biasbios.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "b94de045",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-03-07T18:09:38.233068Z",
     "iopub.status.busy": "2024-03-07T18:09:38.232522Z",
     "iopub.status.idle": "2024-03-07T18:09:42.527852Z",
     "shell.execute_reply": "2024-03-07T18:09:42.525902Z"
    },
    "papermill": {
     "duration": 4.358771,
     "end_time": "2024-03-07T18:09:42.532301",
     "exception": false,
     "start_time": "2024-03-07T18:09:38.173530",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "from json import dumps\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import torch\n",
    "import transformers\n",
    "from tqdm.autonotebook import tqdm\n",
    "\n",
    "import loader\n",
    "from models import BinningCalibrator\n",
    "import utils\n",
    "\n",
    "split_ratio_test = 0.3\n",
    "split_ratio_postproc = 0.1  # among all training data\n",
    "\n",
    "# These seeds control the randomness for the post-process/test split and in\n",
    "# postprocessing.  It does not affect pre-training data nor the randomness in\n",
    "# pre-training, i.e., we assume the pre-trained predictor to be fixed.  Results\n",
    "# will be aggregated over the seeds.\n",
    "seeds = range(33, 38)\n",
    "\n",
    "# This seed controls the randomness during pre-training (fixed)\n",
    "seed_pretrain = 33\n",
    "\n",
    "# Settings for calibration\n",
    "split_ratio_calib = 0.3\n",
    "n_bins_calib = 70\n",
    "\n",
    "max_workers = 4\n",
    "device = torch.device(\"cuda\") if torch.cuda.is_available() else torch.device(\n",
    "    \"cpu\")\n",
    "\n",
    "data_dir = \"data/biasbios\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "08eed660",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-03-07T18:09:42.611674Z",
     "iopub.status.busy": "2024-03-07T18:09:42.610784Z",
     "iopub.status.idle": "2024-03-07T18:09:57.660372Z",
     "shell.execute_reply": "2024-03-07T18:09:57.659741Z"
    },
    "papermill": {
     "duration": 15.083608,
     "end_time": "2024-03-07T18:09:57.661671",
     "exception": false,
     "start_time": "2024-03-07T18:09:42.578063",
     "status": "completed"
    },
    "tags": []
   },
   "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>Group</th>\n",
       "      <th>female</th>\n",
       "      <th>male</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Target</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>accountant</th>\n",
       "      <td>2081</td>\n",
       "      <td>3571</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>architect</th>\n",
       "      <td>2398</td>\n",
       "      <td>7715</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>attorney</th>\n",
       "      <td>12494</td>\n",
       "      <td>20113</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>chiropractor</th>\n",
       "      <td>690</td>\n",
       "      <td>1908</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>comedian</th>\n",
       "      <td>592</td>\n",
       "      <td>2207</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>composer</th>\n",
       "      <td>918</td>\n",
       "      <td>4682</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>dentist</th>\n",
       "      <td>5153</td>\n",
       "      <td>9326</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>dietitian</th>\n",
       "      <td>3689</td>\n",
       "      <td>289</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>dj</th>\n",
       "      <td>211</td>\n",
       "      <td>1274</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>filmmaker</th>\n",
       "      <td>2310</td>\n",
       "      <td>4699</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>interior_designer</th>\n",
       "      <td>1183</td>\n",
       "      <td>280</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>journalist</th>\n",
       "      <td>9873</td>\n",
       "      <td>10077</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>model</th>\n",
       "      <td>6214</td>\n",
       "      <td>1288</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>nurse</th>\n",
       "      <td>17236</td>\n",
       "      <td>1735</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>painter</th>\n",
       "      <td>3543</td>\n",
       "      <td>4193</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>paralegal</th>\n",
       "      <td>1499</td>\n",
       "      <td>268</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>pastor</th>\n",
       "      <td>609</td>\n",
       "      <td>1923</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>personal_trainer</th>\n",
       "      <td>654</td>\n",
       "      <td>778</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>photographer</th>\n",
       "      <td>8689</td>\n",
       "      <td>15635</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>physician</th>\n",
       "      <td>19579</td>\n",
       "      <td>18986</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>poet</th>\n",
       "      <td>3441</td>\n",
       "      <td>3570</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>professor</th>\n",
       "      <td>53290</td>\n",
       "      <td>64820</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>psychologist</th>\n",
       "      <td>11385</td>\n",
       "      <td>6910</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>rapper</th>\n",
       "      <td>136</td>\n",
       "      <td>1271</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>software_engineer</th>\n",
       "      <td>1089</td>\n",
       "      <td>5817</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>surgeon</th>\n",
       "      <td>1972</td>\n",
       "      <td>11301</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>teacher</th>\n",
       "      <td>9768</td>\n",
       "      <td>6428</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>yoga_teacher</th>\n",
       "      <td>1406</td>\n",
       "      <td>257</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "Group              female   male\n",
       "Target                          \n",
       "accountant           2081   3571\n",
       "architect            2398   7715\n",
       "attorney            12494  20113\n",
       "chiropractor          690   1908\n",
       "comedian              592   2207\n",
       "composer              918   4682\n",
       "dentist              5153   9326\n",
       "dietitian            3689    289\n",
       "dj                    211   1274\n",
       "filmmaker            2310   4699\n",
       "interior_designer    1183    280\n",
       "journalist           9873  10077\n",
       "model                6214   1288\n",
       "nurse               17236   1735\n",
       "painter              3543   4193\n",
       "paralegal            1499    268\n",
       "pastor                609   1923\n",
       "personal_trainer      654    778\n",
       "photographer         8689  15635\n",
       "physician           19579  18986\n",
       "poet                 3441   3570\n",
       "professor           53290  64820\n",
       "psychologist        11385   6910\n",
       "rapper                136   1271\n",
       "software_engineer    1089   5817\n",
       "surgeon              1972  11301\n",
       "teacher              9768   6428\n",
       "yoga_teacher         1406    257"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "example from attribute-aware dataset: {\n",
      "  \"bio\": \"Male. He produced scores of films including such as al-Dhareeh (the shrine), 1976, winner of the Cinema Institute Films\\u2019 Award at the Documentary and Short Films Festival, Egypt, 1977; as well as the Kelibia Festival Award, Tunisia, 1978; al-Mahatta (The Station), winner of a major award at Oberhausen Short Film Festival, Germany, 1989; the EU Award at FESPACO Festival, Burkina Faso, 1990; The Silver Sword Award at Damascus festival, 1990; and The Silver Tanit Award, Carthage festival, Tunisia, 1991. Eltayeb has served as head of the Sudanese Film group for several terms and as secretary of the Sudanese Film club. He has written numerous articles on cinema, published in major Sudanese newspapers. He is currently working on a long fiction film, al-Siraj wal-attama (The Lantern and Darkness).\",\n",
      "  \"title\": 9,\n",
      "  \"gender\": 1\n",
      "}\n",
      "example from attribute-blind dataset: {\n",
      "  \"bio\": \"He produced scores of films including such as al-Dhareeh (the shrine), 1976, winner of the Cinema Institute Films\\u2019 Award at the Documentary and Short Films Festival, Egypt, 1977; as well as the Kelibia Festival Award, Tunisia, 1978; al-Mahatta (The Station), winner of a major award at Oberhausen Short Film Festival, Germany, 1989; the EU Award at FESPACO Festival, Burkina Faso, 1990; The Silver Sword Award at Damascus festival, 1990; and The Silver Tanit Award, Carthage festival, Tunisia, 1991. Eltayeb has served as head of the Sudanese Film group for several terms and as secretary of the Sudanese Film club. He has written numerous articles on cinema, published in major Sudanese newspapers. He is currently working on a long fiction film, al-Siraj wal-attama (The Lantern and Darkness).\",\n",
      "  \"title\": 9,\n",
      "  \"gender\": 1\n",
      "}\n"
     ]
    }
   ],
   "source": [
    "## Download BiasBios dataset\n",
    "\n",
    "(dataset, labels, label_names, groups,\n",
    " group_names) = loader.load_biasbios(data_dir, add_sensitive_attribute=True)\n",
    "dataset_u = loader.load_biasbios(data_dir, add_sensitive_attribute=False)[0]\n",
    "display(\n",
    "    loader.dataset_stats(dataset['title'], label_names, dataset['gender'],\n",
    "                         group_names))\n",
    "\n",
    "n_classes = len(label_names)\n",
    "n_groups = len(group_names)\n",
    "\n",
    "print('example from attribute-aware dataset:',\n",
    "      dumps(dataset[seed_pretrain], indent=2))\n",
    "print('example from attribute-blind dataset:',\n",
    "      dumps(dataset_u[seed_pretrain], indent=2))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "f0121d26",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-03-07T18:09:57.745289Z",
     "iopub.status.busy": "2024-03-07T18:09:57.745006Z",
     "iopub.status.idle": "2024-03-07T18:10:03.627451Z",
     "shell.execute_reply": "2024-03-07T18:10:03.626322Z"
    },
    "papermill": {
     "duration": 5.927935,
     "end_time": "2024-03-07T18:10:03.631191",
     "exception": false,
     "start_time": "2024-03-07T18:09:57.703256",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "## Split data into (test + post-processing) and pre-training sets\n",
    "\n",
    "split_dataset = dataset.train_test_split(\n",
    "    test_size=(1 - split_ratio_test) * (1 - split_ratio_postproc),\n",
    "    seed=seed_pretrain,\n",
    ")\n",
    "dataset_ = split_dataset['train']\n",
    "dataset_pretrain = split_dataset['test']\n",
    "\n",
    "# Encode the joint (A, Y) labels by flattening\n",
    "dataset_u = dataset_u.add_column(\n",
    "    'labels_ay',\n",
    "    np.array(dataset_u['gender']) * n_classes + np.array(dataset_u['title']))\n",
    "\n",
    "split_dataset_u = dataset_u.train_test_split(\n",
    "    test_size=(1 - split_ratio_test) * (1 - split_ratio_postproc),\n",
    "    seed=seed_pretrain,\n",
    ")\n",
    "dataset_u_ = split_dataset_u['train']\n",
    "dataset_pretrain_u = split_dataset_u['test']\n",
    "\n",
    "labels_ = np.array(dataset_['title'])\n",
    "groups_ = np.array(dataset_['gender'])\n",
    "\n",
    "n_samples = len(dataset)\n",
    "n_test = int(n_samples * split_ratio_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "21dbaf98",
   "metadata": {
    "papermill": {
     "duration": 0.028978,
     "end_time": "2024-03-07T18:10:03.706508",
     "exception": false,
     "start_time": "2024-03-07T18:10:03.677530",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## BERT model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "ad429d3c",
   "metadata": {},
   "outputs": [],
   "source": [
    "model_name = \"bert-base-uncased\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "63e68f43",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "ae685cd10a904220af08a8b79ceeab93",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Running tokenizer:   0%|          | 0/145566 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "24489e4bcc4d4eb3891dac14948c639e",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Running tokenizer:   0%|          | 0/247857 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "c94bfffb04dc4eef84da5800d455e2b5",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Running tokenizer:   0%|          | 0/145566 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "774487b6ff7f40679142ed24979537f8",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Running tokenizer:   0%|          | 0/247857 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "tokenizer = transformers.AutoTokenizer.from_pretrained(model_name)\n",
    "\n",
    "\n",
    "def tokenize_function(examples):\n",
    "  tokenized_examples = tokenizer(examples[\"bio\"],\n",
    "                                 padding=False,\n",
    "                                 max_length=tokenizer.model_max_length,\n",
    "                                 truncation=True)\n",
    "  return tokenized_examples\n",
    "\n",
    "\n",
    "tokenized_dataset_ = dataset_.map(\n",
    "    tokenize_function,\n",
    "    batched=True,\n",
    "    remove_columns=['bio'],\n",
    "    desc=\"Running tokenizer\",\n",
    ")\n",
    "tokenized_dataset_pretrain = dataset_pretrain.map(\n",
    "    tokenize_function,\n",
    "    batched=True,\n",
    "    remove_columns=['bio'],\n",
    "    desc=\"Running tokenizer\",\n",
    ")\n",
    "tokenized_dataset_u_ = dataset_u_.map(\n",
    "    tokenize_function,\n",
    "    batched=True,\n",
    "    remove_columns=['bio'],\n",
    "    desc=\"Running tokenizer\",\n",
    ")\n",
    "tokenized_dataset_pretrain_u = dataset_pretrain_u.map(\n",
    "    tokenize_function,\n",
    "    batched=True,\n",
    "    remove_columns=['bio'],\n",
    "    desc=\"Running tokenizer\",\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "f5094a2d",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-03-05 10:27:50.767288: I tensorflow/core/util/port.cc:110] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable `TF_ENABLE_ONEDNN_OPTS=0`.\n",
      "2024-03-05 10:27:50.828739: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.\n",
      "To enable the following instructions: AVX2 AVX512F AVX512_VNNI FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.\n",
      "2024-03-05 10:27:51.740701: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT\n"
     ]
    }
   ],
   "source": [
    "data_collator = transformers.DataCollatorWithPadding(tokenizer)\n",
    "\n",
    "\n",
    "def train(model, dataset_train, dataset_test, label_col_name, batch_size,\n",
    "          n_epochs, lr, warmup_ratio, weight_decay, max_grad_norm):\n",
    "\n",
    "  dataloader_train = torch.utils.data.DataLoader(\n",
    "      dataset_train,\n",
    "      shuffle=True,\n",
    "      collate_fn=data_collator,\n",
    "      batch_size=batch_size,\n",
    "  )\n",
    "  dataloader_test = torch.utils.data.DataLoader(\n",
    "      dataset_test,\n",
    "      collate_fn=data_collator,\n",
    "      batch_size=batch_size,\n",
    "  )\n",
    "\n",
    "  no_decay = [\"bias\", \"LayerNorm.weight\"]\n",
    "  optimizer_grouped_parameters = [\n",
    "      {\n",
    "          \"params\": [\n",
    "              p for n, p in model.named_parameters()\n",
    "              if not any(nd in n for nd in no_decay)\n",
    "          ],\n",
    "          \"weight_decay\": weight_decay,\n",
    "      },\n",
    "      {\n",
    "          \"params\": [\n",
    "              p for n, p in model.named_parameters()\n",
    "              if any(nd in n for nd in no_decay)\n",
    "          ],\n",
    "          \"weight_decay\": 0.0\n",
    "      },\n",
    "  ]\n",
    "  optimizer = torch.optim.AdamW(optimizer_grouped_parameters, lr=lr)\n",
    "  scheduler = transformers.get_linear_schedule_with_warmup(\n",
    "      optimizer,\n",
    "      num_warmup_steps=(warmup_ratio * n_epochs * len(dataloader_train)),\n",
    "      num_training_steps=n_epochs * len(dataloader_train))\n",
    "\n",
    "  model_input_args = list(model.forward.__code__.co_varnames)\n",
    "  loss_fn = torch.nn.CrossEntropyLoss()\n",
    "\n",
    "  for epoch in range(n_epochs):\n",
    "\n",
    "    model.train()\n",
    "    for batch in tqdm(dataloader_train, desc=f\"train epoch {epoch+1}\"):\n",
    "      batch = {k: v.to(device) for k, v in batch.items()}\n",
    "      optimizer.zero_grad()\n",
    "      outputs = model(**{\n",
    "          k: v for k, v in batch.items() if k in model_input_args\n",
    "      })\n",
    "      loss = loss_fn(outputs.logits, batch[label_col_name])\n",
    "      loss.backward()\n",
    "      torch.nn.utils.clip_grad_norm_(model.parameters(), max_grad_norm)\n",
    "      optimizer.step()\n",
    "      scheduler.step()\n",
    "\n",
    "    model.eval()\n",
    "    with torch.no_grad():\n",
    "      test_loss = 0\n",
    "      test_acc = 0\n",
    "      for batch in tqdm(dataloader_test, desc=f\"test {epoch+1}\"):\n",
    "        batch = {k: v.to(device) for k, v in batch.items()}\n",
    "        outputs = model(**{\n",
    "            k: v for k, v in batch.items() if k in model_input_args\n",
    "        })\n",
    "        probas = outputs.logits\n",
    "        test_loss += loss_fn(probas, batch[label_col_name]).item()\n",
    "        probas_y = probas.softmax(dim=1).reshape(len(batch['title']), -1,\n",
    "                                                 n_classes).sum(dim=1)\n",
    "        test_acc += (probas_y.argmax(dim=1) == batch['title']).sum().item()\n",
    "      test_loss /= len(dataset_test['title'])\n",
    "      test_acc /= len(dataset_test['title'])\n",
    "      print(\n",
    "          f\"epoch {epoch+1}/{n_epochs}: loss={test_loss:.4f}, acc={test_acc:.4f}\"\n",
    "      )\n",
    "\n",
    "\n",
    "def predict_probas(model, dataset, batch_size):\n",
    "  dataloader = torch.utils.data.DataLoader(\n",
    "      dataset,\n",
    "      collate_fn=data_collator,\n",
    "      batch_size=batch_size,\n",
    "  )\n",
    "  model_input_args = list(model.forward.__code__.co_varnames)\n",
    "  model.eval()\n",
    "  with torch.no_grad():\n",
    "    probas = []\n",
    "    for batch in tqdm(dataloader, desc=\"inference\"):\n",
    "      batch = {\n",
    "          k: v.to(device) for k, v in batch.items() if k in model_input_args\n",
    "      }\n",
    "      outputs = model(**batch)\n",
    "      probas.append(outputs.logits.softmax(dim=1).cpu().numpy())\n",
    "    probas = np.concatenate(probas, axis=0)\n",
    "    return probas"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "87c196cd",
   "metadata": {},
   "source": [
    "## Fine-tune BERT"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "5185128f",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of the model checkpoint at bert-base-uncased were not used when initializing BertForSequenceClassification: ['cls.predictions.bias', 'cls.predictions.decoder.weight', 'cls.predictions.transform.LayerNorm.bias', 'cls.predictions.transform.dense.weight', 'cls.seq_relationship.weight', 'cls.predictions.transform.LayerNorm.weight', 'cls.seq_relationship.bias', 'cls.predictions.transform.dense.bias']\n",
      "- This IS expected if you are initializing BertForSequenceClassification from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n",
      "- This IS NOT expected if you are initializing BertForSequenceClassification from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of BertForSequenceClassification were not initialized from the model checkpoint at bert-base-uncased and are newly initialized: ['classifier.weight', 'classifier.bias']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "18a91353522742b68c073bd29a83cc45",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "train epoch 1:   0%|          | 0/7746 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "You're using a BertTokenizerFast tokenizer. Please note that with a fast tokenizer, using the `__call__` method is faster than using a method to encode the text followed by a call to the `pad` method to get a padded encoding.\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "5f6bf28a5cae44db954e959f2366c4ff",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "test 1:   0%|          | 0/455 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 1/3: loss=0.0156, acc=0.8526\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "80688a0ba688412d8b8e9a3cf604f8d3",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "train epoch 2:   0%|          | 0/7746 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a56602a0975f49baa1c26fe28c7c0341",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "test 2:   0%|          | 0/455 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 2/3: loss=0.0146, acc=0.8614\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "8e7ed7ded06d4880bf24e367742c8f38",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "train epoch 3:   0%|          | 0/7746 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "59e77b8c85104891b78dfa2e1f8aa33a",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "test 3:   0%|          | 0/455 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 3/3: loss=0.0154, acc=0.8605\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of the model checkpoint at bert-base-uncased were not used when initializing BertForSequenceClassification: ['cls.predictions.bias', 'cls.predictions.decoder.weight', 'cls.predictions.transform.LayerNorm.bias', 'cls.predictions.transform.dense.weight', 'cls.seq_relationship.weight', 'cls.predictions.transform.LayerNorm.weight', 'cls.seq_relationship.bias', 'cls.predictions.transform.dense.bias']\n",
