ASR_RNN.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "#  Compatibility imports\n",
    "from __future__ import absolute_import\n",
    "from __future__ import division\n",
    "from __future__ import print_function\n",
    "\n",
    "import time\n",
    "\n",
    "import tensorflow as tf\n",
    "import scipy.io.wavfile as wav\n",
    "import numpy as np\n",
    "import glob\n",
    "\n",
    "from six.moves import xrange as range\n",
    "\n",
    "import soundfile as sf\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "try:\n",
    "    from python_speech_features import mfcc\n",
    "except ImportError:\n",
    "    print(\"Failed to import python_speech_features.\\n Try pip install python_speech_features.\")\n",
    "    raise ImportError\n",
    "\n",
    "from utils import maybe_download as maybe_download\n",
    "from utils import sparse_tuple_from as sparse_tuple_from\n",
    "\n",
    "from keras.preprocessing.sequence import pad_sequences"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "ename": "ModuleNotFoundError",
     "evalue": "No module named 'nltk'",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mModuleNotFoundError\u001b[0m                       Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-4-d2400d7045d1>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0;32mimport\u001b[0m \u001b[0mnltk\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      2\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0mnltk\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mstem\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlancaster\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mLancasterStemmer\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      3\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0mnltk\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mngrams\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mModuleNotFoundError\u001b[0m: No module named 'nltk'"
     ]
    }
   ],
   "source": [
    "import nltk\n",
    "from nltk.stem.lancaster import LancasterStemmer\n",
    "from nltk import ngrams"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "nltk.download('punkt')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Constants\n",
    "SPACE_TOKEN = '<space>'\n",
    "SPACE_INDEX = 0\n",
    "FIRST_INDEX = ord('a') - 1  # 0 is reserved to space\n",
    "\n",
    "# Some configs\n",
    "num_features = 13\n",
    "num_units=50 # Number of units in the LSTM cell\n",
    "# Accounting the 0th indice +  space + blank label = 28 characters\n",
    "num_classes = ord('z') - ord('a') + 1 + 1 + 1\n",
    "\n",
    "# Hyper-parameters\n",
    "num_epochs = 200\n",
    "num_hidden = 50\n",
    "num_layers = 2\n",
    "batch_size = 1\n",
    "initial_learning_rate = 1e-2\n",
    "momentum = 0.9\n",
    "\n",
    "num_examples = 1\n",
    "num_batches_per_epoch = int(num_examples/batch_size)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "ROOT_DIRECTORY = 'TIMIT_full/'\n",
    "\n",
    "WAV_CLASS = '*.wav'\n",
    "TXT_CLASS = '*.txt'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "Clean"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "TRAIN_DIRECTORY = 'train_16k/'\n",
    "VAL_DIRECTORY = 'val_16k/'\n",
    "TEST_DIRECTORY = 'test_16k/'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "Noisy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "TRAIN_DIRECTORY = 'timit_train_16k/'\n",
    "VAL_DIRECTORY = 'timit_val_16k/'\n",
    "TEST_DIRECTORY = 'timit_test_16k/'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "def load_dataset(wav_path,txt_path):\n",
    "\n",
    "    x_s = []\n",
    "    y_s = []\n",
    "    text_s = []\n",
    "    print(wav_path, txt_path)\n",
    "    for e, (f1, f2) in enumerate( zip(sorted(glob.glob(wav_path)),sorted(glob.glob(txt_path))) ) :\n",
    "#         print('FileName:',f1, f2)\n",
    "        \n",
    "        ##### Read Audio features #####\n",
    "        ###############################\n",
    "        audio, fs = sf.read(f1)\n",
    "        inputs = mfcc(audio, samplerate=fs)\n",
    "        \n",
    "        # Tranform in 3D array\n",
    "        train_inputs = np.asarray(inputs[np.newaxis, :])\n",
    "        train_inputs = (train_inputs - np.mean(train_inputs))/np.std(train_inputs)\n",
    "#         print('train_inputs.shape:', train_inputs.shape)\n",
    "        \n",
    "        train_seq_len = [train_inputs.shape[1]]\n",
    "#         print('train_seq_len.len:', len(train_seq_len) )\n",
    "        \n",
    "        x_s.append(train_inputs)\n",
    "#         print('x_s.len', len(x_s))\n",
    "        \n",
    "        ##### Read Labels features #####\n",
    "        ###############################\n",
    "        with open(f2, 'r') as txt_f:\n",
    "            line = txt_f.readlines()[-1] #Only the last line is necessary\n",
    "\n",
    "            targets = preprocess_line(line)\n",
    "            text_s.append(targets)\n",
    "\n",
    "            # Adding blank label\n",
    "            targets = np.hstack([SPACE_TOKEN if x == '' else list(x) for x in targets])\n",
    "\n",
    "            # Transform char into index\n",
    "            targets = np.asarray([SPACE_INDEX if x == SPACE_TOKEN else ord(x) - FIRST_INDEX\n",
    "                                  for x in targets])\n",
    "\n",
    "            # Creating sparse representation to feed the placeholder\n",
    "            train_targets = sparse_tuple_from([targets])\n",
    "#             print('train_targets.shape:', len(train_targets))\n",
    "            y_s.append(train_targets)\n",
    "              \n",
    "    \n",
    "    print('x_s.len', len(x_s),', [-1]x_s.len', x_s[-1].shape)\n",
    "#     x_s = pad_sequences(x_s, maxlen=500, dtype='float', padding='post', truncating='post')\n",
    "#     print('x_s.shape', (x_s.shape))\n",
    "        \n",
    "    \n",
    "    return x_s, y_s,text_s\n",
    "        "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "def preprocess_line(line):\n",
    "    \n",
    "    original = ' '.join(line.strip().lower().split(' ')[2:])\n",
    "    for c in set(original):\n",
    "        if c != ' ' and not c.isalpha():\n",
    "            original = original.replace(c,'')\n",
    "        \n",
    "    targets = original.replace(' ', '  ')\n",
    "    targets = targets.split(' ')\n",
    "\n",
    "    i = 0\n",
    "    while i<len(targets)-1:\n",
    "        if targets[i]==targets[i+1] and targets[i+1]=='':\n",
    "            del targets[i+1]\n",
    "        else:\n",
    "            i +=1\n",
    "            \n",
    "    return targets\n",
    "    \n",
    "    # Get only the words between [a-z] and replace period for none\n",
    "#     original = ' '.join(line.strip().lower().split(' ')[2:]).replace('.', '').replace('\\'','')\n",
    "#     targets = original.replace(' ', '  ')\n",
    "#     return targets.split(' ')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "TIMIT_full/train_16k/*.wav TIMIT_full/train_16k/*.txt\n",
      "x_s.len 4576 , [-1]x_s.len (1, 204, 13)\n",
      "4576 4576\n"
     ]
    }
   ],
   "source": [
    "x_, y_,text_ = load_dataset(ROOT_DIRECTORY+TRAIN_DIRECTORY+WAV_CLASS, ROOT_DIRECTORY+TRAIN_DIRECTORY+TXT_CLASS)  \n",
    "print(len(x_), len(y_))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "TIMIT_full/val_16k/*.wav TIMIT_full/val_16k/*.txt\n",
      "x_s.len 44 , [-1]x_s.len (1, 194, 13)\n",
      "44 44\n"
     ]
    }
   ],
   "source": [
    "val_x_, val_y_, val_text_ = load_dataset(ROOT_DIRECTORY+VAL_DIRECTORY+WAV_CLASS, ROOT_DIRECTORY+VAL_DIRECTORY+TXT_CLASS)  \n",
    "print(len(val_x_), len(val_y_))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "TIMIT_full/test_16k/*.wav TIMIT_full/test_16k/*.txt\n",
      "x_s.len 1680 , [-1]x_s.len (1, 315, 13)\n",
      "1680 1680\n"
     ]
    }
   ],
   "source": [
    "test_x_, test_y_, test_text_ = load_dataset(ROOT_DIRECTORY+TEST_DIRECTORY+WAV_CLASS, ROOT_DIRECTORY+TEST_DIRECTORY+TXT_CLASS)  \n",
    "print(len(test_x_), len(test_y_))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "# THE MAIN CODE! LSTM\n",
    "\n",
    "graph = tf.Graph()\n",
    "with graph.as_default():\n",
    "    # e.g: log filter bank or MFCC features\n",
    "    # Has size [batch_size, max_stepsize, num_features], but the\n",
    "    # batch_size and max_stepsize can vary along each step\n",
    "    inputs = tf.placeholder(tf.float32, [None, None, num_features])\n",
    "\n",
    "    # Here we use sparse_placeholder that will generate a\n",
    "    # SparseTensor required by ctc_loss op.\n",
    "    targets = tf.sparse_placeholder(tf.int32)\n",
    "\n",
    "    # 1d array of size [batch_size]\n",
    "    seq_len = tf.placeholder(tf.int32, [None])\n",
    "\n",
    "    # Defining the cell\n",
    "    # Can be:\n",
    "    #   tf.nn.rnn_cell.RNNCell\n",
    "    #   tf.nn.rnn_cell.GRUCell \n",
    "    cells = []\n",
    "    for _ in range(num_layers):\n",
    "        cell = tf.contrib.rnn.LSTMCell(num_units)  # Or LSTMCell(num_units)\n",
    "        cells.append(cell)\n",
    "    stack = tf.contrib.rnn.MultiRNNCell(cells)\n",
    "\n",
    "    # The second output is the last state and we will no use that\n",
    "    outputs, _ = tf.nn.dynamic_rnn(stack, inputs, seq_len, dtype=tf.float32)\n",
    "\n",
    "    shape = tf.shape(inputs)\n",
    "    batch_s, max_timesteps = shape[0], shape[1]\n",
    "\n",
    "    # Reshaping to apply the same weights over the timesteps\n",
    "    outputs = tf.reshape(outputs, [-1, num_hidden])\n",
    "\n",
    "    # Truncated normal with mean 0 and stdev=0.1\n",
    "    # Tip: Try another initialization\n",
    "    # see https://www.tensorflow.org/versions/r0.9/api_docs/python/contrib.layers.html#initializers\n",
    "    W = tf.Variable(tf.truncated_normal([num_hidden,\n",
    "                                         num_classes],\n",
    "                                        stddev=0.1))\n",
    "    # Zero initialization\n",
    "    # Tip: Is tf.zeros_initializer the same?\n",
    "    b = tf.Variable(tf.constant(0., shape=[num_classes]))\n",
    "\n",
    "    # Doing the affine projection\n",
    "    logits = tf.matmul(outputs, W) + b\n",
    "\n",
    "    # Reshaping back to the original shape\n",
    "    logits = tf.reshape(logits, [batch_s, -1, num_classes])\n",
    "\n",
    "    # Time major\n",
    "    logits = tf.transpose(logits, (1, 0, 2))\n",
    "\n",
    "    loss = tf.nn.ctc_loss(targets, logits, seq_len)\n",
    "    cost = tf.reduce_mean(loss)\n",
    "\n",
    "    optimizer = tf.train.MomentumOptimizer(initial_learning_rate,\n",
    "                                           0.9).minimize(cost)\n",
    "\n",
    "    # Option 2: tf.nn.ctc_beam_search_decoder\n",
    "    # (it's slower but you'll get better results)\n",
    "    decoded, log_prob = tf.nn.ctc_greedy_decoder(logits, seq_len)\n",
    "\n",
    "    # Inaccuracy: label error rate\n",
    "    ler = tf.reduce_mean(tf.edit_distance(tf.cast(decoded[0], tf.int32),\n",
    "                                          targets))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "# THE MAIN CODE! BiLSTM+GRU\n",
    "\n",
    "graph = tf.Graph()\n",
    "with graph.as_default():\n",
    "    # e.g: log filter bank or MFCC features\n",
    "    # Has size [batch_size, max_stepsize, num_features], but the\n",
    "    # batch_size and max_stepsize can vary along each step\n",
    "    inputs = tf.placeholder(tf.float32, [None, None, num_features])\n",
    "\n",
    "    # Here we use sparse_placeholder that will generate a\n",
    "    # SparseTensor required by ctc_loss op.\n",
    "    targets = tf.sparse_placeholder(tf.int32)\n",
    "\n",
    "    # 1d array of size [batch_size]\n",
    "    seq_len = tf.placeholder(tf.int32, [None])\n",
    "\n",
    "    # Defining the cell\n",
    "    # Can be:\n",
    "    #   tf.nn.rnn_cell.RNNCell\n",
    "    #   tf.nn.rnn_cell.GRUCell \n",
    "    cells = []\n",
    "    for _ in range(num_layers):\n",
    "        cell = tf.contrib.rnn.GRUCell(num_units)  # Or LSTMCell(num_units)\n",
    "        cells.append(cell)\n",
    "    stack = tf.contrib.rnn.MultiRNNCell(cells)\n",
    "    \n",
    "    cells_bw = []\n",
