P7 azure training.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "013b16ae-b3b2-4fdc-a158-90d150b40267",
   "metadata": {},
   "source": [
    "# AZURE ML designer\n",
    "## 1 -creation données brutes en CSV (500 lignes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "1c29afa0-e02a-44e9-8463-20fbf4c29d97",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2.8633785247802734\n",
      "le fichier contient  1600000 lignes\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>target</th>\n",
       "      <th>id</th>\n",
       "      <th>date</th>\n",
       "      <th>flag</th>\n",
       "      <th>user</th>\n",
       "      <th>text</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>1467810369</td>\n",
       "      <td>Mon Apr 06 22:19:45 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>_TheSpecialOne_</td>\n",
       "      <td>@switchfoot http://twitpic.com/2y1zl - Awww, t...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>1467810672</td>\n",
       "      <td>Mon Apr 06 22:19:49 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>scotthamilton</td>\n",
       "      <td>is upset that he can't update his Facebook by ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>1467810917</td>\n",
       "      <td>Mon Apr 06 22:19:53 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>mattycus</td>\n",
       "      <td>@Kenichan I dived many times for the ball. Man...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>1467811184</td>\n",
       "      <td>Mon Apr 06 22:19:57 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>ElleCTF</td>\n",
       "      <td>my whole body feels itchy and like its on fire</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>1467811193</td>\n",
       "      <td>Mon Apr 06 22:19:57 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>Karoli</td>\n",
       "      <td>@nationwideclass no, it's not behaving at all....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1599995</th>\n",
       "      <td>4</td>\n",
       "      <td>2193601966</td>\n",
       "      <td>Tue Jun 16 08:40:49 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>AmandaMarie1028</td>\n",
       "      <td>Just woke up. Having no school is the best fee...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1599996</th>\n",
       "      <td>4</td>\n",
       "      <td>2193601969</td>\n",
       "      <td>Tue Jun 16 08:40:49 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>TheWDBoards</td>\n",
       "      <td>TheWDB.com - Very cool to hear old Walt interv...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1599997</th>\n",
       "      <td>4</td>\n",
       "      <td>2193601991</td>\n",
       "      <td>Tue Jun 16 08:40:49 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>bpbabe</td>\n",
       "      <td>Are you ready for your MoJo Makeover? Ask me f...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1599998</th>\n",
       "      <td>4</td>\n",
       "      <td>2193602064</td>\n",
       "      <td>Tue Jun 16 08:40:49 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>tinydiamondz</td>\n",
       "      <td>Happy 38th Birthday to my boo of alll time!!! ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1599999</th>\n",
       "      <td>4</td>\n",
       "      <td>2193602129</td>\n",
       "      <td>Tue Jun 16 08:40:50 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>RyanTrevMorris</td>\n",
       "      <td>happy #charitytuesday @theNSPCC @SparksCharity...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>1600000 rows × 6 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "         target          id                          date      flag  \\\n",
       "0             0  1467810369  Mon Apr 06 22:19:45 PDT 2009  NO_QUERY   \n",
       "1             0  1467810672  Mon Apr 06 22:19:49 PDT 2009  NO_QUERY   \n",
       "2             0  1467810917  Mon Apr 06 22:19:53 PDT 2009  NO_QUERY   \n",
       "3             0  1467811184  Mon Apr 06 22:19:57 PDT 2009  NO_QUERY   \n",
       "4             0  1467811193  Mon Apr 06 22:19:57 PDT 2009  NO_QUERY   \n",
       "...         ...         ...                           ...       ...   \n",
       "1599995       4  2193601966  Tue Jun 16 08:40:49 PDT 2009  NO_QUERY   \n",
       "1599996       4  2193601969  Tue Jun 16 08:40:49 PDT 2009  NO_QUERY   \n",
       "1599997       4  2193601991  Tue Jun 16 08:40:49 PDT 2009  NO_QUERY   \n",
       "1599998       4  2193602064  Tue Jun 16 08:40:49 PDT 2009  NO_QUERY   \n",
       "1599999       4  2193602129  Tue Jun 16 08:40:50 PDT 2009  NO_QUERY   \n",
       "\n",
       "                    user                                               text  \n",
       "0        _TheSpecialOne_  @switchfoot http://twitpic.com/2y1zl - Awww, t...  \n",
       "1          scotthamilton  is upset that he can't update his Facebook by ...  \n",
       "2               mattycus  @Kenichan I dived many times for the ball. Man...  \n",
       "3                ElleCTF    my whole body feels itchy and like its on fire   \n",
       "4                 Karoli  @nationwideclass no, it's not behaving at all....  \n",
       "...                  ...                                                ...  \n",
       "1599995  AmandaMarie1028  Just woke up. Having no school is the best fee...  \n",
       "1599996      TheWDBoards  TheWDB.com - Very cool to hear old Walt interv...  \n",
       "1599997           bpbabe  Are you ready for your MoJo Makeover? Ask me f...  \n",
       "1599998     tinydiamondz  Happy 38th Birthday to my boo of alll time!!! ...  \n",
       "1599999   RyanTrevMorris  happy #charitytuesday @theNSPCC @SparksCharity...  \n",
       "\n",
       "[1600000 rows x 6 columns]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import time\n",
    "\n",
    "start = time.time()\n",
    "data = pd.read_csv('input/sentiment140/training.1600000.processed.noemoticon.csv', delimiter = ',', header=None)\n",
    "print(time.time()-start)\n",
    "print('le fichier contient ', len(data.index),'lignes' )\n",
    "data.columns=['target','id','date','flag','user','text']\n",
    "display(data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "660e30aa-20a3-4eb2-ba3f-256a73b42f81",
   "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>cible</th>\n",
       "      <th>text</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>@BluebirdQueen It is a really nice sunday  aww x</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>@scrappergirl106  Don't feel bad, I'm up, too ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>Wow The Hangover was hilarious!  Nonstop laughs.</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>OMG Next Friday is our prom... Miss my class a...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>Good bye my delicious bread</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>495</th>\n",
       "      <td>0</td>\n",
       "      <td>Back from sleepover! Fun night! Lots of homewo...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>496</th>\n",
       "      <td>0</td>\n",
       "      <td>@hannahnicklin  Oh well, we shall have to meet...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497</th>\n",
       "      <td>0</td>\n",
       "      <td>Fuck fuck fuck fuck never never i fucking hate...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>498</th>\n",
       "      <td>1</td>\n",
       "      <td>Today has been long and tiring, but I've a smi...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>499</th>\n",
       "      <td>0</td>\n",
       "      <td>pizza and singing and drinking don't go good...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>500 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "     cible                                               text\n",
       "0        0   @BluebirdQueen It is a really nice sunday  aww x\n",
       "1        1  @scrappergirl106  Don't feel bad, I'm up, too ...\n",
       "2        1   Wow The Hangover was hilarious!  Nonstop laughs.\n",
       "3        0  OMG Next Friday is our prom... Miss my class a...\n",
       "4        0                       Good bye my delicious bread \n",
       "..     ...                                                ...\n",
       "495      0  Back from sleepover! Fun night! Lots of homewo...\n",
       "496      0  @hannahnicklin  Oh well, we shall have to meet...\n",
       "497      0  Fuck fuck fuck fuck never never i fucking hate...\n",
       "498      1  Today has been long and tiring, but I've a smi...\n",
       "499      0   pizza and singing and drinking don't go good... \n",
       "\n",
       "[500 rows x 2 columns]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#melange\n",
    "data0_shuffled=data.loc[data.target==0,:].sample(frac=1).reset_index(drop=True)\n",
    "data4_shuffled=data.loc[data.target==4,:].sample(frac=1).reset_index(drop=True)\n",
    "\n",
    "data_0_sample=data0_shuffled.iloc[0:250].copy()\n",
    "data_4_sample=data4_shuffled.iloc[0:250].copy()\n",
    "\n",
    "result = pd.concat([data_0_sample,data_4_sample])\n",
    "result['cible']=result['target'].map({0:0,4:1})\n",
    "result=result.sample(frac=1).reset_index(drop=True)\n",
    "\n",
    "res=result.loc[:,['cible','text']].copy()\n",
    "display(res)\n",
    "\n",
    "res.to_csv('./data2AMLD500.csv', index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fa73d72d-1bb2-406a-ba22-de413245e0c1",
   "metadata": {},
   "source": [
    "## AZURE ML SDK"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "2a56297a-d1d7-4dbe-9677-07a7d1dd60af",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Ready to use Azure ML 1.38.0\n"
     ]
    }
   ],
   "source": [
    "import azureml.core\n",
    "print(\"Ready to use Azure ML\", azureml.core.VERSION)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "60ad45a2-65fe-4a89-8fde-af7fa171cdf5",
   "metadata": {},
   "source": [
    "## 2- connexion au workspace"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "3017d18c-864c-473c-a385-dd533915887d",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "workspace nab-ML loaded\n",
      "Compute Resources:\n",
      "instance existance: P7-CPU : ComputeInstance\n"
     ]
    }
   ],
   "source": [
    "from azureml.core import Workspace\n",
    "\n",
    "#connection avec authentification \n",
    "#fichier config recuperé sur le portail et stocké dans P7\n",
    "ws = Workspace.from_config()\n",
    "print(\"workspace\",ws.name, \"loaded\")\n",
    "\n",
    "#recuperaton des cibles de calcul par exemple \n",
    "#grace à l'attribut compute_targets de la classe workspace\n",
    "print(\"Compute Resources:\")\n",
    "for compute_name in ws.compute_targets:\n",
    "    compute = ws.compute_targets[compute_name]\n",
    "    print(\"instance existance:\", compute.name, \":\", compute.type)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "54fe8986-7bf7-493f-ac3f-8cc376de004c",
   "metadata": {},
   "source": [
    "## 3-  recupération du datastore de defaut lié au workspace"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "id": "8cd538ec-7f94-4724-98c0-889d7c61ca31",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "azureml_globaldatasets - Default = False\n",
      "workspacefilestore - Default = False\n",
      "workspaceworkingdirectory - Default = False\n",
      "workspaceartifactstore - Default = False\n",
      "workspaceblobstore - Default = True\n"
     ]
    }
   ],
   "source": [
    "# Get the default datastore\n",
    "default_ds = ws.get_default_datastore()\n",
    "\n",
    "# Enumerate all datastores, indicating which is the default\n",
    "for ds_name in ws.datastores:\n",
    "    print(ds_name, \"- Default =\", ds_name == default_ds.name)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a63648cb-84e5-4b46-aeb3-443663fef5b3",
   "metadata": {},
   "source": [
    "## 4-  Upload des fichiers CSV vers datastore\n",
    "* on passe par la methode upload_directory disponible qu 'avec le dataset.File et non le dataset.Tabular\n",
    "* les methodes from_file ne fonctionnent que pour une url ou dans le datastore"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "id": "b456c969-db8b-4f88-9bbb-2d3f744f9366",
   "metadata": {},
   "outputs": [],
   "source": [
    "#Upload data to a datastore\n",
    "#Now that you have determined the available datastores, you can upload files from your local file system to a datastore \n",
    "#so that it will be accessible to experiments running in the workspace, regardless of where the experiment script is actually being run."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "id": "04defa70-f8ad-4f00-864b-43de44f01036",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Validating arguments.\n",
      "Arguments validated.\n",
      "Uploading file to P7/\n",
      "Filtering files with pattern matching *.csv\n",
      "Uploading an estimated of 1 files\n",
      "Uploading ./input/sentiment140/training.1600000.processed.noemoticon.csv\n",
      "Uploaded ./input/sentiment140/training.1600000.processed.noemoticon.csv, 1 files out of an estimated total of 1\n",
      "Uploaded 1 files\n",
      "Creating new dataset\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "{\n",
       "  \"source\": [\n",
       "    \"('workspaceblobstore', 'P7/')\"\n",
       "  ],\n",
       "  \"definition\": [\n",
       "    \"GetDatastoreFiles\"\n",
       "  ]\n",
       "}"
      ]
     },
     "execution_count": 96,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from azureml.core import Dataset, Datastore\n",
    "\n",
    "\n",
    "target_path='P7/'\n",
    "\n",
    "#upload_directory(src_dir, target, pattern=None, overwrite=False, show_progress=True)\n",
    "#sentiment\n",
    "Dataset.File.upload_directory('./input/sentiment140/', (default_ds, target_path), pattern='*.csv', overwrite=True, show_progress=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "id": "62fd9b53-3333-4f14-9712-821380da1f4c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Validating arguments.\n",
      "Arguments validated.\n",
      "Uploading file to P7/\n",
      "Filtering files with pattern matching *50d.txt\n",
      "Uploading an estimated of 1 files\n",
      "Uploading ./embedding/glove.6B/glove.6B.50d.txt\n",
      "Uploaded ./embedding/glove.6B/glove.6B.50d.txt, 1 files out of an estimated total of 1\n",
      "Uploaded 1 files\n",
      "Creating new dataset\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "{\n",
       "  \"source\": [\n",
       "    \"('workspaceblobstore', 'P7/')\"\n",
       "  ],\n",
       "  \"definition\": [\n",
       "    \"GetDatastoreFiles\"\n",
       "  ]\n",
       "}"
      ]
     },
     "execution_count": 97,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#word embedding\n",
    "Dataset.File.upload_directory('./embedding/glove.6B/', (default_ds, target_path), pattern='*50d.txt', overwrite=True, show_progress=True)\n",
    "\n",
    "#depreciated\n",
    "#depreciated default_ds.upload_files(files=['./input/sentiment140/training.1600000.processed.noemoticon.csv'], # Upload \n",
    "#depreciated                       target_path='P7/', # Put it in a folder path in the datastore\n",
    "#depreciated                       overwrite=True, # Replace existing files of the same name\n",
    "#depreciated                       show_progress=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c6d143c2-aa8c-4625-a42c-9c050333a724",
   "metadata": {},
   "source": [
    "## 5-a Passer le file sentiment en tabular dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "id": "7fe1d7bb-cc76-43b3-b09c-f1cf0b403e5a",
   "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>Column1</th>\n",
       "      <th>Column2</th>\n",
       "      <th>Column3</th>\n",
       "      <th>Column4</th>\n",
       "      <th>Column5</th>\n",
       "      <th>Column6</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>1467810369</td>\n",
       "      <td>Mon Apr 06 22:19:45 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>_TheSpecialOne_</td>\n",
       "      <td>@switchfoot http://twitpic.com/2y1zl - Awww, t...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>1467810672</td>\n",
       "      <td>Mon Apr 06 22:19:49 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>scotthamilton</td>\n",
       "      <td>is upset that he can't update his Facebook by ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>1467810917</td>\n",
       "      <td>Mon Apr 06 22:19:53 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>mattycus</td>\n",
       "      <td>@Kenichan I dived many times for the ball. Man...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>1467811184</td>\n",
       "      <td>Mon Apr 06 22:19:57 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>ElleCTF</td>\n",
       "      <td>my whole body feels itchy and like its on fire</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>1467811193</td>\n",
       "      <td>Mon Apr 06 22:19:57 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>Karoli</td>\n",
       "      <td>@nationwideclass no, it's not behaving at all....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>0</td>\n",
       "      <td>1467811372</td>\n",
       "      <td>Mon Apr 06 22:20:00 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>joy_wolf</td>\n",
       "      <td>@Kwesidei not the whole crew</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>0</td>\n",
       "      <td>1467811592</td>\n",
       "      <td>Mon Apr 06 22:20:03 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>mybirch</td>\n",
       "      <td>Need a hug</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>0</td>\n",
       "      <td>1467811594</td>\n",
       "      <td>Mon Apr 06 22:20:03 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>coZZ</td>\n",
       "      <td>@LOLTrish hey  long time no see! Yes.. Rains a...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>0</td>\n",
       "      <td>1467811795</td>\n",
       "      <td>Mon Apr 06 22:20:05 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>2Hood4Hollywood</td>\n",
       "      <td>@Tatiana_K nope they didn't have it</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>0</td>\n",
       "      <td>1467812025</td>\n",
       "      <td>Mon Apr 06 22:20:09 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>mimismo</td>\n",
       "      <td>@twittera que me muera ?</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Column1     Column2                       Column3   Column4  \\\n",
       "0        0  1467810369  Mon Apr 06 22:19:45 PDT 2009  NO_QUERY   \n",
       "1        0  1467810672  Mon Apr 06 22:19:49 PDT 2009  NO_QUERY   \n",
       "2        0  1467810917  Mon Apr 06 22:19:53 PDT 2009  NO_QUERY   \n",
       "3        0  1467811184  Mon Apr 06 22:19:57 PDT 2009  NO_QUERY   \n",
       "4        0  1467811193  Mon Apr 06 22:19:57 PDT 2009  NO_QUERY   \n",
       "5        0  1467811372  Mon Apr 06 22:20:00 PDT 2009  NO_QUERY   \n",
       "6        0  1467811592  Mon Apr 06 22:20:03 PDT 2009  NO_QUERY   \n",
       "7        0  1467811594  Mon Apr 06 22:20:03 PDT 2009  NO_QUERY   \n",
       "8        0  1467811795  Mon Apr 06 22:20:05 PDT 2009  NO_QUERY   \n",
       "9        0  1467812025  Mon Apr 06 22:20:09 PDT 2009  NO_QUERY   \n",
       "\n",
       "           Column5                                            Column6  \n",
       "0  _TheSpecialOne_  @switchfoot http://twitpic.com/2y1zl - Awww, t...  \n",
       "1    scotthamilton  is upset that he can't update his Facebook by ...  \n",
       "2         mattycus  @Kenichan I dived many times for the ball. Man...  \n",
       "3          ElleCTF    my whole body feels itchy and like its on fire   \n",
       "4           Karoli  @nationwideclass no, it's not behaving at all....  \n",
       "5         joy_wolf                      @Kwesidei not the whole crew   \n",
       "6          mybirch                                        Need a hug   \n",
       "7             coZZ  @LOLTrish hey  long time no see! Yes.. Rains a...  \n",
       "8  2Hood4Hollywood               @Tatiana_K nope they didn't have it   \n",
       "9          mimismo                          @twittera que me muera ?   "
      ]
     },
     "execution_count": 98,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from azureml.core import Dataset\n",
    "\n",
    "#Create a tabular dataset from the path on the datastore (this may take a short while)\n",
    "tab_data_set = Dataset.Tabular.from_delimited_files(path=(default_ds, 'P7/*.csv'),header=False, separator=',', support_multi_line=True, encoding='latin1')\n",
    "#'latin1' pour eviter le sproblèmes avec 'utf8'\n",
    "#« UTF8 », « iso88591 », « latin1 », « ASCII », « UTF16 », « UTF32 », « utf8bom » et « Windows1252\n",
    "\n",
    "# Display the first 10 rows as a Pandas dataframe\n",
    "tab_data_set.take(10).to_pandas_dataframe(on_error='null', out_of_range_datetime='null')\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "466d2f02-7758-4d4f-a3d6-dff1fa04a399",
   "metadata": {},
   "source": [
    "## 5-b passer le file WE glove 50d en file dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "id": "af398a58-6a21-4fd5-bb5c-bd3a27177b85",
   "metadata": {},
   "outputs": [],
   "source": [
    "from azureml.core import Dataset\n",
    "\n",
    "#Create a filedataset from the path on the datastore (this may take a short while)\n",
    "fil_data_set = Dataset.File.from_files(path=(default_ds, 'P7/*.txt'))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b1d22883-07d9-4c7c-887b-6226833c31ee",
   "metadata": {},
   "source": [
    "## 6- Register dataset\n",
    "* on inscrit le dataset pour qu il puisse etre utilisé par des experiences du workspace"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "id": "4074dda6-c904-4679-8e00-f534ad4dd423",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Register the tabular dataset\n",
    "try:\n",
    "    tab_data_set = tab_data_set.register(workspace=ws, \n",
    "                                        name='sentiment dataset',\n",
    "                                        description='sentiment data',\n",
