MAGNET.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "yh4B6G1X0onL"
   },
   "source": [
    "## Installations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1000
    },
    "id": "IvGcgjcrHBiH",
    "outputId": "8ef9d252-59b6-417b-f099-d7adb19edd1c"
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "\n",
       "     <input type=\"file\" id=\"files-d13722cc-e6b9-482d-a8cc-bf8b1b3275fb\" name=\"files[]\" multiple disabled\n",
       "        style=\"border:none\" />\n",
       "     <output id=\"result-d13722cc-e6b9-482d-a8cc-bf8b1b3275fb\">\n",
       "      Upload widget is only available when the cell has been executed in the\n",
       "      current browser session. Please rerun this cell to enable.\n",
       "      </output>\n",
       "      <script>// Copyright 2017 Google LLC\n",
       "//\n",
       "// Licensed under the Apache License, Version 2.0 (the \"License\");\n",
       "// you may not use this file except in compliance with the License.\n",
       "// You may obtain a copy of the License at\n",
       "//\n",
       "//      http://www.apache.org/licenses/LICENSE-2.0\n",
       "//\n",
       "// Unless required by applicable law or agreed to in writing, software\n",
       "// distributed under the License is distributed on an \"AS IS\" BASIS,\n",
       "// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n",
       "// See the License for the specific language governing permissions and\n",
       "// limitations under the License.\n",
       "\n",
       "/**\n",
       " * @fileoverview Helpers for google.colab Python module.\n",
       " */\n",
       "(function(scope) {\n",
       "function span(text, styleAttributes = {}) {\n",
       "  const element = document.createElement('span');\n",
       "  element.textContent = text;\n",
       "  for (const key of Object.keys(styleAttributes)) {\n",
       "    element.style[key] = styleAttributes[key];\n",
       "  }\n",
       "  return element;\n",
       "}\n",
       "\n",
       "// Max number of bytes which will be uploaded at a time.\n",
       "const MAX_PAYLOAD_SIZE = 100 * 1024;\n",
       "\n",
       "function _uploadFiles(inputId, outputId) {\n",
       "  const steps = uploadFilesStep(inputId, outputId);\n",
       "  const outputElement = document.getElementById(outputId);\n",
       "  // Cache steps on the outputElement to make it available for the next call\n",
       "  // to uploadFilesContinue from Python.\n",
       "  outputElement.steps = steps;\n",
       "\n",
       "  return _uploadFilesContinue(outputId);\n",
       "}\n",
       "\n",
       "// This is roughly an async generator (not supported in the browser yet),\n",
       "// where there are multiple asynchronous steps and the Python side is going\n",
       "// to poll for completion of each step.\n",
       "// This uses a Promise to block the python side on completion of each step,\n",
       "// then passes the result of the previous step as the input to the next step.\n",
       "function _uploadFilesContinue(outputId) {\n",
       "  const outputElement = document.getElementById(outputId);\n",
       "  const steps = outputElement.steps;\n",
       "\n",
       "  const next = steps.next(outputElement.lastPromiseValue);\n",
       "  return Promise.resolve(next.value.promise).then((value) => {\n",
       "    // Cache the last promise value to make it available to the next\n",
       "    // step of the generator.\n",
       "    outputElement.lastPromiseValue = value;\n",
       "    return next.value.response;\n",
       "  });\n",
       "}\n",
       "\n",
       "/**\n",
       " * Generator function which is called between each async step of the upload\n",
       " * process.\n",
       " * @param {string} inputId Element ID of the input file picker element.\n",
       " * @param {string} outputId Element ID of the output display.\n",
       " * @return {!Iterable<!Object>} Iterable of next steps.\n",
       " */\n",
       "function* uploadFilesStep(inputId, outputId) {\n",
       "  const inputElement = document.getElementById(inputId);\n",
       "  inputElement.disabled = false;\n",
       "\n",
       "  const outputElement = document.getElementById(outputId);\n",
       "  outputElement.innerHTML = '';\n",
       "\n",
       "  const pickedPromise = new Promise((resolve) => {\n",
       "    inputElement.addEventListener('change', (e) => {\n",
       "      resolve(e.target.files);\n",
       "    });\n",
       "  });\n",
       "\n",
       "  const cancel = document.createElement('button');\n",
       "  inputElement.parentElement.appendChild(cancel);\n",
       "  cancel.textContent = 'Cancel upload';\n",
       "  const cancelPromise = new Promise((resolve) => {\n",
       "    cancel.onclick = () => {\n",
       "      resolve(null);\n",
       "    };\n",
       "  });\n",
       "\n",
       "  // Wait for the user to pick the files.\n",
       "  const files = yield {\n",
       "    promise: Promise.race([pickedPromise, cancelPromise]),\n",
       "    response: {\n",
       "      action: 'starting',\n",
       "    }\n",
       "  };\n",
       "\n",
       "  cancel.remove();\n",
       "\n",
       "  // Disable the input element since further picks are not allowed.\n",
       "  inputElement.disabled = true;\n",
       "\n",
       "  if (!files) {\n",
       "    return {\n",
       "      response: {\n",
       "        action: 'complete',\n",
       "      }\n",
       "    };\n",
       "  }\n",
       "\n",
       "  for (const file of files) {\n",
       "    const li = document.createElement('li');\n",
       "    li.append(span(file.name, {fontWeight: 'bold'}));\n",
       "    li.append(span(\n",
       "        `(${file.type || 'n/a'}) - ${file.size} bytes, ` +\n",
       "        `last modified: ${\n",
       "            file.lastModifiedDate ? file.lastModifiedDate.toLocaleDateString() :\n",
       "                                    'n/a'} - `));\n",
       "    const percent = span('0% done');\n",
       "    li.appendChild(percent);\n",
       "\n",
       "    outputElement.appendChild(li);\n",
       "\n",
       "    const fileDataPromise = new Promise((resolve) => {\n",
       "      const reader = new FileReader();\n",
       "      reader.onload = (e) => {\n",
       "        resolve(e.target.result);\n",
       "      };\n",
       "      reader.readAsArrayBuffer(file);\n",
       "    });\n",
       "    // Wait for the data to be ready.\n",
       "    let fileData = yield {\n",
       "      promise: fileDataPromise,\n",
       "      response: {\n",
       "        action: 'continue',\n",
       "      }\n",
       "    };\n",
       "\n",
       "    // Use a chunked sending to avoid message size limits. See b/62115660.\n",
       "    let position = 0;\n",
       "    do {\n",
       "      const length = Math.min(fileData.byteLength - position, MAX_PAYLOAD_SIZE);\n",
       "      const chunk = new Uint8Array(fileData, position, length);\n",
       "      position += length;\n",
       "\n",
       "      const base64 = btoa(String.fromCharCode.apply(null, chunk));\n",
       "      yield {\n",
       "        response: {\n",
       "          action: 'append',\n",
       "          file: file.name,\n",
       "          data: base64,\n",
       "        },\n",
       "      };\n",
       "\n",
       "      let percentDone = fileData.byteLength === 0 ?\n",
       "          100 :\n",
       "          Math.round((position / fileData.byteLength) * 100);\n",
       "      percent.textContent = `${percentDone}% done`;\n",
       "\n",
       "    } while (position < fileData.byteLength);\n",
       "  }\n",
       "\n",
       "  // All done.\n",
       "  yield {\n",
       "    response: {\n",
       "      action: 'complete',\n",
       "    }\n",
       "  };\n",
       "}\n",
       "\n",
       "scope.google = scope.google || {};\n",
       "scope.google.colab = scope.google.colab || {};\n",
       "scope.google.colab._files = {\n",
       "  _uploadFiles,\n",
       "  _uploadFilesContinue,\n",
       "};\n",
       "})(self);\n",
       "</script> "
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Saving kaggle.json to kaggle.json\n",
      "Downloading jigsaw-toxic-comment-classification-challenge.zip to /content\n",
      " 99% 52.0M/52.6M [00:03<00:00, 24.8MB/s]\n",
      "100% 52.6M/52.6M [00:03<00:00, 17.0MB/s]\n",
      "Archive:  jigsaw-toxic-comment-classification-challenge.zip\n",
      "  inflating: sample_submission.csv.zip  \n",
      "  inflating: test.csv.zip            \n",
      "  inflating: test_labels.csv.zip     \n",
      "  inflating: train.csv.zip           \n",
      "Archive:  train.csv.zip\n",
      "  inflating: train.csv               \n",
      "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
      "Collecting scikit-multilearn\n",
      "  Downloading scikit_multilearn-0.2.0-py3-none-any.whl (89 kB)\n",
      "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m89.4/89.4 kB\u001b[0m \u001b[31m8.4 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
      "\u001b[?25hInstalling collected packages: scikit-multilearn\n",
      "Successfully installed scikit-multilearn-0.2.0\n",
      "--2023-05-07 20:26:43--  https://nlp.stanford.edu/data/glove.6B.zip\n",
      "Resolving nlp.stanford.edu (nlp.stanford.edu)... 171.64.67.140\n",
      "Connecting to nlp.stanford.edu (nlp.stanford.edu)|171.64.67.140|:443... connected.\n",
      "HTTP request sent, awaiting response... 301 Moved Permanently\n",
      "Location: https://downloads.cs.stanford.edu/nlp/data/glove.6B.zip [following]\n",
      "--2023-05-07 20:26:44--  https://downloads.cs.stanford.edu/nlp/data/glove.6B.zip\n",
      "Resolving downloads.cs.stanford.edu (downloads.cs.stanford.edu)... 171.64.64.22\n",
      "Connecting to downloads.cs.stanford.edu (downloads.cs.stanford.edu)|171.64.64.22|:443... connected.\n",
      "HTTP request sent, awaiting response... 200 OK\n",
      "Length: 862182613 (822M) [application/zip]\n",
      "Saving to: ‘glove.6B.zip’\n",
      "\n",
      "glove.6B.zip        100%[===================>] 822.24M  5.02MB/s    in 2m 39s  \n",
      "\n",
      "2023-05-07 20:29:24 (5.17 MB/s) - ‘glove.6B.zip’ saved [862182613/862182613]\n",
      "\n",
      "Archive:  glove.6B.zip\n",
      "  inflating: glove.6B/glove.6B.50d.txt  \n",
      "  inflating: glove.6B/glove.6B.100d.txt  \n",
      "  inflating: glove.6B/glove.6B.200d.txt  \n",
      "  inflating: glove.6B/glove.6B.300d.txt  \n",
      "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
      "Collecting transformers\n",
      "  Downloading transformers-4.28.1-py3-none-any.whl (7.0 MB)\n",
      "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m7.0/7.0 MB\u001b[0m \u001b[31m105.2 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
      "\u001b[?25hRequirement already satisfied: filelock in /usr/local/lib/python3.10/dist-packages (from transformers) (3.12.0)\n",
      "Requirement already satisfied: packaging>=20.0 in /usr/local/lib/python3.10/dist-packages (from transformers) (23.1)\n",
      "Requirement already satisfied: regex!=2019.12.17 in /usr/local/lib/python3.10/dist-packages (from transformers) (2022.10.31)\n",
      "Collecting tokenizers!=0.11.3,<0.14,>=0.11.1\n",
      "  Downloading tokenizers-0.13.3-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (7.8 MB)\n",
      "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m7.8/7.8 MB\u001b[0m \u001b[31m117.7 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
      "\u001b[?25hCollecting huggingface-hub<1.0,>=0.11.0\n",
      "  Downloading huggingface_hub-0.14.1-py3-none-any.whl (224 kB)\n",
      "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m224.5/224.5 kB\u001b[0m \u001b[31m19.3 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
      "\u001b[?25hRequirement already satisfied: tqdm>=4.27 in /usr/local/lib/python3.10/dist-packages (from transformers) (4.65.0)\n",
      "Requirement already satisfied: numpy>=1.17 in /usr/local/lib/python3.10/dist-packages (from transformers) (1.22.4)\n",
      "Requirement already satisfied: requests in /usr/local/lib/python3.10/dist-packages (from transformers) (2.27.1)\n",
      "Requirement already satisfied: pyyaml>=5.1 in /usr/local/lib/python3.10/dist-packages (from transformers) (6.0)\n",
      "Requirement already satisfied: typing-extensions>=3.7.4.3 in /usr/local/lib/python3.10/dist-packages (from huggingface-hub<1.0,>=0.11.0->transformers) (4.5.0)\n",
      "Requirement already satisfied: fsspec in /usr/local/lib/python3.10/dist-packages (from huggingface-hub<1.0,>=0.11.0->transformers) (2023.4.0)\n",
      "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.10/dist-packages (from requests->transformers) (2022.12.7)\n",
      "Requirement already satisfied: urllib3<1.27,>=1.21.1 in /usr/local/lib/python3.10/dist-packages (from requests->transformers) (1.26.15)\n",
      "Requirement already satisfied: idna<4,>=2.5 in /usr/local/lib/python3.10/dist-packages (from requests->transformers) (3.4)\n",
      "Requirement already satisfied: charset-normalizer~=2.0.0 in /usr/local/lib/python3.10/dist-packages (from requests->transformers) (2.0.12)\n",
      "Installing collected packages: tokenizers, huggingface-hub, transformers\n",
      "Successfully installed huggingface-hub-0.14.1 tokenizers-0.13.3 transformers-4.28.1\n"
     ]
    }
   ],
   "source": [
    "!pip install -q kaggle\n",
    "from google.colab import files\n",
    "files.upload()\n",
    "! mkdir ~/.kaggle\n",
    "! cp kaggle.json ~/.kaggle/\n",
    "! chmod 600 ~/.kaggle/kaggle.json\n",
    "\n",
    "!kaggle competitions download -c jigsaw-toxic-comment-classification-challenge\n",
    "!unzip jigsaw-toxic-comment-classification-challenge.zip\n",
    "\n",
    "!unzip train.csv.zip\n",
    "\n",
    "!pip install scikit-multilearn\n",
    "!wget https://nlp.stanford.edu/data/glove.6B.zip\n",
    "!unzip glove.6B.zip -d glove.6B/\n",
    "!pip install transformers"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "DriPlvPN0riP"
   },
   "source": [
    "## Imports"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "QZS8p3SrHW3u",
    "outputId": "31b29109-7c81-44f6-b532-d4d6c6420f58"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[nltk_data] Downloading package stopwords to /root/nltk_data...\n",
      "[nltk_data]   Package stopwords is already up-to-date!\n",
      "[nltk_data] Downloading package wordnet to /root/nltk_data...\n",
      "[nltk_data]   Package wordnet is already up-to-date!\n"
     ]
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import nltk\n",
    "# nltk.download('reuters')\n",
    "nltk.download('stopwords')\n",
    "nltk.download('wordnet')\n",
    "from nltk.corpus import reuters\n",
    "from nltk.corpus import stopwords\n",
    "from nltk import PorterStemmer, WordNetLemmatizer\n",
    "import re\n",
    "import string\n",
    "import tensorflow as tf\n",
    "from tensorflow.keras.preprocessing.text import Tokenizer\n",
    "from tensorflow.keras.preprocessing.sequence import pad_sequences\n",
    "import numpy as np\n",
    "\n",
    "import torch\n",
    "from torch import optim\n",
    "from torch import nn\n",
    "from torch.nn.utils import clip_grad_norm_\n",
    "\n",
    "from sklearn.metrics import hamming_loss, f1_score\n",
    "from sklearn.preprocessing import MultiLabelBinarizer\n",
    "\n",
    "from tqdm import tqdm\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "import xgboost as xgb\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.naive_bayes import MultinomialNB\n",
    "from sklearn.multiclass import OneVsRestClassifier\n",
    "from skmultilearn.problem_transform import BinaryRelevance\n",
    "\n",
    "from sklearn import metrics\n",
    "from sklearn.metrics import roc_auc_score, accuracy_score, classification_report\n",
    "\n",
    "import time\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "from skmultilearn.problem_transform import LabelPowerset\n",
    "from sklearn.multioutput import ClassifierChain\n",
    "import copy\n",
    "from scipy.sparse import csr_matrix\n",
    "from nltk.corpus import stopwords\n",
    "from nltk.stem.snowball import SnowballStemmer\n",
    "import re\n",
    "import sys\n",
    "import warnings\n",
    "import keras\n",
    "from keras.preprocessing.text import Tokenizer\n",
    "from keras.layers import Dense, Input, LSTM, Embedding, Dropout, Activation\n",
    "from keras.layers import Bidirectional, GlobalMaxPool1D\n",
    "from keras.models import Model\n",
    "from keras import initializers, regularizers, constraints, optimizers, layers\n",
    "from gensim.models import Word2Vec\n",
    "from sklearn.svm import SVC\n",
    "import os\n",
    "from typing import Tuple, List\n",
    "from functools import partial\n",
    "\n",
    "from torch.utils.data import Dataset, DataLoader, RandomSampler\n",
    "from torch.nn.utils.rnn import pad_sequence\n",
    "from transformers import BertTokenizer, BertModel, AdamW, get_linear_schedule_with_warmup, BertPreTrainedModel\n",
    "from sklearn.metrics import roc_auc_score\n",
    "\n",
    "\n",
    "if not sys.warnoptions:\n",
    "    warnings.simplefilter(\"ignore\")\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "fypuqi69_Y9P"
   },
   "source": [
    "### **EDA, Preprocessing and finding embeddings of labels**\n",
    "\n",
    "\n",
    "*   Below cell contains code for calculating meta data which will later be used for exploratory data analysis.\n",
    "\n",
    "*   Then there is general preprocessing code which is applied over the text like lemmatization, etc.\n",
    "*   Then there are functions which specifically preprocess the data for it to be fed to classical ML algorithms.\n",
    "\n",
    "\n",
    "*   Then we have the code to find the glove and bert embeddings of the labels. These embeddings will then be used in deep learning algorithms and MAGNET.\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "5ZJ_z_Ue79Ng"
   },
   "outputs": [],
   "source": [
    "stop_words = set(stopwords.words('english'))\n",
    "stop_words.update(['zero','one','two','three','four','five','six','seven','eight','nine','ten','may','also','across','among','beside','however','yet','within'])\n",
    "re_stop_words = re.compile(r\"\\b(\" + \"|\".join(stop_words) + \")\\\\W\", re.I)\n",
    "\n",
    "stemmer = SnowballStemmer(\"english\")\n",
    "\n",
    "def plot_len_of_sentences(text_cols):\n",
    "    comment_len = text_cols.len()\n",
    "\n",
    "    # plot the distribution of comment lengths\n",
    "    plt.figure(figsize=(8,4))\n",
    "    sns.histplot(comment_len, kde=False, bins=50, color=\"red\")\n",
    "    plt.xlabel(\"Comment Length (Number of words)\", fontsize=12)\n",
    "    plt.ylabel(\"Number of Comments\", fontsize=12)\n",
    "    plt.title(\"Distribution of comment Lengths\", fontsize=12)\n",
    "\n",
    "  print(\"Average comment length:\",int(sum(comment_len)/len(comment_len)),\" characters\")\n",
    "\n",
    "def plot_bar_graph(train_df,labels,title):\n",
    "    fig = plt.figure(figsize=(8,6))\n",
    "    ax = fig.add_axes([0,0,1,1])\n",
    "    total_count = []\n",
    "    for label in labels:\n",
    "        total_count.append(len(train_df[train_df[label] == 1]))\n",
    "    ax.bar(labels,total_count, color=['red', 'green', 'blue', 'purple', 'orange', 'yellow'])\n",
    "    for i,data in enumerate(total_count):\n",
    "        plt.text(i-.25, \n",
    "                data/total_count[i]+100, \n",
    "                total_count[i], \n",
    "                fontsize=12)\n",
    "    plt.title(title)\n",
    "    plt.xlabel('Labels')\n",
    "    plt.ylabel('Number of comments')\n",
    "\n",
    "def cleanHtml(sentence):\n",
    "    cleanr = re.compile('<.*?>')\n",
    "    cleantext = re.sub(cleanr, ' ', str(sentence))\n",
    "    return cleantext\n",
    "def cleanPunc(sentence): #function to clean the word of any punctuation or special characters\n",
    "    cleaned = re.sub(r'[?|!|\\'|\"|#]',r'',sentence)\n",
    "    cleaned = re.sub(r'[.|,|)|(|\\|/]',r' ',cleaned)\n",
    "    cleaned = cleaned.strip()\n",
    "    cleaned = cleaned.replace(\"\\n\",\" \")\n",
    "    return cleaned\n",
    "def keepAlpha(sentence):\n",
    "    alpha_sent = \"\"\n",
    "    for word in sentence.split():\n",
    "        alpha_word = re.sub('[^a-z A-Z]+', ' ', word)\n",
    "        alpha_sent += alpha_word\n",
    "        alpha_sent += \" \"\n",
    "    alpha_sent = alpha_sent.strip()\n",
    "    return alpha_sent\n",
    "\n",
    "def removeStopWords(sentence):\n",
    "    global re_stop_words\n",
    "    return re_stop_words.sub(\" \", sentence)\n",
    "\n",
    "def lemmatize(sentence):\n",
    "    lemmatizer = WordNetLemmatizer()\n",
    "    lemSentence = \"\"\n",
    "    for word in sentence.split():\n",
    "        lem = lemmatizer.lemmatize(word)\n",
    "        lemSentence += lem\n",
    "        lemSentence += \" \"\n",
    "    lemSentence = lemSentence.strip()\n",
    "    return lemSentence\n",
    "\n",
    "def preprocess_text(train_df):\n",
    "    train_df['comment_text'] = train_df['comment_text'].str.lower()\n",
    "    train_df['comment_text'] = train_df['comment_text'].apply(cleanHtml)\n",
    "    train_df['comment_text'] = train_df['comment_text'].apply(cleanPunc)\n",
    "    train_df['comment_text'] = train_df['comment_text'].apply(keepAlpha)\n",
    "    train_df['comment_text'] = train_df['comment_text'].apply(removeStopWords)\n",
    "    train_df['comment_text'] = train_df['comment_text'].apply(lemmatize)\n",
    "    return train_df\n",
    "\n",
    "def sample_dataset(train_df):\n",
    "    df = train_df[train_df['clean']==0]\n",
    "    df = df.append(train_df[train_df['clean']==1].sample(frac=1, random_state=200).iloc[:df.shape[0],:])\n",
    "    df = df.sample(frac=1,random_state=200)\n",
    "    return df\n",
    "\n",
    "#ML preprocessing\n",
    "def baseline_ml_preprocessing(X_train,X_test):\n",
    "    word_vectorizer = TfidfVectorizer(\n",
    "        strip_accents='unicode',     \n",
    "        analyzer='word',\n",
    "        max_features=3000,        \n",
    "        token_pattern=r'\\w{1,}',    \n",
    "        ngram_range=(1, 3),         \n",
    "        stop_words='english',\n",
    "        sublinear_tf=True)\n",
    "\n",
    "    word_vectorizer.fit(X_train)\n",
    "    ml_X_train = word_vectorizer.transform(X_train)\n",
    "    ml_X_test = word_vectorizer.transform(X_test)\n",
    "    return ml_X_train, ml_X_test\n",
    "\n",
    "def Sequential_Preprocessing(X_train, X_test, y_train, y_test,max_seq_len):\n",
    "    tokenizer = Tokenizer(num_words=20000,oov_token='<UNK>')\n",
    "    tokenizer.fit_on_texts(X_train)\n",
    "    X_train = tokenizer.texts_to_sequences(X_train)\n",
    "    X_test = tokenizer.texts_to_sequences(X_test)\n",
    "\n",
    "    X_train = torch.from_numpy(pad_sequences(X_train, maxlen=max_seq_len))\n",
    "    X_test = torch.from_numpy(pad_sequences(X_test, maxlen=max_seq_len))\n",
    "\n",
    "    y_train = torch.from_numpy(y_train)\n",
    "    y_test = torch.from_numpy(y_test)\n",
    "\n",
    "    return tokenizer, X_train, y_train, X_test, y_test\n",
    "\n",
    "def create_glove_embedding_matrix(tokenizer, glove_file_path):\n",
    "    glove_embeddings = {}\n",
    "    with open(glove_file_path, 'r', encoding='utf-8') as f:\n",
    "        for line in f:\n",
    "            parts = line.split()\n",
    "            word = parts[0]\n",
    "            vector = np.array(parts[1:], dtype=np.float32)\n",
    "            glove_embeddings[word] = vector\n",
    "\n",
    "    vocab_size = len(tokenizer.index_word) + 1\n",
    "    #File 300 dim expected\n",
    "    embedding_dim = 300\n",
    "    glove_embedding_matrix = torch.zeros(vocab_size, embedding_dim)\n",
    "\n",
    "    unknown_words = 0\n",
    "    for i in range(1, vocab_size):\n",
    "        word = tokenizer.index_word[i]\n",
    "        if word in glove_embeddings.keys():\n",
    "            glove_embedding_matrix[i] = torch.from_numpy(glove_embeddings[word]).float()\n",
    "        else:\n",
    "            unknown_words += 1\n",
    "\n",
    "    print(\"GloVe embedding matrix created!\")\n",
    "    print('Vocabulary size: {}'.format(vocab_size))\n",
    "    print('Total unknown words: {}'.format(unknown_words))\n",
    "\n",
    "    return glove_embeddings, glove_embedding_matrix\n",
    "\n",
    "def create_glove_label_embedding(labels, glove_embeddings):\n",
    "    glove_label_embedding = torch.zeros(len(labels), 300)\n",
    "\n",
    "    for index, label in enumerate(labels):\n",
    "        wrds = label.split('_')\n",
    "        for l in wrds:\n",
    "            if l in glove_embeddings.keys():\n",
    "                glove_label_embedding[index] += torch.from_numpy(glove_embeddings[l])\n",
    "        glove_label_embedding[index] /= len(wrds)\n",
    "\n",
    "    return glove_label_embedding\n",
    "\n",
    "def get_bert_embedding_matrix(tokenizer):\n",
    "    bert_model = BertModel.from_pretrained('bert-base-uncased')\n",
    "    bert_tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')\n",
    "    vocab_size = len(tokenizer.index_word)+1\n",
    "    bert_embedding_matrix = torch.zeros(vocab_size, 768)\n",
    "\n",
    "\n",
    "    tokens_tensor=[]\n",
    "    for i in tqdm(range(1, vocab_size)):\n",
    "        word = tokenizer.index_word[i]\n",
    "        tokens = bert_tokenizer.encode(word,add_special_tokens=False)[0]\n",
    "        tokens_tensor.append(tokens)\n",
    "    batch_size = 512\n",
    "    vocab_batches = [tokens_tensor[i:i+batch_size] for i in range(0, len(tokens_tensor), batch_size)]\n",
    "\n",
    "    with torch.no_grad():\n",
    "      bert_embeddings = []\n",
    "      for i in tqdm(range(len(vocab_batches))):\n",
    "        batch = vocab_batches[i]\n",
    "        inputs = {\n",
    "        'input_ids': torch.tensor([batch]),\n",
    "        'attention_mask': torch.ones(len(batch)).unsqueeze(0)\n",
    "        }\n",
    "        outputs = bert_model(**inputs)\n",
    "        batch_embeddings = outputs[0][0]\n",
    "        bert_embeddings.append(batch_embeddings)\n",
    "        \n",
    "    bert_embedding_matrix = torch.cat(bert_embeddings, dim=0)\n",
    "    return bert_embedding_matrix\n",
    "\n",
    "\n",
    "#Build adjacency matrix based on Co-Occurencies label\n",
    "def create_adjacency_matrix_cooccurance(data_label):\n",
    "  cooccur_matrix = np.zeros((data_label.shape[1], data_label.shape[1]), dtype=float)\n",
    "  for y in data_label:\n",
    "      y = list(y)\n",
    "      for i in range(len(y)):\n",
    "          for j in range(len(y)):\n",
    "            #data_label\n",
    "              if y[i] == 1 and y[j] == 1:\n",
    "                  cooccur_matrix[i, j] += 1\n",
    "  row_sums = data_label.sum(axis=0)\n",
    "\n",
    "  for i in range(cooccur_matrix.shape[0]):\n",
    "    for j in range(cooccur_matrix.shape[0]):\n",
    "      if row_sums[i]!=0:\n",
    "        cooccur_matrix[i][j]=cooccur_matrix[i,j]/row_sums[i]\n",
    "      else:\n",
    "        cooccur_matrix[i][j]=cooccur_matrix[i,j]\n",
    "  \n",
    "  return cooccur_matrix\n",
    "\n",
    "\n",
    "def check_accuracy(model, label_embedding, X, y):\n",
    "  model.eval()\n",
    "  with torch.no_grad():\n",
    "    out = model(X, label_embedding)\n",
    "    y_pred = torch.sigmoid(out.detach()).round().cpu()\n",
    "    f1score = f1_score(y, y_pred, average='micro')\n",
    "    hammingloss = hamming_loss(y, y_pred)\n",
    "  return hammingloss, f1score\n",
    "\n",
    "class dataset(Dataset):\n",
    "  def __init__(self, x, y):\n",
    "    self.x  = x\n",
    "    self.y = y\n",
    "  def __len__(self):\n",
    "    return len(self.x)\n",
    "  \n",
    "  def __getitem__(self, idx):\n",
    "    return self.x[idx], self.y[idx]\n",
    "\n",
    "def create_adjacency_matrix_xavier(data_label):\n",
    "  adj_matrix = torch.empty((data_label.shape[1], data_label.shape[1]))\n",
    "  adj_matrix = nn.init.xavier_uniform_(adj_matrix)\n",
    "  return adj_matrix"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "UeFNhqOw04lK"
   },
   "source": [
    "Since most of the tweets dont contain any hateful content, we add a new column called 'clean'. If this is marked as 1, then this means that the tweet does not contain any offensive content. If clean = 1, then all other labels will be marked as 0."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "fJp6iNspHawY",
    "outputId": "877738be-557e-4376-da1f-c29905343189"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Data shape: (159571, 8)\n",
      "Labels: Index(['toxic', 'severe_toxic', 'obscene', 'threat', 'insult', 'identity_hate',\n",
      "       'clean'],\n",
      "      dtype='object')\n"
     ]
    }
   ],
   "source": [
    "train_df = pd.read_csv('train.csv')\n",
    "print(\"Data shape:\",train_df.shape)\n",
    "#assigning labels\n",
    "\n",
    "#Adding new column of Clean Text\n",
    "arr=[]\n",
    "for i in range(train_df.shape[0]):\n",
    "  if (train_df.iloc[i,2:]==0).all():\n",
    "    arr.append(1)\n",
    "  else:\n",
    "    arr.append(0)\n",
    "train_df['clean'] = pd.Series(np.asarray(arr))\n",
    "labels = train_df.columns[2:]\n",
    "print(\"Labels:\",labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 206
    },
    "id": "gqEhaacKHdac",
    "outputId": "17214870-a2cd-4f55-d31b-64133ecb988c"
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "\n",
       "  <div id=\"df-a3db0ed6-fab8-4641-a6a4-784ccb966c37\">\n",
       "    <div class=\"colab-df-container\">\n",
       "      <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>id</th>\n",
       "      <th>comment_text</th>\n",
       "      <th>toxic</th>\n",
       "      <th>severe_toxic</th>\n",
       "      <th>obscene</th>\n",
       "      <th>threat</th>\n",
       "      <th>insult</th>\n",
       "      <th>identity_hate</th>\n",
       "      <th>clean</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0000997932d777bf</td>\n",
       "      <td>Explanation\\nWhy the edits made under my usern...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>000103f0d9cfb60f</td>\n",
       "      <td>D'aww! He matches this background colour I'm s...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>000113f07ec002fd</td>\n",
       "      <td>Hey man, I'm really not trying to edit war. It...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0001b41b1c6bb37e</td>\n",
       "      <td>\"\\nMore\\nI can't make any real suggestions on ...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0001d958c54c6e35</td>\n",
       "      <td>You, sir, are my hero. Any chance you remember...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>\n",
       "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-a3db0ed6-fab8-4641-a6a4-784ccb966c37')\"\n",
       "              title=\"Convert this dataframe to an interactive table.\"\n",
       "              style=\"display:none;\">\n",
       "        \n",
       "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
       "       width=\"24px\">\n",
       "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
       "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
       "  </svg>\n",
       "      </button>\n",
       "      \n",
       "  <style>\n",
       "    .colab-df-container {\n",
       "      display:flex;\n",
       "      flex-wrap:wrap;\n",
       "      gap: 12px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert {\n",
       "      background-color: #E8F0FE;\n",
       "      border: none;\n",
       "      border-radius: 50%;\n",
       "      cursor: pointer;\n",
       "      display: none;\n",
       "      fill: #1967D2;\n",
       "      height: 32px;\n",
       "      padding: 0 0 0 0;\n",
       "      width: 32px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert:hover {\n",
       "      background-color: #E2EBFA;\n",
       "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
       "      fill: #174EA6;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert {\n",
       "      background-color: #3B4455;\n",
       "      fill: #D2E3FC;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert:hover {\n",
       "      background-color: #434B5C;\n",
       "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
       "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
       "      fill: #FFFFFF;\n",
       "    }\n",
       "  </style>\n",
       "\n",
       "      <script>\n",
       "        const buttonEl =\n",
       "          document.querySelector('#df-a3db0ed6-fab8-4641-a6a4-784ccb966c37 button.colab-df-convert');\n",
       "        buttonEl.style.display =\n",
       "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
       "\n",
       "        async function convertToInteractive(key) {\n",
       "          const element = document.querySelector('#df-a3db0ed6-fab8-4641-a6a4-784ccb966c37');\n",
       "          const dataTable =\n",
       "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
       "                                                     [key], {});\n",
       "          if (!dataTable) return;\n",
       "\n",
       "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
       "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
       "            + ' to learn more about interactive tables.';\n",
       "          element.innerHTML = '';\n",
       "          dataTable['output_type'] = 'display_data';\n",
       "          await google.colab.output.renderOutput(dataTable, element);\n",
       "          const docLink = document.createElement('div');\n",
       "          docLink.innerHTML = docLinkHtml;\n",
       "          element.appendChild(docLink);\n",
       "        }\n",
       "      </script>\n",
       "    </div>\n",
       "  </div>\n",
       "  "
      ],
      "text/plain": [
       "                 id                                       comment_text  toxic  \\\n",
       "0  0000997932d777bf  Explanation\\nWhy the edits made under my usern...      0   \n",
       "1  000103f0d9cfb60f  D'aww! He matches this background colour I'm s...      0   \n",
       "2  000113f07ec002fd  Hey man, I'm really not trying to edit war. It...      0   \n",
       "3  0001b41b1c6bb37e  \"\\nMore\\nI can't make any real suggestions on ...      0   \n",
       "4  0001d958c54c6e35  You, sir, are my hero. Any chance you remember...      0   \n",
       "\n",
       "   severe_toxic  obscene  threat  insult  identity_hate  clean  \n",
       "0             0        0       0       0              0      1  \n",
       "1             0        0       0       0              0      1  \n",
       "2             0        0       0       0              0      1  \n",
       "3             0        0       0       0              0      1  \n",
       "4             0        0       0       0              0      1  "
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 431
    },
    "id": "dn3xv3YIvIrO",
    "outputId": "f09e979f-8e6b-49db-dd55-0a624aa770a6"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average comment length: 394  characters\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 800x400 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_len_of_sentences(train_df.comment_text.str)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 702
    },
    "id": "jIHG4g2bvzel",
    "outputId": "cbe6cab6-5e37-4b8f-9007-8e143ce72098"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 800x600 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#Without clean Text\n",
    "plot_bar_graph(train_df,labels[:-1],'Number of comments per label- Without CLEAN column')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 702
    },
    "id": "2KtM7GiRwrzq",
    "outputId": "7f395082-25b9-466d-b059-5b74aae54a64"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 800x600 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#With clean Text\n",
    "plot_bar_graph(train_df,labels,'Number of comments per label- With CLEAN column')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "drRaBbig1PbA"
   },
   "source": [
    "The above graph shows that most of the tweets dont contain any hate content."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 206
    },
    "id": "Nq_mjjB-Irte",
    "outputId": "af6057d4-32e7-4e9c-c154-a00fce67a893"
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "\n",
       "  <div id=\"df-e0f4bfe4-0540-43ac-a96c-fee0292baba8\">\n",
       "    <div class=\"colab-df-container\">\n",
       "      <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>id</th>\n",
       "      <th>comment_text</th>\n",
       "      <th>toxic</th>\n",
       "      <th>severe_toxic</th>\n",
       "      <th>obscene</th>\n",
       "      <th>threat</th>\n",
       "      <th>insult</th>\n",
       "      <th>identity_hate</th>\n",
       "      <th>clean</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0000997932d777bf</td>\n",
       "      <td>explanation edits made username hardcore metal...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>000103f0d9cfb60f</td>\n",
       "      <td>daww match background colour im seemingly stuc...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>000113f07ec002fd</td>\n",
       "      <td>hey man im really trying edit war guy constant...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0001b41b1c6bb37e</td>\n",
       "      <td>cant make real suggestion improvement wondered...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0001d958c54c6e35</td>\n",
       "      <td>sir hero chance remember page thats on</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>\n",
       "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-e0f4bfe4-0540-43ac-a96c-fee0292baba8')\"\n",
       "              title=\"Convert this dataframe to an interactive table.\"\n",
       "              style=\"display:none;\">\n",
       "        \n",
       "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
       "       width=\"24px\">\n",
       "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
       "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
       "  </svg>\n",
       "      </button>\n",
       "      \n",
       "  <style>\n",
       "    .colab-df-container {\n",