    "    for _ in range(num_layers):\n",
    "        cell_bw = tf.contrib.rnn.GRUCell(num_units)  # Or LSTMCell(num_units)\n",
    "        cells_bw.append(cell_bw)\n",
    "    stack_bw = tf.contrib.rnn.MultiRNNCell(cells)\n",
    "\n",
    "    # The second output is the last state and we will no use that\n",
    "    outputs, _ = tf.nn.bidirectional_dynamic_rnn(stack,stack_bw, inputs, seq_len, dtype=tf.float32)\n",
    "\n",
    "    shape = tf.shape(inputs)\n",
    "    batch_s, max_timesteps = shape[0], shape[1]\n",
    "\n",
    "    # Reshaping to apply the same weights over the timesteps\n",
    "    outputs = tf.reshape(outputs, [-1, num_hidden])\n",
    "\n",
    "    # Truncated normal with mean 0 and stdev=0.1\n",
    "    # Tip: Try another initialization\n",
    "    # see https://www.tensorflow.org/versions/r0.9/api_docs/python/contrib.layers.html#initializers\n",
    "    W = tf.Variable(tf.truncated_normal([num_hidden,\n",
    "                                         num_classes],\n",
    "                                        stddev=0.1))\n",
    "    # Zero initialization\n",
    "    # Tip: Is tf.zeros_initializer the same?\n",
    "    b = tf.Variable(tf.constant(0., shape=[num_classes]))\n",
    "\n",
    "    # Doing the affine projection\n",
    "    logits = tf.matmul(outputs, W) + b\n",
    "\n",
    "    # Reshaping back to the original shape\n",
    "    logits = tf.reshape(logits, [batch_s, -1, num_classes])\n",
    "\n",
    "    # Time major\n",
    "    logits = tf.transpose(logits, (1, 0, 2))\n",
    "\n",
    "    loss = tf.nn.ctc_loss(targets, logits, seq_len)\n",
    "    cost = tf.reduce_mean(loss)\n",
    "\n",
    "    optimizer = tf.train.MomentumOptimizer(initial_learning_rate,\n",
    "                                           0.9).minimize(cost)\n",
    "\n",
    "    # Option 2: tf.nn.ctc_beam_search_decoder\n",
    "    # (it's slower but you'll get better results)\n",
    "    decoded, log_prob = tf.nn.ctc_greedy_decoder(logits, seq_len)\n",
    "\n",
    "    # Inaccuracy: label error rate\n",
    "    ler = tf.reduce_mean(tf.edit_distance(tf.cast(decoded[0], tf.int32),\n",
    "                                          targets))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "batch: 0 4246 (1, 405, 13)\n",
      "200 1976.6066131591797 4.0 2.321467638015747\n",
      "Epoch 1/200, train_cost = 1976.607, train_ler = 4.000, val_cost = 176.181, val_ler = 2.000, time = 2.533\n",
      "batch: 0 3893 (1, 389, 13)\n",
      "200 2697.872573852539 3.850834310054779 2.7401297092437744\n",
      "Epoch 2/200, train_cost = 2697.873, train_ler = 3.851, val_cost = 1231.738, val_ler = 1.874, time = 3.066\n",
      "batch: 0 3300 (1, 247, 13)\n",
      "200 1412.4134826660156 3.7042123675346375 2.5466465950012207\n",
      "Epoch 3/200, train_cost = 1412.413, train_ler = 3.704, val_cost = 517.885, val_ler = 1.745, time = 2.776\n",
      "batch: 0 3214 (1, 208, 13)\n",
      "200 823.6518859863281 3.4106186628341675 2.0286641120910645\n",
      "Epoch 4/200, train_cost = 823.652, train_ler = 3.411, val_cost = 660.005, val_ler = 1.746, time = 2.450\n",
      "batch: 0 635 (1, 303, 13)\n",
      "200 1087.6393280029297 3.5416603088378906 2.3202409744262695\n",
      "Epoch 5/200, train_cost = 1087.639, train_ler = 3.542, val_cost = 262.486, val_ler = 1.660, time = 2.536\n",
      "batch: 0 3850 (1, 300, 13)\n",
      "200 971.4058685302734 3.3425958156585693 2.5970828533172607\n",
      "Epoch 6/200, train_cost = 971.406, train_ler = 3.343, val_cost = 450.334, val_ler = 1.789, time = 2.856\n",
      "batch: 0 2220 (1, 211, 13)\n",
      "200 804.5540161132812 3.329553723335266 2.2298688888549805\n",
      "Epoch 7/200, train_cost = 804.554, train_ler = 3.330, val_cost = 306.305, val_ler = 1.764, time = 2.454\n",
      "batch: 0 243 (1, 307, 13)\n",
      "200 1069.0091857910156 3.9423076510429382 2.74409818649292\n",
      "Epoch 8/200, train_cost = 1069.009, train_ler = 3.942, val_cost = 487.856, val_ler = 1.919, time = 3.043\n",
      "batch: 0 577 (1, 318, 13)\n",
      "200 834.7809753417969 3.4469757676124573 2.6202595233917236\n",
      "Epoch 9/200, train_cost = 834.781, train_ler = 3.447, val_cost = 462.316, val_ler = 2.000, time = 2.865\n",
      "batch: 0 1829 (1, 225, 13)\n",
      "200 953.9486694335938 3.0759366750717163 2.221971035003662\n",
      "Epoch 10/200, train_cost = 953.949, train_ler = 3.076, val_cost = 641.818, val_ler = 1.567, time = 2.651\n",
      "batch: 0 4344 (1, 275, 13)\n",
      "200 1087.9081268310547 3.3144115805625916 2.2933247089385986\n",
      "Epoch 11/200, train_cost = 1087.908, train_ler = 3.314, val_cost = 618.460, val_ler = 1.726, time = 2.639\n",
      "batch: 0 1768 (1, 338, 13)\n",
      "200 1836.1343536376953 3.555531144142151 2.9933888912200928\n",
      "Epoch 12/200, train_cost = 1836.134, train_ler = 3.556, val_cost = 807.608, val_ler = 1.649, time = 3.404\n",
      "batch: 0 3112 (1, 307, 13)\n",
      "200 912.6245269775391 3.3039377331733704 2.223280429840088\n",
      "Epoch 13/200, train_cost = 912.625, train_ler = 3.304, val_cost = 470.786, val_ler = 2.113, time = 2.456\n",
      "batch: 0 3465 (1, 437, 13)\n",
      "200 1167.934829711914 3.600280463695526 2.5559017658233643\n",
      "Epoch 14/200, train_cost = 1167.935, train_ler = 3.600, val_cost = 355.936, val_ler = 1.675, time = 2.863\n",
      "batch: 0 845 (1, 235, 13)\n",
      "200 803.5740966796875 3.4275596737861633 1.8627650737762451\n",
      "Epoch 15/200, train_cost = 803.574, train_ler = 3.428, val_cost = 588.727, val_ler = 1.853, time = 2.131\n",
      "batch: 0 2677 (1, 320, 13)\n",
      "200 887.2008514404297 3.522567331790924 2.10526442527771\n",
      "Epoch 16/200, train_cost = 887.201, train_ler = 3.523, val_cost = 692.207, val_ler = 1.670, time = 2.467\n",
      "batch: 0 80 (1, 302, 13)\n",
      "200 814.6145324707031 3.073035418987274 2.3624207973480225\n",
      "Epoch 17/200, train_cost = 814.615, train_ler = 3.073, val_cost = 464.372, val_ler = 1.562, time = 2.640\n",
      "batch: 0 4338 (1, 291, 13)\n",
      "200 849.1654205322266 3.30342435836792 2.645787239074707\n",
      "Epoch 18/200, train_cost = 849.165, train_ler = 3.303, val_cost = 820.731, val_ler = 1.810, time = 2.931\n",
      "batch: 0 3206 (1, 380, 13)\n",
      "200 717.5823593139648 3.1933318972587585 2.226073980331421\n",
      "Epoch 19/200, train_cost = 717.582, train_ler = 3.193, val_cost = 307.445, val_ler = 1.583, time = 2.540\n",
      "batch: 0 2519 (1, 297, 13)\n",
      "200 781.3793487548828 3.214285731315613 2.3534698486328125\n",
      "Epoch 20/200, train_cost = 781.379, train_ler = 3.214, val_cost = 390.182, val_ler = 1.630, time = 2.713\n",
      "batch: 0 1742 (1, 201, 13)\n",
      "200 1150.6809997558594 3.58342045545578 2.5261383056640625\n",
      "Epoch 21/200, train_cost = 1150.681, train_ler = 3.583, val_cost = 490.094, val_ler = 1.735, time = 2.779\n",
      "batch: 0 2295 (1, 263, 13)\n",
      "200 845.6755065917969 3.5384411215782166 2.616842269897461\n",
      "Epoch 22/200, train_cost = 845.676, train_ler = 3.538, val_cost = 590.313, val_ler = 1.922, time = 2.878\n",
      "batch: 0 4531 (1, 280, 13)\n",
      "200 858.5399627685547 4.416355311870575 2.1451549530029297\n",
      "Epoch 23/200, train_cost = 858.540, train_ler = 4.416, val_cost = 751.157, val_ler = 2.005, time = 2.449\n",
      "batch: 0 1484 (1, 252, 13)\n",
      "200 1487.6480712890625 3.2373300790786743 2.2289342880249023\n",
      "Epoch 24/200, train_cost = 1487.648, train_ler = 3.237, val_cost = 718.557, val_ler = 1.664, time = 2.549\n",
      "batch: 0 2847 (1, 198, 13)\n",
      "200 1161.808822631836 2.953260362148285 2.23447847366333\n",
      "Epoch 25/200, train_cost = 1161.809, train_ler = 2.953, val_cost = 671.011, val_ler = 1.701, time = 2.518\n",
      "batch: 0 4363 (1, 280, 13)\n",
      "200 1272.751220703125 3.6090824007987976 2.4689319133758545\n",
      "Epoch 26/200, train_cost = 1272.751, train_ler = 3.609, val_cost = 617.251, val_ler = 1.664, time = 2.711\n",
      "batch: 0 472 (1, 250, 13)\n",
      "200 1314.013671875 3.373524844646454 3.1686184406280518\n",
      "Epoch 27/200, train_cost = 1314.014, train_ler = 3.374, val_cost = 474.893, val_ler = 1.468, time = 3.528\n",
      "batch: 0 4471 (1, 210, 13)\n",
      "200 942.0341949462891 3.0930579900741577 2.2749762535095215\n",
      "Epoch 28/200, train_cost = 942.034, train_ler = 3.093, val_cost = 450.467, val_ler = 1.585, time = 2.579\n",
      "batch: 0 4023 (1, 499, 13)\n",
      "200 1301.7052612304688 3.3007142543792725 3.0260958671569824\n",
      "Epoch 29/200, train_cost = 1301.705, train_ler = 3.301, val_cost = 358.329, val_ler = 1.699, time = 3.239\n",
      "batch: 0 4282 (1, 309, 13)\n",
      "200 1111.9794311523438 3.4184715151786804 2.753424882888794\n",
      "Epoch 30/200, train_cost = 1111.979, train_ler = 3.418, val_cost = 426.592, val_ler = 1.598, time = 3.143\n",
      "batch: 0 3404 (1, 211, 13)\n",
      "200 891.9694976806641 3.166664242744446 2.442080497741699\n",
      "Epoch 31/200, train_cost = 891.969, train_ler = 3.167, val_cost = 760.339, val_ler = 1.693, time = 2.830\n",
      "batch: 0 1838 (1, 251, 13)\n",
      "200 1056.6273651123047 3.156222641468048 2.4581751823425293\n",
      "Epoch 32/200, train_cost = 1056.627, train_ler = 3.156, val_cost = 544.446, val_ler = 1.625, time = 2.800\n",
      "batch: 0 2416 (1, 249, 13)\n",
      "200 903.9184265136719 3.4672786593437195 2.4298455715179443\n",
      "Epoch 33/200, train_cost = 903.918, train_ler = 3.467, val_cost = 397.002, val_ler = 1.665, time = 2.773\n",
      "batch: 0 1070 (1, 371, 13)\n",
      "200 969.7411804199219 3.016416609287262 3.079531192779541\n",
      "Epoch 34/200, train_cost = 969.741, train_ler = 3.016, val_cost = 461.263, val_ler = 1.495, time = 3.364\n",
      "batch: 0 2387 (1, 195, 13)\n",
      "200 1023.2970275878906 3.2837478518486023 2.368861198425293\n",
      "Epoch 35/200, train_cost = 1023.297, train_ler = 3.284, val_cost = 626.799, val_ler = 2.096, time = 2.672\n",
      "batch: 0 3119 (1, 286, 13)\n",
      "200 1203.3353729248047 3.6202797889709473 2.0581371784210205\n",
      "Epoch 36/200, train_cost = 1203.335, train_ler = 3.620, val_cost = 953.883, val_ler = 1.692, time = 2.459\n",
      "batch: 0 1074 (1, 478, 13)\n",
      "200 1598.1207885742188 3.147355794906616 2.694622755050659\n",
      "Epoch 37/200, train_cost = 1598.121, train_ler = 3.147, val_cost = 803.876, val_ler = 1.595, time = 3.008\n",
      "batch: 0 1559 (1, 203, 13)\n",
      "200 1665.935806274414 3.918747842311859 2.304176092147827\n",
      "Epoch 38/200, train_cost = 1665.936, train_ler = 3.919, val_cost = 1037.473, val_ler = 1.518, time = 2.593\n",
      "batch: 0 2344 (1, 227, 13)\n",
      "200 1825.851318359375 3.7007948756217957 2.5082848072052\n",
      "Epoch 39/200, train_cost = 1825.851, train_ler = 3.701, val_cost = 688.178, val_ler = 2.521, time = 2.830\n",
      "batch: 0 3295 (1, 281, 13)\n",
      "200 1504.2810974121094 3.32763534784317 2.4567935466766357\n",
      "Epoch 40/200, train_cost = 1504.281, train_ler = 3.328, val_cost = 847.894, val_ler = 1.648, time = 2.751\n",
      "batch: 0 3756 (1, 366, 13)\n",
      "200 1503.2043762207031 2.938794493675232 2.4235055446624756\n",
      "Epoch 41/200, train_cost = 1503.204, train_ler = 2.939, val_cost = 810.476, val_ler = 1.557, time = 2.799\n",
      "batch: 0 1396 (1, 243, 13)\n",
      "200 1216.1765594482422 3.6088497638702393 1.7901227474212646\n",
      "Epoch 42/200, train_cost = 1216.177, train_ler = 3.609, val_cost = 612.976, val_ler = 1.460, time = 2.148\n",
      "batch: 0 1743 (1, 174, 13)\n",
      "200 1415.4249114990234 3.5162078142166138 2.2519919872283936\n",
      "Epoch 43/200, train_cost = 1415.425, train_ler = 3.516, val_cost = 514.222, val_ler = 1.619, time = 2.529\n",
      "batch: 0 1308 (1, 167, 13)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "200 1080.8341064453125 4.474288463592529 1.8302841186523438\n",
      "Epoch 44/200, train_cost = 1080.834, train_ler = 4.474, val_cost = 584.812, val_ler = 1.562, time = 2.157\n",
      "batch: 0 748 (1, 291, 13)\n",
      "200 911.7025146484375 3.134920656681061 2.2950332164764404\n",