      "- This IS expected if you are initializing BertForSequenceClassification from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n",
      "- This IS NOT expected if you are initializing BertForSequenceClassification from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of BertForSequenceClassification were not initialized from the model checkpoint at bert-base-uncased and are newly initialized: ['classifier.weight', 'classifier.bias']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "386f626394754fdca731aee1017fc814",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "train epoch 1:   0%|          | 0/7746 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "65ed372f110a4f99b9721c1100ac7ee9",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "test 1:   0%|          | 0/455 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 1/3: loss=0.0158, acc=0.8520\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a40410f68bca466497c3452f19aac4f6",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "train epoch 2:   0%|          | 0/7746 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "beb287e5598f48999924615d3535e5ec",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "test 2:   0%|          | 0/455 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 2/3: loss=0.0153, acc=0.8594\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "f03c402765e748688844e5d5735ec5e8",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "train epoch 3:   0%|          | 0/7746 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "c97fcd74baa24ad6af85ccc811ddf267",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "test 3:   0%|          | 0/455 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 3/3: loss=0.0158, acc=0.8599\n"
     ]
    }
   ],
   "source": [
    "## (Pre-)train predictors\n",
    "\n",
    "batch_size = 32\n",
    "n_epochs = 3\n",
    "lr = 2e-5\n",
    "warmup_ratio = 0.1\n",
    "weight_decay = 0.01\n",
    "max_grad_norm = 1.0\n",
    "\n",
    "# Train attribute-aware p(Y | X) predictor\n",
    "transformers.set_seed(seed_pretrain)\n",
    "model = transformers.AutoModelForSequenceClassification.from_pretrained(\n",
    "    model_name, num_labels=n_classes).to(device)\n",
    "train(\n",
    "    model,\n",
    "    tokenized_dataset_pretrain,\n",
    "    tokenized_dataset_.train_test_split(test_size=0.1,\n",
    "                                        seed=seed_pretrain)['test'],\n",
    "    'title',\n",
    "    batch_size,\n",
    "    n_epochs,\n",
    "    lr,\n",
    "    warmup_ratio,\n",
    "    weight_decay,\n",
    "    max_grad_norm,\n",
    ")\n",
    "\n",
    "# Train attribute-blind p(A, Y | X) predictor\n",
    "transformers.set_seed(seed_pretrain)\n",
    "model_u = transformers.AutoModelForSequenceClassification.from_pretrained(\n",
    "    model_name, num_labels=n_groups * n_classes).to(device)\n",
    "train(\n",
    "    model_u,\n",
    "    tokenized_dataset_pretrain_u,\n",
    "    tokenized_dataset_u_.train_test_split(test_size=0.1,\n",
    "                                          seed=seed_pretrain)['test'],\n",
    "    'labels_ay',\n",
    "    batch_size,\n",
    "    n_epochs,\n",
    "    lr,\n",
    "    warmup_ratio,\n",
    "    weight_decay,\n",
    "    max_grad_norm,\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "843c7fdb",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "8737f866d9154b2c9fca848b3148146e",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "inference:   0%|          | 0/1138 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "You're using a BertTokenizerFast tokenizer. Please note that with a fast tokenizer, using the `__call__` method is faster than using a method to encode the text followed by a call to the `pad` method to get a padded encoding.\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "07f9b7f222fc491993aa92e322e0f083",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "inference:   0%|          | 0/1138 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Precompute P(A, Y | X) on (test + post-processing) set for aware setting\n",
    "probas_y_ = predict_probas(model, tokenized_dataset_,\n",
    "                           128).reshape(-1, n_classes)\n",
    "probas_ay_ = np.einsum(\"ij,ik->ijk\", np.eye(n_groups)[groups_], probas_y_)\n",
    "\n",
    "# Precompute P(A, Y | X) on (test + post-processing) set for blind setting\n",
    "probas_ay_u_ = predict_probas(model_u, tokenized_dataset_u_,\n",
    "                              128).reshape(-1, n_groups, n_classes)\n",
    "\n",
    "postprocess_kwargs = {\n",
    "    'n_test': n_test,\n",
    "    'n_classes': n_classes,\n",
    "    'n_groups': n_groups,\n",
    "    'labels': labels_,\n",
    "    'groups': groups_,\n",
    "    'probas_ay': probas_ay_,\n",
    "    'max_workers': max_workers,\n",
    "}\n",
    "postprocess_u_kwargs = {**postprocess_kwargs, 'probas_ay': probas_ay_u_}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "21cc0b0c",
   "metadata": {},
   "outputs": [
    {
      "name": "stdout",
      "output_type": "stream",
      "text": [
       "Calibration error of attribute-blind predictor: 1.7358485674272057\n"
      ]
     }
   ],
   "source": [
    "print(\n",
    "    'Calibration error of attribute-blind predictor:',\n",
    "    utils.calibration_error(probas_ay_u_.reshape(-1, n_groups * n_classes),\n",
    "                            groups_ * n_classes + labels_,\n",
    "                            n_bins=n_bins_calib))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "493f0b91",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 350x300 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Results for attribute-aware setting:\n"
     ]
    },
    {
     "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 tr th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr:last-of-type th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th colspan=\"2\" halign=\"left\">accuracy</th>\n",
       "      <th colspan=\"2\" halign=\"left\">delta_sp</th>\n",
       "      <th colspan=\"2\" halign=\"left\">delta_sp_rms</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>alpha</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>inf</th>\n",
       "      <td>0.864159</td>\n",
       "      <td>0.000346</td>\n",
       "      <td>0.083156</td>\n",
       "      <td>0.000290</td>\n",
       "      <td>0.023233</td>\n",
       "      <td>0.000082</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.080</th>\n",
       "      <td>0.864006</td>\n",
       "      <td>0.000376</td>\n",
       "      <td>0.080139</td>\n",
       "      <td>0.001456</td>\n",
       "      <td>0.022897</td>\n",
       "      <td>0.000189</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.070</th>\n",
       "      <td>0.863366</td>\n",
       "      <td>0.000465</td>\n",
       "      <td>0.070672</td>\n",
       "      <td>0.002480</td>\n",
       "      <td>0.021871</td>\n",
       "      <td>0.000294</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.060</th>\n",
       "      <td>0.862498</td>\n",
       "      <td>0.000554</td>\n",
       "      <td>0.060735</td>\n",
       "      <td>0.003311</td>\n",
       "      <td>0.020843</td>\n",
       "      <td>0.000343</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.050</th>\n",
       "      <td>0.860990</td>\n",
       "      <td>0.000690</td>\n",
       "      <td>0.050214</td>\n",
       "      <td>0.002938</td>\n",
       "      <td>0.019870</td>\n",
       "      <td>0.000283</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.040</th>\n",
       "      <td>0.859311</td>\n",
       "      <td>0.000800</td>\n",
       "      <td>0.041008</td>\n",
       "      <td>0.002559</td>\n",
       "      <td>0.019087</td>\n",
       "      <td>0.000295</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.030</th>\n",
       "      <td>0.856871</td>\n",
       "      <td>0.000838</td>\n",
       "      <td>0.032499</td>\n",
       "      <td>0.000757</td>\n",
       "      <td>0.017474</td>\n",
       "      <td>0.000430</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.020</th>\n",
       "      <td>0.851782</td>\n",
       "      <td>0.001019</td>\n",
       "      <td>0.023189</td>\n",
       "      <td>0.001714</td>\n",
       "      <td>0.013716</td>\n",
       "      <td>0.000592</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.015</th>\n",
       "      <td>0.846798</td>\n",
       "      <td>0.001245</td>\n",
       "      <td>0.018684</td>\n",
       "      <td>0.001551</td>\n",
       "      <td>0.011105</td>\n",
       "      <td>0.000442</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.010</th>\n",
       "      <td>0.838776</td>\n",
       "      <td>0.001396</td>\n",
       "      <td>0.014442</td>\n",
       "      <td>0.001359</td>\n",
       "      <td>0.007981</td>\n",
       "      <td>0.000404</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.005</th>\n",
       "      <td>0.827880</td>\n",
       "      <td>0.001224</td>\n",
       "      <td>0.009591</td>\n",
       "      <td>0.002083</td>\n",
       "      <td>0.004830</td>\n",
       "      <td>0.000276</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.002</th>\n",
       "      <td>0.817701</td>\n",
       "      <td>0.000589</td>\n",
       "      <td>0.006515</td>\n",
       "      <td>0.002049</td>\n",
       "      <td>0.002729</td>\n",
       "      <td>0.000381</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.001</th>\n",
       "      <td>0.813565</td>\n",
       "      <td>0.000627</td>\n",
       "      <td>0.005730</td>\n",
       "      <td>0.002235</td>\n",
       "      <td>0.002183</td>\n",
       "      <td>0.000498</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       accuracy            delta_sp           delta_sp_rms          \n",
       "           mean       std      mean       std         mean       std\n",
       "alpha                                                               \n",
       "inf    0.864159  0.000346  0.083156  0.000290     0.023233  0.000082\n",
       "0.080  0.864006  0.000376  0.080139  0.001456     0.022897  0.000189\n",
       "0.070  0.863366  0.000465  0.070672  0.002480     0.021871  0.000294\n",
       "0.060  0.862498  0.000554  0.060735  0.003311     0.020843  0.000343\n",
       "0.050  0.860990  0.000690  0.050214  0.002938     0.019870  0.000283\n",
       "0.040  0.859311  0.000800  0.041008  0.002559     0.019087  0.000295\n",
       "0.030  0.856871  0.000838  0.032499  0.000757     0.017474  0.000430\n",
       "0.020  0.851782  0.001019  0.023189  0.001714     0.013716  0.000592\n",
       "0.015  0.846798  0.001245  0.018684  0.001551     0.011105  0.000442\n",
       "0.010  0.838776  0.001396  0.014442  0.001359     0.007981  0.000404\n",
       "0.005  0.827880  0.001224  0.009591  0.002083     0.004830  0.000276\n",
       "0.002  0.817701  0.000589  0.006515  0.002049     0.002729  0.000381\n",
       "0.001  0.813565  0.000627  0.005730  0.002235     0.002183  0.000498"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Results for attribute-blind setting:\n"
     ]
    },
    {
     "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 tr th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr:last-of-type th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th colspan=\"2\" halign=\"left\">accuracy</th>\n",
       "      <th colspan=\"2\" halign=\"left\">delta_sp</th>\n",
       "      <th colspan=\"2\" halign=\"left\">delta_sp_rms</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>alpha</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>inf</th>\n",
       "      <td>0.863750</td>\n",
       "      <td>0.000425</td>\n",
       "      <td>0.082381</td>\n",
       "      <td>0.000363</td>\n",
       "      <td>0.023045</td>\n",
       "      <td>0.000091</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.080</th>\n",
       "      <td>0.863618</td>\n",
       "      <td>0.000439</td>\n",
       "      <td>0.080081</td>\n",
       "      <td>0.002008</td>\n",
       "      <td>0.022793</td>\n",
       "      <td>0.000259</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.070</th>\n",
       "      <td>0.863149</td>\n",
       "      <td>0.000522</td>\n",
       "      <td>0.070174</td>\n",
       "      <td>0.002632</td>\n",
       "      <td>0.021688</td>\n",
       "      <td>0.000322</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.060</th>\n",
       "      <td>0.862283</td>\n",
       "      <td>0.000688</td>\n",
       "      <td>0.060133</td>\n",
       "      <td>0.003053</td>\n",
       "      <td>0.020640</td>\n",
       "      <td>0.000346</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.050</th>\n",
       "      <td>0.860912</td>\n",
       "      <td>0.000717</td>\n",
       "      <td>0.050166</td>\n",
       "      <td>0.002249</td>\n",
       "      <td>0.019697</td>\n",
       "      <td>0.000265</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.040</th>\n",
       "      <td>0.859265</td>\n",
       "      <td>0.000739</td>\n",
       "      <td>0.040163</td>\n",
       "      <td>0.001894</td>\n",
       "      <td>0.018885</td>\n",
       "      <td>0.000200</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.030</th>\n",
       "      <td>0.856901</td>\n",
       "      <td>0.000992</td>\n",
       "      <td>0.032743</td>\n",
       "      <td>0.000630</td>\n",
       "      <td>0.017445</td>\n",
       "      <td>0.000352</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.020</th>\n",
       "      <td>0.850826</td>\n",
       "      <td>0.001289</td>\n",
       "      <td>0.023499</td>\n",
       "      <td>0.001088</td>\n",
       "      <td>0.013798</td>\n",
       "      <td>0.000491</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.015</th>\n",
       "      <td>0.845166</td>\n",
       "      <td>0.001476</td>\n",
       "      <td>0.019011</td>\n",
       "      <td>0.002379</td>\n",
       "      <td>0.011182</td>\n",
       "      <td>0.000465</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.010</th>\n",
       "      <td>0.836539</td>\n",
       "      <td>0.001668</td>\n",
       "      <td>0.013862</td>\n",
       "      <td>0.002396</td>\n",
       "      <td>0.007977</td>\n",
       "      <td>0.000435</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.005</th>\n",
       "      <td>0.825750</td>\n",
       "      <td>0.001509</td>\n",
       "      <td>0.009081</td>\n",
       "      <td>0.002112</td>\n",
       "      <td>0.004793</td>\n",
       "      <td>0.000346</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.002</th>\n",
       "      <td>0.815739</td>\n",
       "      <td>0.001797</td>\n",
       "      <td>0.006014</td>\n",
       "      <td>0.002288</td>\n",
       "      <td>0.002625</td>\n",
       "      <td>0.000417</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.001</th>\n",
       "      <td>0.811533</td>\n",
       "      <td>0.001255</td>\n",
       "      <td>0.005723</td>\n",
       "      <td>0.002381</td>\n",
       "      <td>0.002116</td>\n",
       "      <td>0.000518</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       accuracy            delta_sp           delta_sp_rms          \n",
       "           mean       std      mean       std         mean       std\n",
       "alpha                                                               \n",
       "inf    0.863750  0.000425  0.082381  0.000363     0.023045  0.000091\n",
       "0.080  0.863618  0.000439  0.080081  0.002008     0.022793  0.000259\n",
       "0.070  0.863149  0.000522  0.070174  0.002632     0.021688  0.000322\n",
       "0.060  0.862283  0.000688  0.060133  0.003053     0.020640  0.000346\n",
       "0.050  0.860912  0.000717  0.050166  0.002249     0.019697  0.000265\n",
       "0.040  0.859265  0.000739  0.040163  0.001894     0.018885  0.000200\n",
       "0.030  0.856901  0.000992  0.032743  0.000630     0.017445  0.000352\n",
       "0.020  0.850826  0.001289  0.023499  0.001088     0.013798  0.000491\n",
       "0.015  0.845166  0.001476  0.019011  0.002379     0.011182  0.000465\n",
       "0.010  0.836539  0.001668  0.013862  0.002396     0.007977  0.000435\n",
       "0.005  0.825750  0.001509  0.009081  0.002112     0.004793  0.000346\n",
       "0.002  0.815739  0.001797  0.006014  0.002288     0.002625  0.000417\n",
       "0.001  0.811533  0.001255  0.005723  0.002381     0.002116  0.000518"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Post-processing for statistical parity\n",
    "\n",
    "df_aware = utils.postprocess_and_evaluate(\n",
    "    [\n",
    "        np.inf, 0.08, 0.07, 0.06, 0.05, 0.04, 0.03, 0.02, 0.015, 0.01, 0.005,\n",
    "        0.002, 0.001\n",
    "    ],\n",
    "    seeds,\n",
    "    'sp',\n",
    "    ['accuracy', 'delta_sp', 'delta_sp_rms'],\n",
    "    **postprocess_kwargs,\n",
    ")\n",
    "df_blind = utils.postprocess_and_evaluate(\n",
    "    [\n",
    "        np.inf, 0.08, 0.07, 0.06, 0.05, 0.04, 0.03, 0.02, 0.015, 0.01, 0.005,\n",
    "        0.002, 0.001\n",
    "    ],\n",
    "    seeds,\n",
    "    'sp',\n",
    "    ['accuracy', 'delta_sp', 'delta_sp_rms'],\n",
    "    **postprocess_u_kwargs,\n",
    ")\n",
    "\n",
    "fig, ax = plt.subplots(1, 1, figsize=(3.5, 3))\n",
    "utils.plot_results(ax, df_aware, 'delta_sp', 'accuracy', label='aware')\n",
    "utils.plot_results(ax, df_blind, 'delta_sp', 'accuracy', label='unaware')\n",
    "ax.set_xlabel(\"Violation of statistical parity\")\n",
    "ax.set_ylabel(\"Accuracy\")\n",
    "ax.set_title(\"BiasBios (BERT)\")\n",
    "ax.legend()\n",
    "plt.show()\n",
    "\n",
    "print(\"Results for attribute-aware setting:\")\n",
    "display(df_aware)\n",
    "print(\"Results for attribute-blind setting:\")\n",
    "display(df_blind)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "e650eabd",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 350x300 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Results for attribute-aware setting:\n"
     ]
    },
    {
     "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 tr th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr:last-of-type th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th colspan=\"2\" halign=\"left\">accuracy</th>\n",
       "      <th colspan=\"2\" halign=\"left\">delta_eopp</th>\n",
       "      <th colspan=\"2\" halign=\"left\">delta_eopp_rms</th>\n",
       "      <th colspan=\"2\" halign=\"left\">delta_eo</th>\n",
       "      <th colspan=\"2\" halign=\"left\">delta_eo_rms</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>alpha</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>inf</th>\n",
       "      <td>0.864159</td>\n",
       "      <td>0.000346</td>\n",
       "      <td>0.406091</td>\n",
       "      <td>0.015177</td>\n",
       "      <td>0.122912</td>\n",
       "      <td>0.003319</td>\n",
       "      <td>0.406091</td>\n",
       "      <td>0.015177</td>\n",
       "      <td>0.028399</td>\n",
       "      <td>0.000665</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.250</th>\n",
       "      <td>0.864132</td>\n",
       "      <td>0.000377</td>\n",
       "      <td>0.375186</td>\n",
       "      <td>0.049533</td>\n",
       "      <td>0.119442</td>\n",
       "      <td>0.005083</td>\n",
       "      <td>0.375186</td>\n",
       "      <td>0.049533</td>\n",
       "      <td>0.027784</td>\n",
       "      <td>0.001010</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.200</th>\n",
       "      <td>0.864035</td>\n",
       "      <td>0.000485</td>\n",
       "      <td>0.335496</td>\n",
       "      <td>0.062597</td>\n",
       "      <td>0.113875</td>\n",
       "      <td>0.006392</td>\n",
       "      <td>0.335496</td>\n",
       "      <td>0.062597</td>\n",
       "      <td>0.026760</td>\n",
       "      <td>0.001274</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.150</th>\n",
       "      <td>0.863925</td>\n",
       "      <td>0.000544</td>\n",
       "      <td>0.278861</td>\n",
       "      <td>0.044036</td>\n",
       "      <td>0.102770</td>\n",
       "      <td>0.005066</td>\n",
       "      <td>0.278861</td>\n",
       "      <td>0.044036</td>\n",
       "      <td>0.024497</td>\n",
       "      <td>0.000919</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.100</th>\n",
       "      <td>0.863357</td>\n",
       "      <td>0.000601</td>\n",
       "      <td>0.221248</td>\n",
       "      <td>0.038116</td>\n",
       "      <td>0.089922</td>\n",
       "      <td>0.003718</td>\n",
       "      <td>0.221248</td>\n",
       "      <td>0.038116</td>\n",
       "      <td>0.022087</td>\n",
       "      <td>0.000749</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.050</th>\n",
       "      <td>0.862371</td>\n",
       "      <td>0.000686</td>\n",
       "      <td>0.175577</td>\n",
       "      <td>0.017705</td>\n",
       "      <td>0.073804</td>\n",
       "      <td>0.003400</td>\n",
       "      <td>0.175577</td>\n",
       "      <td>0.017705</td>\n",
