English2Korean.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# English 2 Korean"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Prepare special letter"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "space_letter = ' '\n",
    "special_letters = [space_letter]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## ~~Prepare Korean data~~\n",
    "모든 한국어 글자를 사용하기엔 너무 많음. 데이터에서 나오는 글자만 사용함.\n",
    "(하단 셀 참조)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Prepare Korean letters\n",
    "first_id = ord('가')\n",
    "last_id = ord('힣')\n",
    "\n",
    "korean_letter_list = [chr(num) for num in range(first_id, last_id+1)]\n",
    "\n",
    "## Valid checking\n",
    "try:\n",
    "    assert len(korean_letter_list) == 11172\n",
    "except AssertionError:\n",
    "    print('Wrong!')\n",
    "    raise\n",
    "else:\n",
    "    print('Cool!')\n",
    "    \n",
    "# Extend special letters\n",
    "korean_letter_list.extend(special_letters)\n",
    "\n",
    "# Make a mapping with index\n",
    "idx2korean = dict(zip(range(len(korean_letter_list)), korean_letter_list))\n",
    "korean2idx = dict(zip(korean_letter_list, range(len(korean_letter_list))))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Prepare English data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Prepare English letters\n",
    "first_big_id = ord('A')\n",
    "last_big_id = ord('Z')\n",
    "first_small_id = ord('a')\n",
    "last_small_id = ord('z')\n",
    "\n",
    "big = list(range(first_big_id, last_big_id+1))\n",
    "small = list(range(first_small_id, last_small_id+1))\n",
    "\n",
    "english_letter_list = [chr(num) for num in big+small]\n",
    "    \n",
    "# Extend special letters\n",
    "english_letter_list.extend(special_letters)\n",
    "    \n",
    "# Make a mapping with index\n",
    "idx2english = dict(enumerate(english_letter_list, 1))\n",
    "english2idx = {v: k for k, v in idx2english.items()}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load dataset from csv"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "\n",
    "df = pd.read_csv('kcrong.csv', encoding='utf8')\n",
    "del df['Unnamed: 0']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# Load X\n",
    "roma = df['roma'].apply(lambda x: np.array([english2idx[letter] for letter in x]))\n",
    "roma_df = pd.DataFrame(roma)\n",
    "X = pd.DataFrame(roma_df['roma'].tolist()).values\n",
    "\n",
    "# Replace nan to 0\n",
    "# Get index of nan, and make them 0\n",
    "X[np.isnan(X)] = 0\n",
    "print(X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "X.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Build korean letter set\n",
    "\n",
    "# 한국어의 모든 글자를 사용하기엔 너무 많음. (가~힣 약 11175개)\n",
    "# 따라서 이미 등장한 단어만 사용\n",
    "\n",
    "# 데이터 셋에서 모든 한글 값 가져옴\n",
    "hang = df['hang'].values\n",
    "\n",
    "# 등장한 모든 글자의 리스트를 가져오고, set() 으로 중복 제거\n",
    "korean_letter_list = list(set([letter for sent in hang for letter in sent]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# Make a mapping with index\n",
    "idx2korean = dict(enumerate(korean_letter_list, 1))\n",
    "korean2idx = {v: k for k, v in idx2korean.items()}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "hang = df['hang'].apply(lambda x: np.array([korean2idx[letter] for letter in x]))\n",
    "hang_df = pd.DataFrame(hang)\n",
    "y = pd.DataFrame(hang_df['hang'].tolist()).values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Replace nan to 0\n",
    "# Get index of nan, and make them 0\n",
    "y[np.isnan(y)] = 0\n",
    "print(y)\n",
    "print(y.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# one-hot encode y\n",
    "from keras.utils import np_utils\n",
    "\n",
    "y = np_utils.to_categorical(y, num_classes=len(korean_letter_list)+1).reshape(300, 12, len(korean_letter_list)+1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "y.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3)\n",
    "X_train, X_val, y_train, y_val = train_test_split(X_train, y_train, test_size=0.1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Hyper Parameters"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.layers import recurrent\n",
    "\n",
    "# Parameters for the model and dataset\n",
    "TRAINING_SIZE = 50000\n",
    "VOCAB_SIZE = 12\n",
    "INVERT = True\n",
    "HIDDEN_SIZE = 200\n",
    "BATCH_SIZE = 100\n",
    "LAYERS = 2\n",
    "MAX_EPOCHS = 1000\n",
    "EMBEDDING_OUTPUT_SIZE = 128\n",
    "MAX_SENT_LENGTH = X.shape[1] + 1\n",
    "\n",
    "# Try replacing with LSTM, GRU, or SimpleRNN\n",
    "RNN = recurrent.GRU\n",
    "stop_monitor = 'val_acc'  # variable for early stop: (default = val_loss)\n",
    "stop_delta = 0.0  # minimum delta before early stop (default = 0)\n",
    "stop_epochs = 20    # how many epochs to do after stop condition (default = 0)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Model visualize function"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "from IPython.display import SVG, display\n",
    "from keras.utils.vis_utils import model_to_dot\n",
    "\n",
    "def viz_model(model):\n",
    "    display(SVG(model_to_dot(model, show_shapes=True).create(prog='dot', format='svg')))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Build Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Done\n"
     ]
    }
   ],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, TimeDistributed, Activation, RepeatVector, Embedding, Dropout\n",
    "\n",
    "def build_model(embedding_input_dim,\n",
    "                embedding_output_dim,\n",
    "                embedding_input_length,\n",
    "                maximum_input_length,\n",
    "                hidden_size,\n",
    "                maximum_output_length,\n",
    "                encoder_layer_count,\n",
    "                dropout_rate,\n",
    "                output_classes_count,\n",
    "                rnn_layer,\n",
    "                summary_type=False):\n",
    "    \n",
    "    print('Build model...')\n",
    "    RNN = rnn_layer\n",
    "\n",
    "    model = Sequential()\n",
    "\n",
    "    model.add(Embedding(\n",
    "        input_dim=embedding_input_dim,\n",
    "        output_dim=embedding_output_dim,\n",
    "        input_length=embedding_input_length\n",
    "    ))\n",
    "    \n",
    "    model.add(RNN(hidden_size, return_sequences=True, input_shape=(maximum_input_length, 123)))\n",
    "    model.add(RNN(hidden_size))\n",
    "\n",
    "    model.add(RepeatVector(maximum_output_length)) # 글자수 맥시멈\n",
    "\n",
    "    for _ in range(encoder_layer_count):\n",
    "        model.add(RNN(hidden_size, return_sequences=True))\n",
    "        model.add(Dropout(dropout_rate))\n",
    "\n",
    "    # For each of step of the output sequence, decide which character should be chosen\n",
    "    model.add(TimeDistributed(Dense(output_classes_count)))\n",
    "    model.add(Activation('softmax'))\n",
    "\n",
    "    model.compile(loss='categorical_crossentropy', # default: categorical_crossentropy\n",
    "                  optimizer='adam', # default: adam, can try SGD, RMSprop\n",
    "                  metrics=['accuracy'])\n",
    "    \n",
    "    \n",
    "    \n",
    "    if summary_type is False:\n",
    "        pass\n",
    "    elif str(summary_type).lower() == 'svg':\n",
    "        viz_model(model)\n",
    "    else:\n",
    "        model.summary()\n",
    "        \n",
    "    return model\n",
    "\n",
    "print('Done')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## ~~NO~~ Early Stop"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Early Stop\n",
    "from keras.callbacks import EarlyStopping\n",
    "\n",
    "early_stop = EarlyStopping(\n",
    "    monitor=stop_monitor,\n",
    "    min_delta=stop_delta,\n",
    "    patience=stop_epochs,\n",
    "    verbose=1,\n",
    "    mode='auto',\n",
    ")\n",
    "\n",
    "# callbacks_list = [early_stop]\n",
    "callbacks_list = []"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Train model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "## Create model object\n",
    "model = build_model(summary_type=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "hist = model.fit(\n",
    "    X_train, y_train,\n",
    "    validation_data=(X_val, y_val),\n",
    "    batch_size=BATCH_SIZE,\n",
    "    epochs=MAX_EPOCHS,\n",
    "    callbacks=callbacks_list,\n",
    "    verbose=1\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Visualize history\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "fig, loss_ax = plt.subplots()\n",
    "\n",