      "Epoch 45/200, train_cost = 911.703, train_ler = 3.135, val_cost = 646.696, val_ler = 1.619, time = 2.694\n",
      "batch: 0 4065 (1, 331, 13)\n",
      "200 975.6529998779297 3.1426987051963806 2.74700665473938\n",
      "Epoch 46/200, train_cost = 975.653, train_ler = 3.143, val_cost = 598.065, val_ler = 1.480, time = 3.145\n",
      "batch: 0 876 (1, 220, 13)\n",
      "200 1138.7510833740234 3.1685675978660583 2.552401065826416\n",
      "Epoch 47/200, train_cost = 1138.751, train_ler = 3.169, val_cost = 744.466, val_ler = 1.966, time = 2.881\n",
      "batch: 0 3201 (1, 302, 13)\n",
      "200 1118.7721405029297 3.3717570900917053 2.611130952835083\n",
      "Epoch 48/200, train_cost = 1118.772, train_ler = 3.372, val_cost = 678.736, val_ler = 1.674, time = 2.936\n",
      "batch: 0 2718 (1, 371, 13)\n",
      "200 1122.4728698730469 3.5191876888275146 2.849470615386963\n",
      "Epoch 49/200, train_cost = 1122.473, train_ler = 3.519, val_cost = 522.632, val_ler = 1.831, time = 3.125\n",
      "batch: 0 3946 (1, 400, 13)\n",
      "200 947.634033203125 3.5107816457748413 2.4909567832946777\n",
      "Epoch 50/200, train_cost = 947.634, train_ler = 3.511, val_cost = 398.893, val_ler = 1.678, time = 2.753\n",
      "batch: 0 3889 (1, 284, 13)\n",
      "200 863.6466369628906 3.2820788621902466 2.464533567428589\n",
      "Epoch 51/200, train_cost = 863.647, train_ler = 3.282, val_cost = 527.421, val_ler = 1.702, time = 2.784\n",
      "batch: 0 4263 (1, 238, 13)\n",
      "200 953.5758972167969 3.4223999977111816 2.468881607055664\n",
      "Epoch 52/200, train_cost = 953.576, train_ler = 3.422, val_cost = 426.058, val_ler = 1.875, time = 2.756\n",
      "batch: 0 978 (1, 462, 13)\n",
      "200 981.0610198974609 3.62316358089447 3.220759153366089\n",
      "Epoch 53/200, train_cost = 981.061, train_ler = 3.623, val_cost = 467.227, val_ler = 1.750, time = 3.587\n",
      "batch: 0 3243 (1, 272, 13)\n",
      "200 826.7026672363281 3.7168240547180176 2.3819010257720947\n",
      "Epoch 54/200, train_cost = 826.703, train_ler = 3.717, val_cost = 415.455, val_ler = 1.888, time = 2.672\n",
      "batch: 0 2754 (1, 237, 13)\n",
      "200 805.8516235351562 3.509482264518738 2.3618319034576416\n",
      "Epoch 55/200, train_cost = 805.852, train_ler = 3.509, val_cost = 444.456, val_ler = 1.897, time = 2.608\n",
      "batch: 0 1029 (1, 385, 13)\n",
      "200 928.2310943603516 3.371260106563568 2.395876407623291\n",
      "Epoch 56/200, train_cost = 928.231, train_ler = 3.371, val_cost = 465.667, val_ler = 1.856, time = 2.674\n",
      "batch: 0 1962 (1, 165, 13)\n",
      "200 794.8100433349609 3.85002863407135 1.9019050598144531\n",
      "Epoch 57/200, train_cost = 794.810, train_ler = 3.850, val_cost = 418.203, val_ler = 1.751, time = 2.170\n",
      "batch: 0 2070 (1, 356, 13)\n",
      "200 964.8188781738281 3.4712324738502502 2.090045213699341\n",
      "Epoch 58/200, train_cost = 964.819, train_ler = 3.471, val_cost = 266.364, val_ler = 1.568, time = 2.278\n",
      "batch: 0 1635 (1, 145, 13)\n",
      "200 969.9056091308594 3.5717621445655823 1.806319236755371\n",
      "Epoch 59/200, train_cost = 969.906, train_ler = 3.572, val_cost = 571.821, val_ler = 1.848, time = 2.199\n",
      "batch: 0 1938 (1, 347, 13)\n",
      "200 940.5580596923828 3.0765601992607117 2.6508870124816895\n",
      "Epoch 60/200, train_cost = 940.558, train_ler = 3.077, val_cost = 504.719, val_ler = 1.685, time = 2.965\n",
      "batch: 0 1375 (1, 204, 13)\n",
      "200 1199.3598327636719 3.574558973312378 2.986039161682129\n",
      "Epoch 61/200, train_cost = 1199.360, train_ler = 3.575, val_cost = 420.654, val_ler = 1.695, time = 3.230\n",
      "batch: 0 593 (1, 534, 13)\n",
      "200 741.7869415283203 3.2347100973129272 2.701932430267334\n",
      "Epoch 62/200, train_cost = 741.787, train_ler = 3.235, val_cost = 457.182, val_ler = 1.568, time = 3.131\n",
      "batch: 0 3081 (1, 279, 13)\n",
      "200 738.9409484863281 3.3738521933555603 2.3051223754882812\n",
      "Epoch 63/200, train_cost = 738.941, train_ler = 3.374, val_cost = 457.403, val_ler = 1.893, time = 2.622\n",
      "batch: 0 2258 (1, 466, 13)\n",
      "200 939.4323120117188 3.117289364337921 2.7299187183380127\n",
      "Epoch 64/200, train_cost = 939.432, train_ler = 3.117, val_cost = 485.831, val_ler = 1.538, time = 3.058\n",
      "batch: 0 2342 (1, 218, 13)\n",
      "200 674.9938049316406 3.245004951953888 2.2441821098327637\n",
      "Epoch 65/200, train_cost = 674.994, train_ler = 3.245, val_cost = 313.153, val_ler = 1.723, time = 2.454\n",
      "batch: 0 2901 (1, 212, 13)\n",
      "200 678.7800064086914 3.2196942567825317 2.1725950241088867\n",
      "Epoch 66/200, train_cost = 678.780, train_ler = 3.220, val_cost = 505.483, val_ler = 1.540, time = 2.526\n",
      "batch: 0 2835 (1, 160, 13)\n",
      "200 815.9358825683594 3.2894298434257507 2.1847283840179443\n",
      "Epoch 67/200, train_cost = 815.936, train_ler = 3.289, val_cost = 340.802, val_ler = 1.686, time = 2.421\n",
      "batch: 0 2500 (1, 281, 13)\n",
      "200 827.5182342529297 3.2368515133857727 2.196826219558716\n",
      "Epoch 68/200, train_cost = 827.518, train_ler = 3.237, val_cost = 445.536, val_ler = 1.626, time = 2.558\n",
      "batch: 0 4083 (1, 295, 13)\n",
      "200 658.123161315918 3.2202380895614624 2.2529430389404297\n",
      "Epoch 69/200, train_cost = 658.123, train_ler = 3.220, val_cost = 513.703, val_ler = 1.790, time = 2.548\n",
      "batch: 0 379 (1, 364, 13)\n",
      "200 1044.4662475585938 3.2832974195480347 2.802558183670044\n",
      "Epoch 70/200, train_cost = 1044.466, train_ler = 3.283, val_cost = 283.496, val_ler = 1.576, time = 3.018\n",
      "batch: 0 4193 (1, 243, 13)\n",
      "200 743.5790557861328 3.526913106441498 2.1104180812835693\n",
      "Epoch 71/200, train_cost = 743.579, train_ler = 3.527, val_cost = 497.811, val_ler = 1.905, time = 2.411\n",
      "batch: 0 3355 (1, 355, 13)\n",
      "200 903.4981384277344 3.2154698371887207 2.514247417449951\n",
      "Epoch 72/200, train_cost = 903.498, train_ler = 3.215, val_cost = 598.466, val_ler = 1.541, time = 2.953\n",
      "batch: 0 3001 (1, 191, 13)\n",
      "200 667.9425354003906 3.1602718830108643 2.1375057697296143\n",
      "Epoch 73/200, train_cost = 667.943, train_ler = 3.160, val_cost = 418.095, val_ler = 1.868, time = 2.343\n",
      "batch: 0 1155 (1, 355, 13)\n",
      "200 845.42236328125 3.3385213017463684 2.2578306198120117\n",
      "Epoch 74/200, train_cost = 845.422, train_ler = 3.339, val_cost = 384.361, val_ler = 1.663, time = 2.504\n",
      "batch: 0 2630 (1, 459, 13)\n",
      "200 945.9177551269531 3.126822590827942 2.7894935607910156\n",
      "Epoch 75/200, train_cost = 945.918, train_ler = 3.127, val_cost = 366.787, val_ler = 1.561, time = 3.060\n",
      "batch: 0 3728 (1, 267, 13)\n",
      "200 699.4282150268555 3.296353340148926 2.1045451164245605\n",
      "Epoch 76/200, train_cost = 699.428, train_ler = 3.296, val_cost = 419.726, val_ler = 1.671, time = 2.385\n",
      "batch: 0 534 (1, 292, 13)\n",
      "200 720.2505340576172 3.2679784893989563 2.4648382663726807\n",
      "Epoch 77/200, train_cost = 720.251, train_ler = 3.268, val_cost = 505.378, val_ler = 1.645, time = 2.764\n",
      "batch: 0 4031 (1, 375, 13)\n",
      "200 1106.1537628173828 3.05026638507843 2.828481912612915\n",
      "Epoch 78/200, train_cost = 1106.154, train_ler = 3.050, val_cost = 239.685, val_ler = 1.430, time = 3.031\n",
      "batch: 0 3640 (1, 359, 13)\n",
      "200 813.9181060791016 3.557456135749817 2.633833885192871\n",
      "Epoch 79/200, train_cost = 813.918, train_ler = 3.557, val_cost = 1010.477, val_ler = 1.954, time = 2.933\n",
      "batch: 0 4302 (1, 314, 13)\n",
      "200 1339.1842803955078 3.2296698093414307 2.786325216293335\n",
      "Epoch 80/200, train_cost = 1339.184, train_ler = 3.230, val_cost = 442.955, val_ler = 1.579, time = 3.056\n",
      "batch: 0 289 (1, 419, 13)\n",
      "200 936.2734832763672 3.0649709701538086 2.7721385955810547\n",
      "Epoch 81/200, train_cost = 936.273, train_ler = 3.065, val_cost = 426.576, val_ler = 1.522, time = 3.009\n",
      "batch: 0 1389 (1, 275, 13)\n",
      "200 910.4475708007812 3.335167706012726 2.8957018852233887\n",
      "Epoch 82/200, train_cost = 910.448, train_ler = 3.335, val_cost = 642.120, val_ler = 1.920, time = 3.270\n",
      "batch: 0 461 (1, 255, 13)\n",
      "200 688.7462005615234 3.2316017150878906 2.338453769683838\n",
      "Epoch 83/200, train_cost = 688.746, train_ler = 3.232, val_cost = 234.803, val_ler = 1.682, time = 2.520\n",
      "batch: 0 2053 (1, 410, 13)\n",
      "200 972.7653961181641 3.3574166297912598 3.049185037612915\n",
      "Epoch 84/200, train_cost = 972.765, train_ler = 3.357, val_cost = 480.728, val_ler = 1.688, time = 3.421\n",
      "batch: 0 202 (1, 327, 13)\n",
      "200 602.9742813110352 3.368004262447357 2.0308997631073\n",
      "Epoch 85/200, train_cost = 602.974, train_ler = 3.368, val_cost = 387.062, val_ler = 1.886, time = 2.343\n",
      "batch: 0 3749 (1, 291, 13)\n",
      "200 754.7248992919922 3.4810500741004944 2.2825968265533447\n",
      "Epoch 86/200, train_cost = 754.725, train_ler = 3.481, val_cost = 298.018, val_ler = 1.604, time = 2.528\n",
      "batch: 0 1697 (1, 131, 13)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "200 729.6758346557617 3.4853479266166687 2.0011067390441895\n",
      "Epoch 87/200, train_cost = 729.676, train_ler = 3.485, val_cost = 471.011, val_ler = 1.735, time = 2.329\n",
      "batch: 0 1742 (1, 201, 13)\n",
      "200 645.1705474853516 3.4273809790611267 2.099644899368286\n",
      "Epoch 88/200, train_cost = 645.171, train_ler = 3.427, val_cost = 437.524, val_ler = 1.570, time = 2.371\n",
      "batch: 0 1463 (1, 304, 13)\n",
      "200 743.280517578125 3.3539811968803406 2.20461106300354\n",
      "Epoch 89/200, train_cost = 743.281, train_ler = 3.354, val_cost = 443.201, val_ler = 1.764, time = 2.450\n",
      "batch: 0 3930 (1, 297, 13)\n",
      "200 935.5697784423828 3.225474774837494 2.2508630752563477\n",
      "Epoch 90/200, train_cost = 935.570, train_ler = 3.225, val_cost = 544.854, val_ler = 1.591, time = 2.593\n",
      "batch: 0 695 (1, 247, 13)\n",
      "200 930.0602416992188 3.0405179262161255 2.393343687057495\n",
      "Epoch 91/200, train_cost = 930.060, train_ler = 3.041, val_cost = 540.643, val_ler = 1.727, time = 2.830\n",
      "batch: 0 3921 (1, 298, 13)\n",
      "200 966.4389953613281 3.290788769721985 2.8819077014923096\n",
      "Epoch 92/200, train_cost = 966.439, train_ler = 3.291, val_cost = 388.806, val_ler = 1.651, time = 3.165\n",
      "batch: 0 2669 (1, 500, 13)\n",
      "200 1063.6470642089844 3.310046911239624 3.0837619304656982\n",
      "Epoch 93/200, train_cost = 1063.647, train_ler = 3.310, val_cost = 277.252, val_ler = 1.671, time = 3.324\n",
      "batch: 0 2645 (1, 462, 13)\n",
      "200 694.5392456054688 3.399069130420685 2.6783318519592285\n",
      "Epoch 94/200, train_cost = 694.539, train_ler = 3.399, val_cost = 360.035, val_ler = 1.701, time = 2.954\n",
      "batch: 0 2026 (1, 471, 13)\n",
      "200 1001.4775543212891 3.5611740946769714 2.973294496536255\n",
      "Epoch 95/200, train_cost = 1001.478, train_ler = 3.561, val_cost = 413.813, val_ler = 1.749, time = 3.380\n",
      "batch: 0 2326 (1, 307, 13)\n",
      "200 756.4222259521484 3.5787015557289124 2.35163950920105\n",
      "Epoch 96/200, train_cost = 756.422, train_ler = 3.579, val_cost = 360.775, val_ler = 1.829, time = 2.614\n",
      "batch: 0 1582 (1, 249, 13)\n",
      "200 850.1750335693359 3.02022647857666 2.419616937637329\n",
      "Epoch 97/200, train_cost = 850.175, train_ler = 3.020, val_cost = 396.636, val_ler = 1.485, time = 2.724\n",
      "batch: 0 2688 (1, 350, 13)\n",
      "200 894.0554046630859 3.3690733909606934 2.5144100189208984\n",
      "Epoch 98/200, train_cost = 894.055, train_ler = 3.369, val_cost = 361.904, val_ler = 1.912, time = 2.757\n",
      "batch: 0 3676 (1, 432, 13)\n",
      "200 853.3892135620117 3.303358733654022 2.1800665855407715\n",
      "Epoch 99/200, train_cost = 853.389, train_ler = 3.303, val_cost = 567.217, val_ler = 1.658, time = 2.503\n",
      "batch: 0 333 (1, 345, 13)\n",
