src/second_level/second_level.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "0dbe870e",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "from collections import defaultdict\n",
    "\n",
    "merchant_items = defaultdict(list)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "45f718f5",
   "metadata": {},
   "outputs": [],
   "source": [
    "path = 'data/'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "3309971e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>buyer_id</th>\n",
       "      <th>pav_order_id</th>\n",
       "      <th>created</th>\n",
       "      <th>item_id</th>\n",
       "      <th>count</th>\n",
       "      <th>price_sold</th>\n",
       "      <th>flag_weight_goods</th>\n",
       "      <th>weight</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>95203091</td>\n",
       "      <td>98506637863</td>\n",
       "      <td>2021-07-01 00:03:44</td>\n",
       "      <td>202808329</td>\n",
       "      <td>1.000</td>\n",
       "      <td>79.990</td>\n",
       "      <td>False</td>\n",
       "      <td>11.14</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>95203091</td>\n",
       "      <td>98506637863</td>\n",
       "      <td>2021-07-01 00:03:44</td>\n",
       "      <td>202953905</td>\n",
       "      <td>1.072</td>\n",
       "      <td>44.945</td>\n",
       "      <td>True</td>\n",
       "      <td>11.14</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>95203091</td>\n",
       "      <td>98506637863</td>\n",
       "      <td>2021-07-01 00:03:44</td>\n",
       "      <td>203566452</td>\n",
       "      <td>1.000</td>\n",
       "      <td>69.990</td>\n",
       "      <td>False</td>\n",
       "      <td>11.14</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>95203091</td>\n",
       "      <td>98506637863</td>\n",
       "      <td>2021-07-01 00:03:44</td>\n",
       "      <td>202820143</td>\n",
       "      <td>1.972</td>\n",
       "      <td>41.295</td>\n",
       "      <td>True</td>\n",
       "      <td>11.14</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>95203091</td>\n",
       "      <td>98506637863</td>\n",
       "      <td>2021-07-01 00:03:44</td>\n",
       "      <td>204400422</td>\n",
       "      <td>1.000</td>\n",
       "      <td>269.990</td>\n",
       "      <td>False</td>\n",
       "      <td>11.14</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   buyer_id  pav_order_id              created    item_id  count  price_sold  \\\n",
       "0  95203091   98506637863  2021-07-01 00:03:44  202808329  1.000      79.990   \n",
       "1  95203091   98506637863  2021-07-01 00:03:44  202953905  1.072      44.945   \n",
       "2  95203091   98506637863  2021-07-01 00:03:44  203566452  1.000      69.990   \n",
       "3  95203091   98506637863  2021-07-01 00:03:44  202820143  1.972      41.295   \n",
       "4  95203091   98506637863  2021-07-01 00:03:44  204400422  1.000     269.990   \n",
       "\n",
       "   flag_weight_goods  weight  \n",
       "0              False   11.14  \n",
       "1               True   11.14  \n",
       "2              False   11.14  \n",
       "3               True   11.14  \n",
       "4              False   11.14  "
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pd.read_csv(path+'hist_data.csv', nrows=10000)\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "cd103843",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "b95018e5",
   "metadata": {},
   "outputs": [],
   "source": [
    "def filter_column(ds, col, min_freq, free_col='rating'):\n",
    "    \n",
    "    temp = ds.groupby(col, as_index=False).agg({free_col:'count'})\n",
    "    filtered = temp.loc[temp[free_col] >= min_freq, col].values\n",
    "    ds = ds.loc[(ds[col].isin(filtered))].copy()\n",
    "\n",
    "    return ds\n",
    "\n",
    "df = filter_column(df, col='pav_order_id', min_freq=5, free_col='price_sold')\n",
    "df = filter_column(df, col='item_id', min_freq=50,free_col='price_sold')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "3517601b",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "9 363\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 th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>buyer_id</th>\n",
       "      <th>pav_order_id</th>\n",
       "      <th>created</th>\n",
       "      <th>item_id</th>\n",
       "      <th>count</th>\n",
       "      <th>price_sold</th>\n",
       "      <th>flag_weight_goods</th>\n",
       "      <th>weight</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>95203091</td>\n",
       "      <td>0</td>\n",
       "      <td>2021-07-01 00:03:44</td>\n",
       "      <td>0</td>\n",
       "      <td>1.002</td>\n",
       "      <td>43.945</td>\n",
       "      <td>True</td>\n",
       "      <td>11.14</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>95203091</td>\n",
       "      <td>0</td>\n",
       "      <td>2021-07-01 00:03:44</td>\n",
       "      <td>1</td>\n",
       "      <td>0.968</td>\n",
       "      <td>52.395</td>\n",
       "      <td>True</td>\n",
       "      <td>11.14</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35</th>\n",
       "      <td>95854621</td>\n",
       "      <td>1</td>\n",
       "      <td>2021-07-01 00:16:39</td>\n",
       "      <td>2</td>\n",
       "      <td>3.924</td>\n",
       "      <td>28.145</td>\n",
       "      <td>True</td>\n",
       "      <td>28.23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>55</th>\n",
       "      <td>95854621</td>\n",
       "      <td>1</td>\n",
       "      <td>2021-07-01 00:16:39</td>\n",
       "      <td>3</td>\n",
       "      <td>1.044</td>\n",
       "      <td>27.690</td>\n",
       "      <td>True</td>\n",
       "      <td>28.23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>70</th>\n",
       "      <td>94767798</td>\n",
       "      <td>2</td>\n",
       "      <td>2021-07-01 00:16:53</td>\n",
       "      <td>4</td>\n",
       "      <td>3.000</td>\n",
       "      <td>49.890</td>\n",
       "      <td>False</td>\n",
       "      <td>69.35</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    buyer_id  pav_order_id              created  item_id  count  price_sold  \\\n",
       "6   95203091             0  2021-07-01 00:03:44        0  1.002      43.945   \n",
       "7   95203091             0  2021-07-01 00:03:44        1  0.968      52.395   \n",
       "35  95854621             1  2021-07-01 00:16:39        2  3.924      28.145   \n",
       "55  95854621             1  2021-07-01 00:16:39        3  1.044      27.690   \n",
       "70  94767798             2  2021-07-01 00:16:53        4  3.000      49.890   \n",
       "\n",
       "    flag_weight_goods  weight  \n",
       "6                True   11.14  \n",
       "7                True   11.14  \n",
       "35               True   28.23  \n",
       "55               True   28.23  \n",
       "70              False   69.35  "
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "user2idx = {v: k for k, v in enumerate(df.pav_order_id.unique())}\n",
    "item2idx = {v: k for k, v in enumerate(df.item_id.unique())}\n",
    "print(len(item2idx), len(user2idx))\n",
    "idx2user = {k:v for v, k in user2idx.items()}\n",
    "idx2item = {k:v for v, k in item2idx.items()}\n",
    "\n",
    "df['pav_order_id'] = df.pav_order_id.apply(lambda x: user2idx[x])\n",
    "df['item_id'] = df.item_id.apply(lambda x: item2idx[x])\n",
    "\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "fa20c287",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(363, 2)\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 th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>buyer_id</th>\n",
       "      <th>history_items</th>\n",
       "      <th>target_basket_x</th>\n",
       "      <th>target_basket_y</th>\n",
       "      <th>segment</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>94590204</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]</td>\n",
       "      <td>[0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0]</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]</td>\n",
       "      <td>train_1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>94590324</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]</td>\n",
       "      <td>[0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 1.0, 0.0, 0.0]</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]</td>\n",
       "      <td>train_1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>94599587</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]</td>\n",
       "      <td>[0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0]</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]</td>\n",
       "      <td>train_1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>94600013</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]</td>\n",
       "      <td>[0.0, 0.0, 0.0, 1.0, 0.0, 1.0, 0.0, 0.0, 1.0]</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0]</td>\n",
       "      <td>train_1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>94601062</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]</td>\n",
       "      <td>[0.0, 1.0, 0.0, 0.0, 1.0, 0.0, 0.0, 1.0, 0.0]</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]</td>\n",
       "      <td>train_2</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   buyer_id                                  history_items  \\\n",
       "0  94590204  [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]   \n",
       "1  94590324  [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]   \n",
       "2  94599587  [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]   \n",
       "3  94600013  [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]   \n",
       "4  94601062  [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]   \n",
       "\n",
       "                                 target_basket_x  \\\n",
       "0  [0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0]   \n",
       "1  [0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 1.0, 0.0, 0.0]   \n",
       "2  [0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0]   \n",
       "3  [0.0, 0.0, 0.0, 1.0, 0.0, 1.0, 0.0, 0.0, 1.0]   \n",
       "4  [0.0, 1.0, 0.0, 0.0, 1.0, 0.0, 0.0, 1.0, 0.0]   \n",
       "\n",
       "                                 target_basket_y  segment  \n",
       "0  [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]  train_1  \n",
       "1  [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]  train_1  \n",
       "2  [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]  train_1  \n",
       "3  [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0]  train_1  \n",
       "4  [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]  train_2  "
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import numpy as np\n",
    "\n",
    "def get_grouped_data(x):\n",
    "    \n",
    "    x = x.sort_values(by=['pav_order_id','created'])\n",
    "    last_history = set()\n",
    "    target_basket_x = set()\n",
    "    target_basket_y = set()\n",
    "    last_order = x.pav_order_id.values[-1]\n",
    "    left_of_last_order = len(x.loc[x.pav_order_id==last_order])\n",
    "    \n",
    "    for idx, row in x.iterrows():\n",
    "        if row.pav_order_id != last_order:\n",
    "            last_history.add(int(row.item_id))\n",
    "        else:\n",
    "            if len(target_basket_x) < 0.7 * left_of_last_order:\n",
    "                target_basket_x.add(int(row.item_id))\n",
    "            else:\n",
    "                target_basket_y.add(int(row.item_id))\n",
    "                \n",
    "    last_history_vector = np.zeros(len(item2idx))\n",
    "    tbx_vector = np.zeros(len(item2idx))\n",
    "    tby_vector = np.zeros(len(item2idx))\n",
    "    last_history_vector[list(last_history)] = 1\n",
    "    tbx_vector[list(target_basket_x)] = 1\n",
    "    tby_vector[list(target_basket_y)] = 1\n",
    "                \n",
    "    return last_history_vector, tbx_vector, tby_vector\n",
    "    \n",
    "grouped = df.groupby('buyer_id').apply(lambda x: get_grouped_data(x)).reset_index()\n",
    "print(grouped.shape)\n",
    "grouped.rename({0:'joined'}, axis=1, inplace=True)\n",
    "grouped['history_items'] = grouped.joined.apply(lambda x: x[0])\n",
    "grouped['target_basket_x'] = grouped.joined.apply(lambda x: x[1])\n",
    "grouped['target_basket_y'] = grouped.joined.apply(lambda x: x[2])\n",
    "grouped.drop(['joined'], axis=1, inplace=True)\n",
    "grouped['segment'] = grouped.buyer_id.apply(lambda x: np.random.choice(['train_1','train_2','test_2']))\n",
    "grouped.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "adec31f2",
   "metadata": {},
   "outputs": [],
   "source": [
    "#grouped.to_pickle(path+'grouped.pkl')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "7b777465",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import bottleneck as bn\n",
    "import torch\n",
    "from scipy import sparse as sps\n",
    "\n",
    "\n",
    "def NDCG_binary_at_k_batch(X_pred, heldout_batch, k=20):\n",
    "    \"\"\"\n",
    "    Normalized Discounted Cumulative Gain@k for binary relevance\n",
    "    ASSUMPTIONS: all the 0's in heldout_data indicate 0 relevance\n",
    "    \"\"\"\n",
    "    batch_users = X_pred.shape[0]\n",
    "    heldout_batch = sps.csr_matrix(heldout_batch)\n",
    "    \n",
    "    idx_topk_part = bn.argpartition(-X_pred, k, axis=1)\n",
    "    topk_part = X_pred[np.arange(batch_users)[:, np.newaxis], idx_topk_part[:, :k]]\n",
    "    idx_part = np.argsort(-topk_part, axis=1)\n",
    "\n",
    "    idx_topk = idx_topk_part[np.arange(batch_users)[:, np.newaxis], idx_part]\n",
    "\n",
    "    tp = 1.0 / np.log2(np.arange(2, k + 2))\n",
    "    \n",
    "\n",
    "    DCG = (heldout_batch[np.arange(batch_users)[:, np.newaxis], idx_topk].toarray() * tp).sum(\n",
    "        axis=1\n",
    "    )\n",
    "    IDCG = np.array([(tp[: min(n, k)]).sum() for n in heldout_batch.getnnz(axis=1)])\n",
    "    ndcg = DCG / IDCG\n",
    "    ndcg = ndcg[ndcg > -100]\n",
    "    return ndcg.mean()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "53f33962",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "eabddeaf",
   "metadata": {},
   "outputs": [],
   "source": [
    "train = grouped.loc[grouped.segment=='train_1']\n",
    "\n",
    "train_matrix = sps.csr_matrix(np.vstack((train.target_basket_x.values)))\n",
    "valid_matrix = sps.csr_matrix(np.vstack((train.target_basket_y.values)))\n",
    "\n",
    "train_2 = grouped.loc[grouped.segment=='train_2'].copy()\n",
    "test_2 = grouped.loc[grouped.segment=='test_2'].copy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "d8baa067",
   "metadata": {},
   "outputs": [
    {
     "ename": "KeyboardInterrupt",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-13-7ab614673d3a>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m      3\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0mpathlib\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mPath\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      4\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0mscipy\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0msparse\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 5\u001b[0;31m \u001b[0;32mfrom\u001b[0m \u001b[0msrc\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfirst_level\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfit\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mfit_first_level\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      6\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      7\u001b[0m \u001b[0mconfig_file\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mPath\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcwd\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m/\u001b[0m \u001b[0;34m\"configs\"\u001b[0m \u001b[0;34m/\u001b[0m \u001b[0;34m\"config.yaml\"\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/Downloads/hack-the-cart/src/first_level/fit.py\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m      3\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mjson\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      4\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mtorch\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 5\u001b[0;31m \u001b[0;32mimport\u001b[0m \u001b[0moptuna\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      6\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mrandom\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      7\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mnumpy\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.7/site-packages/optuna/__init__.py\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m      6\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0moptuna\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mdistributions\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      7\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0moptuna\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mexceptions\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 8\u001b[0;31m \u001b[0;32mfrom\u001b[0m \u001b[0moptuna\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mimportance\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      9\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0moptuna\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mintegration\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     10\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0moptuna\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mlogging\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.7/site-packages/optuna/importance/__init__.py\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m      4\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0mtyping\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mOptional\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      5\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 6\u001b[0;31m \u001b[0;32mfrom\u001b[0m \u001b[0moptuna\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mimportance\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_base\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mBaseImportanceEvaluator\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      7\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0moptuna\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mimportance\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_fanova\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mFanovaImportanceEvaluator\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      