDXoiLdUOpnHSOi0HBdGgUbpk2PPTqNnyy0xkaUptpdu8rc/sJOlby5IrRQKOu4ejwoFHXcOBVVp26bLmmI3PfWcXYyc9y7UoY8Td241dTYW9sZcLGjeTAyjU2+R83C0i0IOun309xPTiM3QuX8vTtA2hxcQsNBw4okYDEnUhSqxBClGE68wa2f2F2Dh/G5AURdDYXEbPe1zV8GAGoC+B5As91DmX35PHc0zmUJNQeE/sEUesiZNZfqs8Az207S+D8neiyTGR2bMCqNjWZF1Yf/Zz70XvWIf28B8npOeQoepRsHaezGrPw7nAO9gi1LH5X/Z8v8PXIoH/rBB7tdpKIZV9wMT6BnBzwzkihw8UzgO1snKEhtS1Bhca6nZA7U+Xgnm0MSInCIyeda1F7WXXqMqlGb5vAQntfD4WEWAqNRe2KQCGLqF1qsDNt8SJavvpmnl4VzdJ5EbA3gXZ/raZpegAPjVrM1iPxllk51kmvN6P9N3j44R6W+32zEvplmQQkQgjhZNqU2ZJcUVULQmoAtQEfoB5qAqoe+Nd/JhBYow7/+s8EvM3H3D5kLIE16nDbkLF4o07bTQAm3NWXOoEBPHNXXxTUYYEjB1pz5IDOkgvRcfBoanp6MaD7HaTO/IjbNqzhhbWLedHcjuyGQZy+/9+8+/ZXLGrZg/DHu2Gs4gG1M/g2pS0mL38M+hwy8WTmAYWkM56c+/MYzz/Snfo1/RjR2ZOrP02jBqeZ+84qy1TWv3QGMrx82aPkALYBh6Ogwr54mjZT5V+1/eiTcJonTqwluE9Lh4GFFizUOrPb8qVf88ZWxrTbR80bW/P8N3C0Fg1A0Cn19UGndnNg5RoWDLjf0hvSrmoaY5vF0q5qWp7z2dNmBe3be4YxTa4RdG53gTVHyjoZshFCCCezLrO+M59ueh1qMJEOvD1mLGMa1wCdjsWnY3jK/KVZDXVVXSPq7JcU876hnUMZ0zuMxVt3s2rPAfyBPnEnOOhTiXZxJ1gC3HX3cFoGBRGNiRtRe7nbakjmYt1QAuJP4a/3YETnUB4YOIAWx9dzfqdC8LwIVgEdnryXSiardNY/cr+g//n3g6x97FUA/HTZPPnaKOY+/xLTnriLaJ+epIS2YLsSRA/+JjapBs279CULDxq27UVKpo6qmVGkZIawdN5XlmEXbfikdYd0fKv70fRsDLxqG3BMGx9uGW7RpJ04QJWgmqSdsO1B0IZuLtRuk2/vxIGVaziwcg2pVomwT618ivoNa/LmnAmW47T75n2/4+GXmb/8Znm9PeueJ0dDTNbDYQdjYdGJqjzRwYfASjoG3RlCt8cOOKw3Uh5IQCJEBWK/IJooHZsWLqXfuJH8sXApRiDb6rlKqAmmetTejgxgZEhtPLPVBIuRIbX5L+rsgxrm12phQRBqLZDhT95PYNANhre9n0WPHmBk51C6Jp2j24mLHA0MwRsYN3AALSLXcjm4DZEjWtMzOonAdet5Jaw++/5ahz7+DKaTSbx1NR7DW8+jz8mi9ZVEWmsNNZlQdJDp5UlmlUB8AvUoNfw4UM2PHW274p2dAUBH01Va180madQHLMyoRrO0ZJrWS2Pf9bb8o7TDv0o6/Voksf2kD77GQL5PuwOdoqNT3B6be6YNn/x9LYbQlNzCY/t+WGRJOLUXOmwwj7dtQsPInzjeMJjVVs99um69JRn0wNo1jHw8nMkzHuHvaz589trXNgGFpaJr6gXmvrOKN+dMwGg0WIavtIDi8ReXWoIHa8v35B80aAXa8lvN177GiKN7UF7JLBshKpBX16+map3axEZf5q1bWF1UFI4euOfu4fxn+GP8uGI+f69dgQ51Cm1L81/C+xcu5Yz5yzAZ8ACeef1jslp1JezaUU7+vYOfk2NpN24kexcuZY/VTI2987/mAW9/Bt3lhV9WAkkZnkxbfYk3GzfA1yMOfNOIyqhCTlI29TxiUeKyqXTtKpy4hj46/qbtV/Q6TD2bcK5tGPVrGDC08iA7wZ+M6hlUNqoBSLYph1+OJDJA3xavjBQyvHzZfE8HTh2tSZKnL36ZKTw1+Cr7L/izNdKf1kEXST6+hTa6bvzqWY9ETzW11iM9gRFhkezZfozOPVvaJHJaBws2iax2eRzj169m8v79+KankoSO5Eea5zvlVpsNk5atIzMHln69jZmPvQPAP1e/oWZ1P24kZDB7l56gc7sZdGeIU6by2i98Z91DAtzSjJzCcKdZNqUSkDz55JO88MIL1KpViwMHDvDUU0/x999/3/R1EpAIUbLKag+Js9vtgTrTxNO8vfH6x9CqK6aju5g6/RkqoQYXmcAvbzxKrzYGth82MfC/X6Izv8YXNaAwoNbtSDL//H7uYponneNy/dZ8vWURy+ZFEACMXjafaj4+KKejOP3CNMbfFkq96jrik3LwM5ggOYO45ErkJMZRQ5cBGdmkZxpJj0vEz6jDmJGJzpRzS+9bqeaD4ulBpqc3nsG+0LomVPMh0+CB0UuPvo4fBPmSpnjhRSZ6nYKigCneB8U/A4PehF4H6QYf0oxheO09zXeRF1mVHMvIIWOJMVSjjTGGpJSt9AnviA6Y9epSls6LYO8rz3C+Tiv+rN0eBR0dgqO5s5Oe7GwTRqOBC2dj6NZoXJ42H3v7v/hkpZLq4UPLV96weS502GAm3HMnnSMj8epdj2oDWuR7Hq2HxL+KH5WMkJqpcCP6KnPfWUWvpydwW4gHu2K8OJzgS0L0ZRY4KYC3nibsqJ3OVqECkgceeIAlS5YwYcIEdu3axTPPPMOwYcNo3rw5165dK/C1EpCIsqCsfslrtPa32XGAgV4KZ6q3YPpPP/NLMacUzh4zlodDarPs1GVL7oM1Per/dIryPx77nh0d6jCHEfhgxEgeaFiD1acv898V3+BlvoYBNdHz5UGDubtOICgKEReus+B/a/E1P28yn2f48EdR6jQhKOUaLTyu4mm4yklTQ65lVuX03+vp37IV9WKvcjwmjtYhOow5mWRn6thPHeonXKeKMZnrWYGcPnmJqrp4mgZXQZ8D8VneVE5KwzM7FX1aJkpaNiSlo1dK/n+7ipcRxdMIXgZ03kbw0jYDJg8P9J56dD4e5AT5ktigMYQG4VFZj48ug1RTJXak16anzyUAtmZ3Jdtg5HbddkvybCVdBooCOh2kmLxZmDmSpl6n6aH7m79zOtAtvDpbTmXxeOv7gLx/+Wu9DjHXkuhY4yFGdA5lxv0DURQ911sFMydiDRNfHsqpi8l06NSYZct28OvvkXl6CNav+ppGp84QFdKYAUNH5Xs/7K+fnynzX+bhh3ugz87A38+bC2djePzF/xE2biTR+w5Rp0Nbp/ZQaNdftmyHpYemNFWogOSvv/7i77//5qmnnlIvqNNx4cIFZs+ezaxZswp8rbMCEh+grvl3HbbT5nTlbN+AO9sz6L5u/LTmL5QzcQxq0YRfjp9i45nzTrnm2H/9i9ZGhaNZsHjdukK9NrhdG5qGdeT0rr3U9sni9j5tOHTyBhmB9Tjz9z4uHD5a4DXrt2pB4/Zt8Uq4RMtGVdm5/Sj79p2xHFe3RTMatmvN+QOHuRx5EoB/3RFOWqUqBJLGnuOHuHzyNL37d6ZNi9qcvphEQKqRmklxAJw16cmoVQefTAMnjh+knreRpv4+nEpIYceJSO4ccic1UjIxXk8kySsAxSOVAEMCekXhamwOv+8/hR71SzK0TTAhQV5kZBnxNmRj0nniQQapmQa8DSZiU4xs23fMcnz39i2o4p1FZo4HnvpMMrMNeOmzScr0Js5QDQ9TJtl6I4FKPH76FBLSPPjn6Cm6hjbHR5+GyaTHU5dFusmAl8EECqQrPmSkZOLtpcfTkEmWzgOPnEyyc4x46DLRKQookGEy4GkwoVMUFMBkMuBBFuhAUXTEZurJ9PTFPysdb0MWWTl6PHXZ5uN15OToMeiyLdFHFkY8yAbz+bMVA0ar57MVPUZMkKOgy8lBMSnkZIMhJxtyFDAp5p8mdG470Fx4ilEPPp7g6wn1AqBmZRSdjjgC8fPNwOBnRFfZk0zPSiheRjw9c9B5G8nxNKL3MpDjZSTd04dzno2o63UdLzLx0mWSpnihoH72o3LqU0cfwz/Z7QhpkE3g5aNE6ruwLbMFKTpPOnGA2wx7uOQZjE9Tf1YeDCUFLzzJwiszkx4xh2idfImoVtWozykuUps6uiv8ndOByP3xpLfpAejoce0Ivzfyt/nitg9A5h35kdtCPGyClinzX2bs6J4ALPpqOzMfe4fx61cTUKc2CdHqzBntd62Hwv689rQv+LVbLnBKV7vQwYSrAgPr9+usXpiCVJiAxMPDg9TUVIYOHcpPP/1k2f/VV18RGBjIfffdZ3O8p6cnXl5elsd+fn5cunSpxAOSu4D1Nz1KCFHWKHodGPRg0IHN7+afBquf+ryPFYMendXxikGPTq8Dox48DOpjDz2K0YDOqAcPPVkGTwweinqcXkeWea6A0ZiDzseDTK9KKN4eeFQCfWVP0nwqo/h4YvTSYdIZuZYRRJDXdQD+UHpyRGmJZ2Ym//L5m6bekZxIbc669C78K2cHTTyOcyOrCh4+GexQurAvpw19r+yn7ZjqeH21CY+gBKJSG7O2Zji9miaQcy6BHddrEhYQTVhfIwaDnphrSSxbn4qhRke6Nkmke5MkchT1dvwdVZmdpwLoHpJAk9MXLEMwDzavh1dGCv9Ua8bOmm3ocuMEbW6cwic7HQUdsWFt6DJwlE1JefsAxFGFUuv1ZOLTFFr73HvTHIqb9XwcSf2ZwEo6EjJ1fHmiWqG/6F0VGNxq5dZb5U4BiVNn2QQFBWE0GomJibHZHxMTQ4sWeZdanjJlCq+99pozmwSoY8WKr2fuDh1qP6TOaoflz3Cd1TFWz2H3Gp3VcfbP2ZzP+ljrYwp4rLPaqbN7PYrd4wLaavd6RfvVfB21O9b2vdsc4+g16NSjrB9bdV9YHpv35T62vqb1Yx06vYPXO3xc1GtaPdbn3h/Le8L6fIrVe7ZqE5Cj0+UeDyg623uUjREFMOpMlnYpOp3N79eUIKrrbqDTq9e/qlSnhu665To2x6MjLqcKVQzx6HRq70O2zoiHLjv3mooBgz7H8j6zFKPl+go6opWaVNffwINs8zXt75me66aqBBgT0ZODSWfgmhJEDf0N9OSQo9MTQ3Vq6K9jIAcTenQ6HZ66rNzXZ1UjwDMBPQomnZ4r1KSW7ioGnXp8QnYgAR4J5scG4rMDCfSMV89veZyAXq+AQUeOwcgNpTpVPWMxGMCkN3DDFES1SvHo9Qomg5Hr2dUJ8onFYFAw6Y2cVRoQrL+AERMmDJzNqZ/7WGfgakZ1anhew2B+Pia7Jg2MFyy5EZd09fBT4rmo1KaePprMFB8q+yRajr+i1KEa17ik1CJYfwEdcEHXkrrGFDwzosjWGziW0oJNHt1pYzhGd8NeLmQ2JCW2ColBEGbcx+mUJmzS94As6BVzkGsx50lrdQ/njNXBaKB5cDydA85wwyuYpGpNOHUqi7mtuxNy5EdqhbQg5bcbXLmQxanqjel/bS8xMef56A9/BlUKoenJK9wI8MI77Xua9OrBJR8jYwJj2XrKwGsTP+fxl4ey8J1V6O9/DL86GaR655CQqeNiPNQP1OHlnciQpqe54VUd423V2FzHn2kvv07Dt6bS1cuXujlxjDu9nl0mE6tOXeaBZnVJad+EHy8cIwC1DL3ms9e+Zo9VYKFNqbW2e+FStjceR496sGzZDofH2b9m6byIAodgli3bYe4hOU+CLrNQBci0tjiaMeNsju5LReXUHpLatWsTHR1N9+7d+euvvyz7Z82aRZ8+fejWrZvN8aXVQwKwe8X7tPI9nfs/muya1DLGuOQxUCrXvJxTh9r6aKe26XJOLWrrrxS/zVk1qeVRwGNHbbzZaxydwxCDUZdNNsY8z1/OrEFtz6vqY52Byybb93Q5pza19Zetzl+DWh5XLY8PJtbhj5zadAjwpLduJ4Cl3QDbsruSmelHU99L1FROsTu7o/lxNPWVo3mOv6BrxY2G/lQ7m0hN5SS7sztSx3CdZvqTKIoaEp6qVANDeg3L64+mNAEgxPc8u5TOxPkFE2yoRkDcH9TVneCqqSo1jLHEZwQQ6JnA9fgabNgcxYrkWEY/P4m4eH9C4qPoG3WQqGZN+Kl6DfwTU0inMdnZYMzO5racXbT1PmJpoy7Nm7qnL3K1dQhV7ghk2aZMUnSN6Nk0kSRSCNhwnNCERM42a8KP1WsQHhVFi+grJAYG4h8fz4EAf3YYrvHk6J4Y9ZCdA/tPxNK+WVUAPvtqO+2uGelpMJBdOwDT+Wj2VNYTNqE3l5IN1K1sYsevOZzMqUHYtaNkVjZwwzOA84ZaNDHG8v43C5navjW9yAF0nGnZnD0+Sdyfk4hP+hli9I2ZdjyT+h3asmvhUrauXMNvn39AaEIikXVq8XPNmpzfd4h6HdrSPk7HeUMt6qSe50LGPsY+3IPjZ+Jo0bgKR1ceoVl0Ml9HXWTm/K8BGNk5lBn33InB6Mnf6Hn01TfJQq2IWhl16m/IsMH0GDeSfQuX8tfKNYQNG0yHcSPZunAp282P+5tn5RxeuQYFNYm2mvn/aUlAAOofXLGo9Uo6DxtMu3Ej2b5wKWfNr0kF2g4bTN+nHwdg76fzyAA6m1fy/WflGjoPG0zbcSM5tHApx1auQYdaUj4ZNXk3CzUHR0vi9QXLuW+28J5wL+7UQ+JWQzb2JKlVCFFe6FHz13JQi5/lWO03oNYXqW5+LtG8334OjQHbpGCT+TV61CCkBmpgoi2G541a10QxvzaH3P/hV0PNpYsDLqJ+GXiiziSqjhrc3ECtc2I0P58NNDCfI8f8XCXULzStWNvN648Kd+JOAYlTS8dnZWWxd+9e+vfvb9mn0+no378/O3fudOalhRDCrWg9DKnYBho5mIeRgavkBhOOJvSazMdaF0e7BsSY91/CdmXedHIDEBO2f30moQYjseQGGFqgdB24gBqQpJrPmWD1+zXzY3/UQCXBfD4v1KBLGxC3HoEW4macXqn1ww8/5Ouvv2bPnj3s3r2bZ555Bl9fXxYvXuzsSwshhMhHJupfxY6GWEw47ulQgMuoQUsl8