entity-tagging/2.concat-luong.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from bs4 import BeautifulSoup\n",
    "import re\n",
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "from tqdm import tqdm"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "def process_string(string):\n",
    "    string = re.sub('[^A-Za-z0-9\\-\\/ ]+', ' ', string).split()\n",
    "    return [y.strip() for y in string]\n",
    "\n",
    "def to_title(string):\n",
    "    if string.isupper():\n",
    "        string = string.title()\n",
    "    return string"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "def parse_raw(filename):\n",
    "    with open(filename, 'r') as fopen:\n",
    "        entities = fopen.read()\n",
    "    soup = BeautifulSoup(entities, 'html.parser')\n",
    "    inside_tag = ''\n",
    "    texts, labels = [], []\n",
    "    for sentence in soup.prettify().split('\\n'):\n",
    "        if len(inside_tag):\n",
    "            splitted = process_string(sentence)\n",
    "            texts += splitted\n",
    "            labels += [inside_tag] * len(splitted)\n",
    "            inside_tag = ''\n",
    "        else:\n",
    "            if not sentence.find('</'):\n",
    "                pass\n",
    "            elif not sentence.find('<'):\n",
    "                inside_tag = sentence.split('>')[0][1:]\n",
    "            else:\n",
    "                splitted = process_string(sentence)\n",
    "                texts += splitted\n",
    "                labels += ['OTHER'] * len(splitted)\n",
    "    assert (len(texts)==len(labels)), \"length texts and labels are not same\"\n",
    "    print('len texts and labels: ', len(texts))\n",
    "    return texts,labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "len texts and labels:  34012\n",
      "len texts and labels:  9249\n"
     ]
    }
   ],
   "source": [
    "train_texts, train_labels = parse_raw('data_train.txt')\n",
    "test_texts, test_labels = parse_raw('data_test.txt')\n",
    "train_texts += test_texts\n",
    "train_labels += test_labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array(['OTHER', 'location', 'organization', 'person', 'quantity', 'time'],\n",
       "       dtype='<U12'), array([35613,  1536,  1592,  2358,  1336,   826]))"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.unique(train_labels,return_counts=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "with open('entities-bm-normalize-v3.txt','r') as fopen:\n",
    "    entities_bm = fopen.read().split('\\n')[:-1]\n",
    "entities_bm = [i.split() for i in entities_bm]\n",
    "entities_bm = [[i[0],'TIME' if i[0] in 'jam' else i[1]] for i in entities_bm]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "replace_by = {'organizaiton':'organization','orgnization':'organization',\n",
    "             'othoer': 'OTHER'}\n",
    "\n",
    "with open('NER-part1.txt','r') as fopen:\n",
    "    nexts = fopen.read().split('\\n')[:-1]\n",
    "nexts = [i.split() for i in nexts]\n",
    "for i in nexts:\n",
    "    if len(i) == 2:\n",
    "        label = i[1].lower()\n",
    "        if 'other' in label:\n",
    "            label = label.upper()\n",
    "        if label in replace_by:\n",
    "            label = replace_by[label]\n",
    "        train_labels.append(label)\n",
    "        train_texts.append(i[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "'KN'\n",
      "'KA'\n"
     ]
    }
   ],
   "source": [
    "replace_by = {'LOC':'location','PRN':'person','NORP':'organization','ORG':'organization','LAW':'law',\n",
    "             'EVENT':'event','FAC':'organization','TIME':'time','O':'OTHER','ART':'person','DOC':'law'}\n",
    "for i in entities_bm:\n",
    "    try:\n",
    "        string = process_string(i[0])\n",
    "        if len(string):\n",
    "            train_labels.append(replace_by[i[1]])\n",
    "            train_texts.append(process_string(i[0])[0])  \n",
    "    except Exception as e:\n",
    "        print(e)\n",
    "        \n",
    "assert (len(train_texts)==len(train_labels)), \"length texts and labels are not same\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array(['OTHER', 'event', 'law', 'location', 'organization', 'person',\n",
       "        'quantity', 'time'], dtype='<U12'),\n",
       " array([49712,   234,   185,  2056,  2596,  4397,  1341,  1296]))"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.unique(train_labels,return_counts=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "word2idx = {'PAD': 0,'NUM':1,'UNK':2}\n",
    "tag2idx = {'PAD': 0}\n",
    "char2idx = {'PAD': 0}\n",
    "word_idx = 3\n",
    "tag_idx = 1\n",
    "char_idx = 1\n",
    "\n",
    "def parse_XY(texts, labels):\n",
    "    global word2idx, tag2idx, char2idx, word_idx, tag_idx, char_idx\n",
    "    X, Y = [], []\n",
    "    for no, text in enumerate(texts):\n",
    "        text = text.lower()\n",
    "        tag = labels[no]\n",
    "        for c in text:\n",
    "            if c not in char2idx:\n",
    "                char2idx[c] = char_idx\n",
    "                char_idx += 1\n",
    "        if tag not in tag2idx:\n",
    "            tag2idx[tag] = tag_idx\n",
    "            tag_idx += 1\n",
    "        Y.append(tag2idx[tag])\n",
    "        if text not in word2idx:\n",
    "            word2idx[text] = word_idx\n",
    "            word_idx += 1\n",
    "        X.append(word2idx[text])\n",
    "    return X, np.array(Y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['OTHER', 'event', 'law', 'location', 'organization', 'person',\n",
       "       'quantity', 'time'], dtype='<U12')"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.unique(train_labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "X, Y = parse_XY(train_texts, train_labels)\n",
    "idx2word={idx: tag for tag, idx in word2idx.items()}\n",
    "idx2tag = {i: w for w, i in tag2idx.items()}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "seq_len = 50\n",
    "def iter_seq(x):\n",
    "    return np.array([x[i: i+seq_len] for i in range(0, len(x)-seq_len, 1)])\n",
    "\n",
    "def to_train_seq(*args):\n",
    "    return [iter_seq(x) for x in args]\n",
    "\n",
    "def generate_char_seq(batch):\n",
    "    x = [[len(idx2word[i]) for i in k] for k in batch]\n",
    "    maxlen = max([j for i in x for j in i])\n",
    "    temp = np.zeros((batch.shape[0],batch.shape[1],maxlen),dtype=np.int32)\n",
    "    for i in range(batch.shape[0]):\n",
    "        for k in range(batch.shape[1]):\n",
    "            for no, c in enumerate(idx2word[batch[i,k]]):\n",
    "                temp[i,k,-1-no] = char2idx[c]\n",
    "    return temp"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "import json\n",
    "with open('luong-ner.json','w') as fopen:\n",
    "    fopen.write(json.dumps({'idx2tag':idx2tag,'idx2word':idx2word,\n",
    "           'word2idx':word2idx,'tag2idx':tag2idx,'char2idx':char2idx}))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(61767, 50)"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_seq, Y_seq = to_train_seq(X, Y)\n",
    "X_char_seq = generate_char_seq(X_seq)\n",
    "X_seq.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "from keras.utils import to_categorical\n",
    "Y_seq_3d = np.array([to_categorical(i, num_classes=len(tag2idx)) for i in Y_seq])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.5/dist-packages/sklearn/cross_validation.py:41: DeprecationWarning: This module was deprecated in version 0.18 in favor of the model_selection module into which all the refactored classes and functions are moved. Also note that the interface of the new CV iterators are different from that of this module. This module will be removed in 0.20.\n",
      "  \"This module will be removed in 0.20.\", DeprecationWarning)\n"
     ]
    }
   ],
   "source": [
    "from sklearn.cross_validation import train_test_split\n",
    "train_X, test_X, train_Y, test_Y, train_char, test_char = train_test_split(X_seq, Y_seq_3d, X_char_seq, \n",
    "                                                                           test_size=0.2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "class Model:\n",
    "    def __init__(\n",
    "        self,\n",
    "        dim_word,\n",
    "        dim_char,\n",
    "        dropout,\n",
    "        learning_rate,\n",
    "        hidden_size_char,\n",
    "        hidden_size_word,\n",
    "        num_layers,\n",
    "    ):\n",
    "        def cells(size, reuse = False):\n",
    "            return tf.contrib.rnn.DropoutWrapper(\n",
    "                tf.nn.rnn_cell.LSTMCell(\n",
    "                    size,\n",
    "                    initializer = tf.orthogonal_initializer(),\n",
    "                    reuse = reuse,\n",
    "                ),\n",
    "                output_keep_prob = dropout,\n",
    "            )\n",
    "\n",
    "        def luong(embedded, size):\n",
    "            attention_mechanism = tf.contrib.seq2seq.LuongAttention(\n",
    "                num_units = hidden_size_word, memory = embedded\n",
    "            )\n",
    "            return tf.contrib.seq2seq.AttentionWrapper(\n",
    "                cell = cells(hidden_size_word),\n",
    "                attention_mechanism = attention_mechanism,\n",
    "                attention_layer_size = hidden_size_word,\n",
    "            )\n",
    "\n",
    "        self.word_ids = tf.placeholder(tf.int32, shape = [None, None])\n",
    "        self.char_ids = tf.placeholder(tf.int32, shape = [None, None, None])\n",
    "        self.labels = tf.placeholder(tf.int32, shape = [None, None, None])\n",
    "        self.maxlen = tf.shape(self.word_ids)[1]\n",
    "        self.lengths = tf.count_nonzero(self.word_ids, 1)\n",
    "\n",
    "        self.word_embeddings = tf.Variable(\n",
    "            tf.truncated_normal(\n",
    "                [len(word2idx), dim_word], stddev = 1.0 / np.sqrt(dim_word)\n",
    "            )\n",
    "        )\n",
    "        self.char_embeddings = tf.Variable(\n",
    "            tf.truncated_normal(\n",
    "                [len(char2idx), dim_char], stddev = 1.0 / np.sqrt(dim_char)\n",
    "            )\n",
    "        )\n",
    "\n",
    "        word_embedded = tf.nn.embedding_lookup(\n",
    "            self.word_embeddings, self.word_ids\n",
    "        )\n",
    "        char_embedded = tf.nn.embedding_lookup(\n",
    "            self.char_embeddings, self.char_ids\n",
    "        )\n",
    "        s = tf.shape(char_embedded)\n",
    "        char_embedded = tf.reshape(\n",
    "            char_embedded, shape = [s[0] * s[1], s[-2], dim_char]\n",
    "        )\n",
    "\n",
    "        for n in range(num_layers):\n",
    "            (out_fw, out_bw), (\n",
    "                state_fw,\n",
    "                state_bw,\n",
    "            ) = tf.nn.bidirectional_dynamic_rnn(\n",
    "                cell_fw = cells(hidden_size_char),\n",
    "                cell_bw = cells(hidden_size_char),\n",
    "                inputs = char_embedded,\n",
    "                dtype = tf.float32,\n",
    "                scope = 'bidirectional_rnn_char_%d' % (n),\n",
    "            )\n",
    "            char_embedded = tf.concat((out_fw, out_bw), 2)\n",
    "        output = tf.reshape(\n",
    "            char_embedded[:, -1], shape = [s[0], s[1], 2 * hidden_size_char]\n",
    "        )\n",
    "        word_embedded = tf.concat([word_embedded, output], axis = -1)\n",
    "\n",
    "        for n in range(num_layers):\n",
    "            (out_fw, out_bw), (\n",
    "                state_fw,\n",
    "                state_bw,\n",
    "            ) = tf.nn.bidirectional_dynamic_rnn(\n",
    "                cell_fw = luong(word_embedded, hidden_size_word),\n",
    "                cell_bw = luong(word_embedded, hidden_size_word),\n",
    "                inputs = word_embedded,\n",
    "                dtype = tf.float32,\n",
    "                scope = 'bidirectional_rnn_word_%d' % (n),\n",
    "            )\n",
    "            word_embedded = tf.concat((out_fw, out_bw), 2)\n",
    "\n",
    "        logits = tf.layers.dense(word_embedded, len(idx2tag))\n",
    "        y_t = tf.argmax(self.labels, 2)\n",
    "        log_likelihood, transition_params = tf.contrib.crf.crf_log_likelihood(\n",
    "            logits, y_t, self.lengths\n",
    "        )\n",
    "        self.cost = tf.reduce_mean(-log_likelihood)\n",
    "        self.optimizer = tf.train.AdamOptimizer(\n",
    "            learning_rate = learning_rate\n",
    "        ).minimize(self.cost)\n",
    "        mask = tf.sequence_mask(self.lengths, maxlen = self.maxlen)\n",
    "        self.tags_seq, tags_score = tf.contrib.crf.crf_decode(\n",
    "            logits, transition_params, self.lengths\n",
    "        )\n",
    "        self.tags_seq = tf.identity(self.tags_seq, name = 'logits')\n",
    "\n",
    "        y_t = tf.cast(y_t, tf.int32)\n",