8\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0moptuna\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mimportance\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_mean_decrease_impurity\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mMeanDecreaseImpurityImportanceEvaluator\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.7/site-packages/optuna/importance/_base.py\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m      7\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      8\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0moptuna\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdistributions\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mBaseDistribution\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 9\u001b[0;31m \u001b[0;32mfrom\u001b[0m \u001b[0moptuna\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msamplers\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mintersection_search_space\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     10\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0moptuna\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mstudy\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mStudy\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     11\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0moptuna\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtrial\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mFrozenTrial\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.7/site-packages/optuna/samplers/__init__.py\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m      7\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0moptuna\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msamplers\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_search_space\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mintersection_search_space\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      8\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0moptuna\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msamplers\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_search_space\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mIntersectionSearchSpace\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 9\u001b[0;31m \u001b[0;32mfrom\u001b[0m \u001b[0moptuna\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msamplers\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_tpe\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmulti_objective_sampler\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mMOTPESampler\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     10\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0moptuna\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msamplers\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_tpe\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msampler\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mTPESampler\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     11\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.7/site-packages/optuna/samplers/_tpe/multi_objective_sampler.py\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m      5\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      6\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0moptuna\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_deprecated\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mdeprecated\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 7\u001b[0;31m \u001b[0;32mfrom\u001b[0m \u001b[0moptuna\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msamplers\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_tpe\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msampler\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mTPESampler\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      8\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      9\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.7/site-packages/optuna/samplers/_tpe/sampler.py\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m     21\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0moptuna\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msamplers\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_search_space\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mgroup_decomposed\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0m_GroupDecomposedSearchSpace\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     22\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0moptuna\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msamplers\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_search_space\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mgroup_decomposed\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0m_SearchSpaceGroup\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 23\u001b[0;31m \u001b[0;32mfrom\u001b[0m \u001b[0moptuna\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msamplers\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_tpe\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mparzen_estimator\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0m_ParzenEstimator\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     24\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0moptuna\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msamplers\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_tpe\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mparzen_estimator\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0m_ParzenEstimatorParameters\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     25\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0moptuna\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mstudy\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mStudy\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.7/site-packages/optuna/samplers/_tpe/parzen_estimator.py\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m      7\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mnumpy\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      8\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mscipy\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mspecial\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 9\u001b[0;31m \u001b[0;32mfrom\u001b[0m \u001b[0mscipy\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mstats\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mtruncnorm\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     10\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     11\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0moptuna\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mdistributions\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.7/site-packages/scipy/stats/__init__.py\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m    439\u001b[0m \"\"\"\n\u001b[1;32m    440\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 441\u001b[0;31m \u001b[0;32mfrom\u001b[0m \u001b[0;34m.\u001b[0m\u001b[0mstats\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    442\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0;34m.\u001b[0m\u001b[0mdistributions\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    443\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0;34m.\u001b[0m\u001b[0mmorestats\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.7/site-packages/scipy/stats/stats.py\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m     41\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mscipy\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mspecial\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0mspecial\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     42\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0mscipy\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mlinalg\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 43\u001b[0;31m \u001b[0;32mfrom\u001b[0m \u001b[0;34m.\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mdistributions\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     44\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0;34m.\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mmstats_basic\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     45\u001b[0m from ._stats_mstats_common import (_find_repeats, linregress, theilslopes,\n",
      "\u001b[0;32m/usr/local/lib/python3.7/site-packages/scipy/stats/distributions.py\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m      8\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0;34m.\u001b[0m\u001b[0m_distn_infrastructure\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0;34m(\u001b[0m\u001b[0mrv_discrete\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mrv_continuous\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mrv_frozen\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      9\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 10\u001b[0;31m \u001b[0;32mfrom\u001b[0m \u001b[0;34m.\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0m_continuous_distns\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     11\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0;34m.\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0m_discrete_distns\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     12\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.7/site-packages/scipy/stats/_continuous_distns.py\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m     15\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0mscipy\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0moptimize\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     16\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0mscipy\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mintegrate\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 17\u001b[0;31m \u001b[0;32mfrom\u001b[0m \u001b[0mscipy\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0minterpolate\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     18\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mscipy\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mspecial\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0msc\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     19\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.7/site-packages/scipy/interpolate/__init__.py\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m    176\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0;34m.\u001b[0m\u001b[0mpolyint\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    177\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 178\u001b[0;31m \u001b[0;32mfrom\u001b[0m \u001b[0;34m.\u001b[0m\u001b[0m_cubic\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    179\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    180\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0;34m.\u001b[0m\u001b[0mndgriddata\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/Cellar/python@3.7/3.7.12_1/Frameworks/Python.framework/Versions/3.7/lib/python3.7/importlib/_bootstrap.py\u001b[0m in \u001b[0;36m_find_and_load\u001b[0;34m(name, import_)\u001b[0m\n",
      "\u001b[0;32m/usr/local/Cellar/python@3.7/3.7.12_1/Frameworks/Python.framework/Versions/3.7/lib/python3.7/importlib/_bootstrap.py\u001b[0m in \u001b[0;36m_find_and_load_unlocked\u001b[0;34m(name, import_)\u001b[0m\n",
      "\u001b[0;32m/usr/local/Cellar/python@3.7/3.7.12_1/Frameworks/Python.framework/Versions/3.7/lib/python3.7/importlib/_bootstrap.py\u001b[0m in \u001b[0;36m_load_unlocked\u001b[0;34m(spec)\u001b[0m\n",
      "\u001b[0;32m/usr/local/Cellar/python@3.7/3.7.12_1/Frameworks/Python.framework/Versions/3.7/lib/python3.7/importlib/_bootstrap_external.py\u001b[0m in \u001b[0;36mexec_module\u001b[0;34m(self, module)\u001b[0m\n",
      "\u001b[0;32m/usr/local/Cellar/python@3.7/3.7.12_1/Frameworks/Python.framework/Versions/3.7/lib/python3.7/importlib/_bootstrap_external.py\u001b[0m in \u001b[0;36mget_code\u001b[0;34m(self, fullname)\u001b[0m\n",
      "\u001b[0;32m/usr/local/Cellar/python@3.7/3.7.12_1/Frameworks/Python.framework/Versions/3.7/lib/python3.7/importlib/_bootstrap_external.py\u001b[0m in \u001b[0;36mget_data\u001b[0;34m(self, path)\u001b[0m\n",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
     ]
    }
   ],
   "source": [
    "import yaml\n",
    "import numpy as np\n",
    "from pathlib import Path\n",
    "from scipy import sparse\n",
    "from src.first_level.fit import fit_first_level\n",
    "\n",
    "config_file = Path.cwd() / \"configs\" / \"config.yaml\"\n",
    "with config_file.open(\"r\", encoding=\"utf-8\") as file:\n",
    "    config = yaml.safe_load(file)\n",
    "    \n",
    "config['first_level']['models'][0]['model']['encoder']['input_size'] = len(item2idx)\n",
    "config['first_level']['models'][0]['model']['decoder']['hidden_sizes'][-1] = len(item2idx)\n",
    "\n",
    "    \n",
    "train = {\n",
    "    #\"ALS\": train_matrix,\n",
    "    \"MultVAE\": (\n",
    "        train_matrix,\n",
    "        train_matrix,\n",
    "    ),\n",
    "    'EASE': train_matrix,\n",
    "    'SLIM': train_matrix,\n",
    "\n",
    "}\n",
    "valid = {\n",
    "    #\"ALS\": valid_matrix,\n",
    "    \"MultVAE\": (\n",
    "        train_matrix,\n",
    "        valid_matrix,\n",
    "    ),\n",
    "    'EASE': valid_matrix,\n",
    "    'SLIM': valid_matrix\n",
    "}\n",
    "models = fit_first_level(config[\"first_level\"], train=train, valid=valid)\n",
    "print(models)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "2ff8750c",
   "metadata": {},
   "outputs": [],
   "source": [
    "from src.fi\n",
    "\n",
    "\n",
    "model = SAR()\n",
    "data = sparse.csr_matrix(np.random.randint(low=0, high=2, size=(100, 300)))\n",
    "metrics = model.fit(train=data, valid=data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "2f27c397",
   "metadata": {},
   "outputs": [],
   "source": [
    "import hnswlib\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "from tqdm import tqdm\n",
    "\n",
    "import gzip\n",
    "from math import asin, cos, sqrt\n",
    "from multiprocessing import Pool\n",
    "import os\n",
    "\n",
    "import dill\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import torch\n",
    "from tqdm import tqdm\n",
    "\n",
    "\n",
    "def perform_parallel(data, function, n_jobs: int, verbose: bool = True):\n",
    "\n",
    "    with Pool(n_jobs) as p:\n",
    "        chunksize = max(1, len(data) // n_jobs)\n",
    "\n",
    "        result = list(\n",
    "            tqdm(\n",
    "                p.imap(function, data, chunksize=chunksize), disable=not verbose, total=len(data),\n",
    "            )\n",
    "        )\n",
    "    return result\n",
    "\n",
    "\n",
    "class BasePreprocess:\n",
    "    \"\"\"\n",
    "    Class that does processing data from zeppelin notebook to appropriate form for the model\n",
    "    \"\"\"\n",
    "\n",
    "    def __init__(self):\n",
    "        pass\n",
    "\n",
    "    def process_df(self, df: pd.DataFrame, case=1) -> pd.DataFrame:\n",
    "\n",
    "        self.ci = {v: k for k, v in enumerate(df.columns)}\n",
    "        df = self._preparatory_actions(df)\n",
    "        data = self._process_rows(df, case=case)\n",
    "        columns = list(data[0].keys())\n",
    "        data = pd.DataFrame(data, columns=columns)\n",
    "        data = self._final_actions(data)\n",
    "        #data = data.set_index(CUST_ID_COL)\n",
    "        return data\n",
    "\n",
    "    def _process_rows(self, dataset: pd.DataFrame, case):\n",
    "        if case == 1:\n",
    "            function = self._preprocess_rows1\n",
    "        return perform_parallel(data=dataset.values, function=function, n_jobs=20)\n",
    "\n",
    "    def _preprocess_rows(self, val: pd.Series) -> dict:\n",
    "        raise NotImplementedError\n",
    "\n",
    "    def _preparatory_actions(self, df: pd.DataFrame) -> pd.DataFrame:\n",
    "        return df\n",
    "\n",
    "    def _final_actions(self, df: pd.DataFrame) -> pd.DataFrame:\n",
    "        return df\n",
    "\n",
    "\n",
    "class Prep1(BasePreprocess):\n",
    "    def __init__(self):\n",
    "\n",
    "        super().__init__()\n",
    "\n",
    "    def _preprocess_rows1(self, val: pd.Series) -> dict:\n",
    "\n",
    "        y = val\n",
    "        row = {}\n",
    "        for col in self.ci:\n",
    "            row[col] = val[self.ci[col]]\n",
    "            \n",
    "        row['joined_scores'] = row['joined_scores'].split(';')\n",
    "            \n",
    "        return row\n",
    "    \n",
    "p = Prep1()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "31c734d9",
   "metadata": {},
   "outputs": [],
   "source": [
    "models_keys = [x for x in list(models.keys()) if 'ALS' not in x]\n",
    "\n",
    "\n",
    "def get_score(x, col):\n",
    "    \n",
    "    if col != 'target_basket_y':\n",
    "        col = 'preds_' + col\n",
    "    \n",
    "    vector = np.zeros(x['cands'].shape[0])\n",
    "    \n",
    "    for idx, cand in enumerate(range(vector.shape[0])):\n",
    "        vector[idx] = x[col][x['cands'][idx]]\n",
    "  #  print(vector.mean())\n",
    "    return vector\n",
    "    \n",
    "\n",
    "    \n",
    "def join_scores(x):\n",
    "    l = 0\n",
    "    ans = []\n",
    "    while l < len(x['scores_MultVAE']):\n",
    "        \n",
    "        s = [str(x['cands'][l])]\n",
    "        if 'label' in x:\n",
    "            s.append(str(x['label'][l]))\n",
    "        for key in models_keys:\n",
    "            s.append(str(round(x[f'scores_{key}'][l], 8)))\n",
    "        if len(s) > 0:\n",
    "            ans.append(';'.join(s))\n",
    "        l += 1\n",
    "        \n",
    "    return ans\n",
    "\n",
    "def join_label(x):\n",
    "    \n",
    "    vector = np.zeros(x['cands'].shape[0])\n",
    "    \n",
    "    for idx, cand in enumerate(range(vector.shape[0])):\n",
    "        vector[idx] = x[f'preds_{col}'][x['cands'][idx]]\n",
    "  #  print(vector.mean())\n",
    "    return vector\n",
    "\n",
    "def batched_contstruc_2_lvl(df, models, top_cands=30, num_chunks=10):\n",
    "    \n",
    "    res1 = pd.DataFrame()\n",
    "    res2 = pd.DataFrame()\n",
    "    \n",
    "    for chunk in tqdm(np.array_split(df, num_chunks)):\n",
    "        x, y = construct_2_lvl(chunk, models, top_cands)\n",
    "        res1 = pd.concat([res1, x])\n",
    "        res2 = pd.concat([res2, y])\n",
    "        \n",