       "      <td>0.018878</td>\n",
       "      <td>0.000525</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.010</th>\n",
       "      <td>0.860422</td>\n",
       "      <td>0.000707</td>\n",
       "      <td>0.172309</td>\n",
       "      <td>0.045785</td>\n",
       "      <td>0.063234</td>\n",
       "      <td>0.007231</td>\n",
       "      <td>0.172309</td>\n",
       "      <td>0.045785</td>\n",
       "      <td>0.017132</td>\n",
       "      <td>0.001266</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.001</th>\n",
       "      <td>0.858011</td>\n",
       "      <td>0.000936</td>\n",
       "      <td>0.197289</td>\n",
       "      <td>0.084379</td>\n",
       "      <td>0.065225</td>\n",
       "      <td>0.010897</td>\n",
       "      <td>0.205250</td>\n",
       "      <td>0.076248</td>\n",
       "      <td>0.019960</td>\n",
       "      <td>0.001670</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       accuracy           delta_eopp           delta_eopp_rms            \\\n",
       "           mean       std       mean       std           mean       std   \n",
       "alpha                                                                     \n",
       "inf    0.864159  0.000346   0.406091  0.015177       0.122912  0.003319   \n",
       "0.250  0.864132  0.000377   0.375186  0.049533       0.119442  0.005083   \n",
       "0.200  0.864035  0.000485   0.335496  0.062597       0.113875  0.006392   \n",
       "0.150  0.863925  0.000544   0.278861  0.044036       0.102770  0.005066   \n",
       "0.100  0.863357  0.000601   0.221248  0.038116       0.089922  0.003718   \n",
       "0.050  0.862371  0.000686   0.175577  0.017705       0.073804  0.003400   \n",
       "0.010  0.860422  0.000707   0.172309  0.045785       0.063234  0.007231   \n",
       "0.001  0.858011  0.000936   0.197289  0.084379       0.065225  0.010897   \n",
       "\n",
       "       delta_eo           delta_eo_rms            \n",
       "           mean       std         mean       std  \n",
       "alpha                                             \n",
       "inf    0.406091  0.015177     0.028399  0.000665  \n",
       "0.250  0.375186  0.049533     0.027784  0.001010  \n",
       "0.200  0.335496  0.062597     0.026760  0.001274  \n",
       "0.150  0.278861  0.044036     0.024497  0.000919  \n",
       "0.100  0.221248  0.038116     0.022087  0.000749  \n",
       "0.050  0.175577  0.017705     0.018878  0.000525  \n",
       "0.010  0.172309  0.045785     0.017132  0.001266  \n",
       "0.001  0.205250  0.076248     0.019960  0.001670  "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Results for attribute-blind setting:\n"
     ]
    },
    {
     "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 tr th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr:last-of-type th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th colspan=\"2\" halign=\"left\">accuracy</th>\n",
       "      <th colspan=\"2\" halign=\"left\">delta_eopp</th>\n",
       "      <th colspan=\"2\" halign=\"left\">delta_eopp_rms</th>\n",
       "      <th colspan=\"2\" halign=\"left\">delta_eo</th>\n",
       "      <th colspan=\"2\" halign=\"left\">delta_eo_rms</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>alpha</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>inf</th>\n",
       "      <td>0.863750</td>\n",
       "      <td>0.000425</td>\n",
       "      <td>0.403001</td>\n",
       "      <td>0.016850</td>\n",
       "      <td>0.132246</td>\n",
       "      <td>0.004998</td>\n",
       "      <td>0.403001</td>\n",
       "      <td>0.016850</td>\n",
       "      <td>0.030981</td>\n",
       "      <td>0.001145</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.300</th>\n",
       "      <td>0.863567</td>\n",
       "      <td>0.000470</td>\n",
       "      <td>0.385960</td>\n",
       "      <td>0.033164</td>\n",
       "      <td>0.127737</td>\n",
       "      <td>0.006458</td>\n",
       "      <td>0.385960</td>\n",
       "      <td>0.033164</td>\n",
       "      <td>0.029805</td>\n",
       "      <td>0.001612</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.250</th>\n",
       "      <td>0.863361</td>\n",
       "      <td>0.000546</td>\n",
       "      <td>0.348455</td>\n",
       "      <td>0.042034</td>\n",
       "      <td>0.120953</td>\n",
       "      <td>0.006435</td>\n",
       "      <td>0.349223</td>\n",
       "      <td>0.040509</td>\n",
       "      <td>0.028430</td>\n",
       "      <td>0.001496</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.200</th>\n",
       "      <td>0.863045</td>\n",
       "      <td>0.000652</td>\n",
       "      <td>0.288043</td>\n",
       "      <td>0.039708</td>\n",
       "      <td>0.104640</td>\n",
       "      <td>0.004298</td>\n",
       "      <td>0.290499</td>\n",
       "      <td>0.035483</td>\n",
       "      <td>0.025248</td>\n",
       "      <td>0.000788</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.150</th>\n",
       "      <td>0.862473</td>\n",
       "      <td>0.000730</td>\n",
       "      <td>0.232388</td>\n",
       "      <td>0.039584</td>\n",
       "      <td>0.087793</td>\n",
       "      <td>0.006314</td>\n",
       "      <td>0.232388</td>\n",
       "      <td>0.039584</td>\n",
       "      <td>0.021930</td>\n",
       "      <td>0.001090</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.100</th>\n",
       "      <td>0.861508</td>\n",
       "      <td>0.000776</td>\n",
       "      <td>0.193961</td>\n",
       "      <td>0.036232</td>\n",
       "      <td>0.077909</td>\n",
       "      <td>0.006510</td>\n",
       "      <td>0.195860</td>\n",
       "      <td>0.032922</td>\n",
       "      <td>0.020199</td>\n",
       "      <td>0.001039</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.050</th>\n",
       "      <td>0.859807</td>\n",
       "      <td>0.000824</td>\n",
       "      <td>0.196765</td>\n",
       "      <td>0.071631</td>\n",
       "      <td>0.069040</td>\n",
       "      <td>0.012090</td>\n",
       "      <td>0.196765</td>\n",
       "      <td>0.071631</td>\n",
       "      <td>0.018416</td>\n",
       "      <td>0.002088</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.010</th>\n",
       "      <td>0.857203</td>\n",
       "      <td>0.000886</td>\n",
       "      <td>0.227392</td>\n",
       "      <td>0.099461</td>\n",
       "      <td>0.071207</td>\n",
       "      <td>0.020847</td>\n",
       "      <td>0.227392</td>\n",
       "      <td>0.099461</td>\n",
       "      <td>0.019067</td>\n",
       "      <td>0.003737</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.001</th>\n",
       "      <td>0.854478</td>\n",
       "      <td>0.001108</td>\n",
       "      <td>0.212924</td>\n",
       "      <td>0.080115</td>\n",
       "      <td>0.071500</td>\n",
       "      <td>0.018936</td>\n",
       "      <td>0.212924</td>\n",
       "      <td>0.080115</td>\n",
       "      <td>0.020160</td>\n",
       "      <td>0.003465</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       accuracy           delta_eopp           delta_eopp_rms            \\\n",
       "           mean       std       mean       std           mean       std   \n",
       "alpha                                                                     \n",
       "inf    0.863750  0.000425   0.403001  0.016850       0.132246  0.004998   \n",
       "0.300  0.863567  0.000470   0.385960  0.033164       0.127737  0.006458   \n",
       "0.250  0.863361  0.000546   0.348455  0.042034       0.120953  0.006435   \n",
       "0.200  0.863045  0.000652   0.288043  0.039708       0.104640  0.004298   \n",
       "0.150  0.862473  0.000730   0.232388  0.039584       0.087793  0.006314   \n",
       "0.100  0.861508  0.000776   0.193961  0.036232       0.077909  0.006510   \n",
       "0.050  0.859807  0.000824   0.196765  0.071631       0.069040  0.012090   \n",
       "0.010  0.857203  0.000886   0.227392  0.099461       0.071207  0.020847   \n",
       "0.001  0.854478  0.001108   0.212924  0.080115       0.071500  0.018936   \n",
       "\n",
       "       delta_eo           delta_eo_rms            \n",
       "           mean       std         mean       std  \n",
       "alpha                                             \n",
       "inf    0.403001  0.016850     0.030981  0.001145  \n",
       "0.300  0.385960  0.033164     0.029805  0.001612  \n",
       "0.250  0.349223  0.040509     0.028430  0.001496  \n",
       "0.200  0.290499  0.035483     0.025248  0.000788  \n",
       "0.150  0.232388  0.039584     0.021930  0.001090  \n",
       "0.100  0.195860  0.032922     0.020199  0.001039  \n",
       "0.050  0.196765  0.071631     0.018416  0.002088  \n",
       "0.010  0.227392  0.099461     0.019067  0.003737  \n",
       "0.001  0.212924  0.080115     0.020160  0.003465  "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Post-processing for equal opportunity\n",
    "\n",
    "df_aware = utils.postprocess_and_evaluate(\n",
    "    [np.inf, 0.25, 0.2, 0.15, 0.1, 0.05, 0.01, 0.001],\n",
    "    seeds,\n",
    "    'eopp',\n",
    "    ['accuracy', 'delta_eopp', 'delta_eopp_rms', 'delta_eo', 'delta_eo_rms'],\n",
    "    **postprocess_kwargs,\n",
    ")\n",
    "df_blind = utils.postprocess_and_evaluate(\n",
    "    [np.inf, 0.3, 0.25, 0.2, 0.15, 0.1, 0.05, 0.01, 0.001],\n",
    "    seeds,\n",
    "    'eopp',\n",
    "    ['accuracy', 'delta_eopp', 'delta_eopp_rms', 'delta_eo', 'delta_eo_rms'],\n",
    "    **postprocess_u_kwargs,\n",
    ")\n",
    "\n",
    "fig, ax = plt.subplots(1, 1, figsize=(3.5, 3))\n",
    "utils.plot_results(ax, df_aware, 'delta_eopp_rms', 'accuracy', label='aware')\n",
    "utils.plot_results(ax,\n",
    "                   df_blind,\n",
    "                   'delta_eopp_rms',\n",
    "                   'accuracy',\n",
    "                   label='unaware')\n",
    "ax.set_xlabel(\"Violation of equal opportunity (rms)\")\n",
    "ax.set_ylabel(\"Accuracy\")\n",
    "ax.set_title(\"BiasBios (BERT)\")\n",
    "ax.legend()\n",
    "plt.show()\n",
    "\n",
    "print(\"Results for attribute-aware setting:\")\n",
    "display(df_aware)\n",
    "print(\"Results for attribute-blind setting:\")\n",
    "display(df_blind)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "c68a6673",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXEAAAE6CAYAAADgGGILAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjcuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/bCgiHAAAACXBIWXMAAA9hAAAPYQGoP6dpAABmqUlEQVR4nO3deVxUVf8H8M/MAMMOyjoIAu4oigrKJqIWKBqmVlKWimGJUoioqfEkuCQuaW5JWe7LI+kvyydJIc0tshQ1F8xdcRlEUBgQgVnO749pRoYZYAYHB/T7fr2mnHvPPffcA3znzLnnnsNhjDEQQghplriGLgAhhJCGoyBOCCHNGAVxQghpxiiIE0JIM0ZBnBBCmjEK4oQQ0oxRECeEkGaMgjghhDRjFMQJIaQZoyBOCCHNGAVxYjAbN24Eh8NReTk4OKBfv374+eef1dJzOBykpKQ0SlkOHTqkVpYWLVrA398fmzZtUkvv4eGB6OjoRilLXa5duwY+n48//vhDuS06Olql3DweD66urhg5ciTOnz+vcrym66z+2rhxozJtv379VPaZmpqic+fOmD9/PqqqqgCgzryqvw4dOoTLly/DxMQEp06dei519bIwMnQBCNmwYQM6deoExhjy8/OxevVqREZGYs+ePYiMjFSm++OPP+Dq6tqoZVmwYAH69+8PACgsLMTmzZsRHR0NkUiEjz/+WJlu9+7dsLa2btSyaDJt2jSEhYUhMDBQZbuZmRkOHjwIAJBIJLh69Srmz5+PoKAgXLx4Ea1atVJJX/06q2vbtq3K+zZt2mDbtm0AgAcPHuC7777DZ599hry8PKxdu1blwwQA5s2bh99++01ZFoXOnTvD2toa7777LqZMmYLDhw83rAKIOkaIgWzYsIEBYCdOnFDZXl5ezvh8PnvnnXeeW1l+++03BoDt3LlTZbtUKmUeHh4sMDDwuZWlNrm5uQwA27dvn8r2sWPHMgsLC7X0Bw4cYADYN998o9xW23VqEhoayrp06aKyTSwWs/bt2zMTExP25MkTtWNqK4vCyZMnGQD2+++/13t+oh3qTiFNjqmpKUxMTGBsbKyyvWZ3yoMHDzBp0iR07twZlpaWcHR0xIABA3D06FG1PNPS0uDj4wNLS0tYWVmhU6dO+PTTT+stC5fLhaWlpVpZNHWn5OXl4b333oOjoyP4fD68vLywdOlSyGQyvZQlLS0Nzs7OCAsLqzctANjY2ACAWtmfhZGREbp3746qqioUFxfrfLyvry+8vLzw9ddf661MLzvqTiEGJ5VKIZFIwBjD/fv3sWTJEjx+/BijRo2q87iHDx8CAJKTk+Hs7IyysjLs3r0b/fr1w4EDB9CvXz8AwI4dOzBp0iR8/PHH+OKLL8DlcnH16lXk5uaq5SmTySCRSAAARUVF2LBhA86fP4+1a9fWWZYHDx4gKCgIVVVVmDdvHjw8PPDzzz9j2rRpuHbtGtasWaNzWWrau3cv+vbtCy5Xc9tLUW5Fd8r06dPRokULDBkypM7rrM7IqP6QcOPGDdja2sLBwaHetJr069cPO3fuBGMMHA6nQXmQagz9VYC8vBTdKTVffD6frVmzRi09AJacnFxrfhKJhInFYvbKK6+w4cOHK7d/9NFHzNbWts6yKLoZar64XC5LSkpSS+/u7s7Gjh2rfD9z5kwGgP35558q6SZOnMg4HA67dOmS1mXR5P79+wwAW7hwodq+sWPHaiy7QCBgx44d0+o6Fa/bt28r0yq6U8RiMROLxUwoFLLZs2czAOzrr7/WWM76ulMYY+zbb79lANjFixd1rgeijlrixOA2b94MLy8vAPKbibt370ZcXBykUik++uijOo/9+uuvsXbtWuTm5qKyslK5vVOnTsp/9+7dG6tXr8Y777yDt99+G8HBwbC3t9eY36JFizBgwAAAQHFxMQ4cOICFCxeisrISS5YsqbUcBw8eROfOndG7d2+V7dHR0UhLS8PBgwfRoUMHncpS3b179wAAjo6OGvebmZnhyJEjAOSt7Lt372LFihUYPHgw9u3bp3YjtPp1Vufk5KTy/sKFC2rdMbNmzcKECRPqLXNtFNdw9+5dlZ8TaRgK4sTgvLy84Ofnp3w/aNAg3Lp1C5988gnee+892Nraajxu2bJlmDp1KmJjYzFv3jzY29uDx+Phs88+w8WLF5XpRo8eDYlEgm+//RZvvPEGZDIZevXqhfnz56v1L7dp00alLK+++ioePXqEpUuXIiYmptagU1RUBA8PD7XtLi4uyv26lqW6J0+eAJDfL9CEy+WqlBsABg4cCDc3NyQmJqqNIql5nbVp27YtduzYAcYYbt26hfnz5yM1NRXdunXD22+/Xe/xmiiuQXFN5NnQjU3SJHXr1g1PnjzB5cuXa02zdetW9OvXD2lpaRgyZAj8/f3h5+eH0tJStbTjxo1DdnY2SkpKsHfvXjDG8Nprr+HWrVtalYUxhrNnz9aaxs7ODkKhUG27ogVdvbXdkLIojlfcB9CGubk52rZti7///lvrY2oyNTWFn58fevXqhTfffBMHDhyAk5MTEhISUFZW1qA8FdegzTcQUj8K4qRJOnPmDADUefOMw+GAz+erbDt79qxaq7M6CwsLREREICkpCVVVVbhw4YLWZamtKwMAXnnlFeTm5qo9yLJ582ZwOByNY7J1KYu7uzvMzMxw7dq1esurUFZWhqtXr9ZZbl3Z2dlh4cKFuH//PlatWtWgPK5fvw4ul4uOHTvqrVwvM+pOIQZ3/vx5lREhP/zwA7KysjB8+HB4enrWetxrr72GefPmITk5GaGhobh06RLmzp0LT09PlZEXH3zwAczMzBAcHAyBQID8/HykpqbCxsYGvXr1UsnzypUrOH78OACgpKQEv/76K9atWwc/Pz+EhITUWpYpU6Zg8+bNGDJkCObOnQt3d3fs3bsXa9aswcSJE9GhQwedy1KdiYkJAgMDlWWrSSaTKfcp+sRXrlyJR48eaXzKtfp1Vufq6lrvA1VjxozBsmXL8MUXXyAuLk7nh56OHz+O7t27o0WLFjodR2ph2Puq5GWmaXSKjY0N6969O1u2bBmrqKhQSY8ao1MqKyvZtGnTWKtWrZipqSnr2bMn+/HHH9nYsWOZu7u7Mt2mTZtY//79mZOTEzMxMWEuLi5s5MiR7OzZs8o0mkZtWFhYsM6dO7Pk5GRWUlKiUpaao1MYY+zWrVts1KhRzM7OjhkbG7OOHTuyJUuWMKlUqlNZarNu3TrG4/HYvXv3VLZrGp3i6OjIQkND2e7du1XS1jc6pfpIHE0P+yjs3buXAWBz5sxRK0tdo1NKS0uZubk5W7p0ab3XS7TDYYyx5/qpQQhpkIqKCrRu3RpTp07FjBkzDF2cBlm3bh0mT56M27dvU0tcT6hPnJBmwtTUFHPmzMGyZcvw+PFjQxdHZxKJBIsWLcKsWbMogOsR9YkT0ox8+OGHKC4uxvXr19G1a1dDF0cnt2/fxnvvvYepU6cauigvFOpOIYSQZoy6UwghpBmjIE4IIc0Y9Yk3IplMhnv37sHKyopmayOEaI0xhtLSUri4uNQ6a2X1xAb11VdfMQ8PD8bn81nPnj3ZkSNH6ky/detW1q1bN2ZmZsacnZ1ZdHQ0KywsVEnz6NEjNmnSJObs7Mz4fD7r1KkT27t3r3J/cnKy2vhYJycnlTxkMhlLTk5mAoGAmZqastDQUHb+/Hmdru327dt1jsmlF73oRa+6XtVnlayNQVvi6enpSEhIwJo1axAcHIxvvvkGERERyM3NRevWrdXSHzt2DGPGjMGXX36JyMhI3L17F7GxsRg/fjx2794NAKiqqkJYWBgcHR2xa9cuuLq64vbt27CyslLJq0uXLvj111+V73k8nsr+xYsXY9myZdi4cSM6dOignKDo0qVLannVRpHu9u3bBlnKqzkRi8XIzMxEeHi4XhcxeNlQPeqHoetRJBLBzc1Nu1ijU9NSz3r37s1iY2NVtnXq1InNnDlTY/olS5awNm3aqGxbuXIlc3V1Vb5PS0tjbdq0YVVVVbWeNzk5mfn4+NS6XyaTMWdnZ5W5mysqKpiNjU2t8yhrUlJSwgCoPe1H1FVVVbEff/yxzp8bqR/Vo34Yuh51iR0Ga4lXVVUhJycHM2fOVNkeHh6O7OxsjccEBQUhKSkJGRkZiIiIQEFBAXbt2qWycsmePXsQGBiIuLg4/PTTT3BwcMCoUaMwY8YMldb2lStX4OLiAj6fD39/fyxYsABt2rQBIF+5JD8/H+Hh4cr0fD4foaGhyM7OrnUu5crKSpU5rUUiEQD5p7pYLNaxhl4uivqheno2VI/6Yeh61OW8BgvihYWFkEqlapPQOzk5IT8/X+MxQUFB2LZtG6KiolBRUQGJRIKhQ4eqzKZ2/fp1HDx4EO+++y4yMjJw5coVxMXFQSKRYPbs2QAAf39/bN68GR06dMD9+/eVq4JfuHABdnZ2yvNrKltd04WmpqZizpw5atszMzNhbm6uXcW85LKysgxdhBcC1aN+GKoey8vLtU5r8NEpNUdtsDrW3cvNzUV8fDxmz56NgQMHQigUYvr06YiNjcW6desAyEeEODo6Yu3ateDxePD19cW9e/ewZMkSZRCPiIhQ5tm1a1cEBgaibdu22LRpExITExtUNkC+4kn14xX9WuHh4dQnXg+xWIysrCyEhYVRX+4zoHrUD0PXo+JbvDYMFsQVq7DUbHUXFBSotYAVUlNTERwcjOnTpwOQT9ZvYWGBkJAQzJ8/HwKBAAKBAMbGxipdJ15eXsjPz0dVVRVMTEzU8rWwsEDXrl1x5coVAICzszMAID8/HwKBQKuyAfIul5rzWwPy1cbpD0o7VFf6QfWoH4aqR13OabCHfUxMTODr66v2dSUrKwtBQUEajykvL1cbM6kI1uzf2QOCg4Nx9epVyGQyZZrLly9DIBBoDOCAvC/74sWLyoDt6ekJZ2dnlbJVVVXh8OHDtZaNEEIMwaBPbCYmJuK7777D+vXrcfHiRUyZMgV5eXmIjY0FIO+eGDNmjDJ9ZGQkfvjhB6SlpeH69ev4/fffER8fj969eyvXMpw4cSKKioowefJkXL58GXv37sWCBQsQFxenzGfatGk4fPgwbty4gT///BNvvvkmRCIRxo4dC0DejZKQkIAFCxZg9+7dOH/+PKKjo2Fubo5Ro0Y9xxoihJC6GbRPPCoqCkVFRZg7dy6EQiG8vb2RkZEBd3d3AIBQKEReXp4yfXR0NEpLS7F69WpMnToVtra2GDBgABYtWqRM4+bmhszMTEyZMgXdunVDq1atMHnyZJX5l+/cuYN33nkHhYWFcHBwQEBAAI4fP648LwB88sknePLkCSZNmoRHjx7B398fmZmZWo8RJ4SQ54FmMWxEIpEINjY2KCkpoRub9RCLxcjIyMDgwYOpL/cZPO96lEhl+P7kHQDASD9XGPGe75f7SrEUS/ZfQmmFBJE+AgS2tQeP++xTXNSsx8Y6T210iR0GH51CyMvO0IHwWUhlDNcflKG0QoLWLc0aPbhVt++8ECl7cpEvqgAApJ+8DYGNKZIjO2OQt6Ceo5veeRqq+fy2EEKalH3nhQhdcgjfHbuB9JO38d66v9Bn0UHsOy98LueeuPWUMrAq5JdUYOLWU3orw/4L95/LeZ4FtcQJMTBDtma1JZHKVN7vv5CPj7afRs2+WEVwWz2qBwZ2cW6UskhlDCl7ctXODchnjeIAmPO/XPTv6FhnPda8purEYgkqpMCyvf/UeZ6UPbkIaVf7z6v6t6rG+oZFQZwQA5FIZdh/IR/zfs5Fvkg+XUP6ydtwtjbFZ695NSgISqQySJn8/xxu7UFKV//967by3zLGsDTzUq3BDQA+3X0ehWVV4DbCFMw3CsvUWsY1yyAsqcCiff/A096y1nQnbj7UuJ3L4UAmk+Gfm1zcL63UmEZxnnxRBT7ccgoOVurPhwBAb8+Wyn+P8lef1E8fKIgTYiCzf7qA7X/lqW3PF1UgbvtpjOrdGt6tbHTKUyqV4nwBB49P3VObmfNZ5Nx6pPz3g9JKiCokdaYveSJG5oX7tQa3Z3HnkXaPpJ+9U4KHj2ufgySvSHM+HvYWAAAmlSCAexWOKEYBbPGXrBNkGnqgKyVSrcrTWCiIk5dWXV+nG5tUxvDz2Xt1ptl7ToiOzlY6tWalMgYZk/8fnGcbeFa9i8DP3Vb577/vlGh1vIutKXxcdfsQ0kZLC2OcrPahUpturjZoY28BjkyCtnd/BABcazUMjCsPez1aay5blJ8b2IUfIb42DdYmT89zj7XEHPEY7Jf1Vkn/8YB28K/W4q7uedykpiBOXggNGeGhSG8I1x+UadWavVX0GG3+bRlqg/NvhwYHTPnvhhrp56Zxu4d9EXbm1F93r3d3QUAbu2cqgyZSGUPGuXzcF1VovEIOAGcbU8wY1En+QSSpBKf0ITgVIvRyug3m3gfg1v4txejSz2A/jYdpjdyd8RBpxssxUZyA/bLeyvP0ae9g0HsYFMQJMYDSegK4wuNKaf3Lc1XDGAOXA3C5XJ2O06S2D8LAtvYQ2Jgiv6TuIKp2g1YqAU5vkf+7x2iA17DwY8QDUoZ2xsStp8ABVMqgOFtyZGfwjXlA7h7glxlA6b/fes5sAaxcgIELAK9I9cxlUnl6MNQMy1wOIGNAsvEW/FrpBxm4SB7SCTwmBTT1qDTw+nRFQZy8tEb6uRrs3K1bmiH95O1600X6CHRqzYrFYlgUnEVET5dGe9iHx+UgObL+INqYrdNB3gKkvddTZfw2IP/wUI7fzt0DfD+mRgkhD+i7ogHfcYDAR3Vf4RWg9J5aAFfgcgAXFGGgaS6G9nDDoKpM4HQtif3GNfDqdENBnLwQGjJMz5AP1TS4NVsPJuOCx5FfW2Nen1ZBtCaZVB4kK0VAC0/AMwR4hgfGB3k5IKxjX/x18xEKSivhaMVHb48W8voSVypb1LW6sBtw7AxwqtVTRbFW517tcwtcgR1Q222VOrpr9I2COGn2mvoTdZo0qDWrp+4IfRnkLUD/jo7aPY5es1vj9BbA2gVoP1C9NawDHoBAxZsSAH//++9/W9R1qigGHt0A7No93Waq3Y1YrpkNUNcN5x6jtcpHH+iJTdKsPa8n9xqDojXrZG2qst3ZxhRp7/Vssh9A1fGNefjPa52x6M1utd/gU3Rr1AyqIiGQswEQ/q1+zLOq1HJRhcpSeUtc8bJrB5ja1n1L2NRWnq76cTVfz/EDllriRO8aMnSvIQ+p6OvJPX1pSPeFTq1ZfZBqd0NVb6rdKFT370/pyn4g/HP9dkHYtn76raUu3m8CHn1Ut5m1BHaNA1O7ufnvu9eWa74paiAUxIneNWToXkMeUrn+QLsn95bsv4Q2DrU/uacvDX0iT9GarZemPmVdA582gU2f6u3WYIDoHvDb54B9e/2dl8nkLeZa+7g58u4cTXXoPRxSBoj3TIGZuNp4dGsXYNBCoPNQ/ZVTDyiIk2ZL22F62qZr0mrrUx60qMkFFRVad2tov6akVjhcoMtweXdNbQYtrPVDkHV6DZnXgCHetjB6UgRYOgHuQc/1hqW2KIgTvWvI0L2GDI1rrGF6TU5tQ+VEQvn2kZu1D+TP8YYbAPk3hoZ2azyrHqMBz1Bg/yygtNq9EetW2rWoOVz5g0FNfH57CuJE7xrSN9yQoXE6D9NrjNEd1fuYG+Mxfm36lPfNlI/y4PIAmQQcJgVkEkCqx371htaVZ4j8G4NICM3XUEe3hj54Dwc6DgYOzJG39r3fbLxzGQgFcfJMDLmgQVN46KTR+5i16lO+Kw9S9u3BkUrRuug8OGceAnqcAKvBD65wefIun+/HALX9lOro1tALYz4waEHj5W9gNMSQNGsvwjC9OhmqT1mfOg+Vd/lY1fhZWLvo1hVENKKWOGn2nvswveoau49Zxz5lJhEjT/gLunSPAIyaUF9u56HyLp8XuFvDUCiIkxcCj8tRDiMMaGOnOYDrY4ie2okb+U9I1z5lGQPj8ACukcGf6FTD5T0dRuhR90yCRHsG705Zs2YNPD09YWpqCl9fXxw9erTO9Nu2bYOPjw/Mzc0hEAgwbtw4FBUVqaQpLi5GXFwcBAIBTE1N4eXlhYyMDOX+1NRU9OrVC1ZWVnB0dMSwYcNw6dIllTyio6PB4XBUXgEBAfq78BeEYs6Sv28X4/j1Ivk81gZgxONilH9rjPJvrblfPncPsLI7cPwrect2y+vAcm/59qZM0acMAGrTMj2nPmV94RnJ+9b9xjW9D5hmzKBBPD09HQkJCUhKSsLp06cREhKCiIgI5OWpr3YCAMeOHcOYMWMQExODCxcuYOfOnThx4gTGjx+vTFNVVYWwsDDcvHkTu3btwqVLl/Dtt9+iVatWyjSHDx9GXFwcjh8/jqysLEgkEoSHh+Px48cq5xs0aBCEQqHyVf2DgBh2oVyd1PXY9/djmn4gpz5lUgeDfhwuW7YMMTExyiC8fPly7N+/H2lpaUhNTVVLf/z4cXh4eCA+Ph4A4OnpiQkTJmDx4sXKNOvXr8fDhw+RnZ2tHG/s7u6uks++fftU3m/YsAGOjo7IyclB3759ldv5fD6cnbVf57CyshKVlU/X5BOJ5DebxGIxxOLal4lqjvZfuI+Pd/xd60K5q972wcAuTpoPlqk/fCMWi8FhUogrn2jc32AyKYx++QSa5oeG4sHqX2ZA4jng+bZmuTr+6bWPADwHgPvbfKBSBNZlxNPFDar9bil+z16037fnzdD1qMt5DRbEq6qqkJOTg5kzZ6psDw8PR3Z2tsZjgoKCkJSUhIyMDERERKCgoAC7du3CkCFDlGn27NmDwMBAxMXF4aeffoKDgwNGjRqFGTNm1Po4d0mJfLmpli1Vl1g6dOgQHB0dYWtri9DQUHz++edwdHSs9ZpSU1MxZ84cte2ZmZkwNzev9bimQKpDL4iMAfNO85RryFTH/v3vf344g4obUmjqmvYo/E3lvSJNawCX049oXxAtWFYI0a609m8GHDCg9B5urY9BmenzG8lyy76/zsdwmBStiyoAmCAvVwR2cX+tabOysp6hdETBUPVYXq7dOqKAAYN4YWEhpFIpnJxUW2tOTk7Iz8/XeExQUBC2bduGqKgoVFRUQCKRYOjQoVi1apUyzfXr13Hw4EG8++67yMjIwJUrVxAXFweJRILZs2er5ckYQ2JiIvr06QNvb2/l9oiICLz11ltwd3fHjRs38Nlnn2HAgAHIyckBn6958ddZs2YhMTFR+V4kEsHNzQ3h4eGwtrbWqX6etx0ntJ/v5EbhY5RU3aojBQfFVUAury08NSwtVlV1WeW9n3sLSKVSXLx4EV5eXnpd4Jdzrwp4UH86Txc7MBfv+hPqSZeeg3U/SCaRj/8G5KNPNLTmxWIxsrKyEBYW1miLQrwMDF2Pim/x2jD43QVOjTl5GWNq2xRyc3MRHx+P2bNnY+DAgRAKhZg+fTpiY2Oxbt06AIBMJoOjoyPWrl0LHo8HX19f3Lt3D0uWLNEYxD/66COcPXsWx44dU9keFRWl/Le3tzf8/Pzg7u6OvXv3YsSIERrLx+fzNQZ4Y2PjJv8HpUvgfFyl3ZOJj6tkGvO94faGyvsAv9byoXEFv6CLbwSM9Dk07mY74O9t9SbjdRup/8e+6zxhQ/70jAH/8fUnQ/P4nWsODFWPupzTYEHc3t4ePB5PrdVdUFCg1jpXSE1NRXBwMKZPnw4A6NatGywsLBASEoL58+dDIBBAIBDA2NhYJXh4eXkhPz8fVVVVMDExUW7/+OOPsWfPHhw5cgSurnXP9yEQCODu7o4rV6409JKbNF3mO9H7nCU8buMNjTP0Y9+ENDKDjU4xMTGBr6+vWp9TVlYWgoKCNB5TXl6utvirIlizf5d5Cg4OxtWrVyGTPW0tXr58GQKBQBnAGWP46KOP8MMPP+DgwYPw9PSst7xFRUW4ffs2BIJm/gRgLRRzlmjzUsxZUtujNBwAgn/nLNEmv0b1Ig3RI0QDgw4xTExMxHfffYf169fj4sWLmDJlCvLy8hAbGwtA3sc8ZswYZfrIyEj88MMPSEtLw/Xr1/H7778jPj4evXv3houLCwBg4sSJKCoqwuTJk3H58mXs3bsXCxYsQFxcnDKfuLg4bN26Fdu3b4eVlRXy8/ORn5+PJ0+eAADKysowbdo0/PHHH7h58yYOHTqEyMhI2NvbY/jw4c+xhpomxZwlQK1hseFzlsgkwMkN8pe+FjCgIXrkBWbQPvGoqCgUFRVh7ty5EAqF8Pb2RkZGhnJIoFAoVBkzHh0djdLSUqxevRpTp06Fra0tBgwYgEWLFinTuLm5ITMzE1OmTEG3bt3QqlUrTJ48GTNmzFCmSUtLAwD069dPpTwbNmxAdHQ0eDwezp07h82bN6O4uBgCgQD9+/dHeno6rKysGrFGmo8GLZRrSPTYN3lBcRh7huWmSZ1EIhFsbGxQUlLS5EenNFSlWKqXOUvEYjEyMjIweFA4jM/tkG/U92LATWyh4cagrMfBg+nG5jMwdD3qEjtevN9i8lxpNWeJLhpjfhNCXmAUxEmTISg+AaM1M56uwtJcliAjxIAoiJMmgfPPz+h1Y5X6joYsQVYbxQRMhLxAKIiTpxowGsQIwCg/l3/fyRq2RJlMCl7mLADqo100LkH2vL2AfefkxUG/neSpxl5qrDaFV8CpY36TmkuQPXfUeidNmMHnEyfkhViCjBADoZY4eaqxlxqrjY5LkBFCnqIgTp4yVN+vZwiYlQtQeq+WR/lpfhNCakPdKcTwuDxIwxcA0LR0A81vQkhdKIiTJoF1eg0nPD8GrGqspETzmxBSJ+pOIU2G0LYXJCNnwvjwAprfhBAtURAnTQuX93QYoUcfCuCE1IO6UwghpBmjIE4IIc0YdaeQpoVL85sQogtqiRNCSDNGQZwQQpoxCuKEENKMURAnhJBmjII4IYQ0YwYP4mvWrIGnpydMTU3h6+uLo0eP1pl+27Zt8PHxgbm5OQQCAcaNG4eioiKVNMXFxYiLi4NAIICpqSm8vLyQkZGh03kZY0hJSYGLiwvMzMzQr18/XLhwQT8XTQghemLQIJ6eno6EhAQkJSXh9OnTCAkJQUREBPLy8jSmP3bsGMaMGYOYmBhcuHABO3fuxIkTJzB+/HhlmqqqKoSFheHmzZvYtWsXLl26hG+//RatWrXS6byLFy/GsmXLsHr1apw4cQLOzs4ICwtDaWlp41VIMyKRyrD9zzxs/zMPkoas5iOVACc3yF8NWFGIEPIvZkC9e/dmsbGxKts6derEZs6cqTH9kiVLWJs2bVS2rVy5krm6uirfp6WlsTZt2rCqqqoGn1cmkzFnZ2e2cOFC5f6KigpmY2PDvv76a+0ujjFWUlLCALCSkhKtj2kuxBIp23b8Ftt2/BYTS6S6ZyARM3ZivfwlEbOqqir2448/1vlzI/WjetQPQ9ejLrHDYA/7VFVVIScnBzNnzlTZHh4ejuzsbI3HBAUFISkpCRkZGYiIiEBBQQF27dqFIUOGKNPs2bMHgYGBiIuLw08//QQHBweMGjUKM2bMAI/H0+q8N27cQH5+PsLDw5X7+Xw+QkNDkZ2djQkTJmgsX2VlJSorK5XvRSL5SjRisRhisViH2mn6JFIZpFIpAPn1MZmOX+pkEnD+PZ5JxBBLmTIv0nCK+qN6fDaGrkddzmuwIF5YWAipVAonJyeV7U5OTsjPz9d4TFBQELZt24aoqChUVFRAIpFg6NChWLXq6Srp169fx8GDB/Huu+8iIyMDV65cQVxcHCQSCWbPnq3VeRX/15Tm1q1btV5Tamoq5syZo7Y9MzMT5ubmddRG8yNlwPkC+VzfFgVnwdO8mkOtODIxOt/LgpH0Ce5eLkChlRfA4SIrK6sRSvvyoXrUD0PVY3l5udZpDf7YPYej+tfPGFPbppCbm4v4+HjMnj0bAwcOhFAoxPTp0xEbG4t169YBAGQyGRwdHbF27VrweDz4+vri3r17WLJkCWbPnq3TeXUpGwDMmjULiYmJyvcikQhubm4IDw+HtbV1HbXQ/EikMjw+dQ8AENHTBUY87VvinH9+Bi/zP8rFkT0eHobMSoCT9m/C+61PYWxs3ChlfhmIxWJkZWUhLCyM6vEZGLoeFd/itWGwIG5vbw8ej6fW6i4oKFBrASukpqYiODgY06dPBwB069YNFhYWCAkJwfz58yEQCCAQCGBsbAwe7+kUpl5eXsjPz0dVVZVW53V2li9MkJ+fD4FAoFXZAHmXC5/PV9tubGz8wv1BcbgyZR0bGxtrH8Rz9wD/Nw4AU82vNB+9SldBes0XRl2H67m0L58X8XfOEAxVj7qc02CjU0xMTODr66v2dSUrKwtBQUEajykvLweXq1pkRSBhTB4UgoODcfXqVchkT0dMXL58GQKBACYmJlqd19PTE87OzippqqqqcPjw4VrLpjdSSfN4ySTg/PuCTMtjxJXALzNQM4ADAOffbbzMT+XpDH19jfEipBEYtDslMTERo0ePhp+fHwIDA7F27Vrk5eUhNjYWgLx74u7du9i8eTMAIDIyEh988AHS0tKU3SkJCQno3bs3XFxcAAATJ07EqlWrMHnyZHz88ce4cuUKFixYgPj4eK3Py+FwkJCQgAULFqB9+/Zo3749FixYAHNzc4waNapxK0WbVd+bAA5jaHPnkfzfRi2AOrqZlAqvAKX3as8TkO8/MOfpwhAvEpqdkTQCgwbxqKgoFBUVYe7cuRAKhfD29kZGRgbc3d0BAEKhUGXsdnR0NEpLS7F69WpMnToVtra2GDBgABYtWqRM4+bmhszMTEyZMgXdunVDq1atMHnyZMyYMUPr8wLAJ598gidPnmDSpEl49OgR/P39kZmZCSsrq+dQM02flAEXSs1RLOZBYmGMAAdJ/Tc3K7Xs59M2HSEEHKbohyB6JxKJYGNjg5KSEu1vbDaDr937LtxHys8XkS96OpxSYM1H8mteGNSl9nsGuHkM2PJ6/ScY/ZN8abYXDa/x20xisRgZGRkYPHgw9Yk/A0PXoy6xw+CjU0gNz+EP/VnsOy/ExO1n1Hq180WVmLj9DNLe64lB3gKNx8IzRL56vUgITf3iDACsW4FDiyMTorWmHTFIvRr0yHsDSWUMKXtyNYRfeQDmAJjzv1z07+gIHldT3woHnPCF4O4aK/93tZwYOAAYql6ZBx7jAFpely5DGwl5EVEQb+a+P3nnuZ3r+oMy5Isqat3PAAhLKrBk/yW0cbCsJVUPuPZYBt/chbCovK/cWs53xP+ZR4HzuCd4OlzTKP/WWqcl5EVEQZxorbRCu/76+tLdcX4V9xxC0O3ySphIynBLEAGhbU9cuJALb30UlJCXCAXxZm6kn+tzO1frlmZIP3m73nSRPgIEtLGrO5FMAo5xDwCAR/dhEEsZzB9cQERPF7ohR4gOKIg3c8+zTziwrT0ENqbIL6nQ2C/OAeBsY4rAtva19IlXx306tpzLBQMDjyO/HurnJkR79NdCtMbjcpAc2RnAvw/mVKN4nxzZWYsATgjRFwriRCeDvAVIe68nnKxNVbY725jWPbywJp6R/AlGv3FNflglIU0Z/fUQnQ3yFqB/R0cs2X8JpRUSRPoItOxCIYToGwVx0iA8Lkc5jDCgjR0FcEIMhLpTCCGkGaMgTgghzRgFcUIIacZ0DuIeHh6YO3euyhSxhBBCDEPnID516lT89NNPaNOmDcLCwrBjxw6VFd4JIYQ8PzoH8Y8//hg5OTnIyclB586dER8fD4FAgI8++ginTp1qjDISQgipRYP7xH18fLBixQrcvXsXycnJ+O6779CrVy/4+Phg/fr1oLUmXmxGPC5G+bfGKP/W9Jg8IQbU4HHiYrEYu3fvxoYNG5CVlYWAgADExMTg3r17SEpKwq+//ort27frs6yEEEJq0DmInzp1Chs2bMB///tf8Hg8jB49Gl9++SU6deqkTBMeHo6+ffvqtaCEEELU6RzEe/XqhbCwMKSlpWHYsGEapw3t3Lkz3n77bb0UkBBCSO107sy8fv069u3bh7feeqvWeZ8tLCywYcMGrfJbs2YNPD09YWpqCl9fXxw9erTO9Nu2bYOPjw/Mzc0hEAgwbtw4FBUVKfdv3LgRHA5H7VVR8XRFGg8PD41p4uLilGmio6PV9gcEBGh1TYQQ8rzoHMQLCgrw559/qm3/888/cfLkSZ3ySk9PR0JCApKSknD69GmEhIQgIiKi1jHox44dw5gxYxATE4MLFy5g586dOHHiBMaPH6+SztraGkKhUOVlavp01r0TJ06o7MvKygIAvPXWWyr5DBo0SCVdRkaGTtdHCCGNTecgHhcXh9u31Vd3uXv3rkpLVhvLli1DTEwMxo8fDy8vLyxfvhxubm5IS0vTmP748ePw8PBAfHw8PD090adPH0yYMEHtw4PD4cDZ2VnlVZ2Dg4PKvp9//hlt27ZFaGioSjo+n6+SrmXLljpdHyGENDad+8Rzc3PRs2dPte09evRAbm6u1vlUVVUhJycHM2fOVNkeHh6O7OxsjccEBQUhKSkJGRkZiIiIQEFBAXbt2oUhQ4aopCsrK4O7uzukUim6d++OefPmoUePHrWWY+vWrUhMTASHozoT36FDh+Do6AhbW1uEhobi888/h6OjY63XVFlZqfLgk0gkAiAfySMWi2uvDKKsH6qnZ0P1qB+GrkddzqtzEOfz+bh//z7atGmjsl0oFMLISPvsCgsLIZVK4eTkpLLdyckJ+fn5Go8JCgrCtm3bEBUVhYqKCkgkEgwdOhSrVq1SpunUqRM2btyIrl27QiQSYcWKFQgODsbff/+N9u3bq+X5448/ori4GNHR0SrbIyIi8NZbb8Hd3R03btzAZ599hgEDBiAnJwd8Pl9j+VJTUzFnzhy17ZmZmTA3N6+vSgig7Noiz4bqUT8MVY/l5eVap+UwHZ/Kefvtt5Gfn4+ffvoJNjY2AIDi4mIMGzYMjo6O+P7777XK5969e2jVqhWys7MRGBio3P75559jy5Yt+Oeff9SOyc3NxauvvoopU6Zg4MCBEAqFmD59Onr16oV169ZpPI9MJkPPnj3Rt29frFy5Um3/wIEDYWJigv/97391llcoFMLd3R07duzAiBEjNKbR1BJ3c3NDYWEhrK2t68z/ZScWi5GVlYWwsDBaKPkZUD3qh6HrUSQSwd7eHiUlJfXGDp1b4kuXLkXfvn3h7u6u7KI4c+YMnJycsGXLFq3zsbe3B4/HU2t1FxQUqLXOFVJTUxEcHIzp06cDALp16wYLCwuEhIRg/vz5EAjUlwbjcrno1asXrly5orbv1q1b+PXXX/HDDz/UW16BQAB3d3eN+Sjw+XyNrXRjY2P6g9IS1ZV+UD3qh6HqUZdz6nxjs1WrVjh79iwWL16Mzp07w9fXFytWrMC5c+fg5uamdT4mJibw9fVV+7qSlZWFoKAgjceUl5eDy1UtMo/HA4BaH/NnjOHMmTMaA/yGDRvg6Oio1qeuSVFREW7fvq0xH0IIMZQGPXZvYWGBDz/88JlPnpiYiNGjR8PPzw+BgYFYu3Yt8vLyEBsbCwCYNWsW7t69i82bNwMAIiMj8cEHHyAtLU3ZnZKQkIDevXvDxcUFADBnzhwEBASgffv2EIlEWLlyJc6cOYOvvvpK5dwymQwbNmzA2LFj1fryy8rKkJKSgjfeeAMCgQA3b97Ep59+Cnt7ewwfPvyZr5sQQvSlwXOn5ObmIi8vD1VVVSrbhw4dqnUeUVFRKCoqwty5cyEUCuHt7Y2MjAy4u7sDkPdDVx8zHh0djdLSUqxevRpTp06Fra0tBgwYgEWLFinTFBcX48MPP0R+fj5sbGzQo0cPHDlyBL1791Y596+//oq8vDy8//77auXi8Xg4d+4cNm/ejOLiYggEAvTv3x/p6emwsrLS+voIIaSx6Xxj8/r16xg+fDjOnTsHDoej7MZQDM+TSqX6L2UzJRKJYGNjo9XNiZedWCxGRkYGBg8eTH25z4DqUT8MXY+6xA6d+8QnT54MT09P3L9/H+bm5rhw4QKOHDkCPz8/HDp0qKFlJoQQ0gA6d6f88ccfOHjwIBwcHMDlcsHlctGnTx+kpqYiPj4ep0+fboxyEkII0UDnlrhUKoWlpSUA+TDBe/fuAQDc3d1x6dIl/ZaOEEJInXRuiXt7e+Ps2bNo06YN/P39sXjxYpiYmGDt2rVqT3ESQghpXDoH8f/85z94/PgxAGD+/Pl47bXXEBISAjs7O6Snp+u9gIQQQmqncxAfOHCg8t9t2rRBbm4uHj58iBYtWqhNIEUIIaRx6dQnLpFIYGRkhPPnz6tsb9myJQVwQggxAJ2CuJGRkXKKV0IIIYan8+iU//znP5g1axYePnzYGOUhhBCiA537xFeuXImrV6/CxcUF7u7usLCwUNl/6tQpvRWOEEJI3XQO4sOGDWuEYhBCCGkInYN4cnJyY5SDEEJIA+jcJ04IIaTp0LklzuVy6xxOSCNXCCHk+dE5iO/evVvlvVgsxunTp7Fp0yaNiwQTQghpPDoH8ddff11t25tvvokuXbogPT0dMTExeikYIYSQ+umtT9zf3x+//vqrvrIjRGsSqQzb/8zD9j/zIJHKDF0cQp4rvQTxJ0+eYNWqVXB1ddVHdoQQQrSkc3dKzYmuGGMoLS2Fubk5tm7dqtfCEUIIqZvOQfzLL79UCeJcLhcODg7w9/dHixYt9Fo4QgghddO5OyU6Ohpjx45VvkaPHo1BgwY1OICvWbMGnp6eMDU1ha+vL44ePVpn+m3btsHHxwfm5uYQCAQYN24cioqKlPs3btwIDoej9qqoqFCmSUlJUdvv7Oysch7GGFJSUuDi4gIzMzP069cPFy5caNA1EkJIY9E5iG/YsAE7d+5U275z505s2rRJp7zS09ORkJCApKQknD59GiEhIYiIiEBeXp7G9MeOHcOYMWMQExODCxcuYOfOnThx4gTGjx+vks7a2hpCoVDlZWpqqpKmS5cuKvvPnTunsn/x4sVYtmwZVq9ejRMnTsDZ2RlhYWEoLS3V6RoJIaQx6RzEFy5cCHt7e7Xtjo6OWLBggU55LVu2DDExMRg/fjy8vLywfPlyuLm5IS0tTWP648ePw8PDA/Hx8fD09ESfPn0wYcIEnDx5UiWdomVd/VWTkZGRyn4HBwflPsYYli9fjqSkJIwYMQLe3t7YtGkTysvLsX37dp2ukRBCGpPOfeK3bt2Cp6en2nZ3d/daW9CaVFVVIScnBzNnzlTZHh4ejuzsbI3HBAUFISkpCRkZGYiIiEBBQQF27dqFIUOGqKQrKytTznvevXt3zJs3Dz169FBJc+XKFbi4uIDP58Pf3x8LFixQrhF648YN5OfnIzw8XJmez+cjNDQU2dnZmDBhgsbyVVZWorKyUvleJBIBkD8QJRaLtayZl5OifhpSTxKpTPmksFgsBpO9vLNJPEs9kqcMXY+6nFfnIO7o6IizZ8/Cw8NDZfvff/8NOzs7rfMpLCyEVCqFk5OTynYnJyfk5+drPCYoKAjbtm1DVFQUKioqIJFIMHToUKxatUqZplOnTti4cSO6du0KkUiEFStWIDg4GH///Tfat28PQD6mffPmzejQoQPu37+P+fPnIygoCBcuXICdnZ3y/JrKduvWrVqvKTU1VeNTq5mZmTA3N9euYl5yWVlZOh8jZcD5AvnNdouCs+DRIlMNqkeizlD1WF5ernVanYP422+/jfj4eFhZWaFv374AgMOHD2Py5Ml4++23dc1ObR4Wxlitc7Pk5uYiPj4es2fPxsCBAyEUCjF9+nTExsZi3bp1AICAgAAEBAQojwkODkbPnj2xatUqrFy5EgAQERGh3N+1a1cEBgaibdu22LRpExITExtUNgCYNWuWyvEikQhubm4IDw+HtbV1fVXxUhOLxcjKykJYWBiMjY11OrZSLMX5X6+itEICu45O8G9jBx735Yzkz1KP5ClD16PiW7w2dA7i8+fPx61bt/DKK6/AyEh+uEwmw5gxY3TqE7e3twePx1NrdRcUFKi1gBVSU1MRHByM6dOnAwC6desGCwsLhISEYP78+RAIBGrHcLlc9OrVC1euXKm1LBYWFujatasyjaIPPT8/XyXPusoGyLtc+Hy+2nZjY2P6g9KSrnW177wQKXtykS+Sjz7aeeouBDamSI7sjEHe6r8PLwv6ndMPQ9WjLufUufPQxMQE6enpuHTpErZt24YffvgB165dw/r162FiYqJTPr6+vmpfV7KyshAUFKTxmPLycnC5qkXm8XgA5K1kTRhjOHPmjMYAr1BZWYmLFy8q03h6esLZ2VmlbFVVVTh8+HCtZSPP377zQkzcekoZwBXySyowcesp7DsvNFDJCHl+dG6JK7Rv317Zx9xQiYmJGD16NPz8/BAYGIi1a9ciLy8PsbGxAOTdE3fv3sXmzZsBAJGRkfjggw+Qlpam7E5JSEhA79694eLiAgCYM2cOAgIC0L59e4hEIqxcuRJnzpzBV199pTzvtGnTEBkZidatW6OgoADz58+HSCTC2LFjAci7URISErBgwQLldS5YsADm5uYYNWrUM11zY2rO84ZIpDJImfz/HG791yGVMaTsyYWmj24GgANgzv9y0b+j40vRtWLEe3lv5r7sdA7ib775Jvz8/NRGlSxZsgR//fWXxjHktYmKikJRURHmzp0LoVAIb29vZGRkwN3dHQAgFApVRrxER0ejtLQUq1evxtSpU2Fra4sBAwZg0aJFyjTFxcX48MMPkZ+fDxsbG/To0QNHjhxB7969lWnu3LmDd955B4WFhXBwcEBAQACOHz+uPC8AfPLJJ3jy5AkmTZqER48ewd/fH5mZmbCystK1yp6b70/eMXQRGkwqleJ8AQePT91Tfruqy/UHZWot8OoYAGFJBZbsv4Q2DpZ6LGnTNMq/taGLQAyEw2rrh6iFg4MDDh48iK5du6psP3fuHF599VXcv39frwVszkQiEWxsbFBSUvJcbmxu/1P7IZ5NjVQqxfnz5+Dt3bXOIM6RSdDmzm78/tAKK2+0qjffKD83+LjZ6rGkTZMiiIvFYmRkZGDw4MHUJ/4MDF2PusQOnVviZWVlGvu+jY2NdbqjSvRvpF/znUVSLBbDouAsInq61P1HI5OAY9QCEgtj4Eb9+Ub6CBDQRvuhr4Q0NzoHcW9vb6Snp2P27Nkq23fs2IHOnTvrrWBEd825X5TJuOBx5NdQ93VwAQ4HAQ4SCKz5yBdVauwX5wBwtjFFYFv7l6JPnLy8dA7in332Gd544w1cu3YNAwYMAAAcOHAA27dvx65du/ReQEI04XGA5Ne8MHH7GXAAlUCuCNnJkZ0pgJMXns5Nt6FDh+LHH3/E1atXMWnSJEydOhV3797FwYMH1Z7iJKQxDerihLT3esLJWnVyM2cbU6S91/OlHidOXh4NGmI4ZMgQ5XwlxcXF2LZtGxISEvD333/TavekccmkQOEVoFIEtPDEoM4h6N/REUv2X0JphQSRPgLqQiEvlQaPEz948CDWr1+PH374Ae7u7njjjTeUj74T0ihy9wC/zABK78nfn94CWLvAKHwh2jjIJzgLeIkfuScvJ52C+J07d7Bx40asX78ejx8/xsiRIyEWi/F///d/dFOTNK7cPcD3Y4CatzFFQnB3jYVrj2W44/yqQYpGiCFpHcQHDx6MY8eO4bXXXsOqVaswaNAg8Hg8fP31141ZPgIAUomhS9D4ZBJwmBSQSQBpjZa0TCpvgdfxfKbvxYW45xAiP14/638T0ixoHcQzMzMRHx+PiRMnPvPj9kRHp7cYugSNjiOVonXReXDOPARqPuxTeOVpF4qmY8FgUXEf3S6vBMe4B1DHTJMvPJ/3DF0C8pxp3WQ5evQoSktL4efnB39/f6xevRoPHjxozLIRIlep3UNkJpKyRi4IIU2P1i3xwMBABAYGYsWKFdixYwfWr1+PxMREyGQyZGVlwc3NrUnPK9Ks9Rht6BI0OiYRI0/4C7p0jwCMajyx2cJTq28jtwQR8Og+DOC+xN0pMp1m0SAvAJ1Hp5ibm+P999/H+++/j0uXLmHdunVYuHAhZs6cibCwMOzZs6cxyvly4zV4EFHzIWNgHB7ANVK/Xs8QwNoFEAmhqV+cgYNyUycU2PX69/iXOYjTsmwvm2f6be/YsSMWL16MO3fu4L///a++ykSIKi4PGKSYqbJmf7f8fY7XDPmHACEvGb008Xg8HoYNG4Zhw4bpIztC1HUeCozcrDpOHACsXcAZtBB9Ow81XNkIMaCX4Hs6eWF0Hgq0HwgcmCO/2en9pryrhUstcPLyoiBOmhcuD7D/d4irRx8K4OSl9xLfASKEkOaPgjghhDRjFMQJIaQZoz5x0rzwjAC/cYYuBSFNBrXECSGkGTN4EF+zZg08PT1hamoKX19fHD16tM7027Ztg4+PD8zNzSEQCDBu3DgUFRUp92/cuBEcDkftVVFRoUyTmpqKXr16wcrKCo6Ojhg2bBguXbqkcp7o6Gi1PAICAvR78YQQ8owMGsTT09ORkJCApKQknD59GiEhIYiIiEBeXp7G9MeOHcOYMWMQExODCxcuYOfOnThx4gTGjx+vks7a2hpCoVDlZWr6dAmvw4cPIy4uDsePH0dWVhYkEgnCw8Px+PFjlXwGDRqkkkdGRob+K4EQQp6BQfvEly1bhpiYGGUQXr58Ofbv34+0tDSkpqaqpT9+/Dg8PDwQHx8PAPD09MSECROwePFilXQcDgfOzs61nnffvn0q7zds2ABHR0fk5OSgb9++yu18Pr/OfAghxNAMFsSrqqqQk5ODmTNnqmwPDw9Hdna2xmOCgoKQlJSEjIwMREREoKCgALt27VKu96lQVlYGd3d3SKVSdO/eHfPmzUOPHj1qLUtJSQkAoGXLlirbDx06BEdHR9ja2iI0NBSff/45HB0da82nsrISlZWVyvcikXwKVbFYDLGYJiaqi6J+qJ6eDdWjfhi6HnU5L4cxZpC5K+/du4dWrVrh999/R1BQkHL7ggULsGnTJrU+aoVdu3Zh3LhxqKiogEQiwdChQ7Fr1y4YG8unLz1+/DiuXr2Krl27QiQSYcWKFcjIyMDff/+tcTELxhhef/11PHr0SKU/Pj09HZaWlnB3d8eNGzfw2WefQSKRICcnB3w+X2PZUlJSMGfOHLXt27dvh7m5uU71Qwh5eZWXl2PUqFEoKSmBtbV1nWkNHsSzs7MRGBio3P75559jy5Yt+Oeff9SOyc3NxauvvoopU6Zg4MCBEAqFmD59Onr16lXrIs0ymQw9e/ZE3759sXLlSrX9cXFx2Lt3L44dOwZXV9dayysUCuHu7o4dO3ZgxIgRGtNoaom7ubmhsLCw3h/Ey04sFiMrKwthYWHKD2SiO6pH/TB0PYpEItjb22sVxA3WnWJvbw8ej4f8/HyV7QUFBXByctJ4TGpqKoKDgzF9+nQAQLdu3WBhYYGQkBDMnz8fAoFA7Rgul4tevXrhypUravs+/vhj7NmzB0eOHKkzgAOAQCCAu7u7xnwU+Hy+xla6sbEx/UFpiepKP6ge9cNQ9ajLOQ02OsXExAS+vr7IyspS2Z6VlaXSvVJdeXk5uDVWbeH9ux5jbV8oGGM4c+aMSoBnjOGjjz7CDz/8gIMHD8LT07Pe8hYVFeH27dsaPygIIcRQDDo6JTExEaNHj4afnx8CAwOxdu1a5OXlITY2FgAwa9Ys3L17F5s3bwYAREZG4oMPPkBaWpqyOyUhIQG9e/eGi4sLAGDOnDkICAhA+/btIRKJsHLlSpw5cwZfffWV8rxxcXHYvn07fvrpJ1hZWSm/DdjY2MDMzAxlZWVISUnBG2+8AYFAgJs3b+LTTz+Fvb09hg8f/pxriRBCamfQIB4VFYWioiLMnTsXQqEQ3t7eyMjIgLu7OwB5P3T1MePR0dEoLS3F6tWrMXXqVNja2mLAgAFYtGiRMk1xcTE+/PBD5Ofnw8bGBj169MCRI0fQu3dvZZq0tDQAQL9+/VTKs2HDBkRHR4PH4+HcuXPYvHkziouLIRAI0L9/f6Snp9M6ooSQJsVgNzZfBiKRCDY2NlrdnHjZicViZGRkYPDgwdSX+wyoHvXD0PWoS+ww+GP3hBBCGo6COCGENGMUxAkhpBmjIE4IIc0YBXFCCGnGKIgTQkgzRkGcEEKaMQrihBDSjFEQJ4SQZoxWuyfkBSKVSmFkZISKigpIpVJDF6fZEovFjV6PJiYmahP6NQQFcUJeAIwx5Ofn49GjR3B2dsbt27fB4XAMXaxmizHW6PXI5XLh6ekJExOTZ8qHgjghL4D8/HwUFxfDwcEBMpkMVlZWemnlvaxkMhnKyspgaWnZKPUok8lw7949CIVCtG7d+pk+KCiIE9LMSaVSFBcXw9HRES1atIBIJIKpqSkF8Wcgk8lQVVXVqPXo4OCAe/fuQSKRPNMkW/RTJqSZUyyqS+u4Ni+KbpRn7XOnljghLwh99N1KpDJ8f/IOAGCknyuMeNTOayz66munnxAhhDRj1BInhChJZQzXH5ShtEKC1i3NENjWHjwujXJpyiiIE0IAAPvOC5GyJxf5ogoAQPrJ2xDYmCI5sjMGedMC4U0VdacQQrDvvBATt55SBnCF/JIKTNx6CvvOCw1UssanuDHcXFEQJ+QlJZHKIJHKUCmWImVPLjQttqvYNud/uagUSyGRyvRejn379qFPnz6wtbWFnZ0dXnvtNVy7dg0A8MYbb+Djjz9Wpk1ISACHw8GFCxfk1yCRwMrKCvv37683LwC4efMmOBwOvv/+e/Tr1w+mpqbYunUrAPlC6V5eXjA1NUXnzp3x3Xff6f1aG4PBg/iaNWvg6ekJU1NT+Pr64ujRo3Wm37ZtG3x8fGBubg6BQIBx48ahqKhIuX/jxo3gcDhqr4oK1RZGfedljCElJQUuLi4wMzNDv379lL84hLwIvj95B9+fvIMl+y+ptcCrYwCEJRVYsv+ScuSKPj1+/BiJiYk4ceIEDhw4AC6Xi+HDh0Mmk6Ffv344dOiQMu3hw4dhb2+Pw4cPAwBOnDiBiooKBAcH15tXdTNmzEB8fDwuXryIgQMH4ttvv0VSUhI+//xzXLx4EfPnz8eCBQuwadMmvV+vvhk0iKenpyMhIQFJSUk4ffo0QkJCEBERgby8PI3pjx07hjFjxiAmJgYXLlzAzp07ceLECYwfP14lnbW1NYRCocrL1NRUp/MuXrwYy5Ytw+rVq3HixAk4OzsjLCwMpaWljVMZhBhIaYVEr+l09cYbb2DEiBFo3749unfvjnXr1uHcuXPIzc1VNp4KCwvx6NEjXLhwAQkJCcrAfujQIfj6+sLS0rLevKpLSEjAiBEj4OnpCRcXF8ybNw9Lly5VbhsxYgQmTZqEb7/9tlGuWZ8MGsSXLVuGmJgYjB8/Hl5eXli+fDnc3NyQlpamMf3x48fh4eGB+Ph4eHp6ok+fPpgwYQJOnjypko7D4cDZ2Vnlpct5GWNYvnw5kpKSMGLECHh7e2PTpk0oLy/H9u3bG6cySN2kEuDkBvlL2jjB5GUz0s8VI/1cEemj3U3LSB8BRvq56r0c165dw6hRo9CmTRtYW1vD09MTAJCXlwdvb2/Y2dnh8OHDOHr0KHx8fDB06FBlS/zQoUMIDQ3VKq/q/Pz8lP9+8OABbt++jZiYGFhaWsLS0hLW1tb44osvVLpimiqDjU6pqqpCTk4OZs6cqbI9PDwc2dnZGo8JCgpCUlISMjIyEBERgYKCAuzatQtDhgxRSVdWVgZ3d3dIpVJ0794d8+bNQ48ePbQ+740bN5Cfn4/w8HDlfj6fj9DQUGRnZ2PChAkay1dZWYnKykrle5FIBEB+46S53zxpbIr6qbWeZBJw/n2yjUnEgExTD+7LSSwWgzEGmUwGxuT1onhfF8XIQX/PlnC2NsV9UYXGfnEOAGcbU/h7tgSXg3rz1VVkZCRcXV3xzTffwMXFBTKZDN26dUNFRQUYYwgJCcFvv/0GExMThIaGonPnzpBKpfj777+RnZ2N+Ph4ZZnqyksmkynTmZmZKf8tkcgbBd988w38/f0ByOvv8ePHsLa21vv1Kih+XmKxGDweT2WfLvHCYEG8sLAQUqkUTk5OKtudnJyQn5+v8ZigoCBs27YNUVFRqKiogEQiwdChQ7Fq1Splmk6dOmHjxo3o2rUrRCIRVqxYgeDgYPz9999o3769VudV/F9Tmlu3btV6TampqZgzZ47a9szMTHokWktZWVkat3OYFK2LzgMA8oS/gHF4GtO9jIyMjODs7IyysjJUVVUBgM7dftNf8cC03f+obVeMEJ82wAOPy/Tflfjw4UNcvHgRX3zxBXr16gUA+OOPPwAAT548gUgkQkBAADZu3AgTExPMmjULpaWlCAgIwMKFC/HkyRPl37o2eZWVlQGQ950rGllmZmZwcXHBP//8g8jISLUyKtLpW1VVFZ48eYIjR44oP0gUysvLtc7H4OPEaz56yhir9XHU3NxcxMfHY/bs2Rg4cCCEQiGmT5+O2NhYrFu3DgAQEBCAgIAA5THBwcHo2bMnVq1ahZUrV+p0Xl3KBgCzZs1CYmKi8r1IJIKbmxvCw8NhbW1d63FE3vLIyspCWFiY5smAZBJwzjwEAHTpHgFwDf6r22RUVFTg9u3bsLS0BJ/PR2lpKaysrHR6rHt4L2uYmZlh7s+5yBc9/TbpbGOKz4Z4YZC3cx1HN5ylpSXs7Oywfft2tGvXDnl5eUhOTgYgD67W1tYYOHAgZs6cCSMjIwwcOBDW1tZ49dVXMW3aNPTs2ROurq5a56XoO7ewsFD5m0xOTkZCQgIcHBwwaNAgVFRU4Pfff8eTJ09U/qb1qaKiAmZmZujbt6/KPTtAtw8Og/0l2Nvbg8fjqbW6CwoK1FrACqmpqQgODsb06dMBAN26dYOFhQVCQkIwf/58CATqfXtcLhe9evXClStXtD6vog89Pz9fJc+6ygbIu1z4fL7admNj42eapexlUmtdSTmA4iunkTHAoyCuIJVKweFwwOVylYFb8V4Xg7u54BUvJyzZfwmlFRJE+gga/YlNLpeLHTt2ID4+Ht26dUPHjh2xcuVK9OvXD1wuF1wuF926dYO9vT3c3d1ha2sLAOjXrx+kUilCQ0OV16lNXtXTVq+fDz/8EJaWlliyZAlmzJgBCwsLeHl5ITExsdFmMVT8vDT9zusSLwz2l2BiYgJfX19kZWVh+PDhyu1ZWVl4/fXXNR5TXl4OIyPVIiv6khR9gTUxxnDmzBl07dpV6/N6enrC2dkZWVlZKn3phw8fxqJFixp4xYQ0fTwuB20c5K3VgDZ2z+WR+1dffVVt9Ej1v2cOh4OCggKV/d27d9f4N19fXh4eHrXGilGjRmHUqFEA5P3VIpGoWXyDNmhzJjExEaNHj4afnx8CAwOxdu1a5OXlITY2FoC8e+Lu3bvYvHkzAPlNiw8++ABpaWnK7pSEhAT07t0bLi4uAIA5c+YgICAA7du3h0gkwsqVK3HmzBl89dVXWp+Xw+EgISEBCxYsQPv27dG+fXssWLAA5ubmyh8yIS8iIx4Xo/xbG7oYRAcGDeJRUVEoKirC3LlzIRQK4e3tjYyMDLi7uwMAhEKhytCg6OholJaWYvXq1Zg6dSpsbW0xYMAAldZxcXExPvzwQ+Tn58PGxgY9evTAkSNH0Lt3b63PCwCffPIJnjx5gkmTJuHRo0fw9/dHZmYmrKysnkPNEEKIdjistu8W5JmJRCLY2NigpKSkWXwtMySxWIyMjAwMHjy4lj5xCXB6i/zfPUZTn3g1FRUVuHHjhnK9RkU3AK3s03DVu1Maqx6r/9w03djUNnbQT5kQQpoxas6Q5kEmBQqvAJUioIUn4BkCcGmsOCEUxEnTl7sH+GUGUHpP/v70FsDaBRi0COg81LBlI8TAqDuFNG25e4DvxzwN4AoioXx77h7DlIuQJoJa4i+a5jo5lEwCDpMCMon8wR5A3oXyywyg1pmuOcC+mUD7gS9O14qhb9jSDeRmh35CLxrFH2Azw5HK50bhnHn49MnMwivqLXAVDBDdBQ7MAezbP5dyNjq/cYYuAWlmKIiTpqtSy/kjtE1H6kc3kJsdCuIvmh6jDV2CBmESMfKEv8gntzL6d5x4C0/tvll4vwl49GncAr4M6AZys0RB/EXTXPswZUw+vSzX6Ok1eIbIg4hICM394hz5fmotPjvFDeSa9ay4gTxyMwXyWlRVVcHExMRg56fRKaTp4vLkrUAAT2e2hur7QQspgDeUVCJ/iSvruYEM+Q1kcaXeb5x7eHhg+fLlKtu6d++OlJQUAPJ5jL777jsMHz4c5ubmaN++PfbseToiSSqVIiYmBp6enjAzM0PHjh2xYsUK5f5z586By+WisLAQAPDo0SNwuVy89dZbyjSpqakIDAxUya9t27YQCATw8vJSyQ+QT/8xbNgwpKamwsXFBR06dAAA3L17F1FRUWjRogXs7Ozw+uuv4+bNm/qqqlpRECdNW+eh8lagVY1phq1dqHX4rE5vkb8OzNH+BrIBbpzPmTMHI0eOxNmzZzF48GC8++67ePhQPre8TCaDq6srvv/+e+Tm5mL27Nn49NNP8f333wOAyvJuAHDkyBHY2dnhyJEjyvyrL/GmyG/Hjh04fvw4/vOf/6jkp3DgwAFcvHgRWVlZ+Pnnn1FeXo7+/fvD0tISR44cwbFjx2BpaYlBgwYpF+poLBTESdPXeSgQfwYIiJP3+Y/+CUg4RwFcX5r4DeTo6Gi88847aNeuHRYsWIDHjx/jr7/+AiCfd3vOnDno1asXPD098e677yI6OloZdDkcDvr27auysPLYsWMhk8mQm5sLiUSC7Oxs9OvXTy0/d3d3tfwULCws8N1336FLly7w9vbGjh07wOVy8d1336Fr167w8vLChg0bkJeXpzx3Y2mmHajkpcPlPR1G6NGHulD0QXETvInfQO7WrZvy3xYWFrCyslKZX/zrr7/Gd999h1u3buHJkyeoqqpC9+7dlfv79euHtWvXAgAOHz6MefPm4caNGzh8+DBKSkrw5MkTBAcHq+V38+ZNVFRUqOUHAF27dlXpB8/JycHVq1fVZjmtqKho9MWWKYgT8rJqAjeQuVyu2iINNRcJrjmrJYfDUS5e/P3332PKlClYunQpAgMDYWVlhSVLluDPP/9Upu/Xrx8mT56Mq1ev4vz58wgJCcG1a9dw+PBhFBcXw9fXVxl8Ffl98cUX6Nq1K5ydnbF06VKV/AD5h0l1MpkMvr6+2LZtm9o1Ojg46FgruqEgTsjLTnED+fsxkN8wrh5UG/cGsoODA4RCofK9SCTCjRs3tD7+6NGjCAoKwqRJk5TbarZ8Ff3i8+fPh4+PD6ytrREaGorU1FQ8evRI2R9ePb+JEycqp6LVpiXds2dPpKenw9HR8blPO0194oQQg91AHjBgALZs2YKjR4/i/PnzGDt2rHLJRW20a9cOJ0+exP79+3H58mV89tlnOHHihEoaRb/41q1blX3f3bp1Q1VVFQ4cOKDcVjO/q1evYvbs2Wr5afLuu+/C3t4er7/+Oo4eParsrpk8eTLu3Lmj9fU0BAVxQoicAW4gz5o1C3379sVrr72GwYMHY9iwYWjbtq3Wx8fGxmLEiBGIioqCv78/ioqKVFrlCv3794dUKlUGbA6Hg5CQEABAnz591PJ755138Oqrr9aaX03m5uY4cuQIWrdujREjRsDLywvvv/8+njx50ugtc1rZpxHRyj7ao5V9Gk6vK/tQPQNoXiv7vJw/IUKIZjwjmoSrmaHuFEIIacYMHsTXrFmj/Drh6+uLo0eP1pl+27Zt8PHxgbm5OQQCAcaNG4eioiKNaXfs2AEOh4Nhw4apbPfw8ACHw1F7xcXFKdNER0er7Q8ICHjm6yWEEH0yaBBPT09HQkICkpKScPr0aYSEhCAiIgJ5eXka0x87dgxjxoxBTEwMLly4gJ07d+LEiRMYP368Wtpbt25h2rRpypsX1Z04cQJCoVD5ysrKAgCV+RQAYNCgQSrpMjIy9HDVpEEUX/P9xr20/bSEaGLQIL5s2TLExMRg/Pjx8PLywvLly+Hm5oa0tDSN6Y8fPw4PDw/Ex8fD09MTffr0wYQJE3Dy5EmVdFKpFO+++y7mzJmDNm3aqOXj4OAAZ2dn5evnn39G27ZtVcaLAgCfz1dJ17JlS/1dPCF6RmMUmhd9/bwM1qSpqqpCTk4OZs6cqbI9PDwc2dnZGo8JCgpCUlISMjIyEBERgYKCAuzatQtDhgxRSTd37lw4ODggJiam3u6ZqqoqbN26FYmJieBwVGfKO3ToEBwdHWFra4vQ0FB8/vnncHR0rDWvyspKVFZWKt+LRPK5JsRisdpTaESVon6onhqGMYaysjLlo+CMMeVTjUR3igDbmPVYWVkJxhgYY2q/97r8HRgsiBcWFkIqlcLJyUllu5OTE/Lz8zUeExQUhG3btiEqKgoVFRWQSCQYOnQoVq1apUzz+++/Y926dThz5oxW5fjxxx9RXFyM6Ohole0RERF466234O7ujhs3buCzzz7DgAEDkJOTAz6frzGv1NRUzJkzR217ZmYmzM3NtSrPy07RtUV0Y2VlhcrKSlRUVMDExKTRZ857WdR2v+1ZMcbw4MEDPHz4EFeuXFHbX15ernVeBu9crNn6ZYypbVPIzc1FfHw8Zs+ejYEDB0IoFGL69OmIjY3FunXrUFpaivfeew/ffvst7O3ttTr/unXrEBERARcXF5XtUVFRyn97e3vDz88P7u7u2Lt3L0aMGKExr1mzZiExMVH5XiQSwc3NDeHh4TROvB5isRhZWVkICwvTPE6c1IkxhoKCApSUlKCiogKmpqa1/h2R+jHGGr0ejYyM4Ofnp/H3XfEtXqt89FkoXdjb24PH46m1ugsKCtRa5wqpqakIDg7G9OnTAcgfnbWwsEBISAjmz5+P+/fv4+bNm4iMjFQeo/gqZGRkhEuXLqk8DXbr1i38+uuv+OGHH+otr0AggLu7u8ZPTQU+n6+xlW5sbEyBSUtUVw3n6uoKe3t7HDhwAH379qV6fAZisRhHjhxp1Ho0MTGp9UEiXc5psCBuYmICX19fZGVlYfjw4crtWVlZeP311zUeU15eDiMj1SIr5llgjKFTp044d+6cyv7//Oc/KC0txYoVK+Dm5qayb8OGDXB0dFTrU9ekqKgIt2/fhkAgqDctIYbC4/EgkUhgampKQfwZNKd6NGh3SmJiIkaPHg0/Pz8EBgZi7dq1yMvLQ2xsLAB598Tdu3exefNmAEBkZCQ++OADpKWlKbtTEhIS0Lt3b2V3iLe3t8o5bG1tNW6XyWTYsGEDxo4dq/bBUFZWhpSUFLzxxhsQCAS4efMmPv30U9jb26t84BBCiKEZNIhHRUWhqKgIc+fOhVAohLe3NzIyMuDu7g4AEAqFKmPGo6OjUVpaitWrV2Pq1KmwtbXFgAEDsGjRotpOUatff/0VeXl5eP/999X28Xg8nDt3Dps3b0ZxcTEEAgH69++P9PR0tUnfCSHEkGgCrEZEE2Bpr94JsIhWqB71w9D1SBNgNRGKz0dd7jS/rMRiMcrLyyESiSj4PAOqR/0wdD0qYoY2bWwK4o2otLQUANRuqBJCiDZKS0thY2NTZxrqTmlEMpkM9+7dg5WVFY3ZrYdiTP3t27ep6+kZUD3qh6HrkTGG0tJSuLi41DufObXEGxGXy4Wrq6uhi9GsWFtbU/DRA6pH/TBkPdbXAlcw+FS0hBBCGo6COCGENGMUxEmTwOfzkZycXOvkYkQ7VI/60ZzqkW5sEkJIM0YtcUIIacYoiBNCSDNGQZwQQpoxCuKEENKMURAnz82aNWvg6ekJU1NT+Pr61rn+6Q8//ICwsDA4ODjA2toagYGB2L9//3MsbdOlSz0eO3YMwcHBsLOzg5mZGTp16oQvv/zyOZa2adKlDqv7/fffYWRkhO7duzduAXXBCHkOduzYwYyNjdm3337LcnNz2eTJk5mFhQW7deuWxvSTJ09mixYtYn/99Re7fPkymzVrFjM2NmanTp16ziVvWnStx1OnTrHt27ez8+fPsxs3brAtW7Ywc3Nz9s033zznkjcdutahQnFxMWvTpg0LDw9nPj4+z6ewWqAhhuS58Pf3R8+ePZGWlqbc5uXlhWHDhiE1NVWrPLp06YKoqCjMnj27sYrZ5OmjHkeMGAELCwts2bKlsYrZpDW0Dt9++220b98ePB4PP/74o9aLsTc26k4hja6qqgo5OTkIDw9X2R4eHo7s7Gyt8pDJZCgtLUXLli0bo4jNgj7q8fTp08jOzkZoaGhjFLHJa2gdbtiwAdeuXUNycnJjF1FnNAEWaXSFhYWQSqVqC2A7OTmpLZRdm6VLl+Lx48cYOXJkYxSxWXiWenR1dcWDBw8gkUiQkpKC8ePHN2ZRm6yG1OGVK1cwc+ZMHD16VG0px6ag6ZWIvLBqTsfLGNNqit7//ve/SElJwU8//QRHR8fGKl6z0ZB6PHr0KMrKynD8+HHMnDkT7dq1wzvvvNOYxWzStK1DqVSKUaNGYc6cOejQocPzKp5OKIiTRmdvbw8ej6fW0ikoKFBrEdWUnp6OmJgY7Ny5E6+++mpjFrPJe5Z69PT0BAB07doV9+/fR0pKyksZxHWtw9LSUpw8eRKnT5/GRx99BEDetccYg5GRETIzMzFgwIDnUvbaUJ84aXQmJibw9fVFVlaWyvasrCwEBQXVetx///tfREdHY/v27RgyZEhjF7PJa2g91sQYQ2Vlpb6L1yzoWofW1tY4d+4czpw5o3zFxsaiY8eOOHPmDPz9/Z9X0WtnwJEx5CWiGNa1bt06lpubyxISEpiFhQW7efMmY4yxmTNnstGjRyvTb9++nRkZGbGvvvqKCYVC5au4uNhQl9Ak6FqPq1evZnv27GGXL19mly9fZuvXr2fW1tYsKSnJUJdgcLrWYU3JyclNaoghdaeQ5yIqKgpFRUWYO3cuhEIhvL29kZGRAXd3dwCAUChEXl6eMv0333wDiUSCuLg4xMXFKbePHTsWGzdufN7FbzJ0rUeZTIZZs2bhxo0bMDIyQtu2bbFw4UJMmDDBUJdgcLrWYVNH48QJIaQZoz5xQghpxiiIE0JIM0ZBnBBCmjEK4oQQ0oxRECeEkGaMgjghhDRjFMQJIaQZoyBOCCHNGAVxopOUlBSdlqa6efMmOBzOM0+gr6989OGff/5BQEAATE1Nm9YyXbXgcDj48ccfATyfety4cSNsbW31nm90dDSGDRtWZ5p+/fohISFB7+duyiiIEwBAZGRkrbME/vHHH+BwODh16hSmTZuGAwcONGpZNP2xurm5KR+RNrTk5GRYWFjg0qVLjV4X+taU6pHoBwVxAgCIiYnBwYMHcevWLbV969evR/fu3dGzZ09YWlrCzs7uuZePx+PB2dm5SUzKf+3aNfTp0wfu7u4GqYtn0ZTqkegHBXECAHjttdfg6OioNrlUeXm5ck5vQL07RSaTYe7cuXB1dQWfz0f37t2xb9++Ws8jlUoRExMDT09PmJmZoWPHjlixYoVyf0pKCjZt2oSffvoJHA4HHA4Hhw4d0tgNcPjwYfTu3Rt8Ph8CgQAzZ86ERCJR7u/Xrx/i4+PxySefoGXLlnB2dkZKSkqd9VDf9XA4HOTk5GDu3LngcDi15scYw+LFi9GmTRuYmZnBx8cHu3btUkmTkZGBDh06wMzMDP3798fGjRvB4XBQXFyssa4BYPny5fDw8FC+P3HiBMLCwmBvbw8bGxuEhobi1KlTtV5fzXqMjo5W1nP116FDhwDIlzP75JNP0KpVK1hYWMDf31+5T2Hjxo1o3bo1zM3NMXz4cBQVFdV6foVz585hwIABMDMzg52dHT788EOUlZUp90ulUiQmJsLW1hZ2dnb45JNPUHOap8ePH2PMmDGwtLSEQCDA0qVL1c6zZs0atG/fHqampnBycsKbb75Zb9maHcNOokiakunTpzMPDw8mk8mU2zZu3Mj4fD57+PAhY0x9Gs5ly5Yxa2tr9t///pf9888/7JNPPmHGxsbs8uXLjDHGbty4wQCw06dPM8YYq6qqYrNnz2Z//fUXu379Otu6dSszNzdn6enpjDHGSktL2ciRI9mgQYOU089WVlaq5XPnzh1mbm7OJk2axC5evMh2797N7O3tWXJysrJsoaGhzNramqWkpLDLly+zTZs2MQ6HwzIzM2utg/quRygUsi5durCpU6cyoVDISktLNebz6aefsk6dOrF9+/axa9eusQ0bNjA+n88OHTrEGGMsLy+P8fl8NnnyZPbPP/+wrVu3MicnJwaAPXr0SGNdM8bYl19+ydzd3ZXvDxw4wLZs2cJyc3NZbm4ui4mJYU5OTkwkEinTAGC7d+/W+PMoLi5Wmep38uTJzNHRkQmFQsYYY6NGjWJBQUHsyJEj7OrVq2zJkiWMz+cr6+P48eOMw+Gw1NRUdunSJbZixQpma2vLbGxsaq3jx48fMxcXFzZixAh27tw5duDAAebp6cnGjh2rTLNo0SJmY2PDdu3apbwuKysr9vrrryvTTJw4kbm6urLMzEx29uxZ9tprrzFLS0s2efJkxhhjJ06cYDwej23fvp3dvHmTnTp1iq1YsaLWcjVXFMSJ0sWLFxkAdvDgQeW2vn37snfeeUf5vmZgcXFxYZ9//rlKPr169WKTJk1ijKkHDU0mTZrE3njjDeX7sWPHqvyxasrn008/ZR07dlT5wPnqq6+YpaUlk0qljDF5EO/Tp49a2WbMmFFrWeq7HsYY8/HxUfmwqKmsrIyZmpqy7Oxsle0xMTHKupw1axbz8vJSKf+MGTN0DuI1SSQSZmVlxf73v/8pt9UVxKv7v//7P8bn89nRo0cZY4xdvXqVcTgcdvfuXZV0r7zyCps1axZjjLF33nmHDRo0SGV/VFRUnUF87dq1rEWLFqysrEy5be/evYzL5bL8/HzGGGMCgYAtXLhQuV8sFjNXV1fl70VpaSkzMTFhO3bsUKYpKipiZmZmyiD+f//3f8za2lrlA+1FRN0pRKlTp04ICgrC+vXrAcj7fo8ePYr3339fY3qRSIR79+4hODhYZXtwcDAuXrxY63m+/vpr+Pn5wcHBAZaWlvj22291nr/54sWLCAwMVFkXMTg4GGVlZbhz545yW7du3VSOEwgEKCgo0Ov11JSbm4uKigqEhYXB0tJS+dq8eTOuXbumLH9AQIBK+QMDA7U+h0JBQQFiY2PRoUMH2NjYwMbGBmVlZTrX5+nTpzFmzBh89dVX6NOnDwDg1KlTYIyhQ4cOKtdx+PBhleuoWe76ruPixYvw8fGBhYWFcltwcDBkMhkuXbqEkpISCIVClXyMjIzg5+enfH/t2jVUVVWppGnZsiU6duyofB8WFgZ3d3e0adMGo0ePxrZt21BeXq5TvTQHdHeDqIiJicFHH32Er776Chs2bIC7uzteeeWVOo/RZeHe77//HlOmTMHSpUsRGBgIKysrLFmyBH/++adO5dR0DvZvn2n17cbGxmpllclkdebd0AWdFRT57927F61atVLZx+fzVcpaFy6Xq5ZOLBarvI+OjsaDBw+wfPlyuLu7g8/nIzAwEFVVVVqXNz8/H0OHDkVMTIzy3ofiOng8HnJycsDj8VSOsbS01Po6aqqrPrWtZ23Oa2VlhVOnTuHQoUPIzMzE7NmzkZKSghMnTjTKEEhDoZY4UTFy5EjweDxs374dmzZtwrhx42r9w7K2toaLiwuOHTumsj07OxteXl4ajzl69CiCgoIwadIk9OjRA+3atVO26hRMTEwglUrrLGfnzp2RnZ2t8secnZ0NKysrtcCprYZcT21l4/P5yMvLQ7t27VRebm5uyjTHjx9XOa7mewcHB+Tn56tcY83x3UePHkV8fDwGDx6MLl26gM/no7CwUOuyVlRU4PXXX0enTp2wbNkylX09evSAVCpFQUGB2nU4OztrfR01de7cGWfOnMHjx4+V237//XdwuVzlNwqBQKCSj0QiQU5OjvJ9u3btYGxsrJLm0aNHuHz5ssq5jIyM8Oqrr2Lx4sU4e/Ysbt68iYMHD2pZO80DtcSJCktLS0RFReHTTz9FSUkJoqOj60w/ffp0JCcno23btujevTs2bNiAM2fOYNu2bRrTt2vXDps3b8b+/fvh6emJLVu24MSJE8rV2AHAw8MD+/fvx6VLl2BnZwcbGxu1fCZNmoTly5fj448/xkcffYRLly4hOTkZiYmJ4HIb3jbR9Xo0sbKywrRp0zBlyhTIZDL06dMHIpEI2dnZsLS0xNixYxEbG4ulS5ciMTEREyZMQE5OjtrIoH79+uHBgwdYvHgx3nzzTezbtw+//PILrK2tlWnatWuHLVu2wM/PDyKRCNOnT4eZmZnWZZ0wYQJu376NAwcO4MGDB8rtLVu2RIcOHfDuu+9izJgxWLp0KXr06IHCwkIcPHgQXbt2xeDBgxEfH4+goCAsXrwYw4YNQ2ZmZp2jkwDg3XffRXJyMsaOHYuUlBQ8ePAAH3/8MUaPHq1ccX7y5MlYuHAh2rdvDy8vLyxbtkw5ageQ/57GxMRg+vTpsLOzg5OTE5KSklR+9j///DOuX7+Ovn37okWLFsjIyIBMJlPpcnkhGKYrnjRl2dnZDAALDw9X21fzZptUKmVz5sxhrVq1YsbGxszHx4f98ssvyv01b6RVVFSw6OhoZmNjw2xtbdnEiRPZzJkzVfIsKChgYWFhzNLSkgFgv/32m8YbcocOHWK9evViJiYmzNnZmc2YMYOJxWLl/tDQUOVNLoXXX39dZRRETfVdD2P139hkjDGZTMZWrFjBOnbsyIyNjZmDgwMbOHAgO3z4sDLN//73P9auXTvG5/NZSEgIW79+vcqNTcYYS0tLY25ubszCwoKNGTOGff755yo3Nk+dOsX8/PwYn89n7du3Zzt37mTu7u7syy+/VKZBHTc23d3dGQC112+//cYYezqayMPDgxkbGzNnZ2c2fPhwdvbsWWX+69atY66urszMzIxFRkayL774os4bm4wxdvbsWda/f39mamrKWrZsyT744AOVkT5isZhNnjyZWVtbM1tbW5aYmMjGjBmjcsO7tLSUvffee8zc3Jw5OTmxxYsXq/zMjx49ykJDQ1mLFi2YmZkZ69atm3IU1IuE1tgkpIk4dOgQ+vfvj0ePHr1QfbakcVGfOCGENGMUxAkhpBmj7hRCCGnGqCVOCCHNGAVxQghpxiiIE0JIM0ZBnBBCmjEK4oQQ0oxRECeEkGaMgjghhDRjFMQJIaQZ+3+17FzczG0A8AAAAABJRU5ErkJggg==",