       "      display:flex;\n",
       "      flex-wrap:wrap;\n",
       "      gap: 12px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert {\n",
       "      background-color: #E8F0FE;\n",
       "      border: none;\n",
       "      border-radius: 50%;\n",
       "      cursor: pointer;\n",
       "      display: none;\n",
       "      fill: #1967D2;\n",
       "      height: 32px;\n",
       "      padding: 0 0 0 0;\n",
       "      width: 32px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert:hover {\n",
       "      background-color: #E2EBFA;\n",
       "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
       "      fill: #174EA6;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert {\n",
       "      background-color: #3B4455;\n",
       "      fill: #D2E3FC;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert:hover {\n",
       "      background-color: #434B5C;\n",
       "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
       "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
       "      fill: #FFFFFF;\n",
       "    }\n",
       "  </style>\n",
       "\n",
       "      <script>\n",
       "        const buttonEl =\n",
       "          document.querySelector('#df-e0f4bfe4-0540-43ac-a96c-fee0292baba8 button.colab-df-convert');\n",
       "        buttonEl.style.display =\n",
       "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
       "\n",
       "        async function convertToInteractive(key) {\n",
       "          const element = document.querySelector('#df-e0f4bfe4-0540-43ac-a96c-fee0292baba8');\n",
       "          const dataTable =\n",
       "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
       "                                                     [key], {});\n",
       "          if (!dataTable) return;\n",
       "\n",
       "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
       "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
       "            + ' to learn more about interactive tables.';\n",
       "          element.innerHTML = '';\n",
       "          dataTable['output_type'] = 'display_data';\n",
       "          await google.colab.output.renderOutput(dataTable, element);\n",
       "          const docLink = document.createElement('div');\n",
       "          docLink.innerHTML = docLinkHtml;\n",
       "          element.appendChild(docLink);\n",
       "        }\n",
       "      </script>\n",
       "    </div>\n",
       "  </div>\n",
       "  "
      ],
      "text/plain": [
       "                 id                                       comment_text  toxic  \\\n",
       "0  0000997932d777bf  explanation edits made username hardcore metal...      0   \n",
       "1  000103f0d9cfb60f  daww match background colour im seemingly stuc...      0   \n",
       "2  000113f07ec002fd  hey man im really trying edit war guy constant...      0   \n",
       "3  0001b41b1c6bb37e  cant make real suggestion improvement wondered...      0   \n",
       "4  0001d958c54c6e35             sir hero chance remember page thats on      0   \n",
       "\n",
       "   severe_toxic  obscene  threat  insult  identity_hate  clean  \n",
       "0             0        0       0       0              0      1  \n",
       "1             0        0       0       0              0      1  \n",
       "2             0        0       0       0              0      1  \n",
       "3             0        0       0       0              0      1  \n",
       "4             0        0       0       0              0      1  "
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_df = preprocess_text(train_df)\n",
    "train_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 241
    },
    "id": "Y5ly2dWsNhWU",
    "outputId": "d8a87805-1081-414c-f08f-fc252cded76a"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "<ipython-input-3-ecde23a7a0f1>:79: FutureWarning: The frame.append method is deprecated and will be removed from pandas in a future version. Use pandas.concat instead.\n",
      "  df = df.append(train_df[train_df['clean']==1].sample(frac=1, random_state=200).iloc[:df.shape[0],:])\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "\n",
       "  <div id=\"df-1bf04ba8-8617-4c40-ab68-b19a6590dc65\">\n",
       "    <div class=\"colab-df-container\">\n",
       "      <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>id</th>\n",
       "      <th>comment_text</th>\n",
       "      <th>toxic</th>\n",
       "      <th>severe_toxic</th>\n",
       "      <th>obscene</th>\n",
       "      <th>threat</th>\n",
       "      <th>insult</th>\n",
       "      <th>identity_hate</th>\n",
       "      <th>clean</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>134065</th>\n",
       "      <td>cd04ee4935924644</td>\n",
       "      <td>got ill start on</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>128040</th>\n",
       "      <td>acd929b3f6c5984b</td>\n",
       "      <td>barnstar anti vandalism barnstar professing lo...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>136471</th>\n",
       "      <td>da225210fb20b417</td>\n",
       "      <td>homey yo beat u up</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>112748</th>\n",
       "      <td>5b33d2be26a63953</td>\n",
       "      <td>encouraged feedback let clear bit first versio...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21589</th>\n",
       "      <td>38dd995c105311e6</td>\n",
       "      <td>shes insane zealot</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>\n",
       "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-1bf04ba8-8617-4c40-ab68-b19a6590dc65')\"\n",
       "              title=\"Convert this dataframe to an interactive table.\"\n",
       "              style=\"display:none;\">\n",
       "        \n",
       "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
       "       width=\"24px\">\n",
       "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
       "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
       "  </svg>\n",
       "      </button>\n",
       "      \n",
       "  <style>\n",
       "    .colab-df-container {\n",
       "      display:flex;\n",
       "      flex-wrap:wrap;\n",
       "      gap: 12px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert {\n",
       "      background-color: #E8F0FE;\n",
       "      border: none;\n",
       "      border-radius: 50%;\n",
       "      cursor: pointer;\n",
       "      display: none;\n",
       "      fill: #1967D2;\n",
       "      height: 32px;\n",
       "      padding: 0 0 0 0;\n",
       "      width: 32px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert:hover {\n",
       "      background-color: #E2EBFA;\n",
       "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
       "      fill: #174EA6;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert {\n",
       "      background-color: #3B4455;\n",
       "      fill: #D2E3FC;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert:hover {\n",
       "      background-color: #434B5C;\n",
       "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
       "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
       "      fill: #FFFFFF;\n",
       "    }\n",
       "  </style>\n",
       "\n",
       "      <script>\n",
       "        const buttonEl =\n",
       "          document.querySelector('#df-1bf04ba8-8617-4c40-ab68-b19a6590dc65 button.colab-df-convert');\n",
       "        buttonEl.style.display =\n",
       "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
       "\n",
       "        async function convertToInteractive(key) {\n",
       "          const element = document.querySelector('#df-1bf04ba8-8617-4c40-ab68-b19a6590dc65');\n",
       "          const dataTable =\n",
       "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
       "                                                     [key], {});\n",
       "          if (!dataTable) return;\n",
       "\n",
       "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
       "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
       "            + ' to learn more about interactive tables.';\n",
       "          element.innerHTML = '';\n",
       "          dataTable['output_type'] = 'display_data';\n",
       "          await google.colab.output.renderOutput(dataTable, element);\n",
       "          const docLink = document.createElement('div');\n",
       "          docLink.innerHTML = docLinkHtml;\n",
       "          element.appendChild(docLink);\n",
       "        }\n",
       "      </script>\n",
       "    </div>\n",
       "  </div>\n",
       "  "
      ],
      "text/plain": [
       "                      id                                       comment_text  \\\n",
       "134065  cd04ee4935924644                                   got ill start on   \n",
       "128040  acd929b3f6c5984b  barnstar anti vandalism barnstar professing lo...   \n",
       "136471  da225210fb20b417                                 homey yo beat u up   \n",
       "112748  5b33d2be26a63953  encouraged feedback let clear bit first versio...   \n",
       "21589   38dd995c105311e6                                 shes insane zealot   \n",
       "\n",
       "        toxic  severe_toxic  obscene  threat  insult  identity_hate  clean  \n",
       "134065      0             0        0       0       0              0      1  \n",
       "128040      0             0        0       0       0              0      1  \n",
       "136471      0             0        0       1       0              0      0  \n",
       "112748      0             0        0       0       0              0      1  \n",
       "21589       1             0        0       0       1              0      0  "
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Get sampled Data by removing Rows which create Bias\n",
    "train_df = sample_dataset(train_df)\n",
    "train_df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "10EIdqHp1XRy"
   },
   "source": [
    "## **Removing bias from the data** \n",
    "\n",
    "Since majority of the tweets were not offensive, this means that the training data is quite biased. Therefore we took 16000 tweets which contained some offensive content, and an equal number of tweets which were clean.\n",
    "\n",
    "Doing this, we ensure that our training data is not biased against a particular label."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 702
    },
    "id": "SNAq7V8QI7H0",
    "outputId": "f44bcd81-567d-4e34-d07a-a7d53b835e00"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 800x600 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#modified dataset\n",
    "plot_bar_graph(train_df,labels,'Processed Balanced Data with columns')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "3VXc27RQItFs",
    "outputId": "6987a3c0-dc16-47cf-9f3f-83d26498d5b2"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(32450,) (32450, 7)\n",
      "(25960,) (6490,) (25960, 7) (6490, 7)\n"
     ]
    }
   ],
   "source": [
    "X = train_df['comment_text'].values\n",
    "y = train_df.iloc[:,2:].values\n",
    "print(X.shape, y.shape)\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2,random_state=200, stratify=y[:,-1])\n",
    "print(X_train.shape, X_test.shape, y_train.shape, y_test.shape )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "YFf_9Ndk1CYW",
    "outputId": "5c166f2f-42fb-4ccd-fb7b-6a0b93048773"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "X_train shape for baseline models: (25960, 3000)\n",
      "X_test shape for baseline models: (6490, 3000)\n"
     ]
    }
   ],
   "source": [
    "ml_X_train, ml_X_test = baseline_ml_preprocessing(X_train,X_test)\n",
    "print(\"X_train shape for baseline models:\",ml_X_train.shape)\n",
    "print(\"X_test shape for baseline models:\",ml_X_test.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "PjSJbPFS1d1x"
   },
   "source": [
    "**Training using classical ML algorithms**\n",
    "\n",
    "Now we have our train and test data ready to be used in ML algorithms. The are prefixed with ml_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "sv-csWp6TcQ4"
   },
   "outputs": [],
   "source": [
    "scores_df = pd.DataFrame(columns=['model_name','micro_avg_f1_score','hamming_loss','accuracy'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "5gGD87321gru"
   },
   "source": [
    "We test the following algorithms - \n",
    "\n",
    "\n",
    "\n",
    "*   Binary Relevance\n",
    "*   Classifier Chain\n",
    "*   One Vs Rest\n",
    "*   Label PowerSet\n",
    "*   Hierarchical SVM\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "W8mBeuS52IvS"
   },
   "outputs": [],
   "source": [
    "class HierarchicalSVM:\n",
    "    def __init__(self):\n",
    "        self.classifiers = {}#Dictionary to store the classifiers for each label\n",
    "\n",
    "    def train(self, X_train, Y_train):\n",
    "        num_labels = Y_train.shape[1]\n",
    "        # Train the classifiers for each label\n",
    "        for label_idx in range(num_labels):\n",
    "            label = str(label_idx)\n",
    "            Y_label = Y_train[:, label_idx]\n",
    "            classifier = SVC(kernel='linear')\n",
    "            classifier.fit(X_train, Y_label)\n",
    "            self.classifiers[label] = classifier\n",
    "\n",
    "    def predict(self, X_test):\n",
    "        num_labels = len(self.classifiers)\n",
    "        num_samples = X_test.shape[0]\n",
    "        predictions = np.zeros((num_samples, num_labels))\n",
    "        for label_idx, classifier in self.classifiers.items():\n",
    "            label_predictions = classifier.predict(X_test)\n",
    "            predictions[:, int(label_idx)] = label_predictions\n",
    "        return predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "Ktlq-F-gPEHz",
    "outputId": "123aa9a4-72e0-4372-cfc3-8e8dbc7b5b5a"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "====================BR============================\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "<ipython-input-16-e632f2dc34b0>:6: FutureWarning: The frame.append method is deprecated and will be removed from pandas in a future version. Use pandas.concat instead.\n",
      "  scores_df = scores_df.append(arr,ignore_index=True)\n",
      "/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classification.py:1344: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in samples with no predicted labels. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n",
      "/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classification.py:1344: UndefinedMetricWarning: Recall and F-score are ill-defined and being set to 0.0 in samples with no true labels. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.84      0.89      0.86      2878\n",
      "           1       0.20      0.55      0.29       108\n",
      "           2       0.69      0.88      0.78      1335\n",
      "           3       0.11      0.71      0.19        17\n",
      "           4       0.58      0.78      0.66      1166\n",
      "           5       0.26      0.69      0.37        97\n",
      "\n",
      "   micro avg       0.68      0.85      0.76      5601\n",
      "   macro avg       0.45      0.75      0.53      5601\n",
      "weighted avg       0.72      0.85      0.78      5601\n",
      " samples avg       0.33      0.37      0.34      5601\n",
      "\n",
      "====================Classifier Chain============================\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "<ipython-input-16-e632f2dc34b0>:15: FutureWarning: The frame.append method is deprecated and will be removed from pandas in a future version. Use pandas.concat instead.\n",
      "  scores_df = scores_df.append(arr,ignore_index=True)\n",
      "/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classification.py:1344: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in samples with no predicted labels. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n",
      "/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classification.py:1344: UndefinedMetricWarning: Recall and F-score are ill-defined and being set to 0.0 in samples with no true labels. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.84      0.89      0.86      2878\n",
      "           1       0.18      0.54      0.28       100\n",
      "           2       0.72      0.87      0.79      1413\n",
      "           3       0.10      0.69      0.18        16\n",
      "           4       0.63      0.73      0.68      1349\n",
      "           5       0.30      0.75      0.43       104\n",
      "\n",
      "   micro avg       0.70      0.84      0.76      5860\n",
      "   macro avg       0.46      0.74      0.53      5860\n",
      "weighted avg       0.74      0.84      0.78      5860\n",
      " samples avg       0.34      0.36      0.34      5860\n",
      "\n",
      "====================One Vs Rest============================\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "<ipython-input-16-e632f2dc34b0>:24: FutureWarning: The frame.append method is deprecated and will be removed from pandas in a future version. Use pandas.concat instead.\n",
      "  scores_df = scores_df.append(arr,ignore_index=True)\n",
      "/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classification.py:1344: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in samples with no predicted labels. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n",
      "/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classification.py:1344: UndefinedMetricWarning: Recall and F-score are ill-defined and being set to 0.0 in samples with no true labels. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.84      0.89      0.86      2878\n",
      "           1       0.20      0.55      0.29       108\n",
      "           2       0.69      0.88      0.78      1335\n",
      "           3       0.11      0.71      0.19        17\n",
      "           4       0.58      0.78      0.66      1166\n",
      "           5       0.26      0.69      0.37        97\n",
      "\n",
      "   micro avg       0.68      0.85      0.76      5601\n",
      "   macro avg       0.45      0.75      0.53      5601\n",
      "weighted avg       0.72      0.85      0.78      5601\n",
      " samples avg       0.33      0.37      0.34      5601\n",
      "\n",
      "====================Label PowerSet============================\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "<ipython-input-16-e632f2dc34b0>:32: FutureWarning: The frame.append method is deprecated and will be removed from pandas in a future version. Use pandas.concat instead.\n",
      "  scores_df = scores_df.append(arr,ignore_index=True)\n",
      "/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classification.py:1344: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in samples with no predicted labels. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n",
      "/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classification.py:1344: UndefinedMetricWarning: Recall and F-score are ill-defined and being set to 0.0 in samples with no true labels. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.76      0.91      0.83      2550\n",
      "           1       0.13      0.58      0.21        64\n",
      "           2       0.62      0.88      0.73      1200\n",
      "           3       0.05      0.83      0.09         6\n",
      "           4       0.55      0.78      0.64      1098\n",
      "           5       0.18      0.68      0.29        71\n",
      "\n",
      "   micro avg       0.62      0.87      0.72      4989\n",
      "   macro avg       0.38      0.78      0.46      4989\n",
      "weighted avg       0.66      0.87      0.75      4989\n",
      " samples avg       0.29      0.34      0.30      4989\n",
      "\n",
      "====================Hierarchical SVM============================\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.85      0.89      0.87      2900\n",
      "           1       0.06      0.71      0.11        24\n",
      "           2       0.74      0.88      0.80      1426\n",
      "           3       0.11      0.63      0.19        19\n",
      "           4       0.63      0.76      0.69      1290\n",
      "           5       0.31      0.73      0.43       111\n",
      "\n",
      "   micro avg       0.71      0.86      0.78      5770\n",
      "   macro avg       0.45      0.77      0.52      5770\n",
      "weighted avg       0.76      0.86      0.80      5770\n",
      " samples avg       0.35      0.38      0.35      5770\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "<ipython-input-16-e632f2dc34b0>:39: FutureWarning: The frame.append method is deprecated and will be removed from pandas in a future version. Use pandas.concat instead.\n",
      "  scores_df = scores_df.append(arr,ignore_index=True)\n",
      "/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classification.py:1344: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in samples with no predicted labels. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n",
      "/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classification.py:1344: UndefinedMetricWarning: Recall and F-score are ill-defined and being set to 0.0 in samples with no true labels. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n"
     ]
    }
   ],
   "source": [
    "print(\"====================BR============================\")\n",
    "model = BinaryRelevance(LogisticRegression(solver='sag'))\n",
    "model.fit(ml_X_train,  y_train[:,:-1])\n",
    "y_pred = model.predict(ml_X_test)\n",
    "arr={'model_name':'BR','micro_avg_f1_score':f1_score(y_test[:,:-1], y_pred, average=\"micro\"),'hamming_loss':hamming_loss(y_test[:,:-1], y_pred),'accuracy':accuracy_score(y_test[:,:-1],y_pred)}\n",
    "scores_df = scores_df.append(arr,ignore_index=True)\n",
    "print(classification_report(y_pred,y_test[:,:-1]))\n",
    "\n",
    "print(\"====================Classifier Chain============================\")\n",
    "\n",
    "model = ClassifierChain(LogisticRegression(solver='sag'))\n",
    "model.fit(ml_X_train, y_train[:,:-1])\n",
    "y_pred = model.predict(ml_X_test)\n",
    "arr={'model_name':'Classifier_Chain','micro_avg_f1_score':f1_score(y_test[:,:-1], y_pred, average=\"micro\"),'hamming_loss':hamming_loss(y_test[:,:-1], y_pred),'accuracy':accuracy_score(y_test[:,:-1],y_pred)}\n",
    "scores_df = scores_df.append(arr,ignore_index=True)\n",
    "print(classification_report(y_pred,y_test[:,:-1]))\n",
    "\n",
    "print(\"====================One Vs Rest============================\")\n",
    "\n",
    "model = OneVsRestClassifier(LogisticRegression())\n",
    "model.fit(ml_X_train, y_train[:,:-1])\n",
    "y_pred = model.predict(ml_X_test)\n",
    "arr={'model_name':'One_vs_rest','micro_avg_f1_score':f1_score(y_test[:,:-1], y_pred, average=\"micro\"),'hamming_loss':hamming_loss(y_test[:,:-1], y_pred),'accuracy':accuracy_score(y_test[:,:-1],y_pred)}\n",
    "scores_df = scores_df.append(arr,ignore_index=True)\n",
    "print(classification_report(y_pred,y_test[:,:-1]))\n",
    "\n",
    "print(\"====================Label PowerSet============================\")\n",
    "model = LabelPowerset(LogisticRegression(max_iter=1000))\n",
    "model.fit(ml_X_train, y_train[:,:-1])\n",
    "y_pred = model.predict(ml_X_test)\n",
    "arr={'model_name':'Label_Powerset','micro_avg_f1_score':f1_score(y_test[:,:-1], y_pred, average=\"micro\"),'hamming_loss':hamming_loss(y_test[:,:-1], y_pred),'accuracy':accuracy_score(y_test[:,:-1],y_pred)}\n",
    "scores_df = scores_df.append(arr,ignore_index=True)\n",
    "print(classification_report(y_pred,y_test[:,:-1]))\n",
    "print(\"====================Hierarchical SVM============================\")\n",
    "model = HierarchicalSVM()\n",
    "model.train(ml_X_train, y_train[:,:-1])\n",
    "y_pred = model.predict(ml_X_test)\n",
    "arr={'model_name':'Hierarchical_SVM','micro_avg_f1_score':f1_score(y_test[:,:-1], y_pred, average=\"micro\"),'hamming_loss':hamming_loss(y_test[:,:-1], y_pred),'accuracy':accuracy_score(y_test[:,:-1],y_pred)}\n",
    "scores_df = scores_df.append(arr,ignore_index=True)\n",
    "print(classification_report(y_pred,y_test[:,:-1]))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "9fZc_R161qYv"
   },
   "source": [
    "**Summarizing the results of classical ML algorithms**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 206
    },
    "id": "9xvICzxDoAK0",
    "outputId": "bda10ddc-3264-41a5-ff3d-8c0c4c397e34"
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "\n",
       "  <div id=\"df-b80468d2-959f-4241-bd09-0cb371491ef6\">\n",
       "    <div class=\"colab-df-container\">\n",
       "      <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>model_name</th>\n",
       "      <th>micro_avg_f1_score</th>\n",
       "      <th>hamming_loss</th>\n",
       "      <th>accuracy</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>BR</td>\n",
       "      <td>0.759005</td>\n",
       "      <td>0.077838</td>\n",
       "      <td>0.661787</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Classifier_Chain</td>\n",
       "      <td>0.764568</td>\n",
       "      <td>0.077607</td>\n",
       "      <td>0.668259</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>One_vs_rest</td>\n",
       "      <td>0.759005</td>\n",
       "      <td>0.077838</td>\n",
       "      <td>0.661787</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Label_Powerset</td>\n",
       "      <td>0.722106</td>\n",
       "      <td>0.085388</td>\n",
       "      <td>0.653621</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Hierarchical_SVM</td>\n",
       "      <td>0.775302</td>\n",
       "      <td>0.073549</td>\n",
       "      <td>0.671957</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>\n",
       "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-b80468d2-959f-4241-bd09-0cb371491ef6')\"\n",
       "              title=\"Convert this dataframe to an interactive table.\"\n",
       "              style=\"display:none;\">\n",
       "        \n",
       "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
       "       width=\"24px\">\n",
       "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
       "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
       "  </svg>\n",
       "      </button>\n",
       "      \n",
       "  <style>\n",
       "    .colab-df-container {\n",
       "      display:flex;\n",
       "      flex-wrap:wrap;\n",
       "      gap: 12px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert {\n",
       "      background-color: #E8F0FE;\n",
       "      border: none;\n",
       "      border-radius: 50%;\n",
       "      cursor: pointer;\n",
       "      display: none;\n",
       "      fill: #1967D2;\n",
       "      height: 32px;\n",
       "      padding: 0 0 0 0;\n",
       "      width: 32px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert:hover {\n",
       "      background-color: #E2EBFA;\n",
       "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
       "      fill: #174EA6;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert {\n",
       "      background-color: #3B4455;\n",
       "      fill: #D2E3FC;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert:hover {\n",
       "      background-color: #434B5C;\n",
       "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
       "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
       "      fill: #FFFFFF;\n",
       "    }\n",
       "  </style>\n",
       "\n",
       "      <script>\n",
       "        const buttonEl =\n",
       "          document.querySelector('#df-b80468d2-959f-4241-bd09-0cb371491ef6 button.colab-df-convert');\n",
       "        buttonEl.style.display =\n",
       "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
       "\n",
       "        async function convertToInteractive(key) {\n",
       "          const element = document.querySelector('#df-b80468d2-959f-4241-bd09-0cb371491ef6');\n",
       "          const dataTable =\n",
       "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
       "                                                     [key], {});\n",
       "          if (!dataTable) return;\n",
       "\n",
       "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
       "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
       "            + ' to learn more about interactive tables.';\n",
       "          element.innerHTML = '';\n",
       "          dataTable['output_type'] = 'display_data';\n",
       "          await google.colab.output.renderOutput(dataTable, element);\n",
       "          const docLink = document.createElement('div');\n",
       "          docLink.innerHTML = docLinkHtml;\n",
       "          element.appendChild(docLink);\n",
       "        }\n",
       "      </script>\n",
       "    </div>\n",
       "  </div>\n",
       "  "
      ],
      "text/plain": [
       "         model_name  micro_avg_f1_score  hamming_loss  accuracy\n",
       "0                BR            0.759005      0.077838  0.661787\n",
       "1  Classifier_Chain            0.764568      0.077607  0.668259\n",
       "2       One_vs_rest            0.759005      0.077838  0.661787\n",
       "3    Label_Powerset            0.722106      0.085388  0.653621\n",
       "4  Hierarchical_SVM            0.775302      0.073549  0.671957"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "scores_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "AIDIGpsUoANk"
   },
   "outputs": [],
   "source": [
    "#DL preprocessing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "syMIcvhN3DSi",
    "outputId": "14d6f50a-379f-44a4-ea28-c4202c3a5ba4"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "torch.Size([25960, 100]) torch.Size([6490, 100]) torch.Size([25960, 7]) torch.Size([6490, 7])\n"
     ]
    }
   ],
   "source": [
    "max_seq_len = 100\n",
    "tokenizer, X_train, y_train, X_test, y_test = Sequential_Preprocessing(X_train, X_test, y_train, y_test,max_seq_len)\n",
    "print(X_train.shape, X_test.shape, y_train.shape, y_test.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "FKtUXEhk1uca"
   },
   "source": [
    "Finding the glove and bert embeddings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "g2n1qO3O6jbv",
    "outputId": "844b9bb6-14ff-44e6-d4df-413e6690d7d9"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "GloVe embedding matrix created!\n",
      "Vocabulary size: 50685\n",
      "Total unknown words: 16419\n"
     ]
    }
   ],
   "source": [
    "glove_embeddings, glove_embedding_matrix = create_glove_embedding_matrix(tokenizer,glove_file_path='glove.6B/glove.6B.300d.txt')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "EtlGRy9d64KT",
    "outputId": "5a8444b0-4d67-44cd-ac23-0c686cbb661e"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "torch.Size([7, 300])\n"
     ]
    }
   ],
   "source": [
    "glove_label_embeddings = create_glove_label_embedding(labels,glove_embeddings)\n",
    "print(glove_label_embeddings.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "WvK2JHGOzGRf"
   },
   "outputs": [],
   "source": [
    "# word2vec_model = Word2Vec.load('path/to/word2vec_model')\n",
    "\n",
    "# word2vec_embedding = {word: word2vec_model.wv[word] for word in word2vec_model.wv.vocab}\n",
    "\n",
    "# # Define the vocabulary size and embedding dimensions\n",
    "# vocab_size = len(tokenizer.index_word)+1\n",
    "# embedding_dim = word2vec_model.vector_size\n",
    "\n",
    "# # Create the embedding matrix\n",
    "# word2vec_embedding_matrix = np.zeros((vocab_size, embedding_dim))\n",
    "# for i, word in enumerate(word2vec_embedding):\n",
    "#     word = tokenizer.index_word[i]\n",
    "#     if word in word2vec_embedding.keys():\n",
    "#       word2vec_embedding_matrix[i] = word2vec_embedding[word]\n",
    "#     else:\n",
    "#       continue\n",
    "# print(\"Word2Vec Embedding matrix created!\")\n",
    "# print(word2vec_embedding_matrix.shape)\n",
    "\n",
    "# word2vec_label_embedding = torch.zeros(len(labels),embedding_dim)\n",
    "\n",
    "# for index, label in enumerate(labels):\n",
    "#   wrds = label.split('_')\n",
    "#   for l in wrds:\n",
    "#     if l in word2vec_embedding.keys():\n",
    "#         word2vec_label_embedding[index] +=  torch.from_numpy(word2vec_embedding[l])\n",
    "#   word2vec_label_embedding[index]=word2vec_label_embedding[index]/len(wrds) \n",
    "\n",
    "# print(\"Word2Vec label embedding:\")\n",
    "# print(word2vec_label_embedding)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 269,
     "referenced_widgets": [
      "b84711d90cfe4f889c662ad03eb14716",
      "a976238571174f8d9192a31ff3d5c3d5",
      "88efcfea3974450c928c974f091744b3",
      "d1542aa04b344713a1134fd0302ba162",
      "b2692d83f09a486c9c3186ec73f64f1d",
      "2d36b2e4ce9247f8a7d556ea2d4fbbaf",
      "b244f49621604e20b90dc0659faa598b",
      "fca86cc61473462eaedaabf3fe3de8e8",
      "27de5ddcf9ad41d0abdb7fc40fd63ef4",
      "6dc757df6fc24eb6ab09ed6bc5363dc8",
      "48bbcd8bdd4e4f22892233bb9c22499c",
      "f7e6d2cac5424ab1b2b060715cd36a07",
      "3a4d463189de4c2199170550c9fcb720",
      "0d6e9828ec294966a09cc9878a0a2969",
      "871cf9dc73a44dffa439696c5b5091e3",
      "e45af4b8cb9f4c9991f18a8828a7015a",
      "2e784340cdd3474485b839f01ea8a735",
      "cf024c1e21af49bf85bac3f215ce2bec",
      "8fdcf4cf179947a5a412857c634b1105",
      "ac665d6cc49844529c2fabeadf40329e",
      "690d419a190a40779b69afe877924db8",
      "0e8de91b02a44247af2544b56d47c72d",
      "3d25d6797f1744a89f8a6b012b5ca18c",
      "bbf6d4148a6047c3843455648933c6c3",
      "ffd62e74ed4a42068989fdb81bcbeaa4",
      "e5dd935d0fd74ae0981f665bdc46ec99",
      "e0ce6cd8bb094835acfe968f0984ae60",
      "cf242fad209e41938ebdd7665a8aa2da",
      "d801e0df47184335bdfedf763e694d17",
      "a2897cf3cb854c33b457160eefd86cc7",
      "66bdaf101c87445db8a031b92014de63",
      "0bb8614697824a8580229d406e3f4bd5",
      "e2bdcc18437647eb82c9df4e2795e9cd",
      "21c623bb43e942448d07235afa5e8cb1",
      "0a477918c72c4786ab284f2a0e98d163",
      "96becd5034d94569b31b6810ea9083c7",
      "8ad3c9319e8144cd95cd91e61c28ea9a",
      "3c9858ad885d46d994580c555b371378",
      "540f7163e5594bb9ac1e14f74fd146b5",
      "ae8f4f6d59034045b45aaddf0af1b93b",
      "ce5fa2e065544e4eaf1d789fa0f09de4",
      "3267d5f59bac457fbea59039d2a3d741",
      "637ffa573b1849fe8aa7589d24dc7351",
      "4d0b3d5f7da44efebf8f40a26d2f878a"
     ]
    },
    "id": "oBR5ARr4_nbh",
    "outputId": "cc3efcda-7048-4371-d892-8d090de44903"
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b84711d90cfe4f889c662ad03eb14716",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Downloading (…)lve/main/config.json:   0%|          | 0.00/570 [00:00<?, ?B/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "f7e6d2cac5424ab1b2b060715cd36a07",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Downloading pytorch_model.bin:   0%|          | 0.00/440M [00:00<?, ?B/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of the model checkpoint at bert-base-uncased were not used when initializing BertModel: ['cls.predictions.transform.dense.weight', 'cls.seq_relationship.bias', 'cls.predictions.transform.LayerNorm.bias', 'cls.predictions.bias', 'cls.predictions.decoder.weight', 'cls.predictions.transform.dense.bias', 'cls.seq_relationship.weight', 'cls.predictions.transform.LayerNorm.weight']\n",
      "- This IS expected if you are initializing BertModel from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n",
      "- This IS NOT expected if you are initializing BertModel from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "3d25d6797f1744a89f8a6b012b5ca18c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Downloading (…)solve/main/vocab.txt:   0%|          | 0.00/232k [00:00<?, ?B/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "21c623bb43e942448d07235afa5e8cb1",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Downloading (…)okenizer_config.json:   0%|          | 0.00/28.0 [00:00<?, ?B/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50684/50684 [00:15<00:00, 3232.66it/s]\n",
      "100%|██████████| 99/99 [01:55<00:00,  1.17s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "torch.Size([50684, 768])\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "bert_embeddding_matrix = get_bert_embedding_matrix(tokenizer)\n",
    "print(bert_embeddding_matrix.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "lcscY1c8JJtg",
    "outputId": "df1f8069-ab7e-47f2-fd65-60e61994704b"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Device mapping:\n",
      "/job:localhost/replica:0/task:0/device:GPU:0 -> device: 0, name: Tesla T4, pci bus id: 0000:00:04.0, compute capability: 7.5\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From <ipython-input-22-ea7676897c82>:4: is_gpu_available (from tensorflow.python.framework.test_util) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Use `tf.config.list_physical_devices('GPU')` instead.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "GPU is Available? => True\n"
     ]
    }
   ],
   "source": [
    "#Checking GPU for tensorflow\n",
    "sess = tf.compat.v1.Session(config=tf.compat.v1.ConfigProto(log_device_placement=True))\n",
    "keras.backend.set_session(sess)\n",
    "print(\"GPU is Available? =>\",tf.test.is_gpu_available())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "egDcQ85dMJ47"
   },
   "outputs": [],
   "source": [
    "VOCAB_SIZE = len(tokenizer.index_word)+1\n",
    "#Preparing for KERAS Model\n",
    "glove_embd_mat= np.array(glove_embedding_matrix.to('cpu'))\n",
    "glove_embd_mat = np.vstack((np.zeros((1, glove_embd_mat.shape[1])), glove_embd_mat))\n",
    "bert_embd_mat= np.array(bert_embeddding_matrix.to('cpu').detach().numpy())\n",
    "bert_embd_mat = np.vstack((np.zeros((1, bert_embd_mat.shape[1])), bert_embd_mat))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Oh_kBwjc15kp"
   },
   "source": [
    "### Running deep learning algorithms\n",
    "\n",
    "The algorithms used are - \n",
    "* BiLSTM using glove embeddings\n",
    "* BiLSTM using BERT embeddings\n",
    "* CNN using BERT embeddings\n",
    "* CNN usin BERT embeddings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "ew-oCvcKw-nL",
