notebooks/6_recommendation_engine.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "b3aae6be",
   "metadata": {},
   "outputs": [],
   "source": [
    "# https://www.kaggle.com/datasets/hernan4444/anime-recommendation-database-2020?select=rating_complete.csv"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "4fbfe3b6",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "2bfa68af",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'\\nChapter 6: Building a Recommendation Engine \\n    Overview of Siamese BERT Architectures \\n    Fine-Tuning BERT for Classifying + Tagging Items\\n    Fine-Tuning Siamese BERT for Recommendations \\n'"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "'''\n",
    "Chapter 6: Building a Recommendation Engine \n",
    "    Overview of Siamese BERT Architectures \n",
    "    Fine-Tuning Siamese BERT for Recommendations \n",
    "'''"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "5e067f76",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(16206, 19)\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>MAL_ID</th>\n",
       "      <th>Name</th>\n",
       "      <th>Score</th>\n",
       "      <th>Genres</th>\n",
       "      <th>English name</th>\n",
       "      <th>Type</th>\n",
       "      <th>Episodes</th>\n",
       "      <th>Premiered</th>\n",
       "      <th>Producers</th>\n",
       "      <th>Studios</th>\n",
       "      <th>Source</th>\n",
       "      <th>Duration</th>\n",
       "      <th>Rating</th>\n",
       "      <th>Ranked</th>\n",
       "      <th>Members</th>\n",
       "      <th>Favorites</th>\n",
       "      <th>Completed</th>\n",
       "      <th>On-Hold</th>\n",
       "      <th>synopsis</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>Cowboy Bebop</td>\n",
       "      <td>8.78</td>\n",
       "      <td>Action, Adventure, Comedy, Drama, Sci-Fi, Space</td>\n",
       "      <td>Cowboy Bebop</td>\n",
       "      <td>TV</td>\n",
       "      <td>26</td>\n",
       "      <td>Spring 1998</td>\n",
       "      <td>Bandai Visual</td>\n",
       "      <td>Sunrise</td>\n",
       "      <td>Original</td>\n",
       "      <td>24 min. per ep.</td>\n",
       "      <td>R - 17+ (violence &amp; profanity)</td>\n",
       "      <td>28.0</td>\n",
       "      <td>1251960</td>\n",
       "      <td>61971</td>\n",
       "      <td>718161</td>\n",
       "      <td>71513</td>\n",
       "      <td>In the year 2071, humanity has colonized sever...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>5</td>\n",
       "      <td>Cowboy Bebop: Tengoku no Tobira</td>\n",
       "      <td>8.39</td>\n",
       "      <td>Action, Drama, Mystery, Sci-Fi, Space</td>\n",
       "      <td>Cowboy Bebop:The Movie</td>\n",
       "      <td>Movie</td>\n",
       "      <td>1</td>\n",
       "      <td>Unknown</td>\n",
       "      <td>Sunrise, Bandai Visual</td>\n",
       "      <td>Bones</td>\n",
       "      <td>Original</td>\n",
       "      <td>1 hr. 55 min.</td>\n",
       "      <td>R - 17+ (violence &amp; profanity)</td>\n",
       "      <td>159.0</td>\n",
       "      <td>273145</td>\n",
       "      <td>1174</td>\n",
       "      <td>208333</td>\n",
       "      <td>1935</td>\n",
       "      <td>other day, another bounty—such is the life of ...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   MAL_ID                             Name Score  \\\n",
       "0       1                     Cowboy Bebop  8.78   \n",
       "1       5  Cowboy Bebop: Tengoku no Tobira  8.39   \n",
       "\n",
       "                                            Genres            English name  \\\n",
       "0  Action, Adventure, Comedy, Drama, Sci-Fi, Space            Cowboy Bebop   \n",
       "1            Action, Drama, Mystery, Sci-Fi, Space  Cowboy Bebop:The Movie   \n",
       "\n",
       "    Type Episodes    Premiered               Producers  Studios    Source  \\\n",
       "0     TV       26  Spring 1998           Bandai Visual  Sunrise  Original   \n",
       "1  Movie        1      Unknown  Sunrise, Bandai Visual    Bones  Original   \n",
       "\n",
       "          Duration                          Rating Ranked  Members  Favorites  \\\n",
       "0  24 min. per ep.  R - 17+ (violence & profanity)   28.0  1251960      61971   \n",
       "1    1 hr. 55 min.  R - 17+ (violence & profanity)  159.0   273145       1174   \n",
       "\n",
       "   Completed  On-Hold                                           synopsis  \n",
       "0     718161    71513  In the year 2071, humanity has colonized sever...  \n",
       "1     208333     1935  other day, another bounty—such is the life of ...  "
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pre_merged_anime = pd.read_csv('../data/pre_merged_anime.csv').dropna()\n",
    "print(pre_merged_anime.shape)\n",
    "pre_merged_anime.head(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "86f7a7a5",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "fa6ef8ce",
   "metadata": {},
   "outputs": [],
   "source": [
    "import re\n",
    "import string\n",
    "\n",
    "def clean_text(text):\n",
    "    # Remove non-printable characters\n",
    "    text = ''.join(filter(lambda x: x in string.printable, text))\n",
    "    \n",
    "    # Replace multiple whitespace characters with a single space\n",
    "    text = re.sub(r'\\s{2,}', ' ', text).strip()\n",
    "\n",
    "    return text.strip()\n",
    "\n",
    "\n",
    "def get_anime_description(anime_row):\n",
    "    \"\"\"\n",
    "    Generates a custom description for an anime title based on various features from the input data.\n",
    "\n",
    "    :param anime_row: A row from the MyAnimeList dataset containing relevant anime information.\n",
    "    :return: A formatted string containing a custom description of the anime.\n",
    "    \"\"\"\n",
    "\n",
    "    # Determine the type of anime (TV show or other)\n",
    "    anime_type = \"TV Show\" if anime_row[\"Type\"] == \"TV\" else anime_row[\"Type\"]\n",
    "\n",
    "    # Create a formatted description using the relevant features from the input data\n",
    "    description = (\n",
    "        f\"{anime_row['Name']} is a {anime_type}.\\n\"\n",
    "        f\"Synopsis: {anime_row['synopsis']}\\n\"\n",
    "        f\"It was produced by {anime_row['Producers']} and it is from {anime_row['Studios']} Studio.\\n\"\n",
    "        f\"Its source is {anime_row['Source']}.\\n\"\n",
    "        f\"It premiered in {anime_row['Premiered']}.\\n\"\n",
    "        f\"Its genres are {anime_row['Genres']}\\n\"\n",
    "    )\n",
    "\n",
    "    return clean_text(description)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "bfbd1501",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Cowboy Bebop is a TV Show.\n",
      "Synopsis: In the year 2071, humanity has colonized several of the planets and moons of the solar system leaving the now uninhabitable surface of planet Earth behind. The Inter Solar System Police attempts to keep peace in the galaxy, aided in part by outlaw bounty hunters, referred to as \"Cowboys.\" The ragtag team aboard the spaceship Bebop are two such individuals. Mellow and carefree Spike Spiegel is balanced by his boisterous, pragmatic partner Jet Black as the pair makes a living chasing bounties and collecting rewards. Thrown off course by the addition of new members that they meet in their travelsEin, a genetically engineered, highly intelligent Welsh Corgi; femme fatale Faye Valentine, an enigmatic trickster with memory loss; and the strange computer whiz kid Edward Wongthe crew embarks on thrilling adventures that unravel each member's dark and mysterious past little by little. Well-balanced with high density action and light-hearted comedy, Cowboy Bebop is a space Western classic and an homage to the smooth and improvised music it is named after.\n",
      "It was produced by Bandai Visual and it is from Sunrise Studio.\n",
      "Its source is Original.\n",
      "It premiered in Spring 1998.\n",
      "Its genres are Action, Adventure, Comedy, Drama, Sci-Fi, Space\n"
     ]
    }
   ],
   "source": [
    "print(clean_text(get_anime_description(pre_merged_anime.iloc[0])))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "id": "376e7066",
   "metadata": {},
   "outputs": [],
   "source": [
    "pre_merged_anime['generated_description'] = pre_merged_anime.apply(get_anime_description, axis=1)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "632bd750",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(5, 2)\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>status</th>\n",
       "      <th>description</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>Currently Watching</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2</td>\n",
       "      <td>Completed</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>On Hold</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4</td>\n",
       "      <td>Dropped</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>6</td>\n",
       "      <td>Plan to Watch</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   status         description\n",
       "0       1  Currently Watching\n",
       "1       2           Completed\n",
       "2       3             On Hold\n",
       "3       4             Dropped\n",
       "4       6       Plan to Watch"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "watching_status = pd.read_csv('../data/anime/watching_status.csv').dropna()\n",
    "print(watching_status.shape)\n",
    "watching_status"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "a4455cc3",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(57633278, 3)\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>user_id</th>\n",
       "      <th>anime_id</th>\n",
       "      <th>rating</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>430</td>\n",
       "      <td>9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>1004</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   user_id  anime_id  rating\n",
       "0        0       430       9\n",
       "1        0      1004       5"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rating_complete = pd.read_csv('../data/anime/rating_complete.csv').dropna()\n",
    "print(rating_complete.shape)\n",
    "rating_complete.head(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "b4edc75e",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "4bd861f5",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "\n",
    "rating_complete_train, rating_complete_test = \\\n",
    "              np.split(rating_complete.sample(frac=1, random_state=42), \n",
    "                       [int(.9*len(rating_complete))])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "id": "5aa7d252",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>user_id</th>\n",
       "      <th>anime_id</th>\n",
       "      <th>rating</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>41605545</th>\n",
       "      <td>255419</td>\n",
       "      <td>4059</td>\n",
       "      <td>7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>33600014</th>\n",
       "      <td>206382</td>\n",
       "      <td>10793</td>\n",
       "      <td>7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>39884414</th>\n",
       "      <td>245012</td>\n",
       "      <td>34246</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>45251042</th>\n",
       "      <td>277697</td>\n",
       "      <td>5030</td>\n",
       "      <td>7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>53777062</th>\n",
       "      <td>329602</td>\n",
       "      <td>227</td>\n",
       "      <td>7</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          user_id  anime_id  rating\n",
       "41605545   255419      4059       7\n",
       "33600014   206382     10793       7\n",
       "39884414   245012     34246       4\n",
       "45251042   277697      5030       7\n",
       "53777062   329602       227       7"
      ]
     },
     "execution_count": 52,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rating_complete_train.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "3cf2f8c5",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(51869950, 3)\n",
      "(5763328, 3)\n"
     ]
    }
   ],
   "source": [
    "print(rating_complete_train.shape)\n",
    "print(rating_complete_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "795a356e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "count     16865.000000\n",
       "mean       3075.597391\n",
       "std        9275.477582\n",
       "min           1.000000\n",
       "25%          36.000000\n",
       "50%         240.000000\n",
       "75%        1506.000000\n",
       "max      164244.000000\n",
       "Name: user_id, dtype: float64"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Group the DataFrame by anime_id and count the number of rows corresponding to each anime\n",
    "anime_counts = rating_complete_train.groupby('anime_id').count()\n",
    "\n",
    "anime_counts['user_id'].describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "05860602",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "dbef82bf",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "count    309426.000000\n",
       "mean        167.632810\n",
       "std         229.910583\n",
       "min           1.000000\n",
       "25%          39.000000\n",
       "50%         102.000000\n",
       "75%         214.000000\n",
       "max       13955.000000\n",
       "Name: anime_id, dtype: float64"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Group the DataFrame by anime_id and count the number of rows corresponding to each anime\n",
    "user_counts = rating_complete_train.groupby('user_id').count()\n",
    "\n",
    "user_counts['anime_id'].describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "e5328d92",
   "metadata": {},
   "outputs": [],
   "source": [
    "# removing animes from training set with less than 240 reviews for higher quality data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "b5149611",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "86194d2b",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "id": "21d33830",
   "metadata": {},
   "source": [
    "# Using content embeddings to recommend titles"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "32e93863",
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# !pip install dask dask-ml"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "1749491a",
   "metadata": {},
   "outputs": [],
   "source": [
    "score_map = {10: 1, 9: 1, 8: 0, 7: 0, 6: -1, 5: -1, 4:-1, 3:-1, 2:-1, 1:-1, 0:-1}\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "b86ba91d",
   "metadata": {},
   "outputs": [],
   "source": [
    "from tqdm.auto import tqdm\n",
    "from torch.utils.data import DataLoader\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "id": "ff97d138",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import itertools\n",
    "import random\n",
    "\n",
    "def jaccard_similarity(set1, set2):\n",
    "    intersection = len(set1.intersection(set2))\n",
    "    union = len(set1.union(set2))\n",
    "    if union:\n",
    "        return intersection / union\n",
    "\n",
    "# Set a rating threshold for positive examples\n",
    "rating_threshold = 9\n",
    "\n",
    "# Filter positive ratings\n",
    "positive_ratings = rating_complete_train[rating_complete_train['rating'] >= rating_threshold]\n",
    "\n",
    "# Create a dictionary of user_id and their positively rated animes\n",
    "user_anime_dict = positive_ratings.groupby('user_id')['anime_id'].apply(set).to_dict()\n",
    "\n",
    "\n",
    "# Compute Jaccard similarity between all anime pairs\n",
    "anime_pairs = itertools.combinations(sorted(set(pre_merged_anime['MAL_ID'].unique()) & set(rating_complete_train['anime_id'].unique())), 2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 255,
   "id": "96bed0c5",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6cdf7d298a3b4d32aec0467143c3b834",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "'''\n",
    "This script prepares training examples where each example consists of a pair of animes. \n",
    "The label for each pair is based on the Jaccard similarity of the users who like \n",
    "each of the animes in the pair. A high Jaccard similarity (>= 0.3) suggests that the \n",
    "animes are similar (positive example), while a low Jaccard similarity (< 0.1) \n",
    "suggests that the animes are dissimilar (negative example). \n",
    "\n",
    "The script maintains a ratio of at most 5 negative examples for each positive example.\n",
    "\n",
    "'''\n",
    "# Importing necessary libraries\n",
    "from random import sample, seed\n",
    "from sentence_transformers import InputExample\n",
    "\n",
    "# Initializing empty lists for storing positive and negative pairs examples\n",
    "positive_pairs_examples, negative_pairs_examples = [], []\n",
    "\n",
    "# Setting the seed value for random function to ensure the reproducibility of results\n",
    "seed(42)\n",
    "\n",
    "# Looping over each pair of animes\n",
    "for a1, a2 in tqdm(anime_pairs):\n",
    "    # If any of the animes in the pair is not found in the pre_merged_anime dataframe index, we skip the current iteration\n",
    "    if a1 not in pre_merged_anime.index or a2 not in pre_merged_anime.index:\n",
    "        continue\n",
    "\n",
    "    # Calculating Jaccard similarity between the sets of users who like anime a1 and anime a2\n",
    "    j = jaccard_similarity(user_anime_dict.get(a1, set()), user_anime_dict.get(a2, set())) \n",
    "\n",
    "    # If the Jaccard similarity is non-zero (i.e., there are some common users)\n",
    "    if j:\n",
    "        # If Jaccard similarity is greater than or equal to 0.3, \n",
    "        # then the pair is considered as a positive example and added to the positive_pairs_examples list\n",
    "        if j >= 0.3:\n",
    "            positive_pairs_examples.append(InputExample(texts=[pre_merged_anime.loc[a1]['generated_description'], pre_merged_anime.loc[a2]['generated_description']], label=1.))\n",
    "\n",
    "        # If Jaccard similarity is less than 0.1, \n",
    "        # then the pair is considered as a negative example and added to the negative_pairs_examples list\n",
    "        elif j < 0.1:\n",
    "            negative_pairs_examples.append(InputExample(texts=[pre_merged_anime.loc[a1]['generated_description'], pre_merged_anime.loc[a2]['generated_description']], label=0.))\n",
    "\n",
    "            # Ensuring that the number of negative examples does not exceed five times the number of positive examples. \n",
    "            # If it does, we shuffle the negative examples and truncate the list to maintain the 5:1 ratio.\n",
    "            if len(negative_pairs_examples) > len(positive_pairs_examples) * 5:\n",
    "                random.shuffle(negative_pairs_examples)\n",
    "                negative_pairs_examples = negative_pairs_examples[:len(positive_pairs_examples)]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "id": "d5749609",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'guid': '',\n",
       " 'texts': ['Sunabouzu is a TV Show.\\nSynopsis: The Great Kanto Desert, a sweltering wasteland of nothing but ruins and sand, is all that remains of post-apocalyptic Japan. The once fair population has been left to cling to the inhospitable dunes for survival. At least, that is the case for normal people. For those who have spent a little too long in the Kanto sun, the desert offers a wondrous opportunity to make a name for themselves. One such person is the masked handyman \"Sunabouzu,\" or Desert Punk, who has forged a legendary reputation for always finishing his jobs, no matter the nature or cost. Cunning and ruthless, he has become a force of crude destruction to the other desert people. However, the \"Vixen of the Desert,\" Junko Asagiri, discovers that Sunabouzu is not without his weaknesseshe is easily swayed by his insatiable lust for large-breasted desert babes. Following their chaotic adventures through the Kanto Desert, Sunabouzu features a bizarre cast of personalities who entertain themselves with senseless violence and perversion in a world long destroyed by their forefathers. And just like them, they have not learned a damn thing.\\nIt was produced by GDH, Pony Canyon, CBC and it is from Gonzo Studio.\\nIts source is Manga.\\nIt premiered in Fall 2004.\\nIts genres are Action, Adventure, Comedy, Ecchi, Sci-Fi, Shounen',\n",