      "text/plain": [
       "<Figure size 350x300 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Results for attribute-aware setting:\n"
     ]
    },
    {
     "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 tr th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr:last-of-type th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th colspan=\"2\" halign=\"left\">accuracy</th>\n",
       "      <th colspan=\"2\" halign=\"left\">delta_eopp</th>\n",
       "      <th colspan=\"2\" halign=\"left\">delta_eopp_rms</th>\n",
       "      <th colspan=\"2\" halign=\"left\">delta_eo</th>\n",
       "      <th colspan=\"2\" halign=\"left\">delta_eo_rms</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>alpha</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>inf</th>\n",
       "      <td>0.864159</td>\n",
       "      <td>0.000346</td>\n",
       "      <td>0.406091</td>\n",
       "      <td>0.015177</td>\n",
       "      <td>0.122912</td>\n",
       "      <td>0.003319</td>\n",
       "      <td>0.406091</td>\n",
       "      <td>0.015177</td>\n",
       "      <td>0.028399</td>\n",
       "      <td>0.000665</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.250</th>\n",
       "      <td>0.864132</td>\n",
       "      <td>0.000377</td>\n",
       "      <td>0.375186</td>\n",
       "      <td>0.049533</td>\n",
       "      <td>0.119442</td>\n",
       "      <td>0.005083</td>\n",
       "      <td>0.375186</td>\n",
       "      <td>0.049533</td>\n",
       "      <td>0.027784</td>\n",
       "      <td>0.001010</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.200</th>\n",
       "      <td>0.864035</td>\n",
       "      <td>0.000485</td>\n",
       "      <td>0.335496</td>\n",
       "      <td>0.062597</td>\n",
       "      <td>0.113875</td>\n",
       "      <td>0.006392</td>\n",
       "      <td>0.335496</td>\n",
       "      <td>0.062597</td>\n",
       "      <td>0.026760</td>\n",
       "      <td>0.001274</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.150</th>\n",
       "      <td>0.863925</td>\n",
       "      <td>0.000544</td>\n",
       "      <td>0.278861</td>\n",
       "      <td>0.044036</td>\n",
       "      <td>0.102770</td>\n",
       "      <td>0.005066</td>\n",
       "      <td>0.278861</td>\n",
       "      <td>0.044036</td>\n",
       "      <td>0.024497</td>\n",
       "      <td>0.000919</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.100</th>\n",
       "      <td>0.863357</td>\n",
       "      <td>0.000601</td>\n",
       "      <td>0.221248</td>\n",
       "      <td>0.038116</td>\n",
       "      <td>0.089922</td>\n",
       "      <td>0.003718</td>\n",
       "      <td>0.221248</td>\n",
       "      <td>0.038116</td>\n",
       "      <td>0.022087</td>\n",
       "      <td>0.000749</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.050</th>\n",
       "      <td>0.862368</td>\n",
       "      <td>0.000702</td>\n",
       "      <td>0.179116</td>\n",
       "      <td>0.018658</td>\n",
       "      <td>0.074000</td>\n",
       "      <td>0.003591</td>\n",
       "      <td>0.179116</td>\n",
       "      <td>0.018658</td>\n",
       "      <td>0.018936</td>\n",
       "      <td>0.000561</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.010</th>\n",
       "      <td>0.859184</td>\n",
       "      <td>0.001075</td>\n",
       "      <td>0.176425</td>\n",
       "      <td>0.040046</td>\n",
       "      <td>0.063330</td>\n",
       "      <td>0.008319</td>\n",
       "      <td>0.177138</td>\n",
       "      <td>0.038814</td>\n",
       "      <td>0.016521</td>\n",
       "      <td>0.001592</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.001</th>\n",
       "      <td>0.854520</td>\n",
       "      <td>0.001105</td>\n",
       "      <td>0.219278</td>\n",
       "      <td>0.077400</td>\n",
       "      <td>0.064882</td>\n",
       "      <td>0.015405</td>\n",
       "      <td>0.219278</td>\n",
       "      <td>0.077400</td>\n",
       "      <td>0.016831</td>\n",
       "      <td>0.002309</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       accuracy           delta_eopp           delta_eopp_rms            \\\n",
       "           mean       std       mean       std           mean       std   \n",
       "alpha                                                                     \n",
       "inf    0.864159  0.000346   0.406091  0.015177       0.122912  0.003319   \n",
       "0.250  0.864132  0.000377   0.375186  0.049533       0.119442  0.005083   \n",
       "0.200  0.864035  0.000485   0.335496  0.062597       0.113875  0.006392   \n",
       "0.150  0.863925  0.000544   0.278861  0.044036       0.102770  0.005066   \n",
       "0.100  0.863357  0.000601   0.221248  0.038116       0.089922  0.003718   \n",
       "0.050  0.862368  0.000702   0.179116  0.018658       0.074000  0.003591   \n",
       "0.010  0.859184  0.001075   0.176425  0.040046       0.063330  0.008319   \n",
       "0.001  0.854520  0.001105   0.219278  0.077400       0.064882  0.015405   \n",
       "\n",
       "       delta_eo           delta_eo_rms            \n",
       "           mean       std         mean       std  \n",
       "alpha                                             \n",
       "inf    0.406091  0.015177     0.028399  0.000665  \n",
       "0.250  0.375186  0.049533     0.027784  0.001010  \n",
       "0.200  0.335496  0.062597     0.026760  0.001274  \n",
       "0.150  0.278861  0.044036     0.024497  0.000919  \n",
       "0.100  0.221248  0.038116     0.022087  0.000749  \n",
       "0.050  0.179116  0.018658     0.018936  0.000561  \n",
       "0.010  0.177138  0.038814     0.016521  0.001592  \n",
       "0.001  0.219278  0.077400     0.016831  0.002309  "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Results for attribute-blind setting:\n"
     ]
    },
    {
     "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 tr th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr:last-of-type th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th colspan=\"2\" halign=\"left\">accuracy</th>\n",
       "      <th colspan=\"2\" halign=\"left\">delta_eopp</th>\n",
       "      <th colspan=\"2\" halign=\"left\">delta_eopp_rms</th>\n",
       "      <th colspan=\"2\" halign=\"left\">delta_eo</th>\n",
       "      <th colspan=\"2\" halign=\"left\">delta_eo_rms</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>alpha</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>inf</th>\n",
       "      <td>0.863750</td>\n",
       "      <td>0.000425</td>\n",
       "      <td>0.403001</td>\n",
       "      <td>0.016850</td>\n",
       "      <td>0.132246</td>\n",
       "      <td>0.004998</td>\n",
       "      <td>0.403001</td>\n",
       "      <td>0.016850</td>\n",
       "      <td>0.030981</td>\n",
       "      <td>0.001145</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.300</th>\n",
       "      <td>0.863567</td>\n",
       "      <td>0.000470</td>\n",
       "      <td>0.385960</td>\n",
       "      <td>0.033164</td>\n",
       "      <td>0.127737</td>\n",
       "      <td>0.006458</td>\n",
       "      <td>0.385960</td>\n",
       "      <td>0.033164</td>\n",
       "      <td>0.029805</td>\n",
       "      <td>0.001612</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.250</th>\n",
       "      <td>0.863361</td>\n",
       "      <td>0.000546</td>\n",
       "      <td>0.348455</td>\n",
       "      <td>0.042034</td>\n",
       "      <td>0.120953</td>\n",
       "      <td>0.006435</td>\n",
       "      <td>0.349223</td>\n",
       "      <td>0.040509</td>\n",
       "      <td>0.028430</td>\n",
       "      <td>0.001496</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.200</th>\n",
       "      <td>0.863045</td>\n",
       "      <td>0.000652</td>\n",
       "      <td>0.288043</td>\n",
       "      <td>0.039708</td>\n",
       "      <td>0.104640</td>\n",
       "      <td>0.004298</td>\n",
       "      <td>0.290499</td>\n",
       "      <td>0.035483</td>\n",
       "      <td>0.025248</td>\n",
       "      <td>0.000788</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.150</th>\n",
       "      <td>0.862473</td>\n",
       "      <td>0.000730</td>\n",
       "      <td>0.232388</td>\n",
       "      <td>0.039584</td>\n",
       "      <td>0.087793</td>\n",
       "      <td>0.006314</td>\n",
       "      <td>0.232388</td>\n",
       "      <td>0.039584</td>\n",
       "      <td>0.021930</td>\n",
       "      <td>0.001090</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.100</th>\n",
       "      <td>0.861459</td>\n",
       "      <td>0.000719</td>\n",
       "      <td>0.201746</td>\n",
       "      <td>0.045609</td>\n",
       "      <td>0.078005</td>\n",
       "      <td>0.006733</td>\n",
       "      <td>0.203645</td>\n",
       "      <td>0.042595</td>\n",
       "      <td>0.020222</td>\n",
       "      <td>0.001044</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.050</th>\n",
       "      <td>0.859548</td>\n",
       "      <td>0.000703</td>\n",
       "      <td>0.196362</td>\n",
       "      <td>0.072471</td>\n",
       "      <td>0.068145</td>\n",
       "      <td>0.012652</td>\n",
       "      <td>0.196362</td>\n",
       "      <td>0.072471</td>\n",
       "      <td>0.017922</td>\n",
       "      <td>0.001855</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.010</th>\n",
       "      <td>0.853746</td>\n",
       "      <td>0.000770</td>\n",
       "      <td>0.205752</td>\n",
       "      <td>0.095283</td>\n",
       "      <td>0.068547</td>\n",
       "      <td>0.021570</td>\n",
       "      <td>0.214078</td>\n",
       "      <td>0.085093</td>\n",
       "      <td>0.017200</td>\n",
       "      <td>0.003262</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.001</th>\n",
       "      <td>0.848928</td>\n",
       "      <td>0.001201</td>\n",
       "      <td>0.208269</td>\n",
       "      <td>0.072931</td>\n",
       "      <td>0.073613</td>\n",
       "      <td>0.023006</td>\n",
       "      <td>0.211253</td>\n",
       "      <td>0.068755</td>\n",
       "      <td>0.017799</td>\n",
       "      <td>0.003534</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       accuracy           delta_eopp           delta_eopp_rms            \\\n",
       "           mean       std       mean       std           mean       std   \n",
       "alpha                                                                     \n",
       "inf    0.863750  0.000425   0.403001  0.016850       0.132246  0.004998   \n",
       "0.300  0.863567  0.000470   0.385960  0.033164       0.127737  0.006458   \n",
       "0.250  0.863361  0.000546   0.348455  0.042034       0.120953  0.006435   \n",
       "0.200  0.863045  0.000652   0.288043  0.039708       0.104640  0.004298   \n",
       "0.150  0.862473  0.000730   0.232388  0.039584       0.087793  0.006314   \n",
       "0.100  0.861459  0.000719   0.201746  0.045609       0.078005  0.006733   \n",
       "0.050  0.859548  0.000703   0.196362  0.072471       0.068145  0.012652   \n",
       "0.010  0.853746  0.000770   0.205752  0.095283       0.068547  0.021570   \n",
       "0.001  0.848928  0.001201   0.208269  0.072931       0.073613  0.023006   \n",
       "\n",
       "       delta_eo           delta_eo_rms            \n",
       "           mean       std         mean       std  \n",
       "alpha                                             \n",
       "inf    0.403001  0.016850     0.030981  0.001145  \n",
       "0.300  0.385960  0.033164     0.029805  0.001612  \n",
       "0.250  0.349223  0.040509     0.028430  0.001496  \n",
       "0.200  0.290499  0.035483     0.025248  0.000788  \n",
       "0.150  0.232388  0.039584     0.021930  0.001090  \n",
       "0.100  0.203645  0.042595     0.020222  0.001044  \n",
       "0.050  0.196362  0.072471     0.017922  0.001855  \n",
       "0.010  0.214078  0.085093     0.017200  0.003262  \n",
       "0.001  0.211253  0.068755     0.017799  0.003534  "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Post-processing for equalized odds\n",
    "\n",
    "df_aware = utils.postprocess_and_evaluate(\n",
    "    [np.inf, 0.25, 0.2, 0.15, 0.1, 0.05, 0.01, 0.001],\n",
    "    seeds,\n",
    "    'eo',\n",
    "    ['accuracy', 'delta_eopp', 'delta_eopp_rms', 'delta_eo', 'delta_eo_rms'],\n",
    "    **postprocess_kwargs,\n",
    ")\n",
    "df_blind = utils.postprocess_and_evaluate(\n",
    "    [np.inf, 0.3, 0.25, 0.2, 0.15, 0.1, 0.05, 0.01, 0.001],\n",
    "    seeds,\n",
    "    'eo',\n",
    "    ['accuracy', 'delta_eopp', 'delta_eopp_rms', 'delta_eo', 'delta_eo_rms'],\n",
    "    **postprocess_u_kwargs,\n",
    ")\n",
    "\n",
    "fig, ax = plt.subplots(1, 1, figsize=(3.5, 3))\n",
    "utils.plot_results(ax, df_aware, 'delta_eo', 'accuracy', label='aware')\n",
    "utils.plot_results(ax, df_blind, 'delta_eo', 'accuracy', label='unaware')\n",
    "ax.set_xlabel(\"Violation of equalized odds\")\n",
    "ax.set_ylabel(\"Accuracy\")\n",
    "ax.set_title(\"BiasBios (BERT)\")\n",
    "ax.legend()\n",
    "plt.show()\n",
    "\n",
    "print(\"Results for attribute-aware setting:\")\n",
    "display(df_aware)\n",
    "print(\"Results for attribute-blind setting:\")\n",
    "display(df_blind)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8fc53533",
   "metadata": {},
   "source": [
    "## Fine-tune BERT and calibrate"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "64e8e531",
   "metadata": {},
   "outputs": [],
   "source": [
    "split_dataset = tokenized_dataset_pretrain_u.train_test_split(\n",
    "    test_size=split_ratio_calib, seed=seed_pretrain)\n",
    "tokenized_dataset_train_u = split_dataset['train']\n",
    "tokenized_dataset_calib = split_dataset['test']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "22855621",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of the model checkpoint at bert-base-uncased were not used when initializing BertForSequenceClassification: ['cls.predictions.bias', 'cls.predictions.decoder.weight', 'cls.predictions.transform.LayerNorm.weight', 'cls.seq_relationship.weight', 'cls.predictions.transform.dense.bias', 'cls.predictions.transform.dense.weight', 'cls.predictions.transform.LayerNorm.bias', 'cls.seq_relationship.bias']\n",
      "- This IS expected if you are initializing BertForSequenceClassification from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n",
      "- This IS NOT expected if you are initializing BertForSequenceClassification from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of BertForSequenceClassification were not initialized from the model checkpoint at bert-base-uncased and are newly initialized: ['classifier.weight', 'classifier.bias']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "8144689b31f34b66a32549f2d1c3a76a",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "train epoch 1:   0%|          | 0/5422 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "You're using a BertTokenizerFast tokenizer. Please note that with a fast tokenizer, using the `__call__` method is faster than using a method to encode the text followed by a call to the `pad` method to get a padded encoding.\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "1230b8e2d9c5446a865266e87cfac353",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "test 1:   0%|          | 0/455 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 1/3: loss=0.0164, acc=0.8488\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "f80fe23f354b4e8fb0a11321d1a4a52a",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "train epoch 2:   0%|          | 0/5422 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "54923e0d53a34eaa80ca9f575934f0c3",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "test 2:   0%|          | 0/455 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 2/3: loss=0.0156, acc=0.8536\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "bd4b9e3232a047b5b8953aaf5f6cce73",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "train epoch 3:   0%|          | 0/5422 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "4555e744dd7f4d8e86c6d61d67916262",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "test 3:   0%|          | 0/455 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 3/3: loss=0.0161, acc=0.8590\n"
     ]
    }
   ],
   "source": [
    "## (Pre-)train predictors\n",
    "\n",
    "batch_size = 32\n",
    "n_epochs = 3\n",
    "lr = 2e-5\n",
    "warmup_ratio = 0.1\n",
    "weight_decay = 0.01\n",
    "max_grad_norm = 1.0\n",
    "\n",
    "# Train attribute-blind p(A, Y | X) predictor\n",
    "transformers.set_seed(seed_pretrain)\n",
    "model_u = transformers.AutoModelForSequenceClassification.from_pretrained(\n",
    "    model_name, num_labels=n_groups * n_classes).to(device)\n",
    "train(\n",
    "    model_u,\n",
    "    tokenized_dataset_train_u,\n",
    "    tokenized_dataset_u_.train_test_split(test_size=0.1,\n",
    "                                          seed=seed_pretrain)['test'],\n",
    "    'labels_ay',\n",
    "    batch_size,\n",
    "    n_epochs,\n",
    "    lr,\n",
    "    warmup_ratio,\n",
    "    weight_decay,\n",
    "    max_grad_norm,\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "3e44a744",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "9e5e0cd2171343359712860019916a77",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "inference:   0%|          | 0/1138 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "ea0cf2f0930e44f9a09d27e69af20fe7",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "inference:   0%|          | 0/581 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Precompute P(A, Y | X) on (test + post-processing) set for blind setting\n",
    "raw_probas_ay_u_ = predict_probas(model_u, tokenized_dataset_u_, 128)\n",
    "raw_probas_ay_u_calib = predict_probas(model_u, tokenized_dataset_calib, 128)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "6436838f",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Train p(A, Y | X) calibrator\n",
    "calib = BinningCalibrator(n_bins_calib, random_state=seed_pretrain)\n",
    "calib.fit(raw_probas_ay_u_calib, np.array(tokenized_dataset_calib['labels_ay']));"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "3701b23d",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Calibrate P(A, Y | X) on (test + post-processing) set\n",
    "probas_ay_u_ = calib.predict_proba(raw_probas_ay_u_).reshape(\n",
    "    -1, n_groups, n_classes)\n",
    "\n",
    "postprocess_u_kwargs = {\n",
    "    'n_test': n_test,\n",
    "    'n_classes': n_classes,\n",
    "    'n_groups': n_groups,\n",
    "    'labels': labels_,\n",
    "    'groups': groups_,\n",
    "    'probas_ay': probas_ay_u_,\n",
    "    'max_workers': max_workers,\n",
    "}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "e3669ff5",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "60ce5722ebdc4fad92a8a1659656dbd7",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/65 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 350x300 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Results for attribute-blind setting:\n"
     ]
    },
    {
     "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 tr th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr:last-of-type th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th colspan=\"2\" halign=\"left\">accuracy</th>\n",