    "acc_ax = loss_ax.twinx()\n",
    "\n",
    "loss_ax.plot(hist.history['loss'], 'y', label='train loss')\n",
    "loss_ax.plot(hist.history['val_loss'], 'g', label='val loss')\n",
    "acc_ax.plot(hist.history['acc'], 'b', label='train acc')\n",
    "acc_ax.plot(hist.history['val_acc'], 'r', label='val acc')\n",
    "\n",
    "loss_ax.set_xlabel('epoch')\n",
    "loss_ax.set_ylabel('loss')\n",
    "acc_ax.set_ylabel('accuracy')\n",
    "\n",
    "loss_ax.legend(loc='upper left')\n",
    "acc_ax.legend(loc='lower left')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.evaluate(X_test, y_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "OVERFITTING...."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Let's insert Dropout "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from IPython.display import SVG\n",
    "from keras.utils.vis_utils import model_to_dot\n",
    "\n",
    "SVG(model_to_dot(model, show_shapes=True).create(prog='dot', format='svg'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "hist = model.fit(\n",
    "    X_train, y_train,\n",
    "    validation_data=(X_val, y_val),\n",
    "    batch_size=BATCH_SIZE,\n",
    "    epochs=MAX_EPOCHS,\n",
    "    callbacks=callbacks_list,\n",
    "    verbose=1\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# Visualize history\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "fig, loss_ax = plt.subplots()\n",
    "\n",
    "acc_ax = loss_ax.twinx()\n",
    "\n",
    "loss_ax.plot(hist.history['loss'], 'y', label='train loss')\n",
    "loss_ax.plot(hist.history['val_loss'], 'g', label='val loss')\n",
    "acc_ax.plot(hist.history['acc'], 'b', label='train acc')\n",
    "acc_ax.plot(hist.history['val_acc'], 'r', label='val acc')\n",
    "\n",
    "loss_ax.set_xlabel('epoch')\n",
    "loss_ax.set_ylabel('loss')\n",
    "acc_ax.set_ylabel('accuracy')\n",
    "\n",
    "loss_ax.legend(loc='upper left')\n",
    "acc_ax.legend(loc='lower left')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.evaluate(X_test, y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def number_to_english(data):\n",
    "    tmp_list = []\n",
    "    for letter in data:\n",
    "        try:\n",
    "            tmp_list.append(idx2english[int(letter)])\n",
    "        except KeyError:\n",
    "            break\n",
    "    return \"\".join(tmp_list)\n",
    "\n",
    "def english_to_number(data):\n",
    "    tmp_list = []\n",
    "    for letter in data:\n",
    "        try:\n",
    "            tmp_list.append(english2idx[letter])\n",
    "        except KeyError:\n",
    "            print(\"I didn't learn %s\" % letter)\n",
    "\n",
    "    return tmp_list\n",
    "            \n",
    "def number_to_korean(data):\n",
    "    tmp_list = []\n",
    "    for letter in data:\n",
    "        try:\n",
    "            tmp_list.append(idx2korean[int(letter)])\n",
    "        except KeyError:\n",
    "            break\n",
    "    return \"\".join(tmp_list)\n",
    "\n",
    "def argmax_data(data):\n",
    "    return [np.argmax(letter) for letter in data]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Check the result !"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "check_length = 10\n",
    "\n",
    "input_x_values = X_train[[range(check_length)]]\n",
    "result = model.predict(input_x_values)\n",
    "\n",
    "for input_x, result_y in zip(input_x_values, result):\n",
    "    print(number_to_english(input_x))\n",
    "    print(number_to_korean(argmax_data(result_y)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "query = \"air conditioner\"\n",
    "\n",
    "# convert to index\n",
    "idx_list = np.array(english_to_number(query), dtype='float')\n",
    "\n",
    "# Padding trick\n",
    "idx_list.resize(1, 24)\n",
    "\n",
    "# Get result\n",
    "result = model.predict(idx_list, verbose=1)\n",
    "\n",
    "# Print the query\n",
    "print(\"Query: \",query)\n",
    "\n",
    "# Print the result as Readable for person\n",
    "print(\"Answer: \", number_to_korean(argmax_data(result[0])))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "query = \"air\"\n",
    "\n",
    "# convert to index\n",
    "idx_list = np.array(english_to_number(query), dtype='float')\n",
    "\n",
    "# Padding trick\n",
    "idx_list.resize(1, 24)\n",
    "\n",
    "# Get result\n",
    "result = model.predict(idx_list, verbose=1)\n",
    "\n",
    "# Print the query\n",
    "print(\"Query: \",query)\n",
    "\n",
    "# Print the result as Readable for person\n",
    "print(\"Answer: \", number_to_korean(argmax_data(result[0])))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Save the Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.save('english2korean.h5')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load the Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "number_to_english(X_train[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.utils import plot_model\n",
    "from keras.models import load_model\n",
    "\n",
    "loaded_model = load_model('english2korean.h5')\n",
    "\n",
    "# plot_model(model, to_file='model.png')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "query = \"Abel\"\n",
    "\n",
    "# convert to index\n",
    "idx_list = np.array(english_to_number(query), dtype='float')\n",
    "\n",
    "# Padding trick\n",
    "idx_list.resize(1, 24)\n",
    "\n",
    "# Get result\n",
    "result = loaded_model.predict(idx_list, verbose=1)\n",
    "\n",
    "# Print the query\n",
    "print(\"Query: \",query)\n",
    "\n",
    "# Print the result as Readable for person\n",
    "print(\"Answer: \", number_to_korean(argmax_data(result[0])))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Character-level Embedding\n",
    "\n",
    "글자 `가-힣` 단계에서 output을 뽑는 위 모델과는 다르게, 자음/모음을 출력값으로 가진 뒤에 조합하는 모델"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "space = [' ']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Prepare English data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Prepare English letters\n",
    "first_big_id = ord('A')\n",
    "last_big_id = ord('Z')\n",
    "first_small_id = ord('a')\n",
    "last_small_id = ord('z')\n",
    "\n",
    "big = list(range(first_big_id, last_big_id+1))\n",
    "small = list(range(first_small_id, last_small_id+1))\n",
    "\n",
    "english_letter_list = [chr(num) for num in big+small] + space\n",
    "    \n",
    "# Make a mapping with index\n",
    "idx2english = dict(enumerate(english_letter_list, 1))\n",
    "english2idx = {v: k for k, v in idx2english.items()}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Prepare Korean data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "consonants = [chr(letter) for letter in range(ord('ㄱ'), ord('ㅎ')+1)]\n",
    "vowel = [chr(letter) for letter in range(ord('ㅏ'), ord('ㅣ')+1)]\n",
    "\n",
    "korean_letter_list = consonants + vowel + space\n",
    "\n",
    "# Make a mapping with index\n",
    "idx2korean = dict(enumerate(korean_letter_list, 1))\n",
    "korean2idx = {v: k for k, v in idx2korean.items()}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>roman</th>\n",
       "      <th>hangul</th>\n",
       "      <th>roma</th>\n",
       "      <th>hang</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Fiery Cross Reef</td>\n",
       "      <td>파이어리 크로스 리프</td>\n",
       "      <td>Fiery Cross Reef</td>\n",
       "      <td>파이어리 크로스 리프</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Rosetta</td>\n",
       "      <td>로제타</td>\n",
       "      <td>Rosetta</td>\n",
       "      <td>로제타</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Lynch Loretta Elizabeth</td>\n",
       "      <td>린치 로레타 엘리자베스</td>\n",
       "      <td>Lynch Loretta Elizabeth</td>\n",
       "      <td>린치 로레타 엘리자베스</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Crabbe Tim</td>\n",
       "      <td>크래브 팀</td>\n",
       "      <td>Crabbe Tim</td>\n",
       "      <td>크래브 팀</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Schumacher Nick</td>\n",
       "      <td>슈매커 닉</td>\n",
       "      <td>Schumacher Nick</td>\n",
       "      <td>슈매커 닉</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                     roman        hangul                     roma  \\\n",
       "0         Fiery Cross Reef   파이어리 크로스 리프         Fiery Cross Reef   \n",
       "1                  Rosetta           로제타                  Rosetta   \n",
       "2  Lynch Loretta Elizabeth  린치 로레타 엘리자베스  Lynch Loretta Elizabeth   \n",
       "3               Crabbe Tim         크래브 팀               Crabbe Tim   \n",
