Fine Tune BERT for Text Classification with TensorFlow.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "NLP-with-BERT",
      "provenance": [],
      "collapsed_sections": [],
      "machine_shape": "hm"
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "zGCJYkQj_Uu2"
      },
      "source": [
        "<h2 align=center> Fine-Tune BERT for Text Classification with TensorFlow</h2>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "MYXXV5n3Ab-4"
      },
      "source": [
        "### Contents"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "XhK-SYGyAjxe"
      },
      "source": [
        "\n",
        "- **[1]()**: Download and Import the Quora Insincere Questions Dataset\n",
        "- **[2]()**: Create tf.data.Datasets for Training and Evaluation\n",
        "- **[3]()**: Download a Pre-trained BERT Model from TensorFlow Hub\n",
        "- **[4]()**: Tokenize and Preprocess Text for BERT\n",
        "- **[5]()**: Wrap a Python Function into a TensorFlow op for Eager Execution\n",
        "- **[6]()**: Create a TensorFlow Input Pipeline with `tf.data`\n",
        "- **[7]()**: Add a Classification Head to the BERT `hub.KerasLayer`\n",
        "- **[8]()**: Fine-Tune BERT for Text Classification\n",
        "- **[9]()**: Evaluate the BERT Text Classification Model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8V9c8vzSL3aj",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "cf1b258d-991e-40ee-e15c-8e403ea84e2b"
      },
      "source": [
        "!nvidia-smi"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Thu Nov 26 11:54:29 2020       \n",
            "+-----------------------------------------------------------------------------+\n",
            "| NVIDIA-SMI 455.38       Driver Version: 418.67       CUDA Version: 10.1     |\n",
            "|-------------------------------+----------------------+----------------------+\n",
            "| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |\n",
            "| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |\n",
            "|                               |                      |               MIG M. |\n",
            "|===============================+======================+======================|\n",
            "|   0  Tesla P100-PCIE...  Off  | 00000000:00:04.0 Off |                    0 |\n",
            "| N/A   50C    P0    31W / 250W |      0MiB / 16280MiB |      0%      Default |\n",
            "|                               |                      |                 ERR! |\n",
            "+-------------------------------+----------------------+----------------------+\n",
            "                                                                               \n",
            "+-----------------------------------------------------------------------------+\n",
            "| Processes:                                                                  |\n",
            "|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |\n",
            "|        ID   ID                                                   Usage      |\n",
            "|=============================================================================|\n",
            "|  No running processes found                                                 |\n",
            "+-----------------------------------------------------------------------------+\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Obch3rAuBVf0"
      },
      "source": [
        "### Install TensorFlow and TensorFlow Model Garden"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "azgGrYd0vd3n"
      },
      "source": [
        "import tensorflow as tf"
      ],
      "execution_count": 2,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "AFRTC-zwUy6D",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "8ff39ea0-ee93-4374-e4a3-0002ffaa052f"
      },
      "source": [
        "!git clone --depth 1 -b v2.3.0 https://github.com/tensorflow/models.git"
      ],
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "fatal: destination path 'models' already exists and is not an empty directory.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3H2G0571zLLs"
      },
      "source": [
        "# install requirements to use tensorflow/models repository\n",
        "!pip install -Uqr models/official/requirements.txt"
      ],
      "execution_count": 4,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "IMsEoT3Fg4Wg"
      },
      "source": [
        "## 1: Download and Import the Quora Insincere Questions Dataset"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "GmqEylyFYTdP"
      },
      "source": [
        "import random\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "from sklearn.model_selection import train_test_split\n",