       "  'Rockman.EXE Stream is a TV Show.\\nSynopsis: comet burns in the sky, its brilliant light shining down upon the people of Earth. At that time, a giant army of viruses invade the real world, crushing everything in their path. Netto and Enzan use cross fusion to intervene, but are whisked away to an unknown location, only to be confronted by the one responsible for the attack - Duo! (Source: Official Site)\\nIt was produced by Unknown and it is from Xebec Studio.\\nIts source is Game.\\nIt premiered in Fall 2004.\\nIts genres are Action, Adventure, Game, Mecha, School, Sci-Fi, Shounen'],\n",
       " 'label': 1.0}"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "positive_pairs_examples[0].__dict__"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "id": "c49514cd",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'guid': '',\n",
       " 'texts': [\"Basilisk: Kouga Ninpou Chou is a TV Show.\\nSynopsis: For centuries, the Iga and Kouga ninja clans have engaged in a bitter war. But when a ceasefire is ordered by the powerful warlord Ieyasu Tokugawa, the two clans are forced to put down their arms. Years later, Gennosuke Kouga, heir of the Kouga clan, and Oboro Iga, heir of the Iga clan, have fallen in love. Through marriage, both heirs aim to bring peace to the clans. But their hopes are dashed when flames of rivalry between their clans are reignited, and they are dragged into another war. Ieyasu's two grandsons have both claimed to be the next heir to the shogunate. To resolve this dispute, both the Kouga and Iga are ordered to send their 10 best warriors to fight in a bloody battle royale, with each clan representing one of the potential shogunate heirs. Two scrolls with the names of the fighters are given and are to be marked in blood upon the given fighter's death. The prize for winning is the favor of the Tokugawa shogunate for a thousand years. Torn between their love for each other and duty to their clans, Gennosuke and Oboro must ultimately decide the fate of their clans.\\nIt was produced by GDH, Kodansha, Sotsu Music Publishing and it is from Gonzo Studio.\\nIts source is Manga.\\nIt premiered in Spring 2005.\\nIts genres are Action, Adventure, Historical, Supernatural, Romance, Samurai, Fantasy\",\n",
       "  'Eikyuu Kazoku is a Movie.\\nSynopsis: It started as a sociological experiment. Six different people were brainwashed to think that they were a family and then put in arbitrary situations to see how they would react; however, when a taping of the families conducts is sold in a desperate attempt to make a little capital, it instantly becomes a success. Now the family is a first rated show and their every move is broadcasted around the world without them knowing it. But when a clogged toilet sets off a chain of events that ultimately frees the family from the room they are confined in the broadcasters, not willing to lose their source of income, set out to hunt down and recapture every member of the Eternal family. (Source: ANN)\\nIt was produced by Beyond C. and it is from Studio 4C Studio.\\nIts source is Original.\\nIt premiered in Unknown.\\nIts genres are Psychological, Comedy'],\n",
       " 'label': 0.0}"
      ]
     },
     "execution_count": 48,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "negative_pairs_examples[0].__dict__"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 256,
   "id": "91b0b305",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1567, 4918)"
      ]
     },
     "execution_count": 256,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(positive_pairs_examples), len(negative_pairs_examples)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 257,
   "id": "c178c550",
   "metadata": {},
   "outputs": [],
   "source": [
    "# downsample negative pairs\n",
    "if len(negative_pairs_examples) > len(positive_pairs_examples):\n",
    "    random.shuffle(negative_pairs_examples)\n",
    "    negative_pairs_examples = negative_pairs_examples[:len(positive_pairs_examples)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 258,
   "id": "529e0951",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1567, 1567)"
      ]
     },
     "execution_count": 258,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(positive_pairs_examples), len(negative_pairs_examples)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 259,
   "id": "4f7af0d3",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "# Assuming you have your positive_pairs and negative_pairs already\n",
    "train_positive_pairs, val_positive_pairs = train_test_split(positive_pairs_examples, test_size=0.2, random_state=42)\n",
    "train_negative_pairs, val_negative_pairs = train_test_split(negative_pairs_examples, test_size=0.2, random_state=42)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "ac54ca74",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "id": "81c7f340",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "SentenceTransformer(\n",
       "  (0): Transformer({'max_seq_length': 128, 'do_lower_case': False}) with Transformer model: RobertaModel \n",
       "  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False})\n",
       ")"
      ]
     },
     "execution_count": 51,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sentence_transformers import SentenceTransformer, evaluation, losses\n",
    "\n",
    "SentenceTransformer('paraphrase-distilroberta-base-v1')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "id": "c4bd3786",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2023-04-24 08:42:35 - Load pretrained SentenceTransformer: paraphrase-distilroberta-base-v1\n",
      "2023-04-24 08:42:36 - Use pytorch device: cuda\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "SentenceTransformer(\n",
       "  (0): Transformer({'max_seq_length': 384, 'do_lower_case': False}) with Transformer model: RobertaModel \n",
       "  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False})\n",
       ")"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Load a pre-trained SBERT model\n",
    "model = SentenceTransformer('paraphrase-distilroberta-base-v1')\n",
    "model.max_seq_length = 384     # Change context window to 384 tokens. It is normally 128\n",
    "model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "ebf771f3",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Token indices sequence length is longer than the specified maximum sequence length for this model (689 > 512). Running this sequence through the model will result in indexing errors\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1000x600 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "lengths = []\n",
    "for c in pre_merged_anime['generated_description']:\n",
    "    lengths.append(model.tokenizer(c, return_tensors='pt')['input_ids'].shape[1])\n",
    "    \n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# Assuming 'correlations' is a dictionary containing your data\n",
    "\n",
    "# Set up the plot and histogram\n",
    "plt.figure(figsize=(10, 6))\n",
    "plt.hist(lengths, bins=30, edgecolor='black', color='steelblue', alpha=0.7)\n",
    "\n",
    "# Customize the plot's appearance\n",
    "plt.title(\"Histogram of Token Lengths\", fontsize=16)\n",
    "plt.xlabel(\"Lengths\", fontsize=14)\n",
    "plt.ylabel(\"Frequency\", fontsize=14)\n",
    "plt.xticks(fontsize=12)\n",
    "plt.yticks(fontsize=12)\n",
    "plt.grid(axis='y', alpha=0.5, linestyle='--')\n",
    "\n",
    "# Display the histogram\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 262,
   "id": "fef14de9",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Combine positive and negative pairs for the validation set\n",
    "all_examples_val = val_positive_pairs + val_negative_pairs\n",
    "\n",
    "# Create a DataLoader for the validation examples\n",
    "val_dataloader = DataLoader(\n",
    "    all_examples_val,\n",
    "    batch_size=16,\n",
    "    shuffle=True\n",
    ")\n",
    "\n",
    "# Define the loss function\n",
    "loss = losses.CosineSimilarityLoss(model=model)\n",
    "\n",
    "# Extract sentences and scores from the validation examples\n",
    "val_sentences1 = [ex.texts[0] for ex in all_examples_val]\n",
    "val_sentences2 = [ex.texts[1] for ex in all_examples_val]\n",
    "val_scores = [ex.label for ex in all_examples_val]\n",
    "\n",
    "# Create the evaluator\n",
    "evaluator = evaluation.EmbeddingSimilarityEvaluator(val_sentences1, val_sentences2, val_scores)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "id": "a98cad44",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2023-04-24 08:42:39 - Load pretrained SentenceTransformer: paraphrase-distilroberta-base-v1\n",
      "2023-04-24 08:42:40 - Use pytorch device: cuda\n",
      "2023-04-24 08:42:40 - EmbeddingSimilarityEvaluator: Evaluating the model on  dataset:\n",
      "2023-04-24 08:42:42 - Cosine-Similarity :\tPearson: -0.3144\tSpearman: -0.3086\n",
      "2023-04-24 08:42:42 - Manhattan-Distance:\tPearson: -0.3988\tSpearman: -0.3878\n",
      "2023-04-24 08:42:42 - Euclidean-Distance:\tPearson: -0.3999\tSpearman: -0.3912\n",
      "2023-04-24 08:42:42 - Dot-Product-Similarity:\tPearson: -0.1789\tSpearman: -0.1800\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "-0.18002233783315705"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "SentenceTransformer('paraphrase-distilroberta-base-v1').evaluate(evaluator)  # original architecture cuts off so much context becasue limit is 128"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "id": "e3294b38",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2023-04-24 08:42:42 - EmbeddingSimilarityEvaluator: Evaluating the model on  dataset:\n",
      "2023-04-24 08:42:45 - Cosine-Similarity :\tPearson: -0.2338\tSpearman: -0.2173\n",
      "2023-04-24 08:42:45 - Manhattan-Distance:\tPearson: -0.3344\tSpearman: -0.3127\n",
      "2023-04-24 08:42:45 - Euclidean-Distance:\tPearson: -0.3290\tSpearman: -0.3069\n",
      "2023-04-24 08:42:45 - Dot-Product-Similarity:\tPearson: -0.0850\tSpearman: -0.0716\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "-0.07156819860087184"
      ]
     },
     "execution_count": 47,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.evaluate(evaluator)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "id": "52262c78",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "9bf3ecfc69fb46309e283dffe5d2b848",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Epoch:   0%|          | 0/5 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a55339c97fa64b92ba4efbd97cd9fc29",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Iteration:   0%|          | 0/145 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2023-04-24 08:43:51 - EmbeddingSimilarityEvaluator: Evaluating the model on  dataset after epoch 0:\n",
      "2023-04-24 08:43:54 - Cosine-Similarity :\tPearson: 0.7982\tSpearman: 0.7847\n",
      "2023-04-24 08:43:54 - Manhattan-Distance:\tPearson: 0.6105\tSpearman: 0.6440\n",
      "2023-04-24 08:43:54 - Euclidean-Distance:\tPearson: 0.6301\tSpearman: 0.6594\n",
      "2023-04-24 08:43:54 - Dot-Product-Similarity:\tPearson: 0.7791\tSpearman: 0.7686\n",
      "2023-04-24 08:43:54 - Save model to anime_encoder\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "2f6acfd71c7b40c2bcce245ecbe09fb1",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Iteration:   0%|          | 0/145 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2023-04-24 08:45:03 - EmbeddingSimilarityEvaluator: Evaluating the model on  dataset after epoch 1:\n",
      "2023-04-24 08:45:06 - Cosine-Similarity :\tPearson: 0.9244\tSpearman: 0.8531\n",
      "2023-04-24 08:45:06 - Manhattan-Distance:\tPearson: 0.8117\tSpearman: 0.8083\n",
      "2023-04-24 08:45:06 - Euclidean-Distance:\tPearson: 0.8231\tSpearman: 0.8114\n",
      "2023-04-24 08:45:06 - Dot-Product-Similarity:\tPearson: 0.9214\tSpearman: 0.8489\n",
      "2023-04-24 08:45:06 - Save model to anime_encoder\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a17049cd23cc49529e692bba22d28414",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Iteration:   0%|          | 0/145 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2023-04-24 08:46:13 - EmbeddingSimilarityEvaluator: Evaluating the model on  dataset after epoch 2:\n",
      "2023-04-24 08:46:16 - Cosine-Similarity :\tPearson: 0.9442\tSpearman: 0.8569\n",
      "2023-04-24 08:46:16 - Manhattan-Distance:\tPearson: 0.8382\tSpearman: 0.8139\n",
      "2023-04-24 08:46:16 - Euclidean-Distance:\tPearson: 0.8468\tSpearman: 0.8156\n",
      "2023-04-24 08:46:16 - Dot-Product-Similarity:\tPearson: 0.9386\tSpearman: 0.8559\n",
      "2023-04-24 08:46:16 - Save model to anime_encoder\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "ee8e1cb79e4142f194cdb0da82bdc5bd",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Iteration:   0%|          | 0/145 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2023-04-24 08:47:24 - EmbeddingSimilarityEvaluator: Evaluating the model on  dataset after epoch 3:\n",
      "2023-04-24 08:47:27 - Cosine-Similarity :\tPearson: 0.9517\tSpearman: 0.8605\n",
      "2023-04-24 08:47:27 - Manhattan-Distance:\tPearson: 0.8725\tSpearman: 0.8249\n",
      "2023-04-24 08:47:27 - Euclidean-Distance:\tPearson: 0.8786\tSpearman: 0.8250\n",
      "2023-04-24 08:47:27 - Dot-Product-Similarity:\tPearson: 0.9514\tSpearman: 0.8579\n",
      "2023-04-24 08:47:27 - Save model to anime_encoder\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6d8a8ee33c1e43bb9fcfd07b43174f4b",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Iteration:   0%|          | 0/145 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2023-04-24 08:48:36 - EmbeddingSimilarityEvaluator: Evaluating the model on  dataset after epoch 4:\n",
      "2023-04-24 08:48:39 - Cosine-Similarity :\tPearson: 0.9524\tSpearman: 0.8601\n",
      "2023-04-24 08:48:39 - Manhattan-Distance:\tPearson: 0.8707\tSpearman: 0.8197\n",
      "2023-04-24 08:48:39 - Euclidean-Distance:\tPearson: 0.8764\tSpearman: 0.8193\n",
      "2023-04-24 08:48:39 - Dot-Product-Similarity:\tPearson: 0.9502\tSpearman: 0.8580\n"
     ]
    }
   ],
   "source": [
    "# Combine positive and negative pairs\n",
    "train_examples = train_positive_pairs + train_negative_pairs\n",
    "\n",
    "# Create a DataLoader for the examples\n",
    "train_dataloader = DataLoader(\n",
    "    train_examples,\n",
    "    batch_size=16,\n",
    "    shuffle=True\n",
    ")\n",
    "\n",
    "# Define the loss function\n",
    "loss = losses.CosineSimilarityLoss(model=model)\n",
    "\n",
    "# Train the model\n",
    "num_epochs = 5\n",
    "warmup_steps = int(len(train_dataloader) * num_epochs * 0.1)  # 10% of training data for warm-up\n",
    "\n",
    "# Configure the training\n",
    "model.fit(\n",
    "    train_objectives=[(train_dataloader, loss)],\n",
    "    epochs=num_epochs,\n",
    "    warmup_steps=warmup_steps,\n",
    "    evaluator=evaluator,\n",
    "    output_path=\"anime_encoder\"\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "id": "7338061f",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2023-04-24 08:48:39 - EmbeddingSimilarityEvaluator: Evaluating the model on  dataset:\n",
      "2023-04-24 08:48:42 - Cosine-Similarity :\tPearson: 0.9524\tSpearman: 0.8601\n",
      "2023-04-24 08:48:42 - Manhattan-Distance:\tPearson: 0.8707\tSpearman: 0.8197\n",
      "2023-04-24 08:48:42 - Euclidean-Distance:\tPearson: 0.8764\tSpearman: 0.8193\n",
      "2023-04-24 08:48:42 - Dot-Product-Similarity:\tPearson: 0.9502\tSpearman: 0.8580\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "0.860054093114362"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.evaluate(evaluator)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "id": "902e9e94",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2023-04-24 08:48:42 - Save model to anime_encoder\n"
     ]
    }
   ],
   "source": [
    "# Save the fine-tuned model\n",
    "model.save(\"anime_encoder\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "6570a51e",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "id": "6a04d767",
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "torch.cuda.empty_cache()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "id": "d4ce640b",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2023-04-24 09:02:34 - Load pretrained SentenceTransformer: all-mpnet-base-v2\n",
      "2023-04-24 09:02:35 - Use pytorch device: cuda\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "SentenceTransformer(\n",
       "  (0): Transformer({'max_seq_length': 384, 'do_lower_case': False}) with Transformer model: MPNetModel \n",
       "  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False})\n",
       "  (2): Normalize()\n",
       ")"
      ]
     },
     "execution_count": 62,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sentence_transformers import SentenceTransformer, evaluation, losses\n",
    "\n",
    "# Load a pre-trained SBERT model\n",
    "model = SentenceTransformer('all-mpnet-base-v2')\n",
    "model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "id": "bba6f93f",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Combine positive and negative pairs\n",
    "train_examples = train_positive_pairs + train_negative_pairs\n",
    "\n",
    "# Create a DataLoader for the examples\n",
    "train_dataloader = DataLoader(\n",
    "    train_examples,\n",
    "    batch_size=8,\n",
    "    shuffle=True\n",
    ")\n",
    "\n",
    "# Define the loss function\n",
    "loss = losses.CosineSimilarityLoss(model=model)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "id": "e3e47a03",
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "\n",
    "device_index = 1  # change this to the index of the GPU you want to use\n",
    "torch.cuda.set_device(device_index)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "id": "cb557fbd",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2023-04-24 09:08:04 - EmbeddingSimilarityEvaluator: Evaluating the model on  dataset:\n",
      "2023-04-24 09:08:15 - Cosine-Similarity :\tPearson: 0.0581\tSpearman: 0.0738\n",
      "2023-04-24 09:08:15 - Manhattan-Distance:\tPearson: 0.0274\tSpearman: 0.0483\n",
      "2023-04-24 09:08:15 - Euclidean-Distance:\tPearson: 0.0479\tSpearman: 0.0738\n",
      "2023-04-24 09:08:15 - Dot-Product-Similarity:\tPearson: 0.0581\tSpearman: 0.0738\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "0.07383366675802173"
      ]
     },
     "execution_count": 74,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.evaluate(evaluator)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "id": "fd76411d",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "637e0334983c4842b2ab973584d06757",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Epoch:   0%|          | 0/5 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "468994ccb4ad4ec09fa049424f04ecae",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Iteration:   0%|          | 0/289 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2023-04-24 09:11:43 - EmbeddingSimilarityEvaluator: Evaluating the model on  dataset after epoch 0:\n",
      "2023-04-24 09:11:53 - Cosine-Similarity :\tPearson: 0.5348\tSpearman: 0.5376\n",
      "2023-04-24 09:11:53 - Manhattan-Distance:\tPearson: 0.5117\tSpearman: 0.5292\n",
      "2023-04-24 09:11:53 - Euclidean-Distance:\tPearson: 0.5194\tSpearman: 0.5376\n",