    "    return res1, res2\n",
    "\n",
    "def construct_2_lvl(df, models, top_cands=30):\n",
    "    \n",
    "    matrix = np.vstack((df.target_basket_x.values))\n",
    "\n",
    "    \n",
    "    for key in models_keys:\n",
    "\n",
    "        model = models[key]\n",
    "        recs = []\n",
    "        preds = model.predict(matrix)\n",
    "        for user in range(matrix.shape[0]):\n",
    "            if key == 'EASE':\n",
    "                recs.append(preds[user].tolist()[0])\n",
    "            else:\n",
    "                recs.append(preds[user].tolist())\n",
    "\n",
    "        df[f'preds_{key}'] = recs\n",
    "        if 'target_basket_y' in df.columns:\n",
    "            metric = NDCG_binary_at_k_batch(preds, np.vstack((df.target_basket_y.values)), 20)\n",
    "            print(key, metric)\n",
    "    \n",
    "    df['cands'] = df.preds_EASE.apply(lambda x: np.argsort(np.array(x))[-top_cands:])\n",
    "        \n",
    "    print('all models predicted')\n",
    "    for key in models_keys:\n",
    "        print(key)\n",
    "        df[f'scores_{key}'] = df.apply(lambda x: get_score(x, col=key), axis=1)\n",
    "\n",
    "    print('we got all scores')\n",
    "\n",
    "    if 'target_basket_y' in df.columns:\n",
    "        df['label'] = df.apply(lambda x: get_score(x, col='target_basket_y'), axis=1)\n",
    "\n",
    "    print('joining scores')\n",
    "    df['joined_scores'] = df.apply(lambda x: join_scores(x), axis=1)\n",
    "    print('exploding')\n",
    "    \n",
    "    user_col = 'buyer_id' if 'buyer_id' in df.columns else  'pav_order_id'\n",
    "    parsed = df[[user_col,'joined_scores']].explode('joined_scores')\n",
    "\n",
    "    #parsed['joined_scores'] = parsed['joined_scores'].apply(lambda x: x.split(\";\"))\n",
    "    parsed = p.process_df(parsed, case=1)\n",
    "    \n",
    "    \n",
    "    \n",
    "    final_cols = ['cands']\n",
    "    if 'label' in df.columns:\n",
    "        final_cols.append('label')\n",
    "    final_cols.extend(models_keys)\n",
    "    \n",
    "    for idx, col in enumerate(final_cols):\n",
    "        parsed[col] = parsed.joined_scores.apply(lambda x: x[idx])\n",
    "        if col == 'cands':\n",
    "            parsed[col] = parsed[col].astype(np.int32)\n",
    "        else:\n",
    "            parsed[col] = parsed[col].astype(np.float32)\n",
    "\n",
    "    parsed.drop(['joined_scores'], axis=1, inplace=True)\n",
    "    return parsed, df\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "9f7f453e",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|          | 0/1 [00:00<?, ?it/s]\n",
      "Reading dataset in memory: 0it [00:00, ?it/s]\u001b[A\n",
      "Reading dataset in memory: 429it [00:00, 4287.48it/s]\u001b[A\n",
      "Reading dataset in memory: 899it [00:00, 4529.66it/s]\u001b[A\n",
      "Reading dataset in memory: 1413it [00:00, 4807.73it/s]\u001b[A\n",
      "Reading dataset in memory: 1894it [00:00, 4680.99it/s]\u001b[A\n",
      "Reading dataset in memory: 2365it [00:00, 4689.13it/s]\u001b[A\n",
      "Reading dataset in memory: 2881it [00:00, 4843.73it/s]\u001b[A\n",
      "Reading dataset in memory: 3406it [00:00, 4975.37it/s]\u001b[A\n",
      "Reading dataset in memory: 3937it [00:00, 5079.78it/s]\u001b[A\n",
      "Reading dataset in memory: 4464it [00:00, 5138.66it/s]\u001b[A\n",
      "Reading dataset in memory: 4979it [00:01, 5083.36it/s]\u001b[A\n",
      "Reading dataset in memory: 5488it [00:01, 4990.77it/s]\u001b[A\n",
      "Reading dataset in memory: 5988it [00:01, 4983.17it/s]\u001b[A\n",
      "Reading dataset in memory: 6487it [00:01, 4976.46it/s]\u001b[A\n",
      "Reading dataset in memory: 6985it [00:01, 4805.74it/s]\u001b[A\n",
      "Reading dataset in memory: 7467it [00:01, 4687.16it/s]\u001b[A\n",
      "Reading dataset in memory: 7937it [00:01, 4595.94it/s]\u001b[A\n",
      "Reading dataset in memory: 8435it [00:01, 4704.85it/s]\u001b[A\n",
      "Reading dataset in memory: 8918it [00:01, 4739.68it/s]\u001b[A\n",
      "Reading dataset in memory: 9393it [00:01, 4726.30it/s]\u001b[A\n",
      "Reading dataset in memory: 9867it [00:02, 4661.25it/s]\u001b[A\n",
      "Reading dataset in memory: 10334it [00:02, 4620.72it/s]\u001b[A\n",
      "Reading dataset in memory: 11182it [00:02, 4774.69it/s]\u001b[A\n",
      "\n",
      "Predict with MultVAE:   0%|          | 0/88 [00:00<?, ?it/s]\u001b[A\n",
      "Predict with MultVAE:   3%|▎         | 3/88 [00:00<00:03, 25.05it/s]\u001b[A\n",
      "Predict with MultVAE:   8%|▊         | 7/88 [00:00<00:02, 31.30it/s]\u001b[A\n",
      "Predict with MultVAE:  12%|█▎        | 11/88 [00:00<00:02, 33.03it/s]\u001b[A\n",
      "Predict with MultVAE:  17%|█▋        | 15/88 [00:00<00:02, 34.17it/s]\u001b[A\n",
      "Predict with MultVAE:  22%|██▏       | 19/88 [00:00<00:01, 35.16it/s]\u001b[A\n",
      "Predict with MultVAE:  26%|██▌       | 23/88 [00:00<00:01, 35.99it/s]\u001b[A\n",
      "Predict with MultVAE:  31%|███       | 27/88 [00:00<00:01, 36.57it/s]\u001b[A\n",
      "Predict with MultVAE:  35%|███▌      | 31/88 [00:00<00:01, 36.84it/s]\u001b[A\n",
      "Predict with MultVAE:  40%|███▉      | 35/88 [00:00<00:01, 36.94it/s]\u001b[A\n",
      "Predict with MultVAE:  44%|████▍     | 39/88 [00:01<00:01, 37.03it/s]\u001b[A\n",
      "Predict with MultVAE:  49%|████▉     | 43/88 [00:01<00:01, 36.59it/s]\u001b[A\n",
      "Predict with MultVAE:  53%|█████▎    | 47/88 [00:01<00:01, 36.27it/s]\u001b[A\n",
      "Predict with MultVAE:  58%|█████▊    | 51/88 [00:01<00:01, 36.36it/s]\u001b[A\n",
      "Predict with MultVAE:  62%|██████▎   | 55/88 [00:01<00:00, 36.38it/s]\u001b[A\n",
      "Predict with MultVAE:  67%|██████▋   | 59/88 [00:01<00:00, 36.47it/s]\u001b[A\n",
      "Predict with MultVAE:  72%|███████▏  | 63/88 [00:01<00:00, 36.36it/s]\u001b[A\n",
      "Predict with MultVAE:  76%|███████▌  | 67/88 [00:01<00:00, 35.54it/s]\u001b[A\n",
      "Predict with MultVAE:  81%|████████  | 71/88 [00:01<00:00, 35.40it/s]\u001b[A\n",
      "Predict with MultVAE:  85%|████████▌ | 75/88 [00:02<00:00, 35.15it/s]\u001b[A\n",
      "Predict with MultVAE:  90%|████████▉ | 79/88 [00:02<00:00, 35.04it/s]\u001b[A\n",
      "Predict with MultVAE:  94%|█████████▍| 83/88 [00:02<00:00, 34.95it/s]\u001b[A\n",
      "Predict with MultVAE: 100%|██████████| 88/88 [00:02<00:00, 35.65it/s]\u001b[A\n",
      "/usr/local/lib/python3.7/site-packages/ipykernel_launcher.py:28: RuntimeWarning: invalid value encountered in true_divide\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "MultVAE 0.08734730214698307\n",
      "EASE 0.15325907465829652\n",
      "SLIM 0.04455862678729088\n",
      "all models predicted\n",
      "MultVAE\n",
      "EASE\n",
      "SLIM\n",
      "we got all scores\n",
      "joining scores\n",
      "exploding\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "  0%|          | 0/335460 [00:00<?, ?it/s]\u001b[A\n",
      "  0%|          | 1/335460 [00:04<411:18:50,  4.41s/it]\u001b[A\n",
      "  0%|          | 2/335460 [00:05<237:37:57,  2.55s/it]\u001b[A\n",
      "  0%|          | 3/335460 [00:07<201:22:33,  2.16s/it]\u001b[A\n",
      "  0%|          | 4/335460 [00:07<126:36:35,  1.36s/it]\u001b[A\n",
      "  0%|          | 5/335460 [00:09<85:51:34,  1.09it/s] \u001b[A\n",
      "  0%|          | 6/335460 [00:10<134:51:01,  1.45s/it]\u001b[A\n",
      "  0%|          | 7/335460 [00:11<131:44:07,  1.41s/it]\u001b[A\n",
      "  0%|          | 8/335460 [00:12<100:25:12,  1.08s/it]\u001b[A\n",
      "  0%|          | 9/335460 [00:14<80:57:31,  1.15it/s] \u001b[A\n",
      "  0%|          | 10/335460 [00:14<109:41:51,  1.18s/it]\u001b[A\n",
      "  0%|          | 11/335460 [00:14<81:01:20,  1.15it/s] \u001b[A\n",
      "  0%|          | 12/335460 [00:15<60:55:26,  1.53it/s]\u001b[A\n",
      "  0%|          | 13/335460 [00:15<82:56:57,  1.12it/s]\u001b[A\n",
      "  6%|▌         | 20635/335460 [00:16<00:33, 9320.47it/s]\u001b[A\n",
      " 13%|█▎        | 43820/335460 [00:16<00:13, 22422.96it/s]\u001b[A\n",
      " 16%|█▌        | 52037/335460 [00:17<00:20, 14060.82it/s]\u001b[A\n",
      " 20%|█▉        | 66745/335460 [00:17<00:12, 21358.93it/s]\u001b[A\n",
      " 45%|████▌     | 150958/335460 [00:17<00:02, 78687.78it/s]\u001b[A\n",
      " 55%|█████▌    | 184504/335460 [00:18<00:02, 50937.59it/s]\u001b[A\n",
      " 70%|███████   | 234823/335460 [00:18<00:01, 78138.28it/s]\u001b[A\n",
      " 80%|████████  | 268369/335460 [00:20<00:01, 54705.90it/s]\u001b[A\n",
      "100%|██████████| 335460/335460 [00:20<00:00, 16617.78it/s]\u001b[A\n",
      "100%|██████████| 1/1 [01:44<00:00, 104.01s/it]\n"
     ]
    }
   ],
   "source": [
    "train_boosting, _ = batched_contstruc_2_lvl(train_2, models, num_chunks=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "3975fdca",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|          | 0/3 [00:00<?, ?it/s]\n",
      "Reading dataset in memory: 0it [00:00, ?it/s]\u001b[A\n",
      "Reading dataset in memory: 217it [00:00, 2168.44it/s]\u001b[A\n",
      "Reading dataset in memory: 551it [00:00, 2854.87it/s]\u001b[A\n",
      "Reading dataset in memory: 981it [00:00, 3511.30it/s]\u001b[A\n",
      "Reading dataset in memory: 1388it [00:00, 3729.30it/s]\u001b[A\n",
      "Reading dataset in memory: 1849it [00:00, 4045.35it/s]\u001b[A\n",
      "Reading dataset in memory: 2313it [00:00, 4246.98it/s]\u001b[A\n",
      "Reading dataset in memory: 2770it [00:00, 4351.39it/s]\u001b[A\n",
      "Reading dataset in memory: 3227it [00:00, 4418.71it/s]\u001b[A\n",
      "Reading dataset in memory: 3719it [00:00, 4074.23it/s]\u001b[A\n",
      "\n",
      "Predict with MultVAE:   0%|          | 0/30 [00:00<?, ?it/s]\u001b[A\n",
      "Predict with MultVAE:   3%|▎         | 1/30 [00:00<00:05,  5.29it/s]\u001b[A\n",
      "Predict with MultVAE:  17%|█▋        | 5/30 [00:00<00:01, 18.48it/s]\u001b[A\n",
      "Predict with MultVAE:  30%|███       | 9/30 [00:00<00:00, 24.73it/s]\u001b[A\n",
      "Predict with MultVAE:  43%|████▎     | 13/30 [00:00<00:00, 28.43it/s]\u001b[A\n",
      "Predict with MultVAE:  57%|█████▋    | 17/30 [00:00<00:00, 30.64it/s]\u001b[A\n",
      "Predict with MultVAE:  70%|███████   | 21/30 [00:00<00:00, 32.45it/s]\u001b[A\n",
      "Predict with MultVAE:  83%|████████▎ | 25/30 [00:00<00:00, 33.19it/s]\u001b[A\n",
      "Predict with MultVAE: 100%|██████████| 30/30 [00:01<00:00, 29.94it/s]\u001b[A\n",
      "/usr/local/lib/python3.7/site-packages/ipykernel_launcher.py:28: RuntimeWarning: invalid value encountered in true_divide\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "MultVAE 0.08796852828677772\n",
      "EASE 0.15155855652276434\n",
      "SLIM 0.046032334222727046\n",
      "all models predicted\n",
      "MultVAE\n",
      "EASE\n",
      "SLIM\n",
      "we got all scores\n",
      "joining scores\n",
      "exploding\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "  0%|          | 0/111570 [00:00<?, ?it/s]\u001b[A\n",
      " 15%|█▍        | 16735/111570 [00:28<02:41, 587.57it/s]\u001b[A\n",
      " 15%|█▌        | 16794/111570 [00:28<02:42, 583.33it/s]\u001b[A\n",
      " 15%|█▌        | 16818/111570 [01:05<09:43, 162.42it/s]\u001b[A\n",
      " 15%|█▌        | 16818/111570 [01:05<09:43, 162.42it/s]\u001b[A\n",
      " 15%|█▌        | 16819/111570 [01:05<09:48, 160.92it/s]\u001b[A\n",
      " 15%|█▌        | 16830/111570 [01:39<22:51, 69.05it/s] \u001b[A\n",
      " 15%|█▌        | 16830/111570 [01:39<22:51, 69.05it/s]\u001b[A\n",
      "100%|██████████| 111570/111570 [01:39<00:00, 1119.47it/s][A\n",
      " 33%|███▎      | 1/3 [02:01<04:02, 121.44s/it]\n",
      "Reading dataset in memory: 0it [00:00, ?it/s]\u001b[A\n",
      "Reading dataset in memory: 461it [00:00, 4603.28it/s]\u001b[A\n",
      "Reading dataset in memory: 461it [00:29, 4603.28it/s]\u001b[A\n",
      "Reading dataset in memory: 697it [00:29, 19.04it/s]  \u001b[A\n",
      "Reading dataset in memory: 1143it [00:29, 40.56it/s]\u001b[A\n",
      "Reading dataset in memory: 1595it [00:29, 71.21it/s]\u001b[A\n",
      "Reading dataset in memory: 2002it [00:29, 109.99it/s]\u001b[A\n",
      "Reading dataset in memory: 2461it [00:29, 171.23it/s]\u001b[A\n",
      "Reading dataset in memory: 2922it [00:29, 256.25it/s]\u001b[A\n",
      "Reading dataset in memory: 3719it [00:30, 123.66it/s]\u001b[A\n",
      "\n",
      "Predict with MultVAE:   0%|          | 0/30 [00:00<?, ?it/s]\u001b[A\n",
      "Predict with MultVAE:   3%|▎         | 1/30 [00:00<00:05,  5.29it/s]\u001b[A\n",
      "Predict with MultVAE:  13%|█▎        | 4/30 [00:00<00:01, 15.62it/s]\u001b[A\n",
      "Predict with MultVAE:  27%|██▋       | 8/30 [00:00<00:00, 22.42it/s]\u001b[A\n",
      "Predict with MultVAE:  37%|███▋      | 11/30 [00:00<00:00, 20.64it/s]\u001b[A\n",
      "Predict with MultVAE:  47%|████▋     | 14/30 [00:00<00:00, 20.42it/s]\u001b[A\n",
      "Predict with MultVAE:  57%|█████▋    | 17/30 [00:00<00:00, 19.54it/s]\u001b[A\n",
      "Predict with MultVAE:  67%|██████▋   | 20/30 [00:01<00:00, 21.98it/s]\u001b[A\n",
      "Predict with MultVAE:  77%|███████▋  | 23/30 [00:01<00:00, 23.30it/s]\u001b[A\n",
      "Predict with MultVAE:  87%|████████▋ | 26/30 [00:01<00:00, 24.04it/s]\u001b[A\n",
      "Predict with MultVAE: 100%|██████████| 30/30 [00:01<00:00, 22.46it/s]\u001b[A\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "MultVAE 0.08980913768819553\n",
      "EASE 0.15261857326839218\n",
      "SLIM 0.04244653822070713\n",
      "all models predicted\n",
      "MultVAE\n",
      "EASE\n",
      "SLIM\n",
      "we got all scores\n",
      "joining scores\n",
      "exploding\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "  0%|          | 0/111570 [00:00<?, ?it/s]\u001b[A\n",
      "  5%|▌         | 5579/111570 [00:42<13:30, 130.81it/s]\u001b[A\n",
      "  5%|▌         | 5593/111570 [00:42<13:31, 130.54it/s]\u001b[A\n",
      "  5%|▌         | 5602/111570 [01:22<39:01, 45.25it/s] \u001b[A\n",
      "  5%|▌         | 5602/111570 [01:22<39:01, 45.25it/s]\u001b[A\n",
      "  5%|▌         | 5603/111570 [01:22<39:17, 44.96it/s]\u001b[A\n",
      " 10%|▉         | 11007/111570 [02:06<18:11, 92.11it/s]\u001b[A\n",
      " 10%|▉         | 11020/111570 [02:06<18:12, 92.07it/s]\u001b[A\n",
      " 70%|██████▉   | 78093/111570 [02:06<00:17, 1947.33it/s]\u001b[A\n",
      " 85%|████████▍ | 94827/111570 [02:47<00:16, 1001.57it/s]\u001b[A\n",
      "100%|██████████| 111570/111570 [02:47<00:00, 667.41it/s]\u001b[A\n",
      " 67%|██████▋   | 2/3 [05:41<02:59, 179.57s/it]\n",
      "Reading dataset in memory: 0it [00:00, ?it/s]\u001b[A\n",
      "Reading dataset in memory: 534it [00:00, 5338.12it/s]\u001b[A\n",
      "Reading dataset in memory: 1068it [00:00, 5167.19it/s]\u001b[A\n",
      "Reading dataset in memory: 1595it [00:00, 5210.96it/s]\u001b[A\n",
      "Reading dataset in memory: 2117it [00:00, 5040.67it/s]\u001b[A\n",
      "Reading dataset in memory: 2746it [00:00, 5478.09it/s]\u001b[A\n",
      "Reading dataset in memory: 3719it [00:00, 5591.22it/s]\u001b[A\n",
      "\n",
      "Predict with MultVAE:   0%|          | 0/30 [00:00<?, ?it/s]\u001b[A\n",
      "Predict with MultVAE:   3%|▎         | 1/30 [00:00<00:04,  6.44it/s]\u001b[A\n",
      "Predict with MultVAE:  13%|█▎        | 4/30 [00:00<00:01, 17.60it/s]\u001b[A\n",
      "Predict with MultVAE:  27%|██▋       | 8/30 [00:00<00:00, 25.99it/s]\u001b[A\n",
      "Predict with MultVAE:  40%|████      | 12/30 [00:00<00:00, 30.93it/s]\u001b[A\n",
      "Predict with MultVAE:  57%|█████▋    | 17/30 [00:00<00:00, 36.18it/s]\u001b[A\n",
      "Predict with MultVAE:  70%|███████   | 21/30 [00:00<00:00, 37.27it/s]\u001b[A\n",
      "Predict with MultVAE:  83%|████████▎ | 25/30 [00:00<00:00, 37.47it/s]\u001b[A\n",
      "Predict with MultVAE: 100%|██████████| 30/30 [00:00<00:00, 33.45it/s]\u001b[A\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "MultVAE 0.08558744114761602\n",
      "EASE 0.14645585147858\n",
      "SLIM 0.045238885039640256\n",
      "all models predicted\n",
      "MultVAE\n",
      "EASE\n",
      "SLIM\n",
      "we got all scores\n",
      "joining scores\n",
      "exploding\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "  0%|          | 0/111570 [00:00<?, ?it/s]\u001b[A\n",
      "  0%|          | 1/111570 [00:48<1503:08:25, 48.50s/it]\u001b[A\n",
      "  0%|          | 2/111570 [00:48<621:11:34, 20.04s/it] \u001b[A\n",
      " 10%|█         | 11157/111570 [00:48<03:35, 465.64it/s]\u001b[A\n",
      " 16%|█▌        | 17338/111570 [01:37<07:23, 212.27it/s]\u001b[A\n",
      " 16%|█▌        | 17338/111570 [01:38<07:23, 212.27it/s]\u001b[A\n",
      " 16%|█▌        | 17351/111570 [01:38<07:24, 211.81it/s]\u001b[A\n",
      " 40%|███▉      | 44625/111570 [02:27<02:41, 415.32it/s]\u001b[A\n",
      " 40%|███▉      | 44625/111570 [02:27<02:41, 415.32it/s]\u001b[A\n",
      " 40%|████      | 44680/111570 [02:27<02:41, 415.35it/s]\u001b[A\n",
      " 55%|█████▍    | 61109/111570 [02:28<01:01, 820.20it/s]\u001b[A\n",
      " 80%|███████▉  | 89249/111570 [02:28<00:12, 1806.31it/s]\u001b[A\n",
      " 90%|████████▉ | 100405/111570 [03:17<00:16, 678.62it/s]\u001b[A\n",
      "100%|██████████| 111570/111570 [03:17<00:00, 564.24it/s]\u001b[A\n",
      "100%|██████████| 3/3 [09:21<00:00, 187.09s/it]\n"
     ]
    }
   ],
   "source": [
    "test_boosting, test2_mod = batched_contstruc_2_lvl(test_2, models, num_chunks=3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "fc9974cd",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.7533032212768079"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from  matplotlib import pyplot as plt\n",
    "import seaborn as sns\n",
    "%matplotlib inline \n",
    "\n",
    "\n",
    "def decode_scores(x):\n",
    "    vector = np.zeros(len(item2idx))-2\n",
    "    for el in x.joined:\n",
    "        t1,t2 = el.split('_')\n",
    "        vector[int(float(t1))] = float(t2)\n",
    "    return vector\n",
    "    \n",
    "    \n",
    "    \n",
    "def plotImp(model, X, num=20, fig_size=(40, 20)):\n",
    "    feature_imp = pd.DataFrame({\"Value\": model.feature_importances_, \"Feature\": X})\n",
    "    feature_imp = feature_imp.sort_values(by=\"Value\", ascending=False)[0:num]\n",
    "\n",
    "    plt.figure(figsize=fig_size)\n",
    "    sns.set(font_scale=5)\n",
    "    sns.barplot(\n",
    "        x=\"Value\", y=\"Feature\", data=feature_imp,\n",
    "    )\n",
    "    plt.title(\"LightGBM Features (avg over folds)\")\n",
    "    plt.tight_layout()\n",
    "    plt.savefig(path)\n",
    "    plt.show()\n",
    "\n",
    "    feature_imp.set_index(\"Value\", inplace=True)\n",
    "\n",
    "    \n",
    "import xgboost\n",