        "\n",
        "import tensorflow_hub as hub\n",
        "import sys\n",
        "sys.path.append('models')\n",
        "from official.nlp.data import classifier_data_lib\n",
        "from official.nlp.bert import tokenization\n",
        "from official.nlp import optimization"
      ],
      "execution_count": 5,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ZuX1lB8pPJ-W",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "d4496964-4c5f-4f4d-a169-ef0985f948e1"
      },
      "source": [
        "print(\"TF Version: \", tf.__version__)\n",
        "print(\"Eager mode: \", tf.executing_eagerly())\n",
        "print(\"Hub version: \", hub.__version__)\n",
        "print(\"GPU is\", \"available\" if tf.config.experimental.list_physical_devices(\"GPU\") else \"NOT AVAILABLE\")"
      ],
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "TF Version:  2.3.0\n",
            "Eager mode:  True\n",
            "Hub version:  0.10.0\n",
            "GPU is available\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "QtbwpWgyEZg7"
      },
      "source": [
        "A downloadable copy of the [Quora Insincere Questions Classification data](https://www.kaggle.com/c/quora-insincere-questions-classification/data) can be found [https://archive.org/download/fine-tune-bert-tensorflow-train.csv/train.csv.zip](https://archive.org/download/fine-tune-bert-tensorflow-train.csv/train.csv.zip). Decompress and read the data into a pandas DataFrame."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "yeHE98KiMvDd",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 419
        },
        "outputId": "62a49897-72c7-48b5-fe2d-83dca99cd987"
      },
      "source": [
        "df = pd.read_csv('https://archive.org/download/fine-tune-bert-tensorflow-train.csv/train.csv.zip', \n",
        "                 compression='zip', low_memory=False)\n",
        "df"
      ],
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>qid</th>\n",
              "      <th>question_text</th>\n",
              "      <th>target</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>00002165364db923c7e6</td>\n",
              "      <td>How did Quebec nationalists see their province...</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>000032939017120e6e44</td>\n",
              "      <td>Do you have an adopted dog, how would you enco...</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>0000412ca6e4628ce2cf</td>\n",
              "      <td>Why does velocity affect time? Does velocity a...</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>000042bf85aa498cd78e</td>\n",
              "      <td>How did Otto von Guericke used the Magdeburg h...</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>0000455dfa3e01eae3af</td>\n",
              "      <td>Can I convert montra helicon D to a mountain b...</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>...</th>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1306117</th>\n",
              "      <td>ffffcc4e2331aaf1e41e</td>\n",
              "      <td>What other technical skills do you need as a c...</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1306118</th>\n",
              "      <td>ffffd431801e5a2f4861</td>\n",
              "      <td>Does MS in ECE have good job prospects in USA ...</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1306119</th>\n",
              "      <td>ffffd48fb36b63db010c</td>\n",
              "      <td>Is foam insulation toxic?</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1306120</th>\n",
              "      <td>ffffec519fa37cf60c78</td>\n",
              "      <td>How can one start a research project based on ...</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1306121</th>\n",
              "      <td>ffffed09fedb5088744a</td>\n",
              "      <td>Who wins in a battle between a Wolverine and a...</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>1306122 rows × 3 columns</p>\n",
              "</div>"
            ],
            "text/plain": [
              "                          qid  ... target\n",
              "0        00002165364db923c7e6  ...      0\n",
              "1        000032939017120e6e44  ...      0\n",
              "2        0000412ca6e4628ce2cf  ...      0\n",
              "3        000042bf85aa498cd78e  ...      0\n",
              "4        0000455dfa3e01eae3af  ...      0\n",
              "...                       ...  ...    ...\n",
              "1306117  ffffcc4e2331aaf1e41e  ...      0\n",
              "1306118  ffffd431801e5a2f4861  ...      0\n",
              "1306119  ffffd48fb36b63db010c  ...      0\n",
