result_collection.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "8e1de5be",
   "metadata": {},
   "source": [
    "# Result collection\n",
    "\n",
    "The notebook is part of the domain-specific study, specifically recording the evaluation of different datasets.\n",
    "\n",
    "## Logitics\n",
    "- [Section 1: Scifact](#section_1)\n",
    "- [Section 2: SCIDOCS](#section_2)\n",
    "- [Section 3: NFCorpus](#section_3)\n",
    "- [Section 4: SciQ](#section_4)\n",
    "- [Section 5: RAG](#section_5)\n",
    "- [Section 6: Others](#section_6)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "a3564bc2",
   "metadata": {},
   "outputs": [],
   "source": [
    "from helpers.helper import *"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "55201805",
   "metadata": {},
   "outputs": [],
   "source": [
    "from helpers.evaluate import Evaluator"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "5eb74fb5",
   "metadata": {},
   "outputs": [],
   "source": [
    "from datasets import load_dataset\n",
    "from transformers import AutoTokenizer, AutoModelForSeq2SeqLM\n",
    "import torch\n",
    "import json\n",
    "import random\n",
    "import re\n",
    "\n",
    "# for evaluation\n",
    "from beir.retrieval.evaluation import EvaluateRetrieval\n",
    "import csv\n",
    "import pickle as pkl\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "from pyserini.search.lucene import LuceneSearcher"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "4a44d4ab",
   "metadata": {},
   "outputs": [],
   "source": [
    "evaluator = EvaluateRetrieval()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c9095192",
   "metadata": {},
   "source": [
    "## <a id='section_1'> Section 1: SciFact </a>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 483,
   "id": "dfa4b988",
   "metadata": {},
   "outputs": [],
   "source": [
    "SCIFACT_DIR = os.path.join(ROOT_DIR, 'retrieval/runs_scifact')\n",
    "scifact_qrels_test = load_qrels(os.path.join(ROOT_DIR, 'retrieval/task_topics/scifact/qrels/test.tsv'))\n",
    "scifact_qrels_train = load_qrels(os.path.join(ROOT_DIR, 'retrieval/task_topics/scifact/qrels/train.tsv'))\n",
    "scifact_qrels_reformat = {('scifact-' + k) : v for k, v in list(scifact_qrels_train.items()) + list(scifact_qrels_test.items())}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 484,
   "id": "dabf7e63",
   "metadata": {},
   "outputs": [],
   "source": [
    "scifact_evaluator = Evaluator(task='scifact', golden_qrels=scifact_qrels_reformat, data_dir=SCIFACT_DIR, evaluator=evaluator)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 485,
   "id": "79e62178",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 28281/28281 [00:00<00:00, 219083.64it/s]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "{'NDCG@5': 0.64652,\n",
       " 'NDCG@20': 0.69168,\n",
       " 'Recall@5': 0.7192,\n",
       " 'Recall@20': 0.85271}"
      ]
     },
     "execution_count": 485,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "scifact_bm25_chunk_dict = scifact_evaluator.evaluate_bm25('run.bm25.hits-100.txt')\n",
    "scifact_bm25_chunk_dict"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "c2f43bb7",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 141500/141500 [00:00<00:00, 214327.40it/s]\n",
      "100%|██████████| 141500/141500 [00:00<00:00, 208412.73it/s]\n",
      "100%|██████████| 141500/141500 [00:00<00:00, 212808.90it/s]\n",
      "100%|██████████| 141500/141500 [00:00<00:00, 208013.46it/s]\n",
      "100%|██████████| 141500/141500 [00:00<00:00, 213482.14it/s]\n",
      "100%|██████████| 141500/141500 [00:00<00:00, 207890.03it/s]\n",
      "100%|██████████| 141500/141500 [00:00<00:00, 210664.61it/s]\n",
      "100%|██████████| 141500/141500 [00:00<00:00, 205477.71it/s]\n",
      "100%|██████████| 141500/141500 [00:00<00:00, 207377.33it/s]\n",
      "100%|██████████| 141500/141500 [00:00<00:00, 205727.86it/s]\n",
      "100%|██████████| 141500/141500 [00:00<00:00, 215848.66it/s]\n",
      "100%|██████████| 141500/141500 [00:00<00:00, 201015.15it/s]\n"
     ]
    }
   ],
   "source": [
    "# all is the results including all the queries\n",
    "# scifact_all_chunk_dict = scifact_evaluator.evaluate(lambda encoder: f'run.all.chunk.{encoder}.hits-500.txt', add_task_prefix=True)\n",
    "# single\n",
    "scifact_whole_chunk_dict = scifact_evaluator.evaluate(lambda encoder: f'run.whole.chunk.{encoder}.hits-500.txt')\n",
    "scifact_whole_prop_dict = scifact_evaluator.evaluate(lambda encoder: f'run.whole.prop.{encoder}.hits-500.txt')\n",
    "scifact_whole_conv_dict = scifact_evaluator.evaluate(lambda encoder: f'rrf/conv.tsv')\n",
    "\n",
    "# ablation study\n",
    "scifact_rrf_whole_prop_dict = scifact_evaluator.evaluate(lambda encoder: f'rrf/qid_cid_max.tsv')\n",
    "scifact_rrf_whole_conv_dict = scifact_evaluator.evaluate(lambda encoder: f'rrf/qid_cid_conv.tsv')\n",
    "scifact_rrf_prop_conv_dict = scifact_evaluator.evaluate(lambda encoder: f'run.rrf.chunk.{encoder}.hits-500.txt')\n",
    "\n",
    "# complete rrf\n",
    "scifact_whole_rrf_chunk_dict = scifact_evaluator.evaluate(lambda encoder: f'rrf/qid_cid_max_conv.tsv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "648bb139",
   "metadata": {},
   "outputs": [],
   "source": [
    "scifact_ndcg = [scifact_whole_chunk_dict, scifact_whole_prop_dict, scifact_whole_conv_dict, scifact_whole_rrf_chunk_dict]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "b4dcd9cc",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "whole-chunk | whole-prop | conv | rrf |\n",
      "NDCG@5\n",
      "ance\n",
      "41.5|43.3|43.5|46.8\n",
      "NDCG@5\n",
      "contriever\n",
      "64.5|64.0|63.2|66.4\n",
      "NDCG@5\n",
      "dpr\n",
      "31.8|36.1|37.1|39.4\n",
      "NDCG@5\n",
      "gtr\n",
      "58.3|60.6|58.4|60.9\n",
      "NDCG@5\n",
      "simcse\n",
      "27.1|32.8|32.7|34.8\n",
      "NDCG@5\n",
      "tasb\n",
      "60.1|60.7|61.7|62.7\n",
      "avg\n",
      "47.2|49.6|49.4|51.8\n",
      "==============\n",
      "NDCG@20\n",
      "ance\n",
      "45.3|46.4|47.3|49.9\n",
      "NDCG@20\n",
      "contriever\n",
      "68.5|68.9|67.5|71.0\n",
      "NDCG@20\n",
      "dpr\n",
      "37.7|40.5|41.3|43.6\n",
      "NDCG@20\n",
      "gtr\n",
      "62.2|63.3|62.0|64.5\n",
      "NDCG@20\n",
      "simcse\n",
      "31.2|37.2|36.9|39.8\n",
      "NDCG@20\n",
      "tasb\n",
      "65.6|64.4|65.0|66.9\n",
      "avg\n",
      "51.8|53.5|53.4|56.0\n",
      "==============\n",
      "Recall@5\n",
      "ance\n",
      "51.2|51.4|50.2|54.4\n",
      "Recall@5\n",
      "contriever\n",
      "74.2|72.1|71.1|74.6\n",
      "Recall@5\n",
      "dpr\n",
      "39.4|44.4|43.0|47.3\n",
      "Recall@5\n",
      "gtr\n",
      "65.6|68.9|67.1|68.9\n",
      "Recall@5\n",
      "simcse\n",
      "32.8|41.2|40.1|42.6\n",
      "Recall@5\n",
      "tasb\n",
      "67.1|71.4|70.5|72.4\n",
      "avg\n",
      "55.0|58.2|57.0|60.0\n",
      "==============\n",
      "Recall@20\n",
      "ance\n",
      "63.1|61.3|62.9|64.6\n",
      "Recall@20\n",
      "contriever\n",
      "86.8|87.6|84.9|89.4\n",
      "Recall@20\n",
      "dpr\n",
      "59.0|58.8|56.6|60.9\n",
      "Recall@20\n",
      "gtr\n",
      "78.1|77.5|78.6|80.3\n",
      "Recall@20\n",
      "simcse\n",
      "47.0|56.0|54.7|59.0\n",
      "Recall@20\n",
      "tasb\n",
      "85.3|83.3|81.2|85.9\n",
      "avg\n",
      "69.9|70.7|69.8|73.3\n",
      "==============\n"
     ]
    }
   ],
   "source": [
    "print('whole-chunk | whole-prop | conv | rrf |')\n",
    "scifact_evaluator.display_metrics(scifact_ndcg)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "0d92d21b",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "NDCG@5\n",
      "ance\n",
      "45.8|46.8|44.9|46.8\n",
      "NDCG@5\n",
      "contriever\n",
      "65.6|66.3|65.0|66.4\n",
      "NDCG@5\n",
      "dpr\n",
      "35.0|38.3|38.0|39.4\n",
      "NDCG@5\n",
      "gtr\n",
      "60.9|60.6|61.5|60.9\n",
      "NDCG@5\n",
      "simcse\n",
      "32.3|33.1|33.9|34.8\n",
      "NDCG@5\n",
      "tasb\n",
      "61.4|63.2|62.1|62.7\n",
      "avg\n",
      "50.2|51.4|50.9|51.8\n",
      "==============\n",
      "NDCG@20\n",
      "ance\n",
      "48.9|50.4|48.6|49.9\n",
      "NDCG@20\n",
      "contriever\n",
      "69.9|70.8|69.5|71.0\n",
      "NDCG@20\n",
      "dpr\n",
      "40.8|42.4|42.1|43.6\n",
      "NDCG@20\n",
      "gtr\n",
      "64.5|64.7|64.4|64.5\n",
      "NDCG@20\n",
      "simcse\n",
      "37.0|37.4|39.0|39.8\n",
      "NDCG@20\n",
      "tasb\n",
      "66.7|67.3|65.3|66.9\n",
      "avg\n",
      "54.6|55.5|54.8|56.0\n",
      "==============\n",
      "Recall@5\n",
      "ance\n",
      "54.6|53.5|50.6|54.4\n",
      "Recall@5\n",
      "contriever\n",
      "75.2|74.8|73.4|74.6\n",
      "Recall@5\n",
      "dpr\n",
      "43.0|46.7|45.9|47.3\n",
      "Recall@5\n",
      "gtr\n",
      "68.0|68.5|70.0|68.9\n",
      "Recall@5\n",
      "simcse\n",
      "41.0|41.8|42.0|42.6\n",
      "Recall@5\n",
      "tasb\n",
      "69.9|72.1|72.9|72.4\n",
      "avg\n",
      "58.6|59.6|59.1|60.0\n",
      "==============\n",
      "Recall@20\n",
      "ance\n",
      "64.5|65.2|62.2|64.6\n",
      "Recall@20\n",
      "contriever\n",
      "88.7|89.2|87.7|89.4\n",
      "Recall@20\n",
      "dpr\n",
      "62.1|59.9|59.1|60.9\n",
      "Recall@20\n",
      "gtr\n",
      "79.3|81.6|79.5|80.3\n",
      "Recall@20\n",
      "simcse\n",
      "56.2|55.4|59.3|59.0\n",
      "Recall@20\n",
      "tasb\n",
      "87.4|85.5|83.0|85.9\n",
      "avg\n",
      "73.0|72.8|71.8|73.3\n",
      "==============\n"
     ]
    }
   ],
   "source": [
    "scifact_ablation = [scifact_rrf_whole_prop_dict, scifact_rrf_whole_conv_dict, scifact_rrf_prop_conv_dict, scifact_whole_rrf_chunk_dict]\n",
    "scifact_evaluator.display_metrics(scifact_ablation)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5a862428",
   "metadata": {},
   "source": [
    "## <a id='section_2'> Section 2: SciDocs </a>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "562db988",
   "metadata": {},
   "outputs": [],
   "source": [
    "SCIDOCS_DIR = os.path.join(ROOT_DIR, 'retrieval/runs_scidocs')\n",
    "scidocs_qrels = load_qrels(os.path.join(ROOT_DIR, 'retrieval/task_topics/scidocs/qrels/test.tsv'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "502c62a3",
   "metadata": {},
   "outputs": [],
   "source": [
    "scidocs_evaluator = Evaluator(task='scidocs', golden_qrels=scidocs_qrels, data_dir=SCIDOCS_DIR, evaluator=evaluator)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "feb4d59b",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20600/20600 [00:00<00:00, 214749.30it/s]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "{'NDCG@5': 0.14767,\n",
       " 'NDCG@20': 0.19924,\n",
       " 'Recall@5': 0.13204,\n",
       " 'Recall@20': 0.23932}"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "scidocs_bm25_chunk_dict = scidocs_evaluator.evaluate_bm25('run.bm25.hits-100.txt')\n",
    "scidocs_bm25_chunk_dict"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "0c96d67f",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 103000/103000 [00:00<00:00, 213220.40it/s]\n",
      "100%|██████████| 103000/103000 [00:00<00:00, 208013.13it/s]\n",
      "100%|██████████| 103000/103000 [00:00<00:00, 215966.92it/s]\n",
      "100%|██████████| 103000/103000 [00:00<00:00, 205985.76it/s]\n",
      "100%|██████████| 103000/103000 [00:00<00:00, 214246.09it/s]\n",
      "100%|██████████| 103000/103000 [00:00<00:00, 206877.56it/s]\n",
      "100%|██████████| 103000/103000 [00:00<00:00, 218452.89it/s]\n",
      "100%|██████████| 103000/103000 [00:00<00:00, 202341.71it/s]\n",
      "100%|██████████| 103000/103000 [00:00<00:00, 214710.02it/s]\n",
      "100%|██████████| 103000/103000 [00:00<00:00, 206100.05it/s]\n",
      "100%|██████████| 103000/103000 [00:00<00:00, 216315.34it/s]\n",
      "100%|██████████| 103000/103000 [00:00<00:00, 201015.15it/s]\n"
     ]
    }
   ],
   "source": [
    "# scidocs_all_chunk_dict = scidocs_evaluator.evaluate(lambda encoder: f'run.all.chunk.{encoder}.hits-500.txt')\n",
    "# single granularity\n",
    "scidocs_whole_chunk_dict = scidocs_evaluator.evaluate(lambda encoder: f'run.whole.chunk.{encoder}.hits-500.txt')\n",
    "scidocs_whole_prop_dict = scidocs_evaluator.evaluate(lambda encoder: f'run.whole.prop.{encoder}.hits-500.txt')\n",
    "scidocs_whole_conv_dict = scidocs_evaluator.evaluate(lambda encoder: f'rrf/conv.tsv')\n",
    "\n",
    "# ablation study\n",
    "scidocs_rrf_whole_prop_dict = scidocs_evaluator.evaluate(lambda encoder: f'rrf/qid_cid_max.tsv')\n",
    "scidocs_rrf_whole_conv_dict = scidocs_evaluator.evaluate(lambda encoder: f'rrf/qid_cid_conv.tsv')\n",