      "200 786.4797744750977 3.2028775811195374 2.1732919216156006\n",
      "Epoch 100/200, train_cost = 786.480, train_ler = 3.203, val_cost = 617.487, val_ler = 1.735, time = 2.470\n",
      "batch: 0 209 (1, 336, 13)\n",
      "200 941.3127899169922 3.491055488586426 2.6519010066986084\n",
      "Epoch 101/200, train_cost = 941.313, train_ler = 3.491, val_cost = 377.301, val_ler = 1.602, time = 2.949\n",
      "batch: 0 2682 (1, 337, 13)\n",
      "200 1070.0460052490234 2.9903395771980286 2.5265462398529053\n",
      "Epoch 102/200, train_cost = 1070.046, train_ler = 2.990, val_cost = 389.720, val_ler = 1.596, time = 2.828\n",
      "batch: 0 132 (1, 334, 13)\n",
      "200 974.0543060302734 3.351192891597748 2.3972315788269043\n",
      "Epoch 103/200, train_cost = 974.054, train_ler = 3.351, val_cost = 278.794, val_ler = 1.615, time = 2.608\n",
      "batch: 0 3529 (1, 338, 13)\n",
      "200 724.1512222290039 3.340107262134552 2.2950241565704346\n",
      "Epoch 104/200, train_cost = 724.151, train_ler = 3.340, val_cost = 372.397, val_ler = 1.631, time = 2.585\n",
      "batch: 0 1164 (1, 613, 13)\n",
      "200 851.7591094970703 3.558222234249115 2.830301523208618\n",
      "Epoch 105/200, train_cost = 851.759, train_ler = 3.558, val_cost = 381.343, val_ler = 1.723, time = 3.106\n",
      "batch: 0 1685 (1, 247, 13)\n",
      "200 925.6828918457031 3.322372019290924 2.8697803020477295\n",
      "Epoch 106/200, train_cost = 925.683, train_ler = 3.322, val_cost = 591.530, val_ler = 1.878, time = 3.170\n",
      "batch: 0 627 (1, 284, 13)\n",
      "200 1059.2347412109375 3.677083373069763 2.28330135345459\n",
      "Epoch 107/200, train_cost = 1059.235, train_ler = 3.677, val_cost = 405.811, val_ler = 1.850, time = 2.525\n",
      "batch: 0 368 (1, 356, 13)\n",
      "200 1152.138442993164 3.191713869571686 2.880213499069214\n",
      "Epoch 108/200, train_cost = 1152.138, train_ler = 3.192, val_cost = 353.470, val_ler = 1.700, time = 3.114\n",
      "batch: 0 1395 (1, 166, 13)\n",
      "200 995.7127990722656 3.4783414602279663 2.3379669189453125\n",
      "Epoch 109/200, train_cost = 995.713, train_ler = 3.478, val_cost = 379.528, val_ler = 1.652, time = 2.618\n",
      "batch: 0 804 (1, 203, 13)\n",
      "200 924.9768371582031 3.2922077775001526 2.456047773361206\n",
      "Epoch 110/200, train_cost = 924.977, train_ler = 3.292, val_cost = 391.341, val_ler = 1.725, time = 2.693\n",
      "batch: 0 1696 (1, 425, 13)\n",
      "200 747.0757598876953 3.3724400401115417 2.4239320755004883\n",
      "Epoch 111/200, train_cost = 747.076, train_ler = 3.372, val_cost = 532.565, val_ler = 1.773, time = 2.800\n",
      "batch: 0 3751 (1, 261, 13)\n",
      "200 726.4080963134766 3.476797878742218 1.904397964477539\n",
      "Epoch 112/200, train_cost = 726.408, train_ler = 3.477, val_cost = 543.551, val_ler = 1.788, time = 2.257\n",
      "batch: 0 3340 (1, 212, 13)\n",
      "200 873.4980773925781 3.6673789620399475 2.0650367736816406\n",
      "Epoch 113/200, train_cost = 873.498, train_ler = 3.667, val_cost = 312.959, val_ler = 1.689, time = 2.253\n",
      "batch: 0 3608 (1, 204, 13)\n",
      "200 974.1660842895508 2.9408661127090454 2.4330482482910156\n",
      "Epoch 114/200, train_cost = 974.166, train_ler = 2.941, val_cost = 601.539, val_ler = 1.890, time = 2.774\n",
      "batch: 0 3827 (1, 215, 13)\n",
      "200 928.4499206542969 3.583887040615082 1.8315541744232178\n",
      "Epoch 115/200, train_cost = 928.450, train_ler = 3.584, val_cost = 479.713, val_ler = 1.816, time = 2.123\n",
      "batch: 0 4419 (1, 246, 13)\n",
      "200 884.3919830322266 3.344751000404358 2.2203896045684814\n",
      "Epoch 116/200, train_cost = 884.392, train_ler = 3.345, val_cost = 419.851, val_ler = 1.556, time = 2.541\n",
      "batch: 0 1590 (1, 288, 13)\n",
      "200 912.9544982910156 3.5234691500663757 2.3008663654327393\n",
      "Epoch 117/200, train_cost = 912.954, train_ler = 3.523, val_cost = 412.310, val_ler = 1.783, time = 2.543\n",
      "batch: 0 3363 (1, 245, 13)\n",
      "200 798.2507476806641 3.3695932626724243 2.29217267036438\n",
      "Epoch 118/200, train_cost = 798.251, train_ler = 3.370, val_cost = 400.923, val_ler = 1.622, time = 2.549\n",
      "batch: 0 1115 (1, 172, 13)\n",
      "200 1000.0244903564453 3.5369293689727783 2.805795431137085\n",
      "Epoch 119/200, train_cost = 1000.024, train_ler = 3.537, val_cost = 711.429, val_ler = 1.858, time = 3.156\n",
      "batch: 0 137 (1, 269, 13)\n",
      "200 987.0102081298828 3.1967172026634216 2.9878158569335938\n",
      "Epoch 120/200, train_cost = 987.010, train_ler = 3.197, val_cost = 519.099, val_ler = 1.763, time = 3.356\n",
      "batch: 0 2952 (1, 333, 13)\n",
      "200 738.7866668701172 3.1343782544136047 2.308424711227417\n",
      "Epoch 121/200, train_cost = 738.787, train_ler = 3.134, val_cost = 349.926, val_ler = 1.672, time = 2.571\n",
      "batch: 0 1464 (1, 441, 13)\n",
      "200 842.1257858276367 3.4977386593818665 2.587502956390381\n",
      "Epoch 122/200, train_cost = 842.126, train_ler = 3.498, val_cost = 320.182, val_ler = 1.700, time = 2.832\n",
      "batch: 0 2343 (1, 409, 13)\n",
      "200 1026.6136779785156 3.645156979560852 2.763430595397949\n",
      "Epoch 123/200, train_cost = 1026.614, train_ler = 3.645, val_cost = 517.786, val_ler = 1.954, time = 3.016\n",
      "batch: 0 436 (1, 306, 13)\n",
      "200 869.878662109375 3.445665419101715 2.5717885494232178\n",
      "Epoch 124/200, train_cost = 869.879, train_ler = 3.446, val_cost = 361.300, val_ler = 1.686, time = 2.819\n",
      "batch: 0 4332 (1, 304, 13)\n",
      "200 716.5878448486328 3.3988357186317444 2.265545606613159\n",
      "Epoch 125/200, train_cost = 716.588, train_ler = 3.399, val_cost = 337.470, val_ler = 1.721, time = 2.477\n",
      "batch: 0 2337 (1, 265, 13)\n",
      "200 740.5728149414062 3.1290042996406555 1.8831090927124023\n",
      "Epoch 126/200, train_cost = 740.573, train_ler = 3.129, val_cost = 570.835, val_ler = 1.534, time = 2.280\n",
      "batch: 0 1475 (1, 230, 13)\n",
      "200 748.135986328125 3.070965886116028 2.3292853832244873\n",
      "Epoch 127/200, train_cost = 748.136, train_ler = 3.071, val_cost = 504.621, val_ler = 1.818, time = 2.702\n",
      "batch: 0 773 (1, 247, 13)\n",
      "200 753.5363159179688 3.627027690410614 2.2562410831451416\n",
      "Epoch 128/200, train_cost = 753.536, train_ler = 3.627, val_cost = 484.568, val_ler = 1.683, time = 2.562\n",
      "batch: 0 3964 (1, 239, 13)\n",
      "200 891.9786224365234 3.4786853790283203 2.8864283561706543\n",
      "Epoch 129/200, train_cost = 891.979, train_ler = 3.479, val_cost = 421.623, val_ler = 1.636, time = 3.311\n",
      "batch: 0 1471 (1, 232, 13)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "200 801.8080291748047 3.2008960247039795 2.525789260864258\n",
      "Epoch 130/200, train_cost = 801.808, train_ler = 3.201, val_cost = 316.187, val_ler = 1.591, time = 2.764\n",
      "batch: 0 2532 (1, 459, 13)\n",
      "200 922.5453186035156 3.5179367065429688 2.6285247802734375\n",
      "Epoch 131/200, train_cost = 922.545, train_ler = 3.518, val_cost = 363.887, val_ler = 1.920, time = 2.896\n",
      "batch: 0 1950 (1, 231, 13)\n",
      "200 588.6575012207031 3.4102272391319275 2.096043109893799\n",
      "Epoch 132/200, train_cost = 588.658, train_ler = 3.410, val_cost = 468.216, val_ler = 1.864, time = 2.349\n",
      "batch: 0 574 (1, 243, 13)\n",
      "200 685.1742706298828 3.607030928134918 2.2034292221069336\n",
      "Epoch 133/200, train_cost = 685.174, train_ler = 3.607, val_cost = 273.556, val_ler = 1.714, time = 2.448\n",
      "batch: 0 1372 (1, 302, 13)\n",
      "200 966.7866821289062 3.5479076504707336 2.806178092956543\n",
      "Epoch 134/200, train_cost = 966.787, train_ler = 3.548, val_cost = 513.828, val_ler = 1.811, time = 3.143\n",
      "batch: 0 1214 (1, 409, 13)\n",
      "200 857.2324829101562 3.1952959299087524 2.7939295768737793\n",
      "Epoch 135/200, train_cost = 857.232, train_ler = 3.195, val_cost = 271.334, val_ler = 1.671, time = 3.055\n",
      "batch: 0 1760 (1, 136, 13)\n",
      "200 562.1968231201172 3.376903772354126 1.7544500827789307\n",
      "Epoch 136/200, train_cost = 562.197, train_ler = 3.377, val_cost = 276.724, val_ler = 1.752, time = 1.958\n",
      "batch: 0 2054 (1, 197, 13)\n",
      "200 713.3578033447266 3.342688739299774 2.1899373531341553\n",
      "Epoch 137/200, train_cost = 713.358, train_ler = 3.343, val_cost = 432.081, val_ler = 1.664, time = 2.417\n",
      "batch: 0 3132 (1, 278, 13)\n",
      "200 584.50732421875 3.084475576877594 1.8436999320983887\n",
      "Epoch 138/200, train_cost = 584.507, train_ler = 3.084, val_cost = 343.566, val_ler = 1.556, time = 2.048\n",
      "batch: 0 3458 (1, 259, 13)\n",
      "200 690.8200988769531 3.311854839324951 2.1314454078674316\n",
      "Epoch 139/200, train_cost = 690.820, train_ler = 3.312, val_cost = 273.916, val_ler = 1.773, time = 2.351\n",
      "batch: 0 3949 (1, 346, 13)\n",
      "200 714.6661605834961 3.227605700492859 2.031221628189087\n",
      "Epoch 140/200, train_cost = 714.666, train_ler = 3.228, val_cost = 425.374, val_ler = 1.465, time = 2.331\n",
      "batch: 0 746 (1, 249, 13)\n",
      "200 584.6147766113281 3.458850920200348 1.8253581523895264\n",
      "Epoch 141/200, train_cost = 584.615, train_ler = 3.459, val_cost = 451.948, val_ler = 1.674, time = 2.179\n",
      "batch: 0 4217 (1, 427, 13)\n",
      "200 778.2845458984375 3.3865036368370056 2.1914920806884766\n",
      "Epoch 142/200, train_cost = 778.285, train_ler = 3.387, val_cost = 500.962, val_ler = 1.682, time = 2.448\n",
      "batch: 0 1458 (1, 299, 13)\n",
      "200 647.7715835571289 3.126525044441223 2.1256866455078125\n",
      "Epoch 143/200, train_cost = 647.772, train_ler = 3.127, val_cost = 392.228, val_ler = 1.660, time = 2.415\n",
      "batch: 0 1632 (1, 418, 13)\n",
      "200 771.0590209960938 3.3672494292259216 2.9484798908233643\n",
      "Epoch 144/200, train_cost = 771.059, train_ler = 3.367, val_cost = 285.605, val_ler = 1.782, time = 3.198\n",
      "batch: 0 1314 (1, 233, 13)\n",
      "200 788.1681518554688 3.1989908814430237 2.565067768096924\n",
      "Epoch 145/200, train_cost = 788.168, train_ler = 3.199, val_cost = 562.668, val_ler = 1.698, time = 2.877\n",
      "batch: 0 1687 (1, 434, 13)\n",
      "200 1020.2733154296875 3.404837131500244 2.553098440170288\n",
      "Epoch 146/200, train_cost = 1020.273, train_ler = 3.405, val_cost = 357.970, val_ler = 1.565, time = 2.781\n",
      "batch: 0 392 (1, 378, 13)\n",
      "200 731.1850891113281 3.4278473258018494 2.325397491455078\n",
      "Epoch 147/200, train_cost = 731.185, train_ler = 3.428, val_cost = 563.021, val_ler = 1.895, time = 2.768\n",
      "batch: 0 452 (1, 311, 13)\n",
      "200 663.3822784423828 3.7615384459495544 2.101315498352051\n",
      "Epoch 148/200, train_cost = 663.382, train_ler = 3.762, val_cost = 423.596, val_ler = 1.596, time = 2.436\n",
      "batch: 0 2975 (1, 219, 13)\n",
      "200 898.852783203125 3.199380040168762 2.2968060970306396\n",
      "Epoch 149/200, train_cost = 898.853, train_ler = 3.199, val_cost = 696.450, val_ler = 1.938, time = 2.640\n",
      "batch: 0 2790 (1, 181, 13)\n",
      "200 594.6305313110352 3.6661754846572876 1.6475653648376465\n",
      "Epoch 150/200, train_cost = 594.631, train_ler = 3.666, val_cost = 351.883, val_ler = 1.719, time = 1.865\n",
      "batch: 0 1900 (1, 237, 13)\n",
      "200 740.9550323486328 2.885807991027832 1.9230625629425049\n",
      "Epoch 151/200, train_cost = 740.955, train_ler = 2.886, val_cost = 376.550, val_ler = 1.590, time = 2.131\n",
      "batch: 0 2227 (1, 443, 13)\n",
      "200 750.8201141357422 3.616071403026581 2.2482361793518066\n",
      "Epoch 152/200, train_cost = 750.820, train_ler = 3.616, val_cost = 497.062, val_ler = 1.923, time = 2.505\n",
      "batch: 0 4207 (1, 192, 13)\n",
      "200 990.8834838867188 3.415183126926422 2.044846534729004\n",
      "Epoch 153/200, train_cost = 990.883, train_ler = 3.415, val_cost = 464.561, val_ler = 1.670, time = 2.303\n",
      "batch: 0 1980 (1, 357, 13)\n",
      "200 1150.2170715332031 3.4874051213264465 2.700972557067871\n",