    "        self.prediction = tf.boolean_mask(self.tags_seq, mask)\n",
    "        mask_label = tf.boolean_mask(y_t, mask)\n",
    "        correct_pred = tf.equal(self.prediction, mask_label)\n",
    "        correct_index = tf.cast(correct_pred, tf.float32)\n",
    "        self.accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.5/dist-packages/tensorflow/python/ops/gradients_impl.py:112: UserWarning: Converting sparse IndexedSlices to a dense Tensor of unknown shape. This may consume a large amount of memory.\n",
      "  \"Converting sparse IndexedSlices to a dense Tensor of unknown shape. \"\n"
     ]
    }
   ],
   "source": [
    "tf.reset_default_graph()\n",
    "sess = tf.InteractiveSession()\n",
    "\n",
    "dim_word = 128\n",
    "dim_char = 256\n",
    "dropout = 0.8\n",
    "learning_rate = 1e-3\n",
    "hidden_size_char = 64\n",
    "hidden_size_word = 64\n",
    "num_layers = 2\n",
    "batch_size = 32\n",
    "\n",
    "model = Model(dim_word,dim_char,dropout,learning_rate,hidden_size_char,hidden_size_word,num_layers)\n",
    "sess.run(tf.global_variables_initializer())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "train minibatch loop: 100%|██████████| 1545/1545 [25:34<00:00,  1.01it/s, accuracy=0.996, cost=0.498]\n",
      "test minibatch loop: 100%|██████████| 387/387 [02:57<00:00,  2.42it/s, accuracy=1, cost=0.0022]   \n",
      "train minibatch loop:   0%|          | 0/1545 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "time taken: 1711.55144906044\n",
      "epoch: 0, training loss: 5.998380, training acc: 0.964804, valid loss: 0.442472, valid acc: 0.999482\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "train minibatch loop: 100%|██████████| 1545/1545 [25:22<00:00,  1.05it/s, accuracy=1, cost=0.0289]    \n",
      "test minibatch loop: 100%|██████████| 387/387 [03:08<00:00,  2.51it/s, accuracy=1, cost=0.00058]   \n",
      "train minibatch loop:   0%|          | 0/1545 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "time taken: 1711.1502566337585\n",
      "epoch: 1, training loss: 0.322114, training acc: 0.998480, valid loss: 0.262932, valid acc: 1.000661\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "train minibatch loop: 100%|██████████| 1545/1545 [25:18<00:00,  1.07it/s, accuracy=1, cost=0.0204]    \n",
      "test minibatch loop: 100%|██████████| 387/387 [03:18<00:00,  2.31it/s, accuracy=1, cost=0.00113]   "
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "time taken: 1716.2329258918762\n",
      "epoch: 2, training loss: 0.185446, training acc: 0.999362, valid loss: 0.231912, valid acc: 1.000928\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "import time\n",
    "\n",
    "for e in range(3):\n",
    "    lasttime = time.time()\n",
    "    train_acc, train_loss, test_acc, test_loss = 0, 0, 0, 0\n",
    "    pbar = tqdm(\n",
    "        range(0, len(train_X), batch_size), desc = 'train minibatch loop'\n",
    "    )\n",
    "    for i in pbar:\n",
    "        batch_x = train_X[i : min(i + batch_size, train_X.shape[0])]\n",
    "        batch_char = train_char[i : min(i + batch_size, train_X.shape[0])]\n",
    "        batch_y = train_Y[i : min(i + batch_size, train_X.shape[0])]\n",
    "        acc, cost, _ = sess.run(\n",
    "            [model.accuracy, model.cost, model.optimizer],\n",
    "            feed_dict = {\n",
    "                model.word_ids: batch_x,\n",
    "                model.char_ids: batch_char,\n",
    "                model.labels: batch_y\n",
    "            },\n",
    "        )\n",
    "        assert not np.isnan(cost)\n",
    "        train_loss += cost\n",
    "        train_acc += acc\n",
    "        pbar.set_postfix(cost = cost, accuracy = acc)\n",
    "    \n",
    "    pbar = tqdm(\n",
    "        range(0, len(test_X), batch_size), desc = 'test minibatch loop'\n",
    "    )\n",
    "    for i in pbar:\n",
    "        batch_x = test_X[i : min(i + batch_size, test_X.shape[0])]\n",
    "        batch_char = test_char[i : min(i + batch_size, test_X.shape[0])]\n",
    "        batch_y = test_Y[i : min(i + batch_size, test_X.shape[0])]\n",
    "        acc, cost = sess.run(\n",
    "            [model.accuracy, model.cost],\n",
    "            feed_dict = {\n",
    "                model.word_ids: batch_x,\n",
    "                model.char_ids: batch_char,\n",
    "                model.labels: batch_y\n",
    "            },\n",
    "        )\n",
    "        assert not np.isnan(cost)\n",
    "        test_loss += cost\n",
    "        test_acc += acc\n",
    "        pbar.set_postfix(cost = cost, accuracy = acc)\n",
    "    \n",
    "    train_loss /= len(train_X) / batch_size\n",
    "    train_acc /= len(train_X) / batch_size\n",
    "    test_loss /= len(test_X) / batch_size\n",
    "    test_acc /= len(test_X) / batch_size\n",
    "\n",
    "    print('time taken:', time.time() - lasttime)\n",
    "    print(\n",
    "        'epoch: %d, training loss: %f, training acc: %f, valid loss: %f, valid acc: %f\\n'\n",
    "        % (e, train_loss, train_acc, test_loss, test_acc)\n",
    "    )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "def pred2label(pred):\n",
    "    out = []\n",
    "    for pred_i in pred:\n",
    "        out_i = []\n",
    "        for p in pred_i:\n",
    "            out_i.append(idx2tag[p])\n",
    "        out.append(out_i)\n",
    "    return out"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "validation minibatch loop: 100%|██████████| 387/387 [02:44<00:00,  2.60it/s]\n"
     ]
    }
   ],
   "source": [
    "real_Y, predict_Y = [], []\n",
    "\n",
    "pbar = tqdm(\n",
    "    range(0, len(test_X), batch_size), desc = 'validation minibatch loop'\n",
    ")\n",
    "for i in pbar:\n",
    "    batch_x = test_X[i : min(i + batch_size, test_X.shape[0])]\n",
    "    batch_char = test_char[i : min(i + batch_size, test_X.shape[0])]\n",
    "    batch_y = test_Y[i : min(i + batch_size, test_X.shape[0])]\n",
    "    predicted = pred2label(sess.run(model.tags_seq,\n",
    "            feed_dict = {\n",
    "                model.word_ids: batch_x,\n",
    "                model.char_ids: batch_char,\n",
    "            },\n",
    "    ))\n",
    "    real = pred2label(np.argmax(batch_y, axis = 2))\n",
    "    predict_Y.extend(predicted)\n",
    "    real_Y.extend(real)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "       OTHER       1.00      1.00      1.00    497138\n",
      "       event       0.99      0.99      0.99      2331\n",
      "         law       0.99      0.99      0.99      1872\n",
      "    location       0.99      1.00      1.00     20671\n",
      "organization       0.99      1.00      0.99     25942\n",
      "      person       0.99      1.00      1.00     43511\n",
      "    quantity       1.00      1.00      1.00     13376\n",
      "        time       1.00      1.00      1.00     12859\n",
      "\n",
      " avg / total       1.00      1.00      1.00    617700\n",
      "\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import classification_report\n",
    "print(classification_report(np.array(real_Y).ravel(), np.array(predict_Y).ravel()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['Placeholder',\n",
       " 'Placeholder_1',\n",
       " 'Placeholder_2',\n",
       " 'Variable',\n",
       " 'Variable_1',\n",
       " 'bidirectional_rnn_char_0/fw/lstm_cell/kernel',\n",
       " 'bidirectional_rnn_char_0/fw/lstm_cell/bias',\n",
       " 'bidirectional_rnn_char_0/bw/lstm_cell/kernel',\n",
       " 'bidirectional_rnn_char_0/bw/lstm_cell/bias',\n",
       " 'bidirectional_rnn_char_1/fw/lstm_cell/kernel',\n",
       " 'bidirectional_rnn_char_1/fw/lstm_cell/bias',\n",
       " 'bidirectional_rnn_char_1/bw/lstm_cell/kernel',\n",
       " 'bidirectional_rnn_char_1/bw/lstm_cell/bias',\n",
       " 'memory_layer/kernel',\n",
       " 'memory_layer_1/kernel',\n",
       " 'bidirectional_rnn_word_0/fw/attention_wrapper/lstm_cell/kernel',\n",
       " 'bidirectional_rnn_word_0/fw/attention_wrapper/lstm_cell/bias',\n",
       " 'bidirectional_rnn_word_0/fw/attention_wrapper/attention_layer/kernel',\n",
       " 'bidirectional_rnn_word_0/bw/attention_wrapper/lstm_cell/kernel',\n",
       " 'bidirectional_rnn_word_0/bw/attention_wrapper/lstm_cell/bias',\n",
       " 'bidirectional_rnn_word_0/bw/attention_wrapper/attention_layer/kernel',\n",
       " 'memory_layer_2/kernel',\n",
       " 'memory_layer_3/kernel',\n",
       " 'bidirectional_rnn_word_1/fw/attention_wrapper/lstm_cell/kernel',\n",
       " 'bidirectional_rnn_word_1/fw/attention_wrapper/lstm_cell/bias',\n",
       " 'bidirectional_rnn_word_1/fw/attention_wrapper/attention_layer/kernel',\n",
       " 'bidirectional_rnn_word_1/bw/attention_wrapper/lstm_cell/kernel',\n",
       " 'bidirectional_rnn_word_1/bw/attention_wrapper/lstm_cell/bias',\n",
       " 'bidirectional_rnn_word_1/bw/attention_wrapper/attention_layer/kernel',\n",
       " 'dense/kernel',\n",
       " 'dense/bias',\n",
       " 'transitions',\n",
       " 'logits']"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "saver = tf.train.Saver(tf.trainable_variables())\n",
    "saver.save(sess, 'concat-bidirectional-luong/model.ckpt')\n",
    "\n",
    "strings = ','.join(\n",
    "    [\n",
    "        n.name\n",
    "        for n in tf.get_default_graph().as_graph_def().node\n",
    "        if ('Variable' in n.op\n",
    "        or 'Placeholder' in n.name\n",
    "        or 'logits' in n.name\n",
    "        or 'alphas' in n.name)\n",
    "        and 'Adam' not in n.name\n",
    "        and 'beta' not in n.name\n",
    "        and 'OptimizeLoss' not in n.name\n",
    "        and 'Global_Step' not in n.name\n",
    "    ]\n",
    ")\n",
    "strings.split(',')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "def freeze_graph(model_dir, output_node_names):\n",
    "\n",
    "    if not tf.gfile.Exists(model_dir):\n",
    "        raise AssertionError(\n",
    "            \"Export directory doesn't exists. Please specify an export \"\n",
    "            'directory: %s' % model_dir\n",
    "        )\n",
    "\n",
    "    checkpoint = tf.train.get_checkpoint_state(model_dir)\n",
    "    input_checkpoint = checkpoint.model_checkpoint_path\n",
    "\n",
    "    absolute_model_dir = '/'.join(input_checkpoint.split('/')[:-1])\n",
    "    output_graph = absolute_model_dir + '/frozen_model.pb'\n",
    "    clear_devices = True\n",
    "    with tf.Session(graph = tf.Graph()) as sess:\n",
    "        saver = tf.train.import_meta_graph(\n",
    "            input_checkpoint + '.meta', clear_devices = clear_devices\n",
    "        )\n",
    "        saver.restore(sess, input_checkpoint)\n",
    "        output_graph_def = tf.graph_util.convert_variables_to_constants(\n",
    "            sess,\n",
    "            tf.get_default_graph().as_graph_def(),\n",
    "            output_node_names.split(','),\n",
    "        )\n",
    "        with tf.gfile.GFile(output_graph, 'wb') as f:\n",
    "            f.write(output_graph_def.SerializeToString())\n",
    "        print('%d ops in the final graph.' % len(output_graph_def.node))\n",
    "        \n",
    "def load_graph(frozen_graph_filename):\n",
    "    with tf.gfile.GFile(frozen_graph_filename, 'rb') as f:\n",
    "        graph_def = tf.GraphDef()\n",
    "        graph_def.ParseFromString(f.read())\n",
    "    with tf.Graph().as_default() as graph:\n",
    "        tf.import_graph_def(graph_def)\n",
    "    return graph"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Restoring parameters from concat-bidirectional-luong/model.ckpt\n",
      "INFO:tensorflow:Froze 29 variables.\n",
      "INFO:tensorflow:Converted 29 variables to const ops.\n",
      "1856 ops in the final graph.\n"
     ]
    }
   ],
   "source": [
    "freeze_graph('concat-bidirectional-luong', strings)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "g = load_graph('concat-bidirectional-luong/frozen_model.pb')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['import/Placeholder',\n",
       " 'import/Placeholder_1',\n",
       " 'import/Placeholder_2',\n",
       " 'import/count_nonzero/zeros',\n",
       " 'import/count_nonzero/NotEqual',\n",
       " 'import/count_nonzero/ToInt64',\n",
       " 'import/count_nonzero/Sum/reduction_indices',\n",
       " 'import/count_nonzero/Sum',\n",
       " 'import/Variable',\n",
       " 'import/Variable/read',\n",
       " 'import/Variable_1',\n",
       " 'import/Variable_1/read',\n",
       " 'import/embedding_lookup/axis',\n",
       " 'import/embedding_lookup',\n",
       " 'import/embedding_lookup/Identity',\n",
       " 'import/embedding_lookup_1/axis',\n",
       " 'import/embedding_lookup_1',\n",
       " 'import/embedding_lookup_1/Identity',\n",
       " 'import/Shape_1',\n",
       " 'import/strided_slice_1/stack',\n",
       " 'import/strided_slice_1/stack_1',\n",
       " 'import/strided_slice_1/stack_2',\n",
       " 'import/strided_slice_1',\n",
       " 'import/strided_slice_2/stack',\n",