    "outputId": "d557dac9-e921-4fb9-a673-6764ed4bcc21"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Layer lstm will not use cuDNN kernels since it doesn't meet the criteria. It will use a generic GPU kernel as fallback when running on GPU.\n",
      "WARNING:tensorflow:Layer lstm will not use cuDNN kernels since it doesn't meet the criteria. It will use a generic GPU kernel as fallback when running on GPU.\n",
      "WARNING:tensorflow:Layer lstm will not use cuDNN kernels since it doesn't meet the criteria. It will use a generic GPU kernel as fallback when running on GPU.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "=====================Glove BiLSTM===========================\n",
      "Epoch 1/3\n",
      "102/102 [==============================] - 91s 828ms/step - loss: 0.3949 - accuracy: 0.6428\n",
      "Epoch 2/3\n",
      "102/102 [==============================] - 77s 756ms/step - loss: 0.2204 - accuracy: 0.8665\n",
      "Epoch 3/3\n",
      "102/102 [==============================] - 79s 778ms/step - loss: 0.1794 - accuracy: 0.9057\n",
      "26/26 [==============================] - 2s 45ms/step\n",
      "         model_name  micro_avg_f1_score  hamming_loss  accuracy\n",
      "0                BR            0.768250      0.076117  0.667180\n",
      "1  Classifier_Chain            0.769028      0.076451  0.672111\n",
      "2       One_vs_rest            0.765692      0.076117  0.667180\n",
      "3    Label_Powerset            0.734849      0.082357  0.661941\n",
      "4  Hierarchical_SVM            0.776746      0.073472  0.672265\n",
      "5         CNN_glove            0.815412      0.078670  0.657473\n",
      "6      BiLSTM_glove            0.828584      0.075083  0.667951\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.90      0.88      0.89      3113\n",
      "           1       0.17      0.71      0.28        72\n",
      "           2       0.81      0.82      0.81      1687\n",
      "           3       0.00      0.00      0.00         0\n",
      "           4       0.72      0.72      0.72      1567\n",
      "           5       0.03      0.80      0.06        10\n",
      "           6       0.91      0.92      0.91      3229\n",
      "\n",
      "   micro avg       0.81      0.85      0.83      9678\n",
      "   macro avg       0.51      0.69      0.52      9678\n",
      "weighted avg       0.85      0.85      0.85      9678\n",
      " samples avg       0.83      0.84      0.82      9678\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "<ipython-input-55-3ae34c0b4e40>:17: FutureWarning: The frame.append method is deprecated and will be removed from pandas in a future version. Use pandas.concat instead.\n",
      "  scores_df = scores_df.append(arr,ignore_index=True)\n",
      "/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classification.py:1344: UndefinedMetricWarning: Recall and F-score are ill-defined and being set to 0.0 in labels with no true samples. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n",
      "/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classification.py:1344: UndefinedMetricWarning: Recall and F-score are ill-defined and being set to 0.0 in samples with no true labels. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n"
     ]
    }
   ],
   "source": [
    "print(\"=====================Glove BiLSTM===========================\")\n",
    "inp = Input(shape=(max_seq_len,))\n",
    "x = Embedding(VOCAB_SIZE+1, 300, weights=[glove_embd_mat])(inp)\n",
    "x = Bidirectional(LSTM(50, return_sequences=True, dropout=0.1, recurrent_dropout=0.1))(x)\n",
    "x = GlobalMaxPool1D()(x)\n",
    "x = Dense(50, activation=\"relu\")(x)\n",
    "x = Dropout(0.1)(x)\n",
    "x = Dense(7, activation=\"sigmoid\")(x)\n",
    "model = Model(inputs=inp, outputs=x)\n",
    "model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
    "\n",
    "model.fit(np.asarray(X_train),np.asarray(y_train), batch_size=256, epochs=3)\n",
    "y_pred = model.predict(np.asarray(X_test),batch_size=256, verbose=1)\n",
    "y_pred= np.where(y_pred > 0.5, 1, 0)\n",
    "\n",
    "arr={'model_name':'BiLSTM_glove','micro_avg_f1_score':f1_score(y_test, y_pred, average=\"micro\"),'hamming_loss':hamming_loss(y_test, y_pred),'accuracy':accuracy_score(y_test,y_pred)}\n",
    "scores_df = scores_df.append(arr,ignore_index=True)\n",
    "print(scores_df)\n",
    "print(classification_report(y_pred,y_test))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "cH3E0KFT7JO4",
    "outputId": "a9f3c0d2-d7fb-4ac9-99b2-2512b4c9d8e2"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "=====================Bert BiLSTM===========================\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Layer lstm will not use cuDNN kernels since it doesn't meet the criteria. It will use a generic GPU kernel as fallback when running on GPU.\n",
      "WARNING:tensorflow:Layer lstm will not use cuDNN kernels since it doesn't meet the criteria. It will use a generic GPU kernel as fallback when running on GPU.\n",
      "WARNING:tensorflow:Layer lstm will not use cuDNN kernels since it doesn't meet the criteria. It will use a generic GPU kernel as fallback when running on GPU.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/4\n",
      "102/102 [==============================] - 107s 952ms/step - loss: 0.3838 - accuracy: 0.6665\n",
      "Epoch 2/4\n",
      "102/102 [==============================] - 93s 909ms/step - loss: 0.1867 - accuracy: 0.8998\n",
      "Epoch 3/4\n",
      "102/102 [==============================] - 87s 856ms/step - loss: 0.1513 - accuracy: 0.9327\n",
      "Epoch 4/4\n",
      "102/102 [==============================] - 84s 827ms/step - loss: 0.1276 - accuracy: 0.9455\n",
      "26/26 [==============================] - 2s 69ms/step\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.88      0.89      0.88      3027\n",
      "           1       0.37      0.52      0.43       206\n",
      "           2       0.83      0.84      0.83      1671\n",
      "           3       0.00      0.00      0.00         0\n",
      "           4       0.68      0.73      0.70      1463\n",
      "           5       0.15      0.75      0.24        51\n",
      "           6       0.93      0.90      0.91      3332\n",
      "\n",
      "   micro avg       0.81      0.85      0.83      9750\n",
      "   macro avg       0.55      0.66      0.57      9750\n",
      "weighted avg       0.84      0.85      0.85      9750\n",
      " samples avg       0.83      0.84      0.82      9750\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classification.py:1344: UndefinedMetricWarning: Recall and F-score are ill-defined and being set to 0.0 in labels with no true samples. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n",
      "/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classification.py:1344: UndefinedMetricWarning: Recall and F-score are ill-defined and being set to 0.0 in samples with no true labels. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n"
     ]
    }
   ],
   "source": [
    "print(\"=====================Bert BiLSTM===========================\")\n",
    "inp = Input(shape=(max_seq_len,))\n",
    "x = Embedding(VOCAB_SIZE, 768, weights=[bert_embd_mat])(inp)\n",
    "x = Bidirectional(LSTM(50, return_sequences=True, dropout=0.1, recurrent_dropout=0.1))(x)\n",
    "x = GlobalMaxPool1D()(x)\n",
    "x = Dense(50, activation=\"relu\")(x)\n",
    "x = Dropout(0.1)(x)\n",
    "x = Dense(7, activation=\"sigmoid\")(x)\n",
    "model = Model(inputs=inp, outputs=x)\n",
    "model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
    "\n",
    "model.fit(np.asarray(X_train),np.asarray(y_train), batch_size=256, epochs=4)\n",
    "y_pred = model.predict(np.asarray(X_test),batch_size=256, verbose=1)\n",
    "y_pred= np.where(y_pred > 0.5, 1, 0)\n",
    "\n",
    "arr={'model_name':'BiLSTM_bert','micro_avg_f1_score':f1_score(y_test, y_pred, average=\"micro\"),'hamming_loss':hamming_loss(y_test, y_pred),'accuracy':accuracy_score(y_test,y_pred)}\n",
    "scores_df = scores_df.append(arr,ignore_index=True)\n",
    "print(scores_df)\n",
    "print(classification_report(y_pred,y_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "WugrHu17ESSL",
    "outputId": "fd97a433-6534-4ef2-8634-09cae0005715"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "=====================Glove CNN===========================\n",
      "Epoch 1/3\n",
      "102/102 [==============================] - 32s 198ms/step - loss: 0.3643 - accuracy: 0.6813\n",
      "Epoch 2/3\n",
      "102/102 [==============================] - 15s 150ms/step - loss: 0.2018 - accuracy: 0.8738\n",
      "Epoch 3/3\n",
      "102/102 [==============================] - 11s 108ms/step - loss: 0.1478 - accuracy: 0.9311\n",
      "26/26 [==============================] - 0s 7ms/step\n",
      "         model_name  micro_avg_f1_score  hamming_loss  accuracy\n",
      "0                BR            0.768250      0.076117  0.667180\n",
      "1  Classifier_Chain            0.769028      0.076451  0.672111\n",
      "2       One_vs_rest            0.765692      0.076117  0.667180\n",
      "3    Label_Powerset            0.734849      0.082357  0.661941\n",
      "4  Hierarchical_SVM            0.776746      0.073472  0.672265\n",
      "5         CNN_glove            0.815412      0.078670  0.657473\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.86      0.89      0.87      2977\n",
      "           1       0.13      0.72      0.22        53\n",
      "           2       0.70      0.88      0.78      1353\n",
      "           3       0.02      1.00      0.04         2\n",
      "           4       0.62      0.75      0.68      1305\n",
      "           5       0.26      0.72      0.38        93\n",
      "           6       0.92      0.89      0.90      3358\n",
      "\n",
      "   micro avg       0.77      0.86      0.82      9141\n",
      "   macro avg       0.50      0.83      0.55      9141\n",
      "weighted avg       0.82      0.86      0.83      9141\n",
      " samples avg       0.80      0.84      0.80      9141\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "<ipython-input-54-76d74db9ed5d>:19: FutureWarning: The frame.append method is deprecated and will be removed from pandas in a future version. Use pandas.concat instead.\n",
      "  scores_df = scores_df.append(arr,ignore_index=True)\n",
      "/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classification.py:1344: UndefinedMetricWarning: Recall and F-score are ill-defined and being set to 0.0 in samples with no true labels. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n"
     ]
    }
   ],
   "source": [
    "print(\"=====================Glove CNN===========================\")\n",
    "inp = Input(shape=(max_seq_len,))\n",
    "\n",
    "x = Embedding(VOCAB_SIZE+1, 300, weights=[glove_embd_mat])(inp)\n",
    "x = layers.Conv1D(filters=128, kernel_size=5, activation='relu')(x)\n",
    "x = GlobalMaxPool1D()(x)\n",
    "x = Dense(64, activation=\"relu\")(x)\n",
    "x = Dropout(0.1)(x)\n",
    "x = Dense(32, activation='relu')(x)\n",
    "x = Dense(7, activation=\"sigmoid\")(x)\n",
    "model = Model(inputs=inp, outputs=x)\n",
    "model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
    "\n",
    "model.fit(np.asarray(X_train),np.asarray(y_train), batch_size=256, epochs=3)\n",
    "y_pred = model.predict(np.asarray(X_test),batch_size=256, verbose=1)\n",
    "y_pred= np.where(y_pred > 0.5, 1, 0)\n",
    "\n",
    "arr={'model_name':'CNN_glove','micro_avg_f1_score':f1_score(y_test, y_pred, average=\"micro\"),'hamming_loss':hamming_loss(y_test, y_pred),'accuracy':accuracy_score(y_test,y_pred)}\n",
    "scores_df = scores_df.append(arr,ignore_index=True)\n",
    "print(scores_df)\n",
    "print(classification_report(y_pred,y_test))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "SCaVzFxvCLmj"
   },
   "outputs": [],
   "source": [
    "print(\"=====================Bert CNN===========================\")\n",
    "inp = Input(shape=(max_seq_len,))\n",
    "\n",
    "x = Embedding(VOCAB_SIZE, 768, weights=[bert_embd_mat])(inp)\n",
    "x = layers.Conv1D(filters=128, kernel_size=5, activation='relu')(x)\n",
    "x = GlobalMaxPool1D()(x)\n",
    "x = Dense(64, activation=\"relu\")(x)\n",
    "x = Dropout(0.1)(x)\n",
    "x = Dense(32, activation='relu')(x)\n",
    "x = Dense(7, activation=\"sigmoid\")(x)\n",
    "model = Model(inputs=inp, outputs=x)\n",
    "model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
    "\n",
    "model.fit(np.asarray(X_train),np.asarray(y_train), batch_size=256, epochs=3)\n",
    "y_pred = model.predict(np.asarray(X_test),batch_size=256, verbose=1)\n",
    "y_pred= np.where(y_pred > 0.5, 1, 0)\n",
    "\n",
    "arr={'model_name':'CNN_bert','micro_avg_f1_score':f1_score(y_test, y_pred, average=\"micro\"),'hamming_loss':hamming_loss(y_test, y_pred),'accuracy':accuracy_score(y_test,y_pred)}\n",
    "scores_df = scores_df.append(arr,ignore_index=True)\n",
    "print(scores_df)\n",
    "print(classification_report(y_pred,y_test))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "LX67ly932c2E"
   },
   "source": [
    "Summary of scores for all classical as well as deep learning models - "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 269
    },
    "id": "vWftT7OgSKBf",
    "outputId": "dac51064-d5ae-4c1a-f035-61621ff1dcc1"
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "\n",
       "  <div id=\"df-511feb40-e225-49ab-bf5b-70afcb06f1fe\">\n",
       "    <div class=\"colab-df-container\">\n",
       "      <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>model_name</th>\n",
       "      <th>micro_avg_f1_score</th>\n",
       "      <th>hamming_loss</th>\n",
       "      <th>accuracy</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>BR</td>\n",
       "      <td>0.768250</td>\n",
       "      <td>0.076117</td>\n",
       "      <td>0.667180</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Classifier_Chain</td>\n",
       "      <td>0.769028</td>\n",
       "      <td>0.076451</td>\n",
       "      <td>0.672111</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>One_vs_rest</td>\n",
       "      <td>0.765692</td>\n",
       "      <td>0.076117</td>\n",
       "      <td>0.667180</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Label_Powerset</td>\n",
       "      <td>0.734849</td>\n",
       "      <td>0.082357</td>\n",
       "      <td>0.661941</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Hierarchical_SVM</td>\n",
       "      <td>0.776746</td>\n",
       "      <td>0.073472</td>\n",
       "      <td>0.672265</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>CNN_glove</td>\n",
       "      <td>0.815412</td>\n",
       "      <td>0.078670</td>\n",
       "      <td>0.657473</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>BiLSTM_glove</td>\n",
       "      <td>0.828584</td>\n",
       "      <td>0.075083</td>\n",
       "      <td>0.667951</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>\n",
       "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-511feb40-e225-49ab-bf5b-70afcb06f1fe')\"\n",
       "              title=\"Convert this dataframe to an interactive table.\"\n",
       "              style=\"display:none;\">\n",
       "        \n",
       "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
       "       width=\"24px\">\n",
       "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
       "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
       "  </svg>\n",
       "      </button>\n",
       "      \n",
       "  <style>\n",
       "    .colab-df-container {\n",
       "      display:flex;\n",
       "      flex-wrap:wrap;\n",
       "      gap: 12px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert {\n",
       "      background-color: #E8F0FE;\n",
       "      border: none;\n",
       "      border-radius: 50%;\n",
       "      cursor: pointer;\n",
       "      display: none;\n",
       "      fill: #1967D2;\n",
       "      height: 32px;\n",
       "      padding: 0 0 0 0;\n",
       "      width: 32px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert:hover {\n",
       "      background-color: #E2EBFA;\n",
       "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
       "      fill: #174EA6;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert {\n",
       "      background-color: #3B4455;\n",
       "      fill: #D2E3FC;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert:hover {\n",
       "      background-color: #434B5C;\n",
       "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
       "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
       "      fill: #FFFFFF;\n",
       "    }\n",
       "  </style>\n",
       "\n",
       "      <script>\n",
       "        const buttonEl =\n",
       "          document.querySelector('#df-511feb40-e225-49ab-bf5b-70afcb06f1fe button.colab-df-convert');\n",
       "        buttonEl.style.display =\n",
       "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
       "\n",
       "        async function convertToInteractive(key) {\n",
       "          const element = document.querySelector('#df-511feb40-e225-49ab-bf5b-70afcb06f1fe');\n",
       "          const dataTable =\n",
       "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
       "                                                     [key], {});\n",
       "          if (!dataTable) return;\n",
       "\n",
       "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
       "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
       "            + ' to learn more about interactive tables.';\n",
       "          element.innerHTML = '';\n",
       "          dataTable['output_type'] = 'display_data';\n",
       "          await google.colab.output.renderOutput(dataTable, element);\n",
       "          const docLink = document.createElement('div');\n",
       "          docLink.innerHTML = docLinkHtml;\n",
       "          element.appendChild(docLink);\n",
       "        }\n",
       "      </script>\n",
       "    </div>\n",
       "  </div>\n",
       "  "
      ],
      "text/plain": [
       "         model_name  micro_avg_f1_score  hamming_loss  accuracy\n",
       "0                BR            0.768250      0.076117  0.667180\n",
       "1  Classifier_Chain            0.769028      0.076451  0.672111\n",
       "2       One_vs_rest            0.765692      0.076117  0.667180\n",
       "3    Label_Powerset            0.734849      0.082357  0.661941\n",
       "4  Hierarchical_SVM            0.776746      0.073472  0.672265\n",
       "5         CNN_glove            0.815412      0.078670  0.657473\n",
       "6      BiLSTM_glove            0.828584      0.075083  0.667951"
      ]
     },
     "execution_count": 56,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "scores_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "Glgo1Vi7Ey8E"
   },
   "outputs": [],
   "source": [
    "# import torch\n",
    "# import torch.nn as nn\n",
    "# import torch.optim as optim\n",
    "\n",
    "# class GraphConvLayer(nn.Module):\n",
    "#     def __init__(self, input_dim, output_dim):\n",
    "#         super(GraphConvLayer, self).__init__()\n",
    "#         self.linear = nn.Linear(input_dim, output_dim)\n",
    "\n",
    "#     def forward(self, x, adj):\n",
    "#         x = self.linear(x)\n",
    "#         print(x.shape, adj.shape)\n",
    "#         # x = torch.matmul(adj, x.unsqueeze(2)).squeeze(2)\n",
    "#         x = torch.matmul(adj, x)\n",
    "#         return x\n",
    "\n",
    "# class GraphNeuralNetwork(nn.Module):\n",
    "#     def __init__(self, input_dim, hidden_dim, output_dim,embeddings):\n",
    "#         super(GraphNeuralNetwork, self).__init__()\n",
    "#         self.embedding = nn.Embedding.from_pretrained(embeddings)\n",
    "#         self.conv1 = GraphConvLayer(input_dim, hidden_dim)\n",
    "#         self.conv2 = GraphConvLayer(hidden_dim, output_dim)\n",
    "#         self.relu = nn.ReLU()\n",
    "#         self.softmax = nn.Softmax(dim=1)\n",
    "\n",
    "#     def forward(self, x, adj):\n",
    "#         x = self.embedding(x)\n",
    "#         x = self.relu(self.conv1(x, adj))\n",
    "#         x = self.softmax(self.conv2(x, adj))\n",
    "#         return x"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "wsNbWlRPL3m1",
    "outputId": "7cd0443e-eabb-40e6-e794-2842c91277dc"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[1.0000, 0.1064, 0.5177, 0.0284, 0.4815, 0.0860, 0.0000],\n",
       "        [1.0000, 1.0000, 0.9509, 0.0675, 0.8611, 0.1972, 0.0000],\n",
       "        [0.9395, 0.1836, 1.0000, 0.0350, 0.7304, 0.1226, 0.0000],\n",
       "        [0.9380, 0.2372, 0.6361, 1.0000, 0.6577, 0.2022, 0.0000],\n",
       "        [0.9339, 0.1777, 0.7806, 0.0387, 1.0000, 0.1481, 0.0000],\n",
       "        [0.9213, 0.2248, 0.7235, 0.0656, 0.8180, 1.0000, 0.0000],\n",
       "        [0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 1.0000]],\n",
       "       dtype=torch.float64)"
      ]
     },
     "execution_count": 182,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# adj_matrix = create_adjacency_matrix_cooccurance(y_train.numpy())\n",
    "# adj_matrix = torch.tensor(adj_matrix)\n",
    "# adj_matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "AaWA_zMCE2aV"
   },
   "outputs": [],
   "source": [
    "\n",
    "# input_dim = 300  # Dimension of input features\n",
    "# hidden_dim = 250  # Hidden dimension in the GNN\n",
    "# output_dim = y_train.shape[1]  # Dimension of output labels\n",
    "# device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n",
    "# model = GraphNeuralNetwork(input_dim, hidden_dim, output_dim, glove_embedding_matrix).to(device)\n",
    "\n",
    "# # Define loss function and optimizer\n",
    "# loss_function = nn.BCEWithLogitsLoss()\n",
    "# optimizer = optim.Adam(model.parameters(), lr=0.01)\n",
    "\n",
    "# train_data = DataLoader(dataset(X_train, y_train), batch_size=256)\n",
    "# # Training loop\n",
    "# num_epochs = 10\n",
    "# for epoch in range(num_epochs):\n",
    "#     tl=0\n",
    "#     for i, (X,y) in enumerate(train_data):\n",
    "#       optimizer.zero_grad()\n",
    "#       print(X.shape, y.shape)\n",
    "#       output = model(X.to(device), adj_matrix.to(device))\n",
    "#       loss = loss_function(output, y_train).to(device)\n",
    "#       tl += loss.item()\n",
    "#       loss.backward()\n",
    "#       optimizer.step()\n",
    "#     print(f\"Epoch {epoch+1}/{num_epochs}, Loss: {tl}\")\n",
    "\n",
    "# # Once trained, you can use the trained model for prediction\n",
    "# # output = model(X_test, adj_matrix)\n",
    "# # y_ = torch.sigmoid(output)  # Apply sigmoid activation for multilabel classification\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "Pz3mNci4wEbP"
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "rFfoFUOWjchi"
   },
   "outputs": [],
   "source": [
    "# adj_matrix = torch.randn(7, 7)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "grHZPqiJPPEp"
   },
   "outputs": [],
   "source": [
    "def train(model, X_train, label_embedding, y_train,\n",
    "          total_epoch=250, batch_size=250, learning_rate=0.001,\n",
    "          save_path='./model.pt', state=None):\n",
    "  \n",
    "    device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n",
    "    model = model.to(device)\n",
    "    label_embedding = label_embedding.to(device)\n",
    "    criterion = nn.BCEWithLogitsLoss()\n",
    "    optimizer = optim.Adam(model.parameters(), lr=learning_rate)\n",
    "    train_data = DataLoader(dataset(X_train, y_train), batch_size=batch_size)\n",
    "  \n",
    "    best_loss = float('inf')\n",
    "    num_increasing_epochs = 0\n",
    "  \n",
    "    for epoch in range(1, total_epoch + 1):\n",
    "        running_loss = 0\n",
    "        y_pred = []\n",
    "        epoch_time = 0\n",
    "        model.train()\n",
    "      \n",
    "        for index, (X, y) in enumerate(train_data):\n",
    "            optimizer.zero_grad()\n",
    "            out = model(X.to(device), label_embedding)\n",
    "            loss = criterion(out, y.to(device).float())\n",
    "            loss.backward()\n",
    "            nn.utils.clip_grad_norm_(model.parameters(), max_norm=10)\n",
    "            optimizer.step()\n",
    "            y_pred.append(torch.sigmoid(out.detach()).round().cpu())\n",
    "            running_loss += loss.item()\n",
    "    \n",
    "        y_pred = torch.vstack(y_pred)\n",
    "        f1score = f1_score(y_train, y_pred, average='micro')\n",
    "        hammingloss = hamming_loss(y_train, y_pred)\n",
    "        print(f'epoch:{epoch} loss:{running_loss:.5f} hamming_loss:{hammingloss:.5f} micro_f1score:{f1score:.5f}')\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "gnEgDAiS2rUo"
   },
   "source": [
    "**The MAGNET CLASS**\n",
    "\n",
    "Here we will initialize the adjacency matrix using three methods - \n",
    "* Cooccurance intialization\n",
    "* Xaviour initialization\n",
    "* Random initialization\n",
    "\n",
    "Then we will be training the model using glove embeddings. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "ZCYRX5MHT8lO"
   },
   "outputs": [],
   "source": [
    "class MAGNET(nn.Module):\n",
    "  def __init__(self, input_size, hidden_size, adjacency, embeddings, heads=4, slope=0.01, dropout=0.5):\n",
    "    super(MAGNET, self).__init__()\n",
    "    self.embedding = nn.Embedding.from_pretrained(embeddings)\n",
    "    self.biLSTM = nn.LSTM(input_size,hidden_size,batch_first=True,bidirectional=True)\n",
    "    self.adjacency = nn.Parameter(adjacency)\n",
    "    self.dropout = nn.Dropout(dropout)\n",
    "    self.edge_weights = nn.Linear(hidden_size*2*2, 1, bias=False)\n",
    "    self.activation = nn.LeakyReLU(slope)\n",
    "    self.softmax = nn.Softmax(dim=1)\n",
    "    self.tanh = nn.Tanh()\n",
    "    self.heads = heads\n",
    "    self.transform_dim1 = nn.Linear(input_size, hidden_size*2, bias=False)\n",
    "    self.transform_dim2 = nn.Linear(hidden_size*2, hidden_size*2, bias=False)\n",
    "    self.transform_dimensions = [self.transform_dim1, self.transform_dim2]\n",
    " \n",
    "  def forward(self, token, label_embedding):\n",
    "      #BILSTM part\n",
    "      features = self.embedding(token)\n",
    "      out, (h, _) = self.biLSTM(features)\n",
    "      embedding = torch.cat([h[-2, :, :], h[-1, :, :]], dim=1)\n",
    "      embedding = self.dropout(embedding)\n",
    "\n",
    "      #GAT PART\n",
    "      for td in self.transform_dimensions: #Two Multiheaded GAT layers\n",
    "        outputs = []\n",
    "        for head in range(self.heads):\n",
    "          label_embed = td(label_embedding)   \n",
    "          n, embed_size = label_embed.shape\n",
    "\n",
    "          label_embed_combinations = label_embed.unsqueeze(1).expand(-1, n, -1)\n",
    "          label_embed_combinations = torch.cat([label_embed_combinations, label_embed.unsqueeze(0).expand(n, -1, -1)], dim=2)\n",
    "          e = self.activation(self.edge_weights(label_embed_combinations).squeeze(2))\n",
    "\n",
    "          attention_coefficients = self.tanh(torch.mul(e,self.adjacency))\n",
    "\n",
    "          new_h = torch.matmul(attention_coefficients.to(label_embed.dtype), label_embed)\n",
    "          outputs.append(new_h)\n",
    "        outputs = self.activation(torch.mean(torch.stack(outputs, dim=0),dim=0))\n",
    "\n",
    "        label_embedding = outputs\n",
    "      attention_features = self.dropout(label_embedding)\n",
    "      attention_features = attention_features.transpose(0, 1) \n",
    "      predicted_labels = torch.matmul(embedding, attention_features)\n",
    "      return predicted_labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "mRnu0sbvYtfp"
   },
   "outputs": [],
   "source": [
    "save_path = './glove_magnet_model.pth'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "wwSOrRqw2xRX"
   },
   "source": [
    "**Initializing using cooccurance matrix.**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "tVECE46ej14p",
    "outputId": "d806a2fd-86a0-4846-cc24-69520faa5d6b"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[1.0000, 0.1064, 0.5177, 0.0284, 0.4815, 0.0860, 0.0000],\n",
       "        [1.0000, 1.0000, 0.9509, 0.0675, 0.8611, 0.1972, 0.0000],\n",
       "        [0.9395, 0.1836, 1.0000, 0.0350, 0.7304, 0.1226, 0.0000],\n",
       "        [0.9380, 0.2372, 0.6361, 1.0000, 0.6577, 0.2022, 0.0000],\n",
       "        [0.9339, 0.1777, 0.7806, 0.0387, 1.0000, 0.1481, 0.0000],\n",
       "        [0.9213, 0.2248, 0.7235, 0.0656, 0.8180, 1.0000, 0.0000],\n",
       "        [0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 1.0000]],\n",
       "       dtype=torch.float64)"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "adj_matrix = create_adjacency_matrix_cooccurance(y_train.numpy())\n",
    "adj_matrix = torch.tensor(adj_matrix)\n",
    "adj_matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "TsidK00lw0-g"
   },
   "outputs": [],
   "source": [
    "model = MAGNET(300, 250, adj_matrix, glove_embedding_matrix,heads=8)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "M2Pk7zRyjm_6",
    "outputId": "209cf9f1-d335-4f8e-8bbd-529e584396a7"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch:1 loss:42.00155 hamming_loss:0.21614 micro_f1score:0.31529\n",
      "epoch:2 loss:26.34696 hamming_loss:0.10193 micro_f1score:0.76203\n",
      "epoch:3 loss:23.10602 hamming_loss:0.08924 micro_f1score:0.79374\n",
      "epoch:4 loss:21.62392 hamming_loss:0.08436 micro_f1score:0.80539\n",
      "epoch:5 loss:20.82383 hamming_loss:0.08149 micro_f1score:0.81237\n",
      "epoch:6 loss:19.96281 hamming_loss:0.07824 micro_f1score:0.82001\n",
      "epoch:7 loss:19.44956 hamming_loss:0.07678 micro_f1score:0.82343\n",
      "epoch:8 loss:18.98440 hamming_loss:0.07506 micro_f1score:0.82763\n",
      "epoch:9 loss:18.26785 hamming_loss:0.07254 micro_f1score:0.83354\n",
      "epoch:10 loss:17.87156 hamming_loss:0.07102 micro_f1score:0.83690\n",
      "epoch:11 loss:16.98031 hamming_loss:0.06738 micro_f1score:0.84586\n",
      "epoch:12 loss:16.91331 hamming_loss:0.06763 micro_f1score:0.84598\n",
      "epoch:13 loss:16.10156 hamming_loss:0.06499 micro_f1score:0.85160\n",
      "epoch:14 loss:15.34780 hamming_loss:0.06219 micro_f1score:0.85844\n",
      "epoch:15 loss:14.77733 hamming_loss:0.06005 micro_f1score:0.86325\n",
      "epoch:16 loss:14.56726 hamming_loss:0.05904 micro_f1score:0.86575\n",
      "epoch:17 loss:14.12070 hamming_loss:0.05745 micro_f1score:0.86943\n",
      "epoch:18 loss:13.89544 hamming_loss:0.05594 micro_f1score:0.87284\n",
      "epoch:19 loss:13.54217 hamming_loss:0.05474 micro_f1score:0.87575\n",
      "epoch:20 loss:15.74666 hamming_loss:0.06126 micro_f1score:0.86093\n",
      "epoch:21 loss:13.03494 hamming_loss:0.05337 micro_f1score:0.87929\n",
      "epoch:22 loss:12.66858 hamming_loss:0.05210 micro_f1score:0.88236\n",
      "epoch:23 loss:12.33123 hamming_loss:0.05074 micro_f1score:0.88537\n",
      "epoch:24 loss:11.91892 hamming_loss:0.04893 micro_f1score:0.88953\n",
      "epoch:25 loss:11.55608 hamming_loss:0.04751 micro_f1score:0.89307\n",
      "epoch:26 loss:11.07241 hamming_loss:0.04598 micro_f1score:0.89661\n",
      "epoch:27 loss:10.67259 hamming_loss:0.04428 micro_f1score:0.90035\n",
      "epoch:28 loss:10.44831 hamming_loss:0.04342 micro_f1score:0.90236\n",
      "epoch:29 loss:10.49969 hamming_loss:0.04306 micro_f1score:0.90326\n",
      "epoch:30 loss:10.40132 hamming_loss:0.04342 micro_f1score:0.90248\n",
      "epoch:31 loss:9.97144 hamming_loss:0.04121 micro_f1score:0.90749\n",
      "epoch:32 loss:9.74523 hamming_loss:0.04043 micro_f1score:0.90932\n",
      "epoch:33 loss:9.50815 hamming_loss:0.03931 micro_f1score:0.91179\n",
      "epoch:34 loss:9.17357 hamming_loss:0.03781 micro_f1score:0.91526\n",
      "epoch:35 loss:8.97913 hamming_loss:0.03713 micro_f1score:0.91691\n",
      "epoch:36 loss:8.88504 hamming_loss:0.03724 micro_f1score:0.91673\n",
      "epoch:37 loss:8.71159 hamming_loss:0.03616 micro_f1score:0.91900\n",
      "epoch:38 loss:8.53744 hamming_loss:0.03555 micro_f1score:0.92045\n",
      "epoch:39 loss:8.41028 hamming_loss:0.03539 micro_f1score:0.92076\n",
      "epoch:40 loss:8.16300 hamming_loss:0.03420 micro_f1score:0.92363\n",
      "epoch:41 loss:8.04597 hamming_loss:0.03378 micro_f1score:0.92457\n",
      "epoch:42 loss:7.94433 hamming_loss:0.03278 micro_f1score:0.92669\n",
      "epoch:43 loss:7.44690 hamming_loss:0.03111 micro_f1score:0.93049\n",
      "epoch:44 loss:7.27314 hamming_loss:0.03026 micro_f1score:0.93242\n",
      "epoch:45 loss:7.01415 hamming_loss:0.02935 micro_f1score:0.93447\n",
      "epoch:46 loss:6.96678 hamming_loss:0.02884 micro_f1score:0.93564\n",
      "epoch:47 loss:6.99887 hamming_loss:0.02897 micro_f1score:0.93532\n",
      "epoch:48 loss:7.14776 hamming_loss:0.02936 micro_f1score:0.93443\n",
      "epoch:49 loss:6.52055 hamming_loss:0.02690 micro_f1score:0.93999\n",
      "epoch:50 loss:6.52035 hamming_loss:0.02660 micro_f1score:0.94065\n"
     ]
    }
   ],
   "source": [
    "#Coocurance initialization\n",
    "train(model, X_train, glove_label_embeddings,y_train,total_epoch=50, save_path=save_path)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "X3Z_aVdr-S19"
   },
   "outputs": [],
   "source": [
    "torch.save(model,save_path)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "W24AhoVHccPp",
    "outputId": "1351d1f4-79b0-458c-ea47-4db207be5bcb"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 6490/6490 [00:48<00:00, 132.56it/s]\n"
     ]
    }
   ],
   "source": [
    "y_pred = []\n",
    "device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n",
    "for i in tqdm(range(X_test.shape[0])):\n",
    "  x = X_test[i].view(1,100)\n",
    "  out = model(x.to(device), glove_label_embeddings.to(device))\n",
    "  y_pred.append(torch.sigmoid(out.detach()).round().cpu())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "8I8DK0HtdW62"
   },
   "outputs": [],
   "source": [
    "y_pred = torch.vstack(y_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "GvLGZavIdJRd",
    "outputId": "cdf40b99-a1ce-4825-ee8d-7ac5201bafd1"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'model_name': 'MAGNET_Cooccurance', 'micro_avg_f1_score': 0.7934177464649947, 'hamming_loss': 0.09229583975346688, 'accuracy': 0.6140215716486903}\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.85      0.86      0.85      3001\n",
      "           1       0.37      0.36      0.36       298\n",
      "           2       0.75      0.79      0.77      1607\n",
      "           3       0.18      0.43      0.25        44\n",
      "           4       0.67      0.68      0.68      1536\n",
      "           5       0.44      0.46      0.45       251\n",
      "           6       0.90      0.87      0.88      3339\n",
      "\n",
      "   micro avg       0.79      0.80      0.79     10076\n",
      "   macro avg       0.59      0.64      0.61     10076\n",
      "weighted avg       0.79      0.80      0.80     10076\n",
      " samples avg       0.81      0.80      0.78     10076\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classification.py:1344: UndefinedMetricWarning: Recall and F-score are ill-defined and being set to 0.0 in samples with no true labels. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n"
     ]
    }
   ],
   "source": [
    "arr={'model_name':'MAGNET_Cooccurance','micro_avg_f1_score':f1_score(y_test, y_pred, average=\"micro\"),'hamming_loss':hamming_loss(y_test, y_pred),'accuracy':accuracy_score(y_test,y_pred)}\n",
    "# scores_df = scores_df.append(arr,ignore_index=True)\n",
    "print(arr)\n",
    "print(classification_report(y_pred,y_test))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "9kYKW9Er22ye"
   },
   "source": [
    "**Xaviour's initialization**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "pOZZyRLDese9",
    "outputId": "cf7b5f9e-c596-4203-8eea-27e4a40724c4"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "<ipython-input-64-f66315f0aa9a>:2: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor).\n",
      "  adj_matrix = torch.tensor(adj_matrix)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "tensor([[ 0.5724, -0.2907, -0.5935, -0.1678, -0.3640,  0.6305,  0.1413],\n",
       "        [-0.3767, -0.4491, -0.1317, -0.3240, -0.4961, -0.2636, -0.3426],\n",
       "        [ 0.1393, -0.3582, -0.0219, -0.0592, -0.1004, -0.5549, -0.3266],\n",
       "        [ 0.4378, -0.0328, -0.4516, -0.3266,  0.4856,  0.3628, -0.0195],\n",
       "        [-0.3437,  0.2951,  0.6149,  0.4478,  0.4687,  0.0273, -0.4245],\n",
       "        [ 0.1250, -0.0474, -0.3595,  0.6054,  0.3584,  0.4017, -0.0841],\n",
       "        [-0.0154,  0.3555,  0.1391,  0.5880,  0.0148,  0.4679, -0.5166]])"
      ]
     },
     "execution_count": 64,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "adj_matrix = create_adjacency_matrix_xavier(y_train.numpy())\n",
    "adj_matrix = torch.tensor(adj_matrix)\n",
    "adj_matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "UVfju6CZe8GH"
   },
   "outputs": [],
   "source": [
    "model = MAGNET(300, 250, adj_matrix, glove_embedding_matrix,heads=8)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "coBdedeNPZKY",
    "outputId": "64b2c32d-02a1-4f7f-cbab-4b5ff1f52ab6"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch:1 loss:33.04614 hamming_loss:0.13466 micro_f1score:0.67637\n",
      "epoch:2 loss:25.92534 hamming_loss:0.09942 micro_f1score:0.76756\n",
      "epoch:3 loss:24.17178 hamming_loss:0.09243 micro_f1score:0.78498\n",
      "epoch:4 loss:23.23519 hamming_loss:0.08903 micro_f1score:0.79357\n",
      "epoch:5 loss:22.04953 hamming_loss:0.08525 micro_f1score:0.80332\n",
      "epoch:6 loss:21.23165 hamming_loss:0.08241 micro_f1score:0.81045\n",
      "epoch:7 loss:20.69443 hamming_loss:0.08056 micro_f1score:0.81434\n",
      "epoch:8 loss:19.82300 hamming_loss:0.07780 micro_f1score:0.82185\n",
      "epoch:9 loss:19.78283 hamming_loss:0.07704 micro_f1score:0.82328\n",
      "epoch:10 loss:19.27693 hamming_loss:0.07553 micro_f1score:0.82715\n",
      "epoch:11 loss:18.45850 hamming_loss:0.07228 micro_f1score:0.83496\n",
      "epoch:12 loss:18.12489 hamming_loss:0.07101 micro_f1score:0.83771\n",
      "epoch:13 loss:17.47616 hamming_loss:0.06777 micro_f1score:0.84581\n",
