Hate_Speech_Detection_Electra_TSA.ipynb

{
  "cells": [
    {
      "cell_type": "code",
      "execution_count": 2,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Opwj7Tg4r9HB",
        "outputId": "16591470-de5b-48aa-aee0-6e763e3f190b"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Requirement already satisfied: tensorflow-text==2.4.1 in /usr/local/lib/python3.7/dist-packages (2.4.1)\n",
            "Requirement already satisfied: tensorflow-hub>=0.8.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow-text==2.4.1) (0.12.0)\n",
            "Requirement already satisfied: tensorflow<2.5,>=2.4.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow-text==2.4.1) (2.4.4)\n",
            "Requirement already satisfied: numpy~=1.19.2 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (1.19.5)\n",
            "Requirement already satisfied: flatbuffers~=1.12.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (1.12)\n",
            "Requirement already satisfied: keras-preprocessing~=1.1.2 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (1.1.2)\n",
            "Requirement already satisfied: gast==0.3.3 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (0.3.3)\n",
            "Requirement already satisfied: six~=1.15.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (1.15.0)\n",
            "Requirement already satisfied: termcolor~=1.1.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (1.1.0)\n",
            "Requirement already satisfied: grpcio~=1.32.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (1.32.0)\n",
            "Requirement already satisfied: wheel~=0.35 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (0.37.1)\n",
            "Requirement already satisfied: wrapt~=1.12.1 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (1.12.1)\n",
            "Requirement already satisfied: protobuf>=3.9.2 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (3.17.3)\n",
            "Requirement already satisfied: astunparse~=1.6.3 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (1.6.3)\n",
            "Requirement already satisfied: typing-extensions~=3.7.4 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (3.7.4.3)\n",
            "Requirement already satisfied: tensorflow-estimator<2.5.0,>=2.4.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (2.4.0)\n",
            "Requirement already satisfied: absl-py~=0.10 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (0.15.0)\n",
            "Requirement already satisfied: tensorboard~=2.4 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (2.8.0)\n",
            "Requirement already satisfied: google-pasta~=0.2 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (0.2.0)\n",
            "Requirement already satisfied: h5py~=2.10.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (2.10.0)\n",
            "Requirement already satisfied: opt-einsum~=3.3.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (3.3.0)\n",
            "Requirement already satisfied: tensorboard-plugin-wit>=1.6.0 in /usr/local/lib/python3.7/dist-packages (from tensorboard~=2.4->tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (1.8.1)\n",
            "Requirement already satisfied: werkzeug>=0.11.15 in /usr/local/lib/python3.7/dist-packages (from tensorboard~=2.4->tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (1.0.1)\n",
            "Requirement already satisfied: google-auth<3,>=1.6.3 in /usr/local/lib/python3.7/dist-packages (from tensorboard~=2.4->tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (1.35.0)\n",
            "Requirement already satisfied: google-auth-oauthlib<0.5,>=0.4.1 in /usr/local/lib/python3.7/dist-packages (from tensorboard~=2.4->tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (0.4.6)\n",
            "Requirement already satisfied: markdown>=2.6.8 in /usr/local/lib/python3.7/dist-packages (from tensorboard~=2.4->tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (3.4.1)\n",
            "Requirement already satisfied: tensorboard-data-server<0.7.0,>=0.6.0 in /usr/local/lib/python3.7/dist-packages (from tensorboard~=2.4->tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (0.6.1)\n",
            "Requirement already satisfied: requests<3,>=2.21.0 in /usr/local/lib/python3.7/dist-packages (from tensorboard~=2.4->tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (2.23.0)\n",
            "Requirement already satisfied: setuptools>=41.0.0 in /usr/local/lib/python3.7/dist-packages (from tensorboard~=2.4->tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (57.4.0)\n",
            "Requirement already satisfied: rsa<5,>=3.1.4 in /usr/local/lib/python3.7/dist-packages (from google-auth<3,>=1.6.3->tensorboard~=2.4->tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (4.9)\n",
            "Requirement already satisfied: cachetools<5.0,>=2.0.0 in /usr/local/lib/python3.7/dist-packages (from google-auth<3,>=1.6.3->tensorboard~=2.4->tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (4.2.4)\n",
            "Requirement already satisfied: pyasn1-modules>=0.2.1 in /usr/local/lib/python3.7/dist-packages (from google-auth<3,>=1.6.3->tensorboard~=2.4->tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (0.2.8)\n",
            "Requirement already satisfied: requests-oauthlib>=0.7.0 in /usr/local/lib/python3.7/dist-packages (from google-auth-oauthlib<0.5,>=0.4.1->tensorboard~=2.4->tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (1.3.1)\n",
            "Requirement already satisfied: importlib-metadata>=4.4 in /usr/local/lib/python3.7/dist-packages (from markdown>=2.6.8->tensorboard~=2.4->tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (4.12.0)\n",
            "Requirement already satisfied: zipp>=0.5 in /usr/local/lib/python3.7/dist-packages (from importlib-metadata>=4.4->markdown>=2.6.8->tensorboard~=2.4->tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (3.8.1)\n",
            "Requirement already satisfied: pyasn1<0.5.0,>=0.4.6 in /usr/local/lib/python3.7/dist-packages (from pyasn1-modules>=0.2.1->google-auth<3,>=1.6.3->tensorboard~=2.4->tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (0.4.8)\n",
            "Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.7/dist-packages (from requests<3,>=2.21.0->tensorboard~=2.4->tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (1.24.3)\n",
            "Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.7/dist-packages (from requests<3,>=2.21.0->tensorboard~=2.4->tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (2.10)\n",
            "Requirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.7/dist-packages (from requests<3,>=2.21.0->tensorboard~=2.4->tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (3.0.4)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.7/dist-packages (from requests<3,>=2.21.0->tensorboard~=2.4->tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (2022.6.15)\n",
            "Requirement already satisfied: oauthlib>=3.0.0 in /usr/local/lib/python3.7/dist-packages (from requests-oauthlib>=0.7.0->google-auth-oauthlib<0.5,>=0.4.1->tensorboard~=2.4->tensorflow<2.5,>=2.4.0->tensorflow-text==2.4.1) (3.2.0)\n",
            "Colab only includes TensorFlow 2.x; %tensorflow_version has no effect.\n",
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Requirement already satisfied: transformers in /usr/local/lib/python3.7/dist-packages (4.21.1)\n",
            "Requirement already satisfied: pyyaml>=5.1 in /usr/local/lib/python3.7/dist-packages (from transformers) (6.0)\n",
            "Requirement already satisfied: filelock in /usr/local/lib/python3.7/dist-packages (from transformers) (3.7.1)\n",
            "Requirement already satisfied: huggingface-hub<1.0,>=0.1.0 in /usr/local/lib/python3.7/dist-packages (from transformers) (0.8.1)\n",
            "Requirement already satisfied: requests in /usr/local/lib/python3.7/dist-packages (from transformers) (2.23.0)\n",
            "Requirement already satisfied: packaging>=20.0 in /usr/local/lib/python3.7/dist-packages (from transformers) (21.3)\n",
            "Requirement already satisfied: numpy>=1.17 in /usr/local/lib/python3.7/dist-packages (from transformers) (1.19.5)\n",
            "Requirement already satisfied: tqdm>=4.27 in /usr/local/lib/python3.7/dist-packages (from transformers) (4.64.0)\n",
            "Requirement already satisfied: importlib-metadata in /usr/local/lib/python3.7/dist-packages (from transformers) (4.12.0)\n",
            "Requirement already satisfied: tokenizers!=0.11.3,<0.13,>=0.11.1 in /usr/local/lib/python3.7/dist-packages (from transformers) (0.12.1)\n",
            "Requirement already satisfied: regex!=2019.12.17 in /usr/local/lib/python3.7/dist-packages (from transformers) (2022.6.2)\n",
            "Requirement already satisfied: typing-extensions>=3.7.4.3 in /usr/local/lib/python3.7/dist-packages (from huggingface-hub<1.0,>=0.1.0->transformers) (3.7.4.3)\n",
            "Requirement already satisfied: pyparsing!=3.0.5,>=2.0.2 in /usr/local/lib/python3.7/dist-packages (from packaging>=20.0->transformers) (3.0.9)\n",
            "Requirement already satisfied: zipp>=0.5 in /usr/local/lib/python3.7/dist-packages (from importlib-metadata->transformers) (3.8.1)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.7/dist-packages (from requests->transformers) (2022.6.15)\n",
            "Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.7/dist-packages (from requests->transformers) (2.10)\n",
            "Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.7/dist-packages (from requests->transformers) (1.24.3)\n",
            "Requirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.7/dist-packages (from requests->transformers) (3.0.4)\n"
          ]
        }
      ],
      "source": [
        "!pip3 install tensorflow-text==2.4.1\n",
        "\n",
        "# import tensorflow as tf\n",
        "# print(tf.__version__)\n",
        "\n",
        "import tensorflow as tf\n",
        "%tensorflow_version 2.x\n",
        "# import tensorflow as tf\n",
        "!pip3 install transformers\n",
        "import os\n",
        "import shutil\n",
        "import tensorflow_hub as hub\n",
        "import tensorflow_text as text\n",
        "import matplotlib.pyplot as plt\n",
        "import pandas as pd\n",
        "import numpy as np\n",
        "from sklearn.model_selection import GroupKFold\n",
        "import matplotlib.pyplot as plt\n",
        "from tqdm.notebook import tqdm\n",
        "from scipy.stats import spearmanr\n",
        "from math import floor, ceil\n",
        "import pickle\n",
        "# import tensorflow as tf\n",
        "from tensorflow.keras.preprocessing import text, sequence\n",
        "from tensorflow.keras.models import Sequential\n",
        "from tensorflow.keras.layers import Conv1D, GlobalMaxPooling1D, MaxPooling1D, Dense, Dropout, Activation, Embedding\n",
        "from sklearn.model_selection import train_test_split\n",
        "from sklearn.metrics import accuracy_score, f1_score, confusion_matrix\n",
        "from sklearn.model_selection import train_test_split\n",
        "from sklearn.metrics import confusion_matrix,classification_report\n",
        "import shutil\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 3,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "-cE_2ZehujcQ",
        "outputId": "81c1cc19-c83a-4e2f-e0a1-5dd605afb099"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Examples:\n",
            "    Total: 31962\n",
            "    hate: 2242 (7.01% of total)\n",
            "\n",
            "Examples:\n",
            "    Total: 31962\n",
            "    Neither: 29720 (92.99% of total)\n",
            "\n"
          ]
        }
      ],
      "source": [
        "#---------Data Read & Transformation---------\n",
        "#Data Reading for Davidson Dataset\n",
        "nRowsRead = None # specify 'None' to read complete file\n",
        "tweets_train = pd.read_csv('/TSA.csv')\n",
        "tweets_train['label'].value_counts()\n",
        "# sample = tweets_train.sample(n = 100, random_state = 155)\n",
        "# sample['text'] = sample['tweet'].apply(lambda x: p.clean(x))\n",
        "# nRow, nCol = df0.shape\n",
        "# print(df0)\n",
        "# print('There are {} rows and {} columns'.format(nRow, nCol))\n",
        "\n",
        "#Doing Transformation\n",
        "c=tweets_train['label']\n",
        "tweets_train.rename(columns={'tweet' : 'text',\n",
        "                   'label' : 'category'}, \n",
        "                    inplace=True)\n",
        "a=tweets_train['text']\n",
        "b=tweets_train['category'].map({0: 'neither', 1: 'hate'})\n",
        "df= pd.concat([a,b,c], axis=1)\n",
        "# df.rename(columns={'class' : 'label'}, \n",