      "2023-04-24 09:11:53 - Dot-Product-Similarity:\tPearson: 0.5348\tSpearman: 0.5376\n",
      "2023-04-24 09:11:53 - Save model to anime_encoder_bigger\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "dd9249bab8eb4e968defb8771efa0f69",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Iteration:   0%|          | 0/289 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2023-04-24 09:15:22 - EmbeddingSimilarityEvaluator: Evaluating the model on  dataset after epoch 1:\n",
      "2023-04-24 09:15:32 - Cosine-Similarity :\tPearson: 0.6965\tSpearman: 0.6860\n",
      "2023-04-24 09:15:32 - Manhattan-Distance:\tPearson: 0.6738\tSpearman: 0.6825\n",
      "2023-04-24 09:15:32 - Euclidean-Distance:\tPearson: 0.6766\tSpearman: 0.6860\n",
      "2023-04-24 09:15:32 - Dot-Product-Similarity:\tPearson: 0.6965\tSpearman: 0.6860\n",
      "2023-04-24 09:15:32 - Save model to anime_encoder_bigger\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "56d44562a0b54dd7ad3dce4d28ac4e60",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Iteration:   0%|          | 0/289 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2023-04-24 09:19:01 - EmbeddingSimilarityEvaluator: Evaluating the model on  dataset after epoch 2:\n",
      "2023-04-24 09:19:11 - Cosine-Similarity :\tPearson: 0.7802\tSpearman: 0.7341\n",
      "2023-04-24 09:19:11 - Manhattan-Distance:\tPearson: 0.7705\tSpearman: 0.7296\n",
      "2023-04-24 09:19:11 - Euclidean-Distance:\tPearson: 0.7725\tSpearman: 0.7341\n",
      "2023-04-24 09:19:11 - Dot-Product-Similarity:\tPearson: 0.7802\tSpearman: 0.7341\n",
      "2023-04-24 09:19:11 - Save model to anime_encoder_bigger\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "40fb4596b0a04a4080b5423084e1a568",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Iteration:   0%|          | 0/289 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2023-04-24 09:22:38 - EmbeddingSimilarityEvaluator: Evaluating the model on  dataset after epoch 3:\n",
      "2023-04-24 09:22:47 - Cosine-Similarity :\tPearson: 0.8430\tSpearman: 0.7949\n",
      "2023-04-24 09:22:47 - Manhattan-Distance:\tPearson: 0.8344\tSpearman: 0.7910\n",
      "2023-04-24 09:22:47 - Euclidean-Distance:\tPearson: 0.8364\tSpearman: 0.7949\n",
      "2023-04-24 09:22:47 - Dot-Product-Similarity:\tPearson: 0.8430\tSpearman: 0.7949\n",
      "2023-04-24 09:22:47 - Save model to anime_encoder_bigger\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "bb05d9d063ad41e0a02c0d43b2c88036",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Iteration:   0%|          | 0/289 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2023-04-24 09:26:13 - EmbeddingSimilarityEvaluator: Evaluating the model on  dataset after epoch 4:\n",
      "2023-04-24 09:26:23 - Cosine-Similarity :\tPearson: 0.8896\tSpearman: 0.8260\n",
      "2023-04-24 09:26:23 - Manhattan-Distance:\tPearson: 0.8781\tSpearman: 0.8220\n",
      "2023-04-24 09:26:23 - Euclidean-Distance:\tPearson: 0.8808\tSpearman: 0.8260\n",
      "2023-04-24 09:26:23 - Dot-Product-Similarity:\tPearson: 0.8896\tSpearman: 0.8260\n",
      "2023-04-24 09:26:23 - Save model to anime_encoder_bigger\n"
     ]
    }
   ],
   "source": [
    "\n",
    "# Train the model\n",
    "num_epochs = 5\n",
    "warmup_steps = int(len(train_dataloader) * num_epochs * 0.1)  # 10% of training data for warm-up\n",
    "\n",
    "# Configure the training\n",
    "model.fit(\n",
    "    train_objectives=[(train_dataloader, loss)],\n",
    "    epochs=num_epochs,\n",
    "    evaluator=evaluator,\n",
    "    output_path=\"anime_encoder_bigger\"\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "id": "acc89d2d",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2023-04-24 09:26:24 - Save model to anime_encoder_bigger\n"
     ]
    }
   ],
   "source": [
    "model.save('anime_encoder_bigger')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "2387b088",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "d619bdf5",
   "metadata": {},
   "outputs": [],
   "source": [
    "import openai, os\n",
    "from openai.embeddings_utils import get_embeddings, get_embedding\n",
    "\n",
    "openai.api_key = os.environ.get('OPENAI_API_KEY')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 267,
   "id": "20ca4b96",
   "metadata": {},
   "outputs": [],
   "source": [
    "# del embeddings['anime_encoder']\n",
    "embeddings = {}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 268,
   "id": "de11cf7a",
   "metadata": {},
   "outputs": [],
   "source": [
    "embedders = ('text-embedding-ada-002', 'paraphrase-distilroberta-base-v1', 'anime_encoder', 'anime_encoder_bigger')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 269,
   "id": "56b52afb",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "761581a222804bdba7a3e8673410a800",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/163 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2023-04-25 08:31:36 - Load pretrained SentenceTransformer: paraphrase-distilroberta-base-v1\n",
      "2023-04-25 08:31:38 - Use pytorch device: cuda\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "7936f980b16e4e3a958b13212e6bbb21",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Batches:   0%|          | 0/127 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2023-04-25 08:32:05 - Load pretrained SentenceTransformer: anime_encoder\n",
      "2023-04-25 08:32:06 - Use pytorch device: cuda\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "eab4765e49294374a1034436e247d971",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Batches:   0%|          | 0/127 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2023-04-25 08:32:50 - Load pretrained SentenceTransformer: anime_encoder_bigger\n",
      "2023-04-25 08:32:51 - Use pytorch device: cuda\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "cd53c87617b64c5fa7a5e4e3cb4661ad",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Batches:   0%|          | 0/127 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2023-04-25 08:34:18 - Load pretrained SentenceTransformer: anime_encoder_even_bigger\n",
      "2023-04-25 08:34:21 - Use pytorch device: cuda\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6fa6dced74144672bea4782e944a6bff",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Batches:   0%|          | 0/127 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from tqdm.auto import tqdm\n",
    "\n",
    "'''\n",
    "Given a user id\n",
    "\n",
    "for each anime that the user gave a 9 or a 10 for:\n",
    "    get k other relevant animes (semantic search given an embedder)\n",
    "    for each relevant anime:\n",
    "        if the anime was rating in the training set for that user, ignore\n",
    "        elif the anime is not present in the testing set for that user, ignore\n",
    "        else:\n",
    "            check rating in testing for that user\n",
    "            case rating:\n",
    "                9,10 -> Promoter  ( + 2 points)\n",
    "                7,8  -> Passive   ( + 1 point )\n",
    "                1-6  -> Detractor ( - 1 point )\n",
    "                Not in the set    (   0 points)\n",
    "\n",
    "'''\n",
    "\n",
    "from sentence_transformers import SentenceTransformer\n",
    "\n",
    "docs = pre_merged_anime['generated_description'].tolist()\n",
    "for embedder in embedders:\n",
    "    if embedder in embeddings:\n",
    "        continue\n",
    "    if 'ada' in embedder:\n",
    "        ada_embeddings = None\n",
    "        for s in tqdm(range(0, len(docs), 100)):\n",
    "            if ada_embeddings is None:\n",
    "                ada_embeddings = np.array(get_embeddings(docs[s:s+100], engine=embedder))\n",
    "            else:\n",
    "                ada_embeddings = np.vstack([ada_embeddings, np.array(get_embeddings(docs[s:s+100], engine=embedder))])\n",
    "    \n",
    "        embeddings[embedder] = ada_embeddings\n",
    "    else:\n",
    "        e = SentenceTransformer(embedder)\n",
    "        embeddings[embedder] = e.encode(docs, batch_size=128, show_progress_bar=True)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "99b69b6b",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 270,
   "id": "9e460e63",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "from sklearn.metrics.pairwise import cosine_similarity\n",
    "cosine_sim_matrices = {embedder: cosine_similarity(embeddings[embedder]) for embedder in embedders}\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 271,
   "id": "4fb9110b",
   "metadata": {},
   "outputs": [],
   "source": [
    "import functools\n",
    "from typing import List, Tuple\n",
    "\n",
    "@functools.lru_cache(maxsize=None)\n",
    "def get_similar_animes(anime_id: int, embedder: str, k: int = 10, verbose: bool = False) -> List[Tuple[int, float]]:\n",
    "    # Get the index of the anime in the pre_merged_anime dataframe\n",
    "    try:\n",
    "        idx = pre_merged_anime[pre_merged_anime['MAL_ID'] == anime_id].index[0]\n",
    "    except:\n",
    "        return []\n",
    "    if verbose:\n",
    "        print(f'idx: {idx}')\n",
    "    cosine_sim_matrix = cosine_sim_matrices[embedder]\n",
    "\n",
    "    # Get the cosine similarity scores for the given anime_id\n",
    "    sim_scores = list(enumerate(cosine_sim_matrix[idx]))\n",
    "\n",
    "    # Sort the animes based on the similarity scores\n",
    "    sim_scores = sorted(sim_scores, key=lambda x: x[1], reverse=True)\n",
    "\n",
    "    # Get the indices and scores of the k most similar animes\n",
    "    k_indices_scores = sim_scores[1:k+1]\n",
    "\n",
    "    # Separate indices and scores\n",
    "    k_indices = [i[0] for i in k_indices_scores]\n",
    "    k_scores = [round(i[1], 3) for i in k_indices_scores]\n",
    "\n",
    "    # Get the MAL_IDs of the k most similar animes\n",
    "    k_anime_ids = pre_merged_anime['MAL_ID'].iloc[k_indices].tolist()\n",
    "\n",
    "    # Return the MAL_IDs and their corresponding similarity scores\n",
    "    return list(zip(k_anime_ids, k_scores))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "id": "62f9364d",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>MAL_ID</th>\n",
       "      <th>Name</th>\n",
       "      <th>Score</th>\n",
       "      <th>Genres</th>\n",
       "      <th>English name</th>\n",
       "      <th>Type</th>\n",
       "      <th>Episodes</th>\n",
       "      <th>Premiered</th>\n",
       "      <th>Producers</th>\n",
       "      <th>Studios</th>\n",
       "      <th>Source</th>\n",
       "      <th>Duration</th>\n",
       "      <th>Rating</th>\n",
       "      <th>Ranked</th>\n",
       "      <th>Members</th>\n",
       "      <th>Favorites</th>\n",
       "      <th>Completed</th>\n",
       "      <th>On-Hold</th>\n",
       "      <th>synopsis</th>\n",
       "      <th>generated_description</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1497</th>\n",
       "      <td>1723</td>\n",
       "      <td>Clannad Movie</td>\n",
       "      <td>7.25</td>\n",
       "      <td>Drama, Romance, Fantasy, School</td>\n",
       "      <td>Clannad The Motion Picture</td>\n",
       "      <td>Movie</td>\n",
       "      <td>1</td>\n",
       "      <td>Unknown</td>\n",
       "      <td>Unknown</td>\n",
       "      <td>Toei Animation</td>\n",
       "      <td>Visual novel</td>\n",
       "      <td>1 hr. 33 min.</td>\n",
       "      <td>PG-13 - Teens 13 or older</td>\n",
       "      <td>2566.0</td>\n",
       "      <td>143560</td>\n",
       "      <td>458</td>\n",
       "      <td>104445</td>\n",
       "      <td>1474</td>\n",
       "      <td>Tomoya Okazaki is a cynical delinquent who lac...</td>\n",
       "      <td>Clannad Movie is a Movie.\\nSynopsis: Tomoya Ok...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3057</th>\n",
       "      <td>4059</td>\n",
       "      <td>Clannad: Mou Hitotsu no Sekai, Tomoyo-hen</td>\n",
       "      <td>8.0</td>\n",
       "      <td>Drama, Romance, School, Slice of Life</td>\n",
       "      <td>Clannad:Another World, Tomoyo Chapter</td>\n",
       "      <td>Special</td>\n",
       "      <td>1</td>\n",
       "      <td>Unknown</td>\n",
       "      <td>Pony Canyon</td>\n",
       "      <td>Kyoto Animation</td>\n",
       "      <td>Visual novel</td>\n",
       "      <td>25 min.</td>\n",
       "      <td>PG-13 - Teens 13 or older</td>\n",
       "      <td>534.0</td>\n",
       "      <td>255959</td>\n",
       "      <td>873</td>\n",
       "      <td>230911</td>\n",
       "      <td>875</td>\n",
       "      <td>Clannad: Mou Hitotsu no Sekai, Tomoyo-hen is s...</td>\n",
       "      <td>Clannad: Mou Hitotsu no Sekai, Tomoyo-hen is a...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      MAL_ID                                       Name Score  \\\n",
       "1497    1723                              Clannad Movie  7.25   \n",
       "3057    4059  Clannad: Mou Hitotsu no Sekai, Tomoyo-hen   8.0   \n",
       "\n",
       "                                     Genres  \\\n",
       "1497        Drama, Romance, Fantasy, School   \n",
       "3057  Drama, Romance, School, Slice of Life   \n",
       "\n",
       "                               English name     Type Episodes Premiered  \\\n",
       "1497             Clannad The Motion Picture    Movie        1   Unknown   \n",
       "3057  Clannad:Another World, Tomoyo Chapter  Special        1   Unknown   \n",
       "\n",
       "        Producers          Studios        Source       Duration  \\\n",
       "1497      Unknown   Toei Animation  Visual novel  1 hr. 33 min.   \n",
       "3057  Pony Canyon  Kyoto Animation  Visual novel        25 min.   \n",
       "\n",
       "                         Rating  Ranked  Members  Favorites  Completed  \\\n",
       "1497  PG-13 - Teens 13 or older  2566.0   143560        458     104445   \n",
       "3057  PG-13 - Teens 13 or older   534.0   255959        873     230911   \n",
       "\n",
       "      On-Hold                                           synopsis  \\\n",
       "1497     1474  Tomoya Okazaki is a cynical delinquent who lac...   \n",
       "3057      875  Clannad: Mou Hitotsu no Sekai, Tomoyo-hen is s...   \n",
       "\n",
       "                                  generated_description  \n",
       "1497  Clannad Movie is a Movie.\\nSynopsis: Tomoya Ok...  \n",
       "3057  Clannad: Mou Hitotsu no Sekai, Tomoyo-hen is a...  "
      ]
     },
     "execution_count": 54,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pre_merged_anime[pre_merged_anime['MAL_ID'].isin((1723, 4059))]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "d0e312e2",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "idx: 3057\n",
      "text-embedding-ada-002\n",
      "[(6351, 0.921), (1723, 0.908), (2167, 0.905)]\n",
      "\n",
      "idx: 3057\n",
      "paraphrase-distilroberta-base-v1\n",
      "[(17835, 0.594), (33970, 0.589), (1723, 0.586)]\n",
      "\n",
      "idx: 3057\n",
      "anime_encoder\n",
      "[(2167, 0.989), (907, 0.988), (18599, 0.987)]\n",
      "\n",
      "idx: 3057\n",
      "anime_encoder_bigger\n",
      "[(39741, 0.99), (323, 0.989), (34299, 0.987)]\n"
     ]
    }
   ],
   "source": [
    "anime_id = 4059\n",
    "k = 10\n",
    "for e in embedders:\n",
    "    similar_animes = get_similar_animes(anime_id, e, k=3, verbose=True)\n",
    "    print(e)\n",
    "    print(similar_animes)\n",
    "    print()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "283b4836",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 273,
   "id": "ceb734ef",
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_promoted_animes_from_user(user_id):\n",
    "    return rating_complete_train[(rating_complete_train['user_id'] == user_id) & (rating_complete_train['rating'] >= 9)]['anime_id'].tolist()\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "e9a9f528",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 275,
   "id": "49110f5f",
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_relevant_animes(reference_promoted_anime_id, embedder, k=5, verbose=True):\n",
    "    preds = []\n",
    "    similar_animes = get_similar_animes(reference_promoted_anime_id, embedder, k=k)\n",
    "\n",
    "    for similar_anime_id, cosine_score in similar_animes:\n",
    "        if cosine_score >= 1:\n",
    "            continue  # ignoring exact matches\n",
    "        if verbose:\n",
    "            print(f'Recommending (ID {similar_anime_id}) Cos_Sim {cosine_score:.3f}')\n",
    "            print(f'User gave it a rating of: {user_rating}')\n",
    "        preds.append((similar_anime_id, cosine_score))\n",
    "    return preds\n",
    "                     "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 276,
   "id": "76340dea",
   "metadata": {},
   "outputs": [],
   "source": [
    "user_preds = {}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 277,
   "id": "ea91b958",
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "from collections import defaultdict\n",
    "import concurrent.futures\n",
    "from functools import partial\n",
    "from tqdm import tqdm\n",
    "\n",
    "def process_user(user_id, embedders, K, verbose=False):\n",
    "    user_predictions = {e: defaultdict(float) for e in embedders}\n",
    "    for promoted_anime_id in set(get_promoted_animes_from_user(user_id)):\n",
    "        for embedder in embedders:\n",
    "            relevant_animes = get_relevant_animes(promoted_anime_id, embedder, verbose=False, k=K)\n",
    "            for anime_id, cosine_score in relevant_animes:\n",
    "                if verbose:\n",
    "                    print(embedder, user_id, promoted_anime_id, anime_id, cosine_score)\n",
    "                if anime_id not in user_predictions[embedder] or cosine_score > user_predictions[embedder][anime_id]:\n",
    "                    user_predictions[embedder][anime_id] = cosine_score\n",
    "    return user_id, user_predictions\n",
    "\n",
    "# Prepare the data\n",
    "user_ids = rating_complete_test['user_id'].unique()[len(user_preds):]\n",
    "\n",
    "# Set the number of workers and the value for K\n",
    "num_workers = min(8, len(user_ids))\n",
    "K = 3\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 279,
   "id": "96ec3d21",
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# Parallelize the code using multiple threads\n",
    "with concurrent.futures.ThreadPoolExecutor(max_workers=num_workers) as executor:\n",
    "    # Create a partial function with fixed arguments for embedders and K\n",
    "    process_user_partial = partial(process_user, embedders=embedders, K=K)\n",
    "    \n",
    "    # Run the tasks in parallel and update the user_preds dictionary\n",
    "    for user_id, user_predictions in tqdm(executor.map(process_user_partial, user_ids), total=len(user_ids)):\n",
    "        user_preds[user_id] = user_predictions\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 145,
   "id": "87725fe4",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "285789"
      ]
     },
     "execution_count": 145,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(user_preds)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 146,
   "id": "75afa953",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(205282,\n",
       " {'text-embedding-ada-002': defaultdict(float,\n",
       "              {5: 0.914,\n",
       "               17205: 0.882,\n",
       "               4037: 0.878,\n",
       "               21507: 0.911,\n",
       "               33775: 0.874,\n",
       "               21195: 0.873,\n",
       "               1: 0.914,\n",
       "               20359: 0.955,\n",
       "               24627: 0.951,\n",
       "               1562: 0.898,\n",
       "               23277: 0.96,\n",
       "               36885: 0.942,\n",
       "               29317: 0.937,\n",
       "               4106: 0.927,\n",
       "               28791: 0.853,\n",