      "epoch:14 loss:17.28810 hamming_loss:0.06711 micro_f1score:0.84722\n",
      "epoch:15 loss:16.68447 hamming_loss:0.06477 micro_f1score:0.85265\n",
      "epoch:16 loss:17.29733 hamming_loss:0.06752 micro_f1score:0.84640\n",
      "epoch:17 loss:17.88668 hamming_loss:0.06965 micro_f1score:0.84125\n",
      "epoch:18 loss:15.96795 hamming_loss:0.06234 micro_f1score:0.85870\n",
      "epoch:19 loss:15.19938 hamming_loss:0.05969 micro_f1score:0.86478\n",
      "epoch:20 loss:14.36214 hamming_loss:0.05642 micro_f1score:0.87261\n",
      "epoch:21 loss:14.21532 hamming_loss:0.05597 micro_f1score:0.87350\n",
      "epoch:22 loss:13.52723 hamming_loss:0.05307 micro_f1score:0.88038\n",
      "epoch:23 loss:13.48493 hamming_loss:0.05283 micro_f1score:0.88092\n",
      "epoch:24 loss:13.49624 hamming_loss:0.05342 micro_f1score:0.87974\n",
      "epoch:25 loss:13.28949 hamming_loss:0.05271 micro_f1score:0.88143\n",
      "epoch:26 loss:12.71347 hamming_loss:0.04958 micro_f1score:0.88849\n",
      "epoch:27 loss:12.25193 hamming_loss:0.04755 micro_f1score:0.89340\n",
      "epoch:28 loss:11.82840 hamming_loss:0.04603 micro_f1score:0.89691\n",
      "epoch:29 loss:11.44363 hamming_loss:0.04445 micro_f1score:0.90030\n",
      "epoch:30 loss:11.19343 hamming_loss:0.04350 micro_f1score:0.90247\n",
      "epoch:31 loss:10.59212 hamming_loss:0.04143 micro_f1score:0.90742\n",
      "epoch:32 loss:10.66802 hamming_loss:0.04119 micro_f1score:0.90779\n",
      "epoch:33 loss:10.37156 hamming_loss:0.04064 micro_f1score:0.90916\n",
      "epoch:34 loss:10.16064 hamming_loss:0.03939 micro_f1score:0.91206\n",
      "epoch:35 loss:9.72657 hamming_loss:0.03768 micro_f1score:0.91584\n",
      "epoch:36 loss:9.41752 hamming_loss:0.03752 micro_f1score:0.91618\n",
      "epoch:37 loss:8.88864 hamming_loss:0.03494 micro_f1score:0.92203\n",
      "epoch:38 loss:8.55753 hamming_loss:0.03331 micro_f1score:0.92581\n",
      "epoch:39 loss:8.66529 hamming_loss:0.03383 micro_f1score:0.92454\n",
      "epoch:40 loss:8.34255 hamming_loss:0.03270 micro_f1score:0.92711\n",
      "epoch:41 loss:8.02438 hamming_loss:0.03145 micro_f1score:0.92997\n",
      "epoch:42 loss:8.10450 hamming_loss:0.03139 micro_f1score:0.93007\n",
      "epoch:43 loss:8.25567 hamming_loss:0.03204 micro_f1score:0.92861\n",
      "epoch:44 loss:7.87600 hamming_loss:0.03068 micro_f1score:0.93163\n",
      "epoch:45 loss:7.80026 hamming_loss:0.03007 micro_f1score:0.93311\n",
      "epoch:46 loss:7.51957 hamming_loss:0.02898 micro_f1score:0.93549\n",
      "epoch:47 loss:7.40309 hamming_loss:0.02827 micro_f1score:0.93703\n",
      "epoch:48 loss:7.08078 hamming_loss:0.02738 micro_f1score:0.93905\n",
      "epoch:49 loss:17.37140 hamming_loss:0.05475 micro_f1score:0.87899\n",
      "epoch:50 loss:12.06504 hamming_loss:0.04157 micro_f1score:0.90759\n"
     ]
    }
   ],
   "source": [
    "#Xavier Initialization\n",
    "train(model, X_train, glove_label_embeddings,y_train,total_epoch=50, save_path=save_path)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "RWyUpdSCmg4L",
    "outputId": "04746dda-4088-451b-ed32-cee7e66c6534"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch:1 loss:7.99771 hamming_loss:0.03299 micro_f1score:0.92643\n",
      "epoch:2 loss:7.75173 hamming_loss:0.03184 micro_f1score:0.92900\n",
      "epoch:3 loss:7.31711 hamming_loss:0.03035 micro_f1score:0.93244\n",
      "epoch:4 loss:7.19163 hamming_loss:0.02941 micro_f1score:0.93448\n",
      "epoch:5 loss:6.99891 hamming_loss:0.02856 micro_f1score:0.93654\n",
      "epoch:6 loss:6.90241 hamming_loss:0.02820 micro_f1score:0.93724\n",
      "epoch:7 loss:6.96195 hamming_loss:0.02832 micro_f1score:0.93705\n",
      "epoch:8 loss:6.91062 hamming_loss:0.02807 micro_f1score:0.93750\n",
      "epoch:9 loss:6.78423 hamming_loss:0.02742 micro_f1score:0.93915\n",
      "epoch:10 loss:6.62988 hamming_loss:0.02698 micro_f1score:0.93999\n",
      "epoch:11 loss:6.60798 hamming_loss:0.02633 micro_f1score:0.94138\n",
      "epoch:12 loss:6.48195 hamming_loss:0.02613 micro_f1score:0.94188\n",
      "epoch:13 loss:6.30166 hamming_loss:0.02558 micro_f1score:0.94312\n",
      "epoch:14 loss:6.03807 hamming_loss:0.02411 micro_f1score:0.94635\n",
      "epoch:15 loss:6.12789 hamming_loss:0.02502 micro_f1score:0.94435\n",
      "epoch:16 loss:6.05874 hamming_loss:0.02434 micro_f1score:0.94593\n",
      "epoch:17 loss:6.18370 hamming_loss:0.02446 micro_f1score:0.94559\n",
      "epoch:18 loss:6.02611 hamming_loss:0.02398 micro_f1score:0.94665\n",
      "epoch:19 loss:6.12102 hamming_loss:0.02444 micro_f1score:0.94566\n",
      "epoch:20 loss:5.84406 hamming_loss:0.02362 micro_f1score:0.94753\n",
      "epoch:21 loss:5.41376 hamming_loss:0.02208 micro_f1score:0.95099\n",
      "epoch:22 loss:5.44650 hamming_loss:0.02176 micro_f1score:0.95169\n",
      "epoch:23 loss:5.46210 hamming_loss:0.02184 micro_f1score:0.95144\n",
      "epoch:24 loss:5.29211 hamming_loss:0.02117 micro_f1score:0.95299\n",
      "epoch:25 loss:5.35080 hamming_loss:0.02138 micro_f1score:0.95248\n",
      "epoch:26 loss:5.07956 hamming_loss:0.02035 micro_f1score:0.95478\n",
      "epoch:27 loss:4.91959 hamming_loss:0.01928 micro_f1score:0.95718\n",
      "epoch:28 loss:5.65834 hamming_loss:0.02143 micro_f1score:0.95236\n",
      "epoch:29 loss:5.12640 hamming_loss:0.02002 micro_f1score:0.95552\n",
      "epoch:30 loss:5.52373 hamming_loss:0.02089 micro_f1score:0.95353\n"
     ]
    }
   ],
   "source": [
    "train(model, X_train, glove_label_embeddings,y_train,total_epoch=30, save_path=save_path)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "v-V01XOwe_9i",
    "outputId": "d6cc284b-aae6-4064-8034-a6f52da7f3d5"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 6490/6490 [00:51<00:00, 127.01it/s]\n"
     ]
    }
   ],
   "source": [
    "y_pred = []\n",
    "device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n",
    "for i in tqdm(range(X_test.shape[0])):\n",
    "  x = X_test[i].view(1,100)\n",
    "  out = model(x.to(device), glove_label_embeddings.to(device))\n",
    "  y_pred.append(torch.sigmoid(out.detach()).round().cpu())\n",
    "\n",
    "y_pred = torch.vstack(y_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "-HfiB4sLfV5R",
    "outputId": "dc03d664-4e35-4fcf-86c2-e5b6c3875f45"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'model_name': 'MAGNET_xavier', 'micro_avg_f1_score': 0.7839896450440584, 'hamming_loss': 0.09550957517059212, 'accuracy': 0.598151001540832}\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.86      0.84      0.85      3097\n",
      "           1       0.20      0.41      0.27       144\n",
      "           2       0.72      0.80      0.76      1546\n",
      "           3       0.38      0.35      0.36       118\n",
      "           4       0.60      0.69      0.64      1364\n",
      "           5       0.45      0.37      0.40       319\n",
      "           6       0.89      0.88      0.88      3278\n",
      "\n",
      "   micro avg       0.77      0.80      0.78      9866\n",
      "   macro avg       0.59      0.62      0.60      9866\n",
      "weighted avg       0.78      0.80      0.79      9866\n",
      " samples avg       0.80      0.79      0.77      9866\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classification.py:1344: UndefinedMetricWarning: Recall and F-score are ill-defined and being set to 0.0 in samples with no true labels. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n"
     ]
    }
   ],
   "source": [
    "arr={'model_name':'MAGNET_xavier','micro_avg_f1_score':f1_score(y_test, y_pred, average=\"micro\"),'hamming_loss':hamming_loss(y_test, y_pred),'accuracy':accuracy_score(y_test,y_pred)}\n",
    "scores_df = scores_df.append(arr,ignore_index=True)\n",
    "print(arr)\n",
    "print(classification_report(y_pred,y_test))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "yaEYI6gk274R"
   },
   "source": [
    "**Random initialization**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "SmiMU9P5gxJK"
   },
   "outputs": [],
   "source": [
    "#Random initialisation\n",
    "adj_matrix = torch.randn(7, 7)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "Q4GFUzgDhZrF"
   },
   "outputs": [],
   "source": [
    "model = MAGNET(300, 250, adj_matrix, glove_embedding_matrix,heads=8)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "bzcvaVwcPXAh",
    "outputId": "d122beb4-7a9c-47e1-ef3a-8590f4273dff"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch:1 loss:37.90182 hamming_loss:0.16219 micro_f1score:0.57687\n",
      "epoch:2 loss:26.94985 hamming_loss:0.10223 micro_f1score:0.75872\n",
      "epoch:3 loss:24.64162 hamming_loss:0.09434 micro_f1score:0.78041\n",
      "epoch:4 loss:26.02718 hamming_loss:0.09832 micro_f1score:0.77030\n",
      "epoch:5 loss:22.37064 hamming_loss:0.08602 micro_f1score:0.80094\n",
      "epoch:6 loss:21.84500 hamming_loss:0.08408 micro_f1score:0.80572\n",
      "epoch:7 loss:20.93736 hamming_loss:0.08116 micro_f1score:0.81278\n",
      "epoch:8 loss:20.16739 hamming_loss:0.07819 micro_f1score:0.82031\n",
      "epoch:9 loss:19.74241 hamming_loss:0.07667 micro_f1score:0.82429\n",
      "epoch:10 loss:19.50452 hamming_loss:0.07606 micro_f1score:0.82564\n",
      "epoch:11 loss:18.60078 hamming_loss:0.07295 micro_f1score:0.83292\n",
      "epoch:12 loss:18.16968 hamming_loss:0.07081 micro_f1score:0.83823\n",
      "epoch:13 loss:17.68715 hamming_loss:0.06938 micro_f1score:0.84189\n",
      "epoch:14 loss:17.06294 hamming_loss:0.06735 micro_f1score:0.84686\n",
      "epoch:15 loss:16.47658 hamming_loss:0.06425 micro_f1score:0.85395\n",
      "epoch:16 loss:16.00995 hamming_loss:0.06292 micro_f1score:0.85722\n",
      "epoch:17 loss:15.54502 hamming_loss:0.06142 micro_f1score:0.86096\n",
      "epoch:18 loss:15.15812 hamming_loss:0.05940 micro_f1score:0.86554\n",
      "epoch:19 loss:14.72547 hamming_loss:0.05751 micro_f1score:0.86995\n",
      "epoch:20 loss:14.62250 hamming_loss:0.05716 micro_f1score:0.87105\n",
      "epoch:21 loss:14.99857 hamming_loss:0.05848 micro_f1score:0.86767\n",
      "epoch:22 loss:14.67181 hamming_loss:0.05680 micro_f1score:0.87189\n",
      "epoch:23 loss:13.69551 hamming_loss:0.05409 micro_f1score:0.87829\n",
      "epoch:24 loss:12.69253 hamming_loss:0.04968 micro_f1score:0.88831\n",
      "epoch:25 loss:12.09417 hamming_loss:0.04740 micro_f1score:0.89355\n",
      "epoch:26 loss:24.68010 hamming_loss:0.08216 micro_f1score:0.81316\n",
      "epoch:27 loss:14.52971 hamming_loss:0.05486 micro_f1score:0.87662\n",
      "epoch:28 loss:12.14052 hamming_loss:0.04667 micro_f1score:0.89533\n",
      "epoch:29 loss:11.59295 hamming_loss:0.04457 micro_f1score:0.89994\n",
      "epoch:30 loss:11.09934 hamming_loss:0.04278 micro_f1score:0.90434\n",
      "epoch:31 loss:10.78875 hamming_loss:0.04186 micro_f1score:0.90645\n",
      "epoch:32 loss:10.34669 hamming_loss:0.03992 micro_f1score:0.91081\n",
      "epoch:33 loss:10.43032 hamming_loss:0.04014 micro_f1score:0.91028\n",
      "epoch:34 loss:10.40769 hamming_loss:0.03996 micro_f1score:0.91076\n",
      "epoch:35 loss:10.18736 hamming_loss:0.03970 micro_f1score:0.91145\n",
      "epoch:36 loss:10.34242 hamming_loss:0.03991 micro_f1score:0.91081\n",
      "epoch:37 loss:9.34928 hamming_loss:0.03573 micro_f1score:0.92021\n",
      "epoch:38 loss:8.67304 hamming_loss:0.03325 micro_f1score:0.92600\n",
      "epoch:39 loss:8.61073 hamming_loss:0.03294 micro_f1score:0.92666\n",
      "epoch:40 loss:8.18495 hamming_loss:0.03144 micro_f1score:0.92996\n",
      "epoch:41 loss:7.95696 hamming_loss:0.03087 micro_f1score:0.93133\n",
      "epoch:42 loss:7.67605 hamming_loss:0.02969 micro_f1score:0.93406\n",
      "epoch:43 loss:7.57434 hamming_loss:0.02917 micro_f1score:0.93513\n",
      "epoch:44 loss:7.18684 hamming_loss:0.02761 micro_f1score:0.93868\n",
      "epoch:45 loss:7.07899 hamming_loss:0.02671 micro_f1score:0.94067\n",
      "epoch:46 loss:7.10115 hamming_loss:0.02671 micro_f1score:0.94068\n",
      "epoch:47 loss:6.87766 hamming_loss:0.02606 micro_f1score:0.94215\n",
      "epoch:48 loss:6.54583 hamming_loss:0.02485 micro_f1score:0.94485\n",
      "epoch:49 loss:6.60289 hamming_loss:0.02542 micro_f1score:0.94360\n",
      "epoch:50 loss:6.40245 hamming_loss:0.02425 micro_f1score:0.94620\n"
     ]
    }
   ],
   "source": [
    "#Random initialisation\n",
    "train(model, X_train, glove_label_embeddings,y_train,total_epoch=50, save_path=save_path)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "N0pUCedTw6qu",
    "outputId": "adcffa6e-0e47-4810-ce9e-afe586fbbe68"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 6490/6490 [00:44<00:00, 144.80it/s]\n"
     ]
    }
   ],
   "source": [
    "y_pred = []\n",
    "device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n",
    "for i in tqdm(range(X_test.shape[0])):\n",
    "  x = X_test[i].view(1,100)\n",
    "  out = model(x.to(device), glove_label_embeddings.to(device))\n",
    "  y_pred.append(torch.sigmoid(out.detach()).round().cpu())\n",
    "\n",
    "y_pred = torch.vstack(y_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "ow2oGKCk0bRV",
    "outputId": "62a6b5b9-b908-4cbf-ed18-3751ed9affc8"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'model_name': 'MAGNET_random', 'micro_avg_f1_score': 0.7803653884762096, 'hamming_loss': 0.09632401496808277, 'accuracy': 0.588135593220339}\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.88      0.81      0.84      3290\n",
      "           1       0.32      0.42      0.36       217\n",
      "           2       0.69      0.82      0.75      1431\n",
      "           3       0.39      0.51      0.44        82\n",
      "           4       0.60      0.70      0.65      1344\n",
      "           5       0.47      0.47      0.47       260\n",
      "           6       0.84      0.89      0.86      3079\n",
      "\n",
      "   micro avg       0.76      0.80      0.78      9703\n",
      "   macro avg       0.60      0.66      0.63      9703\n",
      "weighted avg       0.77      0.80      0.78      9703\n",
      " samples avg       0.78      0.79      0.76      9703\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classification.py:1344: UndefinedMetricWarning: Recall and F-score are ill-defined and being set to 0.0 in samples with no true labels. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n"
     ]
    }
   ],
   "source": [
    "arr={'model_name':'MAGNET_random','micro_avg_f1_score':f1_score(y_test, y_pred, average=\"micro\"),'hamming_loss':hamming_loss(y_test, y_pred),'accuracy':accuracy_score(y_test,y_pred)}\n",
    "scores_df = scores_df.append(arr,ignore_index=True)\n",
    "print(arr)\n",
    "print(classification_report(y_pred,y_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "q77_T0qlnpYC",
    "outputId": "eb3e59b8-3ccc-4269-a019-aea90faee907"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Parameter containing:\n",
       "tensor([[ 0.6937, -0.2006,  0.2486, -0.1822,  0.1479, -0.1086,  0.0293],\n",
       "        [ 0.8458,  0.9264,  0.8186,  0.2366,  0.7374, -0.1675,  0.0992],\n",
       "        [ 0.7063,  0.1879,  0.9135, -0.0044,  0.6248, -0.0269, -0.2106],\n",
       "        [ 0.6145, -0.0421,  0.3925,  1.1011,  0.3909, -0.0373, -0.0110],\n",
       "        [ 0.7057,  0.0351,  0.6755, -0.0243,  0.8600,  0.0699,  0.2344],\n",
       "        [ 0.6129, -0.2898,  0.3176,  0.0868,  0.5169,  1.1826,  0.0220],\n",
       "        [-0.0602, -0.0384, -0.0482,  0.0138, -0.0936, -0.0930,  1.1234]],\n",
       "       device='cuda:0', dtype=torch.float64, requires_grad=True)"
      ]
     },
     "execution_count": 110,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.adjacency"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "Ycz4FO2LOsrf"
   },
   "outputs": [],
   "source": [
    "new_h = model.adjacency @ glove_label_embeddings.to(device).double()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "9icqL-6M3C1u"
   },
   "source": [
    "### Interesting observation\n",
    "\n",
    "The attention mechanism focusses on the closeness between labels. To demonstrate this we use the following method - \n",
    "* We first reduced the dimension of original glove embeddings and the embeddings obtained after training through the model, of all the labels using PCA to 2.\n",
    "* Then we plotted original embeddings and trained embeddings on a graph.\n",
    "* We observed that in the first graph, labels were positioned as per the glove embedding.\n",
    "* But after training through the model, the labels which were similar to each other came closer\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "_4Ind0t3jPXO"
   },
   "outputs": [],
   "source": [
    "from sklearn.decomposition import PCA\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 435
    },
    "id": "rW9wejEcgZsa",
    "outputId": "5763d12b-0ec1-4d0a-fa36-86bc6052773f"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.decomposition import PCA\n",
    "pca = PCA(2)\n",
    "Xt = pca.fit_transform(new_h.cpu().detach().numpy())\n",
    "plot = plt.scatter(Xt[:,0], Xt[:,1])\n",
    "for i, label in enumerate(labels):\n",
    "    plt.annotate(label, (Xt[i][0], Xt[i][1]), textcoords=\"offset points\", xytext=(0, 5), ha='center')\n",
    "\n",
    "plt.xlim(-6, 20)\n",
    "plt.ylim(-10, 10)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 435
    },
    "id": "PHUDou4Ni05S",
    "outputId": "bdd93e77-6390-4f92-a668-510e5d6fba66"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.decomposition import PCA\n",
    "pca = PCA(2)\n",
    "Xt = pca.fit_transform(glove_label_embeddings)\n",
    "plot = plt.scatter(Xt[:,0], Xt[:,1])\n",
    "for i, label in enumerate(labels):\n",
    "    plt.annotate(label, (Xt[i][0], Xt[i][1]), textcoords=\"offset points\", xytext=(0, 5), ha='center')\n",
    "\n",
    "plt.xlim(-6, 20)\n",
    "plt.ylim(-10, 10)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 513
    },
    "id": "ig2_QITtCyTT",
    "outputId": "e4f5cf97-3b7e-409b-8a6f-d70a2c786e34"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(adj_matrix, cmap='hot', interpolation='nearest')\n",
    "plt.colorbar()\n",
    "plt.xticks(np.arange(len(labels)), labels, rotation=45)\n",
    "plt.title('Initialized Adjacency Matrix HeatMap')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 513
    },
    "id": "Rq4lE8wtBiz8",
    "outputId": "0ea78d63-731f-445e-c6cf-f8797c0bed38"
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAe4AAAHwCAYAAABgy4y9AAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjcuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/bCgiHAAAACXBIWXMAAA9hAAAPYQGoP6dpAABn0klEQVR4nO3dd1xV9f8H8NcFWSrDgaAynYiKA5PIkSZKzqzc5kDFEZqKlWnOTNFvZtrPvS333iMzV+UEZ+5EJRV3oCgg975/fxgnrlwM5F7uAV7Px+M8lHM/5573ne/7mUcjIgIiIiLKFSzMHQARERFlHhM3ERFRLsLETURElIswcRMREeUiTNxERES5CBM3ERFRLsLETURElIswcRMREeUiTNxERES5CBM3AQAaNGiABg0a5Mi5vLy80L17d+Xvffv2QaPRYN++fTlyfnp9Y8aMgUajMXcYRPkaE3caixcvhkajwfHjx80diupptVqUKlUKGo0GO3bsMHc4eU7qe1Gj0eDXX39Nd7uIwN3dHRqNBi1atHitc0yYMAEbN27MZqTZ06BBA1SpUsXgbdeuXYNGo8HkyZNNGkNGz0NOvAZEr4OJm17LL7/8gtu3b8PLywvLli3L1n3Vr18fz549Q/369Y0UXd5ha2uL5cuXp9u/f/9+/PXXX7CxsXnt+36dxD1ixAg8e/bstc+pRv/1PJjyNSB6HUzcKpSSkoLk5GRzh/FKS5cuRc2aNTF48GBs3LgRCQkJr31fFhYWsLW1hYUF344va9asGdasWYOUlBS9/cuXL4e/vz9cXV1zJI7U17dAgQKwtbXNkXOqhVpeA6JU/KZ8DTdv3kSPHj3g4uICGxsbVK5cGQsXLtQrk5ycjFGjRsHf3x+Ojo4oVKgQ6tWrh7179+qVS9scOHXqVJQtWxY2NjY4d+6c0p945coVdO/eHU5OTnB0dERISAiePn2aLq6lS5fC398fdnZ2KFq0KDp06ICYmJh05ebOnYuyZcvCzs4OtWvXxsGDB7P0+J89e4YNGzagQ4cOaNeuHZ49e4ZNmzalKyci+Prrr+Hm5oaCBQuiYcOG+OOPP9KVM9THffDgQbRt2xYeHh6wsbGBu7s7Bg8ebLC2d+HCBbRr1w7Ozs6ws7NDxYoV8eWXX+qVycxrlhrH6tWrMX78eLi5ucHW1haNGjXClStX0p33yJEjaNasGYoUKYJChQrBz88P06ZNAwAsWrQIGo0GJ06cSHfchAkTYGlpiZs3bxp+gtPo2LEjHjx4gN27dyv7kpOTsXbtWnTq1MngMZMnT8Zbb72FYsWKwc7ODv7+/li7dq1eGY1Gg4SEBCxZskRpDk4dd5D6vjt37hw6deqEIkWKoG7dunq3pUp9nC8/lxMmTIBGo8H27dv/8zFm1d9//41BgwbB3d0dNjY2KFeuHCZNmgSdTmfU5yGVqV6D1PP3798fy5YtQ8WKFWFrawt/f38cOHDgNZ8dyheEFIsWLRIAcuzYsQzLxMbGipubm7i7u8tXX30ls2bNklatWgkA+e6775Ry9+7dk5IlS0p4eLjMmjVL/ve//0nFihXFyspKTpw4oZSLjo4WAOLr6ytlypSRiRMnynfffSfXr1+X0aNHCwCpUaOGfPDBBzJz5kzp1auXAJDPP/9cL66vv/5aNBqNtG/fXmbOnCljx46V4sWLi5eXlzx69EgpN3/+fAEgb731lnz//fcyaNAgcXJykjJlysjbb7+dqedp5cqVotFo5MaNGyIi8s4770izZs3SlRsxYoQAkGbNmsn06dOlR48eUqpUKSlevLh069ZNKbd3714BIHv37lX2DRgwQJo1ayYTJkyQOXPmSM+ePcXS0lLatGmjd45Tp06Jg4ODFCtWTIYNGyZz5syRzz//XKpWrZrl1yw1jho1aoi/v7989913MmbMGClYsKDUrl1b77w//fSTWFtbi6enp4wePVpmzZoln3zyiQQFBYmISHx8vNjZ2cmQIUPSPS++vr7yzjvvvPI5TvtefOutt6RLly7KbRs3bhQLCwu5efOmeHp6SvPmzfWOdXNzk48//limT58uU6ZMkdq1awsA2bp1q1Lmxx9/FBsbG6lXr578+OOP8uOPP8rvv/8uIqK873x9feW9996TmTNnyowZM/RuS6tFixbi6OiovB9Onz4t1tbW0rNnz1c+RhGRt99+W3x8fOTevXvptqioKAEg33zzjVI+ISFB/Pz8pFixYjJ8+HCZPXu2dO3aVTQajQwcONCoz4OpXwMREQBSpUoVKV68uHz11VcyadIk8fT0FDs7Ozlz5sx/Pn+UPzFxp5GZxN2zZ08pWbKk3L9/X29/hw4dxNHRUZ4+fSoiIikpKZKUlKRX5tGjR+Li4iI9evRQ9qUmbgcHB7l7965e+dQvybTlRUTef/99KVasmPL3tWvXxNLSUsaPH69X7syZM1KgQAFlf3JyspQoUUKqV6+uF9vcuXMFQKYTd4sWLaROnTp6xxcoUEAv/rt374q1tbU0b95cdDqdsn/48OEC4D8Td+rzmFZERIRoNBq5fv26sq9+/fpib2+vt09E9M6Z2dcsNY5KlSrpPT/Tpk0TAMoXaUpKinh7e4unp6fej6KXz9uxY0cpVaqUaLVaZV9qMlq0aFG6x5dW2vfi9OnTxd7eXomzbdu20rBhQxERg0nj5ecuOTlZqlSpku7HQqFChfReh1Sp77uOHTtmeFtat2/flqJFi0rjxo0lKSlJatSoIR4eHhIXF/fKxyjyInEDeOWWNnGPGzdOChUqJJcuXdK7ny+++EIsLS2VHw/GeB5y4jVIfYzHjx9X9l2/fl1sbW3l/fffN/icEbGpPAtEBOvWrUPLli0hIrh//76yBQcHIy4uDlFRUQAAS0tLWFtbAwB0Oh0ePnyIlJQU1KpVSymT1ocffghnZ2eD5+3bt6/e3/Xq1cODBw8QHx8PAFi/fj10Oh3atWunF5OrqyvKly+vNM8fP34cd+/eRd++fZXYAKB79+5wdHTM1HPw4MED7Nq1Cx07dtSLPbWJOdXPP/+M5ORkDBgwQK9pddCgQZk6j52dnfL/hIQE3L9/H2+99RZERGl+vnfvHg4cOIAePXrAw8ND7/jUc2blNUsVEhKi9/zUq1cPAHD16lUAwIkTJxAdHY1BgwbBycnJ4HkBoGvXrrh165Ze98iyZctgZ2eHDz/8MFPPAwClO2Lr1q14/Pgxtm7dmmETLaD/3D169AhxcXGoV6+ewffdq7z8vsuIq6srZsyYgd27d6NevXo4efIkFi5cCAcHh0wd7+Xlhd27d6fbli5dmq7smjVrUK9ePRQpUkTvtQwKCoJWq9VrYjbW8wCY9jUIDAyEv7+/8reHhwfee+897Nq1C1qtNsuxUt5XwNwB5Cb37t3D33//jblz52Lu3LkGy9y9e1f5/5IlS/Dtt9/iwoULeP78ubLf29s73XGG9qV6OSkVKVIEwIsvBAcHB1y+fBkigvLlyxs83srKCgBw/fp1AEhXzsrKCmXKlMnw/GmtWrUKz58/R40aNfT6fQMCArBs2TKEhYW98lzOzs5K/K9y48YNjBo1Cps3b8ajR4/0bouLiwPwbyLNaDoRkPXXDHj18w0Af/7553+eFwAaN26MkiVLYtmyZWjUqBF0Oh1WrFiB9957D/b29q88Ni1nZ2cEBQVh+fLlePr0KbRaLdq0aZNh+a1bt+Lrr7/GyZMnkZSUpOzP6vzrV70nX9ahQwcsXboU27ZtQ+/evdGoUaNMH1uoUCEEBQWl23/t2rV0+y5fvozTp09n+CM37WtprOcBMO1rYOhzW6FCBTx9+hT37t3j4DdKh4k7C1IHv3z00Ufo1q2bwTJ+fn4AXgwU6969O1q3bo3PPvsMJUqUgKWlJSIiIpQv/rTS/kJ/maWlpcH9IqLElTqf2lDZwoULv/qBZUHq1K86deoYvP3q1auZ/hGQEa1Wi8aNG+Phw4cYOnQofHx8UKhQIdy8eRPdu3dPNwjpVbLymqX6r+c7sywtLdGpUyfMmzcPM2fOxG+//YZbt27ho48+ytL9AECnTp0QGhqK2NhYNG3aNF1NP9XBgwfRqlUr1K9fHzNnzkTJkiVhZWWFRYsWGZzS9Cqvek++7MGDB8r6B+fOnYNOpzPJLAGdTofGjRvj888/N3h7hQoVABj3eUhljteAyBAm7ixwdnaGvb09tFqtwRpCWmvXrkWZMmWwfv16vV/Zo0ePNnpcZcuWhYjA29tb+eIyxNPTE8CLWss777yj7H/+/Dmio6NRrVq1V54nOjoav//+O/r374+3335b7zadTocuXbpg+fLlGDFihN650ibye/fupatBv+zMmTO4dOkSlixZgq5duyr7047qBaDc79mzZzO8r6y8ZplVtmxZ5bz/dZ9du3bFt99+iy1btmDHjh1wdnZGcHBwls/5/vvvo0+fPjh8+DBWrVqVYbl169bB1tYWu3bt0ptfvGjRonRljbkCWlhYGB4/foyIiAgMGzYMU6dORXh4uNHuP1XZsmXx5MmT/3zeTfE8mOI1AF58Rl526dIlFCxYMMOWBcrf2MedBZaWlvjwww+xbt06g8ni3r17emUB/VrakSNHcOjQIaPH9cEHH8DS0hJjx45NVysUETx48AAAUKtWLTg7O2P27Nl688QXL16Mv//++z/Pk1rb/vzzz9GmTRu9rV27dnj77beVMkFBQbCyssL//d//6cU0derU/zyPoedORJSpVqmcnZ1Rv359LFy4EDdu3Ej3uFPvK7OvWWbVrFkT3t7emDp1arrn7eXn38/PD35+fpg/fz7WrVuHDh06oECBrP9eLly4MGbNmoUxY8agZcuWGZaztLSERqPR6xu9du2awQVGChUqlKnX/b+sXbsWq1atwsSJE/HFF1+gQ4cOGDFiBC5dupTt+35Zu3btcOjQIezatSvdbX///bcy19oUz4MpXgMAOHTokF7fd0xMDDZt2oQmTZpk2PpD+Rtr3AYsXLgQO3fuTLd/4MCBmDhxIvbu3YuAgACEhobC19cXDx8+RFRUFH7++Wc8fPgQANCiRQusX78e77//Ppo3b47o6GjMnj0bvr6+ePLkiVHjLVu2LL7++msMGzYM165dQ+vWrWFvb4/o6Ghs2LABvXv3xqeffgorKyt8/fXX6NOnD9555x20b98e0dHRWLRoUaaat5ctW4bq1avD3d3d4O2tWrXCgAEDEBUVhZo1a+LTTz9FREQEWrRogWbNmuHEiRPYsWMHihcv/srz+Pj4oGzZsvj0009x8+ZNODg4YN26dQZr6t9//z3q1q2LmjVronfv3vD29sa1a9ewbds2nDx5EgAy/ZplloWFBWbNmoWWLVuievXqCAkJQcmSJXHhwgX88ccf6ZJK165d8emnnwLAazWTp8qoqT+t5s2bY8qUKXj33XfRqVMn3L17FzNmzEC5cuVw+vRpvbL+/v74+eefMWXKFJQqVQre3t4ICAjIUkx3795Fv3790LBhQ/Tv3x8AMH36dOzduxfdu3fHr7/+atQm888++wybN29GixYt0L17d/j7+yMhIQFnzpzB2rVrce3aNRQvXtxkz4OxXwPgxViJ4OBgfPLJJ7CxscHMmTMBAGPHjn2NZ4jyhRwfx65iqdM/MtpiYmJEROTOnTsSFhYm7u7uYmVlJa6urtKoUSOZO3eucl86nU4mTJggnp6eYmNjIzVq1JCtW7dKt27dxNPTUymXOh0s7ZSXVKlTb+7du2cwzujoaL3969atk7p160qhQoWkUKFC4uPjI2FhYXLx4kW9cjNnzhRvb2+xsbGRWrVqyYEDB+Ttt99+5XSwyMhIASAjR47MsMy1a9cEgAwePFhERLRarYwdO1ZKliwpdnZ20qBBAzl79qx4enr+53Swc+fOSVBQkBQuXFiKFy8uoaGhcurUKYNTqc6ePSvvv/++ODk5ia2trVSsWDFdnJl5zVLjWLNmjd6xqa/Ry+f99ddfpXHjxmJvby+FChUSPz8/+b//+790z8vt27fF0tJSKlSokOFz97LMTE0UMTwVacGCBVK+fHmxsbERHx8fWbRokcFpXBcuXJD69euLnZ2d3hS9jN53aW9L9cEHH4i9vb1cu3ZNr9ymTZsEgEyaNOmV8b/99ttSuXJlg7dl9Nl4/PixDBs2TMqVKyfW1tZSvHhxeeutt2Ty5MmSnJxstOchJ14DABIWFiZLly5VyteoUUPvs0D0Mo1IFkfcEBnZnj17EBQUhIMHDyordOUl9+/fR8mSJTFq1CiMHDnS3OGQimg0GoSFhWH69OnmDoVyEfZxk9ndvn0bAP6zCT23Wrx4MbRaLbp06WLuUIgoD2AfN5lNQkICli1bhmnTpsHNze2VI+Jzo19++QXnzp3D+PHj0bp1a3h5eZk7JCLKA1jjJrO5d+8eBgwYADs7O6xbty7PXR3sq6++Qnh4OKpXr47/+7//M3c4RJRHsI+biIgoF8lbVRwiIqI8jombiIgoF8nxwWk6nQ63bt2Cvb29UZdcJCIi0xMRPH78GKVKlTLpuJTExES9FR5fl7W1NWxtbY0QkXrkeOK+detWhitvERFR7hATEwM3NzeT3HdiYiK8vb0RGxub7ftydXVFdHR0nkreOZ64Uy9nGBNTAg4Oeaelvo5j9t9gavNb5i6nnGu8E2/uCIxvibkDMIGvzR2ACdQ0dwBGlAhgFJClS9NmVXJyMmJjYxETE53p67obEh8fD3d3byQnJzNxZ0dq87iDg0WeStx58VIADnmsJyMvLlpguq9O87E2dwAmkPkLpOYeOdHV6eDgkK3EnVflxe8yIiLKE1L+2bJzfN7DxE1ERCrFxG1I3mmrJiIiygeYuImISKVSjLBlzYEDB9CyZUuUKlUKGo0GGzdufGX59evXo3HjxnB2doaDgwMCAwOxa9euLJ83K5i4iYhIpbTIXtLWZvmMCQkJqFatGmbMmJGp8gcOHEDjxo2xfft2REZGomHDhmjZsiVOnDiR5XNnFvu4iYiI/tG0aVM0bdo00+WnTp2q9/eECROwadMmbNmyBTVq1DBydC8wcRMRkUoZZ3BafLz+Ig42NjawsbHJxv1mTKfT4fHjxyhatKhJ7h9gUzkREamWcfq43d3d4ejoqGwREREmi3jy5Ml48uQJ2rVrZ7JzsMZNRER5WkxMjN5CLqaqbS9fvhxjx47Fpk2bUKJECZOcA2DiJiIi1TJOU3lOrMC2cuVK9OrVC2vWrEFQUJBJz8XETUREKqXF64wM1z/e9FasWIEePXpg5cqVaN68ucnPx8RNREQqlTodLDvHZ82TJ09w5coV5e/o6GicPHkSRYsWhYeHB4YNG4abN2/ihx9+APCiebxbt26YNm0aAgIClCua2dnZwdHRMRuxZ4yD04iIiP5x/Phx1KhRQ5nKFR4ejho1amDUqFEAgNu3b+PGjRtK+blz5yIlJQVhYWEoWbKksg0cONBkMbLGTUREKpXza5U3aNAAIpLh7YsXL9b7e9++fVk+R3YxcRMRkUrxIiOGsKmciIgoF2GNm4iIVIo1bkOYuImISKVyflR5bsCmciIiolzktRL3jBkz4OXlBVtbWwQEBODo0aPGjouIiPK9nL8ed26Q5cS9atUqhIeHY/To0YiKikK1atUQHByMu3fvmiI+IiLKt5i4Dcly4p4yZQpCQ0MREhICX19fzJ49GwULFsTChQtNER8RERGlkaXBacnJyYiMjMSwYcOUfRYWFggKCsKhQ4cMHpOUlISkpCTl75evi0pERGQYR5UbkqUa9/3796HVauHi4qK338XFRVmf9WURERF610F1d3d//WiJiCgfYVO5ISYfVT5s2DDExcUpW0xMjKlPSUREeULqdLDX3fLmdLAsNZUXL14clpaWuHPnjt7+O3fuwNXV1eAxNjY2JrtoORERUX6TpRq3tbU1/P39sWfPHmWfTqfDnj17EBgYaPTgiIgoP2NTuSFZXjktPDwc3bp1Q61atVC7dm1MnToVCQkJCAkJMUV8RESUb3FwmiFZTtzt27fHvXv3MGrUKMTGxqJ69erYuXNnugFrREREZHyvtVZ5//790b9/f2PHQkRElAZr3IbwIiNERKRSTNyG8CIjREREuQhr3EREpFK8rKchTNxERKRSWmQv+ebNxM2mciIiolyENW4iIlIpDk4zhImbiIhUionbECZuIiJSKQ5OM4R93ERERLkIa9xERKRSbCo3hImbiIhUionbEDaVExER5SKscRMRkUqxxm0IEzcREakUE7chbConIiLKRVjjJiIileI8bkOYuImISKVSAFhm8/i8h03lREREuQhr3EREpFKscRvCxE1ERCrFxG2IGRN3MABr853eyE6Kv7lDML4/+po7AqOaUcXcERifx1pzR2B87m3MHYHx9bcxdwTGEy/A58k5dTYOTjOEfdxERES5CBM3ERGpVIoRtqw5cOAAWrZsiVKlSkGj0WDjxo3/ecy+fftQs2ZN2NjYoFy5cli8eHGWz5sVTNxERKRSOZ+4ExISUK1aNcyYMSNT5aOjo9G8eXM0bNgQJ0+exKBBg9CrVy/s2rUry+fOLA5OIyIi+kfTpk3RtGnTTJefPXs2vL298e233wIAKlWqhF9//RXfffcdgoODTRIja9xERKRSxqlxx8fH621JSUlGi/DQoUMICgrS2xccHIxDhw4Z7RwvY+ImIiKV0hphA9zd3eHo6KhsERERRoswNjYWLi4uevtcXFwQHx+PZ8+eGe08abGpnIiI8rSYmBg4ODgof9vY5O75eUzcRESkUsaZx+3g4KCXuI3J1dUVd+7c0dt3584dODg4wM7OziTnZOImIiKVSgGgyebxphUYGIjt27fr7du9ezcCAwNNdk72cRMREf3jyZMnOHnyJE6ePAngxXSvkydP4saNGwCAYcOGoWvXrkr5vn374urVq/j8889x4cIFzJw5E6tXr8bgwYNNFiNr3EREpFI5X+M+fvw4GjZsqPwdHh4OAOjWrRsWL16M27dvK0kcALy9vbFt2zYMHjwY06ZNg5ubG+bPn2+yqWAAEzcREalWzifuBg0aQEQyvN3QqmgNGjTAiRMnsnyu18XETUREKqX+Pm5zYB83ERFRLsIaNxERqZQW2atx583LejJxExGRSmW3qZtN5URERGRmrHETEZFKscZtCBM3ERGpFBO3IWwqJyIiykVY4yYiIpXK7qjwvDmqPMs17gMHDqBly5YoVaoUNBoNNm7caIKwiIiIUoyw5T1ZTtwJCQmoVq0aZsyYYYp4iIiI/sHEbUiWm8qbNm2Kpk2bmiIWIiIi+g8m7+NOSkpCUlKS8nd8fLypT0lERHlCCoCML/jx39jH/VoiIiLg6OiobO7u7qY+JRER5QlsKjfE5Il72LBhiIuLU7aYmBhTn5KIiCjPMnlTuY2NDWxsbEx9GiIiynO0yF5Tuc5YgagK53ETEZFKMXEbkuXE/eTJE1y5ckX5Ozo6GidPnkTRokXh4eFh1OCIiIhIX5YT9/Hjx9GwYUPl7/DwcABAt27dsHjxYqMFRkRE+V0KsjcUizVuAECDBg0gkp2mCyIiosxg4jaEFxkhIiLKRTg4jYiIVIo1bkOYuImISKW0yF7yzZvdukzcRESkUikANNk4Pm8mbvZxExER5SKscRMRkUqxxm0IEzcREakUE7chbConIiLKRVjjJiIidRJd9irNebPCzcRNREQqpUP2ZoPlzWncbConIiLKTVjjJiIiddL+s2Xn+DyIiZuIiNSJidsgNpUTERHlIqxxExGROnFwmkFM3EREpE5sKjeIiZuIiNSJNW6D2MdNRESUizBxExGROunwb3P562yvWeOeMWMGvLy8YGtri4CAABw9evSV5adOnYqKFSvCzs4O7u7uGDx4MBITE1/v5JnAxE1EROqUnaT9mv3jq1atQnh4OEaPHo2oqChUq1YNwcHBuHv3rsHyy5cvxxdffIHRo0fj/PnzWLBgAVatWoXhw4dn/eSZZMY+7vPmPb2RxWrmmTsEo3OVouYOwai24qG5QzA6/0PmjsD4Hps7ABNokGTuCIwnxdwBmNiUKVMQGhqKkJAQAMDs2bOxbds2LFy4EF988UW68r///jvq1KmDTp06AQC8vLzQsWNHHDlyxGQxssZNRETqpDPCBiA+Pl5vS0oy/EsqOTkZkZGRCAoKUvZZWFggKCgIhw4Z/pX81ltvITIyUmlOv3r1KrZv345mzZpl77G/Qt6p8hIRUd5ipOlg7u7uertHjx6NMWPGpCt+//59aLVauLi46O13cXHBhQsXDJ6iU6dOuH//PurWrQsRQUpKCvr27ZtXm8qJiIhMLyYmBg4ODsrfNjY2Rrvvffv2YcKECZg5cyYCAgJw5coVDBw4EOPGjcPIkSONdp60mLiJiEidjFTjdnBw0EvcGSlevDgsLS1x584dvf137tyBq6urwWNGjhyJLl26oFevXgCAqlWrIiEhAb1798aXX34JCwvj90izj5uIiNTJSH3cmWVtbQ1/f3/s2bPn3xB0OuzZsweBgYEGj3n69Gm65GxpaQkAEJGsBZBJrHETERH9Izw8HN26dUOtWrVQu3ZtTJ06FQkJCcoo865du6J06dKIiIgAALRs2RJTpkxBjRo1lKbykSNHomXLlkoCNzYmbiIiUiczrFXevn173Lt3D6NGjUJsbCyqV6+OnTt3KgPWbty4oVfDHjFiBDQaDUaMGIGbN2/C2dkZLVu2xPjx47MR+KtpxFR1+QzEx8fD0dERcXG14eCQd343xGp+N3cIRpfX5nGP1eS9edyjh5g7AuMb+K25IzC+U+YOwIhSAPwGIC4uLlP9xq9DyRNRgEPhbNzPE8CxpmljNYe8kzmJiChv4dXBDOLgNCIiolyENW4iIlIn1rgNYuImIiJ14vW4DWJTORERUS7CGjcREakTm8oNYuImIiJ1YuI2iE3lREREuQhr3EREpE4cnGYQEzcREamTDtlr7s6jiZtN5URERLkIa9xERKRObCo3iImbiIjUiaPKDcpSU3lERATeeOMN2Nvbo0SJEmjdujUuXrxoqtiIiCg/0xphy4OylLj379+PsLAwHD58GLt378bz58/RpEkTJCQkmCo+IiIiSiNLTeU7d+7U+3vx4sUoUaIEIiMjUb9+faMGRkRE+Rz7uA3KVh93XFwcAKBo0aIZlklKSkJSUpLyd3x8fHZOSURE+QX7uA167elgOp0OgwYNQp06dVClSpUMy0VERMDR0VHZ3N3dX/eURERE+d5rJ+6wsDCcPXsWK1eufGW5YcOGIS4uTtliYmJe95RERJSfcHCaQa/VVN6/f39s3boVBw4cgJub2yvL2tjYwMbG5rWCIyKifEyQvX5qMVYg6pKlxC0iGDBgADZs2IB9+/bB29vbVHERERGRAVlK3GFhYVi+fDk2bdoEe3t7xMbGAgAcHR1hZ2dnkgCJiCif4uA0g7LUxz1r1izExcWhQYMGKFmypLKtWrXKVPEREVF+pTPClgdluamciIiIzIdrlRMRkTqxqdwgJm4iIlInJm6DmLiJiEiduOSpQa+9AAsRERHlPNa4iYhIndhUbhATNxERqZMO2Uu+bConIiIic2ONm4iI1ImD0wxi4iYiInViH7dBbConIiLKRVjjJiIidWJTuUFM3EREpE5sKjeITeVERES5CGvcRESkTqxxG8QaNxERqZOZrsc9Y8YMeHl5wdbWFgEBATh69Ogry//9998ICwtDyZIlYWNjgwoVKmD79u2vd/JMYI2biIjUyQwrp61atQrh4eGYPXs2AgICMHXqVAQHB+PixYsoUaJEuvLJyclo3LgxSpQogbVr16J06dK4fv06nJycshH4qzFxExER/WPKlCkIDQ1FSEgIAGD27NnYtm0bFi5ciC+++CJd+YULF+Lhw4f4/fffYWVlBQDw8vIyaYxsKiciInUyUlN5fHy83paUlGTwdMnJyYiMjERQUJCyz8LCAkFBQTh06JDBYzZv3ozAwECEhYXBxcUFVapUwYQJE6DVmq6DnYmbiIjUSWuEDYC7uzscHR2VLSIiwuDp7t+/D61WCxcXF739Li4uiI2NNXjM1atXsXbtWmi1Wmzfvh0jR47Et99+i6+//jpbD/1VzNdUHv/qzv7cZr+5AzCB9gMfmjsEoxoti80dgvH1627uCIwuzNwBmMBUcwdgRMkAfjN3EFkUExMDBwcH5W8bGxuj3bdOp0OJEiUwd+5cWFpawt/fHzdv3sQ333yD0aNHG+08abGPm4iI1MlI08EcHBz0EndGihcvDktLS9y5c0dv/507d+Dq6mrwmJIlS8LKygqWlpbKvkqVKiE2NhbJycmwtrZ+/fgzwKZyIiJSpxyeDmZtbQ1/f3/s2bPn3xB0OuzZsweBgYEGj6lTpw6uXLkCne7fk126dAklS5Y0SdIGmLiJiIgU4eHhmDdvHpYsWYLz58+jX79+SEhIUEaZd+3aFcOGDVPK9+vXDw8fPsTAgQNx6dIlbNu2DRMmTEBYmOk6fdhUTkRE6mSGldPat2+Pe/fuYdSoUYiNjUX16tWxc+dOZcDajRs3YGHxb53X3d0du3btwuDBg+Hn54fSpUtj4MCBGDp0aDYCfzUmbiIiUiczLXnav39/9O/f3+Bt+/btS7cvMDAQhw8ffr2TvQY2lRMREeUirHETEZE6CbJ3TW0xViDqwsRNRETqxKuDGcTETURE6pSNK3wpx+dB7OMmIiLKRVjjJiIidWJTuUFM3EREpE5M3AaxqZyIiCgXYY2biIjUiYPTDGLiJiIidWJTuUFsKiciIspFWOMmIiJ10iF7tWY2lRMREeUg9nEbxKZyIiKiXIQ1biIiUicOTjOIiZuIiNSJTeUGMXETEZE6scZtUJb6uGfNmgU/Pz84ODjAwcEBgYGB2LFjh6liIyIiopdkqcbt5uaGiRMnonz58hARLFmyBO+99x5OnDiBypUrmypGIiLKj1jjNihLibtly5Z6f48fPx6zZs3C4cOHmbiJiMi42Mdt0Gv3cWu1WqxZswYJCQkIDAzMsFxSUhKSkpKUv+Pj41/3lERERPlelhP3mTNnEBgYiMTERBQuXBgbNmyAr69vhuUjIiIwduzYbAVJRET5EFdOMyjLC7BUrFgRJ0+exJEjR9CvXz9069YN586dy7D8sGHDEBcXp2wxMTHZCpiIiPIJrRG2PCjLNW5ra2uUK1cOAODv749jx45h2rRpmDNnjsHyNjY2sLGxyV6UREREBMAI87h1Op1eHzYREZFRcHCaQVlK3MOGDUPTpk3h4eGBx48fY/ny5di3bx927dplqviIiCi/0iJ7V9RgUzlw9+5ddO3aFbdv34ajoyP8/Pywa9cuNG7c2FTxERERURpZStwLFiwwVRxERET62FRuENcqJyIidWJTuUFM3EREpE5M3AZl5ykhIiKiHMYaNxERqZMge/3UYqxA1IWJm4iI1EkLQJPN4/MgNpUTERHlIqxxExGROrHGbRATNxERqRPncRvEpnIiIqJchDVuIiJSJzaVG8QaNxERqZPOCNtrmDFjBry8vGBra4uAgAAcPXo0U8etXLkSGo0GrVu3fr0TZxITNxER0T9WrVqF8PBwjB49GlFRUahWrRqCg4Nx9+7dVx537do1fPrpp6hXr57JY2TiJiIiddIaYcuiKVOmIDQ0FCEhIfD19cXs2bNRsGBBLFy4MOMwtVp07twZY8eORZkyZbJ+0ixi4iYiInXSIXtJ+5+m8vj4eL0tKSnJ4OmSk5MRGRmJoKAgZZ+FhQWCgoJw6NChDMP86quvUKJECfTs2TNbDzezmLiJiEidjNTH7e7uDkdHR2WLiIgweLr79+9Dq9XCxcVFb7+LiwtiY2MNHvPrr79iwYIFmDdvXrYealZwVDkREeVpMTExcHBwUP62sbExyv0+fvwYXbp0wbx581C8eHGj3GdmMHETEZE6ZXc61z/HOzg46CXujBQvXhyWlpa4c+eO3v47d+7A1dU1Xfk///wT165dQ8uWLZV9Ot2Lan6BAgVw8eJFlC1bNhsPwDA2lRMRkTrl8OA0a2tr+Pv7Y8+ePco+nU6HPXv2IDAwMF15Hx8fnDlzBidPnlS2Vq1aoWHDhjh58iTc3d2z+ogzhTVuIiKif4SHh6Nbt26oVasWateujalTpyIhIQEhISEAgK5du6J06dKIiIiAra0tqlSpone8k5MTAKTbb0zmS9wOmhdbHmGRFxfFbWjuAIzrU013c4dgdJOPmDsC49s729wRGN8QcwdgRI8BLMipk+mQvZXTXuNruX379rh37x5GjRqF2NhYVK9eHTt37lQGrN24cQMWFuZtrGaNm4iI1MlIfdxZ1b9/f/Tv39/gbfv27XvlsYsXL369k2YB+7iJiIhyEda4iYhInczQVJ4bMHETEZE6ZTfx5tHEzaZyIiKiXIQ1biIiUictAMnG8Xm0xs3ETURE6sSmcoOYuImISJ1Y4zaIfdxERES5CGvcRESkTqxxG8TETURE6sQ+boPYVE5ERJSLsMZNRETqpEP2msqzc6yKMXETEZE6ZXfJ0zyauNlUTkRElIuwxk1EROqkBWvcBjBxExGROjFxG8SmciIiolyENW4iIlInDk4ziImbiIjUiU3lBjFxExGROjFxG5StPu6JEydCo9Fg0KBBRgqHiIiIXuW1a9zHjh3DnDlz4OfnZ8x4iIiIXhDk2VpzdrxWjfvJkyfo3Lkz5s2bhyJFihg7JiIiImiNsOVFr5W4w8LC0Lx5cwQFBRk7HiIiInqFLDeVr1y5ElFRUTh27FimyiclJSEpKUn5Oz4+PqunJCKifCi7tWbWuAHExMRg4MCBWLZsGWxtbTN1TEREBBwdHZXN3d39tQIlIqL8RWeELS/KUuKOjIzE3bt3UbNmTRQoUAAFChTA/v378f3336NAgQLQatP/vhk2bBji4uKULSYmxmjBExER5TdZaipv1KgRzpw5o7cvJCQEPj4+GDp0KCwtLdMdY2NjAxsbm+xFSURE+Q6byg3LUuK2t7dHlSpV9PYVKlQIxYoVS7efiIgoO7Lb3M2mciIiIjK7bC95um/fPiOEQUREpI9N5YZxrXIiIlIlHbKXfPNqUzkTNxERqRL7uA1jHzcREVEuwho3ERGpEvu4DWPiJiIiVWLiNoxN5URERLkIa9xERKRKHJxmGGvcRESkSua6HveMGTPg5eUFW1tbBAQE4OjRoxmWnTdvHurVq4ciRYqgSJEiCAoKemV5Y2DiJiIi+seqVasQHh6O0aNHIyoqCtWqVUNwcDDu3r1rsPy+ffvQsWNH7N27F4cOHYK7uzuaNGmCmzdvmixGJm4iIlIlc1zWc8qUKQgNDUVISAh8fX0xe/ZsFCxYEAsXLjRYftmyZfj4449RvXp1+Pj4YP78+dDpdNizZ89rnD1zmLiJiEiVUldOe90tNXHHx8frbUlJSQbPl5ycjMjISAQFBSn7LCwsEBQUhEOHDmUq5qdPn+L58+coWrRoVh9upjFxExFRnubu7g5HR0dli4iIMFju/v370Gq1cHFx0dvv4uKC2NjYTJ1r6NChKFWqlF7yNzaOKiciIlUy1jzumJgYODg4KPttbGyyE1aGJk6ciJUrV2Lfvn2wtbU1yTkAJm4iIlIpY00Hc3Bw0EvcGSlevDgsLS1x584dvf137tyBq6vrK4+dPHkyJk6ciJ9//hl+fn6vG3KmsKmciIhUKaeng1lbW8Pf319vYFnqQLPAwMAMj/vf//6HcePGYefOnahVq1YWz5p1rHETERH9Izw8HN26dUOtWrVQu3ZtTJ06FQkJCQgJCQEAdO3aFaVLl1b6ySdNmoRRo0Zh+fLl8PLyUvrCCxcujMKFC5skRiZuIiJSJXOsVd6+fXvcu3cPo0aNQmxsLKpXr46dO3cqA9Zu3LgBC4t/G6tnzZqF5ORktGnTRu9+Ro8ejTFjxmQj+owxcRMRkSqZa8nT/v37o3///gZv27dvn97f165de82zvD72cRMREeUirHETEZEq8bKehpkvcW8VoKCY7fTG1t3cAZhA2+7mjsC43jB3AKYw0twBGF8fyTvfC6lGazTmDsFoDK85ZhqC7DWV57130gtsKiciIspF2FRORESqxKZyw5i4iYhIlZi4DWNTORERUS7CGjcREamSueZxqx0TNxERqRKbyg1j4iYiIlVi4jaMfdxERES5CGvcRESkSuzjNoyJm4iIVEmH7DV359XEzaZyIiKiXIQ1biIiUiU2lRvGxE1ERKrEUeWGsamciIgoF2GNm4iIVIk1bsOYuImISJXYx20Ym8qJiIhyEda4iYhIldhUbhgTNxERqRITt2FM3EREpEqC7PVTi7ECURn2cRMREeUiWUrcY8aMgUaj0dt8fHxMFRsREeVjWiNseVGWm8orV66Mn3/++d87KMDWdiIiMj5OBzMsy1m3QIECcHV1NUUsRERE9B+y3Md9+fJllCpVCmXKlEHnzp1x48aNV5ZPSkpCfHy83kZERPRf2FRuWJYSd0BAABYvXoydO3di1qxZiI6ORr169fD48eMMj4mIiICjo6Oyubu7ZztoIiLK+5i4DctS4m7atCnatm0LPz8/BAcHY/v27fj777+xevXqDI8ZNmwY4uLilC0mJibbQRMREeVX2RpZ5uTkhAoVKuDKlSsZlrGxsYGNjU12TkNERPkQB6cZlq153E+ePMGff/6JkiVLGiseIiIiAGwqz0iWEvenn36K/fv349q1a/j999/x/vvvw9LSEh07djRVfERERJRGlprK//rrL3Ts2BEPHjyAs7Mz6tati8OHD8PZ2dlU8RERUT6lQ/ZqzXm1qTxLiXvlypWmioOIiEgP+7gN47JnRESkSrw6mGG8yAgREVEuwho3ERGpEpvKDWPiJiIiVWJTuWFsKiciIkpjxowZ8PLygq2tLQICAnD06NFXll+zZg18fHxga2uLqlWrYvv27SaNj4mbiIhUyRwLsKxatQrh4eEYPXo0oqKiUK1aNQQHB+Pu3bsGy//+++/o2LEjevbsiRMnTqB169Zo3bo1zp49+xpnzxwmbiIiUiWdEbasmjJlCkJDQxESEgJfX1/Mnj0bBQsWxMKFCw2WnzZtGt5991189tlnqFSpEsaNG4eaNWti+vTpr3H2zGHiJiKiPO3lS0snJSUZLJecnIzIyEgEBQUp+ywsLBAUFIRDhw4ZPObQoUN65QEgODg4w/LGwMRNRESqlLpy2utuqTVud3d3vctLR0REGDzf/fv3odVq4eLiorffxcUFsbGxBo+JjY3NUnlj4KhyIiJSJWONKo+JiYGDg4OyP7dfsZKJm4iIVMlY87gdHBz0EndGihcvDktLS9y5c0dv/507d+Dq6mrwGFdX1yyVNwY2lRMREQGwtraGv78/9uzZo+zT6XTYs2cPAgMDDR4TGBioVx4Adu/enWF5Y2CNm4iIVMkcC7CEh4ejW7duqFWrFmrXro2pU6ciISEBISEhAICuXbuidOnSSj/5wIED8fbbb+Pbb79F8+bNsXLlShw/fhxz587NRuSvxsRNRESqZI4lT9u3b4979+5h1KhRiI2NRfXq1bFz505lANqNGzdgYfFvY/Vbb72F5cuXY8SIERg+fDjKly+PjRs3okqVKtmI/NWYuImIiNLo378/+vfvb/C2ffv2pdvXtm1btG3b1sRR/YuJm4iIVIlrlRvGxE1ERKrExG0YR5UTERHlIqxxExGRKgmyNzhNjBWIyjBxExGRKrGp3DCzJW7/zoCluU5uAnXNHYAJvBln7giMq5W5AzCBPj+ZOwLj02o05g7B6ObLCXOHYDTx8U8wybGeucPI11jjJiIiVWKN2zAmbiIiUiVzLMCSGzBxExGRKrHGbRingxEREeUirHETEZEqsancMCZuIiJSJTaVG8amciIiolyENW4iIlIlHbJXa2ZTORERUQ5iH7dhbConIiLKRVjjJiIiVdIie7XLvDo4jYmbiIhUiYnbMDaVExER5SKscRMRkSpxcJphTNxERKRKbCo3jImbiIhUiTVuw9jHTURElIuwxk1ERKrEldMMY+ImIiJV0gLQZPP4vIhN5URERLkIa9xERKRKHJxmWJZr3Ddv3sRHH32EYsWKwc7ODlWrVsXx48dNERsREeVjWiNseVGWatyPHj1CnTp10LBhQ+zYsQPOzs64fPkyihQpYqr4iIiIKI0sJe5JkybB3d0dixYtUvZ5e3sbPSgiIiIOTjMsS03lmzdvRq1atdC2bVuUKFECNWrUwLx58155TFJSEuLj4/U2IiKi/6IzwpYXZSlxX716FbNmzUL58uWxa9cu9OvXD5988gmWLFmS4TERERFwdHRUNnd392wHTURElF9pREQyW9ja2hq1atXC77//ruz75JNPcOzYMRw6dMjgMUlJSUhKSlL+jo+Ph7u7O8oBsHz9uFXH09wBmECcuQMwslbmDsAErps7ABPIi82b8+WEuUMwmvj4J3B0rIe4uDg4ODiY6BzxcHR0hD+yN/UpBUAkYNJYzSFLz0nJkiXh6+urt69SpUpYt25dhsfY2NjAxsbm9aIjIqJ8S5C95u5M10pzmSwl7jp16uDixYt6+y5dugRPz7xY3yQiInPKbutLXmy9AbLYxz148GAcPnwYEyZMwJUrV7B8+XLMnTsXYWFhpoqPiIiI0shS4n7jjTewYcMGrFixAlWqVMG4ceMwdepUdO7c2VTxERFRPsUFWAzLcr9/ixYt0KJFC1PEQkREpNAhe/O4OR2MiIiIFA8fPkTnzp3h4OAAJycn9OzZE0+ePHll+QEDBqBixYqws7ODh4cHPvnkE8TFZW0ODy8yQkREqqT2wWmdO3fG7du3sXv3bjx//hwhISHo3bs3li9fbrD8rVu3cOvWLUyePBm+vr64fv06+vbti1u3bmHt2rWZPm+W5nEbQ+r8PM7jVj/O41Y/zuPOHTiPO6vnME6e0AK4AtPM4z5//jx8fX1x7Ngx1KpVCwCwc+dONGvWDH/99RdKlSqVqftZs2YNPvroIyQkJKBAgczVpdlUTkREedrLy26nXRTsdR06dAhOTk5K0gaAoKAgWFhY4MiRI5m+n9QfFZlN2gATNxERqZSx1ip3d3fXW3o7IiIi27HFxsaiRIkSevsKFCiAokWLIjY2NlP3cf/+fYwbNw69e/fO0rnZx01ERKqU3VHhqcfHxMToNZW/ajXPL774ApMmTXrl/Z4/fz6bkb1oBWjevDl8fX0xZsyYLB3LxE1ERHmag4NDpvu4hwwZgu7du7+yTJkyZeDq6oq7d+/q7U9JScHDhw/h6ur6yuMfP36Md999F/b29tiwYQOsrKwyFVsqJm4iIlIlY9W4s8LZ2RnOzs7/WS4wMBB///03IiMj4e/vDwD45ZdfoNPpEBAQkOFx8fHxCA4Oho2NDTZv3gxbW9ssx8g+biIiUiU1r5xWqVIlvPvuuwgNDcXRo0fx22+/oX///ujQoYMyovzmzZvw8fHB0aNHAbxI2k2aNEFCQgIWLFiA+Ph4xMbGIjY2Flpt5qNljZuIiFRJi+xd4cvUK6ctW7YM/fv3R6NGjWBhYYEPP/wQ33//vXL78+fPcfHiRTx9+hQAEBUVpYw4L1eunN59RUdHw8vLK1PnZeImIiJ6DUWLFs1wsRUA8PLyQtqlUho0aABjLJ3CxE1ERKpkjj7u3ICJm4iIVEntTeXmwsFpREREuQhr3EREpEo6ZK/GnaMX4shBTNxERKRK2b0ed15N3GwqJyIiykVY4yYiIlXSgjVuQ5i4iYhIldhUbliOJ+7Uyed5bZh+irkDMIG89pgSzR2ACSSbOwATMOUyleYSH//E3CEYTXx8AgAYZSGR/5LdMzBxG8njx48BAFdz+sQmdsXcAdB/ijR3AJRv/ehYz9whGN3jx4/h6Ohokvu2traGq6trpq9r/Squrq6wtrY2QlTqoZGc+NmUhk6nw61bt2Bvbw+NJjuNIK8WHx8Pd3f3dNdhzc34mNQvrz0egI8pt8ipxyQiePz4MUqVKgULC9ONb05MTERycvbblKytrV/rClxqluM1bgsLC7i5ueXY+bJyHdbcgo9J/fLa4wH4mHKLnHhMpqppp2Vra5vnEq6xcDoYERFRLsLETURElIvk2cRtY2OD0aNHw8bGxtyhGA0fk/rltccD8DHlFnnxMZFhOT44jYiIiF5fnq1xExER5UVM3ERERLkIEzcREVEuwsRNRESUizBxE5mJTqdT1nvmGFEiyiwmbtLzcgJRe0JJG5/aY9XpXlxaJzHxxeVOLCwscP78eQAw6fK/piQiykZEOSNfJe7UL5fLly/j4sWLZo5GfXQ6nZJAzp07B0C9CUW5ypzu3+vMaTQaVScQCwsLXL16FWFhYbh27RrWrl2LKlWq4PTp0+YOLctSn+dnz55Bo9FAo9EgMjJS+SFCL6j1/Zj2c0O5T75J3CICjUaD9evXo2nTpti5cyf++usvc4elGjqdTrlgwJgxY9C3b1+sXr3azFEZlvpa/vLLL+jXrx86d+6MwYMHA1DvD41U9+7dw+bNm9GlSxd89NFHWLx4Mfz8/FT7BZ8RjUaDW7duoXr16jh37hx27dqFhg0bIi4uztyhqUbaH8IJCQl48uSJWV7n1HNeu3YN58+fx6NHj0x6cRDKAZKP/PTTT1KwYEGZPn263L1719zhqIJOp9P7e8SIEVKsWDH56aef5MaNG2aK6r+tX79eChcuLP3795dJkyaJq6urBAYGyoMHD8wdWoZSn+tvv/1WLCwspHbt2nL+/Pl0t+cW58+fl86dO0vRokXF2tpa1q1bJyK573GYQtrnYPz48dKqVSvx8PCQIUOGyPbt23M8nnXr1kmJEiWkTJkyUqJECdm6datotdocj4OMI88m7vv37yv/1+l0kpKSIp06dZKwsDC9cnzz/uvSpUtSrVo12bx5s95+tX0Rx8bGSvXq1WXKlCkiInLz5k0pXbq09O3bV6+c2uJOjWfWrFkyYcIEKVeunLRr106OHTuWrszL/1eLb775Rtq3b6/8/eOPP4pGoxF7e3uJiooSEZGUlBRzhac6w4cPl2LFismKFStk6dKlEhAQIL6+vnLz5k2Tnzv1/XP16lXx9vaW6dOny969e+Xjjz8Wa2trWbx4sTx//tzkcZDx5cnEPWrUKBk+fLgkJyfr7Q8MDJSRI0eKSPovlz///DPH4lODjh07yo4dO/T2HTt2TBwcHOTIkSPpyiclJSn/N3dCuXr1qlSsWFESExOVpN2nTx/l9m3btpkxuvQyer4OHDggZcqUkbZt20pkZKTefrVauXKl2NjYSGhoqIiInDp1SubPny89evSQYsWKKbEzIbxokahRo4bs379fRET27NkjdnZ2snDhQhHJmR84e/fulSVLlsjgwYP19g8dOlSsra1lyZIlfK1yoTyZuJcsWSJnzpwREZGnT58q+xs3bixBQUHK36kfnL/++ksmTZqUb5L3n3/+KWPHjk33wyYyMlK8vb31atypLRLr16+XefPm5WicGYmLi5M33nhDFixYIJ6entKnTx/lsVy9elVatmwp+/btM3OUL6Qm7d9++01mzpwpI0eOlPPnz0tCQoKIvEjSZcuWlfbt28uGDRtk7NixotFo5Pbt22b/gWSIVquVjRs3SqFChfR+LJ09e1Y6d+4sxYoVk99++03Zv23bNr3ugLzs5UR89uxZKVu2rDx9+lTp2pk1a5aIvPheWrZsmclr3p06dRKNRiN16tRR3nOpPv/8cylUqJDMmTOHyTuXyZOJO9WePXtk8ODByhfHzp07xd3dXQYMGKBXbujQoVK1atV82e89c+ZMWbRokfJ3nTp1xM/PTy5evKjsS0xMlJYtW8rHH3+c4/EZSl5xcXHSoUMHKVSokLRu3VrvtqFDh8obb7wht27dyqkQM5Qa+7p168TR0VFatGghFSpUkICAAJk6darExcWJiMivv/4q/v7+Uq1aNfH29pbjx4+bM2yDXm7CX79+vRQqVEh69uyp7P/jjz+kc+fOUqRIEVm2bJkMGzZMihYtKtevXzdHyDnq2bNnyv83btwoDx48kAsXLoifn59899134uTkJDNmzFDKHDp0SNq2bWvy1zopKUk++eQTsbW1TdfCJiISFhYmJUqUkL///tukcZBx5bnEnfYLZvbs2WJvby+fffaZ3LhxQ5KSkmTq1Kni6ekpdevWlb59+0qbNm3EyclJTpw4Yb6gc1DaWsH9+/elQ4cOUq5cOVm6dKmyz9fXVypUqCDDhw+X//3vf9KgQQOpXLlyjv8qT30tf/rpJwkPD5fQ0FA5e/asiLxoovX19ZXg4GCZMWOGbNmyRT7++GNxdHSUkydP5micr3Lw4EEpWbKkLFiwQERetO4UKFBAqlatKhMnTpT4+HgREbl+/bqcPXtWbt++bc5w/1PasSOGkvfFixelX79+4urqKn5+fnr993nVrl27pFy5ciIiEh4eLuXLl5fY2FgREenWrZtoNBr56quvlPIJCQnSrFkzadGihVHH2GQ0PkKr1Urnzp3FyclJ9uzZk+64O3fuGC0Gyhl5LnGLvPg1m5iYKCIiixYtklKlSsmgQYPk9u3botVq5fDhw9K+fXt5//33pU+fPnLu3DkzR5zzrl69KiIiJ06ckL59+4qPj4+SvFNSUqRHjx7yzjvvSL169aRnz55KU3RODzzatm2b2NnZSbNmzcTPz08KFSokK1asEBGRo0ePSocOHcTd3V2qVasmjRs3llOnTuVofK+i1Wpl9uzZ8sknn4jIiy6KMmXKSEhIiHTp0kVKlCghkydPzjW1nVOnTomDg4Ns3bpV2WcoeYuIXLt2Te7du5fTIZrF+fPnxc/PT0qXLi2Ojo5y+fJl5bb79+9LixYtpFixYvLVV1/JiBEjpFGjRlK5cmXlM2WM5J2aqPfs2SMDBgyQtm3byoIFC/S6wzp16iROTk7yyy+/ZPt8ZF55LnHfvHlTPDw89Po4FyxYIKVKlZLBgwfLtWvX9Mrnx1HlixcvVmoIIiJnzpyR0NBQ8fHxkR9++EHZ/+zZM71+sZyuccfHx8uIESNk7ty5yr5BgwaJtbW1/PjjjyIikpycLI8ePZJHjx6l68NTg6tXr8q5c+fk6dOn0qBBA+nRo4eIiDx+/FiZnjNlyhRV9mcb0qpVKylRooTs3LlT2ZeavNP2eec3H3/8sWg0GilfvrzynZL675MnTyQ8PFzeeecdeffdd2Xw4MHKZ8mYn6n169eLk5OTtGvXToYMGSIWFhYydOhQva6KLl26iEajUQbMUe6U5xL3gwcPxNnZWak9pkpN3p999lmunjtrDDt27JCqVavq/Ro/ffq0hIaGSqVKlWT58uXpjsnp5+nEiRPi5OQkNWrUkC1btujdlpq8ly9fnm6AnTmlfY4MDfzz9fWVw4cPi8iLWlrz5s2lT58+6X5MqkVGTa8dOnSQIkWK6CXvjRs3ikajkYEDB+ZkiGaX+rz8+uuvsmnTJnnjjTekcuXK8ujRIxFJ/z5IW1EwZtKOiooST09PmTNnjoi8+NHt6OgoGo1GQkJCJCYmRikbGhoqFy5cMNq5Kefl+sRtKKG0bNlS/ve//4mI/qjyhQsXiq2trXz55Zf5ZhSloefn3LlzYm9vn65Z+dSpU9K3b18pUqSI7N69O6dCNCg5OVkZEZvaP5z2sQwZMkQ0Go2sXbvWXCHqSY1t586d0qNHDwkMDJSvvvpKfv/9dxEROXz4sJQtW1Z+/PFHiYuLkzFjxkibNm3k8ePH5gz7P+3fv195n6R9/jt27Kgk79T9W7duzRcjyF9upUv799mzZ6V69epSuXJlZfyCyIuKQ9rFgYz5Q1in08n27dtl1KhRIiJy48YN8fT0lPDwcNm2bZtYWlrK4MGD882smfwg1yduEZHdu3dLvXr1pH///rJ06VKpVauWtGvXzmDZH374QS5dupTDEZrftGnT5LPPPpM1a9bIvHnzpE6dOkrtL61Tp07JxIkTc7wv29AXmU6nk3bt2kmRIkUMNu0NHz5cVYli48aNUrhwYRk0aJDMnDlTypQpI3Xq1JHLly/Ls2fPpFWrVuLl5SVly5aVYsWK6c3dViOtViv16tWTYsWKKdMr075O9evXl4oVK6ZbsCcvS5ukFyxYIAMGDJDu3bvrdc2dO3dOatSoIeXLl5ft27dLo0aNJDAw0Ojdcmlfi9u3b8upU6ckOTlZWrVqJSEhIZKYmCjJycni6+srGo1GwsLCuDhOHqERyWWLJKeh0+mg0+mwefNmbNmyBU+fPsW5c+fw/PlzXLp0CTVq1IC3tzd8fHxgb2+P7t27w8XFxdxh57jExER06tQJIoJz586hQIECOH/+PDw9PdGwYUOUKVMGnp6e8Pb2RmBgICwtLQEAWq1W+b8pyT9rjx85cgRHjhzB06dPUalSJbz33nsAgLZt2+KXX37Bhg0bUL9+fZPH8zpiY2PRqlUrdOnSBQMGDIBWq4WLiwu6d++O//3vf7CwsMCzZ8+wdetWPHv2DHXq1EHZsmXNHXY6qa9FfHw8HBwckJCQgNatW+PKlSvYvHkzqlatqpT9+OOPMX/+fLi5ueHMmTMoVKiQGSPPWV988QWWLVuG+vXrw8bGBosXL8aSJUvQpUsXAC/WBe/Vqxdu3bqFUqVKYceOHbCyslKe3+xIvY+4uDjY2NhAo9HAxsYGAPDo0SM0adIE4eHh6NixIxITE/HFF1+gbt26qFy5MipVqpTtx04qYMYfDa/tVc1McXFxsnr1aqlRo4ZMmDBBhg0bJoGBgVKjRo18U9N+1S/7u3fvyr179yQgIEA8PDxkxIgR4u/vLx4eHhIcHGy2Pv+1a9eKk5OTtGnTRpo2bSrlypVTBjs9f/5c2rVrJ66urvLzzz+bJT5D0j5XDx48kJo1a8rdu3flzz//lFKlSimri4mI/PLLL/LkyRNzhJlpaZv7+/TpI7/++quIvBgk2LBhQ/H29pbTp08r3UyfffaZHDp0SPVT2Ixt0aJF4uHhoUx127Fjh2g0GrGyspL/+7//0yt76dIl5Xk1Rvdc2m6Jhg0bSu3ataV69eqya9cuefLkifz111/i6Ogoo0aNkjNnzsiXX34pZcuW1Wu2p9wv1yXu1Dfu77//LuPHj5dJkybJqlWr9MocOnRIbG1tlQEYSUlJegsk5GVpk/auXbtk1apVsnLlynRTjkaPHi2dOnUSkX+/UFKPzenkfeHCBfHw8JCZM2eKyIuBcg4ODso0qlRNmjSRMmXK6I1bMLclS5bI/Pnz5dGjR+Lh4SHLli2TcuXKSWhoqNIseenSJfnggw9yxUjetWvXSsGCBSUiIkJvDMTjx4+lQYMG4uzsLL1795bOnTuLo6OjMq0wv3j69Kl88803ykyHLVu2iIODg8ybN0/GjBmT4Rrgxmwm37p1q9jZ2UlERIQcPnxYOnToILa2tsoPicWLF4tGo5GyZcuKq6ursoY85R25LnGLvFiJqnDhwhIUFCT+/v5iY2MjvXr1Uj4scXFx4uvrq6raWU4bMmSIuLi4SNWqVcXa2loaNGgga9asUW6fO3euuLm5yePHj/W+ZMwxPe6nn34Sf39/EXkx/9fDw0NvatHRo0dF5MUPir/++ivH43tZ6g+bS5cuia2trURERIjIi1XbLCwspHnz5nrlhw8fLtWrV1dF7K9y6tQpKV26tDIYMNWVK1eU/4eFhUnz5s0lODhYVXPmTcXQj9gLFy5IdHS0REdHi6+vr0ydOlVEXgxAtLS0NNmgSZ1OJ4mJifLee+/J6NGjRUQkJiZGypUrJ71799Yre+rUKTl8+HCOXMyEcl6uS9xXr14VNzc3pUkqPj5etm/fLkWKFNF78/r4+MiIESPMFaZZ/fDDD+Li4iKRkZESHx8vt27dkiZNmkjDhg2VSwoePHhQXFxc5OHDh2aO9sV63cHBwXLixAlxd3eX3r17K7XVY8eOySeffKK6EbFHjhyRyZMny9ChQ5V9Z86ckfbt20vJkiVl0aJFsnDhQhkwYIDY29urajW3jGzbtk0qV64sIi9G9S9atEgaNWokJUuWlM6dOyvlUlJSlAWO8rK0F9YxtEjOvn37pHr16sp789SpUzJo0CBZvny5UZrFJ0yYIMOHD9f7Mf3kyROpXLmyHDlyRP7++28pVaqU3vfe3Llz883CN/mZahO3oZqfTqeTkydPSpkyZdJ9kW/ZskUKFiwomzZtEp1OJwMHDtRbbzsve/m5Gj58uDRp0kS5nKnIi1GntWvXVtb2vnz5srRo0cJso8ejoqLk1KlT8vz5c7l48aK4urpKgQIF0i3iMXDgQAkODlbFD4xU9+7dk5YtW4qdnZ106dJF77aoqCiltaNmzZrSsmVLOX36tJkizZpff/1VfHx8pFOnTuLv7y+tWrWSfv36yerVq0Wj0ei12ORlL9eWIyIipG7dutKyZUuZPXu28nnbunWraDQa2bp1q1y8eFFatGghbdq0UY7LbvL+/vvvRaPRSEREhF7Nv127dtKjRw/x8PCQfv36KT8w4uLipGnTpnprolPepNrELfJiPmLql8WKFSskNDRUaZ7csGGDXtm7d+9KhQoVlAUI1LQwR05JXchj0KBBUqdOHWV/au3o559/loIFC8rly5f1vlRyKnmnvehGiRIlZNSoUcrFQNatWycajUa+/PJLiYqKknPnzsmQIUPEyclJmYqkJps2bZImTZpkGN/du3clJSVFlau5ifz7Wjx48ECpoSUmJsrMmTOlTZs28tlnnymPKz4+XurUqSN79+41V7g5ZsGCBVKmTBmZMGGCiIjMmzdPihQpIhEREdKkSRN58803pX///spnpnfv3qLRaKRMmTJSvXp1o33vpL4+8+bNEwsLCxk3bpzymZ0+fbq4u7vLW2+9pXfMsGHDpGLFihIdHW2UGEi9VJu4k5OTpUOHDvLWW2/J4MGDRaPRyJw5c0Sr1Ur79u2lRYsWepcP1Gq1EhgYqAxwyg8roq1fv15Wr14tIqJchEPkxcA9jUaTboRr6opp5rxy1t69e6Vw4cKycOFCvQtWiLxo5itRooS4uLhIlSpVxM/Pz+wXf0lJSVHeS4mJiXrNpz/99JO888478sYbbygXP0ktn3qMmt+H69evlzfffFM8PT1l8ODBGbZQjRo1Sry8vOTGjRs5HGHOu3XrlgwaNEgCAgJk7NixMnToUGWe+tOnT2XChAlSq1YtCQsLU2re+/btk/379yvJPLs17ZffP0uXLlWSt8iLi5T07dtXqlWrJq1bt5bRo0dLx44d89XFkvI71SZuEZFHjx5JQECAaDQa6devn7J/y5Yt0rBhQwkODpZly5ZJZGSkfPrpp1KsWDHV9YWayuPHj+Xjjz+WAgUKSOvWraVgwYJ6/agRERFibW0tERER8ueff8rVq1elWbNm0qhRI7Ouzz506FBlcZzUL6e0Nf7r16/LkSNH5PTp0+kSe056eQT4li1bJDg4WJo3b64MRhN58WOoadOmEhAQIH/88YeIqHf9+7Q/Io4dOybOzs4ycuRIGT9+vHh6ekrr1q31LkCxefNmCQ0NleLFi+eLkcmpr9udO3fkk08+kbp164qnp6feQkXx8fESEREhtWrV0qt5pzJG61Xq67R7927lssSLFy8WCwsLGTt2rIi8+PzPmjVL3nvvPWnYsGG+vVhSfqXqxJ2cnCzvvPOOVK9eXRo3bqx3AYytW7dK165dxdbWVnx8fMTHxydffLmk9fDhQ2VVpG+//VZE/v3yefDggUyfPl3s7e2ldOnSUq5cOaldu7ZRr0iUVSkpKdK4cWPp2LGjsi9tMrl27ZoqlqI9efKkaDQaGT58uIi8aCWws7OT3r17S9euXcXGxkZCQkKU8jt27JAWLVpIxYoVVbWSW6qVK1fqxXXlyhX55ptvlBqcyItE7u/vr5e858yZI6GhofkqIaS+H2NjY2Xw4MHi6OiYbv31x48fy6RJk8TT01P53BnbunXrxM7OTsaNG6dM85o7d65e8jYUN+UPqk7cIi+aJ2/fvi3NmzeXhg0b6iVvEVGmZZizdmYu9+/fl86dO0u7du3E0dFRaTZP69q1a7J3717Zu3ev0ZryMiv1yyQuLk4551dffSU+Pj5K03KqmzdvyqeffqqKxJeYmChz584VW1tbGTNmjGzevFn5gn7+/Lns3LlTHBwcpFu3bsoxmzZtkrZt26qufzEmJkbq1q2rNHM/fPhQSpcuLXZ2djJgwAC9skeOHJGaNWvKBx98oCy+oqY58zkl9X17//59GTx4sPj7+8vXX3+tVyY+Pl5+/PFHk4wPuXjxonh7eyvdfmnNmTNHLCwsZPz48ar4kUvmofrEnerPP/+U5s2bS6NGjWTJkiUiIvLFF19I3759zRyZeWm1WomNjZX+/fuLg4NDuuSd9pJ+Ijk/EG3r1q0SEhIi+/btE51OJz/99JO88cYbEhoaqiTv5ORkGTVqlHh7e+tdxSgnGWqBmD17ttja2oqzs7NMmTJF77adO3eKvb29cplOEVHtymipyff06dPy8OFDOXTokHh4eEjdunXT9YkeO3ZMvL29pVOnTvkyaadKff/evXtXBg4cKLVr15bx48cbLGvsz9Tu3bulQoUKeleNS/v+XLp0qWg0Gvnmm2+Mel7KPXJN4hZ5MYf7/ffflypVqsgbb7whDg4OBi+UkR9du3ZNBgwYII6OjsplOd977z359NNPzRbTunXrpFChQjJ27Fi5fPmysn/hwoVSt25d8fb2luDgYGnUqJEUKVLE7F0dN27cUH74rFq1Sjp16iQLFiwQR0dH6dWrV7ryP/30U7rxF2oVFxcnVatWlY4dO8qDBw/k0KFD4u7uLt27d083XS0yMjLfrYhmSGryvnPnjgwaNEjeeustvXn7prJhwwZxd3dXErdWq1Vi2bt3r5w/f15Wr16dr7owSF+uStwiIn/99ZcsWLBAxo4dy2vKvuTatWvK5S6rVKkiFSpUMNu0uD/++EPc3d1l4cKFyr7nz59LdHS06HQ6uXLlikyfPl0++ugj+frrr80+5z7tLIZBgwaJRqORRYsWiU6nkwULFoiVlZXBBX327NmTa96Hx44dk1q1akmPHj3k4cOH8uuvvyrJW41T7kzp5T7hjPqI09a8u3fvLqGhoSbvT7569arY2dkpYyzSGjRokIwcOZJX+crncl3ipldLSEiQvXv3yoIFC3K8TzutI0eOiL+/v/z555/y9OlTmTFjhrz99tvi5eUlQUFBqrwwRUazGJ49eybz58+XAgUK5PrV+KKioqR69ep6ybtMmTLy4YcfKqPi85PMXHgoNVE/fPhQabI29eDO1B+LqfPpz507J59//rk4OTmpYhwImRcTdx5nrgEsBw4cEDc3NwkJCZFy5crJe++9J1988YUsXbpUypcvL0uXLjVLXK/y8iyGtDE+ffpU5s+fL3Z2djJ48GAzRpl9aZP3o0ePZO/evVKlSpV8t671Dz/8IM2aNROR/x6VndPr+Wu1Wlm9erUUKVJE3NzcpFy5clKxYkWzdyeROuTq63HnZTqdDhYWFgZvk1dc0/dVx5lCSkoKLC0tlesDA4CjoyMAYMWKFThw4ACKFSuGkJAQ5frTgYGBGDJkCNq0aZNjcWZWUlISHj16hF69euHp06fo0aMHPvroI+X27777DpMmTcKZM2fg7Oxsxkiz58SJE+jduzfKlCmDuXPnwtraGnZ2duYOK0ft3r0bwcHB2LNnDxo2bJhhubSft59//hne3t45di31W7du4fr169BoNPD29oaLi0uOnJdUzry/G8iQtL/oZ8yYIZ988om8++67smbNmleOuk5ba9i/f3+6KVfGtHLlSr2/N2zYINWqVRMfHx8JCAiQdevWGbyU6ogRI8TDw0N106ZelnYWQ+oUxFGjRkm3bt3kwYMHZo7OOI4ePSr169c360p6OSXtJWt1Op3yd0hIiHTr1k0SEhIM1rrT7ps1a5ZoNBo5cuRIzgRNlAEmbhX7/PPPxdnZWcaNGye9evWSsmXLSkhIiMH1r9N+wUyfPl1KlCghkZGRJokrJiZGbGxspEmTJiIicuLECbG1tZURI0bIkiVLpHXr1uLj4yMTJkxQktyCBQukZ8+eUqJEiVzT3Jd2FkOtWrXE0dExz81iyC/XqU/18sVqUtf9Th1zkfZzlPb/s2fPliJFiuSbC62QujFxq9SePXukbNmycvz4ceXvAgUKKFO90nr5C8bJyUlWrVpl0vj2798vHh4e0qJFC9m2bZuMHDlS7/bPP/9cfHx8lHWeV69eLb169cp1A2s4iyH3Snt1PJEX70EXFxeZM2eO3ij6wMDAdFd5e/kz5eDgYJJrbBO9DiZulXh5esemTZvkzTffFJEXzdL29vbKSkpPnjyR/fv3p7smck5/wRw8eFDc3NxEo9FI9+7dRUT/cTRt2lQaNGig/J3fandkXo8ePVL+v3XrVlmyZImMGzdOypYtKwEBAdKnTx+5evWqfPvtt9KmTRtl3vTLrVdFixZl0iZVYeJWmcmTJ0t0dLQsX75c6tWrJz/99JM4ODjI9OnTlTIbNmyQjz/+WK9vctasWWJvby/r1q0zaXwv9wMePHhQ/Pz8pGrVqhIXFyci/ybv7777Tt544418vQIXmcf+/fulePHicvfuXRkyZIiUL19e+bycP39eVqxYIZUqVZJ33nlHatWqJRqNRm+JUa1WK+fPnxeNRmPy1iuirGLiNrO0A9Hmz58vGo1Gzp49K0+ePJHy5cuLRqPRW5/92bNn0qxZM+ncubOSRH/77Tdxc3Mzea0g9XynTp2SHTt2yMaNG+XOnTty4MABKV++vDRq1EgePXqkN/CnXr16rGlTjjt58qS0atVKihUrJkWKFFEGdab9vGm1Wlm5cqUMHjxYbG1txd/fP92KcWofREn5ExO3SuzcuVNmzZql9+t+586d4unpKU2bNpU9e/bI6tWrpUmTJlKlShW9eaXPnz/Xu6SnKa1Zs0aKFSsm1atXF41GI3Xr1pWpU6fKgQMHpEKFCuLj4yMffvihDBw4UOzt7XMsLqJ69erJ999/r/w9dOhQ0Wg04urqqlxkJfVzY6hrytPTU/bs2WPwdiI14TxuFTh58iTeeustJCcnY8mSJejcuTMA4OnTpzhy5Ag+//xz3L17Fy4uLvD29sbSpUthZWUFrVYLjUaTY/O2T5w4gSZNmmDixIn44IMPkJSUhKFDhyImJgYffPABqlWrhvDwcERGRmL//v1wd3eHl5dXjsRG+ZtWq8XWrVvx7rvvwsbGBgBw5MgRxMbGYsmSJTh27Bh++uknVKpUCcnJybC2tk53Hx988AGsrKywYsWKHF0LgSjLzP3LIT96uZ/4wYMHMmfOHClZsqR06tTJ4DHR0dHy8OFD5VhzrIi2bNky8fX1lbi4OCWO27dvS6dOnaRBgwby+PFj+eWXX8TX19dsV/ki+vrrr/VWtzt+/Li0aNFC3Nzc9NbEX7hwofz1119K8/n7778vvXr1Ym2bVI+JO4el7WPT6XTKyPDUa0AXKlRIwsLClDJJSUmvvI+ctGLFCilbtqwy5zX1x0N0dLRoNBr55ZdfRCR/XsOZ1GP69Omi0Wj0LtIRGRkpLVu2FGdnZ1m5cqUEBQVJrVq1lM/SlStXRKPRmGztAyJjKmDuGn9+knY50m+//RanTp1CVFQU+vTpgwYNGiA0NBQAMGLECFhYWOD777+HtbV1umVMzdWM98Ybb+Cvv/7CjBkzMG7cOBQo8OLto9FoULlyZWXJzPy2dCaZz9GjR1G4cGH4+vpiyJAhaNCgAfr06YPChQsjNDQUOp0OERERqFmzJiZOnIhJkybhyy+/RIUKFfD777/DwsICOp0OZcuWxaNHj5TleonUjIk7B6Um3GHDhmHBggUYNWoUfH19MW3aNOzatQsrV65Eu3btoNFoMGLECPz999/44YcfVNPfVrZsWSxYsAA9evRASkoKevbsCQcHB8ybNw9xcXHw8PAwd4iUj0RHR6NHjx6oU6cOnj9/jsWLF6Nbt24oUKAAOnfuDJ1Ohz59+gAAIiIi4OvriyVLluCvv/5C6dKlodFolLX2AcDBwcGcD4co88xd5c9vjhw5Ij4+PsrSmQcPHhQrKytZsmSJUiYxMVGmTp0qzZo1M1uzeEZ0Op2sWLFC7O3txcPDQypUqCBubm5sYiSzWLlypbi6uoq1tbWyhkHq+Ivk5GRZuHChWFtby5dffpnuWLV9togyizVuE3u5mTslJQW2trYICAjA6tWr0bNnT3z//ffo2rUrEhISsH//fjRq1Ai9e/fGJ598Ao1Gk+NX/HoVjUaDDh06IDAwEBcuXIBWq4Wfnx/c3NzMHRrlI6mfiZIlS8LJyQlOTk74+eefUa5cOfj5+QEArKys8NFHH0Gj0aBHjx5wd3dXauCA+bqciLKLidvEUr8cYmNj4erqioSEBCQmJmLlypXo27cvJk6ciL59+wIAfv/9dyxfvhzly5dH+fLlAby4pKAav2A8PT3h6elp7jAon0lN2KmfiTfffBNnz57FihUr8N133yE5ORkDBw5E1apVAbxI3t26dYOzszOCg4PNGTqR0agvI+RBCxYsQOPGjQEAjRs3hpubGzp16oTx48cjLCwMAJCYmIhp06YhKSlJ71q/GV13myi/SdvytG3bNqxfvx6HDx+GpaUlPvroI/Tr1w8nTpzAjBkzcPr0aQBA69atsXv3bjRv3hwFChRASkqKOR8CkVFwAZYcEB0djXr16mHcuHEICQlBZGQkwsLCcP/+fYwZMwaPHj3Cli1bcOvWLZw8eRIFChRQVfM4kbmJiPIjdsiQIVi6dCksLCxQvHhxNGrUCFOnTgXw4kfy3LlzlWNu3bqF6OhoWFlZmSt0IqNj4jaytF8wwIsVnZ4+fYp+/frB2toaCxcuxPPnz3Hp0iWMGzcOp06dQokSJVC+fHnMmjULVlZWSElJUaZaEdG/Ll++jJCQEMycORNWVlbYtWsXZs6cifr162P+/PkAgM2bN+OPP/7AgwcPMHHiRKWmzc8U5RVM3CZy584duLi4KH/v2rULTZs2xY4dO/T62u7fvw8nJyflS4VfMESGLVy4EGvXroWLiwvmz58PS0tLxMXFYcWKFZg8eTIaNmyIefPmpTtOq9UqU76I8gK2xZrAnDlz0K5dO0yYMAHPnz9HSkoKgoOD0a1bN8yfPx+PHj2CTqcDABQrVkxJ1CLCpE1kwJMnT3DhwgWcPXsWV65cURKxo6MjOnbsiM8++wwHDx5E27Zt0x3LpE15DRO3EbzcaFGuXDnUr18f06ZNQ4MGDTBhwgTEx8ejZcuWOH/+PO7duwcLC4t0zeociEb0QuoP21SFCxfGgAED0KtXL5w+fRojR45UbktN3r1791ZWQiPKy9hUnk1pB5Hdv38fBQoUgIODAywsLPDgwQOMGzcOx48fx7Vr1zBmzBgMHDgQzZo1w5o1a8wcOZE6pf1MnT59GvHx8XBzc4OXlxeePHmCyZMnY9WqVWjfvj3GjBmjHJeQkICCBQuqbu0DImNju2w2pJ1j/fXXX+PAgQP4888/8eabb+KDDz7Ahx9+iClTpuDx48eYMmUKVqxYgWfPnuHBgwfpattEpP+Z+vLLL7F69WpYWFjg6dOneO+99/D5558rUyhTbxs1ahQAoFChQunugyhPyuGV2vKkESNGSLFixWT9+vWydetWadiwoRQpUkRu3LihV+7y5cuybds25apaL1/ekyg/S/t5mDJliri4uMi+fftERKR3797i5OQkv/76q4iI3Lp1S8aOHStFihSRBQsWmCVeInNhjTubbty4gZ9//hmrV6/GO++8g127diEyMhKTJ0+Gu7u73ojWcuXKoVy5cgA4epwo1dWrV1GmTBloNBpotVpoNBr89ttv+Oyzz/D2229j06ZNWLVqFSZOnIg6deogMTERJUuWRK9eveDm5oZu3bqZ+yEQ5Si2J2XRywNfEhMTcf36dVSuXBlbtmxBmzZtMGnSJISGhiIxMRHz58/HtWvX0t0PkzYR0K9fP/Tr1w9RUVEAXowAT0pKwp07d1C/fn389ttv+OijjzBp0iT07dsXycnJmDt3Lg4ePIhSpUqhR48esLS0hFarNfMjIco5TNxZkHbAy4YNG3D9+nUULVoUlSpVwuzZs9GlSxd88803ytrjFy9exO7duxETE2POsIlUq02bNvjzzz8xefJkREZGAnhxPfeyZcviww8/RJMmTTBz5kzl4iDx8fHYsGGDsqRpKk75ovyEiTuTJM2Al+HDh2PAgAHYvHkzihcvjnLlymHs2LHo3bu3krQTEhIwfPhwJCQkoE6dOuYMnUiVnj9/jkaNGmHp0qU4evQopk2bhsOHDwN4sayph4cHvL298dFHH0FE8PDhQ3Tp0gXJycnK54woP2J7bSaljgAfN24c5s2bh+3bt6NChQoAXiy4kpCQgB9++AHPnj1DgQIFcPLkSdy/fx9RUVHK3FKOdCV6QafTKeuH29vbo1WrVpg3bx4SExMxatQoVK1aFf3798eECRNQsmRJlC1bFs+fP4dWq1UuLMIV0Si/4jzuLHj48CHat2+P7t27o3Pnzrh58yYuXbqEFStWoH79+ti1axe0Wi2ePXsGX19fjB07luskE73Cp59+ijVr1qBdu3aIjY3F6tWr0bx5c4wfPx6VKlXCrVu3sHTpUgCAq6srOnfuDEtLS36mKF9j4s6CR48eoUqVKggJCVH63qKjo6HT6XD79m2MGDECffv21ZujzVoB0QsTJ05E27ZtlcvWHjlyBC1btsS6detQr149AMC+ffvQrl07BAYGYty4cfDz80t3P/xMUX7HttssKFKkCL766ivMnDkTLVu2hKenJ8aPH49jx46hQYMGOHToEAD9pUv5BUMEXLp0CSdPnoSXl5eyz8rKCtbW1rC3twfwIiE3aNAAy5cvx7Zt2zBlyhQcPHgw3X3xM0X5Hduasqhnz55o3LgxkpKSUL58eQAv+utiY2Px5ptvmjk6InWqUKECVqxYAY1Gg23btsHT0xMlSpRAfHw8Ll68iOrVq0Or1cLCwgJvvvkmvLy88MMPP6BMmTJKbZyIXmBTeTY8efIEJ0+exKRJk3D9+nVERUWx343oFWJjYxEYGIgGDRpg3Lhx+OGHHzBu3Dj89NNPSoKOi4vD0KFD0bx5czRr1ow1bKKXMHG/JhHB/v378e233+L58+fYsmULrKys2P9G9B+ioqLQp08f1KhRAx9++CG2b9+O6dOnY/jw4ShSpAi2b9+Ox48f4/Dhw8pqavxMEf2LiTsbkpKScO7cOVSrVg0WFhYc6UqUSSdOnEDv3r3h7++PLl264MKFC5g+fTqsra3h4uKCdevWwcrKihfjITKAidtIOE+bKGuioqLQu3dv1KxZE+PGjYOLiwuSk5NhZWUFjUbDH8JEGWDiJiKzOXHiBEJDQ+Hl5YVJkyYpU8X4Q5goY/xkEJHZ1KhRAzNnzoSDgwO8vb2V/UzaRBljjZuIzC61L5s1baL/xsRNRKrAgWhEmcOftkSkCkzaRJnDxE1ERJSLMHETERHlIkzcREREuQgTNxERUS7CxE1ERJSLMHETERHlIkzcREREuQgTNxERUS7CxE1ERJSL/D/ikvgXWkzRFwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(model.adjacency.detach().cpu().numpy(), cmap='hot', interpolation='nearest')\n",
    "plt.colorbar()\n",
    "plt.xticks(np.arange(len(labels)), labels, rotation=45)\n",
    "plt.title('Learned Adjacency Matrix HeatMap')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "A8WZUwOA3Kv3"
   },
   "source": [
    "**Summary of all the algorithms used till now (classical + DL)**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 426
    },
    "id": "zl--5dhPM_m1",
    "outputId": "8a242411-a25f-4963-d161-1ad92c5856ab"
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "\n",
       "  <div id=\"df-b8335c31-fe42-4c9c-8052-f971ec9dee16\">\n",
       "    <div class=\"colab-df-container\">\n",
       "      <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>model_name</th>\n",
       "      <th>micro_avg_f1_score</th>\n",
       "      <th>hamming_loss</th>\n",
       "      <th>accuracy</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>BR</td>\n",
       "      <td>0.76825</td>\n",
       "      <td>0.076117</td>\n",
       "      <td>0.66718</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Classifier_Chain</td>\n",
       "      <td>0.769028</td>\n",
       "      <td>0.076451</td>\n",
       "      <td>0.672111</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>One_vs_rest</td>\n",
       "      <td>0.765692</td>\n",
       "      <td>0.076117</td>\n",
       "      <td>0.66718</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>One_vs_rest</td>\n",
       "      <td>0.765692</td>\n",
       "      <td>0.076117</td>\n",
       "      <td>0.66718</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Hierarchical_SVM</td>\n",
       "      <td>0.776746</td>\n",
       "      <td>0.073472</td>\n",
       "      <td>0.672265</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>BiLSTM_glove</td>\n",
       "      <td>0.828584</td>\n",
       "      <td>0.075083</td>\n",
       "      <td>0.667951</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>BiLSTM_bert</td>\n",
       "      <td>0.83385</td>\n",
       "      <td>0.076097</td>\n",
       "      <td>0.68041</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>CNN_glove</td>\n",
       "      <td>0.815412</td>\n",
       "      <td>0.07867</td>\n",
       "      <td>0.657473</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>CNN_bert</td>\n",
       "      <td>0.82195</td>\n",
       "      <td>0.079621</td>\n",
       "      <td>0.67003</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>MAGNET_Cooccurance</td>\n",
       "      <td>0.79341</td>\n",
       "      <td>0.09229</td>\n",
       "      <td>0.61402</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>MAGNET_xavier</td>\n",
       "      <td>0.783989</td>\n",
       "      <td>0.095509</td>\n",
       "      <td>0.598151</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>MAGNET_random</td>\n",
       "      <td>0.780365</td>\n",
       "      <td>0.096324</td>\n",
       "      <td>0.588135</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>\n",
       "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-b8335c31-fe42-4c9c-8052-f971ec9dee16')\"\n",
       "              title=\"Convert this dataframe to an interactive table.\"\n",
       "              style=\"display:none;\">\n",
       "        \n",
       "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
       "       width=\"24px\">\n",
       "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
       "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
       "  </svg>\n",
       "      </button>\n",
       "      \n",
       "  <style>\n",
       "    .colab-df-container {\n",
       "      display:flex;\n",
       "      flex-wrap:wrap;\n",
       "      gap: 12px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert {\n",
       "      background-color: #E8F0FE;\n",
       "      border: none;\n",
       "      border-radius: 50%;\n",
       "      cursor: pointer;\n",
       "      display: none;\n",
       "      fill: #1967D2;\n",
       "      height: 32px;\n",
       "      padding: 0 0 0 0;\n",
       "      width: 32px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert:hover {\n",
       "      background-color: #E2EBFA;\n",
       "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
       "      fill: #174EA6;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert {\n",
       "      background-color: #3B4455;\n",
       "      fill: #D2E3FC;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert:hover {\n",
       "      background-color: #434B5C;\n",
       "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
       "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
       "      fill: #FFFFFF;\n",
       "    }\n",
       "  </style>\n",
       "\n",
       "      <script>\n",
       "        const buttonEl =\n",
       "          document.querySelector('#df-b8335c31-fe42-4c9c-8052-f971ec9dee16 button.colab-df-convert');\n",
       "        buttonEl.style.display =\n",
       "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
       "\n",
       "        async function convertToInteractive(key) {\n",
       "          const element = document.querySelector('#df-b8335c31-fe42-4c9c-8052-f971ec9dee16');\n",
       "          const dataTable =\n",
       "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
       "                                                     [key], {});\n",
       "          if (!dataTable) return;\n",
       "\n",
       "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
       "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
       "            + ' to learn more about interactive tables.';\n",
       "          element.innerHTML = '';\n",
       "          dataTable['output_type'] = 'display_data';\n",
       "          await google.colab.output.renderOutput(dataTable, element);\n",
       "          const docLink = document.createElement('div');\n",
       "          docLink.innerHTML = docLinkHtml;\n",
       "          element.appendChild(docLink);\n",
       "        }\n",
       "      </script>\n",
       "    </div>\n",
       "  </div>\n",
       "  "
      ],
      "text/plain": [
       "            model_name micro_avg_f1_score hamming_loss  accuracy\n",
       "0                   BR            0.76825     0.076117   0.66718\n",
       "1     Classifier_Chain           0.769028     0.076451  0.672111\n",
       "2          One_vs_rest           0.765692     0.076117   0.66718\n",
       "3          One_vs_rest           0.765692     0.076117   0.66718\n",
       "4     Hierarchical_SVM           0.776746     0.073472  0.672265\n",
       "5         BiLSTM_glove           0.828584     0.075083  0.667951\n",
       "6          BiLSTM_bert            0.83385     0.076097   0.68041\n",
       "7            CNN_glove           0.815412      0.07867  0.657473\n",
       "8             CNN_bert            0.82195     0.079621   0.67003\n",
       "9   MAGNET_Cooccurance            0.79341      0.09229   0.61402\n",
       "10       MAGNET_xavier           0.783989     0.095509  0.598151\n",
       "11       MAGNET_random           0.780365     0.096324  0.588135"
      ]
     },
     "execution_count": 129,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "scores_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "Sn6nT6jnE_Vs"
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "EgOt9QqjSmZ0",
    "outputId": "74bbfa26-7602-464e-e095-7437af6b7962"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Data shape: (159571, 8)\n",
      "Labels: Index(['toxic', 'severe_toxic', 'obscene', 'threat', 'insult', 'identity_hate',\n",
      "       'clean'],\n",
      "      dtype='object')\n"
     ]
    }
   ],
   "source": [
    "train_df = pd.read_csv('train.csv')\n",
    "print(\"Data shape:\",train_df.shape)\n",
    "#assigning labels\n",
    "\n",
    "#Adding new column of Clean Text\n",
    "arr=[]\n",
    "for i in range(train_df.shape[0]):\n",
    "  if (train_df.iloc[i,2:]==0).all():\n",
    "    arr.append(1)\n",
    "  else:\n",
    "    arr.append(0)\n",
    "train_df['clean'] = pd.Series(np.asarray(arr))\n",
    "labels = train_df.columns[2:]\n",
    "print(\"Labels:\",labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 441
    },
    "id": "k3TI-WZkTRWH",
    "outputId": "c3234da5-bcd5-4c4f-c522-b7d8d6adb69b"
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "\n",
       "  <div id=\"df-2becfe59-e90e-4976-9ad0-46d00a216670\">\n",
       "    <div class=\"colab-df-container\">\n",
       "      <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>id</th>\n",
       "      <th>comment_text</th>\n",
       "      <th>toxic</th>\n",
       "      <th>severe_toxic</th>\n",
       "      <th>obscene</th>\n",
       "      <th>threat</th>\n",
       "      <th>insult</th>\n",
       "      <th>identity_hate</th>\n",
       "      <th>clean</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0000997932d777bf</td>\n",
       "      <td>explanation edits made username hardcore metal...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>000103f0d9cfb60f</td>\n",
       "      <td>daww match background colour im seemingly stuc...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>000113f07ec002fd</td>\n",
       "      <td>hey man im really trying edit war guy constant...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0001b41b1c6bb37e</td>\n",
       "      <td>cant make real suggestion improvement wondered...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0001d958c54c6e35</td>\n",
       "      <td>sir hero chance remember page thats on</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>\n",
       "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-2becfe59-e90e-4976-9ad0-46d00a216670')\"\n",
       "              title=\"Convert this dataframe to an interactive table.\"\n",
       "              style=\"display:none;\">\n",
       "        \n",
       "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
       "       width=\"24px\">\n",
       "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
       "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
       "  </svg>\n",
       "      </button>\n",
       "      \n",
       "  <style>\n",
       "    .colab-df-container {\n",
       "      display:flex;\n",
       "      flex-wrap:wrap;\n",
       "      gap: 12px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert {\n",
       "      background-color: #E8F0FE;\n",
       "      border: none;\n",
       "      border-radius: 50%;\n",
       "      cursor: pointer;\n",
       "      display: none;\n",
       "      fill: #1967D2;\n",
       "      height: 32px;\n",
       "      padding: 0 0 0 0;\n",
       "      width: 32px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert:hover {\n",
       "      background-color: #E2EBFA;\n",
       "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
       "      fill: #174EA6;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert {\n",
       "      background-color: #3B4455;\n",
       "      fill: #D2E3FC;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert:hover {\n",
       "      background-color: #434B5C;\n",
       "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
       "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
       "      fill: #FFFFFF;\n",
       "    }\n",
       "  </style>\n",
       "\n",
       "      <script>\n",
       "        const buttonEl =\n",
       "          document.querySelector('#df-2becfe59-e90e-4976-9ad0-46d00a216670 button.colab-df-convert');\n",
       "        buttonEl.style.display =\n",
       "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
       "\n",
       "        async function convertToInteractive(key) {\n",
       "          const element = document.querySelector('#df-2becfe59-e90e-4976-9ad0-46d00a216670');\n",
       "          const dataTable =\n",
       "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
       "                                                     [key], {});\n",
       "          if (!dataTable) return;\n",
       "\n",
       "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
       "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
       "            + ' to learn more about interactive tables.';\n",
       "          element.innerHTML = '';\n",
       "          dataTable['output_type'] = 'display_data';\n",
       "          await google.colab.output.renderOutput(dataTable, element);\n",
       "          const docLink = document.createElement('div');\n",
       "          docLink.innerHTML = docLinkHtml;\n",
       "          element.appendChild(docLink);\n",
       "        }\n",
       "      </script>\n",
       "    </div>\n",
       "  </div>\n",
       "  "
      ],
      "text/plain": [
       "                 id                                       comment_text  toxic  \\\n",
       "0  0000997932d777bf  explanation edits made username hardcore metal...      0   \n",
       "1  000103f0d9cfb60f  daww match background colour im seemingly stuc...      0   \n",
       "2  000113f07ec002fd  hey man im really trying edit war guy constant...      0   \n",
       "3  0001b41b1c6bb37e  cant make real suggestion improvement wondered...      0   \n",
       "4  0001d958c54c6e35             sir hero chance remember page thats on      0   \n",
       "\n",
       "   severe_toxic  obscene  threat  insult  identity_hate  clean  \n",
       "0             0        0       0       0              0      1  \n",
       "1             0        0       0       0              0      1  \n",
       "2             0        0       0       0              0      1  \n",
       "3             0        0       0       0              0      1  \n",
       "4             0        0       0       0              0      1  "
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_df = preprocess_text(train_df)\n",
    "train_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 444
    },
    "id": "wvpVc2_zTbl2",
    "outputId": "80ad6d24-dacd-401a-a2a5-3790648505fb"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "<ipython-input-4-ecde23a7a0f1>:79: FutureWarning: The frame.append method is deprecated and will be removed from pandas in a future version. Use pandas.concat instead.\n",
      "  df = df.append(train_df[train_df['clean']==1].sample(frac=1, random_state=200).iloc[:df.shape[0],:])\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "\n",
       "  <div id=\"df-280a3298-5677-4d06-b241-c59489a20305\">\n",
       "    <div class=\"colab-df-container\">\n",
       "      <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>id</th>\n",
       "      <th>comment_text</th>\n",
       "      <th>toxic</th>\n",
       "      <th>severe_toxic</th>\n",
       "      <th>obscene</th>\n",
       "      <th>threat</th>\n",
       "      <th>insult</th>\n",
       "      <th>identity_hate</th>\n",
       "      <th>clean</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>134065</th>\n",
       "      <td>cd04ee4935924644</td>\n",
       "      <td>got ill start on</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>128040</th>\n",
       "      <td>acd929b3f6c5984b</td>\n",
       "      <td>barnstar anti vandalism barnstar professing lo...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>136471</th>\n",
       "      <td>da225210fb20b417</td>\n",
       "      <td>homey yo beat u up</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>112748</th>\n",
       "      <td>5b33d2be26a63953</td>\n",
       "      <td>encouraged feedback let clear bit first versio...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21589</th>\n",
       "      <td>38dd995c105311e6</td>\n",
       "      <td>shes insane zealot</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>\n",
       "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-280a3298-5677-4d06-b241-c59489a20305')\"\n",
       "              title=\"Convert this dataframe to an interactive table.\"\n",
       "              style=\"display:none;\">\n",
       "        \n",
       "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
       "       width=\"24px\">\n",
       "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
       "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
       "  </svg>\n",
       "      </button>\n",
       "      \n",
       "  <style>\n",
       "    .colab-df-container {\n",
       "      display:flex;\n",
       "      flex-wrap:wrap;\n",
       "      gap: 12px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert {\n",
       "      background-color: #E8F0FE;\n",
       "      border: none;\n",
       "      border-radius: 50%;\n",
       "      cursor: pointer;\n",
       "      display: none;\n",
       "      fill: #1967D2;\n",
       "      height: 32px;\n",
       "      padding: 0 0 0 0;\n",
       "      width: 32px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert:hover {\n",
       "      background-color: #E2EBFA;\n",
       "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
       "      fill: #174EA6;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert {\n",
       "      background-color: #3B4455;\n",
       "      fill: #D2E3FC;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert:hover {\n",
       "      background-color: #434B5C;\n",
       "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
       "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
       "      fill: #FFFFFF;\n",
       "    }\n",
       "  </style>\n",
       "\n",
       "      <script>\n",
       "        const buttonEl =\n",
       "          document.querySelector('#df-280a3298-5677-4d06-b241-c59489a20305 button.colab-df-convert');\n",
       "        buttonEl.style.display =\n",
       "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
       "\n",
       "        async function convertToInteractive(key) {\n",
       "          const element = document.querySelector('#df-280a3298-5677-4d06-b241-c59489a20305');\n",
       "          const dataTable =\n",
       "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
       "                                                     [key], {});\n",
       "          if (!dataTable) return;\n",
       "\n",
       "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
       "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
       "            + ' to learn more about interactive tables.';\n",
       "          element.innerHTML = '';\n",
       "          dataTable['output_type'] = 'display_data';\n",
       "          await google.colab.output.renderOutput(dataTable, element);\n",
       "          const docLink = document.createElement('div');\n",
       "          docLink.innerHTML = docLinkHtml;\n",
       "          element.appendChild(docLink);\n",
       "        }\n",
       "      </script>\n",
       "    </div>\n",
       "  </div>\n",
       "  "
      ],
      "text/plain": [
       "                      id                                       comment_text  \\\n",
       "134065  cd04ee4935924644                                   got ill start on   \n",
       "128040  acd929b3f6c5984b  barnstar anti vandalism barnstar professing lo...   \n",
       "136471  da225210fb20b417                                 homey yo beat u up   \n",
       "112748  5b33d2be26a63953  encouraged feedback let clear bit first versio...   \n",
       "21589   38dd995c105311e6                                 shes insane zealot   \n",
       "\n",
       "        toxic  severe_toxic  obscene  threat  insult  identity_hate  clean  \n",
       "134065      0             0        0       0       0              0      1  \n",
       "128040      0             0        0       0       0              0      1  \n",
       "136471      0             0        0       1       0              0      0  \n",
       "112748      0             0        0       0       0              0      1  \n",
       "21589       1             0        0       0       1              0      0  "
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Get sampled Data by removing Rows which create Bias\n",
    "train_df = sample_dataset(train_df)\n",
    "train_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 113,
     "referenced_widgets": [
      "1526fec20e314de0b608ed99a1e52bfc",
      "0c54c577b5a84e869ff08fa88e6a56d4",
      "4aa3e6c3b5a04342bda5fd9555abb341",
      "ab1cf86a4bbd490fb019944902b93a39",
      "81898db93e7a41e8ad52f6221f1ede10",
      "587ac2218ae942cb808e986eef900c27",
      "2a5ca9a345cc421588bd3b6b8b0051cd",
      "06b2188431f343d493036470c6dc4513",
      "30b73218a3c54f7caf19418394b7359b",
      "d54cb516fe6e4e9eb5842a4d2e24a113",
      "81f75aaefd384d879a4e6e0eb237c328",
      "4e9614212fac49bab0817aa3bf82c00e",
      "4bab7951efa942c69cf422c1e4744653",
      "1b68dabc2e5343b0adb4b2ab541f05b3",
      "a1022c8d33bf497f9505fee450b7b4d2",
      "f4b4e6cfe4574bfa87772778139502ae",
      "4fbb4aa295bd4950a80ddf130ab3d18f",
      "245bdbec3f984357adbd45ff79df6cdc",
      "8efc2adc3f674a9eb36dbf8dcb00df23",
      "12d58448ad6247bea10a071373312c81",
      "afcddcf35c314552b17a37bc0203f9ea",
      "17b2f163b596404e938f8d72000e9a86",
      "9d345ae3647d4ee9b4bd8a4cc0ed666c",
      "43a4866d6a8149eeac7e8d17315be499",
      "861aa32b2b254bafbfa1e0199831a02e",
      "82ac79a730ba4734a9919cfe7f4cc4fd",
      "0104af4daf4043f2a9a3b9c42b26d37d",
      "94604076a83b4fe9aadadec982606cb8",
      "0b49a0eca87c44ebbdfed7c2c0be0df5",
      "6be8209420ff47e9aaef3e8d0066ac60",
      "020578f9dde14e7ca5d80ae3108ef265",
      "2ca6e53967f74ad0912247bddac46cda",
      "e495e1aacd3342d1af26f54ae5bd1ea0"
     ]
    },
    "id": "-KVlh0KjZwO-",
    "outputId": "f5441956-d08e-4a18-f6c1-64afc49a8c0a"
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "1526fec20e314de0b608ed99a1e52bfc",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Downloading (…)solve/main/vocab.txt:   0%|          | 0.00/232k [00:00<?, ?B/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "4e9614212fac49bab0817aa3bf82c00e",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Downloading (…)okenizer_config.json:   0%|          | 0.00/28.0 [00:00<?, ?B/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "9d345ae3647d4ee9b4bd8a4cc0ed666c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Downloading (…)lve/main/config.json:   0%|          | 0.00/570 [00:00<?, ?B/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "device = torch.device('cuda')\n",
    "if torch.cuda.is_available():\n",
    "    device = torch.device('cuda:0')\n",
    "tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "Dg7s2vE-aEqc"
   },
   "outputs": [],
   "source": [
    "train_df, test_df = train_test_split(train_df, test_size=0.2,random_state=200,stratify=train_df.iloc[:,-1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "_L-AFlCma1sU"
   },
   "outputs": [],
   "source": [
    "class ToxicDataset(Dataset):\n",
    "    \n",
