5.TEXT_SIM/5.3.2.Quora_CNN.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 모듈 불러오기"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import json\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import tensorflow as tf\n",
    "import matplotlib.pyplot as plt\n",
    "from tensorflow.keras import layers\n",
    "from tensorflow.keras.callbacks import EarlyStopping, ModelCheckpoint"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 시각화 함수"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_graphs(history, string):\n",
    "    plt.plot(history.history[string])\n",
    "    plt.plot(history.history['val_'+string], '')\n",
    "    plt.xlabel(\"Epochs\")\n",
    "    plt.ylabel(string)\n",
    "    plt.legend([string, 'val_'+string])\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 학습 데이터 경로 정의"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "DATA_IN_PATH = './data_in/'\n",
    "DATA_OUT_PATH = './data_out/'\n",
    "TRAIN_Q1_DATA_FILE = 'train_q1.npy'\n",
    "TRAIN_Q2_DATA_FILE = 'train_q2.npy'\n",
    "TRAIN_LABEL_DATA_FILE = 'train_label.npy'\n",
    "DATA_CONFIGS = 'data_configs.json'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 랜덤 시드 고정"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "SEED_NUM = 1234\n",
    "tf.random.set_seed(SEED_NUM)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 파일 로드"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "q1_data = np.load(open(DATA_IN_PATH + TRAIN_Q1_DATA_FILE, 'rb'))\n",
    "q2_data = np.load(open(DATA_IN_PATH + TRAIN_Q2_DATA_FILE, 'rb'))\n",
    "labels = np.load(open(DATA_IN_PATH + TRAIN_LABEL_DATA_FILE, 'rb'))\n",
    "prepro_configs = json.load(open(DATA_IN_PATH + DATA_CONFIGS, 'r'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 모델 하이퍼파라메터 정의"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_name = 'cnn_similarity'\n",
    "BATCH_SIZE = 1024\n",
    "NUM_EPOCHS = 100\n",
    "VALID_SPLIT = 0.1\n",
    "MAX_LEN = 31\n",
    "\n",
    "kargs = {'model_name': model_name,\n",
    "         'vocab_size': prepro_configs['vocab_size'],\n",
    "         'word_embedding_dimension': 100,\n",
    "         'conv_num_filters': 300,\n",
    "         'conv_window_size': 3,\n",
    "         'max_pool_seq_len': MAX_LEN,\n",
    "         'sent_embedding_dimension': 128,\n",
    "         'dropout_rate': 0.2,\n",
    "         'hidden_dimension': 200,\n",
    "         'output_dimension':1}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 모델 선언 및 컴파일"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "class SentenceEmbedding(layers.Layer):\n",
    "    def __init__(self, **kargs):\n",
    "        super(SentenceEmbedding, self).__init__()\n",
    "\n",
    "        self.conv = layers.Conv1D(kargs['conv_num_filters'], kargs['conv_window_size'], \n",
    "                                activation=tf.keras.activations.relu, \n",
    "                                padding='same')\n",
    "        self.max_pool = layers.MaxPool1D(kargs['max_pool_seq_len'], 1)\n",
    "        self.dense = layers.Dense(kargs['sent_embedding_dimension'], \n",
    "                              activation=tf.keras.activations.relu)\n",
    "\n",
    "    def call(self, x):\n",
    "        x = self.conv(x)\n",
    "        x = self.max_pool(x)\n",
    "        x = self.dense(x)\n",
    "        \n",
    "        return tf.squeeze(x, 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "class SentenceSimilarityModel(tf.keras.Model):\n",
    "    def __init__(self, **kargs):\n",
    "        super(SentenceSimilarityModel, self).__init__(name=kargs['model_name'])\n",
    "        \n",
    "        self.word_embedding = layers.Embedding(kargs['vocab_size'], kargs['word_embedding_dimension'])\n",
    "        self.base_encoder = SentenceEmbedding(**kargs)\n",
    "        self.hypo_encoder = SentenceEmbedding(**kargs)\n",
    "        self.dense = layers.Dense(kargs['hidden_dimension'], \n",
    "                                           activation=tf.keras.activations.relu)\n",
    "        self.logit = layers.Dense(1, activation=tf.keras.activations.sigmoid)\n",
    "        self.dropout = layers.Dropout(kargs['dropout_rate'])\n",
    "        \n",
    "    def call(self, x):\n",
    "        x1, x2 = x\n",
    "        b_x = self.word_embedding(x1)\n",
    "        h_x = self.word_embedding(x2)\n",
    "        b_x = self.dropout(b_x)\n",
    "        h_x = self.dropout(h_x)\n",
    "        \n",
    "        b_x = self.base_encoder(b_x)\n",
    "        h_x = self.hypo_encoder(h_x)\n",
    "        \n",
    "        e_x = tf.concat([b_x, h_x], -1)\n",
    "        e_x = self.dense(e_x)\n",
    "        e_x = self.dropout(e_x)\n",
    "        \n",
    "        return self.logit(e_x)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = SentenceSimilarityModel(**kargs)\n",