    "scidocs_rrf_prop_conv_dict = scidocs_evaluator.evaluate(lambda encoder: f'run.rrf.chunk.{encoder}.hits-500.txt')\n",
    "\n",
    "# rrf\n",
    "scidocs_whole_rrf_chunk_dict = scidocs_evaluator.evaluate(lambda encoder: f'rrf/qid_cid_max_conv.tsv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "0123d58a",
   "metadata": {},
   "outputs": [],
   "source": [
    "scidocs_ndcg = [scidocs_whole_chunk_dict, scidocs_whole_prop_dict, scidocs_whole_conv_dict, scidocs_whole_rrf_chunk_dict]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "c174fb22",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "whole-chunk | whole-prop | conv | rrf\n",
      "NDCG@5\n",
      "ance\n",
      "9.3|9.2|7.5|9.6\n",
      "NDCG@5\n",
      "contriever\n",
      "13.5|14.5|13.5|15.5\n",
      "NDCG@5\n",
      "dpr\n",
      "7.3|7.8|6.4|8.2\n",
      "NDCG@5\n",
      "gtr\n",
      "13.6|13.2|11.6|13.6\n",
      "NDCG@5\n",
      "simcse\n",
      "7.6|8.2|7.3|9.1\n",
      "NDCG@5\n",
      "tasb\n",
      "13.8|14.3|12.6|14.0\n",
      "avg\n",
      "10.9|11.2|9.8|11.7\n",
      "==============\n",
      "NDCG@20\n",
      "ance\n",
      "13.1|12.9|11.9|14.1\n",
      "NDCG@20\n",
      "contriever\n",
      "18.5|19.4|18.3|20.7\n",
      "NDCG@20\n",
      "dpr\n",
      "10.4|10.6|10.0|11.5\n",
      "NDCG@20\n",
      "gtr\n",
      "18.9|18.0|16.2|19.2\n",
      "NDCG@20\n",
      "simcse\n",
      "9.7|11.1|10.5|12.2\n",
      "NDCG@20\n",
      "tasb\n",
      "19.3|18.1|17.2|19.6\n",
      "avg\n",
      "15.0|15.0|14.0|16.2\n",
      "==============\n",
      "Recall@5\n",
      "ance\n",
      "8.6|7.9|6.7|8.6\n",
      "Recall@5\n",
      "contriever\n",
      "12.4|12.8|12.1|13.8\n",
      "Recall@5\n",
      "dpr\n",
      "6.7|6.8|5.8|7.0\n",
      "Recall@5\n",
      "gtr\n",
      "12.7|12.1|10.2|12.4\n",
      "Recall@5\n",
      "simcse\n",
      "6.9|7.1|6.9|8.4\n",
      "Recall@5\n",
      "tasb\n",
      "12.4|13.0|11.4|12.5\n",
      "avg\n",
      "10.0|10.0|8.8|10.5\n",
      "==============\n",
      "Recall@20\n",
      "ance\n",
      "16.5|15.4|15.9|17.9\n",
      "Recall@20\n",
      "contriever\n",
      "22.9|23.1|22.2|24.7\n",
      "Recall@20\n",
      "dpr\n",
      "13.3|12.6|13.5|13.9\n",
      "Recall@20\n",
      "gtr\n",
      "23.4|22.3|20.1|24.1\n",
      "Recall@20\n",
      "simcse\n",
      "11.2|13.2|13.3|15.0\n",
      "Recall@20\n",
      "tasb\n",
      "23.8|21.0|21.1|24.3\n",
      "avg\n",
      "18.5|17.9|17.7|20.0\n",
      "==============\n"
     ]
    }
   ],
   "source": [
    "print('whole-chunk | whole-prop | conv | rrf')\n",
    "scidocs_evaluator.display_metrics(scidocs_ndcg)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "edad65f4",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "NDCG@5\n",
      "ance\n",
      "10.0|9.0|8.8|9.6\n",
      "NDCG@5\n",
      "contriever\n",
      "14.8|14.6|14.7|15.5\n",
      "NDCG@5\n",
      "dpr\n",
      "8.2|7.5|8.0|8.2\n",
      "NDCG@5\n",
      "gtr\n",
      "13.4|13.8|12.6|13.6\n",
      "NDCG@5\n",
      "simcse\n",
      "8.7|8.5|8.5|9.1\n",
      "NDCG@5\n",
      "tasb\n",
      "13.8|13.8|13.6|14.0\n",
      "avg\n",
      "11.5|11.2|11.0|11.7\n",
      "==============\n",
      "NDCG@20\n",
      "ance\n",
      "13.7|13.4|13.4|14.1\n",
      "NDCG@20\n",
      "contriever\n",
      "20.0|20.1|20.0|20.7\n",
      "NDCG@20\n",
      "dpr\n",
      "11.2|11.2|11.2|11.5\n",
      "NDCG@20\n",
      "gtr\n",
      "18.9|19.5|18.0|19.2\n",
      "NDCG@20\n",
      "simcse\n",
      "11.5|11.6|11.9|12.2\n",
      "NDCG@20\n",
      "tasb\n",
      "19.2|19.6|18.8|19.6\n",
      "avg\n",
      "15.7|15.9|15.5|16.2\n",
      "==============\n",
      "Recall@5\n",
      "ance\n",
      "9.1|8.3|7.9|8.6\n",
      "Recall@5\n",
      "contriever\n",
      "13.2|13.5|13.1|13.8\n",
      "Recall@5\n",
      "dpr\n",
      "7.0|6.6|7.0|7.0\n",
      "Recall@5\n",
      "gtr\n",
      "12.4|12.6|11.3|12.4\n",
      "Recall@5\n",
      "simcse\n",
      "7.9|8.1|8.0|8.4\n",
      "Recall@5\n",
      "tasb\n",
      "12.2|12.2|12.1|12.5\n",
      "avg\n",
      "10.3|10.2|9.9|10.5\n",
      "==============\n",
      "Recall@20\n",
      "ance\n",
      "16.6|17.6|17.6|17.9\n",
      "Recall@20\n",
      "contriever\n",
      "24.3|24.9|23.9|24.7\n",
      "Recall@20\n",
      "dpr\n",
      "13.3|14.1|13.7|13.9\n",
      "Recall@20\n",
      "gtr\n",
      "23.7|24.7|22.4|24.1\n",
      "Recall@20\n",
      "simcse\n",
      "13.8|14.4|15.2|15.0\n",
      "Recall@20\n",
      "tasb\n",
      "23.5|24.3|22.9|24.3\n",
      "avg\n",
      "19.2|20.0|19.3|20.0\n",
      "==============\n"
     ]
    }
   ],
   "source": [
    "scidocs_ablation = [scidocs_rrf_whole_prop_dict, scidocs_rrf_whole_conv_dict, scidocs_rrf_prop_conv_dict, scidocs_whole_rrf_chunk_dict]\n",
    "scidocs_evaluator.display_metrics(scidocs_ablation)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "434f1717",
   "metadata": {},
   "source": [
    "## <a id='section_3'> Section 3: NFCorpus </a>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "465b702f",
   "metadata": {},
   "outputs": [],
   "source": [
    "NFCORPUS_DIR = os.path.join(ROOT_DIR, 'retrieval/runs_nfcorpus')\n",
    "nfcorpus_qrels = load_qrels(os.path.join(ROOT_DIR, 'retrieval/task_topics/nfcorpus/origin/agg_qrels.tsv'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "67e4bfda",
   "metadata": {},
   "outputs": [],
   "source": [
    "nfcorpus_evaluator = Evaluator(task='nfcorpus', golden_qrels=nfcorpus_qrels, data_dir=NFCORPUS_DIR, evaluator=evaluator)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "57a25b79",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 64602/64602 [00:00<00:00, 222209.45it/s]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "{'NDCG@5': 0.37759,\n",
       " 'NDCG@20': 0.30657,\n",
       " 'Recall@5': 0.10251,\n",
       " 'Recall@20': 0.19169}"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "nfcorpus_bm25_chunk_dict = nfcorpus_evaluator.evaluate_bm25('run.bm25.hits-100.txt')\n",
    "nfcorpus_bm25_chunk_dict"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "67dc6da8",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 323500/323500 [00:01<00:00, 211398.68it/s]\n",
      "100%|██████████| 323500/323500 [00:01<00:00, 206857.65it/s]\n",
      "100%|██████████| 323500/323500 [00:01<00:00, 207608.95it/s]\n",
      "100%|██████████| 323500/323500 [00:01<00:00, 263999.13it/s]\n",
      "100%|██████████| 323500/323500 [00:01<00:00, 290537.28it/s]\n",
      "100%|██████████| 323500/323500 [00:01<00:00, 282452.00it/s]\n",
      "100%|██████████| 323500/323500 [00:01<00:00, 221238.22it/s]\n",
      "100%|██████████| 323500/323500 [00:01<00:00, 217378.93it/s]\n",
      "100%|██████████| 323500/323500 [00:01<00:00, 207786.35it/s]\n",
      "100%|██████████| 323500/323500 [00:01<00:00, 203216.30it/s]\n",
      "100%|██████████| 323500/323500 [00:01<00:00, 212441.31it/s]\n",
      "100%|██████████| 323500/323500 [00:01<00:00, 202287.20it/s]\n"
     ]
    }
   ],
   "source": [
    "# nfcorpus_all_chunk_dict = nfcorpus_evaluator.evaluate(lambda encoder: f'run.all.chunk.{encoder}.hits-500.txt')\n",
    "\n",
    "nfcorpus_whole_chunk_dict = nfcorpus_evaluator.evaluate(lambda encoder: f'run.whole.chunk.{encoder}.hits-500.txt')\n",
    "nfcorpus_whole_prop_dict = nfcorpus_evaluator.evaluate(lambda encoder: f'run.whole.prop.{encoder}.hits-500.txt')\n",
    "nfcorpus_whole_conv_dict = nfcorpus_evaluator.evaluate(lambda encoder: f'rrf/conv.tsv')\n",
    "\n",
    "nfcorpus_rrf_whole_prop_dict = nfcorpus_evaluator.evaluate(lambda encoder: f'rrf/qid_cid_max.tsv')\n",
    "nfcorpus_rrf_whole_conv_dict = nfcorpus_evaluator.evaluate(lambda encoder: f'rrf/qid_cid_conv.tsv')\n",
    "nfcorpus_rrf_prop_conv_dict = nfcorpus_evaluator.evaluate(lambda encoder: f'run.rrf.chunk.{encoder}.hits-500.txt')\n",
    "\n",
    "nfcorpus_whole_rrf_chunk_dict = nfcorpus_evaluator.evaluate(lambda encoder: f'rrf/qid_cid_max_conv.tsv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "6d090fa3",
   "metadata": {},
   "outputs": [],
   "source": [
    "nfcorpus_ndcg = [nfcorpus_whole_chunk_dict, nfcorpus_whole_prop_dict, nfcorpus_whole_conv_dict, nfcorpus_whole_rrf_chunk_dict]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "959dd369",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "whole-chunk | whole-prop | conv | rrf\n",
      "NDCG@5\n",
      "ance\n",
      "29.9|29.4|30.3|31.9\n",
      "NDCG@5\n",
      "contriever\n",
      "42.2|43.0|41.4|44.0\n",
      "NDCG@5\n",
      "dpr\n",
      "25.1|25.2|26.5|27.7\n",
      "NDCG@5\n",
      "gtr\n",
      "42.1|42.3|41.5|43.3\n",
      "NDCG@5\n",
      "simcse\n",
      "16.2|20.0|22.8|22.3\n",
      "NDCG@5\n",
      "tasb\n",
      "42.3|42.5|40.9|43.6\n",
      "avg\n",
      "33.0|33.7|33.9|35.5\n",
      "==============\n",
      "NDCG@20\n",
      "ance\n",
      "24.4|24.0|24.5|25.9\n",
      "NDCG@20\n",
      "contriever\n",
      "34.9|35.5|34.9|37.1\n",
      "NDCG@20\n",
      "dpr\n",
      "20.7|20.6|21.4|22.9\n",
      "NDCG@20\n",
      "gtr\n",
      "34.1|34.4|33.6|35.6\n",
      "NDCG@20\n",
      "simcse\n",
      "13.3|16.4|18.3|18.1\n",
      "NDCG@20\n",
      "tasb\n",
      "34.1|34.4|33.1|35.2\n",
      "avg\n",
      "26.9|27.6|27.6|29.1\n",
      "==============\n",
      "Recall@5\n",
      "ance\n",
      "7.8|7.7|7.6|8.4\n",
      "Recall@5\n",
      "contriever\n",
      "11.5|11.8|11.1|12.2\n",
      "Recall@5\n",
      "dpr\n",
      "6.7|6.8|6.8|7.3\n",
      "Recall@5\n",
      "gtr\n",
      "11.3|10.9|10.4|11.3\n",
      "Recall@5\n",
      "simcse\n",
      "4.1|5.2|5.8|5.8\n",
      "Recall@5\n",
      "tasb\n",
      "11.2|11.4|10.7|11.6\n",
      "avg\n",
      "8.7|9.0|8.7|9.4\n",
      "==============\n",
      "Recall@20\n",
      "ance\n",
      "14.9|15.1|14.9|15.9\n",
      "Recall@20\n",
      "contriever\n",
      "21.5|21.9|21.4|22.9\n",
      "Recall@20\n",
      "dpr\n",
      "12.8|12.7|13.2|14.3\n",
      "Recall@20\n",
      "gtr\n",
      "20.4|20.5|20.3|21.5\n",
      "Recall@20\n",
      "simcse\n",
      "8.3|10.0|11.0|11.0\n",
      "Recall@20\n",
      "tasb\n",
      "20.7|20.8|19.6|21.1\n",
      "avg\n",
      "16.4|16.8|16.8|17.8\n",
      "==============\n"
     ]
    }
   ],
   "source": [
    "print('whole-chunk | whole-prop | conv | rrf')\n",
    "nfcorpus_evaluator.display_metrics(nfcorpus_ndcg)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "5482b811",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "NDCG@5\n",
      "ance\n",
      "30.7|32.0|30.8|31.9\n",
      "NDCG@5\n",
      "contriever\n",
      "43.6|43.0|43.2|44.0\n",
      "NDCG@5\n",
      "dpr\n",
      "26.5|27.5|26.6|27.7\n",
      "NDCG@5\n",
      "gtr\n",
      "43.2|43.0|42.4|43.3\n",
      "NDCG@5\n",
      "simcse\n",
      "19.6|21.4|22.8|22.3\n",
      "NDCG@5\n",
      "tasb\n",
      "42.9|42.9|42.7|43.6\n",
      "avg\n",
      "34.4|35.0|34.8|35.5\n",
      "==============\n",
      "NDCG@20\n",
      "ance\n",
      "25.2|26.2|25.1|25.9\n",
      "NDCG@20\n",
      "contriever\n",
      "36.2|36.6|36.3|37.1\n",
      "NDCG@20\n",
      "dpr\n",
      "21.9|22.8|22.2|22.9\n",
      "NDCG@20\n",
      "gtr\n",
      "35.2|35.5|34.9|35.6\n",
      "NDCG@20\n",
      "simcse\n",
      "16.0|17.4|18.6|18.1\n",
      "NDCG@20\n",
      "tasb\n",
      "34.7|34.9|34.5|35.2\n",
      "avg\n",
      "28.2|28.9|28.6|29.1\n",
      "==============\n",
      "Recall@5\n",
      "ance\n",
      "7.7|8.4|8.0|8.4\n",
      "Recall@5\n",
      "contriever\n",
      "12.1|11.5|11.9|12.2\n",
      "Recall@5\n",
      "dpr\n",
      "7.0|7.5|7.3|7.3\n",
      "Recall@5\n",
      "gtr\n",
      "11.4|11.2|10.7|11.3\n",
      "Recall@5\n",
      "simcse\n",
      "5.1|5.5|6.0|5.8\n",
      "Recall@5\n",
      "tasb\n",
      "11.5|11.5|11.4|11.6\n",
      "avg\n",
      "9.1|9.3|9.2|9.4\n",
      "==============\n",
      "Recall@20\n",
      "ance\n",
      "15.5|15.9|15.6|15.9\n",
      "Recall@20\n",
      "contriever\n",
      "22.3|22.8|22.4|22.9\n",
      "Recall@20\n",
      "dpr\n",
      "13.5|14.1|14.0|14.3\n",
      "Recall@20\n",
      "gtr\n",
      "21.0|21.3|21.1|21.5\n",
      "Recall@20\n",
      "simcse\n",
      "9.9|10.5|11.2|11.0\n",
      "Recall@20\n",
      "tasb\n",
      "21.0|21.0|20.8|21.1\n",
      "avg\n",
      "17.2|17.6|17.5|17.8\n",
      "==============\n"
     ]
    }
   ],
   "source": [
    "nfcorpus_ablation = [nfcorpus_rrf_whole_prop_dict, nfcorpus_rrf_whole_conv_dict, nfcorpus_rrf_prop_conv_dict, nfcorpus_whole_rrf_chunk_dict]\n",
