translate-keras.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import io\n",
    "import re\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "import itertools\n",
    "import pickle\n",
    "from collections import Counter, defaultdict\n",
    "import spacy\n",
    "from sklearn.model_selection import train_test_split\n",
    "from nltk.tokenize.treebank import TreebankWordDetokenizer as Detok\n",
    "from nltk.translate.bleu_score import corpus_bleu\n",
    "from sklearn.manifold import TSNE\n",
    "from sklearn.decomposition import PCA\n",
    "from scipy.spatial.distance import cosine\n",
    "\n",
    "import tensorflow as tf\n",
    "from keras.preprocessing.sequence import pad_sequences\n",
    "from keras.utils import to_categorical\n",
    "from keras import Input, layers\n",
    "from keras.activations import softmax\n",
    "from keras.models import Sequential, Model, load_model\n",
    "from keras.optimizers import Adam\n",
    "from keras.callbacks import ReduceLROnPlateau, EarlyStopping, ModelCheckpoint, Callback\n",
    "from keras import backend as K\n",
    "\n",
    "import multiprocessing\n",
    "from joblib import Parallel, delayed\n",
    "num_cores = multiprocessing.cpu_count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "%reload_ext autoreload\n",
    "%autoreload 2\n",
    "%config InlineBackend.figure_format = 'retina'\n",
    "%load_ext rpy2.ipython"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "data_path = '../data'\n",
    "model_path = '../model'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(154883, 2)"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Translation data\n",
    "# http://www.manythings.org/anki/\n",
    "with io.open(os.path.join(data_path, 'fra-eng/fra.txt'), encoding='utf-8') as f:\n",
    "    lines = f.read().split('\\n')\n",
    "\n",
    "pairs = [l.split('\\t') for l in lines if l != '']\n",
    "pairs = [[re.sub(u\"\\u202f|\\u2009\", \" \", s) for s in p] for p in pairs]\n",
    "pairs = [[re.sub(u\"\\u2019\", \"'\", s) for s in p] for p in pairs]\n",
    "\n",
    "# Convert to datafarme for ease of processing\n",
    "pairs = pd.DataFrame(pairs, columns=['en', 'fr'])\n",
    "pairs.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(154699, 2)"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# De-dupe the language pairs\n",
    "pairs['en'] = pairs['en'].str.lower().str.strip()\n",
    "pairs['fr'] = pairs['fr'].str.lower().str.strip()\n",
    "pairs.drop_duplicates(inplace=True)\n",
    "pairs.reset_index(drop=True, inplace=True)\n",
    "pairs.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "% of unique English sentences: 0.695731711259\n",
      "% of unique French sentences: 0.949605362672\n"
     ]
    }
   ],
   "source": [
    "print('% of unique English sentences: {}\\n% of unique French sentences: {}'.\n",
    "      format(1.0 * len(pairs['en'].unique()) / pairs.shape[0],\n",
    "             1.0 * len(pairs['fr'].unique()) / pairs.shape[0]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# Given that there are far more duplicates in English,\n",
    "# use French as the source and English as the target in translations"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Extract features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Tokenize\n",
    "spacy_tokenizer_en = spacy.load('en', disable=['tagger', 'parser', 'ner', 'textcat'])\n",
    "spacy_tokenizer_fr = spacy.load('fr', disable=['tagger', 'parser', 'ner', 'textcat'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def tokenize(msgs, tokenizer, batch_size=10000, n_threads=num_cores):\n",
    "    msgs = [unicode(msg.lower()) for msg in msgs]\n",
    "    toks = []\n",
    "    for msg in tokenizer.pipe(msgs, batch_size=batch_size, n_threads=n_threads):\n",
    "        toks.append([tok.text for tok in msg])\n",
    "    return toks"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>en</th>\n",
       "      <th>fr</th>\n",
       "      <th>toks_en</th>\n",
       "      <th>toks_fr</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>go .</td>\n",
       "      <td>va !</td>\n",
       "      <td>[go, .]</td>\n",
       "      <td>[va, !]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>run !</td>\n",
       "      <td>cours !</td>\n",
       "      <td>[run, !]</td>\n",
       "      <td>[cours, !]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>run !</td>\n",
       "      <td>courez !</td>\n",
       "      <td>[run, !]</td>\n",