    "from sklearn.metrics import roc_auc_score as auc\n",
    "\n",
    "\n",
    "FEATURES = models_keys.copy()\n",
    "\n",
    "model = xgboost.XGBClassifier()\n",
    "model.fit(train_boosting[FEATURES], train_boosting['label'])\n",
    "test_boosting['preds'] = model.predict_proba(test_boosting[FEATURES])[:,1]\n",
    "test_boosting['joined'] = test_boosting.apply(lambda x: f'{x.cands}_{x.preds}', axis=1)\n",
    "test_grouped = test_boosting.groupby('buyer_id').agg({'joined':list}).reset_index()\n",
    "test_grouped['preds_boosting'] = test_grouped.apply(lambda x: decode_scores(x), axis=1)\n",
    "test_grouped.head()                                   \n",
    "auc(test_boosting['label'], test_boosting['preds'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "48841f07",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAACswAAAUsCAYAAACZ8SBTAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjQuMywgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/MnkTPAAAACXBIWXMAAAsTAAALEwEAmpwYAADmDElEQVR4nOzddZhtV3k/8O+9USIQg6Ah6Iu2aJDiDqW4k+BupQKlLUUKlJYWSosE+RES3CkOwQKhUDxIAwsNBGiQuEHs/v7Y58LkzD4jZ87I3ffzeZ55YNbee+01M9tuzne/a9OWLVsCAAAAAAAAAAAAAEO1eb0HAAAAAAAAAAAAAACrSWAWAAAAAAAAAAAAgEETmAUAAAAAAAAAAABg0ARmAQAAAAAAAAAAABg0gVkAAAAAAAAAAAAABk1gFgAAAAAAAAAAAIBB23G9BwAAAAAAW1XVpiQHJLlGkkskuWiSPZNsSXJmkpOSHJfke621X63TMAHYgKrqikkeMqfpv1trH1+v8QAbQ1XtluSgJFdNsne6z0dPT3JquufKHyb5eWtty3qNkelV1Y5J/ibJTqOmE1prr1rHIQEAABuYwCwAAAAwM1V1eJKHjre31jat8xiOaK09bK3GMK2q6vuQ/tattaPWeixrqaqumuSuo68bJtljidv9KMlnkrwnyZGttXNmMJajktxyBV1sSXJ2krOSnJDkp0mOSfL5JJ9srf12heM7PD3n2BznJNm/tXbKSvazyBiekeSFC6zyZ621D67W/ueM4/As/LtYDQ9vrR2+xvsElqCqdkjy5iQ3HjWdk+7lC2A7VVWXTfKPSR6Q5CILrHpuuufPFT9LzsKE59Hnttaes4ZjeE6SZ481f6a1dqu1GsNStdbOq6r9kzx5a1tVndRae8c6DgsAANigNq/3AAAAAADY2Krq1lX1NzPuc3NV3bOqjk7Skrw4ya2zxLDsyBWTPDzJB5IcX1XPqqq9ZznOKWxKsluS/ZJcK8ldkvxdkg8mOaGq/rOqLrmK+985yT1Wsf8keeAq98/AVNUlquoN6z0OBu/v84ewbJL8R2vth+s1GGB9VdVN07209PAsHJZNku/M4sUr1tVz0s1EsdWhVXXpdRoLAACwgQnMAgAAANCrqi5ZVW9O8qkkV59hvzdN8s10lWFvNqNuL5HkuUm+V1UPn1Gfs3axdJWvvldVB6/ifu67Wh1X1dWT/NFq9c+wjILxT0wXij9kvcfDcFXVDZP8w5ymXyd5/joNB1hnVXX5JB9Osu8SN/n6Kg6HNdBaOyldNeGt9klyeFWt2UwnAADAtkFgFgAAAIALqaodquop6UJuD5phvztV1cuSfC7JNZewyflJfpPkJ0l+nuT0JWyzX5LDquqIqtpl6sGurj2TvLGq/m6V+r99Ve21Sn0/YJX6ZWCq6qAkX07y8iR7re9oGLKq2inJ4Ul2nNP8j62109ZnRMAG8K/pXlTqsyXJCemeL08etR2zBmNi9R2aZG5l8dsnefQ6jQUAANigBGYBAAAAGPfvSf4jyUVn1WFV7Zmu0teTkkyq9PSbJK9OF8q8UpJdW2sXb60d2Fq7bGvtoukCsXdMF4Q4foFdPiTJO6pqxwXWWW8vqKqZBZLn2CnJPVeh3yS5/yr1y4BU1Q2SfCHJ9dZ7LGwXnprkGnO+/366ewmwHaqqfZLcq2fRWekq/e/VWrvU6Plyn3TB2v+3lmNkdbTWzknyt2PNL6yqpVYaBgAAtgMb+QMDAAAAANbHzIKySVJVuyf5dJLrT1jlR+mmzn5Ta+3chfpqrZ2Y5MgkR1bVM9IFOF+Q5Ao9q98tyb8k+asph77VT1prBy620qjK4U7pfn/7JDkwXWDwnpkcHHxVVX2ytfbLFY5x3H2TvH6WHVbVdZPULPtcDa01U++uvz2iWANroKouk+RZY83PWuxeAgzaDZPs0NP+F62114w3qkY9OO9K8rX84dl7nyQvTPKYdRsRAACwoQjMAgAAAIPSWntYkoet8zC4sMPTH5bdkuRFSZ7TWvvtcjttrV2Q5K1V9b50U7A+pGe1v6iq97fWPrPc/qcYz7lJzk1XweyEJMemq6r7/Kp6YJLXJbnI2GZ7Jnlmuopn0zo186cdvl1V7d1aO7lvgyk9YIZ9AczCi9MFtLf6VpK3r9NYgI3hwJ62c5O8aY3HwTporW2pqn9I8qE5zY+sqte21r68XuMCAAA2Dm/5AwAAALBqquqvk9ynZ9G5Se7fWnvGNGHZuVprZ7XWHpouNDtuU5IXV9W6Vh1trb01XTXcPg+tqt1W0P0PkvzvWNtO6SrbztL9xr4f3yfAmqmq62f+dfWfWmtb1mM8wIaxd0/bD1prZ635SFgXrbUPJzlmTtPmdLNOAAAACMwCAAAAsDqq6nJJnjth8aNba++c8S6fnOSrPe3XT3LnGe9r2VprH0jy/p5Feya5zQq776uoeN8V9vl7VXWTzK/Y9tZZ9Q8wheeMfX98klnfV4Btz849bbOsuM+24cVj39+6qm6xLiMBAAA2FIFZAAAAAFbLvyTpq5z6utbaEbPeWWvt/CRPmLD48bPe35ReN6H9Bivsty8we9uq2meF/W71gLHvz07yvhn1DbAsVXWDJHcdaz58dB8Atm99n32eu+ajYL29O8mpY23PWYdxAAAAG4zALAAAAAAzV1XXSPLAnkXHp6sEuypaa19K8tGeRXeqqr4petfalya077+STltr38uFp55Nkp2S3HMl/SZJVW3O/Gq1H0hyxkr7BpjSP4x9vyXJYesxEAA2ntba2UneMtZ866q6+XqMBwAA2Dh2XO8BAAAAAGx0VbVnkusluUqSvdMFEU9O8p0kXxx9ILvWY7pSkusmuVyS3dNV/PxNku8m+Vprbb0raT1xQvs/rsHv69Akdxpr2zHJLZP81yrvezG/mdB+wQz6fnuS64y13TeTq9ou1a2SXGqs7a0r7HO7VFWV5JpJLptkzyTnp6t+dly68/aXq7z/HZJUkisnufRoDLumq7x3VpITkvw4ybdaa79dzbEMRVX9Sbq/6b5JTkzyzSRfaa2dN0Vfu+YP95qLJ7lIumv7r5J8P90x8rsZDX1837sluX6SK6W7z+2e5Lfp7nU/SfLN1tqvVmPfy1VVl8/86rKfbK0dN+P9XDJ/OF/3Tff3SLrfy4lJfpbuXFnV83atVdWmdNeI66Z7meOi6X7mn6U7tn+4jsNbltHPcs1017390v0dz093XP843Tl14vqNcHWNqsxfL8kV84fn1zOT/DzJsUn+t7W2ZQ3GMbPr5EYwupdeJ931cr90v9tz0v1sP0ry1dba6es2wAmq6mLpZjS4croxb0l3f/lmkq+31mbxLLrQ/tfjPvO6zJ9l4olJjp7xfgAAgG2IwCwAAAAwKFV1eJKHjjUf0Vp72DL72ZTk3kkeleR2SXaYsOpvq+o9SV7UWvvGnO2PSfLHc1dsrW1azhh6xnSxJE9K8vB0HzZPclpVvT/Jv7fWvraEfp+T5NmLrPbQqhr/vf6ktXZgT397Jjmkp4+fJjl8sfHMwEeSnJIuhPqJJEcm+Vhr7WdrsO/F7DGh/ecz6PsdSV441nbbqtqntXbSCvp9wNj3p6b7HY+HaOlRVddOF9a4d5JLLLLut9OFkV+1wr/Z3D73TFft+R7pQuO7LWGzc6vq6CRvTvKWxcKzVXVguvDZYmPpC2bdurV21Nh6h2cG1/GV9FdVt0ry6bHmv22t/fNo+R2TvDhdCGzciVX1liQvWSzEOQpe3TfJw5LcOsnOC6z+u6r6eLpKqv+10qDb6J7y4NHXjTL5Prd1/e8neW+S17bWfrCSfa/QYzJ/9rw3zqLjqrpNumvendK9kLKUbb6f5D1JXt1aW8p58OV0obW5jm6t3WKZwx3vd78kv0gXipzr4Nbam5ew/WXTVYB/cJLLLLDed5O8Mt3Pe86o7R7pjo25nttae85Sxz9LVXXLJE9Icpt0gcaF1v1Gkrclec1Sr7tV9a9J/nqs+awk+7fWVlR9vaq+nvkvv/y/1tqjl7j9RdM9uz44Xeh5oWfPk6rqfUle2Vr7yhRjvVXW4Dq5yBgOzBLuP0luOeEe1Pss27OfTemC+o9Jcot0QfJJLqiq/0lX3fTw1tqZSxjfqpjz75lHp/v3zKSZR0+oqlcn+beVHsNj+1/X+0xr7atV9b+58DF4z6q6xEZ5CQQAAFh7k/5hBAAAALDdqqo/TvLFJO9Mcscs/OHurkkelORrVfWSqloo7LSSMR2SrmrV87NwWDbpPsQ/OMlXqupVo6qFa+nO6SpXjnvLWlQTG1XXvV6Si7fW7t9ae90GCcsm80MwW/3PSjturf0oyZfHmndMcs9p+6yqnZLca6z5PatV5XJIqurSVfW2JN9IF5hdMCw7cq0kL0jyk6p6+ihMOe3+L1JVz0oXxn51uvNyKWHZpAvc3SZdZbbvV9X4MbBdq6oHJ/lw+kNgSVdF8cmZX+l6vJ87pKtU/tZ095rF7h+7pAtsvSfJMaOqjVOpqkcm+UGSVyS5aRYJMY1cJcnTk7SqOqKqLj7t/qdVVTsmecRY85YkH11hv3esqm8l+WS6YNmSwrIjV0nyN0m+V1X/UVUXWWT91/e03WxUOXclHpD5YdlT0x0vE1XVDqMXZ76X7u87MSw7crUk/5nkf6vqoOmGujqq6kZV9aUkRyW5XxYJy478cbqXTX5aVX+/xOe4vr/hbpl/v1yWqrpW+p8TDlvCtpur6inpKnW+ON1z0GIvau2T7iWsL1fVe6vqgOWNeMHxzOQ6uRFU1V3SVeR9f7pr8EJh2aT77PWmSV6e7n7+uFFwdU1V1XXTPRe+M8kdsvBnwpdM9/LcN6vq5jPa/0a5z3xk7Pud0x33AADAdkpgFgAAAGCOqnpQki8kueEyN92c5C+SHDmq7jUrm6vq0CRvSBdsWI5NSR6b5JOjaVDXyl0mtL99rQbQWvvxak8tO6XxKpdJckKSz8yo/77f8f1W0N/t04Va5nrrCvrbLlTV7dNNcXz/LB5Y6rNHkn9J8qlpwiJVdaUkX0ry3PSH15fjskneXVVPW2E/g1BVN04Xllvsv62fnQnnyiig+JIkH0sXDprGHyX5TFX97XI2qqpNo33/vywtTNhnc5KHpHsp42pT9jGtO6cLds319WkrBY7+Fi9LF7i91grHtmOSpyT5dFVNqiaedFUnx6s2b0pXgXElHtLT9rbW2tmTNqiqvdNVYn92ksWCvuOunOToqrrvMrebuTmh389l+c9vW+2e7qWk/14sONpaOzbdi1XjDp5y31v1/Q1ba+0LC200qi58ZJL/SLLXlPu+R7og/ooDrLO4Tm4EoxdPXpnkQ+mC4tPYN8mhST48Ot/WxGhWiC8kuf4yN71CkqOq6v4r2PdGu88c2dM2/uIFAACwHRGYBQAAABgZfTj8xiwcGjk7yc9G/9vnlukqUO0yo2H9e5LHTVh2apKfJjllkT5umu7D+rVyx562k9JV2txuVdVt0x+G+dfW2vkz2s070lVbnOs2VTUeel2qB4x9/6skn5qyr+1CVd0nXbhmod/579JVfv11koWC3bdIF0hbSnXarfu/TLopshcK/12Q5DfpKhGekPnhvT4vrKrlBm+GZrckb878Kp593tVaO3W8cVQ1+G3pXrBYyGnpru8LTY29Q5J/qqp/X8J4tvqrRfZ9Qbrj8rh0x8a5C6x7QLoQ2EpD2ctx9562lVSXfW2SJy2yzunpztefprvvLuZG6YKXvVprp6Sbcnzc1IHZqqr0B0X7KqFu3WaPdJUXb7VA11vSXfd/nfn3lqSr1PjWJH+61LHO2qgS+tvShX53XGDVc9I9vy32N7xBks8vIaTXV/X1NlV1qUW261VVm9PNWDBu4t9wtN0lk3w2yW0XWO2CJCem+/kXqhC/d5IPjO5j01rxdXIjGF3XjkxXIX4hZyc5Pgtfq5Ouku7Ro7/Xqqqqh6Q7bhb6t8iZ6Y6Hc3qWbU73ot6tphzCRrvPHJ35/2676ui6CQAAbIcEZgEAAACSVNUfJTki/f+95OQkz0xyldbabq21y7XWdktXGfDv04UQ5rplpq9ENe6Pxr7/QboKdge01vZqrV2+tbZ3kkun+1D/pxP6ecgCUye/J8khc76O7lnn6LF1Dkny1PGVqurSmV/9L0k+31rrC9sM3qjS1kOTvC/zj69j0k1VOxOttePTVRSba8ck91xuX1W1a+aH0945w3Dv4IyuI29Mf1Dop0n+Osk1W2u7ttYu21q7RLpAyy3TVQbsC65WkreOgpZLcUT6p5M/Nck/JTkoyR6ttYu31g5srV0qXcDpKummov/qhH53SPJvE5b9Ohe+NvzThPXGryGHJPnOYj/QBvLEJFec8/356QJVL0ryr+mC0ltDOZOmUH9ekklhtCNHy/ZtrV1sdH3fM134+sFJPjlhu6dW1aLTS1fVVZO8oGfRWaPxH5Rk99baJVprVxgdG3skuXG6ly7O69n2Cunuj6tuNKX5XXsWfWzK/g7O5Gm535vuurl/a+2io/P18q21vdJVe79TkjdlctDrSVV1xQnLkv4Q5DWq6npLG/08fS9jHNta66uCutVr04V7+3w0XcXR3Vpr+4+uVbuN2j48tu4OSR61rNHO1isy+Zz6Wbrp3au1tsvo+W2vJBdNV319fKr2rS6T5KNVtf8C+31b5ofwdkjywKUOfMxtR/ud6/x0wcVeo/vCO5JcvWfxOen+xrdLd17vN/r5d013HXtKku/3bLdjkiOqatqKy7O4Ti7H+P3nkPQH0r/bs96kZ9nN6X6vN5uwz+8keXK6Z/HdWmsHjK7VF0/ysMx/Dtvqmkk+tJqzPlTVddL93fuq2/863flwYGttj9ba5ZLsmuTaSV6YLkS71c7pnk2Wu/8Nd59prf02Xah83N2m7RMAANi2LfS2LQAAAMB2YfTB+BvSX4npo0ke2jfdc2vtB+mq+706yevSX/luVi5IVz3tn1tr8z5Mbq39X5JXVdXb0gUFbtXTx6PTTdM+vu03000fnySpqtslufnYaj9qrb1pCeO85oT2bSkUN5VRmGvHdNM675vkwHQBgAem//fy8yT3bq0tVO1tGm9PV1V4rvulmxp3Oe6SLlQ014adOnm9jao1vjNd+GSuC9IFUZ7bWpsXrhudz59N8tnRFMZvyPyQym2S/GO6gP5CY7hr+qsMfi3JXVprv+zbbhRm/0GSH1TV69IFgV6a+YGbW1bVpUbXm7nbn5kuPLh1HLdK8nc9+1nKNWQj22fO//9Wkvu11r47d4WqumiShyb5zPjGVXXHJM/o6fcXSQ5urX26b6ettZOSvCXJW6rq7ulC0RcbW+0VVfXV0fV8ksenC0HNdUKS27TWeq/RrbVz0k09/8WqekO6sOT4tOKPr6rnttbOWmDfs3CDJOMBxvOTfHm5HY1eCPjnnkVnJ7l/a+0Dk7ZtrZ2cLqT7sap6cZJ358IBwaQLTt43yb9M6OaT6UL0B4y1H5zufF2y0b2nrzrtQtVl7535FcST7ud/eGvt7eMLRqGz9yV5X1XdM921ao/ljHXWqup+6Z5t+rw0yd/3HZettdPTXa/fWVV3SfezjFcFv3y63+Fd+jpvrZ1WVe9O9zeb6+AkL1nqzzDHIT1tHx2/3o55XuY/ryVdYPNBrbXj+jZqrf04ycuq6lXpApTPy4Wv97sleXdVXb+1tljl1HEruk4u1/j9Z9T/lTP/RaFfLuMe9NfpQvHjLkh3H/63Cc/iv0kXNn5DuhD7oZk/a8X1krw4i1euXbbRv2eOyPzrfNIFlR8yup/MHfOWJN9O8ndV9Zp0QfBJQfql2Kj3mS9k/gwYf5ouxAsAAGxnBGYBAAAAuqDJH/e0vy/JfftCbnO11k4chU+OyAqmVF7Ewa21RcOKrbVTquq+Sf43yfg07uMfFK+GvipnSRfG21ZdvqpmXR33m0nu1Vr70Yz7TboQ0L/nwtVsb11V+43CHEs1Hqb6aZLPr3Rwq20V/lbXba0ds4T1npnkqmNtW9IF7pcU0mmt/XQUqnxbukqOcz2tqg5trf1sgS7+sqftpCR/Niks2zOGLUn+s6oum+RpY4s3pbuOHL6UvgbsB0luN+FFitOSvGy8fTRtfF/Vv58kuWlr7RdL2XFr7X1VdbMkn0pXzXCri6QLZ955gc3v29P2pEkhpp59/09VPSLzqzfume646KvqOEvjLwIkXRXV8SqfS3G/zK/mmSR/sVBYdlxr7ZhRiPkbmV9B/M6ZEJhtrV1QVYcnedbYogdW1dOWWcn7lunCnXOdl67a9TyjqqR9VaB/m+TOrbVFQ4yttfdW1W2TfDzzX6xYE1V1kXT3uj5/2VqbtOxCWmsfHp1Tn8j8Y+LOVXXv1tq7J2x+WOYHZq9bVddorR27lP0nSVXtnuRePYsWCj1fPf0B/A8nuXtfoHPc6Pn2BVX1o3Sh07nH8FXTVYudFPpezLKvkxtBVV0q88/LpAvnP7C19s7F+hjdR4+oqh+mC6qOnyOPrarXtda+suIBX9j9M39miiT5YLpnzsX+PXNcVd0+3QsBN5lyDBv1PtP3uz6oqnZcyrkCAAAMS98UgwAAAADbm77AwfHpgm4Lfri81Sjc8qisTiXVQ5cSlp0zlt8keVXPostV1XiIdtb2m9C+nKDmkJ2VrlLojVprP1yNHYyq0R091rxj5ldbm2gU3vnTsea3jUIgjBlVl31sz6J/XW5V1VHF4UOSjB8fO6U/ELt1DJdKf2Xpf15qGHPMpLDZgVP0NTR/0xcCW8QDklxurO38dGHmZf19WmvfTledcdydqqrv5Y+tFVUv3bNo0pT0k/b9X5lTkXyOWyynnyn1VT1cVjXWOR7Y0/btJK9Zbkejv8fHexYduMimh6cL1c91ySS3W+YQ+iqTfniBkPy9Mj/cn3TVWJdc8bO19qUkT1nq+qvgoek/pt+w1LDsVqMKqA9KV0F03HMW2PSoJD/uae/7myzkXumq08/1myTvX2Cbv878AP6P073otawA4OgZs6/S5lOrqm/2haWY5jq5Efx55v8tkuQFSwnLztVa+1ySJ/Qs2pSFj6tp/VVP26+TPGwZ/545Pd39armVhTf6fabvXnGRJNdeQZ8AAMA2SmAWAAAA2K5V1Y2SXKNn0dNaa6cup6/RdMWPm8nA/uD8dAHL5ZoUshivQjdrkyrNnbnK+91WbEpy5axsutulmDeddrqKikt193RTMs+15ND2duhRSfYaazsxyfOn6Ww0BfYLehY9uqrG/y5b3Szzw1PnpKtqOs0Y/i/JcT2LxqdJ3t78JNNVuOsLMh3WWvvWNINorX0kyWd7Fv35hE32yfzjI0n2n2L3b0hXgfDQdFWI75Xk5VP0s1w37Gn7/nI7qaod01898ZUreCngCz1tC54rrbUfpwtcjhuvWDrRKKB2n55FEyuTJnlET9uxSV661P1u1Vo7Ismnl7vdjDy6p+2kJE+eprPW2mfTHdPjrlVVN5+wzZb0V9x+UFX1nW+T9AVs3zwp5FhV+6d/RoNnrmDK+hcmOW2s7ZLpqpYu17TXyXU1qr7cd358N8nzpumztfbmdFVmx925qmb2TF5VV0ty/Z5F/9haO3E5fbXWfprpft4Ne58ZPdOc3rPooGn7BAAAtl07rvcAAAAAANZZ39Shv0zynmk6a619tqq+ldlVLPpca+2EKbb7Zrqw7Q5j7Zdc+ZAWtMeE9qmnOx1NDX/8tNuPuUJr7bgZ9TWNi6SrYvegqvpoksev0njenW6647l//1tX1X6jCsSLGa+++N3W2jGzGtwA3bun7QOjSm3Teme6StE7z2nbI8mNk3xqfOXW2jtHlYGvOOfrgtbaKSsYwy8zv0rmrivobwiOXm6osqqulKSv8uubVziWN2d+xb3bTlj3pPTfE/4h/QGxiVprL07y4uVss1JVtXO6Y3rcsgOzrbXzqmqfdBV/554vU933R/qquS7lXDksya3H2u5RVbstMfh4j8x/UeXX6Q/opaouluT2PYte3Vrrq666FK/I/J9hVY2ChtfrWXREa2089Lkc/5GuIuh46O+emV+5favDkzw7Fy5Qc0C6c3PRir1Vden0n7cLhZ7vmmS88uvZWUFItbV2alV9KPPv/7dLF15cjmVfJzeIP0ly8Z72Q5dbtXfMv2d+1f7N6V5O+s8V9DvX3Xvazknylin7e326l352WsY2G/0+8/3Mv25cbcb7AAAAtgEqzAIAAADbu76Qx/uWOnXpBNN+ON3nf6bZaDT+vtDIpOqUs/K7Ce1909tu7+6U5CtVdatZdzyaBnm86t8O6Sp0Laiq9k5yh7Fm1WUnGE1X3Vf5clK4aklGVWa/3LNo4pTErbWzWmvfbq29v7X20tbaSoM4v+1p296LMPRVEl3MzXraLkjy3yscy7zgdJIDqurA8cZRBfRjetZ/eFW9vW+bDeaA9H+ecdw0nbXWLmit/aS19unW2utaa3/fWusLvS7VtOfKuzP/Xr1HuoDmUvRVJn3TAs8wt8j8MFvSBfSn9b5MMX37CvVWfM3yg50X0lr7fvrPy4mB4FE1zr5zcamVgh+c+cf211tr31hgm75ryldaa2cvcZ+T9P0cE+85C5jmOrkR9P2sW5K8aSWdttY+mf4Xv2YZNO8b+1GttZOm6ay19uskn1jmNhv9PvPjnrYD13oQAADA+tve/+MmAAAAsB0bBd36qv71hdSW40sr3H6uH6xg29Mzf0ro8Ypkszapoua2HJj9SWvtwKWsWFU7pasIune6KWgryU3TVf66bM8m+yb5QFXdqrX21dkM9/fenq4y3Fz3S/KaRba7Vy5c1TTZhgKzrbXlTIM9CzdM/3m1kvDdVt9JV/Furr6p5GdmNI341ZPcOf2V17b3Igxfn2KbvnDfSSusWJgkP0r3ksL48XeT9AdJj0j/lN33S3LvqvrvJB9JNw32MRusQuSBE9p/tZaDGFdV+yW5TfrDkYueK621s6vqbUkeM7bo4CxSgbiq9s/8lxuSrmrtJDfqafv5aLryqYwq9n49k0Osq6HvOD4zybdm0PcXMj+Qeq2q2rm1ds6EbQ7L/PvtfarqSa21SS8SbdUXel7ob5j0/65ndc8Zd/mquuQyZzuY5jq5EfQdV23a0OmY/0lX1Xqx/U2rr6+V/h2+nO5ZYDk28n2m735x4BruHwAA2CAEZgEAAIDt2ZXTX2ltoapeS/G1FW4/10oCEH0fQq922O2UCe2XXuX9bgijqn7npgvu/CzJV5O8paqemuSRSf41yZ5jm+2R5O1Vde0ZVIeb6z1JXpkLT6d7q6q6+Khy2CQPGPv+q6Oqe/S7xoT2D1bVauzvkrPqqKoukW4q+qsluWaSP0oXAB4P2s+11oHkjWahc2eSvmNkv6paraDQpGPkNUmemC7IP26HdBUKb5HkhUl+XVWfTHJkkiNbaz9fjYEuw6UmtK9JYLaqdk8XrLpKunPlWklukO45YqUOy/zA7O2q6hKjauGTPDDzP+P5Smvt2wts0/e3X+kzT9I996xlYPaKPW1fb62dP4O++16a2jFd2PGHE7Z5b7rnn73mtO2V5K7pqgj3qqrrJLn2WPPvssBMBaOXva7Us+g+q3hNuVSS5QRmp7lObgR9x9VXZtT3l5Pcd6ztMlW10wpntUhV7ZHuJa1xKw2QHzPFNhv5PtP3b6rt4t8nAADAhQnMAgAAANuzAya0n7iSTltrp1TVublwUHFaswxQroUfTWi/6pqOYoMZVZJ8dVV9KclnMj80e6Ukf5Xk+TPc50lV9YlcuDrYDukqyL66b5tRgHJ8iuBtprrsOtl3jfe3z3I3qKprJLlxupDfVZJcIV34b1uu/LxeTp5imw1xjLTWfldV90pyVJKLL9LHxdOF5x+QJFX1zSQfSvK+JF9ah+qze/S0nddaO2uWOxlVjP2TdNXnr5YuQHdg+sNoM9Fa+2JVHZsLB6t3TBeI/Y8FNn1IT9vrF9ld33PPip55RtY6INkXcvvFjPqeFAy9ZCYEZltrv62qtyZ5/Niig7NAYDb9f8P3L1LRdNn3gBlY7j6nuU5uBGt9XG1OcokkKw2Kjleu3WqlLxQct9wNNvh95rSeNs9BAACwHRKYBQAAALZn46HFrU6dQd+nJtlvBv2sdMrutTapst3Vp+2wtfazLLOqZVU9J8mzp93nammtfb2qHp/kTT2LH19V/zyDadrnenvmT6d7v0wIzKarfja36vKWUR9MtlA11tWwpPDlKPj3pCQPzmyqYNI5ZYptNswx0lo7tqpulOSN6YKhS/VHo6+/TfLTqnpTkkNH1+e1sFtP22JT3S9JVW1Od118RJLbpL/y/Gp7fboK5HMdkgmB2aq6ZpLrjjX/NgtUJh3pe+6ZxTPPKTPoYzn6AtSz+DmSyT/LYsG6wzI/MHuXqtqnLwBbVTukC0X39bOQtb6eJMsP/Z+yGoNYAxvxuFqKvnEnyekr7Heq7TfwfabvnnGRGfUNAABsQ1Z7Cj4AAACAjWznCe0XzKDvc2bQx7boe+n/QPqgquoLPG13WmtvTnJsz6JLJzloxrv7r8w/Fm9ZVZOqfj1g7Puj1zAQt63aa433t2C4o6o2VdVfpauE+OxMF5b9vySvTPLdKbYdtCmnfN9r1uNYxILHSGvtx621myW5R5KPZ/n3vAOS/F2SVlXPHk0Rv9r67h8rvs9W1UFJvpqukvbts/yw7O+SfDiLB1UX88bMf0Hm+lV1tQnrH9LT9l+ttVMW2U/fc8+2+MzT93OcOaO+J1X2X/DFndbaV5J8a6x553Rh7D63T1e1dq6fpzsnF7LXIstXw7JChVNeJzeCDXdcLdGuy9znUvVVZF2SDXqf+W1P2+aqmvT7AwAABkqFWQAAAGB7Nqlq1EWz8ulkL7rC7bdJrbXzquqoJHccW7RLklsk+eiaD2pjeleSZ/W03zjJ52e1k9baqVX10SR3m9O8Q5J7J3nV3HWr6rKZXwnsrbMay4BNqnL5sCRrGhqqqp3TVap80DI2OyvJd5J8PclXkny2tfadUX9HpZuWnpX5XZKdxto+n+TQVdpf77Tx41pr70vyvqraP8k9k9w1ya3TH07ts1uS56R7IeIerbVzpxjrUvUFMsd/p8tSVfdJF1RdaljqgnRTlH8jXcj2C0k+31r7bVU9LMs77y6ktfbLqvpwLnytTpKDkzxzbNyb01WOHvf6Jeyq77lnFs8ra/3Ms5pTq0/6WZYSnHx9kpeMtR2csfvtyEN62t6whLDppHvO65Ictci205rZc8kGd1rmB5I3wnG1mL4gaLLysa+46NIGu89MelFye33JEQAAtlsCswAAAMD2bN4UuSMrCn6Mwiyz+oB9W/TBzA/MJslDIzC71TcmtF9mFfb19swPYd0v8wM898uFK52dly7Yy8ImXUc+0DcN9yr71ywc2js2yf+kO/6+m6Ql+WlrbcuE9YcyQ9laVEBdyEmZP2X2ia21N63HYMa11n6Z7nrwqlEVv5smuU26UNNBWTyYepckr0ny8FUc5hk9bVNPpV1VN0z3QsCkz0hOTPK5JF9Ld658J8n3W2uTgmmzOFcOy/xr9YOq6h/GztFbJ7ns2HrHJ/nEEvbRd03aFgOzfT/HxWbU96R+zlrCtm9K8i+58Dlz06q6Qmvtx1sbqmrPdJU3xy0l9DzpvnLsRrmmbMNOyvzA7EY4rhZz+oT2lZ6Xs/rZN8p9pi+ke1ZrbRZVtgEAgG2IwCwAAACwPTthQvuVMn9a3eW4QmYzxeq26r+SvDTzp7a+d1VdqrX2f2s+oo2nL/yVrE7o6P3ppuWdGy67RVVdorX2qzltDxzb7hOttd+swniGZlJ4aZ8Fls1cVd02yZN7Fm1JcniS580NbC3R1IHEVTB+PVmOpVayWy0npZteeq591mMgi2mt/S7Jp0dfqao9ktwqXVXAu2f+FPJbPayqXtZa+9oqDa3vmrlDVe0yGvOSjcJab07/5yPfSvL0JEcuM0Q1i3PlQ0l+mWT/OW1XSBcs++85bYf0bHvEEsfb99xzpSWPcLIrzqCP5eibBeDSM+r7chPaf7HYhq21X1fVB5Lca07zpnQvMrxgTtu9M/+Y+Vxr7ftLGN9C9xxWZq2Pq99O2OdyHT+h/VIr7HfvFW7fax3vM33PApOexwEAgAEbSoUAAAAAgGn8KP3Bg+ussN/rrXD7bVpr7WdJ3tuzaKckz1/j4WxU+09on0WlsQtprZ2R5MNjzTukmxo3SVJVV0xyg7F13jrrsQzUpKDKtdZ0FMmz0h/Uf3xr7RFThGWTycfpauureDtpKuWlmBS+WSt9x8g1qmrDv1jRWjujtfbB1trj0lU1vXeSSS89HLyKQzl1Qvs0FRAflOQqPe1HJblha+2jU1QcXPG50lo7L12F0nH33vp/quoic7+f4/Al7ubLPW3XqKqVnF/J2j/3/Kin7bpVtZJg/Vbj98IkObO1Nuklq3F9VWLH/2Z9oeelVJdNktNGX+PW+p4zRH3H1Q1n1HffcfWjBSq8L1lr7cwkP+9ZtNLz8tor3H5J1vA+03e/6DuXAACAgROYBQAAALZbow+pP9+z6OYr7PpmK9x+CP59QvvDq2qlv98h6AtrJZPDlyv19p62+8z5//cfW/bbdJWCWdx/T2j/k5V2XFU3Hn1daqFwZVXtn/7r1sdaa6+ect97pL+y3lr8N+Xzetp2X0F/s6iguRKf62nbO8k1VtJpVe1UVXetqmuPpnhfzrb7VNWyfqettfNba+9JVwnw7J5VZhUs6/OTCe2XnaKv+/S0nZvkIcutVjtH9TZWLfd8Oayn7Z5z/v/dkuwxtvyzrbUfLrH/vuvVTklutMTt56mq/TLh519FfT/H7kn+aAZ936Sn7TvL2P4jmR/2u25VHZgkVXWpdOfQXGcmecdSOh89u/b9/Dddxhh7VdVlqup2VXWVUSXm7U3f7/WqVbXvSjodXQcO6ll07Er6HfOlnrapz+uRFQVuN+B9pq/K73Er6A8AANhGCcwCAAAA27uP97TdqqomTcm7oFGVtgevbEjbvtba59NNLz1uU5J3VNX49OTbmwdOaD9mlfb3oXSBnLluVVVbp3C+3/j6rTVVt5agtfbLJN/rWXSflVQ7rKrd0lUG/kK6qcDPrqrvVdWLe1a/bvqry75t2v0nuWW6SsTjZlHBcTHjx2qS7DdNR1V1pSQrCjvNwNET2seD6st1nyQfSPLNJKdV1UlV9bWquu34ilX15Kp6bVUdXVW/SnLitPtvrX0v/df3S0zT3xJNmqp+mnvJ9XvaPt9am+qFhdF5fosJi5d1vrTWjk3yxbHmA6vquqP//6CezfpCtpMck+TXPe0PXUYffduuxXVhrs+lvxL1Sn6OVNU1018J9NNL7aO1dn6SN/Qs2hp8fkDmfzb3zlE1+KXqu6ZcvKpus4w++jwr3XPx99Ldc35RVZ+vqouvsN9tRd/vdVP6KwIvx53TX4V6ycfVEvT9e+aGVTXpBa0Fja5rd1vmNhv9PtN3v+h7fgMAAAZOYBYAAADY3r0x8ysYbU7yl1P294isfzhrpZY7FfUkT05/dahLJvnY1mprs1ZVB6X7O2xIVXX/JFfuWXRmJgfrVqS1dla6YN1cOya5e1VdOcl1xpa9dTXGMWAf6Wk7MBeuDLlcj05XhXSrXdJVJv5Fz7qTrjmTpjReiidNaN9xCduu9Bpyck9b3zmzFH3VRNfaV5L8pqf9cVV1kWk6HFUc/uux5r3TVdj8Vs8m90vyqHQV0LeG324/zb5HftXTds4K+lvQKEzYdzxPE5jtO19Wcq7cJ919rc9Szpdxr+9pu9eoivCdxtrPSPKupXY8CnO+rmfR/atq2dV6R8H+Jy53u5VqrZ2Y5Ms9ix5SVX3Tri/Vkye0f2yZ/fT+DUf/e98lrr+Qj05on/bZdWvl27nT3W9Kcqkke7bW+kLWQ/SN9F8LHl9V05zLW006ro5cQZ/j3pv+6uyPn7K/u6a/yvxCNvp9pu9lyEkvYwAAAAMmMAsAAABs11prJ6c/HPiUqlrWlOpVdYUk/zqTga2vvg/cl/3fkVprP07yjAmLr5bki1V1u+X2O0lV7VVV/5bk8+n/UHzdjapdvmLC4resYDrwpXh7T9s9Mz9QeHr6q3ox2cuSnN/T/tJppnKuqkukq/Q37rz0/x1Pn9DVFZe779H+H5j5wbytljJNd981ZDnT0x/X07ZvVS1reumqumiSpy5nm9XQWjs3/ef9xZP8+5TdPjL902V/vLXWFzLqq2R4r6q6zJT7/+Oeth9O2ddSfaen7ZpT9NN3vkx7rlw8yb8tsMo009q/NfNfNrlnkrsk2Xms/R2ttb6KzAt5VeaH2vdI8tpREHs5XpTkCsvcZlb6zp29k7x8ms6q6pbpXlQY9/0kn1pOX621lu5ZZK6bVtUfJ7nxWPsPs8yXZVprX0/ymZ5Ffzp6KWca/5pkt572N03Z3zZnFCj/z55FV03y7Gn6rKpDktyxZ9EnWms/mKbPPq21E5K8r2fRk6uq714x0SgIP82/ZzbsfWYUpO8bR999BQAAGDiBWQAAAIDk+UnOGmvbnOR9VXXDpXQwCkJ+Ml3oZFvXVxV2z2k6aq39Z5LXTFh8iSQfr6q3VdVVp+k/Sapq/6p6bpIfJ/mrLDw19Kyq5y5bVd01yf+kv7rheUlevMpD+EiS08babp/kwWNt722t/XaVxzIorbUfJnlPz6LLJHlnVS35ulBVu6QLxe7Ts/iw1trPetp/OqG7Ry0jpLp1/3dMfwXKrZZSEbXvGpIs/TryuQntz13i9hlVAzwikyt/rrWXZ/59JkkeW1XLqtBZVddJ8h8TFj9vQvtbMn8K+53ThSQXumb27f+WSW7as6hvSvBZ+u+etmUFwUb6zpcbVtV1l9NJVe2X5INJFqrMuuwKwq2105K8e6z5munub+OWW5k0rbWfJHl1z6I7JXndUo+H0X13zavLzvHO9IfrD66q8erLC6qqqyV5c/o/M3tRa2383FmKw8a+35zktekqt851+JT9v2hC+2ur6ibL6aiqHp35zwJJVxn7lcsd2DbuVekP1f/tcsPIoxfvJgW4/2W5A1uC52f+c+6OSd5WVUuqFjt6Zjg0XUX75drI95nrZf65d36SL0zZHwAAsA0TmAUAAAC2e6NKqH3VHPdNclRV/e0oxDZPVW0eBQ2+mPWrsjZrJ/W0XX0F/T0xyX8tsPz+Sb5bVR+tqoeOKvYtqKr2rar7VdV70oWfnpVkrwU2OT3JIa21ScHCmamqTVW16yjIe9Oq+vuq+lqSDyTZb8JmLx9VpFs1o+q149XHdk1yrbG2vorLLO5vk5za037rJEePAlkLqqr9k3w4ya16Fp+e5B8nbPqNJCf0tF8vyX8sJTQ7Omb/MV0AcKGQ31JCr33XkCS5xhK23RpAPrZn0R2r6kWLVcEcBYM+lOQeS9nfWmit/Sb995kkeXlVvXhU1W9BVXX7dNUu+9Z9b2ttvKrl1v1/N11oftydk7xlVI13UVV1UJJ3ZX7w6PR0gcPVdFRP27WmmCr9Yz1tm5K8famVEEfB8i8nOWiRVad62ST9QdjxF3i+31qbFC5fzDOS/Lyn/eFJPjsKZfeqqitW1fsz+XheE6NqoH+e+QG9JPnXqnppVe2+WD9Vdad0FV77/vb/nYVfIFjIO5KMV/8d/xtekC7Yv2yttQ+ne64Yt2eST46epxa7Vm4ehYv7AtRJ8rzW2qQK5oPUWjslyd/3LNoh3bXyGVW102L9jCrLfjxJ37X1ra21T6xooD1aa8dkflA76cKvRy90Xie/r8r+liQPmXL/G/k+0/dCxNdHLygAAADbmeX+hyQAAACAZauqaSpnLea6ow+GZ+WlSe6e5OZj7bsl+ackT6+qDyT5drqKW3slqSR3y8apYDgrfRUsq6qen+T5W6uPVtXOrbVzFuustXZeVd07XQXVp05YbVO6KWvvmGRLVf0w3e/6F/lDCHGPJAek+71X5n+QPsnRSR6xgqlvL79Kx/BW30jyd6vY/1xvT3LIAst/k2TmIY7tQWvth1X18PRXmr1Okm9V1eFJ3pbk6K3nzijQdI10wfEnpZtSvM+jWmt9Abe01rZU1WHpP46elOT6VfVvST7ZWvt9qLeqLjIa258leWS6qs+LmRT6nuv/0lVuG68o97Kquk9r7bjR/jcl2TwKvo17xehr3NOS3KSq/jnJp1prZ4/62jFdIOZB6aZVnxuU+0GSKy9h3KuqtfbiqrpZ+oO8f5nk/lX1H0ne11r73tYFoyDtzZM8Jsm9JnR/fPqnk5/raUlum2T8BZD7Jbl5VR2aLlT/v3P/JqMKyTdIcnCSh6W/ivdzRkGz1fSFJOekq1i41a5J/ijJ15bRz+Hpft/jn49cJck3q+pFSd41Cm4n+f2xesUkt0vy0CRLreC5X5LvLbrWfJ9OVzV9oRdxll1ddqvW2mlV9ah0wfLxQP1Nk3y9qr6S5Mh0zwTnJ7l0ujD/zXu26bOa980kSWvt/aO/19/0LP7zJPcZnVPvnfsMMDqm75jkEUnuMqH7E5I8aMrqr2mtnV5V70p3vEzyidba8dP0P/LQdMf+gWPtF0l3nD9xdF5/qLX2q60LRy8m3THd7+gGE/r+cJKXrWBs26zW2stG1WHHK8puTvLCJA+tqpenu1b//pm5qvZJ8qdJHpvkTyZ0/50kT5j9qH/vL9KdozXWfsUkX66q16a7dnxl67E9elHgPunOo0vN2WZLlv6svdVGvc/0zRhy1JR9AQAA2ziBWQAAAIB0lcqq6p5JPpNu6uNxe2XhoOFWp6ebvnY8vDE+RepG9o0J7X+f5K+q6pfpwhh7VNVFJ4TdLqS1dkGSvxhVWv33dNV7J9mULty20oDbb9JV0Tts2sDLGjg2yV22hv7WwJFJTs7kUOY7W2vnrdFYBqe19t6q+ockz+tZvGOSR42+tlTVb5L8Nt25sFhl0ee11t6xyDovTFcVrm96+Jukm+J9637PHO1zv0wOvp2WLrD6t2Ptl6+qXbcG5/u01s6pqu9m/rX0+kl+VFUnpAvgXSJdRcu39HTzmnTVqfuq0t4sXSXc80bXowtGffVVAv91utDTVyeNd409LMkn0/0uxl0m3TTrL6qq3yX5Vbpw6H7pDw9tdVKSe7XWTlxox621Y6vqaUn+s2fxpdJVMP7HJOdU1Unpjs89k+yThUNTH0h3XV9VrbWzq+pzSW4ztuh2WUZgtrX2nVHY7ak9i/dJ8s9J/rmqTk93vdxh1L5Q5eXXjsYxHnC9WpLeqr+LjHHLKGD/3AmrnJ/kDcvtd2wfH62qJ6U7z/v+vjfI5DDlXG9NF1oeP9/X6rnn79P9nu/es6zvnNojk++BW/0q3b15pVXpD8vCgdmpQ89J0lo7uaruke6a0vdcdcPRV6rqjHTXit0nrDvXMemq8m/UZ6e18Kgkl09y455lV0vy8nTVwc9Od5/ZK/3VZOf6Qbrj6pTZDfPCWmtnVNVd0/175tJji3dM8vjR1zlV9askF8vkStiTwugL7X/D3WdGLzzctmfRkdP0BwAAbPuW8hYwAAAAwHZhFDa6efqna16Kn6WrvtYXzDp3yj7Xw1czOXy0a7oAwSXSBe7GK1gtqLX2xtE2/y9d4Gc1nJrk2Umu2Fp73QYNfJyXrnLbTVprv1irnbbWzk3y3gVWeetajWWoWmvPT1dd7ncLrLYpycWTXC4Lh2XPT/IXrbVFpz5vrZ2RrlLiKUvY74HpzuFJ/334S+lCVi/M/CqRO2d+Je4+r1lgDJdKF+zdOV2F23lGwe37JPnlAvvYMV0g7nLpD8v+Il248kdLGO+aGFX4vU26CnsL2SXdz7V/Fg7LHp/kFq21ryxx/y/L4gGondNVTj8wXbBuoRDTu5Pcdw2vs33XqDtM0c/Tk3x0kXX2TFfV/DKZHJY9NcnDWmuPSf99sy+ktVRHZHLo9MhJFaeXo7V2aJIHpnvZZxovT/cyUd8xsibPPaOXdu6Txauhbj2nFgvLfi3JjVprX5/B8I5O8sMJy07OwvfjJWmtfSPdSwTfWWTVrVX6FwvLfjrJLVtrJ610bNuy0T31dkneuciqF0n3e10sLPvxJDfeWmF9NY2qKd8qyXcXWG3ndPfhSWHZf0/3rD7N/jfafeY66Z595vplkk9N2R8AALCNE5gFAAAAmKO1dnKSO6ebpneplcXOT/eh8h+31r6W/ll9zprNCFff6APphyZZsFrhyB9P0f+JrbVHJ7lSug/kT1tuHxN8K900t5dtrf1ja23aANBqOjFdkPA6rbWntNZm9bMvx9sntB+f5HNrOZChaq29Jl0F0ZWEMb6eLlzz0mXs91vpqslOG/T6TZKnJLlpa+17o3PomJ717ruEvl6e5CNLWG/iNaS19p10wd2PL6GfcR9Ocv3W2ren2HZVtdZOa63dI1113ROm7OaCJIcmuXZr7X+Xuf8XJblTumqH0/plkke31u7TWlsoHD5r70pyzljbzapqoeqv84xeHrhHuiqI04SwtiR5U5JrtdaOGLV9tme9O1fV7lP0n9baTzL5GrKiyqRj+3l7ugqxb83SX2T5QZK7tdaePAqsrutzT2vtvNbaU9K9NPD9Kbs5MclfJzloVqHG0fPUpL/VW2d17rTWvpuuGvAL01XsnMbp6a7/t1unZ5MNp7V2ZmvtfulC4f83ZTfHJ3l4a+0Oi1UBn6XW2vfTHROHZnkvqJ2b5Bmttb9c4f430n3m9j1tb1/KDBkAAMAwCcwCAAAAjGmtbWmtvT7JFdOFL16e5Ivpplw9d/T1y3RVuJ6ZrpLpo+dU4+oLx5y66gOfoVHI7HpZvNrusgOzc/bxk9EH8vsnuWuSV6erxLbU8NIZ6cJEf5/k6q21P2qtHTqqCraezks35f3P0lXrfVeS56SrgnjJ1tpjlxtwm7FPpQtGjnvHBq3Gu01qrf1va+226QKsr0t3/VjMGUnek+66c4OlVg0d2+93k9woXej9i1n8fDo9XaXNxyQ5sLX2srEQSd+07w+pqksuMo4LkvxZuinlFzonF7yGtNaOb63dIV3g5f2ZH5ac69x0P8udWmt/2lqbNoy6Jlprhye5QroXND6RpVXk/EmSf0lSrbUnjCrWTrPvjyW5VpIHJPlQkrOXsNm5SY5K8ugkV2itTVV9cCVGU5l/eKx5l3TnzHL7+l1r7c+THJTkjemu2wu5IF2A/EXpfv+HtNZ+Nmf5WzP/b7h3ut/XtA7raTsp3bkwM62141prD0pX7fEvR/3/IN25e0G6382xSQ5PcvckV2utfWBOFxviuae19pEkV08X6v9QFq70nXQ/29FJnpjkgNbai1chRDepUvDMQs9J0lo7q7X2d+lmAfj79L/sMO6CdC9YPDXJ5UfX/0lVjbdbrbU3pbtWPzLds/95i2xyTrrn54ckudLoWr/mRoHfJyS5drqXtRY6J3+X5G3pXuj6lxntf6PcZ+7d0/bmGfQLAABsozZt2eK/gQMAAADMUlX9RZKXjDV/ubV20HqMZ6Wq6jLpQn8HpAvFnJ1uKuEfJvl2a60vfLmS/e2W5JrpKtBeLN00t7uM9nt6umpdP0zyQ9Wh2JZU1VXSHduXyx+mQT41XQDuf5P876yP6araK10g8IAke6WbBvmkdKHp45J8fbXPo1H1z5skqST7pAtpnZYuVN5GId+l9rVbup+n8oepxU9J8r0kX9qWKyNW1a7ppo6+croXCS6SLsR0SrpKtF9prU1b5XCxfe+Q5GpJrpHubzT3uvvrdOHJY1pr614tvaruki58Nde7W2v3WWG/O6artLr1d7BnusDoifnD7/+UlexjqKrq1Myfkv5PW2vj4eY1VVW7pDunrpI/nFO/TXdOtSTfnDZ0vi2oqn3S/fxXSHe93CXdMX1KuvD9lzdoNf4NbVQ1+rrprtUXT7JrulD5SUm+k+RbG+FaOW50jfvjdNe5/dI9D5yS7lz4QmttsZcGVrr/Nb/PVNWVMr/K7Tdba1O/8AcAAGz7BGYBAAAAZqyq/iXJ08ea39Na66twBAAsUVVtShdKqznNv02y/7YcmN5WjQLxfQG3P2qtfWutxwOwVVU9M8nzxpofvl5VfwEAgI1hx/UeAAAAAMB6qqqrppsu9PtJftBaW8p0oYu5Zk/b92bQLwBs11prW6rqpUkOndO8a5L7JnndugxqG1JV90lXYfp7rbVfzKDLvmeeCzK/qiPAWjt47Pv/S/KW9RgIAACwcWxe7wEAAAAArLMrJ3l3km8mObOqflpVh03bWVVdNMltexZ9bdo+AYALOSJd6HOuJ67HQLZBT0/y6SQ/r6ozqurrVfUnK+jvXj1t353RC0gAU6mq2+XClciT5OWttXPWYzwAAMDGocIssNZ2SXLDdG/ynr/OYwEAAMid73znTR/5yEe2frspyeWSHPzBD37wpXe9612XPbXzJS95yaeecMIJu441X/CsZz3rh0kOXNFgAYC01nK7293uVccff/wz5zRf9zGPecy9X/Oa13x13Qa2Ddh9991PPfPMM3//bZLr7Lfffock+fly+3rf+9530c2bNz/hggsuuFD7Pvvs86V45gHW0e677/70Ode6bN68+aRXvOIV74trEwAAbC92SHKpJF9O8ru5CwRmgbV2wyRHr/cgAAAAtnrRi16Uo48+Omecccbc5p2OO+64byy3r2OOOSYnnXTSvPab3exmmx/84AcL8ADAjHzoQx/KHe5wh5xwwgm/b9ttt93etY5D2iY87nGPy4tf/OILtZ1//vmPPfnkkx+79957L7mfLVu25Mgjj8x4WDZJXvnKVz4sycNWNlKA6Rx//PE5++wLF7n+m7/5m31uc5vbfHudhgQAAKyfmyf53NwGgVlgrf1fkpx88pm54IIt6z0WgJnbd989cuKJZyy+IsA2yDWOIbvJTf4kH//4xy7Uduihh+ZiF9s3d7nLny2pj//+76Pz3Oc+M+ecc+FZPjdt2pRDDnmk82eDc40Dhmyo17iHPexR+ed/fv7vvz/yyCPzjW98J5e97OXWcVQb2w1ucNMkFw7MnnzyyXnsYx+XF7zgX7PPPvss2scZZ5yef/qn5+Wooz45b9mNbnSTHHDAVQZ5vLGxDfU6x/K94hWHXijMf4lL7J873vFujg+2aa5xwJC5xgGrYfPmTdl7792TUU5tLoFZYK2dnyQXXLBFYBYYLNc3YMhc4xiqgw9+eD71qU/k/PPP/33beeedl+c//zn54Affn7vd7V653vVukP322+9C2/3mN7/ON77x9Xz4wx/MF7/4+d6+7373e+ea17y282cb4G8EDNkQr3F3vvOf5T3veWe+972WJDn//PNz2GGvzTOf+dx1HtnGdbnLXT63ve3t88lPfvxC7d/4xjF5wAPumbvf/d659a1vmytd6SrZaaedfr/8nHPOyY9+9MN87nOfyX/917tzyiknz+t79913z1Of+rRBHmtsGxx7nHDCCfngB99/obYnPOEp2XHHnRwfbPMcw8CQucYBq+j88YZNW7a46ABr6sAkPz7xxDM89ACDdPGL75lf//r09R4GwKpwjWPoXve6V+f1r3/tguvsvvvu2X33PZIkZ5xxRs4668wF17/e9W6Ql7zk5dlxR+8sb3SuccCQDfkad+yx387jHveI31cT3GGHHfKmN70zl7vcAes8so3rxBN/k0c84sE58cQTJ66zww475KIXvVh23XXXnH322Tn99NMu9GLRuB133DEvfOGLc5Ob/MlqDBkWNeTrHEv3ohe9IO9//3t///1BB904L3nJy9dxRDAbrnHAkLnGAath8+ZN2XffPZLkCkmOu9Cy9RgQAAAAwEbzyEc+Nocc8vAF1znzzDPzq1/9Mr/61S8XDcve9rZ3yL/9238KywLAKrrGNa6Vu9/93r///vzzz8/rXvfqdRzRxrfvvvvl5S9/bS55yUtNXOf888/PySeflP/7v1/klFNOXjAsu+eeF81LXvJyYVlgXf385z/Lhz/8gd9/v/POu+Qv//Jv1nFEAADARiQwCwAAADDy2Mc+MS960Utz2ctebuo+9tvv4nn2s5+f5z73n7LzzjvPcHQAQJ/HPe6JufSlL/P77z/xiY/lW9/6xjqOaOO73OUOyGGHvSn3uMe9s3nzdB8Vbdq0Kbe97R3y5je/M9e73g1mPEKA5Xn5y1+a88477/ffP/rRj1/Rv+sAAIBh2rRliynRgTV1YJIfn3jiGbngAtcfYHhMGwIMmWsc25Nzzz03n/3sUfnYxz6cr3zlSznnnN8tuP7OO++S61znern97e+Y293ujtlpp53WaKTMimscMGTbwzXu2GO/nSc84VG/D0tVXT2vfe0RU4dBtyfHH//TfPjDH8gnP3lkfvGLny+6/mUuc9nc9KY3zz3vee8ccMCBqz9AWILt4TrHZF/+8hfzF3/xxN9/f9BBN8mLX/yf2bRp0zqOCmbHNQ4YMtc4YDVs3rwp++67R5JcIclxc5cJzAJr7cAIzAID5h91wJC5xrG9Ou+883LccT/Occf9KKeddlrOPPOMbNmyJRe5yG7Za6+9cvnLH5jLX/4K2WWXXdZ7qKyAaxwwZNvLNe4tb3lDXvnK//z9909/+t/nbne75zqOaNtzyimn5Hvf+25+9asTcuaZZ+bss8/Ozjvvkt133z2XutSlc+CBV8glLrH/eg8T5tlernPMd9555+XhD39QfvzjHyVJ9t57nxxxxFuzzz77rvPIYHZc44Ahc40DVoPALLCRHBiBWWDA/KMOGDLXOGDIXOOAIXONA4bOdQ4YMtc4YMhc44DVsFBg1lxEAAAAAAAAAAAAAAyawCwAAAAAAAAAAAAAgyYwCwAAAAAAAAAAAMCgCcwCAAAAAAAAAAAAMGgCswAAAAAAAAAAAAAMmsAsAAAAAAAAAAAAAIMmMAsAAAAAAAAAAADAoAnMAgAAAAAAAAAAADBoArMAAAAAAAAAAAAADJrALAAAAAAAAAAAAACDJjALAAAAAAAAAAAAwKAJzAIAAAAAAAAAAAAwaAKzAAAAAAAAAAAAAAyawCwAAAAAAAAAAAAAgyYwCwAAAAAAAAAAAMCgCcwCAAAAAAAAAAAAMGgCswAAAAAAAAAAAAAMmsAsAAAAAAAAAAAAAIMmMAsAAAAAAAAAAADAoAnMAgAAAAAAAAAAADBoArMAAAAAAAAAAAAADJrALAAAAAAAAAAAAACDJjALAAAAAAAAAAAAwKAJzAIAAAAAAAAAAAAwaAKzAAAAAAAAAAAAAAyawCwAAAAAAAAAAAAAgyYwCwAAAAAAAAAAAMCgCcwCAAAAAAAAAAAAMGgCswAAAAAAAAAAAAAM2qYtW7as9xiA7cuBSX683oMAAAAAAAAAAAC2b7/93bk5/bTfrvcwmKHNmzdl3333SJIrJDlu7rId12NAAE954X/lNyefud7DAAAAAAAAAAAAtlNvedGDc3oEZrcXm9d7AAAAAAAAAAAAAACwmgRmAQAAAAAAAAAAABg0gVkAAAAAAAAAAAAABk1gFgAAAAAAAAAAAIBBE5gFAAAAAAAAAAAAYNAEZgEAAAAAAAAAAAAYNIFZAAAAAAAAAAAAAAZNYBYAAAAAAAAAAACAQROYBQAAAAAAAAAAAGDQBGYBAAAAAAAAAAAAGDSBWQAAAAAAAAAAAAAGTWAWAAAAAAAAAAAAgEETmAUAAAAAAAAAAABg0ARmAQAAAAAAAAAAABg0gVkAAAAAAAAAAAAABk1gFgAAAAAAAAAAAIBBE5gFAAAAAAAAAAAAYNAEZgEAAAAAAAAAAAAYNIFZAAAAAAAAAAAAAAZNYBYAAAAAAAAAAACAQROYBQAAAAAAAAAAAGDQBGYBAAAAAAAAAAAAGDSBWQAAAAAAAAAAAAAGTWAWAAAAAAAAAAAAgEETmAUAAAAAAAAAAABg0ARmAQAAAAAAAAAAABg0gVkAAAAAAAAAAAAABk1gFgAAAAAAAAAAAIBBE5gFAAAAAAAAAAAAYNAEZgEAAAAAAAAAAAAYNIFZAAAAAAAAAAAAAAZNYBYAAAAAAAAAAACAQROYBQAAAAAAAAAAAGDQBGYBAAAAAAAAAAAAGDSBWQAAAAAAAAAAAAAGTWAWAAAAAAAAAAAAgEETmAUAAAAAAAAAAABg0ARmAQAAAAAAAAAAABg0gVkAAAAAAAAAAAAABk1gFgAAAAAAAAAAAIBBE5gFAAAAAAAAAAAAYNAEZgEAAAAAAAAAAAAYNIFZAAAAAAAAAAAAAAZNYBYAAAAAAAAAAACAQROYBQAAAAAAAAAAAGDQBGYBAAAAAAAAAAAAGDSBWQAAAAAAAAAAAAAGTWAWAAAAAAAAAAAAgEETmAUAAAAAAAAAAABg0ARmAQAAAAAAAAAAABg0gVkAAAAAAAAAAAAABk1gFgAAAAAAAAAAAIBBE5gFAAAAAAAAAAAAYNAEZgEAAAAAAAAAAAAYNIFZAAAAAAAAAAAAAAZNYBYAAAAAAAAAAACAQROYBQAAAAAAAAAAAGDQBGYBAAAAAAAAAAAAGDSBWQAAAAAAAAAAAAAGTWAWAAAAAAAAAAAAgEETmAUAAAAAAAAAAABg0ARmAQAAAAAAAAAAABg0gVkAAAAAAAAAAAAABk1gFgAAAAAAAAAAAIBBE5gFAAAAAAAAAAAAYNAEZgEAAAAAAAAAAAAYNIFZAAAAAAAAAAAAAAZNYBYAAAAAAAAAAACAQROYBQAAAAAAAAAAAGDQBGYBAAAAAAAAAAAAGDSBWQAAAAAAAAAAAAAGTWAWAAAAAAAAAAAAgEETmAUAAAAAAAAAAABg0ARmAQAAAAAAAAAAABg0gVkAAAAAAAAAAAAABk1gFgAAAAAAAAAAAIBBE5gFAAAAAAAAAAAAYNAEZgEAAAAAAAAAAAAYNIFZAAAAAAAAAAAAAAZNYBYAAAAAAAAAAACAQROYBQAAAAAAAAAAAGDQBGYBAAAAAAAAAAAAGDSBWQAAAAAAAAAAAAAGTWAWAAAAAAAAAAAAgEETmAUAAAAAAAAAAABg0ARmAQAAAAAAAAAAABg0gVkAAAAAAAAAAAAABk1gFgAAAAAAAAAAAIBBE5gFAAAAAAAAAAAAYNAEZgEAAAAAAAAAAAAYNIFZAAAAAAAAAAAAAAZNYBYAAAAAAAAAAACAQROYBQAAAAAAAAAAAGDQBGYBAAAAAAAAAAAAGDSBWQAAAAAAAAAAAAAGTWAWAAAAAAAAAAAAgEETmAUAAAAAAAAAAABg0ARmAQAAAAAAAAAAABg0gVkAAAAAAAAAAAAABk1gFgAAAAAAAAAAAIBBE5gFAAAAAAAAAAAAYNAEZgEAAAAAAAAAAAAYNIFZAAAAAAAAAAAAAAZNYBYAAAAAAAAAAACAQROYBQAAAAAAAAAAAGDQBGYBAAAAAAAAAAAAGDSBWQAAAAAAAAAAAAAGTWAWAAAAAAAAAAAAgEETmAUAAAAAAAAAAABg0ARmAQAAAAAAAAAAABg0gVkAAAAAAAAAAAAABk1gFgAAAAAAAAAAAIBBE5gFAAAAAAAAAAAAYNAEZgEAAAAAAAAAAAAYNIFZAAAAAAAAAAAAAAZNYBYAAAAAAAAAAACAQROYBQAAAAAAAAAAAGDQBGYBAAAAAAAAAAAAGDSBWQAAAAAAAAAAAAAGTWAWAAAAAAAAAAAAgEETmAUAAAAAAAAAAABg0ARmAQAAAAAAAAAAABg0gVkAAAAAAAAAAAAABk1gFgAAAAAAAAAAAIBBE5gFAAAAAAAAAAAAYNAEZgEAAAAAAAAAAAAYNIFZAAAAAAAAAAAAAAZNYBYAAAAAAAAAAACAQROYBQAAAAAAAAAAAGDQBGYBAAAAAAAAAAAAGLQd13sAMFRVtTnJ1ZJcJcllk+yRZKckZyU5NclxSVpr7WfrNUYAAAAAAAAAAADYHgjM0quqDk/y0DXc5Tdaa9dZaSdV9Zgkr+5Z9LrW2qNW2v8S9r9rknsluV+SWyW52BK2+WmSI5O8JclRrbUtU+77OUmePc22Uzq1tbbXGu4PAAAAAAAAAAAAprJ5vQcAM/bICe0PrKq9VmunVbVDVf1luqqxb05y9ywhLDtyQJJHJflUkmOq6m6rMkgAAAAAAAAAAADYTgnMMhhVda0kB01YvFuSh63Sfi+b5H+SvDjJ/ivs7o+SvK+q3lxVu694cAAAAAAAAAAAAIDALIPyqEWWP27WO6yqA5J8IckNZtz1g5IcWVUXnXG/AAAAAAAAAAAAsN3Zcb0HALNQVTsnOXis+YtJbnTh1eq2rbVPzmifOyV5b5LLLrDa/yX5dpIfJTkjydlJdk+yV5KrJ7n26Ps+N03yhqq6Z2ttyyzGDAAAAAAAAAAAANsjgVmW44jW2sPWexAT3CPJvmNtT0/ymbG2JySZSWA2yROTXK+n/YIkr0vyqtba1xbqoKp2SfJn6cZ6w55V7p7kYUleP+UYP9Nau9WU2wIAAAAAAAAAAMAgbF7vAcCMPHLs+5+31j6brsrsXHerqsusdGdVtUOSp/UsOiXJLVtrj1ksLJskrbXftdbeleTGSf5hwmrPH1WzBQAAAAAAAAAAAKYgMMs2r6oun+R2Y81Hj/73HWPtOyZ59Ax2e8sklx5r25Lk3q21zy23s9baBa215yd5Qc/iSye56/KHCAAAAAAAAAAAACQCswzDwzP/WP7g6H/fkeSCsWWPrqodV7jPm/S0fai19qkV9vvcJK2nXWAWAAAAAAAAAAAApiQwyzatqjanC8zOdU5GgdnW2s+SfHZs+aWT3GOFu75MT9uRK+wzrbVzkxzas+j6K+0bAAAAAAAAAAAAtlcCs2zrbp/kgLG2T7bWTp3z/Rt7tnvCCvfbd+6ctsI+t/pgT9ulZtQ3AAAAAAAAAAAAbHdWOi09rLdH9rS9a+z7dyZ5WZLd5rTduqqu1lr77pT7/UVP29Wn7Gvcj5K8NckZSU4cff1yRn0DAAAAAAAAAADAdkdglm1WVe2b5O5jzWdlLDDbWju9qt6d5JCxdR+f5M+n3P3/9LQ9rKpeOFbddtlaa1uSPGglfQAAAAAAAAAAAAB/0DetPGwrDkmy81jbe1prp/Ws+/qetodW1W497UvxqSQnjLXtn+Q9VXWxKfsEAAAAAAAAAAAAVoHALNuyR/a09QVjk+SoJD8ea7tYkgdPs+PW2nlJntez6DZJWlU9o6ouOU3fAAAAAAAAAAAAwGwJzLJNqqobJbnWWPNxST7dt35rbUuSw3sWPX4Fw3hVko/0tO+f5IVJfl5Vn6mqv6mq61eV8w0AAAAAAAAAAADWwY7rPQC2KQ+tqoeuQr8Pb60dvsxt+qrLHjEKxk5yRJLnJNk0p+26VXXj1tr/LHP/aa1dUFX3SfLeJHfoWWVzkluMvpLktKr6YpL/Hn39T2vtjOXud5luWVUL/U6m9dzW2nNWoV8AAAAAAAAAAACYORUv2eZU1e5JHjDWvCVdIHai1tpPknyqZ9ETph1La+2sJHdJ8qwkv11k9YsmuX260O7Hk5xSVV+sqhdW1a2rSoAdAAAAAAAAAAAAVoHALNui+yXZc6ztM621Hy9h29f3tN23qvaddjCttfNba89LcvUkr05y1hI33SHJQUmekS7I+6uqel1V3WTasQAAAAAAAAAAAADzCcyyLXpkT1tfELbPe5KcOta2a5JHrGhESVprx7XWHpfkkkkOHu3rtGV0sfdoHJ+vqiOr6torHRMAAAAAAAAAAAAgMMs2pqoqyZ+MNZ+R5N1L2b61dnaSt/UselxVbVrh8Lbu4/TW2ptba/dOsl+Smyd5ZpKPj8a6FLdP8pWqetosxgQAAAAAAAAAAADbsx3XewBsU45orT1sncfwqJ62d7TWzlxGH69P8tixtismuVOSj0w7sD6ttXOTfG709YKq2jHJdZPcIsmtR/+754TNd07yoqo6oLX25CmH8JnW2q2m3BYAAAAAAAAAAAAGQYVZthlVtVOSh/Qsev1y+mmtfTHJsT2LnjDNuJa57/Naa19urb24tXbXJPumC86+PMmpEzZ7UlX9+WqPDQAAAAAAAAAAAIZKhVm2JXdNcome9o9W1XL72qWn7S5VdfnW2k+WPbIpjSrQHpXkqKr62yRPT/KMJDuNrfpPVfXu1trP1mpsAAAAAAAAAAAAMBQqzLItedSE9t2n+OoLi29O8tjZDnnpWmtntNaeleT2Sc4aW7xbkiev/agAAAAAAAAAAABg2ycwyzahqi6T5I5rsKtHVtXOS125qi5aVZeb5QBaa59J8rc9i+43y/0AAAAAAAAAAADA9kJglm3Fw5LssAb7uUSS+0xaWFW7VNVXquqnVXV2klOTHFtVm2Y8jtdkfpXZA6tqvxnvBwAAAAAAAAAAAAavb1p62FBGYdRH9Cx6UGvtrSvs+1tJrjXW/Pgkb+lbv7X2u6r641z43NkjyTWS/O9KxjK2n99W1bFJbjC26FJJfjOr/QAAAAAAAAAAAMD2QIVZtgW3TnLFsbbTk7xvBn0f1tN2s6q69gLb/LSn7W4zGMu4LT1tzlkAAAAAAAAAAABYJuE7tgWP7Gl7b2vtrBn0/aYk5/a0P2GBbb7U0/bYqtp5BuNJklTVLumq1o775az2AQAAAAAAAAAAANsLgVk2tKraK8m9eha9cRb9t9Z+neQDPYsOrqo9J2z2wZ62yyd55izGNPKQJLuPtf2stXbCDPcBAAAAAAAAAAAA2wWBWTa6g5PsOtb2iySfmuE+Dutp2yPJIRPWf2+SU3ran1lVj1vpYKrqykn+uWfR+1baNwAAAAAAAAAAAGyPBGbZ6B7Z0/bW1toFM9zHR9OFcMc9vm/l1tpZSV7Ys2hTkkOr6o1Vtf80A6mqWyT5TJJ9xhadn+QV0/QJAAAAAAAAAAAA2zuBWTasqrpekuv0LHrjLPfTWjs/yRt6Fl1rFGDt8x9JvjVh2cFJflxVr66qm1XVDgvtv6p2r6o7V9V7khyV5NI9q726tfadhfoBAAAAAAAAAAAA+u243gNgm/LQqnroKvb/mdbareZ831dd9tuttW+swr4PS/KMnvYnJPnseGNr7XdVda8kX0qyd892F0nymNHX6VX1tSQ/THJikt8l2SvJvkkOTHKDJDstMLZvJXnaEn+Ocbesqi1TbrsUP2mtHbiK/QMAAAAAAAAAAMCKCcyyIVXVRZI8qGfRm1Zjf62171fV55LcbGzRvapq/9baL3u2+UFV3SbJkUkuvkD3eya55ehr2UNL8qettbOm2BYAAAAAAAAAAABIsnm9BwAT3DtdFda5Lkjy5lXc52E9bTsledSkDVprx6SrEDuvCu0MfDDJzVprx69C3wAAAAAAAAAAALDdEJhlo3pkT9tnWms/W8V9viPJGT3tj6mqHSZt1Fr7aZJbJzkkyfdmMI7vJXlAa+3PWmu/mUF/AAAAAAAAAAAAsF0TmGXDqaorJbllz6I3reZ+W2tnpgvNjjsgyV0X2faC1tqbklwtXXj2VUm+m2TLEnd/Qrqf705Jrt5ae/tSxw0AAAAAAAAAAAAsbNOWLUvN8wHLVVV7JblqkgOTXCLJbkl2SfK7JKcn+UmS1lr74ToNcT0cmOTHT3nhf+U3J5+53mMBAAAAAAAAAAC2U2950YPz61+fvt7DYIY2b96UfffdI0mukOS4uct2XI8BwfaitXZKki+NvgAAAAAAAAAAAIB1sHm9BwAAAAAAAAAAAAAAq0lgFgAAAAAAAAAAAIBBE5gFAAAAAAAAAAAAYNAEZgEAAAAAAAAAAAAYNIFZAAAAAAAAAAAAAAZNYBYAAAAAAAAAAACAQROYBQAAAAAAAAAAAGDQBGYBAAAAAAAAAAAAGDSBWQAAAAAAAAAAAAAGTWAWAAAAAAAAAAAAgEETmAUAAAAAAAAAAABg0ARmAQAAAAAAAAAAABg0gVkAAAAAAAAAAAAABk1gFgAAAAAAAAAAAIBBE5gFAAAAAAAAAAAAYNAEZgEAAAAAAAAAAAAYNIFZAAAAAAAAAAAAAAZNYBYAAAAAAAAAAACAQROYBQAAAAAAAAAAAGDQBGYBAAAAAAAAAAAAGDSBWQAAAAAAAAAAAAAGTWAWAAAAAAAAAAAAgEETmAUAAAAAAAAAAABg0ARmAQAAAAAAAAAAABg0gVkAAAAAAAAAAAAABk1gFgAAAAAAAAAAAIBBE5gFAAAAAAAAAAAAYNAEZgEAAAAAAAAAAAAYNIFZAAAAAAAAAAAAAAZNYBYAAAAAAAAAAACAQROYBQAAAAAAAAAAAGDQBGYBAAAAAAAAAAAAGDSBWQAAAAAAAAAAAAAGTWAWAAAAAAAAAAAAgEETmAUAAAAAAAAAAABg0ARmAQAAAAAAAAAAABg0gVkAAAAAAAAAAAAABk1gFgAAAAAAAAAAAIBBE5gFAAAAAAAAAAAAYNAEZgEAAAAAAAAAAAAYNIFZAAAAAAAAAAAAAAZNYBYAAAAAAAAAAACAQROYBQAAAAAAAAAAAGDQBGYBAAAAAAAAAAAAGDSBWQAAAAAAAAAAAAAGTWAWAAAAAAAAAAAAgEETmAUAAAAAAAAAAABg0ARmAQAAAAAAAAAAABg0gVkAAAAAAAAAAAAABk1gFgAAAAAAAAAAAIBBE5gFAAAAAAAAAAAAYNAEZgEAAAAAAAAAAAAYNIFZAAAAAAAAAAAAAAZNYBYAAAAAAAAAAACAQROYBQAAAAAAAAAAAGDQBGYBAAAAAAAAAAAAGDSBWQAAAAAAAAAAAAAGTWAWAAAAAAAAAAAAgEETmAUAAAAAAAAAAABg0ARmAQAAAAAAAAAAABg0gVkAAAAAAAAAAAAABk1gFgAAAAAAAAAAAIBBE5gFAAAAAAAAAAAAYNAEZgEAAAAAAAAAAAAYNIFZAAAAAAAAAAAAAAZNYBYAAAAAAAAAAACAQROYBQAAAAAAAAAAAGDQBGYBAAAAAAAAAAAAGDSBWQAAAAAAAAAAAAAGTWAWAAAAAAAAAAAAgEETmAUAAAAAAAAAAABg0ARmAQAAAAAAAAAAABg0gVkAAAAAAAAAAAAABk1gFgAAAAAAAAAAAIBBE5gFAAAAAAAAAAAAYNAEZgEAAAAAAAAAAAAYNIFZAAAAAAAAAAAAAAZNYBYAAAAAAAAAAACAQROYBQAAAAAAAAAAAGDQBGYBAAAAAAAAAAAAGDSBWQAAAAAAAAAAAAAGTWAWAAAAAAAAAAAAgEETmAUAAAAAAAAAAABg0ARmAQAAAAAAAAAAABg0gVkAAAAAAAAAAAAABk1gFgAAAAAAAAAAAIBBE5gFAAAAAAAAAAAAYNAEZgEAAAAAAAAAAAAYNIFZAAAAAAAAAAAAAAZNYBYAAAAAAAAAAACAQROYBQAAAAAAAAAAAGDQBGYBAAAAAAAAAAAAGDSBWQAAAAAAAAAAAAAGTWAWAAAAAAAAAAAAgEETmAUAAAAAAAAAAABg0ARmAQAAAAAAAAAAABg0gVkAAAAAAAAAAAAABk1gFgAAAAAAAAAAAIBBE5gFAAAAAAAAAAAAYNAEZgEAAAAAAAAAAAAYNIFZAAAAAAAAAAAAAAZNYBYAAAAAAAAAAACAQROYBQAAAAAAAAAAAGDQBGYBAAAAAAAAAAAAGDSBWQAAAAAAAAAAAAAGTWAWAAAAAAAAAAAAgEETmAUAAAAAAAAAAABg0ARmAQAAAAAAAAAAABg0gVkAAAAAAAAAAAAABk1gFgAAAAAAAAAAAIBBE5gFAAAAAAAAAAAAYNAEZgEAAAAAAAAAAAAYNIFZAAAAAAAAAAAAAAZNYBYAAAAAAAAAAACAQROYBQAAAAAAAAAAAGDQBGYBAAAAAAAAAAAAGDSBWQAAAAAAAAAAAAAGTWAWAAAAAAAAAAAAgEETmAUAAAAAAAAAAABg0ARmAQAAAAAAAAAAABg0gVkAAAAAAAAAAAAABk1gFgAAAAAAAAAAAIBBE5gFAAAAAAAAAAAAYNAEZgEAAAAAAAAAAAAYNIFZAAAAAAAAAAAAAAZNYBYAAAAAAAAAAACAQROYBQAAAAAAAAAAAGDQBGYB4P+zd+9xll51ne+/1elIQhJuIRAkQMLFHwSBKCgwCATwikw0CAmRoNGAAjo6BoGj6CCoIDqHow56UAIEgRhuMowDqHOAcAdRcxSVWSKQcBVCCLkREpLU/LF32buf3t1dl/1Uda1+v1+v/ar9rP08a62qrtRfn9cKAAAAAAAAAADQNcEsAAAAAAAAAAAAAF0TzAIAAAAAAAAAAADQNcEsAAAAAAAAAAAAAF1bWl5e3uo9AAeX45N8aqs3AQAAAAAAAAAAHNy+ft03ctWVX9/qbbBAO3Ys5eijj0ySE5JcPPvZzq3YEMBll12dm24S7AP9OeaYo3LppVdt9TYARuFvHNAzf+OAnvkbB/TO3zmgZ/7GAT3zNw7YbDu2egMAAAAAAAAAAAAAMCbBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABd27nVGwAOTkcffeRWbwFgNMccc9RWbwH264brr8vlV1y/1dsAAAAAAAAAgE0hmAW2xEdf+uxcf+VlW70NADho3f9Z5yYRzAIAAAAAAABwcNix1RsAAAAAAAAAAAAAgDEJZgEAAAAAAAAAAADommAWAAAAAAAAAAAAgK4JZgEAAAAAAAAAAADommAWAAAAAAAAAAAAgK4JZgEAAAAAAAAAAADommAWAAAAAAAAAAAAgK4JZgEAAAAAAAAAAADommAWAAAAAAAAAAAAgK4JZgEAAAAAAAAAAADommAWAAAAAAAAAAAAgK4JZgEAAAAAAAAAAADommAWAAAAAAAAAAAAgK4JZgEAAAAAAAAAAADommAWAAAAAAAAAAAAgK4JZgEAAAAAAAAAAADommAWAAAAAAAAAAAAgK4JZgEAAAAAAAAAAADommAWAAAAAAAAAAAAgK4JZgEAAAAAAAAAAADommAWAAAAAAAAAAAAgK4JZgEAAAAAAAAAAADommAWAAAAAAAAAAAAgK4JZgEAAAAAAAAAAADommAWAAAAAAAAAAAAgK4JZgEAAAAAAAAAAADommAWAAAAAAAAAAAAgK4JZgEAAAAAAAAAAADommAWAAAAAAAAAAAAgK4JZgEAAAAAAAAAAADommAWAAAAAAAAAAAAgK4JZgEAAAAAAAAAAADommAWAAAAAAAAAAAAgK4JZgEAAAAAAAAAAADommAWAAAAAAAAAAAAgK4JZgEAAAAAAAAAAADommAWAAAAAAAAAAAAgK4JZgEAAAAAAAAAAADommAWAAAAAAAAAAAAgK4JZgEAAAAAAAAAAADommAWAAAAAAAAAAAAgK4JZgEAAAAAAAAAAADommAWAAAAAAAAAAAAgK4JZgEAAAAAAAAAAADommAWAAAAAAAAAAAAgK4JZgEAAAAAAAAAAADommAWAAAAAAAAAAAAgK4JZgEAAAAAAAAAAADommAWAAAAAAAAAAAAgK4JZgEAAAAAAAAAAADommAWAAAAAAAAAAAAgK4JZgEAAAAAAAAAAADommAWAAAAAAAAAAAAgK4JZgEAAAAAAAAAAADommAWAAAAAAAAAAAAgK7t3OoNzFNVO5LcJsnhSb7cWrt2i7cEAAAAAAAAAAAAwDa15cFsVd05ySOTPCzJg5Mcm+QWM7c8Kcn5g2dem+TLSV7TWvvIJm0VAAAAAAAAAAAAgG1oy4LZqvqOJM9O8kNJdkyHlwa3Le/l8QclOT7Jz1bVXyb5pdba34+xTwAAAAAAAAAAAAC2tx37v2Wxqurwqnp1kg8lOTXJIZmEskuZBLIrr709v5TkuOnlUpLvT/LXVfWMMfcNAAAAAAAAAAAAwPa0qcFsVd0jyV8n+dGsIZIduGOSQ6fvV545NMlvV9XLF7dbAAAAAAAAAAAAAHqwacFsVd01yYVJTszuoeyKpZnXvtxlcL0yz1KSs6rqRYvYLwAAAAAAAAAAAAB92JRgtqoOT/Lfk9xhOrQSyq4Esl9I8tokv7KK6a5K0rJnWLsSzf5iVT1yg1sGAAAAAAAAAAAAoBObdcLsc5N8a/Y8UfYDSR7ZWjuutfak1toL9jdRa+0fWmv3SnJqkn/N7uHsSjT70qo6ZGG7BwAAAAAAAAAAAGDbGj2YrapvTvKfs/upsknyh0ke1lq7cD3zttbekuR+Sc7PnqfN3i3JD61nXgAAAAAAAAAAAAD6snMT1viZJN+UXae/Lid5Y2vtZzc6cWvt60nOrKodSZ6Q3U+wfUqSP9voGgAAAAAAAAAAAABsb6OfMJvk9Owesn41ydMWvMZPJfnc9P1KmPvwqjpkwesAAAAAAAAAAAAAsM2MGsxW1Z2S3HV6uXK67Etaa19Z5DqttauT/P50jRU3S3LSItcBAAAAAAAAAAAAYPsZ+4TZ+80Ze8tIa71pzliNtBYAAAAAAAAAAAAA28TYwexdBtc3ttb+boyFWmufTHLVYPjWY6wFAAAAAAAAAAAAwPYxdjB7i8H1pSOvd/l+1gcAAAAAAAAAAADgIDN2MLvZjhxc37gluwAAAAAAAAAAAADggDF2MHvl4ProsRaqqsOT3GY/6wMAAAAAAAAAAABwkBk7mP304PrQqjpppLUeNWfskpHWAgAAAAAAAAAAAGCbGDuY/Yc5Y48Zaa0nzRn7x5