        "#                     inplace=True)\n",
        "\n",
        "neither,hate = np.bincount(df['label'])\n",
        "total = hate + neither\n",
        "print('Examples:\\n    Total: {}\\n    hate: {} ({:.2f}% of total)\\n'.format(\n",
        "    total, hate, 100 * hate / total))\n",
        "# print('Examples:\\n    Total: {}\\n    Ofensive: {} ({:.2f}% of total)\\n'.format(\n",
        "#     total, ofensive, 100 * ofensive / total))\n",
        "print('Examples:\\n    Total: {}\\n    Neither: {} ({:.2f}% of total)\\n'.format(\n",
        "    total, neither, 100 * neither / total))\n",
        "\n",
        "#Splitting Dataset\n",
        "X_train_, X_test, y_train_, y_test = train_test_split(\n",
        "    df.index.values,\n",
        "    df.label.values,\n",
        "    test_size=0.10,\n",
        "    random_state=42,\n",
        "    stratify=df.label.values,    \n",
        ")\n",
        "\n",
        "X_train, X_val, y_train, y_val = train_test_split(\n",
        "    df.loc[X_train_].index.values,\n",
        "    df.loc[X_train_].label.values,\n",
        "    test_size=0.10,\n",
        "    random_state=42,\n",
        "    stratify=df.loc[X_train_].label.values,  \n",
        ")\n",
        "\n",
        "df['data_type'] = ['not_set']*df.shape[0]\n",
        "df.loc[X_train, 'data_type'] = 'train'\n",
        "df.loc[X_val, 'data_type'] = 'val'\n",
        "df.loc[X_test, 'data_type'] = 'test'\n",
        "\n",
        "df_train = df.loc[df[\"data_type\"]==\"train\"]\n",
        "df_val = df.loc[df[\"data_type\"]==\"val\"]\n",
        "df_test = df.loc[df[\"data_type\"]==\"test\"]"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 4,
      "metadata": {
        "id": "oDOmWZQ2oMcC"
      },
      "outputs": [],
      "source": [
        "\n",
        "\n",
        "\n",
        "#--------- Choosing Transfomer Model ---------\n",
        "#creating tensors\n",
        "train_ds = tf.data.Dataset.from_tensor_slices((df_train.text.values, df_train.label.values))\n",
        "val_ds = tf.data.Dataset.from_tensor_slices((df_val.text.values, df_val.label.values))\n",
        "test_ds = tf.data.Dataset.from_tensor_slices((df_test.text.values, df_test.label.values))\n",
        "\n",
        "train_ds = train_ds.shuffle(len(df_train)).batch(32, drop_remainder=False)\n",
        "val_ds = val_ds.shuffle(len(df_val)).batch(32, drop_remainder=False)\n",
        "test_ds = test_ds.shuffle(len(df_test)).batch(32, drop_remainder=False)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 5,
      "metadata": {
        "id": "B-y27QSeoUSg"
      },
      "outputs": [],
      "source": [
        "#Reading bert_en_uncased_L-2_H-128_A-2 in this file to fine tune\n",
        "\n",
        "bert_model_name = 'electra_small'  #@param [\"bert_en_uncased_L-12_H-768_A-12\", \"bert_en_cased_L-12_H-768_A-12\", \"bert_multi_cased_L-12_H-768_A-12\", \"small_bert/bert_en_uncased_L-2_H-128_A-2\", \"small_bert/bert_en_uncased_L-2_H-256_A-4\", \"small_bert/bert_en_uncased_L-2_H-512_A-8\", \"small_bert/bert_en_uncased_L-2_H-768_A-12\", \"small_bert/bert_en_uncased_L-4_H-128_A-2\", \"small_bert/bert_en_uncased_L-4_H-256_A-4\", \"small_bert/bert_en_uncased_L-4_H-512_A-8\", \"small_bert/bert_en_uncased_L-4_H-768_A-12\", \"small_bert/bert_en_uncased_L-6_H-128_A-2\", \"small_bert/bert_en_uncased_L-6_H-256_A-4\", \"small_bert/bert_en_uncased_L-6_H-512_A-8\", \"small_bert/bert_en_uncased_L-6_H-768_A-12\", \"small_bert/bert_en_uncased_L-8_H-128_A-2\", \"small_bert/bert_en_uncased_L-8_H-256_A-4\", \"small_bert/bert_en_uncased_L-8_H-512_A-8\", \"small_bert/bert_en_uncased_L-8_H-768_A-12\", \"small_bert/bert_en_uncased_L-10_H-128_A-2\", \"small_bert/bert_en_uncased_L-10_H-256_A-4\", \"small_bert/bert_en_uncased_L-10_H-512_A-8\", \"small_bert/bert_en_uncased_L-10_H-768_A-12\", \"small_bert/bert_en_uncased_L-12_H-128_A-2\", \"small_bert/bert_en_uncased_L-12_H-256_A-4\", \"small_bert/bert_en_uncased_L-12_H-512_A-8\", \"small_bert/bert_en_uncased_L-12_H-768_A-12\", \"albert_en_base\", \"electra_small\", \"electra_base\", \"experts_pubmed\", \"experts_wiki_books\", \"talking-heads_base\"]\n",
        "\n",
        "map_name_to_handle = {\n",
        "    'bert_en_uncased_L-12_H-768_A-12':\n",
        "        'https://tfhub.dev/tensorflow/bert_en_uncased_L-12_H-768_A-12/3',\n",
        "    'bert_en_cased_L-12_H-768_A-12':\n",
        "        'https://tfhub.dev/tensorflow/bert_en_cased_L-12_H-768_A-12/3',\n",
        "    'bert_multi_cased_L-12_H-768_A-12':\n",
        "        'https://tfhub.dev/tensorflow/bert_multi_cased_L-12_H-768_A-12/3',\n",
        "    'small_bert/bert_en_uncased_L-2_H-128_A-2':\n",
        "        'https://tfhub.dev/tensorflow/small_bert/bert_en_uncased_L-2_H-128_A-2/1',\n",
        "    'small_bert/bert_en_uncased_L-2_H-256_A-4':\n",
        "        'https://tfhub.dev/tensorflow/small_bert/bert_en_uncased_L-2_H-256_A-4/1',\n",
        "    'small_bert/bert_en_uncased_L-2_H-512_A-8':\n",
        "        'https://tfhub.dev/tensorflow/small_bert/bert_en_uncased_L-2_H-512_A-8/1',\n",
        "    'small_bert/bert_en_uncased_L-2_H-768_A-12':\n",
        "        'https://tfhub.dev/tensorflow/small_bert/bert_en_uncased_L-2_H-768_A-12/1',\n",
        "    'small_bert/bert_en_uncased_L-4_H-128_A-2':\n",
        "        'https://tfhub.dev/tensorflow/small_bert/bert_en_uncased_L-4_H-128_A-2/1',\n",
        "    'small_bert/bert_en_uncased_L-4_H-256_A-4':\n",
        "        'https://tfhub.dev/tensorflow/small_bert/bert_en_uncased_L-4_H-256_A-4/1',\n",
        "    'small_bert/bert_en_uncased_L-4_H-512_A-8':\n",
        "        'https://tfhub.dev/tensorflow/small_bert/bert_en_uncased_L-4_H-512_A-8/1',\n",
        "    'small_bert/bert_en_uncased_L-4_H-768_A-12':\n",
        "        'https://tfhub.dev/tensorflow/small_bert/bert_en_uncased_L-4_H-768_A-12/1',\n",
        "    'small_bert/bert_en_uncased_L-6_H-128_A-2':\n",
        "        'https://tfhub.dev/tensorflow/small_bert/bert_en_uncased_L-6_H-128_A-2/1',\n",
        "    'small_bert/bert_en_uncased_L-6_H-256_A-4':\n",
        "        'https://tfhub.dev/tensorflow/small_bert/bert_en_uncased_L-6_H-256_A-4/1',\n",
        "    'small_bert/bert_en_uncased_L-6_H-512_A-8':\n",
        "        'https://tfhub.dev/tensorflow/small_bert/bert_en_uncased_L-6_H-512_A-8/1',\n",
        "    'small_bert/bert_en_uncased_L-6_H-768_A-12':\n",
        "        'https://tfhub.dev/tensorflow/small_bert/bert_en_uncased_L-6_H-768_A-12/1',\n",
        "    'small_bert/bert_en_uncased_L-8_H-128_A-2':\n",
        "        'https://tfhub.dev/tensorflow/small_bert/bert_en_uncased_L-8_H-128_A-2/1',\n",
        "    'small_bert/bert_en_uncased_L-8_H-256_A-4':\n",
        "        'https://tfhub.dev/tensorflow/small_bert/bert_en_uncased_L-8_H-256_A-4/1',\n",
        "    'small_bert/bert_en_uncased_L-8_H-512_A-8':\n",
        "        'https://tfhub.dev/tensorflow/small_bert/bert_en_uncased_L-8_H-512_A-8/1',\n",
        "    'small_bert/bert_en_uncased_L-8_H-768_A-12':\n",
        "        'https://tfhub.dev/tensorflow/small_bert/bert_en_uncased_L-8_H-768_A-12/1',\n",
        "    'small_bert/bert_en_uncased_L-10_H-128_A-2':\n",
        "        'https://tfhub.dev/tensorflow/small_bert/bert_en_uncased_L-10_H-128_A-2/1',\n",
        "    'small_bert/bert_en_uncased_L-10_H-256_A-4':\n",
        "        'https://tfhub.dev/tensorflow/small_bert/bert_en_uncased_L-10_H-256_A-4/1',\n",
        "    'small_bert/bert_en_uncased_L-10_H-512_A-8':\n",
        "        'https://tfhub.dev/tensorflow/small_bert/bert_en_uncased_L-10_H-512_A-8/1',\n",
        "    'small_bert/bert_en_uncased_L-10_H-768_A-12':\n",
        "        'https://tfhub.dev/tensorflow/small_bert/bert_en_uncased_L-10_H-768_A-12/1',\n",
        "    'small_bert/bert_en_uncased_L-12_H-128_A-2':\n",
        "        'https://tfhub.dev/tensorflow/small_bert/bert_en_uncased_L-12_H-128_A-2/1',\n",
        "    'small_bert/bert_en_uncased_L-12_H-256_A-4':\n",
        "        'https://tfhub.dev/tensorflow/small_bert/bert_en_uncased_L-12_H-256_A-4/1',\n",
        "    'small_bert/bert_en_uncased_L-12_H-512_A-8':\n",
        "        'https://tfhub.dev/tensorflow/small_bert/bert_en_uncased_L-12_H-512_A-8/1',\n",
        "    'small_bert/bert_en_uncased_L-12_H-768_A-12':\n",
        "        'https://tfhub.dev/tensorflow/small_bert/bert_en_uncased_L-12_H-768_A-12/1',\n",
        "    'albert_en_base':\n",
        "        'https://tfhub.dev/tensorflow/albert_en_base/2',\n",
        "    'electra_small':\n",
        "        'https://tfhub.dev/google/electra_small/2',\n",
        "    'electra_base':\n",
        "        'https://tfhub.dev/google/electra_base/2',\n",
        "    'experts_pubmed':\n",
        "        'https://tfhub.dev/google/experts/bert/pubmed/2',\n",
        "    'experts_wiki_books':\n",
        "        'https://tfhub.dev/google/experts/bert/wiki_books/2',\n",
        "    'talking-heads_base':\n",
        "        'https://tfhub.dev/tensorflow/talkheads_ggelu_bert_en_base/1',\n",
        "}"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 6,
      "metadata": {
        "id": "qgKHJlcwoghu"
      },
      "outputs": [],
      "source": [
        "map_model_to_preprocess = {\n",
        "    'bert_en_uncased_L-12_H-768_A-12':\n",
        "        'https://tfhub.dev/tensorflow/bert_en_uncased_preprocess/3',\n",
        "    'bert_en_cased_L-12_H-768_A-12':\n",
        "        'https://tfhub.dev/tensorflow/bert_en_cased_preprocess/3',\n",
        "    'small_bert/bert_en_uncased_L-2_H-128_A-2':\n",
        "        'https://tfhub.dev/tensorflow/bert_en_uncased_preprocess/3',\n",
        "    'small_bert/bert_en_uncased_L-2_H-256_A-4':\n",
        "        'https://tfhub.dev/tensorflow/bert_en_uncased_preprocess/3',\n",
        "    'small_bert/bert_en_uncased_L-2_H-512_A-8':\n",
        "        'https://tfhub.dev/tensorflow/bert_en_uncased_preprocess/3',\n",
        "    'small_bert/bert_en_uncased_L-2_H-768_A-12':\n",
        "        'https://tfhub.dev/tensorflow/bert_en_uncased_preprocess/3',\n",
        "    'small_bert/bert_en_uncased_L-4_H-128_A-2':\n",
        "        'https://tfhub.dev/tensorflow/bert_en_uncased_preprocess/3',\n",
        "    'small_bert/bert_en_uncased_L-4_H-256_A-4':\n",
        "        'https://tfhub.dev/tensorflow/bert_en_uncased_preprocess/3',\n",
        "    'small_bert/bert_en_uncased_L-4_H-512_A-8':\n",
        "        'https://tfhub.dev/tensorflow/bert_en_uncased_preprocess/3',\n",
        "    'small_bert/bert_en_uncased_L-4_H-768_A-12':\n",
        "        'https://tfhub.dev/tensorflow/bert_en_uncased_preprocess/3',\n",
        "    'small_bert/bert_en_uncased_L-6_H-128_A-2':\n",
        "        'https://tfhub.dev/tensorflow/bert_en_uncased_preprocess/3',\n",
        "    'small_bert/bert_en_uncased_L-6_H-256_A-4':\n",
        "        'https://tfhub.dev/tensorflow/bert_en_uncased_preprocess/3',\n",
        "    'small_bert/bert_en_uncased_L-6_H-512_A-8':\n",
        "        'https://tfhub.dev/tensorflow/bert_en_uncased_preprocess/3',\n",
        "    'small_bert/bert_en_uncased_L-6_H-768_A-12':\n",
        "        'https://tfhub.dev/tensorflow/bert_en_uncased_preprocess/3',\n",
        "    'small_bert/bert_en_uncased_L-8_H-128_A-2':\n",
        "        'https://tfhub.dev/tensorflow/bert_en_uncased_preprocess/3',\n",
        "    'small_bert/bert_en_uncased_L-8_H-256_A-4':\n",
        "        'https://tfhub.dev/tensorflow/bert_en_uncased_preprocess/3',\n",
        "    'small_bert/bert_en_uncased_L-8_H-512_A-8':\n",
        "        'https://tfhub.dev/tensorflow/bert_en_uncased_preprocess/3',\n",
        "    'small_bert/bert_en_uncased_L-8_H-768_A-12':\n",
        "        'https://tfhub.dev/tensorflow/bert_en_uncased_preprocess/3',\n",
        "    'small_bert/bert_en_uncased_L-10_H-128_A-2':\n",
        "        'https://tfhub.dev/tensorflow/bert_en_uncased_preprocess/3',\n",
        "    'small_bert/bert_en_uncased_L-10_H-256_A-4':\n",
        "        'https://tfhub.dev/tensorflow/bert_en_uncased_preprocess/3',\n",
        "    'small_bert/bert_en_uncased_L-10_H-512_A-8':\n",
        "        'https://tfhub.dev/tensorflow/bert_en_uncased_preprocess/3',\n",
        "    'small_bert/bert_en_uncased_L-10_H-768_A-12':\n",
        "        'https://tfhub.dev/tensorflow/bert_en_uncased_preprocess/3',\n",
        "    'small_bert/bert_en_uncased_L-12_H-128_A-2':\n",
        "        'https://tfhub.dev/tensorflow/bert_en_uncased_preprocess/3',\n",
        "    'small_bert/bert_en_uncased_L-12_H-256_A-4':\n",
        "        'https://tfhub.dev/tensorflow/bert_en_uncased_preprocess/3',\n",
        "    'small_bert/bert_en_uncased_L-12_H-512_A-8':\n",
        "        'https://tfhub.dev/tensorflow/bert_en_uncased_preprocess/3',\n",
