notebook.ipynb

{
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "id": "ehxfuJUan9yT",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 419
        },
        "outputId": "b148ccab-f884-441e-a398-3dc121d3189d"
      },
      "source": [
        "!pip install bert-for-tf2\n",
        "#!pip install tf-nightly\n",
        "import tensorflow_hub as hub\n",
        "import tensorflow as tf\n",
        "import bert\n",
        "from bert import bert_tokenization\n",
        "from tensorflow.keras import layers\n",
        "\n",
        "FullTokenizer = bert.bert_tokenization.FullTokenizer\n",
        "import math"
      ],
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Collecting bert-for-tf2\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/35/5c/6439134ecd17b33fe0396fb0b7d6ce3c5a120c42a4516ba0e9a2d6e43b25/bert-for-tf2-0.14.4.tar.gz (40kB)\n",
            "\r\u001b[K     |████████                        | 10kB 17.7MB/s eta 0:00:01\r\u001b[K     |████████████████▏               | 20kB 1.7MB/s eta 0:00:01\r\u001b[K     |████████████████████████▎       | 30kB 2.2MB/s eta 0:00:01\r\u001b[K     |████████████████████████████████| 40kB 1.8MB/s \n",
            "\u001b[?25hCollecting py-params>=0.9.6\n",
            "  Downloading https://files.pythonhosted.org/packages/a4/bf/c1c70d5315a8677310ea10a41cfc41c5970d9b37c31f9c90d4ab98021fd1/py-params-0.9.7.tar.gz\n",
            "Collecting params-flow>=0.8.0\n",
            "  Downloading https://files.pythonhosted.org/packages/a9/95/ff49f5ebd501f142a6f0aaf42bcfd1c192dc54909d1d9eb84ab031d46056/params-flow-0.8.2.tar.gz\n",
            "Requirement already satisfied: numpy in /usr/local/lib/python3.6/dist-packages (from params-flow>=0.8.0->bert-for-tf2) (1.18.5)\n",
            "Requirement already satisfied: tqdm in /usr/local/lib/python3.6/dist-packages (from params-flow>=0.8.0->bert-for-tf2) (4.41.1)\n",
            "Building wheels for collected packages: bert-for-tf2, py-params, params-flow\n",
            "  Building wheel for bert-for-tf2 (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "  Created wheel for bert-for-tf2: filename=bert_for_tf2-0.14.4-cp36-none-any.whl size=30114 sha256=c8815685412dd1ed1d507d9fca4e8b11ce3a14a5540272bf2b6af29441627729\n",
            "  Stored in directory: /root/.cache/pip/wheels/cf/3f/4d/79d7735015a5f523648df90d871ce8e89a7df8185f7703eeab\n",
            "  Building wheel for py-params (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "  Created wheel for py-params: filename=py_params-0.9.7-cp36-none-any.whl size=7302 sha256=152f2f2571bcf76fcc2148653b041dd41424aa5b049a285a6afee32ab8de247f\n",
            "  Stored in directory: /root/.cache/pip/wheels/67/f5/19/b461849a50aefdf4bab47c4756596e82ee2118b8278e5a1980\n",
            "  Building wheel for params-flow (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "  Created wheel for params-flow: filename=params_flow-0.8.2-cp36-none-any.whl size=19473 sha256=229e78e97dd5209dcb7f6bc2582fe0e53d507594a2cc995004bc65845ca5f717\n",
            "  Stored in directory: /root/.cache/pip/wheels/08/c8/7f/81c86b9ff2b86e2c477e3914175be03e679e596067dc630c06\n",
            "Successfully built bert-for-tf2 py-params params-flow\n",
            "Installing collected packages: py-params, params-flow, bert-for-tf2\n",
            "Successfully installed bert-for-tf2-0.14.4 params-flow-0.8.2 py-params-0.9.7\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "n1WSmpsw_Bjv",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import bert\n",
        "import tensorflow as tf\n",
        "import tensorflow_hub as hub\n",
        "from tensorflow.keras.optimizers import Adam\n",
        "from tensorflow.keras.callbacks import ModelCheckpoint, TensorBoard\n",