    "nfcorpus_evaluator.display_metrics(nfcorpus_ablation)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "72fe6d02",
   "metadata": {},
   "source": [
    "## <a id='section_4'> Section 4: SciQ </a>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "08d0f134",
   "metadata": {},
   "outputs": [],
   "source": [
    "SCIQ_DIR = os.path.join(ROOT_DIR, 'retrieval/runs_sciq')\n",
    "sciq_qrels = load_qrels(os.path.join(ROOT_DIR, 'retrieval/task_topics/sciq/qrels.tsv'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "22500617",
   "metadata": {},
   "outputs": [],
   "source": [
    "sciq_evaluator = Evaluator(task='sciq', golden_qrels=sciq_qrels, data_dir=SCIQ_DIR, evaluator=evaluator)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "a54ffe5f",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 25600/25600 [00:00<00:00, 213810.59it/s]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "{'NDCG@5': 0.9191,\n",
       " 'NDCG@20': 0.92225,\n",
       " 'Recall@5': 0.97656,\n",
       " 'Recall@20': 0.98828}"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sciq_bm25_chunk_dict = sciq_evaluator.evaluate_bm25('run.bm25.hits-100.txt')\n",
    "sciq_bm25_chunk_dict"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "316f1eea",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 128000/128000 [00:00<00:00, 221265.61it/s]\n",
      "100%|██████████| 128000/128000 [00:00<00:00, 213905.41it/s]\n",
      "100%|██████████| 128000/128000 [00:00<00:00, 221034.32it/s]\n",
      "100%|██████████| 128000/128000 [00:00<00:00, 215633.50it/s]\n",
      "100%|██████████| 128000/128000 [00:00<00:00, 220569.37it/s]\n",
      "100%|██████████| 128000/128000 [00:00<00:00, 215838.87it/s]\n",
      "100%|██████████| 128000/128000 [00:00<00:00, 222421.83it/s]\n",
      "100%|██████████| 128000/128000 [00:00<00:00, 214371.76it/s]\n",
      "100%|██████████| 128000/128000 [00:00<00:00, 220107.99it/s]\n",
      "100%|██████████| 128000/128000 [00:00<00:00, 215707.49it/s]\n",
      "100%|██████████| 128000/128000 [00:00<00:00, 217544.07it/s]\n",
      "100%|██████████| 128000/128000 [00:00<00:00, 213554.68it/s]\n"
     ]
    }
   ],
   "source": [
    "# sciq_all_chunk_dict = sciq_evaluator.evaluate(lambda encoder: f'run.all.chunk.{encoder}.hits-500.txt')\n",
    "# single granularity\n",
    "sciq_whole_chunk_dict = sciq_evaluator.evaluate(lambda encoder: f'run.whole.chunk.{encoder}.hits-500.txt')\n",
    "sciq_whole_prop_dict = sciq_evaluator.evaluate(lambda encoder: f'run.whole.prop.{encoder}.hits-500.txt')\n",
    "sciq_whole_conv_dict = sciq_evaluator.evaluate(lambda encoder: f'rrf/conv.tsv')\n",
    "\n",
    "# ablation study\n",
    "sciq_rrf_whole_prop_dict = sciq_evaluator.evaluate(lambda encoder: f'rrf/qid_cid_max.tsv')\n",
    "sciq_rrf_whole_conv_dict = sciq_evaluator.evaluate(lambda encoder: f'rrf/qid_cid_conv.tsv')\n",
    "sciq_rrf_prop_conv_dict = sciq_evaluator.evaluate(lambda encoder: f'run.rrf.chunk.{encoder}.hits-500.txt')\n",
    "\n",
    "# rrf\n",
    "sciq_whole_rrf_chunk_dict = sciq_evaluator.evaluate(lambda encoder: f'rrf/qid_cid_max_conv.tsv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "9fd42a1f",
   "metadata": {},
   "outputs": [],
   "source": [
    "sciq_ndcg = [sciq_whole_chunk_dict, sciq_whole_prop_dict, sciq_whole_conv_dict, sciq_whole_rrf_chunk_dict]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "f5e0bff7",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "chunk | prop | conv | rrf | rrf_w_chunk\n",
      "NDCG@5\n",
      "ance\n",
      "66.4|62.3|66.1|74.4\n",
      "NDCG@5\n",
      "contriever\n",
      "67.2|79.7|83.6|85.2\n",
      "NDCG@5\n",
      "dpr\n",
      "60.6|63.6|67.7|73.6\n",
      "NDCG@5\n",
      "gtr\n",
      "83.3|85.8|88.5|92.9\n",
      "NDCG@5\n",
      "simcse\n",
      "62.3|75.6|80.9|84.0\n",
      "NDCG@5\n",
      "tasb\n",
      "84.8|85.6|85.3|90.5\n",
      "avg\n",
      "70.8|75.4|78.7|83.5\n",
      "==============\n",
      "NDCG@20\n",
      "ance\n",
      "69.1|66.4|69.1|76.8\n",
      "NDCG@20\n",
      "contriever\n",
      "70.0|81.0|84.6|86.7\n",
      "NDCG@20\n",
      "dpr\n",
      "64.1|67.9|70.7|76.1\n",
      "NDCG@20\n",
      "gtr\n",
      "84.4|86.5|89.0|93.0\n",
      "NDCG@20\n",
      "simcse\n",
      "67.3|78.5|83.2|85.5\n",
      "NDCG@20\n",
      "tasb\n",
      "86.3|86.3|86.6|91.0\n",
      "avg\n",
      "73.5|77.8|80.5|84.9\n",
      "==============\n",
      "Recall@5\n",
      "ance\n",
      "80.1|73.4|77.3|87.5\n",
      "Recall@5\n",
      "contriever\n",
      "80.9|92.6|93.4|93.4\n",
      "Recall@5\n",
      "dpr\n",
      "75.0|76.6|80.5|87.1\n",
      "Recall@5\n",
      "gtr\n",
      "94.1|96.1|96.1|98.4\n",
      "Recall@5\n",
      "simcse\n",
      "72.7|85.9|88.7|93.0\n",
      "Recall@5\n",
      "tasb\n",
      "93.0|96.1|93.4|97.3\n",
      "avg\n",
      "82.6|86.8|88.2|92.8\n",
      "==============\n",
      "Recall@20\n",
      "ance\n",
      "89.1|87.5|87.9|95.7\n",
      "Recall@20\n",
      "contriever\n",
      "90.2|96.9|96.5|98.0\n",
      "Recall@20\n",
      "dpr\n",
      "85.9|91.0|90.6|95.3\n",
      "Recall@20\n",
      "gtr\n",
      "97.7|98.4|97.7|98.8\n",
      "Recall@20\n",
      "simcse\n",
      "89.5|96.1|95.7|97.7\n",
      "Recall@20\n",
      "tasb\n",
      "97.7|98.4|97.7|98.8\n",
      "avg\n",
      "91.7|94.7|94.3|97.4\n",
      "==============\n"
     ]
    }
   ],
   "source": [
    "print('chunk | prop | conv | rrf | rrf_w_chunk')\n",
    "sciq_evaluator.display_metrics(sciq_ndcg)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "b97b51b0",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "NDCG@5\n",
      "ance\n",
      "69.0|71.3|67.8|74.4\n",
      "NDCG@5\n",
      "contriever\n",
      "78.0|81.6|83.3|85.2\n",
      "NDCG@5\n",
      "dpr\n",
      "66.6|71.0|69.5|73.6\n",
      "NDCG@5\n",
      "gtr\n",
      "87.2|88.4|89.0|92.9\n",
      "NDCG@5\n",
      "simcse\n",
      "76.1|77.9|80.7|84.0\n",
      "NDCG@5\n",
      "tasb\n",
      "86.7|88.3|85.2|90.5\n",
      "avg\n",
      "77.3|79.7|79.3|83.5\n",
      "==============\n",
      "NDCG@20\n",
      "ance\n",
      "72.0|73.9|70.1|76.8\n",
      "NDCG@20\n",
      "contriever\n",
      "80.1|83.3|84.8|86.7\n",
      "NDCG@20\n",
      "dpr\n",
      "69.9|73.1|72.2|76.1\n",
      "NDCG@20\n",
      "gtr\n",
      "87.5|88.5|89.3|93.0\n",
      "NDCG@20\n",
      "simcse\n",
      "78.0|79.6|82.2|85.5\n",
      "NDCG@20\n",
      "tasb\n",
      "87.0|88.8|85.9|91.0\n",
      "avg\n",
      "79.1|81.2|80.7|84.9\n",
      "==============\n",
      "Recall@5\n",
      "ance\n",
      "85.2|86.3|82.0|87.5\n",
      "Recall@5\n",
      "contriever\n",
      "91.4|92.6|93.4|93.4\n",
      "Recall@5\n",
      "dpr\n",
      "83.6|86.3|84.4|87.1\n",
      "Recall@5\n",
      "gtr\n",
      "98.0|98.4|98.0|98.4\n",
      "Recall@5\n",
      "simcse\n",
      "90.6|91.4|92.2|93.0\n",
      "Recall@5\n",
      "tasb\n",
      "97.7|97.3|96.1|97.3\n",
      "avg\n",
      "91.1|92.1|91.0|92.8\n",
      "==============\n",
      "Recall@20\n",
      "ance\n",
      "94.9|94.9|89.8|95.7\n",
      "Recall@20\n",
      "contriever\n",
      "98.0|97.7|98.0|98.0\n",
      "Recall@20\n",
      "dpr\n",
      "94.5|93.4|93.4|95.3\n",
      "Recall@20\n",
      "gtr\n",
      "98.8|98.8|98.8|98.8\n",
      "Recall@20\n",
      "simcse\n",
      "97.3|96.9|97.3|97.7\n",
      "Recall@20\n",
      "tasb\n",
      "98.8|98.8|98.4|98.8\n",
      "avg\n",
      "97.1|96.7|96.0|97.4\n",
      "==============\n"
     ]
    }
   ],
   "source": [
    "sciq_ablation = [sciq_rrf_whole_prop_dict, sciq_rrf_whole_conv_dict, sciq_rrf_prop_conv_dict, sciq_whole_rrf_chunk_dict]\n",
    "sciq_evaluator.display_metrics(sciq_ablation)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "id": "c7ca80d0",
   "metadata": {},
   "outputs": [],
   "source": [
    "fiqa_evaluator = Evaluator(task='fiqa', golden_qrels=fiqa_qrels_reformat, data_dir=FIQA_DIR, evaluator=evaluator)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "id": "6aa0512d-04c7-4f46-a114-4468dc7adbc9",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 690500/690500 [00:03<00:00, 207996.99it/s]\n",
      "100%|██████████| 690500/690500 [00:03<00:00, 206471.99it/s]\n",
      "100%|██████████| 690500/690500 [00:02<00:00, 285159.90it/s]\n",
      "100%|██████████| 690500/690500 [00:03<00:00, 216149.84it/s]\n",
      "100%|██████████| 690500/690500 [00:03<00:00, 203327.08it/s]\n",
      "100%|██████████| 690500/690500 [00:03<00:00, 206729.09it/s]\n",
      "100%|██████████| 690500/690500 [00:03<00:00, 209990.35it/s]\n",
      "100%|██████████| 690500/690500 [00:03<00:00, 204529.80it/s]\n",
      "100%|██████████| 690500/690500 [00:03<00:00, 206224.26it/s]\n",
      "100%|██████████| 690500/690500 [00:03<00:00, 200304.40it/s]\n",
      "100%|██████████| 690500/690500 [00:03<00:00, 205857.61it/s]\n",
      "100%|██████████| 690500/690500 [00:03<00:00, 205459.38it/s]\n"
     ]
    }
   ],
   "source": [
    "# single granularity\n",
    "fiqa_whole_chunk_dict = fiqa_evaluator.evaluate(lambda encoder: f'run.whole.chunk.{encoder}.hits-500.txt')\n",
    "fiqa_whole_prop_dict = fiqa_evaluator.evaluate(lambda encoder: f'run.whole.prop.{encoder}.hits-500.txt')\n",
    "fiqa_whole_conv_dict = fiqa_evaluator.evaluate(lambda encoder: f'rrf/conv.tsv')\n",
    "\n",
    "# rrf\n",
    "fiqa_whole_rrf_chunk_dict = fiqa_evaluator.evaluate(lambda encoder: f'rrf/qid_cid_max_conv.tsv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "id": "94446277-1a54-4f2e-a60c-2c9c8be384dd",
   "metadata": {},
   "outputs": [],
   "source": [
    "fiqa_ndcg = [fiqa_whole_chunk_dict, fiqa_whole_prop_dict, fiqa_whole_conv_dict, fiqa_whole_rrf_chunk_dict]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "id": "2ffcbe8a",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "NDCG@5\n",
      "ance\n",
      "14.6|8.5|8.4|12.8\n",
      "NDCG@5\n",
      "contriever\n",
      "25.0|18.8|18.6|24.7\n",
      "NDCG@5\n",
      "dpr\n",
      "12.0|8.4|9.2|11.2\n",
      "NDCG@5\n",
      "gtr\n",
      "34.4|22.8|19.6|30.8\n",
      "NDCG@5\n",
      "simcse\n",
      "8.4|7.6|9.3|10.6\n",
      "NDCG@5\n",
      "tasb\n",
      "26.7|15.3|13.9|22.8\n",
      "avg\n",
      "20.2|13.6|13.2|18.8\n",
      "==============\n",
      "NDCG@20\n",
      "ance\n",
      "18.2|10.9|11.3|16.2\n",
      "NDCG@20\n",
      "contriever\n",
      "29.9|23.1|22.9|29.8\n",
      "NDCG@20\n",
      "dpr\n",
      "14.6|10.9|11.8|14.9\n",
      "NDCG@20\n",
      "gtr\n",
      "39.6|27.4|24.2|36.1\n",
      "NDCG@20\n",
      "simcse\n",
      "10.9|9.7|11.6|13.8\n",
      "NDCG@20\n",
      "tasb\n",
      "31.5|19.7|17.8|27.9\n",
      "avg\n",
      "24.1|16.9|16.6|23.1\n",
      "==============\n",
      "Recall@5\n",
      "ance\n",
      "15.6|9.2|9.2|14.6\n",
      "Recall@5\n",
      "contriever\n",
      "26.7|20.6|20.3|26.8\n",
      "Recall@5\n",
      "dpr\n",
      "13.8|9.0|10.1|12.5\n",
      "Recall@5\n",
      "gtr\n",
      "35.6|24.6|21.5|33.2\n",
      "Recall@5\n",
      "simcse\n",
      "9.3|8.7|10.4|12.2\n",
      "Recall@5\n",
      "tasb\n",
      "28.9|17.2|15.9|26.0\n",
      "avg\n",
      "21.7|14.9|14.6|20.9\n",
      "==============\n",
      "Recall@20\n",
      "ance\n",
      "26.4|16.4|17.8|24.8\n",
      "Recall@20\n",
      "contriever\n",
      "41.7|33.3|33.0|41.9\n",
      "Recall@20\n",
      "dpr\n",
      "21.6|16.5|17.8|23.1\n",
      "Recall@20\n",
      "gtr\n",
      "51.0|38.5|34.8|48.9\n",
      "Recall@20\n",
      "simcse\n",
      "16.9|15.5|17.6|21.8\n",
      "Recall@20\n",
      "tasb\n",
      "42.9|30.5|27.5|41.3\n",
      "avg\n",
      "33.4|25.1|24.7|33.6\n",
      "==============\n"
     ]
    }
   ],
   "source": [
    "fiqa_evaluator.display_metrics(fiqa_ndcg)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "id": "777dae85-4cdb-4e5a-9a3e-d5eb78740131",
   "metadata": {},
   "outputs": [],
   "source": [
    "ARGUANA_DIR = os.path.join(ROOT_DIR, 'retrieval/runs_arguana')\n",
    "arguana_qrels = load_qrels(os.path.join(ROOT_DIR, 'retrieval/task_topics/arguana/qrels/test.tsv'))\n",
    "arguana_qrels_reformat = {('arguana-' + k) : v for k, v in arguana_qrels.items()}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "id": "6e2e1a56-dfdd-4940-a1a8-cd469e11db39",
   "metadata": {},
   "outputs": [],
   "source": [
    "arguana_evaluator = Evaluator(task='arguana', golden_qrels=arguana_qrels_reformat, data_dir=ARGUANA_DIR, encoder_suffix='_chunk', evaluator=evaluator)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "id": "71c16855-ca46-4118-85c4-e65bb6131df6",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 703000/703000 [00:03<00:00, 211796.44it/s]\n",