        "    'small_bert/bert_en_uncased_L-12_H-768_A-12':\n",
        "        'https://tfhub.dev/tensorflow/bert_en_uncased_preprocess/3',\n",
        "    'bert_multi_cased_L-12_H-768_A-12':\n",
        "        'https://tfhub.dev/tensorflow/bert_multi_cased_preprocess/3',\n",
        "    'albert_en_base':\n",
        "        'https://tfhub.dev/tensorflow/albert_en_preprocess/3',\n",
        "    'electra_small':\n",
        "        'https://tfhub.dev/tensorflow/bert_en_uncased_preprocess/3',\n",
        "    'electra_base':\n",
        "        'https://tfhub.dev/tensorflow/bert_en_uncased_preprocess/3',\n",
        "    'experts_pubmed':\n",
        "        'https://tfhub.dev/tensorflow/bert_en_uncased_preprocess/3',\n",
        "    'experts_wiki_books':\n",
        "        'https://tfhub.dev/tensorflow/bert_en_uncased_preprocess/3',\n",
        "    'talking-heads_base':\n",
        "        'https://tfhub.dev/tensorflow/bert_en_uncased_preprocess/3',\n",
        "}\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 7,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "sSJQttuFo3TF",
        "outputId": "7989ea65-ecbe-426c-a348-05a5da336867"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "BERT model selected           : https://tfhub.dev/google/electra_small/2\n",
            "Preprocess model auto-selected: https://tfhub.dev/tensorflow/bert_en_uncased_preprocess/3\n",
            "Weight for class 0: 0.54\n",
            "Weight for class 1: 7.13\n"
          ]
        }
      ],
      "source": [
        "tfhub_handle_encoder = map_name_to_handle[bert_model_name]\n",
        "tfhub_handle_preprocess = map_model_to_preprocess[bert_model_name]\n",
        "\n",
        "print(f'BERT model selected           : {tfhub_handle_encoder}')\n",
        "print(f'Preprocess model auto-selected: {tfhub_handle_preprocess}')\n",
        "\n",
        "bert_preprocess_model = hub.KerasLayer(tfhub_handle_preprocess)\n",
        "\n",
        "weight_for_0 = (1 / neither)*(total)/2.0 \n",
        "# weight_for_1 = (1 / ofensive)*(total)/3.0\n",
        "weight_for_1 = (1 / hate)*(total)/2.0\n",
        "\n",
        "#defining classweight to give into the models\n",
        "class_weight = {0: weight_for_0, 1: weight_for_1}\n",
        "\n",
        "print('Weight for class 0: {:.2f}'.format(weight_for_0))\n",
        "print('Weight for class 1: {:.2f}'.format(weight_for_1))\n",
        "# print('Weight for class 2: {:.2f}'.format(weight_for_2))"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 25,
      "metadata": {
        "id": "7UA4PZdlpGdw",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "48b52668-d244-46c0-81df-d32c4fd7b40d"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "WARNING:tensorflow:11 out of the last 11 calls to <function recreate_function.<locals>.restored_function_body at 0x7f9da6dfc320> triggered tf.function retracing. Tracing is expensive and the excessive number of tracings could be due to (1) creating @tf.function repeatedly in a loop, (2) passing tensors with different shapes, (3) passing Python objects instead of tensors. For (1), please define your @tf.function outside of the loop. For (2), @tf.function has experimental_relax_shapes=True option that relaxes argument shapes that can avoid unnecessary retracing. For (3), please refer to https://www.tensorflow.org/guide/function#controlling_retracing and https://www.tensorflow.org/api_docs/python/tf/function for  more details.\n",
            "WARNING:tensorflow:11 out of the last 11 calls to <function recreate_function.<locals>.restored_function_body at 0x7f9da6dfc9e0> triggered tf.function retracing. Tracing is expensive and the excessive number of tracings could be due to (1) creating @tf.function repeatedly in a loop, (2) passing tensors with different shapes, (3) passing Python objects instead of tensors. For (1), please define your @tf.function outside of the loop. For (2), @tf.function has experimental_relax_shapes=True option that relaxes argument shapes that can avoid unnecessary retracing. For (3), please refer to https://www.tensorflow.org/guide/function#controlling_retracing and https://www.tensorflow.org/api_docs/python/tf/function for  more details.\n",
            "WARNING:tensorflow:11 out of the last 11 calls to <function recreate_function.<locals>.restored_function_body at 0x7f9e41c83170> triggered tf.function retracing. Tracing is expensive and the excessive number of tracings could be due to (1) creating @tf.function repeatedly in a loop, (2) passing tensors with different shapes, (3) passing Python objects instead of tensors. For (1), please define your @tf.function outside of the loop. For (2), @tf.function has experimental_relax_shapes=True option that relaxes argument shapes that can avoid unnecessary retracing. For (3), please refer to https://www.tensorflow.org/guide/function#controlling_retracing and https://www.tensorflow.org/api_docs/python/tf/function for  more details.\n"
          ]
        }
      ],
      "source": [
        "#Defining Model Architecture\n",
        "def CNN_model():\n",
        "    #Trasnformer Encoding before feeding to CNN\n",
        "    text_in = tf.keras.layers.Input(shape=(), dtype=tf.string, name='text')\n",
        "    preprocess_layer = hub.KerasLayer(tfhub_handle_preprocess, name='preprocessing')\n",
        "    encoder_in = preprocess_layer(text_in)\n",
        "    encode = hub.KerasLayer(tfhub_handle_encoder, trainable=True, name='BERT_encoder')\n",
        "    outputs = encode(encoder_in)\n",
        "    network_layer= sequence_output = outputs[\"sequence_output\"]\n",
        "    # 2 CNNLayers\n",
        "    network_layer = tf.keras.layers.Conv1D(32, (2), activation='relu')(network_layer) #CNN Layer 1\n",
        "    network_layer = tf.keras.layers.Conv1D(64, (2), activation='relu')(network_layer) #CNN Layer 2\n",
        "    network_layer = tf.keras.layers.GlobalMaxPool1D()(network_layer) #Global Maxpooling Layer\n",
        "    network_layer = tf.keras.layers.Dense(256, activation=\"relu\")(network_layer) #DenseLayer \n",
        "    network_layer = tf.keras.layers.Dropout(0.1)(network_layer) #Dropout Layer\n",
        "    network_layer = tf.keras.layers.Dense(3, activation=\"softmax\", name='classifier')(network_layer) #ClassificationLayer\n",
        "    return tf.keras.Model(text_in, network_layer)\n",
        "\n",
        "#building model\n",
        "cnn_model = CNN_model()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 26,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "uDlSSQ9ypK47",
        "outputId": "1908d3a2-ee33-4361-d70b-dd9f6b3cf5f7"
      },
      "outputs": [
        {
          "metadata": {
            "tags": null
          },
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Requirement already satisfied: tf-models-official in /usr/local/lib/python3.7/dist-packages (2.9.2)\n",
            "Requirement already satisfied: gin-config in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (0.5.0)\n",
            "Requirement already satisfied: tensorflow~=2.9.0 in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (2.9.1)\n",
            "Requirement already satisfied: google-api-python-client>=1.6.7 in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (1.12.11)\n",
            "Requirement already satisfied: oauth2client in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (4.1.3)\n",
            "Requirement already satisfied: sacrebleu in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (2.2.0)\n",
            "Requirement already satisfied: pycocotools in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (2.0.4)\n",
            "Requirement already satisfied: pandas>=0.22.0 in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (1.3.5)\n",
            "Requirement already satisfied: six in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (1.15.0)\n",
            "Requirement already satisfied: tensorflow-text~=2.9.0 in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (2.9.0)\n",
            "Requirement already satisfied: tf-slim>=1.1.0 in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (1.1.0)\n",
            "Requirement already satisfied: Cython in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (0.29.32)\n",
            "Requirement already satisfied: kaggle>=1.3.9 in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (1.5.12)\n",
            "Requirement already satisfied: numpy>=1.20 in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (1.21.6)\n",
            "Requirement already satisfied: py-cpuinfo>=3.3.0 in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (8.0.0)\n",
            "Requirement already satisfied: scipy>=0.19.1 in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (1.7.3)\n",
            "Requirement already satisfied: matplotlib in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (3.2.2)\n",
            "Requirement already satisfied: opencv-python-headless in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (4.6.0.66)\n",
            "Requirement already satisfied: sentencepiece in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (0.1.97)\n",
            "Requirement already satisfied: tensorflow-hub>=0.6.0 in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (0.12.0)\n",
            "Requirement already satisfied: psutil>=5.4.3 in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (5.4.8)\n",
            "Requirement already satisfied: tensorflow-datasets in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (4.6.0)\n",
            "Requirement already satisfied: tensorflow-model-optimization>=0.4.1 in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (0.7.3)\n",
            "Requirement already satisfied: seqeval in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (1.2.2)\n",
            "Requirement already satisfied: Pillow in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (7.1.2)\n",
            "Requirement already satisfied: pyyaml<6.0,>=5.1 in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (5.4.1)\n",
            "Requirement already satisfied: tensorflow-addons in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (0.17.1)\n",
            "Requirement already satisfied: httplib2<1dev,>=0.15.0 in /usr/local/lib/python3.7/dist-packages (from google-api-python-client>=1.6.7->tf-models-official) (0.17.4)\n",
            "Requirement already satisfied: uritemplate<4dev,>=3.0.0 in /usr/local/lib/python3.7/dist-packages (from google-api-python-client>=1.6.7->tf-models-official) (3.0.1)\n",
            "Requirement already satisfied: google-auth<3dev,>=1.16.0 in /usr/local/lib/python3.7/dist-packages (from google-api-python-client>=1.6.7->tf-models-official) (1.35.0)\n",
            "Requirement already satisfied: google-api-core<3dev,>=1.21.0 in /usr/local/lib/python3.7/dist-packages (from google-api-python-client>=1.6.7->tf-models-official) (1.31.6)\n",
            "Requirement already satisfied: google-auth-httplib2>=0.0.3 in /usr/local/lib/python3.7/dist-packages (from google-api-python-client>=1.6.7->tf-models-official) (0.0.4)\n",
            "Requirement already satisfied: googleapis-common-protos<2.0dev,>=1.6.0 in /usr/local/lib/python3.7/dist-packages (from google-api-core<3dev,>=1.21.0->google-api-python-client>=1.6.7->tf-models-official) (1.56.4)\n",
            "Requirement already satisfied: pytz in /usr/local/lib/python3.7/dist-packages (from google-api-core<3dev,>=1.21.0->google-api-python-client>=1.6.7->tf-models-official) (2022.1)\n",
            "Requirement already satisfied: setuptools>=40.3.0 in /usr/local/lib/python3.7/dist-packages (from google-api-core<3dev,>=1.21.0->google-api-python-client>=1.6.7->tf-models-official) (57.4.0)\n",
            "Requirement already satisfied: requests<3.0.0dev,>=2.18.0 in /usr/local/lib/python3.7/dist-packages (from google-api-core<3dev,>=1.21.0->google-api-python-client>=1.6.7->tf-models-official) (2.23.0)\n",
            "Requirement already satisfied: protobuf<4.0.0dev,>=3.12.0 in /usr/local/lib/python3.7/dist-packages (from google-api-core<3dev,>=1.21.0->google-api-python-client>=1.6.7->tf-models-official) (3.17.3)\n",
            "Requirement already satisfied: packaging>=14.3 in /usr/local/lib/python3.7/dist-packages (from google-api-core<3dev,>=1.21.0->google-api-python-client>=1.6.7->tf-models-official) (21.3)\n",
            "Requirement already satisfied: rsa<5,>=3.1.4 in /usr/local/lib/python3.7/dist-packages (from google-auth<3dev,>=1.16.0->google-api-python-client>=1.6.7->tf-models-official) (4.9)\n",
            "Requirement already satisfied: cachetools<5.0,>=2.0.0 in /usr/local/lib/python3.7/dist-packages (from google-auth<3dev,>=1.16.0->google-api-python-client>=1.6.7->tf-models-official) (4.2.4)\n",