       "      <th colspan=\"2\" halign=\"left\">delta_sp</th>\n",
       "      <th colspan=\"2\" halign=\"left\">delta_sp_rms</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>alpha</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>inf</th>\n",
       "      <td>0.856201</td>\n",
       "      <td>0.000412</td>\n",
       "      <td>0.080079</td>\n",
       "      <td>0.000399</td>\n",
       "      <td>0.022518</td>\n",
       "      <td>0.000081</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.080</th>\n",
       "      <td>0.856174</td>\n",
       "      <td>0.000419</td>\n",
       "      <td>0.079654</td>\n",
       "      <td>0.001322</td>\n",
       "      <td>0.022471</td>\n",
       "      <td>0.000179</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.070</th>\n",
       "      <td>0.855618</td>\n",
       "      <td>0.000439</td>\n",
       "      <td>0.071421</td>\n",
       "      <td>0.002682</td>\n",
       "      <td>0.021639</td>\n",
       "      <td>0.000271</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.060</th>\n",
       "      <td>0.854761</td>\n",
       "      <td>0.000548</td>\n",
       "      <td>0.060696</td>\n",
       "      <td>0.003459</td>\n",
       "      <td>0.020846</td>\n",
       "      <td>0.000260</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.050</th>\n",
       "      <td>0.853227</td>\n",
       "      <td>0.000578</td>\n",
       "      <td>0.050088</td>\n",
       "      <td>0.003026</td>\n",
       "      <td>0.020260</td>\n",
       "      <td>0.000179</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.040</th>\n",
       "      <td>0.851273</td>\n",
       "      <td>0.000739</td>\n",
       "      <td>0.041452</td>\n",
       "      <td>0.002077</td>\n",
       "      <td>0.019705</td>\n",
       "      <td>0.000104</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.030</th>\n",
       "      <td>0.848423</td>\n",
       "      <td>0.000862</td>\n",
       "      <td>0.031869</td>\n",
       "      <td>0.000890</td>\n",
       "      <td>0.018322</td>\n",
       "      <td>0.000159</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.020</th>\n",
       "      <td>0.841464</td>\n",
       "      <td>0.000794</td>\n",
       "      <td>0.023544</td>\n",
       "      <td>0.002389</td>\n",
       "      <td>0.014171</td>\n",
       "      <td>0.000387</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.015</th>\n",
       "      <td>0.834723</td>\n",
       "      <td>0.001319</td>\n",
       "      <td>0.018824</td>\n",
       "      <td>0.002135</td>\n",
       "      <td>0.011233</td>\n",
       "      <td>0.000511</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.010</th>\n",
       "      <td>0.824125</td>\n",
       "      <td>0.001653</td>\n",
       "      <td>0.013807</td>\n",
       "      <td>0.001823</td>\n",
       "      <td>0.008082</td>\n",
       "      <td>0.000403</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.005</th>\n",
       "      <td>0.809332</td>\n",
       "      <td>0.001244</td>\n",
       "      <td>0.009062</td>\n",
       "      <td>0.001059</td>\n",
       "      <td>0.004660</td>\n",
       "      <td>0.000304</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.002</th>\n",
       "      <td>0.798082</td>\n",
       "      <td>0.001043</td>\n",
       "      <td>0.005993</td>\n",
       "      <td>0.001597</td>\n",
       "      <td>0.002580</td>\n",
       "      <td>0.000352</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.001</th>\n",
       "      <td>0.793579</td>\n",
       "      <td>0.001665</td>\n",
       "      <td>0.005616</td>\n",
       "      <td>0.001518</td>\n",
       "      <td>0.002089</td>\n",
       "      <td>0.000490</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       accuracy            delta_sp           delta_sp_rms          \n",
       "           mean       std      mean       std         mean       std\n",
       "alpha                                                               \n",
       "inf    0.856201  0.000412  0.080079  0.000399     0.022518  0.000081\n",
       "0.080  0.856174  0.000419  0.079654  0.001322     0.022471  0.000179\n",
       "0.070  0.855618  0.000439  0.071421  0.002682     0.021639  0.000271\n",
       "0.060  0.854761  0.000548  0.060696  0.003459     0.020846  0.000260\n",
       "0.050  0.853227  0.000578  0.050088  0.003026     0.020260  0.000179\n",
       "0.040  0.851273  0.000739  0.041452  0.002077     0.019705  0.000104\n",
       "0.030  0.848423  0.000862  0.031869  0.000890     0.018322  0.000159\n",
       "0.020  0.841464  0.000794  0.023544  0.002389     0.014171  0.000387\n",
       "0.015  0.834723  0.001319  0.018824  0.002135     0.011233  0.000511\n",
       "0.010  0.824125  0.001653  0.013807  0.001823     0.008082  0.000403\n",
       "0.005  0.809332  0.001244  0.009062  0.001059     0.004660  0.000304\n",
       "0.002  0.798082  0.001043  0.005993  0.001597     0.002580  0.000352\n",
       "0.001  0.793579  0.001665  0.005616  0.001518     0.002089  0.000490"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Post-processing for statistical parity\n",
    "\n",
    "df_blind = utils.postprocess_and_evaluate(\n",
    "    [\n",
    "        np.inf, 0.08, 0.07, 0.06, 0.05, 0.04, 0.03, 0.02, 0.015, 0.01, 0.005,\n",
    "        0.002, 0.001\n",
    "    ],\n",
    "    seeds,\n",
    "    'sp',\n",
    "    ['accuracy', 'delta_sp', 'delta_sp_rms'],\n",
    "    **postprocess_u_kwargs,\n",
    ")\n",
    "\n",
    "fig, ax = plt.subplots(1, 1, figsize=(3.5, 3))\n",
    "utils.plot_results(ax, df_blind, 'delta_sp', 'accuracy', label='unaware')\n",
    "ax.set_xlabel(\"Violation of statistical parity\")\n",
    "ax.set_ylabel(\"Accuracy\")\n",
    "ax.set_title(\"BiasBios (BERT, cal.)\")\n",
    "ax.legend()\n",
    "plt.show()\n",
    "\n",
    "print(\"Results for attribute-blind setting:\")\n",
    "display(df_blind)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "29c2241c",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "4c04523bf326411caed18467b9679f9e",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/55 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 350x300 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Results for attribute-blind setting:\n"
     ]
    },
    {
     "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 tr th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr:last-of-type th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th colspan=\"2\" halign=\"left\">accuracy</th>\n",
       "      <th colspan=\"2\" halign=\"left\">delta_eopp</th>\n",
       "      <th colspan=\"2\" halign=\"left\">delta_eopp_rms</th>\n",
       "      <th colspan=\"2\" halign=\"left\">delta_eo</th>\n",
       "      <th colspan=\"2\" halign=\"left\">delta_eo_rms</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>alpha</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>inf</th>\n",
       "      <td>0.856201</td>\n",
       "      <td>0.000412</td>\n",
       "      <td>0.859382</td>\n",
       "      <td>0.006759</td>\n",
       "      <td>0.372853</td>\n",
       "      <td>0.001254</td>\n",
       "      <td>0.859382</td>\n",
       "      <td>0.006759</td>\n",
       "      <td>0.090766</td>\n",
       "      <td>0.000399</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.800</th>\n",
       "      <td>0.856001</td>\n",
       "      <td>0.000411</td>\n",
       "      <td>0.805001</td>\n",
       "      <td>0.024732</td>\n",
       "      <td>0.364264</td>\n",
       "      <td>0.003524</td>\n",
       "      <td>0.805001</td>\n",
       "      <td>0.024732</td>\n",
       "      <td>0.085535</td>\n",
       "      <td>0.001062</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.500</th>\n",
       "      <td>0.851816</td>\n",
       "      <td>0.000271</td>\n",
       "      <td>0.577219</td>\n",
       "      <td>0.024269</td>\n",
       "      <td>0.257541</td>\n",
       "      <td>0.005524</td>\n",
       "      <td>0.579388</td>\n",
       "      <td>0.025368</td>\n",
       "      <td>0.063110</td>\n",
       "      <td>0.002365</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.300</th>\n",
       "      <td>0.845615</td>\n",
       "      <td>0.000727</td>\n",
       "      <td>0.369879</td>\n",
       "      <td>0.024024</td>\n",
       "      <td>0.165975</td>\n",
       "      <td>0.006183</td>\n",
       "      <td>0.753323</td>\n",
       "      <td>0.030323</td>\n",
       "      <td>0.059263</td>\n",
       "      <td>0.001235</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.250</th>\n",
       "      <td>0.843897</td>\n",
       "      <td>0.000659</td>\n",
       "      <td>0.321323</td>\n",
       "      <td>0.023465</td>\n",
       "      <td>0.146300</td>\n",
       "      <td>0.004235</td>\n",
       "      <td>0.793344</td>\n",
       "      <td>0.029298</td>\n",
       "      <td>0.060820</td>\n",
       "      <td>0.000810</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.200</th>\n",
       "      <td>0.841820</td>\n",
       "      <td>0.000614</td>\n",
       "      <td>0.272476</td>\n",
       "      <td>0.029137</td>\n",
       "      <td>0.124053</td>\n",
       "      <td>0.005527</td>\n",
       "      <td>0.799225</td>\n",
       "      <td>0.022594</td>\n",
       "      <td>0.061012</td>\n",
       "      <td>0.000847</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.150</th>\n",
       "      <td>0.838459</td>\n",
       "      <td>0.000301</td>\n",
       "      <td>0.225314</td>\n",
       "      <td>0.040660</td>\n",
       "      <td>0.100598</td>\n",
       "      <td>0.007214</td>\n",
       "      <td>0.800179</td>\n",
       "      <td>0.021908</td>\n",
       "      <td>0.060653</td>\n",
       "      <td>0.001081</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.100</th>\n",
       "      <td>0.834167</td>\n",
       "      <td>0.000714</td>\n",
       "      <td>0.176605</td>\n",
       "      <td>0.042031</td>\n",
       "      <td>0.076713</td>\n",
       "      <td>0.007699</td>\n",
       "      <td>0.800179</td>\n",
       "      <td>0.021908</td>\n",
       "      <td>0.061203</td>\n",
       "      <td>0.000602</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.050</th>\n",
       "      <td>0.827103</td>\n",
       "      <td>0.001196</td>\n",
       "      <td>0.137344</td>\n",
       "      <td>0.020677</td>\n",
       "      <td>0.052093</td>\n",
       "      <td>0.006318</td>\n",
       "      <td>0.801108</td>\n",
       "      <td>0.022548</td>\n",
       "      <td>0.062064</td>\n",
       "      <td>0.001990</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.010</th>\n",
       "      <td>0.819096</td>\n",
       "      <td>0.001757</td>\n",
       "      <td>0.109834</td>\n",
       "      <td>0.024795</td>\n",
       "      <td>0.037845</td>\n",
       "      <td>0.003840</td>\n",
       "      <td>0.802063</td>\n",
       "      <td>0.021952</td>\n",
       "      <td>0.067667</td>\n",
       "      <td>0.003738</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.001</th>\n",
       "      <td>0.817040</td>\n",
       "      <td>0.001722</td>\n",
       "      <td>0.127311</td>\n",
       "      <td>0.025638</td>\n",
       "      <td>0.038456</td>\n",
       "      <td>0.004429</td>\n",
       "      <td>0.803543</td>\n",
       "      <td>0.020597</td>\n",
       "      <td>0.068213</td>\n",
       "      <td>0.003336</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       accuracy           delta_eopp           delta_eopp_rms            \\\n",
       "           mean       std       mean       std           mean       std   \n",
       "alpha                                                                     \n",
       "inf    0.856201  0.000412   0.859382  0.006759       0.372853  0.001254   \n",
       "0.800  0.856001  0.000411   0.805001  0.024732       0.364264  0.003524   \n",
       "0.500  0.851816  0.000271   0.577219  0.024269       0.257541  0.005524   \n",
       "0.300  0.845615  0.000727   0.369879  0.024024       0.165975  0.006183   \n",
       "0.250  0.843897  0.000659   0.321323  0.023465       0.146300  0.004235   \n",
       "0.200  0.841820  0.000614   0.272476  0.029137       0.124053  0.005527   \n",
       "0.150  0.838459  0.000301   0.225314  0.040660       0.100598  0.007214   \n",
       "0.100  0.834167  0.000714   0.176605  0.042031       0.076713  0.007699   \n",
       "0.050  0.827103  0.001196   0.137344  0.020677       0.052093  0.006318   \n",
       "0.010  0.819096  0.001757   0.109834  0.024795       0.037845  0.003840   \n",
       "0.001  0.817040  0.001722   0.127311  0.025638       0.038456  0.004429   \n",
       "\n",
       "       delta_eo           delta_eo_rms            \n",
       "           mean       std         mean       std  \n",
       "alpha                                             \n",
       "inf    0.859382  0.006759     0.090766  0.000399  \n",
       "0.800  0.805001  0.024732     0.085535  0.001062  \n",
       "0.500  0.579388  0.025368     0.063110  0.002365  \n",
       "0.300  0.753323  0.030323     0.059263  0.001235  \n",
       "0.250  0.793344  0.029298     0.060820  0.000810  \n",
       "0.200  0.799225  0.022594     0.061012  0.000847  \n",
       "0.150  0.800179  0.021908     0.060653  0.001081  \n",
       "0.100  0.800179  0.021908     0.061203  0.000602  \n",
       "0.050  0.801108  0.022548     0.062064  0.001990  \n",
       "0.010  0.802063  0.021952     0.067667  0.003738  \n",
       "0.001  0.803543  0.020597     0.068213  0.003336  "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Post-processing for equal opportunity\n",
    "\n",
    "df_blind = utils.postprocess_and_evaluate(\n",
    "    [np.inf, 0.8, 0.5, 0.3, 0.25, 0.2, 0.15, 0.1, 0.05, 0.01, 0.001],\n",
    "    seeds,\n",
    "    'eopp',\n",
    "    ['accuracy', 'delta_eopp', 'delta_eopp_rms', 'delta_eo', 'delta_eo_rms'],\n",
    "    **postprocess_u_kwargs,\n",
    ")\n",
    "\n",
    "fig, ax = plt.subplots(1, 1, figsize=(3.5, 3))\n",
    "utils.plot_results(ax,\n",
    "                   df_blind,\n",
    "                   'delta_eopp_rms',\n",
    "                   'accuracy',\n",
    "                   label='unaware')\n",
    "ax.set_xlabel(\"Violation of equal opportunity (rms)\")\n",
    "ax.set_ylabel(\"Accuracy\")\n",
    "ax.set_title(\"BiasBios (BERT, cal.)\")\n",
    "ax.legend()\n",
    "plt.show()\n",
    "\n",
    "print(\"Results for attribute-blind setting:\")\n",
    "display(df_blind)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "746c29e3",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 350x300 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Results for attribute-blind setting:\n"
     ]
    },
    {
     "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 tr th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr:last-of-type th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th colspan=\"2\" halign=\"left\">accuracy</th>\n",
       "      <th colspan=\"2\" halign=\"left\">delta_eopp</th>\n",
       "      <th colspan=\"2\" halign=\"left\">delta_eopp_rms</th>\n",
       "      <th colspan=\"2\" halign=\"left\">delta_eo</th>\n",
       "      <th colspan=\"2\" halign=\"left\">delta_eo_rms</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>alpha</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>inf</th>\n",
       "      <td>0.856201</td>\n",
       "      <td>0.000412</td>\n",
       "      <td>0.859382</td>\n",
       "      <td>0.006759</td>\n",
       "      <td>0.372853</td>\n",
       "      <td>0.001254</td>\n",
       "      <td>0.859382</td>\n",
       "      <td>0.006759</td>\n",
       "      <td>0.090766</td>\n",
       "      <td>0.000399</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.800</th>\n",
       "      <td>0.856001</td>\n",
       "      <td>0.000411</td>\n",
       "      <td>0.805001</td>\n",
       "      <td>0.024732</td>\n",
       "      <td>0.364264</td>\n",
       "      <td>0.003524</td>\n",
       "      <td>0.805001</td>\n",
       "      <td>0.024732</td>\n",
       "      <td>0.085523</td>\n",
       "      <td>0.001063</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.500</th>\n",
       "      <td>0.851814</td>\n",
       "      <td>0.000274</td>\n",
       "      <td>0.577219</td>\n",
       "      <td>0.024269</td>\n",
       "      <td>0.257508</td>\n",
       "      <td>0.005500</td>\n",
       "      <td>0.582205</td>\n",
       "      <td>0.028011</td>\n",
       "      <td>0.063123</td>\n",
       "      <td>0.002379</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.300</th>\n",
       "      <td>0.845093</td>\n",
       "      <td>0.000587</td>\n",
       "      <td>0.377735</td>\n",
       "      <td>0.029784</td>\n",
       "      <td>0.165059</td>\n",
       "      <td>0.006222</td>\n",
       "      <td>0.462466</td>\n",
       "      <td>0.025308</td>\n",
       "      <td>0.044513</td>\n",
       "      <td>0.001613</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.250</th>\n",
       "      <td>0.843182</td>\n",
       "      <td>0.000674</td>\n",
       "      <td>0.341663</td>\n",
       "      <td>0.022674</td>\n",
       "      <td>0.144997</td>\n",
       "      <td>0.006943</td>\n",
       "      <td>0.413338</td>\n",
       "      <td>0.042185</td>\n",
       "      <td>0.040524</td>\n",
       "      <td>0.002058</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.200</th>\n",
       "      <td>0.841416</td>\n",
       "      <td>0.000690</td>\n",
       "      <td>0.307541</td>\n",
       "      <td>0.025356</td>\n",
       "      <td>0.127835</td>\n",
       "      <td>0.007207</td>\n",
       "      <td>0.360343</td>\n",
       "      <td>0.038568</td>\n",
       "      <td>0.036794</td>\n",
       "      <td>0.001869</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.150</th>\n",
       "      <td>0.838071</td>\n",
       "      <td>0.000388</td>\n",
       "      <td>0.251290</td>\n",
       "      <td>0.029484</td>\n",
       "      <td>0.102666</td>\n",
       "      <td>0.005007</td>\n",
       "      <td>0.313076</td>\n",
       "      <td>0.028181</td>\n",
       "      <td>0.031788</td>\n",
       "      <td>0.001554</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.100</th>\n",
       "      <td>0.833318</td>\n",
       "      <td>0.000575</td>\n",
       "      <td>0.173101</td>\n",
       "      <td>0.041253</td>\n",
       "      <td>0.075173</td>\n",
       "      <td>0.008051</td>\n",
       "      <td>0.257253</td>\n",
       "      <td>0.032722</td>\n",
       "      <td>0.027212</td>\n",
       "      <td>0.001822</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.050</th>\n",
       "      <td>0.825855</td>\n",
       "      <td>0.001298</td>\n",
       "      <td>0.136332</td>\n",
       "      <td>0.031932</td>\n",
       "      <td>0.051252</td>\n",
       "      <td>0.008171</td>\n",
       "      <td>0.220383</td>\n",
       "      <td>0.032116</td>\n",
       "      <td>0.021707</td>\n",
       "      <td>0.001722</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.010</th>\n",
       "      <td>0.811208</td>\n",
       "      <td>0.001564</td>\n",
       "      <td>0.106063</td>\n",
       "      <td>0.043276</td>\n",
       "      <td>0.037212</td>\n",
       "      <td>0.009782</td>\n",
       "      <td>0.188214</td>\n",
       "      <td>0.053383</td>\n",
       "      <td>0.018450</td>\n",
       "      <td>0.001924</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.001</th>\n",
       "      <td>0.799831</td>\n",
       "      <td>0.001414</td>\n",
       "      <td>0.097795</td>\n",
       "      <td>0.026004</td>\n",
       "      <td>0.038196</td>\n",
       "      <td>0.007396</td>\n",
       "      <td>0.208036</td>\n",
       "      <td>0.063534</td>\n",
       "      <td>0.018834</td>\n",
       "      <td>0.001613</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       accuracy           delta_eopp           delta_eopp_rms            \\\n",
       "           mean       std       mean       std           mean       std   \n",
       "alpha                                                                     \n",
       "inf    0.856201  0.000412   0.859382  0.006759       0.372853  0.001254   \n",
       "0.800  0.856001  0.000411   0.805001  0.024732       0.364264  0.003524   \n",
       "0.500  0.851814  0.000274   0.577219  0.024269       0.257508  0.005500   \n",
       "0.300  0.845093  0.000587   0.377735  0.029784       0.165059  0.006222   \n",
       "0.250  0.843182  0.000674   0.341663  0.022674       0.144997  0.006943   \n",
       "0.200  0.841416  0.000690   0.307541  0.025356       0.127835  0.007207   \n",
       "0.150  0.838071  0.000388   0.251290  0.029484       0.102666  0.005007   \n",
       "0.100  0.833318  0.000575   0.173101  0.041253       0.075173  0.008051   \n",
       "0.050  0.825855  0.001298   0.136332  0.031932       0.051252  0.008171   \n",
       "0.010  0.811208  0.001564   0.106063  0.043276       0.037212  0.009782   \n",
       "0.001  0.799831  0.001414   0.097795  0.026004       0.038196  0.007396   \n",
       "\n",
       "       delta_eo           delta_eo_rms            \n",
       "           mean       std         mean       std  \n",
       "alpha                                             \n",
       "inf    0.859382  0.006759     0.090766  0.000399  \n",
       "0.800  0.805001  0.024732     0.085523  0.001063  \n",