       "      <td>[courez, !]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>fire !</td>\n",
       "      <td>au feu !</td>\n",
       "      <td>[fire, !]</td>\n",
       "      <td>[au, feu, !]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>help !</td>\n",
       "      <td>à l' aide !</td>\n",
       "      <td>[help, !]</td>\n",
       "      <td>[à, l', aide, !]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       en           fr    toks_en           toks_fr\n",
       "0    go .         va !    [go, .]           [va, !]\n",
       "1   run !      cours !   [run, !]        [cours, !]\n",
       "2   run !     courez !   [run, !]       [courez, !]\n",
       "3  fire !     au feu !  [fire, !]      [au, feu, !]\n",
       "4  help !  à l' aide !  [help, !]  [à, l', aide, !]"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Tokenize English sentences\n",
    "pairs['toks_en'] = tokenize(pairs['en'].tolist(), spacy_tokenizer_en)\n",
    "\n",
    "# Tokenize French sentences\n",
    "pairs['toks_fr'] = tokenize(pairs['fr'].tolist(), spacy_tokenizer_fr)\n",
    "\n",
    "pairs['en'] = [' '.join(s) for s in pairs['toks_en']]\n",
    "pairs['fr'] = [' '.join(s) for s in pairs['toks_fr']]\n",
    "pairs.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(153981, 4)"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# For ease of training, only keep sentences shorter than 20 words\n",
    "max_len = 20\n",
    "pairs = pairs[(pairs['toks_en'].str.len() <= max_len) & (pairs['toks_fr'].str.len() <= max_len)]\n",
    "pairs.reset_index(drop=True, inplace=True)\n",
    "pairs.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "146163"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Consolidate references\n",
    "# (Because one source can have multiple targets)\n",
    "ref_dict = defaultdict(list)\n",
    "for source, target in zip(pairs['fr'], pairs['en']):\n",
    "    ref_dict[source].append(target)\n",
    "len(ref_dict)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "trn    0.799514\n",
       "val    0.200486\n",
       "Name: grp, dtype: float64"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Split into training and validation sets based on the distinct English sentences\n",
    "# (because there are more duplicates in English)\n",
    "pairs_en_uni = pairs.drop_duplicates(subset=['en']).copy()\n",
    "\n",
    "# Shuffle\n",
    "pairs_en_uni = pairs_en_uni.sample(frac=1, replace=False, random_state=0)\n",
    "\n",
    "# Split\n",
    "test_size = 0.2\n",
    "pairs_en_uni['grp'] = 'trn'\n",
    "pairs_en_uni['grp'][:int(test_size * pairs_en_uni.shape[0])] = 'val'\n",
    "\n",
    "# Join back to the original, duplicated data\n",
    "pairs = pd.merge(pairs, pairs_en_uni[['en', 'grp']])\n",
    "pairs['grp'].value_counts(normalize=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(153981, 5)"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pairs.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(123110, 123110, 30871, 30871)"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Extract tokens\n",
    "X_trn = pairs.query(\"grp == 'trn'\")['toks_fr'].tolist()\n",
    "y_trn = pairs.query(\"grp == 'trn'\")['toks_en'].tolist()\n",
    "X_val = pairs.query(\"grp == 'val'\")['toks_fr'].tolist()\n",
    "y_val = pairs.query(\"grp == 'val'\")['toks_en'].tolist()\n",
    "\n",
    "len(X_trn), len(y_trn), len(X_val), len(y_val)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1025954, 927369)"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Create mappers (using training data)\n",
    "X_toks_trn = list(itertools.chain(*X_trn))\n",
    "y_toks_trn = list(itertools.chain(*y_trn))\n",
    "\n",
    "len(X_toks_trn), len(y_toks_trn)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "UNK = '_unk_'\n",
    "PAD = '_pad_'\n",
    "\n",
    "def create_mapper(toks, max_vocab=100000, min_freq=1, UNK=UNK, PAD=PAD):\n",
    "    \"\"\"Create mappers between tokens and numerical indices.\n",
    "\n",
    "    Parameters\n",
    "    ----------\n",
    "    toks : A list containing all raw tokens (before deduping).\n",
    "\n",
    "    max_vocab : The maximum vocabulary size.\n",
    "\n",
    "    min_freq : The minimum frequency for a token to be included in the vocabulary.\n",
    "\n",
    "    UNK : Special token for unknown word (default to '_unk_').\n",
    "\n",
    "    PAD : Special token for paddings (default to '_pad_').\n",
    "\n",
    "    Returns\n",