xZP7lCOFth4ogYm1sGQwfw4ESEcc3pA8v3331O9enXeeOMNatWqxf79+wkPD+fq1avOvrQQQogCFCffQ+u5SUPtkcmyehyD2lNiQP1yUQA/1J6STNThAX8kKBGOOTWH5FZJDokQQpQflVBXJM4gN5ARrlVhckiEEEIITRpqHoonuUm+xWVATdoV5YcEJEIIIUpNGuraQdarHBdHJdS8FVF+SEAihBCi1GSh5goYUYMTz4IPz5c2jVmUHxKQCCGEKDVZqIGEHjWXxKOAYw0FPOfWCZCiWCQgEUIIUWqyyK2lolV81eiBINReE1/UGTmV8jmP1DgpfyQgEUIIUWq04m5ZqLM7rPmg1jTxQg1KYlGHdhwlr2rVakG+yMoL+e8ohBCiVGmzbdJR65NoeSRaXokH6rBOImr1Wm8cD98oqD0lgajTV51NW7dHOIcEJEIIIUpVEmq5ea0irCfq0Iy2jo/Wg5KBWgn2BurwjUZPbil8LZfkZlU+K3Fr04whdyhJOIcEJEIIIUpVunkDddjGA3WYJgE1QMk0/9QqycaZj9fySawXCtRySQpKcNWh9rJ4FXBMYTha8FCUHAlIhBBCuEw6ub0daeSud2OdX5KFmk+iBSTWgYEWkGST/4wdH9QholudlaNDAhJncvpaNkIIIUR+MsybQm6vSayD45JQAxZvcgMDPblDNlmoX2iOStJ7oSbL5jdjp7C0a8mUY+eQHhIhhBAulWDeCvqSN6Hmk2jVXbVeFT25AYmjImt6cvNRbpUWCMmUY+eQHhIhhBAuFVfI41JQAxEvcnNPtOGbLBz/he2BOpyTQW4wUdzeDQlInEt6SIQQQpQJGajBiwE18RVyA4x01MDDfnqw0XxsNrk9KsWlBT8SkDiHBCRCCCHKjDhypwxbD9lkYFvTROOBmtCaw60HE5I34lwSkAghhCgzsoEY1CRV64DEhJr46mhqr7Z+jpYIeyuspxqLkiU5JEIIIcoUbTaOFmBoZeQdJa4q5A7XgAQk7kx6SIQQQpRJ1j0e1sM2Wi+Jj/lxlvn5ksr/kIDEOSQgEUIIUSZZ1wTRpv7GoZZ39zM/dw21hwTUYR37YKIowwTWdUhEyZOARAghRJlkPWSjSUQtoJYFRKPmlWjsAxJvoCqFDzCsr2NEvkBLmuSQCCGEKJO0pFbrcu7ZwGXzT/uqrdnYBhFG8z6d1WYif1pPjB61B8aEGgCJkiEBnhBCiDIpx+6nRushcXS8HrWEvIHcGiVagOFnd7wHag8K2A4NaeQLtGRJD4kQQogyKb+AJD8m1ECkEuqXn3XlVUfDNh7m/daJs5JD4jwS4AkhhCiTihqQaD0k6ajBRg4FBxh6cqu/WveQyOJ6ziE9JEIIIcokLSgobECiHZ9t3rR8Ee0vc/vARFuYTytHn5PPcaJkSEAihBCiTNJ6LIrSQ5KDmmNiPTvHekjGmtZDYiS3N0XWsnEeGbIRQghRJhU1INFKzGegTgdOw7bXwz7Y0GE7ZIP59fZTjUuLF+pU5fJKAhIhhBBlkkLemS8FySG3cmsi6iJ91rVMHK11k4NtDol1QFLaQUklyndAIkM2QgghyixtKKWwx2oBiTYt2DrA0AISR7VItB6UHKvf9ZR+gmt5Hi6SHhIhhBBlUlF7SLJQVwm2rlGiTQXWAhv7L0UtONHZHZNfj0pRVAICinB8eZ/dIwGJEEKIMkkLCorSQxLvYJ/1EIzWA6EFINrCfFqCq/36ObfSY+FJ7gwee5XI+wVd3pNqJSARQghRJhU1IHHEPodEK5KmlaTPQu0lsa9bYl1GvjLFCxK0oMcRPxzni2g9OuWRBCRCCCHKpJIISLQAQ8sPMQBBqD0U1gGJ9bCOYvVaD/OxnrfQBkccrUyMVRvLIwlIhBBClEkl1UOiyUYNLDJReye0gCSF3B4TxW6zXqDPA7W3pCisgw4DuYmy+eWn3Greijsrr+9LCCFEOVfUpNabnSMTNcDIIHdYJgd1inA6tsFLDmovhnVA4mXeCss6STUA8AH8yZ3p4+gLOr/95UF5fV9CCCHKOS1gKGxhtPzOodUX0XJIsrAtLZ+GWkgtx8HxBvOxWs9GJoUfvrHPW0k079dhO2SjLQSoXVeGbIQQQgg3UhI9JNYl4bVejyzUwCLb6rhrqFOGravDaj0jWahfpnrU3hVHBb4cfdlaJ8ZCbhCiBR9ar0tVcoeMJIdECCGEcDNaj0FJDNloPR7ZqD0i2gJ8Gq3XxHpBvxyr12qBRCZqLokHtl+wgYCv1WPrnBTtd5PVY+tgRRsacvR+PYFqxXrn7kcCEiGEEGVWSeWQaAFJBrmJrNkFHK8FL9ZBhMaAmtzqa/da654N66nDBqtzQt6F/6yrydq3SRvOKQ+1SaR0vBBCiDJJ68VwVOq9sKyHaxRyy8onUnCgYz3UA7kBRja5AY2/3fHWQYPe7rVaD4l1aXrtOC1HxVFAouWwlIcqrtJDIoQQokxSgKvcWlKr9bBLFmoPSbb5Z2Y+17SebqzVKdGCgXTUarDp5AYgWi+HfUBi30Nisnts/VptnxaEGaye1wIWe96UfH0UZ5KARAghRIVlH1xcLeTx1kmwWg8F5t+vouahWAcV+fWQWOeMWM/c0Z4zYJvIqvUIaY8Lqlnig1q0rayQgEQIIUSFVdTiatbHWw/RWFd7BdvAwTr4sGYddFj3kNiXsrcub59F3pk22owcf9SS89bnL0u5JRKQCCGEqLCsezsKwzogySF3erD9eayHVqyHYDR6q9dZ1xixT3LVfmo9JZivaf3lrQUsBmwTQ8taTokEJEIIISqsok4dtq59kgrEYhuMaEGDdc0QrdCZNeuAREtMtZ7qm4XttazPkUluD4n9qsPa+zCQ/3o47koCEiGEEBVWcWqZWA/ZZJK3N0OTQd4hGx1qoqmWrJqD7Ywbk9VP+0q01r0vWo+I9cwc62Ee6zyUskICEiGEEBWaVvSsMBzlnOTks1+r2mpdZdULNdfDC8fBhv25TOQNdrSeD+uCakZya6MYsR3iKSskIBFCCFGhXUMNHgrD0fo51jNurGlBjnVviBG1Ron9a7Ktfmr7ravH2gck1jNwsqx+167haJjI3UlAIoQQQhSSo54Q62qv1rRAQmf1u4f591QK7iGxHkrKtjvWOrjRph1b95A4Gs4pCyQgEUIIIQrJujCa/T77YZ9Mchfe0wISI2pvTBK2wzFagGOd3Ko9b51PArlBiCe504C116SRO2RjnaNSFhbkk4BECCGEKCStOqx9FVf7xfi0fVrSq/UU30zzOVLJOwRkPbMmG9veEeugResVsZ6No1Wb1da2se6VqYr7V22VtWyEEEKIInCUm6EFDvZSUBfZs16rxvr19kNA1sdpQzDW+zQx5K67ox1rn5yrPaf1ynjiuBy+u5AeEiGEEOIWZeI4UMkwb9aJqo4CEuseEushmixsh2006UA0kEBuQKPlpsRjmwxrNLfP45beofNJD4kQQghxi2Lz2Z+JGpBoeR+OAhLrYMN6eCaJ3OGXgmqKWA/xKMB1q2trAUlRa624ggQkQgghhJNoC/ZpeR/