              "1306120  ffffec519fa37cf60c78  ...      0\n",
              "1306121  ffffed09fedb5088744a  ...      0\n",
              "\n",
              "[1306122 rows x 3 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 7
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ZTT2U-_BFru-",
        "outputId": "cd7c0943-cf86-432d-d65c-cdb3dcfb7d50"
      },
      "source": [
        "df.target.value_counts()"
      ],
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0    1225312\n",
              "1      80810\n",
              "Name: target, dtype: int64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 8
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ELjswHcFHfp3"
      },
      "source": [
        "## 2: Create tf.data.Datasets for Training and Evaluation"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "fScULIGPwuWk",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "4799e656-5dba-44a4-e89c-7c9fc31c2b57"
      },
      "source": [
        "train ,validation = train_test_split(df, train_size=0.60, random_state=42, stratify=df.target.values)\n",
        "validation, test = train_test_split(validation, random_state=42, test_size=0.5, stratify=validation.target.values)\n",
        "print(\"Train Shape\\tValid Shape\\tTest Shape\")\n",
        "train.shape, validation.shape, test.shape"
      ],
      "execution_count": 61,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train Shape\tValid Shape\tTest Shape\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "((783673, 3), (261224, 3), (261225, 3))"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 61
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "qQYMGT5_qLPX",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "9975ed7e-9e42-481a-98ea-9565e722092c"
      },
      "source": [
        " with tf.device('/cpu:0'):\n",
        "   train_data = tf.data.Dataset.from_tensor_slices((train['question_text'].values, train['target'].values))\n",
        "   valid_data = tf.data.Dataset.from_tensor_slices((validation['question_text'].values, validation['target'].values))\n",
        "   test_data = tf.data.Dataset.from_tensor_slices((test['question_text'].values, test['target'].values))\n",
        "\n",
        "   for text, label in train_data.take(1):\n",
        "     print(test)\n",
        "     print(label)"
      ],
      "execution_count": 62,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "                          qid  ... target\n",
            "22004    0452e4e5b5c34944bc36  ...      0\n",
            "1013706  c6a65b5d62175723a176  ...      0\n",
            "659002   8114ca3a08a56ade83d9  ...      0\n",
            "1299258  fea6446c17da271b7426  ...      0\n",
            "850519   a6a69b2770aa2dcf3073  ...      0\n",
            "...                       ...  ...    ...\n",
            "503308   628c8ecf3b45b64c0a65  ...      0\n",
            "1065936  d0e1a71eee5f806706f2  ...      0\n",
            "319173   3e8d066aed1996792729  ...      0\n",
            "286674   3823c5055028979fa0ab  ...      0\n",
            "131093   19a5635a21c045709a26  ...      0\n",
            "\n",
            "[261225 rows x 3 columns]\n",
            "tf.Tensor(0, shape=(), dtype=int64)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "e2-ReN88Hvy_"
      },
      "source": [
        "## 3: Download a Pre-trained BERT Model from TensorFlow Hub"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "EMb5M86b4-BU"
      },
      "source": [
        "\"\"\"\n",
        "Each line of the dataset is composed of the review text and its label\n",
        "- Data preprocessing consists of transforming text to BERT input features:\n",
        "input_word_ids, input_mask, segment_ids\n",
        "- In the process, tokenizing the text is done with the provided BERT model tokenizer\n",
        "\"\"\"\n",
        "labels_list = [0, 1]\n",
        "max_seq_len = 150\n",
        "batch = 64\n",
        "\n",
        "#: https://tfhub.dev/tensorflow/bert_en_uncased_L-12_H-768_A-12/2\n",
        "bert_layer = hub.KerasLayer('https://tfhub.dev/tensorflow/bert_en_uncased_L-12_H-768_A-12/2', trainable=False)\n",
        "vocab = bert_layer.resolved_object.vocab_file.asset_path.numpy()\n",
        "do_lower_case = bert_layer.resolved_object.do_lower_case.numpy()\n",
        "tokenizer = tokenization.FullTokenizer(vocab, do_lower_case)"
      ],
      "execution_count": 11,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "wEUezMK-zkkI",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "1c352aa8-bd79-4bcc-8019-bc6a0b980330"
      },
      "source": [
        "tokenizer.wordpiece_tokenizer.tokenize(\"hi, how, are you\")"
      ],