        "\n",
        "from tensorflow.keras import backend as K"
      ],
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8oWSPjznpZ3z",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import pandas as pd\n",
        "import numpy as np\n",
        "\n",
        "from sklearn.model_selection import train_test_split\n",
        "from sklearn.metrics import classification_report\n",
        "from sklearn.utils import class_weight\n",
        "\n",
        "df = pd.read_csv('./data/clean_train.csv', index_col = False)\n",
        "df.dropna(inplace = True)\n",
        "df.reset_index(drop=True,inplace=True)"
      ],
      "execution_count": 5,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "APSTQ8cNRYAD",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        },
        "outputId": "fe8b8b4f-59ab-43e2-87fb-a3368d04972b"
      },
      "source": [
        "X_data = df['text'].to_numpy()\n",
        "y_data = df['target'].to_numpy()\n",
        "y_data = y_data.reshape(-1, 1)\n",
        "\n",
        "X_train, X_test, y_train, y_test = train_test_split(X_data, y_data, test_size=0.2, random_state=777)\n",
        "class_weights = class_weight.compute_class_weight('balanced', np.unique(df['target'].to_numpy()), df['target'].to_numpy())\n",
        "class_weights = dict(enumerate(class_weights))\n",
        "print(class_weights)"
      ],
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "{0: 1.086939150585321, 1: 0.9259385317041625}\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "eSEAd6MG7b_o",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "MAX_SEQ_LEN = 65"
      ],
      "execution_count": 7,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zCRVIOPUQyhk",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def bert_encode(texts, tokenizer, sequence_length=512):\n",
        "    all_tokens = []\n",
        "    all_masks = []\n",
        "    all_segments = []\n",
        "\n",
        "    for text in texts:\n",
        "        try:\n",
        "            #print(text) #Our Deeds are the Reason of this #earthquake May ALLAH Forgive us all\n",
        "            text = tokenizer.tokenize(text)\n",
        "            #print(text) #['our', 'deeds', 'are', 'the', 'reason', 'of', 'this', '#', 'earthquake', 'may', 'allah', 'forgive', 'us', 'all']\n",
        "            #print(len(text)) #14\n",
        "            text = text[:sequence_length-2]\n",
        "            #print(len(text)) #14\n",
        "            #print(text) #['our', 'deeds', 'are', 'the', 'reason', 'of', 'this', '#', 'earthquake', 'may', 'allah', 'forgive', 'us', 'all']\n",
        "            input_sequence = [\"[CLS]\"] + text + [\"[SEP]\"]\n",
        "            #print(input_sequence) #['[CLS]', 'our', 'deeds', 'are', 'the', 'reason', 'of', 'this', '#', 'earthquake', 'may', 'allah', 'forgive', 'us', 'all', '[SEP]']\n",
        "            #print(len(input_sequence)) #16\n",
        "            pad_len = sequence_length - len(input_sequence)\n",
        "            tokens = tokenizer.convert_tokens_to_ids(input_sequence)\n",
        "            #print(tokens) #[101, 2256, 15616, 2024, 1996, 3114, 1997, 2023, 1001, 8372, 2089, 16455, 9641, 2149, 2035, 102]\n",
        "            #print(len(tokens)) #16\n",
        "            tokens += [0] * pad_len\n",
        "            #print(tokens) #[101, 2256, 15616, 2024, 1996, 3114, 1997, 2023, 1001, 8372, 2089, 16455, 9641, 2149, 2035, 102, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]\n",
        "            #print(len(tokens)) #160\n",
        "            pad_masks = [1] * len(input_sequence) + [0] * pad_len\n",
        "            #print(pad_masks) #[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]\n",