       "4          Schumacher Nick         슈매커 닉          Schumacher Nick   \n",
       "\n",
       "           hang  \n",
       "0   파이어리 크로스 리프  \n",
       "1           로제타  \n",
       "2  린치 로레타 엘리자베스  \n",
       "3         크래브 팀  \n",
       "4         슈매커 닉  "
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "\n",
    "df = pd.read_csv('full_data.csv', encoding='utf8')\n",
    "df['roma'] = df['roman']\n",
    "df['hang'] = df['hangul']\n",
    "df.head()\n",
    "# del df['Unnamed: 0']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Load X"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[  6.  35.  31. ...,   0.   0.   0.]\n",
      " [ 18.  41.  45. ...,   0.   0.   0.]\n",
      " [ 12.  51.  40. ...,   0.   0.   0.]\n",
      " ..., \n",
      " [ 33.  27.  30. ...,   0.   0.   0.]\n",
      " [ 33.  27.  44. ...,   0.   0.   0.]\n",
      " [ 33.  27.  44. ...,   0.   0.   0.]]\n",
      "(11189, 28)\n"
     ]
    }
   ],
   "source": [
    "roma = df['roma'].apply(lambda x: np.array([english2idx[letter] for letter in x]))\n",
    "roma_df = pd.DataFrame(roma)\n",
    "X = pd.DataFrame(roma_df['roma'].tolist()).values\n",
    "\n",
    "# Replace nan to 0\n",
    "# Get index of nan, and make them 0\n",
    "X[np.isnan(X)] = 0\n",
    "print(X)\n",
    "print(X.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Load y"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# from https://github.com/Kcrong/separate-korean/blob/master/separater.py\n",
    "class Separater:\n",
    "    # BASE_CODE, CHOSUNG, JUNGSUNG = 44032, 588, 28\n",
    "    _BASE_CODE = 44032\n",
    "    _CHOSUNG = 588\n",
    "    _JUNGSUNG = 28\n",
    "\n",
    "    _CHOSUNG_LIST = [u'ㄱ', u'ㄲ', u'ㄴ', u'ㄷ', u'ㄸ', u'ㄹ', u'ㅁ', u'ㅂ', u'ㅃ', u'ㅅ', u'ㅆ', u'ㅇ', u'ㅈ', u'ㅉ', u'ㅊ', u'ㅋ',\n",
    "                     u'ㅌ', u'ㅍ', u'ㅎ']\n",
    "    _JUNGSUNG_LIST = [u'ㅏ', u'ㅐ', u'ㅑ', u'ㅒ', u'ㅓ', u'ㅔ', u'ㅕ', u'ㅖ', u'ㅗ', u'ㅘ', u'ㅙ', u'ㅚ', u'ㅛ', u'ㅜ', u'ㅝ', u'ㅞ',\n",
    "                      u'ㅟ', u'ㅠ', u'ㅡ', u'ㅢ', u'ㅣ']\n",
    "    _JONGSUNG_LIST = [u' ', u'ㄱ', u'ㄲ', u'ㄳ', u'ㄴ', u'ㄵ', u'ㄶ', u'ㄷ', u'ㄹ', u'ㄺ', u'ㄻ', u'ㄼ', u'ㄽ', u'ㄾ', u'ㄿ', u'ㅀ',\n",
    "                      u'ㅁ', u'ㅂ', u'ㅄ', u'ㅅ', u'ㅆ', u'ㅇ', u'ㅈ', u'ㅊ', u'ㅋ', u'ㅌ', u'ㅍ', u'ㅎ']\n",
    "\n",
    "    ch = _BASE_CODE + (0 * _CHOSUNG + 2 * _JUNGSUNG)\n",
    "\n",
    "    def __init__(self, word):\n",
    "        self.word = word\n",
    "\n",
    "    @property\n",
    "    def sep_all(self):\n",
    "        try:\n",
    "            _all_ = list()\n",
    "\n",
    "            for ch in self.word:\n",
    "                for tmp in ch:\n",
    "                    chbase = ord(tmp) - Separater._BASE_CODE\n",
    "\n",
    "                    ch_1 = chbase // Separater._CHOSUNG\n",
    "                    _all_.append(Separater._CHOSUNG_LIST[ch_1])\n",
    "\n",
    "                    ch_2 = (chbase - (Separater._CHOSUNG * ch_1)) // Separater._JUNGSUNG\n",
    "                    _all_.append(Separater._JUNGSUNG_LIST[ch_2])\n",
    "\n",
    "                    ch_3 = (chbase - (Separater._CHOSUNG * ch_1) - (Separater._JUNGSUNG * ch_2))\n",
    "                    _all_.append(Separater._JONGSUNG_LIST[ch_3])\n",
    "\n",
    "            while True:\n",
    "                try:\n",
    "                    _all_.remove(' ')\n",
    "                except ValueError:\n",
    "                    break\n",
    "        except:\n",
    "            print(self.word)\n",
    "            raise\n",
    "                    \n",
    "        return _all_\n",
    "\n",
    "    @property\n",
    "    def chosung(self):\n",
    "        all_chosung = list()\n",
    "        for ch in self.word:\n",
    "            for tmp in ch:\n",
    "                all_chosung.append(Separater._CHOSUNG_LIST[((ord(tmp) - Separater._BASE_CODE) // Separater._CHOSUNG)])\n",
    "        return all_chosung\n",
    "\n",
    "    @property\n",
    "    def jungsung(self):\n",
    "        all_jungsung = list()\n",
    "\n",
    "        for ch in self.word:\n",
    "            for tmp in ch:\n",
    "                chbase = ord(tmp) - Separater._BASE_CODE\n",
    "                jungsung_idx = (chbase - (Separater._CHOSUNG * (chbase // Separater._CHOSUNG))) // Separater._JUNGSUNG\n",
    "                all_jungsung.append(Separater._JUNGSUNG_LIST[jungsung_idx])\n",
    "        return all_jungsung\n",
    "\n",
    "    @property\n",
    "    def jongsung(self):\n",
    "        all_jongsung = list()\n",
    "\n",
    "        for ch in self.word:\n",
    "            for tmp in ch:\n",
    "                chbase = ord(tmp) - Separater._BASE_CODE\n",
    "                c_idx = chbase // Separater._CHOSUNG\n",
    "\n",
    "                jongsung_idx = (chbase - (Separater._CHOSUNG * c_idx) - (\n",
    "                    Separater._JUNGSUNG * ((chbase - (Separater._CHOSUNG * c_idx)) // Separater._JUNGSUNG)))\n",
    "                all_jongsung.append(Separater._JONGSUNG_LIST[jongsung_idx])\n",
    "        return all_jongsung"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "def sep_encode_all(sent):\n",
    "    return [korean2idx[let] for letter in sent if letter != ' ' for let in Separater(letter).sep_all]\n",
    "\n",
    "def sep_all(sent):\n",
    "    return [let for letter in sent if letter != ' ' for let in Separater(letter).sep_all]\n",
    "\n",
    "def encode_all(sep_sent):\n",
    "    return [korean2idx[let] for let in sep_sent]\n",
    "\n",
    "assert encode_all(sep_all('안녕 하세요')) == sep_encode_all('안녕 하세요')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[ 29.  31.  23. ...,   0.   0.   0.]\n",
      " [  9.  39.  24. ...,   0.   0.   0.]\n",
      " [  9.  51.   4. ...,   0.   0.   0.]\n",
      " ..., \n",
      " [  1.  31.   7. ...,   0.   0.   0.]\n",
      " [  1.  31.   4. ...,   0.   0.   0.]\n",
      " [  1.  31.   4. ...,   0.   0.   0.]]\n",
      "(11189, 30)\n"
     ]
    }
   ],
   "source": [
    "# Load y\n",
    "hang = df['hang'].apply(lambda x: sep_encode_all(x))\n",
    "\n",
    "# Padding Trick. Add nan automatically\n",
    "hang_df = pd.DataFrame(hang)\n",
    "y = pd.DataFrame(hang_df['hang'].tolist()).values\n",
    "\n",
    "# nan to 0\n",
    "y[np.isnan(y)] = 0\n",
    "\n",
    "print(y)\n",
    "print(y.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "# one-hot encoding\n",
    "from keras.utils import np_utils\n",
    "\n",
    "y = np_utils.to_categorical(y, num_classes=len(korean_letter_list)+1).reshape(y.shape[0], y.shape[1], len(korean_letter_list)+1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(11189, 30, 53)"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Split train & test & validation set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3)\n",
    "X_train, X_val, y_train, y_val = train_test_split(X_train, y_train, test_size=0.1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Build & Train model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Done\n"
     ]
    }
   ],
   "source": [
    "from IPython.display import SVG, display\n",
    "from keras.utils.vis_utils import model_to_dot\n",
    "\n",
    "def viz_model(model):\n",
    "    display(SVG(model_to_dot(model, show_shapes=True).create(prog='dot', format='svg')))\n",
    "\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, TimeDistributed, Activation, RepeatVector, Embedding, Dropout\n",
    "\n",
    "def build_model(embedding_input_dim,\n",
    "                embedding_output_dim,\n",
    "                embedding_input_length,\n",
    "                maximum_input_length,\n",
    "                hidden_size,\n",
    "                maximum_output_length,\n",
    "                encoder_layer_count,\n",
    "                dropout_rate,\n",
    "                output_classes_count,\n",
    "                rnn_layer,\n",
    "                summary_type=False):\n",
    "    \n",
    "    print('Build model...')\n",
    "    RNN = rnn_layer\n",
    "\n",
    "    model = Sequential()\n",
    "\n",
    "    model.add(Embedding(\n",
    "        input_dim=embedding_input_dim,\n",
    "        output_dim=embedding_output_dim,\n",
    "        input_length=embedding_input_length\n",
    "    ))\n",
    "    \n",
    "    model.add(RNN(hidden_size, return_sequences=True, input_shape=(maximum_input_length, 123)))\n",
    "    model.add(RNN(hidden_size))\n",
    "\n",
    "    model.add(RepeatVector(maximum_output_length)) # 글자수 맥시멈\n",
    "\n",
    "    for _ in range(encoder_layer_count):\n",
    "        model.add(RNN(hidden_size, return_sequences=True))\n",