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "['hi', '##,', 'how', '##,', 'are', 'you']"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 12
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "qeHAKdTKWBnl",
        "outputId": "c347e54b-b790-4510-b4ff-231c0a3525e5"
      },
      "source": [
        "tokenizer.wordpiece_tokenizer.tokenize(\"hi, How, are you\")"
      ],
      "execution_count": 13,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "['hi', '##,', '[UNK]', 'are', 'you']"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 13
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5AFsmTO5JSmc",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "de5a0fc1-f705-433c-90ab-2c908656c49d"
      },
      "source": [
        "tokenizer.convert_tokens_to_ids(tokenizer.wordpiece_tokenizer.tokenize(\"hi, how, are you\"))"
      ],
      "execution_count": 14,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[7632, 29623, 2129, 29623, 2024, 2017]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 14
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "9QinzNq6OsP1"
      },
      "source": [
        "## 4: Tokenize and Preprocess Text for BERT"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "3FTqJ698zZ1e"
      },
      "source": [
        "<div align=\"center\">\n",
        "    <img width=\"512px\" src='https://drive.google.com/uc?id=1-SpKFELnEvBMBqO7h3iypo8q9uUUo96P' />\n",
        "    <p style=\"text-align: center;color:gray\"> BERT Tokenizer</p>\n",
        "</div>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "cWYkggYe6HZc"
      },
      "source": [
        "We'll need to transform our data into a format BERT understands. This involves two steps. First, we create InputExamples using `classifier_data_lib`'s constructor `InputExample` provided in the BERT library."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "m-21A5aNJM0W"
      },
      "source": [
        "# This provides a function to convert row to input features and label\n",
        "\n",
        "def to_feature(text, label, label_list=labels_list, max_seq_length=max_seq_len, tokenizer=tokenizer):\n",
        "   \n",
        "   example = classifier_data_lib.InputExample(guid=None,\n",
        "                                              text_a=text.numpy(),\n",
        "                                              text_b=None,\n",
        "                                              label=label.numpy()\n",
        "                                              )\n",
        "   feature = classifier_data_lib.convert_single_example(0, example, label_list, max_seq_length, tokenizer)\n",
        "   \n",
        "   return (feature.input_ids, feature.input_mask, feature.segment_ids, feature.label_id)"
      ],
      "execution_count": 55,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "A_HQSsHwWCsK"
      },
      "source": [
        "We want to use [`Dataset.map`](https://www.tensorflow.org/api_docs/python/tf/data/Dataset#map) to apply this function to each element of the dataset. [`Dataset.map`](https://www.tensorflow.org/api_docs/python/tf/data/Dataset#map) runs in graph mode.\n",
        "\n",
        "- Graph tensors do not have a value.\n",
        "- In graph mode we can only use TensorFlow Ops and functions.\n",
        "\n",
        "So we can't `.map` this function directly: We need to wrap it in a [`tf.py_function`](https://www.tensorflow.org/api_docs/python/tf/py_function). The [`tf.py_function`](https://www.tensorflow.org/api_docs/python/tf/py_function) will pass regular tensors (with a value and a `.numpy()` method to access it), to the wrapped python function."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "zaNlkKVfWX0Q"
      },
      "source": [
        "## 5: Wrap a Python Function into a TensorFlow op for Eager Execution"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "AGACBcfCWC2O"
      },
      "source": [
        "def to_feature_map(text, label):\n",
        "  input_ids, input_mask, segment_ids, label_id = tf.py_function(to_feature, inp=[text, label],\n",
        "                                                                Tout=[tf.int32, tf.int32, tf.int32, tf.int32]\n",
        "                                                                )\n",
        "  input_ids.set_shape([max_seq_len])\n",
        "  input_mask.set_shape([max_seq_len])\n",
        "  segment_ids.set_shape([max_seq_len])\n",
        "  label_id.set_shape([])\n",
        "\n",
        "  x = {\n",
        "      'input_word_ids': input_ids,\n",
        "       'input_mask':input_mask,\n",
        "       'input_type_ids':segment_ids\n",
        "  }\n",
        "  return (x, label_id) "