       "               1200: 0.848,\n",
       "               22729: 0.927,\n",
       "               33051: 0.878,\n",
       "               29223: 0.887,\n",
       "               39587: 0.944,\n",
       "               42203: 0.947,\n",
       "               38414: 0.915,\n",
       "               39940: 0.938,\n",
       "               32282: 0.937,\n",
       "               35788: 0.93,\n",
       "               11597: 0.943,\n",
       "               28025: 0.929,\n",
       "               5081: 0.943,\n",
       "               31758: 0.942,\n",
       "               9260: 0.927,\n",
       "               31757: 0.942,\n",
       "               19291: 0.926,\n",
       "               40351: 0.925,\n",
       "               1565: 0.924,\n",
       "               22507: 0.936,\n",
       "               187: 0.948,\n",
       "               185: 0.948,\n",
       "               39332: 0.872,\n",
       "               1469: 0.87,\n",
       "               40708: 0.868,\n",
       "               1735: 0.939,\n",
       "               34566: 0.912,\n",
       "               28755: 0.904,\n",
       "               11483: 0.912,\n",
       "               2283: 0.839,\n",
       "               6162: 0.838,\n",
       "               33218: 0.902,\n",
       "               41379: 0.895,\n",
       "               147: 0.887,\n",
       "               30143: 0.875,\n",
       "               32805: 0.864,\n",
       "               159: 0.853,\n",
       "               527: 0.926,\n",
       "               22297: 0.961,\n",
       "               31389: 0.938,\n",
       "               27821: 0.957,\n",
       "               11977: 0.953,\n",
       "               38256: 0.919,\n",
       "               11981: 0.908,\n",
       "               31: 0.926,\n",
       "               2759: 0.924,\n",
       "               32: 0.917,\n",
       "               30: 0.917,\n",
       "               32379: 0.907,\n",
       "               10218: 0.883,\n",
       "               34055: 0.872,\n",
       "               34034: 0.905,\n",
       "               30484: 0.936,\n",
       "               10863: 0.907,\n",
       "               11577: 0.903,\n",
       "               34438: 0.975,\n",
       "               2904: 0.969,\n",
       "               34439: 0.943,\n",
       "               41074: 0.968,\n",
       "               1313: 0.936,\n",
       "               2961: 0.924,\n",
       "               13411: 0.88,\n",
       "               13561: 0.871,\n",
       "               17949: 0.868,\n",
       "               22043: 0.938,\n",
       "               35972: 0.934,\n",
       "               18393: 0.892,\n",
       "               40054: 0.902,\n",
       "               27411: 0.877,\n",
       "               16662: 0.876,\n",
       "               7791: 0.953,\n",
       "               9617: 0.917,\n",
       "               9734: 0.894,\n",
       "               1692: 0.873,\n",
       "               2785: 0.872,\n",
       "               38483: 0.868,\n",
       "               269: 0.893,\n",
       "               834: 0.881,\n",
       "               4835: 0.875,\n",
       "               9969: 0.965,\n",
       "               34096: 0.947,\n",
       "               28977: 0.965,\n",
       "               2789: 0.929,\n",
       "               6901: 0.888,\n",
       "               9346: 0.886,\n",
       "               37994: 0.895,\n",
       "               38410: 0.854,\n",
       "               30916: 0.854,\n",
       "               2620: 0.941,\n",
       "               1264: 0.859,\n",
       "               3817: 0.855,\n",
       "               2713: 0.917,\n",
       "               1710: 0.92,\n",
       "               2109: 0.895,\n",
       "               886: 0.87,\n",
       "               1485: 0.865,\n",
       "               22247: 0.864,\n",
       "               14353: 0.86,\n",
       "               30659: 0.848,\n",
       "               6880: 0.843,\n",
       "               2110: 0.915,\n",
       "               2619: 0.941,\n",
       "               24527: 0.853,\n",
       "               8532: 0.888,\n",
       "               33740: 0.872,\n",
       "               9023: 0.864,\n",
       "               1889: 0.963,\n",
       "               934: 0.963,\n",
       "               10491: 0.926,\n",
       "               31704: 0.913,\n",
       "               31772: 0.898,\n",
       "               34134: 0.896,\n",
       "               36632: 0.88,\n",
       "               11843: 0.879,\n",
       "               11013: 0.878,\n",
       "               2589: 0.871,\n",
       "               42585: 0.865,\n",
       "               4531: 0.86,\n",
       "               22265: 0.93,\n",
       "               36704: 0.905,\n",
       "               30415: 0.888,\n",
       "               34822: 0.892,\n",
       "               634: 0.89,\n",
       "               1145: 0.887,\n",
       "               6351: 0.917,\n",
       "               2167: 0.914,\n",
       "               1723: 0.953,\n",
       "               22877: 0.873,\n",
       "               28497: 0.87,\n",
       "               40595: 0.867,\n",
       "               23199: 0.93,\n",
       "               27831: 0.91,\n",
       "               27833: 0.908,\n",
       "               4197: 0.924,\n",
       "               2819: 0.873,\n",
       "               9330: 0.872,\n",
       "               31561: 0.845,\n",
       "               411: 0.837,\n",
       "               38910: 0.836,\n",
       "               38003: 0.941,\n",
       "               31478: 0.923,\n",
       "               42250: 0.902,\n",
       "               38883: 0.928,\n",
       "               28891: 0.924,\n",
       "               40372: 0.907,\n",
       "               38972: 0.924,\n",
       "               31933: 0.915,\n",
       "               14719: 0.893,\n",
       "               36480: 0.917,\n",
       "               33032: 0.888,\n",
       "               38360: 0.883,\n",
       "               35073: 0.938,\n",
       "               37675: 0.911,\n",
       "               34161: 0.876,\n",
       "               15417: 0.899,\n",
       "               21899: 0.919,\n",
       "               32937: 0.956,\n",
       "               34626: 0.93,\n",
       "               32380: 0.91,\n",
       "               2335: 0.904,\n",
       "               40456: 0.902,\n",
       "               47778: 0.902,\n",
       "               35760: 0.949,\n",
       "               25777: 0.944,\n",
       "               40028: 0.92,\n",
       "               36838: 0.928,\n",
       "               6945: 0.902,\n",
       "               37086: 0.94,\n",
       "               4550: 0.886,\n",
       "               242: 0.884,\n",
       "               33031: 0.869,\n",
       "               11371: 0.861,\n",
       "               914: 0.861,\n",
       "               4181: 0.914,\n",
       "               4059: 0.905,\n",
       "               5114: 0.963,\n",
       "               664: 0.896,\n",
       "               430: 0.895,\n",
       "               10719: 0.951,\n",
       "               10897: 0.942,\n",
       "               14027: 0.872,\n",
       "               16498: 0.936,\n",
       "               37374: 0.863,\n",
       "               685: 0.857,\n",
       "               1491: 0.85,\n",
       "               1218: 0.881,\n",
       "               23701: 0.88,\n",
       "               11553: 0.874,\n",
       "               5962: 0.93,\n",
       "               8410: 0.926,\n",
       "               13231: 0.915,\n",
       "               288: 0.944,\n",
       "               1669: 0.933,\n",
       "               1670: 0.922,\n",
       "               32182: 0.964,\n",
       "               36616: 0.911,\n",
       "               39651: 0.883,\n",
       "               849: 0.945,\n",
       "               4382: 0.938,\n",
       "               5118: 0.903,\n",
       "               37999: 0.974,\n",
       "               43608: 0.94,\n",
       "               43609: 0.915,\n",
       "               2836: 0.86,\n",
       "               335: 0.853,\n",
       "               31647: 0.853,\n",
       "               24439: 0.95,\n",
       "               30915: 0.898,\n",
       "               48466: 0.867,\n",
       "               35376: 0.964,\n",
       "               31234: 0.919,\n",
       "               30991: 0.908,\n",
       "               42205: 0.978,\n",
       "               3861: 0.895,\n",
       "               1557: 0.866,\n",
       "               249: 0.938,\n",
       "               450: 0.924,\n",
       "               452: 0.917,\n",
       "               842: 0.959,\n",
       "               31098: 0.938,\n",
       "               5308: 0.899,\n",
       "               31240: 0.944,\n",
       "               11741: 0.933,\n",
       "               40776: 0.942,\n",
       "               23447: 0.901,\n",
       "               12119: 0.889,\n",
       "               4483: 0.888,\n",
       "               30831: 0.956,\n",
       "               28701: 0.961,\n",
       "               6922: 0.931,\n",
       "               2108: 0.92,\n",
       "               36702: 0.951,\n",
       "               38524: 0.949,\n",
       "               30813: 0.899,\n",
       "               6347: 0.889,\n",
       "               17827: 0.881,\n",
       "               35566: 0.882,\n",
       "               2236: 0.878,\n",
       "               38585: 0.891,\n",
       "               661: 0.887,\n",
       "               16023: 0.884,\n",
       "               6746: 0.91,\n",
       "               26449: 0.881,\n",
       "               6973: 0.876,\n",
       "               18679: 0.872,\n",
       "               32915: 0.91,\n",
       "               186: 0.939,\n",
       "               18: 0.915,\n",
       "               11177: 0.92,\n",
       "               4563: 0.878,\n",
       "               2702: 0.905,\n",
       "               71: 0.924,\n",
       "               73: 0.909,\n",
       "               38149: 0.906,\n",
       "               41006: 0.951,\n",
       "               28817: 0.89,\n",
       "               8765: 0.881,\n",
       "               8709: 0.876,\n",
       "               11737: 0.911,\n",
       "               33506: 0.904,\n",
       "               11266: 0.882,\n",
       "               822: 0.922,\n",
       "               3467: 0.879,\n",
       "               26351: 0.878,\n",
       "               29589: 0.878,\n",
       "               27655: 0.927,\n",
       "               413: 0.868,\n",
       "               8516: 0.934,\n",
       "               36250: 0.889,\n",
       "               35507: 0.889,\n",
       "               3901: 0.887,\n",
       "               19991: 0.857,\n",
       "               6277: 0.849,\n",
       "               19365: 0.876,\n",
       "               1167: 0.867,\n",
       "               1098: 0.865,\n",
       "               36456: 0.934,\n",
       "               31964: 0.93,\n",
       "               39565: 0.924,\n",
       "               31722: 0.928,\n",
       "               34577: 0.909,\n",
       "               39701: 0.903,\n",
       "               33028: 0.929,\n",
       "               32189: 0.924,\n",
       "               34152: 0.898,\n",
       "               10370: 0.93,\n",
       "               20583: 0.924,\n",
       "               33486: 0.93,\n",
       "               36896: 0.927,\n",
       "               1686: 0.888,\n",
       "               762: 0.881,\n",
       "               4667: 0.929,\n",
       "               1672: 0.904,\n",
       "               36756: 0.86,\n",
       "               33964: 0.847,\n",
       "               34777: 0.817,\n",
       "               25647: 0.881,\n",
       "               13535: 0.865,\n",
       "               19919: 0.86,\n",
       "               36432: 0.95,\n",
       "               4654: 0.945,\n",
       "               10249: 0.92,\n",
       "               28069: 0.892,\n",
       "               37975: 0.877,\n",
       "               3229: 0.876,\n",
       "               31637: 0.955,\n",
       "               127: 0.89,\n",
       "               2025: 0.883,\n",
       "               35134: 0.932,\n",
       "               4246: 0.921,\n",
       "               35136: 0.911,\n",
       "               29067: 0.942,\n",
       "               39326: 0.88,\n",
       "               1222: 0.875,\n",
       "               27601: 0.929,\n",
       "               23237: 0.925,\n",
       "               14741: 0.919,\n",
       "               676: 0.879,\n",
       "               9154: 0.852,\n",
       "               33689: 0.852,\n",
       "               918: 0.952,\n",
       "               6033: 0.944,\n",
       "               813: 0.944,\n",
       "               223: 0.938,\n",
       "               844: 0.911,\n",
       "               911: 0.895,\n",
       "               7430: 0.888,\n",
       "               6811: 0.938,\n",
       "               308: 0.879,\n",
       "               2523: 0.873,\n",
       "               1691: 0.87,\n",
       "               26243: 0.933,\n",
       "               31378: 0.923,\n",
       "               31756: 0.903,\n",
       "               30654: 0.923,\n",
       "               19759: 0.89,\n",
       "               33513: 0.888,\n",
       "               15039: 0.95,\n",
       "               38963: 0.917,\n",
       "               37993: 0.884,\n",
       "               777: 0.953,\n",
       "               11077: 0.912,\n",
       "               7728: 0.878,\n",
       "               9253: 0.936,\n",
       "               6956: 0.952,\n",
       "               10521: 0.941,\n",
       "               33094: 0.913,\n",
       "               356: 0.935,\n",
       "               15489: 0.939,\n",
       "               16574: 0.902,\n",
       "               21639: 0.898,\n",
       "               1668: 0.944,\n",
       "               552: 0.936,\n",
       "               2398: 0.928,\n",
       "               22145: 0.921,\n",
       "               6793: 0.905,\n",
       "               23317: 0.902,\n",
       "               20507: 0.934,\n",
       "               30885: 0.921,\n",
       "               20767: 0.911,\n",
       "               33161: 0.933,\n",
       "               14813: 0.933,\n",
       "               39547: 0.936,\n",
       "               9120: 0.922,\n",
       "               37716: 0.905,\n",
       "               10479: 0.878,\n",
       "               37514: 0.918,\n",
       "               41084: 0.896,\n",
       "               36862: 0.892,\n",
       "               30694: 0.926,\n",
       "               27899: 0.946,\n",
       "               36511: 0.919,\n",
       "               37799: 0.9,\n",
       "               9735: 0.931,\n",
       "               30100: 0.87,\n",
       "               1607: 0.862,\n",
       "               1269: 0.857,\n",
       "               35466: 0.941,\n",
       "               37921: 0.875,\n",
       "               33782: 0.873,\n",
       "               34414: 0.909,\n",
       "               34533: 0.874,\n",
       "               776: 0.868,\n",
       "               12391: 0.883,\n",
       "               20001: 0.879,\n",
       "               5343: 0.876,\n",
       "               35815: 0.914,\n",
       "               160: 0.87,\n",
       "               2596: 0.886,\n",
       "               44387: 0.881,\n",
       "               511: 0.879,\n",
       "               9790: 0.91,\n",
       "               8407: 0.899,\n",
       "               6098: 0.885,\n",
       "               31553: 0.907,\n",
       "               3196: 0.875,\n",
       "               25879: 0.913,\n",
       "               5030: 0.877,\n",
       "               42826: 0.875,\n",
       "               614: 0.874,\n",
       "               21681: 0.883,\n",
       "               28677: 0.869,\n",
       "               37970: 0.869,\n",
       "               31181: 0.942,\n",
       "               250: 0.911,\n",
       "               31593: 0.883,\n",
       "               18677: 0.879,\n",
       "               22865: 0.876,\n",
       "               30736: 0.924,\n",
       "               34642: 0.909,\n",
       "               28447: 0.906,\n",
       "               26: 0.844,\n",
       "               31988: 0.963,\n",
       "               31989: 0.951,\n",
       "               35082: 0.933,\n",
       "               1575: 0.969,\n",
       "               37476: 0.923,\n",
       "               15611: 0.921,\n",
       "               9539: 0.919,\n",
       "               31592: 0.949,\n",
       "               24075: 0.934,\n",
       "               36023: 0.921,\n",
       "               38431: 0.887,\n",
       "               2962: 0.925,\n",
       "               2397: 0.91,\n",
       "               28979: 0.946,\n",
       "               9181: 0.96,\n",
       "               3455: 0.93,\n",
       "               3652: 0.945,\n",
       "               11771: 0.953,\n",
       "               16894: 0.949,\n",
       "               32871: 0.931,\n",
       "               3011: 0.884,\n",
       "               553: 0.879,\n",
       "               12445: 0.878,\n",
       "               37055: 0.915,\n",
       "               34904: 0.888,\n",
       "               34742: 0.877,\n",
       "               23321: 0.914,\n",
       "               41109: 0.911,\n",
       "               21235: 0.902,\n",
       "               1519: 0.91,\n",
       "               4901: 0.872,\n",
       "               8440: 0.865,\n",
       "               898: 0.92,\n",
       "               502: 0.914,\n",
       "               891: 0.93,\n",
       "               666: 0.923,\n",
       "               3603: 0.922,\n",
       "               665: 0.909,\n",
       "               13663: 0.93,\n",
       "               5667: 0.923,\n",
       "               903: 0.923,\n",
       "               896: 0.912,\n",
       "               899: 0.909,\n",
       "               901: 0.923,\n",
       "               904: 0.92,\n",
       "               22777: 0.931,\n",
       "               39617: 0.913,\n",
       "               40642: 0.868,\n",
       "               47616: 0.867,\n",
       "               14093: 0.911,\n",
       "               17115: 0.903,\n",
       "               1564: 0.892,\n",
       "               28907: 0.955,\n",
       "               28869: 0.873,\n",
       "               36196: 0.872,\n",
       "               18671: 0.919,\n",
       "               19021: 0.915,\n",
       "               9062: 0.913,\n",
       "               10067: 0.891,\n",
       "               51: 0.871,\n",
       "               13469: 0.911,\n",
       "               15391: 0.881,\n",
       "               26055: 0.95,\n",
       "               2899: 0.936,\n",
       "               15547: 0.92,\n",
       "               21479: 0.92,\n",
       "               17074: 0.934,\n",
       "               36796: 0.924,\n",
       "               31588: 0.869,\n",
       "               3545: 0.861,\n",
       "               30276: 0.856,\n",
       "               37510: 0.964,\n",
       "               37504: 0.874,\n",
       "               39247: 0.94,\n",
       "               35145: 0.911,\n",
       "               21033: 0.876,\n",
       "               34103: 0.944,\n",
       "               16592: 0.924,\n",
       "               24833: 0.923,\n",
       "               32863: 0.898,\n",
       "               39988: 0.892,\n",
       "               24823: 0.858,\n",
       "               33219: 0.854,\n",
       "               6215: 0.853,\n",
       "               48413: 0.916,\n",
       "               14833: 0.888,\n",