      "100%|██████████| 699000/699000 [00:03<00:00, 208939.57it/s]\n",
      "100%|██████████| 703000/703000 [00:03<00:00, 208890.91it/s]\n",
      "100%|██████████| 703000/703000 [00:03<00:00, 208387.38it/s]\n",
      "100%|██████████| 703000/703000 [00:03<00:00, 212079.46it/s]\n",
      "100%|██████████| 703000/703000 [00:03<00:00, 207932.42it/s]\n",
      "100%|██████████| 703000/703000 [00:03<00:00, 212770.38it/s]\n",
      "100%|██████████| 699000/699000 [00:03<00:00, 209023.43it/s]\n",
      "100%|██████████| 703000/703000 [00:03<00:00, 210437.05it/s]\n",
      "100%|██████████| 703000/703000 [00:03<00:00, 209527.46it/s]\n",
      "100%|██████████| 703000/703000 [00:03<00:00, 212374.73it/s]\n",
      "100%|██████████| 703000/703000 [00:03<00:00, 208532.13it/s]\n"
     ]
    }
   ],
   "source": [
    "# single granularity\n",
    "arguana_whole_chunk_dict = arguana_evaluator.evaluate(lambda encoder: f'run.whole.chunk.{encoder}.hits-500.txt')\n",
    "arguana_whole_prop_dict = arguana_evaluator.evaluate(lambda encoder: f'run.whole.prop.{encoder}.hits-500.txt')\n",
    "arguana_whole_conv_dict = arguana_evaluator.evaluate(lambda encoder: f'rrf/conv.tsv')\n",
    "\n",
    "# rrf\n",
    "arguana_whole_rrf_chunk_dict = arguana_evaluator.evaluate(lambda encoder: f'rrf/qid_cid_max_conv.tsv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "id": "5e255617-0147-4834-a922-71059af938e1",
   "metadata": {},
   "outputs": [],
   "source": [
    "arguana_ndcg = [arguana_whole_chunk_dict, arguana_whole_prop_dict, arguana_whole_conv_dict, arguana_whole_rrf_chunk_dict]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "id": "03470172-6999-4863-a5b1-f8e27951be72",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "NDCG@5\n",
      "ance\n",
      "12.0|11.7|10.1|12.4\n",
      "NDCG@5\n",
      "contriever\n",
      "25.9|24.8|24.1|28.7\n",
      "NDCG@5\n",
      "dpr\n",
      "9.0|8.4|6.1|8.2\n",
      "NDCG@5\n",
      "gtr\n",
      "31.4|25.6|20.4|29.4\n",
      "NDCG@5\n",
      "simcse\n",
      "16.4|12.5|6.3|12.7\n",
      "NDCG@5\n",
      "tasb\n",
      "27.9|18.8|12.9|22.6\n",
      "avg\n",
      "20.5|17.0|13.3|19.0\n",
      "==============\n",
      "NDCG@20\n",
      "ance\n",
      "20.5|21.3|18.6|21.8\n",
      "NDCG@20\n",
      "contriever\n",
      "36.0|35.9|34.5|39.2\n",
      "NDCG@20\n",
      "dpr\n",
      "16.6|16.9|12.2|16.3\n",
      "NDCG@20\n",
      "gtr\n",
      "40.7|36.9|30.0|39.4\n",
      "NDCG@20\n",
      "simcse\n",
      "25.9|20.9|12.3|22.4\n",
      "NDCG@20\n",
      "tasb\n",
      "37.8|30.5|20.8|33.6\n",
      "avg\n",
      "29.6|27.1|21.4|28.8\n",
      "==============\n",
      "Recall@5\n",
      "ance\n",
      "23.0|21.3|19.5|24.2\n",
      "Recall@5\n",
      "contriever\n",
      "48.7|47.0|45.9|54.2\n",
      "Recall@5\n",
      "dpr\n",
      "17.3|15.6|11.7|15.9\n",
      "Recall@5\n",
      "gtr\n",
      "58.7|46.2|38.3|55.1\n",
      "Recall@5\n",
      "simcse\n",
      "31.5|21.5|12.1|24.8\n",
      "Recall@5\n",
      "tasb\n",
      "52.3|35.2|25.5|43.2\n",
      "avg\n",
      "38.6|31.1|25.5|36.2\n",
      "==============\n",
      "Recall@20\n",
      "ance\n",
      "52.9|54.7|49.1|57.0\n",
      "Recall@20\n",
      "contriever\n",
      "82.4|84.5|81.6|89.8\n",
      "Recall@20\n",
      "dpr\n",
      "43.7|45.7|33.3|44.3\n",
      "Recall@20\n",
      "gtr\n",
      "89.5|84.1|71.6|89.0\n",
      "Recall@20\n",
      "simcse\n",
      "64.3|50.9|33.7|58.7\n",
      "Recall@20\n",
      "tasb\n",
      "85.2|75.0|53.1|80.5\n",
      "avg\n",
      "69.7|65.8|53.7|69.9\n",
      "==============\n"
     ]
    }
   ],
   "source": [
    "arguana_evaluator.display_metrics(arguana_ndcg)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8e1e74c6",
   "metadata": {},
   "source": [
    "## <a id='section_5'> Section 5: RAG </a>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "70a61d49",
   "metadata": {},
   "outputs": [],
   "source": [
    "def readjsonl(_path):\n",
    "    with open(_path, 'r') as f:\n",
    "        lines = [json.loads(line) for line in f.readlines()]\n",
    "        \n",
    "    return lines"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 239,
   "id": "6716e7ee",
   "metadata": {},
   "outputs": [],
   "source": [
    "# lines = readjsonl(os.path.join(ROOT_DIR, 'context_qa_rrf.log'))\n",
    "lines = readjsonl(os.path.join(ROOT_DIR, 'context_llama3.log'))\n",
    "# lines = readjsonl(os.path.join(ROOT_DIR, 'context_llama.log'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 241,
   "id": "edb89b53",
   "metadata": {},
   "outputs": [],
   "source": [
    "def print_result(lines, em=10, context_type='max_words'):\n",
    "    # for llama3 -  we always need to remove the original MCQA setupabs\n",
    "    filtered_lines = [line for line in lines if not (line['dataset'] == 'sciq' and 'f1' not in line)]\n",
    "    filtered_lines = [line for line in filtered_lines if context_type in line and line[context_type] == em]\n",
    "    \n",
    "    result_dict = {}\n",
    "    for line in filtered_lines:\n",
    "        dataset = line['dataset']\n",
    "        retrieval_type = line['retrieval']\n",
    "        perf = line['em']\n",
    "        _encoder = line['encoder']\n",
    "        if dataset not in result_dict:\n",
    "            result_dict[dataset] = {}\n",
    "        \n",
    "        if retrieval_type not in result_dict[dataset]:\n",
    "            result_dict[dataset][retrieval_type] = {}\n",
    "        result_dict[dataset][retrieval_type][_encoder] = perf\n",
    "        \n",
    "    \n",
    "    for dataset in ['sciq', 'scifact']:\n",
    "        print(dataset)\n",
    "        if context_type != 'max_words':\n",
    "            retrieval_types = ['chunk', 'rrf_chunk']\n",
    "        else:\n",
    "            retrieval_types = ['qid_max', 'rrf_qid_max']\n",
    "        for retrieval_type in retrieval_types:\n",
    "            avg = np.mean([result_dict[dataset][retrieval_type][_encoder] for _encoder in encoders]).item()\n",
    "            print(f\"{retrieval_type:<15}\", '|'.join([str(result_dict[dataset][retrieval_type][_encoder]) for _encoder in encoders] + [str(round(avg, 1))]))\n",
    "\n",
    "#             print(f\"{retrieval_type:<15}\", '|'.join([str(result_dict[dataset][retrieval_type][_encoder]) for _encoder in encoders]))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 221,
   "id": "cfb74e71",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sciq\n",
      "chunk           52.7|53.9|52.0|59.8|54.7|60.5|55.6\n",
      "rrf_chunk       54.7|61.7|57.4|64.1|61.3|64.5|60.6\n",
      "scifact\n",
      "chunk           51.7|63.4|51.2|61.0|50.0|62.8|56.7\n",
      "rrf_chunk       57.6|64.0|51.7|62.8|48.3|62.2|57.8\n"
     ]
    }
   ],
   "source": [
    "print_result(lines, em=1, context_type='top_passages')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 223,
   "id": "929b0fbe",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sciq\n",
      "chunk           59.4|63.3|57.4|64.8|58.2|66.4|61.6\n",
      "rrf_chunk       62.9|66.0|62.5|66.0|66.4|67.6|65.2\n",
      "scifact\n",
      "chunk           65.1|75.6|59.9|72.1|61.6|74.4|68.1\n",
      "rrf_chunk       69.2|70.9|65.7|73.8|62.8|70.3|68.8\n"
     ]
    }
   ],
   "source": [
    "print_result(lines, em=3, context_type='top_passages')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 244,
   "id": "9ddcbbd1-6822-43e5-9d31-c66fd10b88f3",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sciq\n",
      "qid_max         53.9|56.2|55.5|60.2|56.2|56.2|56.4\n",
      "rrf_qid_max     55.5|57.4|59.0|60.9|59.0|58.2|58.3\n",
      "scifact\n",
      "qid_max         47.1|49.4|49.4|41.9|43.0|50.0|46.8\n",
      "rrf_qid_max     45.9|47.7|52.3|45.9|45.3|52.3|48.2\n"
     ]
    }
   ],
   "source": [
    "print_result(lines, 50)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 246,
   "id": "ea436cbf-b6af-48ad-a8e1-bf2b4064320a",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sciq\n",
      "qid_max         60.5|62.9|60.2|63.7|60.9|60.9|61.5\n",
      "rrf_qid_max     59.8|62.5|60.9|65.2|63.3|62.9|62.4\n",
      "scifact\n",
      "qid_max         61.6|67.4|56.4|66.3|60.5|69.8|63.7\n",
      "rrf_qid_max     66.9|71.5|59.9|63.4|62.2|68.0|65.3\n"
     ]
    }
   ],
   "source": [
    "print_result(lines, 200)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 249,
   "id": "9162ce6f",
   "metadata": {},
   "outputs": [],
   "source": [
    "def print_latex(lines, datasets, ems, context_type='max_words'):\n",
    "    filtered_lines = [line for line in lines if not (line['dataset'] == 'sciq' and 'f1' not in line)]\n",
    "\n",
    "    filtered_lines = [line for line in filtered_lines if context_type in line and line[context_type] in ems]\n",
    "    # from outer to inner: retriever - setup - dataset - em200/500\n",
    "    result_dict = {}\n",
    "    \n",
    "    for line in filtered_lines:\n",
    "        dataset = line['dataset']\n",
    "        retrieval_type = line['retrieval']\n",
    "        perf = line['em']\n",
    "        _encoder = line['encoder']\n",
    "        max_words = line[context_type]\n",
    "        if _encoder not in result_dict:\n",
    "            result_dict[_encoder] = {}\n",
    "        \n",
    "        if retrieval_type not in result_dict[_encoder]:\n",
    "            result_dict[_encoder][retrieval_type] = {}\n",
    "            \n",
    "        if dataset not in result_dict[_encoder][retrieval_type]:\n",
    "            result_dict[_encoder][retrieval_type][dataset] = {}\n",
    "        \n",
    "        if max_words not in result_dict[_encoder][retrieval_type][dataset]:\n",
    "            result_dict[_encoder][retrieval_type][dataset][max_words] = {}\n",
    "            \n",
    "        result_dict[_encoder][retrieval_type][dataset][max_words] = perf\n",
    "    \n",
    "    lines = []\n",
    "    for _encoder in encoders:\n",
    "        for retrieval_type in ['qid_max', 'rrf_qid_max']:\n",
    "            values = []\n",
    "            for dataset in datasets:\n",
    "                for max_words in ems:\n",
    "                    values.append(str(result_dict[_encoder][retrieval_type][dataset][max_words]))\n",
    "            lines.append(_encoder + ' ' + retrieval_type + ' ' + ' & '.join(values))\n",
    "    return lines"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 233,
   "id": "9aa7a4b6",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['ance chunk 51.7 & 65.1 & 52.7 & 59.4',\n",
       " 'ance rrf_chunk 57.6 & 69.2 & 54.7 & 62.9',\n",
       " 'contriever chunk 63.4 & 75.6 & 53.9 & 63.3',\n",
       " 'contriever rrf_chunk 64.0 & 70.9 & 61.7 & 66.0',\n",
       " 'dpr chunk 51.2 & 59.9 & 52.0 & 57.4',\n",
       " 'dpr rrf_chunk 51.7 & 65.7 & 57.4 & 62.5',\n",
       " 'gtr chunk 61.0 & 72.1 & 59.8 & 64.8',\n",
       " 'gtr rrf_chunk 62.8 & 73.8 & 64.1 & 66.0',\n",
       " 'simcse chunk 50.0 & 61.6 & 54.7 & 58.2',\n",
       " 'simcse rrf_chunk 48.3 & 62.8 & 61.3 & 66.4',\n",
       " 'tasb chunk 62.8 & 74.4 & 60.5 & 66.4',\n",
       " 'tasb rrf_chunk 62.2 & 70.3 & 64.5 & 67.6']"
      ]
     },
     "execution_count": 233,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print_latex(lines, ['scifact', 'sciq'], [1, 3], context_type='top_passages')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 250,
   "id": "bf9a87af-b613-4b69-9774-0eed7337c420",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['ance qid_max 47.1 & 61.6 & 53.9 & 60.5',\n",
       " 'ance rrf_qid_max 45.9 & 66.9 & 55.5 & 59.8',\n",
       " 'contriever qid_max 49.4 & 67.4 & 56.2 & 62.9',\n",
       " 'contriever rrf_qid_max 47.7 & 71.5 & 57.4 & 62.5',\n",
       " 'dpr qid_max 49.4 & 56.4 & 55.5 & 60.2',\n",
       " 'dpr rrf_qid_max 52.3 & 59.9 & 59.0 & 60.9',\n",
       " 'gtr qid_max 41.9 & 66.3 & 60.2 & 63.7',\n",
       " 'gtr rrf_qid_max 45.9 & 63.4 & 60.9 & 65.2',\n",
       " 'simcse qid_max 43.0 & 60.5 & 56.2 & 60.9',\n",
       " 'simcse rrf_qid_max 45.3 & 62.2 & 59.0 & 63.3',\n",
       " 'tasb qid_max 50.0 & 69.8 & 56.2 & 60.9',\n",
       " 'tasb rrf_qid_max 52.3 & 68.0 & 58.2 & 62.9']"
      ]
     },
     "execution_count": 250,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print_latex(lines, ['scifact', 'sciq'], [50, 200])"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c13292e3",
   "metadata": {},
   "source": [
    "## Section: 6: Others"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b6a1d958",
   "metadata": {},
   "source": [
    "### Section 6.1: Proposition-level MixGR"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "6d73168e-de42-42c0-a4f3-454fd6bc577b",
   "metadata": {},