       " 'import/strided_slice_2/stack_1',\n",
       " 'import/strided_slice_2/stack_2',\n",
       " 'import/strided_slice_2',\n",
       " 'import/mul',\n",
       " 'import/strided_slice_3/stack',\n",
       " 'import/strided_slice_3/stack_1',\n",
       " 'import/strided_slice_3/stack_2',\n",
       " 'import/strided_slice_3',\n",
       " 'import/Reshape/shape/2',\n",
       " 'import/Reshape/shape',\n",
       " 'import/Reshape',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/Rank',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/range/start',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/range/delta',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/range',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/concat/values_0',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/concat/axis',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/concat',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/transpose',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/Shape',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/strided_slice/stack',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/strided_slice/stack_1',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/strided_slice/stack_2',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/strided_slice',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/DropoutWrapperZeroState/LSTMCellZeroState/ExpandDims/dim',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/DropoutWrapperZeroState/LSTMCellZeroState/ExpandDims',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/DropoutWrapperZeroState/LSTMCellZeroState/Const',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/DropoutWrapperZeroState/LSTMCellZeroState/concat/axis',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/DropoutWrapperZeroState/LSTMCellZeroState/concat',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/DropoutWrapperZeroState/LSTMCellZeroState/zeros/Const',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/DropoutWrapperZeroState/LSTMCellZeroState/zeros',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/DropoutWrapperZeroState/LSTMCellZeroState/ExpandDims_2/dim',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/DropoutWrapperZeroState/LSTMCellZeroState/ExpandDims_2',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/DropoutWrapperZeroState/LSTMCellZeroState/Const_2',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/DropoutWrapperZeroState/LSTMCellZeroState/concat_1/axis',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/DropoutWrapperZeroState/LSTMCellZeroState/concat_1',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/DropoutWrapperZeroState/LSTMCellZeroState/zeros_1/Const',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/DropoutWrapperZeroState/LSTMCellZeroState/zeros_1',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/Shape_1',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/strided_slice_1/stack',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/strided_slice_1/stack_1',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/strided_slice_1/stack_2',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/strided_slice_1',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/time',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/TensorArray',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/TensorArray_1',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/TensorArrayUnstack/Shape',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/TensorArrayUnstack/strided_slice/stack',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/TensorArrayUnstack/strided_slice/stack_1',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/TensorArrayUnstack/strided_slice/stack_2',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/TensorArrayUnstack/strided_slice',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/TensorArrayUnstack/range/start',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/TensorArrayUnstack/range/delta',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/TensorArrayUnstack/range',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/TensorArrayUnstack/TensorArrayScatter/TensorArrayScatterV3',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/Maximum/x',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/Maximum',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/Minimum',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/iteration_counter',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/Enter',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/Enter_1',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/Enter_2',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/Enter_3',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/Enter_4',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/Merge',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/Merge_1',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/Merge_2',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/Merge_3',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/Merge_4',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/Less/Enter',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/Less',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/Less_1/Enter',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/Less_1',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/LogicalAnd',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/LoopCond',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/Switch',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/Switch_1',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/Switch_2',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/Switch_3',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/Switch_4',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/Identity',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/Identity_1',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/Identity_2',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/Identity_3',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/Identity_4',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/add/y',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/add',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/TensorArrayReadV3/Enter',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/TensorArrayReadV3/Enter_1',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/TensorArrayReadV3',\n",
       " 'import/bidirectional_rnn_char_0/fw/lstm_cell/kernel',\n",
       " 'import/bidirectional_rnn_char_0/fw/lstm_cell/kernel/read',\n",
       " 'import/bidirectional_rnn_char_0/fw/lstm_cell/bias',\n",
       " 'import/bidirectional_rnn_char_0/fw/lstm_cell/bias/read',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/lstm_cell/concat/axis',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/lstm_cell/concat',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/lstm_cell/MatMul/Enter',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/lstm_cell/MatMul',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/lstm_cell/BiasAdd/Enter',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/lstm_cell/BiasAdd',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/lstm_cell/split/split_dim',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/lstm_cell/split',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/lstm_cell/add/y',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/lstm_cell/add',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/lstm_cell/Sigmoid',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/lstm_cell/mul',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/lstm_cell/Sigmoid_1',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/lstm_cell/Tanh',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/lstm_cell/mul_1',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/lstm_cell/add_1',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/lstm_cell/Sigmoid_2',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/lstm_cell/Tanh_1',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/lstm_cell/mul_2',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/dropout/keep_prob',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/dropout/Shape',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/dropout/random_uniform/min',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/dropout/random_uniform/max',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/dropout/random_uniform/RandomUniform',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/dropout/random_uniform/sub',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/dropout/random_uniform/mul',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/dropout/random_uniform',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/dropout/add',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/dropout/Floor',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/dropout/div',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/dropout/mul',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/TensorArrayWrite/TensorArrayWriteV3/Enter',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/TensorArrayWrite/TensorArrayWriteV3',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/add_1/y',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/add_1',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/NextIteration',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/NextIteration_1',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/NextIteration_2',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/NextIteration_3',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/NextIteration_4',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/while/Exit_2',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/TensorArrayStack/TensorArraySizeV3',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/TensorArrayStack/range/start',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/TensorArrayStack/range/delta',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/TensorArrayStack/range',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/TensorArrayStack/TensorArrayGatherV3',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/Rank_1',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/range_1/start',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/range_1/delta',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/range_1',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/concat_2/values_0',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/concat_2/axis',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/concat_2',\n",
       " 'import/bidirectional_rnn_char_0/fw/fw/transpose_1',\n",
       " 'import/bidirectional_rnn_char_0/bw/ReverseV2/axis',\n",
       " 'import/bidirectional_rnn_char_0/bw/ReverseV2',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/Rank',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/range/start',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/range/delta',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/range',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/concat/values_0',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/concat/axis',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/concat',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/transpose',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/Shape',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/strided_slice/stack',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/strided_slice/stack_1',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/strided_slice/stack_2',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/strided_slice',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/DropoutWrapperZeroState/LSTMCellZeroState/ExpandDims/dim',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/DropoutWrapperZeroState/LSTMCellZeroState/ExpandDims',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/DropoutWrapperZeroState/LSTMCellZeroState/Const',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/DropoutWrapperZeroState/LSTMCellZeroState/concat/axis',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/DropoutWrapperZeroState/LSTMCellZeroState/concat',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/DropoutWrapperZeroState/LSTMCellZeroState/zeros/Const',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/DropoutWrapperZeroState/LSTMCellZeroState/zeros',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/DropoutWrapperZeroState/LSTMCellZeroState/ExpandDims_2/dim',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/DropoutWrapperZeroState/LSTMCellZeroState/ExpandDims_2',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/DropoutWrapperZeroState/LSTMCellZeroState/Const_2',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/DropoutWrapperZeroState/LSTMCellZeroState/concat_1/axis',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/DropoutWrapperZeroState