      ],
      "execution_count": 56,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "dhdO6MjTbtn1"
      },
      "source": [
        "## 6: Create a TensorFlow Input Pipeline with `tf.data`"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "LHRdiO3dnPNr"
      },
      "source": [
        "with tf.device('/cpu:0'):\n",
        "  # train\n",
        "  train_data = (train_data.map(to_feature_map,\n",
        "                               num_parallel_calls=tf.data.experimental.AUTOTUNE\n",
        "                               )\n",
        "  .shuffle(1000)\n",
        "  .batch(batch, drop_remainder=True)\n",
        "  .prefetch(tf.data.experimental.AUTOTUNE)\n",
        "  )\n",
        "\n",
        "  # valid\n",
        "  valid_data = (valid_data.map(to_feature_map,\n",
        "                               num_parallel_calls=tf.data.experimental.AUTOTUNE\n",
        "                               )\n",
        "  .batch(batch, drop_remainder=True)\n",
        "  .prefetch(tf.data.experimental.AUTOTUNE)\n",
        "  )\n",
        "  \n",
        "  # test\n",
        "  test_data = (test_data.map(to_feature_map,\n",
        "                               num_parallel_calls=tf.data.experimental.AUTOTUNE\n",
        "                               )\n",
        "  .batch(batch, drop_remainder=True)\n",
        "  .prefetch(tf.data.experimental.AUTOTUNE)\n",
        "  )\n",
        "  "
      ],
      "execution_count": 63,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "KLUWnfx-YDi2"
      },
      "source": [
        "The resulting `tf.data.Datasets` return `(features, labels)` pairs, as expected by [`keras.Model.fit`](https://www.tensorflow.org/api_docs/python/tf/keras/Model#fit):"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "B0Z2cy9GHQ8x",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "4c73f55c-136b-4ddc-8126-d31cd50938ee"
      },
      "source": [
        "# train data spec\n",
        "train_data.element_spec"
      ],
      "execution_count": 64,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "({'input_mask': TensorSpec(shape=(64, 150), dtype=tf.int32, name=None),\n",
              "  'input_type_ids': TensorSpec(shape=(64, 150), dtype=tf.int32, name=None),\n",
              "  'input_word_ids': TensorSpec(shape=(64, 150), dtype=tf.int32, name=None)},\n",
              " TensorSpec(shape=(64,), dtype=tf.int32, name=None))"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 64
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DGAH-ycYOmao",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "22d8f456-6092-4c98-cce8-007802c6d64f"
      },
      "source": [
        "# valid data spec\n",
        "valid_data.element_spec"
      ],
      "execution_count": 65,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "({'input_mask': TensorSpec(shape=(64, 150), dtype=tf.int32, name=None),\n",
              "  'input_type_ids': TensorSpec(shape=(64, 150), dtype=tf.int32, name=None),\n",
              "  'input_word_ids': TensorSpec(shape=(64, 150), dtype=tf.int32, name=None)},\n",
              " TensorSpec(shape=(64,), dtype=tf.int32, name=None))"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 65
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "GZxe-7yhPyQe"
      },
      "source": [
        "## 7: Add a Classification Head to the BERT Layer"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "9THH5V0Dw2HO"
      },
      "source": [
        "<div align=\"center\">\n",
        "    <img width=\"512px\" src='https://drive.google.com/uc?id=1fnJTeJs5HUpz7nix-F9E6EZdgUflqyEu' />\n",
        "    <p style=\"text-align: center;color:gray\"> BERT Layer</p>\n",
        "</div>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "G9il4gtlADcp"
      },
      "source": [
        "# Building the model\n",
        "def create_model():\n",
        "  input_word_ids = tf.keras.layers.Input(shape=(max_seq_len,), dtype=tf.int32,\n",
        "                                       name=\"input_word_ids\")\n",
        "  input_mask = tf.keras.layers.Input(shape=(max_seq_len,), dtype=tf.int32,\n",
        "                                   name=\"input_mask\")\n",
        "  segment_ids = tf.keras.layers.Input(shape=(max_seq_len,), dtype=tf.int32,\n",
        "                                    name=\"segment_ids\")\n",
        "  pooled_output, sequence_output = bert_layer([input_word_ids, input_mask, segment_ids])\n",
        "\n",
        "  drop = tf.keras.layers.Dropout(0.3)(pooled_output)\n",
        "  dense_1 = tf.keras.layers.Dense(512, activation='relu')(drop)\n",
        "  dense_2 = tf.keras.layers.Dense(1024, activation='relu')(dense_1)\n",