    "        model.add(Dropout(dropout_rate))\n",
    "\n",
    "    # For each of step of the output sequence, decide which character should be chosen\n",
    "    model.add(TimeDistributed(Dense(output_classes_count)))\n",
    "    model.add(Activation('softmax'))\n",
    "\n",
    "    model.compile(loss='categorical_crossentropy', # default: categorical_crossentropy\n",
    "                  optimizer='adam', # default: adam, can try SGD, RMSprop\n",
    "                  metrics=['accuracy'])\n",
    "    \n",
    "    \n",
    "    \n",
    "    if summary_type is False:\n",
    "        pass\n",
    "    elif str(summary_type).lower() == 'svg':\n",
    "        viz_model(model)\n",
    "    else:\n",
    "        model.summary()\n",
    "        \n",
    "    return model\n",
    "\n",
    "print('Done')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Build model...\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding_4 (Embedding)      (None, 28, 200)           10800     \n",
      "_________________________________________________________________\n",
      "gru_13 (GRU)                 (None, 28, 200)           240600    \n",
      "_________________________________________________________________\n",
      "gru_14 (GRU)                 (None, 200)               240600    \n",
      "_________________________________________________________________\n",
      "repeat_vector_4 (RepeatVecto (None, 30, 200)           0         \n",
      "_________________________________________________________________\n",
      "gru_15 (GRU)                 (None, 30, 200)           240600    \n",
      "_________________________________________________________________\n",
      "dropout_7 (Dropout)          (None, 30, 200)           0         \n",
      "_________________________________________________________________\n",
      "gru_16 (GRU)                 (None, 30, 200)           240600    \n",
      "_________________________________________________________________\n",
      "dropout_8 (Dropout)          (None, 30, 200)           0         \n",
      "_________________________________________________________________\n",
      "time_distributed_4 (TimeDist (None, 30, 53)            10653     \n",
      "_________________________________________________________________\n",
      "activation_4 (Activation)    (None, 30, 53)            0         \n",
      "=================================================================\n",
      "Total params: 983,853\n",
      "Trainable params: 983,853\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "from keras.layers import recurrent\n",
    "\n",
    "model = build_model(\n",
    "    embedding_input_dim=len(english_letter_list)+1,  # Add 1 because of padding\n",
    "    embedding_output_dim=200,\n",
    "    embedding_input_length=X.shape[1],\n",
    "    maximum_input_length=X.shape[1],\n",
    "    hidden_size = 200,\n",
    "    maximum_output_length=y.shape[1],  # Add 1 because of last chr. 마지막 패딩 출력을 위해,\n",
    "    encoder_layer_count=2,  # Make 2 layers for encoding layer\n",
    "    dropout_rate=0.2, # Make no-dropout\n",
    "    output_classes_count=len(korean_letter_list)+1,  # Add 1 because of padding\n",
    "    rnn_layer=recurrent.GRU,\n",
    "    summary_type=True\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 7048 samples, validate on 784 samples\n",
      "Epoch 1/100\n",
      "7048/7048 [==============================] - 8s - loss: 1.3135 - acc: 0.7031 - val_loss: 1.0997 - val_acc: 0.7193\n",
      "Epoch 2/100\n",
      "7048/7048 [==============================] - 7s - loss: 1.0515 - acc: 0.7196 - val_loss: 0.9905 - val_acc: 0.7239\n",
      "Epoch 3/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.9907 - acc: 0.7260 - val_loss: 0.9744 - val_acc: 0.7291\n",
      "Epoch 4/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.9309 - acc: 0.7352 - val_loss: 0.9301 - val_acc: 0.7321\n",
      "Epoch 5/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.9099 - acc: 0.7360 - val_loss: 0.9001 - val_acc: 0.7363\n",
      "Epoch 6/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.8994 - acc: 0.7380 - val_loss: 0.9040 - val_acc: 0.7369\n",
      "Epoch 7/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.8948 - acc: 0.7384 - val_loss: 0.8963 - val_acc: 0.7372\n",
      "Epoch 8/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.8856 - acc: 0.7397 - val_loss: 0.8816 - val_acc: 0.7389\n",
      "Epoch 9/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.8814 - acc: 0.7404 - val_loss: 0.8859 - val_acc: 0.7397\n",
      "Epoch 10/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.8784 - acc: 0.7411 - val_loss: 0.8715 - val_acc: 0.7414\n",
      "Epoch 11/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.8690 - acc: 0.7425 - val_loss: 0.8685 - val_acc: 0.7407\n",
      "Epoch 12/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.8655 - acc: 0.7426 - val_loss: 0.8924 - val_acc: 0.7406\n",
      "Epoch 13/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.8556 - acc: 0.7440 - val_loss: 0.8641 - val_acc: 0.7432\n",
      "Epoch 14/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.8449 - acc: 0.7458 - val_loss: 0.8458 - val_acc: 0.7460\n",
      "Epoch 15/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.8333 - acc: 0.7477 - val_loss: 0.8263 - val_acc: 0.7496\n",
      "Epoch 16/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.8086 - acc: 0.7518 - val_loss: 0.7932 - val_acc: 0.7558\n",
      "Epoch 17/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.7808 - acc: 0.7567 - val_loss: 0.8149 - val_acc: 0.7457\n",
      "Epoch 18/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.7666 - acc: 0.7593 - val_loss: 0.7673 - val_acc: 0.7601\n",
      "Epoch 19/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.7552 - acc: 0.7617 - val_loss: 0.7626 - val_acc: 0.7606\n",
      "Epoch 20/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.7277 - acc: 0.7680 - val_loss: 0.7417 - val_acc: 0.7694\n",
      "Epoch 21/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.7140 - acc: 0.7729 - val_loss: 0.7343 - val_acc: 0.7689\n",
      "Epoch 22/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.6930 - acc: 0.7797 - val_loss: 0.6915 - val_acc: 0.7839\n",
      "Epoch 23/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.6740 - acc: 0.7860 - val_loss: 0.6848 - val_acc: 0.7863\n",
      "Epoch 24/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.6538 - acc: 0.7910 - val_loss: 0.6818 - val_acc: 0.7869\n",
      "Epoch 25/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.6417 - acc: 0.7955 - val_loss: 0.7144 - val_acc: 0.7825\n",
      "Epoch 26/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.6237 - acc: 0.8013 - val_loss: 0.6559 - val_acc: 0.7977\n",
      "Epoch 27/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.5987 - acc: 0.8084 - val_loss: 0.6343 - val_acc: 0.8014\n",
      "Epoch 28/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.5778 - acc: 0.8134 - val_loss: 0.6295 - val_acc: 0.8026\n",
      "Epoch 29/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.5590 - acc: 0.8192 - val_loss: 0.6153 - val_acc: 0.8120\n",
      "Epoch 30/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.5350 - acc: 0.8267 - val_loss: 0.5717 - val_acc: 0.8214\n",
      "Epoch 31/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.5138 - acc: 0.8334 - val_loss: 0.5589 - val_acc: 0.8278\n",
      "Epoch 32/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.4951 - acc: 0.8392 - val_loss: 0.5488 - val_acc: 0.8283\n",
      "Epoch 33/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.4765 - acc: 0.8443 - val_loss: 0.5425 - val_acc: 0.8358\n",
      "Epoch 34/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.4599 - acc: 0.8497 - val_loss: 0.5286 - val_acc: 0.8383\n",
      "Epoch 35/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.4471 - acc: 0.8540 - val_loss: 0.5232 - val_acc: 0.8373\n",
      "Epoch 36/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.4293 - acc: 0.8591 - val_loss: 0.5072 - val_acc: 0.8484\n",
      "Epoch 37/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.4117 - acc: 0.8645 - val_loss: 0.4890 - val_acc: 0.8506\n",
      "Epoch 38/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.3922 - acc: 0.8711 - val_loss: 0.4760 - val_acc: 0.8554\n",
      "Epoch 39/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.3684 - acc: 0.8789 - val_loss: 0.4724 - val_acc: 0.8599\n",
      "Epoch 40/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.3571 - acc: 0.8820 - val_loss: 0.4785 - val_acc: 0.8540\n",