/LSTMCellZeroState/concat_1',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/DropoutWrapperZeroState/LSTMCellZeroState/zeros_1/Const',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/DropoutWrapperZeroState/LSTMCellZeroState/zeros_1',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/Shape_1',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/strided_slice_1/stack',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/strided_slice_1/stack_1',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/strided_slice_1/stack_2',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/strided_slice_1',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/time',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/TensorArray',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/TensorArray_1',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/TensorArrayUnstack/Shape',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/TensorArrayUnstack/strided_slice/stack',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/TensorArrayUnstack/strided_slice/stack_1',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/TensorArrayUnstack/strided_slice/stack_2',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/TensorArrayUnstack/strided_slice',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/TensorArrayUnstack/range/start',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/TensorArrayUnstack/range/delta',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/TensorArrayUnstack/range',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/TensorArrayUnstack/TensorArrayScatter/TensorArrayScatterV3',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/Maximum/x',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/Maximum',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/Minimum',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/iteration_counter',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/Enter',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/Enter_1',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/Enter_2',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/Enter_3',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/Enter_4',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/Merge',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/Merge_1',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/Merge_2',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/Merge_3',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/Merge_4',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/Less/Enter',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/Less',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/Less_1/Enter',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/Less_1',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/LogicalAnd',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/LoopCond',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/Switch',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/Switch_1',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/Switch_2',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/Switch_3',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/Switch_4',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/Identity',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/Identity_1',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/Identity_2',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/Identity_3',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/Identity_4',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/add/y',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/add',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/TensorArrayReadV3/Enter',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/TensorArrayReadV3/Enter_1',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/TensorArrayReadV3',\n",
       " 'import/bidirectional_rnn_char_0/bw/lstm_cell/kernel',\n",
       " 'import/bidirectional_rnn_char_0/bw/lstm_cell/kernel/read',\n",
       " 'import/bidirectional_rnn_char_0/bw/lstm_cell/bias',\n",
       " 'import/bidirectional_rnn_char_0/bw/lstm_cell/bias/read',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/lstm_cell/concat/axis',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/lstm_cell/concat',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/lstm_cell/MatMul/Enter',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/lstm_cell/MatMul',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/lstm_cell/BiasAdd/Enter',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/lstm_cell/BiasAdd',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/lstm_cell/split/split_dim',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/lstm_cell/split',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/lstm_cell/add/y',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/lstm_cell/add',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/lstm_cell/Sigmoid',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/lstm_cell/mul',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/lstm_cell/Sigmoid_1',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/lstm_cell/Tanh',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/lstm_cell/mul_1',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/lstm_cell/add_1',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/lstm_cell/Sigmoid_2',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/lstm_cell/Tanh_1',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/lstm_cell/mul_2',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/dropout/keep_prob',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/dropout/Shape',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/dropout/random_uniform/min',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/dropout/random_uniform/max',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/dropout/random_uniform/RandomUniform',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/dropout/random_uniform/sub',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/dropout/random_uniform/mul',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/dropout/random_uniform',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/dropout/add',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/dropout/Floor',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/dropout/div',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/dropout/mul',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/TensorArrayWrite/TensorArrayWriteV3/Enter',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/TensorArrayWrite/TensorArrayWriteV3',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/add_1/y',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/add_1',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/NextIteration',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/NextIteration_1',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/NextIteration_2',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/NextIteration_3',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/NextIteration_4',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/while/Exit_2',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/TensorArrayStack/TensorArraySizeV3',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/TensorArrayStack/range/start',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/TensorArrayStack/range/delta',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/TensorArrayStack/range',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/TensorArrayStack/TensorArrayGatherV3',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/Rank_1',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/range_1/start',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/range_1/delta',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/range_1',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/concat_2/values_0',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/concat_2/axis',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/concat_2',\n",
       " 'import/bidirectional_rnn_char_0/bw/bw/transpose_1',\n",
       " 'import/ReverseV2/axis',\n",
       " 'import/ReverseV2',\n",
       " 'import/concat/axis',\n",
       " 'import/concat',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/Rank',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/range/start',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/range/delta',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/range',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/concat/values_0',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/concat/axis',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/concat',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/transpose',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/Shape',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/strided_slice/stack',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/strided_slice/stack_1',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/strided_slice/stack_2',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/strided_slice',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/DropoutWrapperZeroState/LSTMCellZeroState/ExpandDims/dim',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/DropoutWrapperZeroState/LSTMCellZeroState/ExpandDims',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/DropoutWrapperZeroState/LSTMCellZeroState/Const',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/DropoutWrapperZeroState/LSTMCellZeroState/concat/axis',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/DropoutWrapperZeroState/LSTMCellZeroState/concat',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/DropoutWrapperZeroState/LSTMCellZeroState/zeros/Const',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/DropoutWrapperZeroState/LSTMCellZeroState/zeros',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/DropoutWrapperZeroState/LSTMCellZeroState/ExpandDims_2/dim',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/DropoutWrapperZeroState/LSTMCellZeroState/ExpandDims_2',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/DropoutWrapperZeroState/LSTMCellZeroState/Const_2',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/DropoutWrapperZeroState/LSTMCellZeroState/concat_1/axis',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/DropoutWrapperZeroState/LSTMCellZeroState/concat_1',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/DropoutWrapperZeroState/LSTMCellZeroState/zeros_1/Const',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/DropoutWrapperZeroState/LSTMCellZeroState/zeros_1',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/Shape_1',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/strided_slice_1/stack',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/strided_slice_1/stack_1',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/strided_slice_1/stack_2',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/strided_slice_1',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/time',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/TensorArray',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/TensorArray_1',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/TensorArrayUnstack/Shape',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/TensorArrayUnstack/strided_slice/stack',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/TensorArrayUnstack/strided_slice/stack_1',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/TensorArrayUnstack/strided_slice/stack_2',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/TensorArrayUnstack/strided_slice',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/TensorArrayUnstack/range/start',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/TensorArrayUnstack/range/delta',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/TensorArrayUnstack/range',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/TensorArrayUnstack/TensorArrayScatter/TensorArrayScatterV3',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/Maximum/x',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/Maximum',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/Minimum',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/iteration_counter',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/Enter',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/Enter_1',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/Enter_2',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/Enter_3',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/Enter_4',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/Merge',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/Merge_1',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/Merge_2',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/Merge_3',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/Merge_4',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/Less/Enter',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/Less',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/Less_1/Enter',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/Less_1',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/LogicalAnd',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/LoopCond',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/Switch',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/Switch_1',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/Switch_2',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/Switch_3',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/Switch_4',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/Identity',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/Identity_1',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/Identity_2',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/Identity_3',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/Identity_4',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/add/y',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/add',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/TensorArrayReadV3/Enter',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/TensorArrayReadV3/Enter_1',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/TensorArrayReadV3',\n",