        "  dense_3 = tf.keras.layers.Dense(1024, activation='relu')(dense_2)\n",
        "  drop = tf.keras.layers.Dropout(0.4)(dense_3)\n",
        "  \n",
        "  output = tf.keras.layers.Dense(1, activation='sigmoid', name='output')(drop)\n",
        "\n",
        "  model = tf.keras.Model(\n",
        "      {\n",
        "        'input_word_ids': input_word_ids,\n",
        "       'input_mask': input_mask,\n",
        "       'input_type_ids': segment_ids\n",
        "      },\n",
        "      output\n",
        "  )\n",
        "  return model"
      ],
      "execution_count": 33,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "S6maM-vr7YaJ"
      },
      "source": [
        "## 8: Fine-Tune BERT for Text Classification"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ptCtiiONsBgo",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "380f9447-188e-468b-d9be-dc7b689d509d"
      },
      "source": [
        "model = create_model()\n",
        "model.compile(\n",
        "    optimizer=tf.keras.optimizers.Adam(learning_rate=2e-4),\n",
        "    loss=tf.keras.losses.BinaryCrossentropy(),\n",
        "    metrics=['binary_accuracy']\n",
        ")\n",
        "model.summary()"
      ],
      "execution_count": 34,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"functional_3\"\n",
            "__________________________________________________________________________________________________\n",
            "Layer (type)                    Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            "input_word_ids (InputLayer)     [(None, 150)]        0                                            \n",
            "__________________________________________________________________________________________________\n",
            "input_mask (InputLayer)         [(None, 150)]        0                                            \n",
            "__________________________________________________________________________________________________\n",
            "segment_ids (InputLayer)        [(None, 150)]        0                                            \n",
            "__________________________________________________________________________________________________\n",
            "keras_layer (KerasLayer)        [(None, 768), (None, 109482241   input_word_ids[0][0]             \n",
            "                                                                 input_mask[0][0]                 \n",
            "                                                                 segment_ids[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "dropout_2 (Dropout)             (None, 768)          0           keras_layer[2][0]                \n",
            "__________________________________________________________________________________________________\n",
            "dense (Dense)                   (None, 512)          393728      dropout_2[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "dense_1 (Dense)                 (None, 1024)         525312      dense[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "dense_2 (Dense)                 (None, 1024)         1049600     dense_1[0][0]                    \n",
            "__________________________________________________________________________________________________\n",
            "dropout_3 (Dropout)             (None, 1024)         0           dense_2[0][0]                    \n",
            "__________________________________________________________________________________________________\n",
            "output (Dense)                  (None, 1)            1025        dropout_3[0][0]                  \n",
            "==================================================================================================\n",
            "Total params: 111,451,906\n",
            "Trainable params: 1,969,665\n",
            "Non-trainable params: 109,482,241\n",
            "__________________________________________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6GJaFnkbMtPL",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 724
        },
        "outputId": "47dbcf26-345b-4e8e-de84-6403a246baa4"
      },
      "source": [
        "tf.keras.utils.plot_model(model, show_shapes=True, dpi=80)"
      ],
      "execution_count": 35,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<IPython.core.display.Image object>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 35
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "OcREcgPUHr9O",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "b92fee51-77d4-4bfd-8e72-854df5ddb3d7"
      },
      "source": [
        "# Train model\n",