      "Epoch 41/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.3468 - acc: 0.8865 - val_loss: 0.5138 - val_acc: 0.8450\n",
      "Epoch 42/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.3665 - acc: 0.8808 - val_loss: 0.4515 - val_acc: 0.8677\n",
      "Epoch 43/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.3234 - acc: 0.8939 - val_loss: 0.4452 - val_acc: 0.8742\n",
      "Epoch 44/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.3080 - acc: 0.8987 - val_loss: 0.4348 - val_acc: 0.8763\n",
      "Epoch 45/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.3034 - acc: 0.9013 - val_loss: 0.4502 - val_acc: 0.8744\n",
      "Epoch 46/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.3155 - acc: 0.8978 - val_loss: 0.4370 - val_acc: 0.8779\n",
      "Epoch 47/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.2802 - acc: 0.9083 - val_loss: 0.4562 - val_acc: 0.8804\n",
      "Epoch 48/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.2652 - acc: 0.9134 - val_loss: 0.4411 - val_acc: 0.8808\n",
      "Epoch 49/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.2630 - acc: 0.9141 - val_loss: 0.4323 - val_acc: 0.8853\n",
      "Epoch 50/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.2510 - acc: 0.9181 - val_loss: 0.4213 - val_acc: 0.8862\n",
      "Epoch 51/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.2435 - acc: 0.9203 - val_loss: 0.4144 - val_acc: 0.8902\n",
      "Epoch 52/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.2446 - acc: 0.9207 - val_loss: 0.4099 - val_acc: 0.8878\n",
      "Epoch 53/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.2252 - acc: 0.9264 - val_loss: 0.4190 - val_acc: 0.8933\n",
      "Epoch 54/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.2125 - acc: 0.9309 - val_loss: 0.4395 - val_acc: 0.8877\n",
      "Epoch 55/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.2254 - acc: 0.9263 - val_loss: 0.4486 - val_acc: 0.8902\n",
      "Epoch 56/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.2093 - acc: 0.9314 - val_loss: 0.4142 - val_acc: 0.8976\n",
      "Epoch 57/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.2011 - acc: 0.9339 - val_loss: 0.4190 - val_acc: 0.8966\n",
      "Epoch 58/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.1919 - acc: 0.9374 - val_loss: 0.4181 - val_acc: 0.8951\n",
      "Epoch 59/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.1920 - acc: 0.9376 - val_loss: 0.4355 - val_acc: 0.8960\n",
      "Epoch 60/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.1828 - acc: 0.9402 - val_loss: 0.4337 - val_acc: 0.8979\n",
      "Epoch 61/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.1733 - acc: 0.9432 - val_loss: 0.4386 - val_acc: 0.8980\n",
      "Epoch 62/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.1879 - acc: 0.9389 - val_loss: 0.4356 - val_acc: 0.8986\n",
      "Epoch 63/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.1786 - acc: 0.9421 - val_loss: 0.4368 - val_acc: 0.8954\n",
      "Epoch 64/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.1676 - acc: 0.9460 - val_loss: 0.4359 - val_acc: 0.8961\n",
      "Epoch 65/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.1767 - acc: 0.9429 - val_loss: 0.4176 - val_acc: 0.9024\n",
      "Epoch 66/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.1558 - acc: 0.9495 - val_loss: 0.4365 - val_acc: 0.8979\n",
      "Epoch 67/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.1624 - acc: 0.9475 - val_loss: 0.4367 - val_acc: 0.8981\n",
      "Epoch 68/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.1565 - acc: 0.9493 - val_loss: 0.4102 - val_acc: 0.9045\n",
      "Epoch 69/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.1359 - acc: 0.9554 - val_loss: 0.4230 - val_acc: 0.9054\n",
      "Epoch 70/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.1288 - acc: 0.9577 - val_loss: 0.4617 - val_acc: 0.9013\n",
      "Epoch 71/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.1364 - acc: 0.9555 - val_loss: 0.4549 - val_acc: 0.8975\n",
      "Epoch 72/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.1471 - acc: 0.9527 - val_loss: 0.4191 - val_acc: 0.9032\n",
      "Epoch 73/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.1404 - acc: 0.9545 - val_loss: 0.4387 - val_acc: 0.9011\n",
      "Epoch 74/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.1355 - acc: 0.9558 - val_loss: 0.4179 - val_acc: 0.9082\n",
      "Epoch 75/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.1239 - acc: 0.9598 - val_loss: 0.4403 - val_acc: 0.9020\n",
      "Epoch 76/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.1141 - acc: 0.9629 - val_loss: 0.4498 - val_acc: 0.9057\n",
      "Epoch 77/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.1034 - acc: 0.9663 - val_loss: 0.4470 - val_acc: 0.9051\n",
      "Epoch 78/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.0950 - acc: 0.9692 - val_loss: 0.4661 - val_acc: 0.9079\n",
      "Epoch 79/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.0911 - acc: 0.9700 - val_loss: 0.4636 - val_acc: 0.9072\n",
      "Epoch 80/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.1156 - acc: 0.9625 - val_loss: 0.4743 - val_acc: 0.9017\n",
      "Epoch 81/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.1236 - acc: 0.9600 - val_loss: 0.4458 - val_acc: 0.9053\n",
      "Epoch 82/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.1178 - acc: 0.9616 - val_loss: 0.4694 - val_acc: 0.9019\n",
      "Epoch 83/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.1050 - acc: 0.9657 - val_loss: 0.4440 - val_acc: 0.9101\n",
      "Epoch 84/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.1093 - acc: 0.9646 - val_loss: 0.4773 - val_acc: 0.9071\n",
      "Epoch 85/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.1070 - acc: 0.9655 - val_loss: 0.4616 - val_acc: 0.9035\n",
      "Epoch 86/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.1051 - acc: 0.9655 - val_loss: 0.4724 - val_acc: 0.9069\n",
      "Epoch 87/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.0956 - acc: 0.9691 - val_loss: 0.4920 - val_acc: 0.9051\n",
      "Epoch 88/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.1122 - acc: 0.9645 - val_loss: 0.4615 - val_acc: 0.9090\n",
      "Epoch 89/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.0868 - acc: 0.9721 - val_loss: 0.4536 - val_acc: 0.9095\n",
      "Epoch 90/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.0835 - acc: 0.9729 - val_loss: 0.4559 - val_acc: 0.9102\n",
      "Epoch 91/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.0910 - acc: 0.9709 - val_loss: 0.4714 - val_acc: 0.9088\n",
      "Epoch 92/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.1148 - acc: 0.9637 - val_loss: 0.5031 - val_acc: 0.8997\n",
      "Epoch 93/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.0966 - acc: 0.9691 - val_loss: 0.4721 - val_acc: 0.9057\n",
      "Epoch 94/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.0898 - acc: 0.9710 - val_loss: 0.4558 - val_acc: 0.9085\n",
      "Epoch 95/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.0732 - acc: 0.9764 - val_loss: 0.4704 - val_acc: 0.9105\n",
      "Epoch 96/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.0682 - acc: 0.9781 - val_loss: 0.4851 - val_acc: 0.9111\n",
      "Epoch 97/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.0665 - acc: 0.9786 - val_loss: 0.4708 - val_acc: 0.9098\n",
      "Epoch 98/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.0782 - acc: 0.9751 - val_loss: 0.4789 - val_acc: 0.9115\n",
      "Epoch 99/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.0721 - acc: 0.9770 - val_loss: 0.4962 - val_acc: 0.9073\n",
      "Epoch 100/100\n",
      "7048/7048 [==============================] - 7s - loss: 0.0877 - acc: 0.9724 - val_loss: 0.5604 - val_acc: 0.8997\n"
     ]
    }
   ],
   "source": [
    "## Params\n",
    "\n",
    "BATCH_SIZE=64\n",
    "MAX_EPOCHS=100\n",
    "\n",
    "hist = model.fit(\n",
    "    X_train, y_train,\n",
    "    validation_data=(X_val, y_val),\n",
    "    batch_size=BATCH_SIZE,\n",
    "    epochs=MAX_EPOCHS,\n",
    "    verbose=1\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[0.092786252633938054,\n",
       " 0.084843806592614368,\n",
       " 0.10074032412653748,\n",
       " 0.096561266618879107,\n",
       " 0.072834398291971233,\n",
       " 0.072016498135248463,\n",
       " 0.05865065379662357,\n",
       " 0.062743712846659097,\n",
       " 0.058412685411031888,\n",
       " 0.061961973126170043,\n",
       " 0.085761299020654089,\n",
       " 0.066430493952467543,\n",