        "            #print(len(pad_masks)) #160\n",
        "            segment_ids = [0] * sequence_length\n",
        "            #print(segment_ids) #[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]\n",
        "            #print(len(segment_ids)) #160\n",
        "\n",
        "            all_tokens.append(tokens)\n",
        "            all_masks.append(pad_masks)\n",
        "            all_segments.append(segment_ids)\n",
        "        except:\n",
        "            print(text)\n",
        "            return 0\n",
        "\n",
        "    return [np.array(all_tokens, dtype=np.int32), np.array(all_masks, dtype=np.int32), np.array(all_segments, dtype=np.int32)]"
      ],
      "execution_count": 8,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "uef3ysDY-5WL",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "base_model = hub.load('https://tfhub.dev/tensorflow/bert_en_uncased_L-24_H-1024_A-16/1')\n",
        "\n",
        "\n",
        "bert_layer = hub.KerasLayer(base_model)\n",
        "\n",
        "vocab_file = bert_layer.resolved_object.vocab_file.asset_path.numpy()\n",
        "do_lower_case = bert_layer.resolved_object.do_lower_case.numpy()\n",
        "tokenizer = bert_tokenization.FullTokenizer(vocab_file, do_lower_case)"
      ],
      "execution_count": 9,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vxz41zLdUKEO",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "X_train_en = bert_encode(X_train, tokenizer=tokenizer, sequence_length=MAX_SEQ_LEN)\n",
        "X_test_en = bert_encode(X_test, tokenizer=tokenizer, sequence_length=MAX_SEQ_LEN)"
      ],
      "execution_count": 10,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Mwg79thB_Wc7",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 780
        },
        "outputId": "e17fea9d-ab76-43da-f435-b06f63fe2313"
      },
      "source": [
        "cnn_filters = 100\n",
        "DNN_UNITS = 64\n",
        "#checkpoint_filepath = './models/cnn_rnn'\n",
        "# model_checkpoint_callback = tf.keras.callbacks.ModelCheckpoint(\n",
        "#     filepath=checkpoint_filepath,\n",
        "#     save_weights_only=True,\n",
        "#     monitor='acc',\n",
        "#     mode='max',\n",
        "#     save_best_only=True)\n",
        "\n",
        "input_word_ids = tf.keras.layers.Input(shape=(MAX_SEQ_LEN,), dtype=tf.int32,name=\"input_ids\")\n",
        "input_mask = tf.keras.layers.Input(shape=(MAX_SEQ_LEN,), dtype=tf.int32,name=\"input_masks\")\n",
        "input_segment = tf.keras.layers.Input(shape=(MAX_SEQ_LEN,), dtype=tf.int32,name=\"segment_ids\")\n",
        "pooled_output, sequence_output = bert_layer([input_word_ids, input_mask, input_segment])\n",
        "\n",
        "\n",
        "\n",
        "cnn_l1 =  tf.keras.layers.Conv1D(filters=cnn_filters,kernel_size=2,padding=\"same\",activation=\"relu\")(sequence_output)\n",
        "lstm1 = tf.keras.layers.LSTM(32, return_state= False)(cnn_l1)\n",
        "flat1 = tf.keras.layers.Flatten()(lstm1)\n",
        "\n",
        "cnn_l2 = tf.keras.layers.Conv1D(filters=cnn_filters,kernel_size=3,padding=\"same\",activation=\"relu\")(sequence_output)\n",
        "lstm2 = tf.keras.layers.LSTM(32, return_state= False)(cnn_l2)\n",
        "flat2 = tf.keras.layers.Flatten()(lstm2)\n",
        "\n",
        "cnn_l3 = tf.keras.layers.Conv1D(filters=cnn_filters, kernel_size=4,padding=\"same\",activation=\"relu\")(sequence_output)\n",
        "lstm3 = tf.keras.layers.LSTM(32, return_state= False)(cnn_l3)\n",
        "flat3 = tf.keras.layers.Flatten()(lstm3)\n",
        "\n",
        "concat = tf.concat([flat1, flat2, flat3], axis=-1)\n",
        "dense_1 = tf.keras.layers.Dense(units=DNN_UNITS, activation=\"relu\")(concat)\n",