       "               37210: 0.887,\n",
       "               33018: 0.898,\n",
       "               2216: 0.888,\n",
       "               30012: 0.887,\n",
       "               9883: 0.943,\n",
       "               9884: 0.932,\n",
       "               38201: 0.926,\n",
       "               20899: 0.95,\n",
       "               35247: 0.942,\n",
       "               36999: 0.921,\n",
       "               4565: 0.93,\n",
       "               4107: 0.945,\n",
       "               10622: 0.92,\n",
       "               27787: 0.961,\n",
       "               9926: 0.889,\n",
       "               41103: 0.887,\n",
       "               2001: 0.93,\n",
       "               8348: 0.9,\n",
       "               10087: 0.933,\n",
       "               23847: 0.933,\n",
       "               46431: 0.921,\n",
       "               14967: 0.951,\n",
       "               35712: 0.887,\n",
       "               482: 0.943,\n",
       "               550: 0.921,\n",
       "               40145: 0.909,\n",
       "               18897: 0.86,\n",
       "               31414: 0.857,\n",
       "               37491: 0.928,\n",
       "               36789: 0.906,\n",
       "               10178: 0.866,\n",
       "               7472: 0.916,\n",
       "               892: 0.915,\n",
       "               893: 0.914,\n",
       "               897: 0.894,\n",
       "               121: 0.963,\n",
       "               6421: 0.893,\n",
       "               33247: 0.879,\n",
       "               21781: 0.878,\n",
       "               31422: 0.875,\n",
       "               24415: 0.949,\n",
       "               32870: 0.922,\n",
       "               2994: 0.889,\n",
       "               789: 0.857}),\n",
       "  'paraphrase-distilroberta-base-v1': defaultdict(float,\n",
       "              {4240: 0.622,\n",
       "               2361: 0.584,\n",
       "               12149: 0.583,\n",
       "               21507: 0.712,\n",
       "               40496: 0.652,\n",
       "               37202: 0.652,\n",
       "               1365: 0.571,\n",
       "               548: 0.559,\n",
       "               10624: 0.549,\n",
       "               1554: 0.624,\n",
       "               18897: 0.576,\n",
       "               7592: 0.575,\n",
       "               31914: 0.627,\n",
       "               23277: 0.626,\n",
       "               36885: 0.625,\n",
       "               4106: 0.742,\n",
       "               2213: 0.541,\n",
       "               3588: 0.521,\n",
       "               4503: 0.565,\n",
       "               33352: 0.56,\n",
       "               10336: 0.554,\n",
       "               11785: 0.577,\n",
       "               39587: 0.574,\n",
       "               22817: 0.568,\n",
       "               35788: 0.638,\n",
       "               36949: 0.633,\n",
       "               32828: 0.632,\n",
       "               40316: 0.65,\n",
       "               23385: 0.648,\n",
       "               34006: 0.645,\n",
       "               31758: 0.721,\n",
       "               11597: 0.691,\n",
       "               5081: 0.691,\n",
       "               31757: 0.721,\n",
       "               28927: 0.559,\n",
       "               1565: 0.71,\n",
       "               40351: 0.64,\n",
       "               31592: 0.745,\n",
       "               187: 0.645,\n",
       "               22507: 0.585,\n",
       "               186: 0.665,\n",
       "               30740: 0.531,\n",
       "               392: 0.599,\n",
       "               19315: 0.485,\n",
       "               1735: 0.686,\n",
       "               2243: 0.671,\n",
       "               4437: 0.65,\n",
       "               33670: 0.453,\n",
       "               27463: 0.447,\n",
       "               7259: 0.437,\n",
       "               33204: 0.659,\n",
       "               7724: 0.63,\n",
       "               17741: 0.629,\n",
       "               40590: 0.638,\n",
       "               21437: 0.55,\n",
       "               6428: 0.545,\n",
       "               9107: 0.672,\n",
       "               11741: 0.698,\n",
       "               31389: 0.647,\n",
       "               27821: 0.762,\n",
       "               11977: 0.735,\n",
       "               33003: 0.656,\n",
       "               576: 0.653,\n",
       "               32: 0.726,\n",
       "               3784: 0.709,\n",
       "               40956: 0.66,\n",
       "               30: 0.726,\n",
       "               2759: 0.7,\n",
       "               32379: 0.634,\n",
       "               10218: 0.581,\n",
       "               12115: 0.578,\n",
       "               527: 0.71,\n",
       "               1564: 0.674,\n",
       "               30484: 0.666,\n",
       "               34502: 0.544,\n",
       "               10568: 0.58,\n",
       "               34438: 0.755,\n",
       "               2904: 0.629,\n",
       "               34439: 0.58,\n",
       "               1132: 0.647,\n",
       "               41074: 0.714,\n",
       "               2961: 0.611,\n",
       "               36511: 0.643,\n",
       "               2257: 0.573,\n",
       "               2303: 0.563,\n",
       "               8536: 0.63,\n",
       "               34392: 0.597,\n",
       "               14513: 0.59,\n",
       "               9982: 0.589,\n",
       "               18393: 0.584,\n",
       "               16211: 0.597,\n",
       "               47250: 0.588,\n",
       "               1347: 0.576,\n",
       "               35639: 0.673,\n",
       "               7791: 0.66,\n",
       "               27989: 0.656,\n",
       "               11757: 0.51,\n",
       "               35073: 0.639,\n",
       "               22729: 0.49,\n",
       "               269: 0.681,\n",
       "               834: 0.618,\n",
       "               7696: 0.597,\n",
       "               41380: 0.592,\n",
       "               9969: 0.598,\n",
       "               34096: 0.523,\n",
       "               22249: 0.592,\n",
       "               41456: 0.579,\n",
       "               2789: 0.573,\n",
       "               17395: 0.53,\n",
       "               693: 0.519,\n",
       "               2346: 0.513,\n",
       "               2620: 0.678,\n",
       "               31631: 0.658,\n",
       "               1623: 0.63,\n",
       "               927: 0.598,\n",
       "               5829: 0.568,\n",
       "               21855: 0.618,\n",
       "               28171: 0.632,\n",
       "               3437: 0.626,\n",
       "               8123: 0.57,\n",
       "               33047: 0.58,\n",
       "               14353: 0.535,\n",
       "               39417: 0.673,\n",
       "               1835: 0.583,\n",
       "               2109: 0.573,\n",
       "               1710: 0.561,\n",
       "               2619: 0.678,\n",
       "               532: 0.607,\n",
       "               32407: 0.598,\n",
       "               61: 0.596,\n",
       "               1889: 0.698,\n",
       "               42249: 0.573,\n",
       "               15959: 0.555,\n",
       "               31704: 0.638,\n",
       "               34134: 0.589,\n",
       "               2402: 0.579,\n",
       "               29854: 0.672,\n",
       "               4654: 0.636,\n",
       "               5958: 0.672,\n",
       "               2051: 0.527,\n",
       "               28701: 0.67,\n",
       "               36704: 0.682,\n",
       "               33845: 0.68,\n",
       "               35191: 0.58,\n",
       "               37458: 0.599,\n",
       "               10271: 0.597,\n",
       "               1559: 0.593,\n",
       "               1723: 0.719,\n",
       "               6351: 0.595,\n",
       "               2167: 0.587,\n",
       "               25397: 0.676,\n",
       "               18679: 0.612,\n",
       "               23199: 0.665,\n",
       "               48470: 0.573,\n",
       "               525: 0.551,\n",
       "               4197: 0.714,\n",
       "               8664: 0.678,\n",
       "               3637: 0.67,\n",
       "               9736: 0.537,\n",
       "               11829: 0.502,\n",
       "               1818: 0.501,\n",
       "               38003: 0.613,\n",
       "               14967: 0.554,\n",
       "               38363: 0.554,\n",
       "               40679: 0.684,\n",
       "               38883: 0.678,\n",
       "               28891: 0.667,\n",
       "               14719: 0.572,\n",
       "               31933: 0.571,\n",
       "               41120: 0.557,\n",
       "               11933: 0.633,\n",
       "               306: 0.596,\n",
       "               618: 0.583,\n",
       "               38297: 0.599,\n",
       "               14833: 0.607,\n",
       "               2473: 0.554,\n",
       "               18497: 0.53,\n",
       "               3731: 0.518,\n",
       "               32937: 0.781,\n",
       "               32380: 0.623,\n",
       "               37210: 0.616,\n",
       "               38759: 0.553,\n",
       "               934: 0.636,\n",
       "               33191: 0.544,\n",
       "               25777: 0.68,\n",
       "               28623: 0.604,\n",
       "               355: 0.595,\n",
       "               37126: 0.82,\n",
       "               37988: 0.791,\n",
       "               40907: 0.769,\n",
       "               32212: 0.648,\n",
       "               1844: 0.631,\n",
       "               3778: 0.649,\n",
       "               20185: 0.582,\n",
       "               2742: 0.564,\n",
       "               33654: 0.653,\n",
       "               2771: 0.653,\n",
       "               4744: 0.619,\n",
       "               5114: 0.795,\n",
       "               430: 0.669,\n",
       "               6421: 0.657,\n",
       "               11887: 0.72,\n",
       "               41006: 0.752,\n",
       "               14811: 0.69,\n",
       "               36702: 0.633,\n",
       "               35760: 0.669,\n",
       "               35828: 0.561,\n",
       "               5734: 0.559,\n",
       "               40542: 0.534,\n",
       "               3806: 0.647,\n",
       "               6045: 0.64,\n",
       "               38787: 0.64,\n",
       "               5027: 0.657,\n",
       "               41102: 0.61,\n",
       "               182: 0.634,\n",
       "               288: 0.647,\n",
       "               10360: 0.667,\n",
       "               10528: 0.617,\n",
       "               32182: 0.712,\n",
       "               31442: 0.619,\n",
       "               24701: 0.573,\n",
       "               12445: 0.654,\n",
       "               849: 0.629,\n",
       "               28423: 0.592,\n",
       "               688: 0.608,\n",
       "               45572: 0.591,\n",
       "               10119: 0.588,\n",
       "               34338: 0.463,\n",
       "               36407: 0.462,\n",
       "               1638: 0.458,\n",
       "               24439: 0.621,\n",
       "               33478: 0.609,\n",
       "               35376: 0.729,\n",
       "               66: 0.699,\n",
       "               15379: 0.651,\n",
       "               42205: 0.733,\n",
       "               38481: 0.6,\n",
       "               43814: 0.575,\n",
       "               2916: 0.62,\n",
       "               404: 0.59,\n",
       "               6682: 0.647,\n",
       "               842: 0.752,\n",
       "               646: 0.663,\n",
       "               28979: 0.683,\n",
       "               31240: 0.574,\n",
       "               31610: 0.553,\n",
       "               40776: 0.751,\n",
       "               20583: 0.667,\n",
       "               33031: 0.61,\n",
       "               38276: 0.577,\n",
       "               584: 0.575,\n",
       "               30831: 0.781,\n",
       "               5821: 0.565,\n",
       "               22297: 0.762,\n",
       "               2108: 0.561,\n",
       "               27943: 0.552,\n",
       "               16498: 0.68,\n",
       "               38524: 0.649,\n",
       "               6347: 0.687,\n",
       "               17549: 0.67,\n",
       "               3228: 0.664,\n",
       "               35507: 0.687,\n",
       "               34504: 0.611,\n",
       "               1222: 0.655,\n",
       "               36214: 0.634,\n",
       "               195: 0.633,\n",
       "               27833: 0.646,\n",
       "               15565: 0.56,\n",
       "               11859: 0.547,\n",
       "               1455: 0.539,\n",
       "               27631: 0.531,\n",
       "               27831: 0.646,\n",
       "               1840: 0.605,\n",
       "               37657: 0.602,\n",
       "               15059: 0.641,\n",
       "               18: 0.598,\n",
       "               34314: 0.579,\n",
       "               20931: 0.572,\n",
       "               2702: 0.641,\n",
       "               73: 0.725,\n",
       "               72: 0.664,\n",
       "               71: 0.619,\n",
       "               7590: 0.634,\n",
       "               31221: 0.63,\n",
       "               3372: 0.617,\n",
       "               6385: 0.607,\n",
       "               6275: 0.606,\n",
       "               32105: 0.644,\n",
       "               1705: 0.608,\n",
       "               36847: 0.6,\n",
       "               185: 0.665,\n",
       "               6864: 0.599,\n",
       "               31018: 0.597,\n",
       "               1603: 0.648,\n",
       "               3294: 0.63,\n",
       "               180: 0.624,\n",
       "               16592: 0.61,\n",
       "               1136: 0.541,\n",
       "               5682: 0.523,\n",
       "               10465: 0.539,\n",
       "               6112: 0.528,\n",
       "               38573: 0.523,\n",
       "               36456: 0.653,\n",
       "               22: 0.604,\n",
       "               33929: 0.604,\n",
       "               34104: 0.65,\n",
       "               47391: 0.642,\n",
       "               23067: 0.632,\n",
       "               38186: 0.641,\n",
       "               850: 0.635,\n",
       "               5962: 0.669,\n",
       "               18097: 0.616,\n",
       "               16894: 0.774,\n",
       "               38408: 0.644,\n",
       "               1557: 0.619,\n",
       "               38390: 0.616,\n",
       "               7666: 0.622,\n",
       "               762: 0.613,\n",
       "               4667: 0.573,\n",
       "               17247: 0.564,\n",
       "               356: 0.547,\n",
       "               2494: 0.621,\n",
       "               36902: 0.587,\n",
       "               36266: 0.605,\n",
       "               36432: 0.75,\n",
       "               12531: 0.638,\n",
       "               20541: 0.614,\n",
       "               5034: 0.6,\n",
       "               2734: 0.607,\n",
       "               34411: 0.596,\n",
       "               40748: 0.595,\n",
       "               1891: 0.509,\n",
       "               2576: 0.505,\n",
       "               29067: 0.64,\n",
       "               32949: 0.587,\n",
       "               36864: 0.561,\n",
       "               35608: 0.646,\n",
       "               39783: 0.637,\n",
       "               34822: 0.622,\n",
       "               819: 0.609,\n",
       "               8768: 0.587,\n",
       "               32901: 0.605,\n",
       "               1604: 0.598,\n",
       "               11359: 0.756,\n",
       "               5658: 0.737,\n",
       "               41491: 0.73,\n",
       "               2253: 0.598,\n",
       "               1775: 0.63,\n",
       "               1356: 0.587,\n",
       "               452: 0.679,\n",
       "               727: 0.635,\n",
       "               496: 0.659,\n",
       "               2924: 0.621,\n",
       "               29974: 0.641,\n",
       "               26243: 0.63,\n",
       "               34389: 0.616,\n",
       "               157: 0.605,\n",
       "               31793: 0.597,\n",
       "               4789: 0.559,\n",
       "               15039: 0.549,\n",
       "               777: 0.751,\n",
       "               27: 0.662,\n",
       "               41265: 0.601,\n",
       "               9253: 0.666,\n",
       "               11577: 0.659,\n",
       "               10521: 0.755,\n",
       "               31715: 0.62,\n",
       "               42250: 0.582,\n",
       "               34662: 0.703,\n",
       "               25519: 0.655,\n",
       "               16574: 0.633,\n",
       "               34866: 0.642,\n",
       "               1668: 0.647,\n",
       "               1669: 0.638,\n",
       "               43756: 0.635,\n",
       "               34299: 0.707,\n",
       "               25687: 0.689,\n",
       "               6793: 0.567,\n",
       "               31812: 0.525,\n",
       "               30885: 0.679,\n",
       "               20507: 0.65,\n",
       "               553: 0.644,\n",
       "               14813: 0.657,\n",
       "               10460: 0.63,\n",
       "               15911: 0.618,\n",
       "               9120: 0.648,\n",
       "               8215: 0.626,\n",
       "               4211: 0.615,\n",
       "               28831: 0.584,\n",
       "               39324: 0.582,\n",
       "               10516: 0.577,\n",
       "               6033: 0.748,\n",
       "               14837: 0.716,\n",
       "               905: 0.643,\n",
       "               27899: 0.66,\n",
       "               39567: 0.657,\n",
       "               28977: 0.675,\n",
       "               12347: 0.587,\n",
       "               3218: 0.586,\n",
       "               19855: 0.601,\n",
       "               4155: 0.583,\n",
       "               32189: 0.58,\n",
       "               7472: 0.675,\n",
       "               35843: 0.558,\n",
       "               131: 0.556,\n",
       "               3407: 0.594,\n",
       "               42145: 0.574,\n",
       "               35247: 0.721,\n",
       "               37964: 0.585,\n",
       "               9922: 0.584,\n",
       "               34414: 0.538,\n",
       "               35889: 0.547,\n",
       "               3509: 0.513,\n",
       "               2704: 0.573,\n",
       "               344: 0.606,\n",
       "               6895: 0.601,\n",
       "               3023: 0.653,\n",
       "               38440: 0.647,\n",
       "               1866: 0.646,\n",
       "               8407: 0.695,\n",
       "               2044: 0.672,\n",
       "               17729: 0.61,\n",
       "               10162: 0.58,\n",
       "               38797: 0.558,\n",
       "               102: 0.539,\n",
       "               5940: 0.566,\n",
       "               690: 0.551,\n",
       "               6324: 0.633,\n",
       "               11837: 0.631,\n",
       "               8410: 0.619,\n",
       "               240: 0.605,\n",
       "               30191: 0.601,\n",
       "               12899: 0.589,\n",
       "               32268: 0.63,\n",
       "               23709: 0.588,\n",
       "               22359: 0.567,\n",
       "               1531: 0.55,\n",
       "               35790: 0.638,\n",
       "               40783: 0.587,\n",
       "               33045: 0.574,\n",
       "               32595: 0.658,\n",
       "               30736: 0.652,\n",
       "               28833: 0.581,\n",
       "               25939: 0.556,\n",
       "               35997: 0.541,\n",
       "               1936: 0.539,\n",
       "               31989: 0.896,\n",
       "               31988: 0.777,\n",
       "               35082: 0.689,\n",
       "               5163: 0.629,\n",
       "               1575: 0.629,\n",
       "               34440: 0.581,\n",