    "\n",
    "model.compile(optimizer=tf.keras.optimizers.Adam(1e-3),\n",
    "              loss=tf.keras.losses.BinaryCrossentropy(),\n",
    "              metrics=[tf.keras.metrics.BinaryAccuracy(name='accuracy')])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Callback 선언"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "./data_out/cnn_similarity -- Folder create complete \n",
      "\n"
     ]
    }
   ],
   "source": [
    "# overfitting을 막기 위한 ealrystop 추가j\n",
    "earlystop_callback = EarlyStopping(monitor='val_accuracy', min_delta=0.0001, patience=1)\n",
    "# min_delta: the threshold that triggers the termination (acc should at least improve 0.0001)\n",
    "# patience: no improvment epochs (patience = 1, 1번 이상 상승이 없으면 종료)\\\n",
    "\n",
    "checkpoint_path = DATA_OUT_PATH + model_name + '/weights.h5'\n",
    "checkpoint_dir = os.path.dirname(checkpoint_path)\n",
    "\n",
    "# Create path if exists\n",
    "if os.path.exists(checkpoint_dir):\n",
    "    print(\"{} -- Folder already exists \\n\".format(checkpoint_dir))\n",
    "else:\n",
    "    os.makedirs(checkpoint_dir, exist_ok=True)\n",
    "    print(\"{} -- Folder create complete \\n\".format(checkpoint_dir))\n",
    "    \n",
    "\n",
    "cp_callback = ModelCheckpoint(\n",
    "    checkpoint_path, monitor='val_accuracy', verbose=1, save_best_only=True, save_weights_only=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 모델 학습"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 268673 samples, validate on 29853 samples\n",
      "Epoch 1/100\n",
      "268288/268673 [============================>.] - ETA: 0s - loss: 0.5500 - accuracy: 0.7169\n",
      "Epoch 00001: val_accuracy improved from -inf to 0.68703, saving model to ./data_out/cnn_similarity/weights.h5\n",
      "268673/268673 [==============================] - 33s 121us/sample - loss: 0.5500 - accuracy: 0.7170 - val_loss: 0.5428 - val_accuracy: 0.6870\n",
      "Epoch 2/100\n",
      "268288/268673 [============================>.] - ETA: 0s - loss: 0.4452 - accuracy: 0.7903\n",
      "Epoch 00002: val_accuracy improved from 0.68703 to 0.70814, saving model to ./data_out/cnn_similarity/weights.h5\n",
      "268673/268673 [==============================] - 32s 118us/sample - loss: 0.4452 - accuracy: 0.7903 - val_loss: 0.5730 - val_accuracy: 0.7081\n",
      "Epoch 3/100\n",
      "268288/268673 [============================>.] - ETA: 0s - loss: 0.3545 - accuracy: 0.8397\n",
      "Epoch 00003: val_accuracy improved from 0.70814 to 0.81345, saving model to ./data_out/cnn_similarity/weights.h5\n",
      "268673/268673 [==============================] - 31s 114us/sample - loss: 0.3545 - accuracy: 0.8397 - val_loss: 0.4269 - val_accuracy: 0.8135\n",
      "Epoch 4/100\n",
      "268288/268673 [============================>.] - ETA: 0s - loss: 0.2715 - accuracy: 0.8819\n",
      "Epoch 00004: val_accuracy did not improve from 0.81345\n",
      "268673/268673 [==============================] - 31s 114us/sample - loss: 0.2716 - accuracy: 0.8819 - val_loss: 0.5208 - val_accuracy: 0.7964\n"
     ]
    }
   ],
   "source": [
    "history = model.fit((q1_data, q2_data), labels, batch_size=BATCH_SIZE, epochs=NUM_EPOCHS,\n",
    "                    validation_split=VALID_SPLIT, callbacks=[earlystop_callback, cp_callback])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 결과 플롯"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_graphs(history, 'accuracy')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_graphs(history, 'loss')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 테스트 데이터 불러오기"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "TEST_Q1_DATA_FILE = 'test_q1.npy'\n",
    "TEST_Q2_DATA_FILE = 'test_q2.npy'\n",
    "TEST_ID_DATA_FILE = 'test_id.npy'\n",
    "\n",
    "test_q1_data = np.load(open(DATA_IN_PATH + TEST_Q1_DATA_FILE, 'rb'))\n",
    "test_q2_data = np.load(open(DATA_IN_PATH + TEST_Q2_DATA_FILE, 'rb'))\n",
    "test_id_data = np.load(open(DATA_IN_PATH + TEST_ID_DATA_FILE, 'rb'), allow_pickle=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 베스트 모델 불러오기"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "SAVE_FILE_NM = 'weights.h5'\n",
    "model.load_weights(os.path.join(DATA_OUT_PATH, model_name, SAVE_FILE_NM))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 테스트 데이터 예측하기"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "predictions = model.predict((test_q1_data, test_q2_data), batch_size=BATCH_SIZE)\n",
    "predictions = predictions.squeeze(-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "output = pd.DataFrame( data={\"test_id\":test_id_data, \"is_duplicate\": list(predictions)} )\n",
    "output.to_csv(\"cnn_predict.csv\", index=False, quoting=3)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}