    "    -------\n",
    "    stoi : A dictionary that maps tokens to indices.\n",
    "\n",
    "    itos : A list that maps indices to tokens.\n",
    "    \"\"\"\n",
    "    \n",
    "    toks_freq = Counter(toks)\n",
    "    \n",
    "    itos = [s for s, c in toks_freq.most_common(max_vocab) if c >= min_freq]\n",
    "    \n",
    "    if PAD:\n",
    "        if UNK:\n",
    "            itos.insert(0, UNK)\n",
    "            itos.insert(0, PAD)  # Note the index for UNK is 1 and the index for PAD is 0\n",
    "\n",
    "            stoi = defaultdict(lambda: 1, {v: k for k, v in enumerate(itos)})\n",
    "        else:\n",
    "            itos.insert(0, PAD)  # Note the index for PAD is 0\n",
    "            stoi = {v: k for k, v in enumerate(itos)}\n",
    "    else:\n",
    "        if UNK:\n",
    "            itos.insert(0, UNK)  # Note the index for UNK is 0\n",
    "            stoi = defaultdict(lambda: 0, {v: k for k, v in enumerate(itos)})\n",
    "        else:\n",
    "            stoi = {v: k for k, v in enumerate(itos)}\n",
    "    \n",
    "    return stoi, itos"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(21553, 12543)"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_stoi, X_itos = create_mapper(X_toks_trn)\n",
    "y_stoi, y_itos = create_mapper(y_toks_trn)\n",
    "\n",
    "len(X_itos), len(y_itos)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['_pad_', '_unk_', u'.', u'je', u'de', u'?', u'pas', u'vous', u'est', u'que']"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_itos[:10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['_pad_', '_unk_', u'.', u'i', u'you', u'to', u'the', u'?', u'a', u\"n't\"]"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_itos[:10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Add GO (only needed for the target mappers)\n",
    "GO = '_go_'\n",
    "\n",
    "y_stoi[GO] = len(y_stoi)\n",
    "y_itos.append(GO)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Encode text features\n",
    "def encode_msg(toks, stoi, max_len, padding_pos='post', truncating='pre', padding_token=PAD):\n",
    "    # Index\n",
    "    toks_id = [[stoi[tok] for tok in msg] for msg in toks]\n",
    "    \n",
    "    # Pad\n",
    "    toks_id_pad = pad_sequences(toks_id, max_len, padding=padding_pos, truncating=truncating, value=stoi[padding_token])\n",
    "    \n",
    "    return toks_id_pad"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((123110, 20), (123110, 20), (30871, 20), (30871, 20))"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_ix_trn = encode_msg(X_trn, X_stoi, max_len)\n",
    "y_ix_trn = encode_msg(y_trn, y_stoi, max_len)\n",
    "X_ix_val = encode_msg(X_val, X_stoi, max_len)\n",
    "y_ix_val = encode_msg(y_val, y_stoi, max_len)\n",
    "\n",
    "X_ix_trn.shape, y_ix_trn.shape, X_ix_val.shape, y_ix_val.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array([u'va', u'!', u'_pad_', u'_pad_', u'_pad_', u'_pad_', u'_pad_',\n",
       "        u'_pad_', u'_pad_', u'_pad_', u'_pad_', u'_pad_', u'_pad_',\n",
       "        u'_pad_', u'_pad_', u'_pad_', u'_pad_', u'_pad_', u'_pad_',\n",
       "        u'_pad_'], dtype='<U22'),\n",
       " array([u'go', u'.', u'_pad_', u'_pad_', u'_pad_', u'_pad_', u'_pad_',\n",
       "        u'_pad_', u'_pad_', u'_pad_', u'_pad_', u'_pad_', u'_pad_',\n",
       "        u'_pad_', u'_pad_', u'_pad_', u'_pad_', u'_pad_', u'_pad_',\n",
       "        u'_pad_'], dtype='<U17'))"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Sanity check\n",
    "np.array(X_itos)[X_ix_trn[0]], np.array(y_itos)[y_ix_trn[0]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array([u'\\xe0', u\"l'\", u'aide', u'!', u'_pad_', u'_pad_', u'_pad_',\n",
       "        u'_pad_', u'_pad_', u'_pad_', u'_pad_', u'_pad_', u'_pad_',\n",
       "        u'_pad_', u'_pad_', u'_pad_', u'_pad_', u'_pad_', u'_pad_',\n",
       "        u'_pad_'], dtype='<U22'),\n",
       " array([u'help', u'!', u'_pad_', u'_pad_', u'_pad_', u'_pad_', u'_pad_',\n",
       "        u'_pad_', u'_pad_', u'_pad_', u'_pad_', u'_pad_', u'_pad_',\n",
       "        u'_pad_', u'_pad_', u'_pad_', u'_pad_', u'_pad_', u'_pad_',\n",
       "        u'_pad_'], dtype='<U17'))"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.array(X_itos)[X_ix_val[0]], np.array(y_itos)[y_ix_val[0]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Save processed data\n",
    "data = {\n",
    "    # Mappers\n",
    "    'X_stoi': X_stoi,\n",
    "    'X_itos': X_itos,\n",