        "\n",
        "\n",
        "pred = tf.keras.layers.Dense(1, activation='sigmoid')(dense_1)\n",
        "\n",
        "model = tf.keras.models.Model(inputs=[input_word_ids,input_mask,input_segment], outputs=pred)\n",
        "model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
        "model.summary()"
      ],
      "execution_count": 11,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"model\"\n",
            "__________________________________________________________________________________________________\n",
            "Layer (type)                    Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            "input_ids (InputLayer)          [(None, 65)]         0                                            \n",
            "__________________________________________________________________________________________________\n",
            "input_masks (InputLayer)        [(None, 65)]         0                                            \n",
            "__________________________________________________________________________________________________\n",
            "segment_ids (InputLayer)        [(None, 65)]         0                                            \n",
            "__________________________________________________________________________________________________\n",
            "keras_layer (KerasLayer)        [(None, 1024), (None 335141889   input_ids[0][0]                  \n",
            "                                                                 input_masks[0][0]                \n",
            "                                                                 segment_ids[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "conv1d (Conv1D)                 (None, None, 100)    204900      keras_layer[0][1]                \n",
            "__________________________________________________________________________________________________\n",
            "conv1d_1 (Conv1D)               (None, None, 100)    307300      keras_layer[0][1]                \n",
            "__________________________________________________________________________________________________\n",
            "conv1d_2 (Conv1D)               (None, None, 100)    409700      keras_layer[0][1]                \n",
            "__________________________________________________________________________________________________\n",
            "lstm (LSTM)                     (None, 32)           17024       conv1d[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "lstm_1 (LSTM)                   (None, 32)           17024       conv1d_1[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "lstm_2 (LSTM)                   (None, 32)           17024       conv1d_2[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "flatten (Flatten)               (None, 32)           0           lstm[0][0]                       \n",
            "__________________________________________________________________________________________________\n",
            "flatten_1 (Flatten)             (None, 32)           0           lstm_1[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "flatten_2 (Flatten)             (None, 32)           0           lstm_2[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "tf_op_layer_concat (TensorFlowO [(None, 96)]         0           flatten[0][0]                    \n",
            "                                                                 flatten_1[0][0]                  \n",
            "                                                                 flatten_2[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "dense (Dense)                   (None, 64)           6208        tf_op_layer_concat[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "dense_1 (Dense)                 (None, 1)            65          dense[0][0]                      \n",