59ZAodo+th3ZuVuTMemhHu+Zl8+9a+XptJWN3HraRIRshhBDCibS8Di3YsM7zcFRy3r43w9E+HDzvKIdF25eOGpC487CNBCRCCCFEKdAW47OvYWLfQ2IffFhPAXZEez6/ZFvtOW1KsLty57YJIYQQ5YYJtafCmn2wkWk+xj64iCX/oRb7IRtr1rN09Lh3kTQJSIQQQohSkIw6Ddiao2m/jhJkCyptn4MaxDgKWKx7T9w9IJEhGyGEEKKU2OeB2OeQFNcNHA/ZaFOHtaBEu74/7hecSA+JEEII4SLaLBxnrcxr30OirTjsgftVbpUeEiGEEMKFnLkqr3VxNW3zRB0Ccrf1bSQgEUIIIcqpTHLzVrSeEi0gcbchEglIhBBCiHIqE7WUPOT2kGiF0ipMD8krr7zC9u3bSUlJIS4uzlmXEUIIIUQhaQmu4H6l5J0WkHh6erJy5Uo+//xzZ11CCCGEEEWgBSQZ2K5z4w6cNoT02muvATBq1ChnXUIIIYQQRZBl3tJRi6Z5kbc2iqu4VU6Lp6cnXl5elsd+fn4ubI0QQghRvmizbrRqsGnmzR24VVLrlClTSExMtGyXLl1ydZOEEEKIciMbNQDR1tVJxH2GbIoUkMycORNFUQrcmjdvXuzGzJw5E39/f8tWt27dYp9LCCGEELZSgGvm3+OABBe2xV6Rhmw++OADvvrqqwKPOXPmTLEbk5mZSWZmfssHCSGEEOJWabNrUl3airyKFJBcv36d69evO6stQgghhKignJbUWr9+fapWrUqDBg0wGAyEhoYCcOrUKVJS3CWnVwghhBDuwGkByRtvvMHo0aMtj/fv3w/A7bffzp9//umsywohhBCiDNLhfsXaLPz8/EhMTMTf35+kpCRXN0cIIYQQhVCc72+3mvYrhBBCiIpJAhIhhBBCuJwEJEIIIYRwOQlIhBBCCOFyEpAIIYQQwuUkIBFCCCGEy0lAIoQQQgiXk4BECCGEEC4nAYkQQgghXM5ppeNLkp+fn6ubIIQQQohCKs73tlsHJNobunTpkotbIoQQQoii8vPzK3TpeLdeywagTp06TlnHxs/Pj0uXLlG3bt0KvU6O3AeV3AeV3AeV3AeV3AeV3IdcRbkXfn5+REdHF/rcbt1DAhTpzRRHUlJShf+AgdwHjdwHldwHldwHldwHldyHXIW5F0W9V5LUKoQQQgiXk4BECCGEEC5XYQOSjIwMXnvtNTIyMlzdFJeS+6CS+6CS+6CS+6CS+6CS+5DLmffC7ZNahRBCCFH+VdgeEiGEEEK4DwlIhBBCCOFyEpAIIYQQwuUkIBFCCCGEy0lAIoQQQgiXq5AByZNPPklUVBRpaWn89ddfdOnSxdVNcqrp06ejKIrNduzYMcvzXl5ezJkzh+vXr5OUlMSqVauoUaOGC1tcMnr37s3PP//MpUuXUBSFQYMG5Tnm9ddfJzo6mtTUVH7//XdCQkJsnq9SpQrLli0jISGBuLg4Fi5ciK+vb2m9hRJxs/uwePHiPJ+PdevW2RxTHu7Dyy+/zO7du0lMTCQmJoY1a9bQrFkzm2MK82+hfv36rF27lpSUFGJiYnj33XcxGAyl+VZuSWHuw+bNm/N8Jj7//HObY8r6fZgwYQIHDhwgISGBhIQEduzYQXh4uOX5ivBZ0NzsXpTm50GpSNsDDzygpKenK6NHj1ZatmypzJs3T4mNjVWqV6/u8rY5a5s+fbpy6NAhpWbNmpatWrVqluc/++wz5dy5c0rfvn2Vjh07Kjt27FC2bdvm8nbf6hYeHq7MmDFDue+++xRFUZRBgwbZPP/iiy8qcXFxyr333qu0bdtW+fHHH5XTp08rXl5elmN+/fVXZd++fUpYWJjSs2dP5cSJE8ry5ctd/t5K8j4sXrxY+fXXX20+H4GBgTbHlIf7sG7dOmXUqFFKq1atlHbt2ilr165Vzp49q/j4+FiOudm/Bb1erxw8eFD57bfflNDQUCU8PFy5evWq8tZbb7n8/ZXkfdi8ebMyb948m8+En59fuboPd999tzJw4EAlJCREadq0qfLmm28qGRkZSqtWrSrMZ6Gw96IUPw+uvxmluf3111/K7NmzLY91Op1y8eJF5aWXXnJ525y1TZ8+Xdm3b5/D5/z9/ZWMjAxlyJAhln3NmzdXFEVRunbt6vK2l9Tm6Is4Ojpaee6552zuRVpamvLggw8qgNKiRQtFURSlU6dOlmMGDBigmEwmpXbt2i5/TyV1HxYvXqysWbMm39eUx/sAKEFBQYqiKErv3r0t//1v9m8hPDxcyc7OVmrUqGE55vHHH1fi4+MVDw8Pl7+nkrgPoH4BffTRR/m+pjzeB0C5ceOG8uijj1bYz4Kje1Gan4cKNWTj4eFBp06d2LBhg2Wfoihs2LCB7t27u7Blzte0aVMuXbrE6dOnWbZsGfXr1wegU6dOeHp62tyTyMhIzp07V67vSaNGjahdu7bN+05MTGTXrl2W9929e3fi4uLYu3ev5ZgNGzaQk5ND165dS73NznT77bcTExPD8ePH+eyzz6hatarlufJ6HwICAgCIjY0FCvdvoXv37hw6dIirV69ajlm/fj0BAQG0bt26FFtfcuzvg2bEiBFcu3aNQ4cO8fbbb1OpUiXLc+XtPuj1eh588EF8fX3ZuXNnhf0sQN57oSmNz4Pbr/ZbkoKCgjAajcTExNjsj4mJoUWLFi5qlfPt2rWL0aNHExkZSe3atZk+fTpbt26lTZs21KpVi4yMDBISEmxeExMTQ61atVzUYufT3pujz4L2XK1atWz+gQGYTCZiY2PL1b2JiIhg9erVREVF0aRJE95++23WrVtH9+7dycnJKZf3QafT8fHHH7Nt2zaOHDkCUKh/C7Vq1XL4mdGeK2sc3QeAb775hnPnzhEdHU27du2YNWsWzZs3Z8iQIUD5uQ9t2rRh586deHt7k5yczODBgzl27Bjt27evcJ+F/O4FlN7noUIFJBVVRESE5fdDhw6xa9cuzp07xwMPPEBaWpoLWybcwXfffWf5/fDhwxw8eJAzZ85w++23s2nTJhe2zHnmzp1LmzZt6NWrl6ub4lL53YcFCxZYfj98+DCXL19m06ZNNG7cmDNnzpR2M50mMjKS9u3bExAQwNChQ/n666/p06ePq5vlEvndi2PHjpXa56FCDdlcv36d7OxsatasabO/Zs2aXLlyxUWtKn0JCQmcOHGCkJAQrly5gpeXl6XbVlPe74n23gr6LFy5ciVPVr3BYKBq1arl+t5ERUVx7do1y4yj8nYfZs+ezd13303fvn25dOmSZX9h/i1cuXLF4WdGe64sye8+OLJr1y4Am89EebgPWVlZnD59mn/++YdXXnmFAwcOMGnSpAr3WYD874Ujzvo8VKiAJCsri71799K/f3/LPp1OR//+/W3Gyso7X19fmjRpwuXLl9m7dy+ZmZk296RZs2YEBweX63sSFRXF5cuXbd63n58fXbt2tbzvnTt3UqVKFTp27Gg5pl+/fuj1ess/yPKobt26VKtWjcuXLwPl6z7Mnj2bwYMH069fP86ePWvzXGH+LezcuZO2bdtSvXp1yzF33nknCQkJHD16tFTeQ0ko6D440r59ewCbz0R5uA/29Ho9Xl5eFeqzkB/tXjjizM+Dy7N5S3N74IEHlLS0NOWRRx5RWrRooXzxxRdKbGysTXZwedvee+895bbbblOCg4OV7t27K7/99pty9epVJSgoSAF1etvZs2eV22+/XenYsaOyfft2Zfv27S5v961uvr6+SmhoqBIaGqooiqI888wzSmhoqFK/fn0F1Gm/sbGxyj333KO0adNGWbNmjcNpv3v37lW6dOmi9OjRQ4mMjCxz010Lug++vr7Ku+++q3Tt2lUJDg5W+vXrp+zZs0eJjIxUPD09y9V9mDt3rhIXF6fcdtttNtMXvb29Lcfc7N+CNr0xIiJCadeunXLXXXcpMTExZWqq583uQ+PGjZWpU6cqHTt2VIKDg5V77rlHOXXqlPLHH3+Uq/vw9ttvK71791aCg4OVNm3aKG+//bZiMpmUO+64o8J8FgpzL0r58+D6m1Ha28SJE5WzZ88q6enpyl9//aWEhYW5vE3O3L799lvl0qVLSnp6unLhwgXl22+/VRo3bmx53svLS5kzZ45y48YNJTk5Wfnhhx+UmjVrurzdt7r16dNHcWTx4sWWY15//XXl8uXLSlpamvL7778rTZs2tTlHlSpVlOXLlyuJiYlKfHy8smjRIsXX19fl762k7oO3t7cSERGhxMTEKBkZGUpUVJQyb968PAF6ebgP+Rk1apTlmML8W2jQoIHyv//9T0lJSVGuXr2qvPfee4rBYHD5+yup+1CvXj3ljz/+UK5fv66kpaUpJ06cUGbNmmVTd6I83IeFCxcqUVFRSnp6uhITE6P8/vvvlmCkonwWCnMvSvPzoDP/IoQQQgjhMhUqh0QIIYQQ7kkCEiGEEEK4nAQkQgghhHA5CUiEEEII4XISkAghhBDC5SQgEUIIIYTLSUAihBBCCJeTgEQIIYQQLicBiRBCCCFcTgISIYQQQricBCRCCCGEcLn/BxupL6dcCLLkAAAAAElFTkSuQmCC",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "idx = id_avg.argsort()\n",
    "\n",
    "plt.title('Features')\n",
    "plt.plot(range(len(idx)), id_avg[idx], color='red', label='CIFAR-10')\n",
    "plt.fill_between(\n",
    "  range(len(idx)),\n",
    "  id_avg[idx] - 2*id_std[idx],\n",
    "  id_avg[idx] + 2*id_std[idx],\n",
    "  color='red', alpha=.1\n",
    ")\n",
    "\n",
    "for k, (avg, std) in ood_features.items():\n",
    "  plt.scatter(range(len(idx)), avg[idx], s=1, label=k)\n",
    "  # plt.fill_between(range(len(idx)), avg[idx] - 2*std[idx], avg[idx] + 2*std[idx], alpha=.1)\n",
    "\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "weight = extractor.fc.weight.data.cpu()\n",
    "contrib = id_avg * weight\n",
    "thresh = np.percentile(contrib, 90)\n",
    "mask = contrib > thresh\n",
    "extractor.fc.weight.data = (mask * weight).to(device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "id_avg = (id_avg * weight).flatten()\n",
    "id_std = (id_std * weight).flatten()\n",
    "idx = id_avg.argsort()\n",
    "\n",
    "plt.title('Contribution')\n",
    "plt.plot(range(len(idx)), id_avg[idx], color='red', label='CIFAR-10')\n",
    "plt.fill_between(\n",
    "  range(len(idx)),\n",
    "  id_avg[idx] - 2*id_std[idx],\n",
    "  id_avg[idx] + 2*id_std[idx],\n",
    "  color='red', alpha=.1\n",
    ")\n",
    "\n",
    "for k, (avg, std) in ood_features.items():\n",
    "  avg, std = (avg * weight).flatten(), (std * weight).flatten()\n",
    "  plt.scatter(range(len(idx)), avg[idx], s=1, label=k)\n",
    "  # plt.fill_between(range(len(idx)), avg[idx] - 2*std[idx], avg[idx] + 2*std[idx], alpha=.1)\n",
    "\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "fe0623e0c9a64952bfb3cff581f8b089",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/20 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "curves.reset()\n",
    "\n",
    "res_id = run(extractor, dataloaders['CIFAR-10'], False, curves)\n",
    "\n",
    "id_avg, id_std = res_id['features']\n",
    "id_curves = res_id['curves']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "65a8f246c8d9412894290d253800b293",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/20 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "0a8515d38b7e4f33ac7e211b8ab08654",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/20 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "957eef5f489b4d61b0c11f89d39396f0",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/20 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "af1cd78fe3284545a9935a1a65520702",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/18 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "46c7f226ea5b4cf291d9b875638255e2",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/12 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "199a516bb28c4abab481d23a193e7966",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/642 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "ood_features = {}\n",