       "               15611: 0.623,\n",
       "               37476: 0.619,\n",
       "               21729: 0.61,\n",
       "               23301: 0.668,\n",
       "               1120: 0.663,\n",
       "               18277: 0.61,\n",
       "               37451: 0.603,\n",
       "               32962: 0.597,\n",
       "               2397: 0.632,\n",
       "               37601: 0.59,\n",
       "               37171: 0.641,\n",
       "               21773: 0.619,\n",
       "               41392: 0.627,\n",
       "               4936: 0.626,\n",
       "               3652: 0.698,\n",
       "               2031: 0.633,\n",
       "               32870: 0.717,\n",
       "               32869: 0.69,\n",
       "               1530: 0.611,\n",
       "               1214: 0.606,\n",
       "               38083: 0.584,\n",
       "               35251: 0.576,\n",
       "               25537: 0.678,\n",
       "               298: 0.603,\n",
       "               23321: 0.582,\n",
       "               38472: 0.575,\n",
       "               1338: 0.598,\n",
       "               32136: 0.561,\n",
       "               1519: 0.519,\n",
       "               37991: 0.508,\n",
       "               901: 0.814,\n",
       "               502: 0.813,\n",
       "               7436: 0.782,\n",
       "               3603: 0.634,\n",
       "               2559: 0.634,\n",
       "               9181: 0.639,\n",
       "               197: 0.608,\n",
       "               11013: 0.579,\n",
       "               893: 0.814,\n",
       "               6714: 0.799,\n",
       "               899: 0.788,\n",
       "               898: 0.694,\n",
       "               897: 0.692,\n",
       "               813: 0.748,\n",
       "               894: 0.652,\n",
       "               22699: 0.631,\n",
       "               2810: 0.611,\n",
       "               39617: 0.576,\n",
       "               23623: 0.561,\n",
       "               14093: 0.635,\n",
       "               33253: 0.566,\n",
       "               1451: 0.623,\n",
       "               38097: 0.607,\n",
       "               2335: 0.605,\n",
       "               4548: 0.599,\n",
       "               41619: 0.583,\n",
       "               37450: 0.582,\n",
       "               45560: 0.608,\n",
       "               13469: 0.668,\n",
       "               14073: 0.581,\n",
       "               7619: 0.564,\n",
       "               28735: 0.556,\n",
       "               665: 0.55,\n",
       "               16700: 0.662,\n",
       "               11111: 0.636,\n",
       "               24347: 0.591,\n",
       "               5994: 0.584,\n",
       "               37283: 0.575,\n",
       "               6375: 0.569,\n",
       "               31181: 0.721,\n",
       "               36999: 0.613,\n",
       "               5941: 0.582,\n",
       "               27837: 0.575,\n",
       "               8937: 0.555,\n",
       "               37510: 0.712,\n",
       "               28999: 0.562,\n",
       "               9015: 0.557,\n",
       "               687: 0.547,\n",
       "               35363: 0.537,\n",
       "               34103: 0.697,\n",
       "               36029: 0.608,\n",
       "               19759: 0.65,\n",
       "               5956: 0.621,\n",
       "               21435: 0.605,\n",
       "               24823: 0.6,\n",
       "               1890: 0.578,\n",
       "               18275: 0.575,\n",
       "               956: 0.608,\n",
       "               4581: 0.606,\n",
       "               3420: 0.599,\n",
       "               1452: 0.597,\n",
       "               877: 0.607,\n",
       "               19585: 0.605,\n",
       "               5629: 0.553,\n",
       "               38062: 0.549,\n",
       "               10622: 0.614,\n",
       "               4028: 0.564,\n",
       "               153: 0.627,\n",
       "               27787: 0.626,\n",
       "               36816: 0.567,\n",
       "               16822: 0.566,\n",
       "               3252: 0.56,\n",
       "               15689: 0.616,\n",
       "               10087: 0.698,\n",
       "               23847: 0.657,\n",
       "               32648: 0.613,\n",
       "               8424: 0.608,\n",
       "               10897: 0.765,\n",
       "               9519: 0.663,\n",
       "               35712: 0.65,\n",
       "               550: 0.679,\n",
       "               28771: 0.673,\n",
       "               916: 0.55,\n",
       "               34076: 0.549,\n",
       "               25879: 0.62,\n",
       "               30503: 0.597,\n",
       "               1091: 0.585,\n",
       "               2114: 0.584,\n",
       "               6634: 0.569,\n",
       "               36817: 0.602,\n",
       "               30757: 0.581,\n",
       "               36943: 0.578,\n",
       "               1122: 0.561,\n",
       "               6945: 0.533,\n",
       "               10017: 0.71,\n",
       "               121: 0.795,\n",
       "               1293: 0.627,\n",
       "               3655: 0.567,\n",
       "               24415: 0.774,\n",
       "               2605: 0.588}),\n",
       "  'anime_encoder': defaultdict(float,\n",
       "              {2598: 0.995,\n",
       "               3194: 0.988,\n",
       "               764: 0.988,\n",
       "               4898: 0.996,\n",
       "               4038: 0.994,\n",
       "               2161: 0.994,\n",
       "               1609: 0.987,\n",
       "               3154: 0.987,\n",
       "               31052: 0.987,\n",
       "               16201: 0.99,\n",
       "               9513: 0.994,\n",
       "               8074: 0.995,\n",
       "               33836: 0.984,\n",
       "               34076: 0.986,\n",
       "               34176: 0.977,\n",
       "               2084: 0.993,\n",
       "               1818: 0.992,\n",
       "               2816: 0.994,\n",
       "               23321: 0.982,\n",
       "               26055: 0.976,\n",
       "               33352: 0.956,\n",
       "               31706: 0.989,\n",
       "               15085: 0.989,\n",
       "               29854: 0.99,\n",
       "               36949: 0.97,\n",
       "               120: 0.979,\n",
       "               22: 0.965,\n",
       "               9496: 0.987,\n",
       "               38094: 0.985,\n",
       "               7619: 0.985,\n",
       "               31758: 0.993,\n",
       "               2985: 0.987,\n",
       "               1920: 0.986,\n",
       "               31757: 0.993,\n",
       "               15689: 0.995,\n",
       "               1180: 0.99,\n",
       "               687: 0.989,\n",
       "               1563: 0.99,\n",
       "               185: 0.938,\n",
       "               12365: 0.902,\n",
       "               186: 0.9,\n",
       "               210: 0.982,\n",
       "               392: 0.992,\n",
       "               249: 0.995,\n",
       "               1498: 0.995,\n",
       "               2916: 0.994,\n",
       "               17913: 0.687,\n",
       "               36144: 0.678,\n",
       "               21431: 0.938,\n",
       "               169: 0.928,\n",
       "               2603: 0.927,\n",
       "               2001: 0.995,\n",
       "               2284: 0.994,\n",
       "               1123: 0.985,\n",
       "               8: 0.984,\n",
       "               1708: 0.983,\n",
       "               11741: 0.987,\n",
       "               32867: 0.984,\n",
       "               34973: 0.965,\n",
       "               10308: 0.989,\n",
       "               3594: 0.992,\n",
       "               10490: 0.989,\n",
       "               31: 0.963,\n",
       "               32: 0.961,\n",
       "               27: 0.958,\n",
       "               30: 0.961,\n",
       "               2759: 0.953,\n",
       "               159: 0.993,\n",
       "               11521: 0.992,\n",
       "               269: 0.992,\n",
       "               1482: 0.995,\n",
       "               30484: 0.984,\n",
       "               28221: 0.971,\n",
       "               34438: 0.981,\n",
       "               34439: 0.976,\n",
       "               15201: 0.976,\n",
       "               1313: 0.99,\n",
       "               1132: 0.99,\n",
       "               31318: 0.987,\n",
       "               7059: 0.981,\n",
       "               36214: 0.98,\n",
       "               2739: 0.984,\n",
       "               347: 0.992,\n",
       "               5862: 0.979,\n",
       "               2471: 0.978,\n",
       "               4485: 0.978,\n",
       "               13663: 0.996,\n",
       "               2980: 0.993,\n",
       "               7791: 0.993,\n",
       "               1230: 0.982,\n",
       "               833: 0.99,\n",
       "               7711: 0.982,\n",
       "               911: 0.99,\n",
       "               33185: 0.988,\n",
       "               6028: 0.989,\n",
       "               35843: 0.992,\n",
       "               9969: 0.992,\n",
       "               14075: 0.996,\n",
       "               30901: 0.987,\n",
       "               41025: 0.971,\n",
       "               36927: 0.97,\n",
       "               5114: 0.965,\n",
       "               4066: 0.964,\n",
       "               23945: 0.972,\n",
       "               2620: 0.995,\n",
       "               607: 0.993,\n",
       "               11785: 0.993,\n",
       "               8915: 0.994,\n",
       "               4550: 0.969,\n",
       "               23325: 0.974,\n",
       "               37982: 0.968,\n",
       "               37215: 0.855,\n",
       "               19489: 0.793,\n",
       "               14537: 0.778,\n",
       "               2109: 0.979,\n",
       "               4962: 0.976,\n",
       "               2819: 0.976,\n",
       "               2619: 0.995,\n",
       "               178: 0.994,\n",
       "               918: 0.994,\n",
       "               1002: 0.99,\n",
       "               36106: 0.99,\n",
       "               2906: 0.99,\n",
       "               31704: 0.976,\n",
       "               34134: 0.973,\n",
       "               3358: 0.973,\n",
       "               24133: 0.982,\n",
       "               24873: 0.985,\n",
       "               31716: 0.981,\n",
       "               5074: 0.994,\n",
       "               26243: 0.989,\n",
       "               35248: 0.988,\n",
       "               24909: 0.984,\n",
       "               40938: 0.947,\n",
       "               28819: 0.946,\n",
       "               2248: 0.939,\n",
       "               27441: 0.99,\n",
       "               28977: 0.994,\n",
       "               2167: 0.982,\n",
       "               10087: 0.99,\n",
       "               37202: 0.975,\n",
       "               27831: 0.985,\n",
       "               23199: 0.993,\n",
       "               2723: 0.993,\n",
       "               25429: 0.994,\n",
       "               16894: 0.992,\n",
       "               25183: 0.991,\n",
       "               22789: 0.985,\n",
       "               14653: 0.992,\n",
       "               5027: 0.984,\n",
       "               14719: 0.992,\n",
       "               36726: 0.984,\n",
       "               32105: 0.995,\n",
       "               32370: 0.992,\n",
       "               907: 0.979,\n",
       "               330: 0.979,\n",
       "               28497: 0.962,\n",
       "               28121: 0.962,\n",
       "               32937: 0.961,\n",
       "               18411: 0.987,\n",
       "               15751: 0.996,\n",
       "               17505: 0.995,\n",
       "               34979: 0.861,\n",
       "               9510: 0.857,\n",
       "               8538: 0.85,\n",
       "               33184: 0.979,\n",
       "               33475: 0.978,\n",
       "               35757: 0.976,\n",
       "               3972: 0.985,\n",
       "               1333: 0.974,\n",
       "               4565: 0.974,\n",
       "               2494: 0.993,\n",
       "               2037: 0.993,\n",
       "               1554: 0.993,\n",
       "               16512: 0.978,\n",
       "               14811: 0.986,\n",
       "               18195: 0.976,\n",
       "               25777: 0.979,\n",
       "               35760: 0.99,\n",
       "               48: 0.972,\n",
       "               1082: 0.976,\n",
       "               63: 0.967,\n",
       "               3464: 0.993,\n",
       "               3455: 0.996,\n",
       "               2924: 0.992,\n",
       "               1542: 0.993,\n",
       "               3561: 0.993,\n",
       "               8375: 0.993,\n",
       "               495: 0.99,\n",
       "               5940: 0.991,\n",
       "               32182: 0.97,\n",
       "               31442: 0.945,\n",
       "               11061: 0.944,\n",
       "               849: 0.969,\n",
       "               853: 0.956,\n",
       "               4382: 0.956,\n",
       "               30384: 0.468,\n",
       "               37962: 0.422,\n",
       "               40904: 0.413,\n",
       "               31629: 0.974,\n",
       "               6129: 0.974,\n",
       "               290: 0.972,\n",
       "               24439: 0.982,\n",
       "               20541: 0.975,\n",
       "               5958: 0.975,\n",
       "               280: 0.993,\n",
       "               34148: 0.993,\n",
       "               2246: 0.988,\n",
       "               1573: 0.988,\n",
       "               1524: 0.987,\n",
       "               31227: 0.795,\n",
       "               28869: 0.784,\n",
       "               19111: 0.783,\n",
       "               20583: 0.93,\n",
       "               40372: 0.917,\n",
       "               35111: 0.913,\n",
       "               10588: 0.979,\n",
       "               12883: 0.99,\n",
       "               11285: 0.99,\n",
       "               37210: 0.975,\n",
       "               25157: 0.973,\n",
       "               22297: 0.95,\n",
       "               1460: 0.938,\n",
       "               1459: 0.936,\n",
       "               3609: 0.988,\n",
       "               16498: 0.991,\n",
       "               34933: 0.961,\n",
       "               17729: 0.96,\n",
       "               2093: 0.984,\n",
       "               36840: 0.965,\n",
       "               28677: 0.965,\n",
       "               36943: 0.963,\n",
       "               20507: 0.995,\n",
       "               11499: 0.99,\n",
       "               18679: 0.992,\n",
       "               14345: 0.993,\n",
       "               33506: 0.982,\n",
       "               1771: 0.941,\n",
       "               1942: 0.941,\n",
       "               5355: 0.94,\n",
       "               18: 0.863,\n",
       "               4903: 0.993,\n",
       "               73: 0.988,\n",
       "               71: 0.987,\n",
       "               29974: 0.986,\n",
       "               25859: 0.99,\n",
       "               23269: 0.989,\n",
       "               33338: 0.993,\n",
       "               6385: 0.992,\n",
       "               1003: 0.992,\n",
       "               395: 0.927,\n",
       "               3931: 0.925,\n",
       "               5231: 0.968,\n",
       "               14131: 0.967,\n",
       "               516: 0.986,\n",
       "               25397: 0.981,\n",
       "               8516: 0.98,\n",
       "               3901: 0.986,\n",
       "               20709: 0.985,\n",
       "               1468: 0.984,\n",
       "               6375: 0.989,\n",
       "               36456: 0.92,\n",
       "               35075: 0.915,\n",
       "               40935: 0.915,\n",
       "               26441: 0.988,\n",
       "               32189: 0.977,\n",
       "               31580: 0.973,\n",
       "               11751: 0.97,\n",
       "               10360: 0.978,\n",
       "               5962: 0.978,\n",
       "               35180: 0.98,\n",
       "               17074: 0.969,\n",
       "               291: 0.968,\n",
       "               11703: 0.993,\n",
       "               32696: 0.99,\n",
       "               11977: 0.989,\n",
       "               1118: 0.988,\n",
       "               2390: 0.988,\n",
       "               2702: 0.988,\n",
       "               31414: 0.983,\n",
       "               23311: 0.982,\n",
       "               298: 0.993,\n",
       "               4654: 0.976,\n",
       "               6707: 0.972,\n",
       "               481: 0.971,\n",
       "               6774: 0.959,\n",
       "               32673: 0.958,\n",
       "               16706: 0.958,\n",
       "               12293: 0.98,\n",
       "               15613: 0.985,\n",
       "               29067: 0.988,\n",
       "               22123: 0.988,\n",
       "               20847: 0.988,\n",
       "               11179: 0.975,\n",
       "               14741: 0.969,\n",
       "               38329: 0.967,\n",
       "               36649: 0.987,\n",
       "               5420: 0.995,\n",
       "               6895: 0.992,\n",
       "               6033: 0.952,\n",
       "               11359: 0.95,\n",
       "               2781: 0.939,\n",
       "               1337: 0.992,\n",
       "               20: 0.991,\n",
       "               1775: 0.995,\n",
       "               182: 0.996,\n",
       "               19: 0.97,\n",
       "               7334: 0.97,\n",
       "               13261: 0.969,\n",
       "               777: 0.99,\n",
       "               9253: 0.984,\n",
       "               17651: 0.978,\n",
       "               10521: 0.986,\n",
       "               31715: 0.984,\n",
       "               18119: 0.983,\n",
       "               6633: 0.963,\n",
       "               4719: 0.962,\n",
       "               32979: 0.96,\n",
       "               15489: 0.985,\n",
       "               355: 0.984,\n",
       "               605: 0.994,\n",
       "               2747: 0.994,\n",
       "               3399: 0.994,\n",
       "               14813: 0.943,\n",
       "               14967: 0.941,\n",
       "               31884: 0.984,\n",
       "               225: 0.98,\n",
       "               223: 0.978,\n",
       "               14837: 0.977,\n",
       "               36511: 0.982,\n",
       "               36838: 0.991,\n",
       "               7472: 0.994,\n",
       "               32214: 0.991,\n",
       "               37614: 0.985,\n",
       "               17513: 0.981,\n",
       "               18523: 0.987,\n",
       "               447: 0.99,\n",
       "               1472: 0.99,\n",
       "               31680: 0.984,\n",
       "               30503: 0.984,\n",
       "               878: 0.989,\n",
       "               8407: 0.989,\n",
       "               7088: 0.989,\n",
       "               30127: 0.984,\n",
       "               15911: 0.982,\n",
       "               24031: 0.982,\n",
       "               14713: 0.993,\n",
       "               9181: 0.995,\n",
       "               11597: 0.995,\n",
       "               5081: 0.994,\n",
       "               4334: 0.994,\n",
       "               8311: 0.996,\n",
       "               5753: 0.893,\n",
       "               22877: 0.892,\n",
       "               6747: 0.888,\n",
       "               12413: 0.993,\n",
       "               10821: 0.989,\n",
       "               24705: 0.988,\n",
       "               32595: 0.987,\n",
       "               31339: 0.986,\n",
       "               31147: 0.97,\n",
       "               4744: 0.977,\n",
       "               17777: 0.986,\n",
       "               1575: 0.97,\n",
       "               15611: 0.986,\n",
       "               9493: 0.984,\n",
       "               31592: 0.993,\n",