      "Epoch 154/200, train_cost = 1150.217, train_ler = 3.487, val_cost = 756.938, val_ler = 1.837, time = 3.103\n",
      "batch: 0 2132 (1, 261, 13)\n",
      "200 922.4645690917969 3.3538646697998047 1.912257194519043\n",
      "Epoch 155/200, train_cost = 922.465, train_ler = 3.354, val_cost = 485.331, val_ler = 1.475, time = 2.166\n",
      "batch: 0 1558 (1, 195, 13)\n",
      "200 834.4515686035156 2.9820067286491394 2.247253179550171\n",
      "Epoch 156/200, train_cost = 834.452, train_ler = 2.982, val_cost = 613.843, val_ler = 1.709, time = 2.547\n",
      "batch: 0 1835 (1, 151, 13)\n",
      "200 1071.4449310302734 3.1325510144233704 2.250062942504883\n",
      "Epoch 157/200, train_cost = 1071.445, train_ler = 3.133, val_cost = 393.315, val_ler = 1.490, time = 2.589\n",
      "batch: 0 1828 (1, 224, 13)\n",
      "200 1010.7692260742188 3.1673588156700134 2.098546028137207\n",
      "Epoch 158/200, train_cost = 1010.769, train_ler = 3.167, val_cost = 392.341, val_ler = 1.583, time = 2.342\n",
      "batch: 0 1035 (1, 228, 13)\n",
      "200 1021.5614929199219 3.065264940261841 2.26025652885437\n",
      "Epoch 159/200, train_cost = 1021.561, train_ler = 3.065, val_cost = 594.535, val_ler = 1.560, time = 2.616\n",
      "batch: 0 4085 (1, 264, 13)\n",
      "200 1048.1328735351562 3.193244755268097 2.3179585933685303\n",
      "Epoch 160/200, train_cost = 1048.133, train_ler = 3.193, val_cost = 350.425, val_ler = 1.465, time = 2.607\n",
      "batch: 0 1398 (1, 220, 13)\n",
      "200 749.2127838134766 3.6969268321990967 2.244776725769043\n",
      "Epoch 161/200, train_cost = 749.213, train_ler = 3.697, val_cost = 385.157, val_ler = 1.794, time = 2.483\n",
      "batch: 0 949 (1, 394, 13)\n",
      "200 958.7665405273438 3.167336583137512 2.009167194366455\n",
      "Epoch 162/200, train_cost = 958.767, train_ler = 3.167, val_cost = 346.261, val_ler = 1.652, time = 2.231\n",
      "batch: 0 470 (1, 279, 13)\n",
      "200 650.2854537963867 3.6956464648246765 2.0851898193359375\n",
      "Epoch 163/200, train_cost = 650.285, train_ler = 3.696, val_cost = 374.078, val_ler = 1.824, time = 2.296\n",
      "batch: 0 2170 (1, 289, 13)\n",
      "200 736.5242614746094 3.2656503915786743 1.966540813446045\n",
      "Epoch 164/200, train_cost = 736.524, train_ler = 3.266, val_cost = 600.780, val_ler = 2.696, time = 2.196\n",
      "batch: 0 839 (1, 275, 13)\n",
      "200 865.9882507324219 3.1231002807617188 1.9387760162353516\n",
      "Epoch 165/200, train_cost = 865.988, train_ler = 3.123, val_cost = 310.368, val_ler = 1.638, time = 2.173\n",
      "batch: 0 1964 (1, 507, 13)\n",
      "200 1158.8372344970703 3.2499294877052307 3.4242405891418457\n",
      "Epoch 166/200, train_cost = 1158.837, train_ler = 3.250, val_cost = 567.763, val_ler = 1.670, time = 3.705\n",
      "batch: 0 2606 (1, 276, 13)\n",
      "200 965.5323944091797 2.8581807017326355 2.700395107269287\n",
      "Epoch 167/200, train_cost = 965.532, train_ler = 2.858, val_cost = 528.100, val_ler = 1.606, time = 2.937\n",
      "batch: 0 389 (1, 340, 13)\n",
      "200 1235.243637084961 3.3907647132873535 2.5480504035949707\n",
      "Epoch 168/200, train_cost = 1235.244, train_ler = 3.391, val_cost = 382.657, val_ler = 1.445, time = 2.822\n",
      "batch: 0 3060 (1, 359, 13)\n",
      "200 938.7341003417969 3.340705096721649 2.155292272567749\n",
      "Epoch 169/200, train_cost = 938.734, train_ler = 3.341, val_cost = 635.897, val_ler = 1.714, time = 2.474\n",
      "batch: 0 2255 (1, 292, 13)\n",
      "200 1003.9232788085938 3.0251117944717407 2.368800640106201\n",
      "Epoch 170/200, train_cost = 1003.923, train_ler = 3.025, val_cost = 479.430, val_ler = 1.440, time = 2.654\n",
      "batch: 0 3303 (1, 319, 13)\n",
      "200 1530.4370727539062 3.531842827796936 2.9453048706054688\n",
      "Epoch 171/200, train_cost = 1530.437, train_ler = 3.532, val_cost = 473.550, val_ler = 1.528, time = 3.192\n",
      "batch: 0 2384 (1, 281, 13)\n",
      "200 1251.3586883544922 3.496545374393463 2.2771620750427246\n",
      "Epoch 172/200, train_cost = 1251.359, train_ler = 3.497, val_cost = 503.707, val_ler = 1.909, time = 2.578\n",
      "batch: 0 3347 (1, 255, 13)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "200 1010.2695617675781 3.1062905192375183 2.722501039505005\n",
      "Epoch 173/200, train_cost = 1010.270, train_ler = 3.106, val_cost = 543.491, val_ler = 1.885, time = 2.989\n",
      "batch: 0 1483 (1, 352, 13)\n",
      "200 636.3333740234375 3.287994861602783 1.8621227741241455\n",
      "Epoch 174/200, train_cost = 636.333, train_ler = 3.288, val_cost = 401.065, val_ler = 1.712, time = 2.142\n",
      "batch: 0 599 (1, 223, 13)\n",
      "200 838.6638946533203 3.356306731700897 2.0338690280914307\n",
      "Epoch 175/200, train_cost = 838.664, train_ler = 3.356, val_cost = 631.428, val_ler = 1.596, time = 2.370\n",
      "batch: 0 2616 (1, 389, 13)\n",
      "200 921.0880279541016 3.156750500202179 2.421438217163086\n",
      "Epoch 176/200, train_cost = 921.088, train_ler = 3.157, val_cost = 542.675, val_ler = 1.508, time = 2.802\n",
      "batch: 0 2492 (1, 426, 13)\n",
      "200 875.5821151733398 3.114104688167572 2.6204404830932617\n",
      "Epoch 177/200, train_cost = 875.582, train_ler = 3.114, val_cost = 508.982, val_ler = 1.881, time = 2.884\n",
      "batch: 0 2722 (1, 301, 13)\n",
      "200 980.7071533203125 3.512890875339508 2.246704578399658\n",
      "Epoch 178/200, train_cost = 980.707, train_ler = 3.513, val_cost = 362.832, val_ler = 1.531, time = 2.503\n",
      "batch: 0 386 (1, 306, 13)\n",
      "200 922.7407531738281 3.1892377734184265 2.599810838699341\n",
      "Epoch 179/200, train_cost = 922.741, train_ler = 3.189, val_cost = 533.903, val_ler = 1.776, time = 2.955\n",
      "batch: 0 2200 (1, 332, 13)\n",
      "200 841.7040710449219 3.381267786026001 2.607173204421997\n",
      "Epoch 180/200, train_cost = 841.704, train_ler = 3.381, val_cost = 355.370, val_ler = 1.686, time = 2.863\n",
      "batch: 0 4253 (1, 203, 13)\n",
      "200 748.3225402832031 3.173994719982147 1.9438176155090332\n",
      "Epoch 181/200, train_cost = 748.323, train_ler = 3.174, val_cost = 284.212, val_ler = 1.575, time = 2.149\n",
      "batch: 0 1710 (1, 243, 13)\n",
      "200 842.822380065918 3.070037305355072 2.6748647689819336\n",
      "Epoch 182/200, train_cost = 842.822, train_ler = 3.070, val_cost = 375.235, val_ler = 1.624, time = 2.926\n",
      "batch: 0 2085 (1, 539, 13)\n",
      "200 892.1311645507812 3.454465329647064 3.0106396675109863\n",
      "Epoch 183/200, train_cost = 892.131, train_ler = 3.454, val_cost = 375.088, val_ler = 1.690, time = 3.243\n",
      "batch: 0 1839 (1, 512, 13)\n",
      "200 947.8085021972656 3.3238232135772705 3.069230318069458\n",
      "Epoch 184/200, train_cost = 947.809, train_ler = 3.324, val_cost = 427.144, val_ler = 1.504, time = 3.382\n",
      "batch: 0 166 (1, 354, 13)\n",
      "200 887.3853302001953 3.174908459186554 2.7723817825317383\n",
      "Epoch 185/200, train_cost = 887.385, train_ler = 3.175, val_cost = 336.664, val_ler = 1.766, time = 2.993\n",
      "batch: 0 1636 (1, 148, 13)\n",
      "200 741.0855407714844 3.6010481119155884 1.968149185180664\n",
      "Epoch 186/200, train_cost = 741.086, train_ler = 3.601, val_cost = 410.707, val_ler = 1.541, time = 2.211\n",
      "batch: 0 1330 (1, 357, 13)\n",
      "200 885.0859222412109 3.36457622051239 2.9699316024780273\n",
      "Epoch 187/200, train_cost = 885.086, train_ler = 3.365, val_cost = 255.942, val_ler = 1.619, time = 3.164\n",
      "batch: 0 441 (1, 344, 13)\n",
      "200 777.0547943115234 3.466794013977051 2.95223331451416\n",
      "Epoch 188/200, train_cost = 777.055, train_ler = 3.467, val_cost = 357.994, val_ler = 1.856, time = 3.199\n",
      "batch: 0 3257 (1, 265, 13)\n",
      "200 635.6185455322266 3.333333373069763 2.1729824542999268\n",
      "Epoch 189/200, train_cost = 635.619, train_ler = 3.333, val_cost = 418.920, val_ler = 1.673, time = 2.505\n",
      "batch: 0 2870 (1, 247, 13)\n",
      "200 599.5002365112305 3.450206458568573 2.0631136894226074\n",
      "Epoch 190/200, train_cost = 599.500, train_ler = 3.450, val_cost = 373.980, val_ler = 1.814, time = 2.350\n",
      "batch: 0 2238 (1, 284, 13)\n",
      "200 709.2025909423828 3.399350643157959 2.19909930229187\n",
      "Epoch 191/200, train_cost = 709.203, train_ler = 3.399, val_cost = 373.581, val_ler = 1.754, time = 2.448\n",
      "batch: 0 4291 (1, 265, 13)\n",
      "200 918.9464416503906 3.168922483921051 2.3982369899749756\n",
      "Epoch 192/200, train_cost = 918.946, train_ler = 3.169, val_cost = 318.354, val_ler = 1.530, time = 2.696\n",
      "batch: 0 282 (1, 330, 13)\n",
      "200 899.7098541259766 3.6156409978866577 3.1792471408843994\n",
      "Epoch 193/200, train_cost = 899.710, train_ler = 3.616, val_cost = 430.054, val_ler = 1.780, time = 3.390\n",
      "batch: 0 4222 (1, 462, 13)\n",
      "200 853.8982696533203 3.368207275867462 2.634108304977417\n",
      "Epoch 194/200, train_cost = 853.898, train_ler = 3.368, val_cost = 359.875, val_ler = 1.527, time = 2.929\n",
      "batch: 0 3128 (1, 242, 13)\n",
      "200 914.9635925292969 3.046634614467621 2.249533176422119\n",
      "Epoch 195/200, train_cost = 914.964, train_ler = 3.047, val_cost = 335.529, val_ler = 1.580, time = 2.488\n",
      "batch: 0 2058 (1, 306, 13)\n",
      "200 1006.0300903320312 3.3742964267730713 2.714378595352173\n",
      "Epoch 196/200, train_cost = 1006.030, train_ler = 3.374, val_cost = 443.330, val_ler = 1.630, time = 2.941\n",
      "batch: 0 483 (1, 376, 13)\n",
      "200 888.9201354980469 3.2829847931861877 2.3818418979644775\n",
      "Epoch 197/200, train_cost = 888.920, train_ler = 3.283, val_cost = 405.142, val_ler = 1.561, time = 2.628\n",
      "batch: 0 3993 (1, 412, 13)\n",
      "200 814.172119140625 3.057568907737732 2.659984588623047\n",
      "Epoch 198/200, train_cost = 814.172, train_ler = 3.058, val_cost = 307.846, val_ler = 1.669, time = 2.887\n",
      "batch: 0 3913 (1, 277, 13)\n",
      "200 907.0110015869141 3.382103741168976 2.0771677494049072\n",
      "Epoch 199/200, train_cost = 907.011, train_ler = 3.382, val_cost = 564.683, val_ler = 1.844, time = 2.404\n",
      "batch: 0 2797 (1, 264, 13)\n",
      "200 712.0077667236328 3.2328500151634216 2.306542158126831\n",
      "Epoch 200/200, train_cost = 712.008, train_ler = 3.233, val_cost = 394.546, val_ler = 1.694, time = 2.638\n",
      "Original: ['she', '', 'had', '', 'your', '', 'dark', '', 'suit', '', 'in', '', 'greasy', '', 'wash', '', 'water', '', 'all', '', 'year']\n",
      "Decoded: e  \n"
     ]
    },
    {
     "data": {
      "application/javascript": [
       "/* Put everything inside the global mpl namespace */\n",
       "window.mpl = {};\n",
       "\n",
       "\n",
       "mpl.get_websocket_type = function() {\n",
       "    if (typeof(WebSocket) !== 'undefined') {\n",
       "        return WebSocket;\n",
       "    } else if (typeof(MozWebSocket) !== 'undefined') {\n",
       "        return MozWebSocket;\n",
       "    } else {\n",
       "        alert('Your browser does not have WebSocket support.' +\n",
       "              'Please try Chrome, Safari or Firefox ≥ 6. ' +\n",
       "              'Firefox 4 and 5 are also supported but you ' +\n",
       "              'have to enable WebSockets in about:config.');\n",
       "    };\n",
       "}\n",
       "\n",
       "mpl.figure = function(figure_id, websocket, ondownload, parent_element) {\n",
       "    this.id = figure_id;\n",
       "\n",
       "    this.ws = websocket;\n",
       "\n",
       "    this.supports_binary = (this.ws.binaryType != undefined);\n",
       "\n",
       "    if (!this.supports_binary) {\n",
       "        var warnings = document.getElementById(\"mpl-warnings\");\n",
       "        if (warnings) {\n",
       "            warnings.style.display = 'block';\n",
       "            warnings.textContent = (\n",
       "                \"This browser does not support binary websocket messages. \" +\n",
       "                    \"Performance may be slow.\");\n",
       "        }\n",
       "    }\n",
       "\n",
       "    this.imageObj = new Image();\n",
       "\n",
       "    this.context = undefined;\n",