            "Requirement already satisfied: pyasn1-modules>=0.2.1 in /usr/local/lib/python3.7/dist-packages (from google-auth<3dev,>=1.16.0->google-api-python-client>=1.6.7->tf-models-official) (0.2.8)\n",
            "Requirement already satisfied: python-dateutil in /usr/local/lib/python3.7/dist-packages (from kaggle>=1.3.9->tf-models-official) (2.8.2)\n",
            "Requirement already satisfied: python-slugify in /usr/local/lib/python3.7/dist-packages (from kaggle>=1.3.9->tf-models-official) (6.1.2)\n",
            "Requirement already satisfied: certifi in /usr/local/lib/python3.7/dist-packages (from kaggle>=1.3.9->tf-models-official) (2022.6.15)\n",
            "Requirement already satisfied: tqdm in /usr/local/lib/python3.7/dist-packages (from kaggle>=1.3.9->tf-models-official) (4.64.0)\n",
            "Requirement already satisfied: urllib3 in /usr/local/lib/python3.7/dist-packages (from kaggle>=1.3.9->tf-models-official) (1.24.3)\n",
            "Requirement already satisfied: pyparsing!=3.0.5,>=2.0.2 in /usr/local/lib/python3.7/dist-packages (from packaging>=14.3->google-api-core<3dev,>=1.21.0->google-api-python-client>=1.6.7->tf-models-official) (3.0.9)\n",
            "Requirement already satisfied: pyasn1<0.5.0,>=0.4.6 in /usr/local/lib/python3.7/dist-packages (from pyasn1-modules>=0.2.1->google-auth<3dev,>=1.16.0->google-api-python-client>=1.6.7->tf-models-official) (0.4.8)\n",
            "Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.7/dist-packages (from requests<3.0.0dev,>=2.18.0->google-api-core<3dev,>=1.21.0->google-api-python-client>=1.6.7->tf-models-official) (2.10)\n",
            "Requirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.7/dist-packages (from requests<3.0.0dev,>=2.18.0->google-api-core<3dev,>=1.21.0->google-api-python-client>=1.6.7->tf-models-official) (3.0.4)\n",
            "Requirement already satisfied: tensorflow-estimator<2.10.0,>=2.9.0rc0 in /usr/local/lib/python3.7/dist-packages (from tensorflow~=2.9.0->tf-models-official) (2.9.0)\n",
            "Requirement already satisfied: typing-extensions>=3.6.6 in /usr/local/lib/python3.7/dist-packages (from tensorflow~=2.9.0->tf-models-official) (3.7.4.3)\n",
            "Requirement already satisfied: h5py>=2.9.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow~=2.9.0->tf-models-official) (2.10.0)\n",
            "Requirement already satisfied: opt-einsum>=2.3.2 in /usr/local/lib/python3.7/dist-packages (from tensorflow~=2.9.0->tf-models-official) (3.3.0)\n",
            "Requirement already satisfied: absl-py>=1.0.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow~=2.9.0->tf-models-official) (1.2.0)\n",
            "Requirement already satisfied: wrapt>=1.11.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow~=2.9.0->tf-models-official) (1.12.1)\n",
            "Requirement already satisfied: google-pasta>=0.1.1 in /usr/local/lib/python3.7/dist-packages (from tensorflow~=2.9.0->tf-models-official) (0.2.0)\n",
            "Requirement already satisfied: grpcio<2.0,>=1.24.3 in /usr/local/lib/python3.7/dist-packages (from tensorflow~=2.9.0->tf-models-official) (1.32.0)\n",
            "Requirement already satisfied: astunparse>=1.6.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow~=2.9.0->tf-models-official) (1.6.3)\n",
            "Requirement already satisfied: keras<2.10.0,>=2.9.0rc0 in /usr/local/lib/python3.7/dist-packages (from tensorflow~=2.9.0->tf-models-official) (2.9.0)\n",
            "Requirement already satisfied: flatbuffers<2,>=1.12 in /usr/local/lib/python3.7/dist-packages (from tensorflow~=2.9.0->tf-models-official) (1.12)\n",
            "Requirement already satisfied: tensorflow-io-gcs-filesystem>=0.23.1 in /usr/local/lib/python3.7/dist-packages (from tensorflow~=2.9.0->tf-models-official) (0.26.0)\n",
            "Requirement already satisfied: gast<=0.4.0,>=0.2.1 in /usr/local/lib/python3.7/dist-packages (from tensorflow~=2.9.0->tf-models-official) (0.3.3)\n",
            "Requirement already satisfied: termcolor>=1.1.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow~=2.9.0->tf-models-official) (1.1.0)\n",
            "Requirement already satisfied: libclang>=13.0.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow~=2.9.0->tf-models-official) (14.0.6)\n",
            "Requirement already satisfied: keras-preprocessing>=1.1.1 in /usr/local/lib/python3.7/dist-packages (from tensorflow~=2.9.0->tf-models-official) (1.1.2)\n",
            "Requirement already satisfied: tensorboard<2.10,>=2.9 in /usr/local/lib/python3.7/dist-packages (from tensorflow~=2.9.0->tf-models-official) (2.9.1)\n",
            "Requirement already satisfied: wheel<1.0,>=0.23.0 in /usr/local/lib/python3.7/dist-packages (from astunparse>=1.6.0->tensorflow~=2.9.0->tf-models-official) (0.37.1)\n",
            "Requirement already satisfied: tensorboard-data-server<0.7.0,>=0.6.0 in /usr/local/lib/python3.7/dist-packages (from tensorboard<2.10,>=2.9->tensorflow~=2.9.0->tf-models-official) (0.6.1)\n",
            "Requirement already satisfied: tensorboard-plugin-wit>=1.6.0 in /usr/local/lib/python3.7/dist-packages (from tensorboard<2.10,>=2.9->tensorflow~=2.9.0->tf-models-official) (1.8.1)\n",
            "Requirement already satisfied: werkzeug>=1.0.1 in /usr/local/lib/python3.7/dist-packages (from tensorboard<2.10,>=2.9->tensorflow~=2.9.0->tf-models-official) (1.0.1)\n",
            "Requirement already satisfied: google-auth-oauthlib<0.5,>=0.4.1 in /usr/local/lib/python3.7/dist-packages (from tensorboard<2.10,>=2.9->tensorflow~=2.9.0->tf-models-official) (0.4.6)\n",
            "Requirement already satisfied: markdown>=2.6.8 in /usr/local/lib/python3.7/dist-packages (from tensorboard<2.10,>=2.9->tensorflow~=2.9.0->tf-models-official) (3.4.1)\n",
            "Requirement already satisfied: requests-oauthlib>=0.7.0 in /usr/local/lib/python3.7/dist-packages (from google-auth-oauthlib<0.5,>=0.4.1->tensorboard<2.10,>=2.9->tensorflow~=2.9.0->tf-models-official) (1.3.1)\n",
            "Requirement already satisfied: importlib-metadata>=4.4 in /usr/local/lib/python3.7/dist-packages (from markdown>=2.6.8->tensorboard<2.10,>=2.9->tensorflow~=2.9.0->tf-models-official) (4.12.0)\n",
            "Requirement already satisfied: zipp>=0.5 in /usr/local/lib/python3.7/dist-packages (from importlib-metadata>=4.4->markdown>=2.6.8->tensorboard<2.10,>=2.9->tensorflow~=2.9.0->tf-models-official) (3.8.1)\n",
            "Requirement already satisfied: oauthlib>=3.0.0 in /usr/local/lib/python3.7/dist-packages (from requests-oauthlib>=0.7.0->google-auth-oauthlib<0.5,>=0.4.1->tensorboard<2.10,>=2.9->tensorflow~=2.9.0->tf-models-official) (3.2.0)\n",
            "Requirement already satisfied: dm-tree~=0.1.1 in /usr/local/lib/python3.7/dist-packages (from tensorflow-model-optimization>=0.4.1->tf-models-official) (0.1.7)\n",
            "Requirement already satisfied: cycler>=0.10 in /usr/local/lib/python3.7/dist-packages (from matplotlib->tf-models-official) (0.11.0)\n",
            "Requirement already satisfied: kiwisolver>=1.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->tf-models-official) (1.4.4)\n",
            "Requirement already satisfied: text-unidecode>=1.3 in /usr/local/lib/python3.7/dist-packages (from python-slugify->kaggle>=1.3.9->tf-models-official) (1.3)\n",
            "Requirement already satisfied: portalocker in /usr/local/lib/python3.7/dist-packages (from sacrebleu->tf-models-official) (2.5.1)\n",
            "Requirement already satisfied: tabulate>=0.8.9 in /usr/local/lib/python3.7/dist-packages (from sacrebleu->tf-models-official) (0.8.10)\n",
            "Requirement already satisfied: regex in /usr/local/lib/python3.7/dist-packages (from sacrebleu->tf-models-official) (2022.6.2)\n",
            "Requirement already satisfied: colorama in /usr/local/lib/python3.7/dist-packages (from sacrebleu->tf-models-official) (0.4.5)\n",
            "Requirement already satisfied: lxml in /usr/local/lib/python3.7/dist-packages (from sacrebleu->tf-models-official) (4.9.1)\n",
            "Requirement already satisfied: scikit-learn>=0.21.3 in /usr/local/lib/python3.7/dist-packages (from seqeval->tf-models-official) (1.0.2)\n",
            "Requirement already satisfied: joblib>=0.11 in /usr/local/lib/python3.7/dist-packages (from scikit-learn>=0.21.3->seqeval->tf-models-official) (1.1.0)\n",
            "Requirement already satisfied: threadpoolctl>=2.0.0 in /usr/local/lib/python3.7/dist-packages (from scikit-learn>=0.21.3->seqeval->tf-models-official) (3.1.0)\n",
            "Requirement already satisfied: typeguard>=2.7 in /usr/local/lib/python3.7/dist-packages (from tensorflow-addons->tf-models-official) (2.7.1)\n",
            "Requirement already satisfied: importlib-resources in /usr/local/lib/python3.7/dist-packages (from tensorflow-datasets->tf-models-official) (5.9.0)\n",
            "Requirement already satisfied: dill in /usr/local/lib/python3.7/dist-packages (from tensorflow-datasets->tf-models-official) (0.3.5.1)\n",
            "Requirement already satisfied: tensorflow-metadata in /usr/local/lib/python3.7/dist-packages (from tensorflow-datasets->tf-models-official) (1.9.0)\n",
            "Requirement already satisfied: toml in /usr/local/lib/python3.7/dist-packages (from tensorflow-datasets->tf-models-official) (0.10.2)\n",
            "Requirement already satisfied: promise in /usr/local/lib/python3.7/dist-packages (from tensorflow-datasets->tf-models-official) (2.3)\n",
            "Requirement already satisfied: etils[epath] in /usr/local/lib/python3.7/dist-packages (from tensorflow-datasets->tf-models-official) (0.7.1)\n"
          ]
        },
        {
          "metadata": {
            "tags": null
          },
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "WARNING:tensorflow:AutoGraph could not transform <function Model.make_train_function.<locals>.train_function at 0x7f9be9635cb0> and will run it as-is.\n",
            "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
            "Cause: closure mismatch, requested ('self', 'step_function'), but source function had ()\n",
            "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n"
          ]
        },
        {
          "metadata": {
            "tags": null
          },
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Training model with https://tfhub.dev/google/electra_small/2\n",
            "Epoch 1/20\n",
            "WARNING: AutoGraph could not transform <function Model.make_train_function.<locals>.train_function at 0x7f9be9635cb0> and will run it as-is.\n",
            "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
            "Cause: closure mismatch, requested ('self', 'step_function'), but source function had ()\n",
            "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n",
            "809/809 [==============================] - ETA: 0s - loss: 0.9984 - accuracy: 0.2973"
          ]
        },
        {
          "metadata": {
            "tags": null
          },
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "WARNING:tensorflow:AutoGraph could not transform <function Model.make_test_function.<locals>.test_function at 0x7f9be6b44950> and will run it as-is.\n",
            "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
            "Cause: closure mismatch, requested ('self', 'step_function'), but source function had ()\n",
            "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "WARNING: AutoGraph could not transform <function Model.make_test_function.<locals>.test_function at 0x7f9be6b44950> and will run it as-is.\n",
            "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
            "Cause: closure mismatch, requested ('self', 'step_function'), but source function had ()\n",
            "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n",
            "809/809 [==============================] - 283s 331ms/step - loss: 0.9981 - accuracy: 0.2976 - val_loss: 0.4141 - val_accuracy: 0.8370\n",
            "Epoch 2/20\n",
            "809/809 [==============================] - 266s 329ms/step - loss: 0.3980 - accuracy: 0.9050 - val_loss: 0.2180 - val_accuracy: 0.9253\n",
            "Epoch 3/20\n",
            "809/809 [==============================] - 266s 329ms/step - loss: 0.3215 - accuracy: 0.9529 - val_loss: 0.1824 - val_accuracy: 0.9409\n",
            "Epoch 4/20\n",
            "809/809 [==============================] - 266s 329ms/step - loss: 0.2585 - accuracy: 0.9729 - val_loss: 0.1198 - val_accuracy: 0.9639\n",
            "Epoch 5/20\n",
            "809/809 [==============================] - 267s 329ms/step - loss: 0.1969 - accuracy: 0.9818 - val_loss: 0.1361 - val_accuracy: 0.9645\n",
            "Epoch 6/20\n",