       "               6682: 0.992,\n",
       "               10620: 0.99,\n",
       "               2397: 0.991,\n",
       "               34299: 0.991,\n",
       "               6374: 0.991,\n",
       "               934: 0.984,\n",
       "               3342: 0.983,\n",
       "               1810: 0.983,\n",
       "               31845: 0.979,\n",
       "               28623: 0.966,\n",
       "               19703: 0.979,\n",
       "               36754: 0.978,\n",
       "               10080: 0.977,\n",
       "               28297: 0.977,\n",
       "               33: 0.986,\n",
       "               5760: 0.986,\n",
       "               902: 0.992,\n",
       "               901: 0.992,\n",
       "               12279: 0.991,\n",
       "               9919: 0.994,\n",
       "               469: 0.995,\n",
       "               893: 0.992,\n",
       "               906: 0.99,\n",
       "               7435: 0.989,\n",
       "               899: 0.987,\n",
       "               37426: 0.983,\n",
       "               1536: 0.983,\n",
       "               38671: 0.967,\n",
       "               1932: 0.983,\n",
       "               28979: 0.988,\n",
       "               2130: 0.996,\n",
       "               1604: 0.994,\n",
       "               15225: 0.985,\n",
       "               9289: 0.977,\n",
       "               31933: 0.989,\n",
       "               104: 0.989,\n",
       "               13145: 0.895,\n",
       "               10527: 0.893,\n",
       "               5973: 0.889,\n",
       "               21437: 0.98,\n",
       "               1707: 0.965,\n",
       "               1842: 0.965,\n",
       "               34881: 0.986,\n",
       "               31793: 0.988,\n",
       "               21353: 0.988,\n",
       "               16592: 0.977,\n",
       "               4186: 0.975,\n",
       "               24823: 0.784,\n",
       "               36904: 0.77,\n",
       "               13863: 0.768,\n",
       "               16524: 0.988,\n",
       "               10330: 0.982,\n",
       "               3588: 0.983,\n",
       "               7768: 0.983,\n",
       "               20899: 0.983,\n",
       "               32936: 0.989,\n",
       "               4411: 0.991,\n",
       "               24543: 0.99,\n",
       "               727: 0.993,\n",
       "               17831: 0.98,\n",
       "               8769: 0.988,\n",
       "               15379: 0.988,\n",
       "               6512: 0.988,\n",
       "               5476: 0.99,\n",
       "               58: 0.99,\n",
       "               1000: 0.99,\n",
       "               4262: 0.982,\n",
       "               1094: 0.989,\n",
       "               1566: 0.988,\n",
       "               34096: 0.991,\n",
       "               558: 0.977,\n",
       "               8630: 0.976,\n",
       "               100: 0.985,\n",
       "               2333: 0.984,\n",
       "               331: 0.984,\n",
       "               892: 0.989,\n",
       "               121: 0.992,\n",
       "               24415: 0.992,\n",
       "               2404: 0.993,\n",
       "               2253: 0.993,\n",
       "               2254: 0.993}),\n",
       "  'anime_encoder_bigger': defaultdict(float,\n",
       "              {696: 0.994,\n",
       "               285: 0.992,\n",
       "               538: 0.992,\n",
       "               3457: 0.995,\n",
       "               4752: 0.995,\n",
       "               4151: 0.995,\n",
       "               15201: 0.997,\n",
       "               38850: 0.995,\n",
       "               513: 0.995,\n",
       "               29854: 0.994,\n",
       "               24439: 0.994,\n",
       "               31147: 0.992,\n",
       "               35376: 0.982,\n",
       "               35220: 0.981,\n",
       "               34547: 0.984,\n",
       "               242: 0.996,\n",
       "               25: 0.995,\n",
       "               29: 0.994,\n",
       "               25867: 0.993,\n",
       "               22729: 0.992,\n",
       "               31174: 0.99,\n",
       "               27833: 0.991,\n",
       "               31706: 0.994,\n",
       "               24455: 0.99,\n",
       "               29976: 0.994,\n",
       "               28883: 0.994,\n",
       "               43349: 0.625,\n",
       "               30842: 0.624,\n",
       "               36903: 0.615,\n",
       "               4280: 0.994,\n",
       "               393: 0.993,\n",
       "               757: 0.993,\n",
       "               9260: 0.993,\n",
       "               4106: 0.992,\n",
       "               2890: 0.994,\n",
       "               21: 0.992,\n",
       "               120: 0.989,\n",
       "               1065: 0.99,\n",
       "               2014: 0.985,\n",
       "               3297: 0.984,\n",
       "               15045: 0.983,\n",
       "               62: 0.992,\n",
       "               57: 0.992,\n",
       "               2104: 0.996,\n",
       "               813: 0.994,\n",
       "               1565: 0.986,\n",
       "               6033: 0.994,\n",
       "               570: 0.992,\n",
       "               21339: 0.992,\n",
       "               212: 0.992,\n",
       "               10389: 0.989,\n",
       "               39569: 0.99,\n",
       "               38826: 0.986,\n",
       "               1123: 0.996,\n",
       "               8: 0.996,\n",
       "               250: 0.994,\n",
       "               161: 0.991,\n",
       "               4548: 0.991,\n",
       "               27831: 0.992,\n",
       "               30091: 0.992,\n",
       "               24703: 0.991,\n",
       "               10572: 0.989,\n",
       "               17651: 0.989,\n",
       "               6758: 0.988,\n",
       "               165: 0.994,\n",
       "               31: 0.994,\n",
       "               159: 0.995,\n",
       "               967: 0.994,\n",
       "               19291: 0.992,\n",
       "               31592: 0.995,\n",
       "               12189: 0.988,\n",
       "               7769: 0.987,\n",
       "               13599: 0.991,\n",
       "               2904: 0.996,\n",
       "               6573: 0.995,\n",
       "               1142: 0.995,\n",
       "               961: 0.994,\n",
       "               116: 0.994,\n",
       "               101: 0.996,\n",
       "               1230: 0.993,\n",
       "               16524: 0.994,\n",
       "               879: 0.993,\n",
       "               38040: 0.995,\n",
       "               597: 0.993,\n",
       "               572: 0.993,\n",
       "               13535: 0.996,\n",
       "               9919: 0.995,\n",
       "               416: 0.99,\n",
       "               7472: 0.99,\n",
       "               431: 0.989,\n",
       "               7791: 0.997,\n",
       "               550: 0.993,\n",
       "               482: 0.993,\n",
       "               28211: 0.993,\n",
       "               38464: 0.993,\n",
       "               1931: 0.994,\n",
       "               462: 0.993,\n",
       "               2454: 0.993,\n",
       "               16498: 0.994,\n",
       "               14227: 0.994,\n",
       "               15809: 0.994,\n",
       "               2789: 0.994,\n",
       "               10330: 0.991,\n",
       "               37370: 0.738,\n",
       "               37994: 0.674,\n",
       "               29860: 0.505,\n",
       "               1554: 0.994,\n",
       "               1050: 0.994,\n",
       "               11785: 0.995,\n",
       "               5684: 0.994,\n",
       "               2259: 0.993,\n",
       "               7645: 0.993,\n",
       "               32673: 0.989,\n",
       "               31952: 0.989,\n",
       "               31646: 0.992,\n",
       "               28537: 0.991,\n",
       "               28497: 0.989,\n",
       "               30947: 0.989,\n",
       "               592: 0.995,\n",
       "               1672: 0.994,\n",
       "               1892: 0.994,\n",
       "               2619: 0.993,\n",
       "               2156: 0.993,\n",
       "               10049: 0.994,\n",
       "               6747: 0.994,\n",
       "               622: 0.994,\n",
       "               34658: 0.994,\n",
       "               16700: 0.994,\n",
       "               22199: 0.994,\n",
       "               28977: 0.993,\n",
       "               22043: 0.992,\n",
       "               23209: 0.993,\n",
       "               21881: 0.992,\n",
       "               23289: 0.994,\n",
       "               5914: 0.991,\n",
       "               3588: 0.993,\n",
       "               6500: 0.99,\n",
       "               12119: 0.995,\n",
       "               14355: 0.99,\n",
       "               28617: 0.99,\n",
       "               38254: 0.978,\n",
       "               35828: 0.978,\n",
       "               38776: 0.977,\n",
       "               1210: 0.985,\n",
       "               16067: 0.985,\n",
       "               2562: 0.985,\n",
       "               27829: 0.991,\n",
       "               23309: 0.994,\n",
       "               855: 0.994,\n",
       "               198: 0.994,\n",
       "               197: 0.994,\n",
       "               1807: 0.994,\n",
       "               1570: 0.993,\n",
       "               1498: 0.992,\n",
       "               26443: 0.993,\n",
       "               31318: 0.989,\n",
       "               31478: 0.99,\n",
       "               31173: 0.98,\n",
       "               31610: 0.971,\n",
       "               122: 0.992,\n",
       "               558: 0.991,\n",
       "               37447: 0.966,\n",
       "               36906: 0.985,\n",
       "               33889: 0.963,\n",
       "               32360: 0.993,\n",
       "               31452: 0.992,\n",
       "               11759: 0.993,\n",
       "               18001: 0.993,\n",
       "               464: 0.991,\n",
       "               2013: 0.991,\n",
       "               34299: 0.99,\n",
       "               32105: 0.984,\n",
       "               32483: 0.983,\n",
       "               33668: 0.981,\n",
       "               39198: 0.987,\n",
       "               38860: 0.987,\n",
       "               38691: 0.984,\n",
       "               15417: 0.994,\n",
       "               18411: 0.994,\n",
       "               17505: 0.994,\n",
       "               37117: 0.428,\n",
       "               40323: 0.427,\n",
       "               37450: 0.426,\n",
       "               34636: 0.986,\n",
       "               34280: 0.984,\n",
       "               32878: 0.983,\n",
       "               20689: 0.984,\n",
       "               21995: 0.983,\n",
       "               11371: 0.99,\n",
       "               6067: 0.994,\n",
       "               98: 0.994,\n",
       "               178: 0.995,\n",
       "               6702: 0.995,\n",
       "               16405: 0.858,\n",
       "               10719: 0.852,\n",
       "               24133: 0.844,\n",
       "               39799: 0.982,\n",
       "               36587: 0.978,\n",
       "               37958: 0.977,\n",
       "               30383: 0.987,\n",
       "               32608: 0.987,\n",
       "               6166: 0.996,\n",
       "               1808: 0.994,\n",
       "               79: 0.994,\n",
       "               13231: 0.987,\n",
       "               10847: 0.986,\n",
       "               9712: 0.986,\n",
       "               582: 0.991,\n",
       "               1675: 0.991,\n",
       "               37964: 0.985,\n",
       "               38797: 0.982,\n",
       "               39798: 0.98,\n",
       "               1441: 0.993,\n",
       "               8950: 0.993,\n",
       "               40056: 0.973,\n",
       "               41389: 0.973,\n",
       "               39469: 0.973,\n",
       "               37520: 0.978,\n",
       "               40004: 0.975,\n",
       "               39456: 0.972,\n",
       "               33487: 0.989,\n",
       "               33506: 0.988,\n",
       "               37716: 0.987,\n",
       "               29163: 0.991,\n",
       "               28387: 0.993,\n",
       "               27727: 0.991,\n",
       "               516: 0.994,\n",
       "               235: 0.994,\n",
       "               4999: 0.996,\n",
       "               4196: 0.994,\n",
       "               28677: 0.994,\n",
       "               28927: 0.994,\n",
       "               29785: 0.993,\n",
       "               38481: 0.897,\n",
       "               38843: 0.895,\n",
       "               42517: 0.891,\n",
       "               31374: 0.992,\n",
       "               33031: 0.989,\n",
       "               11597: 0.993,\n",
       "               30736: 0.981,\n",
       "               33095: 0.981,\n",
       "               6772: 0.991,\n",
       "               540: 0.99,\n",
       "               9232: 0.989,\n",
       "               916: 0.992,\n",
       "               2110: 0.992,\n",
       "               298: 0.992,\n",
       "               34055: 0.993,\n",
       "               34076: 0.989,\n",
       "               35062: 0.989,\n",
       "               34012: 0.872,\n",
       "               32951: 0.871,\n",
       "               34104: 0.985,\n",
       "               522: 0.993,\n",
       "               14293: 0.993,\n",
       "               2560: 0.992,\n",
       "               33089: 0.991,\n",
       "               33654: 0.991,\n",
       "               29067: 0.993,\n",
       "               16732: 0.993,\n",
       "               22319: 0.994,\n",
       "               19315: 0.994,\n",
       "               21507: 0.993,\n",
       "               31240: 0.991,\n",
       "               32189: 0.99,\n",
       "               186: 0.992,\n",
       "               74: 0.992,\n",
       "               41674: 0.992,\n",
       "               3782: 0.993,\n",
       "               31552: 0.992,\n",
       "               13185: 0.993,\n",
       "               16890: 0.99,\n",
       "               22735: 0.99,\n",
       "               37446: 0.988,\n",
       "               37245: 0.985,\n",
       "               35757: 0.984,\n",
       "               28619: 0.994,\n",
       "               23251: 0.991,\n",
       "               37765: 0.909,\n",
       "               28755: 0.907,\n",
       "               38664: 0.905,\n",
       "               9513: 0.995,\n",
       "               7592: 0.995,\n",
       "               185: 0.992,\n",
       "               627: 0.992,\n",
       "               22: 0.991,\n",
       "               1827: 0.994,\n",
       "               934: 0.992,\n",
       "               849: 0.992,\n",
       "               27655: 0.992,\n",
       "               25731: 0.992,\n",
       "               850: 0.992,\n",
       "               9181: 0.992,\n",
       "               3731: 0.992,\n",
       "               10087: 0.993,\n",
       "               348: 0.993,\n",
       "               1195: 0.992,\n",
       "               884: 0.989,\n",
       "               5129: 0.988,\n",
       "               4725: 0.994,\n",
       "               34403: 0.985,\n",
       "               34525: 0.985,\n",
       "               25013: 0.993,\n",
       "               16049: 0.993,\n",
       "               19775: 0.993,\n",
       "               15613: 0.98,\n",
       "               18893: 0.979,\n",
       "               7334: 0.992,\n",
       "               21729: 0.992,\n",
       "               10884: 0.991,\n",
       "               30230: 0.988,\n",
       "               29589: 0.994,\n",
       "               30503: 0.995,\n",
       "               32648: 0.993,\n",
       "               31229: 0.992,\n",
       "               16918: 0.994,\n",
       "               15199: 0.994,\n",
       "               2760: 0.993,\n",
       "               32979: 0.981,\n",
       "               28221: 0.989,\n",
       "               31338: 0.979,\n",
       "               21325: 0.994,\n",
       "               21681: 0.994,\n",
       "               20509: 0.994,\n",
       "               1889: 0.991,\n",
       "               11319: 0.99,\n",
       "               11933: 0.989,\n",
       "               22101: 0.994,\n",
       "               21039: 0.992,\n",
       "               584: 0.991,\n",
       "               5630: 0.99,\n",
       "               24211: 0.987,\n",
       "               21845: 0.986,\n",
       "               11235: 0.991,\n",
       "               18497: 0.993,\n",
       "               20541: 0.991,\n",
       "               36094: 0.986,\n",
       "               36028: 0.983,\n",
       "               36649: 0.98,\n",
       "               1606: 0.993,\n",
       "               18679: 0.993,\n",
       "               22821: 0.978,\n",
       "               30470: 0.978,\n",
       "               142: 0.995,\n",
       "               1573: 0.995,\n",
       "               7724: 0.995,\n",
       "               953: 0.995,\n",
       "               22663: 0.992,\n",
       "               160: 0.992,\n",
       "               3358: 0.992,\n",
       "               2732: 0.992,\n",
       "               334: 0.994,\n",
       "               269: 0.994,\n",
       "               33012: 0.98,\n",
       "               36296: 0.98,\n",
       "               31621: 0.988,\n",
       "               33094: 0.988,\n",
       "               26351: 0.988,\n",
       "               14913: 0.992,\n",
       "               3493: 0.991,\n",
       "               21167: 0.99,\n",
       "               15489: 0.996,\n",
       "               2702: 0.992,\n",
       "               1175: 0.991,\n",
       "               225: 0.993,\n",
       "               3359: 0.994,\n",
       "               6324: 0.993,\n",
       "               24833: 0.995,\n",
       "               26441: 0.994,\n",
       "               46431: 0.815,\n",
       "               18753: 0.753,\n",
       "               39547: 0.712,\n",
       "               14289: 0.995,\n",
       "               34565: 0.989,\n",
       "               34591: 0.989,\n",
       "               35629: 0.987,\n",
       "               223: 0.993,\n",
       "               26243: 0.993,\n",
       "               35946: 0.995,\n",
       "               15335: 0.995,\n",
       "               2418: 0.994,\n",
       "               36023: 0.985,\n",
       "               32182: 0.984,\n",
       "               35079: 0.984,\n",
       "               30276: 0.993,\n",
       "               24765: 0.993,\n",
       "               25681: 0.992,\n",
       "               31442: 0.982,\n",
       "               34102: 0.988,\n",
       "               31670: 0.98,\n",
       "               32696: 0.995,\n",
       "               31109: 0.994,\n",
       "               31680: 0.989,\n",
       "               812: 0.993,\n",
       "               1655: 0.993,\n",
       "               32023: 0.989,\n",
       "               31933: 0.988,\n",
       "               13601: 0.992,\n",
       "               19769: 0.992,\n",
       "               8407: 0.995,\n",
       "               30385: 0.987,\n",
       "               32603: 0.993,\n",
       "               30651: 0.99,\n",
       "               31793: 0.99,\n",
       "               21855: 0.992,\n",
       "               18153: 0.992,\n",
       "               5597: 0.994,\n",
       "               335: 0.993,\n",
       "               668: 0.993,\n",
       "               34712: 0.988,\n",
       "               35712: 0.987,\n",
       "               11617: 0.994,\n",
       "               7088: 0.994,\n",
       "               5081: 0.994,\n",
       "               3905: 0.994,\n",
       "               3277: 0.994,\n",