       " 'import/bidirectional_rnn_char_1/fw/lstm_cell/kernel',\n",
       " 'import/bidirectional_rnn_char_1/fw/lstm_cell/kernel/read',\n",
       " 'import/bidirectional_rnn_char_1/fw/lstm_cell/bias',\n",
       " 'import/bidirectional_rnn_char_1/fw/lstm_cell/bias/read',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/lstm_cell/concat/axis',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/lstm_cell/concat',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/lstm_cell/MatMul/Enter',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/lstm_cell/MatMul',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/lstm_cell/BiasAdd/Enter',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/lstm_cell/BiasAdd',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/lstm_cell/split/split_dim',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/lstm_cell/split',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/lstm_cell/add/y',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/lstm_cell/add',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/lstm_cell/Sigmoid',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/lstm_cell/mul',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/lstm_cell/Sigmoid_1',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/lstm_cell/Tanh',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/lstm_cell/mul_1',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/lstm_cell/add_1',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/lstm_cell/Sigmoid_2',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/lstm_cell/Tanh_1',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/lstm_cell/mul_2',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/dropout/keep_prob',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/dropout/Shape',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/dropout/random_uniform/min',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/dropout/random_uniform/max',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/dropout/random_uniform/RandomUniform',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/dropout/random_uniform/sub',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/dropout/random_uniform/mul',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/dropout/random_uniform',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/dropout/add',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/dropout/Floor',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/dropout/div',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/dropout/mul',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/TensorArrayWrite/TensorArrayWriteV3/Enter',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/TensorArrayWrite/TensorArrayWriteV3',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/add_1/y',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/add_1',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/NextIteration',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/NextIteration_1',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/NextIteration_2',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/NextIteration_3',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/NextIteration_4',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/while/Exit_2',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/TensorArrayStack/TensorArraySizeV3',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/TensorArrayStack/range/start',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/TensorArrayStack/range/delta',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/TensorArrayStack/range',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/TensorArrayStack/TensorArrayGatherV3',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/Rank_1',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/range_1/start',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/range_1/delta',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/range_1',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/concat_2/values_0',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/concat_2/axis',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/concat_2',\n",
       " 'import/bidirectional_rnn_char_1/fw/fw/transpose_1',\n",
       " 'import/bidirectional_rnn_char_1/bw/ReverseV2/axis',\n",
       " 'import/bidirectional_rnn_char_1/bw/ReverseV2',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/Rank',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/range/start',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/range/delta',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/range',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/concat/values_0',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/concat/axis',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/concat',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/transpose',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/Shape',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/strided_slice/stack',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/strided_slice/stack_1',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/strided_slice/stack_2',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/strided_slice',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/DropoutWrapperZeroState/LSTMCellZeroState/ExpandDims/dim',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/DropoutWrapperZeroState/LSTMCellZeroState/ExpandDims',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/DropoutWrapperZeroState/LSTMCellZeroState/Const',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/DropoutWrapperZeroState/LSTMCellZeroState/concat/axis',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/DropoutWrapperZeroState/LSTMCellZeroState/concat',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/DropoutWrapperZeroState/LSTMCellZeroState/zeros/Const',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/DropoutWrapperZeroState/LSTMCellZeroState/zeros',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/DropoutWrapperZeroState/LSTMCellZeroState/ExpandDims_2/dim',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/DropoutWrapperZeroState/LSTMCellZeroState/ExpandDims_2',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/DropoutWrapperZeroState/LSTMCellZeroState/Const_2',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/DropoutWrapperZeroState/LSTMCellZeroState/concat_1/axis',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/DropoutWrapperZeroState/LSTMCellZeroState/concat_1',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/DropoutWrapperZeroState/LSTMCellZeroState/zeros_1/Const',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/DropoutWrapperZeroState/LSTMCellZeroState/zeros_1',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/Shape_1',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/strided_slice_1/stack',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/strided_slice_1/stack_1',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/strided_slice_1/stack_2',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/strided_slice_1',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/time',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/TensorArray',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/TensorArray_1',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/TensorArrayUnstack/Shape',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/TensorArrayUnstack/strided_slice/stack',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/TensorArrayUnstack/strided_slice/stack_1',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/TensorArrayUnstack/strided_slice/stack_2',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/TensorArrayUnstack/strided_slice',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/TensorArrayUnstack/range/start',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/TensorArrayUnstack/range/delta',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/TensorArrayUnstack/range',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/TensorArrayUnstack/TensorArrayScatter/TensorArrayScatterV3',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/Maximum/x',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/Maximum',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/Minimum',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/iteration_counter',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/Enter',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/Enter_1',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/Enter_2',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/Enter_3',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/Enter_4',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/Merge',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/Merge_1',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/Merge_2',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/Merge_3',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/Merge_4',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/Less/Enter',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/Less',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/Less_1/Enter',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/Less_1',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/LogicalAnd',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/LoopCond',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/Switch',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/Switch_1',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/Switch_2',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/Switch_3',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/Switch_4',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/Identity',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/Identity_1',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/Identity_2',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/Identity_3',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/Identity_4',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/add/y',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/add',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/TensorArrayReadV3/Enter',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/TensorArrayReadV3/Enter_1',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/TensorArrayReadV3',\n",
       " 'import/bidirectional_rnn_char_1/bw/lstm_cell/kernel',\n",
       " 'import/bidirectional_rnn_char_1/bw/lstm_cell/kernel/read',\n",
       " 'import/bidirectional_rnn_char_1/bw/lstm_cell/bias',\n",