            "==================================================================================================\n",
            "Total params: 336,121,134\n",
            "Trainable params: 979,245\n",
            "Non-trainable params: 335,141,889\n",
            "__________________________________________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9RQiFwDvQExZ",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 758
        },
        "outputId": "c88a926c-df85-4740-90bb-80c81c9e4a5d"
      },
      "source": [
        "tf.keras.utils.plot_model(model)"
      ],
      "execution_count": 13,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<IPython.core.display.Image object>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 13
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5iFFlHXh-6zT",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 104
        },
        "outputId": "5eafcc39-2a86-452e-aabf-3ae8ccf50406"
      },
      "source": [
        "model.fit(X_train_en, y_train, class_weight = class_weights, epochs=2, batch_size=32)"
      ],
      "execution_count": 14,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/2\n",
            "1361/1361 [==============================] - 1225s 900ms/step - loss: 0.5593 - accuracy: 0.7081\n",
            "Epoch 2/2\n",
            "1361/1361 [==============================] - 1225s 900ms/step - loss: 0.5078 - accuracy: 0.7421\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<tensorflow.python.keras.callbacks.History at 0x7f35b1afa128>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 14
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7PxQESh1pyGv",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 225
        },
        "outputId": "3ac251af-24a6-45a3-cb19-d729de95123c"
      },
      "source": [
        "%%time\n",
        "y_prob = model.predict(X_test_en, verbose = 1)\n",
        "y_pred = (y_prob > 0.5).astype(int)\n",
        "print(classification_report(y_test, y_pred, digits = 4))"
      ],
      "execution_count": 15,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "341/341 [==============================] - 298s 875ms/step\n",
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0     0.7030    0.8026    0.7495      5020\n",
            "           1     0.8076    0.7097    0.7555      5862\n",
            "\n",
            "    accuracy                         0.7525     10882\n",
            "   macro avg     0.7553    0.7561    0.7525     10882\n",
            "weighted avg     0.7594    0.7525    0.7527     10882\n",
            "\n",
            "CPU times: user 3min 37s, sys: 1min 57s, total: 5min 34s\n",
            "Wall time: 5min 3s\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9wRQcTiSDcV6",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 337
        },
        "outputId": "d72aea8d-02ec-4019-d738-6b3468f0baeb"
      },
      "source": [
        "from sklearn.metrics import confusion_matrix\n",
        "import seaborn as sns\n",
        "cm = confusion_matrix(y_test, y_pred)\n",
        "sns.heatmap(cm/np.sum(cm), annot=True, cmap='Blues')"
      ],
      "execution_count": 16,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/statsmodels/tools/_testing.py:19: FutureWarning: pandas.util.testing is deprecated. Use the functions in the public API at pandas.testing instead.\n",
            "  import pandas.util.testing as tm\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.axes._subplots.AxesSubplot at 0x7f35aec215f8>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 16