       "    this.message = undefined;\n",
       "    this.canvas = undefined;\n",
       "    this.rubberband_canvas = undefined;\n",
       "    this.rubberband_context = undefined;\n",
       "    this.format_dropdown = undefined;\n",
       "\n",
       "    this.image_mode = 'full';\n",
       "\n",
       "    this.root = $('<div/>');\n",
       "    this._root_extra_style(this.root)\n",
       "    this.root.attr('style', 'display: inline-block');\n",
       "\n",
       "    $(parent_element).append(this.root);\n",
       "\n",
       "    this._init_header(this);\n",
       "    this._init_canvas(this);\n",
       "    this._init_toolbar(this);\n",
       "\n",
       "    var fig = this;\n",
       "\n",
       "    this.waiting = false;\n",
       "\n",
       "    this.ws.onopen =  function () {\n",
       "            fig.send_message(\"supports_binary\", {value: fig.supports_binary});\n",
       "            fig.send_message(\"send_image_mode\", {});\n",
       "            if (mpl.ratio != 1) {\n",
       "                fig.send_message(\"set_dpi_ratio\", {'dpi_ratio': mpl.ratio});\n",
       "            }\n",
       "            fig.send_message(\"refresh\", {});\n",
       "        }\n",
       "\n",
       "    this.imageObj.onload = function() {\n",
       "            if (fig.image_mode == 'full') {\n",
       "                // Full images could contain transparency (where diff images\n",
       "                // almost always do), so we need to clear the canvas so that\n",
       "                // there is no ghosting.\n",
       "                fig.context.clearRect(0, 0, fig.canvas.width, fig.canvas.height);\n",
       "            }\n",
       "            fig.context.drawImage(fig.imageObj, 0, 0);\n",
       "        };\n",
       "\n",
       "    this.imageObj.onunload = function() {\n",
       "        fig.ws.close();\n",
       "    }\n",
       "\n",
       "    this.ws.onmessage = this._make_on_message_function(this);\n",
       "\n",
       "    this.ondownload = ondownload;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_header = function() {\n",
       "    var titlebar = $(\n",
       "        '<div class=\"ui-dialog-titlebar ui-widget-header ui-corner-all ' +\n",
       "        'ui-helper-clearfix\"/>');\n",
       "    var titletext = $(\n",
       "        '<div class=\"ui-dialog-title\" style=\"width: 100%; ' +\n",
       "        'text-align: center; padding: 3px;\"/>');\n",
       "    titlebar.append(titletext)\n",
       "    this.root.append(titlebar);\n",
       "    this.header = titletext[0];\n",
       "}\n",
       "\n",
       "\n",
       "\n",
       "mpl.figure.prototype._canvas_extra_style = function(canvas_div) {\n",
       "\n",
       "}\n",
       "\n",
       "\n",
       "mpl.figure.prototype._root_extra_style = function(canvas_div) {\n",
       "\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_canvas = function() {\n",
       "    var fig = this;\n",
       "\n",
       "    var canvas_div = $('<div/>');\n",
       "\n",
       "    canvas_div.attr('style', 'position: relative; clear: both; outline: 0');\n",
       "\n",
       "    function canvas_keyboard_event(event) {\n",
       "        return fig.key_event(event, event['data']);\n",
       "    }\n",
       "\n",
       "    canvas_div.keydown('key_press', canvas_keyboard_event);\n",
       "    canvas_div.keyup('key_release', canvas_keyboard_event);\n",
       "    this.canvas_div = canvas_div\n",
       "    this._canvas_extra_style(canvas_div)\n",
       "    this.root.append(canvas_div);\n",
       "\n",
       "    var canvas = $('<canvas/>');\n",
       "    canvas.addClass('mpl-canvas');\n",
       "    canvas.attr('style', \"left: 0; top: 0; z-index: 0; outline: 0\")\n",
       "\n",
       "    this.canvas = canvas[0];\n",
       "    this.context = canvas[0].getContext(\"2d\");\n",
       "\n",
       "    var backingStore = this.context.backingStorePixelRatio ||\n",
       "\tthis.context.webkitBackingStorePixelRatio ||\n",
       "\tthis.context.mozBackingStorePixelRatio ||\n",
       "\tthis.context.msBackingStorePixelRatio ||\n",
       "\tthis.context.oBackingStorePixelRatio ||\n",
       "\tthis.context.backingStorePixelRatio || 1;\n",
       "\n",
       "    mpl.ratio = (window.devicePixelRatio || 1) / backingStore;\n",
       "\n",
       "    var rubberband = $('<canvas/>');\n",
       "    rubberband.attr('style', \"position: absolute; left: 0; top: 0; z-index: 1;\")\n",
       "\n",
       "    var pass_mouse_events = true;\n",
       "\n",
       "    canvas_div.resizable({\n",
       "        start: function(event, ui) {\n",
       "            pass_mouse_events = false;\n",
       "        },\n",
       "        resize: function(event, ui) {\n",
       "            fig.request_resize(ui.size.width, ui.size.height);\n",
       "        },\n",
       "        stop: function(event, ui) {\n",
       "            pass_mouse_events = true;\n",
       "            fig.request_resize(ui.size.width, ui.size.height);\n",
       "        },\n",
       "    });\n",
       "\n",
       "    function mouse_event_fn(event) {\n",
       "        if (pass_mouse_events)\n",
       "            return fig.mouse_event(event, event['data']);\n",
       "    }\n",
       "\n",
       "    rubberband.mousedown('button_press', mouse_event_fn);\n",
       "    rubberband.mouseup('button_release', mouse_event_fn);\n",
       "    // Throttle sequential mouse events to 1 every 20ms.\n",
       "    rubberband.mousemove('motion_notify', mouse_event_fn);\n",
       "\n",
       "    rubberband.mouseenter('figure_enter', mouse_event_fn);\n",
       "    rubberband.mouseleave('figure_leave', mouse_event_fn);\n",
       "\n",
       "    canvas_div.on(\"wheel\", function (event) {\n",
       "        event = event.originalEvent;\n",
       "        event['data'] = 'scroll'\n",
       "        if (event.deltaY < 0) {\n",
       "            event.step = 1;\n",
       "        } else {\n",
       "            event.step = -1;\n",
       "        }\n",
       "        mouse_event_fn(event);\n",
       "    });\n",
       "\n",
       "    canvas_div.append(canvas);\n",
       "    canvas_div.append(rubberband);\n",
       "\n",
       "    this.rubberband = rubberband;\n",
       "    this.rubberband_canvas = rubberband[0];\n",
       "    this.rubberband_context = rubberband[0].getContext(\"2d\");\n",
       "    this.rubberband_context.strokeStyle = \"#000000\";\n",
       "\n",
       "    this._resize_canvas = function(width, height) {\n",
       "        // Keep the size of the canvas, canvas container, and rubber band\n",
       "        // canvas in synch.\n",
       "        canvas_div.css('width', width)\n",
       "        canvas_div.css('height', height)\n",
       "\n",
       "        canvas.attr('width', width * mpl.ratio);\n",
       "        canvas.attr('height', height * mpl.ratio);\n",
       "        canvas.attr('style', 'width: ' + width + 'px; height: ' + height + 'px;');\n",
       "\n",
       "        rubberband.attr('width', width);\n",
       "        rubberband.attr('height', height);\n",
       "    }\n",
       "\n",
       "    // Set the figure to an initial 600x600px, this will subsequently be updated\n",
       "    // upon first draw.\n",
       "    this._resize_canvas(600, 600);\n",
       "\n",
       "    // Disable right mouse context menu.\n",
       "    $(this.rubberband_canvas).bind(\"contextmenu\",function(e){\n",
       "        return false;\n",
       "    });\n",
       "\n",
       "    function set_focus () {\n",
       "        canvas.focus();\n",
       "        canvas_div.focus();\n",
       "    }\n",
       "\n",
       "    window.setTimeout(set_focus, 100);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_toolbar = function() {\n",
       "    var fig = this;\n",
       "\n",
       "    var nav_element = $('<div/>')\n",
       "    nav_element.attr('style', 'width: 100%');\n",
       "    this.root.append(nav_element);\n",
       "\n",
       "    // Define a callback function for later on.\n",
       "    function toolbar_event(event) {\n",
       "        return fig.toolbar_button_onclick(event['data']);\n",
       "    }\n",
       "    function toolbar_mouse_event(event) {\n",
       "        return fig.toolbar_button_onmouseover(event['data']);\n",
       "    }\n",
       "\n",
       "    for(var toolbar_ind in mpl.toolbar_items) {\n",
       "        var name = mpl.toolbar_items[toolbar_ind][0];\n",
       "        var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
       "        var image = mpl.toolbar_items[toolbar_ind][2];\n",
       "        var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
       "\n",
       "        if (!name) {\n",
       "            // put a spacer in here.\n",
       "            continue;\n",
       "        }\n",
       "        var button = $('<button/>');\n",
       "        button.addClass('ui-button ui-widget ui-state-default ui-corner-all ' +\n",
       "                        'ui-button-icon-only');\n",
       "        button.attr('role', 'button');\n",
       "        button.attr('aria-disabled', 'false');\n",
       "        button.click(method_name, toolbar_event);\n",
       "        button.mouseover(tooltip, toolbar_mouse_event);\n",
       "\n",
       "        var icon_img = $('<span/>');\n",
       "        icon_img.addClass('ui-button-icon-primary ui-icon');\n",
       "        icon_img.addClass(image);\n",
       "        icon_img.addClass('ui-corner-all');\n",
       "\n",
       "        var tooltip_span = $('<span/>');\n",
       "        tooltip_span.addClass('ui-button-text');\n",
       "        tooltip_span.html(tooltip);\n",
       "\n",
       "        button.append(icon_img);\n",
       "        button.append(tooltip_span);\n",
       "\n",
       "        nav_element.append(button);\n",
       "    }\n",
       "\n",
       "    var fmt_picker_span = $('<span/>');\n",
       "\n",
       "    var fmt_picker = $('<select/>');\n",
       "    fmt_picker.addClass('mpl-toolbar-option ui-widget ui-widget-content');\n",
       "    fmt_picker_span.append(fmt_picker);\n",
       "    nav_element.append(fmt_picker_span);\n",
       "    this.format_dropdown = fmt_picker[0];\n",
       "\n",
       "    for (var ind in mpl.extensions) {\n",
       "        var fmt = mpl.extensions[ind];\n",
       "        var option = $(\n",
       "            '<option/>', {selected: fmt === mpl.default_extension}).html(fmt);\n",
       "        fmt_picker.append(option)\n",
       "    }\n",
       "\n",
       "    // Add hover states to the ui-buttons\n",
       "    $( \".ui-button\" ).hover(\n",
       "        function() { $(this).addClass(\"ui-state-hover\");},\n",
       "        function() { $(this).removeClass(\"ui-state-hover\");}\n",
       "    );\n",
       "\n",
       "    var status_bar = $('<span class=\"mpl-message\"/>');\n",
       "    nav_element.append(status_bar);\n",
       "    this.message = status_bar[0];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.request_resize = function(x_pixels, y_pixels) {\n",
       "    // Request matplotlib to resize the figure. Matplotlib will then trigger a resize in the client,\n",
       "    // which will in turn request a refresh of the image.\n",
       "    this.send_message('resize', {'width': x_pixels, 'height': y_pixels});\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.send_message = function(type, properties) {\n",
       "    properties['type'] = type;\n",
       "    properties['figure_id'] = this.id;\n",
       "    this.ws.send(JSON.stringify(properties));\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.send_draw_message = function() {\n",
       "    if (!this.waiting) {\n",
       "        this.waiting = true;\n",
       "        this.ws.send(JSON.stringify({type: \"draw\", figure_id: this.id}));\n",
       "    }\n",
       "}\n",
       "\n",
       "\n",
       "mpl.figure.prototype.handle_save = function(fig, msg) {\n",
       "    var format_dropdown = fig.format_dropdown;\n",
       "    var format = format_dropdown.options[format_dropdown.selectedIndex].value;\n",
       "    fig.ondownload(fig, format);\n",
       "}\n",
       "\n",