    "ood_metrics = {}\n",
    "\n",
    "for nm in ood_dls:\n",
    "  curves = id_curves.clone()\n",
    "\n",
    "  res_ood = run(extractor, dataloaders[nm], True, curves)\n",
    "\n",
    "  ood_features[nm] = res_ood['features']\n",
    "  ood_metrics[nm] = compute_metrics(res_ood['curves'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>FPR@95</th>\n",
       "      <th>AUROC</th>\n",
       "      <th>AUPR_In</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>SVHN</th>\n",
       "      <td>0.297000</td>\n",
       "      <td>0.946658</td>\n",
       "      <td>0.952062</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>LSUN</th>\n",
       "      <td>0.003800</td>\n",
       "      <td>0.998986</td>\n",
       "      <td>0.998997</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>LSUN_resize</th>\n",
       "      <td>0.044700</td>\n",
       "      <td>0.990291</td>\n",
       "      <td>0.991147</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>iSUN</th>\n",
       "      <td>0.051541</td>\n",
       "      <td>0.989713</td>\n",
       "      <td>0.991639</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>dtd</th>\n",
       "      <td>0.459220</td>\n",
       "      <td>0.869672</td>\n",
       "      <td>0.888168</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>places365</th>\n",
       "      <td>0.451291</td>\n",
       "      <td>0.901546</td>\n",
       "      <td>0.333267</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Avg.</th>\n",
       "      <td>0.217925</td>\n",
       "      <td>0.949478</td>\n",
       "      <td>0.859213</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "               FPR@95     AUROC   AUPR_In\n",
       "SVHN         0.297000  0.946658  0.952062\n",
       "LSUN         0.003800  0.998986  0.998997\n",
       "LSUN_resize  0.044700  0.990291  0.991147\n",
       "iSUN         0.051541  0.989713  0.991639\n",
       "dtd          0.459220  0.869672  0.888168\n",
       "places365    0.451291  0.901546  0.333267\n",
       "Avg.         0.217925  0.949478  0.859213"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pd.DataFrame(ood_metrics)\n",
    "df['Avg.'] = df.mean(axis=1)\n",
    "result_dfs['CIFAR-10_dice'] = df.T\n",
    "\n",
    "df.T[['FPR@95', 'AUROC', 'AUPR_In']]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>FPR@95</th>\n",
       "      <th>AUROC</th>\n",
       "      <th>AUPR_In</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>SVHN</th>\n",
       "      <td>0.875400</td>\n",
       "      <td>0.818515</td>\n",
       "      <td>0.863067</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>LSUN</th>\n",
       "      <td>0.147800</td>\n",
       "      <td>0.974348</td>\n",
       "      <td>0.976312</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>LSUN_resize</th>\n",
       "      <td>0.707300</td>\n",
       "      <td>0.801376</td>\n",
       "      <td>0.814240</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>iSUN</th>\n",
       "      <td>0.746106</td>\n",
       "      <td>0.789539</td>\n",