    "                                        tags = {'format':'CSV'},\n",
    "                                        create_new_version=True)\n",
    "except Exception as ex:\n",
    "    print(ex)\n",
    "    \n",
    "    \n",
    "# Register the file dataset\n",
    "try:\n",
    "    fil_data_set = fil_data_set.register(workspace=ws, \n",
    "                                        name='glove50d dataset',\n",
    "                                        description='glove 50d word emb data',\n",
    "                                        tags = {'format':'txt'},\n",
    "                                        create_new_version=True)\n",
    "except Exception as ex:\n",
    "    print(ex)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c249be31-de24-4009-8b4b-b0ad4453bb19",
   "metadata": {},
   "source": [
    "* liste des datasets inscrits dans le workspace nab-ML"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "id": "78aaf373-4dff-49d6-915d-5b386cc1a861",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Datasets:\n",
      "\t glove50d dataset version 1\n",
      "\t sentiment dataset version 1\n",
      "\t P7_sent_ultralight version 1\n",
      "\t P7_sent_light version 1\n"
     ]
    }
   ],
   "source": [
    "print(\"Datasets:\")\n",
    "for dataset_name in list(ws.datasets.keys()):\n",
    "    dataset = Dataset.get_by_name(ws, dataset_name)\n",
    "    print(\"\\t\", dataset.name, 'version', dataset.version)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "303f9a30-cb00-4b1e-bc6d-489c503169c3",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "281.4914824962616\n"
     ]
    }
   ],
   "source": [
    "import time\n",
    "from azureml.core import Dataset\n",
    "dataset=Dataset.get_by_name(ws, 'sentiment dataset')\n",
    "start = time.time()\n",
    "data=dataset.to_pandas_dataframe(on_error='null', out_of_range_datetime='null')\n",
    "print(time.time()-start)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "c0d8e8fe-69e4-4a94-b780-ee955b40e550",
   "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>Column1</th>\n",
       "      <th>Column2</th>\n",
       "      <th>Column3</th>\n",
       "      <th>Column4</th>\n",
       "      <th>Column5</th>\n",
       "      <th>Column6</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>1467810369</td>\n",
       "      <td>Mon Apr 06 22:19:45 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>_TheSpecialOne_</td>\n",
       "      <td>@switchfoot http://twitpic.com/2y1zl - Awww, t...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>1467810672</td>\n",
       "      <td>Mon Apr 06 22:19:49 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>scotthamilton</td>\n",
       "      <td>is upset that he can't update his Facebook by ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>1467810917</td>\n",
       "      <td>Mon Apr 06 22:19:53 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>mattycus</td>\n",
       "      <td>@Kenichan I dived many times for the ball. Man...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>1467811184</td>\n",
       "      <td>Mon Apr 06 22:19:57 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>ElleCTF</td>\n",
       "      <td>my whole body feels itchy and like its on fire</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>1467811193</td>\n",
       "      <td>Mon Apr 06 22:19:57 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>Karoli</td>\n",
       "      <td>@nationwideclass no, it's not behaving at all....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1599995</th>\n",
       "      <td>4</td>\n",
       "      <td>2193601966</td>\n",
       "      <td>Tue Jun 16 08:40:49 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>AmandaMarie1028</td>\n",
       "      <td>Just woke up. Having no school is the best fee...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1599996</th>\n",
       "      <td>4</td>\n",
       "      <td>2193601969</td>\n",
       "      <td>Tue Jun 16 08:40:49 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>TheWDBoards</td>\n",
       "      <td>TheWDB.com - Very cool to hear old Walt interv...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1599997</th>\n",
       "      <td>4</td>\n",
       "      <td>2193601991</td>\n",
       "      <td>Tue Jun 16 08:40:49 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>bpbabe</td>\n",
       "      <td>Are you ready for your MoJo Makeover? Ask me f...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1599998</th>\n",
       "      <td>4</td>\n",
       "      <td>2193602064</td>\n",
       "      <td>Tue Jun 16 08:40:49 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>tinydiamondz</td>\n",
       "      <td>Happy 38th Birthday to my boo of alll time!!! ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1599999</th>\n",
       "      <td>4</td>\n",
       "      <td>2193602129</td>\n",
       "      <td>Tue Jun 16 08:40:50 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>RyanTrevMorris</td>\n",
       "      <td>happy #charitytuesday @theNSPCC @SparksCharity...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>1600000 rows × 6 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "         Column1     Column2                       Column3   Column4  \\\n",
       "0              0  1467810369  Mon Apr 06 22:19:45 PDT 2009  NO_QUERY   \n",
       "1              0  1467810672  Mon Apr 06 22:19:49 PDT 2009  NO_QUERY   \n",
       "2              0  1467810917  Mon Apr 06 22:19:53 PDT 2009  NO_QUERY   \n",
       "3              0  1467811184  Mon Apr 06 22:19:57 PDT 2009  NO_QUERY   \n",
       "4              0  1467811193  Mon Apr 06 22:19:57 PDT 2009  NO_QUERY   \n",
       "...          ...         ...                           ...       ...   \n",
       "1599995        4  2193601966  Tue Jun 16 08:40:49 PDT 2009  NO_QUERY   \n",
       "1599996        4  2193601969  Tue Jun 16 08:40:49 PDT 2009  NO_QUERY   \n",
       "1599997        4  2193601991  Tue Jun 16 08:40:49 PDT 2009  NO_QUERY   \n",
       "1599998        4  2193602064  Tue Jun 16 08:40:49 PDT 2009  NO_QUERY   \n",
       "1599999        4  2193602129  Tue Jun 16 08:40:50 PDT 2009  NO_QUERY   \n",
       "\n",
       "                 Column5                                            Column6  \n",
       "0        _TheSpecialOne_  @switchfoot http://twitpic.com/2y1zl - Awww, t...  \n",
       "1          scotthamilton  is upset that he can't update his Facebook by ...  \n",
       "2               mattycus  @Kenichan I dived many times for the ball. Man...  \n",
       "3                ElleCTF    my whole body feels itchy and like its on fire   \n",
       "4                 Karoli  @nationwideclass no, it's not behaving at all....  \n",
       "...                  ...                                                ...  \n",
       "1599995  AmandaMarie1028  Just woke up. Having no school is the best fee...  \n",
       "1599996      TheWDBoards  TheWDB.com - Very cool to hear old Walt interv...  \n",
       "1599997           bpbabe  Are you ready for your MoJo Makeover? Ask me f...  \n",
       "1599998     tinydiamondz  Happy 38th Birthday to my boo of alll time!!! ...  \n",
       "1599999   RyanTrevMorris  happy #charitytuesday @theNSPCC @SparksCharity...  \n",
       "\n",
       "[1600000 rows x 6 columns]"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "30d28bce-6f2f-445b-8da1-f8697d126ef0",
   "metadata": {},
   "source": [
    "## 7-preprocessing  CLEANING / LEMMATIZATION / STEMMING"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "5a7160e9-a78a-4b14-a563-a327c9573092",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "276.4624447822571\n"
     ]
    }
   ],
   "source": [
    "#https://inside-machinelearning.com/preprocessing-nlp-tutoriel-pour-nettoyer-rapidement-un-texte/\n",
    "#https://towardsdatascience.com/text-normalization-with-spacy-and-nltk-1302ff430119\n",
    "\n",
    "import regex as re #module regex\n",
    "import nltk\n",
    "import string\n",
    "from nltk.stem import WordNetLemmatizer\n",
    "#tous les dictionnaires: https://www.nltk.org/nltk_data/\n",
    "from nltk.corpus import words\n",
    "\n",
    "\n",
    " #DL des stopwords NLTK\n",
    "#nltk.download('stopwords')\n",
    " #DL des punctuations NLTK\n",
    "#nltk.download('punkt')\n",
    " #DL du dictionnaire de mots anglais NTLK\n",
    "#nltk.download('words')\n",
    " #DL pour la lemmatization\n",
    "#nltk.download('wordnet')\n",
    "\n",
    "\n",
    "#stopwords anglais\n",
    "#attention les stopwords n'est pas une liste absolue\n",
    "stopwords = nltk.corpus.stopwords.words('english')\n",
    "more_stopwords = ['us', 'im', 'c', \"youd\", \"hadnt\", \n",
    "                  \"hasnt\", \"havent\", \"isnt\",\"might\", \"mightnt\", \"mustnt\", \"neednt\", \n",
    "                  \"shant\", \"shouldnt\", \"wasnt\", \"werent\", \"wont\", \"wouldnt\",\n",
    "                  \"youre\",\"yourent\", \"youve\", \"youvent\", \"youll\", \"shes\",\"shesnt\" \"thatll\",\n",
    "                  'cant','wont',\"dont\", \"shouldve\",'aint',\"arent\",\"couldnt\",\"didnt\", \"doesnt\",\"thats\",\"shoulda\"]\n",
    "stopwords = stopwords + more_stopwords\n",
    "\n",
    "\n",
    "#mots anglais du dictionnaire NLTK\n",
    "#tous les dictionnaires: https://www.nltk.org/nltk_data/\n",
    "#set() permet d'avoir un mot unique\n",
    "words = set(nltk.corpus.words.words())\n",
    "#words = set(nltk.corpus.brown.words())\n",
    "\n",
    "wnl = WordNetLemmatizer()\n",
    "\n",
    "def Preprocess_npl(listofdoc):\n",
    "    '''cleaning + lemmatization\n",
    "    sub=substitute'''\n",
    "    preprocess_list_1 = []\n",
    "    preprocess_list_2 = []\n",
    "    preprocess_list_3 = []\n",
    "    for doc in listofdoc :\n",
    "        #remove text in square brackets: .(any caracter) * (0 or more times) ?\\] (0 or 1 time ']')\n",
    "        doc = re.sub('\\[.*?\\]', '', doc)\n",
    "        #remove links\n",
    "        doc = re.sub('https?://\\S+|www\\.\\S+', '', doc)\n",
    "        doc = re.sub('http?://\\S+|www\\.\\S+', '', doc)\n",
    "        ##remove text in angle brackets\n",
    "        doc = re.sub('<.*?>+', '', doc)\n",
    "        #remove punctuation\n",
    "        #!\"#$%&'()*+,-./:;<=>?@[\\]^_`{|}~\n",
    "        #doc = re.sub('[%s]' % re.escape(string.punctuation), '', doc)\n",
    "        #remove back line\n",
    "        doc = re.sub('\\n', '', doc)\n",
    "        #word cracter digit word caracter\n",
    "        #doc = re.sub('\\w*\\d\\w*', '', doc)\n",
    "        #remove special carcter comme  ♫\n",
    "        doc = re.sub('\\p{S}', '', doc)\n",
    "    \n",
    "        #la fonction join permet de creer un separateur\n",
    "        #lowercase & remove punctuation : example: Don't->dont\n",
    "        doc_wo_punct = \"\".join([i.lower() for i in doc if i not in string.punctuation])\n",
    "        #remove number\n",
    "        doc_wo_num = ''.join(i for i in doc_wo_punct if not i.isdigit())\n",
    "        #tokenization\n",
    "        token_doc = nltk.tokenize.word_tokenize(doc_wo_num)\n",
    "        #remove stopwords\n",
    "        words_wo_stopwords = [w for w in token_doc if w not in stopwords]\n",
    "        \n",
    "        #lemmatization\n",
    "        #attention ici on cree un generateur et non une liste: (wnl.lemmatize(w) for w in words_wo_stopwords) different de [wnl.lemmatize(w) for w in words_wo_stopwords]\n",
    "        #la difference et qu on recupere à la demande\n",
    "        #https://stackoverflow.com/questions/47789/generator-expressions-vs-list-comprehensions\n",
    "        words_lemmatize = [wnl.lemmatize(w) for w in words_wo_stopwords]\n",
    "        \n",
    "        #on ne retourne que les mots appartenant au dictionnaire words  et ne possédant pas autre chose que des carcatère de l'alphabet\n",
    "        doc_clean_1 = ' '.join(w for w in words_wo_stopwords )\n",
    "        doc_clean_2 = ' '.join(w for w in words_lemmatize)\n",
    "        doc_clean_3 = ' '.join(w for w in words_lemmatize if w.lower() in words or not w.isalpha())\n",
    "        \n",
    "        preprocess_list_1.append(doc_clean_1)\n",
    "        preprocess_list_2.append(doc_clean_2)\n",
    "        preprocess_list_3.append(doc_clean_3)\n",
    "    return preprocess_list_1, preprocess_list_2, preprocess_list_3\n",
    "\n",
    "\n",
    "start = time.time()\n",
    "data.columns=['target','id','date','flag','user','text']\n",
    "data['text_sw'],data['text_swlem'],data['text_swlemdic'] =Preprocess_npl(data.text)\n",
    "print(time.time()-start)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "034d0fc5-0577-49a2-a03f-be769de4ac38",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "88.36992263793945\n"
     ]
    }
   ],
   "source": [
    "#https://stackoverflow.com/questions/10554052/what-are-the-major-differences-and-benefits-of-porter-and-lancaster-stemming-alg\n",
    "from nltk.stem.porter import PorterStemmer #less aggressive\n",
    "from nltk.stem.snowball import SnowballStemmer #kind of Porter2 (probably first choice) :0.78sec pour 1000\n",
    "from nltk.stem.lancaster import LancasterStemmer #most aggressive\n",
    "from nltk.stem import WordNetLemmatizer #fast\n",
    "#nltk.download('wordnet')\n",
    "\n",
    "def stemm_data(data,colin,colout,typ,algo):\n",
    "    '''type: stemmer/lemmatizer\n",
    "        stemmer algo: \n",
    "            nltk.PorterStemmer()\n",
    "            nltk.SnowballStemmer(\"english\")\n",
    "            nltk.LancasterStemmer()\n",
    "        lemme algo:\n",
    "            nltk.WordNetLemmatizer()\n",
    "            '''\n",
    "    if typ=='stemmer':\n",
    "        stemmer = algo\n",
    "        def stemm_text(text):\n",
    "            text = ' '.join(stemmer.stem(word) for word in text.split(' '))\n",
    "            return text\n",
    "    elif typ=='lemmatizer':\n",
    "        lemmatizer = algo\n",
    "        def stemm_text(text):\n",
    "            text = ' '.join(lemmatizer.lemmatize(word) for word in text.split(' '))\n",
    "            return text\n",
    "        \n",
    "    data[colout] = data[colin].apply(stemm_text)\n",
    "    return data\n",
    "\n",
    "\n",
    "#start = time.time()\n",
    "#revfilt=stemm_data(data=revfilt,colin='text4',colout='lemm',typ='lemmatizer',algo=nltk.WordNetLemmatizer())\n",
    "#print(time.time()-start)\n",
    "\n",
    "start = time.time()\n",
    "data=stemm_data(data=data,colin='text_swlemdic',colout='text_stemm',typ='stemmer',algo=nltk.SnowballStemmer(\"english\"))\n",
    "print(time.time()-start)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "id": "20c9de61-5df3-44fb-b55f-8814be230af1",
   "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>target</th>\n",
       "      <th>id</th>\n",
       "      <th>date</th>\n",
       "      <th>flag</th>\n",
       "      <th>user</th>\n",
       "      <th>text</th>\n",
       "      <th>text_sw</th>\n",
       "      <th>text_swlem</th>\n",
       "      <th>text_swlemdic</th>\n",
       "      <th>text_stemm</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>1467810369</td>\n",
       "      <td>Mon Apr 06 22:19:45 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>_TheSpecialOne_</td>\n",
       "      <td>@switchfoot http://twitpic.com/2y1zl - Awww, t...</td>\n",
       "      <td>switchfoot awww bummer got david carr third day</td>\n",
       "      <td>switchfoot awww bummer got david carr third day</td>\n",
       "      <td>bummer got carr third day</td>\n",
       "      <td>bummer got carr third day</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>1467810672</td>\n",
       "      <td>Mon Apr 06 22:19:49 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>scotthamilton</td>\n",
       "      <td>is upset that he can't update his Facebook by ...</td>\n",
       "      <td>upset update facebook texting cry result schoo...</td>\n",
       "      <td>upset update facebook texting cry result schoo...</td>\n",
       "      <td>upset update cry result school today also blah</td>\n",
       "      <td>upset updat cri result school today also blah</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>1467810917</td>\n",
       "      <td>Mon Apr 06 22:19:53 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>mattycus</td>\n",
       "      <td>@Kenichan I dived many times for the ball. Man...</td>\n",
       "      <td>kenichan dived many times ball managed save re...</td>\n",
       "      <td>kenichan dived many time ball managed save res...</td>\n",
       "      <td>many time ball save rest go bound</td>\n",
       "      <td>mani time ball save rest go bound</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>1467811184</td>\n",
       "      <td>Mon Apr 06 22:19:57 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>ElleCTF</td>\n",
       "      <td>my whole body feels itchy and like its on fire</td>\n",
       "      <td>whole body feels itchy like fire</td>\n",
       "      <td>whole body feel itchy like fire</td>\n",
       "      <td>whole body feel itchy like fire</td>\n",
       "      <td>whole bodi feel itchi like fire</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>1467811193</td>\n",
       "      <td>Mon Apr 06 22:19:57 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>Karoli</td>\n",
       "      <td>@nationwideclass no, it's not behaving at all....</td>\n",
       "      <td>nationwideclass behaving mad see</td>\n",
       "      <td>nationwideclass behaving mad see</td>\n",
       "      <td>mad see</td>\n",
       "      <td>mad see</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1599995</th>\n",
       "      <td>4</td>\n",
       "      <td>2193601966</td>\n",
       "      <td>Tue Jun 16 08:40:49 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>AmandaMarie1028</td>\n",
       "      <td>Just woke up. Having no school is the best fee...</td>\n",
       "      <td>woke school best feeling ever</td>\n",
       "      <td>woke school best feeling ever</td>\n",
       "      <td>woke school best feeling ever</td>\n",
       "      <td>woke school best feel ever</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1599996</th>\n",
       "      <td>4</td>\n",
       "      <td>2193601969</td>\n",
       "      <td>Tue Jun 16 08:40:49 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>TheWDBoards</td>\n",
       "      <td>TheWDB.com - Very cool to hear old Walt interv...</td>\n",
       "      <td>thewdbcom cool hear old walt interviews â «</td>\n",
       "      <td>thewdbcom cool hear old walt interview â «</td>\n",
       "      <td>cool hear old walt interview «</td>\n",
       "      <td>cool hear old walt interview «</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1599997</th>\n",
       "      <td>4</td>\n",
       "      <td>2193601991</td>\n",
       "      <td>Tue Jun 16 08:40:49 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>bpbabe</td>\n",
       "      <td>Are you ready for your MoJo Makeover? Ask me f...</td>\n",
       "      <td>ready mojo makeover ask details</td>\n",
       "      <td>ready mojo makeover ask detail</td>\n",
       "      <td>ready mojo ask detail</td>\n",
       "      <td>readi mojo ask detail</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1599998</th>\n",
       "      <td>4</td>\n",
       "      <td>2193602064</td>\n",
       "      <td>Tue Jun 16 08:40:49 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>tinydiamondz</td>\n",
       "      <td>Happy 38th Birthday to my boo of alll time!!! ...</td>\n",
       "      <td>happy th birthday boo alll time tupac amaru sh...</td>\n",
       "      <td>happy th birthday boo alll time tupac amaru sh...</td>\n",
       "      <td>happy th birthday boo time</td>\n",
       "      <td>happi th birthday boo time</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1599999</th>\n",
       "      <td>4</td>\n",
       "      <td>2193602129</td>\n",
       "      <td>Tue Jun 16 08:40:50 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>RyanTrevMorris</td>\n",
       "      <td>happy #charitytuesday @theNSPCC @SparksCharity...</td>\n",
       "      <td>happy charitytuesday thenspcc sparkscharity sp...</td>\n",
       "      <td>happy charitytuesday thenspcc sparkscharity sp...</td>\n",
       "      <td>happy</td>\n",
       "      <td>happi</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>1600000 rows × 10 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "         target          id                          date      flag  \\\n",
       "0             0  1467810369  Mon Apr 06 22:19:45 PDT 2009  NO_QUERY   \n",
       "1             0  1467810672  Mon Apr 06 22:19:49 PDT 2009  NO_QUERY   \n",
       "2             0  1467810917  Mon Apr 06 22:19:53 PDT 2009  NO_QUERY   \n",
       "3             0  1467811184  Mon Apr 06 22:19:57 PDT 2009  NO_QUERY   \n",
       "4             0  1467811193  Mon Apr 06 22:19:57 PDT 2009  NO_QUERY   \n",
       "...         ...         ...                           ...       ...   \n",
       "1599995       4  2193601966  Tue Jun 16 08:40:49 PDT 2009  NO_QUERY   \n",
       "1599996       4  2193601969  Tue Jun 16 08:40:49 PDT 2009  NO_QUERY   \n",
       "1599997       4  2193601991  Tue Jun 16 08:40:49 PDT 2009  NO_QUERY   \n",
       "1599998       4  2193602064  Tue Jun 16 08:40:49 PDT 2009  NO_QUERY   \n",
       "1599999       4  2193602129  Tue Jun 16 08:40:50 PDT 2009  NO_QUERY   \n",
       "\n",
       "                    user                                               text  \\\n",
       "0        _TheSpecialOne_  @switchfoot http://twitpic.com/2y1zl - Awww, t...   \n",
       "1          scotthamilton  is upset that he can't update his Facebook by ...   \n",
       "2               mattycus  @Kenichan I dived many times for the ball. Man...   \n",
       "3                ElleCTF    my whole body feels itchy and like its on fire    \n",
       "4                 Karoli  @nationwideclass no, it's not behaving at all....   \n",
       "...                  ...                                                ...   \n",