HWAgAAAAAAAAAAAGCbGDWYba1dkl2nzC4nWUryn6rq8EWuU1XfmeRx0zVWfKa19plFrgMAAAAAAAAAAADA9jP2CbNJ8oZMQtkVt03yu4uavKpun+SCmTWWMgln37yoNQAAAAAAAAAAAADYvjYjmP2DJDdM36+cMvvkqnr2Rieuqrsl+askx2f302WT5I83Oj8AAAAAAAAAAAAA29/owWxr7eIkL8uuE2BXotkXVNX5VXWntc5ZVUdV1TlJ/jbJt2ZXLLtyuuwbW2sf2+jeAQAAAAAAAAAAANj+dm7SOr+U5NFJ7jy9XolmT0/yI1X1viTvTfLFOc/evqpOSnKbTOLYhyT5/iRHZvcId8VXk5yz2O0DAAAAAAAAAAAAsF1tSjDbWruyqh6b5MJMQtdkVzR7aJKTp68VSzNf/+ucKeeFsktJbkxyVmvtc4vYNwAAAAAAAAAAAADb347NWqi1dlGSH0zy5ewevK6Es0sz47OW5rxWnpu95/okZ7fW/nyM/QMAAAAAAAAAAACwPW1aMJskrbX3Jfn2JO/L7nHscvaMYOd9Nu+epSSfTvLQ1tqfLHrPAAAAAAAAAAAAAGxvmxrMJklr7XNJTk7y9CT/mj1Plt1XHLti5ZmrkvxOkm9rrX1kpC0DAAAAAAAAAAAAsI3t3IpFW2s3JXlpVf1RklOS/EiS70py/CoevyrJB5O8PckrWmtXjbVPAAAAAAAAAAAAALa/LQlmV7TWlpO8ZfpKVX1zknsnOTrJrZPcKsl1Sb4yfX0myd9Pg1sAAAAAAAAAAAAA2K9NCWaramkax+5Ta+3zST6/CVsCAAAAAAAAAAAA4CCxY+wFquoJSS6pqudX1QljrwcAAAAAAAAAAAAAs0YPZpP8bJLjkjwnycer6h1V9fBNWBcAAAAAAAAAAAAAxg1mq+rYJA9OspxkabreyUm+Mea6AAAAAAAAAAAAALBi7BNmH5RJKJtMotkk+d+ttQ+MvC4AAAAAAAAAAAAAJBk/mD1h5v1SJtHs+0ZeEwAAAAAAAAAAAAD+3djB7CFzxr408poAAAAAAAAAAAAA8O/GDmY/NWfsdiOvCQAAAAAAAAAAAAD/buxg9j1Jrpu+X06ylOT7qmpp5HUBAAAAAAAAAAAAIMnIwWxr7dIk52USyq64U5JfHHNdAAAAAAAAAAAAAFgx9gmzySSO/dtMotmVU2ZfWFU/vwlrAwAAAAAAAAAAAHCQGz2Yba1dk+QRSf48u6LZHUleXFUXVdVZVXXC2PsAAAAAAAAAAAAA4OC0czMWaa1dneSHquqnk7woyVGZxLP3S/LyJKmqS5NclOTSJJcn+WqSGze47vM38jwAAAAAAAAAAAAA29/owWxVfW3OmisnzS5NX0lyuyTfu+DlBbMAAAAAAAAAAAAAB7nNOGH2sL2Mr0Szw7FFGc4NAAAAAAAAAAAAwEFoM4LZZPXx6qIi10WGtwAAAAAAAAAAAABsY5sVzCYiVgAAAAAAAAAAAAC2wGYEs+/J4k6OBQAAAAAAAAAAAIA1GT2Yba2dPPYaAAAAAAAAAAAAALA3O7Z6AwAAAAAAAAAAAAAwJsEsAAAAAAAAAAAAAF0TzAIAAAAAAAAAAADQNcEsAAAAAAAAAAAAAF0TzAIAAAAAAAAAAADQNcEsAAAAAAAAAAAAAF3bOfYCVXXnsdfYm9bap7dqbQAAAAAAAAAAAAAODKMHs0kuTrK8CesMLWdzvj8AAAAAAAAAAAAADmCbFZQubdI6AAAAAAAAAAAAALCbzQpmN/uEWYEuAAAAAAAAAAAAAEk2L5gdizAWAAAAAAAAAAAAgH3arGB2jLB1efpamn79RJK3jrAOAAAAAAAAAAAAANvYZgSzJyxgjpsluXmSY5PcLckDk/xgkltnVzR7tyTfaK09awHrAQAAAAAAAAAAANCJ0YPZ1tolI0z7B1V1syQ/l+T5Sb4pk2j2GVW1s7V2zghrAgAAAAAAAAAAALAN7djqDaxXa+261trvJPnuJF/LrpNmf76qztzSzQEAAAAAAAAAAABwwNi2weyK1tr7k/xMJrHsSjT7u1V1my3dGAAAAAAAAAAAAAAHhG0fzCZJa+1Pknwkk1g2SW6d5D9t3Y4AAAAAAAAAAAAAOFB0EcxO/cH068opsz+5hXsBAAAAAAAAAAAA4ADRUzD7tsH1cVV1ry3ZCQAAAAAAAAAAAAAHjG6C2dbal5N8eTB8363YCwAAAAAAAAAAAAAHjm6C2amrB9ffvCW7AAAAAAAAAAAAAOCA0Vswe+vB9c23ZBcAAAAAAAAAAAAAHDC6CWar6q5JbjkY/tpW7AUAAAAAAAAAAACAA0c3wWySH5sz9plN3wUAAAAAAAAAAAAAB5QugtmqqiTPSLI8+OiiLdgOAAAAAAAAAAAAAAeQbR/MVtWDkvxVkiMGH32qtfaJLdgSAAAAAAAAAAAAAAeQnWMvUFV3XtBUhyT5pkzC2NsmuUeSRyf5vkzC35XTZZem789b0LoAAAAAAAAAAAAAbGOjB7NJLs6umHUMK4HsrCuTvGTENQEAAAAAAAAAAADYJjYjmE0mUetYZmPZlXj2ma21r25k0qq6OMld9vLxTUmOa619YSNr7GXdFyV51l4+/qPW2lMXveZGVNXxST41GH53a+3kBa7x4iS/MBj+RpI7tNYuW9Q6c9Y9Ocm7BsPXZfJv/+VVznF+kjPmfPSk1tprNrbD3db5tSTPXdR8q3BFa+1Wm7geAAAAAAAAAAAArNuOTVpnecTXipUo9/dba+eO++1kR5LHLXrSqlpK8oRFz3ugq6pbV9Uf7uOWef+ehyY5faQtrfiJOWNvXkMse+skp+7l46eve1cAAAAAAAAAAADAmmxWMDumpenrM0me2FobnkQ6ltNGmPM/JLnzCPMekKpqqarOStKSPG1v97XW/jnJB+d8dOZIW0tVHZnkR+Z89LI1TPPEJIft5bMHV9X91rwxAAAAAAAAAAAAYM02K5hdGuF1U5JLkrwmk3j1W1prf7pJ30+SPKSq7rjgOc9Y8HwHrKq6b5L3JnllkmNW8ci8U2YfXFV3W+jGdnl8kiMGY59I8q41zHH2fj53yiwAAAAAAAAAAABsgp2bsMYJC5xrOckNSa5JcmVrbXmBc6/VUiZR5e8uYrKqOmQ638Hi95M8ZA33vy6Tn/VRg/EzkzxvQXuaddacsXNX+ztXVfdPctJg+MNJHjhz/cSqemZr7cp17RAAAAAAAAAAAABYldGD2dbaJWOvsYmWMwllV5yWBQWzSR6V5HYLmqs7rbVrquqCJE8ZfLTwYLaq7prkoYPhG5Kct4ZphqfLfibJf8vuwewRSX58Oj6Gd7fWTh5pbgAAAAAAAAAAANg2dmz1BraZ9w+uH1RVd1rQ3GcMrr+yoHl7cu6csbtX1YMWvM6PZ/cwOkn+vLX2b6t5uKoOT/Kjg+H3JPmfSa4fjD9tXTsEAAAAAAAAAAAAVk0wuzYXDK6XMjlldkOq6mZJTh0Mv3Gj8/amtfbXST4656MnLWqNqlrKJJgdetkapnlcklsOxt7bWrsiyV8Oxu9VVSevYW4AAAAAAAAAAABgjboLZqvq5lX1wKr6qao6asHTvzXJVYOxDQezSR6d3QPLTyX54ALm7dG8U2ZPq6pDFzT/I5LcZTD26ewZuu7L2YPr5Ux+d5I9o+skefoa5gYAAAAAAAAAAADWaOfYC1TVOwdDv9lae8cI6xyayemjd8uuEPijWWx4+vUkb0ly5szYd1bVXVprl2xg3jMG1/OiSiZeneRFSQ6bGbttkh9I8j8WMP9Zc8Ze3lq7aTUPV9XdkzxsMPyR1tpnp+/fkuSaJEfMfP7DVXWH1toX1rpZAAAAAAAAAAAAYP8244TZk5M8fObr7cdYpLX2jSRHJzkkydJ0+O4jLDUvZl33KbNVdWSSxwyGz1/vfL1rrV2e5M/mfHTmnLE1mf5bPHYwfGOSV6xhmrOz6/dvxZtW3rTWrsme+z80yVPWsAYAAAAAAAAAAACwBqOfMLvJrk5ym+wKFo8eYY2/SnJ5klvPjJ2e5HfWOd8PJzl85vqjrbV/rKoHrGWSqrowkyB51gmttYvXOM+vJXnuYPgnWmvnrWWeOfOeleSV+7lneTB0SWvt+Dm3npvkRwdj/7Gqbtlau2Ldm5yEz0cMxv5i5nTYfaqqQ5L8+JyP3jS4Pi/JkwZjT6mq32yt3biatQAAAAAAAAAAAIDV24wTZjfFNFa87WB4GD9u2PQk2+EJofevqruuc8ozBtd/us55DiYXJvnXwdhhSR63wXnPmjP2sjU8/+gkdxiMva+19onB2LuSXDIYOy7JKWtYCwAAAAAAAAAAAFilboLZTP6X9sNA9tqR1rpgzthpa52kqo5O8j0zQ8sRzO5Xa205ySvmfDQ8tXXVqupuSR46GP5CkreuYZqz54ydNxyY7v9Vc+59+hrWAgAAAAAAAAAAAFZp53oeqqqzk9xxnWs+tqruvs5nh3YkOSrJiUm+N5PgdNZnF7TO0LuSfDHJ7WfGTk/yW2uc53FJDp25/mBr7eKNbe2A9Y0k18xcH549g+1r9nM967wkz8/uv8MPq6o7t9Y+vY79nTVn7JWttRtW83BVHZvkBwfDX0vy+r08cl6SX02yNDP2qKr6ltbav6xmTQAAAAAAAAAAAGB11hXMJvlqJv+r+mGgujdLM19Pnb4WbWnOfv5mhHXSWruxqt6Y5Gdmhk+qqnu01j6+hqnOGFyfv/HdHZhaa69N8tqV66q6MMnDB/ccuYb5vlBVb0tyyszwUpInJnnhWvZWVUtJfmwwvJzk3DVM8+PZ87+nN7bWrpp3c2vtU9OfwSNmhpeSPDXJOWtYFwAAAAAAAAAAANiPdQWzrbU3VdVfZXKq61ot7f+WdVkevP9wa+2TI62VJBdk92A2SU5L8purebiq7pjkoTNDNyR5w2K2dtA4N7sHs0lyZtYYzCZ5ZJI7D8b+v9bap9Ywx0/OGXvlfp55ZXYPZpPkrKp6Tmvt2jWsvTcPr6rVRu1r8bzW2q+NMC8AAAAAAAAAAACMYscGnv3ZJNdlEqfu6zW0v/vX+1qxlORrSZ6yge9tNd6f5LODsdPX8Pzp2f3n/47W2pc2vKuDy9uSfH4wdmJVffsa5zlrztjLVvtwVT0sybcMhj+V5N37efRNSYYn0N46e548DAAAAAAAAAAAAGzAuoPZ1tq/Jvmd7Doxdmkvr6G93beo12eTnNJa+6f1fm+r0VpbTvL6wfB9quqeq5xiGEWev/FdHVxaazcmOW/OR2eudo6qOirJYwfDlyZ5yxq2cvacsVdNf0f2qrX2tSSvm/PR09awNgAAAAAAAAAAALAfOzf4/G8mOSnJUfu45+GZnAC7NP3aknxxg+smyU1JbkhybZKvZBLK/n2St09DxM1wQZJzBmOnJXn+vh6qqrsnecDM0NeTvHmxWztovDzJL2X3OPuMqnrmNKjdn9OT3Hww9qrW2vWrWbyqbpHkcYPh5SSvWs3zSV6R5MmDsQdU1Xe01j6yyjkAAAAAAAAAAACAfdhQMNtauy7JKfu6p6puGgz9Rmuti9NUW2sfqapPJLnbzPB+g9nsebrs/2ytXbXQzR0kWmufrKp3JnnUzPCxSb47yV+uYoqz5oydu4YtnJE9g9sLW2sXr+bh1toHq6olqcFHT0/yE2vYBwAAAAAAAAAAALAXGz1hluR1SX555vreVXXv1to/7eOZYTDbRUC8hc7N7sFskjwp+wlmpyf9PmQw/J7WWlvD2sPTYZPklWt4fuX+3xqMnV5Vz2itfWWNc816d2vt5A08DwAAAAAAAAAAAF3YsUnrLG3SOlvhgjljp+3t5qq6X5J7zQxdkeRti97UQebNSS4bjP1wVR2xn+fOmjP2stUuWlX3TfKAwfBVSd602jmm/iTJjYOxw+OEWQAAAAAAAAAAAFiIzThh9lWD609swpqbprX20ar6WHaPYE9L8ty9PDI8XfbPWmvXjbK5g0Rr7bqqek2Sn58ZPiLJY5O8et4zVbUjk1NoZ12e5I1rWPrsOWNHJPlSVa1hmiTz4/WnVtWLW2vLa50MAAAAAAAAAAAA2GX0YLa1djCcknlBkufNXN+zqu7bWvuH2ZuqainJEwbPnj/25g4S52b3YDZJzsxegtkkj0xy58HYq1trX1/NYlV1s+n8QzsyiWYX4e5JvifJXy1oPgAAAAAAAAAAADgozTvVkrW7YM7YaXPGHpzkLjPX/5bkXaPs6CDTWvvHJB8eDD+qqu6wl0fOmjP2sjUseWqS26zh/vV6+iasAQAAAAAAAAAAAF0TzC5Aa+1fklw0GJ4XzJ4xuH59a+3GcXa1bods9QY24NzB9SHZ82eeqrpFkscOhj80jW5X6+w17m29HlNVd9qktQAAAAAAAAAAAKBLO7d6A2OoquOSnNNaO2cTl70gybfNXN+jqr6ttXbRdE+HJHn84JnzR97TeoLowxe+i81zQZL/J8mRM2NPSvLiwX2nZ8/vc9Wny1bV8UkeNRi+KcmdW2ufW+08c+a9fZLPZvf/Lg9J8lNJfnW98wIAAAAAAAAAAMDBbsuD2aq6ZZLDpnvZkWRpP4/syCQi3JnkZpmEj0cluW2Se2YSMj5wes9mBrOvT/Kiwdjjs+vk2Uckuf3MZ59orX14gesvzxlbz7/v7Ta6ka3SWru6ql6X3U9/PamqTmyt/fPM2I8NHr0qyevWsNRPZs/f0ws3EssmSWvti1X1tiSnDD56clU9v7X2jY3MDwAAAAAAAAAAAAerTQ9mq+qhSZ6Q5AeS3CnrOwV1f5YyPyAdTWvt4qr6UJIHzQw/PskvT9+fNnjkggVv4cY5Y0fOGdufu250I1vsZdk9mE2SMzI9obWq7pLkIYPPz2+tXbOayatqR5Kz5nz0mrVtc69ekT2D2WOTPDZri3oBAAAAAAAAAACAqTFi1bmq6vZV9WdJLkzy1CTHZ3IK7NKCX1tpGMHevapOqqqdmQSPs85f8NpfmzN29FomqKqbJTlpIbvZItNTez86GD595v2PZs/fk5etYYnvzST0nnVtkjetYY59eWuSL84Zf/qC5gcAAAAAAAAAAICDzqYEs1V1qyTvTfJD2RW2Lo/02kqvT3LTYOxHkjwyu8erf99a++cFr33ZnLF7rXGOxyQ5YgF7WY3hz2mRXj64vkdVfdv0/RMGn13UWvvbNcz95Dlj/6O1duUa5tir1toNSV4956OHVdWJi1gDAAAAAAAAAAAADjabdcLsf0ty9+n7/YWtqz0tdn+nzF663s2uV2vtC0neMxg+NcnjBmOLPl02ST4+Z+y01T5cVYcnef7itrNf35izh50LmvvVSa4bjJ1aVXdNct/B+B+vdtKqOibJKXM+es3atrdfr9jLuFNmAQAAAAAAAAAAYB1GD2ar6t5JfjS7ItnZEHZvseu+Pp93Qm2mXz+Q5P/O5CTb4xb8razWBYPreyc5Y+Z6ec49i/C/5ow9pKrO3N+DVXVYkj9NspknmF49Z+y2i5i4tfaVJG8eDJ+SPWPXr2Vt8fKTkhw6GLs0yV+saYP70Vr7WJIPzVu/qo5c5FoAAAAAAAAAAABwMFjUiZ778tTsClwz/bqU5KpMYsUPJvlikq8n+ZnsOo11OclLk7w+k7D38CS3T1JJHpnkAdnzpNpLW2vPHOsbWaU3JXlJdv/ZzkaO72+tfXqEdf8mycVJjh+Mv6Kqjk7yktbajcOHqur7M4mMNzOWTeafAHx6kt9b0PznJnnCzPX9kvz04J7XtdauXMOcZ88Ze11r7Ya1bm4VXpHkQYOxWyQ5M5P/Llbj4VW1r9OcN+qS1trxI84PAAAAAAAAAAAAC7EZwewPZffTZZcziTsf01r70uyNVXVUJsHsyv33bK29e96kVfXATMLB+2VXhHtKVZ3TWnvxwr+LVWqtfbmq3pHk+/Zyy1pONF3LustV9atJXj346NAkv5vkmVX1F0k+lcnP6vhMwuMTBve/MMmzkhwyxj5nfHzO2Iur6kFJPprJ7+YtW2vPWOf870zyySR3nRm75+Cel612sqp6cOZHxa9Z+9ZW5YJM/t1uPhh/WlYfzAIAAAAAAAAAAACZnNw6mqo6Lslxg+Frkjx+GMtOvTvJTdP3S0m+q6qOmDd3a+3DSR6c5K3ZFeIuJfmNqvqWBWx/Iy7Yy/gNSd4w4rqvTfLmvXx2x0xOSP2NJL8+fT+MZc9N8pzRdre7tyUZnni7I5NTYX8zyfOSnFNVt1vP5K215UxOad2bf2qtfXANU847Xfbj09/DhWutXZXkjXM+um9VPWSMNQEAAAAAAAAAAKBXowazSe49834lav2T1tol826eRoIfmxnameS79jZ5a+3rSR6fyYmkmc5/syS/s4E9L8Kbk1w3Z/x/tda+PNai00j0tExOJr1p33fv5rokz2ytPWU6x+haax/L5DTb/TlpA8u8MntGuSvWcrrskUlOn/PRWKfLrthb8Pv0kdcFAAAAAAAAAACArowdzN5xztjr9vPM32YS16540L5unkazT555ZinJY6rqfqvd5KK11q5I8hdzPjp/E9a+obX2C0m+Lcl5Sb6yj9uvzuRU2RNba/917L0NtdZ+NZP48/J93HbSBub/fCYn2Q5dl+TVa5jqtCRHzhkfO5h9T5JPzBl/XFUdM/LaAAAAAAAAAAAA0I2l5eXxDhStqmdkctrrciYh601JbtFa+9o+nnlmkhdNn0mSt7bWTlnFWm9P8n3Ty+Ukf9Ba+7kNbL8LVXVIkgcmuWuSY5McmuSyTE7y/XBr7fot3F6SpKoOS/Ifktwrya0y+T35apKLk1zUWvu3rdobozg+yac++tJn5/orL9vqvQDAQev+zzo3l1561VZvg23mmGOO8nsDdMvfOKBn/sYBvfN3DuiZv3FAz/yNA8awY8dSjj76yCQ5IZMG8d/tHHntwwbXX91XLDv1TzPvl5Lcd5VrnZ9JMLsS5z5qlc91rbV2Y5IPTF8HpOkpwe+cvgAAAAAAAAAAAAAWasfI8980uL5yFc/86+D6uKo6fBXPDWPLe1bVbVfxHAAAAAAAAAAAAAAdGzuYvXZw/U2reObiTE6JXbGUpPb3UGvtc0muHgx/+yrWAwAAAAAAAAAAAKBjYwezlw6ub7m/B1pr1yf5/GD4Xqtc7/LB9e1W+RwAAAAAAAAAAAAAnRo7mP3M4PrmVXWHVTx3cSYny65YbTC7NLi+7SqfAwAAAAAAAAAAAKBTYwezH50z9vBVPPeJ6dfl6df7r3K92wyuD1/lcwAAAAAAAAAAAAB0atRgtrV2eZKPD4afvIpH//fM+6UkD6mqQ/b1QFXdK8nNB8NXrGItAAAAAAAAAAAAADo29gmzSfK2TKLX5enXR1TV/qLZiwbXRyU5dT/PPHHO2GWr2iEAAAAAAAAAAAAA3dqMYPa1M+9Xotn/t6p+o6oO28szH0hy3eCZ366qW867uapOTPIL03tn/cO6dw0AAAAAAAAAAABAF0YPZltrf5NJALtiOckhSX4pycVV9YKquufgmauSvD2TUHbFXZK8p6q+Y/beqjolyTuSHD5Y+vLW2scW810AAAAAAAAAAAAAsF3t3KR1zkny/uwKdFdOjb1dkmdncprs8wbPvCzJD89cLyW5T5IPVdUlSb6Q5IQkt59+tjxz33KSP130NwEAAAAAAAAAAADA9jP6CbNJ0lr76yT/JbufGLs88/6Tc555e5J3Z1cAOxvEHp/kQUmOze6x7IpvJPn9BWwdAAAAAAAAAAAAgG1uU4LZJGmtvTDJr+3l4z2C2amfTvKVmevlmdcwpM3M2K+11j6+kf0CAAAAAAAAAAAA0IdNC2aTpLX2/CSPSPJ32f202bnBbGvtX5KcmuTKwf3JnqfKrnh5a+23NrhVAAAAAAAAAAAAADqxqcFskrTW3tNa+44kD0/yR0n+ubX2hX3c/94kJyV5bybR7Morg+srk/zn1tpPjbd7AAAAAAAAAAAAALabnVu18DSEfe8q770kycOr6uQkj0ty/yRHJ7kqyaeT/GWSN7TWLhtntwAAAAAAAAAAAABsV1sWzK5Ha+3CJBdu8TYAAAAAAAAAAAAA2EZ2bPUGAAAAAAAAAAAAAGBMglkAAAAAAAAAAAAAuiaYBQAAAAAAAAAAAKBrO7d6Ayuq6tAk357k2CS3SXJ0ksOTvLm19o+De09M0lprN276RgEAAAAAAAAAAADYVrY0mK2q2yZ5apLvTvKdSW4257ZPJPnHwdgrk9y1ql6X5Ldba58edaMAAAAAAAAAAAAAbFtbEsxW1R2S/EqSs5IcNh1emnPr8l6mOD6TE2ifluTsqnpJkue01q5f7E4BAAAAAAAAAAAA2O52bPaCVfXIJBdlcrLs4ZmEskuZxLGzr709f3iSY6b3LGVyKu05Sd5fVXcZdfMAAAAAAAAAAAAAbDubGsxW1c8l+cskt8uekeyKeSfNzpqNYleeXUpy/yTvrKpjF7ZhAAAAAAAAAAAAALa9TQtmq+rHkvxukkMyP5Jdee3PvFNkV+Y6Icnbq+rQ9e8UAAAAAAAAAAAAgJ5sSjBbVScmeen0chjKXpzkhUm+J8mdVjHdh5I8J8nns3tguzLvfZP88ga2CwAAAAAAAAAAAEBHNuuE2T9Mclh2Ra1LSb6e5GlJ7tFae05r7R2ttc/tb6LW2hWttRcm+ZYkL8juAe7ydO7/q6qOXeQ3AAAAAAAAAAAAAMD2NHowW1WPSPKw7B7LXpfk0a21P2qt3bSeeVtr17bWfiXJY5JcP/j4m5Kctb4dAwAAAAAAAAAAANCTzThh9qdn3i9lEs7+TGvt3YuYvLX2F0mePJ17dp0nLWJ+AAAAAAAAAAAAALa3UYPZqjokyaMziWRXgta/ba29YpHrtNZem+Sd2RXkJsk9q+rWi1wHAAAAAAAAAAAAgO1n7BNm75fkyJnr5SQvHWmtl8wZ+46R1gIAAAAAAAAAAABgmxg7mK05Y28baa13ZdfpsivuONJaAAAAAAAAAAAAAGwTYwezdxhcX9ta+7cxFmqtXZHky4PhW42xFgAAAAAAAAAAAADbx9jB7OGD62HQumjXDq5vPvJ6AAAAAAAAAAAAABzgxg5mrxtcHznyescMrocBLQAAAAAAAAAAAAAHmbGD2csG17eqqiPGWKiqjsvmn2gLAAAAAAAAAAAAwAFu7GD2E4PrpSQPHWmt/7iK9QEAAAAAAAAAAAA4yIwdzP5dkhsHYz+26EWq6pAkP5dkeWb4piQXLXotAAAAAAAAAAAAALaXUYPZ1trVST6Yycmyy9Ovj6+qb13wUuckqen7lbU+3Fr72oLXAQAAAAAAAAAAAGCbGfuE2SR59eD6kCQXVNURi5i8qk5J8sLsfrpsklywiPkBAAAAAAAAAAAA2N42I5j9kyRfmL5fiVpPTPKuqjpmIxNX1bOSvCF7fh9XJDlvI3MDAAAAAAAAAAAA0IfRg9nW2nVJnplkaTq0Es0+IEmrqmdX1Z1WO19VHVlVZ1XV/5/JybKHzsy5NH3/6621qxexfwAAAAAAAAAAAAC2t52bsUhr7fyqekySJ2QStC5nErfeKskLkrygqj6T5ItzHn98VX1nktsk+dYk98kk9B0GuCvvP5Dk90b4NgAAAAAAAAAAAADYhjYlmJ06O8lxSb4ru6LZZFf4eufpa3ZsKckpM3MsZXfLg88+meRxrbWbFrRnAAAAAAAAAAAAALa5HZu1UGvt2iQ/kORPs3v4upzdA9qhpZnX3u5fSvLXSU5urc07pRYAAAAAAAAAAACAg9SmBbNJ0lq7prX2xCRPTfK1zD8xdhjDLu9lPDPPvyTJQ1trn134pgEAAAAAAAAAAADY1jY1mF3RWvvjJHdJ8twkX8qep8gmew9kV143JHlVkvu01n6utfaNsfcNAAAAAAAAAAAAwPazc6sWbq19JcmvV9WLknzX9PXQJCcmuU2Sm83cvpzkiiSfSfK+JO9N8o7W2qWbumkAAAAAAAAAAAAAtp0tC2ZXtNauT/LO6evfVdXhSW6V5Lokl7fWhqfNAgAAAAAAAAAAAMB+bXkwuzettWuTXLvV+wAAAAAAAAAAAABge9ux1RsAAAAAAAAAAAAAgDEJZgEAAAAAAAAAAADo2s61PlBVnxwMLbfW7rag/QAAAAAAAAAAAADAQq05mE1yfJLlJEvT6+WF7QYAAAAAAAAAAAAAFmzHBp4VygIAAAAAAAAAAABwwFvPCbMLUVV/Nhhabq39yJZsBgAAAAAAAAAAAIBubVkwm+SHs+uU2qU4sRYAAAAAAAAAAACAEezY6g1kEssCAAAAAAAAAAAAwCgOhGDWybIAAAAAAAAAAAAAjOZACGYBAAAAAAAAAAAAYDSCWQAAAAAAAAAAAAC6JpgFAAAAAAAAAAAAoGuCWQAAAAAAAAAAAAC6JpgFAAAAAAAAAAAAoGuCWQAAAAAAAAAAAAC6JpgFAAAAAAAAAAAAoGuCWQAAAAAAAAAAAAC6JpgFAAAAAAAAAAAAoGuCWQAAAAAAAAAAAAC6JpgFAAAAAAAAAAAAoGuCWQAAAAAAAAAAAAC6JpgFAAAAAAAAAAAAoGuCWQAAAAAAAAAAAAC6JpgFAAAAAAAAAAAAoGuCWQAAAAAAAAAAAAC6tnMRk1TVfzmQ5lnRWnv+IucDAAAAAAAAAAAAYPvZSDC7NPP1uRucY6Pz7I1gFgAAAAAAAAAAAOAgt5ATZrN7+HogzJMkywucCwAAAAAAAAAAAIBtalHB7Hri1Hlx7KIi10WGtwAAAAAAAAAAAABsYxsJZjcatzoBFgAAAAAAAAAAAIDRrTeYdYIrAAAAAAAAAAAAANvCeoLZExa+CwAAAAAAAAAAAAAYyZqD2dbaJWNsBAAAAAAAAAAAAADGsGOrNwAAAAAAAAAAAAAAYxLMAgAAAAAAAAAAANA1wSwAAAAAAAAAAAAAXRPMAgAAAAAAAAAAANA1wSwAAAAAAAAAAAAAXRPMAgAAAAAAAAAAANA1wSwAAAAAAAAAAAAAXRPMAgAAAAAAAAAAANA1wSwAAAAAAAAAAAAAXRPMAgAAAAAAAAAAANA1wSwAAAAAAAAAAAAAXRPMAgAAAAAAAAAAANA1wSwAAAAAAAAAAAAAXRPMAgAAAAAAAAAAANA1wSwAAAAAAAAAAAAAXRPMAgAAAAAAAAAAANA1wSwAAAAAAAAAAAAAXRPMAgAAAAAAAAAAANA1wSwAAAAAAAAAAAAAXRPMAgAAAAAAAAAAANA1wSwAAAAAAAAAAAAAXRPMAgAAAAAAAAAAANA1wSwAAAAAAAAAAAAAXRPMAgAAAAAAAAAAANA1wSwAAAAAAAAAAAAAXRPMAgAAAAAAAAAAANA1wSwAAAAAAAAAAAAAXRPMAgAAAAAAAAAAANA1wSwAAAAAAAAAAAAAXRPMAgAAAAAAAAAAANA1wSwAAAAAAAAAAAAAXRPMAgAAAAAAAAAAANA1wSwAAAAAAAAAAAAAXRPMAgAAAAAAAAAAANA1wSwAAAAAAAAAAAAAXRPMAgAAAAAAAAAAANA1wSwAAAAAAAAAAAAAXRPMAgAAAAAAAAAAANA1wSwAAAAAAAAAAAAAXRPMAgAAAAAAAAAAANA1wSwAAAAAAAAAAAAAXRPMAgAAAAAAAAAAANA1wSwAAAAAAAAAAAAAXRPMAgAAAAAAAAAAANA1wSwAAAAAAAAAAAAAXRPMAgAAAAAAAAAAANC1nVu9AeDgdJ+nvmirtwAAB7Ubrr9uq7cAAAAAAAAAAJtGMAtsicsuuzo33bS81dsAWLhjjjkql1561VZvAwAAAAAAAACAGTu2egMAAAAAAAAAAAAAMCbBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0DXBLAAAAAAAAAAAAABdE8wCAAAAAAAAAAAA0LWdW70B4OB09NFHbvUWAJIk111/fa684rqt3gYAAAAAAAAAACMSzAJb4hff8Lx8+eqvbPU2AHLeT/xeEsEsAAAAAAAAAEDPdmz1BgAAAAAAAAAAAABgTIJZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALommAUAAAAAAAAAAACga4JZAAAAAAAAAAAAALq2c6s3AFV1ZJL7JLlrkmOSHJHkpiTXJPlSkk8m+afW2jVbtkkAAAAAAAAAAABg2xLMsiWq6vgkT0xyapKTkhyyn0duqKqLkvz3JK9trV2yjjVPTvKu4Xj7P+3dd7xtd1kn/s9NQhqhE0IRCGB4pEmAIVJEyiCRpiBFIlUdHUEUGEDHCoioQ7HLDE5oIh0RBGVUOkgv0nkUpEvoJQTS7++PffB3su+69/Sz91rn/X69zgv3s9b6rufeJI/n7PPZ39W9b6NrbUZVPTbJY+bKb+ju26xx3YlJPnGIUz7S3dfdUnMHv/fbk5xykMOndfcLduK+AAAAAAAAAAAAsJ0OW3QD7C1VddWqem6SjyX5nSQ3ydph2WQW7r5pkick+XhV/eVKkJTkOlV1g+1etKqulYOHZQEAAAAAAAAAAGA0BGbZNVV17yQfTPKTWV9I9mAOT3L/JO+vqgdsR28TcO8dWPM+O7AmAAAAAAAAAAAA7DqBWXZFVf1MkhckueQ2LnuJJM+uqkdv45pjtROB2dN2YE0AAAAAAAAAAADYdUcsugGmr6puluR/J9l3kFMuSNJJPpTki0nOTLI/s3DtCUm+P8lJh7j+iVX18e5+6Xb2PTLXrqqTu/tftmOxqrpBkuttx1oAAAAAAAAAAACwaAKz7KiqOizJU5NcbODwF5L8TpIXdveX1ljne5LcP8mvJLnUwCmnV9Xru/urW2x5LPbnwADxvZP8yzatb3dZAAAAAAAAAAAAJuOwRTfA5J2a5EYD9dckqe7+s7XCsknS3Z/t7t9LUkneNnDKZZL86pY6HZePZxY4Xu3e27j+feZe75UgMgAAAAAAAAAAABMkMMtOG9qp9KNJ7trd39joYt39hSR3WFlj3oOqamgn2ym6IMmL52rXqqqbbHXhqrpZkmusKn0zyT9tdV0AAAAAAAAAAABYFIFZdtrNB2q/0d3f2eyC3X1mkp8bOHT5JDfb7Loj9IKB2nbsMjsfcn5pkrO3YV0AAAAAAAAAAABYCIFZdtpVBmr/uNVFu/tNSf5l4NCWd1gdkbck+fRc7V5bWbCqDsuBodvnb2VNAAAAAAAAAAAAWDSBWXba/L9jFyY5a5vWfuVA7UrbtPbS6+79SV40V75GVd10C8veNskVV73+QpLXbGE9AAAAAAAAAAAAWLgjFt0Ak/cfSa6x6vVhSa6d5KPbsPY/Jblmkq8k+erK/759G9YdkxcmedRc7SeSvHOT65029/pF3X1BVW1yOQAAAAAAAAAAAFg8gVl22tty0cBskjwsyYO3unB3vzHJG7e6zph197uq6mNJvndV+Z45MES7pqo6MsmPz5Wfv4X2AAAAAAAAAAAAYCkctugGmLznDdR+vqp+cdc7ma4Xzr2+elXdbBPr/EiSy6x6/Ynufuvm2wIAAAAAAAAAAIDlIDDLTvu7JO8YqP9JVb2pqu5eVRfb7aYm5gUDtZ/YxDqnzb22uywAAAAAAAAAAACTIDDLjuru/Unun+SbA4d/MMlLk3yxqv6qqu5fVVfZ1QYnoLs/mORDc+V7VtW+9a5RVRdP8qNz5aHdgQEAAAAAAAAAAGB0jlh0A0xfd/9rVd0hs91mLzdwyqWT3HflK1X1iSRvSfLPK18f7O4Ld6fb0XpBksevev09SW6R2d/fevxokmNXvf5Ad8+HcAEAAAAAAAAAAGCU7DDLrujutye5UZJXr+P0a2QWnn1qkvcl+UpVvayqfqmqvncH2xyzFw7UfmID158299rusgAAAAAAAAAAAEyGwCy7prs/090/nOSuSd68gUsvneTHkvxxkn+rqg9V1a9V1Qk70OYodfe/JXnPXPkeVbXmf+NVdZkkp64q7U/y/G1sDwAAAAAAAAAAABZKYJZd192v7O5bJbl2kt9I8o4kF25giesmeUKSf6+q36+qY3egzTF6wdzrKyf5wXVcd48kR656/dbu/tS2dQUAAAAAAAAAAAALJjDLwnT3v3X3E7r7B5KckOSeSf4syfsz2+V0Lccm+ZUk76+qG+9cp6Pxwhz493bvdVx32tzr521POwAAAAAAAAAAALAcBGZZCt395e7+6+7+xe6+YZLLJblrkqckeU8OHaC9VpI3VNVtdr7T5dXdn07y1rnyPavqoP+dV9WVktxmVen8JC/a/u4AAAAAAAAAAABgcQRmWUrd/bXufmV3P6q7b5LZDrQPTPIPGQ7PHpfkr6vqxF1scxm9YO71CUlufYjz752LzoFXd/eXtr0rAAAAAAAAAAAAWCCBWUahu7/U3X/Z3T+S5PuTvHbgtMsm+ZPd7WzpvDjJhXO1ex/i/NPmXj9/e9sBAAAAAAAAAACAxROYZXS6+4NJfjjJnw4cvmtVXWeXW1oa3X1GkjfMle9RVYfPn1tV10zyA6tK30nyNzvYHgAAAAAAAAAAACyEwCw7rqoOq6orVNVlt2vN7r4wycOTvGvg8KF2VN0LXjD3+vgktx047z5zr1/Z3WfuTEsAAAAAAAAAAACwOAKz7IiquntVfbCqvpjkvCRfSPL47bzHSmj2TwYO3XQ77zNCf53Z3/lqQyHi0+ZeP39n2gEAAAAAAAAAAIDFEphlpxyb5HqZ7W763X/PTtmB+7xnoHalHbjPaHT3V5K8eq7841V1xHdfVNX1klx/1fGvJ/n7ne8OAAAAAAAAAAAAdp/ALDvl0wO1m1TVVbb5PvsHav69Tl449/pySW636vX87rIv7e5zdrYlAAAAAAAAAAAAWAzBQnbKe5JcMFfbl+Sh23yfGw3UvrDN9xijv0kyH4C916r/+95zx56/s+0AAAAAAAAAAADA4gjMsiO6+6wkrxs49PCqut423urBA7V3buP6o9Td30zyqrny3avqiKq6UZKTVtXPSPLaXWsOAAAAAAAAAAAAdpnALDvpGQO1o5P8XVVde6uLV9Ujktxy4NDLt7r2RLxg7vXlktw6F91pNkle2N0X7k5LAAAAAAAAAAAAsPsEZtlJL0rywYH61ZO8o6p+rqoO3+iiVXWxqnpMkqcMHH5rd79ro2tO1CuSnDVXu3uSe87Vnrc77QAAAAAAAAAAAMBiCMyyY7r7giQPSXLBwOFLJXlakq6qR1XV1dZar6quXlUPySyE+9gk++ZOuTDJI7bU9IR097czC82u9sAkJ616/fHufsfudQUAAAAAAAAAAAC774hFN8C0dfebqupXkjz5IKdcK8mTkjypqj6Z5ANJPpXkzMwCsZdLcvkkN0xyzTVu99jufvtGe6yq/Ru95hB+qruftY3rbdULk9xn1evj5o4/fxd7AQAAAAAAAAAAgIUQmGXHdfdTquroJL+zxqknrnxtxp+vY/296FVJvpHZjr5DnreLvQAAAAAAAAAAAMBCHLboBtgbuvsJSe6Z5CvbvPTZSR7e3Q/t7u3cKXYSuvucJC87yOH3dfdHdrEdAAAAAAAAAAAAWAiBWXZNd/91kusk+YMk39nicvuTvCTJyd39x1vtbeJecJC63WUBAAAAAAAAAADYE45YdAPsLd39pSSPrKrHJblHkrsnuXmSy6/j8v1J3p/kb5M8u7s/vmONTsurk3w5F/073p/k+YtpBwAAAAAAAAAAAHaXwCwL0d3fTPLMla9U1YlJrpnkakkuneSYJIcn+XaSryb5eJIPdfdXt3DP1yfZt4W2t6S7H5vksZu47pPZQt/dfX6S4zd7/coaD0ryoK2sAQAAAAAAAAAAAIsiMMtSWAmFfnLBbQAAAAAAAAAAAAATdNiiGwAAAAAAAAAAAACAnSQwCwAAAAAAAAAAAMCkCcwCAAAAAAAAAAAAMGkCswAAAAAAAAAAAABMmsAsAAAAAAAAAAAAAJMmMAsAAAAAAAAAAADApAnMAgAAAAAAAAAAADBpArMAAAAAAAAAAAAATJrALAAAAAAAAAAAAACTJjALAAAAAAAAAAAAwKQJzAIAAAAAAAAAAAAwaQKzAAAAAAAAAAAAAEyawCwAAAAAAAAAAAAAkyYwCwAAAAAAAAAAAMCkCcwCAAAAAAAAAAAAMGkCswAAAAAAAAAAAABMmsAsAAAAAAAAAAAAAJMmMAsAAAAAAAAAAADApAnMAgAAAAAAAAAAADBpArMAAAAAAAAAAAAATJrALAAAAAAAAAAAAACTJjALAAAAAAAAAAAAwKQJzAIAAAAAAAAAAAAwaQKzAAAAAAAAAAAAAEyawCwAAAAAAAAAAAAAkyYwCwAAAAAAAAAAAMCkCcwCAAAAAAAAAAAAMGkCswAAAAAAAAAAAABMmsAsAAAAAAAAAAAAAJMmMAsAAAAAAAAAAADApAnMAgAAAAAAAAAAADBpArMAAAAAAAAAAAAATJrALAAAAAAAAAAAAACTJjALAAAAAAAAAAAAwKQJzAIAAAAAAAAAAAAwaQKzAAAAAAAAAAAAAEyawCwAAAAAAAAAAAAAkyYwCwAAAAAAAAAAAMCkCcwCAAAAAAAAAAAAMGkCswAAAAAAAAAAAABMmsAsAAAAAAAAAAAAAJMmMAsAAAAAAAAAAADApAnMAgAAAAAAAAAAADBpArMAAAAAAAAAAAAATJrALAAAAAAAAAAAAACTJjALAAAAAAAAAAAAwKQJzAIAAAAAAAAAAAAwaQKzAAAAAAAAAAAAAEyawCwAAAAAAAAAAAAAkyYwCwAAAAAAAAAAAMCkCcwCAAAAAAAAAAAAMGkCswAAAAAAAAAAAABMmsAsAAAAAAAAAAAAAJMmMAsAAAAAAAAAAADApAnMAgAAAAAAAAAAADBpArMAAAAAAAAAAAAATJrALAAAAAAAAAAAAACTJjALAAAAAAAAAAAAwKQJzAIAAAAAAAAAAAAwaQKzAAAAAAAAAAAAAEyawCwAAAAAAAAAAAAAkyYwCwAAAAAAAAAAAMCkCcwCAAAAAAAAAAAAMGkCswAAAAAAAAAAAABMmsAsAAAAAAAAAAAAAJMmMAsAAAAAAAAAAADApAnMAgAAAAAAAAAAADBpArMAAAAAAAAAAAAATJrALAAAAAAAAAAAAACTJjALAAAAAAAAAAAAwKQJzAIAAAAAAAAAAAAwaQKzAAAAAAAAAAAAAEyawCwAAAAAAAAAAAAAkyYwCwAAAAAAAAAAAMCkCcwCAAAAAAAAAAAAMGkCswAAAAAAAAAAAABMmsAsAAAAAAAAAAAAAJMmMAsAAAAAAAAAAADApAnMAgAAAAAAAAAAADBpArMAAAAAAAAAAAAATJrALAAAAAAAAAAAAACTJjALAAAAAAAAAAAAwKQJzAIAAAAAAAAAAAAwaQKzAAAAAAAAAAAAAEyawCwAAAAAAAAAAAAAkyYwCwAAAAAAAAAAAMCkCcwCAAAAAAAAAAAAMGkCswAAAAAAAAAAAABMmsAsAAAAAAAAAAAAAJMmMAsAAAAAAAAAAADApAnMAgAAAAAAAAAAADBpArMAAAAAAAAAAAAATJrALAAAAAAAAAAAAACTJjALAAAAAAAAAAAAwKQJzAIAAAAAAAAAAAAwaQKzAAAAAAAAAAAAAEyawCwAAAAAAAAAAAAAkyYwCwAAAAAAAAAAAMCkCcwCAAAAAAAAAAAAMGkCswAAAAAAAAAAAABMmsAsAAAAAAAAAAAAAJMmMAsAAAAAAAAAAADApAnMAgAAAAAAAAAAADBpArMAAAAAAAAAAAAATJrALAAAAAAAAAAAAACTJjALAAAAAAAAAAAAwKQJzAIAAAAAAAAAAAAwaQKzAAAAAAAAAAAAAEyawCwAAAAAAAAAAAAAkyYwCwAAAAAAAAAAAMCkCcwCAAAAAAAAAAAAMGkCswAAAAAAAAAAAABMmsAsAAAAAAAAAAAAAJMmMAsAAAAAAAAAAADApAnMAgAAAAAAAAAAADBpArMAAAAAAAAAAAAATJrALAAAAAAAAAAAAACTJjALAAAAAAAAAAAAwKQJzAIAAAAAAAAAAAAwaQKzAAAAAAAAAAAAAEyawCwAAAAAAAAAAAAAkyYwCwAAAAAAAAAAAMCkCcwCAAAAAAAAAAAAMGkCswAAAAAAAAAAAABMmsAsAAAAAAAAAAAAAJMmMAsAAAAAAAAAAADApAnMAgAAAAAAAAAAADBpArMAAAAAAAAAAAAATJrALAAAAAAAAAAAAACTJjALAAAAAAAAAAAAwKQdsegGgL3pyfd6zKJbAEiSnHPuuYtuAQAAAAAAAACAHSYwCyzEV77yrVx44f5FtwEAAAAAAAAAAMAecNiiGwAAAAAAAAAAAACAnSQwCwAAAAAAAAAAAMCkCcwCAAAAAAAAAAAAMGkCswAAAAAAAAAAAABMmsAsAAAAAAAAAAAAAJMmMAsAAAAAAAAAAADApAnMAgAAAAAAAAAAADBpArMAAAAAAAAAAAAATJrALAAAAAAAAAAAAACTJjALAAAAAAAAAAAAwKQJzAIAAAAAAAAAAAAwaQKzAAAAAAAAAAAAAEyawCwAAAAAAAAAAAAAkyYwCwAAAAAAAAAAAMCkCcwCAAAAAAAAAAAAMGkCswAAAAAAAAAAAABMmsAsAAAAAAAAAAAAAJMmMAsAAAAAAAAAAADApAnMAgAAAAAAAAAAADBpArMAAAAAAAAAAAAATJrALAAAAAAAAAAAAACTJjALAAAAAAAAAAAAwKQJzAIAAAAAAAAAAAAwaQKzAAAAAAAAAAAAAEyawCwAAAAAAAAAAAAAkyYwCwAAAAAAAAAAAMCkCcwCAAAAAAAAAAAAMGkCswAAAAAAAAAAAABMmsAsAAAAAAAAAAAAAJMmMAsAAAAAAAAAAADApAnMAgAAAAAAAAAAADBpArMAAAAAAAAAAAAATJrALAAAAAAAAAAAAACTJjALAAAAAAAAAAAAwKQJzAIAAAAAAAAAAAAwaUcsugFgzzk8SQ47bN+i+wDYMWYcMGVmHDBlZhwwZWYcMHXmHDBlZhwwZWYcsN1WzZXD54/t279//+52A+x1P5jkTYtuAgAAAAAAAAAAgMm6VZI3ry4IzAK77agkN03y+SQXLLgXAAAAAAAAAAAApuPwJFdK8s4k56w+IDALAAAAAAAAAAAAwKQdtugGAAAAAAAAAAAAAGAnCcwCAAAAAAAAAAAAMGkCswAAAAAAAAAAAABMmsAsAAAAAAAAAAAAAJMmMAsAAAAAAAAAAADApAnMAgAAAAAAAAAAADBpArMAAAAAAAAAAAAATJrALAAAAAAAAAAAAACTdsSiGwAWr6quneS0JLdM8n1JLpfkYkm+keTjSd6W5G+6+w0La3JOVd04yb0y6/laSS6bZF+Sryf5tyRvTvKS7n73onoElsMYZ9zBVNVVknSSi6+Ufqq7n7W4joBFG9uMq6rjkvxoZv3eNMkJmX0fd2SSryX5SpIPJXljkld198cX1CqwBEY44y6V5N5JfijJTZIcn+TSSc5O8vnM5turk7y0uz+/oDaBJTG2GbeWqjoqyduT3HBV+XHd/djFdAQs2hjmXFXdPsk/beOSD+7u/7ON6wFLagwzbl5V7Uty2yR3TnKzzH6/eumVw19J8pnMfr/6d939ukX0CCyHZZxxVfXYJI/Z6ft0976dvgewePv279+/6B6ABamqqyX5syR3ySxsupb3JXl4d79+J/s6lKq6fpKnJrnVOi95Q5KHdff7dq4rYBmNccatpapeluTHVpUEZmGPGtuMq6rLJ/nNJA9Kcsl1XnZhkpcneXx3v3eHWgOW0Ahn3CWS/HaS/5bkuHVccl6SFyb59e7+9E72Biyfsc249aqqP0zy8LmywCzsQWOac1X16CRP3MYlBWZh4sY041arqvtk9nPrSeu85F+SPLq7X71jTQFLZ5lnnMAssJ0OW3QDwGJU1V2SfDDJXbO+b3aS2Q4Rr62qx+1YY4dQVT+b5N1Zf1g2SW6d5J1V9XM70xWwjMY449ZSVY/MRcOywB41thlXVXdM8oEkv5T1h2WT2c+rd0/y9qr6HzvRG7B8RjjjfiCzGffwrC8sm8x25Lhfko9U1YN2pjNgGY1txq1XVZ2a5GGL7gNYvBHOuZMXcE9gpEY441JVl6qqv03y/Kw/LJvM5uM/VtXv70hjwNIZ44wD2CyBWdiDquruSf4mySUOcsr+zHb0GrIvyW9V1XZ+6npNVfWLSf4is8f1Drkgs76HXCzJ06rqF3aiN2C5jHHGrWXl099PWnQfwOKNbcZV1b2SvCLJFQ9x2oVJzj/E8YsleUpV/el29gYsnxHOuFtk9gjfqx/itHMPcezYJM+sql/e1saApTS2GbdeVXV8kmdl/b9QBSZqpHPuRrt8P2CkxjjjVp749ObMwm8Hc14O/vvVfUl+par8bgImbowzDmArBGZhj6mq6yZ5TpIj5g59NclvJPm+JEevfJ2U5NeSnDGw1KNXAlw7rqpum+QPBw59OskjklwjyVFJjklygyS/l+QbA+f/UVXdcqf6BBZvjDNuLVV1v8z+TH75CHvc2GbcSpDsuUkOnzu0P8nLk9wryZUyC8QemeRySe6c2Z/xgoElH1pVj9qxhoGFGuGMu2KSv82Bv0i4MLOde+6Q5FLdfVRmM+4GSX4zyZcHlvv9qrrzDrYLLNjYZtwGPSOH/nAUsAeMcc5V1TFJrr0b9wLGbaQz7sgkr0xy/YHDb09yWpITuvvIzPq+YZLfTfLtgfMfVVX33qlegcUa44zbQU9bdAPA7ti3f//BPjAETE1V7cvsh6Cbzh16e5K7d/fnD3Ld8ZmFHX547tCXk1y7u7+23b2uuvcxSTrJVecOvSLJ/bt7KBibqrpGkpckufHcoU5yg+4+b7t7BRZrjDNuLVX1PzN7k+pgYdmf6u5n7V5HwKKMbcZV1eFJ3ptZQGy1M5Lcu7vftMb1N0ry15l9MGq1c5Jcp7s/sV29Aos3thm3cu+XJfmxufJXktyju99wiOuukNl8+8G5Q59NclJ3n72dfQKLN8YZt15V9dAkh3oKwOO6+7G71A6wIGOdc1V1ykqPqz21uz2pDvhPI55xT07yyLnyhUl+NcmTunswJLLy+9W/zywgt9p/JKnu/tZ29woszlhn3FZV1bOTPGCu/IrM/sxDm3kAE2OHWdhb7psDv9n5tyR3Otg3O0nS3V9K8qNJ3jJ36PKZ/WC1kx6ZA8Oyb0lyr4OFZZNkJURxh8z+fKtVkp/d1g6BZTHGGTeoqi5VVS/MbMdsO8sCyfhm3ANyYFj260lus1ZYNkm6+71Jbp3Zm/GrHZXk8dvRILBURjXjqurkHBiWPT/JXQ4Vlk2S7v5iZrtpf2zu0PfkwDfqgWkY1Yxbr6q6XhKP5wWS8c65Gw3U3rkL9wXGZXQzbmW3yIcPHPrZ7n7iwcKyyX/+fvV2Sb40d+jKSca+cyRwoNHNuK2qql/Jge/BvT/JfYRlYe8QmIW9ZegRtv+9u7+61oUru9zcI8k35w49uKoutR3NzVt5XMgvzpXPz2xHxXPWur67v5JZz+fPHXp0VZl/MD2jmnEHU1U/nOQDSTziCFhtbDPugQO1R3d3r3eB7v5MkqGdfX6sqo7edGfAMhrbjPulgdpTu/tt67m4u7+Z5DcHDp22pa6AZTW2GbemqjoqyfMzeyTnd525oHaAxRvrnDt5oCYwC8wb44x7fJLD52p/1t3PWM/FKyG53xk4NPR+HzBuY5xxm1ZVt8/syZ6rfTPJPbv72wtoCVgQgTHYI6rqZkluOFd+W3e/br1rdPcZSZ46Vz4uyU9ssb2D+fEkV5irvaS7/3W9C3T3B5K8ZK58YpLbb601YJmMdMZdRFWdWFUvTvKPOXBn7ST54m70ASyfsc24qjohya3myl9M8uyNrtXdL8uBTww4Lgc+yhwYqRHOuH2Z7RC72v4kf7zBpV6RZH7XilNW1gcmYmwzbgOemIs+TeAL8RQA2JNGPudOnnv9rSQf2eF7AiMyxhlXVVfNgU9E+WyS/7nBpZ6V5Oy52ikrmx0BEzDGGbcVVXV8kufkwJzcT3f3/O8ggIkTmIW9424Dtb/axDrPHKjt1Dc8dxuoLXvPwGLcbaA2inlRVZesqv+T5F+T3HPglAuS/HKS/72TfQBL7W4DtWWecTfOgT9rvqa7z9vkeq8ZqF19k2sBy+duA7VlnnHXTHKJudoHuvvfN7JId5+V5Mtz5WOTLOUOHMCm3W2gtswzbk1Vdccc+ESon86Bj+0F9oa7DdSWfs6tPIHu++fK7+nuC3fqnsAo3W2gtuwz7r45cHfZJ678DLpuK09GecOq0rlJvprke7fWHrBE7jZQW/YZtxXPSHLFudpfdvdfL6IZYLEEZmHvOHWg9vcbXWRld9f5XwTeqqrmf2G4JStvWM3vAntOhgMTa3l9ku/M1eZ3BALGbVQzbs41k/z3JBcbOPbxJLfp7ift4P2B5Te2GXe9gdpWduk5Y6B2/BbWA5bLqGZcd388s50yTsrsFwu/nuRPNrnccQO1+V1ngXEb1Yxby8qTBJ6VZPVu2E/t7g3/mYDJGOucu3ZmH1Za7Z07dC9gvMY44+Z3lz07s+/fNuOnk1wjySW7+6juvlJ3f3grzQFLZYwzblOq6v5J7jJX/lyShy2gHWAJHLHoBoCdV1XH5sDt9D/f3Z/Y5JJvzSzg9V0XS3LTJK/d5HpDKsnl5mrv6e75x3+sqbvPrar3JLnlqvIJVXXNje4CBCyfkc64tZyb5A+T/HZ3f3sX7wssmZHOuL9K8q4kV05ypZX/fesW1hvabfHrW1gPWBIjnXFZ2XnsYytfL9/MGlV1YpKLz5W/3N1nbq07YFmMdcat4ZlJrrDq9UeTPGoX7w8skZHPuZMHagKzwH8a44xbCaedMlf+x83+nNnd/7H1roBlNMYZt1lVdXxmv3Od99+7++u73A6wJARmYW+4Xi6680OSfGAL631ooHaTbO83PNcfqG2151vO1W6SAz/tBIzPGGfcwexP8rwkv7mFH0qBaRndjOvuMzK8K+xmzb9xl8w+/Q2M3+hm3Db6bwO1f9r1LoCdNKkZV1W/lOSOq0rnJblvd88/1QnYO8Y85240UBOYBVYb44w7JQc+YfjN27g+MB1jnHGb9eQcuFHby7r77xbRDLAcBGZhb7juQK23sN6/DdS+dwvrDRljz8BiTGFenJvkRUme1N3v3+F7AeMyhRm3aVV1hSS3mitfmOTtC2gH2H57csZV1Q8l+eWBQ3+2270AO2oyM66qbpDkf82Vf6u737Mb9weW1pjn3Mlzr7/y3afRVdV1ktwrsw04rpPk8pn9HHpGkk8neXWSV3oPDyZvjDNu6MMAvl8Dhoxxxm1YVZ2S5P5z5W8nefjudwMsE4FZ2BuuMlD77BbW+/xA7cQtrDdkjD0DizH2efGJJFfr7i/s4D2A8Rr7jNuqR2T2+KbV3tTdX1xEM8C221MzrqqOSvLQJL+bA2fbs7v7LbvfFbCDJjHjqurozJ6EcvSq8huTPHGn7w0svTHPuZPnXr+rqm6V5Ak58EOb33Wtla/bJnlCVb0qya929/t2qEdgscY4404aqF3kSZtVdXiSU5PcObPdH6+W2fd5X0ryH0nekORvfTAKJm+MM24z/igH7qT7e939qQX0AiwRgVnYG644UNvKY3KHQl3Hb2G9IWPsGViMUc+L7v5Gkm/s1PrA6I16xm1FVZ2U5GEDh56x270AO2ayM66qjsjsUZiXTXKDJD+c5L5Jrjxw+huTPHj3ugN2yVRm3JOTXH/V628kuX93X7gL9waW2yjnXFVdOckV5so3y+x7so24Y5I7VNUvd/cfbEtzwDIZ44y7xkDtP5Kkqg5L8jNJfivJ9wycd5kk105ymySPqarXJnm04CxM1hhn3IZU1T2S3Hyu/PkkT1lAO8CSOWzRDQC7Yuibka2Es84cqF1mC+sNGWPPwGKYF8CU7ckZV1UXS/KsJMfMHfpgkufuekPATpnyjHtNknMyeyP+H5M8OsNh2f+T5NTu/s4u9gbsjtHPuKq6c5JfmCs/pLs/vZP3BUZjrHPu5IHapQ5y7jlrrHV4kqdU1V9sqSNgGY1xxp0wf8/uPqeqjk/yT0n+IsNh2SG3S/LOqhr6MDswfmOcceu28iGBxw4c+m3vwQGJwCzsFfNBg2T4m5b1+tZA7bgtrDdkjD0Di2FeAFO2V2fcnyW5xVzt/CQ/290XLKAfYGdMecZdbY3jZyX58e5+cHefvRsNAbtu1DOuqq6Y5Jlz5ed19/N26p7A6Ix1zp18iGNnZvbo3lslOa67j05yZJJrZrYz48F2of3Zqvr1bewRWLwxzrj5ANzZVXWFJG/ILAC7UYcl+aOq+sMtdwYsmzHOuI24Vy76pJQk+ViS0xfQC7CEBGZhbzhqoHbeFtYbuvZiW1hvyBh7BhbDvACmbM/NuKp6fJKfGzj06939tt3uB9hRk5xxK7tYXGWN0y6e5KVV9fqquu0utAXsvtHOuKral9lu/6tDF59O8pCduB8wWmOdczc6SP01Saq7H9Hdb+7us5Kku8/r7k909zO6+9ZJfjzJNweuf3xVbSaQBiynMc64+XDaBUlekuQ6q2pnZfbBgNtktiPtUZk9DeVHM3uq04UD6z68qn5+m3sFFmuMM24jfmOg9nvdff6udwIspSMW3QCwK44cqG3lm4Hd+IZnjD0Di2FeAFO2p2ZcVT0uw29m/WV3P3G3+wF23FRn3JVW3Xd/Zn+mg/Vx6ySvrao/TvIob9zDpIx5xj0syamrXl+Y5AHdvZVHdALTM9Y5d/JA7RmZPdFkKCh2Ed39N1X1sSRvTnLJVYf2JXlikv+yHU0CCzfGGTcfgLviytd3vTbJfbv7jLnzPp/kFUlesbKb7N8kuercOX9cVW/s7g9vZ8PAwoxxxq1LVd0+B+4u+x9J/moB7QBLyg6zsDfsH6jt28J6hw/UtvvRuGPsGVgM8wKYsj0x46pqX1X9QZLfGjj8sswefQlMz1Rn3HlJ7p3kGkmO7O4jk1w2yZ2SPPsgPT0syekruzoC0zDKGVdVN0zy+3PlJ3X3G7b7XsDojXLOJflvmf2M+YQkz8ssPPHz6wnLfld3f2BlnXk3qaq7bEuXwKKNccYdKpz2t0lOHQjLXkR3vzvJzTML0a52ZJLHbqk7YJmMccat18MHan/U3efudiPA8rLDLOwNQ//Pfyv//Q9de84W1hsyxp6BxTAvgCmb/IyrqiMz283nvgOH/ybJfey4CJM1yRnX3V9M8uK52teSvCrJq6rqyUlemuSkuUsfmOQtSf5iN/oEdtzoZlxVHZNZeGz17mTvSfKb23kfYDJGN+eSZOUDAFv+EEB3v7iq3pHklLlDd0/yyq2uDyzcGGfchRkOrX0yyf3W+/5ad3+uqu6X5DVzh+5ZVd/b3R/bWpvAEhjjjFtTVV0zsw+sr3ZWkqftdi/AcrPDLOwNZw/Ujt7CekPXbvc3PGPsGVgM8wKYsknPuKq6TJJ/yHBY9tlJ7uWT3zBpk55xB9PdH0xyiyT/PnD496vq4rvcErAzxjjjnpLkuqtefyezx/YOPV4TYIxzbrsNhS9O3fUugJ0wxhl3sPUe391nbmSh7n5tDvxwwb4kd9xMY8DSGeOMW48H5MCdcl/Y3d9cQC/AEhOYhb3hqwO1S2xhvaFrv7GF9YaMsWdgMcwLYMomO+Oq6tpJ3pbkNgOHfz/JT3X3oh7bBOyOyc64tXT3l5PcJwc+Au8yK3Vg/EY146rqrkkePFd+VHd/dLvuAUzOqObcDnndQO0qKx8OBcZtjDPuWwO1c5K8YJPr/dVA7fabXAtYLmOccetx/4Ha/931LoClJzALe8OXB2qX3cJ6Q9eesYX1hoyxZ2AxzAtgyiY546rq1CRvT3LtuUPnJ3lwd/9qd8+HyIDpmeSMW6/ufmeSVwwc+pHd7gXYEaOZcVV1pSTPmCv/fXc/dTvWByZrNHNup3T3J5IM7cJ9hd3uBdh2Y5xxXxmovbO7v73J9d48UDtpk2sBy2WMM+6QquoWSa45V/5Qd79tN/sAxuGIRTcA7IrPDNSuuIX1hq79whbWGzLGnoHFMC+AKZvcjKuqX0ryB0kOnzv0jST37u5/3M1+gIWa3IzbhJcl+dG52o0X0Aew/cY0405Ncvm52p2qaqsfYHpMVT1mrvap7j5xi+sCy2FMc24nfTXJCXO1yy2iEWBbjXHGnZHkOnO1z21hvc8O1I7fwnrA8hjjjFvL3Qdqm91hG5g4O8zC3vDxgdrVt7DeiQO1j21hvSFj7BlYDPMCmLLJzLiqOqyq/izJH+fAsOzHk9xcWBb2nMnMuC340EDNLyBhGsw4YOrMuZmhzYmGHosOjMsYZ9xQz5ueR939rSQXzJUvvdn1gKUyxhm3lrsO1F6yyz0AIyEwC3vD0C/gagvrzT86N0k+uIX1hoyxZ2AxzAtgyiYx46rqqMzenPqFgcOvS3JKd39kp/sAls7oZ1xV7dviEl8fqB25xTWB5TD6GQewhj0/51a+F7z0wKGhx6ID4zLGGffRgdqlNrtYVR2TAz/07gMBMA1jnHEHVVUn5cD+P9TdQ3MRQGAW9oLu/kwO3PL+5Kra7Ay46UDtfZtc62DekwM/tXiTLaw33/P5Gf5GEBiZkc44gHWZwoxbeXP9bzP8SKSnJblDd391J3sAltMYZ1xV3aOq/qSqXllVH05yVlXdbAtLXnagNoZHDwNrGNmMOy/JWVv8OmcD6wITMLI5lySpql+sqtOr6uVV9Zaq+lhVDe1Gtu4lc2CY7Kz4fg5Gb4wzLsnbB2rX2cJ6JwzUztjCesCSGOmMO5TbD9Revov3B0Zm6DEhwDS9Kck9V70+LrMA6js3skhVHZsDv+H5bHf31tq7qO4+q6rem+S/rCp/X1Wd0N0berOpqq6R5Kpz5bd1tzfoYTpGNeMANmi0M66qjkzysiR3mDt0YZJHd/cf7NS9gdEY24y7XZKHDNTetsn1rjtQ+9wm1wKWzyhmXHc/N8lzt7JGVT0oyTPnyr/b3Y/dyrrA0hvFnFvlh3LRfpPkVklescn1bj1Qe2t3n7/J9YDlMrYZ964k30lyzKradavqit29maDrKQO1pd75G9iQsc24Q/mhgdprd/H+wMjYYRb2jr8bqP34Jta5U5Kj52r/sIl11mO+531J7raJde4xUNupnoHFGOOMA1ivUc64lUdT/mUODMuem+Q+wrLAirHNuH8ZqJ22hfV+dKD2ui2sByyXsc04gI0a25wb2unsbltY76cHaq/ewnrAchnVjOvus5P801x5X5L7bXLJOw3UXr/JtYDlM6oZt4b5wOy5Sd6yyz0AIyIwC3vHy5OcPVf7mao6boPrPHyg9uxNdbS2FwzUHrqRRwGs7Go2v/vPhUmes5XGgKUzxhkHsF5jnXG/muQn5mrnJbl7d794B+8LjMvYZtyrMvuZcrXrV9XtNrpQVd0ww4HZl22iL2A5jW3GAWzU2Obc0E6yJ1XVqRtdqKrumAN3Xzw7yTM20xiwlMY245LZh9fnPWJlB8h1q6orJbnPXPnCeMQ5TMkYZ9wBquqqSa48V/5od39nt3oAxkdgFvaI7v5akhfOlY9P8sT1rlFVP5vklnPlD3b3m7bY3qDu/miSN8yVr5/kURtY5jFJrjFX+/vu/tRWegOWyxhnHMB6jXHGVdVNk/z2wKGf7e6/34l7AuM0thnX3Z/N8C4Zf7Tygc11qaqLZ/bLg31zh97c3Rt69B2wvMY24wA2amxzrrvfl+FdZp9SVccM1AdV1QkZDsY+p7u/tNn+gOUythm34mVJPjlXu3KSJ29wnb9IctRc7eUrPxMDEzDSGTfk5IHaB3bx/sAICczC3vKEJOfP1R5cVWsGUFc+Yf3nA4d+fTsaO4THDtR+r6ruvdaFVfXTSX5trnxhkt/chr6A5TPGGQewXqOZcStPA/iLJIfPHTq9u+2EBgwZzYxb8bgcuMvsDZI8p6qOWOviqrpMZrt43HDu0P4kj9iWDoFlMrYZB7BRY5tzjxuoXS/J86pqPhx2gJVdzF6X5Ipzh76YA38fAYzfqGZcd1+Q2VOf5j24qn5jreural9V/WGSu8wduiDJ47ehRWC5jGrGHcT8+2tJ8pFd7gEYmX379+9fdA/ALqqq/5XklwcOPSvJo7v7y3PnH53kkZkFV+d/8feq7r7TOu75oCTPHDh0je7+5Dquf2GS+YDs/iRPSvL47v7W3PmXzOyHtl8aWO5/d/dD1ronME5jnHHrVVWPzWzX7NV+qruftV33AJbbWGZcVZ2W5Hlrrb1Fj+vux+7wPYBdNJYZt+raP03y0IFDb03y0O5+z8A1+5LcM8n/yoFPQkmSx3T30O7cwMiNbcZtxkHu53s22CPGNueq6hU5MAyWJO/J7Hu5tw5cc0SSByX53cx2X5t39+5+2SGbBkZpbDNu5fqXJLnHwKFXrvT80YFrrpnkT5MM9fdH3e0DnjBBY5xxc2s9J8n95sp+fwoc0pq7XgCT85uZbYs/vzX+g5KcVlV/n+SjSc5Ncs3M3jS6zMA6n0nygJ1r8yJ+PrOt9K+9qrYvs2/cHrLy5tbHVmrXzqznYwfWeW+S/7GjnQKLNsYZB7BeY5lxD9/BtYHpGsuM+65HZvbz5x3m6jdP8u6qeneSf07yhSTHJDkxye1z4E5k3/UMYVmYtLHNOICNGtuce0CSN2T2lIDVbpzkLVX1/iRvTHJGZr9ruEaSU5Nc9iDr/YawLEza2GZckvxUkqsluelc/S5J7lxVb0vyliRfyWy2/ZckP5ThJxS/OcNhOmAaxjjjVjtxoPb53W4CGBeBWdhjuvvcqrpjklflwG96jkpy93Us85kkt5//NNFO6e6vVdVtMnvMUc0dPi7JaetY5gNJ7tjdZ29ze8ASGeOMA1ivMcy4qvqeJKfsxNrAtI1hxq220u+PJ3lxkjsOnHKTla+17E/yO5ntygFM1NhmHMBGjW3OrfzO4Q5J/i6zkOy871/5WsuFSX6lu5+8nf0By2VsMy5JuvvMlTn3kiT/de7wvsw+7HnzdSz1xiR37e7ztrlFYEmMccbNuepA7Su73gUwKkOfEAImrrvPTPIjSf53Zm/obMSrk9yiu/912xs7hO7+fGafbHzRJi5/QZJbd/cXtrcrYBmNccYBrNcIZtwP7eDawMSNYMZdRHefleTOSR6V5NubWOJfk9y5u3+ruzf65wVGZmwzDmCjxjbnuvuMzMJiT0ly/iaW+NcktxOWhb1hbDMuSbr765n1/OtJNrqh0PlJfi+zANw3t7k1YMmMccatcqmBmk3UgEMSmIU9qru/1d0PyexRHM9P8o1DnL4/yeuT3LO7f7i7P7sLLR6gu7/Y3T+R2SchX57kO4c4/fwkr8jsB7nTuvtru9EjsBzGOOMA1mvJZ9yVd3h9YOKWfMYdoLv3d/dTklw9yW8n+fc1LjknyWuT3C/Jdbv7VTvcIrBExjbjADZqbHOuu8/t7kcluVaSP0yyVg8XJHlTkgcmuV53v2GHWwSWyNhmXJJ09/nd/buZPUb997P2nPtyZoG56u5fs7Ms7B1jnHErLj5QE5gFDmnf/v37F90DsASq6mJJbpbZlvVXzGx7/a8l+VSSt658CnGpVNWxmfV8lSQnJDk8s54/luTtK7v9AIxyxgGslxkHTNkYZ1xVXS2zXy4cn+SymX3Y80tJPpfkbd19qA9/AnvIGGccwEaMcc5V1TWT3DjJ5TP7Xu7C/P+/d3j3MvYMLMYYZ1ySVNX3JblBZn0fm1mw7ItJPpDk/d19wQLbA5bEWGccwHoIzAIAAAAAAAAAAAAwaYctugEAAAAAAAAAAAAA2EkCswAAAAAAAAAAAABMmsAsAAAAAAAAAAAAAJMmMAsAAAAAAAAAAADApAnMAgAAAAAAAAAAADBpArMAAAAAAAAAAAAATJrALAAAAAAAAAAAAACTJjALAAAAAAAAAAAAwKQJzAIAAAAAAAAAAAAwaQKzAAAAAAAAAAAAAEyawCwAAAAAAAAAAAAAkyYwCwAAAAAAAAAAAMCkCcwCAAAAAAAAAAAAMGlHLLoBAAAAAAD2nqp6YpJHDxz6i+7+77vUw8OS/NHAoZ/s7ufvRg9Dquo2SV43X+/ufbvfDQAAAABMgx1mAQAAAABYhGccpH6vqjpyl3p44EDtq0leukv3BwAAAAB2icAsAAAAAAC7rrs/muSfBw5dJsmddvr+VXWDJDcaOPRX3X3OTt8fAAAAANhdArMAAAAAACzK0w9Sv+8u3Htod9kkOX0X7g0AAAAA7DKBWQAAAAAAFuVFSc4cqN+lqi65UzetqsMzHMp9R3d/YKfuCwAAAAAsjsAsAAAAAAAL0d1nJXnhwKGjk9xzB299apIrDtTtLgsAAAAAEyUwCwAAAADAIj39IPWhHWC3ywMHamclecEO3hMAAAAAWCCBWQAAAAAAFqa735bkwwOHblNVV9nu+1XVpZP86MChF3b3mdt9PwAAAABgOQjMAgAAAACwaEO7zB6W5LQduNdPJDl6oH76DtwLAAAAAFgSArMAAAAAACzac5KcN1C/7w7c64EDtQ9191t34F4AAAAAwJIQmAUAAAAAYKG6+0tJ/nbg0MlVdd3tuk9VXTvJzQcO2V0WAAAAACbuiEU3AAAAAAAASZ6e5B4D9fsl+bVtuscDBmrnZLbD7bpU1eFJKsn3JrlykkskOTqzHXK/neSMJJ9I8oHuPnurDU9RVV0iyY2TXCvJ5ZIcleQ7ST6fpJO8r7vPX1yHAAAAAEzRvv379y+6BwAAAAAA9riqOizJp5J8z9yhTyW5Rndv6c3sqtqX5JNJrjZ36IXdfZ81rr1EktOS3C3JrZMcu45bnpfkTUmem+R5GwnPVtVtkrxuvt7d+9a47pNJrj5X/qnuftZ6771T61XV0Unun1lo+eZJDj/E6d9K8vdJntbdr93ovQAAAABgyGGLbgAAAAAAALr7wiTPGjh09SQ/uA23uG0ODMsmyekHu6Cqjqmq30ryuSRPS3LHrC8smyQXS3K7zHbO/beq+vGNtTsdVfWTSf49yV9k9s/yUGHZJDkuyb2TvKaq3lBV19vhFgEAAADYAwRmAQAAAABYFs9IMrST7H23Ye0HDtQ+keQ1QydX1bWSvCPJ45JcYov3/p4kf11Vj97iOqNSVcdW1fMy22X3Sptc5oeSvKuqhv75AQAAAMC6CcwCAAAAALAUuvsTSV47cOheVXXkZtetquOS3GPg0DO6+4CAblVdJcnrklz/EMtemOTLST6V5IwkZ6+jld+rqpus47zRW/k7/39JTjvEafuTfC3JZ5J85xDnHZ3kWVX18G1rEAAAAIA9R2AWAAAAAIBl8vSB2mWT3HELa94jycXnahckeeZBzn92kqsO1L+R5HeTnJLkuO4+vrtP7O4rJTk2yUlJfjbJuw+y7uFJnrzB3sfqaUluNVC/MMmLktw5yaW6+7LdfbXuPjbJVXLov7+nVNXtd6RbAAAAACbviEU3AAAAAAAAq/xNZruOXmaufr8kL9/kmg8cqP2/7v7cfLGq7pLkvw6c/54kd+ruLwzdYGWn2o8l+VhVPT3JLyb5oyT75k69dVVdqbs/v4H+R6Wqfi7JTw4c+miS+3T3+4au6+7/SHL6yt/fzyT58ySrdxY+LMnzqupGQ//sAAAAAOBQ7DALAAAAAMDS6O6zkzx34NBdquqSG12vqq6e5DYDh04/yCX/Y6D21SR3PVhYdl537+/uP8nwbrL7kpy6nnXGqKoul+QPBg69J8kpBwvLrrby93d6Zn9P35k7fHyS39hyowAAAADsOQKzAAAAAAAsm6cP1I5Oco9NrHX/HLjL6xlJXjl/YlVdKcPh2t9f2f10o/7wIPUTN7HWWDwkycXnat9IcufuPnMjC3X36zMcYH5QVV1hc+0BAAAAsFcJzAIAAAAAsFS6+18y25F03v02sdwDBmrP7u7zB+o/mAPDtecm+b+buG+6+/NJPjlw6DKbWW/ZVdVRSX5h4NATu/uMTS77f5N8Yq52dJKf3+R6AAAAAOxRArMAAAAAACyjoV1mb1NVV17vAlV1iyQnzZX3Jzl96PzufnFmu6PeIMmPJXlEkkd399fXe88BXxioHb2F9ZbZDyY5YaD+3M0u2N0XJHnBwKHbb3ZNAAAAAPamIxbdAAAAAAAADHhekqfkouHSw5L8ZJInr3ONBw7U3tDdHzvYBd397SQfXPnaDmcP1Kb63vwPDtQ+092f2uK6r03yq3O1U6rqqO4+Z4trAwAAALBH2GEWAAAAAICls7Kr618PHLrveq6vqqOT3Hvg0ODustupqvZV1XWr6pFJvm/glKm+N3+rgdrQDrsb9ZGB2lFJbrwNawMAAACwR0z1U+wAAAAAAIzf03NgQPbkqrpud394jWt/LMml52pfy3AId9Oq6gpJrplZMPZ6Sb4/yU2TXOYQl+3bzh6WyHUHav+lqvbv0P2uuEPrAgAAADBBArMAAAAAACyr1yf598wCqavdL8mvrXHtAwdqz+3uszfTSFVdN8nNklw/yUlJrpHkxCQX38x6E3XZid8PAAAAgBETmAUAAAAAYCl19/6qekaS35k79JNV9evdPbhzaVVdMckdBg6dvpH7V9Xlkzw0s11uv3cj1+41VXVskqN2+baX2+X7AQAAADBihy26AQAAAAAAOIRnJblgrnb1JLc8xDX3S3L4XO1d3f2+9dywqvZV1SOTfDzJY7K5sOznkzw1yUc3ce0YXXoB9zxmAfcEAAAAYKTsMAsAAAAAwNLq7s9V1T8kudPcofslefNBLnvgQG1du8tW1ZFJnpnkJ9fdZPLtJB9J8t4k70ryxu7+yMp6r0/yfRtYa6zOOUj9b5O8eIfu+f4dWhcAAACACRKYBQAAAABg2Z2eAwOz96qqX+zu81YXq+rGSa4/d+5ZSZ6/zns9KYcOy344yduSvC+z3WM7yae7e/9Bzh/zk96O2sC5X0uyP8m+ufqnuvuvtq8lAAAAANgcgVkAAAAAAJbdK5N8MckVVtUum1mI9uVz5w7tLvvi7v7mWjepqv+a5BcHDu1P8qwkj+/uT6yn4VWO2eD52+3wLVx77HpP7O4Lq+obSS49d+iyW7g/AAAAAGybMX+yHQAAAACAPWBlF9m/HDh0kZ1gq+qIJKcNnHf6Om/1Wzlwh9QkeXB3//QmwrJJcsImrtmsoV1uj9zMQlV1TJJLbvCyzwzU5nf7BQAAAICFEJgFAAAAAGAMnj5Qu0tVXXzV6zskOX7unI909z+vtXhVnZDkVgOH/qG7n7b+Ni+y5nFJrjxwaKfemz9/oHbxgdp6XGsT17x5oHaDqtpo8PYiqurSVfUjVXWdqlr3rrcAAAAAsJrALAAAAAAAS6+7P5rkLXPlY5PcZdXrn8yB1ru77I0yvLvsC9Z5/ZBbJzl8oD5U2w5nDdQuv8m1fmAT17xpoHZYkntusofvenCSVyX5cJKzquqLVfWOqrJ7LQAAAADrJjALAAAAAMBYDO0y+xNJsrLz6N3mjp2b5C/XufblDlL//DqvH/LQg9SP2MKah/K1gdpJm1xrMyHX1yQ5b6D+8KoaCiOvaeWf6y/MlY/PbAfc3syaAAAAAOxNArMAAAAAAIzFi5J8a652alUdneTUJBefO/by7v7yOtc+8yD1a26gv/9UVacl+ZGDHD5qM2uuwycHardb+ftZt6o6JQfv/aC6+4tJnjdw6AZJfmWj6634tSRXGai/qLuHwrkAAAAAMEhgFgAAAACAUejubyV54Vz52CT/NcmPDVxy+gaW//RB6v+tqjb0XnpVnZrh3XC/65iNrLcBbx6oXTrJI9a7QFWdkOHQ63o9Kcn+gfrjq2ron9GhevmRJL86cOjcJP9rE70BAAAAsIcJzAIAAAAAMCZDQdQfS3Lnudqnkrx6A+u+L8kZA/UbJ/nj9YRmq+roqvrtJK/MoUOxl9hAXxvxyiQXDtQfV1V3X+viqvqBJP+c5FqbbaC7P5TkzwcOHZHkpVX1y1V1xDp6uW+Sl2X49xh/3t2f3GyPAAAAAOxN+/bvH/qgNwAAAAAALKeq+nCS66wqnZvkyLnTHtPdv73BdZ+Q5NcOcvitSZ6c5DXd/Y1V1xyT5OQkd03yM0musI5b/Vt3X/sQfdwmyevm6929b62Fq+pFSe41cGh/kudkFmZ9d3dfsHL+xZP8UJKfTnKPJKvv8bEk3zu3zk9197PW6OHIJG9KcspBTukkf5LkFd39mVXXXTrJbZP8Qma7Bg95b5JbdPfZh+oBAAAAAOYJzAIAAAAAMCpV9cjMwqsHc2GSq3f3Zze47nFJPpLkew5x2v4kX05yVpJjk1w+B3+a2zczC6j+6lz93CSXOljoc4uB2Wsm+XCSow5x2neSfCnJxZKckOH+35JZqPUFc/U1A7MrfVw1yRuSXGONU7/byzGZ/V0e6s/4qSS37e5PrHV/AAAAAJi35iOkAAAAAABgyTwnyXmHOP4PGw3LJkl3fyvJnZJ8/RCn7UtyfJITM9tN9mDvs78jyU2T/F5mIdvVjkxyq432tx7d/e9JHpBD//0ck+RqSa6U4f7fnOSOSc7ZQh+fSXLLzIK3h/LdXo7PocOyH0hyS2FZAAAAADZLYBYAAAAAgFHp7i8mecUhTjl9C2t/IMnNk7x3k0t8OckvJblFd/9rd5+Z5F8GzrvXJtdfU3e/KMntknxsg5een+QpSW7X3d/chj4+n+TWSR6V5BubXObcJI9PctPu/txWewIAAABg7xKYBQAAAABgjJ5+kPoXcugw7Zq6+6NJfiDJA5O8PQfuEDvvzCT/L8nPJTmxu/+0uy9YdfwvB655QFVdcSt9Hkp3vznJdZP8TNb+M3wzyTOS3LC7H9Xdh9qddqN9nN/dT8lsF9mHJXlrkgvXcelHk/xGkmt0929196Z3uwUAAACAJNm3f/9a7/MBAAAAAMDeVVWXTnJKZqHPSyc5MslXM9tN9pNJ3jsXkF06VXWFzP4M10xyySTnJflSkg8nefd2hmTX0ctxSW6U5FpJLp/kmCTfTvL1JJ9L8s7u/spu9QMAAADA3iAwCwAAAAAAAAAAAMCkHbboBgAAAAAAAAAAAABgJwnMAgAAAAAAAAAAADBpArMAAAAAAAAAAAAATJrALAAAAAAAAAAAAACTJjALAAAAAAAAAAAAwKQJzAIAAAAAAAAAAAAwaQKzAAAAAAAAAAAAAEyawCwAAAAAAAAAAAAAkyYwCwAAAAAAAAAAAMCkCcwCAAAAAAAAAAAAMGkCswAAAAAAAAAAAABMmsAsAAAAAAAAAAAAAJMmMAsAAAAAAAAAAADApAnMAgAAAAAAAAAAADBpArMAAAAAAAAAAAAATJrALAAAAAAAAAAAAACT9v8BIE4MKOCKGIMAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 2880x1440 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plotImp(model, FEATURES)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "af2f007a",