            "809/809 [==============================] - 266s 329ms/step - loss: 0.1330 - accuracy: 0.9881 - val_loss: 0.1552 - val_accuracy: 0.9663\n",
            "Epoch 7/20\n",
            "809/809 [==============================] - 266s 328ms/step - loss: 0.1003 - accuracy: 0.9922 - val_loss: 0.1688 - val_accuracy: 0.9649\n",
            "Epoch 8/20\n",
            "809/809 [==============================] - 266s 329ms/step - loss: 0.0876 - accuracy: 0.9933 - val_loss: 0.2019 - val_accuracy: 0.9625\n",
            "Epoch 9/20\n",
            "809/809 [==============================] - 265s 328ms/step - loss: 0.0712 - accuracy: 0.9952 - val_loss: 0.2178 - val_accuracy: 0.9680\n",
            "Epoch 10/20\n",
            "809/809 [==============================] - 266s 329ms/step - loss: 0.0444 - accuracy: 0.9968 - val_loss: 0.1959 - val_accuracy: 0.9684\n",
            "Epoch 11/20\n",
            "809/809 [==============================] - 266s 328ms/step - loss: 0.0317 - accuracy: 0.9973 - val_loss: 0.2400 - val_accuracy: 0.9642\n",
            "Epoch 12/20\n",
            "809/809 [==============================] - 265s 328ms/step - loss: 0.0262 - accuracy: 0.9981 - val_loss: 0.2595 - val_accuracy: 0.9673\n",
            "Epoch 13/20\n",
            "809/809 [==============================] - 266s 329ms/step - loss: 0.0540 - accuracy: 0.9974 - val_loss: 0.2310 - val_accuracy: 0.9687\n",
            "Epoch 14/20\n",
            "809/809 [==============================] - 266s 329ms/step - loss: 0.0229 - accuracy: 0.9984 - val_loss: 0.2382 - val_accuracy: 0.9698\n",
            "Epoch 15/20\n",
            "809/809 [==============================] - 266s 329ms/step - loss: 0.0289 - accuracy: 0.9981 - val_loss: 0.2509 - val_accuracy: 0.9670\n",
            "Epoch 16/20\n",
            "809/809 [==============================] - 266s 328ms/step - loss: 0.0094 - accuracy: 0.9991 - val_loss: 0.2619 - val_accuracy: 0.9698\n",
            "Epoch 17/20\n",
            "809/809 [==============================] - 266s 329ms/step - loss: 0.0231 - accuracy: 0.9990 - val_loss: 0.2660 - val_accuracy: 0.9687\n",
            "Epoch 18/20\n",
            "809/809 [==============================] - 266s 329ms/step - loss: 0.0315 - accuracy: 0.9989 - val_loss: 0.2476 - val_accuracy: 0.9708\n",
            "Epoch 19/20\n",
            "809/809 [==============================] - 267s 330ms/step - loss: 0.0094 - accuracy: 0.9993 - val_loss: 0.2548 - val_accuracy: 0.9705\n",
            "Epoch 20/20\n",
            "809/809 [==============================] - 266s 329ms/step - loss: 0.0099 - accuracy: 0.9992 - val_loss: 0.2543 - val_accuracy: 0.9712\n"
          ]
        }
      ],
      "source": [
        "#defining loss function\n",
        "loss = tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True)\n",
        "\n",
        "#importing latest tfmodel here to avoid package conflicts\n",
        "!pip install tf-models-official\n",
        "from official.nlp import optimization\n",
        "\n",
        "#fine tuning and Training models\n",
        "epochs_count = 20#defining_epochs_count\n",
        "steps_each_epoch = tf.data.experimental.cardinality(train_ds).numpy()#steps_per_epoch\n",
        "number_of_training_steps = steps_each_epoch * epochs_count#defining_Training_steps\n",
        "number_of_warmup_steps = int(0.1*number_of_training_steps)#defining_warmup_steps\n",
        "init_lr = 3e-5 #deifning learning rate\n",
        "\n",
        "#creating_optimiser\n",
        "optimizer = optimization.create_optimizer(init_lr=init_lr,\n",
        "                                          num_train_steps=number_of_training_steps,\n",
        "                                          num_warmup_steps=number_of_warmup_steps,\n",
        "                                          optimizer_type='adamw')\n",
        "#model compiling\n",
        "cnn_model.compile(optimizer=optimizer,\n",
        "                          loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),\n",
        "                          metrics=tf.keras.metrics.SparseCategoricalAccuracy('accuracy'))\n",
        "\n",
        "\n",
        "print(f'Training model with {tfhub_handle_encoder}')\n",
        "#training_model\n",
        "cnn_history = cnn_model.fit(x=train_ds,\n",
        "                            validation_data=val_ds,\n",
        "                            epochs=epochs_count,\n",
        "                            class_weight=class_weight)\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 27,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 657
        },
        "id": "DBDQzvumpWwK",
        "outputId": "ef04ea2b-30f9-41fb-ffbd-ad3a6e0947b9"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "dict_keys(['loss', 'accuracy', 'val_loss', 'val_accuracy'])\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.legend.Legend at 0x7f9be64d6410>"
            ]
          },
          "metadata": {},
          "execution_count": 27
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 864x720 with 2 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ],
      "source": [
        "#creating graphs for accuracy and loss\n",
        "#fining_values\n",
        "history_dict = cnn_history.history\n",
        "print(history_dict.keys())\n",
        "acc = history_dict['accuracy']\n",
        "val_acc = history_dict['val_accuracy']\n",
        "loss = history_dict['loss']\n",
        "val_loss = history_dict['val_loss']\n",
        "epochs = range(1, len(acc) + 1)\n",
        "\n",
        "#defining_figures\n",
        "fig = plt.figure(figsize=(12, 10))\n",
        "fig.tight_layout()\n",
        "#1st_figure\n",
        "plt.subplot(2, 1, 1)\n",
        "plt.plot(epochs, loss, 'r', label='Training-loss')#training_loss\n",
        "plt.plot(epochs, val_loss, 'b', label='Validation-loss')#validation_loss\n",
        "plt.title('Training & validation Loss')\n",
        "plt.xlabel('Epochs-Count')\n",
        "plt.ylabel('Loss')\n",
        "plt.legend()\n",
        "#2nd_Figure\n",
        "plt.subplot(2, 1, 2)\n",
        "plt.plot(epochs, acc, 'r', label='Training-accuracy')#training_accuracy\n",
        "plt.plot(epochs, val_acc, 'b', label='Validation-accuracy')#validation_accuracy\n",
        "plt.title('Training & Validation Accuracy')\n",
        "plt.xlabel('Epochs-Count')\n",
        "plt.ylabel('Acc')\n",
        "plt.legend(loc='lower right')\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 28,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "_Xu9MJTvpbDM",
        "outputId": "b0ef23c4-8861-4f56-d72f-8b6b720d7106"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "WARNING:tensorflow:AutoGraph could not transform <function Model.make_predict_function.<locals>.predict_function at 0x7f9be6423680> and will run it as-is.\n",
            "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
            "Cause: closure mismatch, requested ('self', 'step_function'), but source function had ()\n",
            "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "WARNING: AutoGraph could not transform <function Model.make_predict_function.<locals>.predict_function at 0x7f9be6423680> and will run it as-is.\n",
            "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
            "Cause: closure mismatch, requested ('self', 'step_function'), but source function had ()\n",
            "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "WARNING:tensorflow:11 out of the last 11 calls to <function recreate_function.<locals>.restored_function_body at 0x7f9e412e5e60> triggered tf.function retracing. Tracing is expensive and the excessive number of tracings could be due to (1) creating @tf.function repeatedly in a loop, (2) passing tensors with different shapes, (3) passing Python objects instead of tensors. For (1), please define your @tf.function outside of the loop. For (2), @tf.function has experimental_relax_shapes=True option that relaxes argument shapes that can avoid unnecessary retracing. For (3), please refer to https://www.tensorflow.org/guide/function#controlling_retracing and https://www.tensorflow.org/api_docs/python/tf/function for  more details.\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       0.98      0.99      0.98      2973\n",
            "           1       0.82      0.76      0.79       224\n",
            "\n",
            "    accuracy                           0.97      3197\n",
            "   macro avg       0.90      0.88      0.89      3197\n",
            "weighted avg       0.97      0.97      0.97      3197\n",
            "\n",
            "[[2935   38]\n",
            " [  53  171]]\n"
          ]
        }
      ],
      "source": [
        "#predicting on test dataset\n",
        "text = []\n",
        "test_labels = []\n",
        "predictions = []\n",
        "for text, label in test_ds.take(-1):\n",
        "  text = text.numpy()#reading_text\n",
        "  test_labels.append(label.numpy())#reading_text_label\n",
        "  predictions.append(cnn_model.predict(text))#predicting\n",
        "\n",
        "from itertools import chain\n",
        "flatten_list = list(chain.from_iterable(predictions))\n",
        "\n",
        "y_pred = np.argmax(flatten_list, axis=-1)#predicted_test_labels\n",
        "y_test = np.array(list(chain.from_iterable(test_labels)))#true_test_labels\n",
        "\n",
        "print(classification_report(y_test, y_pred))#priting_classfication_report\n",
        "print(confusion_matrix(y_test, y_pred)) #priting_confusion_matrix\n"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "user_input = [\"I cannot tolerate how these bitches conspire against each other\", \"You are welcome to come to my country\",\"I hate black people\"]\n",
        "user_input_prediction = cnn_model.predict(user_input)\n",
        "\n",
        "result = []\n",
        "for values in user_input_prediction:\n",
        "   max = -1\n",
        "   type_comment = -1\n",
        "   for index,value in enumerate(values): \n",
        "      if (value > max):\n",
        "         max = value\n",
        "         type_comment = index\n",
        "   result.append(type_comment) \n",
        "for speech, output in zip(user_input, result):\n",
        "   print(f'{speech[:100]} -> {output}')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "zpPVs8c3SPOr",
        "outputId": "09f48adc-7c88-4c83-b115-22191c5974d1"
      },
      "execution_count": 29,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "I cannot tolerate how these bitches conspire against each other -> 0\n",
            "You are welcome to come to my country -> 0\n",
            "I hate black people -> 1\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 8,
      "metadata": {
        "id": "WFGQ-5gVrfF4"
      },
      "outputs": [],
      "source": [
        "#building MLP archictecture\n",
        "def create_MLP_model(output_bias=None):\n",
        "    if output_bias is not None:\n",
        "        output_bias = tf.keras.initializers.Constant(output_bias)\n",
        "    #input layer    \n",
        "    text_input = tf.keras.layers.Input(shape=(), dtype=tf.string, name='text')\n",
        "    #preprocessing layer \n",
        "    preprocessing_layer = hub.KerasLayer(tfhub_handle_preprocess, name='preprocessing')\n",
        "    #giving input to bert model from the pre processed layer\n",
        "    encoder_inputs = preprocessing_layer(text_input)\n",
        "    #selecting the encoder\n",
        "    encoder = hub.KerasLayer(tfhub_handle_encoder, trainable=True, name='BERT_encoder')\n",
        "    #collecting the output of the bert encoder\n",
        "    outputs = encoder(encoder_inputs)\n",
        "    #collecting the pooled outputs\n",
        "    net = outputs['pooled_output']\n",
        "    #feeding the output to the multilayer perception\n",
        "    #1 DENSE LAYER\n",
        "    net = tf.keras.layers.Dense(512, activation=\"relu\")(net)\n",
        "    #1 DROPUT LAYER\n",
        "    net = tf.keras.layers.Dropout(0.2)(net)\n",
        "    #1 OUTPUT LAYER WITH SOFTMAX\n",
        "    net = tf.keras.layers.Dense(3, activation=\"softmax\", name='classifier', bias_initializer=output_bias)(net)\n",
        "    return tf.keras.Model(text_input, net)\n",
        "\n",
        "mlp_model = create_MLP_model()\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 9,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "H2GEwY4TrlsF",
        "outputId": "22a589b8-d52a-4b76-ad7c-838e8d8c2e85"
      },
      "outputs": [
        {
          "metadata": {
            "tags": null
          },
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Collecting tf-models-official\n",