       "\n",
       "mpl.figure.prototype.handle_resize = function(fig, msg) {\n",
       "    var size = msg['size'];\n",
       "    if (size[0] != fig.canvas.width || size[1] != fig.canvas.height) {\n",
       "        fig._resize_canvas(size[0], size[1]);\n",
       "        fig.send_message(\"refresh\", {});\n",
       "    };\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_rubberband = function(fig, msg) {\n",
       "    var x0 = msg['x0'] / mpl.ratio;\n",
       "    var y0 = (fig.canvas.height - msg['y0']) / mpl.ratio;\n",
       "    var x1 = msg['x1'] / mpl.ratio;\n",
       "    var y1 = (fig.canvas.height - msg['y1']) / mpl.ratio;\n",
       "    x0 = Math.floor(x0) + 0.5;\n",
       "    y0 = Math.floor(y0) + 0.5;\n",
       "    x1 = Math.floor(x1) + 0.5;\n",
       "    y1 = Math.floor(y1) + 0.5;\n",
       "    var min_x = Math.min(x0, x1);\n",
       "    var min_y = Math.min(y0, y1);\n",
       "    var width = Math.abs(x1 - x0);\n",
       "    var height = Math.abs(y1 - y0);\n",
       "\n",
       "    fig.rubberband_context.clearRect(\n",
       "        0, 0, fig.canvas.width, fig.canvas.height);\n",
       "\n",
       "    fig.rubberband_context.strokeRect(min_x, min_y, width, height);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_figure_label = function(fig, msg) {\n",
       "    // Updates the figure title.\n",
       "    fig.header.textContent = msg['label'];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_cursor = function(fig, msg) {\n",
       "    var cursor = msg['cursor'];\n",
       "    switch(cursor)\n",
       "    {\n",
       "    case 0:\n",
       "        cursor = 'pointer';\n",
       "        break;\n",
       "    case 1:\n",
       "        cursor = 'default';\n",
       "        break;\n",
       "    case 2:\n",
       "        cursor = 'crosshair';\n",
       "        break;\n",
       "    case 3:\n",
       "        cursor = 'move';\n",
       "        break;\n",
       "    }\n",
       "    fig.rubberband_canvas.style.cursor = cursor;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_message = function(fig, msg) {\n",
       "    fig.message.textContent = msg['message'];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_draw = function(fig, msg) {\n",
       "    // Request the server to send over a new figure.\n",
       "    fig.send_draw_message();\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_image_mode = function(fig, msg) {\n",
       "    fig.image_mode = msg['mode'];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.updated_canvas_event = function() {\n",
       "    // Called whenever the canvas gets updated.\n",
       "    this.send_message(\"ack\", {});\n",
       "}\n",
       "\n",
       "// A function to construct a web socket function for onmessage handling.\n",
       "// Called in the figure constructor.\n",
       "mpl.figure.prototype._make_on_message_function = function(fig) {\n",
       "    return function socket_on_message(evt) {\n",
       "        if (evt.data instanceof Blob) {\n",
       "            /* FIXME: We get \"Resource interpreted as Image but\n",
       "             * transferred with MIME type text/plain:\" errors on\n",
       "             * Chrome.  But how to set the MIME type?  It doesn't seem\n",
       "             * to be part of the websocket stream */\n",
       "            evt.data.type = \"image/png\";\n",
       "\n",
       "            /* Free the memory for the previous frames */\n",
       "            if (fig.imageObj.src) {\n",
       "                (window.URL || window.webkitURL).revokeObjectURL(\n",
       "                    fig.imageObj.src);\n",
       "            }\n",
       "\n",
       "            fig.imageObj.src = (window.URL || window.webkitURL).createObjectURL(\n",
       "                evt.data);\n",
       "            fig.updated_canvas_event();\n",
       "            fig.waiting = false;\n",
       "            return;\n",
       "        }\n",
       "        else if (typeof evt.data === 'string' && evt.data.slice(0, 21) == \"data:image/png;base64\") {\n",
       "            fig.imageObj.src = evt.data;\n",
       "            fig.updated_canvas_event();\n",
       "            fig.waiting = false;\n",
       "            return;\n",
       "        }\n",
       "\n",
       "        var msg = JSON.parse(evt.data);\n",
       "        var msg_type = msg['type'];\n",
       "\n",
       "        // Call the  \"handle_{type}\" callback, which takes\n",
       "        // the figure and JSON message as its only arguments.\n",
       "        try {\n",
       "            var callback = fig[\"handle_\" + msg_type];\n",
       "        } catch (e) {\n",
       "            console.log(\"No handler for the '\" + msg_type + \"' message type: \", msg);\n",
       "            return;\n",
       "        }\n",
       "\n",
       "        if (callback) {\n",
       "            try {\n",
       "                // console.log(\"Handling '\" + msg_type + \"' message: \", msg);\n",
       "                callback(fig, msg);\n",
       "            } catch (e) {\n",
       "                console.log(\"Exception inside the 'handler_\" + msg_type + \"' callback:\", e, e.stack, msg);\n",
       "            }\n",
       "        }\n",
       "    };\n",
       "}\n",
       "\n",
       "// from http://stackoverflow.com/questions/1114465/getting-mouse-location-in-canvas\n",
       "mpl.findpos = function(e) {\n",
       "    //this section is from http://www.quirksmode.org/js/events_properties.html\n",
       "    var targ;\n",
       "    if (!e)\n",
       "        e = window.event;\n",
       "    if (e.target)\n",
       "        targ = e.target;\n",
       "    else if (e.srcElement)\n",
       "        targ = e.srcElement;\n",
       "    if (targ.nodeType == 3) // defeat Safari bug\n",
       "        targ = targ.parentNode;\n",
       "\n",
       "    // jQuery normalizes the pageX and pageY\n",
       "    // pageX,Y are the mouse positions relative to the document\n",
       "    // offset() returns the position of the element relative to the document\n",
       "    var x = e.pageX - $(targ).offset().left;\n",
       "    var y = e.pageY - $(targ).offset().top;\n",
       "\n",
       "    return {\"x\": x, \"y\": y};\n",
       "};\n",
       "\n",
       "/*\n",
       " * return a copy of an object with only non-object keys\n",
       " * we need this to avoid circular references\n",
       " * http://stackoverflow.com/a/24161582/3208463\n",
       " */\n",
       "function simpleKeys (original) {\n",
       "  return Object.keys(original).reduce(function (obj, key) {\n",
       "    if (typeof original[key] !== 'object')\n",
       "        obj[key] = original[key]\n",
       "    return obj;\n",
       "  }, {});\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.mouse_event = function(event, name) {\n",
       "    var canvas_pos = mpl.findpos(event)\n",
       "\n",
       "    if (name === 'button_press')\n",
       "    {\n",
       "        this.canvas.focus();\n",
       "        this.canvas_div.focus();\n",
       "    }\n",
       "\n",
       "    var x = canvas_pos.x * mpl.ratio;\n",
       "    var y = canvas_pos.y * mpl.ratio;\n",
       "\n",
       "    this.send_message(name, {x: x, y: y, button: event.button,\n",
       "                             step: event.step,\n",
       "                             guiEvent: simpleKeys(event)});\n",
       "\n",
       "    /* This prevents the web browser from automatically changing to\n",
       "     * the text insertion cursor when the button is pressed.  We want\n",
       "     * to control all of the cursor setting manually through the\n",
       "     * 'cursor' event from matplotlib */\n",
       "    event.preventDefault();\n",
       "    return false;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._key_event_extra = function(event, name) {\n",
       "    // Handle any extra behaviour associated with a key event\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.key_event = function(event, name) {\n",
       "\n",
       "    // Prevent repeat events\n",
       "    if (name == 'key_press')\n",
       "    {\n",
       "        if (event.which === this._key)\n",
       "            return;\n",
       "        else\n",
       "            this._key = event.which;\n",
       "    }\n",
       "    if (name == 'key_release')\n",
       "        this._key = null;\n",
       "\n",
       "    var value = '';\n",
       "    if (event.ctrlKey && event.which != 17)\n",
       "        value += \"ctrl+\";\n",
       "    if (event.altKey && event.which != 18)\n",
       "        value += \"alt+\";\n",
       "    if (event.shiftKey && event.which != 16)\n",
       "        value += \"shift+\";\n",
       "\n",
       "    value += 'k';\n",
       "    value += event.which.toString();\n",
       "\n",
       "    this._key_event_extra(event, name);\n",
       "\n",
       "    this.send_message(name, {key: value,\n",
       "                             guiEvent: simpleKeys(event)});\n",
       "    return false;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.toolbar_button_onclick = function(name) {\n",
       "    if (name == 'download') {\n",
       "        this.handle_save(this, null);\n",
       "    } else {\n",
       "        this.send_message(\"toolbar_button\", {name: name});\n",
       "    }\n",
       "};\n",
       "\n",
       "mpl.figure.prototype.toolbar_button_onmouseover = function(tooltip) {\n",
       "    this.message.textContent = tooltip;\n",
       "};\n",
       "mpl.toolbar_items = [[\"Home\", \"Reset original view\", \"fa fa-home icon-home\", \"home\"], [\"Back\", \"Back to  previous view\", \"fa fa-arrow-left icon-arrow-left\", \"back\"], [\"Forward\", \"Forward to next view\", \"fa fa-arrow-right icon-arrow-right\", \"forward\"], [\"\", \"\", \"\", \"\"], [\"Pan\", \"Pan axes with left mouse, zoom with right\", \"fa fa-arrows icon-move\", \"pan\"], [\"Zoom\", \"Zoom to rectangle\", \"fa fa-square-o icon-check-empty\", \"zoom\"], [\"\", \"\", \"\", \"\"], [\"Download\", \"Download plot\", \"fa fa-floppy-o icon-save\", \"download\"]];\n",
       "\n",
       "mpl.extensions = [\"eps\", \"jpeg\", \"pdf\", \"png\", \"ps\", \"raw\", \"svg\", \"tif\"];\n",
       "\n",
       "mpl.default_extension = \"png\";var comm_websocket_adapter = function(comm) {\n",
       "    // Create a \"websocket\"-like object which calls the given IPython comm\n",
       "    // object with the appropriate methods. Currently this is a non binary\n",
       "    // socket, so there is still some room for performance tuning.\n",
       "    var ws = {};\n",
       "\n",
       "    ws.close = function() {\n",
       "        comm.close()\n",
       "    };\n",
       "    ws.send = function(m) {\n",
       "        //console.log('sending', m);\n",
       "        comm.send(m);\n",
       "    };\n",
       "    // Register the callback with on_msg.\n",
       "    comm.on_msg(function(msg) {\n",
       "        //console.log('receiving', msg['content']['data'], msg);\n",
       "        // Pass the mpl event to the overriden (by mpl) onmessage function.\n",
       "        ws.onmessage(msg['content']['data'])\n",
       "    });\n",
       "    return ws;\n",
       "}\n",
       "\n",
       "mpl.mpl_figure_comm = function(comm, msg) {\n",
       "    // This is the function which gets called when the mpl process\n",
       "    // starts-up an IPython Comm through the \"matplotlib\" channel.\n",
       "\n",
       "    var id = msg.content.data.id;\n",
       "    // Get hold of the div created by the display call when the Comm\n",
       "    // socket was opened in Python.\n",
       "    var element = $(\"#\" + id);\n",
       "    var ws_proxy = comm_websocket_adapter(comm)\n",
       "\n",
       "    function ondownload(figure, format) {\n",
       "        window.open(figure.imageObj.src);\n",
       "    }\n",
       "\n",
       "    var fig = new mpl.figure(id, ws_proxy,\n",
       "                           ondownload,\n",
       "                           element.get(0));\n",
       "\n",
       "    // Call onopen now - mpl needs it, as it is assuming we've passed it a real\n",