   "outputs": [],
   "source": [
    "# {em200: {sciq: {qid_max:, rrf_qid_max:}}}\n",
    "simcse_pid_dict = {}\n",
    "\n",
    "for line in lines:\n",
    "    if line['encoder'] == 'simcse' and (line['retrieval'] == 'qid_max' or line['retrieval'] == 'rrf_qid_max') and (line['max_words'] == 200 or line['max_words'] == 500):\n",
    "        dataset = line['dataset']\n",
    "        max_words = 'EM@' + str(line['max_words'])\n",
    "        retrieval = line['retrieval']\n",
    "        em = line['em']\n",
    "\n",
    "        if dataset not in simcse_pid_dict:\n",
    "            simcse_pid_dict[dataset] = {}\n",
    "\n",
    "        if max_words not in simcse_pid_dict[dataset]:\n",
    "            simcse_pid_dict[dataset][max_words] = {}\n",
    "\n",
    "        simcse_pid_dict[dataset][max_words][retrieval] = em"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 470,
   "id": "11dd4463-8537-48c7-84ab-874c5112325a",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABKUAAAG4CAYAAACUzVSpAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjUuMywgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/NK7nSAAAACXBIWXMAAA9hAAAPYQGoP6dpAABsl0lEQVR4nO3dd3gV1f7+/TuNVELvRCIIoUjkAIaqIFVEelE6SlUBe0HBxvGriMgBDgiIFOkoXQWBo9IRBUFM6CUQQguQHlL380d+mSebJDs7bULg/bquXNeUNWs+s4/kbG7WWuNgsVgsAgAAAAAAAEzkWNgFAAAAAAAA4P5DKAUAAAAAAADTEUoBAAAAAADAdIRSAAAAAAAAMB2hFAAAAAAAAExHKAUAAAAAAADTEUoBAAAAAADAdIRSAAAAAAAAMJ1zYRdQmFJSUhQaGqrixYvLwcGhsMsBAAAAAAAo8iwWi6KiolS5cmU5OmY9Huq+DqVCQ0Pl4+NT2GUAAAAAAADccy5evKiqVatmef6+DqWKFy8uKfVD8vb2LuRqAAAAAAAAir7IyEj5+PgYuUtW7utQKm3Knre3N6EUAAAAAABAPspuqSQWOgcAAAAAAIDpCKUAAAAAAABgOkIpAAAAAAAAmI5QCgAAAAAAAKYjlAIAAAAAAIDpCKUAAAAAAABgOkIpAAAAAAAAmI5QCgAAAAAAAKYjlAIAAAAAAIDpCKUAAAAAAABgOkIpAAAAAAAAmM65sAu4l6SkpCguLk5JSUmFXQqAfOTs7Cx3d3c5OpLjAwAAAEB+IZTKBzdv3lRYWJji4uIkSQ4ODoVcEYD8ZLFYJEnu7u4qW7asSpcuXcgVAQAAAEDRRyiVR1evXtWVK1fk5uam8uXLy9PTU05OTgRTwD3CYrEoOTlZMTExioyM1MWLF5WYmKgKFSoUdmkAAAAAUKQRSuXBzZs3deXKFZUqVUpVq1Zlag9wD/P29laFChUUEhKiK1euyMXFhRFTAAAAAJAHpCh5EBYWJjc3NwIp4D7h6OioqlWrys3NTWFhYYVdDgAAAAAUaYyUyqW0Rc3Lly8vR0dH3bhxQzExMYVdlmk8PT1VpkwZU+6Vm6mQfn5+On78eLZ9DRkyRIsWLbKrz2HDhmnBggVWx9LWGsqJoKAgLVu2THv27NGpU6d08+ZNJSYmytPTU1WrVpW/v7+6dOmiXr16ydXVNUd9V6lSRaGhoXa1ffTRR3XgwIFs26WkpGjhwoVaunSpgoKCdOvWLZUoUUIPPfSQ+vbtqzFjxsjFxSVHdRZljo6O8vb21rVr15SSkkIgDQAAAAC55GDJzd+q7xGRkZEqUaKEIiIi5O3tnaNrY2JidPr0aT344INKTEzUnDlz7qu37jk7O2v06NGmBFMFGUqVLFlSly9flpubm83+4uPjVbFiRYWHh1sdz8kfn0OHDumdd97Rtm3b7Gpfrlw5vfzyy3rrrbfsDn3yO5S6fv26unbtqv3792fZpm7dulq/fr1q1qxp133vBZGRkTp37pweeugheXp6FnY5AAAAAHBXsTdvYaRULiUlJcnBwUFOTk4KDw9XUlKSevTooXLlyhV2aQXu+vXrWrdunWJiYkwbLZVmyZIldrUrWbKkzfNOTk5KTk5WeHi4Vq1apSFDhthsv3r1aiOQSrvWXklJSRo7dqy+/vpr47qSJUuqadOmatiwoSpVqiQvLy9FRETo1KlT2r9/v/766y9dv35dEyZM0LJly7Rq1SrVr1/f7nuWKlVKM2bMsNmmfPnyNs/fvn1bnTp10sGDByVJFStW1KBBg1SrVi1dvHhRS5cu1dmzZxUUFKROnTrpjz/+UKlSpeyusShLe5nB/RREAwAAoGj4cvGfhV0C8uC1IY0LuwRTEUrlUfqRN+XKlVOlSpUKsZp738CBA/OlHx8fH1ksFgUHB2vx4sXZhlKLFy+WJPn6+kqSzp8/b9d9bty4oe7du2v37t3G9RMmTNCAAQNsjs46efKkPvjgA61atUrHjh3TY489po0bN+rxxx+3677u7u55/qy++OILI5CqW7eufvvtN6vQ9e2331bHjh21e/dunTlzRuPHj9ecOXPydM+igrdrAgAAAEDesRgK7lvPPvusJGnnzp0KCQnJsl1oaKh27NhhdY09EhMT9dRTTxmB1MiRI3X8+HENGzYs2+mCtWrV0ooVK7R161ZVrlxZERER6tatm4KCguy+f14kJibqyy+/lJQawHz77bcZRgF6eHho5cqVcnd3lyQtXLhQV69eNaU+AAAAAEDRRyiF+9aIESPk6Oio5ORkzZ8/P8t28+fPV1JSkhwdHTVixAi7+3/hhRd04MABOTk5acqUKZo7d26OFy5v166dfvnlF1WqVEnh4eF69tlnczR1MLd++OEH3bp1S5LUrFkzNWrUKNN2VapUUZcuXSRJCQkJWrlyZYHXBgAAAAC4NxBK4b5Vo0YNNW3aVJK0YsWKLNstX75cUmo4U716dbv63rFjh7755htJqdPc3njjjVzX6efnp61bt8rd3V1Hjx7V1KlTc92XvbZs2WJst2/f3mbbTp06Gds///xzgdUEAAAAALi3EErhvjZo0CBJqWs47dmzJ8P5vXv36sSJE1Zt7fHee+9Jkho1aqSPP/44w/kDBw7o6aefVtmyZeXq6qoqVapo8ODBOnv2rJKTk9W+fXu1b99ely5dkiQ9/PDDevPNNyVJs2bNUkpKis37h4eHq3nz5ipdurRcXFxUqlQp+fv768UXX7RrCmD6NgEBATbbNmvWzNhO+6wAAAAAAMgOoRTua4MGDZKXl5ckZTqFL+2Yl5eX3QuH79u3T3v27JGDg4Pmz58vJycnq/OLFi1SixYt9OOPP+rGjRtKSEhQaGiolixZogYNGmjbtm3avn27tm/frujoaOO6V199VV5eXrpw4YK2b99us4bY2Fjt27dPt27dUlJSksLDw3X06FF99dVX8vf315gxY2y+Oe7cuXPGdo0aNWzeq3r16sYzBgcHZxuYAQAAAAAgEUqhiHFwcLDrZ+bMmXb15+npqc6dO0uS1q9fr/j4eONcfHy81q1bJ0l6+umn5enpaVefa9eulSS1bNlSDRo0sDp3+PBhjR49WklJSSpWrJgGDRqkOXPmaObMmeratatiYmL0zDPPZNpvyZIljVFJtkKpcuXKqW/fvvr3v/+tb775RrNnz9Zrr72mmjVrSpKSk5M1a9Ysm4u2R0VFGdsVK1a0+bwuLi7GZ5OcnGx1LQAAAAAAWXEu7AKAwjZs2DCtWrVK4eHhWrVqlQYPHixJxrG0Nvbatm2bJKlbt24Zzk2cOFHx8fEqVqyYNm/erDZt2hjnxowZo/nz59tcTL1hw4batm1bltPkFi5cqLZt22YYnSVJU6dO1eeff67x48crJSVFa9as0Zw5czR69OgMbWNjY41tDw+PrB/2/0m/gHt4eLhKlCiR7TUAAAAAgPsboRSKlCVLltjVrkmTJnb32b59e1WrVk3BwcH69ttvjVBq8eLFkiRfX1+1a9fO7v7OnDkjSWrRooXV8fj4eCOwev75560CqTTDhw/X6tWrjXZ3KleunCTp5s2bmZ7v0KGDzdreeust3bp1S5999pkk6fPPP880lAIAAAAAoKARSqFIsXddp5x69tlnNXnyZO3YsUOhoaGyWCzasWOHcc5eMTExxjpQfn5+VucCAwON6YHdu3fPso+uXbtmGUolJCRIsh6ZlFMTJ07Uf//7X0VHR+vcuXM6duyY6tSpY9XGw8NDkZGRklJHTWU38in9tMeSJUvmujYAAAAAwP2DNaUASSNGjJCjo6OSkpI0f/58ff3110pOTpajo6PN6XR3un79urGdtoB6mhs3bhjbVatWzbIPW+fOnz8v6f8fMZUbHh4eVmtd/fPPPxnaFC9e3Ni+cuWKzf4SExMVExMjSXJ2dra6FgAAAACArBBKAUp9w1zTpk0lScuWLdOKFSskSc2aNVP16tXt7qdMmTLG9rVr16zOpV8oPSIiIss+0kYoZWbnzp1GXXlRunRpYzuzqYAPPvigsZ02HTErZ8+eVXJysiTpgQcekKMjv1YAAAAAANlj+h7w/wwaNEh79+7VyZMnrY7lRPHixeXh4aHY2FhdvHhRVapUMc75+fnJwcFBFotFu3btUvPmzTPtY8+ePZke37Bhg44fPy5XV1f17NkzR3XdKX0QlT6gSlO3bl3t3r1bknTgwAE99dRTWfa1b98+Y/vOKYtAUfPl4j8LuwTk0WtDGhd2CQAAALATQxqA/2fQoEFWU+68vLxytYaVr6+vJBlrUqUpU6aM/P39JUmzZs3KdLTU2bNnjVFa6Z06dcpYkHzAgAE2p/hlJy4uTocPHzb269atm6FNp06djO3t27fb7G/z5s3GdseOHXNdFwAAAADg/kIoBfw/np6eGj58uPz9/eXv76+RI0daTbmz1xNPPCHJOqxJ8+qrr0qSLl68qObNm2vjxo2KiopSRESEli9frlatWikqKspof+nSJX366acKCAjQlStXVLduXc2YMSOXT5hq0qRJxmLs1apVU7169TK06dy5s0qVKiVJ2rt3rw4ePJhpX5cuXdKmTZskpS6+npNF4QEAAAAA9zdCKSCdadOm6ciRIzpy5IimTp2aqz66du0qSdq9e7fViCRJGjJkiPr06SNJCgoKUrdu3eTt7a2SJUtqwIABCgkJ0dtvv220b9u2rd59912Fh4erRYsW2r59e5ZB2TvvvKO//vrLZm1TpkzR559/buy/8cYbmbZzcXHRa6+9JkmyWCwaPHiw1SLuUupb+fr166e4uDhJ0tChQ1WhQgWb9wcAAAAAIA1rSqFIWbp0qd1t+/TpI1dX1wKsJnMdOnRQ/fr1dfToUb311lvaunWr1fkVK1bI19dXs2fPNt5aJ6W+Ue+DDz7QyJEjNXnyZEmSk5OTGjRooJdeekmDBw+Wk5NTlvddv369Jk+erDp16qh58+aqU6eOSpcurfj4eJ0+fVqbNm2yWi+re/fuGjNmTJb9vfHGG1q/fr0OHjyooKAg+fv7a8iQIapZs6YuXryoJUuW6OzZs5JSF4r/5JNPcvV5AQAAAADuT4RSKFJysvD4k08+WSihlCR99NFH6tmzp7Zt26YpU6bozTffNM45OTnp888/14cffqi9e/cqPDxcVapUUUBAgJycnJSSkqI1a9aoXLlyeuSRR+Tt7Z2jex87dkzHjh3L8ryTk5PGjBmjKVOm2OzHzc1NmzdvVteuXbV//35duXLFCMvSq1u3rtavX2/15kEAAAAAALJDKJWP7pzedK+6X54zL3r06KFu3bppw4YNGj9+vHx8fDKst+Th4aF27dpluNbR0TFXb9f79ttv9csvv2j//v06ceKEbt26pfDwcFksFpUoUUIPPfSQWrRooRdffFEPPvigXX2WK1dOe/bs0cKFC7V06VIFBgYqPDxc3t7eqlmzpvr06aOxY8fKxcUlx/UCAAAAAO5vhFL5wNPTU87Ozlq3bl1hl2IaZ2fnXC0CnhsWi+Wu6uvcuXN2tVu2bJmaNGmiwMBADRw4UMePH9eHH36Yo3stWrRIDg4OGjJkSLZtAwICFBAQkKP+7eHo6Khhw4Zp2LBh+d43AAAAAOD+RSiVD8qUKaPRo0dbrQ90r/P09GS6VjY8PT21ZcsWde7cWX///bc++ugjff/995o0aZK6desmR8es3zPw888/a/Lkyfr111/l5uam+vXrq2HDhiZWDwAAAABAwSKUyidlypQhpEEGVatW1f79+zVkyBB99913CgwMVM+ePVW2bFm1atVK9erVU8WKFeXi4qKrV68qMDBQe/bs0YULFySlrv80cuRI1atXr5CfBAAAAACA/EUoBRQwd3d3rV69Wtu2bdM777yjQ4cOKSwsTGvWrNGaNWsyvcbBwUFt27bVBx98oJYtW5pcMQAAAAAABY9QCjBJ+/bt1b59ex04cEDr16/Xr7/+qtDQUIWFhSk5OVllypTRgw8+qDZt2qhPnz6qX79+YZcMAAAAAECBIZQCTFZQC5IDAAAAAFCUZL3SMgAAAAAAAFBACKUAAAAAAABgunwLpa5fv66vvvpKHTt2VI0aNeTh4SF3d3f5+PioQ4cO+ve//629e/cqOTnZrv6OHj2q1157Tf7+/ipdurS8vLzk5+enAQMGaMuWLflVNgAAAAAAAApBnteUSklJ0cyZMzVhwgRFR0dnOB8SEqKQkBBt27ZNEydO1B9//KHGjRtn2V9SUpLef/99TZ48WSkpKVbnTp48qZMnT2r58uV6+umntWDBApUrVy6vjwAAAAAAkDRq1/LCLgF55KdahV0CYLc8hVJJSUl65plntHbtWqvj1atXV5UqVSRJV65c0ZkzZzIETFkZNWqUFixYYOy7uLiobt268vLy0vHjx3Xjxg1J0g8//KD27dtrz5498vT0zMtjAAAAAAAAwGR5mr43ZMgQI5BydnbW66+/rvPnz+vMmTPauXOndu7cqZMnT+rWrVv67rvv1LlzZzk5OWXZ37x586wCqa5du+rcuXM6fPiwdu/ercuXL2vmzJlydk7N0o4cOaKRI0fm5REAAAAAAABQCHI9Umr58uVavjx1aKe7u7s2bdqktm3bZtrW29tbvXv3Vu/evbPsLzY2Vh988IGx37p1a61du9YqxHJxcdGYMWPk7u6u4cOHS5JWrFih119/XQ0bNsztowAAAAAAAMBkuRopFR0drVdeecXYnzp1apaBlL0WL16sK1euSJIcHBz01VdfZTmqatiwYWrSpIkkyWKxaPLkyXm6NwAAAAAAAMyVq5FSK1as0PXr1yVJtWrV0qhRo/JcyJo1a4ztVq1aqXbt2jbbjxo1Sr///rsk6aefflJ8fLxcXV3zXAcA5BYLgxZ9LAwKAAAAmCdXI6Xmz59vbA8cOFCOjnlamkrR0dHauXOnsf/kk09me02nTp2srt+xY0eeagAAAAAAAIB5cpwmRURE6I8//jD227Rpk+cigoKClJiYaOw3a9Ys22sqVqwoX19fY//IkSN5rgMAAAAAAADmyHEo9ccff8hisRj79evXlyTt3btXQ4cOVY0aNeTm5qbSpUvL399fr7zyig4fPmyzz2PHjlnt16hRw65a0re7sw8AAAAAAADcvXK8ptTff/9tbHt6esrNzU2jR4/W3LlzrdrFx8fr1q1bOnr0qGbMmKFhw4Zp1qxZKlasWIY+g4OD//+CnJ1VqVIlu2p54IEHjO3z58/n8EkAAAAAAABQWHIcSt24ccPYLl68uJ5//nktW7ZMkuTk5KT69eurVKlSCgkJ0alTpySlviFv/vz5On/+vLZs2ZLhrXqRkZFWfdq7RpW3t7exHRUVlW37+Ph4xcfHZ3pfAAAAAAAAmCdXa0qluXLlihFI9evXTyEhIfrrr7/0yy+/6OTJkzp8+LAaN25stN++fbs+/vjjDH3GxMQY225ubnbX4u7unmkfWfn0009VokQJ48fHx8fuewEAAAAAACD/5DiUun37doZj/fv31/Lly1WxYkWr44888oh++eUX1a1b1zg2depU3bx506pd+kXOnZ3tH7yVvm1CQkK27cePH6+IiAjj5+LFi3bfCwAAAAAAAPknx9P3PD09rfbd3d01Y8aMLNsXL15c06ZNU8eOHSWljmhavXq1Ro8ebbTx8PAwtjMLvbKSvu2ddWXG1dVVrq6udvefEzdu3LBrtNa9wtPTU2XKlDHlXg4ODjm+xs/PT8ePH8+2ryFDhmjRokV29Tls2DAtWLDA6lj6Rf/tFRQUpGXLlmnPnj06deqUbt68qcTERHl6eqpq1ary9/dXly5d1KtXrxz/93rp0iWtX79eO3bs0NGjRxUSEqK4uDh5eXmpatWqat68uV544QX961//srvPlJQULVy4UEuXLlVQUJBu3bqlEiVK6KGHHlLfvn01ZswYubi42N3fmjVrtGDBAv3111+6ceOGPD095evrq+7du+uVV16xmpYLAAAAALh35TiU8vLystrv1KlTtuFE+/btVb58eV27dk2StHPnTqtQKn2fcXFxdtcSGxubZV1munHjhubMmaOkpKRCq8Fszs7OGj16tGnBVEHZsGGDbt++ne200fj4eK1duzZP9zp06JDeeecdbdu2LdPzkZGRCgoKUlBQkFauXKlXXnlFL7/8st566y27Qp/nnntOS5YsUXJycoZzaaMDAwMD9c0332j48OGaNWtWtiMTr1+/rq5du2r//v1Wx8PCwhQWFqb9+/dr/vz5Wr9+vWrWrGmzr7i4OPXt21c//PCD1fGEhATdunVLf/31l+bPn6/vvvtOTZo0yfZ5AQAAAABFW45DqbJly1rtN2zYMNtrHBwc9K9//Us///yzJOns2bNZ9hkdHa3o6Gi7QqYrV64Y24UZjsTExCgpKUk9evRQuXLlCq0Os1y/fl3r1q1TTEyM6Z/7kiVL7GpXsmRJm+ednJyUnJys8PBwrVq1SkOGDLHZfvXq1QoPD7e61l5JSUkaO3asvv76a+O6kiVLqmnTpmrYsKEqVaokLy8vRURE6NSpU9q/f7/++usvXb9+XRMmTNCyZcu0atUq1a9f3+Z9Tpw4YfT/0EMPqWXLlsaLB65evarNmzdr586dSklJ0bx58xQREaGVK1dm2d/t27fVqVMnHTx4UJJUsWJFDRo0SLVq1dLFixe1dOlSnT17VkFBQerUqZP++OMPlSpVKsv++vTpox9//NF4/gEDBsjf31/Xr1/XqlWrdPToUV28eFFdunTRvn37VKNGDbs/YwAAAABA0ZPjUKpOnTpW+/aGEunb3bp1y+qcn5+f1f6FCxes1qHKSvo1oWrXrm1XHQWpXLlyqlSpUmGXcU8bOHBgvvTj4+Mji8Wi4OBgLV68ONtQavHixZIkX19fSdL58+ftus+NGzfUvXt37d6927h+woQJGjBggM3RWSdPntQHH3ygVatW6dixY3rssce0ceNGPf7441le4+TkpG7duumNN95Qy5YtM5x/5513tGjRIg0fPlzJyclatWqVBgwYoC5dumTa3xdffGEEUnXr1tVvv/1mFbq+/fbb6tixo3bv3q0zZ85o/PjxmjNnTqZ9LV++3AikKlWqpB07dliNrHrnnXfUr18/fffdd7p+/brGjBmjzZs3Z/msAAAAAICiL8cLnderV89qPz4+3q7r0q//dOdfxu8Mug4fPpxtf4mJifrnn3+y7APIzrPPPispdTppSEhIlu1CQ0O1Y8cOq2vskZiYqKeeesoIpEaOHKnjx49r2LBh2U4XrFWrllasWKGtW7eqcuXKioiIULdu3RQUFJTlNWvXrtX69eszDaTSDB06VMOGDTP2s1pPKzExUV9++aWk1JGO3377bYZRgB4eHlq5cqXxFsyFCxfq6tWrmfY3adIkY3v69OkZpvo5OTlp4cKFRqi7ZcsW/fnnn1k+BwAAAACg6MtxKOXj46Pq1asb++fOnbPruvQjSypUqGB1rnr16qpataqxn/aXeFsOHjxotf6UrREkQGZGjBghR0dHJScna/78+Vm2mz9/vpKSkuTo6KgRI0bY3f8LL7ygAwcOyMnJSVOmTNHcuXNzvHB5u3bt9Msvv6hSpUoKDw/Xs88+m+XUQXunjqYP1o4dO5Zpmx9++MEY0disWTM1atQo03ZVqlQxRlolJCRkOh3wn3/+MRadr1atmvr06ZNpX56enlYj4ZYuXWrH0wAAAAAAiqoch1KS1KNHD2M7q0Wb07t69ar+/vtvY79p06YZ2nTt2tXY/u6775SQkGCzz2XLlhnb9erVY/0Z5FiNGjWM/xZXrFiRZbvly5dLSg1n0geytuzYsUPffPONpNRpbm+88Uau6/Tz89PWrVvl7u6uo0ePaurUqbnuS5JKlChhbGf1tsstW7YY2+3bt7fZX6dOnYzttHXj0tu0aZOx3bp1a5t9pZ9K+Msvv9hsCwAAAAAo2nIVSj333HNydEy9NCgoSBs3brTZ/osvvrB6M123bt0ytBk6dKixHRYWprlz52bZX0hIiLHGz53X3s1SUlIKvQ+LxSKLxVKoNdxNBg0aJCl1Dac9e/ZkOL93716dOHHCqq093nvvPUlSo0aN9PHHH2c4f+DAAT399NMqW7asXF1dVaVKFQ0ePFhnz55VcnKy2rdvr/bt2+vSpUuSpIcfflhvvvmmJGnWrFl5+t8g/fTYKlWqZNom/TTBgIAAm/01a9bM2E77rNJLP802qxFXaZo2bSonJydJ0unTp++p/9YAAAAAANZyFUrVq1fPaprN8OHDrUZCpbdy5UpNmzbN2O/YsWOmfzF99NFHrUZLvfvuu5mGBJGRkerfv7+ioqIkpS6a/NJLL+XmMUwTHx+vpUuXatKkSZo2bZoRNOTEpUuX9OWXX2rSpElaunRptiPJMrN371598skn+uSTT7R3794cX5/2HLamuhU1gwYNMt70mNlzpR3z8vKye5H1ffv2ac+ePXJwcND8+fONkCXNokWL1KJFC/3444+6ceOGEhISFBoaqiVLlqhBgwbatm2btm/fru3btys6Otq47tVXX5WXl5cuXLig7du35/aRrZ7zySefzLRN+mm52Y1CrF69uvGMwcHBGYKkM2fO2N2Xi4uLypcvL0mKi4uze0F5AAAAAEDRk6tQSpKmTJmiBx98UJJ0/fp1BQQEaMyYMdq4caN27dql5cuXq0ePHurXr5+xBk7ZsmU1b968LPucPn26sS5OdHS02rZtq5deekkbNmzQ//73P02bNk0NGjTQrl27Uot3dNTcuXONhZbvVjt37tTZs2clSVFRUVq9enWO+1i1apURUJw9e9ZYeNtely5d0rZt25ScnKzk5GRt27Ytx+FY+ucoLA4ODnb9zJw5067+PD091blzZ0nS+vXrrRbuj4+P17p16yRJTz/9tDw9Pe3qc+3atZKkli1bqkGDBlbnDh8+rNGjRyspKUnFihXToEGDNGfOHM2cOVNdu3ZVTEyMnnnmmUz7LVmypDEqKbeh1Pz587Vv3z5JqW/EHDt2bKbt0kJfSapYsaLNPl1cXIzPJjk52epaKTVITpMWONlSsmRJY/vmzZvZtgcAAAAAFE3Oub2wfPny+umnn9SxY0dduHBB8fHxmjVrlmbNmpVp+8qVK2vTpk164IEHsuzT19dX69evV5cuXXTz5k3Fx8dr9uzZmj17doa2Tk5Omj59epavs7+bhIeHG9sWi0WRkZFKSUkxpkBmJyUlJcNf9NP3mdMa0h/LavqWvX3cC4YNG6ZVq1YpPDxcq1at0uDBgyXJOJbWxl5p66xlNk114sSJio+PV7FixbR582a1adPGODdmzBjNnz/f5mLqDRs21LZt2zKdJpedQ4cO6ZVXXjH2p0yZIm9v70zbxsbGGtseHh7Z9p1+Affw8HCrdavS92VPgJz+zYQRERHZtgcAAAAAFE25DqUkqXbt2jp69KjeeecdLVmyxGqqURo3Nzc9//zzev/99zO8dS8zzZs3199//62XX35ZGzZssFqLKk1AQID+85//WK1lczerWbOmgoKC5ODgICl1CpO9gZSUOiKsRo0axigli8WimjVr5qgGHx8fFStWTImJiZJSR7f4+PjkqI+05yhMS5YssatdkyZN7O6zffv2qlatmoKDg/Xtt98aoVTaumW+vr5q166d3f2lTVdr0aKF1fH4+HgjsHr++eetAqk0w4cP1+rVq7N8gUDaSMKcjiC6ePGiunXrppiYGEmp0xafe+65HPUBAACQnS8X/1nYJSCv7HuvDwDkizyFUpLk7e2t2bNn64svvtCOHTsUHBysmzdvqmTJknrooYf02GOP5Xh6XZUqVfT999/r+vXr2rlzp0JCQpSQkKDKlSvr0UcfVa1atfJatqkeeeQRSdKpU6dUsmRJtWrVKsd99O3bVzt27FB4eLhq1qxp9Gkvb29vPf/888ZaUs2bN89ylExW0u75999/W605ZCZ713XKqWeffVaTJ0/Wjh07FBoaKovFYkyRfPbZZ+3uJyYmxghn/fz8rM4FBgYa0wO7d++eZR9du3bNMpRKW0ss/cik7Fy7dk1t27ZVSEiIpNR1pBYsWGDzGg8PD2PaXWxsrNXIp8ykn/aYfvpdWl9p4uLisq03/RsBs7svAAAAAKDoynMolcbDw8Pq1fD5oVy5curVq1e+9lkYHBwc1KBBgwzrC+VEsWLF1L59+zzVUaFCBfXo0SPX16c9R4UKFWyuDVYUjRgxQlOmTFFSUpLmz5+vlJQUJScny9HR0eZ0ujtdv37d2E5bQD3NjRs3jO2qVatm2Yetc2kLf6eNmMrOjRs31KZNG506dUqS1KZNG61fv17Ozrb/6BcvXtwIpa5cuWIzHEpMTDRGYDk7O6t48eJW59OHn9euXcu25vTTREuXLp1tewAAAABA0ZTrhc6Be0mNGjXUtGlTSdKyZcu0YsUKSVKzZs1Uvbr9Y5jLlCljbN8ZwKRfKN3WWknpFwa/086dO426shMeHq42bdooMDBQUurC6z/88INdo6zSXmIgWb89LzNnz541XmbwwAMPZJiamv6Ne9n1lZiYaHxu7u7u8vX1zbZWAAAAAEDRRCgF/D+DBg2SJJ08eVInT560Omav4sWLG9PVLl68aHXOz8/PWFcs7Q2SmdmzZ0+mxzds2KDjx4/L1dVVPXv2tFlHRESE2rZtq7///luS1LRpU/388892T6WtW7eusX3gwAGbbdPe5idlnLIoSQ8//LCxffDgQZt97d+/3wi4HnrooRytvQYAAAAAKFr4Gx/w/wwaNMhqyp2Xl1eu1rBKG92TtiZVmjJlysjf31+SNGvWrExHS509e9YYpZXeqVOnNHr0aEnSgAEDbE7xi4qKUvv27XXo0CFJUuPGjbV161a73qKXJv1U3O3bt9tsu3nzZmO7Y8eOGc6nf0Pmb7/9ZrOvTZs2GduZLQQPAAAAALh3EEoB/4+np6eGDx8uf39/+fv7a+TIkVZT7uz1xBNPSLIOa9K8+uqrklJHUTVv3lwbN25UVFSUIiIitHz5crVq1UpRUVFG+0uXLunTTz9VQECArly5orp162rGjBlZ3js2NlZPPvmk/vjjD0lSw4YNtX379gzrPGWnc+fOKlWqlCRp7969WY5wunTpkhEkubq6Zroo/MMPP6w6depIkoKDg/X9999nWfvSpUuN/YJa1B4AAAAAcHcglALSmTZtmo4cOaIjR45o6tSpueqja9eukqTdu3fr8OHDVueGDBmiPn36SJKCgoLUrVs3eXt7q2TJkhowYIBCQkL09ttvG+3btm2rd999V+Hh4WrRooW2b9+eZVB2+/ZtPfXUU8YbFv39/bV9+/ZcvcHOxcVFr732miTJYrFo8ODBVou4S6khUr9+/Yw36g0dOlQVKlTItL8JEyYY2y+//LKx8Hqa5ORkPffcc7p8+bKk1DcENm7cOMd1AwAAAACKjnx7+x5ghvQjabLTp08fuxb1zm8dOnRQ/fr1dfToUb311lvaunWr1fkVK1bI19dXs2fPNt5aJ6W+Ue+DDz7QyJEjNXnyZEmSk5OTGjRooJdeekmDBw+Wk5NTlvft27evMWXQ09NTI0aM0I8//phtvVmNSHrjjTe0fv16HTx4UEFBQfL399eQIUNUs2ZNXbx4UUuWLNHZs2clpS5m/sknn2R5j/79+2v58uX68ccfFRoaqoCAAA0aNEj+/v4KCwvT8uXLdfToUeNz+O9//5tt3QAAAACAoo1QCkVKThYef/LJJwsllJKkjz76SD179tS2bds0ZcoUvfnmm8Y5Jycnff755/rwww+1d+9ehYeHq0qVKgoICJCTk5NSUlK0Zs0alStXTo888oi8vb3tumf6KXYxMTEaO3asXddlFUq5ublp8+bN6tq1q/bv368rV64YYVl6devW1fr1663ePJiZ7777Tn379tUPP/yg8PBwzZw5M0MbHx8ffffdd1Zv7AMAAAAA3JsIpfLRndOb7lX3y3PmRY8ePdStWzdt2LBB48ePl4+PT4b1ljw8PNSuXbsM1zo6Omb7dj2zlCtXTnv27NHChQu1dOlSBQYGKjw8XN7e3qpZs6b69OmjsWPHysXFJdu+3N3dtWnTJq1Zs0bffPON/vrrL928eVMeHh7y9fVVjx499Morr9gdwgEAAAAAijZCqXzg6ekpZ2dnrVu3rrBLMY2zs3OuFgHPDYvFclf1de7cObvaLVu2TE2aNFFgYKAGDhyo48eP68MPP8zRvRYtWiQHBwcNGTIk27aXLl3KUd/2cnR01LBhwzRs2LB86a9Xr17q1atXvvQFAAAAACi6CKXyQZkyZTR69Gir9YHudZ6entlO17rfeXp6asuWLercubP+/vtvffTRR/r+++81adIkdevWTY6OWb9n4Oeff9bkyZP166+/ys3NTfXr11fDhg1NrB4AAAAAgIJFKJVPypQpQ0iDDKpWrar9+/dryJAh+u677xQYGKiePXuqbNmyatWqlerVq6eKFSvKxcVFV69eVWBgoPbs2aMLFy5ISl1/auTIkapXr14hPwkAAAAAAPmLUAooYO7u7lq9erW2bdumd955R4cOHVJYWJjWrFmjNWvWZHqNg4OD2rZtqw8++EAtW7Y0uWIAAAAAAAoeoRRgkvbt26t9+/Y6cOCA1q9fr19//VWhoaEKCwtTcnKyypQpowcffFBt2rRRnz59VL9+/cIuGQAAAACAAkMoBZgsICBAAQEBhV0GAAAAAACFKuuVlgEAAAAAAIACQigFAAAAAAAA0xFKAQAAAAAAwHSEUgAAAAAAADAdoRQAAAAAAABMRygFAAAAAAAA0xFKAQAAAAAAwHSEUgAAAAAAADAdoRQAAAAAAABMRygFAAAAAAAA0xFKAQAAAAAAwHSEUgAAAAAAADAdoRQAAAAAAABMRygFAAAAAAAA0xFKAQAAAAAAwHSEUgAAAAAAADAdoRQAAAAAAABM51zYBdwrRu1aXtglmGruY/0LuwQAAHCPud++T92L/FSrsEsAABQhjJQCAAAAAACA6QilAAAAAAAAYDpCKQAAAAAAAJiOUApFxsqVK9WmTRuVLVtWrq6uqlKlirp166b//e9/kqSZM2fKwcFBDg4OmjlzptW1ts7dKSdtb9++rf/85z9q27atKlasKFdXVxUvXly1atXSyJEjdeLECZvX9+7d27hXWttvv/1Wbdu2VaVKleTi4iIHBwdJ0iOPPCIHBwe5uLgoJCTEZr+SFBISYlzv7++fbfuspNUXEBAgSQoNDdUrr7yiWrVqycvLS15eXqpXr57effddxcTE2OyrSpUqcnBwUJUqVSRJsbGx+uSTT9SoUSOVKVNGjo6Oxn3Su3btmt588001aNBApUqVkouLi8qVK6fmzZvrk08+UWxsrM37ZvY5z5kzR82bN1e5cuWM/56eeeYZHTp0KDcfEwAAAAAgh1joHHe9pKQkPfvss1qzZo3V8dDQUG3cuFE//PCD3n33XZUvX97Uunbt2qV+/frp0qVLVscTEhJ06tQpnTp1SosWLdL//d//6Y033si2v/j4eHXo0EHbtm3L9PywYcP08ssvKykpSXPnztWkSZNs9vfVV18pKSlJkvT888/b+VS27d27Vz169NC1a9esjgcFBSkoKEjLli3Tzz//rNq1a2fb1/Hjx9WlSxedPn3aZrvly5dr9OjRioqKsjoeFhamsLAw7du3T7NmzdL333+v5s2bZ3vfhISETD/n0NBQrV69WuvWrdPnn3+uV155Jdu+AAAAAAC5RyiFu97zzz9vBFIuLi7q1auXWrZsKUdHRx04cEArV67Uv//9b7Vv3960mv73v//p6aef1u3bt+Xg4KCWLVuqXbt2qlq1qmJjY7Vv3z6tXbtWt2/f1ptvvilXV1eNHTvWZp9jx47Vzp075ePjo759+6p27dqKiYnRzp07JaWGUhMmTFBUVJSWLFliM5RKSUnR0qVLJUmenp75EkpFRUWpT58+unbtmlq2bKkuXbqoTJkyOn78uJYtW6bLly/rwoULateunY4ePapSpUpl2VdCQoJ69Oih06dPq1GjRurWrZuqVKmia9eu6cqVK0a77777ToMHD1ZycrIkqXHjxurevbsqVaqk8+fPa+XKlTp16pQuX76sDh06aPfu3WrQoIHN53j99de1bds2Va5cWf3791ft2rV148YNbdq0Sbt371ZiYqJee+01lS9fXv3785ZJAAAAACgohFK4q/38889asmSJJKl48eL66aef1LJlS+P8Cy+8oFdffVVt27bNcoRRfgsPD9fAgQN1+/ZtFS9eXKtWrVKnTp2s2owZM0aBgYFq3769Ll++rLffflu9e/dWpUqVsux3586d6tSpk9auXSs3Nzfj+MsvvywpNVzq1auXFi1apODgYP3000966qmnMu3rp59+0oULFyRJXbt2lbe3d14fW8ePH5ckffrpp3rnnXeszk2YMEEdOnTQgQMHdOnSJb366qtatGhRln2ljXL66KOP9P7772fa5tatWxo9erQRSH3wwQf68MMPrdpMnDhRgwcP1sqVKxUTE6OBAwfqn3/+sfkc27ZtU5MmTfTzzz+rRIkSxvG33npLn376qd59911ZLBa9/PLL6ty5s1UbAAAAAED+YU0p3NW++OILY/v//u//rAKpNP7+/vrvf/9rWk3/+c9/jNE8c+fOzRBIpalXr57mzp0rSYqLi9OMGTNs9luhQgWtXLnSKpC6U/rRVvPmzcuy3ddff21sv/DCCzbvmxNPP/10hkBKkkqUKKE1a9bIy8tLUur6X2FhYTb7at++fZaBlCTNnj1bN2/elCQ99dRTGQIpKXXk3JIlS1SrVi1JUmBgoDZs2GDzvl5eXlqzZk2mYdP48ePVuXNnSanBWfrPEQAAAACQvwilcNeKi4vTjh07JEllypTR6NGjs2z7zDPPqHr16qbUtWrVKkmSr6+v+vXrZ7Ntly5dVK5cOUkyFmTPSr9+/bId0dSwYUM1atRIkrRlyxZdv349Q5tr165py5YtkqQ6deroscces9lnTrz11ltZnqtatap69OghKXV9rPXr19vsa9y4cTbPpw+X3n333SzbOTs7W63/9P3339vst2fPnsZC65lJ/4zZBVwAAAAAgNxj+h7uWgcOHFBiYqIkqUmTJnJ2tv2f62OPPaazZ88WaE03btzQyZMnJaUGZWnrNtni6emp69ev68yZMzbbtW7d2q4ahg8froMHDyo+Pl7z5s3Te++9Z3V+3rx5SkhIkCQNHTrUrj7t4eXlpRYtWths07ZtW2O65YEDBzR8+PBM2zk5OalNmzZZ9pOSkqKjR49KkkqWLJntfbt27aoXX3xRkrJ9e167du1snm/ZsqW8vLwUHR1t1AAAAAAAyH+EUrhrXbx40diuUaNGtu0feuihgixHknTu3DmlpKRIkg4ePKhBgwbZfW1kZKTN8w888IBd/QwZMkTjx49XeHi4Fi9enCGU+vbbbyVJbm5uGjZsmNW5qKgom6N/ypcvrw4dOmRZn6Oj7cGVfn5+xvbly5ezbFeiRAl5eHhkef7WrVu6ffu2pNQRadmpUqWKESRlNnosqxoz4+joqKpVq+r48eOKiIhQXFyc3N3ds60BAAAAAJAzhFK4a0VFRRnbtgKMNJ6engVZjqTUsCS3kpKSbJ635xklyd3dXX379tW8efN06tQp/frrr3riiSckSb/88otOnTolSercubPKlCljdW1oaKjNIO3RRx/NMpSyJ5gpXry4sR0TE5NlO1vrZklSRESEsW3v5+Lh4aHo6Gib95VkrHuVXV9pIiMjCaUAAAAAoACwphTuWukDjtjY2GzbZxdG2CttJFR2NfXs2VMWiyVHP/ll7NixcnBwkCTNmTPHOJ62sLokm2tw5UZcXFy2bdIHiXkJCdMvQm7P//bp22V33+joaLv7kpQvby4EAAAAAGREKIW7lo+Pj7Gd3XpMknT69Oksz7m6uhrbaestZcXWW+OqVatmbIeGhmZbU0F5+OGH1bRpU0nSDz/8oIiICIWHh2vTpk2SUqcyZrZ2kp+fn83Q7MCBA1ne88KFCzYDO0k6ceKEsV2pUqXcPJokqVSpUsZoquDg4Gzbh4aGGmFT+fLl7a4xMykpKQoJCZGUGo4xSgoAAAAACgahFO5aAQEBcnFxkST9/vvv2U5/27VrV5bnSpcubWxnFybZCmYqVapkrG91+PDhPE3ny6sRI0ZISh3V88033+ibb74xRjMNGTIk3+8XHR2tvXv32mzzyy+/GNsBAQG5vpejo6Pq168vKXXK5L59+2y237hxo7H9r3/9y2bb7N6CuGfPHiPgSqsBAAAAAJD/CKVw13J3dzfeSHfjxg19/fXXWbb9/vvvbb55z9/f39jesWNHlu2OHz9uFaxkpl+/fpKk27dva+LEiTbbFqT+/furbNmykqRFixZp4cKFkqRixYpp5MiRBXLPKVOmZHkuNDRUa9eulZS6ZlT37t3zdK9u3boZ25999lmW7ZKSkjR9+nRjv0+fPjb7Xbt2rc1gMv0z5vUZAAAAAABZI5TCXe2NN94wtt95551MR8z8888/evHFF232U6tWLVWvXl1S6lvzVqxYkaFNaGioevXqle2IrDfffFOVK1eWJM2ePVvvv/++kpOTs2x/48YNffLJJzbfepcbrq6uRkB29OhRBQYGSpKefPLJbKew5dbGjRszDaYiIyPVu3dvY4TRM888YwRmufXSSy8ZI9w2btyoTz75JEObpKQkDR06VMePH5eUOrKpa9euNvuNiopS7969M30b4pQpU4wpkGXLltXw4cPz9AwAAAAAgKzx9j3c1Tp06KBBgwZpyZIlioyMVOvWrdWrVy899thjcnBw0IEDB7Ry5UrFxcWpffv22rZtW5Z9jRs3Tq+88ookafDgwfrhhx/0+OOPy2Kx6K+//tKqVasUFRWlTp06afPmzVn24+3tre+//14dOnRQdHS0Jk2apEWLFunpp59W3bp1Vbx4cUVEROjs2bM6ePCgfv/9dyUmJmr27Nn5/fHopZde0qxZs6zWesrvBc7T1K5dW5GRkXrrrbe0ceNGdevWTaVKldLx48e1fPlyY/RRlSpVNG3atDzfr2TJkpozZ4769eun5ORkTZgwQRs3blSPHj1Uvnx5BQcHa+XKlTp58qSk1AXOv/3222z7TfvvpE6dOhowYID8/Px069YtbdiwQbt375YkOTg4aPr06VYLrgMAAAAA8hehFO56CxYsUHR0tNatW6eEhAStWLHCaqSTo6OjJk6cqHLlytkMpcaOHatdu3ZpzZo1SkpK0vLly7V8+XLjvIuLi7744gs5OzvbDKUkqVmzZtqzZ4+effZZHTt2TBcvXtRXX32VZftixYoVyOglPz8/tWzZUjt37pQk+fr6qlOnTvl+Hyn1zYMLFixQ9+7dtXv3biPASc/Hx0c///yzSpUqlS/37NOnjxITEzVq1ChFR0frwIEDma75ValSJX333Xdq0KBBtn1OnTpVr7/+urZt25bpqC8XFxdNnjxZ/fv3z49HAAAAAABkgel7uOs5Oztr7dq1WrFihVq3bq1SpUqpWLFiqlSpkrp06aKtW7fq448/zrYfR0dHrV69Wl9//bWaNm0qb29vFStWTJUrV1bfvn21e/duvfrqq3bX5e/vr3/++UfLli1Tr1695OvrKy8vLzk5Oal48eLy8/NTz549NXPmTIWEhKhXr155+Riy9MQTTxjbgwYNKpB7pGnWrJn++usvvfzyy6pZs6Y8PT3l4eGhOnXqaPz48QoMDFSdOnXy9Z79+/fX6dOn9cYbb8jf318lS5aUs7OzSpcuraZNm+rf//63Tp06pRYtWtjVX7FixbR161Z99dVXatasmcqUKSMXFxdVqlRJffv21f79+3P03wEAAAAAIHccLBaLpbCLKCyRkZEqUaKEIiIi5O3tnaNrIyIiFBwcrIceekgeHh4FVCFyYubMmRo3bpwkacaMGRo7dmwhV2SORo0a6dChQ3J2dtb58+dVpUqVfO3fwcFBkvToo4/afDPh3ax3795as2aNpNTF7P38/PLUX2xsrE6fPq1q1apZTfEbtWu5jatQFPidrVXYJSCPXhvSuLBLQB7we7To4/do0Xei+snCLgF5xJ/Dou1e+S5jb97CSCmgCDtw4IAOHTokKXX9rfwOpAAAAAAAKCiEUkARNmHCBGM7bRF3AAAAAACKAhY6B4qQwMBAHT9+XBEREVq3bp2xsHvz5s3Vvn37Qq4OAAAAAAD7EUoBRciCBQv05ZdfWh0rUaKEvv7660KqCAAAAACA3GH6HlAEOTo6qmLFiurWrZv27dununXrFnZJAAAAAADkCCOlcM8YO3bsPf/GvalTp2rq1Kmm3vNeeEHn999/X9glAAAAAADuwEgpAAAAAAAAmI5QCgAAAAAAAKYjlAIAAAAAAIDpCKUAAAAAAABgOkIpAAAAAAAAmI5QCgAAAAAAAKYjlAIAAAAAAIDpCKUAAAAAAABgOkIpAAAAAAAAmI5QCgAAAAAAAKYjlAIAAAAAAIDpCKUAAAAAAABgOkIpAAAAAAAAmI5QCgAAAAAAAKYjlAIk/fTTT3JwcJCDg4Nef/31wi4HAAAAAIB7HqEU7nppYVHaz9ChQ+2+dtiwYRmuv9tERERo1qxZ6t27t2rWrKmyZcvKxcVFnp6eqlSpkpo3b64XXnhBmzdvVnJyss2+evfuneF5036KFSumMmXKqHHjxhozZoz++ecfk54QAAAAAICMnAu7gHvFl4v/LOwSTPXakMaFdu8NGzbo9u3bcnNzs9kuPj5ea9euNamqnIuPj9fEiRM1b948RUREZDiflJSk2NhYXblyRfv27dOcOXNUvnx5jRs3Tq+99prc3d1zdL/ExETdvHlTN2/e1MGDBzVnzhy98sor+uKLL/LrkQAAAAAAsBuhFIoMJycnJScnKzw8XKtWrdKQIUNstl+9erXCw8Otrs3KU089JYvFkp/l2nTp0iV169ZNBw8eNI5Vq1ZNLVu2VP369VWmTBklJiYqLCxMR44c0b59+xQaGqpr165pwoQJqlatmgYOHGjzHkOHDlXbtm2N/YSEBJ07d07r1q1TYGCgkpOTNXXqVJUsWVITJkwosGcFAAAAACAzhFIoMnx8fGSxWBQcHKzFixdnG0otXrxYkuTr6ytJOn/+fAFXaJ/Y2Fh17NhRgYGBkqQKFSroyy+/VP/+/W1et2nTJk2ZMkW7du2y6z4NGzbMNLj66KOP9MILL2jevHmSpMmTJ2vcuHHy9vbO4ZMAAAAAAJB7rCmFIuXZZ5+VJO3cuVMhISFZtgsNDdWOHTusrrlbjB492gikfH199fvvv2cbSElSly5dtHPnTs2fP1/lypXL9f0dHR01Y8YMo4/o6Ght3bo11/0BAAAAAJAbhFIoUkaMGCFHR0clJydr/vz5WbabP3++kpKS5OjoqBEjRmTbr62378XGxqpmzZrG+bQRWFlp06aN0fatt96yOnfy5EktX75cUuqUwmXLlqlatWrZ1pfesGHD1LFjxxxdcydXV1c1aNDA2D9+/Hie+gMAAAAAIKcIpVCk1KhRQ02bNpUkrVixIst2acFPs2bNVL169Tzd08PDQ8uWLZOrq6sk6eWXX9bZs2czbfvZZ5/p119/lSQ1btxYn376qdX5adOmGWtbtWnTRs2bN89TbXmR9jySFBcXV2h1AAAAAADuT4RSKHIGDRokKXXU0Z49ezKc37t3r06cOGHVNq8CAgI0ceJESVJERIT69euXYeH0gwcP6sMPP5QkeXt7a+XKlXJycrJq89tvvxnb2S1UXtDSPiNJOR6tBQAAAABAXhFKocgZNGiQvLy8JCnTKXxpx7y8vPI1+HnvvffUunVrSdKBAweMkEpKHWnUr18/xcfHS5KmT5+uGjVqWF0fERGhkydPGvutWrXKt9pyauXKlTp16pSk1DWmWrZsWWi1AAAAAADuT4RSKHI8PT3VuXNnSdL69euNIEiS4uPjtW7dOknS008/LU9Pz3y997Jly1S2bFlJ0pQpU7Rz505J0osvvmiEPH379tXQoUMzXBsSEqKUlBRJqVPnzB6dFB8fr6NHj+qtt96yqq9Tp06qW7euqbUAAAAAAEAohSJp2LBhkqTw8HCtWrXKOL5q1SqFh4dbtclPlStX1rx58+Tg4KCkpCQNGjRICxYs0KJFiySlvk3v66+/zvTaa9euGdvFixfP9l4PPPCAsWD6nT9VqlTJ9vpx48ZZXePm5iZ/f39NmTLFCPIaNWqkpUuX2vHkAAAAAADkL0IpFEnt27c3Rhp9++23xvG0N+P5+vqqXbt2BXLvHj16aPjw4ZKkCxcuGOGXi4uLli5dKm9v7wK5b35ydXXVzJkztX//fpUsWbKwywEAAAAA3IecC7sAILeeffZZTZ48WTt27FBoaKgsFot27NhhnCtIM2bM0O7du3Xs2DHj2Pjx49WiRYssrylfvryxHRUVle09Zs6cmaFdThZuHzp0qNq2bStJSklJ0aVLl7Rr1y79/PPPio+P1xdffKH27dvLz8/P7j4BAAAAAMgvhFIoskaMGKEpU6YoKSlJ8+fPV0pKipKTk+Xo6KgRI0YU6L3d3Nzk4+NjhFJOTk7ZBmFVq1aVo6OjUlJSFB8fr+DgYJvrSnXr1i3DsZyEUg0bNsyw0Pv48eO1detWde3aVcHBwXryySd1+PBhlShRwu5+AQAAAADID0zfQ5FVo0YNNW3aVFLqAuQrVqyQJDVr1kzVq1cv0HtPmzZNW7duNfaTk5M1YMAAJSUlZXlNiRIlVKtWLWM/bZF0s3Xo0MF4c+D58+c1duzYQqkDAAAAAHB/I5RCkZY2cujkyZM6efKk1bGCcvToUb333nuSJC8vLzVr1kyS9Ndff+ntt9+2eW3r1q2N7WXLlhVYjdl5++235evrK0lavny5Dh8+XGi1AAAAAADuT4RSKNIGDRokLy8vY9/LyyvDlLX8FB8fr379+ikuLk6SNHXqVK1bt07lypWTJE2fPl3btm3L8vpXX31VTk5OkqTt27dr//79BVarLc7OznrjjTckpY7yevfddwulDgAAAADA/YtQCkWap6enhg8fLn9/f/n7+2vkyJHy9PQssPu98sorCgwMlJT6Fr6RI0eqQoUKmj9/vhwcHJScnKyhQ4fqxo0bmV5fq1Yt9e/fX1JqGNS/f3+FhIQUWL22jBw5UpUrV5YkbdmyRQcPHiyUOgAAAAAA9ydCKRR506ZN05EjR3TkyBFNnTq1wO6zceNGzZ07V1LqouULFiwwznXt2lWjRo2SJIWGhuq5557Lsp85c+aoXr16kqRz584pICDAWA/LlvRrWOUHFxcXjRkzRpJksViMdaYAAAAAADADoRRgh6tXr2r48OGyWCxycnLS4sWLVbJkSas206ZNU926dSVJmzZt0ldffZVpXx4eHvr555/VqFEjSdLly5fVv39/+fr6atCgQZo8ebLmz5+vBQsWaOrUqRo5cqT8/PzUsWNHo48KFSrky3ONGzdOpUqVkpQ6WurPP//Ml34BAAAAAMiOc2EXABQFAwcO1PXr1yWlrgvVpk2bDG3c3Ny0cuVKNWnSRHFxcXrzzTfVunVr1alTJ0PbKlWqaM+ePZowYYLmzZunyMhIBQcHKzg42GYdlStX1osvvqi33norX57L09NTo0aN0meffWaMltq8eXO+9A0AAAAAgC25Hin122+/ycHBIcc/x48ft/see/fu1ciRI1W3bl15e3vL29tbdevW1ciRI7V3797clg7kyJQpU7R9+3ZJUqNGjfTpp59m2bZ+/fr65JNPJEkxMTHq16+fEhMTM23r6uqqKVOmKDg4WDNnzlTPnj1Vo0YNlSpVSk5OTvL09FSlSpXUtGlTvfTSS/rxxx918eJFvffee3Jxccm353vjjTeMxeJ//vlnHThwIN/6BgAAAAAgKw4Wi8WSmwt/++03PfHEEzm+7tixY6pdu7bNNjExMRo3bpzVmj2Zef755zVjxoxcL2wdGRmpEiVKKCIiQt7e3jm6NiIiQsHBwXrooYfk4eGRq/sDKJpiY2N1+vRpVatWTSVKlDCOj9q1vBCrQn7wO1ursEtAHr02pHFhl4A84Pdo0cfv0aLvRPWThV0C8og/h0XbvfJdxt68JV+m77m5ualVq1Z2tU0bkZGV5ORk9ezZ02pRZ3d3d9WrV0/Ozs4KCgpSZGSkJGnBggW6dOmSfvzxRzk5OeX+AQAAAAAAAGCqfAmlKlSooC1btuRHV5o4caJVIDVixAh99tlnKl26tKTUUVSfffaZ/v3vf0tKnW70/vvvG1OmAAAAAAAAcPe7q96+d+nSJU2bNs3YHzRokObNm2cEUlLqwsyTJk3ShAkTjGPTpk1TaGioqbWmyeXsRwBFGH/uAQAAACDv7qpQasaMGbp9+7YkycPDQ//5z3+ybDtx4kT5+PhIkuLi4jR9+nQzSjQ4OzvLYrEoOTnZ1PsCKHzJycmyWCxyduYFpgAAAACQW3dVKLV27Vpju2/fvlYjpO5UrFgxPffcc8b+unXrCrS2O7m7u0tKnU4I4P6S9uc+7fcAAAAAACDn7ppQ6sSJEzp9+rSx/+STT2Z7TadOnYztU6dO6eRJ894U4ejoKHd3d0VGRiolJcW0+wIoXCkpKYqMjJS7u7scHe+aX6EAAAAAUOTcNX+jOnLkiNV+s2bNsr2mYcOGKlasWJZ9FLSyZcvq9u3bCgkJIZgC7gMpKSkKCQnR7du3VbZs2cIuBwAAAACKtHxZECU8PFx9+/bVgQMHdPXqVTk7O6ts2bJ65JFH1K5dOw0ePFje3t42+zh27JixXaxYMWO9KFvS2p05cyZDH2YoXbq0EhMTdeXKFcXFxcnb21uenp5ycnKSg4ODqbUAKBhpa8fFxMQoMjJSt2/fVsWKFW1OLwYAAAAAZC9fQqmIiAh99913Vseio6N1/vx5bdiwQRMmTNCkSZM0duzYLPsIDg42tqtWrWp3qPPAAw8YodT58+dzXnweVahQQS4uLgoLC9O1a9ckiUAKuMekvW3P3d1dPj4+BFIAAAAAkA/y7dVRvr6+qlKlilxdXRUWFqagoCAlJSVJSg2txo0bp8OHD+ubb77J9PrIyEhju0SJEnbfN/0IrKioKJtt4+PjFR8fn+k986J06dIqXbq0UlJSFBcXZzw3gHuDs7Mza0gBAAAAQD7LdSjl6Oiodu3a6bnnntOTTz6ZYeRAdHS0li5dqokTJyosLEyStGDBAtWqVUtvv/12hv7Sv8XOzc3N7jrSv/0quzfhffrpp/roo4/s7junHB0d5enpWWD9AwAAAAAA3Cty/c/+jz/+uLZt26b+/ftnOpXFy8tLo0eP1qFDh+Tr62sc//jjj3X16tUM7RMTE41tZ2f7s7L0bRMSEmy2HT9+vCIiIoyfixcv2n0fAAAAAAAA5J8Cn4vi4+OjlStXGvuxsbGZTuHz8PAwtm/fvm13/+nbZjdKydXVVd7e3lY/AAAAAAAAMJ8pC6Q0adJErVu3Nva3bduWoY2Xl5exHRcXZ3ffsbGxmfYBAAAAAACAu5dpq/amD6VOnjyZ4XzZsmWN7cuXL9vd75UrV4ztMmXK5K44AAAAAAAAmMq0UKpSpUrGdtrC5+n5+fkZ2zdu3LAaAWVL+nWhateunYcKAQAAAAAAYBbTQqn0IVP69aPS1KlTx2r/8OHD2fZ56dIlXb9+Pcs+AAAAAAAAcHcyLZQKCgoytsuXL5/hfEBAgFxdXY393bt3Z9vnrl27jG03NzcFBATksUoAAAAAAACYwZRQKi4uThs3bjT2mzdvnqGNl5eX2rZta+wvW7Ys237Tt2nbtm22b98DAAAAAADA3cGUUGrixIm6evWqsd+9e/dM2w0dOtTY/vvvv7Vp06Ys+zx06JA2b96c6bUAAAAAAAC4u+UqlNq6datef/11hYSE2GyXmJiod955R1OnTjWONWzYUF27ds20fe/evfXII48Y+6NGjdLx48cztLt8+bIGDhyo5ORkSVKDBg3Uq1ev3DwKAAAAAAAACoFzbi6KjY3Vl19+qf/85z9q0aKFWrVqpYcfflhly5ZVsWLFFBYWpgMHDmjZsmVWb8crXbq0li9fLgcHh0z7dXBw0Pz58/X4448rLi5Oly9fVpMmTfTCCy/o8ccfl7Ozsw4cOKD//ve/xsgrd3d3ff3111n2CQAAAAAAgLtPrkKpNCkpKdq1a5fVguNZqVmzplatWiU/Pz+b7Ro3bqxly5ZpwIABiouLU2RkpCZPnqzJkydnaOvu7q5ly5apcePGuX4GAAAAAAAAmC9X0/dq166t7t27q1SpUtm29fX11eeff66//vpL//rXv+zqv0ePHjp48KDatm2b6QgoBwcHtWvXTocOHVKPHj1yXD8AAAAAAAAKV65GStWuXVvr1q2TJJ05c0bHjh1TSEiIwsPDlZycLG9vb5UvX16PPvqoqlevnqvC6tSpo+3bt+vixYvau3evLl26JEmqUqWKmjdvLh8fn1z1CwAAAAAAgMKXp+l7klSjRg3VqFEjP2rJlI+Pj5555pkC6x8AAAAAAADmy9X0PQAAAAAAACAvCKUAAAAAAABgOkIpAAAAAAAAmI5QCgAAAAAAAKYjlAIAAAAAAIDpCKUAAAAAAABgOkIpAAAAAAAAmI5QCgAAAAAAAKYjlAIAAAAAAIDpCKUAAAAAAABgOkIpAAAAAAAAmI5QCgAAAAAAAKYjlAIAAAAAAIDpCKUAAAAAAABgOkIpAAAAAAAAmI5QCgAAAAAAAKYjlAIAAAAAAIDpCKUAAAAAAABgOkIpAAAAAAAAmI5QCgAAAAAAAKYjlAIAAAAAAIDpCKUAAAAAAABgOkIpAAAAAAAAmI5QCgAAAAAAAKYjlAIAAAAAAIDpCKUAAAAAAABgOkIpAAAAAAAAmI5QCgAAAAAAAKYjlAIAAAAAAIDpCKUAAAAAAABgOkIpAAAAAAAAmI5QCgAAAAAAAKYjlAIAAAAAAIDpCKUAAAAAAABgOkIpAAAAAAAAmI5QCgAAAAAAAKYjlAIAAAAAAIDpCKUAAAAAAABgOkIpAAAAAAAAmI5QCgAAAAAAAKYjlAIAAAAAAIDpCKUAAAAAAABgOufCLgBAqi8X/1nYJSCvqhd2AQAAAABQdDBSCgAAAAAAAKYjlAIAAAAAAIDpCKUAAAAAAABgOkIpAAAAAAAAmI5QCgAAAAAAAKYjlAIAAAAAAIDpCKUAAAAAAABgOkIpAAAAAAAAmI5QCgAAAAAAAKYjlAIAAAAAAIDpCKUAAAAAAABgOkIpAAAAAAAAmI5QCgAAAAAAAKYjlAIAAAAAAIDpCKUAAAAAAABgOkIpAAAAAAAAmI5QCgAAAAAAAKYjlAIAAAAAAIDpCKUAAAAAAABgOkIpAAAAAAAAmI5QCgAAAAAAAKYjlAIAAAAAAIDpCKUAAAAAAABgOkIpAAAAAAAAmI5QCgAAAAAAAKYjlAIAAAAAAIDpCKUAAAAAAABgOufCLgD5Z9Su5YVdAvLAT7UKuwQAAAAAAEzDSCkAAAAAAACYjlAKAAAAAAAApiOUAgAAAAAAgOkIpQAAAAAAAGA6QikAAAAAAACYjlAKAAAAAAAApiOUAgAAAAAAgOkIpQAAAAAAAGA6QikAAAAAAACYjlAKAAAAAAAApiOUAgAAAAAAgOkIpQAAAAAAAGA6QikAAAAAAACYjlAKAAAAAAAApiOUAgAAAAAAgOkIpQAAAAAAAGA6QikAAAAAAACYjlAKAAAAAAAApiOUAgAAAAAAgOkIpQAAAAAAAGA6QikAAAAAAACYjlAKAAAAAAAApiOUAgAAAAAAgOkIpQAAAAAAAGC6Agulzp8/Ly8vLzk4OBg/H374YY76OHr0qF577TX5+/urdOnS8vLykp+fnwYMGKAtW7YUTOEAAAAAAAAocM4F1fGoUaMUExOTq2uTkpL0/vvva/LkyUpJSbE6d/LkSZ08eVLLly/X008/rQULFqhcuXL5UTIAAAAAAABMUiAjpZYuXaqtW7fm+vpRo0bp008/NQIpFxcXPfLII2rRooXKlCljtPvhhx/Uvn37XIdfAAAAAAAAKBz5HkqFhYXp1VdflSTVqVNHlStXztH18+bN04IFC4z9rl276ty5czp8+LB2796ty5cva+bMmXJ2Th3kdeTIEY0cOTL/HgAAAAAAAAAFLt9DqVdffVVhYWGSpDlz5sjFxcXua2NjY/XBBx8Y+61bt9batWtVpUoV45iLi4vGjBmjOXPmGMdWrFihQ4cO5UP1AAAAAAAAMEO+hlJbt27V0qVLJUnPPfecHn/88Rxdv3jxYl25ckWS5ODgoK+++kpOTk6Zth02bJiaNGkiSbJYLJo8eXIeKgcAAAAAAICZ8i2Uio2N1ejRoyVJZcuW1ZQpU3Lcx5o1a4ztVq1aqXbt2jbbjxo1ytj+6aefFB8fn+N7AgAAAAAAwHz5FkpNnDhR586dkyR98cUXVguS2yM6Olo7d+409p988slsr+nUqZPV9Tt27MjRPQEAAAAAAFA48iWUOnjwoKZPny4pdR2oIUOG5LiPoKAgJSYmGvvNmjXL9pqKFSvK19fX2D9y5EiO7wsAAAAAAADz5TmUSkpK0vDhw5WcnKxixYrpq6++ylU/x44ds9qvUaOGXdelb3dnHwAAAAAAALg75TmUmjp1qg4fPixJevvtt7NdByorwcHBxrazs7MqVapk13UPPPCAsX3+/Plc3RsAAAAAAADmcs7LxWfOnNFHH30kSapZs6befffdXPcVGRlpbBcvXlyOjvblZd7e3sZ2VFSUzbbx8fFWi6GnvycAAAAAAADMk6eRUqNGjVJcXJwkafbs2XJzc8t1XzExMcZ2Tvpxd3fPtI/MfPrppypRooTx4+Pjk/NCAQAAAAAAkGe5DqUWLlyo//3vf5KkAQMGqF27dnkqJP0i587O9g/gSt82ISHBZtvx48crIiLC+Ll48WLOCwUAAAAAAECe5Wr63rVr1/TGG29IkkqVKqUvv/wyz4V4eHgY27dv37b7uvRtPT09bbZ1dXWVq6trzosDAAAAAABAvsrVSKlx48bp5s2bkqTPPvtM5cuXz3MhXl5exnbalEB7xMbGZtoHAAAAAAAA7l45DqX27dunVatWSZKaNWumESNG5EshZcuWNbajo6MVHR1t13VXrlwxtsuUKZMvtQAAAAAAAKBg5TiUunr1qrG9b98+OTo6ysHBIcuf4OBgo/1HH31kde78+fPGOT8/P6v7XLhwwa560q8LVbt27Zw+DgAAAAAAAApBnt6+l5/q1KljtX/48OFsr0lMTNQ///yTZR8AAAAAAAC4O+V4oXNXV9ccTZO7deuWUlJSJEnu7u5WC5o7OTkZ29WrV1fVqlUVEhIiSdq9e7f69+9vs++DBw9arT/1+OOP210XAAAAAAAACk+OR0p16tRJYWFhdv/4+PgY17711ltZnpOkrl27GtvfffedEhISbNaybNkyY7tevXqqUaNGTh8HAAAAAAAAheCumb4nSUOHDjW2w8LCNHfu3CzbhoSEaPHixZleCwAAAAAAgLvbXRVKPfroo1ajpd59913t2bMnQ7vIyEj1799fUVFRkqRKlSrppZdeMq1OAAAAAAAA5E2O15QqaNOnT9e+fft0/fp1RUdHq23btho2bJg6dOggLy8v/f3335o5c6bOnTsnSXJ0dNTcuXPl7u5eyJUDAAAAAADAXnddKOXr66v169erS5cuunnzpuLj4zV79mzNnj07Q1snJydNnz5dXbp0KYRKAQAAAAAAkFt31fS9NM2bN9fff/+tXr16ydk589wsICBAu3btYtoeAAAAAABAEVTgI6XOnz+fq+uqVKmi77//XtevX9fOnTsVEhKihIQEVa5cWY8++qhq1aqVv4UCAAAAAADANHfd9L07lStXTr169SrsMgAAAAAAAJCP7srpewAAAAAAALi3EUoBAAAAAADAdIRSAAAAAAAAMB2hFAAAAAAAAExHKAUAAAAAAADTEUoBAAAAAADAdIRSAAAAAAAAMB2hFAAAAAAAAExHKAUAAAAAAADTEUoBAAAAAADAdIRSAAAAAAAAMB2hFAAAAAAAAExHKAUAAAAAAADTEUoBAAAAAADAdIRSAAAAAAAAMB2hFAAAAAAAAExHKAUAAAAAAADTEUoBAAAAAADAdIRSAAAAAAAAMB2hFAAAAAAAAExHKAUAAAAAAADTEUoBAAAAAADAdIRSAAAAAAAAMB2hFAAAAAAAAExHKAUAAAAAAADTEUoBAAAAAADAdIRSAAAAAAAAMB2hFAAAAAAAAExHKAUAAAAAAADTEUoBAAAAAADAdIRSAAAAAAAAMB2hFAAAAAAAAExHKAUAAAAAAADTEUoBAAAAAADAdIRSAAAAAAAAMB2hFAAAAAAAAExHKAUAAAAAAADTEUoBAAAAAADAdIRSAAAAAAAAMB2hFAAAAAAAAExHKAUAAAAAAADTEUoBAAAAAADAdIRSAAAAAAAAMB2hFAAAAAAAAEznXNgFFCaLxSJJioyMLORK8kdCTGxhl4A8uB0XXdglII/4M1j08eew6LtX/j/9fsXv0aKP36NFH38Oiz7+HBZt98p3mbTnSMtdsuJgya7FPSwkJEQ+Pj6FXQYAAAAAAMA95+LFi6patWqW5+/rUColJUWhoaEqXry4HBwcCrscAEVYZGSkfHx8dPHiRXl7exd2OQAAADnG9xkA+cVisSgqKkqVK1eWo2PWK0fd19P3HB0dbSZ2AJBT3t7efIkDAABFGt9nAOSHEiVKZNuGhc4BAAAAAABgOkIpAAAAAAAAmI5QCgDygaurqz744AO5uroWdikAAAC5wvcZAGa7rxc6BwAAAAAAQOFgpBQAAAAAAABMRygFAAAAAAAA0xFKAQAAAAAAwHSEUgAAAAAAADAdoRSA+1br1q3l4OBg/AAAACBzfG8CUBCcC7sAAPeWuLg4HTp0SKdOndKtW7cUExMjd3d3eXt764EHHlCNGjVUvXp1OTqSiQMAgPvb/fy9KTo6Wn/++afOnDmjGzduKCEhQR4eHqpSpYpq166t+vXr35PPDcAaoRSAPLNYLNq4caPmzJmj7du3KykpyWb74sWLq1GjRmrVqpU6deqkRx99tEh86ciPfxWcNm2aXnnllbwXAwAAiqT75XtTZpKSkrRixQotXLhQu3btsvnspUqVUo8ePfTSSy+pYcOGJlYJwEwOFovFUthFACi6goOD9fzzz+uXX37JdR/r169Xt27d8rEq+7Ru3Vo7duww9rP7dUgoZS2nnx8AAPe7++l70522b9+uF154QadPn87xvfv27asZM2aoQoUKOb4WwN2NkVIAcu3s2bN6/PHHdenSpQznihUrpgcffFAlSpRQfHy8bt68qUuXLiklJSVDW8IMAABwr7ufvzd98MEHmjRpUqa1e3p6qnLlyipRooSuXLmiK1euZBhBtXr1au3cuVPr169XkyZNzCobgAkIpQDkSmJiorp06WL1xcrBwUEDBgzQqFGj1LRpUzk7W/+KiY6O1sGDB7V582Z9//33OnPmjNllW/ntt9/ydP3SpUtz/C92fn5+ebonAAAoeu7n700vv/yyZsyYYXXMxcVFI0eO1IABA9S0aVOr0egxMTH64YcfNGfOHKt7XrlyRe3atdPWrVvVrFmzXNUC4O7D9D0AuTJz5kyNGzfO2Hdzc9OaNWv01FNP2d3Hjh07NG3aNA0bNkxdunQpiDLz1Z3T986dOydfX9/CKeYuwPQ9AADscz9+b5KkhQsX6vnnn7c6Vr9+fS1dulT+/v52Xf/SSy8pLi7OOFaxYkUdPXpUZcuWzfd6AZiPUApArjRu3FgHDx409j/99FO98847hVhRwSOUskYoBQCAfe7H701nzpxRgwYNFB0dbRx7+OGH9dtvv6lMmTJ297N582Z1795dCQkJxrFu3bpp/fr1+VkugEJSNF/bAKBQ3bx50+qLlaOjo0aMGFGIFQEAANyd7tfvTWPGjLEKpLy8vPTTTz/lKJCSpE6dOmnq1KlWxzZs2KC1a9fmS50AChdrSgHIsTsX6CxbtmyOv2DkVUJCgn7//XcFBwfr+vXrio2NVfHixVWtWjU9/PDDqlGjhqn15Idz584pMDBQFy5cUEREhJydnVW6dGlVq1ZNTZs2lZeXV77d6178/AAAuBvdj9+bgoKCtGXLFqtjH3/8sXx8fHLV34svvqhFixZZhXtTp05Vz54981QngMJHKAUgx6Kioqz2k5OTTbv3vn379Omnn+p///ufYmNjs2zn6+ur3r17a/To0Vl+0Srs6We3b9/Wjz/+qLVr1+qXX37RlStXsmzr5OSktm3bavz48WrdunWu75nXz+/DDz/URx99lOk1d05vTK9Vq1Z5XlgeAICi6H783jR9+nSr/fLly2vMmDE5rP7/5+joqA8++EBdu3Y1ju3du1d//PGHHn300Vz3C6DwMX0PQI6VLFnSav/GjRs6ffp0gd4zKipKvXv3VvPmzbVp0yabX6wk6fz58/riiy80bNiwAq0rL1q2bKnevXtr+fLlNgMpKfUL7NatW/XEE09ozJgxGV6VnJ178fMDAKAouB+/N905tW7gwIFycXHJU5+dOnXK8Nbj77//Pk99Aih8hFIAcqx69epyc3OzOvb2228X2EijkJAQNWvWTGvWrMlwrnjx4qpdu7YCAgLk5+eXoa672e3btzMcq1y5sh5++GE1bdpU9evXV4kSJTK0mTVrlkaOHGn3fe7Vzw8AgKLgfvvedPz4cYWFhVkd69WrV577dXZ2thopJUl79uzJc78AChfT9wDkmJubm9q2basff/zROLZ27Vq1bdtWkyZNUosWLfLtXgkJCerVq5cCAwOtjnfv3l2vv/66mjVrJicnJ+N4UlKSDh8+rPXr12vJkiX5VkdBeeCBB9SnTx899dRTaty4sby9va3OWywWHT16VHPmzNG8efOMIf8LFy5Uly5d1KNHD5v95/fnN3jwYLVs2VKS9Prrr+vvv/82zm3bti3LOkqVKmWzTgAA7lX32/emO4MiJycnNWjQIF/6btSokb7++mtj/88//1RCQoKKFSuWL/0DMJ+DhXd4A8iF3bt367HHHsv0XLVq1dShQwc1a9ZMAQEBqlOnjhwdczcwc/z48frss8+M/WLFimnRokXq169fttcmJiZq//79WdaZ0zWl7lwzaenSpRmGkdvSsmVLq3+R3LVrl5o3b2715dCWbdu2qUuXLoqPj5ckBQQE6Pfff7d5zd30+QEAcL+6n743vfnmm/riiy+M/Tp16igoKMjeR7DpwIEDatKkidWxoKAg1alTJ1/6B2A+QikAufb+++9r0qRJ2bbz8vJSQECAWrdurU6dOqlx48Z29X/z5k098MADiomJMY7Nmzcv316jnNdQKqfOnTsnX1/fPPUxceJE/fvf/zb2AwMDVbdu3Uzb3m2fHwAA97P75XvTsGHDtGDBAmO/TZs2+t///pcvNVy4cEHVqlWzOrZ79+58HW0GwFysKQUg1z7++GNNnz492/UIoqOj9csvv+j999/Xo48+qocfflgLFixQSkqKzevmz59v9cXqsccey7cvVkXVwIEDrfb37t2bZVs+PwAA7h73y/emmzdvWu1ntj5mbmXW161bt/KtfwDmI5QCkCfjxo3TqVOn9OKLL9r9pSMwMFDDhg1TQECAgoODs2y3devWDPe63z344INW+3/99VeWbfn8AAC4u9wP35uioqKs9j09PfOt78z6ioiIyLf+AZiPUApAnlWtWlWzZs3S1atXtXHjRr366qtq3LhxtotOHjx4UAEBATpz5kyGc0lJSdq/f7+x7+joqCeffDLfa8+Lc+fOyWKx2P1ja+regQMH9O677+qpp55S9erVVbp0abm4uMjBwcHqx9XV1eq6O99uk6YofH4AANyP7vXvTcWLF7faTz96K68y64s3BwNFG2/fA5BvXF1d1aVLF3Xp0kVS6htgjh49qj179mj79u36+eeflZCQYHXNtWvX1KtXLx08eNBqwe8rV65YffHw8/OTl5eXOQ9iol27dmnMmDFWb7HLifDw8EyP3y+fHwAARdW9+r2pdOnSVvv5OZIps77ufHMxgKKFkVIACkyxYsXUqFEjjRs3Ths3blRoaKjefPPNDG+bO3LkiFauXGl17M71CMqXL1/g9Zpt7ty5atWqVa4DKUnGm/judD98fgAA3Evule9Nd4ZSV65cybe+M+srJ29CBnD3YaQUANOUKVNGn3/+uR5//HF1795dycnJxrmlS5dqwIABxv6d6xHca6N8fv31V73wwgtWb65xdnZWy5Yt1aRJE1WrVk3ly5eXm5tbhil77du3z7b/e/3zAwDgXldUvzfVrl3bav/EiROKjY2Vh4dHnvu+cy1NJycn+fn55blfAIWHUAqA6Z5++mkNGTLE6nXBu3fvtmpz53oE0dHRptRmltdff90qkOrcubPmzJmjqlWr2rwuq5FRd7rXPz8AAO4XRe17U4sWLaz2k5OTdfjwYTVv3jzPfR88eNBqv1atWhn+8Q5A0cL0PQCFom/fvlb70dHRVusE3Dn0+9q1a6bUZYaTJ09a/Uvfww8/rLVr12YbSEkZh+dn5V7+/AAAuN8Upe9Nfn5+Klu2rNWxNWvW5LnfpKQkbdy40eoYL3EBij5CKQCFIrM30cXGxhrbFStWtBp6fuLEiXtmtE/6t+NI0vDhw7N9406awMBAu9rdy58fAAD3m6L0vcnBwUE9e/a0OrZ06VIlJSXlqd/Nmzfr6tWrVse6deuWpz4BFD5CKQCFIrNX+pYpU8bYdnZ2VrNmzYz9lJQUbdmyxZTaCtqdX6hyshbCL7/8Ylc7Mz4/R0fr/wtJPx0RAADkn6L2venll1+22r927Zr++9//5rq/lJQUffzxx1bH6tWrp8cffzzXfQK4OxBKASgUf/zxh9V+xYoVM4wWunNI9owZMwq8LjPcGd7c+brnrMTHx1utJ5Gdgv78PD09rfbT/4stAADIP0Xte1PdunXVsWNHq2Pvv/++Ll68mKv+Zs+erT///NPq2Ntvvy0HB4dc1wjg7kAoBSDHwsLCtHz5cqWkpOTq+oSEBM2cOdPq2J1fXCTp+eeft1q4c9euXfr6669zdc+7ScWKFa3271ysNCsTJ07MMMrKloL+/O5cv+LcuXP51jcAAPeK+/V708yZM63+ASsqKkpPPfWU3etjptmyZYtef/11q2OPPfaYBg4cmC91AihchFIAciw6OloDBgxQ/fr1tXTpUsXFxdl97e3btzVw4MAMayMNHjw4Q9uSJUtmGP49ZswYrVy50q57JSYmateuXXbXZpY73z4zZ84cnT592uY1c+fO1RdffJGj+xT051evXj2r/e+//z5H9QEAcD+4X7831axZU9OnT7c69s8//+iJJ57Q0aNH7epj0aJF6tmzp9Wocm9vb33zzTeMkgLuFRYAyKFz585ZJBk/3t7eluHDh1tWr15tCQ0NzfSaS5cuWWbNmmXx9fW1ulaSpWfPnlneKyEhwdK0adNMr9m9e7clKSnJqn1iYqLlzz//tLz33nsWHx8fS6tWrbLsu1WrVlZ9ZufOGs6dO5ftNVlp1qyZVV+VKlWyrF692pKYmGjV7vDhw5a+ffsa7erUqWN1na3ns1gK9vM7dOiQVZ+Ojo6WoUOHWr799lvL5s2bLdu2bTN+/vzzz9x+VAAAFGn36/emNC+++GKGelxcXCwvvfSSZd++fRnax8TEWFatWmVp3bp1huvc3NwsO3bssPveAO5+DhYLK9MCyJnz58/rwQcfzPJ8mTJlVLZsWZUsWVK3b9/W5cuXs3w1cZMmTbR161Z5e3tn2d+lS5fUsWPHTN88V7x4cfn4+Kh48eKKiIjQ+fPndfv2beN8q1at9Ntvv2Xab+vWrbVjxw5jP7tfh3f+i9y5c+cyfRuOPfbu3avWrVsrMTHR6riXl5dq1qwpR0dHhYSEWE3X8/T01K5du9SwYUPjmK3nS1NQn58ktW3b1q7F1+2pEwCAe9H9+r0pfbsJEybo//7v/zI9X7x4cVWuXFnFixfXtWvXdPny5QzfjySpQoUKWrVqlVq1amXXfQEUDUzfA5BjXl5eeuSRR7I8f+PGDZ04cUK///67jhw5kukXK0dHR40ePVrbtm2z+cVKkqpUqaI9e/aoa9euGc5FRUUpKChIv//+u44fP271xepu1rx5c3399ddycXGxOh4dHa2//vpLBw8etAqkSpUqpR9++EH/+te/cnyvgvz8lixZYhWSAQAAa/f79yYHBwd98skn2rJli6pXr55pTSdOnNCff/6pCxcuZBpIdevWTYcOHSKQAu5BhFIAcqxs2bI6fPiwzpw5oy+//FJdunRRqVKl7Lq2UqVKevnll3X48GF99dVXVgty2lKiRAlt2LBBv/76qzp06JDhjTN38vPz04QJE/Ttt9/a1X9hGDJkiHbu3GnzdcZubm56/vnnFRgYqNatW+f6XgX1+VWuXFn79+/Xd999p/79+6tevXoqWbKknJ2dc10rAAD3Er43perYsaOOHz+uRYsWqVWrVnJycrLrujlz5mj9+vWqXLlygdUGoPAwfQ9AvrBYLAoODtbJkyd14cIFRUREKC4uTh4eHsaw7EceeURVq1bNl/vFxMRoz549CgkJUVhYmJKTk+Xt7a0HH3xQ/v7++XYfs5w/f1579uzR5cuXFR8fr5IlS8rPz0/NmzeXh4dHvt/vXvv8AAAoSvjeJEVGRurgwYM6ffq0bt68qdu3b8tisWjx4sU6f/680a5q1arat28f302AexShFAAAAADgrnDixAm1aNFCN27cMI7Vq1dPu3fvVsmSJQuvMAAFgul7AAAAAIC7gp+fn3788UerkeKBgYHq2rVrkVk7FID9CKUAAAAAAHeNJk2aaPXq1VZrVO7atUsDBgxQSkpKIVYGIL8RSgEAAAAA7iqdO3fWnDlzrI6tXbtWY8eOLaSKABQEXo8EAAAAALjrDBs2TO7u7jp58qTV8XPnzunBBx8spKoA5CcWOgcAAAAAAIDpmL4HAAAAAAAA0xFKAQAAAAAAwHSEUgAAAAAAADAdoRQAAAAAAABMRygFAAAAAAAA0xFKAQAAAAAAwHSEUgAAAAAAADAdoRQAAAAAAABMRygFAAAAAAAA0xFKAQAAAAAAwHSEUgAAAAAAADDd/wckI9iBCscFWwAAAABJRU5ErkJggg==",