       "0.500  0.582205  0.028011     0.063123  0.002379  \n",
       "0.300  0.462466  0.025308     0.044513  0.001613  \n",
       "0.250  0.413338  0.042185     0.040524  0.002058  \n",
       "0.200  0.360343  0.038568     0.036794  0.001869  \n",
       "0.150  0.313076  0.028181     0.031788  0.001554  \n",
       "0.100  0.257253  0.032722     0.027212  0.001822  \n",
       "0.050  0.220383  0.032116     0.021707  0.001722  \n",
       "0.010  0.188214  0.053383     0.018450  0.001924  \n",
       "0.001  0.208036  0.063534     0.018834  0.001613  "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Post-processing for equalized odds\n",
    "\n",
    "df_blind = utils.postprocess_and_evaluate(\n",
    "    [np.inf, 0.8, 0.5, 0.3, 0.25, 0.2, 0.15, 0.1, 0.05, 0.01, 0.001],\n",
    "    seeds,\n",
    "    'eo',\n",
    "    ['accuracy', 'delta_eopp', 'delta_eopp_rms', 'delta_eo', 'delta_eo_rms'],\n",
    "    **postprocess_u_kwargs,\n",
    ")\n",
    "\n",
    "fig, ax = plt.subplots(1, 1, figsize=(3.5, 3))\n",
    "utils.plot_results(ax, df_blind, 'delta_eo', 'accuracy', label='unaware')\n",
    "ax.set_xlabel(\"Violation of equalized odds\")\n",
    "ax.set_ylabel(\"Accuracy\")\n",
    "ax.set_title(\"BiasBios (BERT, cal.)\")\n",
    "ax.legend()\n",
    "plt.show()\n",
    "\n",
    "print(\"Results for attribute-blind setting:\")\n",
    "display(df_blind)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "dfcf95c0",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "fair",
   "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": "3.10.10"
  },
  "papermill": {
   "default_parameters": {},
   "duration": 1588.675969,
   "end_time": "2024-03-07T18:36:05.647968",
   "environment_variables": {},
   "exception": null,
   "input_path": "biasbios.ipynb",
   "output_path": "biasbios_.ipynb",
   "parameters": {},
   "start_time": "2024-03-07T18:09:36.971999",
   "version": "2.3.4"
  },
  "vscode": {
   "interpreter": {
    "hash": "aac456e002ecb64114f81d56e2b750c6ea20a76d8646af90a7cd3cc27dac8d4c"
   }
  },
  "widgets": {
   "application/vnd.jupyter.widget-state+json": {
    "state": {
     "074862f4cb0a448981fb52310a0adad3": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "2.0.0",
      "model_name": "HTMLStyleModel",
      "state": {
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "2.0.0",
       "_model_name": "HTMLStyleModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "2.0.0",
       "_view_name": "StyleView",
       "background": null,
       "description_width": "",
       "font_size": null,
       "text_color": null
      }
     },
     "096e5eadf6e54272be7f7d0b0f3d3228": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "2.0.0",
      "model_name": "FloatProgressModel",
      "state": {
       "_dom_classes": [],
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "2.0.0",
       "_model_name": "FloatProgressModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/controls",
       "_view_module_version": "2.0.0",
       "_view_name": "ProgressView",
       "bar_style": "success",
       "description": "",
       "description_allow_html": false,
       "layout": "IPY_MODEL_a134a036991c45899d811dab10be5eef",
       "max": 45,
       "min": 0,
       "orientation": "horizontal",
       "style": "IPY_MODEL_d9433aaf1faa45379f5b9cb675471be9",
       "tabbable": null,
       "tooltip": null,
       "value": 45
      }
     },
     "0e108582dbe041c0b51c327d89d6f369": {
      "model_module": "@jupyter-widgets/base",
      "model_module_version": "2.0.0",
      "model_name": "LayoutModel",
      "state": {
       "_model_module": "@jupyter-widgets/base",
       "_model_module_version": "2.0.0",
       "_model_name": "LayoutModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "2.0.0",
       "_view_name": "LayoutView",
       "align_content": null,
       "align_items": null,
       "align_self": null,
       "border_bottom": null,
       "border_left": null,
       "border_right": null,
       "border_top": null,
       "bottom": null,
       "display": null,
       "flex": null,
       "flex_flow": null,
       "grid_area": null,
       "grid_auto_columns": null,
       "grid_auto_flow": null,
       "grid_auto_rows": null,
       "grid_column": null,
       "grid_gap": null,
       "grid_row": null,
       "grid_template_areas": null,
       "grid_template_columns": null,
       "grid_template_rows": null,
       "height": null,
       "justify_content": null,
       "justify_items": null,
       "left": null,
       "margin": null,
       "max_height": null,
       "max_width": null,
       "min_height": null,
       "min_width": null,
       "object_fit": null,
       "object_position": null,
       "order": null,
       "overflow": null,
       "padding": null,
       "right": null,
       "top": null,
       "visibility": null,
       "width": null
      }
     },
     "144f5045294f40a7b9dd37503b2b65c6": {
      "model_module": "@jupyter-widgets/base",
      "model_module_version": "2.0.0",
      "model_name": "LayoutModel",
      "state": {
       "_model_module": "@jupyter-widgets/base",
       "_model_module_version": "2.0.0",
       "_model_name": "LayoutModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "2.0.0",
       "_view_name": "LayoutView",
       "align_content": null,
       "align_items": null,
       "align_self": null,
       "border_bottom": null,
       "border_left": null,
       "border_right": null,
       "border_top": null,
       "bottom": null,
       "display": null,
       "flex": null,
       "flex_flow": null,
       "grid_area": null,
       "grid_auto_columns": null,
       "grid_auto_flow": null,
       "grid_auto_rows": null,
       "grid_column": null,
       "grid_gap": null,
       "grid_row": null,
       "grid_template_areas": null,
       "grid_template_columns": null,
       "grid_template_rows": null,
       "height": null,
       "justify_content": null,
       "justify_items": null,
       "left": null,
       "margin": null,
       "max_height": null,
       "max_width": null,
       "min_height": null,
       "min_width": null,
       "object_fit": null,
       "object_position": null,
       "order": null,
       "overflow": null,
       "padding": null,
       "right": null,
       "top": null,
       "visibility": null,
       "width": null
      }
     },
     "165f28820a184775b076b222200a30ed": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "2.0.0",
      "model_name": "ProgressStyleModel",
      "state": {
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "2.0.0",
       "_model_name": "ProgressStyleModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "2.0.0",
       "_view_name": "StyleView",
       "bar_color": null,
       "description_width": ""
      }
     },
     "186223a8b408471fb54f2e0b043f36f4": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "2.0.0",
      "model_name": "HTMLStyleModel",
      "state": {
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "2.0.0",
       "_model_name": "HTMLStyleModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "2.0.0",
       "_view_name": "StyleView",
       "background": null,
       "description_width": "",
       "font_size": null,
       "text_color": null
      }
     },
     "1867f79c6207497c8fac6109fb631eee": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "2.0.0",
      "model_name": "HTMLModel",
      "state": {
       "_dom_classes": [],
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "2.0.0",
       "_model_name": "HTMLModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/controls",
       "_view_module_version": "2.0.0",
       "_view_name": "HTMLView",
       "description": "",
       "description_allow_html": false,
       "layout": "IPY_MODEL_64615f48fd4444c5a4069793a040cb63",
       "placeholder": "​",
       "style": "IPY_MODEL_074862f4cb0a448981fb52310a0adad3",
       "tabbable": null,
       "tooltip": null,
       "value": "100%"
      }
     },
     "28c557faa8d244c4b16afc9c4ccb0354": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "2.0.0",
      "model_name": "FloatProgressModel",
      "state": {
       "_dom_classes": [],
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "2.0.0",
       "_model_name": "FloatProgressModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/controls",
       "_view_module_version": "2.0.0",
       "_view_name": "ProgressView",
       "bar_style": "success",
       "description": "",
       "description_allow_html": false,
       "layout": "IPY_MODEL_b8547e5480aa4499b855264f9418f0b3",
       "max": 65,
       "min": 0,
       "orientation": "horizontal",
       "style": "IPY_MODEL_165f28820a184775b076b222200a30ed",
       "tabbable": null,
       "tooltip": null,
       "value": 65
      }
     },
     "2e669b6233ef4dc38831fd0fce9c5279": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "2.0.0",
      "model_name": "HBoxModel",
      "state": {
       "_dom_classes": [],
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "2.0.0",
       "_model_name": "HBoxModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/controls",
       "_view_module_version": "2.0.0",
       "_view_name": "HBoxView",
       "box_style": "",
       "children": [
        "IPY_MODEL_1867f79c6207497c8fac6109fb631eee",
        "IPY_MODEL_a143ad464ade43aeb4a8b05d47fdc8e6",
        "IPY_MODEL_762102dac8d84154aab25bc96e26cb3f"
       ],
       "layout": "IPY_MODEL_144f5045294f40a7b9dd37503b2b65c6",
       "tabbable": null,
       "tooltip": null
      }
     },
     "306a5b50911b45e2b6277ad3dd7b5caa": {
      "model_module": "@jupyter-widgets/base",
      "model_module_version": "2.0.0",
      "model_name": "LayoutModel",
      "state": {
       "_model_module": "@jupyter-widgets/base",
       "_model_module_version": "2.0.0",
       "_model_name": "LayoutModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "2.0.0",
       "_view_name": "LayoutView",
       "align_content": null,
       "align_items": null,
       "align_self": null,
       "border_bottom": null,
       "border_left": null,
       "border_right": null,
       "border_top": null,
       "bottom": null,
       "display": null,
       "flex": null,
       "flex_flow": null,
       "grid_area": null,
       "grid_auto_columns": null,
       "grid_auto_flow": null,
       "grid_auto_rows": null,
       "grid_column": null,
       "grid_gap": null,
       "grid_row": null,
       "grid_template_areas": null,
       "grid_template_columns": null,
       "grid_template_rows": null,
       "height": null,
       "justify_content": null,
       "justify_items": null,
       "left": null,
       "margin": null,
       "max_height": null,
       "max_width": null,
       "min_height": null,
       "min_width": null,
       "object_fit": null,
       "object_position": null,
       "order": null,
       "overflow": null,
       "padding": null,
       "right": null,
       "top": null,
       "visibility": null,
       "width": null
      }
     },
     "38c80e8809f04126bcd3315d0189f014": {
      "model_module": "@jupyter-widgets/base",
      "model_module_version": "2.0.0",
      "model_name": "LayoutModel",
      "state": {
       "_model_module": "@jupyter-widgets/base",
       "_model_module_version": "2.0.0",
       "_model_name": "LayoutModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "2.0.0",
       "_view_name": "LayoutView",
       "align_content": null,
       "align_items": null,
       "align_self": null,
       "border_bottom": null,
       "border_left": null,
       "border_right": null,
       "border_top": null,
       "bottom": null,
       "display": null,
       "flex": null,
       "flex_flow": null,
       "grid_area": null,
       "grid_auto_columns": null,
       "grid_auto_flow": null,
       "grid_auto_rows": null,
       "grid_column": null,
       "grid_gap": null,
       "grid_row": null,
       "grid_template_areas": null,
       "grid_template_columns": null,
       "grid_template_rows": null,
       "height": null,
       "justify_content": null,
       "justify_items": null,
       "left": null,
       "margin": null,
       "max_height": null,
       "max_width": null,
       "min_height": null,
       "min_width": null,
       "object_fit": null,
       "object_position": null,
       "order": null,
       "overflow": null,
       "padding": null,
       "right": null,
       "top": null,
       "visibility": null,
       "width": null
      }
     },
     "3fb74553702f42aab8b7aa1e0eebf3ea": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "2.0.0",
      "model_name": "HTMLModel",
      "state": {
       "_dom_classes": [],
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "2.0.0",
       "_model_name": "HTMLModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/controls",
       "_view_module_version": "2.0.0",
       "_view_name": "HTMLView",
       "description": "",
       "description_allow_html": false,
       "layout": "IPY_MODEL_76256206ee4c4fd2b442abc9023d15a6",
       "placeholder": "​",
       "style": "IPY_MODEL_c48b2b2f96f948bfa650345d5ab7cd70",
       "tabbable": null,
       "tooltip": null,
       "value": "100%"
      }
     },
     "4348abda4bd5484aad5956d2e8936c5c": {
      "model_module": "@jupyter-widgets/base",
      "model_module_version": "2.0.0",
      "model_name": "LayoutModel",
      "state": {
       "_model_module": "@jupyter-widgets/base",
       "_model_module_version": "2.0.0",
       "_model_name": "LayoutModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "2.0.0",
       "_view_name": "LayoutView",
       "align_content": null,
       "align_items": null,
       "align_self": null,
       "border_bottom": null,
       "border_left": null,
       "border_right": null,
       "border_top": null,
       "bottom": null,
       "display": null,
       "flex": null,
       "flex_flow": null,
       "grid_area": null,
       "grid_auto_columns": null,
       "grid_auto_flow": null,
       "grid_auto_rows": null,
       "grid_column": null,
       "grid_gap": null,
       "grid_row": null,
       "grid_template_areas": null,
       "grid_template_columns": null,
       "grid_template_rows": null,
       "height": null,
       "justify_content": null,
       "justify_items": null,
       "left": null,
       "margin": null,
       "max_height": null,
       "max_width": null,
       "min_height": null,
       "min_width": null,
       "object_fit": null,
       "object_position": null,
       "order": null,
       "overflow": null,
       "padding": null,
       "right": null,
       "top": null,
       "visibility": null,
       "width": null
      }
     },
     "529ebf015c964545849d6167f55aba7d": {
      "model_module": "@jupyter-widgets/base",
      "model_module_version": "2.0.0",
      "model_name": "LayoutModel",
      "state": {
       "_model_module": "@jupyter-widgets/base",
       "_model_module_version": "2.0.0",
       "_model_name": "LayoutModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "2.0.0",
       "_view_name": "LayoutView",
       "align_content": null,
       "align_items": null,
       "align_self": null,
       "border_bottom": null,
       "border_left": null,
       "border_right": null,
       "border_top": null,
       "bottom": null,
       "display": null,
       "flex": null,
       "flex_flow": null,
       "grid_area": null,
       "grid_auto_columns": null,
       "grid_auto_flow": null,
       "grid_auto_rows": null,
       "grid_column": null,
       "grid_gap": null,
       "grid_row": null,
       "grid_template_areas": null,
       "grid_template_columns": null,
       "grid_template_rows": null,
       "height": null,
       "justify_content": null,
       "justify_items": null,
       "left": null,
       "margin": null,
       "max_height": null,
       "max_width": null,
       "min_height": null,
       "min_width": null,
       "object_fit": null,
       "object_position": null,
       "order": null,
       "overflow": null,
       "padding": null,
       "right": null,
       "top": null,
       "visibility": null,
       "width": null
      }
     },
     "568e40ce02c244eba0ce4871a92b4668": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "2.0.0",
      "model_name": "HBoxModel",
      "state": {
       "_dom_classes": [],
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "2.0.0",
       "_model_name": "HBoxModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/controls",
       "_view_module_version": "2.0.0",
       "_view_name": "HBoxView",
       "box_style": "",
       "children": [
        "IPY_MODEL_a65ac803b77b4e4fa8bcc2fa6f321add",
        "IPY_MODEL_a615faf8a876468198e43785b3a9a453",
        "IPY_MODEL_cdd5075bd06145f4a05df88f327292d6"
       ],
       "layout": "IPY_MODEL_4348abda4bd5484aad5956d2e8936c5c",
       "tabbable": null,
       "tooltip": null
      }
     },
     "6355021ca6004d8baea8b87eb2fcbce3": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "2.0.0",
      "model_name": "HTMLStyleModel",
      "state": {
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "2.0.0",
       "_model_name": "HTMLStyleModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "2.0.0",
       "_view_name": "StyleView",
       "background": null,
       "description_width": "",
       "font_size": null,
       "text_color": null
      }
     },
     "63e2a42b72b24b4d825855504ee1a639": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "2.0.0",
      "model_name": "HTMLModel",
      "state": {
       "_dom_classes": [],
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "2.0.0",
       "_model_name": "HTMLModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/controls",
       "_view_module_version": "2.0.0",
       "_view_name": "HTMLView",
       "description": "",
       "description_allow_html": false,
       "layout": "IPY_MODEL_306a5b50911b45e2b6277ad3dd7b5caa",
       "placeholder": "​",
       "style": "IPY_MODEL_71c31c284b5940bd994567451ab8868b",
       "tabbable": null,
       "tooltip": null,
       "value": " 65/65 [07:30&lt;00:00,  3.39s/it]"
      }
     },
     "64615f48fd4444c5a4069793a040cb63": {
      "model_module": "@jupyter-widgets/base",
      "model_module_version": "2.0.0",
      "model_name": "LayoutModel",
      "state": {
       "_model_module": "@jupyter-widgets/base",
       "_model_module_version": "2.0.0",
       "_model_name": "LayoutModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "2.0.0",
       "_view_name": "LayoutView",
       "align_content": null,
       "align_items": null,
       "align_self": null,
       "border_bottom": null,
       "border_left": null,
       "border_right": null,
       "border_top": null,
       "bottom": null,
       "display": null,
       "flex": null,
       "flex_flow": null,
       "grid_area": null,
       "grid_auto_columns": null,
       "grid_auto_flow": null,
       "grid_auto_rows": null,
       "grid_column": null,
       "grid_gap": null,
       "grid_row": null,
       "grid_template_areas": null,
       "grid_template_columns": null,
       "grid_template_rows": null,
       "height": null,
       "justify_content": null,
       "justify_items": null,
       "left": null,
       "margin": null,
       "max_height": null,
       "max_width": null,
       "min_height": null,
       "min_width": null,
       "object_fit": null,
       "object_position": null,
       "order": null,
       "overflow": null,
       "padding": null,
       "right": null,
       "top": null,
       "visibility": null,
       "width": null
      }
     },
     "67e34cd961fe45e79b4aa27ac14b18bf": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "2.0.0",
      "model_name": "ProgressStyleModel",
      "state": {
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "2.0.0",
       "_model_name": "ProgressStyleModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "2.0.0",
       "_view_name": "StyleView",
       "bar_color": null,
       "description_width": ""
      }
     },
     "71c31c284b5940bd994567451ab8868b": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "2.0.0",
      "model_name": "HTMLStyleModel",
      "state": {
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "2.0.0",
       "_model_name": "HTMLStyleModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "2.0.0",
       "_view_name": "StyleView",
       "background": null,
       "description_width": "",
       "font_size": null,
       "text_color": null
      }
     },
     "762102dac8d84154aab25bc96e26cb3f": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "2.0.0",
      "model_name": "HTMLModel",
      "state": {
       "_dom_classes": [],
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "2.0.0",
       "_model_name": "HTMLModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/controls",
       "_view_module_version": "2.0.0",
       "_view_name": "HTMLView",
       "description": "",
       "description_allow_html": false,
       "layout": "IPY_MODEL_d176d91e5a6f4656946e5166cabcedd3",
       "placeholder": "​",
       "style": "IPY_MODEL_186223a8b408471fb54f2e0b043f36f4",
       "tabbable": null,
       "tooltip": null,
       "value": " 40/40 [04:17&lt;00:00,  8.47s/it]"
      }
     },
     "76256206ee4c4fd2b442abc9023d15a6": {
      "model_module": "@jupyter-widgets/base",
      "model_module_version": "2.0.0",
      "model_name": "LayoutModel",
      "state": {
       "_model_module": "@jupyter-widgets/base",
       "_model_module_version": "2.0.0",
       "_model_name": "LayoutModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "2.0.0",
       "_view_name": "LayoutView",
       "align_content": null,
       "align_items": null,