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAWQAAAD4CAYAAADbyJysAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAXi0lEQVR4nO3de3RU5bnH8e8zEyPeRYSoBLkoXkAoKhdbj4AKCtWCihe0nsqpGmul2mpVLFaXKC5F6qUFCzke1NYqBbUaFUW81yotURQMyjJECokSqKBgQSHJc/7ImE5CkpmQSbL35vdxvYvZe79773e68NfHd1/G3B0REWl7sbYegIiIVFMgi4gEhAJZRCQgFMgiIgGhQBYRCYislj7BbsdP1G0csp0Nr09u6yFIALXLwpp7jN2OHp925mxZPK3Z58skVcgiIgHR4hWyiEirsvDWmQpkEYmWWLytR7DDFMgiEi0WqGnhJlEgi0i0aMpCRCQgVCGLiASEKmQRkYBQhSwiEhC6y0JEJCA0ZSEiEhCashARCQhVyCIiAaFAFhEJiLgu6omIBIPmkEVEAkJTFiIiAaEKWUQkIFQhi4gEhCpkEZGACPGj0+Gt7UVE6mOx9FuqQ5mNMLPlZlZsZhPq2f4TM1tqZu+Z2Ztm1iuxvpuZbUmsf8/MZqQzdFXIIhItGZqyMLM4MB0YDpQCi8yswN2XJXV71N1nJPqPAu4GRiS2rXD3fk05pypkEYmWzFXIA4Fidy9x963AbGB0cgd335i0uAfgzRm6AllEoqUJgWxmeWZWmNTyko7UGVidtFyaWFf7dGZXmNkKYApwZdKm7ma22MxeN7MT0hm6pixEJFqacFHP3fOB/Oaczt2nA9PN7ALgRuAi4DPgYHf/3MyOBZ4ys951KurtqEIWkWgxS781rgzokrScm1jXkNnAGQDu/o27f574/A6wAjgs1QkVyCISLZmbQ14E9DSz7maWDYwFCmqdyqxn0uJpwMeJ9R0TFwUxsx5AT6Ak1Qk1ZSEi0ZKhuyzcvcLMxgPzgTgwy92LzGwSUOjuBcB4MxsGbAM2UD1dATAYmGRm24Aq4Cfuvj7VORXIIhIplsEn9dx9HjCvzrqbkj5f1cB+TwBPNPV8CmQRiZRMBnJrUyCLSKRYTIEsIhIIqpBFRAJCgSwiEhAKZBGRoAhvHiuQRSRaVCGLiARELBbeB5AVyCISKaqQRUSCIrx5rEAWkWhRhSwiEhAKZBGRgNCj0yIiAaEKWUQkIBTIIiIBoUAWEQkIBbKISFCEN48VyCISLXp0WkQkIDRlISISFOHNYwVypgwf1JOpPz+NeCzGQ88UMvWRN2ptv+SMgVx21iAqq5x/b/6GK6Y8xUcr1zH2lO/w8wtOqOnX55Acvvvj+1ny8Wet/RUkA/721ze4847JVFVWceaYc7j40rxa27du3crEG67jw6Ii9tl3X6b85h46d85l29atTLrlZpYVfUDMjOtumMiAgYMA+N199/BMwVNs/HIjCwsXt8XXCpUwV8jhnWwJkFjMuPeaHzD6moc5+of3cc6wvhzRrWOtPn9+8X0G/Oh3HDduGnc/+lfu/Nn3AZj94vscN24ax42bxsWT5rLysw0K45CqrKzk9smTuH/GA/yl4DlemPcsK4qLa/X5yxNz2XvvvXn2hQVc+KNx3Hv3VACeeHxu9Z9PPcOMBx7kN3fdSVVVFQBDhp7In2bPbd0vE2JmlnYLmpSBbGZHmNn1ZvbbRLvezI5sjcGFxYAjc1lRup6Vn25gW0Ulc19ewukn1P6faNPmb2o+79EuG/ftj3Pu8L7MfWlpSw9XWsgHS5fQpUtXcrt0YZfsbEZ8/zRee/XlWn1efeUVRo0+E4Dhp5zKPxa+jbtTsqKYgYOqK+IOHTqw1157UfTBBwD0/U4/Onbs1LpfJsQiG8hmdj0wm+pZmX8kmgGPmdmElh9eOBzUcW9K135Zs1y2diOdO+6zXb/LzhpE0ZyrmfzTU7nm3me32372yX2Ys+D9Fh2rtJy15eUccOABNcudcnIoLy+v3WdtOQcccCAAWVlZ7LnXXnzxxQYOO/wIXn/1FSoqKigtXc2Hy4ooX6P/UtoRFrO0W9CkmkO+GOjt7tuSV5rZ3UARcEd9O5lZHpAHkNVjJFkHHJ2BoYbfzCf/zswn/855w/syYdxQLr3tiZptA3rlsvnrbSz7ZG0bjlDayhlnjeGTkhVccO4YDjzoIL7T72hi8XhbDyuUglj5pivVlEUVcFA96w9MbKuXu+e7e393778zhPGn6zaS2+k/FXHnTntTtu7LBvvPeWkpPzihV6115wzry5yXlrTYGKXldcrJYc1na2qW15aXk5OTU7tPpxzWJCrfiooKvtq0iX33bU9WVhbXTvgVc558mvum/Z5NmzbRtWu31hx+ZER2ygL4OfCymT1vZvmJ9gLwMnBVyw8vHAo/KuPQ3A50PbA9u2TFOefkvjz35ke1+hyS26Hm88jvHU5x6ec1y2bGmJP6MFeBHGq9j+rDqlUrKS1dzbatW3lh3nMMOfGkWn2GnngSBU//BYAFL85n4KDjMDO2bNnC5s2bAXj7rb8Rj8c55NBDW/07RIFZ+i1oGp2ycPcXzOwwYCDQObG6DFjk7pUtPbiwqKys4hf3PMMzd48jHjcefvZdPvxkLb++5GTe/aiM5978iMvHHMeJAw5hW0UVX2zawqW3PV6z/3/160bp2i9Y+emGNvwW0lxZWVncMPEmLs+7hKqqSs44cwyHHtqT6b+7j969j2LoSSdz5pizmTjhWk4fMZy999mHKVPvAWD9+s+5PO9iYrEYnTrlMPmOKTXHvWfqFObNe5avv97C8JMGc9aYc7j8ip+11dcMvCBWvukyr+9yfwbtdvzElj2BhNKG1ye39RAkgNplNf+xjsOvn5925iy/89RGz2dmI4D7gDjwgLvfUWf7T4ArgErgKyDP3Zcltt1A9XW4SuBKd5+fajy6D1lEIiVTUxZmFgemAyOBXsD5ZtarTrdH3b2Pu/cDpgB3J/btBYwFegMjgPsTx2uUAllEIiUWs7RbCgOBYncvcfetVN8CPDq5g7tvTFrcA/i2Oh8NzHb3b9z9E6A4cbxG6dFpEYmUpkwhJ9+im5Dv7vmJz52B1UnbSoFB9RzjCuBqIBv49ipuZ2BhnX07k4ICWUQipSkX9RLhm5+yY+PHmA5MN7MLgBuBi3b0WJqyEJFIyeBtb2VAl6Tl3MS6hswGztjBfQEFsohETCwWS7ulsAjoaWbdzSyb6ot0BckdzKxn0uJpwMeJzwXAWDPb1cy6Az2pfvVEozRlISKRkqnbkN29wszGA/Opvu1tlrsXmdkkoNDdC4DxZjYM2AZsIDFdkeg3B1gGVABXpPPshgJZRCIlkw+GuPs8YF6ddTclfW7wiWV3nww06YZ7BbKIREqIH9RTIItItIT50WkFsohESojzWIEsItGSxhN4gaVAFpFI0ZSFiEhAhDiPFcgiEi2qkEVEAiLEeaxAFpFo0UU9EZGA0JSFiEhAKJBFRAIixHmsQBaRaFGFLCISECHOYwWyiESL7rIQEQmIWIhLZAWyiERKiPNYgSwi0aKLeiIiARHiKWQFsohEiy7qiYgEhKFAFhEJhBAXyApkEYkWXdQTEQmIEOexAllEokUPhoiIBITushARCYgQF8gKZBGJljBPWcTaegAiIplkTWgpj2U2wsyWm1mxmU2oZ/vVZrbMzJaY2ctm1jVpW6WZvZdoBemMXRWyiERKpm57M7M4MB0YDpQCi8yswN2XJXVbDPR3981mdjkwBTgvsW2Lu/dryjlVIYtIpMQs/ZbCQKDY