    "    def __init__(self, tokenizer: BertTokenizer, dataframe: pd.DataFrame, lazy: bool = False):\n",
    "        self.tokenizer = tokenizer\n",
    "        self.pad_idx = tokenizer.pad_token_id\n",
    "        self.lazy = lazy\n",
    "        if not self.lazy:\n",
    "            self.X = []\n",
    "            self.Y = []\n",
    "            for i, (row) in tqdm(dataframe.iterrows()):\n",
    "                x, y = self.row_to_tensor(self.tokenizer, row)\n",
    "                self.X.append(x)\n",
    "                self.Y.append(y)\n",
    "        else:\n",
    "            self.df = dataframe        \n",
    "    \n",
    "    @staticmethod\n",
    "    def row_to_tensor(tokenizer: BertTokenizer, row: pd.Series) -> Tuple[torch.LongTensor, torch.LongTensor]:\n",
    "        tokens = tokenizer.encode(row[\"comment_text\"], add_special_tokens=True)\n",
    "        if len(tokens) > 120:\n",
    "            tokens = tokens[:119] + [tokens[-1]]\n",
    "        x = torch.LongTensor(tokens)\n",
    "        y = torch.FloatTensor(row[[\"toxic\", \"severe_toxic\", \"obscene\", \"threat\", \"insult\", \"identity_hate\",\"clean\"]])\n",
    "        return x, y\n",
    "        \n",
    "    \n",
    "    def __len__(self):\n",
    "        if self.lazy:\n",
    "            return len(self.df)\n",
    "        else:\n",
    "            return len(self.X)\n",
    "\n",
    "    def __getitem__(self, index: int) -> Tuple[torch.LongTensor, torch.LongTensor]:\n",
    "        if not self.lazy:\n",
    "            return self.X[index], self.Y[index]\n",
    "        else:\n",
    "            return self.row_to_tensor(self.tokenizer, self.df.iloc[index])\n",
    "            \n",
    "\n",
    "def collate_fn(batch: List[Tuple[torch.LongTensor, torch.LongTensor]], device: torch.device) \\\n",
    "        -> Tuple[torch.LongTensor, torch.LongTensor]:\n",
    "    x, y = list(zip(*batch))\n",
    "    x = pad_sequence(x, batch_first=True, padding_value=0)\n",
    "    y = torch.stack(y)\n",
    "    return x.to(device), y.to(device)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "DNV5BeXoa9pE"
   },
   "outputs": [],
   "source": [
    "train_dataset = ToxicDataset(tokenizer, train_df, lazy=True)\n",
    "test_dataset = ToxicDataset(tokenizer, test_df, lazy=True)\n",
    "collate_fn = partial(collate_fn, device=device)\n",
    "BATCH_SIZE = 32\n",
    "train_sampler = RandomSampler(train_dataset)\n",
    "dev_sampler = RandomSampler(test_dataset)\n",
    "train_iterator = DataLoader(train_dataset, batch_size=BATCH_SIZE, sampler=train_sampler, collate_fn=collate_fn)\n",
    "dev_iterator = DataLoader(test_dataset, batch_size=BATCH_SIZE, sampler=dev_sampler, collate_fn=collate_fn)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "6CqGM0OZo3jI"
   },
   "outputs": [],
   "source": [
    "class BertClassifier(nn.Module):\n",
    "    \n",
    "    def __init__(self, bert: BertModel, num_classes: int):\n",
    "        super().__init__()\n",
    "        self.bert = bert\n",
    "        self.classifier = nn.Linear(bert.config.hidden_size, num_classes)\n",
    "        \n",
    "    def forward(self, input_ids, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None,\n",
    "                \n",
    "            labels=None):\n",
    "        outputs = self.bert(input_ids,\n",
    "                               attention_mask=attention_mask,\n",
    "                               token_type_ids=token_type_ids,\n",
    "                               position_ids=position_ids,\n",
    "                               head_mask=head_mask)\n",
    "        cls_output = outputs[1] # batch, hidden\n",
    "        cls_output = self.classifier(cls_output) # batch, 6\n",
    "        cls_output = torch.sigmoid(cls_output)\n",
    "        criterion = nn.BCELoss()\n",
    "        loss = 0\n",
    "        if labels is not None:\n",
    "            loss = criterion(cls_output, labels)\n",
    "        return loss, cls_output"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "t5JMXZ_5o_dG"
   },
   "outputs": [],
   "source": [
    "def train(model, iterator, optimizer, scheduler):\n",
    "    model.train()\n",
    "    total_loss = 0\n",
    "    for x, y in tqdm(iterator):\n",
    "        optimizer.zero_grad()\n",
    "        mask = (x != 0).float()\n",
    "        loss, outputs = model(x, attention_mask=mask, labels=y)\n",
    "        total_loss += loss.item()\n",
    "        loss.backward()\n",
    "        optimizer.step()\n",
    "        scheduler.step()\n",
    "    print(f\"Train loss {total_loss / len(iterator)}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "7SO_mtUGo-gc",
    "outputId": "d0e19f82-275b-4231-8bad-095e29ffcddc"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of the model checkpoint at bert-base-uncased were not used when initializing BertModel: ['cls.predictions.transform.LayerNorm.bias', 'cls.seq_relationship.bias', 'cls.predictions.decoder.weight', 'cls.predictions.transform.dense.weight', 'cls.predictions.bias', 'cls.seq_relationship.weight', 'cls.predictions.transform.LayerNorm.weight', 'cls.predictions.transform.dense.bias']\n",
      "- This IS expected if you are initializing BertModel from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n",
      "- This IS NOT expected if you are initializing BertModel from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n"
     ]
    }
   ],
   "source": [
    "model = BertClassifier(BertModel.from_pretrained('bert-base-uncased'), 7).to(device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "XZmgc0MzpYTM",
    "outputId": "7fc63518-c2cc-4d38-eed9-f70ca8cfc286"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.10/dist-packages/transformers/optimization.py:391: FutureWarning: This implementation of AdamW is deprecated and will be removed in a future version. Use the PyTorch implementation torch.optim.AdamW instead, or set `no_deprecation_warning=True` to disable this warning\n",
      "  warnings.warn(\n"
     ]
    }
   ],
   "source": [
    "no_decay = ['bias', 'LayerNorm.weight']\n",
    "optimizer_grouped_parameters = [\n",
    "{'params': [p for n, p in model.named_parameters() if not any(nd in n for nd in no_decay)], 'weight_decay': 0.01},\n",
    "{'params': [p for n, p in model.named_parameters() if any(nd in n for nd in no_decay)], 'weight_decay': 0.0}\n",
    "]\n",
    "EPOCH_NUM = 8\n",
    "# triangular learning rate, linearly grows untill half of first epoch, then linearly decays \n",
    "warmup_steps = 10 ** 3\n",
    "total_steps = len(train_iterator) * EPOCH_NUM - warmup_steps\n",
    "optimizer = AdamW(optimizer_grouped_parameters, lr=2e-5, eps=1e-8)\n",
    "scheduler = get_linear_schedule_with_warmup(optimizer, warmup_steps, total_steps)\n",
    "# scheduler = WarmupLinearSchedule(optimizer, warmup_steps=warmup_steps, t_total=total_steps)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "m_-ryCQ_pinl",
    "outputId": "7fecafc9-c384-487c-d184-f77b773046fe"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "================================================== EPOCH 0 ==================================================\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 812/812 [08:08<00:00,  1.66it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train loss 0.31736027307147757\n",
      "================================================== EPOCH 1 ==================================================\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 812/812 [08:06<00:00,  1.67it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train loss 0.16124800353177957\n",
      "================================================== EPOCH 2 ==================================================\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 812/812 [08:10<00:00,  1.66it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train loss 0.13076399981039646\n",
      "================================================== EPOCH 3 ==================================================\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 812/812 [08:08<00:00,  1.66it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train loss 0.10927103324495073\n",
      "================================================== EPOCH 4 ==================================================\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 812/812 [08:09<00:00,  1.66it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train loss 0.0929172182928202\n",
      "================================================== EPOCH 5 ==================================================\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 812/812 [08:07<00:00,  1.66it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train loss 0.08077572518156777\n",
      "================================================== EPOCH 6 ==================================================\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 812/812 [08:09<00:00,  1.66it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train loss 0.07278126178382727\n",
      "================================================== EPOCH 7 ==================================================\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 812/812 [08:08<00:00,  1.66it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train loss 0.07109613822809802\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "for i in range(EPOCH_NUM):\n",
    "    print('=' * 50, f\"EPOCH {i}\", '=' * 50)\n",
    "    train(model, train_iterator, optimizer, scheduler)\n",
    "    # evaluate(model, dev_iterator)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "n6wyMilfpupc",
    "outputId": "9b50bf3f-25c1-484e-8486-c487a50fe0af"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 203/203 [00:43<00:00,  4.72it/s]\n"
     ]
    }
   ],
   "source": [
    "model.eval()\n",
    "y_pred=[]\n",
    "columns = ['toxic', 'severe_toxic', 'obscene', 'threat', 'insult', 'identity_hate','clean']\n",
    "for i in tqdm(range(len(test_df) // BATCH_SIZE + 1)):\n",
    "    batch_df = test_df.iloc[i * BATCH_SIZE: (i + 1) * BATCH_SIZE]\n",
    "    texts = []\n",
    "    for text in batch_df[\"comment_text\"].tolist():\n",
    "        text = tokenizer.encode(text, add_special_tokens=True)\n",
    "        if len(text) > 120:\n",
    "            text = text[:119] + [tokenizer.sep_token_id]\n",
    "        texts.append(torch.LongTensor(text))\n",
    "    x = pad_sequence(texts, batch_first=True, padding_value=tokenizer.pad_token_id).to(device)\n",
    "    mask = (x != tokenizer.pad_token_id).float().to(device)\n",
    "    with torch.no_grad():\n",
    "        _, outputs = model(x, attention_mask=mask)\n",
    "    outputs = outputs.cpu().numpy()\n",
    "    y_pred.append(outputs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "PKO0VVTLy2rO",
    "outputId": "ecbe923b-f11b-4248-cfab-b8b54ba13947"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "<ipython-input-73-c5eb8d659cd7>:1: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray.\n",
      "  y_pred = np.vstack(np.asarray(y_pred))\n"
     ]
    }
   ],
   "source": [
    "y_pred = np.vstack(np.asarray(y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "r_2jFRV8v_w8"
   },
   "outputs": [],
   "source": [
    "y_pred = np.where(y_pred>0.5,1,0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "CPDvBS8gzQto"
   },
   "outputs": [],
   "source": [
    "y_test = test_df.iloc[:,2:].values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "xXH7sZDuzZws",
    "outputId": "56f7cea6-215b-4282-c897-7a28bca2d548"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'model_name': 'BERT', 'micro_avg_f1_score': 0.8586138229151489, 'hamming_loss': 0.06407660136473696, 'accuracy': 0.700462249614792}\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.94      0.88      0.91      3237\n",
      "           1       0.47      0.55      0.51       247\n",
      "           2       0.85      0.83      0.84      1751\n",
      "           3       0.48      0.67      0.56        76\n",
      "           4       0.76      0.74      0.75      1600\n",
      "           5       0.56      0.56      0.56       266\n",
      "           6       0.93      0.94      0.93      3191\n",
      "\n",
      "   micro avg       0.86      0.85      0.86     10368\n",
      "   macro avg       0.71      0.74      0.72     10368\n",
      "weighted avg       0.87      0.85      0.86     10368\n",
      " samples avg       0.88      0.86      0.85     10368\n",
      "\n"
     ]
    }
   ],
   "source": [
    "arr={'model_name':'BERT','micro_avg_f1_score':f1_score(y_test, y_pred, average=\"micro\"),'hamming_loss':hamming_loss(y_test, y_pred),'accuracy':accuracy_score(y_test,y_pred)}\n",
    "print(arr)\n",
    "print(classification_report(y_pred,y_test,zero_division=0))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "z2DH_zaC3OKh"
   },
   "source": [
    "# Second dataset\n",
    "\n",
    "For this dataset as well, we perform similar preprocessing which is shown below."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "644ACBOaxck4",
    "outputId": "4a780c49-4bef-471a-f3db-34f5539dca18"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[nltk_data] Downloading package reuters to /root/nltk_data...\n"
     ]
    }
   ],
   "source": [
    "# Load the dataset and extract the documents and labels\n",
    "import nltk\n",
    "\n",
    "nltk.download('reuters')\n",
    "docs = nltk.corpus.reuters.fileids()\n",
    "labels = [nltk.corpus.reuters.categories(doc) for doc in docs]\n",
    "texts = [nltk.corpus.reuters.raw(doc) for doc in docs]\n",
    "mlb = MultiLabelBinarizer()\n",
    "bin_labels = mlb.fit_transform(labels)\n",
    "label_names = mlb.classes_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "51h4DUYayAG_",
    "outputId": "8f89d55e-5941-40ed-9ad2-1ea9af9ff571"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                                                text  acq  alum  barley  bop  \\\n",
      "0  ASIAN EXPORTERS FEAR DAMAGE FROM U.S.-JAPAN RI...    0     0       0    0   \n",
      "1  CHINA DAILY SAYS VERMIN EAT 7-12 PCT GRAIN STO...    0     0       0    0   \n",
      "2  JAPAN TO REVISE LONG-TERM ENERGY DEMAND DOWNWA...    0     0       0    0   \n",
      "3  THAI TRADE DEFICIT WIDENS IN FIRST QUARTER\\n  ...    0     0       0    0   \n",
      "4  INDONESIA SEES CPO PRICE RISING SHARPLY\\n  Ind...    0     0       0    0   \n",
      "\n",
      "   carcass  castor-oil  cocoa  coconut  coconut-oil  ...  sun-oil  sunseed  \\\n",
      "0        0           0      0        0            0  ...        0        0   \n",
      "1        0           0      0        0            0  ...        0        0   \n",
      "2        0           0      0        0            0  ...        0        0   \n",
      "3        0           0      0        0            0  ...        0        0   \n",
      "4        0           0      0        0            0  ...        0        0   \n",
      "\n",
      "   tea  tin  trade  veg-oil  wheat  wpi  yen  zinc  \n",
      "0    0    0      1        0      0    0    0     0  \n",
      "1    0    0      0        0      0    0    0     0  \n",
      "2    0    0      0        0      0    0    0     0  \n",
      "3    0    1      1        0      0    0    0     0  \n",
      "4    0    0      0        1      0    0    0     0  \n",
      "\n",
      "[5 rows x 91 columns]\n"
     ]
    }
   ],
   "source": [
    "df = pd.DataFrame({'text': texts})\n",
    "for i, label in enumerate(label_names):\n",
    "    df[label] = bin_labels[:, i]\n",
    "\n",
    "print(df.head())\n",
    "labels = df.columns[1:]\n",
    "X_train = df['text']\n",
    "# X_test = test_df['comment_text']\n",
    "y_train = df.iloc[:,1:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 647
    },
    "id": "56K4PoDwyOAn",
    "outputId": "46f95432-c45c-42e2-d653-c796b1098a85"
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "\n",
       "  <div id=\"df-0ced64c1-17ab-4ad7-a423-3876a20b55d8\">\n",
       "    <div class=\"colab-df-container\">\n",
       "      <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>text</th>\n",
       "      <th>acq</th>\n",
       "      <th>alum</th>\n",
       "      <th>barley</th>\n",
       "      <th>bop</th>\n",
       "      <th>carcass</th>\n",
       "      <th>castor-oil</th>\n",
       "      <th>cocoa</th>\n",
       "      <th>coconut</th>\n",
       "      <th>coconut-oil</th>\n",
       "      <th>...</th>\n",
       "      <th>sun-oil</th>\n",
       "      <th>sunseed</th>\n",
       "      <th>tea</th>\n",
       "      <th>tin</th>\n",
       "      <th>trade</th>\n",
       "      <th>veg-oil</th>\n",
       "      <th>wheat</th>\n",
       "      <th>wpi</th>\n",
       "      <th>yen</th>\n",
       "      <th>zinc</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>asian exporter fear damage u japan rift mounti...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>china daily say vermin eat pct grain stock sur...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>japan revise long term energy demand downwards...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>thai trade deficit widens first quarter thaila...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>indonesia see cpo price rising sharply indones...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 91 columns</p>\n",
       "</div>\n",
       "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-0ced64c1-17ab-4ad7-a423-3876a20b55d8')\"\n",
       "              title=\"Convert this dataframe to an interactive table.\"\n",
       "              style=\"display:none;\">\n",
       "        \n",
       "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
       "       width=\"24px\">\n",
       "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
       "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
       "  </svg>\n",
       "      </button>\n",
       "      \n",
       "  <style>\n",
       "    .colab-df-container {\n",
       "      display:flex;\n",
       "      flex-wrap:wrap;\n",
       "      gap: 12px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert {\n",
       "      background-color: #E8F0FE;\n",
       "      border: none;\n",
       "      border-radius: 50%;\n",
       "      cursor: pointer;\n",
       "      display: none;\n",
       "      fill: #1967D2;\n",
       "      height: 32px;\n",
       "      padding: 0 0 0 0;\n",
       "      width: 32px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert:hover {\n",
       "      background-color: #E2EBFA;\n",
       "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
       "      fill: #174EA6;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert {\n",
       "      background-color: #3B4455;\n",
       "      fill: #D2E3FC;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert:hover {\n",
       "      background-color: #434B5C;\n",
       "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
       "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
       "      fill: #FFFFFF;\n",
       "    }\n",
       "  </style>\n",
       "\n",
       "      <script>\n",
       "        const buttonEl =\n",
       "          document.querySelector('#df-0ced64c1-17ab-4ad7-a423-3876a20b55d8 button.colab-df-convert');\n",
       "        buttonEl.style.display =\n",
       "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
       "\n",
       "        async function convertToInteractive(key) {\n",
       "          const element = document.querySelector('#df-0ced64c1-17ab-4ad7-a423-3876a20b55d8');\n",
       "          const dataTable =\n",
       "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
       "                                                     [key], {});\n",
       "          if (!dataTable) return;\n",
       "\n",
       "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
       "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
       "            + ' to learn more about interactive tables.';\n",
       "          element.innerHTML = '';\n",
       "          dataTable['output_type'] = 'display_data';\n",
       "          await google.colab.output.renderOutput(dataTable, element);\n",
       "          const docLink = document.createElement('div');\n",
       "          docLink.innerHTML = docLinkHtml;\n",
       "          element.appendChild(docLink);\n",
       "        }\n",
       "      </script>\n",
       "    </div>\n",
       "  </div>\n",
       "  "
      ],
      "text/plain": [
       "                                                text  acq  alum  barley  bop  \\\n",
       "0  asian exporter fear damage u japan rift mounti...    0     0       0    0   \n",
       "1  china daily say vermin eat pct grain stock sur...    0     0       0    0   \n",
       "2  japan revise long term energy demand downwards...    0     0       0    0   \n",
       "3  thai trade deficit widens first quarter thaila...    0     0       0    0   \n",
       "4  indonesia see cpo price rising sharply indones...    0     0       0    0   \n",
       "\n",
       "   carcass  castor-oil  cocoa  coconut  coconut-oil  ...  sun-oil  sunseed  \\\n",
       "0        0           0      0        0            0  ...        0        0   \n",
       "1        0           0      0        0            0  ...        0        0   \n",
       "2        0           0      0        0            0  ...        0        0   \n",
       "3        0           0      0        0            0  ...        0        0   \n",
       "4        0           0      0        0            0  ...        0        0   \n",
       "\n",
       "   tea  tin  trade  veg-oil  wheat  wpi  yen  zinc  \n",
       "0    0    0      1        0      0    0    0     0  \n",
       "1    0    0      0        0      0    0    0     0  \n",
       "2    0    0      0        0      0    0    0     0  \n",
       "3    0    1      1        0      0    0    0     0  \n",
       "4    0    0      0        1      0    0    0     0  \n",
       "\n",
       "[5 rows x 91 columns]"
      ]
     },
     "execution_count": 135,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df['text'] = df['text'].str.lower()\n",
    "df['text'] = df['text'].apply(cleanHtml)\n",
    "df['text'] = df['text'].apply(cleanPunc)\n",
    "df['text'] = df['text'].apply(keepAlpha)\n",
    "df['text'] = df['text'].apply(removeStopWords)\n",
    "df['text'] = df['text'].apply(lemmatize)\n",
    "\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 699
    },
    "id": "fYhPDFiEyhCr",
    "outputId": "beb0036f-d8d2-4e75-c05f-fa21da90e68a"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1500x900 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# filter the DataFrame to include only numeric columns\n",
    "numeric_df = df.select_dtypes(include=['int', 'float'])\n",
    "# count the number of 1's in each column\n",
    "counts = numeric_df.sum()\n",
    "# create a bar chart of the counts with a larger figure size\n",
    "fig = plt.figure(figsize=(15, 9))\n",
    "counts.plot(kind='bar')\n",
    "# add chart labels\n",
    "plt.title('Counts of 1s in Columns')\n",
    "plt.xlabel('Columns')\n",
    "plt.ylabel('Count')\n",
    "# display the chart\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Nx6dqUBy3VUt"
   },
   "source": [
    "Preparing training data for classical ML algorithms"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "J67Du-6tyrq0",
    "outputId": "848f6f95-6a62-45c1-ecb8-99c4f22a683b"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(10788,) (10788, 90)\n",
      "(8630,) (2158,) (8630, 90) (2158, 90)\n",
      "(8630, 3000) (2158, 3000)\n"
     ]
    }
   ],
   "source": [
    "X = df['text'].values\n",
    "y = df.iloc[:,1:].values\n",
    "print(X.shape, y.shape)\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2,random_state=200)\n",
    "print(X_train.shape, X_test.shape, y_train.shape, y_test.shape )\n",
    "#ML preprocessing\n",
    "\n",
    "word_vectorizer = TfidfVectorizer(\n",
    "    strip_accents='unicode',     \n",
    "    analyzer='word',\n",
    "    max_features=3000,        \n",
    "    token_pattern=r'\\w{1,}',    \n",
    "    ngram_range=(1, 3),         \n",
    "    stop_words='english',\n",
    "    sublinear_tf=True)\n",
    "\n",
    "word_vectorizer.fit(X_train)\n",
    "ml_X_train = word_vectorizer.transform(X_train)\n",
    "ml_X_test = word_vectorizer.transform(X_test)\n",
    "\n",
    "print(ml_X_train.shape, ml_X_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "CgCX45bR0q9Q"
   },
   "outputs": [],
   "source": [
    "scores_df = pd.DataFrame(columns=['model_name','micro_avg_f1_score','hamming_loss','accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "WBfSBe8vgSB-",
    "outputId": "3040a3e1-ccf2-4f8a-c97c-949f10bbf844"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.94      0.98      0.96       463\n",
      "           1       0.00      0.00      0.00         0\n",
      "           2       0.00      0.00      0.00         0\n",
      "           3       0.40      1.00      0.57         8\n",
      "           4       0.00      0.00      0.00         0\n",
      "           5       0.00      0.00      0.00         0\n",
      "           6       0.20      1.00      0.33         2\n",
      "           7       0.00      0.00      0.00         0\n",
      "           8       0.00      0.00      0.00         0\n",
      "           9       0.50      1.00      0.67        15\n",
      "          10       0.08      1.00      0.15         1\n",
      "          11       0.00      0.00      0.00         0\n",
      "          12       0.42      1.00      0.59        22\n",
      "          13       0.11      1.00      0.20         1\n",
      "          14       0.00      0.00      0.00         0\n",
      "          15       0.47      1.00      0.64         7\n",
      "          16       0.00      0.00      0.00         0\n",
      "          17       0.73      0.90      0.81       105\n",
      "          18       0.00      0.00      0.00         0\n",
      "          19       0.32      0.82      0.46        11\n",
      "          20       0.00      0.00      0.00         0\n",
      "          21       0.95      0.99      0.97       756\n",
      "          22       0.00      0.00      0.00         0\n",
      "          23       0.00      0.00      0.00         0\n",
      "          24       0.36      1.00      0.53         9\n",
      "          25       0.45      0.93      0.61        15\n",
      "          26       0.73      0.98      0.83        84\n",
      "          27       0.00      0.00      0.00         0\n",
      "          28       0.00      0.00      0.00         0\n",
      "          29       0.00      0.00      0.00         0\n",
      "          30       0.00      0.00      0.00         0\n",
      "          31       0.00      0.00      0.00         0\n",
      "          32       0.00      0.00      0.00         0\n",
      "          33       0.00      0.00      0.00         0\n",
      "          34       0.60      0.78      0.68        60\n",
      "          35       0.00      0.00      0.00         0\n",
      "          36       0.00      0.00      0.00         0\n",
      "          37       0.00      0.00      0.00         0\n",
      "          38       0.08      1.00      0.15         1\n",
      "          39       0.00      0.00      0.00         0\n",
      "          40       0.00      0.00      0.00         0\n",
      "          41       0.00      0.00      0.00         0\n",
      "          42       0.00      0.00      0.00         0\n",
      "          43       0.06      0.50      0.10         2\n",
      "          44       0.00      0.00      0.00         0\n",
      "          45       0.00      0.00      0.00         0\n",
      "          46       0.70      0.92      0.80       112\n",
      "          47       0.47      1.00      0.64        15\n",
      "          48       0.00      0.00      0.00         0\n",
      "          49       0.27      0.89      0.41         9\n",
      "          50       0.00      0.00      0.00         0\n",
      "          51       0.00      0.00      0.00         0\n",
      "          52       0.00      0.00      0.00         0\n",
      "          53       0.00      0.00      0.00         0\n",
      "          54       0.23      0.78      0.35         9\n",
      "          55       0.00      0.00      0.00         0\n",
      "          56       0.00      0.00      0.00         0\n",
      "          57       0.00      0.00      0.00         0\n",
      "          58       0.00      0.00      0.00         0\n",
      "          59       0.00      0.00      0.00         0\n",
      "          60       0.00      0.00      0.00         0\n",
      "          61       0.00      0.00      0.00         0\n",
      "          62       0.00      0.00      0.00         0\n",
      "          63       0.00      0.00      0.00         0\n",
      "          64       0.00      0.00      0.00         0\n",
      "          65       0.00      0.00      0.00         0\n",
      "          66       0.00      0.00      0.00         0\n",
      "          67       0.00      0.00      0.00         0\n",
      "          68       0.00      0.00      0.00         0\n",
      "          69       0.00      0.00      0.00         0\n",
      "          70       0.00      0.00      0.00         0\n",
      "          71       0.55      0.97      0.70        34\n",
      "          72       0.00      0.00      0.00         0\n",
      "          73       0.00      0.00      0.00         0\n",
      "          74       0.00      0.00      0.00         0\n",
      "          75       0.00      0.00      0.00         0\n",
      "          76       0.25      1.00      0.40         5\n",
      "          77       0.00      0.00      0.00         0\n",
      "          78       0.54      1.00      0.70        13\n",
      "          79       0.00      0.00      0.00         0\n",
      "          80       0.00      0.00      0.00         0\n",
      "          81       0.00      0.00      0.00         0\n",
      "          82       0.00      0.00      0.00         0\n",
      "          83       0.00      0.00      0.00         0\n",
      "          84       0.67      0.89      0.77        75\n",
      "          85       0.05      1.00      0.09         1\n",
      "          86       0.59      1.00      0.74        33\n",
      "          87       0.00      0.00      0.00         0\n",
      "          88       0.00      0.00      0.00         0\n",
      "          89       0.00      0.00      0.00         0\n",
      "\n",
      "   micro avg       0.68      0.96      0.80      1868\n",
      "   macro avg       0.13      0.28      0.17      1868\n",
      "weighted avg       0.83      0.96      0.88      1868\n",
      " samples avg       0.75      0.78      0.76      1868\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "<ipython-input-142-ba6a06d0e02b>:5: FutureWarning: The frame.append method is deprecated and will be removed from pandas in a future version. Use pandas.concat instead.\n",
      "  scores_df = scores_df.append(arr,ignore_index=True)\n"
     ]
    }
   ],
   "source": [
    "model = BinaryRelevance(LogisticRegression(solver='sag'))\n",
    "model.fit(ml_X_train, y_train)\n",
    "y_pred = model.predict(ml_X_test)\n",
    "arr={'model_name':'BR','micro_avg_f1_score':f1_score(y_test, y_pred, average=\"micro\"),'hamming_loss':hamming_loss(y_test, y_pred),'accuracy':accuracy_score(y_test,y_pred)}\n",
    "scores_df = scores_df.append(arr,ignore_index=True)\n",
    "print(classification_report(y_pred,y_test, zero_division=0))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "8wwaIgmqchB-",
    "outputId": "5bee3061-aa64-44a1-bc6b-45362bbd774c"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.94      0.98      0.96       463\n",
      "           1       0.00      0.00      0.00         0\n",
      "           2       0.00      0.00      0.00         0\n",
      "           3       0.40      1.00      0.57         8\n",
      "           4       0.00      0.00      0.00         0\n",
      "           5       0.00      0.00      0.00         0\n",
      "           6       0.20      1.00      0.33         2\n",
      "           7       0.00      0.00      0.00         0\n",
      "           8       0.00      0.00      0.00         0\n",
      "           9       0.50      1.00      0.67        15\n",
      "          10       0.08      1.00      0.15         1\n",
      "          11       0.00      0.00      0.00         0\n",
      "          12       0.35      1.00      0.51        18\n",
      "          13       0.11      1.00      0.20         1\n",
      "          14       0.00      0.00      0.00         0\n",
      "          15       0.47      1.00      0.64         7\n",
      "          16       0.00      0.00      0.00         0\n",
      "          17       0.72      0.91      0.81       103\n",
      "          18       0.00      0.00      0.00         0\n",
      "          19       0.32      0.82      0.46        11\n",
      "          20       0.00      0.00      0.00         0\n",
      "          21       0.97      0.98      0.98       781\n",
      "          22       0.00      0.00      0.00         0\n",
      "          23       0.00      0.00      0.00         0\n",
      "          24       0.32      1.00      0.48         8\n",
      "          25       0.65      1.00      0.78        20\n",
      "          26       0.61      1.00      0.76        69\n",
      "          27       0.00      0.00      0.00         0\n",
      "          28       0.00      0.00      0.00         0\n",
      "          29       0.00      0.00      0.00         0\n",
      "          30       0.00      0.00      0.00         0\n",
      "          31       0.00      0.00      0.00         0\n",
      "          32       0.00      0.00      0.00         0\n",
      "          33       0.00      0.00      0.00         0\n",
      "          34       0.76      0.80      0.78        74\n",
      "          35       0.00      0.00      0.00         0\n",
      "          36       0.00      0.00      0.00         0\n",
      "          37       0.00      0.00      0.00         0\n",
      "          38       0.17      1.00      0.29         2\n",
      "          39       0.00      0.00      0.00         0\n",
      "          40       0.00      0.00      0.00         0\n",
      "          41       0.00      0.00      0.00         0\n",
      "          42       0.00      0.00      0.00         0\n",
      "          43       0.00      0.00      0.00         0\n",
      "          44       0.00      0.00      0.00         0\n",
      "          45       0.00      0.00      0.00         0\n",
      "          46       0.73      0.93      0.82       116\n",
      "          47       0.59      1.00      0.75        19\n",
      "          48       0.00      0.00      0.00         0\n",
      "          49       0.30      0.90      0.45        10\n",
      "          50       0.00      0.00      0.00         0\n",
      "          51       0.00      0.00      0.00         0\n",
      "          52       0.00      0.00      0.00         0\n",
      "          53       0.00      0.00      0.00         0\n",
      "          54       0.23      0.78      0.35         9\n",
      "          55       0.00      0.00      0.00         0\n",
      "          56       0.00      0.00      0.00         0\n",
      "          57       0.00      0.00      0.00         0\n",
      "          58       0.00      0.00      0.00         0\n",
      "          59       0.00      0.00      0.00         0\n",
      "          60       0.00      0.00      0.00         0\n",
      "          61       0.00      0.00      0.00         0\n",
      "          62       0.00      0.00      0.00         0\n",
      "          63       0.00      0.00      0.00         0\n",
      "          64       0.00      0.00      0.00         0\n",
      "          65       0.00      0.00      0.00         0\n",
      "          66       0.08      1.00      0.15         1\n",
      "          67       0.00      0.00      0.00         0\n",
      "          68       0.00      0.00      0.00         0\n",
      "          69       0.10      1.00      0.18         1\n",
      "          70       0.00      0.00      0.00         0\n",
      "          71       0.62      0.97      0.76        38\n",
      "          72       0.00      0.00      0.00         0\n",
      "          73       0.00      0.00      0.00         0\n",
      "          74       0.00      0.00      0.00         0\n",
      "          75       0.00      0.00      0.00         0\n",
      "          76       0.30      0.75      0.43         8\n",
      "          77       0.00      0.00      0.00         0\n",
      "          78       0.58      1.00      0.74        14\n",
      "          79       0.00      0.00      0.00         0\n",
      "          80       0.00      0.00      0.00         0\n",
      "          81       0.00      0.00      0.00         0\n",
      "          82       0.00      0.00      0.00         0\n",
      "          83       0.00      0.00      0.00         0\n",
      "          84       0.83      0.85      0.84        98\n",
      "          85       0.05      1.00      0.09         1\n",
      "          86       0.68      0.93      0.78        41\n",
      "          87       0.00      0.00      0.00         0\n",
      "          88       0.00      0.00      0.00         0\n",
      "          89       0.00      0.00      0.00         0\n",
      "\n",
      "   micro avg       0.70      0.96      0.81      1939\n",
      "   macro avg       0.14      0.30      0.17      1939\n",
      "weighted avg       0.85      0.96      0.89      1939\n",
      " samples avg       0.77      0.80      0.78      1939\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "<ipython-input-143-0040459b4f27>:12: FutureWarning: The frame.append method is deprecated and will be removed from pandas in a future version. Use pandas.concat instead.\n",
      "  scores_df = scores_df.append(arr,ignore_index=True)\n"
     ]
    }
   ],
   "source": [
    "import warnings\n",
    "from sklearn.exceptions import ConvergenceWarning\n",
    "\n",
    "# Set the warning filter to ignore ConvergenceWarning\n",