       "1599995  AmandaMarie1028  Just woke up. Having no school is the best fee...   \n",
       "1599996      TheWDBoards  TheWDB.com - Very cool to hear old Walt interv...   \n",
       "1599997           bpbabe  Are you ready for your MoJo Makeover? Ask me f...   \n",
       "1599998     tinydiamondz  Happy 38th Birthday to my boo of alll time!!! ...   \n",
       "1599999   RyanTrevMorris  happy #charitytuesday @theNSPCC @SparksCharity...   \n",
       "\n",
       "                                                   text_sw  \\\n",
       "0          switchfoot awww bummer got david carr third day   \n",
       "1        upset update facebook texting cry result schoo...   \n",
       "2        kenichan dived many times ball managed save re...   \n",
       "3                         whole body feels itchy like fire   \n",
       "4                         nationwideclass behaving mad see   \n",
       "...                                                    ...   \n",
       "1599995                      woke school best feeling ever   \n",
       "1599996        thewdbcom cool hear old walt interviews â «   \n",
       "1599997                    ready mojo makeover ask details   \n",
       "1599998  happy th birthday boo alll time tupac amaru sh...   \n",
       "1599999  happy charitytuesday thenspcc sparkscharity sp...   \n",
       "\n",
       "                                                text_swlem  \\\n",
       "0          switchfoot awww bummer got david carr third day   \n",
       "1        upset update facebook texting cry result schoo...   \n",
       "2        kenichan dived many time ball managed save res...   \n",
       "3                          whole body feel itchy like fire   \n",
       "4                         nationwideclass behaving mad see   \n",
       "...                                                    ...   \n",
       "1599995                      woke school best feeling ever   \n",
       "1599996         thewdbcom cool hear old walt interview â «   \n",
       "1599997                     ready mojo makeover ask detail   \n",
       "1599998  happy th birthday boo alll time tupac amaru sh...   \n",
       "1599999  happy charitytuesday thenspcc sparkscharity sp...   \n",
       "\n",
       "                                          text_swlemdic  \\\n",
       "0                             bummer got carr third day   \n",
       "1        upset update cry result school today also blah   \n",
       "2                     many time ball save rest go bound   \n",
       "3                       whole body feel itchy like fire   \n",
       "4                                               mad see   \n",
       "...                                                 ...   \n",
       "1599995                   woke school best feeling ever   \n",
       "1599996                  cool hear old walt interview «   \n",
       "1599997                           ready mojo ask detail   \n",
       "1599998                      happy th birthday boo time   \n",
       "1599999                                           happy   \n",
       "\n",
       "                                            text_stemm  \n",
       "0                            bummer got carr third day  \n",
       "1        upset updat cri result school today also blah  \n",
       "2                    mani time ball save rest go bound  \n",
       "3                      whole bodi feel itchi like fire  \n",
       "4                                              mad see  \n",
       "...                                                ...  \n",
       "1599995                     woke school best feel ever  \n",
       "1599996                 cool hear old walt interview «  \n",
       "1599997                          readi mojo ask detail  \n",
       "1599998                     happi th birthday boo time  \n",
       "1599999                                          happi  \n",
       "\n",
       "[1600000 rows x 10 columns]"
      ]
     },
     "execution_count": 55,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "36ef5850-0467-4909-bf37-cb616b666aa7",
   "metadata": {},
   "source": [
    "## 8-keras tokenization -> dictionnaire -> sequence -> padding"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "id": "9fe21202-a713-4d86-b10b-6ff4061f467f",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "40222\n",
      "27.48770260810852\n",
      "(1600000, 25)\n"
     ]
    }
   ],
   "source": [
    "from keras.preprocessing.text import Tokenizer\n",
    "from keras.preprocessing.sequence import pad_sequences\n",
    "\n",
    "start = time.time()\n",
    "#initialisation de la tokenization avec les 2500 mots les plus fréquents\n",
    "#ca ne change pas la taille du dictionnaire mais ca impacte le nombre de colonne de X\n",
    "#2500 ca fait dansnotre cas des sequences de 25 mots\n",
    "#None donne des sequences de 52 mots\n",
    "\n",
    "tokenizer = Tokenizer(num_words=2500, lower=True,split=' ')\n",
    "\n",
    "#update du dictionnaire de vocabulaire de tokenization #obligatoire\n",
    "tokenizer.fit_on_texts(data['text_swlemdic'].values)\n",
    "#taille du dictionnaire\n",
    "vocab_size = len(tokenizer.word_index) + 1\n",
    "\n",
    "#transformation des docs en sequences d'entiers correpondant à un mot\n",
    "X = tokenizer.texts_to_sequences(data['text_swlemdic'].values)\n",
    "\n",
    "#padding des sequences de facons à ce qu'elles fassent toute la même taille\n",
    "#par defaut pad en pre (debut de sequence) mais on peut le mettre en post (fin de sequence)\n",
    "X = pad_sequences(X)\n",
    "print(vocab_size)\n",
    "print(time.time()-start)\n",
    "print (X.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "148552d0-5233-4c05-974a-08285eda6a6d",
   "metadata": {},
   "source": [
    "## 9- chargement des modèles préentrainés"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "id": "bea003f7-21dd-400b-ab01-ef3d85a3f041",
   "metadata": {},
   "outputs": [],
   "source": [
    "#on telecharge le dataset\n",
    "finput = Dataset.get_by_name(ws, 'glove50d dataset')\n",
    "loc=finput.download(target_path=None, overwrite=False, ignore_not_found=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "id": "d09ffacd-10a9-4c19-b041-0b0508740236",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "10.373645782470703\n"
     ]
    }
   ],
   "source": [
    "from gensim.models import KeyedVectors #pour lecture word emb\n",
    "\n",
    "#glove\n",
    "start = time.time()\n",
    "#glove50d = KeyedVectors.load_word2vec_format('embedding/glove.6B/glove.6B.50d.txt', binary=False, no_header=True)\n",
    "glove50d = KeyedVectors.load_word2vec_format(loc[0], binary=False, no_header=True)\n",
    "print(time.time()-start)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "id": "7b3ab6d0-e570-4baa-981a-8aec4c7b56a3",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.06283998489379883\n",
      "pct non matché: 0.27\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "# create a weight matrix for words in training docs\n",
    "def we2matrix(tokenizer, we_kv, we_dim, normalization=False):\n",
    "    start = time.time()\n",
    "    #taille du vocabulaire du keras tokenizer fitté sur nos datas\n",
    "    vocab_size = len(tokenizer.word_index) + 1\n",
    "    #initialisation de la matrice à 0\n",
    "    embedding_matrix = np.zeros((vocab_size, we_dim))\n",
    "    #compteur de mot non matché\n",
    "    count_nm=0\n",
    "    \n",
    "    #recuperation des vecteurs pour le dictionnaire du keras tokenizer fitté sur nos datas\n",
    "    for word, i in tokenizer.word_index.items():\n",
    "        try:\n",
    "            embedding_vector = we_kv.get_vector(word, norm=normalization)\n",
    "        except:\n",
    "            count_nm=count_nm+1\n",
    "            #on passe à la prochaine iteration de la boucle\n",
    "            continue\n",
    "        #on modifie la matrice que si on a un terme sinon ca reste à 0\n",
    "        embedding_matrix[i] = embedding_vector\n",
    "            \n",
    "    #pourcentage final de mots non matchés       \n",
    "    nm=count_nm/vocab_size\n",
    "    print(time.time()-start)\n",
    "    print('pct non matché:', \"{:.2f}\".format(nm))\n",
    "    return embedding_matrix\n",
    "\n",
    "glove50d_matrix=we2matrix(tokenizer, glove50d, 50, normalization=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c772332c-dc71-453d-b1a9-1ca040ebd800",
   "metadata": {},
   "source": [
    "## 10 - split train/test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "id": "95e03b10-3628-4a11-af84-aff3d46b64af",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "#creation des dummies pour le label/target\n",
    "Y = pd.get_dummies(data['target']).values\n",
    "\n",
    "#split 1.6M->1000 :test_size = 0.0006263\n",
    "X1, X2, y1, y2 = train_test_split(X,Y, test_size = 0.0006263, random_state = 36, stratify=Y)\n",
    "\n",
    "#split train / test sur les 10000\n",
    "X_train, X_test, y_train, y_test = train_test_split(X2,y2, test_size = 0.2, random_state = 36, stratify=y2)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9db6736e-b4e7-480f-9485-3a81515658fc",
   "metadata": {},
   "source": [
    "## 11- sauvegarde en binaries des dataset necessaires"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "id": "217f58e6-6d0d-46da-8c8b-91040b2a407f",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['./input/joblib\\\\glove50d_matrix.joblib']"
      ]
     },
     "execution_count": 81,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#on serialize en binaries\n",
    "import os\n",
    "from joblib import dump, load\n",
    "\n",
    "folder = './input/joblib'\n",
    "#creation du folder\n",
    "try:\n",
    "    os.mkdir(folder)\n",
    "except FileExistsError:\n",
    "    pass\n",
    "\n",
    "\n",
    "dump(X_train,os.path.join(folder, 'X_train.joblib'))\n",
    "dump(X_test,os.path.join(folder, 'X_test.joblib'))\n",
    "dump(y_train,os.path.join(folder, 'y_train.joblib'))\n",
    "dump(y_test,os.path.join(folder, 'y_test.joblib'))\n",
    "dump(glove50d_matrix,os.path.join(folder, 'glove50d_matrix.joblib'))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "20592072-2a94-464b-9ff2-df60e4d5d94a",
   "metadata": {},
   "source": [
    "## 12-upload des binaries vers le datastore du workspace"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "id": "a59a7cbd-5e01-4d11-a74c-be401e19d49e",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Validating arguments.\n",
      "Arguments validated.\n",
      "Uploading file to P7/train_input/\n",
      "Filtering files with pattern matching *.joblib\n",
      "Uploading an estimated of 5 files\n",
      "Uploading ./input/joblib/X_test.joblib\n",
      "Uploaded ./input/joblib/X_test.joblib, 1 files out of an estimated total of 5\n",
      "Uploading ./input/joblib/X_train.joblib\n",
      "Uploaded ./input/joblib/X_train.joblib, 2 files out of an estimated total of 5\n",
      "Uploading ./input/joblib/y_test.joblib\n",
      "Uploaded ./input/joblib/y_test.joblib, 3 files out of an estimated total of 5\n",
      "Uploading ./input/joblib/y_train.joblib\n",
      "Uploaded ./input/joblib/y_train.joblib, 4 files out of an estimated total of 5\n",
      "Uploading ./input/joblib/glove50d_matrix.joblib\n",
      "Uploaded ./input/joblib/glove50d_matrix.joblib, 5 files out of an estimated total of 5\n",
      "Uploaded 5 files\n",
      "Creating new dataset\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "{\n",
       "  \"source\": [\n",
       "    \"('workspaceblobstore', 'P7/train_input/')\"\n",
       "  ],\n",
       "  \"definition\": [\n",
       "    \"GetDatastoreFiles\"\n",
       "  ]\n",
       "}"
      ]
     },
     "execution_count": 82,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "target_path='P7/train_input/'\n",
    "Dataset.File.upload_directory('./input/joblib/', (default_ds, target_path), pattern='*.joblib', overwrite=True, show_progress=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3cb643ce-dfc6-4240-80ec-8738fb4f5699",
   "metadata": {},
   "source": [
    "## 13- chargement des inputs d'entrainement dans le blobstore et registration"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "id": "1ed6e4b5-e3bf-4c5e-8cd7-874c0850a627",
   "metadata": {},
   "outputs": [],
   "source": [
    "from azureml.core import Dataset\n",
    "#Create a filedataset from the path on the datastore (this may take a short while)\n",
    "fil_data_set_1 = Dataset.File.from_files(path=(default_ds, 'P7/train_input/X_train.joblib'))\n",
    "fil_data_set_2 = Dataset.File.from_files(path=(default_ds, 'P7/train_input/X_test.joblib'))\n",
    "fil_data_set_3 = Dataset.File.from_files(path=(default_ds, 'P7/train_input/y_train.joblib'))\n",
    "fil_data_set_4 = Dataset.File.from_files(path=(default_ds, 'P7/train_input/y_test.joblib'))\n",
    "fil_data_set_5 = Dataset.File.from_files(path=(default_ds, 'P7/train_input/glove50d_matrix.joblib'))\n",
    "\n",
    "\n",
    "# Register the file dataset\n",
    "try:\n",
    "    fil_data_set = fil_data_set_1.register(workspace=ws, \n",
    "                                        name='X_train.joblib',\n",
    "                                        description='LSTM small X_train',\n",
    "                                        tags = {'format':'joblib'},\n",
    "                                        create_new_version=True)\n",
    "except Exception as ex:\n",
    "    print(ex)\n",
    "try:\n",
    "    fil_data_set = fil_data_set_2.register(workspace=ws, \n",
    "                                        name='X_test.joblib',\n",
    "                                        description='LSTM small X_test',\n",
    "                                        tags = {'format':'joblib'},\n",
    "                                        create_new_version=True)\n",
    "except Exception as ex:\n",
    "    print(ex)\n",
    "try:\n",
    "    fil_data_set = fil_data_set_3.register(workspace=ws, \n",
    "                                        name='y_train.joblib',\n",
    "                                        description='LSTM small y_train',\n",
    "                                        tags = {'format':'joblib'},\n",
    "                                        create_new_version=True)\n",
    "except Exception as ex:\n",
    "    print(ex)\n",
    "try:\n",
    "    fil_data_set = fil_data_set_4.register(workspace=ws, \n",
    "                                        name='y_test.joblib',\n",
    "                                        description='LSTM small y_test',\n",
    "                                        tags = {'format':'joblib'},\n",
    "                                        create_new_version=True)\n",
    "except Exception as ex:\n",
    "    print(ex)\n",
    "try:\n",
    "    fil_data_set = fil_data_set_5.register(workspace=ws, \n",
    "                                        name='glove50d_matrix.joblib',\n",
    "                                        description='LSTM glove50d dico matrix',\n",
    "                                        tags = {'format':'joblib'},\n",
    "                                        create_new_version=True)\n",
    "except Exception as ex:\n",
    "    print(ex)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1d135ce9-77de-4485-af1a-8c82e1e8a0e5",
   "metadata": {},
   "source": [
    "## 14- creation d'une cible de calcul"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "id": "77ddaa99-5762-4cbe-bf6d-bca1b14f6d1f",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Creating.......................................................................\n",
      "Running\n"
     ]
    }
   ],
   "source": [
    "from azureml.core import Workspace\n",
    "#pour gerer les clusters\n",
    "from azureml.core.compute import AmlCompute\n",
    "#pour gerer une instance managéee dans ML\n",
    "#gère un environnement de développement ML optimisé basé sur le cloud dans Azure Machine Learning.\n",
    "from azureml.core.compute import ComputeInstance\n",
    "\n",
    "from azureml.core.compute import ComputeTarget\n",
    "\n",
    "# Load the workspace from the saved config file\n",
    "ws = Workspace.from_config()\n",
    "\n",
    "# Specify a name for the compute (unique within the workspace)\n",
    "compute_name = 'P7-CPU'\n",
    "\n",
    "# Define compute configuration\n",
    "#cluster comportant jusqu’à quatre nœuds basé sur l’image de machine virtuelle STANDARD_DS12_v2 \n",
    "# La priorité des machines virtuelles (VM) est définie sur Dedicated\n",
    "# l’alternative consiste à spécifier lowpriority, qui a un coût inférieur mais signifie que les machines virtuelles peuvent être préemptées \n",
    "#si une charge de travail d’une priorité plus élevée a besoin du calcul\n",
    "compute_config = ComputeInstance.provisioning_configuration(vm_size='Standard_DS3_v2')\n",
    "\n",
    "# Create the compute\n",
    "#Cette méthode crée une cible de calcul plutôt que d’en attacher une existante\n",
    "P7_cpu = ComputeTarget.create(ws, compute_name, compute_config)\n",
    "P7_cpu.wait_for_completion(show_output=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "id": "52871ebe-b1f2-4e90-b7c9-2bb807f5cd25",
   "metadata": {},
   "outputs": [],
   "source": [
    "#stop la becane\n",
    "P7_cpu.stop()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "f99e8830-ef7d-43df-86b9-720b1620cbeb",
   "metadata": {},
   "outputs": [],
   "source": [
    "#efface la becane\n",
    "P7_cpu.delete()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "id": "2f63ddd7-0156-470f-88f4-0c6b8d05a783",
   "metadata": {},
   "outputs": [],
   "source": [
    "#demarrer la becane\n",
    "P7_cpu.start()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0bc3f234-c71d-4792-a527-1db27cfc9845",
   "metadata": {},
   "source": [
    "### environnement\n",
    "* encapsulation of the environment where your machine learning training happens\n",
    "* Python packages, Docker image, environment variables, and software settings around your training and scoring scripts\n",
    "* They also specify runtimes (Python, Spark, or Docker)\n",
    "* You can either define your own environment, or use an Azure ML curated environment, These environments are backed by cached Docker images which reduces the run preparation cost"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "a4da095b-4eb9-41e2-9faa-9099bfbe044a",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'2.7.0'"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import tensorflow as tf\n",
    "tf.__version__"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 107,
   "id": "1f3edce1-821d-4ca4-b826-6556d90625b6",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'1.0.1'"
      ]
     },
     "execution_count": 107,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import joblib\n",
    "joblib.__version__"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "59251eab-cb0b-40fa-b62d-68c0d7b8250b",
   "metadata": {},
   "source": [
    "## 15- creation d'un script run configuration"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "52438d96-bea7-4803-b7fd-975faece700f",
   "metadata": {},
   "source": [
    "### a- Create a training folder"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "b6263f46-454e-4fe1-9adc-c297f077c04f",
   "metadata": {},
   "outputs": [],
   "source": [
    "import os, shutil\n",
    "from azureml.core import Dataset\n",
    "\n",
    "# Create a folder for the experiment files\n",
    "training_folder = 'P7-training'\n",
    "os.makedirs(training_folder, exist_ok=True)\n",
    "\n",
    "#on telecharge le dataset\n",
    "loc=Dataset.get_by_name(ws, 'glove50d_matrix.joblib').download(target_path=training_folder, overwrite=True, ignore_not_found=False)\n",
    "loc=Dataset.get_by_name(ws, 'y_test.joblib').download(target_path=training_folder, overwrite=True, ignore_not_found=False)\n",
    "loc=Dataset.get_by_name(ws, 'y_train.joblib').download(target_path=training_folder, overwrite=True, ignore_not_found=False)\n",
    "loc=Dataset.get_by_name(ws, 'X_test.joblib').download(target_path=training_folder, overwrite=True, ignore_not_found=False)\n",
    "loc=Dataset.get_by_name(ws, 'X_train.joblib').download(target_path=training_folder, overwrite=True, ignore_not_found=False)\n",
    "\n",
    "# Copy the data file into the experiment folder\n",
    "#shutil.copy('data/diabetes.csv', os.path.join(training_folder, \"diabetes.csv\"))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f0ac07af-14cb-4a22-b8f3-c55d6359e5a6",
   "metadata": {},
   "source": [
    "### b- Create and writing a training script"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "id": "18fec2ae-70ef-443c-afd3-9052aa7744ea",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Overwriting P7-training/train.py\n"
     ]
    }
   ],
   "source": [
    "%%writefile $training_folder/train.py\n",
    "\n",
    "import os\n",
    "from azureml.core import Run\n",
    "# Get the experiment run context\n",
    "run = Run.get_context()\n",
    "\n",
    "#recuperation des data\n",
    "from joblib import dump, load\n",
    "\n",
    "\n",
    "glove50d_matrix = load('glove50d_matrix.joblib', mmap_mode='r')\n",
    "y_test = load('y_test.joblib', mmap_mode='r')\n",
    "X_test = load('X_test.joblib', mmap_mode='r')\n",
    "y_train = load('y_train.joblib', mmap_mode='r')\n",
    "X_train = load('X_train.joblib', mmap_mode='r')\n",
    "\n",
    "######################\n",
    "\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Embedding, LSTM\n",
    "from keras import backend\n",
    "\n",
    "#clear keras session\n",
    "backend.clear_session()\n",
    "\n",
    "\n",
    "#taille du vocabulaire /dim du we/ we matrix\n",
    "vocab_size=glove50d_matrix.shape[0]\n",
    "embed_dim=50\n",
    "embedding_matrix=glove50d_matrix\n",
    "\n",
    "#taille du vecteur de sortie / nb de neurones\n",
    "lstm_out=50\n",
    "\n",
    "# define model\n",
    "model1 = Sequential()\n",
    "# couche embedding non entrainable\n",