            "  Downloading tf_models_official-2.9.2-py2.py3-none-any.whl (2.1 MB)\n",
            "\u001b[K     |████████████████████████████████| 2.1 MB 7.4 MB/s \n",
            "\u001b[?25hRequirement already satisfied: pycocotools in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (2.0.4)\n",
            "Requirement already satisfied: psutil>=5.4.3 in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (5.4.8)\n",
            "Collecting tensorflow-addons\n",
            "  Downloading tensorflow_addons-0.17.1-cp37-cp37m-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (1.1 MB)\n",
            "\u001b[K     |████████████████████████████████| 1.1 MB 36.5 MB/s \n",
            "\u001b[?25hRequirement already satisfied: tensorflow-datasets in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (4.6.0)\n",
            "Requirement already satisfied: kaggle>=1.3.9 in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (1.5.12)\n",
            "Requirement already satisfied: gin-config in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (0.5.0)\n",
            "Collecting py-cpuinfo>=3.3.0\n",
            "  Downloading py-cpuinfo-8.0.0.tar.gz (99 kB)\n",
            "\u001b[K     |████████████████████████████████| 99 kB 11.8 MB/s \n",
            "\u001b[?25hCollecting sacrebleu\n",
            "  Downloading sacrebleu-2.2.0-py3-none-any.whl (116 kB)\n",
            "\u001b[K     |████████████████████████████████| 116 kB 62.2 MB/s \n",
            "\u001b[?25hRequirement already satisfied: opencv-python-headless in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (4.6.0.66)\n",
            "Requirement already satisfied: tensorflow-hub>=0.6.0 in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (0.12.0)\n",
            "Requirement already satisfied: Cython in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (0.29.32)\n",
            "Collecting tensorflow-model-optimization>=0.4.1\n",
            "  Downloading tensorflow_model_optimization-0.7.3-py2.py3-none-any.whl (238 kB)\n",
            "\u001b[K     |████████████████████████████████| 238 kB 71.8 MB/s \n",
            "\u001b[?25hCollecting seqeval\n",
            "  Downloading seqeval-1.2.2.tar.gz (43 kB)\n",
            "\u001b[K     |████████████████████████████████| 43 kB 2.3 MB/s \n",
            "\u001b[?25hRequirement already satisfied: pandas>=0.22.0 in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (1.3.5)\n",
            "Collecting numpy>=1.20\n",
            "  Downloading numpy-1.21.6-cp37-cp37m-manylinux_2_12_x86_64.manylinux2010_x86_64.whl (15.7 MB)\n",
            "\u001b[K     |████████████████████████████████| 15.7 MB 44.9 MB/s \n",
            "\u001b[?25hCollecting pyyaml<6.0,>=5.1\n",
            "  Downloading PyYAML-5.4.1-cp37-cp37m-manylinux1_x86_64.whl (636 kB)\n",
            "\u001b[K     |████████████████████████████████| 636 kB 69.9 MB/s \n",
            "\u001b[?25hRequirement already satisfied: scipy>=0.19.1 in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (1.7.3)\n",
            "Collecting tensorflow-text~=2.9.0\n",
            "  Downloading tensorflow_text-2.9.0-cp37-cp37m-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (4.6 MB)\n",
            "\u001b[K     |████████████████████████████████| 4.6 MB 45.6 MB/s \n",
            "\u001b[?25hCollecting tf-slim>=1.1.0\n",
            "  Downloading tf_slim-1.1.0-py2.py3-none-any.whl (352 kB)\n",
            "\u001b[K     |████████████████████████████████| 352 kB 77.8 MB/s \n",
            "\u001b[?25hCollecting tensorflow~=2.9.0\n",
            "  Downloading tensorflow-2.9.1-cp37-cp37m-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (511.7 MB)\n",
            "\u001b[K     |████████████████████████████████| 511.7 MB 6.4 kB/s \n",
            "\u001b[?25hRequirement already satisfied: Pillow in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (7.1.2)\n",
            "Collecting sentencepiece\n",
            "  Downloading sentencepiece-0.1.97-cp37-cp37m-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (1.3 MB)\n",
            "\u001b[K     |████████████████████████████████| 1.3 MB 51.5 MB/s \n",
            "\u001b[?25hRequirement already satisfied: six in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (1.15.0)\n",
            "Requirement already satisfied: oauth2client in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (4.1.3)\n",
            "Requirement already satisfied: google-api-python-client>=1.6.7 in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (1.12.11)\n",
            "Requirement already satisfied: matplotlib in /usr/local/lib/python3.7/dist-packages (from tf-models-official) (3.2.2)\n",
            "Requirement already satisfied: google-auth-httplib2>=0.0.3 in /usr/local/lib/python3.7/dist-packages (from google-api-python-client>=1.6.7->tf-models-official) (0.0.4)\n",
            "Requirement already satisfied: httplib2<1dev,>=0.15.0 in /usr/local/lib/python3.7/dist-packages (from google-api-python-client>=1.6.7->tf-models-official) (0.17.4)\n",
            "Requirement already satisfied: google-api-core<3dev,>=1.21.0 in /usr/local/lib/python3.7/dist-packages (from google-api-python-client>=1.6.7->tf-models-official) (1.31.6)\n",
            "Requirement already satisfied: google-auth<3dev,>=1.16.0 in /usr/local/lib/python3.7/dist-packages (from google-api-python-client>=1.6.7->tf-models-official) (1.35.0)\n",
            "Requirement already satisfied: uritemplate<4dev,>=3.0.0 in /usr/local/lib/python3.7/dist-packages (from google-api-python-client>=1.6.7->tf-models-official) (3.0.1)\n",
            "Requirement already satisfied: requests<3.0.0dev,>=2.18.0 in /usr/local/lib/python3.7/dist-packages (from google-api-core<3dev,>=1.21.0->google-api-python-client>=1.6.7->tf-models-official) (2.23.0)\n",
            "Requirement already satisfied: protobuf<4.0.0dev,>=3.12.0 in /usr/local/lib/python3.7/dist-packages (from google-api-core<3dev,>=1.21.0->google-api-python-client>=1.6.7->tf-models-official) (3.17.3)\n",
            "Requirement already satisfied: pytz in /usr/local/lib/python3.7/dist-packages (from google-api-core<3dev,>=1.21.0->google-api-python-client>=1.6.7->tf-models-official) (2022.1)\n",
            "Requirement already satisfied: packaging>=14.3 in /usr/local/lib/python3.7/dist-packages (from google-api-core<3dev,>=1.21.0->google-api-python-client>=1.6.7->tf-models-official) (21.3)\n",
            "Requirement already satisfied: setuptools>=40.3.0 in /usr/local/lib/python3.7/dist-packages (from google-api-core<3dev,>=1.21.0->google-api-python-client>=1.6.7->tf-models-official) (57.4.0)\n",
            "Requirement already satisfied: googleapis-common-protos<2.0dev,>=1.6.0 in /usr/local/lib/python3.7/dist-packages (from google-api-core<3dev,>=1.21.0->google-api-python-client>=1.6.7->tf-models-official) (1.56.4)\n",
            "Requirement already satisfied: cachetools<5.0,>=2.0.0 in /usr/local/lib/python3.7/dist-packages (from google-auth<3dev,>=1.16.0->google-api-python-client>=1.6.7->tf-models-official) (4.2.4)\n",
            "Requirement already satisfied: pyasn1-modules>=0.2.1 in /usr/local/lib/python3.7/dist-packages (from google-auth<3dev,>=1.16.0->google-api-python-client>=1.6.7->tf-models-official) (0.2.8)\n",
            "Requirement already satisfied: rsa<5,>=3.1.4 in /usr/local/lib/python3.7/dist-packages (from google-auth<3dev,>=1.16.0->google-api-python-client>=1.6.7->tf-models-official) (4.9)\n",
            "Requirement already satisfied: tqdm in /usr/local/lib/python3.7/dist-packages (from kaggle>=1.3.9->tf-models-official) (4.64.0)\n",
            "Requirement already satisfied: python-dateutil in /usr/local/lib/python3.7/dist-packages (from kaggle>=1.3.9->tf-models-official) (2.8.2)\n",
            "Requirement already satisfied: certifi in /usr/local/lib/python3.7/dist-packages (from kaggle>=1.3.9->tf-models-official) (2022.6.15)\n",
            "Requirement already satisfied: python-slugify in /usr/local/lib/python3.7/dist-packages (from kaggle>=1.3.9->tf-models-official) (6.1.2)\n",
            "Requirement already satisfied: urllib3 in /usr/local/lib/python3.7/dist-packages (from kaggle>=1.3.9->tf-models-official) (1.24.3)\n",
            "Requirement already satisfied: pyparsing!=3.0.5,>=2.0.2 in /usr/local/lib/python3.7/dist-packages (from packaging>=14.3->google-api-core<3dev,>=1.21.0->google-api-python-client>=1.6.7->tf-models-official) (3.0.9)\n",
            "Requirement already satisfied: pyasn1<0.5.0,>=0.4.6 in /usr/local/lib/python3.7/dist-packages (from pyasn1-modules>=0.2.1->google-auth<3dev,>=1.16.0->google-api-python-client>=1.6.7->tf-models-official) (0.4.8)\n",
            "Requirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.7/dist-packages (from requests<3.0.0dev,>=2.18.0->google-api-core<3dev,>=1.21.0->google-api-python-client>=1.6.7->tf-models-official) (3.0.4)\n",
            "Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.7/dist-packages (from requests<3.0.0dev,>=2.18.0->google-api-core<3dev,>=1.21.0->google-api-python-client>=1.6.7->tf-models-official) (2.10)\n",
            "Requirement already satisfied: wrapt>=1.11.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow~=2.9.0->tf-models-official) (1.12.1)\n",
            "Requirement already satisfied: astunparse>=1.6.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow~=2.9.0->tf-models-official) (1.6.3)\n",
            "Requirement already satisfied: gast<=0.4.0,>=0.2.1 in /usr/local/lib/python3.7/dist-packages (from tensorflow~=2.9.0->tf-models-official) (0.3.3)\n",
            "Collecting tensorflow-estimator<2.10.0,>=2.9.0rc0\n",
            "  Downloading tensorflow_estimator-2.9.0-py2.py3-none-any.whl (438 kB)\n",
            "\u001b[K     |████████████████████████████████| 438 kB 73.2 MB/s \n",
            "\u001b[?25hRequirement already satisfied: grpcio<2.0,>=1.24.3 in /usr/local/lib/python3.7/dist-packages (from tensorflow~=2.9.0->tf-models-official) (1.32.0)\n",
            "Requirement already satisfied: termcolor>=1.1.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow~=2.9.0->tf-models-official) (1.1.0)\n",
            "Requirement already satisfied: typing-extensions>=3.6.6 in /usr/local/lib/python3.7/dist-packages (from tensorflow~=2.9.0->tf-models-official) (3.7.4.3)\n",
            "Collecting keras<2.10.0,>=2.9.0rc0\n",
            "  Downloading keras-2.9.0-py2.py3-none-any.whl (1.6 MB)\n",
            "\u001b[K     |████████████████████████████████| 1.6 MB 50.9 MB/s \n",
            "\u001b[?25hCollecting tensorboard<2.10,>=2.9\n",
            "  Downloading tensorboard-2.9.1-py3-none-any.whl (5.8 MB)\n",
            "\u001b[K     |████████████████████████████████| 5.8 MB 56.9 MB/s \n",
            "\u001b[?25hRequirement already satisfied: h5py>=2.9.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow~=2.9.0->tf-models-official) (2.10.0)\n",
            "Requirement already satisfied: google-pasta>=0.1.1 in /usr/local/lib/python3.7/dist-packages (from tensorflow~=2.9.0->tf-models-official) (0.2.0)\n",
            "Requirement already satisfied: flatbuffers<2,>=1.12 in /usr/local/lib/python3.7/dist-packages (from tensorflow~=2.9.0->tf-models-official) (1.12)\n",
            "Collecting absl-py>=1.0.0\n",
            "  Downloading absl_py-1.2.0-py3-none-any.whl (123 kB)\n",
            "\u001b[K     |████████████████████████████████| 123 kB 68.8 MB/s \n",
            "\u001b[?25hRequirement already satisfied: tensorflow-io-gcs-filesystem>=0.23.1 in /usr/local/lib/python3.7/dist-packages (from tensorflow~=2.9.0->tf-models-official) (0.26.0)\n",
            "Requirement already satisfied: opt-einsum>=2.3.2 in /usr/local/lib/python3.7/dist-packages (from tensorflow~=2.9.0->tf-models-official) (3.3.0)\n",
            "Requirement already satisfied: libclang>=13.0.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow~=2.9.0->tf-models-official) (14.0.6)\n",
            "Requirement already satisfied: keras-preprocessing>=1.1.1 in /usr/local/lib/python3.7/dist-packages (from tensorflow~=2.9.0->tf-models-official) (1.1.2)\n",
            "Requirement already satisfied: wheel<1.0,>=0.23.0 in /usr/local/lib/python3.7/dist-packages (from astunparse>=1.6.0->tensorflow~=2.9.0->tf-models-official) (0.37.1)\n",
            "Requirement already satisfied: werkzeug>=1.0.1 in /usr/local/lib/python3.7/dist-packages (from tensorboard<2.10,>=2.9->tensorflow~=2.9.0->tf-models-official) (1.0.1)\n",
            "Requirement already satisfied: tensorboard-data-server<0.7.0,>=0.6.0 in /usr/local/lib/python3.7/dist-packages (from tensorboard<2.10,>=2.9->tensorflow~=2.9.0->tf-models-official) (0.6.1)\n",
            "Requirement already satisfied: tensorboard-plugin-wit>=1.6.0 in /usr/local/lib/python3.7/dist-packages (from tensorboard<2.10,>=2.9->tensorflow~=2.9.0->tf-models-official) (1.8.1)\n",