       " 'import/bidirectional_rnn_char_1/bw/lstm_cell/bias/read',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/lstm_cell/concat/axis',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/lstm_cell/concat',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/lstm_cell/MatMul/Enter',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/lstm_cell/MatMul',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/lstm_cell/BiasAdd/Enter',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/lstm_cell/BiasAdd',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/lstm_cell/split/split_dim',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/lstm_cell/split',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/lstm_cell/add/y',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/lstm_cell/add',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/lstm_cell/Sigmoid',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/lstm_cell/mul',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/lstm_cell/Sigmoid_1',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/lstm_cell/Tanh',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/lstm_cell/mul_1',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/lstm_cell/add_1',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/lstm_cell/Sigmoid_2',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/lstm_cell/Tanh_1',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/lstm_cell/mul_2',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/dropout/keep_prob',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/dropout/Shape',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/dropout/random_uniform/min',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/dropout/random_uniform/max',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/dropout/random_uniform/RandomUniform',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/dropout/random_uniform/sub',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/dropout/random_uniform/mul',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/dropout/random_uniform',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/dropout/add',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/dropout/Floor',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/dropout/div',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/dropout/mul',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/TensorArrayWrite/TensorArrayWriteV3/Enter',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/TensorArrayWrite/TensorArrayWriteV3',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/add_1/y',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/add_1',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/NextIteration',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/NextIteration_1',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/NextIteration_2',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/NextIteration_3',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/NextIteration_4',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/while/Exit_2',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/TensorArrayStack/TensorArraySizeV3',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/TensorArrayStack/range/start',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/TensorArrayStack/range/delta',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/TensorArrayStack/range',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/TensorArrayStack/TensorArrayGatherV3',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/Rank_1',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/range_1/start',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/range_1/delta',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/range_1',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/concat_2/values_0',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/concat_2/axis',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/concat_2',\n",
       " 'import/bidirectional_rnn_char_1/bw/bw/transpose_1',\n",
       " 'import/ReverseV2_1/axis',\n",
       " 'import/ReverseV2_1',\n",
       " 'import/concat_1/axis',\n",
       " 'import/concat_1',\n",
       " 'import/strided_slice_4/stack',\n",
       " 'import/strided_slice_4/stack_1',\n",
       " 'import/strided_slice_4/stack_2',\n",
       " 'import/strided_slice_4',\n",
       " 'import/strided_slice_5/stack',\n",
       " 'import/strided_slice_5/stack_1',\n",
       " 'import/strided_slice_5/stack_2',\n",
       " 'import/strided_slice_5',\n",
       " 'import/strided_slice_6/stack',\n",
       " 'import/strided_slice_6/stack_1',\n",
       " 'import/strided_slice_6/stack_2',\n",
       " 'import/strided_slice_6',\n",
       " 'import/Reshape_1/shape/2',\n",
       " 'import/Reshape_1/shape',\n",
       " 'import/Reshape_1',\n",
       " 'import/concat_2/axis',\n",
       " 'import/concat_2',\n",
       " 'import/memory_layer/kernel',\n",
       " 'import/memory_layer/kernel/read',\n",
       " 'import/LuongAttention/memory_layer/Tensordot/axes',\n",
       " 'import/LuongAttention/memory_layer/Tensordot/free',\n",
       " 'import/LuongAttention/memory_layer/Tensordot/Shape',\n",
       " 'import/LuongAttention/memory_layer/Tensordot/GatherV2/axis',\n",
       " 'import/LuongAttention/memory_layer/Tensordot/GatherV2',\n",
       " 'import/LuongAttention/memory_layer/Tensordot/GatherV2_1/axis',\n",
       " 'import/LuongAttention/memory_layer/Tensordot/GatherV2_1',\n",
       " 'import/LuongAttention/memory_layer/Tensordot/Const',\n",
       " 'import/LuongAttention/memory_layer/Tensordot/Prod',\n",
       " 'import/LuongAttention/memory_layer/Tensordot/Const_1',\n",
       " 'import/LuongAttention/memory_layer/Tensordot/Prod_1',\n",
       " 'import/LuongAttention/memory_layer/Tensordot/concat/axis',\n",
       " 'import/LuongAttention/memory_layer/Tensordot/concat',\n",
       " 'import/LuongAttention/memory_layer/Tensordot/stack',\n",
       " 'import/LuongAttention/memory_layer/Tensordot/transpose',\n",
       " 'import/LuongAttention/memory_layer/Tensordot/Reshape',\n",
       " 'import/LuongAttention/memory_layer/Tensordot/transpose_1/perm',\n",
       " 'import/LuongAttention/memory_layer/Tensordot/transpose_1',\n",
       " 'import/LuongAttention/memory_layer/Tensordot/Reshape_1/shape',\n",
       " 'import/LuongAttention/memory_layer/Tensordot/Reshape_1',\n",
       " 'import/LuongAttention/memory_layer/Tensordot/MatMul',\n",
       " 'import/LuongAttention/memory_layer/Tensordot/Const_2',\n",
       " 'import/LuongAttention/memory_layer/Tensordot/concat_1/axis',\n",
       " 'import/LuongAttention/memory_layer/Tensordot/concat_1',\n",
       " 'import/LuongAttention/memory_layer/Tensordot',\n",
       " 'import/LuongAttention/Shape_1',\n",
       " 'import/LuongAttention/strided_slice_1/stack',\n",
       " 'import/LuongAttention/strided_slice_1/stack_1',\n",
       " 'import/LuongAttention/strided_slice_1/stack_2',\n",
       " 'import/LuongAttention/strided_slice_1',\n",
       " 'import/memory_layer_1/kernel',\n",
       " 'import/memory_layer_1/kernel/read',\n",
       " 'import/LuongAttention_1/memory_layer/Tensordot/axes',\n",
       " 'import/LuongAttention_1/memory_layer/Tensordot/free',\n",
       " 'import/LuongAttention_1/memory_layer/Tensordot/Shape',\n",
       " 'import/LuongAttention_1/memory_layer/Tensordot/GatherV2/axis',\n",
       " 'import/LuongAttention_1/memory_layer/Tensordot/GatherV2',\n",
       " 'import/LuongAttention_1/memory_layer/Tensordot/GatherV2_1/axis',\n",
       " 'import/LuongAttention_1/memory_layer/Tensordot/GatherV2_1',\n",
       " 'import/LuongAttention_1/memory_layer/Tensordot/Const',\n",
       " 'import/LuongAttention_1/memory_layer/Tensordot/Prod',\n",
       " 'import/LuongAttention_1/memory_layer/Tensordot/Const_1',\n",
       " 'import/LuongAttention_1/memory_layer/Tensordot/Prod_1',\n",
       " 'import/LuongAttention_1/memory_layer/Tensordot/concat/axis',\n",
       " 'import/LuongAttention_1/memory_layer/Tensordot/concat',\n",
       " 'import/LuongAttention_1/memory_layer/Tensordot/stack',\n",
       " 'import/LuongAttention_1/memory_layer/Tensordot/transpose',\n",
       " 'import/LuongAttention_1/memory_layer/Tensordot/Reshape',\n",
       " 'import/LuongAttention_1/memory_layer/Tensordot/transpose_1/perm',\n",
       " 'import/LuongAttention_1/memory_layer/Tensordot/transpose_1',\n",
       " 'import/LuongAttention_1/memory_layer/Tensordot/Reshape_1/shape',\n",
       " 'import/LuongAttention_1/memory_layer/Tensordot/Reshape_1',\n",
       " 'import/LuongAttention_1/memory_layer/Tensordot/MatMul',\n",
       " 'import/LuongAttention_1/memory_layer/Tensordot/Const_2',\n",
       " 'import/LuongAttention_1/memory_layer/Tensordot/concat_1/axis',\n",
       " 'import/LuongAttention_1/memory_layer/Tensordot/concat_1',\n",
       " 'import/LuongAttention_1/memory_layer/Tensordot',\n",
       " 'import/LuongAttention_1/Shape_1',\n",
       " 'import/LuongAttention_1/strided_slice_1/stack',\n",
       " 'import/LuongAttention_1/strided_slice_1/stack_1',\n",
       " 'import/LuongAttention_1/strided_slice_1/stack_2',\n",
       " 'import/LuongAttention_1/strided_slice_1',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/Rank',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/range/start',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/range/delta',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/range',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/concat/values_0',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/concat/axis',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/concat',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/transpose',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/Shape',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/strided_slice/stack',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/strided_slice/stack_1',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/strided_slice/stack_2',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/strided_slice',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/AttentionWrapperZeroState/DropoutWrapperZeroState/LSTMCellZeroState/ExpandDims/dim',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/AttentionWrapperZeroState/DropoutWrapperZeroState/LSTMCellZeroState/ExpandDims',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/AttentionWrapperZeroState/DropoutWrapperZeroState/LSTMCellZeroState/Const',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/AttentionWrapperZeroState/DropoutWrapperZeroState/LSTMCellZeroState/concat/axis',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/AttentionWrapperZeroState/DropoutWrapperZeroState/LSTMCellZeroState/concat',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/AttentionWrapperZeroState/DropoutWrapperZeroState/LSTMCellZeroState/zeros/Const',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/AttentionWrapperZeroState/DropoutWrapperZeroState/LSTMCellZeroState/zeros',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/AttentionWrapperZeroState/DropoutWrapperZeroState/LSTMCellZeroState/ExpandDims_2/dim',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/AttentionWrapperZeroState/DropoutWrapperZeroState/LSTMCellZeroState/ExpandDims_2',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/AttentionWrapperZeroState/DropoutWrapperZeroState/LSTMCellZeroState/Const_2',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/AttentionWrapperZeroState/DropoutWrapperZeroState/LSTMCellZeroState/concat_1/axis',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/AttentionWrapperZeroState/DropoutWrapperZeroState/LSTMCellZeroState/concat_1',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/AttentionWrapperZeroState/DropoutWrapperZeroState/LSTMCellZeroState/zeros_1/Const',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/AttentionWrapperZeroState/DropoutWrapperZeroState/LSTMCellZeroState/zeros_1',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/AttentionWrapperZeroState/assert_equal/Equal',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/AttentionWrapperZeroState/assert_equal/Const',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/AttentionWrapperZeroState/assert_equal/All',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/AttentionWrapperZeroState/assert_equal/Assert/Assert/data_0',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/AttentionWrapperZeroState/assert_equal/Assert/Assert/data_1',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/AttentionWrapperZeroState/assert_equal/Assert/Assert/data_2',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/AttentionWrapperZeroState/assert_equal/Assert/Assert/data_4',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/AttentionWrapperZeroState/assert_equal/Assert/Assert',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/AttentionWrapperZeroState/checked_cell_state',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/AttentionWrapperZeroState/checked_cell_state_1',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/AttentionWrapperZeroState/ExpandDims_4/dim',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/AttentionWrapperZeroState/ExpandDims_4',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/AttentionWrapperZeroState/Const',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/AttentionWrapperZeroState/concat_1/axis',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/AttentionWrapperZeroState/concat_1',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/AttentionWrapperZeroState/zeros_2/Const',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/AttentionWrapperZeroState/zeros_2',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/Shape_1',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/strided_slice_1/stack',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/strided_slice_1/stack_1',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/strided_slice_1/stack_2',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