       "               3839: 0.994,\n",
       "               30901: 0.992,\n",
       "               19157: 0.99,\n",
       "               125: 0.991,\n",
       "               486: 0.99,\n",
       "               25649: 0.991,\n",
       "               1575: 0.996,\n",
       "               15651: 0.992,\n",
       "               15611: 0.991,\n",
       "               4080: 0.991,\n",
       "               3972: 0.991,\n",
       "               2123: 0.992,\n",
       "               28675: 0.991,\n",
       "               3258: 0.991,\n",
       "               5734: 0.994,\n",
       "               4038: 0.994,\n",
       "               12549: 0.993,\n",
       "               16201: 0.993,\n",
       "               856: 0.994,\n",
       "               2927: 0.994,\n",
       "               3467: 0.993,\n",
       "               16894: 0.957,\n",
       "               28869: 0.955,\n",
       "               1555: 0.987,\n",
       "               17729: 0.987,\n",
       "               32836: 0.986,\n",
       "               33948: 0.984,\n",
       "               2251: 0.993,\n",
       "               18277: 0.993,\n",
       "               17513: 0.993,\n",
       "               12679: 0.993,\n",
       "               21563: 0.994,\n",
       "               19369: 0.993,\n",
       "               1516: 0.991,\n",
       "               1583: 0.991,\n",
       "               502: 0.996,\n",
       "               896: 0.996,\n",
       "               906: 0.993,\n",
       "               20899: 0.994,\n",
       "               4081: 0.993,\n",
       "               11665: 0.992,\n",
       "               4192: 0.995,\n",
       "               2476: 0.995,\n",
       "               61: 0.995,\n",
       "               899: 0.995,\n",
       "               897: 0.994,\n",
       "               904: 0.996,\n",
       "               36946: 0.993,\n",
       "               34209: 0.976,\n",
       "               36407: 0.975,\n",
       "               38276: 0.973,\n",
       "               2595: 0.989,\n",
       "               5071: 0.989,\n",
       "               1218: 0.989,\n",
       "               28907: 0.982,\n",
       "               31043: 0.978,\n",
       "               4214: 0.993,\n",
       "               5005: 0.993,\n",
       "               6895: 0.993,\n",
       "               17247: 0.992,\n",
       "               5342: 0.995,\n",
       "               10869: 0.994,\n",
       "               3091: 0.992,\n",
       "               1198: 0.992,\n",
       "               14719: 0.994,\n",
       "               25517: 0.993,\n",
       "               4896: 0.993,\n",
       "               9926: 0.992,\n",
       "               15795: 0.991,\n",
       "               11017: 0.989,\n",
       "               8410: 0.989,\n",
       "               2035: 0.989,\n",
       "               35298: 0.986,\n",
       "               37171: 0.983,\n",
       "               37675: 0.983,\n",
       "               11013: 0.992,\n",
       "               15125: 0.991,\n",
       "               5231: 0.991,\n",
       "               33255: 0.992,\n",
       "               31430: 0.992,\n",
       "               34978: 0.991,\n",
       "               33731: 0.988,\n",
       "               32370: 0.994,\n",
       "               39545: 0.992,\n",
       "               17068: 0.992,\n",
       "               37827: 0.99,\n",
       "               10165: 0.995,\n",
       "               9065: 0.994,\n",
       "               24031: 0.993,\n",
       "               27899: 0.991,\n",
       "               30524: 0.988,\n",
       "               28621: 0.983,\n",
       "               32682: 0.983,\n",
       "               878: 0.995,\n",
       "               4334: 0.994,\n",
       "               38668: 0.991,\n",
       "               6: 0.99,\n",
       "               3783: 0.993,\n",
       "               1034: 0.993,\n",
       "               6098: 0.994,\n",
       "               8841: 0.994,\n",
       "               21033: 0.994,\n",
       "               7588: 0.992,\n",
       "               14749: 0.962,\n",
       "               17909: 0.96,\n",
       "               13659: 0.959,\n",
       "               8516: 0.992,\n",
       "               1532: 0.989,\n",
       "               12321: 0.992,\n",
       "               6707: 0.99,\n",
       "               21329: 0.99,\n",
       "               38999: 0.99,\n",
       "               36470: 0.984,\n",
       "               36726: 0.984,\n",
       "               34798: 0.983,\n",
       "               11783: 0.992,\n",
       "               546: 0.992,\n",
       "               12115: 0.997,\n",
       "               10218: 0.996,\n",
       "               893: 0.996,\n",
       "               903: 0.994,\n",
       "               11771: 0.989,\n",
       "               18689: 0.989,\n",
       "               16518: 0.993,\n",
       "               11061: 0.993})})"
      ]
     },
     "execution_count": 146,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "process_user(205282, embedders, K=3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 147,
   "id": "bb234b56",
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "import json\n",
    "\n",
    "# Convert int64 keys to int or str\n",
    "import json\n",
    "\n",
    "def convert_float32(obj):\n",
    "    if isinstance(obj, dict):\n",
    "        return {key: convert_float32(value) for key, value in obj.items()}\n",
    "    elif isinstance(obj, list):\n",
    "        return [convert_float32(item) for item in obj]\n",
    "    elif isinstance(obj, np.float32):\n",
    "        return float(obj)\n",
    "    else:\n",
    "        return obj\n",
    "\n",
    "user_preds_converted = {int(key): value for key, value in user_preds.items()}\n",
    "user_preds_converted = convert_float32(user_preds_converted)\n",
    "\n",
    "# save this because it was computationally expensive to create\n",
    "with open('user_preds.json', 'w') as f:\n",
    "    json.dump(user_preds_converted, f)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "f93c2928",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "5ceab3a7",
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "import json\n",
    "with open('user_preds.json', 'rU') as f:\n",
    "    user_preds = json.load(f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 149,
   "id": "cd726e5b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "4"
      ]
     },
     "execution_count": 149,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "total_animes_predicted = {e: set() for e in embedders}\n",
    "for user_id, recs in user_preds.items():\n",
    "    for embedder, rec in recs.items():\n",
    "        total_animes_predicted[embedder].update((rec.keys()))\n",
    "len(total_animes_predicted)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 280,
   "id": "ea006049",
   "metadata": {},
   "outputs": [],
   "source": [
    "rating_complete_test['anime_id'].nunique()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 281,
   "id": "f99be65d",
   "metadata": {},
   "outputs": [],
   "source": [
    "{k: len(v) for k, v in total_animes_predicted.items()}\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "4ef280da",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 152,
   "id": "7560413f",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>user_id</th>\n",
       "      <th>anime_id</th>\n",
       "      <th>rating</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>57039029</th>\n",
       "      <td>349749</td>\n",
       "      <td>154</td>\n",
       "      <td>9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>33418394</th>\n",
       "      <td>205282</td>\n",
       "      <td>18001</td>\n",
       "      <td>9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18511337</th>\n",
       "      <td>113697</td>\n",
       "      <td>10110</td>\n",
       "      <td>9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>53862736</th>\n",
       "      <td>330116</td>\n",
       "      <td>37520</td>\n",
       "      <td>7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30501411</th>\n",
       "      <td>187612</td>\n",
       "      <td>209</td>\n",
       "      <td>10</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13315092</th>\n",
       "      <td>81644</td>\n",
       "      <td>5178</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21081788</th>\n",
       "      <td>129703</td>\n",
       "      <td>10721</td>\n",
       "      <td>9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>56882282</th>\n",
       "      <td>348777</td>\n",
       "      <td>11499</td>\n",
       "      <td>8</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>49648910</th>\n",
       "      <td>304381</td>\n",
       "      <td>10702</td>\n",
       "      <td>7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>56755036</th>\n",
       "      <td>348022</td>\n",
       "      <td>6045</td>\n",
       "      <td>8</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5763328 rows × 3 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "          user_id  anime_id  rating\n",
       "57039029   349749       154       9\n",
       "33418394   205282     18001       9\n",
       "18511337   113697     10110       9\n",
       "53862736   330116     37520       7\n",
       "30501411   187612       209      10\n",
       "...           ...       ...     ...\n",
       "13315092    81644      5178       3\n",
       "21081788   129703     10721       9\n",
       "56882282   348777     11499       8\n",
       "49648910   304381     10702       7\n",
       "56755036   348022      6045       8\n",
       "\n",
       "[5763328 rows x 3 columns]"
      ]
     },
     "execution_count": 152,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rating_complete_test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 153,
   "id": "177f46f3",
   "metadata": {},
   "outputs": [],
   "source": [
    "from collections import defaultdict\n",
    "try:\n",
    "    user_anime_test_ratings\n",
    "except:\n",
    "    user_anime_test_ratings = defaultdict(dict)\n",
    "    for index, user_id, anime_id, rating in tqdm(rating_complete_test.to_records()):\n",
    "        user_anime_test_ratings[user_id][anime_id] = rating\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "07d1a5cc",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Importing tqdm for progress bar visualization\n",
    "from tqdm.auto import tqdm\n",
    "\n",
    "# Creating a map to score user ratings.\n",
    "# 10 and 9 are considered positive (1), 8 and 7 are neutral (0), and all others are considered negative (-1)\n",
    "score_map = {10: 1, 9: 1, 8: 0, 7: 0, 6: -1, 5: -1, 4:-1, 3:-1, 2:-1, 1:-1, 0:-1}\n",
    "\n",
    "# Initializing list for k_values and dictionary for counts of each embedder\n",
    "k_values = []\n",
    "embedder_counts = defaultdict(list)\n",
    "\n",
    "# Creating a range of integers from 1 to 10\n",
    "_r = list(range(1, 11))\n",
    "\n",
    "# Looping over the range to get the different k values (k-factor) for recommendations\n",
    "for k_factor in tqdm(_r):\n",
    "    # Looping over each user and their corresponding recommendations\n",
    "    for user_id, user_recs in user_preds.items():\n",
    "        # Looping over each embedder and the corresponding recommendations from the embedder\n",
    "        for embedder, recs_from_embedder in user_recs.items():\n",
    "            # Selecting the top-k recommendations for each user from each embedder\n",
    "            final_recs = sorted(recs_from_embedder.items(), key=lambda x: x[1], reverse=True)[:k_factor]\n",
    "\n",
    "            # Looping over the top-k recommendations\n",
    "            for index, (recommended_anime_id, recommendation_score) in enumerate(final_recs):  \n",
    "                # Fetching the user rating for the recommended anime\n",
    "                user_rating = user_anime_test_ratings.get(int(user_id), {}).get(int(recommended_anime_id))\n",
    "\n",
    "                # Checking if the user has a rating for the recommended anime\n",
    "                if user_rating:\n",
    "                    # The score of each recommendation is adjusted by its rank and added to the embedder_counts dictionary\n",
    "                    embedder_counts[embedder].append((score_map[user_rating] * (k_factor - index) / k_factor))\n",
    "\n",
    "    # Calculating the average Net Promoter Score (NPS) for each embedder and adding it to the k_values list\n",
    "    k_values.append({e: dict(k=k_factor, avg_nps=sum(l) / len(l), predictions_found_in_test_set=len(l)) for e, l in embedder_counts.items()})\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 228,
   "id": "ee16acc0",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1000x500 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# Extract the data\n",
    "models = list(k_values[0].keys())\n",
    "ks = [entry[models[0]][\"k\"] for entry in k_values]\n",
    "model_avg_nps = {model: [entry[model][\"avg_nps\"] for entry in k_values] for model in models}\n",
    "\n",
    "# Create a single figure with modern styling\n",
    "plt.style.use('ggplot')\n",
    "fig, ax = plt.subplots(figsize=(10, 5))\n",
    "\n",
    "# Define color-blind friendly colors, line styles, and marker styles\n",
    "colors = ['#0072B2', '#009E73', '#D55E00', '#CC79A7']\n",
    "line_styles = ['-', '--', '-.', ':']\n",
    "marker_styles = ['o', 's', 'D', '^']\n",
    "\n",
    "# Plot lines for each model\n",
    "for idx, (model, avg_nps) in enumerate(model_avg_nps.items()):\n",
    "    ax.plot(ks, avg_nps, label=model, linewidth=2, marker=marker_styles[idx], linestyle=line_styles[idx], color=colors[idx])\n",
    "\n",
    "# Add labels, title, and legend\n",
    "ax.set_xlabel('# of Recommendedtions shown to user')\n",
    "ax.set_ylabel('Average Rec. Promoted Score (scaled for rank)')\n",
    "ax.set_title('Performance of Embedders and k Values')\n",
    "ax.legend()\n",
    "\n",
    "# Display the graph\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "37c11eb7",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "2e049ec5",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "d7da7ba5",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "5f7046ef",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 239,
   "id": "0a9f803a",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>anime_encoder_avg</th>\n",
       "      <th>anime_encoder_bigger_avg</th>\n",
       "      <th>text-embedding-ada-002_avg</th>\n",
       "      <th>paraphrase-distilroberta-base-v1_avg</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.606310</td>\n",
       "      <td>0.715316</td>\n",
       "      <td>0.707477</td>\n",
       "      <td>0.641283</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.450873</td>\n",
       "      <td>0.617474</td>\n",
       "      <td>0.735131</td>\n",
       "      <td>0.609570</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.466259</td>\n",
       "      <td>0.590638</td>\n",
       "      <td>0.711768</td>\n",
       "      <td>0.615678</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.474589</td>\n",
       "      <td>0.559918</td>\n",
       "      <td>0.703781</td>\n",
       "      <td>0.616045</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>0.470490</td>\n",
       "      <td>0.519009</td>\n",
       "      <td>0.687056</td>\n",
       "      <td>0.616407</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>0.452965</td>\n",
       "      <td>0.500867</td>\n",
       "      <td>0.670499</td>\n",
       "      <td>0.610668</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>0.471801</td>\n",
       "      <td>0.463763</td>\n",
       "      <td>0.660379</td>\n",
       "      <td>0.602164</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>0.441482</td>\n",
       "      <td>0.434208</td>\n",
       "      <td>0.668584</td>\n",
       "      <td>0.606114</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>0.428241</td>\n",
       "      <td>0.429510</td>\n",
       "      <td>0.647485</td>\n",
       "      <td>0.600284</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>0.446724</td>\n",
       "      <td>0.406008</td>\n",
       "      <td>0.649871</td>\n",
       "      <td>0.602507</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>0.441558</td>\n",
       "      <td>0.404528</td>\n",
       "      <td>0.644636</td>\n",
       "      <td>0.609465</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>0.416373</td>\n",
       "      <td>0.370144</td>\n",
       "      <td>0.658037</td>\n",
       "      <td>0.583698</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>0.426342</td>\n",
       "      <td>0.375810</td>\n",
       "      <td>0.639781</td>\n",
       "      <td>0.596703</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>0.423287</td>\n",
       "      <td>0.358482</td>\n",
       "      <td>0.640206</td>\n",
       "      <td>0.587202</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>0.407305</td>\n",
       "      <td>0.367975</td>\n",
       "      <td>0.624091</td>\n",
       "      <td>0.581563</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    anime_encoder_avg  anime_encoder_bigger_avg  text-embedding-ada-002_avg  \\\n",
       "1            0.606310                  0.715316                    0.707477   \n",
       "2            0.450873                  0.617474                    0.735131   \n",
       "3            0.466259                  0.590638                    0.711768   \n",
       "4            0.474589                  0.559918                    0.703781   \n",
       "5            0.470490                  0.519009                    0.687056   \n",
       "6            0.452965                  0.500867                    0.670499   \n",
       "7            0.471801                  0.463763                    0.660379   \n",
       "8            0.441482                  0.434208                    0.668584   \n",
       "9            0.428241                  0.429510                    0.647485   \n",
       "10           0.446724                  0.406008                    0.649871   \n",
       "11           0.441558                  0.404528                    0.644636   \n",
       "12           0.416373                  0.370144                    0.658037   \n",
       "13           0.426342                  0.375810                    0.639781   \n",
       "14           0.423287                  0.358482                    0.640206   \n",
       "15           0.407305                  0.367975                    0.624091   \n",
       "\n",
       "    paraphrase-distilroberta-base-v1_avg  \n",
       "1                               0.641283  \n",
       "2                               0.609570  \n",
       "3                               0.615678  \n",
       "4                               0.616045  \n",
       "5                               0.616407  \n",
       "6                               0.610668  \n",
       "7                               0.602164  \n",
       "8                               0.606114  \n",
       "9                               0.600284  \n",
       "10                              0.602507  \n",
       "11                              0.609465  \n",
       "12                              0.583698  \n",
       "13                              0.596703  \n",
       "14                              0.587202  \n",
       "15                              0.581563  "
      ]
     },
     "execution_count": 239,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pd.DataFrame(graph_scores_by_k).T\n",
    "\n",
    "df"
   ]
  }
 ],
 "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": "3.11.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}