   "metadata": {},
   "outputs": [],
   "source": [
    "final_test = test2_mod.merge(test_grouped)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "3aee7e0d",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "ccff87bb",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.7/site-packages/ipykernel_launcher.py:28: RuntimeWarning: invalid value encountered in true_divide\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "MultVAE 0.08779497629282285\n",
      "EASE 0.15022132572163718\n",
      "SLIM 0.04456921555200555\n",
      "boosting 0.15176205847625243\n"
     ]
    }
   ],
   "source": [
    "for el in models_keys + ['boosting']:\n",
    "    metric = NDCG_binary_at_k_batch(np.vstack((final_test['preds_'+el].values)),\n",
    "                                    np.vstack((final_test.target_basket_y.values)), 20)\n",
    "    print(el, metric)\n",
    "    "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e4d0b61a",
   "metadata": {},
   "source": [
    "## Testing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "ef73ffe5",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "ec74fcb0",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(1081420, 7)\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 th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>buyer_id</th>\n",
       "      <th>pav_order_id</th>\n",
       "      <th>created</th>\n",
       "      <th>item_id</th>\n",
       "      <th>count</th>\n",
       "      <th>price_sold</th>\n",
       "      <th>flag_weight_goods</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>94640077</td>\n",
       "      <td>98519243164</td>\n",
       "      <td>2021-08-30 17:56:31</td>\n",
       "      <td>203053459</td>\n",
       "      <td>1.00</td>\n",
       "      <td>67.62</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>95865222</td>\n",
       "      <td>98512083628</td>\n",
       "      <td>2021-07-26 16:17:21</td>\n",
       "      <td>202967705</td>\n",
       "      <td>1.14</td>\n",
       "      <td>406.80</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>95147155</td>\n",
       "      <td>98519972197</td>\n",
       "      <td>2021-09-02 21:54:18</td>\n",
       "      <td>203551512</td>\n",
       "      <td>1.00</td>\n",
       "      <td>52.77</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>94832207</td>\n",
       "      <td>98518646272</td>\n",
       "      <td>2021-08-28 10:43:23</td>\n",
       "      <td>202801712</td>\n",
       "      <td>1.00</td>\n",
       "      <td>92.89</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>95483101</td>\n",
       "      <td>98510857920</td>\n",
       "      <td>2021-07-20 14:27:08</td>\n",
       "      <td>203416702</td>\n",
       "      <td>2.00</td>\n",
       "      <td>238.99</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   buyer_id  pav_order_id              created    item_id  count  price_sold  \\\n",
       "0  94640077   98519243164  2021-08-30 17:56:31  203053459   1.00       67.62   \n",
       "1  95865222   98512083628  2021-07-26 16:17:21  202967705   1.14      406.80   \n",
       "2  95147155   98519972197  2021-09-02 21:54:18  203551512   1.00       52.77   \n",
       "3  94832207   98518646272  2021-08-28 10:43:23  202801712   1.00       92.89   \n",
       "4  95483101   98510857920  2021-07-20 14:27:08  203416702   2.00      238.99   \n",
       "\n",
       "   flag_weight_goods  \n",
       "0              False  \n",
       "1               True  \n",
       "2              False  \n",
       "3              False  \n",
       "4              False  "
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test_data = pd.read_csv(path+'test.csv')\n",
    "print(test_data.shape)\n",
    "test_data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "ed95c405",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>pav_order_id</th>\n",
       "      <th>target_basket_x</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>4620121489</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>4620121505</td>\n",
       "      <td>[0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>4620121594</td>\n",
       "      <td>[0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 1.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4620121684</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>4620121902</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   pav_order_id                                    target_basket_x\n",
       "0    4620121489  [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...\n",
       "1    4620121505  [0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, ...\n",
       "2    4620121594  [0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 1.0, 0.0, 0.0, ...\n",
       "3    4620121684  [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...\n",
       "4    4620121902  [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ..."
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "def get_preds(x):\n",
    "    \n",
    "    basket = [item2idx[t] for t in x if t in item2idx]\n",
    "    vector = np.zeros(len(item2idx))\n",
    "    vector[basket] = 1\n",
    "    \n",
    "    return vector\n",
    "\n",
    "def get_final_recs(x):\n",
    "    vector = np.argsort(-x)[:100]\n",
    "    vector = [idx2item[t] for t in vector if t in idx2item][:20]\n",
    "    \n",
    "    return vector\n",
    "\n",
    "pred = test_data.groupby(['pav_order_id'])['item_id'].agg([('basket', list)]).reset_index()\n",
    "pred.rename({'basket':'target_basket_x'}, axis=1, inplace=True)\n",
    "pred['target_basket_x'] = pred.target_basket_x.apply(lambda x: get_preds(x))\n",
    "#pred['target_basket_x'] = pred.target_basket_x.apply(lambda x: [0]+[item2idx[t] for t in x if t in item2idx])\n",
    "pred.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "b5f9d8f7",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|          | 0/10 [00:00<?, ?it/s]\n",
      "  0%|          | 0/1 [00:00<?, ?it/s]\u001b[A\n",
      "\n",
      "Reading dataset in memory: 0it [00:00, ?it/s]\u001b[A\u001b[A\n",
      "\n",
      "Reading dataset in memory: 472it [00:00, 4692.60it/s]\u001b[A\u001b[A\n",
      "\n",
      "Reading dataset in memory: 942it [00:00, 4586.25it/s]\u001b[A\u001b[A\n",
      "\n",
      "Reading dataset in memory: 1401it [00:00, 4556.39it/s]\u001b[A\u001b[A\n",
      "\n",
      "Reading dataset in memory: 1885it [00:00, 4663.98it/s]\u001b[A\u001b[A\n",
      "\n",
      "Reading dataset in memory: 2362it [00:00, 4701.00it/s]\u001b[A\u001b[A\n",
      "\n",
      "Reading dataset in memory: 2840it [00:00, 4725.05it/s]\u001b[A\u001b[A\n",
      "\n",
      "Reading dataset in memory: 3332it [00:00, 4786.59it/s]\u001b[A\u001b[A\n",
      "\n",
      "Reading dataset in memory: 3868it [00:00, 4967.78it/s]\u001b[A\u001b[A\n",
      "\n",
      "Reading dataset in memory: 4399it [00:00, 5073.75it/s]\u001b[A\u001b[A\n",
      "\n",
      "Reading dataset in memory: 4907it [00:01, 5055.58it/s]\u001b[A\u001b[A\n",
      "\n",
      "Reading dataset in memory: 5413it [00:01, 5005.91it/s]\u001b[A\u001b[A\n",
      "\n",
      "Reading dataset in memory: 5980it [00:01, 5204.38it/s]\u001b[A\u001b[A\n",
      "\n",
      "Reading dataset in memory: 6637it [00:01, 5613.12it/s]\u001b[A\u001b[A\n",
      "\n",
      "Reading dataset in memory: 7199it [00:01, 5607.15it/s]\u001b[A\u001b[A\n",
      "\n",
      "Reading dataset in memory: 7896it [00:01, 5179.22it/s]\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Predict with MultVAE:   0%|          | 0/62 [00:00<?, ?it/s]\u001b[A\u001b[A\n",
      "\n",
      "Predict with MultVAE:   2%|▏         | 1/62 [00:00<00:10,  5.95it/s]\u001b[A\u001b[A\n",
      "\n",
      "Predict with MultVAE:   8%|▊         | 5/62 [00:00<00:02, 20.73it/s]\u001b[A\u001b[A\n",
      "\n",
      "Predict with MultVAE:  15%|█▍        | 9/62 [00:00<00:01, 27.00it/s]\u001b[A\u001b[A\n",
      "\n",
      "Predict with MultVAE:  19%|█▉        | 12/62 [00:00<00:01, 27.55it/s]\u001b[A\u001b[A\n",
      "\n",
      "Predict with MultVAE:  26%|██▌       | 16/62 [00:00<00:01, 29.46it/s]\u001b[A\u001b[A\n",
      "\n",
      "Predict with MultVAE:  32%|███▏      | 20/62 [00:00<00:01, 31.15it/s]\u001b[A\u001b[A\n",
      "\n",
      "Predict with MultVAE:  39%|███▊      | 24/62 [00:00<00:01, 31.55it/s]\u001b[A\u001b[A\n",
      "\n",
      "Predict with MultVAE:  45%|████▌     | 28/62 [00:00<00:01, 31.26it/s]\u001b[A\u001b[A\n",
      "\n",
      "Predict with MultVAE:  52%|█████▏    | 32/62 [00:01<00:00, 30.94it/s]\u001b[A\u001b[A\n",
      "\n",
      "Predict with MultVAE:  58%|█████▊    | 36/62 [00:01<00:00, 32.20it/s]\u001b[A\u001b[A\n",
      "\n",
      "Predict with MultVAE:  65%|██████▍   | 40/62 [00:01<00:00, 32.97it/s]\u001b[A\u001b[A\n",
      "\n",
      "Predict with MultVAE:  71%|███████   | 44/62 [00:01<00:00, 34.41it/s]\u001b[A\u001b[A\n",
      "\n",
      "Predict with MultVAE:  77%|███████▋  | 48/62 [00:01<00:00, 34.86it/s]\u001b[A\u001b[A\n",
      "\n",
      "Predict with MultVAE:  84%|████████▍ | 52/62 [00:01<00:00, 34.24it/s]\u001b[A\u001b[A\n",
      "\n",
      "Predict with MultVAE:  90%|█████████ | 56/62 [00:01<00:00, 34.30it/s]\u001b[A\u001b[A\n",
      "\n",
      "Predict with MultVAE: 100%|██████████| 62/62 [00:01<00:00, 31.54it/s]\u001b[A\u001b[A\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "all models predicted\n",
      "MultVAE\n",
      "EASE\n",
      "SLIM\n",
      "we got all scores\n",
      "joining scores\n",
      "exploding\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "  0%|          | 0/240750 [00:00<?, ?it/s]\u001b[A\u001b[A\n",
      "\n",
      "  0%|          | 1/240750 [01:05<4381:31:29, 65.52s/it]\u001b[A\u001b[A\n",
      "\n",
      "  5%|▌         | 12038/240750 [01:05<14:34, 261.55it/s]\u001b[A\u001b[A\n",
      "\n",
      " 10%|█         | 24075/240750 [02:12<17:30, 206.34it/s]\u001b[A\u001b[A\n",
      "\n",
      " 10%|█         | 24075/240750 [02:13<17:30, 206.34it/s]\u001b[A\u001b[A\n",
      "\n",
      " 10%|█         | 24093/240750 [03:21<34:49, 103.68it/s]\u001b[A\u001b[A\n",
      "\n",
      " 10%|█         | 24093/240750 [03:21<34:49, 103.68it/s]\u001b[A\u001b[A\n",
      "\n",
      " 10%|█         | 24094/240750 [03:21<34:56, 103.35it/s]\u001b[A\u001b[A\n",
      "\n",
      " 11%|█▏        | 27165/240750 [04:21<44:54, 79.28it/s] \u001b[A\u001b[A\n",
      "\n",
      " 11%|█▏        | 27165/240750 [04:21<44:54, 79.28it/s]\u001b[A\u001b[A\n",
      "\n",
      " 14%|█▎        | 33075/240750 [04:21<23:48, 145.39it/s]\u001b[A\u001b[A\n",
      "\n",
      " 16%|█▌        | 37442/240750 [05:17<15:46, 214.88it/s]\u001b[A\u001b[A\n",
      "\n",
      " 17%|█▋        | 41610/240750 [05:17<24:20, 136.31it/s]\u001b[A\u001b[A\n",
      "\n",
      " 17%|█▋        | 41610/240750 [05:17<24:20, 136.31it/s]\u001b[A\u001b[A\n",
      "\n",
      " 24%|██▍       | 57499/240750 [05:17<08:24, 363.54it/s]\u001b[A\u001b[A\n",
      "\n",
      " 55%|█████▍    | 132408/240750 [05:17<00:57, 1890.36it/s]\u001b[A\u001b[A\n",
      "\n",
      " 60%|█████▉    | 144445/240750 [06:20<02:01, 791.67it/s] \u001b[A\u001b[A\n",
      "\n",
      " 60%|█████▉    | 144445/240750 [06:20<02:01, 791.67it/s]\u001b[A\u001b[A\n",
      "\n",
      " 75%|███████▍  | 180556/240750 [06:20<00:45, 1320.32it/s]\u001b[A\u001b[A\n",
      "\n",
      "100%|██████████| 240750/240750 [06:20<00:00, 632.51it/s] \u001b[A\u001b[A\n",
      "\n",
      "100%|██████████| 1/1 [08:01<00:00, 481.96s/it]\u001b[A\n",
      " 10%|█         | 1/10 [08:20<1:15:07, 500.80s/it]\n",
      "  0%|          | 0/1 [00:00<?, ?it/s]\u001b[A\n",
      "\n",
      "Reading dataset in memory: 0it [00:00, ?it/s]\u001b[A\u001b[A\n",
      "\n",
      "Reading dataset in memory: 555it [00:00, 5547.56it/s]\u001b[A\u001b[A\n",
      "\n",
      "Reading dataset in memory: 967it [00:42, 18.71it/s]  \u001b[A\u001b[A\n",
      "\n",
      "Reading dataset in memory: 967it [00:42, 18.71it/s]\u001b[A\u001b[A\n",
      "\n",
      "Reading dataset in memory: 1389it [00:42, 32.69it/s]\u001b[A\u001b[A\n",
      "\n",
      "Reading dataset in memory: 1866it [00:43, 55.06it/s]\u001b[A\u001b[A\n",
      "\n",
      "Reading dataset in memory: 2330it [00:43, 85.82it/s]\u001b[A\u001b[A\n",
      "\n",
      "Reading dataset in memory: 2954it [00:43, 144.23it/s]\u001b[A\u001b[A\n",
      "\n",
      "Reading dataset in memory: 3527it [00:43, 218.96it/s]\u001b[A\u001b[A\n",
      "\n",
      "Reading dataset in memory: 4061it [00:43, 314.70it/s]\u001b[A\u001b[A\n",
      "\n",
      "Reading dataset in memory: 4624it [00:43, 453.26it/s]\u001b[A\u001b[A\n",
      "\n",
      "Reading dataset in memory: 5252it [00:43, 662.44it/s]\u001b[A\u001b[A\n",
      "\n",
      "Reading dataset in memory: 5818it [00:43, 893.88it/s]\u001b[A\u001b[A\n",
      "\n",
      "Reading dataset in memory: 6346it [00:43, 1155.80it/s]\u001b[A\u001b[A\n",
      "\n",
      "Reading dataset in memory: 6839it [00:44, 1449.80it/s]\u001b[A\u001b[A\n",
      "\n",
      "Reading dataset in memory: 7312it [00:44, 1771.58it/s]\u001b[A\u001b[A\n",
      "\n",
      "Reading dataset in memory: 8002it [00:44, 180.88it/s] \u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Predict with MultVAE:   0%|          | 0/63 [00:00<?, ?it/s]\u001b[A\u001b[A\n",
      "\n",
      "Predict with MultVAE:   2%|▏         | 1/63 [00:00<00:09,  6.27it/s]\u001b[A\u001b[A\n",
      "\n",
      "Predict with MultVAE:   8%|▊         | 5/63 [00:00<00:02, 20.66it/s]\u001b[A\u001b[A\n",
      "\n",
      "Predict with MultVAE:  14%|█▍        | 9/63 [00:00<00:02, 25.86it/s]\u001b[A\u001b[A\n",
      "\n",
      "Predict with MultVAE:  21%|██        | 13/63 [00:00<00:01, 28.99it/s]\u001b[A\u001b[A\n",
      "\n",
      "Predict with MultVAE:  27%|██▋       | 17/63 [00:00<00:01, 31.08it/s]\u001b[A\u001b[A\n",
      "\n",
      "Predict with MultVAE:  35%|███▍      | 22/63 [00:00<00:01, 34.14it/s]\u001b[A\u001b[A\n",
      "\n",
      "Predict with MultVAE:  41%|████▏     | 26/63 [00:00<00:01, 35.13it/s]\u001b[A\u001b[A\n",
      "\n",
      "Predict with MultVAE:  48%|████▊     | 30/63 [00:00<00:00, 36.17it/s]\u001b[A\u001b[A\n",
      "\n",
      "Predict with MultVAE:  56%|█████▌    | 35/63 [00:01<00:00, 37.73it/s]\u001b[A\u001b[A\n",
      "\n",
      "Predict with MultVAE:  62%|██████▏   | 39/63 [00:01<00:00, 37.14it/s]\u001b[A\u001b[A\n",
      "\n",
      "Predict with MultVAE:  68%|██████▊   | 43/63 [00:01<00:00, 37.74it/s]\u001b[A\u001b[A\n",
      "\n",
      "Predict with MultVAE:  76%|███████▌  | 48/63 [00:01<00:00, 40.23it/s]\u001b[A\u001b[A\n",
      "\n",
      "Predict with MultVAE:  84%|████████▍ | 53/63 [00:01<00:00, 40.90it/s]\u001b[A\u001b[A\n",
      "\n",
      "Predict with MultVAE:  92%|█████████▏| 58/63 [00:01<00:00, 42.23it/s]\u001b[A\u001b[A\n",
      "\n",
      "Predict with MultVAE: 100%|██████████| 63/63 [00:01<00:00, 36.18it/s]\u001b[A\u001b[A\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "all models predicted\n",
      "MultVAE\n",
      "EASE\n",
      "SLIM\n",
      "we got all scores\n",
      "joining scores\n",
      "exploding\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "  0%|          | 0/240750 [00:00<?, ?it/s]\u001b[A\u001b[A\n",
      "\n",
      "  0%|          | 1/240750 [00:53<3577:59:13, 53.50s/it]\u001b[A\u001b[A\n",
      "\n",
      "  0%|          | 2/240750 [01:45<3507:23:47, 52.45s/it]\u001b[A\u001b[A\n",
      "\n",
      "  0%|          | 3/240750 [02:52<3968:31:33, 59.34s/it]\u001b[A\u001b[A\n",
      "\n",
      " 10%|█         | 25089/240750 [02:52<13:03, 275.38it/s]\u001b[A\u001b[A\n",
      "\n",
      " 18%|█▊        | 44420/240750 [02:53<05:40, 577.13it/s]\u001b[A\u001b[A\n",
      "\n",
      " 23%|██▎       | 55703/240750 [03:52<08:40, 355.43it/s]\u001b[A\u001b[A\n",
      "\n",
      " 23%|██▎       | 55703/240750 [04:55<16:21, 188.56it/s]\u001b[A\u001b[AProcess ForkPoolWorker-114:\n",
      "Process ForkPoolWorker-115:\n",
      "Process ForkPoolWorker-106:\n",
      "\n",
      "Process ForkPoolWorker-116:\n",
      "Process ForkPoolWorker-119:\n",
      "Process ForkPoolWorker-117:\n",
      "Process ForkPoolWorker-118:\n",
      "Process ForkPoolWorker-105:\n",
      "Process ForkPoolWorker-112:\n",
      "Process ForkPoolWorker-120:\n",
      "Process ForkPoolWorker-104:\n",
      "Process ForkPoolWorker-113:\n",
      "Process ForkPoolWorker-107:\n",
      "Process ForkPoolWorker-111:\n",
      "Process ForkPoolWorker-109:\n",
      "Process ForkPoolWorker-108:\n",
      "Process ForkPoolWorker-102:\n",
      "Process ForkPoolWorker-103:\n",
      "Process ForkPoolWorker-110:\n",
      "Process ForkPoolWorker-101:\n",
      "Traceback (most recent call last):\n",
      "Traceback (most recent call last):\n",
      "Traceback (most recent call last):\n",
      "Traceback (most recent call last):\n",
      "Traceback (most recent call last):\n",
      "Traceback (most recent call last):\n",
      "Traceback (most recent call last):\n",
      "Traceback (most recent call last):\n"
     ]
    }
   ],
   "source": [
    "\n",
    "num_chunks = 10\n",
    "\n",
    "joined = pd.DataFrame()\n",
    "\n",
    "for chunk in tqdm(np.array_split(pred, num_chunks)):\n",
    "    \n",
    "    chunk, _ = batched_contstruc_2_lvl(chunk, models, num_chunks=1)\n",
    "\n",
    "    chunk['preds'] = model.predict_proba(chunk[FEATURES])[:,1]\n",
    "    chunk['joined'] = chunk.apply(lambda x: f'{x.cands}_{x.preds}', axis=1)\n",
    "    chunk = chunk.groupby('pav_order_id').agg({'joined':list}).reset_index()\n",
    "    chunk['preds_boosting'] = chunk.apply(lambda x: decode_scores(x), axis=1)\n",
    "    chunk['preds'] = chunk.preds_boosting.apply(lambda x: get_final_recs(x))\n",
    "    chunk = chunk.set_index('pav_order_id')\n",
    "    chunk = chunk[['preds']].copy()\n",
    "    joined = pd.concat([joined, chunk])\n",
    "\n",
    "joined.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "id": "a5fe2c3c",
   "metadata": {},
   "outputs": [],
   "source": [
    "joined.to_csv(path+'boosting.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "id": "978bbb4e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((80244, 1), (80244, 2))"
      ]
     },
     "execution_count": 66,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "joined.shape, pred.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "5d4dd015",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}