/strided_slice_1',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/time',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/TensorArray',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/TensorArray_1',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/TensorArrayUnstack/Shape',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/TensorArrayUnstack/strided_slice/stack',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/TensorArrayUnstack/strided_slice/stack_1',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/TensorArrayUnstack/strided_slice/stack_2',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/TensorArrayUnstack/strided_slice',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/TensorArrayUnstack/range/start',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/TensorArrayUnstack/range/delta',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/TensorArrayUnstack/range',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/TensorArrayUnstack/TensorArrayScatter/TensorArrayScatterV3',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/Maximum/x',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/Maximum',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/Minimum',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/iteration_counter',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/Enter',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/Enter_1',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/Enter_2',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/Enter_3',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/Enter_4',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/Enter_5',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/Merge',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/Merge_1',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/Merge_2',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/Merge_3',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/Merge_4',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/Merge_5',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/Less/Enter',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/Less',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/Less_1/Enter',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/Less_1',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/LogicalAnd',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/LoopCond',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/Switch',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/Switch_1',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/Switch_2',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/Switch_3',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/Switch_4',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/Switch_5',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/Identity',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/Identity_1',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/Identity_2',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/Identity_3',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/Identity_4',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/Identity_5',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/add/y',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/add',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/TensorArrayReadV3/Enter',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/TensorArrayReadV3/Enter_1',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/TensorArrayReadV3',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/concat/axis',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/concat',\n",
       " 'import/bidirectional_rnn_word_0/fw/attention_wrapper/lstm_cell/kernel',\n",
       " 'import/bidirectional_rnn_word_0/fw/attention_wrapper/lstm_cell/kernel/read',\n",
       " 'import/bidirectional_rnn_word_0/fw/attention_wrapper/lstm_cell/bias',\n",
       " 'import/bidirectional_rnn_word_0/fw/attention_wrapper/lstm_cell/bias/read',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/lstm_cell/concat/axis',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/lstm_cell/concat',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/lstm_cell/MatMul/Enter',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/lstm_cell/MatMul',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/lstm_cell/BiasAdd/Enter',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/lstm_cell/BiasAdd',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/lstm_cell/split/split_dim',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/lstm_cell/split',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/lstm_cell/add/y',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/lstm_cell/add',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/lstm_cell/Sigmoid',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/lstm_cell/mul',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/lstm_cell/Sigmoid_1',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/lstm_cell/Tanh',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/lstm_cell/mul_1',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/lstm_cell/add_1',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/lstm_cell/Sigmoid_2',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/lstm_cell/Tanh_1',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/lstm_cell/mul_2',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/dropout/keep_prob',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/dropout/Shape',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/dropout/random_uniform/min',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/dropout/random_uniform/max',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/dropout/random_uniform/RandomUniform',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/dropout/random_uniform/sub',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/dropout/random_uniform/mul',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/dropout/random_uniform',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/dropout/add',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/dropout/Floor',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/dropout/div',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/dropout/mul',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/Shape',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/strided_slice/stack',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/strided_slice/stack_1',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/strided_slice/stack_2',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/strided_slice',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/assert_equal/Equal/Enter',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/assert_equal/Equal',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/assert_equal/Const',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/assert_equal/All',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/assert_equal/Assert/Assert/data_0',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/assert_equal/Assert/Assert/data_1',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/assert_equal/Assert/Assert/data_2',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/assert_equal/Assert/Assert/data_4',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/assert_equal/Assert/Assert',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/checked_cell_output',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/luong_attention/ExpandDims/dim',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/luong_attention/ExpandDims',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/luong_attention/MatMul/Enter',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/luong_attention/MatMul',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/luong_attention/Squeeze',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/Softmax',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/ExpandDims/dim',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/ExpandDims',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/MatMul/Enter',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/MatMul',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/Squeeze',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/concat_1/axis',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/concat_1',\n",
       " 'import/bidirectional_rnn_word_0/fw/attention_wrapper/attention_layer/kernel',\n",
       " 'import/bidirectional_rnn_word_0/fw/attention_wrapper/attention_layer/kernel/read',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/attention_layer/MatMul/Enter',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/attention_layer/MatMul',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/fw/attention_wrapper/concat_2',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/TensorArrayWrite/TensorArrayWriteV3/Enter',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/TensorArrayWrite/TensorArrayWriteV3',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/add_1/y',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/add_1',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/NextIteration',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/NextIteration_1',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/NextIteration_2',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/NextIteration_3',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/NextIteration_4',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/NextIteration_5',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/while/Exit_2',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/TensorArrayStack/TensorArraySizeV3',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/TensorArrayStack/range/start',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/TensorArrayStack/range/delta',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/TensorArrayStack/range',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/TensorArrayStack/TensorArrayGatherV3',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/Rank_1',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/range_1/start',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/range_1/delta',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/range_1',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/concat_2/values_0',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/concat_2/axis',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/concat_2',\n",
       " 'import/bidirectional_rnn_word_0/fw/fw/transpose_1',\n",
       " 'import/bidirectional_rnn_word_0/bw/ReverseV2/axis',\n",
       " 'import/bidirectional_rnn_word_0/bw/ReverseV2',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/Rank',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/range/start',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/range/delta',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/range',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/concat/values_0',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/concat/axis',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/concat',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/transpose',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/Shape',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/strided_slice/stack',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/strided_slice/stack_1',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/strided_slice/stack_2',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/strided_slice',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/AttentionWrapperZeroState/DropoutWrapperZeroState/LSTMCellZeroState/ExpandDims/dim',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/AttentionWrapperZeroState/DropoutWrapperZeroState/LSTMCellZeroState/ExpandDims',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/AttentionWrapperZeroState/DropoutWrapperZeroState/LSTMCellZeroState/Const',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/AttentionWrapperZeroState/DropoutWrapperZeroState/LSTMCellZeroState/concat/axis',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/AttentionWrapperZeroState/DropoutWrapperZeroState/LSTMCellZeroState/concat',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/AttentionWrapperZeroState/DropoutWrapperZeroState/LSTMCellZeroState/zeros/Const',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/AttentionWrapperZeroState/DropoutWrapperZeroState/LSTMCellZeroState/zeros',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/AttentionWrapperZeroState/DropoutWrapperZeroState/LSTMCellZeroState/ExpandDims_2/dim',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/AttentionWrapperZeroState/DropoutWrapperZeroState/LSTMCellZeroState/ExpandDims_2',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/AttentionWrapperZeroState/DropoutWrapperZeroState/LSTMCellZeroState/Const_2',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