       "      <td>0.817554</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>dtd</th>\n",
       "      <td>0.843617</td>\n",
       "      <td>0.710096</td>\n",
       "      <td>0.765005</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>places365</th>\n",
       "      <td>0.782843</td>\n",
       "      <td>0.781495</td>\n",
       "      <td>0.141730</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Avg.</th>\n",
       "      <td>0.683844</td>\n",
       "      <td>0.812562</td>\n",
       "      <td>0.729651</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "               FPR@95     AUROC   AUPR_In\n",
       "SVHN         0.875400  0.818515  0.863067\n",
       "LSUN         0.147800  0.974348  0.976312\n",
       "LSUN_resize  0.707300  0.801376  0.814240\n",
       "iSUN         0.746106  0.789539  0.817554\n",
       "dtd          0.843617  0.710096  0.765005\n",
       "places365    0.782843  0.781495  0.141730\n",
       "Avg.         0.683844  0.812562  0.729651"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_dfs['CIFAR-100_energy'][['FPR@95', 'AUROC', 'AUPR_In']]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>FPR@95</th>\n",
       "      <th>AUROC</th>\n",
       "      <th>AUPR_In</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>SVHN</th>\n",
       "      <td>0.591500</td>\n",
       "      <td>0.886079</td>\n",
       "      <td>0.903695</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>LSUN</th>\n",
       "      <td>0.009200</td>\n",
       "      <td>0.997392</td>\n",
       "      <td>0.997412</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>LSUN_resize</th>\n",
       "      <td>0.516800</td>\n",
       "      <td>0.893147</td>\n",
       "      <td>0.902909</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>iSUN</th>\n",
       "      <td>0.494006</td>\n",
       "      <td>0.894948</td>\n",
       "      <td>0.909927</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>dtd</th>\n",
       "      <td>0.612234</td>\n",
       "      <td>0.771527</td>\n",
       "      <td>0.793554</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>places365</th>\n",
       "      <td>0.802904</td>\n",
       "      <td>0.774476</td>\n",
       "      <td>0.140147</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Avg.</th>\n",
       "      <td>0.504441</td>\n",
       "      <td>0.869595</td>\n",
       "      <td>0.774607</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "               FPR@95     AUROC   AUPR_In\n",
       "SVHN         0.591500  0.886079  0.903695\n",
       "LSUN         0.009200  0.997392  0.997412\n",
       "LSUN_resize  0.516800  0.893147  0.902909\n",
       "iSUN         0.494006  0.894948  0.909927\n",
       "dtd          0.612234  0.771527  0.793554\n",
       "places365    0.802904  0.774476  0.140147\n",
       "Avg.         0.504441  0.869595  0.774607"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_dfs['CIFAR-100_dice'][['FPR@95', 'AUROC', 'AUPR_In']]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>FPR@95</th>\n",
       "      <th>AUROC</th>\n",
       "      <th>AUPR_In</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>SVHN</th>\n",
       "      <td>0.406300</td>\n",
       "      <td>0.939926</td>\n",
       "      <td>0.953340</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>LSUN</th>\n",
       "      <td>0.038100</td>\n",
       "      <td>0.991504</td>\n",
       "      <td>0.992562</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>LSUN_resize</th>\n",