       "    // web socket which is closed, not our websocket->open comm proxy.\n",
       "    ws_proxy.onopen();\n",
       "\n",
       "    fig.parent_element = element.get(0);\n",
       "    fig.cell_info = mpl.find_output_cell(\"<div id='\" + id + \"'></div>\");\n",
       "    if (!fig.cell_info) {\n",
       "        console.error(\"Failed to find cell for figure\", id, fig);\n",
       "        return;\n",
       "    }\n",
       "\n",
       "    var output_index = fig.cell_info[2]\n",
       "    var cell = fig.cell_info[0];\n",
       "\n",
       "};\n",
       "\n",
       "mpl.figure.prototype.handle_close = function(fig, msg) {\n",
       "    var width = fig.canvas.width/mpl.ratio\n",
       "    fig.root.unbind('remove')\n",
       "\n",
       "    // Update the output cell to use the data from the current canvas.\n",
       "    fig.push_to_output();\n",
       "    var dataURL = fig.canvas.toDataURL();\n",
       "    // Re-enable the keyboard manager in IPython - without this line, in FF,\n",
       "    // the notebook keyboard shortcuts fail.\n",
       "    IPython.keyboard_manager.enable()\n",
       "    $(fig.parent_element).html('<img src=\"' + dataURL + '\" width=\"' + width + '\">');\n",
       "    fig.close_ws(fig, msg);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.close_ws = function(fig, msg){\n",
       "    fig.send_message('closing', msg);\n",
       "    // fig.ws.close()\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.push_to_output = function(remove_interactive) {\n",
       "    // Turn the data on the canvas into data in the output cell.\n",
       "    var width = this.canvas.width/mpl.ratio\n",
       "    var dataURL = this.canvas.toDataURL();\n",
       "    this.cell_info[1]['text/html'] = '<img src=\"' + dataURL + '\" width=\"' + width + '\">';\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.updated_canvas_event = function() {\n",
       "    // Tell IPython that the notebook contents must change.\n",
       "    IPython.notebook.set_dirty(true);\n",
       "    this.send_message(\"ack\", {});\n",
       "    var fig = this;\n",
       "    // Wait a second, then push the new image to the DOM so\n",
       "    // that it is saved nicely (might be nice to debounce this).\n",
       "    setTimeout(function () { fig.push_to_output() }, 1000);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_toolbar = function() {\n",
       "    var fig = this;\n",
       "\n",
       "    var nav_element = $('<div/>')\n",
       "    nav_element.attr('style', 'width: 100%');\n",
       "    this.root.append(nav_element);\n",
       "\n",
       "    // Define a callback function for later on.\n",
       "    function toolbar_event(event) {\n",
       "        return fig.toolbar_button_onclick(event['data']);\n",
       "    }\n",
       "    function toolbar_mouse_event(event) {\n",
       "        return fig.toolbar_button_onmouseover(event['data']);\n",
       "    }\n",
       "\n",
       "    for(var toolbar_ind in mpl.toolbar_items){\n",
       "        var name = mpl.toolbar_items[toolbar_ind][0];\n",
       "        var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
       "        var image = mpl.toolbar_items[toolbar_ind][2];\n",
       "        var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
       "\n",
       "        if (!name) { continue; };\n",
       "\n",
       "        var button = $('<button class=\"btn btn-default\" href=\"#\" title=\"' + name + '\"><i class=\"fa ' + image + ' fa-lg\"></i></button>');\n",
       "        button.click(method_name, toolbar_event);\n",
       "        button.mouseover(tooltip, toolbar_mouse_event);\n",
       "        nav_element.append(button);\n",
       "    }\n",
       "\n",
       "    // Add the status bar.\n",
       "    var status_bar = $('<span class=\"mpl-message\" style=\"text-align:right; float: right;\"/>');\n",
       "    nav_element.append(status_bar);\n",
       "    this.message = status_bar[0];\n",
       "\n",
       "    // Add the close button to the window.\n",
       "    var buttongrp = $('<div class=\"btn-group inline pull-right\"></div>');\n",
       "    var button = $('<button class=\"btn btn-mini btn-primary\" href=\"#\" title=\"Stop Interaction\"><i class=\"fa fa-power-off icon-remove icon-large\"></i></button>');\n",
       "    button.click(function (evt) { fig.handle_close(fig, {}); } );\n",
       "    button.mouseover('Stop Interaction', toolbar_mouse_event);\n",
       "    buttongrp.append(button);\n",
       "    var titlebar = this.root.find($('.ui-dialog-titlebar'));\n",
       "    titlebar.prepend(buttongrp);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._root_extra_style = function(el){\n",
       "    var fig = this\n",
       "    el.on(\"remove\", function(){\n",
       "\tfig.close_ws(fig, {});\n",
       "    });\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._canvas_extra_style = function(el){\n",
       "    // this is important to make the div 'focusable\n",
       "    el.attr('tabindex', 0)\n",
       "    // reach out to IPython and tell the keyboard manager to turn it's self\n",
       "    // off when our div gets focus\n",
       "\n",
       "    // location in version 3\n",
       "    if (IPython.notebook.keyboard_manager) {\n",
       "        IPython.notebook.keyboard_manager.register_events(el);\n",
       "    }\n",
       "    else {\n",
       "        // location in version 2\n",
       "        IPython.keyboard_manager.register_events(el);\n",
       "    }\n",
       "\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._key_event_extra = function(event, name) {\n",
       "    var manager = IPython.notebook.keyboard_manager;\n",
       "    if (!manager)\n",
       "        manager = IPython.keyboard_manager;\n",
       "\n",
       "    // Check for shift+enter\n",
       "    if (event.shiftKey && event.which == 13) {\n",
       "        this.canvas_div.blur();\n",
       "        event.shiftKey = false;\n",
       "        // Send a \"J\" for go to next cell\n",
       "        event.which = 74;\n",
       "        event.keyCode = 74;\n",
       "        manager.command_mode();\n",
       "        manager.handle_keydown(event);\n",
       "    }\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_save = function(fig, msg) {\n",
       "    fig.ondownload(fig, null);\n",
       "}\n",
       "\n",
       "\n",
       "mpl.find_output_cell = function(html_output) {\n",
       "    // Return the cell and output element which can be found *uniquely* in the notebook.\n",
       "    // Note - this is a bit hacky, but it is done because the \"notebook_saving.Notebook\"\n",
       "    // IPython event is triggered only after the cells have been serialised, which for\n",
       "    // our purposes (turning an active figure into a static one), is too late.\n",
       "    var cells = IPython.notebook.get_cells();\n",
       "    var ncells = cells.length;\n",
       "    for (var i=0; i<ncells; i++) {\n",
       "        var cell = cells[i];\n",
       "        if (cell.cell_type === 'code'){\n",
       "            for (var j=0; j<cell.output_area.outputs.length; j++) {\n",
       "                var data = cell.output_area.outputs[j];\n",
       "                if (data.data) {\n",
       "                    // IPython >= 3 moved mimebundle to data attribute of output\n",
       "                    data = data.data;\n",
       "                }\n",
       "                if (data['text/html'] == html_output) {\n",
       "                    return [cell, data, j];\n",
       "                }\n",
       "            }\n",
       "        }\n",
       "    }\n",
       "}\n",
       "\n",
       "// Register the function which deals with the matplotlib target/channel.\n",
       "// The kernel may be null if the page has been refreshed.\n",
       "if (IPython.notebook.kernel != null) {\n",
       "    IPython.notebook.kernel.comm_manager.register_target('matplotlib', mpl.mpl_figure_comm);\n",
       "}\n"
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<img src=\"data:image/png;base64,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\" width=\"640\">"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "No handles with labels found to put in legend.\n"
     ]
    }
   ],
   "source": [
    "# Configuration to control GPU use\n",
    "config = tf.ConfigProto()\n",
    "config.gpu_options.allow_growth = True\n",
    "config.gpu_options.per_process_gpu_memory_fraction = 0.33\n",
    "# sess = tf.InteractiveSession(config=config)\n",
    "\n",
    "num_batches_per_epoch = int(len(x_)/batch_size)\n",
    "\n",
    "with tf.Session(graph=graph,config=config) as session:\n",
    "    # Initializate the weights and biases\n",
    "    tf.global_variables_initializer().run()\n",
    "    \n",
    "    train_epoch = range(num_epochs)\n",
    "    train_edist = []\n",
    "    val_edist = []\n",
    "\n",
    "    for curr_epoch in range(num_epochs):\n",
    "        train_cost = train_ler = 0\n",
    "        start = time.time()\n",
    "\n",
    "        for batch in range(num_batches_per_epoch//1000):\n",
    "            b = int(np.random.randint(len(x_), size=1))\n",
    "            if not batch%100:\n",
    "                print('batch:', batch, b, x_[b].shape)\n",
    "                \n",
    "            feed = {inputs: x_[b],\n",
    "                    targets: y_[b],\n",
    "                    seq_len: [ x_[b].shape[1] ]}\n",
    "\n",
    "            batch_cost, _ = session.run([cost, optimizer], feed)\n",
    "            train_cost += batch_cost*batch_size\n",
    "            train_ler += session.run(ler, feed_dict=feed)*batch_size\n",
    "\n",
    "        train_cost /= num_examples\n",
    "        train_ler /= num_examples\n",
    "        \n",
    "        print(num_epochs, train_cost, train_ler,time.time() - start)\n",
    "        train_edist.append(train_ler)\n",
    "\n",
    "        val_cost, val_ler = 0, 0\n",
    "        for batch in range(len(val_x_)//20):\n",
    "            b = int(np.random.randint(len(val_x_), size=1) )\n",
    "            val_feed = {inputs: val_x_[b],\n",
    "                        targets: val_y_[b],\n",
    "                        seq_len: [val_x_[b].shape[1]]}\n",
    "\n",
    "            val_c, val_l = session.run([cost, ler], feed_dict=val_feed)\n",
    "            val_cost += val_c\n",
    "            val_ler += val_l \n",
    "\n",
    "        log = \"Epoch {}/{}, train_cost = {:.3f}, train_ler = {:.3f}, val_cost = {:.3f}, val_ler = {:.3f}, time = {:.3f}\"\n",
    "        print(log.format(curr_epoch+1, num_epochs, train_cost, train_ler,\n",
    "                         val_cost, val_ler, time.time() - start))\n",
    "        val_edist.append(val_ler)\n",
    "    \n",
    "    \n",
    "    \n",
    "    for batch in range(len(test_x_)//1000):\n",
    "        b = int(batch)\n",
    "        test_feed = {inputs: test_x_[b],\n",
    "                        targets: test_y_[b],\n",
    "                        seq_len: [test_x_[b].shape[1]]}\n",
    "            \n",
    "        # Decoding\n",
    "        d = session.run(decoded[0], feed_dict=test_feed)\n",
    "        str_decoded = ''.join([chr(x) for x in np.asarray(d[1]) + FIRST_INDEX])\n",
    "        # Replacing blank label to none\n",
    "        str_decoded = str_decoded.replace(chr(ord('z') + 1), '')\n",
    "        # Replacing space label to space\n",
    "        str_decoded = str_decoded.replace(chr(ord('a') - 1), ' ')\n",
    "\n",
    "        print('Original:', test_text_[b])\n",
    "        print('Decoded:' , str_decoded)\n",
    "        \n",
    "        %matplotlib notebook\n",
    "        plt.figure()\n",
    "        plt.plot(train_epoch, train_edist, 'r-')\n",
    "        plt.plot(train_epoch, val_edist, 'b-')\n",
    "        plt.xlabel(\"epoch\")\n",
    "        plt.ylabel(\"Accuracy\")\n",
    "        plt.legend(loc='center left', bbox_to_anchor=(1, 0.5))\n",
    "        plt.tight_layout()\n",
    "        plt.subplots_adjust(right=0.8) \n",
    "        plt.show()"
   ]
  },
  {
   "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": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}