3UvcfSswGxid3MHdX3X3zYnFhUBus8benJ1FRIImFrO0m5nlmVlhUstLOlRnYHXScmliXUMuBp5PWm6XOOZCMzsjnbFrykJEIqUpUxbung/kZ+CcFwL9gSFJq7u6e5mZ9QBeMbOl7r6iseMokEUkUjJ4G3IZ0CVpOTexrhYzGwZMBIa4+zffrnf3ssSfJWb2GnA00Ggga8pCRCLFzNJuKSwCeppZdzPLBsYCte6WMLOjgZnAKHdfm7S+vZntmvi8P3A8kHwxsF6qkEUkUjJVILt7hZmNB+YDcWCWuxeZ2SSg0N0LgLuAPYG5iYBf5e6jgCOBmWZWRXXhe0eduzPqpUAWkUiJZ3DOwt3nAfPqrLsp6fOwBvZ7C+jT1PMpkEUkUvT6TRGRgAhxHiuQRSRawvwuCwWyiERKiPO45QP5r49c39KnkBBqP2B8Ww9BAmjL4mnNPobmkEVEAiKuQBYRCYYQ/2CIAllEokWBLCISEJpDFhEJCFXIIiIBEeICWYEsItGSFeJEViCLSKSEOI8VyCISLXp0WkQkIEKcxwpkEYkW3WUhIhIQmXxBfWtTIItIpIQ4jxXIIhItlrFf1Wt9CmQRiRRVyCIiAaFAFhEJCL1cSEQkIOKxth7BjlMgi0ik6Ek9EZGA0ByyiEhAhLhAViCLSLTEdB+yiEgwhLlCDvH1SBGR7WXFLO2WipmNMLPlZlZsZhPq2X61mS0zsyVm9rKZdU3adpGZfZxoF6UzdgWyiESKWfqt8eNYHJgOjAR6AeebWa863RYD/d29L/A4MCWx737AzcAgYCBws5m1TzV2BbKIRErMLO2WwkCg2N1L3H0rMBsYndzB3V91982JxYVAbuLzqcACd1/v7huABcCIlGNvwvcUEQm8plTIZpZnZoVJLS/pUJ2B1UnLpYl1DbkYeH4H9wV0UU9EIqYpVaa75wP5zT2nmV0I9AeGNOc4qpBFJFIyOGVRBnRJWs5NrKvFzIYBE4FR7v5NU/bdbuypOoiIhEkGA3kR0NPMuptZNjAWKEjuYGZHAzOpDuO1SZvmA6eYWfvExbxTEusapSkLEYmUTN2G7O4VZjae6iCNA7PcvcjMJgGF7l4A3AXsCcxNvGVulbuPcvf1ZnYr1aEOMMnd16c6pwJZRCIlkw+GuPs8YF6ddTclfR7WyL6zgFlNOZ8CWUQiRe9DFhEJiDBfGFMgi0ik6H3IIiIBoSkLEZGA0JSFiEhAqEIWEQmI8MaxAllEIiauCllEJBhCnMcKZBGJFgvxpIUCWUQiRRWyiEhA6FenRUQCQhWyiEhA6NFpEZGAiIU3jxXIIhItustCRCQgQjxjoUDOlPcL3+KPv/8NVVVVDB0xmlHnjau1/cOl7/LIjLtZ9Ukx42+YzKATTq7Z9q+1a/jfe29j/bpyMOO6SffS8YCDWvkbSEsY/r0jmXrt2cRjMR566i2mPrig1vZLzv4vLjt3MJVVVfx78zdccdtjfFSyhqysGL+/6Yf0O6ILWfEYf3ruH0yd9WIbfYtwUYW8k6uqrOSh6VO44fZp7Ld/Dr++8iKOOW4wuV171PTZv+MBXHbNzTz3xCPb7T/jrpsZff6P6XPMIL7eshmzML+vSr4Vixn3TjiX0y6fRln5F7z5p2t59vWlfFSypqbPn58v5IHH3wTgtCF9uPPqsxg9/n7GDDuGXbOzGHDu7ezWbhcWP3Ejc54vZNVnKX+WbacX5jlk/ZufASuWF5FzYBc6HZhL1i67cNyQ4bzz9uu1+nQ84CAO7tFzuzdRlf6zhMrKSvocMwiAdrvtzq7t2rXa2KXlDDiqGytW/4uVZZ+zraKSufPf5fShfWv12fTvr2s+77FbNo4D4Di7t8smHo+x267ZbN1WWauvNCyDvzrd6lQhZ8D6z9fRoWNOzfJ+++ewYvkHae27pmwVu++5F/dMupZ15Z9yVL+BjP3xeGLxeEsNV1rJQZ32obR8Q81yWfkGBh7Vbbt+l507mCsvPJHsXbIYcdlvAXjypcWcPrQvnyyYzO7tsrlu6pNs2Li5tYYeasGL2fTtcIVsZv/TyLY8Mys0s8InH3twR0+xU6isrGT5B4v54aVXcetvH2btmjLeWPBsWw9LWtHMOW/Qe9Qt3Hjf00y4ZAQAA3p3o7Kyih6nTOTI027mqv8+iW6dO7TxSMMhzBVyc6Ysbmlog7vnu3t/d+9/1vkN5nZk7NehI5+vK69ZXv+vctp36Jjevvt3oushh9HpwFzi8SyO/e5QPin+qKWGKq3o07VfkpvTvma5c057ytZ92WD/OfPf4QeJKY1zR/bnxbeWUVFRxboNX/H2eyUc2+vgFh9zFFgTWtA0GshmtqSBthTIaWzfnUmPw3ux5tNVrF1TRsW2bSx8fQHHHjc4rX0POawXm7/6io1fVP+n7bL3F9H54O4tOVxpJYVF/+TQgzvS9aAO7JIV55xTj+G515bU6nPIwf/5P+6RJ/SmePU6AErXrGfogMMB2L1dNgP7dmP5ynIkDSFO5FRzyDnAqcCGOusNeKtFRhRC8XgW4356HXdOvJKqqkqGnDKK3G6H8PgfZtC955Ec+90hrFhexD23XsfmTRtZ/Pc3eeKPM5mSP4dYPM4Fl17F7RN+iuN0P/QIThp5Zlt/JcmAysoqfnHnHJ65/wriMePhpxfyYckafn35aby7bBXPvb6Uy88bzImDjmBbRSVfbNzMpb/+AwAz/vwG+bdcyDuPT8QM/vj0Qj74+NM2/kbhEMSpiHSZuze80ez/gAfd/c16tj3q7hekOkHhJxsbPoHstE4461dtPQQJoC2LpzU7TReVfJl25gzosU+g0rvRCtndL25kW8owFhFpdYGK2KbRbW8iEil6Uk9EJCBCPIWsJ/VEJFoyeZOFmY0ws+VmVmxmE+rZPtjM3jWzCjM7u862SjN7L9EK0hm7KmQRiZS6rydoxnHiwHRgOFAKLDKzAndfltRtFTAO+GU9h9ji7v2ack4FsohESganLAYCxe5eUn1cmw2MBmoC2d1XJrZVZeKEmrIQkUhpypRF8mseEi0v6VCdgdVJy6WJdelqlzjmQjM7I50dVCGLSLQ0oUJ293wgv4VG0tXdy8ysB/CKmS119xWN7aAKWUQixZrwTwplQJek5dzEurS4e1nizxLgNeDoVPsokEUkUszSbyksAnqaWXczywbGAmndLWFm7c1s18Tn/YHjSZp7bogCWUQiJVOB7O4VwHhgPvAhMMfdi8xskpmNqj6XDTCzUuAcYKaZFSV2PxIoNLP3gVeBO+rcnVEvzSGLSKRk8kk9d58HzKuz7qakz4uonsqou99bQJ+mnk+BLCKREuYn9RTIIhIpIc5jBbKIREyIE1mBLCKREuYX1CuQRSRSwhvHCmQRiZoQJ7ICWUQiRS+oFxEJiBBPISuQRSRaQpzHCmQRiZZMvaC+LSiQRSRSQpzHCmQRiZYQ57ECWUQiJsSJrEAWkUjRbW8iIgGhOWQRkYCIKZBFRIIivImsQBaRSNGUhYhIQIQ4jxXIIhItqpBFRAJCj06LiAREeONYgSwiERPiAlmBLCLRoif1RESCIrx5rEAWkWgJcR4rkEUkWmIhnkRWIItIpIQ4j4m19QBERILKzEaY2XIzKzazCfVsH2xm75pZhZmdXWfbRWb2caJdlM75VCGLSKRkqkI2szgwHRgOlAKLzKzA3ZcldVsFjAN+WWff/YCbgf6AA+8k9t3Q2DlVIYtIpFgT/klhIFDs7iXuvhWYDYxO7uDuK919CVBVZ99TgQXuvj4RwguAEalOqEAWkUgxa0qzPDMrTGp5SYfqDKxOWi5NrEvHDu2rKQsRiZSmTFm4ez6Q32KDaSJVyCISKRmcsigDuiQt5ybWpWOH9lUgi0ikNGXKIoVFQE8z625m2cBYoCDNYcwHTjGz9mbWHjglsa5RCmQRiRRrQmuMu1cA46kO0g+BOe5eZGaTzGwUgJkNMLNS4BxgppkVJfZdD9xKdagvAiYl1jVKc8giEi0ZfDDE3ecB8+qsuynp8yKqpyPq23cWMKsp51Mgi0ikhPnRaXP3th7DTsPM8hJXdUVq6O+FfEtzyK0rL3UX2Qnp74UACmQRkcBQIIuIBIQCuXVpnlDqo78XAuiinohIYKhCFhEJCAWyiEhAKJBbSapfHpCdj5nNMrO1ZvZBW49FgkGB3AqSfnlgJNALON/MerXtqCQAHiKNl5bLzkOB3DpS/vKA7Hzc/Q0g5QtnZOehQG4dzfnlARHZSSiQRUQCQoHcOprzywMispNQILeO5vzygIjsJBTIraChXx5o21FJWzOzx4C3gcPNrNTMLm7rMUnb0qPTIiIBoQpZRCQgFMgiIgGhQBYRCQgFsohIQCiQRUQCQoEsIhIQCmQRkYD4f6NHR6AU81kZAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "g0nIaa00NBBf",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 176
        },
        "outputId": "1951ca89-55e3-4ef5-d501-8bfba9aeb4b1"
      },
      "source": [
        "model.save('./models/cnn_rnn')"
      ],
      "execution_count": 17,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/resource_variable_ops.py:1817: calling BaseResourceVariable.__init__ (from tensorflow.python.ops.resource_variable_ops) with constraint is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "If using Keras pass *_constraint arguments to layers.\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/resource_variable_ops.py:1817: calling BaseResourceVariable.__init__ (from tensorflow.python.ops.resource_variable_ops) with constraint is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "If using Keras pass *_constraint arguments to layers.\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "INFO:tensorflow:Assets written to: ./models/cnn_rnn/assets\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "INFO:tensorflow:Assets written to: ./models/cnn_rnn/assets\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "nrXAw__8a0IR",
        "colab_type": "text"
      },
      "source": [
        "# Submit"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "TILQlrFRObvZ",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "new_model = tf.keras.models.load_model('./models/rnn')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nIYjVENSfZR5",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "df = pd.read_csv('./data/test_clean.csv', index_col = False)\n",
        "df['text'] = df['text'].fillna(\"\")"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-DpdmXVKg5Zn",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "X_data = df['text'].to_numpy()\n",
        "X_data_en = bert_encode(X_data, tokenizer=tokenizer, sequence_length=80)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VkzWzuc0f3bh",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 52
        },
        "outputId": "415a55dd-5695-47e0-8131-083d5b5543f3"
      },
      "source": [
        "%%time\n",
        "\n",
        "y_prob = model.predict(X_data_en, verbose = 1)\n",
        "y_pred = (y_prob > 0.5).astype(int)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "CPU times: user 4min 6s, sys: 3min 22s, total: 7min 29s\n",
            "Wall time: 19min 12s\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VIbfyazZkJqR",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "submission = pd.read_csv('./data/sample_submit.csv', index_col = False)\n",
        "\n",
        "submission[\"target\"] = y_pred"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-eYPu-0Djk9Y",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "submission.to_csv(\"./results/result1.csv\", index = False)\n",
        "submission.head(10)"
      ],
      "execution_count": null,
      "outputs": []
    }
  ],
  "metadata": {
    "colab": {
      "name": "MODEL.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "nbformat": 4,
  "nbformat_minor": 0
}