      "text/plain": [
       "<Figure size 1200x450 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "from matplotlib.patches import Patch\n",
    "import matplotlib.cm as cm\n",
    "\n",
    "# Data\n",
    "data = {\n",
    "    'SciFact': {\n",
    "        'EM@50': {'query-prop': 43.0, 'MixGR': 45.3},\n",
    "        'EM@200': {'query-prop': 60.5, 'MixGR': 62.2}\n",
    "    },\n",
    "    'SciQ': {\n",
    "        'EM@50': {'query-prop': 56.2, 'MixGR': 59.0},\n",
    "        'EM@200': {'query-prop': 60.9, 'MixGR': 63.3}\n",
    "    },\n",
    "}\n",
    "\n",
    "# Convert data to a DataFrame\n",
    "df = pd.DataFrame([\n",
    "    {\"Dataset\": dataset, \"Metric\": metric, \"Method\": method, \"Score\": score}\n",
    "    for dataset, metrics in data.items()\n",
    "    for metric, methods in metrics.items()\n",
    "    for method, score in methods.items()\n",
    "])\n",
    "\n",
    "# Define colors for methods using viridis palette\n",
    "method_colors = {\n",
    "    'query-prop': '#66c2a5',  # Teal\n",
    "    'MixGR': '#8da0cb'  # Green\n",
    "}\n",
    "\n",
    "# Define hatch patterns for metrics\n",
    "metric_hatches = {\n",
    "    'EM@50': '',\n",
    "    'EM@200': '.'\n",
    "}\n",
    "\n",
    "# Plotting with Matplotlib\n",
    "fig, ax = plt.subplots(figsize=(12, 4.5))\n",
    "\n",
    "# Unique datasets, methods, and metrics\n",
    "datasets = df['Dataset'].unique()\n",
    "methods = df['Method'].unique()\n",
    "metrics = df['Metric'].unique()\n",
    "\n",
    "# Width of a single bar\n",
    "bar_width = 0.2\n",
    "\n",
    "# Calculate the positions for each bar\n",
    "bar_positions = []\n",
    "for i, dataset in enumerate(datasets):\n",
    "    for j, metric in enumerate(metrics):\n",
    "        for k, method in enumerate(methods):\n",
    "            bar_positions.append(i + j * (len(methods) * bar_width) + k * bar_width + 0.1)\n",
    "\n",
    "# Plotting the bars\n",
    "bars = []\n",
    "for i, (dataset, metric, method, score) in enumerate(zip(df['Dataset'], df['Metric'], df['Method'], df['Score'])):\n",
    "    offset = bar_positions[i]\n",
    "    bar = ax.bar(offset, score, color=method_colors[method], width=bar_width, edgecolor='none', hatch=metric_hatches[metric])\n",
    "    bars.append(bar)\n",
    "\n",
    "# Adding labels and title\n",
    "ax.set_xlabel('', fontsize=14)\n",
    "ax.set_ylabel('', fontsize=14)\n",
    "ax.tick_params(axis='y', labelsize=25)\n",
    "ax.set_title('', fontsize=16)\n",
    "ax.set_xticks([r + (len(metrics) * len(methods) * bar_width) / 2 for r in range(len(datasets))])\n",
    "ax.set_xticklabels(datasets, fontsize=30)\n",
    "ax.set_ylim(35, 65)\n",
    "\n",
    "# Custom legend for methods\n",
    "method_patches = [Patch(facecolor=method_colors[method], edgecolor='none', label=method) for method in methods]\n",
    "\n",
    "# Custom legend for metrics\n",
    "hatch_patches = [Patch(facecolor='white', edgecolor='grey', hatch=metric_hatches[metric], label=metric) for metric in metrics]\n",
    "\n",
    "# Combine legends\n",
    "combined_legend = ax.legend(handles=hatch_patches + method_patches, title='', fontsize=21.6, loc=2, borderpad=0.1)\n",
    "ax.add_artist(combined_legend)\n",
    "\n",
    "# Show the plot\n",
    "plt.tight_layout()\n",
    "plt.savefig('extention_to_proposition.pdf')\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "865ce8a6-8d2c-430f-8347-18542bc033d8",
   "metadata": {},
   "source": [
    "### Section 6.2: BM25"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "8df8eaa9-6592-4146-8fc9-9422b0e1b7f3",
   "metadata": {},
   "outputs": [],
   "source": [
    "{\n",
    "    'SciFact': {\n",
    "        'Original': 68.5,\n",
    "        'MixGR': 71.0,\n",
    "        'BM25': 69.2,\n",
    "    },\n",
    "    'SciDocs':  {\n",
    "        'Original': 18.5,\n",
    "        'MixGR': 20.7,\n",
    "        'BM25': 19.9,\n",
    "    },\n",
    "    'NFCorpus': {\n",
    "        'Original': 34.9,\n",
    "        'MixGR': 37.1,\n",
    "        'BM25': 30.65,\n",
    "    },\n",
    "    'SciQ':  {\n",
    "        'Original': 70.0,\n",
    "        'MixGR': 86.7,\n",
    "        'BM25': 92.2,\n",
    "    }\n",
    "}\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 508,
   "id": "081c47b8-5741-4f78-b638-2d30f380d763",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/scratch/92771/ipykernel_558910/3423460148.py:39: UserWarning: The palette list has more values (4) than needed (3), which may not be intended.\n",
      "  sns.barplot(x='Dataset', y='Score', hue='Method', data=df, palette=['#3c5488', '#4dbbd5', '#00a087', '#e69f00'])\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1000x400 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "import pandas as pd\n",
    "\n",
    "# Data from the dictionary\n",
    "data = {\n",
    "    'SciFact': {\n",
    "        'w/o MixGR': 68.5,\n",
    "        'w/ MixGR': 71.0,\n",
    "        'BM25': 69.2,\n",
    "    },\n",
    "    'SciDocs':  {\n",
    "        'w/o MixGR': 18.5,\n",
    "        'w/ MixGR': 20.7,\n",
    "        'BM25': 19.9,\n",
    "    },\n",
    "    'NFCorpus': {\n",
    "        'w/o MixGR': 34.9,\n",
    "        'w/ MixGR': 37.1,\n",
    "        'BM25': 30.65,\n",
    "    },\n",
    "    'SciQ':  {\n",
    "        'w/o MixGR': 70.0,\n",
    "        'w/ MixGR': 86.7,\n",
    "        'BM25': 92.2,\n",
    "    }\n",
    "}\n",
    "\n",
    "# Convert the data into a DataFrame for easier plotting with Seaborn\n",
    "data_list = []\n",
    "for dataset, scores in data.items():\n",
    "    for method, score in scores.items():\n",
    "        data_list.append([dataset, method, score])\n",
    "\n",
    "df = pd.DataFrame(data_list, columns=['Dataset', 'Method', 'Score'])\n",
    "\n",
    "# Plotting the data\n",
    "plt.figure(figsize=(10, 4))\n",
    "sns.barplot(x='Dataset', y='Score', hue='Method', data=df, palette=['#3c5488', '#4dbbd5', '#00a087', '#e69f00'])\n",
    "\n",
    "# Set labels and title\n",
    "plt.ylabel('NDCG@20(%)')\n",
    "\n",
    "plt.xlabel('')\n",
    "\n",
    "\n",
    "# Adjust the legend\n",
    "plt.legend(fontsize=20, title_fontsize='13')\n",
    "\n",
    "plt.savefig(\"contriever_bm25.pdf\", format='pdf')\n",
    "\n",
    "# Show the plot\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0a247910",
   "metadata": {},
   "source": [
    "### Section 6.3: Ablation study"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "id": "5d219465",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'NDCG@5': 0.5017183333333333,\n",
       " 'NDCG@20': 0.5463049999999999,\n",
       " 'Recall@5': 0.5863750000000001,\n",
       " 'Recall@20': 0.7303683333333333}"
      ]
     },
     "execution_count": 45,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "scifact_ablation[0]['avg']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "id": "bb4d5ac2",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'nfcorpus': {'-conv': 0.28202,\n",
       "  '-prop': 0.28897833333333334,\n",
       "  '-whole': 0.28602,\n",
       "  'mixgr': 0.29142333333333337},\n",
       " 'scidocs': {'-conv': 0.15744666666666665,\n",
       "  '-prop': 0.15874,\n",
       "  '-whole': 0.15528666666666666,\n",
       "  'mixgr': 0.16227166666666668},\n",
       " 'scifact': {'-conv': 0.5463049999999999,\n",
       "  '-prop': 0.5552,\n",
       "  '-whole': 0.5483033333333334,\n",
       "  'mixgr': 0.55958},\n",
       " 'sciq': {'-conv': 0.7910150000000001,\n",
       "  '-prop': 0.812065,\n",
       "  '-whole': 0.807265,\n",
       "  'mixgr': 0.848585}}"
      ]
     },
     "execution_count": 51,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# dict_ = {dataset: {'-conv': ...}, ...}\n",
    "ablation_dict = {}\n",
    "for dataset, ablations in {'nfcorpus': nfcorpus_ablation, 'scidocs': scidocs_ablation, 'scifact': scifact_ablation, 'sciq': sciq_ablation}.items():\n",
    "    ablation_dict[dataset] = {}\n",
    "    for type_idx, ablation_type in enumerate(['-conv', '-prop', '-whole', 'mixgr']):\n",
    "        ablation = ablations[type_idx]\n",
    "        ablation_dict[dataset][ablation_type] = ablation['avg']['NDCG@20']\n",
    "        \n",
    "ablation_dict        "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 309,
   "id": "f14f54a8",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/scratch/92771/ipykernel_558910/519703109.py:34: FutureWarning: \n",
      "\n",
      "Passing `palette` without assigning `hue` is deprecated and will be removed in v0.14.0. Assign the `x` variable to `hue` and set `legend=False` for the same effect.\n",
      "\n",
      "  barplot = sns.barplot(data=subset, x='Measurement', y='Score', ax=ax, palette='viridis', width=0.8)\n",
      "/scratch/92771/ipykernel_558910/519703109.py:34: FutureWarning: \n",
      "\n",
      "Passing `palette` without assigning `hue` is deprecated and will be removed in v0.14.0. Assign the `x` variable to `hue` and set `legend=False` for the same effect.\n",
      "\n",
      "  barplot = sns.barplot(data=subset, x='Measurement', y='Score', ax=ax, palette='viridis', width=0.8)\n",
      "/scratch/92771/ipykernel_558910/519703109.py:34: FutureWarning: \n",
      "\n",
      "Passing `palette` without assigning `hue` is deprecated and will be removed in v0.14.0. Assign the `x` variable to `hue` and set `legend=False` for the same effect.\n",
      "\n",
      "  barplot = sns.barplot(data=subset, x='Measurement', y='Score', ax=ax, palette='viridis', width=0.8)\n",
      "/scratch/92771/ipykernel_558910/519703109.py:34: FutureWarning: \n",
      "\n",
      "Passing `palette` without assigning `hue` is deprecated and will be removed in v0.14.0. Assign the `x` variable to `hue` and set `legend=False` for the same effect.\n",
      "\n",
      "  barplot = sns.barplot(data=subset, x='Measurement', y='Score', ax=ax, palette='viridis', width=0.8)\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1200x1000 with 4 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "import pandas as pd\n",
    "\n",
    "# Data conversion to DataFrame\n",
    "data = {\n",
    "    'Dataset': [],\n",
    "    'Measurement': [],\n",
    "    'Score': []\n",
    "}\n",
    "\n",
    "datasets = {\n",
    "    'NFCorpus': {'w/o s-p': 0.28202, 'w/o q-p': 0.288978, 'w/o q-d': 0.28602, 'MixGR': 0.291423},\n",
    "    'SciDocs': {'w/o s-p': 0.157447, 'w/o q-p': 0.15874, 'w/o q-d': 0.155287, 'MixGR': 0.162272},\n",
    "    'SciFact': {'w/o s-p': 0.546305, 'w/o q-p': 0.5552, 'w/o q-d': 0.548303, 'MixGR': 0.55958},\n",
    "    'SciQ': {'w/o s-p': 0.791015, 'w/o q-p': 0.812065, 'w/o q-d': 0.807265, 'MixGR': 0.848585}\n",
    "}\n",
    "\n",
    "for dataset, measurements in datasets.items():\n",
    "    for measurement, score in measurements.items():\n",
    "        data['Dataset'].append(dataset)\n",
    "        data['Measurement'].append(measurement)\n",
    "        data['Score'].append(score)\n",
    "\n",
    "df = pd.DataFrame(data)\n",
    "\n",
    "# Create the figure and axes\n",
    "fig, axes = plt.subplots(2, 2, figsize=(12, 10))\n",
    "axes = axes.flatten()\n",
    "\n",
    "# Plot each dataset in its respective subplot\n",
    "for i, (dataset, ax) in enumerate(zip(datasets, axes)):\n",
    "    subset = df[df['Dataset'] == dataset]\n",
    "    barplot = sns.barplot(data=subset, x='Measurement', y='Score', ax=ax, palette='viridis', width=0.8)\n",
    "    ax.set_title(dataset, fontsize=25)\n",
    "    ax.set_ylim(0.99 * subset['Score'].min(), 1.01 * subset['Score'].max())\n",
    "    ax.set_xlabel('')  # Remove x-axis label\n",
    "    ax.set_ylabel('')  # Remove y-axis label\n",
    "    ax.tick_params(axis='x', length=0, labelsize=20)  # Remove x-axis ticks\n",
    "    ax.tick_params(axis='y', labelsize=20) \n",
    "#     ax.tick_params(labelbottom=False)  # Hide x-axis tick labels\n",
    "\n",
    "# Place a legend at the bottom of the figure\n",
    "# handles, labels = axes[0].get_legend_handles_labels()  # Get legend from the first plot\n",
    "# fig.legend(handles, labels, loc='lower center', bbox_to_anchor=(0.5, 0.0), ncol=4, fontsize=100)\n",
    "\n",
    "# Adjust layout to make space for the legend\n",
    "plt.tight_layout(pad=3.0)\n",
    "fig.subplots_adjust(bottom=0.2)  # Increase the bottom margin to make space for the larger legend\n",
    "plt.savefig('./ablation.pdf', bbox_inches='tight')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "8e1c73c8-26a0-458c-8683-3e838d873232",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.9.18"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}