       "align_self": null,
       "border_bottom": null,
       "border_left": null,
       "border_right": null,
       "border_top": null,
       "bottom": null,
       "display": null,
       "flex": null,
       "flex_flow": null,
       "grid_area": null,
       "grid_auto_columns": null,
       "grid_auto_flow": null,
       "grid_auto_rows": null,
       "grid_column": null,
       "grid_gap": null,
       "grid_row": null,
       "grid_template_areas": null,
       "grid_template_columns": null,
       "grid_template_rows": null,
       "height": null,
       "justify_content": null,
       "justify_items": null,
       "left": null,
       "margin": null,
       "max_height": null,
       "max_width": null,
       "min_height": null,
       "min_width": null,
       "object_fit": null,
       "object_position": null,
       "order": null,
       "overflow": null,
       "padding": null,
       "right": null,
       "top": null,
       "visibility": null,
       "width": null
      }
     },
     "7ca727f0d4af4307a6dc7a7383701a03": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "2.0.0",
      "model_name": "HTMLStyleModel",
      "state": {
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "2.0.0",
       "_model_name": "HTMLStyleModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "2.0.0",
       "_view_name": "StyleView",
       "background": null,
       "description_width": "",
       "font_size": null,
       "text_color": null
      }
     },
     "8afa28d3b88c46039d07d75b7c38e03a": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "2.0.0",
      "model_name": "HBoxModel",
      "state": {
       "_dom_classes": [],
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "2.0.0",
       "_model_name": "HBoxModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/controls",
       "_view_module_version": "2.0.0",
       "_view_name": "HBoxView",
       "box_style": "",
       "children": [
        "IPY_MODEL_3fb74553702f42aab8b7aa1e0eebf3ea",
        "IPY_MODEL_28c557faa8d244c4b16afc9c4ccb0354",
        "IPY_MODEL_63e2a42b72b24b4d825855504ee1a639"
       ],
       "layout": "IPY_MODEL_ec76c1c2d929428fa7abd1c55e7c2ed8",
       "tabbable": null,
       "tooltip": null
      }
     },
     "a134a036991c45899d811dab10be5eef": {
      "model_module": "@jupyter-widgets/base",
      "model_module_version": "2.0.0",
      "model_name": "LayoutModel",
      "state": {
       "_model_module": "@jupyter-widgets/base",
       "_model_module_version": "2.0.0",
       "_model_name": "LayoutModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "2.0.0",
       "_view_name": "LayoutView",
       "align_content": null,
       "align_items": null,
       "align_self": null,
       "border_bottom": null,
       "border_left": null,
       "border_right": null,
       "border_top": null,
       "bottom": null,
       "display": null,
       "flex": null,
       "flex_flow": null,
       "grid_area": null,
       "grid_auto_columns": null,
       "grid_auto_flow": null,
       "grid_auto_rows": null,
       "grid_column": null,
       "grid_gap": null,
       "grid_row": null,
       "grid_template_areas": null,
       "grid_template_columns": null,
       "grid_template_rows": null,
       "height": null,
       "justify_content": null,
       "justify_items": null,
       "left": null,
       "margin": null,
       "max_height": null,
       "max_width": null,
       "min_height": null,
       "min_width": null,
       "object_fit": null,
       "object_position": null,
       "order": null,
       "overflow": null,
       "padding": null,
       "right": null,
       "top": null,
       "visibility": null,
       "width": null
      }
     },
     "a143ad464ade43aeb4a8b05d47fdc8e6": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "2.0.0",
      "model_name": "FloatProgressModel",
      "state": {
       "_dom_classes": [],
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "2.0.0",
       "_model_name": "FloatProgressModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/controls",
       "_view_module_version": "2.0.0",
       "_view_name": "ProgressView",
       "bar_style": "success",
       "description": "",
       "description_allow_html": false,
       "layout": "IPY_MODEL_d3137da017c24040b3d0b3e486dbd5be",
       "max": 40,
       "min": 0,
       "orientation": "horizontal",
       "style": "IPY_MODEL_67e34cd961fe45e79b4aa27ac14b18bf",
       "tabbable": null,
       "tooltip": null,
       "value": 40
      }
     },
     "a615faf8a876468198e43785b3a9a453": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "2.0.0",
      "model_name": "FloatProgressModel",
      "state": {
       "_dom_classes": [],
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "2.0.0",
       "_model_name": "FloatProgressModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/controls",
       "_view_module_version": "2.0.0",
       "_view_name": "ProgressView",
       "bar_style": "success",
       "description": "",
       "description_allow_html": false,
       "layout": "IPY_MODEL_ef8f2573c4e840baaf484178d7d46cd9",
       "max": 65,
       "min": 0,
       "orientation": "horizontal",
       "style": "IPY_MODEL_ca8f45cec802410ca5f1d02842bc8314",
       "tabbable": null,
       "tooltip": null,
       "value": 65
      }
     },
     "a65ac803b77b4e4fa8bcc2fa6f321add": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "2.0.0",
      "model_name": "HTMLModel",
      "state": {
       "_dom_classes": [],
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "2.0.0",
       "_model_name": "HTMLModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/controls",
       "_view_module_version": "2.0.0",
       "_view_name": "HTMLView",
       "description": "",
       "description_allow_html": false,
       "layout": "IPY_MODEL_e0747950912b491ab838c8784c62e4ab",
       "placeholder": "​",
       "style": "IPY_MODEL_cf3f0c10a73149b58552b1b77971d56e",
       "tabbable": null,
       "tooltip": null,
       "value": "100%"
      }
     },
     "b8547e5480aa4499b855264f9418f0b3": {
      "model_module": "@jupyter-widgets/base",
      "model_module_version": "2.0.0",
      "model_name": "LayoutModel",
      "state": {
       "_model_module": "@jupyter-widgets/base",
       "_model_module_version": "2.0.0",
       "_model_name": "LayoutModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "2.0.0",
       "_view_name": "LayoutView",
       "align_content": null,
       "align_items": null,
       "align_self": null,
       "border_bottom": null,
       "border_left": null,
       "border_right": null,
       "border_top": null,
       "bottom": null,
       "display": null,
       "flex": null,
       "flex_flow": null,
       "grid_area": null,
       "grid_auto_columns": null,
       "grid_auto_flow": null,
       "grid_auto_rows": null,
       "grid_column": null,
       "grid_gap": null,
       "grid_row": null,
       "grid_template_areas": null,
       "grid_template_columns": null,
       "grid_template_rows": null,
       "height": null,
       "justify_content": null,
       "justify_items": null,
       "left": null,
       "margin": null,
       "max_height": null,
       "max_width": null,
       "min_height": null,
       "min_width": null,
       "object_fit": null,
       "object_position": null,
       "order": null,
       "overflow": null,
       "padding": null,
       "right": null,
       "top": null,
       "visibility": null,
       "width": null
      }
     },
     "c48b2b2f96f948bfa650345d5ab7cd70": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "2.0.0",
      "model_name": "HTMLStyleModel",
      "state": {
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "2.0.0",
       "_model_name": "HTMLStyleModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "2.0.0",
       "_view_name": "StyleView",
       "background": null,
       "description_width": "",
       "font_size": null,
       "text_color": null
      }
     },
     "ca8f45cec802410ca5f1d02842bc8314": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "2.0.0",
      "model_name": "ProgressStyleModel",
      "state": {
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "2.0.0",
       "_model_name": "ProgressStyleModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "2.0.0",
       "_view_name": "StyleView",
       "bar_color": null,
       "description_width": ""
      }
     },
     "cdd5075bd06145f4a05df88f327292d6": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "2.0.0",
      "model_name": "HTMLModel",
      "state": {
       "_dom_classes": [],
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "2.0.0",
       "_model_name": "HTMLModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/controls",
       "_view_module_version": "2.0.0",
       "_view_name": "HTMLView",
       "description": "",
       "description_allow_html": false,
       "layout": "IPY_MODEL_0e108582dbe041c0b51c327d89d6f369",
       "placeholder": "​",
       "style": "IPY_MODEL_d3616abaa5bc40d3a9ee5f73f6524994",
       "tabbable": null,
       "tooltip": null,
       "value": " 65/65 [07:56&lt;00:00,  2.59s/it]"
      }
     },
     "cf3f0c10a73149b58552b1b77971d56e": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "2.0.0",
      "model_name": "HTMLStyleModel",
      "state": {
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "2.0.0",
       "_model_name": "HTMLStyleModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "2.0.0",
       "_view_name": "StyleView",
       "background": null,
       "description_width": "",
       "font_size": null,
       "text_color": null
      }
     },
     "d176d91e5a6f4656946e5166cabcedd3": {
      "model_module": "@jupyter-widgets/base",
      "model_module_version": "2.0.0",
      "model_name": "LayoutModel",
      "state": {
       "_model_module": "@jupyter-widgets/base",
       "_model_module_version": "2.0.0",
       "_model_name": "LayoutModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "2.0.0",
       "_view_name": "LayoutView",
       "align_content": null,
       "align_items": null,
       "align_self": null,
       "border_bottom": null,
       "border_left": null,
       "border_right": null,
       "border_top": null,
       "bottom": null,
       "display": null,
       "flex": null,
       "flex_flow": null,
       "grid_area": null,
       "grid_auto_columns": null,
       "grid_auto_flow": null,
       "grid_auto_rows": null,
       "grid_column": null,
       "grid_gap": null,
       "grid_row": null,
       "grid_template_areas": null,
       "grid_template_columns": null,
       "grid_template_rows": null,
       "height": null,
       "justify_content": null,
       "justify_items": null,
       "left": null,
       "margin": null,
       "max_height": null,
       "max_width": null,
       "min_height": null,
       "min_width": null,
       "object_fit": null,
       "object_position": null,
       "order": null,
       "overflow": null,
       "padding": null,
       "right": null,
       "top": null,
       "visibility": null,
       "width": null
      }
     },
     "d3137da017c24040b3d0b3e486dbd5be": {
      "model_module": "@jupyter-widgets/base",
      "model_module_version": "2.0.0",
      "model_name": "LayoutModel",
      "state": {
       "_model_module": "@jupyter-widgets/base",
       "_model_module_version": "2.0.0",
       "_model_name": "LayoutModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "2.0.0",
       "_view_name": "LayoutView",
       "align_content": null,
       "align_items": null,
       "align_self": null,
       "border_bottom": null,
       "border_left": null,
       "border_right": null,
       "border_top": null,
       "bottom": null,
       "display": null,
       "flex": null,
       "flex_flow": null,
       "grid_area": null,
       "grid_auto_columns": null,
       "grid_auto_flow": null,
       "grid_auto_rows": null,
       "grid_column": null,
       "grid_gap": null,
       "grid_row": null,
       "grid_template_areas": null,
       "grid_template_columns": null,
       "grid_template_rows": null,
       "height": null,
       "justify_content": null,
       "justify_items": null,
       "left": null,
       "margin": null,
       "max_height": null,
       "max_width": null,
       "min_height": null,
       "min_width": null,
       "object_fit": null,
       "object_position": null,
       "order": null,
       "overflow": null,
       "padding": null,
       "right": null,
       "top": null,
       "visibility": null,
       "width": null
      }
     },
     "d3616abaa5bc40d3a9ee5f73f6524994": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "2.0.0",
      "model_name": "HTMLStyleModel",
      "state": {
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "2.0.0",
       "_model_name": "HTMLStyleModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "2.0.0",
       "_view_name": "StyleView",
       "background": null,
       "description_width": "",
       "font_size": null,
       "text_color": null
      }
     },
     "d7c2d5f92ab74a64b0759e1d98ed0ebe": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "2.0.0",
      "model_name": "HTMLModel",
      "state": {
       "_dom_classes": [],
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "2.0.0",
       "_model_name": "HTMLModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/controls",
       "_view_module_version": "2.0.0",
       "_view_name": "HTMLView",
       "description": "",
       "description_allow_html": false,
       "layout": "IPY_MODEL_38c80e8809f04126bcd3315d0189f014",
       "placeholder": "​",
       "style": "IPY_MODEL_7ca727f0d4af4307a6dc7a7383701a03",
       "tabbable": null,
       "tooltip": null,
       "value": " 45/45 [06:09&lt;00:00,  9.62s/it]"
      }
     },
     "d9433aaf1faa45379f5b9cb675471be9": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "2.0.0",
      "model_name": "ProgressStyleModel",
      "state": {
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "2.0.0",
       "_model_name": "ProgressStyleModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "2.0.0",
       "_view_name": "StyleView",
       "bar_color": null,
       "description_width": ""
      }
     },
     "e0747950912b491ab838c8784c62e4ab": {
      "model_module": "@jupyter-widgets/base",
      "model_module_version": "2.0.0",
      "model_name": "LayoutModel",
      "state": {
       "_model_module": "@jupyter-widgets/base",
       "_model_module_version": "2.0.0",
       "_model_name": "LayoutModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "2.0.0",
       "_view_name": "LayoutView",
       "align_content": null,
       "align_items": null,
       "align_self": null,
       "border_bottom": null,
       "border_left": null,
       "border_right": null,
       "border_top": null,
       "bottom": null,
       "display": null,
       "flex": null,
       "flex_flow": null,
       "grid_area": null,
       "grid_auto_columns": null,
       "grid_auto_flow": null,
       "grid_auto_rows": null,
       "grid_column": null,
       "grid_gap": null,
       "grid_row": null,
       "grid_template_areas": null,
       "grid_template_columns": null,
       "grid_template_rows": null,
       "height": null,
       "justify_content": null,
       "justify_items": null,
       "left": null,
       "margin": null,
       "max_height": null,
       "max_width": null,
       "min_height": null,
       "min_width": null,
       "object_fit": null,
       "object_position": null,
       "order": null,
       "overflow": null,
       "padding": null,
       "right": null,
       "top": null,
       "visibility": null,
       "width": null
      }
     },
     "ec76c1c2d929428fa7abd1c55e7c2ed8": {
      "model_module": "@jupyter-widgets/base",
      "model_module_version": "2.0.0",
      "model_name": "LayoutModel",
      "state": {
       "_model_module": "@jupyter-widgets/base",
       "_model_module_version": "2.0.0",
       "_model_name": "LayoutModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "2.0.0",
       "_view_name": "LayoutView",
       "align_content": null,
       "align_items": null,
       "align_self": null,
       "border_bottom": null,
       "border_left": null,
       "border_right": null,
       "border_top": null,
       "bottom": null,
       "display": null,
       "flex": null,
       "flex_flow": null,
       "grid_area": null,
       "grid_auto_columns": null,
       "grid_auto_flow": null,
       "grid_auto_rows": null,
       "grid_column": null,
       "grid_gap": null,
       "grid_row": null,
       "grid_template_areas": null,
       "grid_template_columns": null,
       "grid_template_rows": null,
       "height": null,
       "justify_content": null,
       "justify_items": null,
       "left": null,
       "margin": null,
       "max_height": null,
       "max_width": null,
       "min_height": null,
       "min_width": null,
       "object_fit": null,
       "object_position": null,
       "order": null,
       "overflow": null,
       "padding": null,
       "right": null,
       "top": null,
       "visibility": null,
       "width": null
      }
     },
     "ef8f2573c4e840baaf484178d7d46cd9": {
      "model_module": "@jupyter-widgets/base",
      "model_module_version": "2.0.0",
      "model_name": "LayoutModel",
      "state": {
       "_model_module": "@jupyter-widgets/base",
       "_model_module_version": "2.0.0",
       "_model_name": "LayoutModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "2.0.0",
       "_view_name": "LayoutView",
       "align_content": null,
       "align_items": null,
       "align_self": null,
       "border_bottom": null,
       "border_left": null,
       "border_right": null,
       "border_top": null,
       "bottom": null,
       "display": null,
       "flex": null,
       "flex_flow": null,
       "grid_area": null,
       "grid_auto_columns": null,
       "grid_auto_flow": null,
       "grid_auto_rows": null,
       "grid_column": null,
       "grid_gap": null,
       "grid_row": null,
       "grid_template_areas": null,
       "grid_template_columns": null,
       "grid_template_rows": null,
       "height": null,
       "justify_content": null,
       "justify_items": null,
       "left": null,
       "margin": null,
       "max_height": null,
       "max_width": null,
       "min_height": null,
       "min_width": null,
       "object_fit": null,
       "object_position": null,
       "order": null,
       "overflow": null,
       "padding": null,
       "right": null,
       "top": null,
       "visibility": null,
       "width": null
      }
     },
     "f203fc46d8e648f9b42b19df3dbe2c07": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "2.0.0",
      "model_name": "HBoxModel",
      "state": {
       "_dom_classes": [],
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "2.0.0",
       "_model_name": "HBoxModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/controls",
       "_view_module_version": "2.0.0",
       "_view_name": "HBoxView",
       "box_style": "",
       "children": [
        "IPY_MODEL_f5f3462911324ddaa4a2961ebf6cacca",
        "IPY_MODEL_096e5eadf6e54272be7f7d0b0f3d3228",
        "IPY_MODEL_d7c2d5f92ab74a64b0759e1d98ed0ebe"
       ],
       "layout": "IPY_MODEL_f35394ce571c4b8aa7c2e1636edbf117",
       "tabbable": null,
       "tooltip": null
      }
     },
     "f35394ce571c4b8aa7c2e1636edbf117": {
      "model_module": "@jupyter-widgets/base",
      "model_module_version": "2.0.0",
      "model_name": "LayoutModel",
      "state": {
       "_model_module": "@jupyter-widgets/base",
       "_model_module_version": "2.0.0",
       "_model_name": "LayoutModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "2.0.0",
       "_view_name": "LayoutView",
       "align_content": null,
       "align_items": null,
       "align_self": null,
       "border_bottom": null,
       "border_left": null,
       "border_right": null,
       "border_top": null,
       "bottom": null,
       "display": null,
       "flex": null,
       "flex_flow": null,
       "grid_area": null,
       "grid_auto_columns": null,
       "grid_auto_flow": null,
       "grid_auto_rows": null,
       "grid_column": null,
       "grid_gap": null,
       "grid_row": null,
       "grid_template_areas": null,
       "grid_template_columns": null,
       "grid_template_rows": null,
       "height": null,
       "justify_content": null,
       "justify_items": null,
       "left": null,
       "margin": null,
       "max_height": null,
       "max_width": null,
       "min_height": null,
       "min_width": null,
       "object_fit": null,
       "object_position": null,
       "order": null,
       "overflow": null,
       "padding": null,
       "right": null,
       "top": null,
       "visibility": null,
       "width": null
      }
     },
     "f5f3462911324ddaa4a2961ebf6cacca": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "2.0.0",
      "model_name": "HTMLModel",
      "state": {
       "_dom_classes": [],
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "2.0.0",
       "_model_name": "HTMLModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/controls",
       "_view_module_version": "2.0.0",
       "_view_name": "HTMLView",
       "description": "",
       "description_allow_html": false,
       "layout": "IPY_MODEL_529ebf015c964545849d6167f55aba7d",
       "placeholder": "​",
       "style": "IPY_MODEL_6355021ca6004d8baea8b87eb2fcbce3",
       "tabbable": null,
       "tooltip": null,
       "value": "100%"
      }
     }
    },
    "version_major": 2,
    "version_minor": 0
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}