/AttentionWrapperZeroState/DropoutWrapperZeroState/LSTMCellZeroState/concat_1/axis',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/AttentionWrapperZeroState/DropoutWrapperZeroState/LSTMCellZeroState/concat_1',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/AttentionWrapperZeroState/DropoutWrapperZeroState/LSTMCellZeroState/zeros_1/Const',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/AttentionWrapperZeroState/DropoutWrapperZeroState/LSTMCellZeroState/zeros_1',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/AttentionWrapperZeroState/assert_equal/Equal',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/AttentionWrapperZeroState/assert_equal/Const',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/AttentionWrapperZeroState/assert_equal/All',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/AttentionWrapperZeroState/assert_equal/Assert/Assert/data_0',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/AttentionWrapperZeroState/assert_equal/Assert/Assert/data_1',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/AttentionWrapperZeroState/assert_equal/Assert/Assert/data_2',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/AttentionWrapperZeroState/assert_equal/Assert/Assert/data_4',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/AttentionWrapperZeroState/assert_equal/Assert/Assert',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/AttentionWrapperZeroState/checked_cell_state',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/AttentionWrapperZeroState/checked_cell_state_1',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/AttentionWrapperZeroState/ExpandDims_4/dim',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/AttentionWrapperZeroState/ExpandDims_4',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/AttentionWrapperZeroState/Const',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/AttentionWrapperZeroState/concat_1/axis',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/AttentionWrapperZeroState/concat_1',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/AttentionWrapperZeroState/zeros_2/Const',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/AttentionWrapperZeroState/zeros_2',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/Shape_1',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/strided_slice_1/stack',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/strided_slice_1/stack_1',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/strided_slice_1/stack_2',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/strided_slice_1',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/time',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/TensorArray',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/TensorArray_1',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/TensorArrayUnstack/Shape',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/TensorArrayUnstack/strided_slice/stack',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/TensorArrayUnstack/strided_slice/stack_1',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/TensorArrayUnstack/strided_slice/stack_2',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/TensorArrayUnstack/strided_slice',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/TensorArrayUnstack/range/start',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/TensorArrayUnstack/range/delta',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/TensorArrayUnstack/range',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/TensorArrayUnstack/TensorArrayScatter/TensorArrayScatterV3',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/Maximum/x',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/Maximum',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/Minimum',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/iteration_counter',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/Enter',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/Enter_1',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/Enter_2',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/Enter_3',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/Enter_4',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/Enter_5',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/Merge',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/Merge_1',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/Merge_2',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/Merge_3',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/Merge_4',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/Merge_5',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/Less/Enter',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/Less',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/Less_1/Enter',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/Less_1',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/LogicalAnd',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/LoopCond',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/Switch',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/Switch_1',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/Switch_2',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/Switch_3',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/Switch_4',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/Switch_5',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/Identity',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/Identity_1',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/Identity_2',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/Identity_3',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/Identity_4',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/Identity_5',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/add/y',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/add',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/TensorArrayReadV3/Enter',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/TensorArrayReadV3/Enter_1',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/TensorArrayReadV3',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/bw/attention_wrapper/concat/axis',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/bw/attention_wrapper/concat',\n",
       " 'import/bidirectional_rnn_word_0/bw/attention_wrapper/lstm_cell/kernel',\n",
       " 'import/bidirectional_rnn_word_0/bw/attention_wrapper/lstm_cell/kernel/read',\n",
       " 'import/bidirectional_rnn_word_0/bw/attention_wrapper/lstm_cell/bias',\n",
       " 'import/bidirectional_rnn_word_0/bw/attention_wrapper/lstm_cell/bias/read',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/bw/attention_wrapper/lstm_cell/concat/axis',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/bw/attention_wrapper/lstm_cell/concat',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/bw/attention_wrapper/lstm_cell/MatMul/Enter',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/bw/attention_wrapper/lstm_cell/MatMul',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/bw/attention_wrapper/lstm_cell/BiasAdd/Enter',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/bw/attention_wrapper/lstm_cell/BiasAdd',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/bw/attention_wrapper/lstm_cell/split/split_dim',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/bw/attention_wrapper/lstm_cell/split',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/bw/attention_wrapper/lstm_cell/add/y',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/bw/attention_wrapper/lstm_cell/add',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/bw/attention_wrapper/lstm_cell/Sigmoid',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/bw/attention_wrapper/lstm_cell/mul',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/bw/attention_wrapper/lstm_cell/Sigmoid_1',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/bw/attention_wrapper/lstm_cell/Tanh',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/bw/attention_wrapper/lstm_cell/mul_1',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/bw/attention_wrapper/lstm_cell/add_1',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/bw/attention_wrapper/lstm_cell/Sigmoid_2',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/bw/attention_wrapper/lstm_cell/Tanh_1',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/bw/attention_wrapper/lstm_cell/mul_2',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/bw/attention_wrapper/dropout/keep_prob',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/bw/attention_wrapper/dropout/Shape',\n",
       " 'import/bidirectional_rnn_word_0/bw/bw/while/bw/attention_wrapper/dropout/random_uniform/min',\n",
       " ...]"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "[n.name for n in g.as_graph_def().node]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "string = 'KUALA LUMPUR: Sempena sambutan Aidilfitri minggu depan, Perdana Menteri Tun Dr Mahathir Mohamad dan Menteri Pengangkutan Anthony Loke Siew Fook menitipkan pesanan khas kepada orang ramai yang mahu pulang ke kampung halaman masing-masing. Dalam video pendek terbitan Jabatan Keselamatan Jalan Raya (JKJR) itu, Dr Mahathir menasihati mereka supaya berhenti berehat dan tidur sebentar  sekiranya mengantuk ketika memandu.'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "def char_str_idx(corpus, dic, UNK = 0):\n",
    "    maxlen = max([len(i) for i in corpus])\n",
    "    X = np.zeros((len(corpus), maxlen))\n",
    "    for i in range(len(corpus)):\n",
    "        for no, k in enumerate(corpus[i][:maxlen][::-1]):\n",
    "            val = dic[k] if k in dic else UNK\n",
    "            X[i, -1 - no] = val\n",
    "    return X\n",
    "\n",
    "def generate_char_seq(batch, idx2word, char2idx):\n",
    "    x = [[len(idx2word[i]) for i in k] for k in batch]\n",
    "    maxlen = max([j for i in x for j in i])\n",
    "    temp = np.zeros((batch.shape[0], batch.shape[1], maxlen), dtype = np.int32)\n",
    "    for i in range(batch.shape[0]):\n",
    "        for k in range(batch.shape[1]):\n",
    "            for no, c in enumerate(idx2word[batch[i, k]].lower()):\n",
    "                temp[i, k, -1 - no] = char2idx[c]\n",
    "    return temp\n",
    "\n",
    "sequence = process_string(string.lower())\n",
    "X_seq = char_str_idx([sequence], word2idx, 2)\n",
    "X_char_seq = generate_char_seq(X_seq, idx2word, char2idx)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.5/dist-packages/tensorflow/python/client/session.py:1702: UserWarning: An interactive session is already active. This can cause out-of-memory errors in some cases. You must explicitly call `InteractiveSession.close()` to release resources held by the other session(s).\n",
      "  warnings.warn('An interactive session is already active. This can '\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "kuala location\n",
      "lumpur location\n",
      "sempena OTHER\n",
      "sambutan event\n",
      "aidilfitri event\n",
      "minggu OTHER\n",
      "depan OTHER\n",
      "perdana OTHER\n",
      "menteri person\n",
      "tun person\n",
      "dr person\n",
      "mahathir person\n",
      "mohamad person\n",
      "dan OTHER\n",
      "menteri organization\n",
      "pengangkutan OTHER\n",
      "anthony person\n",
      "loke person\n",
      "siew person\n",
      "fook person\n",
      "menitipkan OTHER\n",
      "pesanan OTHER\n",
      "khas OTHER\n",
      "kepada OTHER\n",
      "orang OTHER\n",
      "ramai OTHER\n",
      "yang OTHER\n",
      "mahu OTHER\n",
      "pulang OTHER\n",
      "ke OTHER\n",
      "kampung OTHER\n",
      "halaman OTHER\n",
      "masing-masing OTHER\n",
      "dalam OTHER\n",
      "video OTHER\n",
      "pendek OTHER\n",
      "terbitan OTHER\n",
      "jabatan organization\n",
      "keselamatan organization\n",
      "jalan organization\n",
      "raya person\n",
      "jkjr person\n",
      "itu OTHER\n",
      "dr person\n",
      "mahathir person\n",
      "menasihati OTHER\n",
      "mereka OTHER\n",
      "supaya OTHER\n",
      "berhenti OTHER\n",
      "berehat person\n",
      "dan OTHER\n",
      "tidur OTHER\n",
      "sebentar OTHER\n",
      "sekiranya OTHER\n",
      "mengantuk OTHER\n",
      "ketika OTHER\n",
      "memandu OTHER\n"
     ]
    }
   ],
   "source": [
    "word_ids = g.get_tensor_by_name('import/Placeholder:0')\n",
    "char_ids = g.get_tensor_by_name('import/Placeholder_1:0')\n",
    "tags_seq = g.get_tensor_by_name('import/logits:0')\n",
    "test_sess = tf.InteractiveSession(graph = g)\n",
    "predicted = test_sess.run(tags_seq,\n",
    "            feed_dict = {\n",
    "                word_ids: X_seq,\n",
    "                char_ids: X_char_seq,\n",
    "            })[0]\n",
    "\n",
    "for i in range(len(predicted)):\n",
    "    print(sequence[i],idx2tag[predicted[i]])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.5.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}