       "      <td>0.092800</td>\n",
       "      <td>0.981238</td>\n",
       "      <td>0.984919</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>iSUN</th>\n",
       "      <td>0.100616</td>\n",
       "      <td>0.980683</td>\n",
       "      <td>0.986024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>dtd</th>\n",
       "      <td>0.563121</td>\n",
       "      <td>0.864202</td>\n",
       "      <td>0.898692</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>places365</th>\n",
       "      <td>0.397665</td>\n",
       "      <td>0.918054</td>\n",
       "      <td>0.385097</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Avg.</th>\n",
       "      <td>0.266434</td>\n",
       "      <td>0.945935</td>\n",
       "      <td>0.866772</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "               FPR@95     AUROC   AUPR_In\n",
       "SVHN         0.406300  0.939926  0.953340\n",
       "LSUN         0.038100  0.991504  0.992562\n",
       "LSUN_resize  0.092800  0.981238  0.984919\n",
       "iSUN         0.100616  0.980683  0.986024\n",
       "dtd          0.563121  0.864202  0.898692\n",
       "places365    0.397665  0.918054  0.385097\n",
       "Avg.         0.266434  0.945935  0.866772"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_dfs['CIFAR-10_energy'][['FPR@95', 'AUROC', 'AUPR_In']]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>FPR@95</th>\n",
       "      <th>AUROC</th>\n",
       "      <th>AUPR_In</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>SVHN</th>\n",
       "      <td>0.297000</td>\n",
       "      <td>0.946658</td>\n",
       "      <td>0.952062</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>LSUN</th>\n",
       "      <td>0.003800</td>\n",
       "      <td>0.998986</td>\n",
       "      <td>0.998997</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>LSUN_resize</th>\n",
       "      <td>0.044700</td>\n",
       "      <td>0.990291</td>\n",
       "      <td>0.991147</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>iSUN</th>\n",
       "      <td>0.051541</td>\n",
       "      <td>0.989713</td>\n",
       "      <td>0.991639</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>dtd</th>\n",
       "      <td>0.459220</td>\n",
       "      <td>0.869672</td>\n",
       "      <td>0.888168</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>places365</th>\n",
       "      <td>0.451291</td>\n",
       "      <td>0.901546</td>\n",
       "      <td>0.333267</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Avg.</th>\n",
       "      <td>0.217925</td>\n",
       "      <td>0.949478</td>\n",
       "      <td>0.859213</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "               FPR@95     AUROC   AUPR_In\n",
       "SVHN         0.297000  0.946658  0.952062\n",
       "LSUN         0.003800  0.998986  0.998997\n",
       "LSUN_resize  0.044700  0.990291  0.991147\n",
       "iSUN         0.051541  0.989713  0.991639\n",
       "dtd          0.459220  0.869672  0.888168\n",
       "places365    0.451291  0.901546  0.333267\n",
       "Avg.         0.217925  0.949478  0.859213"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_dfs['CIFAR-10_dice'][['FPR@95', 'AUROC', 'AUPR_In']]"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "jax",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.9"
  },
  "orig_nbformat": 4,
  "vscode": {
   "interpreter": {
    "hash": "3fa6689b546259a140f700df87133380835e0a96e9af08af581bf45c0ff0ee88"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}