    "#pas besoin de faire de drop out de regularisation a ce niveau la car on c'est deja pré-entrainé\n",
    "model1.add(Embedding(vocab_size, embed_dim, weights=[embedding_matrix], input_length=X_train.shape[1], trainable=False))\n",
    "\n",
    "#ajout d'une couche LSTM (mémoire) avec regularisation dropout\n",
    "model1.add(LSTM(lstm_out))#, dropout=0.1, recurrent_dropout=0.1))\n",
    "#fully connected de sortie 2D\n",
    "model1.add(Dense(2,activation='softmax'))\n",
    "#complilation\n",
    "model1.compile(loss = 'categorical_crossentropy', optimizer='adam',metrics = ['accuracy'])\n",
    "# summarize the model\n",
    "print('model1',model1.summary())\n",
    "\n",
    "\n",
    "from keras import callbacks\n",
    "callback1=callbacks.EarlyStopping(monitor=\"val_loss\",patience=5,mode=\"auto\",restore_best_weights=True)\n",
    "\n",
    "\n",
    "\n",
    "#Here we train the Network.\n",
    "print('Training a LSTM neural network')\n",
    "batch_size= 32\n",
    "\n",
    "\n",
    "model_info_1=model1.fit(X_train, y_train, batch_size =batch_size,validation_split=0.2, epochs = 50, callbacks=[callback1],  verbose = 2)\n",
    "print(model_info_1.history.keys())\n",
    "      \n",
    "      \n",
    "# Save the trained model in the outputs folder\n",
    "os.makedirs('outputs', exist_ok=True)\n",
    "dump(value=model_info_1.history, filename='outputs/model_info_1.history.joblib')\n",
    "\n",
    "\n",
    "      \n",
    "      \n",
    "run.complete()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d5273f42-b778-4c34-9eb8-493b6ef418e9",
   "metadata": {},
   "source": [
    "## 16-a run the training locally with azure ML\n",
    "\n",
    "* environnement\n",
    "    * encapsulation of the environment where your machine learning training happens\n",
    "    * Python packages, Docker image, environment variables, and software settings around your training and scoring scripts\n",
    "    * They also specify runtimes (Python, Spark, or Docker)\n",
    "    * You can either define your own environment, or use an Azure ML curated environment, These environments are backed by cached Docker images which reduces the run preparation cost"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 134,
   "id": "61ffeae2-617d-4458-92dd-9a71ed78e7d6",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'runId': 'lstm-train_1645566973_f3d523f6',\n",
       " 'target': 'local',\n",
       " 'status': 'Finalizing',\n",
       " 'startTimeUtc': '2022-02-22T21:56:15.742017Z',\n",
       " 'services': {},\n",
       " 'properties': {'_azureml.ComputeTargetType': 'local',\n",
       "  'ContentSnapshotId': 'bc56cd2b-2fb6-4951-9a01-f5c92bc4eb6f'},\n",
       " 'inputDatasets': [],\n",
       " 'outputDatasets': [],\n",
       " 'runDefinition': {'script': 'train.py',\n",
       "  'command': '',\n",
       "  'useAbsolutePath': False,\n",
       "  'arguments': [],\n",
       "  'sourceDirectoryDataStore': None,\n",
       "  'framework': 'Python',\n",
       "  'communicator': 'None',\n",
       "  'target': 'local',\n",
       "  'dataReferences': {},\n",
       "  'data': {},\n",
       "  'outputData': {},\n",
       "  'datacaches': [],\n",
       "  'jobName': None,\n",
       "  'maxRunDurationSeconds': 2592000,\n",
       "  'nodeCount': 1,\n",
       "  'instanceTypes': [],\n",
       "  'priority': None,\n",
       "  'credentialPassthrough': False,\n",
       "  'identity': None,\n",
       "  'environment': {'name': 'user-managed-env',\n",
       "   'version': 'Autosave_2022-02-22T10:43:38Z_38287da1',\n",
       "   'python': {'interpreterPath': 'python',\n",
       "    'userManagedDependencies': True,\n",
       "    'condaDependencies': {'channels': ['anaconda', 'conda-forge'],\n",
       "     'dependencies': ['python=3.6.2', {'pip': ['azureml-defaults']}],\n",
       "     'name': 'project_environment'},\n",
       "    'baseCondaEnvironment': None},\n",
       "   'environmentVariables': {'EXAMPLE_ENV_VAR': 'EXAMPLE_VALUE'},\n",
       "   'docker': {'baseImage': 'mcr.microsoft.com/azureml/openmpi3.1.2-ubuntu18.04:20220113.v1',\n",
       "    'platform': {'os': 'Linux', 'architecture': 'amd64'},\n",
       "    'baseDockerfile': None,\n",
       "    'baseImageRegistry': {'address': None, 'username': None, 'password': None},\n",
       "    'enabled': False,\n",
       "    'arguments': []},\n",
       "   'spark': {'repositories': [], 'packages': [], 'precachePackages': True},\n",
       "   'inferencingStackVersion': None},\n",
       "  'history': {'outputCollection': True,\n",
       "   'directoriesToWatch': ['logs'],\n",
       "   'enableMLflowTracking': True,\n",
       "   'snapshotProject': True},\n",
       "  'spark': {'configuration': {'spark.app.name': 'Azure ML Experiment',\n",
       "    'spark.yarn.maxAppAttempts': '1'}},\n",
       "  'parallelTask': {'maxRetriesPerWorker': 0,\n",
       "   'workerCountPerNode': 1,\n",
       "   'terminalExitCodes': None,\n",
       "   'configuration': {}},\n",
       "  'amlCompute': {'name': None,\n",
       "   'vmSize': None,\n",
       "   'retainCluster': False,\n",
       "   'clusterMaxNodeCount': None},\n",
       "  'aiSuperComputer': {'instanceType': 'D2',\n",
       "   'imageVersion': 'pytorch-1.7.0',\n",
       "   'location': None,\n",
       "   'aiSuperComputerStorageData': None,\n",
       "   'interactive': False,\n",
       "   'scalePolicy': None,\n",
       "   'virtualClusterArmId': None,\n",
       "   'tensorboardLogDirectory': None,\n",
       "   'sshPublicKey': None,\n",
       "   'sshPublicKeys': None,\n",
       "   'enableAzmlInt': True,\n",
       "   'priority': 'Medium',\n",
       "   'slaTier': 'Standard',\n",
       "   'userAlias': None},\n",
       "  'kubernetesCompute': {'instanceType': None},\n",
       "  'tensorflow': {'workerCount': 1, 'parameterServerCount': 1},\n",
       "  'mpi': {'processCountPerNode': 1},\n",
       "  'pyTorch': {'communicationBackend': 'nccl', 'processCount': None},\n",
       "  'hdi': {'yarnDeployMode': 'Cluster'},\n",
       "  'containerInstance': {'region': None, 'cpuCores': 2.0, 'memoryGb': 3.5},\n",
       "  'exposedPorts': None,\n",
       "  'docker': {'useDocker': False,\n",
       "   'sharedVolumes': True,\n",
       "   'shmSize': '2g',\n",
       "   'arguments': []},\n",
       "  'cmk8sCompute': {'configuration': {}},\n",
       "  'commandReturnCodeConfig': {'returnCode': 'Zero',\n",
       "   'successfulReturnCodes': []},\n",
       "  'environmentVariables': {},\n",
       "  'applicationEndpoints': {},\n",
       "  'parameters': []},\n",
       " 'logFiles': {'azureml-logs/60_control_log.txt': 'https://nabml2207185915.blob.core.windows.net/azureml/ExperimentRun/dcid.lstm-train_1645566973_f3d523f6/azureml-logs/60_control_log.txt?sv=2019-07-07&sr=b&sig=cWIsgiiddNA51yVRxuMhQXrzKhG4tjoYSitqk4JFAhY%3D&skoid=c871746f-ec9b-4687-9273-ffe3af167554&sktid=c211c6f1-15e1-48c9-81fb-6ff6d7d1569f&skt=2022-02-22T21%3A37%3A57Z&ske=2022-02-24T05%3A47%3A57Z&sks=b&skv=2019-07-07&st=2022-02-22T21%3A46%3A39Z&se=2022-02-23T05%3A56%3A39Z&sp=r',\n",
       "  'azureml-logs/70_driver_log.txt': 'https://nabml2207185915.blob.core.windows.net/azureml/ExperimentRun/dcid.lstm-train_1645566973_f3d523f6/azureml-logs/70_driver_log.txt?sv=2019-07-07&sr=b&sig=g%2BLOjYOjnVfHPowENlO03TcaPoBkihWZA4EEzvhHZhU%3D&skoid=c871746f-ec9b-4687-9273-ffe3af167554&sktid=c211c6f1-15e1-48c9-81fb-6ff6d7d1569f&skt=2022-02-22T21%3A37%3A57Z&ske=2022-02-24T05%3A47%3A57Z&sks=b&skv=2019-07-07&st=2022-02-22T21%3A46%3A39Z&se=2022-02-23T05%3A56%3A39Z&sp=r',\n",
       "  'logs/azureml/18292_azureml.log': 'https://nabml2207185915.blob.core.windows.net/azureml/ExperimentRun/dcid.lstm-train_1645566973_f3d523f6/logs/azureml/18292_azureml.log?sv=2019-07-07&sr=b&sig=h792r%2FAhz%2FZTpF%2BeXagQHD6eTrma4XJQow7%2FLqIg2Cg%3D&skoid=c871746f-ec9b-4687-9273-ffe3af167554&sktid=c211c6f1-15e1-48c9-81fb-6ff6d7d1569f&skt=2022-02-22T11%3A09%3A04Z&ske=2022-02-23T19%3A19%3A04Z&sks=b&skv=2019-07-07&st=2022-02-22T21%3A46%3A39Z&se=2022-02-23T05%3A56%3A39Z&sp=r'},\n",
       " 'submittedBy': 'nabil ouddane'}"
      ]
     },
     "execution_count": 134,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from azureml.core import Experiment, ScriptRunConfig, Environment\n",
    "#from azureml.widgets import RunDetails\n",
    "\n",
    "from azureml.core import Workspace, Environment\n",
    "\n",
    "ws = Workspace.from_config()\n",
    "\n",
    "\n",
    "#env = Environment.get(workspace=ws, name=\"AzureML-tensorflow-2.7-ubuntu20.04-py38-cuda11-gpu\")\n",
    "#si en local on prend l'environnement local\n",
    "env = Environment(\"user-managed-env\")\n",
    "env.python.user_managed_dependencies = True\n",
    "\n",
    "# Create a Python environment for the experiment (from a .yml file)\n",
    "#env = Environment.from_conda_specification(\"experiment_env\", \"environment.yml\")\n",
    "\n",
    "# Create a script config\n",
    "from azureml.core import ScriptRunConfig\n",
    "\n",
    "# compute_target=my_compute_target sur le cloud /  compute_target='local' en local\n",
    "script_config = ScriptRunConfig(source_directory=training_folder,\n",
    "                      script='train.py',\n",
    "                      compute_target='local',\n",
    "                      environment=env)\n",
    "\n",
    "# Set compute target\n",
    "# Skip this if you are running on your local computer\n",
    "#script_run_config.run_config.target = my_compute_target\n",
    "\n",
    "from azureml.core import Experiment\n",
    "\n",
    "# submit the experiment run\n",
    "experiment_name = 'lstm-train'\n",
    "experiment = Experiment(workspace=ws, name=experiment_name)\n",
    "run = experiment.submit(config=script_config)\n",
    "\n",
    "# Show the running experiment run in the notebook widget\n",
    "#RunDetails(run).show()\n",
    "\n",
    "# Block until the experiment run has completed\n",
    "run.wait_for_completion()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "d9130eba-f15e-4f01-aa8c-1455a23980e3",
   "metadata": {},
   "outputs": [],
   "source": [
    "from azureml.core import Experiment, ScriptRunConfig, Environment\n",
    "from azureml.core import Workspace, Environment\n",
    "from azureml.core.compute import ComputeTarget\n",
    "\n",
    "ws = Workspace.from_config()\n",
    "compute_name = 'P7-cpu'\n",
    "ComputeTarget(workspace=ws, name=compute_name).start()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fc380579-1deb-44b3-ab30-792402d11b59",
   "metadata": {},
   "source": [
    "## 17-a creation et registration d'un custom environment\n",
    "* probleme : obligé d'ajouter le pip joblib dans le dockerfile de l environnement dans l interface\n",
    "    * le conda_dep.add_pip_package(\"joblib==1.0.1\") ne semble rien apporter"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "09922a9e-f168-48b0-b34b-20fe18a22bd8",
   "metadata": {},
   "outputs": [],
   "source": [
    "#https://docs.microsoft.com/en-us/azure/machine-learning/how-to-use-environments#use-a-curated-environment\n",
    "\n",
    "from azureml.core.environment import Environment\n",
    "from azureml.core.conda_dependencies import CondaDependencies\n",
    "\n",
    "env = Environment.get(workspace=ws, name=\"AzureML-tensorflow-2.7-ubuntu20.04-py38-cuda11-gpu\")\n",
    "curated_clone = env.clone(\"customize_curated\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "c6fe9dee-b88a-4b04-ab36-bb76a714201d",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Installs numpy version 1.17.0 conda package\n",
    "#conda_dep.add_conda_package(\"joblib\")\n",
    "conda_dep = CondaDependencies()\n",
    "# Installs pillow package\n",
    "conda_dep.add_pip_package(\"joblib==1.0.1\")\n",
    "\n",
    "# Adds dependencies to PythonSection of myenv\n",
    "curated_clone.python.conda_dependencies=conda_dep"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "84ac64d1-43f5-41d2-8cd9-5909e228ee78",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<azureml.core.environment.ImageBuildDetails at 0x2cda470acd0>"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "curated_clone.register(workspace=ws)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0dde51a2-957b-4063-ace1-18e93443b698",
   "metadata": {},
   "source": [
    "## 17-b run the training sur une instance du cloud with azure ML"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 129,
   "id": "014ea945-e417-4f78-bc67-53492e679e76",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'runId': 'lstm-train_1645565607_a07b442f',\n",
       " 'target': 'P7-CPU',\n",
       " 'status': 'Finalizing',\n",
       " 'startTimeUtc': '2022-02-22T21:47:55.495189Z',\n",
       " 'services': {},\n",
       " 'properties': {'_azureml.ComputeTargetType': 'amlcompute',\n",
       "  'ContentSnapshotId': 'ce5303ad-fee6-4d0c-902e-2498cbaf57d5',\n",
       "  'ProcessInfoFile': 'azureml-logs/process_info.json',\n",
       "  'ProcessStatusFile': 'azureml-logs/process_status.json'},\n",
       " 'inputDatasets': [],\n",
       " 'outputDatasets': [],\n",
       " 'runDefinition': {'script': 'train.py',\n",
       "  'command': '',\n",
       "  'useAbsolutePath': False,\n",
       "  'arguments': [],\n",
       "  'sourceDirectoryDataStore': None,\n",
       "  'framework': 'Python',\n",
       "  'communicator': 'None',\n",
       "  'target': 'P7-CPU',\n",
       "  'dataReferences': {},\n",
       "  'data': {},\n",
       "  'outputData': {},\n",
       "  'datacaches': [],\n",
       "  'jobName': None,\n",
       "  'maxRunDurationSeconds': 2592000,\n",
       "  'nodeCount': 1,\n",
       "  'instanceTypes': [],\n",
       "  'priority': None,\n",
       "  'credentialPassthrough': False,\n",
       "  'identity': None,\n",
       "  'environment': {'name': 'customize_curated',\n",
       "   'version': 'Autosave_2022-02-22T21:33:28Z_252756a3',\n",
       "   'python': {'interpreterPath': 'python',\n",
       "    'userManagedDependencies': True,\n",
       "    'condaDependencies': {'name': 'project_environment',\n",
       "     'dependencies': ['python=3.6.2',\n",
       "      {'pip': ['azureml-defaults', 'joblib==1.0.1']}],\n",
       "     'channels': ['anaconda', 'conda-forge']},\n",
       "    'baseCondaEnvironment': None},\n",
       "   'environmentVariables': {'EXAMPLE_ENV_VAR': 'EXAMPLE_VALUE'},\n",
       "   'docker': {'baseImage': None,\n",
       "    'platform': {'os': 'Linux', 'architecture': 'amd64'},\n",
       "    'baseDockerfile': \"FROM mcr.microsoft.com/azureml/openmpi4.1.0-cuda11.1-cudnn8-ubuntu20.04:20220208.v1\\n\\nENV AZUREML_CONDA_ENVIRONMENT_PATH /azureml-envs/tensorflow-2.7\\n\\n# Create conda environment\\nRUN conda create -p $AZUREML_CONDA_ENVIRONMENT_PATH \\\\\\n    python=3.8 pip=20.2.4\\n\\n# Prepend path to AzureML conda environment\\nENV PATH $AZUREML_CONDA_ENVIRONMENT_PATH/bin:$PATH\\n\\n# Install pip dependencies\\nRUN pip install 'matplotlib~=3.5.0' \\\\\\n                'psutil~=5.8.0' \\\\\\n                'tqdm~=4.62.0' \\\\\\n                'pandas~=1.3.0' \\\\\\n                'scipy~=1.7.0' \\\\\\n                'numpy~=1.21.0' \\\\\\n                'ipykernel~=6.0' \\\\\\n                'azureml-core==1.38.0.post2' \\\\\\n                'azureml-defaults==1.38.0' \\\\\\n                'azureml-mlflow==1.38.0' \\\\\\n                'azureml-telemetry==1.38.0' \\\\\\n                'tensorboard~=2.7.0' \\\\\\n                'tensorflow-gpu~=2.7.0' \\\\\\n                'tensorflow-datasets~=4.5.0' \\\\\\n                'onnxruntime-gpu~=1.9.0' \\\\\\n                'joblib~=1.0.1' \\\\\\n                'horovod~=0.23.0'\\n\\n# This is needed for mpi to locate libpython\\nENV LD_LIBRARY_PATH $AZUREML_CONDA_ENVIRONMENT_PATH/lib:$LD_LIBRARY_PATH\",\n",
       "    'baseImageRegistry': {'address': None, 'username': None, 'password': None},\n",
       "    'enabled': False,\n",
       "    'arguments': []},\n",
       "   'spark': {'repositories': [], 'packages': [], 'precachePackages': True},\n",
       "   'inferencingStackVersion': None},\n",
       "  'history': {'outputCollection': True,\n",
       "   'directoriesToWatch': ['logs'],\n",
       "   'enableMLflowTracking': True,\n",
       "   'snapshotProject': True},\n",
       "  'spark': {'configuration': {'spark.app.name': 'Azure ML Experiment',\n",
       "    'spark.yarn.maxAppAttempts': '1'}},\n",
       "  'parallelTask': {'maxRetriesPerWorker': 0,\n",
       "   'workerCountPerNode': 1,\n",
       "   'terminalExitCodes': None,\n",
       "   'configuration': {}},\n",
       "  'amlCompute': {'name': None,\n",
       "   'vmSize': None,\n",
       "   'retainCluster': False,\n",
       "   'clusterMaxNodeCount': None},\n",
       "  'aiSuperComputer': {'instanceType': 'D2',\n",
       "   'imageVersion': 'pytorch-1.7.0',\n",
       "   'location': None,\n",
       "   'aiSuperComputerStorageData': None,\n",
       "   'interactive': False,\n",
       "   'scalePolicy': None,\n",
       "   'virtualClusterArmId': None,\n",
       "   'tensorboardLogDirectory': None,\n",
       "   'sshPublicKey': None,\n",
       "   'sshPublicKeys': None,\n",
       "   'enableAzmlInt': True,\n",
       "   'priority': 'Medium',\n",
       "   'slaTier': 'Standard',\n",
       "   'userAlias': None},\n",
       "  'kubernetesCompute': {'instanceType': None},\n",
       "  'tensorflow': {'workerCount': 1, 'parameterServerCount': 1},\n",
       "  'mpi': {'processCountPerNode': 1},\n",
       "  'pyTorch': {'communicationBackend': 'nccl', 'processCount': None},\n",
       "  'hdi': {'yarnDeployMode': 'Cluster'},\n",
       "  'containerInstance': {'region': None, 'cpuCores': 2.0, 'memoryGb': 3.5},\n",
       "  'exposedPorts': None,\n",
       "  'docker': {'useDocker': False,\n",
       "   'sharedVolumes': True,\n",
       "   'shmSize': '2g',\n",
       "   'arguments': []},\n",
       "  'cmk8sCompute': {'configuration': {}},\n",
       "  'commandReturnCodeConfig': {'returnCode': 'Zero',\n",
       "   'successfulReturnCodes': []},\n",
       "  'environmentVariables': {},\n",
       "  'applicationEndpoints': {},\n",
       "  'parameters': []},\n",
       " 'logFiles': {'azureml-logs/20_image_build_log.txt': 'https://nabml2207185915.blob.core.windows.net/azureml/ExperimentRun/dcid.lstm-train_1645565607_a07b442f/azureml-logs/20_image_build_log.txt?sv=2019-07-07&sr=b&sig=sxUlO04BcqWMb2%2FGrhZsBiothZqaAz3X%2BiiUsRCHGUc%3D&skoid=c871746f-ec9b-4687-9273-ffe3af167554&sktid=c211c6f1-15e1-48c9-81fb-6ff6d7d1569f&skt=2022-02-22T20%3A22%3A53Z&ske=2022-02-24T04%3A32%3A53Z&sks=b&skv=2019-07-07&st=2022-02-22T21%3A39%3A28Z&se=2022-02-23T05%3A49%3A28Z&sp=r',\n",
       "  'azureml-logs/55_azureml-execution-tvmps_2d2199304725aa11e0be162efeb7e3d744a3c93ee4d5880c51e3fc7979170f07_d.txt': 'https://nabml2207185915.blob.core.windows.net/azureml/ExperimentRun/dcid.lstm-train_1645565607_a07b442f/azureml-logs/55_azureml-execution-tvmps_2d2199304725aa11e0be162efeb7e3d744a3c93ee4d5880c51e3fc7979170f07_d.txt?sv=2019-07-07&sr=b&sig=E9w8VpBDpdL1C%2BPEWreQgLDL%2FQlkESWUn6tOqsCS8DI%3D&skoid=c871746f-ec9b-4687-9273-ffe3af167554&sktid=c211c6f1-15e1-48c9-81fb-6ff6d7d1569f&skt=2022-02-22T20%3A22%3A53Z&ske=2022-02-24T04%3A32%3A53Z&sks=b&skv=2019-07-07&st=2022-02-22T21%3A39%3A28Z&se=2022-02-23T05%3A49%3A28Z&sp=r',\n",
       "  'azureml-logs/65_job_prep-tvmps_2d2199304725aa11e0be162efeb7e3d744a3c93ee4d5880c51e3fc7979170f07_d.txt': 'https://nabml2207185915.blob.core.windows.net/azureml/ExperimentRun/dcid.lstm-train_1645565607_a07b442f/azureml-logs/65_job_prep-tvmps_2d2199304725aa11e0be162efeb7e3d744a3c93ee4d5880c51e3fc7979170f07_d.txt?sv=2019-07-07&sr=b&sig=HZWkYdstGFNDOK1vtgUAxwpPkpEm461cQnU8XR667kI%3D&skoid=c871746f-ec9b-4687-9273-ffe3af167554&sktid=c211c6f1-15e1-48c9-81fb-6ff6d7d1569f&skt=2022-02-22T20%3A22%3A53Z&ske=2022-02-24T04%3A32%3A53Z&sks=b&skv=2019-07-07&st=2022-02-22T21%3A39%3A28Z&se=2022-02-23T05%3A49%3A28Z&sp=r',\n",
       "  'azureml-logs/70_driver_log.txt': 'https://nabml2207185915.blob.core.windows.net/azureml/ExperimentRun/dcid.lstm-train_1645565607_a07b442f/azureml-logs/70_driver_log.txt?sv=2019-07-07&sr=b&sig=tAGIKJcj%2FPDZ6CE8Ilz5Kg5ZdFQoB54JxKPflX5l%2FkQ%3D&skoid=c871746f-ec9b-4687-9273-ffe3af167554&sktid=c211c6f1-15e1-48c9-81fb-6ff6d7d1569f&skt=2022-02-22T20%3A22%3A53Z&ske=2022-02-24T04%3A32%3A53Z&sks=b&skv=2019-07-07&st=2022-02-22T21%3A39%3A28Z&se=2022-02-23T05%3A49%3A28Z&sp=r',\n",
       "  'azureml-logs/75_job_post-tvmps_2d2199304725aa11e0be162efeb7e3d744a3c93ee4d5880c51e3fc7979170f07_d.txt': 'https://nabml2207185915.blob.core.windows.net/azureml/ExperimentRun/dcid.lstm-train_1645565607_a07b442f/azureml-logs/75_job_post-tvmps_2d2199304725aa11e0be162efeb7e3d744a3c93ee4d5880c51e3fc7979170f07_d.txt?sv=2019-07-07&sr=b&sig=Aa4x%2FIxey3BCNI3BROk3WfRQ%2BF%2BJo%2FJji90aCNe3q1s%3D&skoid=c871746f-ec9b-4687-9273-ffe3af167554&sktid=c211c6f1-15e1-48c9-81fb-6ff6d7d1569f&skt=2022-02-22T20%3A22%3A53Z&ske=2022-02-24T04%3A32%3A53Z&sks=b&skv=2019-07-07&st=2022-02-22T21%3A39%3A28Z&se=2022-02-23T05%3A49%3A28Z&sp=r',\n",
       "  'azureml-logs/process_info.json': 'https://nabml2207185915.blob.core.windows.net/azureml/ExperimentRun/dcid.lstm-train_1645565607_a07b442f/azureml-logs/process_info.json?sv=2019-07-07&sr=b&sig=R4Mb%2FW7MHDdI0uRn2SUg93tEsoEAb9KGYN8FTw1ftAY%3D&skoid=c871746f-ec9b-4687-9273-ffe3af167554&sktid=c211c6f1-15e1-48c9-81fb-6ff6d7d1569f&skt=2022-02-22T20%3A22%3A53Z&ske=2022-02-24T04%3A32%3A53Z&sks=b&skv=2019-07-07&st=2022-02-22T21%3A39%3A28Z&se=2022-02-23T05%3A49%3A28Z&sp=r',\n",
       "  'azureml-logs/process_status.json': 'https://nabml2207185915.blob.core.windows.net/azureml/ExperimentRun/dcid.lstm-train_1645565607_a07b442f/azureml-logs/process_status.json?sv=2019-07-07&sr=b&sig=sy6FzsCp4%2BXpr1tEroRC9Q%2FXOom19CV4kuvw%2B7f4amc%3D&skoid=c871746f-ec9b-4687-9273-ffe3af167554&sktid=c211c6f1-15e1-48c9-81fb-6ff6d7d1569f&skt=2022-02-22T20%3A22%3A53Z&ske=2022-02-24T04%3A32%3A53Z&sks=b&skv=2019-07-07&st=2022-02-22T21%3A39%3A28Z&se=2022-02-23T05%3A49%3A28Z&sp=r',\n",