       " 0.07656145227330903,\n",
       " 0.078425144043060765,\n",
       " 0.082060973237138596,\n",
       " 0.062968125168305994,\n",
       " 0.048720408121079514,\n",
       " 0.041397358321507594,\n",
       " 0.05001379251818381,\n",
       " 0.073632024858616266,\n",
       " 0.14807636073444813,\n",
       " 0.11778501607266241,\n",
       " 0.07641093127607082,\n",
       " 0.062228863739602007,\n",
       " 0.055939549282108612,\n",
       " 0.041551248246019495,\n",
       " 0.044343871701927377,\n",
       " 0.070916062056020357,\n",
       " 0.056179544678069945,\n",
       " 0.049725167127696125]"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "hist2.history['loss']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3357/3357 [==============================] - 1s     \n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[0.54901867985973973, 0.91450703211543738]"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.evaluate(X_test, y_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Viz history (Within Dropout)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAa0AAAEKCAYAAAChTwphAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fdf7be68470>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Visualize history\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "fig, loss_ax = plt.subplots()\n",
    "\n",
    "acc_ax = loss_ax.twinx()\n",
    "\n",
    "loss_ax.plot(hist.history['loss'], 'y', label='train loss')\n",
    "loss_ax.plot(hist.history['val_loss'], 'g', label='val loss')\n",
    "acc_ax.plot(hist.history['acc'], 'b', label='train acc')\n",
    "acc_ax.plot(hist.history['val_acc'], 'r', label='val acc')\n",
    "\n",
    "loss_ax.set_xlabel('epoch')\n",
    "loss_ax.set_ylabel('loss')\n",
    "acc_ax.set_ylabel('accuracy')\n",
    "\n",
    "loss_ax.legend(loc='upper left')\n",
    "acc_ax.legend(loc='lower left')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Viz history (Without Dropout)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 128,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAacAAAEKCAYAAAC2bZqoAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nUkqpqKPJSSmlVNTR5KSUUirq/Cey6g/81hSArgAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7ff0833a3940>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Visualize history\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "fig, loss_ax = plt.subplots()\n",
    "\n",
    "acc_ax = loss_ax.twinx()\n",
    "\n",
    "loss_ax.plot(hist.history['loss'], 'y', label='train loss')\n",
    "loss_ax.plot(hist.history['val_loss'], 'g', label='val loss')\n",
    "acc_ax.plot(hist.history['acc'], 'b', label='train acc')\n",
    "acc_ax.plot(hist.history['val_acc'], 'r', label='val acc')\n",
    "\n",
    "loss_ax.set_xlabel('epoch')\n",
    "loss_ax.set_ylabel('loss')\n",
    "acc_ax.set_ylabel('accuracy')\n",
    "\n",
    "loss_ax.legend(loc='upper left')\n",
    "acc_ax.legend(loc='lower left')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Viz history (Within Dropout, Large Dataset)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAa0AAAEKCAYAAAChTwphAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fdf7bdbe438>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Visualize history\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "fig, loss_ax = plt.subplots()\n",
    "\n",
    "acc_ax = loss_ax.twinx()\n",
    "\n",
    "loss_ax.plot(hist.history['loss'] + hist2.history['loss'], 'y', label='train loss')\n",
    "loss_ax.plot(hist.history['val_loss'] + hist2.history['val_loss'], 'g', label='val loss')\n",
    "acc_ax.plot(hist.history['acc'] + hist2.history['acc'], 'b', label='train acc')\n",
    "acc_ax.plot(hist.history['val_acc'] + hist2.history['val_acc'], 'r', label='val acc')\n",
    "\n",
    "loss_ax.set_xlabel('epoch')\n",
    "loss_ax.set_ylabel('loss')\n",
    "acc_ax.set_ylabel('accuracy')\n",
    "\n",
    "loss_ax.legend(loc='upper left')\n",
    "acc_ax.legend(loc='lower left')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Change model \n",
    "Tips from paper (https://arxiv.org/abs/1409.0473)\n",
    "\n",
    "1. non-bidirectional LSTM in encoder.\n",
    "2. reverse the input (np.flip)\n",
    "3. pad from the start. \n",
    "\n",
    "Before:  \n",
    "X --> y  \n",
    "[h, e, l, l, o, PAD, PAD] --> [안, 녕, 하, 세, 요]  \n",
    "\n",
    "After:  \n",
    "X --> y  \n",
    "[PAD, PAD, o, l, l, e, h] >>> [안, 녕, 하, 세, 요]  "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Ready dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([  6.,  35.,  31.,  44.,  51.,  53.,   3.,  44.,  41.,  45.,  45.,\n",
       "        53.,  18.,  31.,  31.,  32.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.])"
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Reverse Input\n",
    "new_X = np.flip(X, axis=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Build model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Done\n"
     ]
    }
   ],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, TimeDistributed, Activation, RepeatVector, Embedding, Dropout\n",
    "\n",
    "def build_model(embedding_input_dim,\n",
    "                embedding_output_dim,\n",
    "                embedding_input_length,\n",
    "                maximum_input_length,\n",
    "                hidden_size,\n",
    "                maximum_output_length,\n",
    "                encoder_layer_count,\n",
    "                dropout_rate,\n",
    "                output_classes_count,\n",
    "                input_rnn_layer,\n",
    "                output_rnn_layer,\n",
    "                summary_type=False):\n",
    "    \n",
    "    print('Build model...')\n",
    "    ENCODER_RNN = input_rnn_layer\n",
    "    DECODER_RNN = output_rnn_layer\n",
    "\n",
    "    model = Sequential()\n",
    "\n",
    "    model.add(Embedding(\n",
    "        input_dim=embedding_input_dim,\n",
    "        output_dim=embedding_output_dim,\n",
    "        input_length=embedding_input_length\n",
    "    ))\n",
    "    \n",
    "    model.add(ENCODER_RNN(hidden_size, return_sequences=True, input_shape=(maximum_input_length, 123)))\n",
    "    model.add(ENCODER_RNN(hidden_size))\n",
    "\n",
    "    model.add(RepeatVector(maximum_output_length)) # 글자수 맥시멈\n",
    "\n",
    "    for _ in range(encoder_layer_count):\n",
    "        model.add(DECODER_RNN(hidden_size, return_sequences=True))\n",
    "        model.add(Dropout(dropout_rate))\n",
    "\n",
    "    # For each of step of the output sequence, decide which character should be chosen\n",
    "    model.add(TimeDistributed(Dense(output_classes_count)))\n",
    "    model.add(Activation('softmax'))\n",
    "\n",
    "    model.compile(loss='categorical_crossentropy', # default: categorical_crossentropy\n",
    "                  optimizer='adam', # default: adam, can try SGD, RMSprop\n",
    "                  metrics=['accuracy'])\n",
    "    \n",
    "    \n",
    "    \n",
    "    if summary_type is False:\n",
    "        pass\n",
    "    elif str(summary_type).lower() == 'svg':\n",
    "        viz_model(model)\n",
    "    else:\n",
    "        model.summary()\n",
    "        \n",
    "    return model\n",
    "\n",
    "print('Done')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'build_model' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-2-44ef79394276>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      1\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0mkeras\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlayers\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mrecurrent\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      2\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 3\u001b[0;31m model = build_model(\n\u001b[0m\u001b[1;32m      4\u001b[0m     \u001b[0membedding_input_dim\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0menglish_letter_list\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m+\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m,\u001b[0m  \u001b[0;31m# Add 1 because of padding\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      5\u001b[0m     \u001b[0membedding_output_dim\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m200\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mNameError\u001b[0m: name 'build_model' is not defined"
     ]
    }
   ],
   "source": [
    "from keras.layers import recurrent\n",
    "\n",
    "model = build_model(\n",
    "    embedding_input_dim=len(english_letter_list)+1,  # Add 1 because of padding\n",