        "h = model.fit(train_data, validation_data=valid_data,\n",
        "              epochs=2\n",
        ")"
      ],
      "execution_count": 66,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/2\n",
            "    2/12244 [..............................] - ETA: 1:13:54 - loss: 0.1812 - binary_accuracy: 0.9531WARNING:tensorflow:Callbacks method `on_train_batch_end` is slow compared to the batch time (batch time: 0.0267s vs `on_train_batch_end` time: 0.3429s). Check your callbacks.\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "WARNING:tensorflow:Callbacks method `on_train_batch_end` is slow compared to the batch time (batch time: 0.0267s vs `on_train_batch_end` time: 0.3429s). Check your callbacks.\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "12244/12244 [==============================] - ETA: 0s - loss: 0.1759 - binary_accuracy: 0.9396WARNING:tensorflow:Callbacks method `on_test_batch_end` is slow compared to the batch time (batch time: 0.0197s vs `on_test_batch_end` time: 0.3425s). Check your callbacks.\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "WARNING:tensorflow:Callbacks method `on_test_batch_end` is slow compared to the batch time (batch time: 0.0197s vs `on_test_batch_end` time: 0.3425s). Check your callbacks.\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\r12244/12244 [==============================] - 5965s 487ms/step - loss: 0.1759 - binary_accuracy: 0.9396 - val_loss: 0.1529 - val_binary_accuracy: 0.9419\n",
            "Epoch 2/2\n",
            "12244/12244 [==============================] - 5962s 487ms/step - loss: 0.1681 - binary_accuracy: 0.9404 - val_loss: 0.1484 - val_binary_accuracy: 0.9427\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "kNZl1lx_cA5Y"
      },
      "source": [
        "## 9: Evaluate the BERT Text Classification Model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "dCjgrUYH_IsE"
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "\n",
        "def plot_graphs(history, metric):\n",
        "  plt.plot(history.history[metric])\n",
        "  plt.plot(history.history['val_'+metric], '')\n",
        "  plt.xlabel(\"Epochs\")\n",
        "  plt.ylabel(metric)\n",
        "  plt.legend([metric, 'val_'+metric])\n",
        "  plt.show()"
      ],
      "execution_count": 67,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "v6lrFRra_KmA",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 279
        },
        "outputId": "3a2d4c8c-87df-43dd-f1e5-fbdb726861b3"
      },
      "source": [
        "plot_graphs(h, 'loss')"
      ],
      "execution_count": 69,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "opu9neBA_98R",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 279
        },
        "outputId": "c73bac78-9e15-408e-cd9c-a8ab3b99be16"
      },
      "source": [
        "plot_graphs(h, 'binary_accuracy')"
      ],
      "execution_count": 70,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hkhtCCgnUbY6",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "94f71c1f-47db-4395-8f1b-d7be651f206b"
      },
      "source": [
        "model.evaluate(test_data)"
      ],
      "execution_count": 71,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "   2/4081 [..............................] - ETA: 24:21 - loss: 0.1205 - binary_accuracy: 0.9609WARNING:tensorflow:Callbacks method `on_test_batch_end` is slow compared to the batch time (batch time: 0.0219s vs `on_test_batch_end` time: 0.3438s). Check your callbacks.\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "WARNING:tensorflow:Callbacks method `on_test_batch_end` is slow compared to the batch time (batch time: 0.0219s vs `on_test_batch_end` time: 0.3438s). Check your callbacks.\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "4081/4081 [==============================] - 1493s 366ms/step - loss: 0.1481 - binary_accuracy: 0.9427\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[0.14811307191848755, 0.9426917433738708]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 71
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "K4B8NQBLd9rN",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "bc50b9a2-4d71-4323-823a-0fa65ebef14c"
      },
      "source": [
        "sample_example = ['may i have your number??']\n",
        "ques = tf.data.Dataset.from_tensor_slices((sample_example, [0]*len(sample_example)))\n",
        "ques = (ques.map(to_feature_map)).batch(1)\n",
        "pred = model.predict(ques)[0][0]\n",
        "t = 0.5\n",
        "['Insincere' if pred>=t else 'Sincere']"
      ],
      "execution_count": 78,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "['Sincere']"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 78
        }
      ]
    }
  ]
}