       "  'logs/azureml/84_azureml.log': 'https://nabml2207185915.blob.core.windows.net/azureml/ExperimentRun/dcid.lstm-train_1645565607_a07b442f/logs/azureml/84_azureml.log?sv=2019-07-07&sr=b&sig=78kz34V66X%2BSstYh%2FcmnHtkMj%2BTE6BLhkCbrAiFZ1LA%3D&skoid=c871746f-ec9b-4687-9273-ffe3af167554&sktid=c211c6f1-15e1-48c9-81fb-6ff6d7d1569f&skt=2022-02-22T11%3A09%3A04Z&ske=2022-02-23T19%3A19%3A04Z&sks=b&skv=2019-07-07&st=2022-02-22T21%3A39%3A28Z&se=2022-02-23T05%3A49%3A28Z&sp=r',\n",
       "  'logs/azureml/job_prep_azureml.log': 'https://nabml2207185915.blob.core.windows.net/azureml/ExperimentRun/dcid.lstm-train_1645565607_a07b442f/logs/azureml/job_prep_azureml.log?sv=2019-07-07&sr=b&sig=k3wp9ASFNKeCa1Lvl36Hz6zBbL994yCMTL5r5CpIBMA%3D&skoid=c871746f-ec9b-4687-9273-ffe3af167554&sktid=c211c6f1-15e1-48c9-81fb-6ff6d7d1569f&skt=2022-02-22T11%3A09%3A04Z&ske=2022-02-23T19%3A19%3A04Z&sks=b&skv=2019-07-07&st=2022-02-22T21%3A39%3A28Z&se=2022-02-23T05%3A49%3A28Z&sp=r',\n",
       "  'logs/azureml/job_release_azureml.log': 'https://nabml2207185915.blob.core.windows.net/azureml/ExperimentRun/dcid.lstm-train_1645565607_a07b442f/logs/azureml/job_release_azureml.log?sv=2019-07-07&sr=b&sig=Lomh3ASsLLvpaS5ypi843n5SZ4N6quXNeanJT23J38U%3D&skoid=c871746f-ec9b-4687-9273-ffe3af167554&sktid=c211c6f1-15e1-48c9-81fb-6ff6d7d1569f&skt=2022-02-22T11%3A09%3A04Z&ske=2022-02-23T19%3A19%3A04Z&sks=b&skv=2019-07-07&st=2022-02-22T21%3A39%3A28Z&se=2022-02-23T05%3A49%3A28Z&sp=r'},\n",
       " 'submittedBy': 'nabil ouddane'}"
      ]
     },
     "execution_count": 129,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#https://docs.microsoft.com/en-us/azure/machine-learning/how-to-use-environments#use-a-curated-environment\n",
    "\n",
    "from azureml.core.environment import Environment\n",
    "from azureml.core.conda_dependencies import CondaDependencies\n",
    "#si en local on prend l'environnement local\n",
    "#env = Environment(\"user-managed-env\")\n",
    "#env.python.user_managed_dependencies = True\n",
    "\n",
    "# Create a Python environment for the experiment (from a .yml file)\n",
    "#env = Environment.from_conda_specification(\"experiment_env\", \"environment.yml\")\n",
    "\n",
    "env = Environment.get(workspace=ws, name=\"customize_curated\")\n",
    "# Create a script config\n",
    "from azureml.core import ScriptRunConfig\n",
    "\n",
    "# compute_target=my_compute_target sur le cloud /  compute_target='local' en local\n",
    "script_config = ScriptRunConfig(source_directory=training_folder,\n",
    "                      script='train.py',\n",
    "                      compute_target=ComputeTarget(workspace=ws, name=compute_name),\n",
    "                      environment=env)\n",
    "\n",
    "# Set compute target\n",
    "# Skip this if you are running on your local computer\n",
    "#script_run_config.run_config.target = ComputeTarget(workspace=ws, name=compute_name)\n",
    "\n",
    "from azureml.core import Experiment\n",
    "\n",
    "# submit the experiment run\n",
    "experiment_name = 'lstm-train'\n",
    "experiment = Experiment(workspace=ws, name=experiment_name)\n",
    "run = experiment.submit(config=script_config)\n",
    "\n",
    "# Show the running experiment run in the notebook widget\n",
    "#RunDetails(run).show()\n",
    "\n",
    "# Block until the experiment run has completed\n",
    "run.wait_for_completion()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 130,
   "id": "5e60d6fe-205c-4e73-ab6f-e8f217689369",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'lstm-train_1645565607_a07b442f'"
      ]
     },
     "execution_count": 130,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "run.id"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e737255e-c77a-43b3-9f9e-02e97ec0a035",
   "metadata": {},
   "source": [
    "## 16-b enregistrment de l'output en local et sortie de l'entrainement du lstm en local"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 135,
   "id": "f8eabff0-ac95-4d77-a043-3ab836ddd035",
   "metadata": {},
   "outputs": [],
   "source": [
    "run.download_file(name='outputs/model_info_1.history.joblib', output_file_path='P7-training/model_info_1.history.joblib')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 136,
   "id": "56211433-fb2f-4916-953d-9e9fb5abe361",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x576 with 4 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "result=load('P7-training/model_info_1.history.joblib', mmap_mode='r')\n",
    "\n",
    "plt.rcParams[\"figure.figsize\"] = [16, 8]\n",
    "plt.rcParams[\"figure.autolayout\"] = True\n",
    "\n",
    "f, axes = plt.subplots(2, 2)\n",
    "\n",
    "# Plot training & validation accuracy values\n",
    "axes[0,0].plot(result['accuracy'])\n",
    "axes[0,0].set_title('Train Model accuracy / ')\n",
    "axes[0,0].set_ylabel('Accuracy')\n",
    "axes[0,0].set_xlabel('Epoch')\n",
    "axes[0,0].legend(['glove50d'], loc='lower right')\n",
    "\n",
    "axes[0,1].plot(result['val_accuracy'])\n",
    "axes[0,1].set_title('Validation Model accuracy')\n",
    "axes[0,1].set_ylabel('Accuracy')\n",
    "axes[0,1].set_xlabel('Epoch')\n",
    "axes[0,1].legend(['glove50d'], loc='lower right')\n",
    "\n",
    "axes[1,0].plot(result['loss'])\n",
    "axes[1,0].set_title('Train Model loss')\n",
    "axes[1,0].set_ylabel('loss')\n",
    "axes[1,0].set_xlabel('Epoch')\n",
    "axes[1,0].legend(['glove50d'], loc='lower left')\n",
    "\n",
    "axes[1,1].plot(result['val_loss'])\n",
    "axes[1,1].set_title('Validation Model loss')\n",
    "axes[1,1].set_ylabel('loss')\n",
    "axes[1,1].set_xlabel('Epoch')\n",
    "axes[1,1].legend(['glove50d'], loc='upper left')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b0f40bf0-80c3-49b3-84f6-e0edd01e5e96",
   "metadata": {},
   "source": [
    "## 17-b sortie de l output de l'entrianement du LSTM sur le cloud"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 132,
   "id": "5d182f5b-1edb-43bc-a092-841d5bc39ecf",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABHgAAAI4CAYAAAARel4VAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjQuMywgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/MnkTPAAAACXBIWXMAAAsTAAALEwEAmpwYAADu2ElEQVR4nOzdd3zV9fXH8dchAcLeO2zCUnZYKoi4cI86cOOsWrXWVqvtr621dtra1lGtew/EUbQ4AQVBlIS9E0YgzAABwkjIOL8/7jf2GgkEyM3NTd7Px+M+cu93nvvlhntyvp9h7o6IiIiIiIiIiMSuGtEOQEREREREREREjo4KPCIiIiIiIiIiMU4FHhERERERERGRGKcCj4iIiIiIiIhIjFOBR0REREREREQkxqnAIyIiIiIiIiIS41TgERHM7EMzuybacRQzs05m5mYWX4Ztx5nZlxURl4iIiFQOQZ7QLXj+pJn9qizbHsF5rjCzT440zopmZveb2Stl3PZzM7sh0jGJSMVRgUckRpnZ7rBHkZntC3t9xeEcy93PcPcXjzCONWa238yal1g+N0ioOh3JceXwmdlwM5sZ7ThEREQOxcw+MrMHDrD8PDPbVJabPMXc/WZ3/105xPS9G0zu/qq7n3a0xz7AuUYF53q3xPJ+wfLPy/ucIlL1qcAjEqPcvX7xA1gLnBO27NXi7Q4nQToKq4HLws7ZB6hbAeet1Cro2oc7C5hUwecUERE5Ei8CV5qZlVh+FfCquxdEIaaKlgUMN7NmYcuuAVZEKZ5KIwo5lEiVoAKPSBUT3BHKNLOfm9km4Hkza2JmH5hZlpllB88Tw/b5tolucZcnM/trsO1qMzvjEKd9Gbg67PU1wEsl4mpkZi8FMWSY2f+ZWY1gXVxwvq1mtopQoaLkvs+a2UYzW29mD5pZXBmvx1vBncCdZjbNzI4JW1fHzP4WxLMzeN91gnUnmNlMM9thZuvMbFzJaxV+vcJeu5n9yMzSgLRg2T+DY+wys1QzGxG2fZyZ/cLMVppZTrC+vZk9bmZ/K/FeJprZTw7yds9EBR4REYkN7wHNgPDvxCbA2cBLZjbEzL4Kvoc3mtljZlbrQAcysxfM7MGw13cH+2wws+tKbHtW0Mp4V/DdfH/Y6mnBzx1Bi+jhB/ieP87MZgd5w2wzOy5s3edm9jszmxF8p39iJVo4l7A/uA5jg/3jgEuBV8M3OsQ5O5vZF8H5PgVKtqgeFpbPzDezUQeJJ3y/g15/MzvGzD41s+1mttnMflH8HkrJa77XOsq+n3/OMLO/m9k24H4z62pmU8xsm4VyxFfNrHHY/u3N7B0L5ZbbimMMYuoTtl1LM9trZi3K8t5FYpkKPCJVU2ugKdARuInQ7/rzwesOwD7gsYPsPxRYTihJ+AvwrNn37rCFmwU0NLNeQXIyFijZ//tRoBHQBTiRUEHo2mDdjYQSugFAMnBRiX1fAAqAbsE2pwFl7TP+IZAEtATm8N2k6a/AIOA4QtfrHqDIzDoG+z0KtAD6A/PKeD6A8wldw97B69nBMZoCrwFvmVlCsO4uQq2fzgQaAtcBewnd2bzM/lcEaw6cEuz/PWbWBmgFzD2MOEVERKLC3fcB4/nuDaJLgGXuPh8oBH5CKBcZDpwM3Hqo45rZGOBnwKmEvv9PKbHJnuCcjQndULrFzM4P1o0MfjYOWkR/VeLYTYH/Ao8QKk49DPzXvtsC53JC+U1LoFYQy8G8xP+uwenAImDDYZzzNSCV0HX6HaGbbMX7tgv2fZBQDvIz4O0yFjpKvf5m1gD4DPgIaEsoP5sc7FdaXlMWQ4FVhPKZ3wMG/DE4Ry+gPXB/EEMc8AGQAXQC2gFvuPt+4A3gyrDjXgZMdvesMsYhErNU4BGpmoqA37h7nrvvc/dt7v62u+919xxCX5onHmT/DHd/2t0LCRUaiosHB1PciudUYCmwvnhFWNHnPnfPcfc1wN8INcOGUEL3D3df5+7bCX2ZF+/bilCScKe773H3LcDfg+Mdkrs/F5wzj1BS0M9CLYJqEEo6fuzu69290N1nBttdDnzm7q+7e35w/eaV5XyBP7r79iB5xd1fCY5R4O5/A2oDPYJtbwD+z92Xe8j8YNtvgJ2EEiqC9/u5u28u5ZxnAh+5ux9GnCIiItH0InBR2E2Pq4NluHuqu88KvjvXAP/m4LlLsUuA5919kbvvISgIFHP3z919obsXufsC4PUyHhdCBaE0d385iOt1YBlwTtg2z7v7irACVv+DHdDdZwJNzawHoff/UolNSj2nmXUABgO/CnK+acD7YfteCUxy90nB+/0USCGUMxzUIa7/2cAmd/+bu+cGedbXwboD5jWHOl9gg7s/Gpxzn7unu/unwXvLIlTcKo5hCKHCz91Bfpjr7sUtrYpvkhXfnLyKUJ4qUuWpwCNSNWW5e27xCzOra2b/tlBXpF2EmiA3ttK7OW0qfuLuxXdd6h/inC8TKoyM4/vJSXOgJqG7LMUyCN1tgdAX9LoS64p1DPbdGDQT3kEoyWh5iHiKmwn/KWgmvAtYExZPcyABWHmAXduXsryswt8LZvYzM1saNK3eQaglU3ET6oOd60X+dwfqSg6enKh7loiIxJTgD/KtwPlm1pXQH+2vAZhZdwt1Kd8UfIf/gRLdj0pxsJwCMxtqZlODbj07gZvLeNziY2eUWBaez0BYDkWo5cqh8icIfb/fBpwEvFti3cHO2RbIDgpZ4euKdQQuLs6fghzkBEI37g7qENf/YLnL0eRQJfOnVmb2hoW65+8i1Do8PIaMA43VFBSb9gKjzKwnoRZGE48wJpGYogKPSNVUshXHTwm1GBnq7g35XxPkg3W7OrwTumcQGmz5TOCdEqu3AvmEEo1iHfhfK5+NhL6ow9cVWwfkAc3dvXHwaOjux3BolwPnEWqe3YhQE14Ive+tQC7Q9QD7rStlOYSadocPIN36ANt8e/0tNN7OPYTuKDZx98aEWuYUX/uDnesV4Dwz60eoafJ7B9rIzGoSuqP1aSnHERERqayKuyhdCXwc1lL1CUItVZKC3OUXlC1vOVhOAaEC0kSgvbs3Ap4MO+6hWsFu4Lu5TPHx1x9g28PxMqHuT5PCbqyV5ZwbgSZmVq/EumLrgJfD8qfG7l7P3f9UhpgOdv3XEepyfyCl5TXFRaiD5VAlr/8fgmV9ghiuLBFDByt9MObim2RXARPCb3yKVGUq8IhUDw0IjbuzI+jL/ZsIned6YHSJO0kEXb3GA783swbBGDd38b9xesYDd5hZooUGWLw3bN+NwCfA38ysoZnVCAbdK0tz6gaEikPbCCUUfwg7bhHwHPCwmbUNWvsMN7PahMbpOcXMLjGzeDNrZmb9g13nARcGraK6Be/5UDEUEJopI97Mfk2oT3qxZ4DfmVmShfQt7lfv7pmExu95GXi7uMvXAZwALHD3XWW4JiIiIpXJS4RuxNxI0D0r0ADYBewOWmHcUsbjjQfGmVlvM6vL93OeBsB2d881syGEbgYVyyLUzb204sUkoLuZXR7kB5cSGm/vgzLGdkDuvprQjZpfHs45g5trKcBvg8GFT+C73cVeIdSV6/Qgz0mw0GQcid8/zfcc7Pp/ALQxszvNrHaQ2w0N1h0wrwm6WK0nNHNanIUGvy7tBld4DLuBncF4QneHrfuGUIHrT2ZWL3hvx5d47xcQKvKUbFkuUmWpwCNSPfwDqEOo1cosQoPilTt3X+nuKaWsvp3Q3ZtVwJeE7qA9F6x7GvgYmE9oIOSSLYCuJjRQ4RIgG5hAGZoXE/pCzyCUUCwh9N7D/QxYSKiIsh34M1DD3dcSaon002D5PKBfsM/fCc16sZlQIvoqB/cxoeu9Iogll+82QX6YUDL6CaFE6llC/1bFXgT6cPDuWZoeXUREYlIwvstMoB7f7UbzM0LFlxxCecKbZTzeh4TynilAevAz3K3AA2aWA/ya0Hdw8b57CY1TOCPo0jSsxLG3ERp/5qeEbh7dA5zt7lvLEtsh4v7S3TccYPmhznk5ocGJtxMqZr0Utu86Qi2Zf0GoeLWOUJGkLH8Dlnr9PTSe46mEikmbCM0aelKw+mB5zY3B+bcBxxD6dz+Y3wIDCbV8/i9h+WFw8/AcQt2v1gKZhGYgC3/vcwi1AJpehvcrUiWYxuMUEam8zGwkobtQHUsbQNnMlgAXufuSCg1OREREpJIys+cIDdz8f9GORaSilNZnUUREoiwYW+fHwDMHKe7UAl5ScUdEREQkxMw6ARcCA6IcikiFUhctEZFKyMx6ATsIdUX7R2nbufv+Mg6WKCIiIlLlmdnvgEXAQ8H4RiLVhrpoiYiIiIiIiIjEOLXgERERERERERGJcVVmDJ7mzZt7p06doh2GiIiIRFlqaupWd28R7TjKm3IdERERgdJznSpT4OnUqRMpKaXNziwiIiLVhZllRDuGSFCuIyIiIlB6rqMuWiIiIiIiIiIiMU4FHhERERERERGRGKcCj4iIiMgRMLMxZrbczNLN7N5StrnEzJaY2WIzey1seQcz+8TMlgbrO1VY4CIiIlIlVZkxeEREREQqipnFAY8DpwKZwGwzm+juS8K2SQLuA45392wzaxl2iJeA37v7p2ZWHyiqwPBFRESkClILHhEREZHDNwRId/dV7r4feAM4r8Q2NwKPu3s2gLtvATCz3kC8u38aLN/t7nsrLnQRERGpilTgERERETl87YB1Ya8zg2XhugPdzWyGmc0yszFhy3eY2TtmNtfMHgpaBH2Pmd1kZilmlpKVlVXub0JERESqDhV4RERERCIjHkgCRgGXAU+bWeNg+QjgZ8BgoAsw7kAHcPen3D3Z3ZNbtGhRASGLiIhIrFKBR0REpBJzd/buL4h2GPJ964H2Ya8Tg2XhMoGJ7p7v7quBFYQKPpnAvKB7VwHwHjAw8iGLiEhVkJObT1GRRzsMqYRU4BEREamEioqcjxdv4uxHv+SY33zMHa/PJW1zTrTDkv+ZDSSZWWczqwWMBSaW2OY9Qq13MLPmhLpmrQr2bWxmxU1yRgNLEBEROYh12/dy3zsLGPi7T7np5VTyCgqjHZJUMhEt8Bxq+tBgitCpQf/zBWZ2Zti6+4L9lpvZ6ZGMU0REpLIoKnImLdzImY9M54cvp7Inr4DLh3Tgs6WbOe0f0/jRq3NYtmlXtMOs9oKWN7cBHwNLgfHuvtjMHjCzc4PNPga2mdkSYCpwt7tvc/dCQt2zJpvZQsCApyv+XYiISCxYvXUPP3trPqP++jlvp65nVI+WfLZ0Mz98OZXcfBV55H/MPTJNu4LBAlcQNn0ocFmJ6UOfAua6+xPBjBKT3L1T8Px1QjNUtAU+A7oHCdEBJScne0pKSkTei4iISKQVBoWdR6eksWLzbrq0qMfto7txTt+2xMfVYPue/Tz35WpemLmG3XkFjDmmNbef3I1j2jaKduiVjpmluntytOMob8p1RESql/QtOTw2JZ2J8zdQM64GVwztyA9P7EKrhgm88c1a7nt3ISd0a85TVyVTp9YBx+qXKqq0XCc+guf8dvrQIIDi6UPDmyA70DB43gjYEDw/D3jD3fOA1WaWHhzvqwjGKyIiUuEKCov4YEGosLMyaw9JLevzyGUDOKtPG+Jq2LfbNa1Xi5+d3oMbR3ThuRmreW7Gaj5avIlTerXijpO70TexcfTexGEqvrlkZofYUkREpPpZunEXj01JZ9KijdSpGceNI7pww4gutGhQ+9ttxg7pQHxcDe6eMJ/rXpjNs+OSqVsrkn/eSyyI5CfgQNOHDi2xzf3AJ2Z2O1APOCVs31kl9i059ShmdhNwE0CHDh3KJWgREZGKUFBYxH/mbeCxqems3rqHHq0a8PjlAznj2NbUqFF64aNR3Zr85NTuXHdCZ16cuYZnv1zNuY/N4KQeLbjj5CQGdGhSge+i7IqKnLnrdjBp4UY+WrSJZ8cl07N1w0PvKCIiUk0sWr+TRyan8cmSzdSvHc+to7py/QldaFqv1gG3v2hQIvE1jLvGz2Pcc7N57trB1K+tIk91Fu1//cuAF9z9b2Y2HHjZzI4t687u/hTwFISaLUcoRhERkXKTX1jEu3PW89jUdNZu30uvNg158sqBnNb74IWdkhrVqckdJydx7fGdeOmrDJ6ZvooL/jWTkd1b8OOTuzGoY9MIvouyKSxyUjOyvy3qbNqVS624GoxIak5Bob62RUREAOauzebRKelMWbaFhgnx/PjkJK47vjON6tY85L7nD2hHXA3jzjfncfWzX/PCdUNomHDo/aRqimSBpyzTh14PjAFw96/MLAFoXsZ9RUREYsb+giLenpPJ41PTyczeR592jXj66mRO6dXyqLoqNUioyY9O6sY1x3XilVkZPD1tFT944iuO79aMH5/cnSGdK7bQU1jkfLN6e6ios3gTWTl51IqvwajuLbi3T09G92qpxFNERAT4ZvV2Hp2SxvS0rTSuW5O7T+/BVcM7Hvb35Dn92hJfw7j99blc9ew3vHTdEBrV0XdtdRTJQZbjCQ2yfDKh4sxs4HJ3Xxy2zYfAm+7+gpn1AiYT6orVG3iN/w2yPBlI0iDLIiISa/IKChmfkskTU9PZsDOXfu0bc+fJSYzq0SIiY9Ds3V/Aa1+v5ckvVrF1dx7DujTljpOTGN6lWcTGvCkoLGLWqu1MWrSRTxZvYuvu/STUrMFJPVpyRp82jO7ZskKbjGuQZRERqazcna9WbuORKWnMWrWd5vVrceOILlw5rCP1jvK78tMlm7n11VR6tG7AK9cPpXHdA3ftkthXWq4TsQJPcNIzgX8AccBz7v57M3sASHH3icFsWU8D9QkNuHyPu38S7PtL4DqgALjT3T882LmU9IiISGWSm1/Im7PX8cTnK9m0K5eBHRrz41O6MzKpeYUMLpybXxgUelayJSePwZ2acMfJSZzQrXzOn19YxIz0rXy4cBOfLNlE9t586taKY3TPlpzZpw2jerSI2mCPKvCIiEhl4+5MS9vKI5PTSM3IpmWD2vzwxK5cPqRDuc6ANXXZFn74SipdW9Tn1RuGljp+j8S2qBR4KpKSHhERqQz27S/ktW/W8u+gsDKkU1N+fEoSx3WNXAuag8nNL2R8SqjQtHFnqNB0x8lJnNj98FsQ5RUUMiN9K5MWbuLTJZvZuS+f+rXjOblXqKhzYvcWJNSM/jStKvCIiEhl4e5MXrqFR6ekMT9zJ20bJXDzqK5cktw+Yt+Z01ZkceNLKXRqVo9XbxxK8/q1D72TxBQVeERERCJo7/4CXpmVwVPTVrF1936Gd2kW6hrVtVm0QwNCxZkJqZn8a+pK1u/YR7/ERtxxchKjex58DKDc/EKmrcjiw0Wb+GzJZnLyCmiQEM+pvVtx5rFtOCGpeaUo6oRTgUdERKKtqMj5ZMkmHp2SzuINu0hsUocfndSNHwxMpFZ8jYiff2b6Vq5/MYV2Terw2g1DadkwIeLnlIqjAo+IiEgE7M4r4OWvMnh6+iq279nPiKTm3D46qcIHNy6r/QVFvDMnk8c/T2fd9n0c264hd4xO4tTerb4t9OzbX8gXK7YwaeEmJi/dzJ79hTSuW5PTerfijD5tOL5r8wpJTo+UCjwiIhIthUXOpIUbeWxKOss359CpWV1+dFI3zh/QjppxFfvd+fWqbVz7wmxaN0zgtRuH0bqRijxVhQo8IiIi5Wj7nv28/s1anp6+ih178zmxewvuODmJQR2bRDu0MskvLOK9uaHp2jO2haZrv3hQIqkZ2UxZtoV9+YU0rVeL049pxRnHtmF412YVnpgeKRV4RCq3ddv3kpKxPdphlKvm9WtzfNfm1KhR8V1x5eDcnS/Tt7J1d17Ez7VrXwEvfbWGlVl76NayPred1I2z+7YhPorfn6kZ27nmudk0q1+L124cRrvGdaIWS7RlZu+lqAg6NKsb7VCOmgo8IiIiRykrJ4+PF2/iw0UbmbVqO4VFzsk9W3L7yUn0b9842uEdkYLCIibO38BjU9JZtXUPzevXZsyxoe5XQzo3jWpSeqRU4BGpvBZm7uTKZ79m5778aIdS7nq0asBto7txZp82xKnQU2m8/s1a7ntnYYWdr2fr0OfgjGMrz+dg7tpsrn7uGxrVqcnrNw6jfdPYL3Acrjlrs7nm2W+oUyuOafecVOm6lx8uFXhERESOwOZduXy8eBOTFm7km9XbKXLo0rweZ/Zpw9n92tCzdcNoh1guCouc1Vv30Ll5vUqTkB4pFXhEKqfwPzIfv3wgjerUjHZI5Wbeuh08NjWd9C276dqiHreN7sY5fdvGZJG8Klm3fS9j/jGNfu0b84cL+kT8fHE1jHaN61TKllwLMndw1bPfUK9WHK/fNIyOzepFO6QKk7JmO+Oen02dWnFk5eTxq7N7c/0JnaMd1lFRgUdERKSMNuzYx0eLQi11UjKycYfurepzxrFtOLNPG7q3qh+VGbGkbFTgEal8iv/AalqvFq/fVDW7iRQWOR8uCo29smxTaOyVW0/qxgVRGHtFQoMcj316Fks27OKjO0eQ2KT6tVopafGGnVz5zNfUjo/jtRuH0qVF/WiHFHGzVm3jurBxiO4aP48Vm3OYds9J1K0VH+3wjpgKPCIiIgeRmb2XDxduYtKijcxduwMINbM+q08bzujTmm4tG0Q3QCkzFXhEKpfigV5bNUzg9Wow0GtRkfPp0s08OiWNRetDsyfdMqorFw1KpHZ8bHcLiSXPfrma332whL9c1JdLkttHO5xKY9mmXVzx9NfUqGG8fuPQKp3fzEjfyvUvziaxSV1eu3EoLRskkJqxnR888RX3ntGTm0/sGu0Qj5gKPCIiIiVkbNvDh4s28eHCjczP3AnAse0acsaxbTjj2NbV4s5WVaQCj0jlUTxVc9vGoeJOdZqq2d2ZunwLj0xOZ966HbRplMDNJ3bl0sHtY378j8oufctuznpkOid0a84z1ySr1W0JaZtzuOzprwHn1RuG0aN11SvyfLEii5teSqFz83q8csNQmtev/e26a577hgWZO5j+89HUrx2brXhU4BEREQFWZe3mw0WhMXUWb9gFQL/ERpzRpw1nHtumSsysUN2pwCNSOUxbkcWNL6XQqVnoD6wWDWofeqcqqHgWp0cmpzF7TTYtGtTmhyO7cMXQjtSppUJPeSsoLOIHT35FxrY9fHLnyGpVVDwcK7N2c/nTs9hfUMSrNwyjd9uqMaYgwJRlm7n55Tl0a1mfV24YStN6tb6zfv66HZz3+Ax+dlp3bhudFKUoj44KPCIiUm2lbc5h0sLQmDrLNuUAMLBDY87s04Yxx7ZWv/wqRgUekeibumwLP3wllS7N6/HqDUNpVr96FnfCuTuzVm3n0SlpzFy5jWb1anHDiC5cNbxjzLYiqIwem5LGXz9ZwaOXDeCcfm2jHU6ltmbrHi5/ehZ78wt55fqhHNuuUbRDOmqfLN7Ej16bQ8/WDXn5+iE0rlvrgNvd8OJsvlm9nek/Hx2TA76rwCMiItWGu7M8KOpMWriR9C27MYPBHZtyRp/WjDm2NW0aVb0BPiVEBR6R6Pp0yWZufTWVHq0b8PJ1Q2lS78B/YFVnKWu288iUdKatyKJx3Zpcf3xnrjm+Ew0TYu8Pzcpk8YadnP/4DE4/pjWPXT4w2uHEhHXb9zL2qVnk5Obz0vVD6d++cbRDOmIfLtzI7a/P5dh2jXjxuiEHLdws3rCTsx75kjtOTuKuU7tXYJTlQwUeERGp8jKz9zIhNZOJ8zawauseahgM6dyUM/u04fRjWtNKzbSrBRV4RKLno0Ubue21uRzTtiEvXTeURnVVsDiYeet28OjkNCYv20KDhHiuPa4T153QudRWB1K6vIJCzntsBlt37+fTn4xUYfEwZGbv5fKnvyZ7z35euG4Igzo2iXZIh+39+Ru488159G/fmBeuHUyDMhRLb3kllelpW/ny5yfF3O+cCjwiIlIl5eYX8vHiTbyVksmMlVsBGNa5GWf1DRV1quuYD9WZCjwi0VH8B1a/xEa8cN0QtUY5DIvW7+SxKel8tHgT9WrFcfVxnbjhhM7q2nYY/vLRMv71+UqevSaZk3u1inY4MWfjzn1c/vTXbNmVy/PXDmFI56bRDqnM3p2byU/Hzye5U1OeHzeYemXs8rh8Uw5j/jmNW07syj1jekY4yvKlAo+IiFQZ7s7iDbsYn7KO9+auZ1duAe0a1+Hi5EQuGpSoMXWqORV4RCree3PXc9f4eSR3bMpz1w7WmDJHaNmmXTw2JZ3/LtxIQnwcVw7rwI0ju9CygVqgHsyctdlc9MRMLhqUyF8u6hftcGLW5l25XP70LDbsyOXZcckc17V5tEM6pPEp6/j52wsY3qUZz1yTTN1ah/d/z+2vz2Xy0s1Mv+ekmCqoqsAjIiIxL3vPft6bt57xKZks3biLWvE1OOPY1lyS3J7hXZpRo4amQRUVeEQq2oTUTO6eMJ9hnZvx7LjD/wNLvi99Sw6PT13Jf+atp2ZcDS4b0oEfnthF48cdwL79hZz1yHTyCor46M4RZeqaI6XLysnjimdmsXb7Xp6+OpkRSS2iHVKpXvt6Lb94dyEjkprz1FXJRzQr3cqs3Zz68Bdcf0JnfnlW7whEGRkq8IiISEwqLApNLzs+ZR2fLt7M/sIi+iY24uLk9pzbt63Gd5DvUYFHpOK88c1a7nt3ISd0O/I/sKR0a7bu4V+fp/POnPXUMOPi5ERuGdVVLVXD3D9xMS/MXMNrNwzluG6Vv8VJLNi2O48rnvmaVVv38OD5x3LhgHbEx9WIdljf8dJXa/j1fxZzUo8WPHHlIBJqHvn/PXeNn8d/F2xk+j0n0TJGxmtUgUdERGLK2m17eSt1HRNSM9m4M5cmdWty/oB2XDyoPb3bNox2eFKJqcAjUjFenpXBr95bxKgeLXjyKP/AkoNbt30vT3yxkrdS1uEOFw5sx49O6kbHZvWiHVpUzVy5lcuf/ppxx3Xi/nOPiXY4VUr2nv1c/+Js5qzdQadmdbn1pG5cMKAdNStBoee5L1fzwAdLOKVXKx6/YgC144/u/541W/dw8sNfcNWwjjHzOVKBR0REKr19+wv5aPFGxs/O5KtV26hhMLJ7Cy5Jbs/JvVoe9Re4VA8q8IhE3vMzVvPb95dwSq+WPH7FQP3/XEE27NjHv79Yyeuz11FY5JzXry0/Gt2Nri3qRzu0CpeTm8+Yf0ynVnwNJt0xQq3HIqCoyPl06WYenZLGovW7SGxSh1tHdeMHg9pF7Xf+qWkr+cOkZYw5pjWPXDaAWvHlU3D6+YQFvDt3PV/cMyomukKqwCMiIpWSuzM/cyfjU9bx/rwN5OQV0KFpXS5JTuQHgxJj4ktWKhcVeEQi6+lpq/j9pKWcfkwrHr1sYLn9gSVlt2VXLk9NW8UrX2eQV1DE2X3bcvvobnRv1SDaoVWYn09YwFup63jr5uEM6hg7Mz7FIndn6vItPDI5nXnrdtCmUQI3n9iVSwe3r9CWe49PTeehj5dzdt82/P3S/uXammjd9r2M/tvnXDq4PQ+e36fcjhspKvCIiEilsm13Hu/OXc9bKZks35xDQs0anNmnDZckt2dIp6YaMFmOmAo8IpFT/AfWWX3a8I+x5fsHlhy+rbvzeGb6al7+ag179hdyxrGtuW10N45p2yjaoUXUlGWbue6FFG4+sSv3nhFb01vHMvfQuIiPTE5j9ppsWjaozU0ju3DF0I4RbUHl7vxzchr/+CyN8/u35a8X94vImEC/fHch41PWMfVnoyr9OFcq8IiISNQVFBYxLS2L8bMz+WzpZgqKnP7tG3Pp4Pac1bcNDTXzhZQDFXhEIuOfn6Xx989WcF7/tvwtQn9gyZHJ3rOf52as5oUZa8jJK+CUXi25fXQS/do3jnZo5S57z35O+8c0mtatxcTbj1f3wChwd2at2s4jk9P4atU2mtevxQ0junDVsI7Uq12+s+i5O3/7ZAWPTU3nokGJ/PkHfYmL0E3AjTv3ceJDn3NB/3b8+aK+ETlHeVGBR0REvqeoyHl6+ipmr8mugLM5C9fvZPOuPJrVq8WFA9txcXL7atWcXCqGCjwi5cvdefjTFTw6JZ0LB7bjoYv6RewPLDk6O/fl88KMNTw3YzU79+VzYvcW3HFytyrVhen21+fy4cKNvPej4zm2XdVuqRQLUtZs55Ep6UxbkUXjujW54YTOXH1cp3K5aefu/OmjZfz7i1WMHdyeP1zQJ+ItvO+fuJiXZ2Uw+a4T6dS88g5irgKPiIh8x568An7y5jw+WbKZbi3rU6sC7sS2a1KHiwYlMrpnSzXrl4hRgUek/Lg7f/5oOU9+sZJLk9vzxwsj/weWHL2c3HxenpXBM9NXs33Pfo7v1ozbRycxrEuzaId2VD5YsIHbXpvLXad2546Tk6IdjoSZt24Hj05OY/KyLTRMiGfc8Z257vhONK5b64iO5+787oOlPDdjNVcO68AD5x5bIf/3bMnJZeRfpnJmnzY8fEn/iJ/vSKnAIyIi31q3fS83vpRC2pbd/OqsXlxzXCfMlLBL1aACj0j5cHd+/9+lPPPlaq4Y2oHfnVcxf2BJ+dm7v4BXZ63l39NWsXV3HkM6N+WO0Ukc361ZzH3vb8nJ5fS/T6ND07q8fctx6iJYSS1av5NHp6Tx8eLN1K8dz9XDO3LDiC40rVf2Qk9RkXP/+4t56asMrj2+E78+u3eFfl5//98lPPvlaj75yYl0a1k5Z6grLdfRb4WISDUza9U2znt8Bht27OPFa4cw7vjOMZfkiYhIZLk7v31/Cc98uZpxx3XiwfNV3IlFdWvFc+PILnz585O4/5zerN22lyuf/ZoLn5jJ1OVbiJWb/e7OL95ZxJ79hfztEo3/VJkd264R/74qmY/uHMGoHi144ouVHP+nKfxh0lK25OQecv+iIueX7y3ipa8yuGlklwov7gDcfGJXEmrG8c/JaRV63vKg3wwRkWrk1a8zuPKZr2lStyb/ue0ETkhqHu2QRGKamY0xs+Vmlm5m95ayzSVmtsTMFpvZayXWNTSzTDN7rGIiFjm0oiLn/95bxAsz13DDCZ35zTkV/weWlK+EmnGMO74zX9wzigfPP5Ytu/K49vnZnPf4DD5ZvKnSF3ompIYmZ7jn9B50a6mx+2JBz9YNeezygXz6k5GMObY1z0xfxYg/T+X+iYvZtPPAhZ7CIufedxbw+jdruXVUV+47o2dU/u9pVr82447rxAcLNrB8U06Fn/9oqIuWiEg1kF9YxAPvL+HlWRmc1KMF/7xsgGaskiqrorpomVkcsAI4FcgEZgOXufuSsG2SgPHAaHfPNrOW7r4lbP0/gRbAdne/7WDnU64jFaGoyLnvnYW8mbKOW0Z15Z7Te6i4UwXtLyji3bmZPD51JWu376VXm4bcProbY45pXelaaq3fsY8xf59Gr7YNeePGYZUuPimbNVv38PjUdN6du54aZlwyOJFbRnWjXeM6QKi4c/db83ln7np+fHISd56SFNX/e7L37GfEX6YyIqk5T1w5KGpxlEZdtEREqqnsPfu5+tlveHlWBj88sQvPXDNYxR2R8jEESHf3Ve6+H3gDOK/ENjcCj7t7NkCJ4s4goBXwSQXFK3JQhUXOzybM582UddwxupuKO1VYrfgaXDq4A1N+eiJ/u7gfefmF3PrqHE7/xzT+M289hUWVoxFAUZHz8wkLKHTnrxf1U3EnhnVqXo+HLu7H1J+N4geDEnlz9jpGPTSVe99ewKqs3dz55jzembuen57anZ+c2j3q//c0qVeL607ozIeLNrF4w86oxnI4VOAREanCVmzO4dzHvyR1bTYPX9KP+87opaltRcpPO2Bd2OvMYFm47kB3M5thZrPMbAyAmdUA/gb87GAnMLObzCzFzFKysrLKMXSR71q8YSfXvTCbd+as565Tu3PXaSruVAfxcTX4waBEPr3rRB65bAAAP35jHqc+/AVvp2ZSUFgU1fhe+TqDL9O38suzetGhWd2oxiLlo33Tuvzxwj58cfdJXD6kA+/MXc/ov33B+/M38PMxPbm9Es2Odv0JnWmYEM/fP10R7VDKLD7aAYiISGR8tmQzP35jLvVqx/PmTcMY0KFJtEMSqY7igSRgFJAITDOzPsCVwCR3zzzYH9Hu/hTwFIS6aEU8Wql25q/bwaNT0vhs6RYa1I7n/nN6M+74ztEOSypYXA3j3H5tObtPGz5evIlHpqTz07fm88/Jadw6qisXDkykVnzFtg1YvXUPf5y0jJHdW3D5kA4Vem6JvLaN6/Db847lRyd147kZa+jUrC5jK9m/c6M6NblpZBf++skK5q/bQb/2jaMd0iGpwCMiUsW4O//6fCV//WQ5fdo14qmrkmndKCHaYYlUReuB9mGvE4Nl4TKBr909H1htZisIFXyGAyPM7FagPlDLzHa7+wEHahYpb6kZ23lkcjpfrMiiUZ2a3HVqd645rhON6qgLb3VWo4ZxRp82jDm2NZ8t3cKjU9K4952FPDolnZtHdeWS5ERqx8dFPI7CIudnb82nZpzxlx/0VWuyKqxlwwTuPaNntMMo1bjjO/Psl6t5+NMVvHjdkGiHc0gq8IiIVCG5+YXcM2EBE+dv4Lz+bfnzD/qSUDPyiZhINTUbSDKzzoQKO2OBy0ts8x5wGfC8mTUn1GVrlbtfUbyBmY0DklXckYowa9U2Hp2Sxoz0bTStV4t7xvTgqmEdaaCx2SSMmXFq71ac0qslX6zI4pHJafzqvUU8NiWNm0/symVDOkQ0v3hm+ipSM7L5+6X9dJNKoqp+7Xh+eGJX/vThMlIztjOoY9Noh3RQKvCIiFQRG3fu46aXUlm0YSf3jOnBLSd21R0vkQhy9wIzuw34GIgDnnP3xWb2AJDi7hODdaeZ2RKgELjb3bdFL2qpjtydGenbeGRKGt+s3k7z+rX5v7N6cfnQDtStpT8HpHRmxqgeLTmxewtmrtzGI5PT+O37S3h86kpuGtmZK4Z2pF7t8v0MLd+Uw98+WcHpx7Ti/P4lhzUTqXhXD+/IM9NX8bdPVvDajcOiHc5BaZp0EZEqYM7abH74cip78wr459gBnNK7VbRDEomaipomvaIp15HD5e58HrS+mLt2B60bJnDziV0YG+HWF1K1fb1qG49OSefL9K00qVuTG0Z04erh5dMKLL+wiPMfn8Gmnbl8/JORNK9fuxwiFjl6z0xfxYP/XcrrNw5jeNdm0Q6n1FxHJXsRkRj3dmom9727kNYNE3j1hqF0b9Ug2iGJiEgUuTufLtnMY1PTWZC5k3aN6/Dg+cdycQWNnyJV29AuzRjapRmpGdk8OiWNhz5ezlPTVnHd8Z0Zd/zRjeP02JR0Fm/YxZNXDlRxRyqVK4d15Onpq/j7pysY1mVYpW0lrwKPiEiMKixy/vzRMp6atorhXZrxrysG0qRerWiHJSIiUVJU5Hy0eBOPTkln6cZddGhalz//oA8XDKj4GZCk6hvUsQkvXDuEBZk7eGRyOn//bAXPTF/FNcd14voTOh92TrIgcwePTU3nggHtGHNsmwhFLXJkEmrG8aOTuvHr/yzmy/StjEhqEe2QDiiiBR4zGwP8k1C/9Gfc/U8l1v8dOCl4WRdo6e6Ng3WFwMJg3Vp3PzeSsYqIxJJdufnc8fpcPl+exTXDO/J/Z/emZpySdxGR6qiwyPlgwQYem5JO2pbddGlej4cv6ce5/doSr+8GibC+iY155ppklmzYxWNT03j883Sen7GaK4d35MYRXcrUEic3v5C7xs+nRf3a3H/OMRUQtcjhu3Rwe578fCUPf7qCE7o1r5SteCJW4DGzOOBx4FRCU4TONrOJ7r6keBt3/0nY9rcDA8IOsc/d+0cqPhGRWLV66x6uf3E2a7ft5Q8X9OHyoR2iHZKIiERBQWER783bwL+mprNq6x66t6rPI5cN4Kw+bYirUfn+8JCqrXfbhvzrikGs2JzDY1PSeXraKl6cuYYrhnbkhyO70LJh6bNhPfzpCtK37ObF64bQqK5mdJPKqXZ8HLeNTuIX7y7k8+VZnNSzZbRD+p5ItuAZAqS7+yoAM3sDOA9YUsr2lwG/iWA8IiIxb3paFj96dQ7xcTV49YahDO0S/UHeRESkYu0vKOKdOZn86/OVrN2+l15tGvLEFQM5/ZjW1FBhR6Kse6sGPHLZAO48JYnHp67khZlreHlWBmMHt+fmE7vStnGd72w/e812np6+isuHduDE7pWz24tIsYuTE3nii3Qe/nQFo3q0qHSteCJZ4GkHrAt7nQkMPdCGZtYR6AxMCVucYGYpQAHwJ3d/L0JxikiE7d1fQJ2acZXuP8BY4u48P2MND/53Cd1bNeDpq5Np37RutMMSEZEKlFdQyPiUTJ78fCXrd+yjX2Ijfn12Mif3aqnvWKl0urSoz98u6cePT07iX5+n89rXa3n9m7VcNKg9t47qSvumddmTV8BPx88nsUkdfnFmr2iHLHJINeNqcMfoJO6esIBPlmzm9GNaRzuk76gsgyyPBSa4e2HYso7uvt7MugBTzGyhu68M38nMbgJuAujQQV0URCobd+e5GWv446SlJHdqwi/O7EXfxMbRDivm5BUU8qv3FjE+JZPTj2nFw5f0p17tyvLft4hI5ePuVargkZtfyOvfrOXJL1ayeVceAzs05vcXHMuJ3Svf3WORkjo0q8ufftCX20Z348kvVjJ+diZvpazjggHtyC8sYl32Xt64cRj1ldtIjLhgQDv+9flK/v7pCk7t1apStZyM5G/ReqB92OvEYNmBjAV+FL7A3dcHP1eZ2eeExudZWWKbp4CnAJKTk71cohaRcpGTm889Exbw4aJNDOvSlLTNuzn3sRmc268td5/eQ61PyigrJ49bXkklJSObO0Z3485TuleqLxERkcomM3sv5z8+g+6tGnD76CSGdWkas0WQPXkFvPp1Bk9NW83W3XkM7dyUhy/pz3Fdm8Xse5LqK7FJXR48vw+3nZTEk1+s5PVv1pJXUMQNJ3RWl3OJKfFxNfjxyUnc+eY8Ply0ibP6Vp5Z38w9MnURM4sHVgAnEyrszAYud/fFJbbrCXwEdPYgGDNrAux19zwzaw58BZwXPkBzScnJyZ6SkhKR9yIih2fZpl3c8soc1m7fy8/H9ODGEV3YnVfAU9NW8fT0VRQVwVXDO3LbSd00rXcp9u4vYN7aHfzsrfls37ufv17cj7P7to12WCIxwcxS3T052nGUN+U6h1ZU5Fz57NfMW7eDerXjycrJY3CnJtxxclKlnfHkQHJy83npqwyemb6K7L35nNCtObeP7qY/gqVK2ZKTy7QVWzmnXxtqx8dFOxyRw1JY5Iz5xzQc+PjOkRU+sH1puU7EWvC4e4GZ3QZ8TGia9OfcfbGZPQCkuPvEYNOxwBv+3UpTL+DfZlYE1CA0Bk+pxR0RqTwmpGbyf+8tpGFCTV6/cRhDOjcFoEFCTX56Wg+uGNqRv3+6gudnrGZ8yjpuO6kb1xzXiYSa1fOLPSc3n/Qtu0nbsjv0c3MOaVt2k5m9D4A2jRKYcPNxHNuuUZQjFRGp/F76ag0zV27jjxf24YIB7Xhz9jqe/GIlVz37Df3bN+bHJydVykExi+3cm8/zM1fz/Iw17NyXz0k9WnDb6CQGdWwS7dBEyl3LBglcNCgx2mGIHJG4Gsadp3TnR6/N4f35Gzh/QLtohwREsAVPRdNdLZHoys0v5P6Ji3lj9jqGd2nGPy/rT8sGpU+HuXxTDn/+aBlTlm2hXeM6/PS07pzfv12V7X60c28+6Vk5pG0OFXNWbM4hfctuNu7M/XabWvE16NqiPkktg0er+gzv0lzThYocJrXgqZ5WZe3mzEemM6xLM54fN/jbIk5eQSETUjP519TQwMR92jXittHdKtW4Cdl79vPsl6t5ceYacvIKOLV3K24f3U3j1omIVGJFRc6Zj0wnr6CIT38ykvi4GhV27tJyHRV4ROSoZWzbw62vzmHxhl386KSu/OSU7mX+D27myq38cdIyFq7fSe82DbnvzJ6MSIrdKTK379n/bSucUMucUFFnS07et9vUqRlHt6CI061VfZJaNiCpZX3aN61b4c07RaoiFXiqn8Ii56InZ7Iqaw+f/GQkrRp+/wZDfmER785dz+NT08nYtpeerUNj9JxxbPSmFt+6O4+np6/i5a8y2JdfyBnHtua2k5Lo3bZhVOIREZHD8/HiTfzw5VQeuqgvFye3P/QO5UQFHhGJiE8Wb+Knb83HgL9f2p+Te7U67GMUFTnvL9jAQx8vJzN7HyO7t+DeMT0rbYLr7mTtziM9aI1TXMRJ37KbbXv2f7td/drx3xZykoJCTreW9WnXuE6luWssUhWpwFP9/OvzdP7y0XL+ObY/5/U/eDP5gsIi3l+wgcempLMyaw/dWtbn9tHdOLtv2worsm/Zlcu/p63i1a8z2F9QxDn92nLbSd1IatWgQs4vIiLlw90557Ev2bkvnyk/HUXNCmrFowKPiJSrgsIiHvp4Of+etoo+7RrxrysGHvXMWHkFhbz8VQaPTklnV24+Fw5I5Kendadt4zrlFPWR25KTy5dpW5kePLbu/l+LnAYJ8XRvFWqF061lfZKC520aJVTacR5EqjIVeKqXZZt2cc6jX3Jq71Y8fvnAMv+/W1jkTFq4kcempLN8cw6dm9fj1lFdOX9Au4gl6Bt27OPJL1byxux1FBY55/dvx49O6kqXFvUjcj4REYm8Kcs2c90LKfzxwj5cNqRDhZxTBR4RKTdbduVy2+tz+Wb1dq4Y2oFfnd27XAdJ3rk3n399ns7zM9dgwLXHd+bWk7rSMKHixqLJzS9k9prtTE/byrQVWSzblANAs3q1OL5bcwZ2aPxtIadFg9oq5IhUIirwVB/7C4o4//EZbMnJ5eM7R9Ksfu3DPkZRkfPJkk08MjmdJRt30b5pHW4d1Y0fDEykVnz5FHrWbd/Lvz5fyYTUdbjDRYMSuXVUNzo0O7obIyIiEn3uzgX/msmWXblMvXtUhcwKpwKPiJSLr1Zu4/bX57Inr4A/XHgsFwyI3OwHmdl7efiTFbwzdz1N6tbk9tFJXDmsY7kl3OHcnRWbdzM9LYtpaVv5etU28gqKqBVXg+ROTRiR1IIRSc3p3aahuleJVHIq8FQfD3+ynEempPPUVYM47ZjWR3Usd2fKsi08MjmN+Zk7adsogVtGdeXi5PZHfBNjzdY9PD41nXfmrifOjEsGJ3LziV1JbKLCjohIVTI9LYurnv2GB847hquHd4r4+VTgEZGjUlTkPDltJX/9eDmdmtfjySsH0b2CxgpYtH4nf/xwKTPSt9GhaV3uPr0HZ/dtc9StZrbuzmNG+lamrdjK9LSsbwdC7tayPiOTWjCie3OGdm5K3Vrx5fE2RKSCqMBTPcxft4MLn5jJ+f3b8bdL+pXbcd2daWlbeXRyGikZ2bRqWJsfjuzKZUM6UKdW2Qo96VtyeGxKOhPnb6BmXA0uH9qBH47sSutGpc8uKSIiscvdufTfs1izbQ/T7jmpXHs3HIgKPCJyxHbuzeeu8fOYvGwLZ/dtw59+0Jf6tSu26FGccP9x0lKWbcqhX2IjfnFmL4Z2aVbmY+QVFJKakf1tQWfxhl0ANK5bkxO6NWdkUgtOSGpeKcb8EZEjpwJP1ZebX8hZj0xn7/5CPrpzJI3qlH8XXnfnq1XbeGRyGrNWbad5/VrcOKILVw7rSL1SvgOXbdrFo1PSmbRwIwnxcVw1vCM3jOhMywYq7IiIVHVfrdzGZU/P4ldn9+b6EzpH9Fwq8IjIEVmYuZNbXk1l865c/u+s3lw9vGNUx5spLHLemZPJ3z5ZwaZduZzSqyU/H9PzgDOPuDsrs3Z/W9CZtWo7+/ILia9hDOzYhBO7h7pdHdO2kaYnF6lCVOCp+h78YAnPfLmal64bwsjuLSJ+vm9Wb+fRKWlMT9tKk7o1uWFEF64e3pEGwdhwi9bv5NEpaXy8eDP1a8dz9fCOXH9C5yMaE0hERGLX5U/PYsXmHKbdc1JEewGowCMih8Xdee2btfx24hKa16/F41cMZECHJtEO61u5+YU8N2M1T0xdyZ79BVw6uD0/OaU7NeNq8GV6qKAzPW0rG3fmAtCleT1GJDVnRFILhnVtVuEtkESk4qjAU7XNWhW6Q3rF0A48eH6fCj33nLXZPDYlnSnLttAwIZ6rhndk2cYcJi/bQoOEeK49vjPXHd+JxnVrVWhcIiJSOaSs2c5FT37FvWf05OYTu0bsPCrwiEiZ7d1fwP+9u4h35q7nxO4t+Mel/WlSr3Imq9v37OfRKWm8MisDMyO/sAh3aJgQzwlBQeeEbs2Pegp3EYkdKvBUXbvzCjjjn9OoYcakO0aU2lUq0hZmhlrsfLJkM43r1uSGEzpz9XGdKnS2RxERqZyufu4blmzYxcx7R0dkchgoPdfRLWwR+Y6VWbu55ZVU0rbs5q5Tu3PbSd0q9axRTevV4jfnHMO44zrx3JeraVqvNiO6N6dvu0bEx0XmP1QREYmOP0xaSmb2Psb/cHjUijsAfRIb8dTVyWzcuY+GCTWjGouIiFQuvz67N3E1LGLFnYPRt5GIfOuDBRv4+YQF1K4Zx0vXDWFEUuTHNSgvHZvV47fnHRvtMEREJEI+X76F175ey00juzC4U9NohwNAm0YalF9ERL6rW8v6UTu3Cjwiwv6CIv4waSkvzFzDwA6NefyKgUpaRUSk0ti5N5+fv72ApJb1uevU7tEOR0REpFJSgUekmtuwYx8/em0Oc9fu4PoTOnPvGT2pqa5NIiJSidz//mK27t7PM1cPJqFmXLTDERERqZRU4BGphoqKnL35hXyzehs/HT+f/ELniSsGckafNtEOTURE5Ds+WrSRd+eu58cnJ9EnsVG0wxEREam0VOARiSEFhUXsySskJy+f3XkF7M4tICf4+f3XoW1ywtZ9+3N/AcUT6PVs3YB/XTGQLi2i11dURETkQLbuzuOX7y7i2HYNuW10t2iHIyIiUqmpwCNSiezbX8iHizYyeekWsvfu/17RZl9+YZmOU792fOiREPrZICGe1g0Tvl3WIPjZpG4tzu7bljq11NxdREQqF3fnF+8sJCe3gNcv6a/uwyIiIoegAo9IlLk789btYHxKJh/M30BOXgFtGyXQtnEdmtarRYemdWkQFGrq1675nQJNyYJN/drx1KsVX6mnNRcRESmL9+at55Mlm7nvjJ50b9Ug2uGIiIhUeirwiETJ1t15vDd3PeNT1rFi827q1IzjzD5tuCQ5kSGdm2KmIo2IiFRPG3fu49f/WUxyxybcMKJLtMMRERGJCSrwiFSggsIivliRxfiUdUxeuoWCImdAh8b88cI+nN23DQ0SakY7RBERkahyd+6ZsICCQuevF/cjTq1SRUREykQFHpEKsCprN2+lZvJ2aiZbcvJoXr8W153QmYsHJZKkZuciIiLfeu2btUxP28rvzjuGTs3rRTscERGRmKECj0iE7Mkr4L8LN/JWyjpmr8kmroZxUo8WXJLcnpN6ttRgkSIiMc7MxgD/BOKAZ9z9TwfY5hLgfsCB+e5+uZn1B54AGgKFwO/d/c2KirsyW7ttL7//71JO6NacK4Z2jHY4IiIiMUUFHpFy5O6kZmQzPmUdHyzYyN79hXRpUY97z+jJhQPa0bJhQrRDFBGRcmBmccDjwKlAJjDbzCa6+5KwbZKA+4Dj3T3bzFoGq/YCV7t7mpm1BVLN7GN331Gx76JyKSxyfvbWfOLM+MtFfTVhgIiIyGFSgUekHGzZlcs7wYDJq7L2UK9WHOf0bcslgxMZ2KGJBkwWEal6hgDp7r4KwMzeAM4DloRtcyPwuLtnA7j7luDniuIN3H2DmW0BWgA7Kib0yun5Gav5Zs12/npxP9o2rhPtcERERGKOCjwiRyi/sIgpy7bwVso6pi7PorDIGdypCTef2JWz+rShXm39eomIVGHtgHVhrzOBoSW26Q5gZjMIdeO6390/Ct/AzIYAtYCVkQu18kvfksNfPl7OKb1a8YOB7aIdjoiISEzSX6Aihyltcw7jU9bx7tz1bN29n5YNanPTyC5cPCiRLi3qRzs8ERGpPOKBJGAUkAhMM7M+xV2xzKwN8DJwjbsXHegAZnYTcBNAhw4dKiDkipdfWMRd4+dTr1Ycf7ywj1q9ioiIHCEVeEQOoajIWbJxF9PTtvLx4k3MW7eD+BrGyb1acung9oxMakG8BkwWEalu1gPtw14nBsvCZQJfu3s+sNrMVhAq+Mw2s4bAf4Ffuvus0k7i7k8BTwEkJyd7OcZfaTzx+UoWZO7kX1cMpEWD2tEOR0REJGYdssBjZucA/y3tzpJIVbRlVy7T07YyLS2LL9O2sm3PfgB6tWnI/53Vi/MHtKN5fSWhIiJVwRHmOrOBJDPrTKiwMxa4vMQ27wGXAc+bWXNCXbZWmVkt4F3gJXefcLTxx7JF63fyyOQ0zu3XljP7tIl2OCIiIjGtLC14LgX+YWZvA8+5+7IIxyRS4XLzC/lm9Xamp2UxPW0ryzblANC8fi1GJDVnRFILRiQ11yxYIiJV02HnOu5eYGa3AR8TGl/nOXdfbGYPACnuPjFYd5qZLSE0Hfrd7r7NzK4ERgLNzGxccMhx7j6v3N9ZJZZXUMhPx8+nab1aPHDeMdEOR0REJOYdssDj7lcGzYgvA14wMweeB15395xIBygSCe7O8s05TFsRKuh8vXo7+wuKqBVXg8Gdm3DvGT0ZkdScXq0bappWEZEq7khzHXefBEwqsezXYc8duCt4hG/zCvBK+b2D2PT3T9NYvjmH58cNpnHdWtEOR0REJOaVaQwed99lZhOAOsCdwAXA3Wb2iLs/GsH4RMpNVk4eM9JD3a6mp20lKycPgKSW9blqWEdGJDVnaOdm1KkVF+VIRUSkoinXqVipGdt5atpKxg5uz0k9W0Y7HBERkSqhLGPwnAtcC3QDXgKGuPsWM6sLLAGU9EillFdQSMqa7FBBZ8VWlmzcBUCTujU5IehyNSKpOW0a1YlypCIiEk3KdSrW3v0F/HT8fNo0qsMvz+oV7XBERESqjLK04PkB8Hd3nxa+0N33mtn1kQlL5PC5O+lbdjMtbSvT07KYtWobuflF1IwzBnVswt2n92BkUguOaatuVyIi8h3KdSrQnz9cxppte3n9xmE0SKgZ7XBERESqjLIUeO4HNha/MLM6QCt3X+PukyMVmEhZLd24i+e+XM30tK1s2pULQJcW9Rg7uEOo21WXZtSvXabeiCIiUj3dj3KdCjEjfSsvfpXBtcd3YnjXZtEOR0REpEopy1+9bwHHhb0uDJYNjkhEImW0v6CIx6am86+p6dSpFcfIoNvVCUnNSWxSN9rhiYhI7FCuUwFy8wu5Z8ICujSvxz2n94x2OCIiIlVOWQo88e6+v/iFu+83M011IFE1f90O7pmwgOWbc7hwQDt+dXZvmtTTx1JERI6Icp0KkJqRzfod+3j66mRNaCAiIhIBZSnwZJnZue4+EcDMzgO2RjYskQPLzS/kH5+l8dS0lbRskMBz45IZ3bNVtMMSEZHYplynAqRmZGMGQzo3jXYoIiIiVVJZCjw3A6+a2WOAAeuAq8tycDMbA/wTiAOecfc/lVj/d+Ck4GVdoKW7Nw7WXQP8X7DuQXd/sSznlKorNWM7d09YwKqsPYwd3J5fnNWLhhqcUUREjt4R5zpSdqkZ2XRv2YBGdfTdLSIiEgmHLPC4+0pgmJnVD17vLsuBzSwOeBw4FcgEZpvZRHdfEnbsn4RtfzswIHjeFPgNkAw4kBrsm13WNyZVx979BTz08XJemLmGto3q8PL1QxiR1CLaYYmISBVxpLmOlF1RkTNnbTbn9Gsb7VBERESqrDJNLWRmZwHHAAlmoeml3f2BQ+w2BEh391XBMd4AzgOWlLL9ZYSKOgCnA5+6+/Zg30+BMcDrZYlXqo6ZK7dy79sLWbt9L1cP78jPx/SknmbEEhGRcnaEuY6UUXrWbnJyCxjUoUm0QxEREamyDvmXspk9Saj71EnAM8BFwDdlOHY7Qk2ci2UCQ0s5R0egMzDlIPu2O8B+NwE3AXTo0KEMIUmsyMnN508fLuPVr9fSqVld3rxpGEO7aDpVEREpf0eR60gZpWaEGmEP6qgCj4iISKTUKMM2x7n71UC2u/8WGA50L+c4xgIT3L3wcHZy96fcPdndk1u0UJedquKLFVmc/vdpvPbNWm44oTMf/nikijsiIhJJFZHrVGupGdk0q1eLjs3qRjsUERGRKqssfV1yg597zawtsA1oU4b91gPtw14nBssOZCzwoxL7jiqx7+dlOKfEsJ1783nwv0t4KzWTbi3r8/YtxzFQTblFRCTyjjTXkTKak5HNwI5NKO7+JiIiIuWvLAWe982sMfAQMIfQoMdPl2G/2UCSmXUmVLAZC1xeciMz6wk0Ab4KW/wx8AczK/7r/jTgvjKcU2LUp0s288t3F7Jtz35+dFJXbh+dRELNuGiHJSIi1cOR5jpSBtt257Fq6x4uGdz+0BuLiIjIETtogcfMagCT3X0H8LaZfQAkuPvOQx3Y3QvM7DZCxZo44Dl3X2xmDwAp7j4x2HQs8Ia7e9i+283sd4SKRAAPFA+4LFXL9j37+e37i/nPvA30bN2AZ68ZTJ/ERtEOS0REqomjyXWkbOas3QFo/B0REZFIO2iBx92LzOxxgunL3T0PyCvrwd19EjCpxLJfl3h9fyn7Pgc8V9ZzSez574KN/Po/i9iVm89PTunOLaO6Uiu+LMNCiYiIlI+jzXXk0FIzsqkZZ/Rppxs4IiIikVSWLlqTzewHwDvhrWxEjlRWTh6//s8iPly0iT7tGvHqxUPp2bphtMMSEZHqS7lOBM3JyOaYto3U9VpERCTCylLg+SFwF1BgZrmAAe7u+otcDou789689fz2/SXs3V/Iz8f05MYRnYmPU6sdERGJKuU6EbK/oIj5mTu4cljHaIciIiJS5R2ywOPuDSoiEKnaNu3M5ZfvLmTysi0M7NCYv1zUj24t60c7LBEREeU6EbRk4y7yCoo0/o6IiEgFOGSBx8xGHmi5u08r/3CkqnF3xqes48EPlpJfVMSvzu7NuOM6EVdD06SKiEjloFwnclIzsgENsCwiIlIRytJF6+6w5wnAECAVGB2RiKTKWL9jH/e+vYDpaVsZ2rkpf7moLx2b1Yt2WCIiIiUp14mQORnZJDapQ6uGCdEORUREpMorSxetc8Jfm1l74B+RCkiqhonzN/DLdxdSVOT87vxjuWJIB2qo1Y6IiFRCynUiw91JydjOsC7Noh2KiIhItVCWFjwlZQK9yjsQqRpycvP5zX8W887c9Qzs0Jh/XDqADs3qRjssERGRw6Fcpxxs2JnL5l156p4lIiJSQcoyBs+jQPGUoTWA/sCcCMYkMSo1Yzt3vjmP9dn7uPOUJG47qZtmyBIRkUpPuU5kFI+/M7CDCjwiIiIVoSwteFLCnhcAr7v7jAjFIzGooLCIx6am8+iUdNo0SuCtm4czqGPTaIclIiJSVsp1IiB1zXbq1oqjZ2tNUiYiIlIRylLgmQDkunshgJnFmVldd98b2dAkFqzbvpcfvzGXOWt3cOGAdvz2vGNokFAz2mGJiIgcDuU6EZC6Npv+7RurNa+IiEgFKcs37mSgTtjrOsBnkQlHYoW78+7cTM7453TStuzmn2P78/Cl/VXcERGRWKRcp5ztyStg6cYcjb8jIiJSgcrSgifB3XcXv3D33WamUXOrsZ378vnVe4uYOH8Dgzs14e+X9iexiT4SIiISs5TrlLP5mTsoLHIGqsAjIiJSYcpS4NljZgPdfQ6AmQ0C9kU2LKmsvlm9nZ+8OY9Nu3L56andufWkbsRp+nMREYltynXK2ZziAZbbq8AjIiJSUcpS4LkTeMvMNgAGtAYujWRQUvnkFxbxyOQ0Hp+aTvumdZlw83AGaFYMERGpGu5EuU65Ss3Ipnur+jSqq67bIiIiFeWQBR53n21mPYEewaLl7p4f2bCkMlmzdQ8/fnMe89ft4KJBidx/7jHUr12W2qCIiEjlp1ynfBUVOXPW7uDMPq2jHYqIiEi1cshBls3sR0A9d1/k7ouA+mZ2a+RDk2hzd95KWcdZj0xnddZuHrt8AH+9uJ+KOyIiUqUo1ylfq7buZue+fAaqpa+IiEiFKsssWje6+47iF+6eDdwYsYikUti5N5/bXpvL3RMWcGy7Rnx050jO7ts22mGJiIhEwhHnOmY2xsyWm1m6md1byjaXmNkSM1tsZq+FLb/GzNKCxzVH+yYqi9Rg/B3NoCUiIlKxytIUI87MzN0dwMzigFqRDUui6auV27hr/DyycvK4Z0wPfjiyqwZSFhGRquyIcp1gu8eBU4FMYLaZTXT3JWHbJAH3Ace7e7aZtQyWNwV+AyQDDqQG+2aX83urcKkZ2TSpW5POzetFOxQREZFqpSwFno+AN83s38HrHwIfRi4kiZb9BUX8/bMVPPnFSjo1q8c7tx5H38TG0Q5LREQk0o401xkCpLv7KgAzewM4D1gSts2NwOPFhRt33xIsPx341N23B/t+CowBXj/K9xJ1KRnZDOrYBDPdHBIREalIZSnw/By4Cbg5eL2A0OwSUoWsytrNnW/OY0HmTsYObs+vzu5NPY21IyIi1cOR5jrtgHVhrzOBoSW26Q5gZjOAOOB+d/+olH3bHXbklcz2PftZlbWHiwYlRjsUERGRaqcss2gVmdnXQFfgEqA58HakA5OK4e68OXsdv31/CbVr1uDJKwcy5tg20Q5LRESkwkQ414kHkoBRQCIwzcz6lHVnM7uJUPGJDh06lFNIkTN3bTD+jgZYFhERqXClFnjMrDtwWfDYCrwJ4O4nVUxoEmnZe/Zz3zsL+WjxJo7r2oyHL+lP60YJ0Q5LRESkQpRDrrMeaB/2OjFYFi4T+DqYdn21ma0gVPBZT6joE77v5yVP4O5PAU8BJCcnexnjiprUjGzia5i6eIuIiETBwVrwLAOmA2e7ezqAmf2kQqKSiJuRvpW7xs9j+579/OLMntxwQhdqaCBlERGpXo4215kNJJlZZ0IFm7HA5SW2eY9QAel5M2tOqMvWKmAl8AczK27qchqhwZhjWmpGNse0bUidWnHRDkVERKTaOViB50JCicpUM/sIeANQBSDG5RUU8vAnK3hq+io6N6/Hs9cM5th2jaIdloiISDQcVa7j7gVmdhvwMaHxdZ5z98Vm9gCQ4u4Tg3WnmdkSoBC42923AZjZ7wgViQAeKB5wOVblFxYxP3MHlw/pGO1QREREqqVSCzzu/h7wnpnVIzQjxJ1ASzN7AnjX3T+pkAil3KRv2c2P35jL4g27uHxoB351Vm/dYRMRkWqrPHIdd58ETCqx7Ndhzx24K3iU3Pc54LmjeAuVytKNu8jNL2JQR42/IyIiEg01DrWBu+9x99fc/RxC/cPnEpptQmKEu/PKrAzOfnQ6G3bs46mrBvGHC/qouCMiIoJynfKSmhEaYHlgx8bRDURERKSaOqx5sN09m9BAf09FJhwpb9t25/Hztxfy2dLNjEhqzt8u7kfLhhpIWURE5ECU6xy51Ixs2jWuQ5tGdaIdioiISLV0WAUeiS3TVmTx07fms3NvPr86uzfXHtdJAymLiIhIRMzJyGZQp6bRDkNERKTaUoGnCsrNL+QvHy3nuRmrSWpZnxevHULvtg2jHZaIiIhUURt27GPDzlxu6tA42qGIiIhUWyrwVDErNudwx+tzWbYph2uGd+S+M3uRUFNj7YiIiEjkFI+/M6ijWvCIiIhEiwo8VYS789JXGfxh0lIaJMTz3LhkRvdsFe2wREREpBpIzcimTs04erZpEO1QREREqi0VeKqArJw87pkwn6nLsxjVowUPXdSPFg1qRzssERERqSbmrM2mX/tG1Iw75AStIiIiEiEq8MS4qcu2cPeE+ezKLeC35x7D1cM7YqaBlEVERKRi7N1fwOINu7j5xC7RDkVERKRaU4EnRuXmF/KnD5fxwsw19GzdgFdvGEaP1moWLSIiIhVrQeZOCoucZI2/IyIiElUq8MSgpRt38eM35rJi826uO74z94zpoYGURUREJCqKB1geoBm0REREokoFnhhSVOS8MHMNf/poGQ0TavLCtYMZ1aNltMMSERGRamxORjbdWtancd1a0Q5FRESkWlOBJ0Zs2ZXLzyYsYNqKLE7p1ZI//6AvzeprIGURERGJHncndW02p/duHe1QREREqr2ITnVgZmPMbLmZpZvZvaVsc4mZLTGzxWb2WtjyQjObFzwmRjLOyu7TJZsZ88/pfL1qG787/1ievjpZxR0RERGJulVb97Bjbz6DOjaJdigiIiLVXsRa8JhZHPA4cCqQCcw2s4nuviRsmyTgPuB4d882s/D+RvvcvX+k4osF+/YX8uB/l/Dq12vp3aYhj1zWn24tNZCyiIiIVA7F4+8MVIFHREQk6iLZRWsIkO7uqwDM7A3gPGBJ2DY3Ao+7ezaAu2+JYDwxZdH6nfz4jbmszNrDjSM687PTe1A7XgMpi4iISOWRuiabxnVr0qV5vWiHIiIiUu1FssDTDlgX9joTGFpim+4AZjYDiAPud/ePgnUJZpYCFAB/cvf3Sp7AzG4CbgLo0KFDuQYfLUVFzjNfruKhj5fTpG4tXrl+KCckNY92WCIiIiLfk7o2m4EdmlCjhkU7FBERkWov2oMsxwNJwCggEZhmZn3cfQfQ0d3Xm1kXYIqZLXT3leE7u/tTwFMAycnJXqGRR8DmXbncNX4eM9K3cVrvVvz5B31pUk8zUoiIiEjls2PvftK37OaCAe2iHYqIiIgQ2QLPeqB92OvEYFm4TOBrd88HVpvZCkIFn9nuvh7A3VeZ2efAAGAlVdgtr6SydGMOf7ywD2MHt8dMd8NERESkcpq7dgcAAzto/B0REZHKIJKzaM0Gksyss5nVAsYCJWfDeo9Q6x3MrDmhLlurzKyJmdUOW3483x27p8qZt24Hc9bu4N4zenLZkA4q7oiIiEillpqRTVwNo1/7RtEORURERIhgCx53LzCz24CPCY2v85y7LzazB4AUd58YrDvNzJYAhcDd7r7NzI4D/m1mRYSKUH8Kn32rKnpx5hrq147nB4MSox2KiIiIyCGlZmRzTNuG1K0V7R7/IiIiAhEeg8fdJwGTSiz7ddhzB+4KHuHbzAT6RDK2yiQrJ4//LtjI5UM7UL+2kiQRERGp3AoKi5i3bgeXDm5/6I1FRESkQkSyi5aU0RvfrGV/YRFXDe8Y7VBEREREDmnZphz25RcyqKPG3xEREaksVOCJsvzCIl75OoMRSc3p2qJ+tMMREREROaTUjGwAFXhEREQqERV4ouyTxZvZvCuPccd1inYoIiIiImWSmpFNm0YJtG1cJ9qhiIiISEAFnih7ceYa2jetw6geLaMdioiIiEiZpGZkM1Ctd0RERCoVFXiiaMmGXXyzZjtXD+tEXA1Niy4iIiKV38ad+1i/Yx+DOqjAIyIiUpmowBNFL321hoSaNbgkWTNQiIiISGyYk7ED0Pg7IiIilY0KPFGyY+9+3pu3ngsGtKNR3ZrRDkdERESkTFIzskmoWYPebRtGOxQREREJowJPlLw5ex25+UVco8GVRUREJIakrs2mb2JjasYpjRQREalM9M0cBYVFzsuzMhjauSk9W+vul4iIiMSG3PxCFq/fSbK6Z4mIiFQ6KvBEwZRlW8jM3qfWOyIiIhJTFmTupKDINf6OiIhIJaQCTxS8OHMNbRolcFrvVtEORURERKTMUjOyARigGbREREQqHRV4Klj6lhy+TN/KlcM6Eq++6yIiIhJDUjOy6dKiHk3r1Yp2KCIiIlKCKgwV7KWvMqgVV4OxgzU1uoiISKwzszFmttzM0s3s3gOsH2dmWWY2L3jcELbuL2a22MyWmtkjZmYVG/3hcXfmrM1mkFrviIiIVErx0Q6gOsnJzeft1EzO7teGZvVrRzscEREROQpmFgc8DpwKZAKzzWyiuy8psemb7n5biX2PA44H+gaLvgROBD6PaNBHYc22vWzfs1/j74iIiFRSasFTgSakZrJnfyHjNLiyiIhIVTAESHf3Ve6+H3gDOK+M+zqQANQCagM1gc0RibKcpKzZDqACj4iISCWlAk8FKSpyXv4qgwEdGtM3sXG0wxEREZGj1w5YF/Y6M1hW0g/MbIGZTTCz9gDu/hUwFdgYPD5296UldzSzm8wsxcxSsrKyyv8dHIY5a7NpmBBP1xb1oxqHiIiIHJgKPBVkevpWVm3dwzXDO0U7FBEREak47wOd3L0v8CnwIoCZdQN6AYmEikKjzWxEyZ3d/Sl3T3b35BYtWlRg2N+XmpHNwI5NqFGjUg8VJCIiUm2pwFNBXpy5hub1a3NmnzbRDkVERETKx3ogfNaExGDZt9x9m7vnBS+fAQYFzy8AZrn7bnffDXwIDI9wvEds5758VmzerQGWRUREKjEVeCpAxrY9TF2+hcuHdqBWvC65iIhIFTEbSDKzzmZWCxgLTAzfwMzC7+ycCxR3w1oLnGhm8WZWk9AAy9/rolVZzF2bDcCgTirwiIiIVFaaRasCvPxVBnFmXDG0Q7RDERERkXLi7gVmdhvwMRAHPOfui83sASDF3ScCd5jZuUABsB0YF+w+ARgNLCQ04PJH7v5+Rb+HspqTkU1cDaOfxhEUERGptFTgibC9+wsYn7KOMce2plXDhGiHIyIiIuXI3ScBk0os+3XY8/uA+w6wXyHww4gHWE5S12bTq00D6tVW6igiIlJZqb9QhL07dz27cgs0NbqIiIjEpILCIuat3aHxd0RERCo5FXgiyN15aWYGvds0ZFBHJUUiIiISe5ZvzmHP/kIGKpcRERGp1FTgiaBZq7azfHMO447rhJmmFBUREZHYMycjGGBZBR4REZFKTQWeCHpx5hoa163Juf3bRjsUERERkSOSkpFNq4a1ade4TrRDERERkYNQgSdC1u/YxydLNnHp4PYk1IyLdjgiIiIiRyQ1I5tBHZuoNbKIiEglpwJPhLw6KwOAq4Z1jHIkIiIiIkdm865cMrP3MVADLIuIiFR6KvBEQG5+IW/MXscpvVqR2KRutMMREREROSIaf0dERCR2qMATAR8s2Mj2Pfu5RlOji4iISAxLzcimdnwNjmnbKNqhiIiIyCGowFPO3J0XZ64hqWV9juvaLNrhiIiIiByx1LXZ9EtsTK14pYwiIiKVnb6ty9mctTtYuH4nV2tqdBEREYlhufmFLFq/k4HqniUiIhITVOApZy99tYYGteO5cEC7aIciIiIicsQWrd9JfqFr/B0REZEYoQJPOdqSk8ukhRu5KDmRerXjox2OiIiIyBFLDQZYHtihcXQDERERkTJRgaccvfb1WvILnauHd4p2KCIiIiJHJTUjm87N69Gsfu1ohyIiIiJloAJPOdlfUMSrX6/lxO4t6Ny8XrTDERERETli7k5qRjYDO6h7loiISKxQgaecfLR4E1k5eYzT1OgiIiIS4zK27WXbnv0af0dERCSGqMBTTl6cuYaOzepyYvcW0Q5FRERE5KgUj7+jAo+IiEjsiGiBx8zGmNlyM0s3s3tL2eYSM1tiZovN7LWw5deYWVrwuCaScR6tRet3kpqRzVXDOlKjhqZGFxERkdiWujabBrXjSWpZP9qhiIiISBlFbKonM4sDHgdOBTKB2WY20d2XhG2TBNwHHO/u2WbWMljeFPgNkAw4kBrsmx2peI/GizPXUKdmHBcnt492KCIiIiJHbU5GNgM6NtGNKxERkRgSyRY8Q4B0d1/l7vuBN4DzSmxzI/B4ceHG3bcEy08HPnX37cG6T4ExEYz1iG3fs5//zN/AhQPb0ahOzWiHIyIiInJUduXms3xzDsnqniUiIhJTIlngaQesC3udGSwL1x3obmYzzGyWmY05jH0xs5vMLMXMUrKyssox9LJ7Y/Za9hcUaWp0ERERqRLmrd2Bu8bfERERiTXRHmQ5HkgCRgGXAU+bWeOy7uzuT7l7srsnt2hR8YMbFxQW8eqstQzv0owerRtU+PlFREREyltqRjY1DPq1bxztUEREROQwRLLAsx4IH5QmMVgWLhOY6O757r4aWEGo4FOWfaPus6VbWL9jH9doanQRERGpIuaszaZn64bUrx2xoRpFREQkAiJZ4JkNJJlZZzOrBYwFJpbY5j1CrXcws+aEumytAj4GTjOzJmbWBDgtWFapvDhzDW0bJXBKr5bRDkVERETkqBUWOXPX7lD3LBERkRgUsVsz7l5gZrcRKszEAc+5+2IzewBIcfeJ/K+QswQoBO52920AZvY7QkUigAfcfXukYj0SKzbn8NWqbdwzpgfxcdHu6SYiIiJy9FZszmF3XoEKPCIiIjEoom1v3X0SMKnEsl+HPXfgruBRct/ngOciGd/ReHHmGmrF12Ds4A7RDkVERGJQfn4+mZmZ5ObmRjuUmJWQkEBiYiI1a2oWy/KSkpENaIBlERE5Ospzysfh5jrqXH0Edu7L55056zmvX1ua1qsV7XBERCQGZWZm0qBBAzp16oSZRTucmOPubNu2jczMTDp37hztcKqMORnZtGhQm8QmdaIdioiIxDDlOUfvSHId9S06AhNSM9mXX6jBlUVE5Ijl5ubSrFkzJT1HyMxo1qyZ7gyWs9SMbAZ1aKLPpYiIHBXlOUfvSHIdFXgOU1GR8/JXaxjUsQnHtmsU7XBERCSGKek5Orp+5WtLTi5rt+9V9ywRESkX+p4+eod7DVXgOUxfrMhizba9ar0jIiIiVcqcjB0ADOqkAo+IiEgsUoHnML341RpaNKjNmGNaRzsUERGRcjdu3DgmTJhQ7sfs3Lkz/fv3p3///sybNw8I9S2/44476NatG3379mXOnDkVFpN835y12dSKr8ExbRtGOxQREZGIqcq5jgZZPgyrt+7h8+VZ3HlKErXiVRsTEREpq4ceeoiLLrroO8s+/PBD0tLSSEtL4+uvv+aWW27h66+/jlKEkpqRTd92jagdHxftUERERGJOZch1VKU4DC99tYaaccblQzU1uoiIxL7f/e539OjRgxNOOIHLLruMv/71r99ZP3nyZAYMGECfPn247rrryMvL46OPPuLiiy/+dpvPP/+cs88+G4BPPvmE4cOHM3DgQC6++GJ279590PP/5z//4eqrr8bMGDZsGDt27GDjxo24O7fddhs9evTglFNOYcuWLeX/5uU78goKWZi5U+PviIhIlVLdch214CmjPXkFTEjJ5Ixj29CyQUK0wxERkSrkt+8vZsmGXeV6zN5tG/Kbc44pdf3s2bN5++23mT9/Pvn5+QwcOJBBgwZ9uz43N5dx48YxefJkunfvztVXX80TTzzBbbfdxk033cSePXuoV68eb775JmPHjmXr1q08+OCDfPbZZ9SrV48///nPPPzww/z6178G4Je//CUPPPAAJ598Mn/605+oXbs269evp3379t+eMzExkfXr1/PVV1+xfPlylixZwubNm+nduzfXXXdduV6f8mBmY4B/AnHAM+7+pxLrxwEPAeuDRY+5+zPBug7AM0B7wIEz3X1NxUT+fYvW72J/YREDVeAREZFyFo08B6pnrqMWPGX0ztz15OQVaHBlERGpEmbMmMF5551HQkICDRo04JxzzvnO+uXLl9O5c2e6d+8OwDXXXMO0adOIj49nzJgxvP/++xQUFPDf//6X8847j1mzZrFkyRKOP/54+vfvz4svvkhGRgYAf/zjH1m2bBmzZ89m+/bt/PnPfz5obNOmTeOyyy4jLi6Otm3bMnr06MhchKNgZnHA48AZQG/gMjPrfYBN33T3/sHjmbDlLwEPuXsvYAgQ1WZKczKyARjYQQUeERGpGqpjrqMWPGXg7rw0cw192jViYIfG0Q5HRESqmEPdgapsxo4dy2OPPUbTpk1JTk6mQYMGuDunnnoqr7/++ve2b9OmDQC1a9fm2muv/bZ5dLt27Vi3bt2322VmZtKuXbuKeRNHbwiQ7u6rAMzsDeA8YMmhdgwKQfHu/imAux+8fXcFSMnYTsdmdWnRoHa0QxERkSom1vIciN1cRy14ymDmym2kbdnN1cM7HvY89CIiIpXR8ccfz/vvv09ubi67d+/mgw8++M76Hj16sGbNGtLT0wF4+eWXOfHEEwE48cQTmTNnDk8//TRjx44FYNiwYcyYMePb7ffs2cOKFSsA2LhxIxC6YfLee+9x7LHHAnDuuefy0ksv4e7MmjWLRo0a0aZNG0aOHMmbb75JYWEhGzduZOrUqZG/IIevHbAu7HVmsKykH5jZAjObYGbFbbS7AzvM7B0zm2tmDwUtgr7HzG4ysxQzS8nKyirfdxBwd1IzdjBIrXdERKQKqY65jlrwlMGLM9fQtF4tzunXNtqhiIiIlIvBgwdz7rnn0rdvX1q1akWfPn1o1KjRt+sTEhJ4/vnnufjiiykoKGDw4MHcfPPNAMTFxXH22Wfzwgsv8OKLLwLQokULXnjhBS677DLy8vIAePDBB+nevTtXXHEFWVlZuDv9+/fnySefBODMM89k0qRJdOvWjbp16/L8888DcMEFFzBlyhR69+5Nhw4dGD58eEVemvL0PvC6u+eZ2Q+BF4HRhPKvEcAAYC3wJjAOeLbkAdz9KeApgOTkZI9EkOu272Pr7jyNvyMiIlVKdcx1zD0iuUKFS05O9pSUlHI/blGR88v3FtK6YR1+fEpSuR9fRESqp6VLl9KrV6+oxrB7927q16/P3r17GTlyJE899RQDBw6MakyH60DX0cxS3T05kuc1s+HA/e5+evD6PgB3/2Mp28cB2929kZkNA/7s7icG664Chrn7jw52zkjlOks27OJ3Hyzh/nOPoUfrBuV+fBERqX4qQ54D1S/XUQueQ6hRw/jjhX2jHYaIiEi5u+mmm1iyZAm5ublcc801MZfwRNlsIMnMOhOaJWsscHn4BmbWxt03Bi/PBZaG7dvYzFq4exahVj3lX7kpo95tG/L6TcOidXoREZGIqW65jgo8IiIi1dRrr70W7RBilrsXmNltwMeEpkl/zt0Xm9kDQIq7TwTuMLNzgQJgO6FuWLh7oZn9DJhsocH9UoGno/E+REREqrLqluuowCMiIhIl7q7B+49CtLuZu/skYFKJZb8Oe34fcF8p+34KqImwiIhUWcpzjt7h5jqaRUtERCQKEhIS2LZtW9SLFLHK3dm2bRsJCQnRDkVERERKUJ5z9I4k11ELHhERkShITEwkMzOTSE19XR0kJCSQmJgY7TBERESkBOU55eNwcx0VeERERKKgZs2adO7cOdphiIiIiJQ75TnRoS5aIiIiIiIiIiIxTgUeEREREREREZEYpwKPiIiIiIiIiEiMs6oyqrWZZQEZETxFc2BrBI8fC3QNdA2K6TroGoCuAegaQOW8Bh3dvUW0gyhvynUqhK6BrgHoGoCuQTFdB10DqJzX4IC5TpUp8ESamaW4e3K044gmXQNdg2K6DroGoGsAugaga1CV6N9S1wB0DUDXAHQNiuk66BpAbF0DddESEREREREREYlxKvCIiIiIiIiIiMQ4FXjK7qloB1AJ6BroGhTTddA1AF0D0DUAXYOqRP+WugagawC6BqBrUEzXQdcAYugaaAweEREREREREZEYpxY8IiIiIiIiIiIxTgUeEREREREREZEYpwJPCWY2xsyWm1m6md17gPW1zezNYP3XZtYpCmFGjJm1N7OpZrbEzBab2Y8PsM0oM9tpZvOCx6+jEWskmdkaM1sYvL+UA6w3M3sk+BwsMLOB0YgzUsysR9i/7zwz22Vmd5bYpkp+DszsOTPbYmaLwpY1NbNPzSwt+NmklH2vCbZJM7NrKi7q8lXKNXjIzJYFn/d3zaxxKfse9HcnVpRyDe43s/Vhn/kzS9n3oN8jsaKUa/Bm2PtfY2bzStm3SnwOqqLqnueAcp1iynWqZ66jPCdEuY5yHaiiuY676xE8gDhgJdAFqAXMB3qX2OZW4Mng+VjgzWjHXc7XoA0wMHjeAFhxgGswCvgg2rFG+DqsAZofZP2ZwIeAAcOAr6MdcwSvRRywCehYHT4HwEhgILAobNlfgHuD5/cCfz7Afk2BVcHPJsHzJtF+P+V4DU4D4oPnfz7QNQjWHfR3J1YepVyD+4GfHWK/Q36PxMrjQNegxPq/Ab+uyp+DqvZQnvPte1Su48p1SrzXapPrKM856HVQrqNcp+T6mMt11ILnu4YA6e6+yt33A28A55XY5jzgxeD5BOBkM7MKjDGi3H2ju88JnucAS4F20Y2qUjoPeMlDZgGNzaxNtIOKkJOBle6eEe1AKoK7TwO2l1gc/nv/InD+AXY9HfjU3be7ezbwKTAmUnFG0oGugbt/4u4FwctZQGKFB1aBSvkclEVZvkdiwsGuQfC9dwnweoUGJUer2uc5oFznMCjXqYKU54Qo11GuA1Uz11GB57vaAevCXmfy/S/8b7cJ/gPYCTSrkOgqWNAsewDw9QFWDzez+Wb2oZkdU7GRVQgHPjGzVDO76QDry/JZqSrGUvp/bFX9c1CslbtvDJ5vAlodYJvq9Jm4jtBd3QM51O9OrLstaLr9XClN2KvL52AEsNnd00pZX9U/B7FKeU4JynWU6wSqe66jPOf7lOso14nJXEcFHjkgM6sPvA3c6e67SqyeQ6gJaz/gUeC9Cg6vIpzg7gOBM4AfmdnIaAcUDWZWCzgXeOsAq6vD5+B7PNQm06MdR7SY2S+BAuDVUjapyr87TwBdgf7ARkLNdquryzj4Ha2q/DmQKkK5jn5PQblOSdU9zwHlOijXKRaTuY4KPN+1Hmgf9joxWHbAbcwsHmgEbKuQ6CqImdUklPC86u7vlFzv7rvcfXfwfBJQ08yaV3CYEeXu64OfW4B3CTVFDFeWz0pVcAYwx903l1xRHT4HYTYXN0sPfm45wDZV/jNhZuOAs4ErggTwe8rwuxOz3H2zuxe6exHwNAd+b9XhcxAPXAi8Wdo2VflzEOOU5wSU6yjXCaNcR3nOt5TrKNeB2M51VOD5rtlAkpl1Dqr5Y4GJJbaZCBSPGn8RMKW0X/5YFPQ1fBZY6u4Pl7JN6+L++GY2hNDnqMokf2ZWz8waFD8nNODaohKbTQSutpBhwM6wpq1VSamV66r+OSgh/Pf+GuA/B9jmY+A0M2sSNGc9LVhWJZjZGOAe4Fx331vKNmX53YlZJcaeuIADv7eyfI/EulOAZe6eeaCVVf1zEOOqfZ4DynVAuU4JynWU5wDKdUC5TpjYzXXKOhpzdXkQmjFgBaGRwX8ZLHuA0C86QAKhJpzpwDdAl2jHXM7v/wRCzTIXAPOCx5nAzcDNwTa3AYsJjZg+Czgu2nGX8zXoEry3+cH7LP4chF8DAx4PPicLgeRoxx2B61CPUBLTKGxZlf8cEEryNgL5hPoUX09o/InJQBrwGdA02DYZeCZs3+uC/xvSgWuj/V7K+RqkE+pvXfz/QvEsO22BScHzA/7uxOKjlGvwcvD7voBQItOm5DUIXn/veyQWHwe6BsHyF4r/Hwjbtkp+Dqri40CfT6pRnhO8R+U6ynWKr0O1y3VK+X6rVnnOQa6Dch3lOjGf61gQoIiIiIiIiIiIxCh10RIRERERERERiXEq8IiIiIiIiIiIxDgVeEREREREREREYpwKPCIiIiIiIiIiMU4FHhERERERERGRGKcCj4hUSmZWaGbzwh73luOxO5nZovI6noiIiMjhUJ4jIpEQH+0ARERKsc/d+0c7CBEREZEIUJ4jIuVOLXhEJKaY2Roz+4uZLTSzb8ysW7C8k5lNMbMFZjbZzDoEy1uZ2btmNj94HBccKs7MnjazxWb2iZnVidqbEhEREUF5jogcHRV4RKSyqlOi6fKlYet2unsf4DHgH8GyR4EX3b0v8CrwSLD8EeALd+8HDAQWB8uTgMfd/RhgB/CDiL4bERERkf9RniMi5c7cPdoxiIh8j5ntdvf6B1i+Bhjt7qvMrCawyd2bmdlWoI275wfLN7p7czPLAhLdPS/sGJ2AT909KXj9c6Cmuz9YAW9NREREqjnlOSISCWrBIyKxyEt5fjjywp4XojHJREREpHJQniMiR0QFHhGJRZeG/fwqeD4TGBs8vwKYHjyfDNwCYGZxZtaoooIUEREROQLKc0TkiKiSKyKVVR0zmxf2+iN3L55CtImZLSB0d+qyYNntwPNmdjeQBVwbLP8x8JSZXU/oDtYtwMZIBy8iIiJyEMpzRKTcaQweEYkpQd/0ZHffGu1YRERERMqT8hwRORrqoiUiIiIiIiIiEuPUgkdEREREREREJMapBY+IiIiIiIiISIxTgUdEREREREREJMapwCMiIiIiIiIiEuNU4BERERERERERiXEq8IiIiIiIiIiIxDgVeEREREREREREYpwKPCIiIiIiIiIiMU4FHhERERERERGRGKcCj4iIiIiIiIhIjFOBR0REREREREQkxqnAIyIRY2Yfmtk10Y6jmJl1MjM3s/gybDvOzL482uOIiIhI5RJ8h3cLnj9pZr8qy7ZHcJ4rzOyTI42zopnZ/Wb2Shm3/dzMbjja44hI+VKBR0S+w8x2hz2KzGxf2OsrDudY7n6Gu794hHGsMbP9Zta8xPK5QbLV6UiOKyIiIrHNzD4yswcOsPw8M9t0ODdg3P1md/9dOcT0vZs/7v6qu592tMc+wLlGBed6t8TyfsHyz8v7nCISG1TgEZHvcPf6xQ9gLXBO2LJXi7eroNYrq4HLws7ZB6hbAecVERGRyutF4EozsxLLrwJedfeCKMRU0bKA4WbWLGzZNcCKKMUjIpWACjwiUibB3aJMM/u5mW0CnjezJmb2gZllmVl28DwxbJ9vm+8Wd3kys78G2642szMOcdqXgavDXl8DvFQirkZm9lIQQ4aZ/Z+Z1QjWxQXn22pmq4CzDrDvs2a20czWm9mDZhZ3BNemrZlNNLPtZpZuZjeGrRtiZilmtsvMNpvZw8HyBDN7xcy2mdkOM5ttZq0O99wiIiLV0HtAM2BE8QIzawKcDbwUfPd+FXy/bjSzx8ys1oEOZGYvmNmDYa/vDvbZYGbXldj2rKAl8S4zW2dm94etnhb83BG0eh5esru3mR0XfN/vDH4eF7buczP7nZnNMLMcM/ukZCvmEvYH12FssH8ccCnwavhGhzhnZzP7Ijjfp0DJVtPDzGxmcB3nm9mog8RTKjM718wWB8f53Mx6ha37eZCD5ZjZcjM7OVh+wPxJRA5OBR4RORytgaZAR+AmQv+HPB+87gDsAx47yP5DgeWEEoi/AM8e4O5buFlAQzPrFSQuY4GSfbofBRoBXYATCRWErg3W3Ugo2RsAJAMXldj3BaAA6BZscxpwwP7kh/AGkAm0Dc7xBzMbHaz7J/BPd28IdAXGB8uvCeJuTyhJvZnQ9RMREZGDcPd9hL5Pw28CXQIsc/f5QCHwE0L5xnDgZODWQx3XzMYAPwNOBZKAU0pssic4Z2NCN41uMbPzg3Ujg5+Ng1bPX5U4dlPgv8AjhL73Hwb+a99tgXM5oRymJVAriOVgXuJ/1+B0YBGw4TDO+RqQSug6/Y5QblK8b7tg3wcJ5X4/A942sxaHiOk7zKw78DpwJ9ACmAS8b2a1zKwHcBsw2N0bBO9hTbBrafmTiByECjwicjiKgN+4e56773P3be7+trvvdfcc4PeEiiylyXD3p929kFDz6jbAoVqtFLfiORVYCqwvXhFW9LnP3XPcfQ3wN0JNtCGU7P3D3de5+3bgj2H7tgLOBO509z3uvgX4e3C8MjOz9sDxwM/dPdfd5wHP8L+EKx/oZmbN3X23u88KW94M6Obuhe6e6u67DufcIiIi1diLwEVmlhC8vjpYRvCdOsvdC4Lc4N8cPD8pdgnwvLsvcvc9wP3hK939c3df6O5F7r6AUOGiLMeFUEEozd1fDuJ6HVgGnBO2zfPuviKsgNX/YAd095lA06BQcjUlWjkf7Jxm1gEYDPwqyOumAe+H7XslMMndJwXv91MghVDudDguBf7r7p+6ez7wV6AOcByhQlxtoLeZ1XT3Ne6+MtivtPxJRA5CBR4RORxZ7p5b/MLM6prZv4OuUbsINU9ufJBuTpuKn7j73uBp/UOc82VCd7TG8f3EpTlQE8gIW5YBtAuetwXWlVhXrGOw78agyfAOQglgy0PEU1JbYHtQ4DpQDNcD3YFlQdPos8Pe18fAG0Ez8L+YWc3DPLeIiEi15O5fAluB882sKzCEUIsUzKy7hbqNbwrykz9QovtRKQ6WN2BmQ81sqoW6he8k1Pq2LMctPnZGiWXh+QKE5UnAXg6dI0Eon7gNOAl4t8S6g52zLZAdFLLC1xXrCFxcnCMFedIJhG7OHY7vxODuRYSucTt3TyfUsud+YIuZvWFmbYNNS8ufROQgVOARkcPhJV7/FOgBDA2a0BY3Tz5Yt6vDO6F7BqHBls8E3imxeiuhOzwdw5Z14H+tfDYS6gIVvq7YOiAPaO7ujYNHQ3c/5jBD3EDo7lmDA8Xg7mnufhmhwtGfgQlmVs/d8939t+7em9BdrLP5blNzERERObjiLkpXAh+7++Zg+ROEWqokBfnJLyhbbnKwvAFCBaSJQHt3bwQ8GXbckjlSSRv4br5SfPz1B9j2cLxMqPvZpLCbZ2U550agiZnVK7Gu2Drg5bAcqbG713P3Px1mfN+JIeia357/5UmvufsJwTZOKFcqNX86zHOLVDsq8IjI0WhAaNyYHUE/799E6DzXA6NL3GUi6Oo1Hvi9mTUws47AXfxvnJ7xwB1mlmihwRfvDdt3I/AJ8Dcza2hmNcysq5mVtal18XHWATOBP1po4OS+QbyvAJjZlWbWIrhjtSPYrcjMTjKzPkFrp12EClVFh3NuERGRau4lQuPk3EjQPSvQgNB3624z6wncUsbjjQfGmVlvM6vL9/OaBoRa7eaa2RBCLYyLZRH6Hu9SyrEnAd3N7HIzizezS4HewAdljO2A3H01oW5ivzyccwY30FKA3wbj4ZzAd7uLvUKoK9fpFpq0IsFCE24kfv80BzUeOMvMTg5aKv+U0A22mWbWw8xGm1ltIJdQTlkEpedPh3lukWpHBR4RORr/INSPeiuhAZE/isRJ3H2lu6eUsvp2QoMergK+JHR37blg3dOEukHNB+bw/RZAVxMaxHAJkA1M4PCbHkNoKvdOhO5SvUtonKLPgnVjgMVmtpvQgIFjg771rYPz7SI0ttAXhO7CiYiISBkE4+vMBOoRallT7GeEii85hHKBN8t4vA8J5TZTgPTgZ7hbgQfMLAf4NWED/watZ34PzAi6NA0rcexthFrr/hTYBtwDnO3uW8sS2yHi/tLdNxxg+aHOeTmhCTC2EypmvRS27zrgPEKtn7IItei5m8P8+9HdlxNqYfUooXzxHOAcd99PaPydPwXLNxFqrXNfsGtp+ZOIHIS5H6o1oYiIiIiIiIiIVGZqwSMiIiIiIiIiEuNU4BERERERERERiXEq8IiIiIiIiIiIxDgVeEREREREREREYlx8tAMoL82bN/dOnTpFOwwRERGJstTU1K3u3iLacZQ35ToiIiICpec6VabA06lTJ1JSSptFWURERKoLM8uIdgyRoFxHREREoPRcR120RERERERERERinAo8IiIiIiIiIiIxTgUeEREREREREZEYV2XG4BERERGpSGY2BvgnEAc84+5/OsA2lwD3Aw7Md/fLzewk4O9hm/UExrr7e4cbQ35+PpmZmeTm5h7BOxCAhIQEEhMTqVmzZrRDEREROSoq8IiIiIgcJjOLAx4HTgUygdlmNtHdl4RtkwTcBxzv7tlm1hLA3acC/YNtmgLpwCdHEkdmZiYNGjSgU6dOmNnRvKVqyd3Ztm0bmZmZdO7cOdrhiIiIHBV10RIRERE5fEOAdHdf5e77gTeA80pscyPwuLtnA7j7lgMc5yLgQ3ffeyRB5Obm0qxZMxV3jpCZ0axZM7WAEhGRKkEFHhEREZHD1w5YF/Y6M1gWrjvQ3cxmmNmsoEtXSWOB10s7iZndZGYpZpaSlZVV2jaHF7l8h66fiIhUFSrwlMGrX2fw8eJN0Q5DREREYks8kASMAi4DnjazxsUrzawN0Af4uLQDuPtT7p7s7sktWrSISJD7CwpZv2Mf+YVFETm+iIiIVAwVeA4hv7CIt1Iy+cmb81iyYVe0wxEREZHKYT3QPux1YrAsXCYw0d3z3X01sIJQwafYJcC77p4f0UgPochh2+48duwt3zDGjRvHhAkTyv2YnTt3pn///vTv35958+YBobF07rjjDrp160bfvn2ZM2dOhcUkIiJSWajAcwg142rw1NWDaFSnJje+lMLW3XnRDklERESibzaQZGadzawWoa5WE0ts8x6h1juYWXNCXbZWha2/jIN0z6ooCTXjqFMzjh1790c7lDJ56KGHmDdvHvPmzaN///4AfPjhh6SlpZGWlsZTTz3FLbfcEt0gRUREokAFnjJo2SCBp69OZtuePG5+OZW8gsJohyQiIiJR5O4FwG2EulctBca7+2Ize8DMzg02+xjYZmZLgKnA3e6+DcDMOhFqAfRFhQd/AE3q1WJf/v+3d9/hUZV5G8e/T3pCGilASKGHDiGELqAigqigYgEL2MCGdfVdXXfVde29YUFUxBVsKIKCDSnSa2ih1yS0EAglkP68f2RwIwIGyOSk3J/rmouZM+ecuecwgSe/eUohR/POrI3zn//8h6ZNm3LOOecwePBgXnrppT88P23aNNq1a0fr1q25+eabyc3N5YcffuCqq676fZ8ZM2ZwySWXAPDTTz/RpUsXEhMTueqqqzh8+PApX//bb79lyJAhGGPo3LkzWVlZ7Ny5E2stI0aMoGnTplxwwQXs2XOiea5FRESqBi2TXkqtokN4+aoE7hq3lH9+s4oXrmyjSflERESqMWvtFGDKcdseK3HfAg+4bscfu5U/T8p8Vv49efUZDye3wJG8Arw9PPDx+t/3fy3qBvP4pS1PeeyiRYuYMGECy5cvJz8/n8TERNq3b//78zk5Odx4441MmzaN+Ph4hgwZwjvvvMOIESMYPnw42dnZ1KhRg88//5xBgwaxd+9ennrqKX755Rdq1KjB888/zyuvvMJjjxVf2kcffZQnn3ySXr168dxzz+Hr60t6ejqxsf8bMRcTE0N6ejrz5s1j3bp1pKSksHv3blq0aMHNN998RtdIRESkolMPntNwcZso7rugCV8uSeOD2VucjiMiIiJSJgzg5WEoKLKnfeycOXMYMGAAfn5+BAUFcemll/7h+XXr1tGgQQPi4+MBGDp0KLNmzcLLy4u+ffsyefJkCgoK+P777xkwYADz588nJSWFbt26kZCQwMcff8y2bdsAePbZZ1m7di2LFi1i3759PP/886fMNmvWLAYPHoynpyd169bl/PPPP+33JyIiUlmoB89puuf8JmzYfZhnpqyhUWQg5zWr5XQkERERkb/safNXDhzNZ1tmNvXDaxDs711GqU5t0KBBvPXWW4SFhZGUlERQUBDWWnr37s348X+enigqKgoAX19fbrrppt+HgkVHR5Oa+r9V69PS0oiOLtMOUiIiIhWeevCcJg8Pw0tXtaVF3WDuGb+MDbsPOR1JRERE5KwF+Xnh5eHB/tOcbLlbt25MnjyZnJwcDh8+zHffffeH55s2bcrWrVvZuHEjAJ988gk9e/YEoGfPnixdupT333+fQYMGAdC5c2fmzJnz+/7Z2dmsX78egJ07dwLFq2ZNnDiRVq1aAdC/f3/Gjh2LtZb58+cTEhJCVFQUPXr04PPPP6ewsJCdO3cyffr0M7w6IiIiFZ8KPGfA38eT94ck4efjya1jF7M/u3KsOiEiIiJyMh7GEBrgzcGcAgoKi0p9XIcOHejfvz9t2rThoosuonXr1oSEhPz+vJ+fHx999BFXXXUVrVu3xsPDg9tvvx0AT09PLrnkEqZOnfr7BMuRkZGMGTOGwYMH06ZNG7p06cLatWsBuO6662jdujWtW7dm7969/POf/wSgX79+NGzYkMaNGzNs2DDefvttAC6//HKaNGlCixYtGDJkCF26dCmTayUiIlIRmeL5/yq/pKQku3jx4nJ9zWXb93PNqPm0j6vJ2Fs64u2pepmIiIjTjDFLrLVJTucoaydq66xZs4bmzZuX2WsczStgw57DRIf6Ex7oW+rjDh8+TGBgIEeOHKFHjx6MGjWKxMTEMsvlbmV9HUVERNzpZG0dVSTOQru4mjw/sDXzNmfy78mrnY4jIiIiclb8vD3x8/Zk/5H80zpu+PDhJCQkkJiYyMCBAytVcUdERKSq0CTLZ+nydjGs332Yd2ZsomntIG7oUt/pSCIiIiJnxBhDzQAfdh44Sk5+IX7enqU6bty4cW5OJiIiIn9FPXjKwEMXNuWC5rV4YnIKczbudTqOiIiIVCNlPdw+NMAbgzntyZYrq6oyXYGIiIgKPGXAw8Pw2qB2NI4M5M5Pl7Jlb7bTkURERKQa8PPzIzMzs0yLFN6eHgT6eZF1JL/KFz+stWRmZuLn5+d0FBERkbOmIVplJNDXi9FDk+j/1mxu+XgR39zZjRB/b6djiYiISBUWExNDWloaGRkZZXreo3mFZGbnkZPhU+phWpWVn58fMTExTscQERE5ayrwlKHYsADevb49141ewN3jl/Hh0CS8tLKWiIiIuIm3tzcNGjQo8/Pm5BfS8elfOL9ZLV4b1KrMzy8iIiJlT9WHMtapYThPXdaKWeszeHbqWqfjiIiIiJw2P29PLm1blx9W7+JQzumtqCUiIiLOUIHHDQZ1jOOmbvX5YPYWPl+03ek4IiIiIqdtYPsYcvKLmLJyp9NRREREpBRU4HGTR/s1p3uTCP45cRULt+xzOo6IiIjIaWkXG0rDiBpMWJLudBQREREpBRV43MTL04O3rk0ktmYAt/93Can7jjgdSURERKTUjDEMbB/Dwq372J6pdoyIiEhFpwKPG4X4ezN6aBIFhUUMG7uYw7kFTkcSERERKbUrEqMxBiYsTXM6ioiIiPwFFXjcrGFkICOvS2TDnsPc91kyRUXW6UgiIiIipRIV4k+3RhFMWJqmNoyIiEgF59YCjzGmrzFmnTFmozHm4ZPsc7UxJsUYs9oYM67E9kJjTLLrNsmdOd2te5NIHrukBb+s2c1LP61zOo6IiIiUgbNs58QZY34yxqxxPV+/3IKfpoHto0nbf5SFWzWnoIiISEXm5a4TG2M8gZFAbyANWGSMmWStTSmxTxPgEaCbtXa/MaZWiVMctdYmuCtfeRvSpR7rdh/i7RmbiK8dxGXtop2OJCIiImeoDNo5Y4GnrbU/G2MCgaJyjH9a+rSsQw2fVUxYkkbnhuFOxxEREZGTcGcPno7ARmvtZmttHvAZMOC4fYYBI621+wGstXvcmMdRxhj+3b8lnRuG8X8TVrBs+36nI4mIiMiZO+N2jjGmBeBlrf3Ztf2wtbbCzmIc4OPFxW2imLJyJ0fyNJ+giIhIReXOAk80kFricZprW0nxQLwxZo4xZr4xpm+J5/yMMYtd2y870QsYY4a79lmckZFRpuHdwdvTg3eua0+dYD+Gf7KEnQeOOh1JREREzszZtHPigSxjzNfGmGXGmBddPYL+pKK0dQYmxpCdV8iPq3c5lkFEREROzelJlr2AJsC5wGDgfWNMqOu5etbaJOBa4DVjTKPjD7bWjrLWJllrkyIjI8sp8tmpWcOH0UOTOJpXyLCxizmaV+h0JBEREXGPk7VzvIDuwINAB6AhcOOJTlBR2jod6ocRG+bPV0u0mpaIiEhF5c4CTzoQW+JxjGtbSWnAJGttvrV2C7Ce4oYQ1tp015+bgRlAOzdmLVfxtYN4Y3ACq3cc5MEvl2OtVqUQERGpZM6mnZMGJLuGdxUAE4FE90c+cx4ehivaxTB3UyY7stQDWUREpCJyZ4FnEdDEGNPAGOMDDAKOXw1rIsXfamGMiaC4y/JmY0xNY4xvie3dgBSqkPOb1eaRi5rx/cqdvDFto9NxRERE5PSccTvHdWyoMeZYl5zzqQTtnIGJMVgL3yw7vo4lIiIix3w4ewt//2qFIx053FbgcX0jNQL4EVgDfGGtXW2MedIY09+1249ApjEmBZgOPGStzQSaA4uNMctd258ruSpFVTGse0MGJsbw6i/rmbJyp9NxREREpJTOpp1jrS2keHjWNGPMSsAA75f/uzg9ceEBdGwQxoQlaep9LCIichI/pexiza6DGGPK/bXdtkw6gLV2CjDluG2PlbhvgQdct5L7zAVauzNbRWCM4ZkrWrFl72Ee+CKZuLAAWkWHOB1LRERESuFM2zmu534G2rg7Y1m7MjGmeDXQ1CwS42o6HUdERKRCKSyyrEw7wMD2MY68vtOTLFd7vl6evHdDEmEBPgwbu5g9h3KcjiQiIiJyQhe1roOft4cmWxYRETmBDXsOkZ1XSEJsqCOvrwJPBRAZ5Mv7Q5PIOpLPDaMXsi0z2+lIIiIiIn8S5OdN35Z1+G75DnLytRKoiIhIScnbswBU4KnuWtYNYfTQJHYdzOHSN2czfe0epyOJiIiI/MnA9jEczCnglzW7nY4iIiJSoSSnZhHs50WDiBqOvL4KPBVIt8YRfHf3OcTUDODmjxfxxrQNFBVpEkMRERGpOLo2iiAqxI8JGqYlIiLyB8mpWbSNDXVkgmVQgafCiQ0LYMIdXbk8IZpXfl7P8E8Wc+BovtOxRERERADw9DBc3i6aWRv2sueg5g4UEREByM4tYP3uQ7RzaHgWqMBTIfn7ePLy1W35d/+WzFiXwYC3ZrNu1yGnY4mIiIgAxcO0CossE5PTnY4iIiJSIaxMP0CRhYS4UMcyqMBTQRljGNq1Pp8N70x2XiGXjZzD5OU7nI4lIiIiQqPIQBJiQ5mwJJ3i1eBFRESqt+TULADaxoQ6lkEFngouqX4Y3999Di3rBnP3+GU8/X0KBYVFTscSERGRam5g+xjW7T7E6h0HnY4iIiLiuOTtWcSFBRAe6OtYBhV4KoFawX6MG9aZoV3q8f5vW7j+gwXsPZzrdCwRERGpxvq3qYuPpwdfabJlERERklOzHFse/RgVeCoJHy8P/j2gFa9c3ZZl27O49M3ZLNu+3+lYIiIiUk2FBHjTu0Vtvk1OJ69AvYtFRKT62nUgh10Hc1TgkdNzRWIME+7oiqeH4Zr35jN+4XanI4mIiEg1NbB9NPuP5DN93R6no4iIiDgmObW484WTEyyDCjyVUqvoECaPOIfOjcJ55OuVPDxhBTn5hU7HEhERkWqmR5NIIgJ9maBhWiIiUo0tS83C29PQIirY0Rwq8FRSNWv48NGNHRhxXmM+W5TKNe/NY0fWUadjiYiISDXi5enBZQl1mb5uD/uy85yOIyIi4ojlqVk0jwrGz9vT0Rwq8FRinh6GB/s05b0b2rMpI5tL3pzN3I17nY4lIiIi1cjA9jHkF1omJac7HUVERKTcFRZZVqYdcHz+HVCBp0ro07IO347oRlgNH67/YAGjZm3CWut0LBEREakGmkcF07JuMF8t1TAtERGpfjbsOUR2XqEKPFJ2GkUGMvGubvRtVYdnpqxlxLhlZOcWOB1LREREqoGBiTGsSj/Iul2HnI4iIiJSrpK3ZwGowCNlK9DXi5HXJvLIRc2Yumonl42cw+aMw07HEhERkSpuQEJdvDwME9SLR0REqpnk1CxC/L1pEFHD6Sgq8FQ1xhhu69mIT27pxN7DuQx4aw4/rd7ldCwREZEqxxjT1xizzhiz0Rjz8En2udoYk2KMWW2MGVdie6ExJtl1m1R+qd0jPNCX85rV4ptl6RQUFjkdR0REpNwkp2bRNjYUY4zTUVTgqaq6NY5g8t3nUD+iBsM/WcLLP62jsEjz8oiIiJQFY4wnMBK4CGgBDDbGtDhunybAI0A3a21L4L4STx+11ia4bv3LKbZbDUyMIeNQLr9pwQcREakmsnMLWL/7UIUYngUq8FRpMTUD+PL2LlzVPoY3f93IzWMWkXVES5iKiIiUgY7ARmvtZmttHvAZMOC4fYYBI621+wGstXvKOWO5Or9ZLWoGePPVEg3TEhGR6mFF2gGKLLRTgUfKg5+3Jy9c2YanL2/F3E17ufSt2azeccDpWCIiIpVdNJBa4nGaa1tJ8UC8MWaOMWa+MaZvief8jDGLXdsvO9mLGGOGu/ZbnJGRUWbh3cHHy4P+bevyc8puDhzJdzqOiIiI2y1PywKgrQo8Ul6MMVzXqR6f39aF/ALLFW/P5aslaVpKXURExL28gCbAucBg4H1jTKjruXrW2iTgWuA1Y0yjE53AWjvKWptkrU2KjIwsh8hnZ2D7GPIKivhu5Q6no4iIiLhd8vYs4sICCKvh43QUQAWeaiUxriaT7z6HhNhQHvxyOcM/WcKegzlOxxIREamM0oHYEo9jXNtKSgMmWWvzrbVbgPUUF3yw1qa7/twMzADauTtweWgdHUJ87UAmaJiWiIhUA8mpWRVm/h1QgafaiQzyZdywzjzarzmz1mfQ+9VZfL1UvXlERERO0yKgiTGmgTHGBxgEHL8a1kSKe+9gjImgeMjWZmNMTWOMb4nt3YCUcsrtVsYYBibGsHR7FpszDjsdR0RExG12Hchh18EcFXjEWZ4ehmE9GjLl3u40rhXIA18s59aPF7NbvXlERERKxVpbAIwAfgTWAF9Ya1cbY540xhxbFetHINMYkwJMBx6y1mYCzYHFxpjlru3PWWurRIEH4PJ20XgYmLBUvXhERKTqSk7dD0BCXKizQUpQgacaaxQZyBe3deGfFzdn9sa99H5lpubmERERKSVr7RRrbby1tpG19mnXtsestZNc96219gFrbQtrbWtr7Weu7XNdj9u6/vzAyfdR1moF+9G9SSTfLE2nqEhtChERqZqWpWbh7WloERXsdJTfqcBTzXl6GG7t3pAf7utB0zpBPPjlcm4es4hdB9SbR0RERM7MwPYx7DiQw7zNmU5HERERcYvk7Vm0iArGz9vT6Si/U4FHAGgQUYPPh3fh8UtbMG9zJr1fnckXi1PVm0dERERO24UtahPk56XJlkVEpEoqLLKsTD9QoebfARV4pAQPD8NN3Rrww709aB4VzP99tYIbP1rEjqyjTkcTERGRSsTP25NL2tRl6qpdHM4tcDqOiIhImdqw5xBH8gor1Pw7oAKPnED9iBp8Nqwz/+7fkoVb9tHn1Vl8vmi7evOIiIhIqV3ZPpqj+YVMWbnT6SgiIiJlKnl7FgBtY0IdzXE8FXjkhDw8DEO71ufH+3rQMjqYv09YyZAPF5Ku3jwiIiJSColxNWkQUUPDtEREpMpJTs0ixN+bBhE1nI7yB24t8Bhj+hpj1hljNhpjHj7JPlcbY1KMMauNMeNKbB9qjNngug11Z045ubjwAMbd2pn/DGjJkm376fPqLMYtUG8eEREROTVjDFe0i2bBln2k7jvidBwREZEyk5yaRdvYUIwxTkf5A7cVeIwxnsBI4CKgBTDYGNPiuH2aAI8A3ay1LYH7XNvDgMeBTkBH4HFjTE13ZZVT8/Aw3NCluDdPm5gQ/vHNSm74YCFp+9VYExERkZO7on0MxsDXS9OdjiIiIlImsnMLWL/7UIWbYBnc24OnI7DRWrvZWpsHfAYMOG6fYcBIa+1+AGvtHtf2PsDP1tp9rud+Bvq6MauUQmxYAJ/e2omnL2/Fsu3FvXn+O38bRUXqzSMiIiJ/Fh3qT5eG4UxYmqbevyIiUiWsSDtAkYV21azAEw2klnic5tpWUjwQb4yZY4yZb4zpexrHigOMMVzXqR4/3t+DdnE1+efEVVz/wQJ1vRYREZETGpgYw/Z9R1i0db/TUURERM5acmoWAG2rWYGnNLyAJsC5wGDgfWNMaGkPNsYMN8YsNsYszsjIcE9COaGYmgF8cktHnr2iNSvSDtDntVl8Mm+revOIiIjIH/RtVYcAH09NtiwiIlVCcup+6oUHEFbDx+kof+LOAk86EFvicYxrW0lpwCRrbb61dguwnuKCT2mOxVo7ylqbZK1NioyMLNPw8teMMQzuGMeP9/egfb2a/Ovb1Vw7ej7bM9WbR0RERIrV8PXiolZRfL9yJ0fzCp2OIyIiclaSU7Mq5Pw74N4CzyKgiTGmgTHGBxgETDpun4kU997BGBNB8ZCtzcCPwIXGmJquyZUvdG2TCig61J+xN3fkhYFtWJ1+kD6vzWLMnC3qzSMiIiIAXNk+hsO5BfyUssvpKCIiImds14Ecdh/MrX4FHmttATCC4sLMGuALa+1qY8yTxpj+rt1+BDKNMSnAdOAha22mtXYf8B+Ki0SLgCdd26SCMsZwdYdYfnqgBx0bhPHE5BQGvT+fLXuznY4mIiIiDuvUIIzoUH++0jAtERGpxJJTi+eTq4jz70DxHDhuY62dAkw5bttjJe5b4AHX7fhjPwQ+dGc+KXtRIf6MuakDXy1J48nvUuj18gx6xEdyVftYLmhRC18vT6cjioiISDnz8DAMTIzmzekb2XngKFEh/k5HEhEROW3LUrPw9jS0iAp2OsoJOT3JslRBxhiuSorllwd6cse5jVi78xB3jVtKp2em8fi3q1iVfkBLpYqIiFQzA9vH4GEM//kuRe0AERGplJK3Z9EiKhg/74rZcUEFHnGb2sF+PNSnGXMePp+Pb+5I9yaRjF+UyiVvzuai139j9G+b2Xs41+mYIiIiUg7qhdfg//o0ZcrKXXw0Z6vTcURERE5LYZFlZfqBCjv/DqjAI+XA08PQMz6SNwe3Y9E/LuA/l7XC18uDp75fQ+dnpjFs7GJ+Wr2L/MIip6OKiIicFmNMX2PMOmPMRmPMwyfZ52pjTIoxZrUxZtxxzwUbY9KMMW+VT2JnDe/RkN4tavPMlDUs2abpFUVEpPJYv/sQR/IKSYgLdTrKSbl1Dh6R44UEeHND53rc0Lke63cf4qslaXy9NJ2fU3YTEejDZQnRXJUUS9M6QU5HFREROSVjjCcwEugNpAGLjDGTrLUpJfZpAjwCdLPW7jfG1DruNP8BZpVXZqcZY3jpqrZc+uZs7vp0Gd/fcw7hgb5OxxIREflLyalZACTE1nQ2yCmoB484Jr52EP/o15x5j5zP6CFJJNUL4+N5W+nz2iwufXM2Y+dtJetIntMxRURETqYjsNFau9lamwd8Bgw4bp9hwEhr7X4Aa+2eY08YY9oDtYGfyilvhRDi783b1yWy70ge936WTGGR5uMREZGKL3l7FqEB3tQPD3A6ykmpwCOO8/b04IIWtXn3hvbMf6QXj13SgoIiy2Pfrqbj09O4a9xSZqzbowagiIhUNNFAaonHaa5tJcUD8caYOcaY+caYvgDGGA/gZeDBU72AMWa4MWaxMWZxRkZGGUZ3VqvoEP4zoCWzN+7l9WkbnI4jIiLyl5anZdE2JhRjjNNRTkpDtKRCCQ/05eZzGnDzOQ1YveMAXy5O49vkdL5fsZPawb5ckRjDle1jaBQZ6HRUERGR0vACmgDnAjHALGNMa+B6YIq1Nu1UDUVr7ShgFEBSUlKV+qbj6qRYFm3dz5u/biAxLpRzmx4/ek1ERKRiyM4tYP3uQ/RpWcfpKKekAo9UWC3rhtCyfwiP9GvGr2v28NWSNEbN2sw7MzaRGBfKVUmxXNImiiA/b6ejiohI9ZQOxJZ4HOPaVlIasMBamw9sMcasp7jg0wXoboy5EwgEfIwxh621J5youSoyxvCfAa1YlX6A+z5P5vt7uhMd6u90LBERkT9ZkXaAIkuFXkELNERLKgFfL08uah3FBzd2YN7D5/PIRc04mFPAI1+vpMPTv3D/58ks2JzpdEwREal+FgFNjDENjDE+wCBg0nH7TKS49w7GmAiKh2xtttZeZ62Ns9bWp3iY1tjqVNw5xt/Hk3eub09hoeXOT5eSW1DodCQREZE/OTbBclsVeETKTq1gP27r2Yif7+/BxLu6MTAxhl/W7OaaUfO5fvQCVqUfcDqiiIhUE9baAmAE8COwBvjCWrvaGPOkMaa/a7cfgUxjTAowHXjIWqtvJUpoEFGDF69qw/LULJ75fo3TcURERP4kOXU/9cIDCKvh43SUUzLWVo3h3ElJSXbx4sVOxxAH5OQXMm7Bdt78dQP7j+Rzebto/nZhPDE1K+7s5iIi4j7GmCXW2iSnc5S1qt7Weeq7FEbP3sIbg9vRv21dp+OIiIj8rtMzv9C5YTivD2rndBTg5G0d9eCRSs/P25Obz2nAzP87jzvPbcSUlTs5/6WZPP19ipZZFxERqST+flEzkurV5OEJK9i455DTcURERADYeeAouw/mVvj5d0AFHqlCgv28+b++zZjx0LkMSKjL6Nlb6PHCdN6buYmcfI3pFxERqci8PT1469pEAnw8uf2/S8nOLXA6koiICMnbs4CKP8EyqMAjVVBUiD8vXtWWqfd2p329mjw7dS29Xp7J10vTKCqqGkMSRUREqqI6IX68PqgdmzMO88jXK6kqUwmIiEjllZyahY+nBy3qBjsd5S+pwCNVVrM6wXx0U0fGDetEeKAPD3yxnIvfnM2s9RlORxMREZGT6NY4ggd6xzNp+Q7+O3+b03FERKSaS07NonndYHy9PJ2O8pdU4JEqr2ujCCbe2Y03BrfjcG4+Qz5cqBW3REREKrA7z23MeU0jefK7lN+XphURESlvhUWWlekHaFcJhmeBCjxSTXh4GPq3rcsvD/TksUtasHrHAS55czb3fbaM1H1HnI4nIiIiJXh4GF69JoFaQX7c9elS9mdr0QQRESl/63cf4kheIW1jQ5yOUioq8Ei14uv1xxW3pq7aRa+XteKWiIhIRRMa4MM71yeScSiX+79I1jx6IiJS7o71Ik2IrelskFJSgUeqpZIrbl3WTituiYiIVERtYkL516UtmLEug5HTNzodR0REqpnk7VmEBnhTPzzA6SilogKPVGtRIf68cOUfV9w6/6UZTFiSRqG+KRQREXHc9Z3iuCyhLq/8sp7ZG/Y6HUdERKqR5NQs2saEYoxxOkqpqMAjwh9X3IoI8uVvXy7nkjdnM3N9hpZoFRERcZAxhmeuaE3jyEDu/WwZuw7kOB1JRESqgcO5Bazfc4iESjLBMqjAI/IHx1bcetO14tbQDxdywwcLteKWiIiIgwJ8vHjn+vYczS/krnFLyS8scjqSiIhUcSvSsrAWEuJCnY5SairwiBzHw8Nwadu6THvgXB6/VCtuiYiIVASNawXy/MA2LNm2n+emrnU6joiIVHHLU4u/5E+ICXU2yGnwcjqASEXl4+XBTd0aMLB9DO/O2MQHs7cwecVOujQMp1/rKC5sWZuIQF+nY4qIiFQbl7aty5Jt+/lg9haS6tXkotZRTkcSEZEqKjl1P/XDA6hZw8fpKKWmHjwif6Hkilu39WhIetZR/vHNSjo+/QvXvj+fT+ZvY88hzQcgIiJSHv7RrzkJsaE89NUKNmccdjqOiIhUUcmpWbStRPPvgAo8IqUWFeLP//Vtxq9/68nUe7sz4rzG7D6Yw78mrqLTM9O45r15fDx3K7sPqtgjIiLiLj5eHoy8LhFvT8Odny7laF6h05FERKSK2XngKLsP5laqCZahlAUeY8y9xphgU+wDY8xSY8yF7g4nUhEZY2geFcwDFzbllwd68tP9Pbjn/CbsP5LH45NW0/nZaVz17lw+nL2FnQeOOh1XRERKQW2dyiU61J/XBrVj3e5D/HPiqnJf8XLPoRwmLEnj/75azqcLtlFYpBU3RUSqkuTtWQCVrsBT2jl4brbWvm6M6QPUBG4APgF+clsykUrAGEN87SDiewdxf+94Nu45xJSVu5iycidPfpfCk9+lkBgXSr/WUfRtVYeYmgFORxYRkRNTW6eS6RkfyT3nN+H1aRvoUL8mgzrGue218gqKWLJtPzPXZzBrfQYpOw8CUMPHky8Wp/Hf+dt54tIWdGoY7rYMIiJSfpJTs/Dx9KBF3WCno5yW0hZ4jOvPfsAn1trVxhhzqgNEqqPGtYK4p1cQ9/RqwuaMw0xdVVzseer7NTz1/RraxobSr1Ud+rWOIjZMxR4RkQrkjNo6xpi+wOuAJzDaWvvcCfa5GngCsMBya+21xph6wDcU96b2Bt601r5bJu+kGrmnVxOWbt/PY5NW0yo6hFbRIWV27u2ZR5i5IYOZ6zKYt2kv2XmFeHkY2teryUN9mtIzPpIWUcFMXbWLZ6as4ZpR87m4TRT/6Nec6FD/MsshIiLlb1lqFs3rBuPr5el0lNNiStOl1RjzERANNADaUtyImWGtbe/eeKWXlJRkFy9e7HQMkRPaujebqat2MXXVTlakFS+31zo6hIta16FfqyjqR9RwOKGISNVhjFlirU06zWNOu61jjPEE1gO9gTRgETDYWptSYp8mwBfA+dba/caYWtbaPcYYH4rbYbnGmEBgFdDVWrvjZK+nts6J7cvO4+I3fsPL0/DdiO6EBHif0XmO5BUwf3Mms9bvZeb6DLbszQYgpqY/PeMj6RkfSZdG4QT5/fn8R/MKeXfmJt6duQlj4I6ejbmtZ0P8vCvXLwYiIgIFhUW0fuInrukQyxP9Wzod54RO1tYpbYHHA0gANltrs4wxYUCMtXZFmSc9Q2r0SGWRuu8IU1ftZMrKXSSnZgHQIiqYfq3rcFHrKBpFBpb6XEVFluy8Ag7nFnA4x/Xnie7nldhW4jlr4e8XNaNnfKSb3q2ISPk7wwLPabd1jDFdgCestX1cjx8BsNY+W2KfF4D11trRpzhPOLAM6KwCz5lZun0/V787j3Ob1mLUDe3x8PjrjubWWtbvPszM9XuYtX4vC7fsI6+wCD9vDzo3DP+9qNMgogal7bietv8Iz05Zy/crdxId6s8/L25O31Z1Sn28iIg4b83Og1z0+m+8dk0Cl7WLdjrOCZ2srVPaIVpdgGRrbbYx5nogkeLuyCJymmLDAhjeoxHDezQiPesoU1fuZOqqXbz003pe+mk9zeoEcV6zWngYOJxTwCFXUSbbVaQ5lFtA9u/bSrdyiI+XB0G+XgT6eVHDp/jPOsF+bNmbzS1jFvHy1W0ZkFAx//ESESknZ9LWiQZSSzxOAzodt088gDFmDsW9gp6w1v7g2hYLfA80Bh46UXHHGDMcGA4QF+e+OWYqu8S4mjx6cXP+PTmFUb9t5vaejU6434Ej+czeuPf3os4u18qX8bUDGdKlHj2bRtKhftgZ97yJqRnAyOsSuX5TJv+evJo7Pl1Kl4bhPN6/Bc3qVK55HEREqqtjX8JXtgmWofQFnneAtsaYtsDfgNHAWKDnqQ76q3HpxpgbgReBdNemt459w2WMKQRWurZvt9b2L2VWkUojOtSfW7s35NbuDdl54Cg/rNrF1JW7eHfmJjyMIdDX6383Py9CA3yIqRnw++Mavl7/K9yUuF/yuBq+Xvh4nXjBvIM5+Qz7eDH3fpbMvuw8burWoJyvgIhIhXFGbZ1S8AKaAOcCMcAsY0xra22WtTYVaGOMqQtMNMZ8Za3dXfJga+0oYBQU9+A5yyxV2o1d67N4235e+GEtCbGhdG4YTmGRZWX6AWauy2Dm+j0kp2ZRZCHIz4vuTSLo0SSSHvGR1C3jOXO6NArnu7vPYdzC7bz803r6vf4bN3Sux/294wkN8CnT1xIRkbKVvD2L0ABv6oVXvjlTS1vgKbDWWmPMAIqLMB8YY2451QGucekjKTEu3RgzqeS4dJfPrbUjTnCKo9bahFLmE6n0okL8ualbA27q1oD8wiK8PIzbu3QH+3nz8c0duWf8Mv49OYW9h3N58MKm6kouItXRabd1KP6CKrbE4xj+96XVMWnAAmttPrDFGLOe4oLPomM7WGt3GGNWAd2Br872jVRXxhieH9iGNTsPMmLcMro0Cue3DRlkHcnHGGgTHcKI8xrTs2kkbWNC8fI88ZcfZcXL04MhXepzaZu6vPLzej6Zv41Jy3fwwIVNubZjHJ6lGEYmIiLlLzk1i7YxoZXyd6LS/s92yDWu/Abge9c49b+awa4jsNFau9lamwd8Bgw486gi1Ye3p0e5/YPi5+3J29clMqhDLCOnb+KRr1dSUFhULq8tIlKBnElbZxHQxBjTwDVp8iBg0nH7TKS49w7GmAiKh2xtNsbEGGP8XdtrAucA68rovVRbgb5evHt9e3LyC5m3KZPzm9Xi9UEJLPlnb74dcQ4PXNiU9vXC3F7cKalmDR/+c1krvr+nO/G1g/jXxFVc8uZsFmzOLLcMIiJSOodzC1i/51ClHJ4Fpe/Bcw1wLXCztXaXMSaO4qFVp1KacekAA40xPSheheJ+V3dlAD9jzGKgAHjOWjvx+AM1Ll2kbHh5evDsFa2JCPTlrekb2ZedxxuD22n1DxGpTk67rWOtLTDGjAB+pHg4+oeu5dWfBBZbaye5nrvQGJMCFFI8106mMaY38LIxxlK8RPtL1tqVJ3kpOQ3xtYNY9OgF+Hp5lGqy5fLSPCqYz4Z3ZspKLasuIlJRrUjLwlpIiAt1OsoZKdUqWgDGmNpAB9fDhdbaPX+x/5VAX2vtra7HNwCdSg7Hcq0acdi1ROhtwDXW2vNdz0Vba9ONMQ2BX4Fe1tpNJ3s9rSwhUjbGzNnCE5NT6NggjNFDkwg+wXKwIiIV2ZmsouU67rTaOuVNbZ2qQ8uqi4hUTG/P2MgLP6xj2b96U7NGxZ0z7WRtnVL1TzXGXA0sBK4CrgYWuAo4p/KX49KttZnW2lzXw9FA+xLPpbv+3AzMANqVJquInJ0buzXg9UEJLNu+n2vem88e1wojIiJV2Rm2dUTOiL+PJ/f3jmfa33pyfrNavPrLenq9PJOpK3dS2i9fRUSk7CVvz6J+eECFLu6cSmkHID8KdLDWDrXWDqF4fp1//cUxfzku3RgTVeJhf2CNa3tNY4yv634E0A04fnJmEXGTAQnRfDC0A9sysxn47ly27s12OpKIiLudSVtH5KzE1Azg7evaM25YJ4L8vLjj06Vc+/4C1u466HQ0EZFqx1pLcmpWpZ1/B0pf4PE4rpty5l8da60tAI6NS18DfHFsXLox5tiS5/cYY1YbY5YD9wA3urY3Bxa7tk+neA4eFXhEylGP+EjGDevM4ZwCrnx3LqvSDzgdSUTEnU67rSNSVro2iuC7u8/hyQEtSdl5kH6v/8bj364i60ie09FERKqNXQdz2HMot1IXeEo7yfIPxpgfgfGux9cAU/7qIGvtlOP3s9Y+VuL+I8AjJzhuLtC6lNlExE0SYkP58vauDP1wIYNGzWfUkPZ0bRThdCwREXc4o7aOSFnRsuoiIs5K3p4FQEJcTWeDnIVSfTNlrX0IGAW0cd1GWWv/7s5gIlIxNK4VyFd3dKFuqB83friIqSt3Oh1JRKTMqa0jFcWJllW/+I3fWJmmnrQiIu6UnJqFj6cHzaOCnI5yxkrd9dhaO8Fa+4Dr9o07Q4lIxRIV4s8Xt3WhVXQwd45byqcLtjkdSUSkzKmtIxXJsWXV37q2HVlH8rlu9Hw27D7kdCwRkSprWWoWzesG4+tVeVc0PGWBxxhzyBhz8AS3Q8YYzf4mUo2EBvjw6a2dOTc+kke/WcUb0zZopQ8RqfTU1pGKzBjDJW3q8uXtXfD19mTohwvZdUCrW4p7HckrYNqa3RQVqZ0n1UdBYREr0w7QrhLPvwN/PVFykLU2+AS3IGttcHmFFJGKwd/Hk1FDkrgiMZpXfl7P45NW6z9/EanU1NaRyiA2LICPbuzAgaP53PjRQg4czXc6klRRBYVF3PnpUm75eDHP/bDW6Tgi5Wb97sMczS+s1BMsg1aHEJHT5O3pwUtXtmV4j4aMnbeNez5bRm5BodOxREREqrRW0SG8e0N7Nu45zG2fLNb/vVLmrLU8MXk1M9Zl0C4ulFGzNjNuwXanY4mUi+TULAAVeESk+vHwMPyjX3MeuagZ363YyS1jFnM4t8DpWCIiIlVa9yaRvHhVG+Zv3sffvliuXrRSpj6YvYX/zt/ObT0a8uVtXTi3aST/+nYVs9ZnOB1NxO2SU/dTM8CbeuEBTkc5KyrwiMgZu61nI168sg3zNmdy7fvzyTyc63QkERGRKu3ydjE87PqC5ekpa5yOI1XED6t28fSUNVzUqg5/79sML08P3hzcjia1Arnr06Ws26UJvqVqW556gLaxoRhjnI5yVlTgEZGzclVSLO9d3551uw5x1bvzSN13xOlIIiIiVdptPRpyY9f6fDB7C6N/2+x0HKnkklOzuO/zZbSNCeXVaxLw8Cj+BTfIz5sPbuyAn48nN49ZRMYhfZEnVdPh3ALW7zlU6YdngQo8IlIGLmhRm09v7cTew7lc+e5c1u7SwjMiIiLuYozhX5e0oF/rOjz1/Rq+TU53OpJUUqn7jnDrx4uICPRl9NAk/Lz/uDx0dKg/HwxNIjM7l1vHLuZonuZ+kqpnRVoW1lb++XdABR4RKSNJ9cP48vauAFz97jwWbd3ncCIREZGqy9PD8MrVCXRsEMaDXy5n7sa9TkeSSubA0XxuHrOIvIIixtzUgYhA3xPu1yYmlNcHtWNFWhYPfJGsuZ+kyqkqEyyDCjwiUoaa1gliwh1diQj05frRC/glZbfTkURERKosP29P3r8hiQYRNRj+yRJSdqgHrZROfmERd366hC17s3n3hvY0rhV0yv37tKzDPy5qztRVu3jhx3XllFKkfCRvz6J+eAChAT5ORzlrKvCISJmKqRnAl7d3oWmdIG777xK+XJzqdCQREZEqKyTAmzE3dSTQ14sbP1pI2n7NhSenZq3l0W9WMmdjJs8NbEPXRhGlOu7W7g24tlMc787cxGcLtXy6VA3WWpJTs6pE7x1QgUdE3CA80JdxwzrTpWE4D321gndnbiInvxBr1aVXRKoOY0xfY8w6Y8xGY8zDJ9nnamNMijFmtTFmnGtbgjFmnmvbCmPMNeWbXKqauqH+fHxzR47mFzL0w4Xsz85zOpJUYG/P2MQXi9O45/zGXNk+ptTHGWP4d/+WdG8SwT8nrmKOhgVKGTlwJN+x+Z12Hshhz6HcKlPgMVXlF66kpCS7ePFip2OISAl5BUU88EUy363YCYAx4O/tSYCPJ36uP/19vPD39iDAxwt/b0/8fTz/tM//7nvh7+OBv7cX/q7tx44J8PHEz8vz95UfRKT6MsYssdYmufk1PIH1QG8gDVgEDLbWppTYpwnwBXC+tXa/MaaWtXaPMSYesNbaDcaYusASoLm1NutUr6m2jvyV+ZszGfLBQlrHhPDprZ3+NGGuyKTlO7hn/DIGJNTltWsSzmhJ6IM5+Vz5zlx2Hsjh6zu60qT2qYd3iZyKtZa+r/1Gdl4B44d1JjYsoFxff8rKndz56VIm3tWtUhV5TtbW8XIijIhUDz5eHrwxqB29mtdi54EcjuYVcjSvkCP5heTkFXKkxP09h/73/NH84udyC4pO+zUjg3yJqelPdKg/MTUDiu/X9Ce2pj/RoQH4+6ixKyJloiOw0Vq7GcAY8xkwAEgpsc8wYKS1dj+AtXaP68/1x3aw1u4wxuwBIoGs8okuVVXnhuG8ek0CI8Yv5Z7xy3jn+vZ46osPcVm8dR8PfrmcjvXDeOHKNmdU3AEI9vPmwxs7cNnIudw0ZhET7+p20gmaRf7KstQs1u0+hIeBq9+bx6e3dqJhZGC5vX5yahY+nh40j6oahUoVeETErTw8DJe3K33335KKiuzvxZ4c15/Fjwv+99i17WheIdm5Bew+mEta1hFWph/gx9W7yC/8Yy/F8Bo+xNQsLv5E1/R33S8u/kTX9CfQV/8sikipRAMlJxlLAzodt088gDFmDuAJPGGt/aHkDsaYjoAPsMl9UaU6ubhNFHsOteDfk1N4fNIq/jOg1Rn/Ii9Vx9a92Qwbu5joUH/eu6E9vl5n94VXTM0ARg9NYtCoeQwbu5jxwzqrx5icka+XpuHn7cEnt3Ti9k+WcM2o+Yy7tVO59QxLTs2iRd3gs/6ZqCj0m4yIVFgeHoYavl7UOMOiS1GRZc+hXNL2HyE96yhp+4+Stv8IafuPsmbnQX5es5u843oJ1QzwLi78hP6v90/JnkDBft5l8dZEpHrwApoA5wIxwCxjTOtjQ7GMMVHAJ8BQa+0JuywaY4YDwwHi4uLKIbJUBTd1a8CuAzm8N2szUSH+3HVeY6cjiYP2Z+dx05hFAHx0Ywdq1iiblYISYkN57ZoE7vh0KX/7cjlvDmqnofJyWnILCpm8fCd9WtahQ/0wPr+tM9e+v4BrRs3nk1s60rJuiFtfv6CwiJVpB7imQ6xbX6c8qcAjIlWWh4ehTogfdUL8ONFkHEVFlr3ZuaTtP0r6/v8VgNKzjrIx4zAz1u8hJ/+Pv3MF+3kRUzOAZnWCaBcXSkJsTZpFBeHtqTnrRaqZdKBkizDGta2kNGCBtTYf2GKMWU9xwWeRMSYY+B541Fo7/2QvYq0dBYyC4jl4yjC/VHF/79uM3QdzePHHddQK8uWqpKrzC4yUXm5BIbd9soT0/UcZN6wT9SNqlOn5+7aK4uG+zXh26lrqhwfwUJ9mZXp+qdp+XbOHA0fzuSKxuLd/41pBfH5bF657fz6DR83nk1s60daN8+Ks332Yo/mFtItz32uUNxV4RKTa8vAw1Aryo1aQH4lxNf/0vLWWfdl5rsLPUdKzinv/bN93hFkb9vL1suLf5Xy9PGgTE0JCbCjt4mrSLi6UqBD/8n47IlK+FgFNjDENKC7sDAKuPW6ficBg4CNjTATFQ7Y2G2N8gG+Asdbar8ovslQnHh6GF65sy97DeTz89Uoignw5r2ktp2NJObLW8n9frWDh1n28MbgdSfXD3PI6w3s0ZGtmNiOnb6JeeA2uVjFRSmnC0nRqBfnSrVH479saRNTg89u6cO3o+Vw3egFjburgts9ucmoWAG1jQt1yfieowCMichLGGMIDfQkP9P3TtwfWWtKzjpKcmsWy7Vks276fj+dt4/3ftgBQO9iXdrE1Xb18QmkdE0KAj/7JFakqrLUFxpgRwI8Uz6/zobV2tTHmSWCxtXaS67kLjTEpQCHwkLU20xhzPdADCDfG3Og65Y3W2uRyfyNSpfl4efDO9Ylc89587vp0KeOHdXbrt+FSsbz6ywa+Td7BQ32a0r9tXbe9jjGGJwe0Im3/Uf7x9UpiavrTtVGE215PqobMw7nMWLeHm89pgNdxPeFjwwL44rYuXPf+AoZ8uJDRQ5Pc8plKTt1PzQBv6oWX78pd7qRl0kVEykheQRFrdh5k2fb9LEvNIjk1i22ZRwDw9DA0qxP0h14+DcJraKy6iBuUxzLpTlBbR87UnoM5XPHOXI7mFfL1nV2pF162w3Sk4vlqSRoPfrmcq5NieH7gma+YdToO5uQz8O257D6Yw9d3dqNxrfJbCUkqn4/mbOHfk1P48b4eNK1z4gmV9xzK4frRC9iWeYRRQ5LoGR9ZphkufHUmdUP9GXNTxzI9b3k4WVtHBR4RETfKPJxLsqvYs2x7FstTsziUWwBAiL83bWNDaRcb+ntPn9CAspn4UKQ6U4FH5M82ZRxm4DtzCfH3ZsIdXbWsdRU2d9Nehn64kI4NwhhzU8dynScwdd8RLn97DgE+XnxzZ1fC9TmTk7j0zdkUWcv393Q/5X77svO4fvQCNu45zMjrEundonaZvP6hnHza/Psn7u3VhPsuiC+Tc5ank7V1NF5ARMSNwgN96dW8Nr2aF/9nVFRk2ZRxuHhYV+p+lm3P4s1fN1DkqrU3jKhBQtyxok9NmkcF46lePiIicpYaRQbywdAOXPv+fG4es4jxwzqf8SqVUnFt3HOI2z9ZQv3wGrx9XftyXwQiNiyA94ckMWjUfIZ/soRPb+2k5dPlT9bvPsTK9AP865IWf7lvWA0fxg/rzJAPF3DHf5fw+qB2XNwm6qwzrEw7gLXFq8FVJfpXXUSkHHl4GJrUDqJJ7SCudi3JeDi3gBVp/+vlM2v9Xr5eWjyBc+voEEZem0hcFRobLCIizmhfryZvXZvIbZ8s5q5xS3l/SJJWgaxC9h7O5aYxi/Dx8uDDGzsQ4u/tSI52cTV59ZoE7vx0KQ99tYLXr0nQkHT5g6+XpuPpYRiQULq5oUICvPnvrZ246aNF3D1+KXmFbbm8XcxZZUhOywKqXoFH/6KLiDgs0NeLro0iuPPcxrw/JIlFj/Zi9t/P47krWrMtM5uL3/yNH1fvcjqmiIhUAb1b1OY/l7VixroMHv1mJVVluobqLie/kFs/XkzGoVxGD+1AbJizXwz1ax3F//VtyuTlO3j1l/WOZpGKpbDI8s2yNM6NjzytoaJBft58fHNHOjUI54EvlvP5ou1nlSN5exYNImpUuekRVOAREalgjDHE1AxgUMc4vr+nOw0janDbJ0t46rsU8guLnI4nIiKV3HWd6nHP+Y35YnEar/6sX74ru6Iiy/2fJ7M8LYvXrmlXYXok3NGzEdckxfLmrxv5akma03Gkgpi7aS+7D+ZyReLp98Cp4evFRzd1oEeTSP4+YSVj5209owzWWpJTsyrMz0pZUoFHRKQCiw0L4Ivbu3Bj1/qMnr2Fq9+bR3rWUadjiYhIJXd/73iuTorhjV838umCbU7HKVPWWhZu2cc945dx16dL2XMox+lIbvX8j2uZumoX/7ioOX1b1XE6zu+MMTx1eSu6NQ7nka9XMG9TptORpAL4emk6wX5e9Gpe64yO9/P2ZNSQ9vRuUZvHvl3N+7M2n/Y5dh7IYc+hXBV4RESk/Pl6efJE/5aMvDaRDbsPc/EbvzF97R6nY4mISCVmjOHpy1tzXtNI/jVxFT9VgaHA2bkFfLpgGxe9/htXvzeP6ev2MG3tbvq/OYflqVlOx3OLcQu2897MzVzfOY5buzdwOs6feHt68PZ17akXXoPb/7uETRmHnY4kDjqcW8APq3ZxSdu6ZzX5tq+XJ29fl8jFbaJ4esoa3py24bSOT3b9e9BWBR4REXHKxW2imHz3OUSF+HPTmEW88MNaCjRkS0REzpC3pwcjr0ukdXQId49fxpJt+5yOdEY2ZRzmiUmr6fzMNB79ZhXGGJ69ojUL/tGLCXd0xdPDcNV786rcMKGZ6zP417erOLdpJE9c2hJjKuZExiH+3nx0Ywe8PAw3j1nEvuw8pyOJQ6au3MnR/EIGJkaf9bm8PT14/ZoErmgXzcs/r+elH9eVek6x5NQsfDw9aB4VdNY5KhoVeEREKpEGETX45s6uDO4Yy9szNnHd6AXsOVi1u56LiIj7BPh48cGNHYgK8eOWjxezKv2A05FKpaCwiB9X7+L60Qvo9fJMPl2wjfOa1eKr27sw5Z5zGNwxjgAfL1rWDWHy3efQPq4mD365nH9PXl0lvhxZs/Mgd326lCa1Annr2kS8KvhqaLFhAYwaksTOAznc9slicgsKnY4kDvh6aTr1wgNIjKtZJufz8vTgpavaMqhDLG9N38gzU9aUqsiTvD2LFnWD8fU6815EFVXF/pdARET+xM/bk2evaMMrV7dlRdoB+r3xG3M37nU6loiIVFIRgb58fHNHvDwMl7w5m/NfmsETk1Yzfe0ejuZVrF/E9x7OZeT0jfR8cQa3fVI85OdvveOZ+3Av3hjcjqT6YX/qyRJWw4ext3Tkpm71+WjOVoZ8uLBS9yLZfTCHm8csooavJx/d1IFAXy+nI5VK+3o1eeXqtizaup//+2qFVnCrZtL2H2He5kyuaBdTpr3NPDwMz1zemqFd6vH+b1t4fNJqiopO/tkqKCxiZfqBKjn/DoBb/zUwxvQFXgc8gdHW2ueOe/5G4EUg3bXpLWvtaNdzQ4F/urY/Za392J1ZRUQqmysSY2gdHcIdny7lug8WcF+veEac3xhPj4rZRVtERCqueuE1+O7u7kxZuZOZ6zMYv3A7Y+ZuxcfLg471w+gZH0mP+EjiaweW+1Agay1Lt2fxybytTFm5i7zCIro2CudflzTngua1S9V7xdvTg8cvbUmLqGAenbiK/m/NZtQNSbSoG1wO76DsZOcWcMvHizhwNJ8vbutCVIi/05FOyyVt6rIt8wgv/riO+uE1uL93vNORpJxMXFb8K/8VZTA863geHoYn+rfEz9uT92ZtJje/iGeuaH3CNvG63Yc4ml9Iu7jQMs9REbitwGOM8QRGAr2BNGCRMWaStTbluF0/t9aOOO7YMOBxIAmwwBLXsfvdlVdEpDJqUjuISSO68c9vVvHqL+tZvG0fr16TQESgr9PRRESkkqkT4sfN5zTg5nMakJNfyMIt+5i1PoOZ6zN4esoanp6yhjrBfvSIj6BnfC3OaRxBSIC32/IczStk8vIdfDxvK6t3HCTQ14vBHWO5oUs9Gtc6s7kzrkqKpUntIG77ZDED35nLi1e14ZI2dcs4uXsUFlnu/WwZKTsOMnpoEq2iQ5yOdEbuPLcRW/Zm8/q0DdSPCODydqe/XLZULtZavl6aTscGYcSGBbjlNYwxPHxRM3y9PXlj2gZyCwp56aq2fyoAL08tHoaqHjynryOw0Vq7GcAY8xkwADi+wHMifYCfrbX7XMf+DPQFxrspq4hIpRXg48XLV7elY4MwHp+0movf+I03ByfSsUGY09FERKSS8vP2pIer184/gR1ZR5m1PoNZGzKYumoXXyxOw8MU/5LUM74WPeIjaBMTWia9SLfuzebTBdv4YnEaB47mE187kP9c1orL20WXyXCkhNhQJo84h9v/u4QR44oLJn+7sGmF7gG7cMs+np6yhuWpWTw5oCXnN6vtdKQzZkzxkJr0/Uf5+1craRARWGV/2ZZiyalZbN6bzW09G7r1dYwxPNA7Hl8vD178cR15hUW8dk07fLz+V+RJTt1PWA0f4txUaHKaOws80UBqicdpQKcT7DfQGNMDWA/cb61NPcmxf+rLZYwZDgwHiIuLK6PYIiKVjzGGQR3jaBMTyl3jljL4/fk81Kcpw7s3xKMCN1hFRKRyqBvqz6COcQzqGEdBYRHL07KYuS6DmRv28tq09bz6y3pCA7w5p3EEPeIj6RkfSe1gv1Kfv7DIMnP9HsbO28aMdRl4eRj6tKzDDV3q0anBn+fVOVu1gv0YP7wzT0xazdszNrFm50FeG9SOEH/39Ug6Exv3HOa5qWv5Zc1uagf78vJVbRnYvvL3ePHx8uDd69vT743fuGf8Mr6/5xyC/CrWtZey8/XSdHy9POjXOqpcXu+u8xrj5+3Jf75LIa9gCW9dm/j7suzJqVm0jQmpsKvOnS2nZ+SaDIy31uYaY24DPgbOL+3B1tpRwCiApKQkzdIlItVei7rBTBrRjYcnrOS5qWtZtGUfL1/dltAAH6ejiYhIFeHl6UH7emG0rxfGAxc2ZV92HrM37mXmuuIePt+t2AlAszpB9HQVe9rXr3nCFWv2Z+fxxeJU/rtgG6n7jlIryJd7ezXh2k5xp1UgOhO+XsWLFrSsG8ITk1Zz+cg5jBrS/oyHf5WljEO5vPbLej5blIq/tycP9WnKzd0a4O9TdVb9CQnw5o3BCVz93nz+NXEVrw1q53QkcYPcgkImLd9Bn5Z1yrWId8s5DfDx8uBfE1cxbOxiRt2QREFRERv2HObi1pVjWOaZcGeBJx2ILfE4hv9NpgyAtTazxMPRwAsljj33uGNnlHlCEZEqKMjPm7eubUfHeWE89X0KF78xm7eubUe7MlqSUkREpKSwGj70b1uX/m3rYq1lzc5DzFyfwaz1GXw4ZwvvzdqMv7cnXRuF/96752BOPmPnbWPS8h3kFRTRqUEYf+/bjD4t6+Bdzkt+X9+5HvG1g7jz0yVcNnIur12TwAUtnBkCdSSvgPdnbWHUrE3kFhRxXac47unVpMrOrde+Xhj39mrCKz+vp0d8JFckVv7eSeUhv7AILw9TKXqhTF+7hwNH890yufJfuaFzPXw9Pfj71yu4acxCbjmnIdZC29jKOX9VaRh3LU9njPGieNhVL4oLNouAa621q0vsE2Wt3em6fznwd2ttZ9cky0uARNeuS4H2x+bkOZGkpCS7ePFit7wXEZHKanlqFnd+upQ9h3L4R7/m3Ni1fqVoDIicDWPMEmttktM5ypraOlIZZecWMG9TJjNdkzVv33fk9+cCfDy5vF00Q7rUp2kd53vN7Mg6ym2fLGFl+gEe6B3PiPMal9sw54LCIr5aksYrP69nz6Fc+rasw//1bUrDyMByeX0nFRZZBr8/n9XpB/j+nu7Uj6jhdKQK7VBOPr1ensmgjnE8UAlWIRs2djHJqVnMe/j8Uq145w7fJqfzwBfL8fH04Gh+IcmP9a70vdtP1tZxWw8ea22BMWYE8CPFy6R/aK1dbYx5ElhsrZ0E3GOM6Q8UAPuAG13H7jPG/IfiohDAk6cq7oiIyIm1jQ1lyj3d+duXyfx7cgoLt+zj+SvbEKxx7iJnzRjTF3id4nbOaGvtcyfY52rgCYpXBV1urb3Wtf0HoDMw21p7SbmFFilnNXy9uKBF7d97xGzdm82sDRl4GEP/hLoV6v+juqH+fHl7Fx75eiWv/LyelB0HeenqtmUysfPJWGuZvm4Pz01dy/rdh0mMC+Xt6xJJql99Fkrw9DC8dk0CF73+G/d8toyvbu/6h0lx5Y/GL9zOnkO5vDtzE1e1j3HbqlRlYV92HtPX7uHmcxo4VtwBGJAQjY+nB/d8toyGETUqfXHnVNzWg6e86VstEZGTs9Yy+rctPPfDWmJq+jPy2sRKu7yqyF8pjx48xhhPinsq96Z4MYhFwGBrbUqJfZoAXwDnW2v3G2NqWWv3uJ7rBQQAt5W2wKO2jkj5sNbywewtPDNlDY1rBfL+kCTqhZd9r5KVaQd4Zsoa5m3OpH54AH/v24y+repU2562P6zaxe3/XcJtPRvyyEXNnY5TIeUWFNLjhelEBvmycc9hLmxRhzcGV9y5i8bM2cITk1P44b7uNKsT7HQcklOzMBR/AVrZnayto9KoiEg1YIxhWI+GfD68M7n5RVzxzlzGLdhOVSnyizigI7DRWrvZWpsHfAYMOG6fYcBIa+1+gGPFHdf9acCh8gorIqVnjOHW7g0Ze3Mn9hzKpf9bc/htQ0aZnT913xHu/WwZl741m3W7D/Hv/i35+YGeXNQ6qtoWdwD6tqrDtZ3ieG/m5jK93lXJxGXp7D6Yy//1acaw7g2ZtHwHyalZTsc6qa+XpdMiKrhCFHcAEmJDq0Rx51RU4BERqUaS6ofx/T3n0LlhOP/4ZiX3f55Mdm6B07FEKqNoILXE4zTXtpLigXhjzBxjzHzXkK7TYowZboxZbIxZnJGhX3hEytM5TSKYdNc5RIX4MfTDhYyatemsvhg5cCSfp79PodfLM/lh1S7uOq8RMx46l6Fd65f7xNIV1b8ubkGTWoE88MVyMg/nOh2nQikssrw3azMt6wbTvUkEt/VsRESgL099l1Ihv7DbsPsQK9IOODK5cnWmf0lERKqZ8EBfxtzYgb/1jmfS8h30f2s2S7btq5CNA5FKzgtoQvHKoIOB940xoadzAmvtKGttkrU2KTIysuwTisgpxYUHMOGOrvRtVYdnpqzl/s+TyckvPK1z5BYU8v6szfR4cTqjZ2+hf0JdZjx0Lg/1aVah5iCqCPx9PHnz2nYcOJrPQ1+tUNukhJ9TdrE5I5vbezbCGEOgrxd/uzCexdv288OqXU7H+5MJS9Px9DAMSFCBpzypwCMiUg15eBju7tWE/97SiQNHCxj4zjy6PPsr/5y4klnrM8grKHI6okhFlw7Elngc49pWUhowyVqbb63dQvGcPU3KKZ+IlJEavl6MvDaRBy+M59vlO7jy3bmkZx39y+OKiizfJqfT6+WZPD1lze8LH7x0VVuiQvzLIXnl1KxOMI/2a86va/cwZu5Wp+NUCNZa3pm5mbiwAC5qVef37VcnxdK0dhDP/bC2QrXdCossE5el0zM+ksggX6fjVCsq8IiIVGNdG0cw7W89eemqtrSNDWHCknSGfLiQ9v/5mRHjlvJtcjoHjuY7HVOkIloENDHGNDDG+ACDgEnH7TOR4t47GGMiKB6ytbkcM4pIGTHGMOL8JoweksS2vUfo/+ZsFmzOPOn+8zdnctnbc7j3s2SC/bz55JaOjL25I82jKsZcJBXdkC71uKB5LZ6dspbVOw44Hcdx8zZnsjw1i+E9Gv5hNSpPD8M/Lm7OtswjjJ231bmAx5m3KZNdB3M0PMsB7lvzT0REKoUQf2+ubB/Dle1jyMkvZM7Gvfycsptf1uzmuxU78fIwdG4YTu8WtendojZ1Q/Wto4i1tsAYMwL4keJl0j+01q42xjwJLLbWTnI9d6ExJgUoBB6y1mYCGGN+A5oBgcaYNOAWa+2PjrwZESm1Xs1r881d3Rj+yWKuG72Axy9twfWd6/0+OfKG3Yd4bupapq3dQ90QP16+qi2Xt4vGw6P6Tp58JowxvHBlWy56fRb3jF/G5LvPIcCn+v7q+u7MzUQE+nBl+5g/PdczPpIe8ZG8+etGrmwfUyGWAP96aRpBfl5c0Ly201GqHS2TLiIiJ1RYZElO3c9PKbv5OWU3mzOyAWgVHUzv5nXo3aI2zaOCqvWKH1Ixlccy6U5QW0ek4jiYk899nyXz69o9DOoQy4jzGzNy+kY+X5RKDR8v7jyvMTd1q4+ft6fTUSu1uRv3ct0HCxjUIZZnr2jjdBxHrEo/wCVvzuahPk2567zGJ9xn3a5DXPT6LG7s2oDHLm1Rzgn/6HBuAR2e+oXL2kXz7BWtHc1SlZ2srVN9y6AiInJKnh6G9vXCaF8vjEcuas6mjMP87Cr2vDZtPa/+sp6Ymv5c0Lw2F7aoTYcGYVoFREREqoVgP29GD0nilZ/X89b0jXy2KBVvT8PQrvW5+/wmhNVwvhdFVdC1cQR39GzE2zM20b1JJP1aRzkdqdy9N2szgb5eXN+53kn3aVoniGs6xPLJ/K0M6VKP+hE1yjHhH/2wahdH8wsZqOFZjlCBR0RESqVRZCCNegZye89GZBzKZdqa4mLPuIXbGTN3KyH+3pzfrBa9W9SmR3wkgb76L0ZERKouDw/Dg32a0io6hFkbMritR0PqhTv3i3VVdX/veOZuyuThCStoExNCTM0ApyOVm+2ZR/h+xQ6GdW9IiP+pV1y7v3c8k5J38NzUtbx7Q/tySvhnXy9No154AO3r1XQsQ3Wm1reIiJy2yCBfBnWMY1DHOI7kFTBrffG8PdPW7uabZen4eHrQtbFr3p7mtakV7Od0ZBEREbfo26oOfUusbCRly9vTgzcGtaPfG79x32fJfDa88x8mGq7KRv22CS8PD24+p8Ff7lsryI/bezbi5Z/Xs3DLPjo2CCuHhH+UnnWUeZszua9XvIbwO6R6/GSIiIjbBPh40bdVHV6+ui2LH72Az4Z35oYu9dickc2j36yi4zPTuGzkHEZO38iWvdlOxxUREZFKJi48gKcvb8Xibft589eNTscpFxmHcvlycRpXJEZTu5RflN3avSF1gv14+vsUiorKf67dicvSsRYub6fhWU5RgUdERMqMl6cHnRuG869LWjDzoXP58b4ePHhhPNZaXvxxHee9NIMbPljAzym7KXSg4SEiIiKV04CEaK5IjObNXzewcMs+p+O43Zi5W8grLGJ4j4alPsbfx5OH+jRledoBJq/Y4cZ0f2atZcLSNDrWDyMuvPoMo6toVOARERG3MMbQtE4QI85vwrcjzmHeI+fzt97xbNh9mGFjF9Pjhem8O3MT+7PznI4qIiIilcCTA1oRFxbAfZ8t48CRfKfjuM2hnHw+mbeNPi3q0DAy8LSOvbxdNK2ig3nhh3Xk5Be6KeGfLU87wOaMbK7Q5MqOUoFHRETKRVSIP3f3asJvfz+Pd65LJDbMn+emrqXTs9N48MvlrEw74HREERERqcACfb14Y3A7Mg7n8vDXK7C2avYGHr9wOwdzCrj93EanfayHh+HRfi1IzzrKh3O2uCHdiX29NA1fLw/6tal+K51VJCrwiIhIufL29OCi1lF8NrwLP97Xg6vaxzBl5U4ufWs2l789h2+WpZFbUH7fOImIiEjl0SYmlAcvbMrUVbv4bFGq03HKXG5BIR/M3kKXhuEkxIae0Tm6NArngua1eXv6JvYezi3bgCeQW1DIpOU7uLBlHYL9Tr3al7iXCjwiIuKYpnWCePry1sz/Ry8ev7QFB47kc//ny+n67K+89OM6dmQddTqiiIiIVDDDujeke5MI/j15NRt2H3I6Tpn6dtkOdh/M5Y4z6L1T0iP9mpGTX8hrv6wvo2QnN31tBllH8jU8qwJQgUdERBwX7OfNTd0a8MsDPRl7c0faxdVk5IyNdH9hOrd/soS5m/ZW2W7YIiIicno8PAwvX92WGj5e3D1+WbnONeNORUWWd2dtomXdYLo3iTirczWKDOS6TnGMX5jKxj3uLYJ9vTSNyCBfujc+u8xy9lTgERGRCsPDw9AjPpLRQ5OY9dB53Nq9AfO3ZHLt+wu48NVZfDJ/G4dzC5yOKSIiIg6rFeTHS1e1Ze2uQzw3da3TccrETym72ZyRze09G2GMOevz3XtBPAE+njwzxX3XZ192HtPX7eGyhLp4eaq84DT9DYiISIUUGxbAIxc1Z/4jvXjxyjb4envwr4mr6PzMNJ6YtJpNGYedjigiIiIOOq9ZLW7u1oAxc7cybc1up+OcFWst78zcRFxYABe1qlMm5wyr4cOI8xrz69o9zNm4t0zOebzJy3eQX2i5IjHGLeeX06MCj4iIVGh+3p5clRTL5BHn8PWdXendojafLthGr5dncsMHC/g5ZTeFRRq+JSIiUh39/aKmtIgK5qGvVrDnYI7Tcc7Y/M37WJ6axfAeDcu0J8zQrvWJqenPU9+vcUt76eulaTSPCqZ5VHCZn1tOnwo8IiJSKRhjSIyryavXJDD34V48eGE8G3YfZtjYxfR4YTrvzNjEvuw8p2OKiIhIOfL18uSNwe04mlfI/V8kU1RJv/R5Z+YmIgJ9uLJ92faE8fP25O99m7Fm50EmLE0r03Nv3HOI5WkHGKjJlSsMFXhERKTSiQzyZcT5TZj99/N457pE4sICeP6HtXR+dhp/+2I5q9IPOB1RREREyknjWoE80b8FczZm8t6szU7HOW2rdxxg1voMburWAD9vzzI//yVtomgXF8pLP67jSF7ZzWU4YWk6nh6G/gl1y+yccnZU4BERkUrLy9ODi1pHMX54Z366vwdXJ8UwddVOLnlzNiPGLWVbZrbTEUVERKQcXJ0Uy8Wto3j5p3Ukp2Y5Hee0vDtzM4G+XlzfuZ5bzm+M4Z8XN2fPoVxGlVEBrLDIMnFZOj2aRFAryK9MzilnTwUeERGpEuJrB/HUZa2Z/49e3HN+Y6at2cMFr8zkiUmrNXRLRESkijPG8MwVrakd7Me9ny3jUE6+05FKZXvmEb5fsYNrO8UR4u/tttdpXy+Mi1tH8d7Mzewug7mK5m/OZOeBHE2uXMGowCMiIlVKsJ83D1zYlBkPncuV7WMYO28rPV+YzsjpGzmaV+h0PKlijDF9jTHrjDEbjTEPn2Sfq40xKcaY1caYcSW2DzXGbHDdhpZfahGRqinE35vXByWQuu8Ij3272uk4pfL+b5vx8vDglnMauP21/t63GYVFlpd/WnfW55qwNI0gPy96t6hdBsmkrKjAIyIiVVLtYD+evaINP97Xg04Nw3nxx3Wc99IMvliUqlW3pEwYYzyBkcBFQAtgsDGmxXH7NAEeAbpZa1sC97m2hwGPA52AjsDjxpia5ZdeRKRqSqofxr294vlmWTpfl/GkwmVt7+FcvlicyuXtoqkd7P5hTnHhAQztWo8vl6SRsuPgGZ8nO7eAH1bt4pI2UW6ZM0jOnAo8IiJSpTWpHcTooUl8PrwztUP8+L8JK+j3+m9MX7sHa1XokbPSEdhord1src0DPgMGHLfPMGCktXY/gLV2j2t7H+Bna+0+13M/A33LKbeISJU24vzGdKwfxr8mrmLr3oo7H9+YOVvJKyxieM+G5faaI85rQoi/N09PSTnjdtAPq3ZxJK9Qw7MqIBV4RESkWujUMJyJd3Zl5LWJ5BQUctOYRVz7/gJWpGU5HU0qr2ggtcTjNNe2kuKBeGPMHGPMfGNM39M4FmPMcGPMYmPM4oyMjDKMLiJSdXl6GF4dlICXpwf3fraMvIIipyP9yeHcAsbO20qfFnVoFBlYbq8bEuDNvb2aMGdjJtPX7fnrA07g62VpxIUFkFRPHU8rGhV4RESk2jDGcHGbKH6+vyf/7t+SdbsP0f+tOdw9fhnbM484HU+qJi+gCXAuMBh43xgTWtqDrbWjrLVJ1tqkyMhI9yQUEamCokP9eX5ga5anHeCVn9c7HedPxi/YzsGcAm4/t1G5v/Z1nerRIKIGz0xZS0Hh6RW/dmQdZe6mTK5IjMYY46aEcqZU4BERkWrHx8uDoV3rM/OhcxlxXmN+TtlFr1dm8OTkFPZrxS0pvXQgtsTjGNe2ktKASdbafGvtFmA9xQWf0hwrIiJnoW+rKAZ3jOPdmZuYtHyH03F+l1tQyOjZm+nSMJyE2NByf30fLw8evqgZG/ccZvyi1L8+oIRvlqVjLVzRTsOzKiIVeEREpNoK8vPmwT5NmfHgeQxMjGHM3C30eHE6b8/YSE6+VtySv7QIaGKMaWCM8QEGAZOO22cixb13MMZEUDxkazPwI3ChMaama3LlC13bRESkDD12SQs61g/j3s+W8fHcrU7HAeDbZTvYfTDXkd47x1zYojYdG4Tx2s/rS72kvLWWr5em0aF+TeLCA9ycUM6EWws8pVk61LXfQGOMNcYkuR7XN8YcNcYku27vujOniIhUb3VC/HhuYBt+uK8HHeuH8cIPxStufblYK27JyVlrC4ARFBdm1gBfWGtXG2OeNMb0d+32I5BpjEkBpgMPWWszrbX7gP9QXCRaBDzp2iYiImXI38eTsbd0pFez2jw+aTUv/7TO0UUWioos787aRIuoYHo0iXAshzGGf17cnMzsPN6esalUx6xIO8CmjGwGanLlCsttBZ7SLB3q2i8IuBdYcNxTm6y1Ca7b7e7KKSIickx87SA+uLED44d1plaQLw99tYKL3/iNGeu04pacmLV2irU23lrbyFr7tGvbY9baSa771lr7gLW2hbW2tbX2sxLHfmitbey6feTUexARqer8vD159/pErkmK5c1fN/KPb1ae9twzZeWnlN1szsjm9nMbOT6HTZuYUC5vF80Hs7eQtv+v5yKcsDQNHy8P+rWJKod0cibc2YOnNEuHQvG3V88DOW7MIiIiUmpdGoUz8a5uvHVtO47kFXLjR4u4/oMFrEo/4HQ0EREROQNenh48N7A1d53XiPELU7nz06XlPhzbWss7MzcRFxZAv1Z1yvW1T+ahPk0xwIs/rjvlfnkFRUxavoMLW9Qm2M+7fMLJaXNngecvl/80xiQCsdba709wfANjzDJjzExjTHc35hQREfkTYwyXtKnLLw/05PFLW5Cy4yCXvDmbez9bRuo+rbglIiJS2RhjeKhPMx6/tAU/pexmyIcLOXC0dPPPlIX5m/exPDWLYT0a4uVZMabDrRvqz63dG/Bt8g6SU7NOut/0dXvIOpKv4VkVnGOfKmOMB/AK8LcTPL0TiLPWtgMeAMYZY4JPcI7hxpjFxpjFGRkZ7g0sIiLVko+XBzd1a8DM/zuPu85rxA+rdtHr5Zk89Z1W3BIREamMburWgNcHJbBs+36ueW8eew6Wz2CSd2duIiLQh6vaV6wiyR3nNiYi0Ienv0856ZD0r5emERHoS3cH5w2Sv+bOAs9fLf8ZBLQCZhhjtgKdgUnGmCRrba61NhPAWrsE2ETxqhN/YK0dZa1NstYmRUZGuultiIiIQLCfNw/1acaMh87lsnZ1+XDOFro+9yv3jF/Gzym7yStwZiy/iIiInL4BCdF8MLQD2/cdYeC7c9myN9utr7d6xwFmrs/gpm4N8PP2dOtrna5AXy/u7x3Poq37+XH1rj89vz87j1/X7uGyhLoVpueRnJg7/3ZOuXSotfaAtTbCWlvfWlsfmA/0t9YuNsZEuiZpxhjTEGhC8ZKiIiIijooK8eeFK9sy9d4eXNYumlkbMhg2djFJT/3M379awZyNe7XyloiISCXQIz6S8cM6k51byJXvzGVlmvvm2ntv5mYCfb24vnM9t73G2bgmKZYmtQJ5buraP31pNXnFDvILLVdoeFaF57YCTymXDj2ZHsAKY0wy8BVwu5YOFRGRiqRpnSCevaI1ix69gI9u7ECv5rX5bsUOrhu9gE7PTOOJSatZsm2/Vt8SERGpwNrGhvLl7V3w8/Zk0Kh5zN6wt8xfY3vmEb5bsYNrO8UR4l8xJyj28vTgHxc3Z2vmET6Zv+0Pz01Ymk7zqGBa1P3TrClSwXi58+TW2inAlOO2PXaSfc8tcX8CMMGd2URERMqCt6cH5zWrxXnNapGTX8iva/cwKXkH4xZuZ8zcrUSH+nNp27r0b1uX5lFBji+JKiIiIn/UKDKQr+/sypAPFnLTmIW8ek0Cl7SpW2bnf/+3zXh6GG7u1qDMzukO58ZH0r1JBG9M28DAxGhCA3zYuOcwy1Oz+OfFzZ2OJ6WgAXQiIiJlxM/bk36to3j3hvYs+ecFvHxVWxrXCuT93zbT743f6P3qLF7/ZYPbx/mLiIjI6akd7McXt3UhITaUu8cvY+y8rWVy3r2Hc/licSpXtIuhTohfmZzTXYwx/KNfcw7m5PPmrxuB4smVPQz0Tyi7gpe4j1t78DgtPz+ftLQ0cnLKZ1b0qsrPz4+YmBi8vStmd0IRkYooyM+bge1jGNg+hszDuUxdtYtJy3fw6i/refWX9bSODuHStlFc0qYudUP9nY4rIiJS7YUEePPJLZ0YMW4Zj327mr2Hcrm/d/xZ9b4dM2creYVFDO/ZsAyTuk/zqGCubh/L2Hlbua5THN8sS6dHfCS1gip2cUqKmaoyN0BSUpJdvHjxH7Zt2bKFoKAgwsPD1SX+DFlryczM5NChQzRoULG7FIqIVAY7Dxzlu+U7mbxiBytckzl2rB/GpQl16deqDuGBvg4nrPyMMUustUlO5yhrJ2rriIhI2SsoLOIf36zki8VpDO4Yx1OXtcLT4/R/nzycW0DXZ6fRtVEE797Q3g1J3WPPwRx6vjiDqBA/Nu/N5s3B7bi0rXrwVCQna+tU6R48OTk51K9fX8Wds2CMITw8nIyMDKejiIhUCVEh/gzr0ZBhPRqyZW82k5fvYNLyHfxr4iqemLSabo0j6N+2Lhe2rE2wn3pOioiIlDcvTw+eH9iGyCBfRk7fxL7sXF4f1O60lzcfv2A7B3MKuP3cRm5K6h61gv24rWdDXvtlA0G+XvRuUdvpSFJKVbrAA6i4UwZ0DUVE3KNBRA3u6dWEu89vzNpdh5i0fAeTl+/gwS+X4/ONB+c1jaR/22h6Na912o1KEREROXPGGB7q04yIQF/+PTmFoR8u5P2hSaX+8iW3oJDRszfTpWE4CbGh7g3rBsN7NOSrJWlc0Ly22iCVSJUv8IiIiFR0xhiaRwXTPCqY/+vTlGWpWUxK3sH3K3fy4+rdGAMB3p74+3gR4ONJgI8n/j6e+Hsfu+/let71+Pf7Xr/ve+w4P+/jtnt74uWpNRdERERO5KZuDQir4cODXy7nmvfm8/FNHagV/Nfz0Xy7bAe7D+bywpVtyyFl2Qvw8eKXB3rirTZCpaICj0NuvPFGLrnkEq688soyPefMmTMJCQkBYMyYMSQkJGCt5d5772XKlCkEBAQwZswYEhMTyyWTiIicHmMMiXE1SYyryb8uacGCzZnM37KP7NwCjuQVkpNfyJG84vtH8wrJOpLPUde2o3mFHM0vJL/w9ObX8/H0wN/Hk2B/L9rH1aRrowi6NAonNizATe9SRESk8hiQEE3NAB9u/+8SBr47l7E3d6JBRI2T7l9UZHl31iZaRAXTo0lEOSYtW+q5U/mowFPFvPjii38q0EydOpUNGzawYcMGFixYwB133MGCBQscSigiIqXl6WHo2jiCro1Pr3GYX1j0ewHoSF4BR/OP3S++Hc3/X4HoaF4hR1zPZxzKZfbGTCYm7wAgNsyfrg0j6No4nC4Nw0v1jaWIiEhV1CM+knHDOnPzmEVc+c5cxtzUkdYxISfc96eU3WzOyOaNwe003YWUq2pT4Pn35NWk7DhYpudsUTeYxy9t+Zf7/ec//+G///0vkZGRxMbG0r79H2dQnzZtGg8++CAFBQV06NCBd955h+nTp/PBBx/w5ZdfAjBjxgxeeuklvvvuO3766Scef/xxcnNzadSoER999BGBgYEnff1vv/2WIUOGYIyhc+fOZGVlsXPnTurUqcPdd9/Nzz//TGxsLD4+Pmd3QUREpELw9vQgxN+DEP/Tn6TZWsuGPYeZtymTuZv2MnXVTj5fnApAo8gadG0UQddG4XRuGE7NGvp/Q0REqo+E2FC+vL0LQz5YyKBR8xg1JIlux30JY63lnZmbiA3zp1+rOg4llepKA+rcbNGiRUyYMIHly5czdepUjl/eNCcnhxtvvJHPP/+clStXUlBQwDvvvMMFF1zAggULyM7OBuDzzz9n0KBB7N27l6eeeopffvmFpUuXkpSUxCuvvPL7+R599FHatGnD/fffT25uLgDp6enExsb+vk9MTAzp6el88803rFu3jpSUFMaOHcvcuXPL4YqIiEhFZowhvnYQQ7vW570bklj22IV8d/c5/KNfM2LDApiwNI07Pl1K4lM/0+/133jquxR+XbubQzn5Tkcvd8aYvsaYdcaYjcaYh0/w/I3GmAxjTLLrdmuJ5543xqxy3a4p3+QiInKmGkUGMuGOrsTUDOCmjxbx3Yodf3h+/uZ9LE/NYniPRprjTspdtenBU5qeNu4wZ84cBgwYgJ+fH35+flx66aV/eH7dunU0aNCA+Ph4AIYOHcrIkSO577776Nu3L5MnT+bKK6/k+++/54UXXmDmzJmkpKTQrVs3APLy8ujSpQsAzz77LHXq1CEvL4/hw4fz/PPP89hjj50026xZsxg8eDCenp7UrVuX888/301XQUREKitPD0Or6BBaRYcwvEcj8guLWJGWxdyNmczdlMnY+dsYPXsLnh6GNjEhdG0UTtdGEbSvV7NKj903xngCI4HeQBqwyBgzyVqbctyun1trRxx37MVAIpAA+AIzjDFTrbVl29VYRETcok6IH1/c1oVbxy7i7vHL2Jedx5Au9QF4d+YmIgJ9uKp9jLMhpVqqNgWeymjQoEG89dZbhIWFkZSURFBQENZaevfuzfjx4/+0f1RUFAC+vr7cdNNNvPTSSwBER0eTmpr6+35paWlER0eXz5sQEZEqxdvTg/b1wmhfL4y7ezUhJ7+Qpdv2M9c1pOvdmZsZOX0TPp4eJNYL/X1IV9vY0Kq2EkdHYKO1djOAMeYzYABwfIHnRFoAs6y1BUCBMWYF0Bf4wl1hRUSkbIUEePPJLZ0YMW4Zj327mr2HcunTqg4z12fwUJ+mVfpLDqm4qlRLqyLq1q0bkydPJicnh8OHD/Pdd9/94fmmTZuydetWNm7cCMAnn3xCz549AejZsydLly7l/fffZ9CgQQB07tyZOXPm/L5/dnY269evB2Dnzp1A8bjPiRMn0qpVKwD69+/P2LFjsdYyf/58QkJCiIqKokePHnz++ecUFhayc+dOpk+f7v4LIiIiVYqftyddG0fwYJ+mfH1nN5Y/fiEf3diBoV3rcSingFd/Wc+V786j7b9/YuiHC3lv5iZWph2gsOj0VvqqgKKB1BKP01zbjjfQGLPCGPOVMebYeOnlQF9jTIAxJgI4D4g9wbEiIlKB+Xl78u71iVydFMMbv25k6IcLqeHjyfWd6jkdTaop9eBxsw4dOtC/f3/atGlD7dq1ad269e/LmAP4+fnx0UcfcdVVV/0+yfLtt98OgKenJ5dccgljxozh448/BiAyMpIxY8YwePDg3+fYeeqpp4iPj+e6664jIyMDay0JCQm8++67APTr148pU6bQuHFjAgIC+OijjwC4/PLL+fXXX2nRogVxcXG/D/USERE5U4G+XpzXrBbnNasFwP7sPBZsyXT18Mnk2alrAXjvhvb0aVnlJ5+cDIy31uYaY24DPgbOt9b+ZIzpAMwFMoB5QOGJTmCMGQ4MB4iLiyuf1CIiUmpenh48P7ANEYG+vD1jE8N7NCQk4PQXORApC8baSv8NGgBJSUn2+AmM16xZQ/PmzR1K9D+HDx8mMDCQI0eO0KNHD0aNGkViYqLTsU5LRbmWIiJSue05mMO8zZmc27TWGa3yVRrGmCXW2iS3nPx/r9EFeMJa28f1+BEAa+2zJ9nfE9hnrf3TmrrGmHHAf621U071midq64iISMWxIi2LZnWC8fHSQBlxr5O1ddSDpxwMHz6clJQUcnJyGDp0aKUr7oiIiJSVWsF+DEioEvPALQKaGGMaAOnAIODakjsYY6KstTtdD/sDa1zbPYFQa22mMaYN0Ab4qdySi4iIW7SJCXU6glRzKvCUg3HjxjkdQURERMqQtbbAGDMC+BHwBD601q42xjwJLLbWTgLuMcb0BwqAfcCNrsO9gd+MMQAHgetdEy6LiIiInLEqX+Cx1uJqQMkZqirD+ERERMqSa0jVlOO2PVbi/iPAIyc4LofilbREREREykyVHhzo5+dHZmamChRnwVpLZmYmfn5+TkcRERERERERkZOo0j14YmJiSEtLIyMjw+kolZqfnx8xMTFOxxARERERERGRk6jSBR5vb28aNGjgdAwREREREREREbeq0kO0RERERERERESqAxV4REREREREREQqORV4REREREREREQqOVNVVpgyxmQA29z4EhHAXjeevzLQNdA1OEbXQdcAdA1A1wAq5jWoZ62NdDpEWVNbp1zoGugagK4B6Boco+ugawAV8xqcsK1TZQo87maMWWytTXI6h5N0DXQNjtF10DUAXQPQNQBdg6pEf5e6BqBrALoGoGtwjK6DrgFUrmugIVoiIiIiIiIiIpWcCjwiIiIiIiIiIpWcCjylN8rpABWAroGuwTG6DroGoGsAugaga1CV6O9S1wB0DUDXAHQNjtF10DWASnQNNAePiIiIiIiIiEglpx48IiIiIiIiIiKVnAo8IiIiIiIiIiKVnAo8xzHG9DXGrDPGbDTGPHyC532NMZ+7nl9gjKnvQEy3McbEGmOmG2NSjDGrjTH3nmCfc40xB4wxya7bY05kdSdjzFZjzErX+1t8gueNMeYN1+dghTEm0Ymc7mKMaVri7zfZGHPQGHPfcftUyc+BMeZDY8weY8yqEtvCjDE/G2M2uP6seZJjh7r22WCMGVp+qcvWSa7Bi8aYta7P+zfGmNCTHHvKn53K4iTX4AljTHqJz3y/kxx7yv9HKouTXIPPS7z/rcaY5JMcWyU+B1VRdW/ngNo6x6itUz3bOmrnFFNbR20dqKJtHWutbq4b4AlsAhoCPsByoMVx+9wJvOu6Pwj43OncZXwNooBE1/0gYP0JrsG5wHdOZ3XzddgKRJzi+X7AVMAAnYEFTmd247XwBHYB9arD5wDoASQCq0psewF42HX/YeD5ExwXBmx2/VnTdb+m0++nDK/BhYCX6/7zJ7oGrudO+bNTWW4nuQZPAA/+xXF/+f9IZbmd6Boc9/zLwGNV+XNQ1W5q5/z+HtXWsWrrHPdeq01bR+2cU14HtXXU1jn++UrX1lEPnj/qCGy01m621uYBnwEDjttnAPCx6/5XQC9jjCnHjG5lrd1prV3qun8IWANEO5uqQhoAjLXF5gOhxpgop0O5SS9gk7V2m9NByoO1dhaw77jNJX/uPwYuO8GhfYCfrbX7rLX7gZ+Bvu7K6U4nugbW2p+stQWuh/OBmHIPVo5O8jkojdL8P1IpnOoauP7fuxoYX66h5GxV+3YOqK1zGtTWqYLUzimmto7aOlA12zoq8PxRNJBa4nEaf/4P//d9XP8AHADCyyVdOXN1y24HLDjB012MMcuNMVONMS3LN1m5sMBPxpglxpjhJ3i+NJ+VqmIQJ/+Hrap/Do6pba3d6bq/C6h9gn2q02fiZoq/1T2Rv/rZqexGuLpuf3iSLuzV5XPQHdhtrd1wkuer+uegslI75zhq66it41Ld2zpq5/yZ2jpq61TKto4KPHJCxphAYAJwn7X24HFPL6W4C2tb4E1gYjnHKw/nWGsTgYuAu4wxPZwO5ARjjA/QH/jyBE9Xh8/Bn9jiPpnW6RxOMcY8ChQAn55kl6r8s/MO0AhIAHZS3G23uhrMqb/RqsqfA6ki1NbRzymorXO86t7OAbV1UFvnmErZ1lGB54/SgdgSj2Nc2064jzHGCwgBMsslXTkxxnhT3OD51Fr79fHPW2sPWmsPu+5PAbyNMRHlHNOtrLXprj/3AN9Q3BWxpNJ8VqqCi4Cl1trdxz9RHT4HJew+1i3d9eeeE+xT5T8TxpgbgUuA61wNwD8pxc9OpWWt3W2tLbTWFgHvc+L3Vh0+B17AFcDnJ9unKn8OKjm1c1zU1lFbpwS1ddTO+Z3aOmrrQOVu66jA80eLgCbGmAauav4gYNJx+0wCjs0afyXw68l++Csj11jDD4A11tpXTrJPnWPj8Y0xHSn+HFWZxp8xpoYxJujYfYonXFt13G6TgCGmWGfgQImurVXJSSvXVf1zcJySP/dDgW9PsM+PwIXGmJqu7qwXurZVCcaYvsD/Af2ttUdOsk9pfnYqrePmnricE7+30vw/UtldAKy11qad6Mmq/jmo5Kp9OwfU1gG1dY6jto7aOYDaOqC2TgmVt61T2tmYq8uN4hUD1lM8M/ijrm1PUvyDDuBHcRfOjcBCoKHTmcv4/Z9DcbfMFUCy69YPuB243bXPCGA1xTOmzwe6Op27jK9BQ9d7W+56n8c+ByWvgQFGuj4nK4Ekp3O74TrUoLgRE1JiW5X/HFDcyNsJ5FM8pvgWiuefmAZsAH4Bwlz7JgGjSxx7s+vfho3ATU6/lzK+BhspHm997N+FY6vs1AWmuO6f8GenMt5Ocg0+cf28r6C4IRN1/DVwPf7T/yOV8Xaia+DaPubYvwMl9q2Sn4OqeDvR55Nq1M5xvUe1ddTWOXYdql1b5yT/v1Wrds4proPaOmrrVPq2jnEFFBERERERERGRSkpDtEREREREREREKjkVeEREREREREREKjkVeEREREREREREKjkVeEREREREREREKjkVeEREREREREREKjkVeESkQjLGFBpjkkvcHi7Dc9c3xqwqq/OJiIiInA61c0TEHbycDiAichJHrbUJTocQERERcQO1c0SkzKkHj4hUKsaYrcaYF4wxK40xC40xjV3b6xtjfjXGrDDGTDPGxLm21zbGfGOMWe66dXWdytMY874xZrUx5idjjL9jb0pEREQEtXNE5OyowCMiFZX/cV2Xrynx3AFrbWvgLeA117Y3gY+ttW2AT4E3XNvfAGZaa9sCicBq1/YmwEhrbUsgCxjo1ncjIiIi8j9q54hImTPWWqcziIj8iTHmsLU28ATbtwLnW2s3G2O8gV3W2nBjzF4gylqb79q+01obYYzJAGKstbklzlEf+Nla28T1+O+At7X2qXJ4ayIiIlLNqZ0jIu6gHjwiUhnZk9w/Hbkl7heiOclERESkYlA7R0TOiAo8IlIZXVPiz3mu+3OBQa771wG/ue5PA+4AMMZ4GmNCyiukiIiIyBlQO0dEzogquSJSUfkbY5JLPP7BWntsCdGaxpgVFH87Ndi17W7gI2PMQ0AGcJNr+73AKGPMLRR/g3UHsNPd4UVEREROQe0cESlzmoNHRCoV19j0JGvtXqeziIiIiJQltXNE5GxoiJaIiIiIiIiISCWnHjwiIiIiIiIiIpWcevCIiIiIiIiIiFRyKvCIiIiIiIiIiFRyKvCIiIiIiIiIiFRyKvCIiIiIiIiIiFRyKvCIiIiIiIiIiFRy/w/4ZrkdrLBnswAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 1152x576 with 4 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "result=load('P7-training/model_info_2.history.joblib', mmap_mode='r')\n",
    "\n",
    "plt.rcParams[\"figure.figsize\"] = [16, 8]\n",
    "plt.rcParams[\"figure.autolayout\"] = True\n",
    "\n",
    "f, axes = plt.subplots(2, 2)\n",
    "\n",
    "# Plot training & validation accuracy values\n",
    "axes[0,0].plot(result['accuracy'])\n",
    "axes[0,0].set_title('Train Model accuracy / ')\n",
    "axes[0,0].set_ylabel('Accuracy')\n",
    "axes[0,0].set_xlabel('Epoch')\n",
    "axes[0,0].legend(['glove50d'], loc='lower right')\n",
    "\n",
    "axes[0,1].plot(result['val_accuracy'])\n",
    "axes[0,1].set_title('Validation Model accuracy')\n",
    "axes[0,1].set_ylabel('Accuracy')\n",
    "axes[0,1].set_xlabel('Epoch')\n",
    "axes[0,1].legend(['glove50d'], loc='lower right')\n",
    "\n",
    "axes[1,0].plot(result['loss'])\n",
    "axes[1,0].set_title('Train Model loss')\n",
    "axes[1,0].set_ylabel('loss')\n",
    "axes[1,0].set_xlabel('Epoch')\n",
    "axes[1,0].legend(['glove50d'], loc='lower left')\n",
    "\n",
    "axes[1,1].plot(result['val_loss'])\n",
    "axes[1,1].set_title('Validation Model loss')\n",
    "axes[1,1].set_ylabel('loss')\n",
    "axes[1,1].set_xlabel('Epoch')\n",
    "axes[1,1].legend(['glove50d'], loc='upper left')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "bde58412-8b43-4d15-9153-0016cebd0be4",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}