    "    embedding_output_dim=200,\n",
    "    embedding_input_length=X.shape[1],\n",
    "    maximum_input_length=X.shape[1],\n",
    "    hidden_size = 200,\n",
    "    maximum_output_length=y.shape[1],  # Add 1 because of last chr. 마지막 패딩 출력을 위해,\n",
    "    encoder_layer_count=2,  # Make 2 layers for encoding layer\n",
    "    dropout_rate=0.2, # Make no-dropout\n",
    "    output_classes_count=len(korean_letter_list)+1,  # Add 1 because of padding\n",
    "    input_rnn_layer=recurrent.LSTM,\n",
    "    output_rnn_layer=recurrent.GRU,\n",
    "    summary_type='svg'\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 7048 samples, validate on 784 samples\n",
      "Epoch 1/100\n",
      "7048/7048 [==============================] - 9s - loss: 1.2900 - acc: 0.7046 - val_loss: 1.2348 - val_acc: 0.7106\n",
      "Epoch 2/100\n",
      "7048/7048 [==============================] - 8s - loss: 1.0133 - acc: 0.7227 - val_loss: 1.0297 - val_acc: 0.7189\n",
      "Epoch 3/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.9858 - acc: 0.7260 - val_loss: 0.9723 - val_acc: 0.7259\n",
      "Epoch 4/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.9382 - acc: 0.7341 - val_loss: 0.9080 - val_acc: 0.7350\n",
      "Epoch 5/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.9057 - acc: 0.7365 - val_loss: 0.9147 - val_acc: 0.7348\n",
      "Epoch 6/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.9051 - acc: 0.7371 - val_loss: 0.9007 - val_acc: 0.7369\n",
      "Epoch 7/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.9032 - acc: 0.7372 - val_loss: 0.9343 - val_acc: 0.7364\n",
      "Epoch 8/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.8944 - acc: 0.7393 - val_loss: 0.8879 - val_acc: 0.7408\n",
      "Epoch 9/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.8801 - acc: 0.7420 - val_loss: 0.8861 - val_acc: 0.7381\n",
      "Epoch 10/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.8731 - acc: 0.7420 - val_loss: 0.8704 - val_acc: 0.7415\n",
      "Epoch 11/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.8589 - acc: 0.7440 - val_loss: 0.8615 - val_acc: 0.7430\n",
      "Epoch 12/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.8562 - acc: 0.7438 - val_loss: 0.8706 - val_acc: 0.7418\n",
      "Epoch 13/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.8500 - acc: 0.7450 - val_loss: 0.8472 - val_acc: 0.7446\n",
      "Epoch 14/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.8379 - acc: 0.7459 - val_loss: 0.8372 - val_acc: 0.7431\n",
      "Epoch 15/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.8236 - acc: 0.7482 - val_loss: 0.8199 - val_acc: 0.7467\n",
      "Epoch 16/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.8107 - acc: 0.7521 - val_loss: 0.8288 - val_acc: 0.7483\n",
      "Epoch 17/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.7904 - acc: 0.7563 - val_loss: 0.7874 - val_acc: 0.7559\n",
      "Epoch 18/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.7820 - acc: 0.7579 - val_loss: 0.7836 - val_acc: 0.7561\n",
      "Epoch 19/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.7732 - acc: 0.7599 - val_loss: 0.7776 - val_acc: 0.7582\n",
      "Epoch 20/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.7549 - acc: 0.7644 - val_loss: 0.7673 - val_acc: 0.7653\n",
      "Epoch 21/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.7453 - acc: 0.7675 - val_loss: 0.7458 - val_acc: 0.7675\n",
      "Epoch 22/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.7290 - acc: 0.7713 - val_loss: 0.7455 - val_acc: 0.7674\n",
      "Epoch 23/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.7133 - acc: 0.7759 - val_loss: 0.7360 - val_acc: 0.7713\n",
      "Epoch 24/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.7016 - acc: 0.7787 - val_loss: 0.7091 - val_acc: 0.7795\n",
      "Epoch 25/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.6859 - acc: 0.7833 - val_loss: 0.7018 - val_acc: 0.7816\n",
      "Epoch 26/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.6699 - acc: 0.7876 - val_loss: 0.6846 - val_acc: 0.7872\n",
      "Epoch 27/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.6504 - acc: 0.7939 - val_loss: 0.6749 - val_acc: 0.7919\n",
      "Epoch 28/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.6321 - acc: 0.7989 - val_loss: 0.6557 - val_acc: 0.7914\n",
      "Epoch 29/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.6158 - acc: 0.8029 - val_loss: 0.6264 - val_acc: 0.8036\n",
      "Epoch 30/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.5937 - acc: 0.8084 - val_loss: 0.6253 - val_acc: 0.8067\n",
      "Epoch 31/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.5723 - acc: 0.8142 - val_loss: 0.5944 - val_acc: 0.8128\n",
      "Epoch 32/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.5535 - acc: 0.8191 - val_loss: 0.5890 - val_acc: 0.8159\n",
      "Epoch 33/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.5366 - acc: 0.8240 - val_loss: 0.5806 - val_acc: 0.8224\n",
      "Epoch 34/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.5113 - acc: 0.8318 - val_loss: 0.5638 - val_acc: 0.8268\n",
      "Epoch 35/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.4991 - acc: 0.8352 - val_loss: 0.5761 - val_acc: 0.8185\n",
      "Epoch 36/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.4803 - acc: 0.8421 - val_loss: 0.5252 - val_acc: 0.8350\n",
      "Epoch 37/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.4609 - acc: 0.8477 - val_loss: 0.5282 - val_acc: 0.8394\n",
      "Epoch 38/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.4489 - acc: 0.8516 - val_loss: 0.5223 - val_acc: 0.8369\n",
      "Epoch 39/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.4366 - acc: 0.8549 - val_loss: 0.5046 - val_acc: 0.8478\n",
      "Epoch 40/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.4108 - acc: 0.8640 - val_loss: 0.4976 - val_acc: 0.8490\n",
      "Epoch 41/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.3985 - acc: 0.8682 - val_loss: 0.4929 - val_acc: 0.8537\n",
      "Epoch 42/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.3779 - acc: 0.8743 - val_loss: 0.4726 - val_acc: 0.8594\n",
      "Epoch 43/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.3648 - acc: 0.8782 - val_loss: 0.4791 - val_acc: 0.8614\n",
      "Epoch 44/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.3586 - acc: 0.8811 - val_loss: 0.4853 - val_acc: 0.8608\n",
      "Epoch 45/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.3514 - acc: 0.8839 - val_loss: 0.4865 - val_acc: 0.8567\n",
      "Epoch 46/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.3360 - acc: 0.8884 - val_loss: 0.4953 - val_acc: 0.8562\n",
      "Epoch 47/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.3220 - acc: 0.8938 - val_loss: 0.4666 - val_acc: 0.8687\n",
      "Epoch 48/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.3002 - acc: 0.8998 - val_loss: 0.4704 - val_acc: 0.8686\n",
      "Epoch 49/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.2918 - acc: 0.9028 - val_loss: 0.4761 - val_acc: 0.8723\n",
      "Epoch 50/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.2887 - acc: 0.9039 - val_loss: 0.4850 - val_acc: 0.8714\n",
      "Epoch 51/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.2805 - acc: 0.9073 - val_loss: 0.4632 - val_acc: 0.8737\n",
      "Epoch 52/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.2638 - acc: 0.9126 - val_loss: 0.4745 - val_acc: 0.8778\n",
      "Epoch 53/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.2555 - acc: 0.9151 - val_loss: 0.4607 - val_acc: 0.8776\n",
      "Epoch 54/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.2390 - acc: 0.9205 - val_loss: 0.4679 - val_acc: 0.8807\n",
      "Epoch 55/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.2287 - acc: 0.9240 - val_loss: 0.4595 - val_acc: 0.8825\n",
      "Epoch 56/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.2261 - acc: 0.9248 - val_loss: 0.4558 - val_acc: 0.8825\n",
      "Epoch 57/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.2168 - acc: 0.9276 - val_loss: 0.4549 - val_acc: 0.8848\n",
      "Epoch 58/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.2188 - acc: 0.9280 - val_loss: 0.4642 - val_acc: 0.8822\n",
      "Epoch 59/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.2064 - acc: 0.9306 - val_loss: 0.4823 - val_acc: 0.8843\n",
      "Epoch 60/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.2024 - acc: 0.9328 - val_loss: 0.4881 - val_acc: 0.8833\n",
      "Epoch 61/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.1951 - acc: 0.9351 - val_loss: 0.4971 - val_acc: 0.8853\n",