            "Requirement already satisfied: markdown>=2.6.8 in /usr/local/lib/python3.7/dist-packages (from tensorboard<2.10,>=2.9->tensorflow~=2.9.0->tf-models-official) (3.4.1)\n",
            "Requirement already satisfied: google-auth-oauthlib<0.5,>=0.4.1 in /usr/local/lib/python3.7/dist-packages (from tensorboard<2.10,>=2.9->tensorflow~=2.9.0->tf-models-official) (0.4.6)\n",
            "Requirement already satisfied: requests-oauthlib>=0.7.0 in /usr/local/lib/python3.7/dist-packages (from google-auth-oauthlib<0.5,>=0.4.1->tensorboard<2.10,>=2.9->tensorflow~=2.9.0->tf-models-official) (1.3.1)\n",
            "Requirement already satisfied: importlib-metadata>=4.4 in /usr/local/lib/python3.7/dist-packages (from markdown>=2.6.8->tensorboard<2.10,>=2.9->tensorflow~=2.9.0->tf-models-official) (4.12.0)\n",
            "Requirement already satisfied: zipp>=0.5 in /usr/local/lib/python3.7/dist-packages (from importlib-metadata>=4.4->markdown>=2.6.8->tensorboard<2.10,>=2.9->tensorflow~=2.9.0->tf-models-official) (3.8.1)\n",
            "Requirement already satisfied: oauthlib>=3.0.0 in /usr/local/lib/python3.7/dist-packages (from requests-oauthlib>=0.7.0->google-auth-oauthlib<0.5,>=0.4.1->tensorboard<2.10,>=2.9->tensorflow~=2.9.0->tf-models-official) (3.2.0)\n",
            "Requirement already satisfied: dm-tree~=0.1.1 in /usr/local/lib/python3.7/dist-packages (from tensorflow-model-optimization>=0.4.1->tf-models-official) (0.1.7)\n",
            "Requirement already satisfied: cycler>=0.10 in /usr/local/lib/python3.7/dist-packages (from matplotlib->tf-models-official) (0.11.0)\n",
            "Requirement already satisfied: kiwisolver>=1.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->tf-models-official) (1.4.4)\n",
            "Requirement already satisfied: text-unidecode>=1.3 in /usr/local/lib/python3.7/dist-packages (from python-slugify->kaggle>=1.3.9->tf-models-official) (1.3)\n",
            "Collecting portalocker\n",
            "  Downloading portalocker-2.5.1-py2.py3-none-any.whl (15 kB)\n",
            "Requirement already satisfied: tabulate>=0.8.9 in /usr/local/lib/python3.7/dist-packages (from sacrebleu->tf-models-official) (0.8.10)\n",
            "Requirement already satisfied: regex in /usr/local/lib/python3.7/dist-packages (from sacrebleu->tf-models-official) (2022.6.2)\n",
            "Collecting colorama\n",
            "  Downloading colorama-0.4.5-py2.py3-none-any.whl (16 kB)\n",
            "Requirement already satisfied: lxml in /usr/local/lib/python3.7/dist-packages (from sacrebleu->tf-models-official) (4.9.1)\n",
            "Requirement already satisfied: scikit-learn>=0.21.3 in /usr/local/lib/python3.7/dist-packages (from seqeval->tf-models-official) (1.0.2)\n",
            "Requirement already satisfied: joblib>=0.11 in /usr/local/lib/python3.7/dist-packages (from scikit-learn>=0.21.3->seqeval->tf-models-official) (1.1.0)\n",
            "Requirement already satisfied: threadpoolctl>=2.0.0 in /usr/local/lib/python3.7/dist-packages (from scikit-learn>=0.21.3->seqeval->tf-models-official) (3.1.0)\n",
            "Requirement already satisfied: typeguard>=2.7 in /usr/local/lib/python3.7/dist-packages (from tensorflow-addons->tf-models-official) (2.7.1)\n",
            "Requirement already satisfied: dill in /usr/local/lib/python3.7/dist-packages (from tensorflow-datasets->tf-models-official) (0.3.5.1)\n",
            "Requirement already satisfied: etils[epath] in /usr/local/lib/python3.7/dist-packages (from tensorflow-datasets->tf-models-official) (0.7.1)\n",
            "Requirement already satisfied: importlib-resources in /usr/local/lib/python3.7/dist-packages (from tensorflow-datasets->tf-models-official) (5.9.0)\n",
            "Requirement already satisfied: toml in /usr/local/lib/python3.7/dist-packages (from tensorflow-datasets->tf-models-official) (0.10.2)\n",
            "Requirement already satisfied: tensorflow-metadata in /usr/local/lib/python3.7/dist-packages (from tensorflow-datasets->tf-models-official) (1.9.0)\n",
            "Requirement already satisfied: promise in /usr/local/lib/python3.7/dist-packages (from tensorflow-datasets->tf-models-official) (2.3)\n",
            "Building wheels for collected packages: py-cpuinfo, seqeval\n",
            "  Building wheel for py-cpuinfo (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "  Created wheel for py-cpuinfo: filename=py_cpuinfo-8.0.0-py3-none-any.whl size=22257 sha256=d61586f74b11cce096d67653f263eb4b889f7a5b5714ab7552c2cd3083c3cb65\n",
            "  Stored in directory: /root/.cache/pip/wheels/d2/f1/1f/041add21dc9c4220157f1bd2bd6afe1f1a49524c3396b94401\n",
            "  Building wheel for seqeval (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "  Created wheel for seqeval: filename=seqeval-1.2.2-py3-none-any.whl size=16180 sha256=e09e93b8c280bf9256dd3022b9f2778ee51835666b83085accc7d2a079519605\n",
            "  Stored in directory: /root/.cache/pip/wheels/05/96/ee/7cac4e74f3b19e3158dce26a20a1c86b3533c43ec72a549fd7\n",
            "Successfully built py-cpuinfo seqeval\n",
            "Installing collected packages: numpy, absl-py, tensorflow-estimator, tensorboard, keras, tensorflow, portalocker, colorama, tf-slim, tensorflow-text, tensorflow-model-optimization, tensorflow-addons, seqeval, sentencepiece, sacrebleu, pyyaml, py-cpuinfo, tf-models-official\n",
            "  Attempting uninstall: numpy\n",
            "    Found existing installation: numpy 1.19.5\n",
            "    Uninstalling numpy-1.19.5:\n",
            "      Successfully uninstalled numpy-1.19.5\n",
            "  Attempting uninstall: absl-py\n",
            "    Found existing installation: absl-py 0.15.0\n",
            "    Uninstalling absl-py-0.15.0:\n",
            "      Successfully uninstalled absl-py-0.15.0\n",
            "  Attempting uninstall: tensorflow-estimator\n",
            "    Found existing installation: tensorflow-estimator 2.4.0\n",
            "    Uninstalling tensorflow-estimator-2.4.0:\n",
            "      Successfully uninstalled tensorflow-estimator-2.4.0\n",
            "  Attempting uninstall: tensorboard\n",
            "    Found existing installation: tensorboard 2.8.0\n",
            "    Uninstalling tensorboard-2.8.0:\n",
            "      Successfully uninstalled tensorboard-2.8.0\n",
            "  Attempting uninstall: keras\n",
            "    Found existing installation: keras 2.8.0\n",
            "    Uninstalling keras-2.8.0:\n",
            "      Successfully uninstalled keras-2.8.0\n",
            "  Attempting uninstall: tensorflow\n",
            "    Found existing installation: tensorflow 2.4.4\n",
            "    Uninstalling tensorflow-2.4.4:\n",
            "      Successfully uninstalled tensorflow-2.4.4\n",
            "  Attempting uninstall: tensorflow-text\n",
            "    Found existing installation: tensorflow-text 2.4.1\n",
            "    Uninstalling tensorflow-text-2.4.1:\n",
            "      Successfully uninstalled tensorflow-text-2.4.1\n",
            "  Attempting uninstall: pyyaml\n",
            "    Found existing installation: PyYAML 6.0\n",
            "    Uninstalling PyYAML-6.0:\n",
            "      Successfully uninstalled PyYAML-6.0\n",
            "Successfully installed absl-py-1.2.0 colorama-0.4.5 keras-2.9.0 numpy-1.21.6 portalocker-2.5.1 py-cpuinfo-8.0.0 pyyaml-5.4.1 sacrebleu-2.2.0 sentencepiece-0.1.97 seqeval-1.2.2 tensorboard-2.9.1 tensorflow-2.9.1 tensorflow-addons-0.17.1 tensorflow-estimator-2.9.0 tensorflow-model-optimization-0.7.3 tensorflow-text-2.9.0 tf-models-official-2.9.2 tf-slim-1.1.0\n"
          ]
        },
        {
          "data": {
            "application/vnd.colab-display-data+json": {
              "pip_warning": {
                "packages": [
                  "absl",
                  "numpy",
                  "tensorboard",
                  "tensorflow",
                  "tensorflow_estimator",
                  "tensorflow_text"
                ]
              }
            }
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "metadata": {
            "tags": null
          },
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/tensorflow_addons/utils/ensure_tf_install.py:67: UserWarning: Tensorflow Addons supports using Python ops for all Tensorflow versions above or equal to 2.7.0 and strictly below 2.10.0 (nightly versions are not supported). \n",
            " The versions of TensorFlow you are currently using is 2.4.4 and is not supported. \n",
            "Some things might work, some things might not.\n",
            "If you were to encounter a bug, do not file an issue.\n",
            "If you want to make sure you're using a tested and supported configuration, either change the TensorFlow version or the TensorFlow Addons's version. \n",
            "You can find the compatibility matrix in TensorFlow Addon's readme:\n",
            "https://github.com/tensorflow/addons\n",
            "  UserWarning,\n",
            "WARNING:tensorflow:AutoGraph could not transform <function Model.make_train_function.<locals>.train_function at 0x7f090a2d8e60> and will run it as-is.\n",
            "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
            "Cause: closure mismatch, requested ('self', 'step_function'), but source function had ()\n",
            "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n"
          ]
        },
        {
          "metadata": {
            "tags": null
          },
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Training model with https://tfhub.dev/google/electra_small/2\n",
            "Epoch 1/20\n",
            "WARNING: AutoGraph could not transform <function Model.make_train_function.<locals>.train_function at 0x7f090a2d8e60> and will run it as-is.\n",
            "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
            "Cause: closure mismatch, requested ('self', 'step_function'), but source function had ()\n",
            "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n",
            "809/809 [==============================] - ETA: 0s - loss: 0.8423 - accuracy: 0.8224"
          ]
        },
        {
          "metadata": {
            "tags": null
          },
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "WARNING:tensorflow:AutoGraph could not transform <function Model.make_test_function.<locals>.test_function at 0x7f08fe0c15f0> and will run it as-is.\n",
            "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
            "Cause: closure mismatch, requested ('self', 'step_function'), but source function had ()\n",
            "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "WARNING: AutoGraph could not transform <function Model.make_test_function.<locals>.test_function at 0x7f08fe0c15f0> and will run it as-is.\n",
            "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
            "Cause: closure mismatch, requested ('self', 'step_function'), but source function had ()\n",
            "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n",
            "809/809 [==============================] - 279s 325ms/step - loss: 0.8421 - accuracy: 0.8224 - val_loss: 0.1639 - val_accuracy: 0.9444\n",
            "Epoch 2/20\n",
            "809/809 [==============================] - 260s 321ms/step - loss: 0.4200 - accuracy: 0.9072 - val_loss: 0.2778 - val_accuracy: 0.9079\n",
            "Epoch 3/20\n",
            "809/809 [==============================] - 260s 321ms/step - loss: 0.3348 - accuracy: 0.9542 - val_loss: 0.1526 - val_accuracy: 0.9527\n",
            "Epoch 4/20\n",
            "809/809 [==============================] - 260s 321ms/step - loss: 0.2532 - accuracy: 0.9734 - val_loss: 0.1361 - val_accuracy: 0.9652\n",
            "Epoch 5/20\n",
            "809/809 [==============================] - 261s 322ms/step - loss: 0.2167 - accuracy: 0.9818 - val_loss: 0.1553 - val_accuracy: 0.9649\n",
            "Epoch 6/20\n",
            "809/809 [==============================] - 258s 319ms/step - loss: 0.1414 - accuracy: 0.9889 - val_loss: 0.1561 - val_accuracy: 0.9684\n",
            "Epoch 7/20\n",
            "809/809 [==============================] - 259s 320ms/step - loss: 0.0916 - accuracy: 0.9924 - val_loss: 0.1919 - val_accuracy: 0.9691\n",
            "Epoch 8/20\n",
            "809/809 [==============================] - 258s 319ms/step - loss: 0.0671 - accuracy: 0.9953 - val_loss: 0.1913 - val_accuracy: 0.9687\n",
            "Epoch 9/20\n",
            "809/809 [==============================] - 258s 319ms/step - loss: 0.0626 - accuracy: 0.9954 - val_loss: 0.1980 - val_accuracy: 0.9698\n",
            "Epoch 10/20\n",