    "warnings.filterwarnings('ignore', category=ConvergenceWarning)\n",
    "\n",
    "model = ClassifierChain(LogisticRegression(solver='sag', max_iter=1000))\n",
    "model.fit(ml_X_train, y_train)\n",
    "y_pred = model.predict(ml_X_test)\n",
    "arr={'model_name':'Classifier_Chain','micro_avg_f1_score':f1_score(y_test, y_pred, average=\"micro\"),'hamming_loss':hamming_loss(y_test, y_pred),'accuracy':accuracy_score(y_test,y_pred)}\n",
    "\n",
    "scores_df = scores_df.append(arr,ignore_index=True)\n",
    "print(classification_report(y_pred,y_test,zero_division=0))\n",
    "\n",
    "# Reset the warning filter to its default behavior\n",
    "warnings.filterwarnings('default', category=ConvergenceWarning)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "7VcNDScCh997",
    "outputId": "95f5d6b5-d0be-44ee-b4fc-878a0cc3949b"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.98      0.83      0.90       564\n",
      "           1       0.38      0.83      0.53         6\n",
      "           2       0.00      0.00      0.00         0\n",
      "           3       0.60      0.92      0.73        13\n",
      "           4       0.33      0.67      0.44         3\n",
      "           5       0.00      0.00      0.00         0\n",
      "           6       0.80      1.00      0.89         8\n",
      "           7       0.00      0.00      0.00         0\n",
      "           8       0.00      0.00      0.00         0\n",
      "           9       0.80      1.00      0.89        24\n",
      "          10       0.58      1.00      0.74         7\n",
      "          11       0.00      0.00      0.00         0\n",
      "          12       0.48      1.00      0.65        25\n",
      "          13       0.56      1.00      0.71         5\n",
      "          14       0.00      0.00      0.00         0\n",
      "          15       0.80      0.71      0.75        17\n",
      "          16       0.00      0.00      0.00         0\n",
      "          17       0.82      0.80      0.81       133\n",
      "          18       0.00      0.00      0.00         0\n",
      "          19       0.39      0.79      0.52        14\n",
      "          20       0.00      0.00      0.00         0\n",
      "          21       0.98      0.94      0.96       821\n",
      "          22       0.00      0.00      0.00         0\n",
      "          23       0.00      0.00      0.00         0\n",
      "          24       0.64      1.00      0.78        16\n",
      "          25       0.81      1.00      0.89        25\n",
      "          26       0.73      0.95      0.83        87\n",
      "          27       0.00      0.00      0.00         0\n",
      "          28       0.00      0.00      0.00         0\n",
      "          29       0.00      0.00      0.00         0\n",
      "          30       0.50      1.00      0.67         1\n",
      "          31       0.50      1.00      0.67         1\n",
      "          32       0.00      0.00      0.00         0\n",
      "          33       0.00      0.00      0.00         0\n",
      "          34       0.82      0.79      0.81        81\n",
      "          35       0.64      1.00      0.78         7\n",
      "          36       0.30      1.00      0.46         3\n",
      "          37       0.00      0.00      0.00         0\n",
      "          38       0.75      1.00      0.86         9\n",
      "          39       0.00      0.00      0.00         0\n",
      "          40       0.00      0.00      0.00         0\n",
      "          41       1.00      1.00      1.00         2\n",
      "          42       0.00      0.00      0.00         0\n",
      "          43       0.28      0.83      0.42         6\n",
      "          44       0.00      0.00      0.00         0\n",
      "          45       0.00      0.00      0.00         0\n",
      "          46       0.80      0.89      0.84       132\n",
      "          47       0.84      0.93      0.89        29\n",
      "          48       0.00      0.00      0.00         0\n",
      "          49       0.30      0.90      0.45        10\n",
      "          50       0.00      0.00      0.00         0\n",
      "          51       0.00      0.00      0.00         0\n",
      "          52       0.00      0.00      0.00         0\n",
      "          53       0.00      0.00      0.00         0\n",
      "          54       0.23      0.88      0.36         8\n",
      "          55       0.67      1.00      0.80         2\n",
      "          56       0.00      0.00      0.00         0\n",
      "          57       0.60      1.00      0.75         3\n",
      "          58       0.00      0.00      0.00         0\n",
      "          59       0.00      0.00      0.00         0\n",
      "          60       0.00      0.00      0.00         0\n",
      "          61       0.00      0.00      0.00         0\n",
      "          62       0.00      0.00      0.00         0\n",
      "          63       0.00      0.00      0.00         0\n",
      "          64       0.00      0.00      0.00         0\n",
      "          65       0.60      1.00      0.75         3\n",
      "          66       0.42      1.00      0.59         5\n",
      "          67       0.00      0.00      0.00         0\n",
      "          68       0.00      0.00      0.00         0\n",
      "          69       0.60      1.00      0.75         6\n",
      "          70       0.00      0.00      0.00         0\n",
      "          71       0.73      0.92      0.81        48\n",
      "          72       0.00      0.00      0.00         0\n",
      "          73       0.00      0.00      0.00         0\n",
      "          74       0.00      0.00      0.00         0\n",
      "          75       0.00      0.00      0.00         0\n",
      "          76       0.25      1.00      0.40         5\n",
      "          77       0.00      0.00      0.00         0\n",
      "          78       0.79      0.90      0.84        21\n",
      "          79       0.00      0.00      0.00         0\n",
      "          80       0.00      0.00      0.00         0\n",
      "          81       0.00      0.00      0.00         0\n",
      "          82       0.00      0.00      0.00         0\n",
      "          83       0.38      1.00      0.55         3\n",
      "          84       0.90      0.73      0.80       124\n",
      "          85       0.24      1.00      0.38         5\n",
      "          86       0.80      0.83      0.82        54\n",
      "          87       0.33      1.00      0.50         2\n",
      "          88       0.00      0.00      0.00         0\n",
      "          89       0.00      0.00      0.00         0\n",
      "\n",
      "   micro avg       0.78      0.88      0.83      2338\n",
      "   macro avg       0.28      0.42      0.32      2338\n",
      "weighted avg       0.89      0.88      0.88      2338\n",
      " samples avg       0.86      0.89      0.86      2338\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "<ipython-input-144-a377907869e2>:5: FutureWarning: The frame.append method is deprecated and will be removed from pandas in a future version. Use pandas.concat instead.\n",
      "  scores_df = scores_df.append(arr,ignore_index=True)\n"
     ]
    }
   ],
   "source": [
    "model = LabelPowerset(LogisticRegression(max_iter=1000))\n",
    "model.fit(ml_X_train, y_train)\n",
    "y_pred = model.predict(ml_X_test)\n",
    "arr={'model_name':'Label_Powerset Chain','micro_avg_f1_score':f1_score(y_test, y_pred, average=\"micro\"),'hamming_loss':hamming_loss(y_test, y_pred),'accuracy':accuracy_score(y_test,y_pred)}\n",
    "scores_df = scores_df.append(arr,ignore_index=True)\n",
    "print(classification_report(y_pred,y_test,zero_division=0))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 143
    },
    "id": "TpKejYqvkDqR",
    "outputId": "54f6b862-def9-45b8-dbf0-6c0466e7e2e9"
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "\n",
       "  <div id=\"df-428c6f42-3205-4a13-b353-3b0002a3e4be\">\n",
       "    <div class=\"colab-df-container\">\n",
       "      <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>model_name</th>\n",
       "      <th>micro_avg_f1_score</th>\n",
       "      <th>hamming_loss</th>\n",
       "      <th>accuracy</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>BR</td>\n",
       "      <td>0.800178</td>\n",
       "      <td>0.004634</td>\n",
       "      <td>0.709917</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Classifier_Chain</td>\n",
       "      <td>0.812240</td>\n",
       "      <td>0.004423</td>\n",
       "      <td>0.736330</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Label_Powerset Chain</td>\n",
       "      <td>0.831524</td>\n",
       "      <td>0.004315</td>\n",
       "      <td>0.812790</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>\n",
       "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-428c6f42-3205-4a13-b353-3b0002a3e4be')\"\n",
       "              title=\"Convert this dataframe to an interactive table.\"\n",
       "              style=\"display:none;\">\n",
       "        \n",
       "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
       "       width=\"24px\">\n",
       "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
       "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
       "  </svg>\n",
       "      </button>\n",
       "      \n",
       "  <style>\n",
       "    .colab-df-container {\n",
       "      display:flex;\n",
       "      flex-wrap:wrap;\n",
       "      gap: 12px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert {\n",
       "      background-color: #E8F0FE;\n",
       "      border: none;\n",
       "      border-radius: 50%;\n",
       "      cursor: pointer;\n",
       "      display: none;\n",
       "      fill: #1967D2;\n",
       "      height: 32px;\n",
       "      padding: 0 0 0 0;\n",
       "      width: 32px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert:hover {\n",
       "      background-color: #E2EBFA;\n",
       "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
       "      fill: #174EA6;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert {\n",
       "      background-color: #3B4455;\n",
       "      fill: #D2E3FC;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert:hover {\n",
       "      background-color: #434B5C;\n",
       "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
       "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
       "      fill: #FFFFFF;\n",
       "    }\n",
       "  </style>\n",
       "\n",
       "      <script>\n",
       "        const buttonEl =\n",
       "          document.querySelector('#df-428c6f42-3205-4a13-b353-3b0002a3e4be button.colab-df-convert');\n",
       "        buttonEl.style.display =\n",
       "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
       "\n",
       "        async function convertToInteractive(key) {\n",
       "          const element = document.querySelector('#df-428c6f42-3205-4a13-b353-3b0002a3e4be');\n",
       "          const dataTable =\n",
       "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
       "                                                     [key], {});\n",
       "          if (!dataTable) return;\n",
       "\n",
       "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
       "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
       "            + ' to learn more about interactive tables.';\n",
       "          element.innerHTML = '';\n",
       "          dataTable['output_type'] = 'display_data';\n",
       "          await google.colab.output.renderOutput(dataTable, element);\n",
       "          const docLink = document.createElement('div');\n",
       "          docLink.innerHTML = docLinkHtml;\n",
       "          element.appendChild(docLink);\n",
       "        }\n",
       "      </script>\n",
       "    </div>\n",
       "  </div>\n",
       "  "
      ],
      "text/plain": [
       "             model_name  micro_avg_f1_score  hamming_loss  accuracy\n",
       "0                    BR            0.800178      0.004634  0.709917\n",
       "1      Classifier_Chain            0.812240      0.004423  0.736330\n",
       "2  Label_Powerset Chain            0.831524      0.004315  0.812790"
      ]
     },
     "execution_count": 145,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "scores_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "YUGGG11WDaFC",
    "outputId": "95ac47e1-4759-41ac-954b-465fc527481e"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "torch.Size([8630, 70]) torch.Size([2158, 70]) torch.Size([8630, 90]) torch.Size([2158, 90])\n",
      "24610\n"
     ]
    }
   ],
   "source": [
    "# from sklearn.model_selection import train_test_split\n",
    "# X_train, X_val, y_train, y_val = train_test_split(X_train, y_train, test_size=0.2,random_state=200)\n",
    "#DL preprocessing\n",
    "\n",
    "tokenizer = Tokenizer(num_words=20000,oov_token='<UNK>')\n",
    "tokenizer.fit_on_texts(X_train)\n",
    "sequences_text_train = tokenizer.texts_to_sequences(X_train)\n",
    "sequences_text_test = tokenizer.texts_to_sequences(X_test)\n",
    "\n",
    "X_train = torch.from_numpy(pad_sequences(sequences_text_train, maxlen=70))\n",
    "X_test = torch.from_numpy(pad_sequences(sequences_text_test, maxlen=70))\n",
    "\n",
    "y_train = torch.from_numpy(y_train)\n",
    "y_test = torch.from_numpy(y_test)\n",
    "\n",
    "print(X_train.shape, X_test.shape, y_train.shape, y_test.shape )\n",
    "print(len(tokenizer.index_word))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "bS-eUxLq2QOC",
    "outputId": "273cfd2a-5cb7-49a3-9a9c-a05ff4593a85"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "VOCAB_SIZE : 24611\n",
      "TOTAL OF UNKNOWN WORD : 4653\n"
     ]
    }
   ],
   "source": [
    "#Load glove embedding\n",
    "glove_embeddings = {}\n",
    "with open('glove.6B/glove.6B.300d.txt', 'r', encoding='utf-8') as f:\n",
    "    for line in f:\n",
    "        parts = line.split()\n",
    "        word = parts[0]\n",
    "        vector = np.array(parts[1:], dtype=np.float32)\n",
    "        glove_embeddings[word] = vector\n",
    "\n",
    "VOCAB_SIZE = len(tokenizer.index_word)+1\n",
    "glove_embedding_matrix = torch.zeros(VOCAB_SIZE, 300)\n",
    "\n",
    "unk = 0\n",
    "for i in range(1, VOCAB_SIZE):\n",
    "  word = tokenizer.index_word[i]\n",
    "  if word in glove_embeddings.keys():\n",
    "    glove_embedding_matrix[i] = torch.from_numpy(glove_embeddings[word]).float()\n",
    "  else:\n",
    "    unk +=1\n",
    "print('VOCAB_SIZE : {}'.format(VOCAB_SIZE))\n",
    "print('TOTAL OF UNKNOWN WORD : {}'.format(unk))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "DTxHZkoOm_Ch"
   },
   "outputs": [],
   "source": [
    "class MLP_Glove(nn.Module):\n",
    "    def __init__(self, input_dim, hidden_dim, output_dim):\n",
    "        super(MLP_Glove, self).__init__()\n",
    "        self.embedding = nn.Embedding.from_pretrained(glove_embedding_matrix)\n",
    "        self.fc1 = nn.Linear(300, hidden_dim)\n",
    "        self.fc2 = nn.Linear(hidden_dim, output_dim)\n",
    "        self.relu = nn.ReLU()\n",
    "        self.sigmoid = nn.Sigmoid()\n",
    "\n",
    "    def forward(self, x):\n",
    "        embedding = self.embedding(x)\n",
    "        out = torch.mean(embedding,dim=1)\n",
    "        out = self.relu(self.fc1(out))\n",
    "        out = self.sigmoid(self.fc2(out))\n",
    "        return out"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "-tyXsFGpnDxZ"
   },
   "outputs": [],
   "source": [
    "from torch.utils.data import DataLoader, TensorDataset\n",
    "train_dataset = TensorDataset(X_train, y_train)\n",
    "train_dataloader = DataLoader(train_dataset, batch_size=256, shuffle=True)\n",
    "device = torch.device('cuda')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "3k8wzRIxnI3m",
    "outputId": "b260c1eb-ae3b-4d9c-ca58-7cb340b915f9"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch [1/100], Loss: 0.6476\n",
      "Epoch [2/100], Loss: 0.6304\n",
      "Epoch [3/100], Loss: 0.6286\n",
      "Epoch [4/100], Loss: 0.6283\n",
      "Epoch [5/100], Loss: 0.6273\n",
      "Epoch [6/100], Loss: 0.6265\n",
      "Epoch [7/100], Loss: 0.6261\n",
      "Epoch [8/100], Loss: 0.6242\n",
      "Epoch [9/100], Loss: 0.6235\n",
      "Epoch [10/100], Loss: 0.6208\n",
      "Epoch [11/100], Loss: 0.6173\n",
      "Epoch [12/100], Loss: 0.6150\n",
      "Epoch [13/100], Loss: 0.6133\n",
      "Epoch [14/100], Loss: 0.6124\n",
      "Epoch [15/100], Loss: 0.6116\n",
      "Epoch [16/100], Loss: 0.6097\n",
      "Epoch [17/100], Loss: 0.6087\n",
      "Epoch [18/100], Loss: 0.6059\n",
      "Epoch [19/100], Loss: 0.6042\n",
      "Epoch [20/100], Loss: 0.6024\n",
      "Epoch [21/100], Loss: 0.6006\n",
      "Epoch [22/100], Loss: 0.6002\n",
      "Epoch [23/100], Loss: 0.5996\n",
      "Epoch [24/100], Loss: 0.5995\n",
      "Epoch [25/100], Loss: 0.5990\n",
      "Epoch [26/100], Loss: 0.5986\n",
      "Epoch [27/100], Loss: 0.5983\n",
      "Epoch [28/100], Loss: 0.5984\n",
      "Epoch [29/100], Loss: 0.5979\n",
      "Epoch [30/100], Loss: 0.5977\n",
      "Epoch [31/100], Loss: 0.5977\n",
      "Epoch [32/100], Loss: 0.5970\n",
      "Epoch [33/100], Loss: 0.5975\n",
      "Epoch [34/100], Loss: 0.5970\n",
      "Epoch [35/100], Loss: 0.5960\n",
      "Epoch [36/100], Loss: 0.5962\n",
      "Epoch [37/100], Loss: 0.5961\n",
      "Epoch [38/100], Loss: 0.5959\n",
      "Epoch [39/100], Loss: 0.5957\n",
      "Epoch [40/100], Loss: 0.5956\n",
      "Epoch [41/100], Loss: 0.5960\n",
      "Epoch [42/100], Loss: 0.5960\n",
      "Epoch [43/100], Loss: 0.5951\n",
      "Epoch [44/100], Loss: 0.5958\n",
      "Epoch [45/100], Loss: 0.5958\n",
      "Epoch [46/100], Loss: 0.5952\n",
      "Epoch [47/100], Loss: 0.5947\n",
      "Epoch [48/100], Loss: 0.5950\n",
      "Epoch [49/100], Loss: 0.5947\n",
      "Epoch [50/100], Loss: 0.5954\n",
      "Epoch [51/100], Loss: 0.5949\n",
      "Epoch [52/100], Loss: 0.5944\n",
      "Epoch [53/100], Loss: 0.5950\n",
      "Epoch [54/100], Loss: 0.5945\n",
      "Epoch [55/100], Loss: 0.5946\n",
      "Epoch [56/100], Loss: 0.5944\n",
      "Epoch [57/100], Loss: 0.5943\n",
      "Epoch [58/100], Loss: 0.5944\n",
      "Epoch [59/100], Loss: 0.5943\n",
      "Epoch [60/100], Loss: 0.5938\n",
      "Epoch [61/100], Loss: 0.5939\n",
      "Epoch [62/100], Loss: 0.5940\n",
      "Epoch [63/100], Loss: 0.5941\n",
      "Epoch [64/100], Loss: 0.5938\n",
      "Epoch [65/100], Loss: 0.5939\n",
      "Epoch [66/100], Loss: 0.5942\n",
      "Epoch [67/100], Loss: 0.5938\n",
      "Epoch [68/100], Loss: 0.5938\n",
      "Epoch [69/100], Loss: 0.5943\n",
      "Epoch [70/100], Loss: 0.5945\n",
      "Epoch [71/100], Loss: 0.5938\n",
      "Epoch [72/100], Loss: 0.5940\n",
      "Epoch [73/100], Loss: 0.5934\n",
      "Epoch [74/100], Loss: 0.5934\n",
      "Epoch [75/100], Loss: 0.5941\n",
      "Epoch [76/100], Loss: 0.5933\n",
      "Epoch [77/100], Loss: 0.5938\n",
      "Epoch [78/100], Loss: 0.5934\n",
      "Epoch [79/100], Loss: 0.5932\n",
      "Epoch [80/100], Loss: 0.5935\n",
      "Epoch [81/100], Loss: 0.5942\n",
      "Epoch [82/100], Loss: 0.5940\n",
      "Epoch [83/100], Loss: 0.5937\n",
      "Epoch [84/100], Loss: 0.5939\n",
      "Epoch [85/100], Loss: 0.5937\n",
      "Epoch [86/100], Loss: 0.5936\n",
      "Epoch [87/100], Loss: 0.5938\n",
      "Epoch [88/100], Loss: 0.5934\n",
      "Epoch [89/100], Loss: 0.5933\n",
      "Epoch [90/100], Loss: 0.5933\n",
      "Epoch [91/100], Loss: 0.5933\n",
      "Epoch [92/100], Loss: 0.5938\n",
      "Epoch [93/100], Loss: 0.5933\n",
      "Epoch [94/100], Loss: 0.5932\n",
      "Epoch [95/100], Loss: 0.5938\n",
      "Epoch [96/100], Loss: 0.5935\n",
      "Epoch [97/100], Loss: 0.5935\n",
      "Epoch [98/100], Loss: 0.5937\n",
      "Epoch [99/100], Loss: 0.5936\n",
      "Epoch [100/100], Loss: 0.5934\n"
     ]
    }
   ],
   "source": [
    "model = MLP_Glove(70, 300, 90).to(device)\n",
    "\n",
    "# Define loss function and optimizer\n",
    "criterion = nn.CrossEntropyLoss()\n",
    "optimizer = torch.optim.Adam(model.parameters(), lr=0.01)\n",
    "\n",
    "# Generate dummy input and target data\n",
    "\n",
    "num_epochs = 100\n",
    "for epoch in range(num_epochs):\n",
    "    total_loss=0\n",
    "    for i,(inputs, targets) in enumerate(train_dataloader):\n",
    "        # Forward pass\n",
    "        # inputs = inputs.to(next(model.parameters()).dtype)\n",
    "        # targets = targets.to(model.device)\n",
    "        output = model(inputs.to(device))\n",
    "        # print(output)\n",
    "        targets = targets.float().to(device)\n",
    "        \n",
    "        # Compute loss\n",
    "        loss = criterion(output, targets).to(device)\n",
    "        total_loss+=loss.item()\n",
    "        \n",
    "        # Backward pass and update weights\n",
    "        optimizer.zero_grad()\n",
    "        loss.backward()\n",
    "        optimizer.step()\n",
    "        \n",
    "    # Print loss every 10 epochs\n",
    "    print('Epoch [{}/{}], Loss: {:.4f}'.format(epoch+1, num_epochs, total_loss/256))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "GlKG0CcJ3L5t"
   },
   "outputs": [],
   "source": [
    "test_dataset = TensorDataset(X_test, y_test)\n",
    "test_dataloader = DataLoader(test_dataset, batch_size=256, shuffle=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "4ItfK2Q53NkP"
   },
   "outputs": [],
   "source": [
    "y_pred = []\n",
    "for i,(inputs, targets) in enumerate(test_dataloader):\n",
    "      # Forward pass\n",
    "      inputs = inputs.to(device)\n",
    "      targets = targets.to(device)\n",
    "      output = model(inputs).detach()\n",
    "      for j in output:\n",
    "        y_pred.append(j.cpu().numpy())\n",
    "\n",
    "# y_test.numpy()\n",
    "y_pred = np.vstack(y_pred)\n",
    "y_pred = np.where(y_pred>0.5,1,0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "hXiFIE3K3Pqc",
    "outputId": "8d563a5b-c458-4253-bc30-0a2384e6a579"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.27      0.22      0.25       580\n",
      "           1       0.00      0.00      0.00        12\n",
      "           2       0.00      0.00      0.00         9\n",
      "           3       0.00      0.00      0.00        52\n",
      "           4       0.00      0.00      0.00        19\n",
      "           5       0.00      0.00      0.00         0\n",
      "           6       0.00      0.00      0.00         9\n",
      "           7       0.00      0.00      0.00         0\n",
      "           8       0.00      0.00      0.00         0\n",
      "           9       0.00      0.00      0.00        26\n",
      "          10       0.00      0.00      0.00        41\n",
      "          11       0.00      0.00      0.00         0\n",
      "          12       0.00      0.00      0.00        84\n",
      "          13       0.00      0.00      0.00         0\n",
      "          14       0.00      0.00      0.00         0\n",
      "          15       0.00      0.00      0.00        30\n",
      "          16       0.00      0.00      0.00         0\n",
      "          17       0.08      0.06      0.07       176\n",
      "          18       0.00      0.00      0.00         0\n",
      "          19       0.00      0.00      0.00        71\n",
      "          20       0.00      0.00      0.00         0\n",
      "          21       0.41      0.37      0.39       864\n",
      "          22       0.00      0.00      0.00         0\n",
      "          23       0.09      0.03      0.05        29\n",
      "          24       0.04      0.03      0.03        36\n",
      "          25       0.06      0.06      0.06        35\n",
      "          26       0.06      0.05      0.05       142\n",
      "          27       0.00      0.00      0.00         0\n",
      "          28       0.00      0.00      0.00         0\n",
      "          29       0.00      0.00      0.00         0\n",
      "          30       0.00      0.00      0.00         2\n",
      "          31       0.00      0.00      0.00         0\n",
      "          32       0.00      0.00      0.00         2\n",
      "          33       0.00      0.00      0.00         0\n",
      "          34       0.03      0.01      0.02       146\n",
      "          35       0.00      0.00      0.00        20\n",
      "          36       0.00      0.00      0.00         0\n",
      "          37       0.00      0.00      0.00         0\n",
      "          38       0.00      0.00      0.00        15\n",
      "          39       0.00      0.00      0.00         0\n",
      "          40       0.00      0.00      0.00         0\n",
      "          41       0.00      0.00      0.00         0\n",
      "          42       0.00      0.00      0.00         0\n",
      "          43       0.00      0.00      0.00        22\n",
      "          44       0.00      0.00      0.00         0\n",
      "          45       0.00      0.00      0.00         0\n",
      "          46       0.14      0.08      0.10       259\n",
      "          47       0.00      0.00      0.00        42\n",
      "          48       0.00      0.00      0.00         0\n",
      "          49       0.00      0.00      0.00         0\n",
      "          50       0.00      0.00      0.00         0\n",
      "          51       0.00      0.00      0.00         0\n",
      "          52       0.00      0.00      0.00         0\n",
      "          53       0.00      0.00      0.00         3\n",
      "          54       0.00      0.00      0.00        59\n",
      "          55       0.00      0.00      0.00         0\n",
      "          56       0.00      0.00      0.00         0\n",
      "          57       0.00      0.00      0.00         0\n",
      "          58       0.00      0.00      0.00         0\n",
      "          59       0.00      0.00      0.00         0\n",
      "          60       0.00      0.00      0.00         1\n",
      "          61       0.00      0.00      0.00         0\n",
      "          62       0.00      0.00      0.00         0\n",
      "          63       0.00      0.00      0.00         0\n",
      "          64       0.00      0.00      0.00         0\n",
      "          65       0.00      0.00      0.00         6\n",
      "          66       0.08      0.03      0.04        33\n",
      "          67       0.00      0.00      0.00         0\n",
      "          68       0.00      0.00      0.00        11\n",
      "          69       0.00      0.00      0.00         0\n",
      "          70       0.00      0.00      0.00         0\n",
      "          71       0.08      0.07      0.07        76\n",
      "          72       0.00      0.00      0.00        26\n",
      "          73       0.00      0.00      0.00         0\n",
      "          74       0.00      0.00      0.00         0\n",
      "          75       0.00      0.00      0.00         0\n",
      "          76       0.00      0.00      0.00        28\n",
      "          77       0.00      0.00      0.00         0\n",
      "          78       0.00      0.00      0.00        21\n",
      "          79       0.00      0.00      0.00         0\n",
      "          80       0.00      0.00      0.00         0\n",
      "          81       0.00      0.00      0.00         0\n",
      "          82       0.00      0.00      0.00         0\n",
      "          83       0.00      0.00      0.00         8\n",
      "          84       0.10      0.06      0.08       161\n",
      "          85       0.00      0.00      0.00        25\n",
      "          86       0.04      0.02      0.02       107\n",
      "          87       0.17      0.08      0.11        13\n",
      "          88       0.08      0.04      0.05        25\n",
      "          89       0.00      0.00      0.00        10\n",
      "\n",
      "   micro avg       0.20      0.15      0.17      3336\n",
      "   macro avg       0.02      0.01      0.02      3336\n",
      "weighted avg       0.18      0.15      0.17      3336\n",
      " samples avg       0.23      0.20      0.21      3336\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "<ipython-input-153-e48a89fd9cc8>:2: FutureWarning: The frame.append method is deprecated and will be removed from pandas in a future version. Use pandas.concat instead.\n",
      "  scores_df = scores_df.append(arr,ignore_index=True)\n"
     ]
    }
   ],
   "source": [
    "arr={'model_name':'MLP_Glove','micro_avg_f1_score':f1_score(y_test, y_pred, average=\"micro\"),'hamming_loss':hamming_loss(y_test, y_pred),'accuracy':accuracy_score(y_test,y_pred)}\n",
    "scores_df = scores_df.append(arr,ignore_index=True)\n",
    "print(classification_report(y_pred,y_test,zero_division=0))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "2SIEWQu23RqW"
   },
   "outputs": [],
   "source": [
    "import keras\n",
    "import tensorflow as tf\n",
    "from keras.preprocessing.text import Tokenizer\n",
    "from keras.layers import Dense, Input, LSTM, Embedding, Dropout, Activation\n",
    "from keras.layers import Bidirectional, GlobalMaxPool1D\n",
    "from keras.models import Model\n",
    "from keras import initializers, regularizers, constraints, optimizers, layers"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "NTjPKRf_3TjF",
    "outputId": "6698d930-e566-404a-dccf-0306a151fc36"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Device mapping:\n",
      "/job:localhost/replica:0/task:0/device:GPU:0 -> device: 0, name: Tesla T4, pci bus id: 0000:00:04.0, compute capability: 7.5\n",
      "\n"
     ]
    }
   ],
   "source": [
    "sess = tf.compat.v1.Session(config=tf.compat.v1.ConfigProto(log_device_placement=True))\n",
    "keras.backend.set_session(sess)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "ndoxWgdg3VnR"
   },
   "outputs": [],
   "source": [
    "embd_mat= np.array(glove_embedding_matrix.to('cpu'))\n",
    "embd_mat = np.vstack((np.zeros((1, embd_mat.shape[1])), embd_mat))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "eXk1fXM63k3l"
   },
   "source": [
    "### Running deep learning algorithms\n",
    "\n",
    "The algorithms used are - \n",
    "* BiLSTM using glove embeddings\n",
    "* BiLSTM using BERT embeddings\n",
    "* CNN using BERT embeddings\n",
    "* CNN usin BERT embeddings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "Wr55BLn49Olk",
    "outputId": "7ca468ec-e313-40fa-9577-00f5e40cf403"
   },
   "outputs": [
    {
     "metadata": {
      "tags": null
     },
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Layer lstm_2 will not use cuDNN kernels since it doesn't meet the criteria. It will use a generic GPU kernel as fallback when running on GPU.\n",
      "WARNING:tensorflow:Layer lstm_2 will not use cuDNN kernels since it doesn't meet the criteria. It will use a generic GPU kernel as fallback when running on GPU.\n",
      "WARNING:tensorflow:Layer lstm_2 will not use cuDNN kernels since it doesn't meet the criteria. It will use a generic GPU kernel as fallback when running on GPU.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "=====================Glove BiLSTM===========================\n",
      "Epoch 1/40\n",
      "34/34 [==============================] - 34s 781ms/step - loss: 0.4786 - accuracy: 0.0787\n",
      "Epoch 2/40\n",
      "34/34 [==============================] - 27s 783ms/step - loss: 0.0861 - accuracy: 0.2504\n",
      "Epoch 3/40\n",
      "34/34 [==============================] - 27s 796ms/step - loss: 0.0553 - accuracy: 0.3308\n",
      "Epoch 4/40\n",
      "34/34 [==============================] - 26s 762ms/step - loss: 0.0532 - accuracy: 0.3468\n",
      "Epoch 5/40\n",
      "34/34 [==============================] - 26s 759ms/step - loss: 0.0513 - accuracy: 0.3762\n",
      "Epoch 6/40\n",
      "34/34 [==============================] - 24s 716ms/step - loss: 0.0492 - accuracy: 0.4168\n",
      "Epoch 7/40\n",
      "34/34 [==============================] - 24s 707ms/step - loss: 0.0467 - accuracy: 0.4590\n",
      "Epoch 8/40\n",
      "34/34 [==============================] - 23s 690ms/step - loss: 0.0438 - accuracy: 0.5023\n",
      "Epoch 9/40\n",
      "34/34 [==============================] - 24s 690ms/step - loss: 0.0407 - accuracy: 0.5378\n",
      "Epoch 10/40\n",
      "34/34 [==============================] - 24s 681ms/step - loss: 0.0384 - accuracy: 0.5600\n",
      "Epoch 11/40\n",
      "34/34 [==============================] - 24s 714ms/step - loss: 0.0360 - accuracy: 0.5851\n",
      "Epoch 12/40\n",
      "34/34 [==============================] - 23s 693ms/step - loss: 0.0343 - accuracy: 0.6028\n",
      "Epoch 13/40\n",
      "34/34 [==============================] - 24s 698ms/step - loss: 0.0330 - accuracy: 0.6130\n",
      "Epoch 14/40\n",
      "34/34 [==============================] - 23s 672ms/step - loss: 0.0318 - accuracy: 0.6280\n",
      "Epoch 15/40\n",
      "34/34 [==============================] - 23s 678ms/step - loss: 0.0302 - accuracy: 0.6451\n",
      "Epoch 16/40\n",
      "34/34 [==============================] - 22s 628ms/step - loss: 0.0290 - accuracy: 0.6540\n",
      "Epoch 17/40\n",
      "34/34 [==============================] - 23s 694ms/step - loss: 0.0277 - accuracy: 0.6676\n",
      "Epoch 18/40\n",
      "34/34 [==============================] - 23s 685ms/step - loss: 0.0264 - accuracy: 0.6848\n",
      "Epoch 19/40\n",
      "34/34 [==============================] - 23s 692ms/step - loss: 0.0251 - accuracy: 0.6950\n",
      "Epoch 20/40\n",
      "34/34 [==============================] - 23s 669ms/step - loss: 0.0242 - accuracy: 0.7065\n",
      "Epoch 21/40\n",
      "34/34 [==============================] - 22s 626ms/step - loss: 0.0232 - accuracy: 0.7163\n",
      "Epoch 22/40\n",
      "34/34 [==============================] - 23s 673ms/step - loss: 0.0220 - accuracy: 0.7263\n",
      "Epoch 23/40\n",
      "34/34 [==============================] - 23s 686ms/step - loss: 0.0211 - accuracy: 0.7349\n",
      "Epoch 24/40\n",
      "34/34 [==============================] - 24s 711ms/step - loss: 0.0202 - accuracy: 0.7437\n",
      "Epoch 25/40\n",
      "34/34 [==============================] - 22s 655ms/step - loss: 0.0195 - accuracy: 0.7489\n",
      "Epoch 26/40\n",
      "34/34 [==============================] - 22s 633ms/step - loss: 0.0186 - accuracy: 0.7601\n",
      "Epoch 27/40\n",
      "34/34 [==============================] - 23s 687ms/step - loss: 0.0178 - accuracy: 0.7711\n",
      "Epoch 28/40\n",
      "34/34 [==============================] - 23s 676ms/step - loss: 0.0170 - accuracy: 0.7825\n",
      "Epoch 29/40\n",
      "34/34 [==============================] - 23s 680ms/step - loss: 0.0163 - accuracy: 0.7848\n",
      "Epoch 30/40\n",
      "34/34 [==============================] - 22s 636ms/step - loss: 0.0157 - accuracy: 0.7928\n",
      "Epoch 31/40\n",
      "34/34 [==============================] - 23s 664ms/step - loss: 0.0148 - accuracy: 0.7976\n",
      "Epoch 32/40\n",
      "34/34 [==============================] - 23s 678ms/step - loss: 0.0143 - accuracy: 0.8042\n",
      "Epoch 33/40\n",
      "34/34 [==============================] - 23s 673ms/step - loss: 0.0136 - accuracy: 0.8118\n",
      "Epoch 34/40\n",
      "34/34 [==============================] - 22s 652ms/step - loss: 0.0129 - accuracy: 0.8196\n",
      "Epoch 35/40\n",
      "34/34 [==============================] - 23s 657ms/step - loss: 0.0124 - accuracy: 0.8225\n",
      "Epoch 36/40\n",
      "34/34 [==============================] - 23s 686ms/step - loss: 0.0120 - accuracy: 0.8262\n",
      "Epoch 37/40\n",
      "34/34 [==============================] - 23s 685ms/step - loss: 0.0115 - accuracy: 0.8309\n",
      "Epoch 38/40\n",
      "34/34 [==============================] - 23s 679ms/step - loss: 0.0109 - accuracy: 0.8349\n",
      "Epoch 39/40\n",
      "34/34 [==============================] - 21s 633ms/step - loss: 0.0105 - accuracy: 0.8390\n",
      "Epoch 40/40\n",
      "34/34 [==============================] - 22s 656ms/step - loss: 0.0101 - accuracy: 0.8433\n",
      "9/9 [==============================] - 1s 41ms/step\n",
      "             model_name  micro_avg_f1_score  hamming_loss  accuracy\n",
      "0                    BR            0.800178      0.004634  0.709917\n",
      "1      Classifier_Chain            0.812240      0.004423  0.736330\n",
      "2  Label_Powerset Chain            0.831524      0.004315  0.812790\n",
      "3             MLP_Glove            0.172472      0.025445  0.159870\n",
      "4          BiLSTM_Glove            0.676641      0.006899  0.568582\n",
      "5          BiLSTM_glove            0.784573      0.005205  0.698795\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.91      0.92      0.92       476\n",
      "           1       0.00      0.00      0.00         1\n",
      "           2       0.00      0.00      0.00         0\n",
      "           3       0.55      0.85      0.67        13\n",
      "           4       0.00      0.00      0.00         0\n",
      "           5       0.00      0.00      0.00         0\n",
      "           6       0.60      1.00      0.75         6\n",
      "           7       0.00      0.00      0.00         0\n",
      "           8       0.00      0.00      0.00         0\n",
      "           9       0.33      0.71      0.45        14\n",
      "          10       0.08      1.00      0.15         1\n",
      "          11       0.00      0.00      0.00         0\n",
      "          12       0.50      0.81      0.62        32\n",
      "          13       0.00      0.00      0.00         0\n",
      "          14       0.00      0.00      0.00         0\n",
      "          15       0.27      0.80      0.40         5\n",
      "          16       0.00      0.00      0.00         0\n",
      "          17       0.75      0.81      0.78       120\n",
      "          18       0.00      0.00      0.00         0\n",
      "          19       0.50      0.67      0.57        21\n",
      "          20       0.00      0.00      0.00         0\n",
      "          21       0.95      0.97      0.96       774\n",
      "          22       0.00      0.00      0.00         0\n",
      "          23       0.00      0.00      0.00         0\n",