      "Epoch 62/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.1839 - acc: 0.9385 - val_loss: 0.4718 - val_acc: 0.8882\n",
      "Epoch 63/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.1821 - acc: 0.9395 - val_loss: 0.4771 - val_acc: 0.8887\n",
      "Epoch 64/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.1808 - acc: 0.9401 - val_loss: 0.4770 - val_acc: 0.8868\n",
      "Epoch 65/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.1731 - acc: 0.9426 - val_loss: 0.4806 - val_acc: 0.8895\n",
      "Epoch 66/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.1658 - acc: 0.9450 - val_loss: 0.4732 - val_acc: 0.8910\n",
      "Epoch 67/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.1641 - acc: 0.9456 - val_loss: 0.4803 - val_acc: 0.8903\n",
      "Epoch 68/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.1588 - acc: 0.9474 - val_loss: 0.4798 - val_acc: 0.8920\n",
      "Epoch 69/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.1665 - acc: 0.9454 - val_loss: 0.4807 - val_acc: 0.8869\n",
      "Epoch 70/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.1566 - acc: 0.9480 - val_loss: 0.4894 - val_acc: 0.8914\n",
      "Epoch 71/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.1312 - acc: 0.9566 - val_loss: 0.4977 - val_acc: 0.8923\n",
      "Epoch 72/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.1293 - acc: 0.9572 - val_loss: 0.5094 - val_acc: 0.8904\n",
      "Epoch 73/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.1438 - acc: 0.9529 - val_loss: 0.4899 - val_acc: 0.8912\n",
      "Epoch 74/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.1381 - acc: 0.9542 - val_loss: 0.5080 - val_acc: 0.8933\n",
      "Epoch 75/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.1822 - acc: 0.9425 - val_loss: 0.5245 - val_acc: 0.8863\n",
      "Epoch 76/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.1370 - acc: 0.9553 - val_loss: 0.4855 - val_acc: 0.8929\n",
      "Epoch 77/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.1119 - acc: 0.9626 - val_loss: 0.5018 - val_acc: 0.8936\n",
      "Epoch 78/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.1108 - acc: 0.9639 - val_loss: 0.5057 - val_acc: 0.8953\n",
      "Epoch 79/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.1035 - acc: 0.9659 - val_loss: 0.5251 - val_acc: 0.8936\n",
      "Epoch 80/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.1263 - acc: 0.9588 - val_loss: 0.5397 - val_acc: 0.8874\n",
      "Epoch 81/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.1307 - acc: 0.9574 - val_loss: 0.5103 - val_acc: 0.8934\n",
      "Epoch 82/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.1142 - acc: 0.9623 - val_loss: 0.5268 - val_acc: 0.8955\n",
      "Epoch 83/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.0993 - acc: 0.9673 - val_loss: 0.5334 - val_acc: 0.8959\n",
      "Epoch 84/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.0882 - acc: 0.9710 - val_loss: 0.5239 - val_acc: 0.8972\n",
      "Epoch 85/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.0825 - acc: 0.9728 - val_loss: 0.5354 - val_acc: 0.8977\n",
      "Epoch 86/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.0803 - acc: 0.9738 - val_loss: 0.5410 - val_acc: 0.8957\n",
      "Epoch 87/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.0861 - acc: 0.9718 - val_loss: 0.5470 - val_acc: 0.8954\n",
      "Epoch 88/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.0889 - acc: 0.9703 - val_loss: 0.5612 - val_acc: 0.8967\n",
      "Epoch 89/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.1345 - acc: 0.9585 - val_loss: 0.5467 - val_acc: 0.8891\n",
      "Epoch 90/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.1307 - acc: 0.9584 - val_loss: 0.5240 - val_acc: 0.8917\n",
      "Epoch 91/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.1802 - acc: 0.9462 - val_loss: 0.5259 - val_acc: 0.8904\n",
      "Epoch 92/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.1227 - acc: 0.9611 - val_loss: 0.5098 - val_acc: 0.8953\n",
      "Epoch 93/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.0879 - acc: 0.9717 - val_loss: 0.5264 - val_acc: 0.8932\n",
      "Epoch 94/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.0758 - acc: 0.9756 - val_loss: 0.5369 - val_acc: 0.8999\n",
      "Epoch 95/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.0651 - acc: 0.9791 - val_loss: 0.5359 - val_acc: 0.9005\n",
      "Epoch 96/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.0730 - acc: 0.9763 - val_loss: 0.5568 - val_acc: 0.8954\n",
      "Epoch 97/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.0741 - acc: 0.9759 - val_loss: 0.5563 - val_acc: 0.8979\n",
      "Epoch 98/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.0750 - acc: 0.9762 - val_loss: 0.5564 - val_acc: 0.8983\n",
      "Epoch 99/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.0773 - acc: 0.9751 - val_loss: 0.5549 - val_acc: 0.8980\n",
      "Epoch 100/100\n",
      "7048/7048 [==============================] - 8s - loss: 0.0700 - acc: 0.9773 - val_loss: 0.5530 - val_acc: 0.9012\n"
     ]
    }
   ],
   "source": [
    "## Params\n",
    "\n",
    "BATCH_SIZE=64\n",
    "MAX_EPOCHS=150\n",
    "\n",
    "hist = model.fit(\n",
    "    X_train, y_train,\n",
    "    validation_data=(X_val, y_val),\n",
    "    batch_size=BATCH_SIZE,\n",
    "    epochs=MAX_EPOCHS,\n",
    "    verbose=1\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAa0AAAEKCAYAAAChTwphAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fdf77d75898>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Visualize history\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "fig, loss_ax = plt.subplots()\n",
    "\n",
    "acc_ax = loss_ax.twinx()\n",
    "\n",
    "loss_ax.plot(hist.history['loss'] + hist2.history['loss'], 'y', label='train loss')\n",
    "loss_ax.plot(hist.history['val_loss'] + hist2.history['val_loss'], 'g', label='val loss')\n",
    "acc_ax.plot(hist.history['acc'] + hist2.history['acc'], 'b', label='train acc')\n",
    "acc_ax.plot(hist.history['val_acc'] + hist2.history['val_acc'], 'r', label='val acc')\n",
    "\n",
    "loss_ax.set_xlabel('epoch')\n",
    "loss_ax.set_ylabel('loss')\n",
    "acc_ax.set_ylabel('accuracy')\n",
    "\n",
    "loss_ax.legend(loc='upper left')\n",
    "acc_ax.legend(loc='lower left')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAa0AAAEKCAYAAAChTwphAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fdf77cb3390>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Visualize history\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "fig, loss_ax = plt.subplots()\n",
    "\n",
    "acc_ax = loss_ax.twinx()\n",
    "\n",
    "loss_ax.plot(data['loss'], 'y', label='train loss')\n",
    "loss_ax.plot(data['val_loss'], 'g', label='val loss')\n",
    "acc_ax.plot(data['acc'], 'b', label='train acc')\n",
    "acc_ax.plot(data['val_acc'], 'r', label='val acc')\n",
    "\n",
    "loss_ax.set_xlabel('epoch')\n",
    "loss_ax.set_ylabel('loss')\n",
    "acc_ax.set_ylabel('accuracy')\n",
    "\n",
    "loss_ax.legend(loc='upper left')\n",
    "acc_ax.legend(loc='lower left')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    },
    {
     "ename": "NameError",
     "evalue": "name 'model' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-1-1b561f1fdad9>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      5\u001b[0m \u001b[0mMAX_EPOCHS\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m150\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      6\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 7\u001b[0;31m hist = model.fit(\n\u001b[0m\u001b[1;32m      8\u001b[0m     \u001b[0mX_train\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my_train\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      9\u001b[0m     \u001b[0mvalidation_data\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mX_val\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my_val\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mNameError\u001b[0m: name 'model' is not defined"
     ]
    }
   ],
   "source": [
    "## Params\n",
    "from keras_tqdm import TQDMNotebookCallback\n",
    "\n",
    "BATCH_SIZE=64\n",
    "MAX_EPOCHS=150\n",
    "\n",
    "hist = model.fit(\n",
    "    X_train, y_train,\n",
    "    validation_data=(X_val, y_val),\n",
    "    batch_size=BATCH_SIZE,\n",
    "    epochs=MAX_EPOCHS,\n",
    "    verbose=1,\n",
    "    callbacks=[TQDMNotebookCallback()]\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.5.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}