            "809/809 [==============================] - 257s 318ms/step - loss: 0.0370 - accuracy: 0.9970 - val_loss: 0.2091 - val_accuracy: 0.9708\n",
            "Epoch 11/20\n",
            "809/809 [==============================] - 257s 318ms/step - loss: 0.0367 - accuracy: 0.9977 - val_loss: 0.2904 - val_accuracy: 0.9698\n",
            "Epoch 12/20\n",
            "809/809 [==============================] - 257s 318ms/step - loss: 0.0358 - accuracy: 0.9981 - val_loss: 0.2548 - val_accuracy: 0.9712\n",
            "Epoch 13/20\n",
            "809/809 [==============================] - 257s 317ms/step - loss: 0.0141 - accuracy: 0.9990 - val_loss: 0.2720 - val_accuracy: 0.9705\n",
            "Epoch 14/20\n",
            "809/809 [==============================] - 257s 318ms/step - loss: 0.0208 - accuracy: 0.9985 - val_loss: 0.2518 - val_accuracy: 0.9698\n",
            "Epoch 15/20\n",
            "809/809 [==============================] - 257s 317ms/step - loss: 0.0147 - accuracy: 0.9994 - val_loss: 0.2831 - val_accuracy: 0.9691\n",
            "Epoch 16/20\n",
            "809/809 [==============================] - 257s 317ms/step - loss: 0.0160 - accuracy: 0.9990 - val_loss: 0.2685 - val_accuracy: 0.9701\n",
            "Epoch 17/20\n",
            "809/809 [==============================] - 257s 317ms/step - loss: 0.0091 - accuracy: 0.9989 - val_loss: 0.2725 - val_accuracy: 0.9698\n",
            "Epoch 18/20\n",
            "809/809 [==============================] - 257s 317ms/step - loss: 0.0084 - accuracy: 0.9994 - val_loss: 0.2857 - val_accuracy: 0.9708\n",
            "Epoch 19/20\n",
            "809/809 [==============================] - 256s 317ms/step - loss: 0.0086 - accuracy: 0.9995 - val_loss: 0.2846 - val_accuracy: 0.9718\n",
            "Epoch 20/20\n",
            "809/809 [==============================] - 257s 317ms/step - loss: 0.0130 - accuracy: 0.9996 - val_loss: 0.2788 - val_accuracy: 0.9705\n"
          ]
        }
      ],
      "source": [
        "#importing latest tfmodel here to avoid package conflicts\n",
        "!pip install tf-models-official\n",
        "from official.nlp import optimization\n",
        "\n",
        "#fine tuning and Training models\n",
        "epochs_count = 20 #defining_epochs_count\n",
        "steps_each_epoch = tf.data.experimental.cardinality(train_ds).numpy() #steps_per_epoch\n",
        "number_of_training_steps = steps_each_epoch * epochs_count #defining_Training_steps\n",
        "number_of_warmup_steps = int(0.1*number_of_training_steps) #defining_warmup_steps\n",
        "init_lr = 3e-5 #deifning learning rate\n",
        "\n",
        "#creating_optimiser\n",
        "optimizer = optimization.create_optimizer(init_lr=init_lr,\n",
        "                                          num_train_steps=number_of_training_steps,\n",
        "                                          num_warmup_steps=number_of_warmup_steps,\n",
        "                                          optimizer_type='adamw')\n",
        "#model compiling\n",
        "mlp_model.compile(optimizer=optimizer,\n",
        "                          loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),\n",
        "                          metrics=tf.keras.metrics.SparseCategoricalAccuracy('accuracy'))\n",
        "\n",
        "\n",
        "print(f'Training model with {tfhub_handle_encoder}')\n",
        "#training_model\n",
        "mlp_history = mlp_model.fit(x=train_ds,\n",
        "                            validation_data=val_ds,\n",
        "                            epochs=epochs_count,\n",
        "                            class_weight=class_weight)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 10,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 657
        },
        "id": "WkeBaeGrrvob",
        "outputId": "f7908a35-c2f1-4982-9551-a9336d14826f"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "dict_keys(['loss', 'accuracy', 'val_loss', 'val_accuracy'])\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.legend.Legend at 0x7f07f9fe8fd0>"
            ]
          },
          "metadata": {},
          "execution_count": 10
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 864x720 with 2 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ],
      "source": [
        "#creating graphs for accuracy and loss\n",
        "#fining_values\n",
        "history_dict = mlp_history.history\n",
        "print(history_dict.keys())\n",
        "acc = history_dict['accuracy']\n",
        "val_acc = history_dict['val_accuracy']\n",
        "loss = history_dict['loss']\n",
        "val_loss = history_dict['val_loss']\n",
        "epochs = range(1, len(acc) + 1)\n",
        "\n",
        "#defining_figures\n",
        "fig = plt.figure(figsize=(12, 10))\n",
        "fig.tight_layout()\n",
        "#1st_figure\n",
        "plt.subplot(2, 1, 1)\n",
        "plt.plot(epochs, loss, 'r', label='Training-loss')#training_loss\n",
        "plt.plot(epochs, val_loss, 'b', label='Validation-loss')#validation_loss\n",
        "plt.title('Training & validation Loss')\n",
        "plt.xlabel('Epochs-Count')\n",
        "plt.ylabel('Loss')\n",
        "plt.legend()\n",
        "#2nd_Figure\n",
        "plt.subplot(2, 1, 2)\n",
        "plt.plot(epochs, acc, 'r', label='Training-accuracy')#training_accuracy\n",
        "plt.plot(epochs, val_acc, 'b', label='Validation-accuracy')#validation_accuracy\n",
        "plt.title('Training & Validation Accuracy')\n",
        "plt.xlabel('Epochs-Count')\n",
        "plt.ylabel('Acc')\n",
        "plt.legend(loc='lower right')"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 11,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "iZaBQe2IugLH",
        "outputId": "4d9ba0ac-d63a-4c3f-c16f-36ce9412b626"
      },
      "outputs": [
        {
          "metadata": {
            "tags": null
          },
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "WARNING:tensorflow:AutoGraph could not transform <function Model.make_predict_function.<locals>.predict_function at 0x7f07f9c51320> and will run it as-is.\n",
            "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
            "Cause: closure mismatch, requested ('self', 'step_function'), but source function had ()\n",
            "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "WARNING: AutoGraph could not transform <function Model.make_predict_function.<locals>.predict_function at 0x7f07f9c51320> and will run it as-is.\n",
            "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
            "Cause: closure mismatch, requested ('self', 'step_function'), but source function had ()\n",
            "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n",
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       0.98      0.99      0.98      2973\n",
            "           1       0.81      0.75      0.77       224\n",
            "\n",
            "    accuracy                           0.97      3197\n",
            "   macro avg       0.89      0.87      0.88      3197\n",
            "weighted avg       0.97      0.97      0.97      3197\n",
            "\n",
            "[[2933   40]\n",
            " [  57  167]]\n"
          ]
        }
      ],
      "source": [
        "\n",
        "#predicting\n",
        "text = []\n",
        "test_label = []\n",
        "predictions = []\n",
        "\n",
        "for text, label in test_ds.take(-1):\n",
        "  text = text.numpy()#reading_text\n",
        "  test_label.append(label.numpy())#reading_text_label\n",
        "  predictions.append(mlp_model.predict(text))#predicting\n",
        "\n",
        "from itertools import chain\n",
        "flatten_list = list(chain.from_iterable(predictions))\n",
        "\n",
        "y_pred = np.argmax(flatten_list, axis=-1)\n",
        "y_test = np.array(list(chain.from_iterable(test_label)))\n",
        "\n",
        "print(classification_report(y_test, y_pred))#priting_classfication_report\n",
        "print(confusion_matrix(y_test, y_pred)) #priting_confusion_matrix"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 12,
      "metadata": {
        "id": "cX-JX46zJ8TG"
      },
      "outputs": [],
      "source": [
        "def MLP_model(output_bias=None):\n",
        "    if output_bias is not None:\n",
        "        output_bias = tf.keras.initializers.Constant(output_bias)\n",
        "    #input layer    \n",
        "    text_input = tf.keras.layers.Input(shape=(), dtype=tf.string, name='text')\n",
        "    #preprocessing layer \n",
        "    preprocessing_layer = hub.KerasLayer(tfhub_handle_preprocess, name='preprocessing')\n",
        "    #giving input to bert model from the pre processed layer\n",
        "    encoder_inputs = preprocessing_layer(text_input)\n",
        "    #selecting the encoder\n",
        "    encoder = hub.KerasLayer(tfhub_handle_encoder, trainable=True, name='BERT_encoder')\n",
        "    #collecting the output of the bert encoder\n",
        "    outputs = encoder(encoder_inputs)\n",
        "    #collecting the pooled outputs\n",
        "    net = outputs['pooled_output']\n",
        "    #feeding the output to the multilayer perception\n",
        "    #1 DENSE LAYER\n",
        "    net = tf.keras.layers.Dense(512, activation=\"relu\")(net)\n",
        "    #1 DROPUT LAYER\n",
        "    net = tf.keras.layers.Dropout(0.2)(net)\n",
        "    #1 OUTPUT LAYER WITH SOFTMAX\n",
        "    net = tf.keras.layers.Dense(3, activation=\"softmax\", name='classifier', bias_initializer=output_bias)(net)\n",
        "    return tf.keras.Model(text_input, net)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 13,
      "metadata": {
        "id": "xKwpxZHRsn-B"
      },
      "outputs": [],
      "source": [
        "#Defining Model Architecture\n",
        "def CNN_model():\n",
        "    #Trasnformer Encoding before feeding to CNN\n",
        "    text_in = tf.keras.layers.Input(shape=(), dtype=tf.string, name='text')\n",
        "    preprocess_layer = hub.KerasLayer(tfhub_handle_preprocess, name='preprocessing')\n",
        "    encoder_in = preprocess_layer(text_in)\n",
        "    encode = hub.KerasLayer(tfhub_handle_encoder, trainable=True, name='BERT_encoder')\n",
        "    outputs = encode(encoder_in)\n",
        "    network_layer= sequence_output = outputs[\"sequence_output\"]\n",
        "    # 2 CNNLayers\n",
        "    network_layer = tf.keras.layers.Conv1D(32, (2), activation='relu')(network_layer) #CNN Layer 1\n",
        "    network_layer = tf.keras.layers.Conv1D(64, (2), activation='relu')(network_layer) #CNN Layer 2\n",
        "    network_layer = tf.keras.layers.GlobalMaxPool1D()(network_layer) #Global Maxpooling Layer\n",
        "    network_layer = tf.keras.layers.Dense(256, activation=\"relu\")(network_layer) #DenseLayer \n",
        "    network_layer = tf.keras.layers.Dropout(0.1)(network_layer) #Dropout Layer\n",
        "    network_layer = tf.keras.layers.Dense(3, activation=\"softmax\", name='classifier')(network_layer) #ClassificationLayer\n",
        "    return tf.keras.Model(text_input, network_layer)\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 14,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "sfuBwiv916Jm",
        "outputId": "7f9a4190-7dfd-4cf1-f3e7-498598a6d05b"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "I dislike all the people who do not come in time to office -> 0\n",
            "You are welcome to come to my country -> 0\n",
            "I hate black people -> 1\n"
          ]
        }
      ],
      "source": [
        "user_input = [\"I dislike all the people who do not come in time to office\", \"You are welcome to come to my country\",\"I hate black people\"]\n",
        "user_input_prediction = mlp_model.predict(user_input)\n",
        "\n",
        "result = []\n",
        "for values in user_input_prediction:\n",
        "   max = -1\n",
        "   type_comment = -1\n",
        "   for index,value in enumerate(values): \n",
        "      if (value > max):\n",
        "         max = value\n",
        "         type_comment = index\n",
        "   result.append(type_comment) \n",
        "for speech, output in zip(user_input, result):\n",
        "   print(f'{speech[:100]} -> {output}')"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Note: Used some slang languages as part of custom inputs to test the hate speech performance."
      ],
      "metadata": {
        "id": "CDrPrFGoGWt1"
      }
    }
  ],
  "metadata": {
    "colab": {
      "collapsed_sections": [],
      "name": "Hate-Speech-Detection.ipynb",
      "provenance": []
    },
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU",
    "gpuClass": "standard"
  },
  "nbformat": 4,
  "nbformat_minor": 0
}