      "          24       0.12      0.50      0.19         6\n",
      "          25       0.55      0.89      0.68        19\n",
      "          26       0.85      0.84      0.85       114\n",
      "          27       0.00      0.00      0.00         0\n",
      "          28       0.00      0.00      0.00         0\n",
      "          29       0.00      0.00      0.00         0\n",
      "          30       0.00      0.00      0.00         0\n",
      "          31       0.00      0.00      0.00         0\n",
      "          32       0.00      0.00      0.00         0\n",
      "          33       0.00      0.00      0.00         0\n",
      "          34       0.72      0.76      0.74        74\n",
      "          35       0.00      0.00      0.00         0\n",
      "          36       0.00      0.00      0.00         0\n",
      "          37       0.00      0.00      0.00         0\n",
      "          38       0.00      0.00      0.00         0\n",
      "          39       0.00      0.00      0.00         0\n",
      "          40       0.14      1.00      0.25         1\n",
      "          41       0.00      0.00      0.00         0\n",
      "          42       0.00      0.00      0.00         0\n",
      "          43       0.11      0.50      0.18         4\n",
      "          44       0.00      0.00      0.00         0\n",
      "          45       0.00      0.00      0.00         0\n",
      "          46       0.77      0.86      0.81       132\n",
      "          47       0.47      0.88      0.61        17\n",
      "          48       0.00      0.00      0.00         0\n",
      "          49       0.00      0.00      0.00         0\n",
      "          50       0.00      0.00      0.00         0\n",
      "          51       0.00      0.00      0.00         0\n",
      "          52       0.00      0.00      0.00         0\n",
      "          53       0.00      0.00      0.00         0\n",
      "          54       0.55      0.81      0.65        21\n",
      "          55       0.00      0.00      0.00         0\n",
      "          56       0.00      0.00      0.00         0\n",
      "          57       0.20      1.00      0.33         1\n",
      "          58       0.00      0.00      0.00         0\n",
      "          59       0.00      0.00      0.00         0\n",
      "          60       0.00      0.00      0.00         0\n",
      "          61       0.00      0.00      0.00         0\n",
      "          62       0.00      0.00      0.00         0\n",
      "          63       0.00      0.00      0.00         0\n",
      "          64       0.00      0.00      0.00         0\n",
      "          65       0.00      0.00      0.00         0\n",
      "          66       0.00      0.00      0.00         0\n",
      "          67       0.00      0.00      0.00         0\n",
      "          68       0.00      0.00      0.00         0\n",
      "          69       0.00      0.00      0.00         0\n",
      "          70       0.00      0.00      0.00         0\n",
      "          71       0.55      0.94      0.69        35\n",
      "          72       0.00      0.00      0.00         0\n",
      "          73       0.00      0.00      0.00         0\n",
      "          74       0.00      0.00      0.00         0\n",
      "          75       0.00      0.00      0.00         0\n",
      "          76       0.35      0.64      0.45        11\n",
      "          77       0.00      0.00      0.00         0\n",
      "          78       0.50      1.00      0.67        12\n",
      "          79       0.00      0.00      0.00         0\n",
      "          80       0.00      0.00      0.00         0\n",
      "          81       0.00      0.00      0.00         0\n",
      "          82       0.00      0.00      0.00         0\n",
      "          83       0.00      0.00      0.00         0\n",
      "          84       0.64      0.75      0.69        85\n",
      "          85       0.33      0.64      0.44        11\n",
      "          86       0.71      0.82      0.76        49\n",
      "          87       0.00      0.00      0.00         0\n",
      "          88       0.00      0.00      0.00         0\n",
      "          89       0.14      0.50      0.22         2\n",
      "\n",
      "   micro avg       0.70      0.89      0.78      2057\n",
      "   macro avg       0.15      0.24      0.17      2057\n",
      "weighted avg       0.83      0.89      0.86      2057\n",
      " samples avg       0.75      0.78      0.76      2057\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "<ipython-input-163-d570d8fba6df>:17: FutureWarning: The frame.append method is deprecated and will be removed from pandas in a future version. Use pandas.concat instead.\n",
      "  scores_df = scores_df.append(arr,ignore_index=True)\n",
      "/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classification.py:1344: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n",
      "/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classification.py:1344: UndefinedMetricWarning: Recall and F-score are ill-defined and being set to 0.0 in labels with no true samples. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n",
      "/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classification.py:1344: UndefinedMetricWarning: Recall and F-score are ill-defined and being set to 0.0 in samples with no true labels. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n"
     ]
    }
   ],
   "source": [
    "print(\"=====================Glove BiLSTM===========================\")\n",
    "inp = Input(shape=(70,))\n",
    "x = Embedding(VOCAB_SIZE+1, 300, weights=[embd_mat])(inp)\n",
    "x = Bidirectional(LSTM(50, return_sequences=True, dropout=0.1, recurrent_dropout=0.1))(x)\n",
    "x = GlobalMaxPool1D()(x)\n",
    "x = Dense(50, activation=\"relu\")(x)\n",
    "x = Dropout(0.1)(x)\n",
    "x = Dense(90, activation=\"sigmoid\")(x)\n",
    "model = Model(inputs=inp, outputs=x)\n",
    "model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
    "\n",
    "model.fit(np.asarray(X_train),np.asarray(y_train), batch_size=256, epochs=40)\n",
    "y_pred = model.predict(np.asarray(X_test),batch_size=256, verbose=1)\n",
    "y_pred= np.where(y_pred > 0.5, 1, 0)\n",
    "\n",
    "arr={'model_name':'BiLSTM_glove','micro_avg_f1_score':f1_score(y_test, y_pred, average=\"micro\"),'hamming_loss':hamming_loss(y_test, y_pred),'accuracy':accuracy_score(y_test,y_pred)}\n",
    "scores_df = scores_df.append(arr,ignore_index=True)\n",
    "print(scores_df)\n",
    "print(classification_report(y_pred,y_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "CLXwtzSX8PIo"
   },
   "outputs": [],
   "source": [
    "print(\"=====================Bert BiLSTM===========================\")\n",
    "inp = Input(shape=(70,))\n",
    "x = Embedding(VOCAB_SIZE, 768, weights=[bert_embd_mat])(inp)\n",
    "x = Bidirectional(LSTM(50, return_sequences=True, dropout=0.1, recurrent_dropout=0.1))(x)\n",
    "x = GlobalMaxPool1D()(x)\n",
    "x = Dense(50, activation=\"relu\")(x)\n",
    "x = Dropout(0.1)(x)\n",
    "x = Dense(7, activation=\"sigmoid\")(x)\n",
    "model = Model(inputs=inp, outputs=x)\n",
    "model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
    "\n",
    "model.fit(np.asarray(X_train),np.asarray(y_train), batch_size=256, epochs=4)\n",
    "y_pred = model.predict(np.asarray(X_test),batch_size=256, verbose=1)\n",
    "y_pred= np.where(y_pred > 0.5, 1, 0)\n",
    "\n",
    "arr={'model_name':'BiLSTM_bert','micro_avg_f1_score':f1_score(y_test, y_pred, average=\"micro\"),'hamming_loss':hamming_loss(y_test, y_pred),'accuracy':accuracy_score(y_test,y_pred)}\n",
    "scores_df = scores_df.append(arr,ignore_index=True)\n",
    "print(scores_df)\n",
    "print(classification_report(y_pred,y_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "iRrhm-fu8tcu",
    "outputId": "ecac48ab-1936-4d6b-8efc-7f519abb1bf1"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "=====================Glove CNN===========================\n",
      "Epoch 1/50\n",
      "34/34 [==============================] - 14s 290ms/step - loss: 0.4044 - accuracy: 0.0090\n",
      "Epoch 2/50\n",
      "34/34 [==============================] - 6s 172ms/step - loss: 0.0594 - accuracy: 0.3910\n",
      "Epoch 3/50\n",
      "34/34 [==============================] - 7s 193ms/step - loss: 0.0409 - accuracy: 0.5372\n",
      "Epoch 4/50\n",
      "34/34 [==============================] - 5s 157ms/step - loss: 0.0360 - accuracy: 0.5767\n",
      "Epoch 5/50\n",
      "34/34 [==============================] - 4s 122ms/step - loss: 0.0322 - accuracy: 0.6249\n",
      "Epoch 6/50\n",
      "34/34 [==============================] - 5s 133ms/step - loss: 0.0289 - accuracy: 0.6592\n",
      "Epoch 7/50\n",
      "34/34 [==============================] - 5s 133ms/step - loss: 0.0255 - accuracy: 0.6903\n",
      "Epoch 8/50\n",
      "34/34 [==============================] - 3s 102ms/step - loss: 0.0224 - accuracy: 0.7229\n",
      "Epoch 9/50\n",
      "34/34 [==============================] - 3s 89ms/step - loss: 0.0200 - accuracy: 0.7502\n",
      "Epoch 10/50\n",
      "34/34 [==============================] - 2s 71ms/step - loss: 0.0177 - accuracy: 0.7698\n",
      "Epoch 11/50\n",
      "34/34 [==============================] - 3s 74ms/step - loss: 0.0158 - accuracy: 0.7898\n",
      "Epoch 12/50\n",
      "34/34 [==============================] - 4s 115ms/step - loss: 0.0140 - accuracy: 0.8094\n",
      "Epoch 13/50\n",
      "34/34 [==============================] - 2s 71ms/step - loss: 0.0124 - accuracy: 0.8277\n",
      "Epoch 14/50\n",
      "34/34 [==============================] - 2s 73ms/step - loss: 0.0110 - accuracy: 0.8450\n",
      "Epoch 15/50\n",
      "34/34 [==============================] - 2s 52ms/step - loss: 0.0098 - accuracy: 0.8613\n",
      "Epoch 16/50\n",
      "34/34 [==============================] - 1s 36ms/step - loss: 0.0087 - accuracy: 0.8700\n",
      "Epoch 17/50\n",
      "34/34 [==============================] - 2s 52ms/step - loss: 0.0078 - accuracy: 0.8746\n",
      "Epoch 18/50\n",
      "34/34 [==============================] - 2s 65ms/step - loss: 0.0071 - accuracy: 0.8824\n",
      "Epoch 19/50\n",
      "34/34 [==============================] - 1s 40ms/step - loss: 0.0064 - accuracy: 0.8860\n",
      "Epoch 20/50\n",
      "34/34 [==============================] - 2s 52ms/step - loss: 0.0059 - accuracy: 0.8955\n",
      "Epoch 21/50\n",
      "34/34 [==============================] - 1s 36ms/step - loss: 0.0054 - accuracy: 0.8920\n",
      "Epoch 22/50\n",
      "34/34 [==============================] - 2s 56ms/step - loss: 0.0052 - accuracy: 0.8921\n",
      "Epoch 23/50\n",
      "34/34 [==============================] - 1s 32ms/step - loss: 0.0047 - accuracy: 0.9006\n",
      "Epoch 24/50\n",
      "34/34 [==============================] - 1s 44ms/step - loss: 0.0043 - accuracy: 0.9012\n",
      "Epoch 25/50\n",
      "34/34 [==============================] - 2s 46ms/step - loss: 0.0040 - accuracy: 0.9027\n",
      "Epoch 26/50\n",
      "34/34 [==============================] - 2s 70ms/step - loss: 0.0038 - accuracy: 0.9000\n",
      "Epoch 27/50\n",
      "34/34 [==============================] - 2s 52ms/step - loss: 0.0037 - accuracy: 0.9013\n",
      "Epoch 28/50\n",
      "34/34 [==============================] - 2s 51ms/step - loss: 0.0034 - accuracy: 0.9050\n",
      "Epoch 29/50\n",
      "34/34 [==============================] - 2s 48ms/step - loss: 0.0032 - accuracy: 0.9054\n",
      "Epoch 30/50\n",
      "34/34 [==============================] - 1s 42ms/step - loss: 0.0031 - accuracy: 0.9058\n",
      "Epoch 31/50\n",
      "34/34 [==============================] - 1s 37ms/step - loss: 0.0030 - accuracy: 0.9085\n",
      "Epoch 32/50\n",
      "34/34 [==============================] - 1s 13ms/step - loss: 0.0029 - accuracy: 0.9046\n",
      "Epoch 33/50\n",
      "34/34 [==============================] - 1s 27ms/step - loss: 0.0027 - accuracy: 0.9024\n",
      "Epoch 34/50\n",
      "34/34 [==============================] - 1s 28ms/step - loss: 0.0026 - accuracy: 0.9086\n",
      "Epoch 35/50\n",
      "34/34 [==============================] - 1s 25ms/step - loss: 0.0025 - accuracy: 0.9070\n",
      "Epoch 36/50\n",
      "34/34 [==============================] - 1s 37ms/step - loss: 0.0025 - accuracy: 0.9081\n",
      "Epoch 37/50\n",
      "34/34 [==============================] - 3s 77ms/step - loss: 0.0024 - accuracy: 0.9097\n",
      "Epoch 38/50\n",
      "34/34 [==============================] - 1s 31ms/step - loss: 0.0024 - accuracy: 0.9049\n",
      "Epoch 39/50\n",
      "34/34 [==============================] - 1s 25ms/step - loss: 0.0021 - accuracy: 0.9088\n",
      "Epoch 40/50\n",
      "34/34 [==============================] - 1s 33ms/step - loss: 0.0021 - accuracy: 0.9092\n",
      "Epoch 41/50\n",
      "34/34 [==============================] - 2s 48ms/step - loss: 0.0020 - accuracy: 0.9074\n",
      "Epoch 42/50\n",
      "34/34 [==============================] - 1s 22ms/step - loss: 0.0020 - accuracy: 0.9067\n",
      "Epoch 43/50\n",
      "34/34 [==============================] - 1s 27ms/step - loss: 0.0019 - accuracy: 0.9126\n",
      "Epoch 44/50\n",
      "34/34 [==============================] - 1s 23ms/step - loss: 0.0019 - accuracy: 0.9068\n",
      "Epoch 45/50\n",
      "34/34 [==============================] - 1s 21ms/step - loss: 0.0019 - accuracy: 0.9095\n",
      "Epoch 46/50\n",
      "34/34 [==============================] - 1s 23ms/step - loss: 0.0019 - accuracy: 0.9097\n",
      "Epoch 47/50\n",
      "34/34 [==============================] - 1s 30ms/step - loss: 0.0018 - accuracy: 0.9116\n",
      "Epoch 48/50\n",
      "34/34 [==============================] - 1s 34ms/step - loss: 0.0018 - accuracy: 0.9090\n",
      "Epoch 49/50\n",
      "34/34 [==============================] - 1s 44ms/step - loss: 0.0016 - accuracy: 0.9078\n",
      "Epoch 50/50\n",
      "34/34 [==============================] - 1s 38ms/step - loss: 0.0017 - accuracy: 0.9089\n",
      "9/9 [==============================] - 0s 4ms/step\n",
      "             model_name  micro_avg_f1_score  hamming_loss  accuracy\n",
      "0                    BR            0.800178      0.004634  0.709917\n",
      "1      Classifier_Chain            0.812240      0.004423  0.736330\n",
      "2  Label_Powerset Chain            0.831524      0.004315  0.812790\n",
      "3             MLP_Glove            0.172472      0.025445  0.159870\n",
      "4          BiLSTM_Glove            0.676641      0.006899  0.568582\n",
      "5          BiLSTM_glove            0.784573      0.005205  0.698795\n",
      "6             CNN_glove            0.515425      0.009139  0.434662\n",
      "7             CNN_glove            0.799916      0.004897  0.715014\n",
      "8             CNN_glove            0.817610      0.004479  0.727525\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.91      0.93      0.92       465\n",
      "           1       0.08      1.00      0.14         1\n",
      "           2       0.70      1.00      0.82         7\n",
      "           3       0.35      0.70      0.47        10\n",
      "           4       0.17      1.00      0.29         1\n",
      "           5       0.00      0.00      0.00         0\n",
      "           6       0.80      1.00      0.89         8\n",
      "           7       0.00      0.00      0.00         0\n",
      "           8       0.00      0.00      0.00         1\n",
      "           9       0.50      1.00      0.67        15\n",
      "          10       0.42      0.83      0.56         6\n",
      "          11       0.00      0.00      0.00         0\n",
      "          12       0.79      0.93      0.85        44\n",
      "          13       0.22      1.00      0.36         2\n",
      "          14       0.00      0.00      0.00         0\n",
      "          15       0.60      0.82      0.69        11\n",
      "          16       0.00      0.00      0.00         0\n",
      "          17       0.69      0.84      0.76       107\n",
      "          18       0.00      0.00      0.00         0\n",
      "          19       0.39      0.85      0.54        13\n",
      "          20       0.00      0.00      0.00         0\n",
      "          21       0.94      0.96      0.95       772\n",
      "          22       0.17      0.50      0.25         2\n",
      "          23       0.27      0.75      0.40         4\n",
      "          24       0.00      0.00      0.00         1\n",
      "          25       0.87      0.96      0.92        28\n",
      "          26       0.81      0.94      0.87        98\n",
      "          27       0.00      0.00      0.00         0\n",
      "          28       0.00      0.00      0.00         0\n",
      "          29       0.33      1.00      0.50         2\n",
      "          30       0.50      1.00      0.67         1\n",
      "          31       0.50      1.00      0.67         1\n",
      "          32       0.00      0.00      0.00         0\n",
      "          33       0.00      0.00      0.00         0\n",
      "          34       0.77      0.82      0.79        73\n",
      "          35       0.45      0.83      0.59         6\n",
      "          36       0.10      1.00      0.18         1\n",
      "          37       0.00      0.00      0.00         0\n",
      "          38       0.42      1.00      0.59         5\n",
      "          39       0.00      0.00      0.00         0\n",
      "          40       0.29      1.00      0.44         2\n",
      "          41       1.00      1.00      1.00         2\n",
      "          42       0.00      0.00      0.00         0\n",
      "          43       0.22      0.67      0.33         6\n",
      "          44       0.11      1.00      0.20         1\n",
      "          45       0.33      0.67      0.44         3\n",
      "          46       0.73      0.86      0.79       126\n",
      "          47       0.59      1.00      0.75        19\n",
      "          48       0.00      0.00      0.00         0\n",
      "          49       0.40      0.80      0.53        15\n",
      "          50       0.00      0.00      0.00         0\n",
      "          51       0.00      0.00      0.00         0\n",
      "          52       0.00      0.00      0.00         0\n",
      "          53       1.00      1.00      1.00         1\n",
      "          54       0.58      0.64      0.61        28\n",
      "          55       0.33      1.00      0.50         1\n",
      "          56       0.50      1.00      0.67         1\n",
      "          57       0.60      0.75      0.67         4\n",
      "          58       0.00      0.00      0.00         0\n",
      "          59       0.00      0.00      0.00         0\n",
      "          60       0.33      1.00      0.50         1\n",
      "          61       0.00      0.00      0.00         0\n",
      "          62       0.00      0.00      0.00         0\n",
      "          63       0.00      0.00      0.00         0\n",
      "          64       0.00      0.00      0.00         0\n",
      "          65       0.80      1.00      0.89         4\n",
      "          66       0.33      1.00      0.50         4\n",
      "          67       0.00      0.00      0.00         0\n",
      "          68       0.00      0.00      0.00         0\n",
      "          69       0.20      1.00      0.33         2\n",
      "          70       0.00      0.00      0.00         0\n",
      "          71       0.52      0.94      0.67        33\n",
      "          72       0.00      0.00      0.00         0\n",
      "          73       0.25      0.50      0.33         2\n",
      "          74       0.20      1.00      0.33         1\n",
      "          75       0.17      0.50      0.25         2\n",
      "          76       0.60      0.63      0.62        19\n",
      "          77       0.22      1.00      0.36         2\n",
      "          78       0.67      0.94      0.78        17\n",
      "          79       0.00      0.00      0.00         0\n",
      "          80       0.00      0.00      0.00         0\n",
      "          81       0.00      0.00      0.00         0\n",
      "          82       0.00      0.00      0.00         0\n",
      "          83       0.38      1.00      0.55         3\n",
      "          84       0.62      0.78      0.69        79\n",
      "          85       0.48      0.91      0.62        11\n",
      "          86       0.77      0.90      0.83        48\n",
      "          87       0.83      1.00      0.91         5\n",
      "          88       0.38      0.83      0.53         6\n",
      "          89       0.14      1.00      0.25         1\n",
      "\n",
      "   micro avg       0.74      0.91      0.82      2134\n",
      "   macro avg       0.30      0.56      0.37      2134\n",
      "weighted avg       0.82      0.91      0.86      2134\n",
      " samples avg       0.79      0.80      0.79      2134\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "<ipython-input-167-14bdbcadc155>:19: FutureWarning: The frame.append method is deprecated and will be removed from pandas in a future version. Use pandas.concat instead.\n",
      "  scores_df = scores_df.append(arr,ignore_index=True)\n",
      "/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classification.py:1344: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n",
      "/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classification.py:1344: UndefinedMetricWarning: Recall and F-score are ill-defined and being set to 0.0 in labels with no true samples. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n",
      "/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classification.py:1344: UndefinedMetricWarning: Recall and F-score are ill-defined and being set to 0.0 in samples with no true labels. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n"
     ]
    }
   ],
   "source": [
    "print(\"=====================Glove CNN===========================\")\n",
    "inp = Input(shape=(70,))\n",
    "\n",
    "x = Embedding(VOCAB_SIZE+1, 300, weights=[embd_mat])(inp)\n",
    "x = layers.Conv1D(filters=128, kernel_size=5, activation='relu')(x)\n",
    "x = GlobalMaxPool1D()(x)\n",
    "x = Dense(64, activation=\"relu\")(x)\n",
    "x = Dropout(0.1)(x)\n",
    "x = Dense(32, activation='relu')(x)\n",
    "x = Dense(90, activation=\"sigmoid\")(x)\n",
    "model = Model(inputs=inp, outputs=x)\n",
    "model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
    "\n",
    "model.fit(np.asarray(X_train),np.asarray(y_train), batch_size=256, epochs=50)\n",
    "y_pred = model.predict(np.asarray(X_test),batch_size=256, verbose=1)\n",
    "y_pred= np.where(y_pred > 0.5, 1, 0)\n",
    "\n",
    "arr={'model_name':'CNN_glove','micro_avg_f1_score':f1_score(y_test, y_pred, average=\"micro\"),'hamming_loss':hamming_loss(y_test, y_pred),'accuracy':accuracy_score(y_test,y_pred)}\n",
    "scores_df = scores_df.append(arr,ignore_index=True)\n",
    "print(scores_df)\n",
    "print(classification_report(y_pred,y_test))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "2-K8IPnl8ygV"
   },
   "outputs": [],
   "source": [
    "print(\"=====================Bert CNN===========================\")\n",
    "inp = Input(shape=(max_seq_len,))\n",
    "\n",
    "x = Embedding(VOCAB_SIZE, 768, weights=[bert_embd_mat])(inp)\n",
    "x = layers.Conv1D(filters=128, kernel_size=5, activation='relu')(x)\n",
    "x = GlobalMaxPool1D()(x)\n",
    "x = Dense(64, activation=\"relu\")(x)\n",
    "x = Dropout(0.1)(x)\n",
    "x = Dense(32, activation='relu')(x)\n",
    "x = Dense(7, activation=\"sigmoid\")(x)\n",
    "model = Model(inputs=inp, outputs=x)\n",
    "model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
    "\n",
    "model.fit(np.asarray(X_train),np.asarray(y_train), batch_size=256, epochs=3)\n",
    "y_pred = model.predict(np.asarray(X_test),batch_size=256, verbose=1)\n",
    "y_pred= np.where(y_pred > 0.5, 1, 0)\n",
    "\n",
    "arr={'model_name':'CNN_bert','micro_avg_f1_score':f1_score(y_test, y_pred, average=\"micro\"),'hamming_loss':hamming_loss(y_test, y_pred),'accuracy':accuracy_score(y_test,y_pred)}\n",
    "scores_df = scores_df.append(arr,ignore_index=True)\n",
    "print(scores_df)\n",
    "print(classification_report(y_pred,y_test))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "JEtUvv-17Djh"
   },
   "outputs": [],
   "source": [
    "adj_matrix = create_adjacency_matrix_cooccurance(y_train.numpy())\n",
    "adj_matrix = torch.tensor(adj_matrix)\n",
    "adj_matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "ltSyz_-77UG9"
   },
   "outputs": [],
   "source": [
    "save_path = './train_result.pt'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "7wdlf7nH7WgF"
   },
   "outputs": [],
   "source": [
    "glove_label_embedding = torch.zeros(len(labels),300)\n",
    "\n",
    "for index, label in enumerate(labels):\n",
    "  wrds = label.split('-')\n",
    "  for l in wrds:\n",
    "    if l in glove_embeddings.keys():\n",
    "        glove_label_embedding[index] +=  torch.from_numpy(glove_embeddings[l])\n",
    "  glove_label_embedding[index]=glove_label_embedding[index]/len(wrds)  \n",
    "\n",
    "print(glove_label_embedding)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "5GwsNwWJ7s6r"
   },
   "outputs": [],
   "source": [
    "model = MAGNET(300, 250, adj_matrix, glove_embedding_matrix,heads=8)\n",
    "train(model, X_train, glove_label_embedding,y_train,total_epoch=15, save_path=save_path)"
   ]
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "provenance": []
  },
  "gpuClass": "standard",
  "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": "3.8.10"
  },
  "widgets": {
   "application/vnd.jupyter.widget-state+json": {
    "0a477918c72c4786ab284f2a0e98d163": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HTMLModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HTMLModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HTMLView",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_540f7163e5594bb9ac1e14f74fd146b5",
      "placeholder": "​",
      "style": "IPY_MODEL_ae8f4f6d59034045b45aaddf0af1b93b",
      "value": "Downloading (…)okenizer_config.json: 100%"
     }
    },
    "0bb8614697824a8580229d406e3f4bd5": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "0d6e9828ec294966a09cc9878a0a2969": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "FloatProgressModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "FloatProgressModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "ProgressView",
      "bar_style": "success",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_8fdcf4cf179947a5a412857c634b1105",
      "max": 440473133,
      "min": 0,
      "orientation": "horizontal",
      "style": "IPY_MODEL_ac665d6cc49844529c2fabeadf40329e",
      "value": 440473133
     }
    },
    "0e8de91b02a44247af2544b56d47c72d": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "DescriptionStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "DescriptionStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "description_width": ""
     }
    },
    "21c623bb43e942448d07235afa5e8cb1": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HBoxModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HBoxModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HBoxView",
      "box_style": "",
      "children": [
       "IPY_MODEL_0a477918c72c4786ab284f2a0e98d163",
       "IPY_MODEL_96becd5034d94569b31b6810ea9083c7",
       "IPY_MODEL_8ad3c9319e8144cd95cd91e61c28ea9a"
      ],
      "layout": "IPY_MODEL_3c9858ad885d46d994580c555b371378"
     }
    },
    "27de5ddcf9ad41d0abdb7fc40fd63ef4": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "ProgressStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "ProgressStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "bar_color": null,
      "description_width": ""
     }
    },
    "2d36b2e4ce9247f8a7d556ea2d4fbbaf": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "2e784340cdd3474485b839f01ea8a735": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "3267d5f59bac457fbea59039d2a3d741": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "ProgressStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "ProgressStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "bar_color": null,
      "description_width": ""
     }
    },
    "3a4d463189de4c2199170550c9fcb720": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HTMLModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HTMLModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HTMLView",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_2e784340cdd3474485b839f01ea8a735",
      "placeholder": "​",
      "style": "IPY_MODEL_cf024c1e21af49bf85bac3f215ce2bec",
      "value": "Downloading pytorch_model.bin: 100%"
     }
    },
    "3c9858ad885d46d994580c555b371378": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "3d25d6797f1744a89f8a6b012b5ca18c": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HBoxModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HBoxModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HBoxView",
      "box_style": "",
      "children": [
       "IPY_MODEL_bbf6d4148a6047c3843455648933c6c3",
       "IPY_MODEL_ffd62e74ed4a42068989fdb81bcbeaa4",
       "IPY_MODEL_e5dd935d0fd74ae0981f665bdc46ec99"
      ],
      "layout": "IPY_MODEL_e0ce6cd8bb094835acfe968f0984ae60"
     }
    },
    "48bbcd8bdd4e4f22892233bb9c22499c": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "DescriptionStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "DescriptionStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "description_width": ""
     }
    },
    "4d0b3d5f7da44efebf8f40a26d2f878a": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "DescriptionStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "DescriptionStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "description_width": ""
     }
    },
    "540f7163e5594bb9ac1e14f74fd146b5": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "637ffa573b1849fe8aa7589d24dc7351": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "66bdaf101c87445db8a031b92014de63": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "ProgressStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "ProgressStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "bar_color": null,
      "description_width": ""
     }
    },
    "690d419a190a40779b69afe877924db8": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "6dc757df6fc24eb6ab09ed6bc5363dc8": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "871cf9dc73a44dffa439696c5b5091e3": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HTMLModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HTMLModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HTMLView",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_690d419a190a40779b69afe877924db8",
      "placeholder": "​",
      "style": "IPY_MODEL_0e8de91b02a44247af2544b56d47c72d",
      "value": " 440M/440M [00:01&lt;00:00, 319MB/s]"
     }
    },
    "88efcfea3974450c928c974f091744b3": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "FloatProgressModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "FloatProgressModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "ProgressView",
      "bar_style": "success",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_fca86cc61473462eaedaabf3fe3de8e8",
      "max": 570,
      "min": 0,
      "orientation": "horizontal",
      "style": "IPY_MODEL_27de5ddcf9ad41d0abdb7fc40fd63ef4",
      "value": 570
     }
    },
    "8ad3c9319e8144cd95cd91e61c28ea9a": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HTMLModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HTMLModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HTMLView",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_637ffa573b1849fe8aa7589d24dc7351",
      "placeholder": "​",
      "style": "IPY_MODEL_4d0b3d5f7da44efebf8f40a26d2f878a",
      "value": " 28.0/28.0 [00:00&lt;00:00, 2.05kB/s]"
     }
    },
    "8fdcf4cf179947a5a412857c634b1105": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "96becd5034d94569b31b6810ea9083c7": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "FloatProgressModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "FloatProgressModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "ProgressView",
      "bar_style": "success",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_ce5fa2e065544e4eaf1d789fa0f09de4",
      "max": 28,
      "min": 0,
      "orientation": "horizontal",
      "style": "IPY_MODEL_3267d5f59bac457fbea59039d2a3d741",
      "value": 28
     }
    },
    "a2897cf3cb854c33b457160eefd86cc7": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "a976238571174f8d9192a31ff3d5c3d5": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HTMLModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HTMLModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HTMLView",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_2d36b2e4ce9247f8a7d556ea2d4fbbaf",
      "placeholder": "​",
      "style": "IPY_MODEL_b244f49621604e20b90dc0659faa598b",
      "value": "Downloading (…)lve/main/config.json: 100%"
     }
    },
    "ac665d6cc49844529c2fabeadf40329e": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "ProgressStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "ProgressStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "bar_color": null,
      "description_width": ""
     }
    },
    "ae8f4f6d59034045b45aaddf0af1b93b": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "DescriptionStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "DescriptionStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "description_width": ""
     }
    },
    "b244f49621604e20b90dc0659faa598b": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "DescriptionStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "DescriptionStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "description_width": ""
     }
    },
    "b2692d83f09a486c9c3186ec73f64f1d": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "b84711d90cfe4f889c662ad03eb14716": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HBoxModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HBoxModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HBoxView",
      "box_style": "",
      "children": [
       "IPY_MODEL_a976238571174f8d9192a31ff3d5c3d5",
       "IPY_MODEL_88efcfea3974450c928c974f091744b3",
       "IPY_MODEL_d1542aa04b344713a1134fd0302ba162"
      ],
      "layout": "IPY_MODEL_b2692d83f09a486c9c3186ec73f64f1d"
     }
    },
    "bbf6d4148a6047c3843455648933c6c3": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HTMLModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HTMLModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HTMLView",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_cf242fad209e41938ebdd7665a8aa2da",
      "placeholder": "​",
      "style": "IPY_MODEL_d801e0df47184335bdfedf763e694d17",
      "value": "Downloading (…)solve/main/vocab.txt: 100%"
     }
    },
    "ce5fa2e065544e4eaf1d789fa0f09de4": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "cf024c1e21af49bf85bac3f215ce2bec": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "DescriptionStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "DescriptionStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "description_width": ""
     }
    },
    "cf242fad209e41938ebdd7665a8aa2da": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "d1542aa04b344713a1134fd0302ba162": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HTMLModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HTMLModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HTMLView",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_6dc757df6fc24eb6ab09ed6bc5363dc8",
      "placeholder": "​",
      "style": "IPY_MODEL_48bbcd8bdd4e4f22892233bb9c22499c",
      "value": " 570/570 [00:00&lt;00:00, 37.5kB/s]"
     }
    },
    "d801e0df47184335bdfedf763e694d17": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "DescriptionStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "DescriptionStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "description_width": ""
     }
    },
    "e0ce6cd8bb094835acfe968f0984ae60": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "e2bdcc18437647eb82c9df4e2795e9cd": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "DescriptionStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "DescriptionStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "description_width": ""
     }
    },
    "e45af4b8cb9f4c9991f18a8828a7015a": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "e5dd935d0fd74ae0981f665bdc46ec99": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HTMLModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HTMLModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HTMLView",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_0bb8614697824a8580229d406e3f4bd5",
      "placeholder": "​",
      "style": "IPY_MODEL_e2bdcc18437647eb82c9df4e2795e9cd",
      "value": " 232k/232k [00:00&lt;00:00, 465kB/s]"
     }
    },
    "f7e6d2cac5424ab1b2b060715cd36a07": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HBoxModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HBoxModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HBoxView",
      "box_style": "",
      "children": [
       "IPY_MODEL_3a4d463189de4c2199170550c9fcb720",
       "IPY_MODEL_0d6e9828ec294966a09cc9878a0a2969",
       "IPY_MODEL_871cf9dc73a44dffa439696c5b5091e3"
      ],
      "layout": "IPY_MODEL_e45af4b8cb9f4c9991f18a8828a7015a"
     }
    },
    "fca86cc61473462eaedaabf3fe3de8e8": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "ffd62e74ed4a42068989fdb81bcbeaa4": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "FloatProgressModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "FloatProgressModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "ProgressView",
      "bar_style": "success",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_a2897cf3cb854c33b457160eefd86cc7",
      "max": 231508,
      "min": 0,
      "orientation": "horizontal",
      "style": "IPY_MODEL_66bdaf101c87445db8a031b92014de63",
      "value": 231508
     }
    }
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}