    "    'y_stoi': y_stoi,\n",
    "    'y_itos': y_itos,\n",
    "    \n",
    "    # Original data\n",
    "    'X_trn': X_trn,\n",
    "    'X_val': X_val,\n",
    "    'y_trn': y_trn,\n",
    "    'y_val': y_val,\n",
    "    \n",
    "    # Processed data\n",
    "    'X_ix_trn': X_ix_trn,\n",
    "    'y_ix_trn': y_ix_trn,\n",
    "    'X_ix_val': X_ix_val,\n",
    "    'y_ix_val': y_ix_val,\n",
    "    \n",
    "    # Other\n",
    "    'ref_dict': ref_dict,\n",
    "    'pairs': pairs\n",
    "}\n",
    "\n",
    "pickle.dump(data, open(os.path.join(data_path, 'translate-data.pkl'), 'wb'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Load processed data (in a new session)\n",
    "data = pickle.load(open(os.path.join(data_path, 'translate-data.pkl'), 'rb'))\n",
    "\n",
    "X_stoi, X_itos, y_stoi, y_itos = data['X_stoi'], data['X_itos'], data['y_stoi'], data['y_itos']\n",
    "X_trn, X_val, y_trn, y_val = data['X_trn'], data['X_val'], data['y_trn'], data['y_val']\n",
    "X_ix_trn, X_ix_val, y_ix_trn, y_ix_val = data['X_ix_trn'], data['X_ix_val'], data['y_ix_trn'], data['y_ix_val']\n",
    "ref_dict = data['ref_dict']\n",
    "pairs = data['pairs']\n",
    "\n",
    "max_len = 20\n",
    "\n",
    "UNK = '_unk_'\n",
    "PAD = '_pad_'\n",
    "GO = '_go_'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Load pre-trained embeddings\n",
    "X_ft_embed = pickle.load(open(os.path.join(data_path, 'X_ft_embed.pkl'), 'rb'))\n",
    "y_ft_embed = pickle.load(open(os.path.join(data_path, 'y_ft_embed.pkl'), 'rb'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Create models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "K.clear_session()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Encoder"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Define hidden size\n",
    "hidden_size_enc = 256\n",
    "hidden_size_dec = 256"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Define layers\n",
    "txt_embed_size = 300\n",
    "\n",
    "X_embed_layer = layers.Embedding(input_dim=len(X_itos), output_dim=txt_embed_size, weights=[X_ft_embed], name='X_embed_layer')\n",
    "lstm_enc = layers.LSTM(hidden_size_enc, return_sequences=True, return_state=True, name='lstm_enc')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Input tensor\n",
    "X = Input(shape=(max_len, ), name='X')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Encode\n",
    "def encode(X):\n",
    "    X_emb = X_embed_layer(X)\n",
    "    a_enc, h_enc, c_enc = lstm_enc(X_emb)\n",
    "    return a_enc, h_enc, c_enc"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Decoder without attention"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Decoder"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Define layers\n",
    "y_embed_layer = layers.Embedding(input_dim=len(y_itos), output_dim=txt_embed_size, weights=[y_ft_embed], name='y_embed_layer')\n",
    "lstm_dec_wo_attn = layers.LSTM(hidden_size_dec, return_sequences=True, return_state=True, name='lstm_dec_wo_attn')\n",
    "densor_dec = layers.Dense(len(y_itos), activation='softmax', name='densor_dec')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Decoder input used in training (i.e., \"GO\" + `y` shifted by 1)\n",
    "in_dec = Input(shape=(max_len, ), name='in_dec')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def decode_wo_attn(in_dec, h_enc, c_enc):\n",
    "    in_dec_emb = y_embed_layer(in_dec)\n",
    "    \n",
    "    # Feed into decoder LSTM\n",
    "    a_dec, h_dec, c_dec = lstm_dec_wo_attn(in_dec_emb, initial_state=[h_enc, c_enc])\n",
    "    \n",
    "    # Feed the output into a dense layer with softmax\n",
    "    out_dec = densor_dec(a_dec)\n",
    "    return out_dec, h_dec, c_dec"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Put together"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Encode\n",
    "a_enc, h_enc, c_enc = encode(X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Decode\n",
    "out_dec, _, _ = decode_wo_attn(in_dec, h_enc, c_enc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "X (InputLayer)                  (None, 20)           0                                            \n",
      "__________________________________________________________________________________________________\n",
      "in_dec (InputLayer)             (None, 20)           0                                            \n",
      "__________________________________________________________________________________________________\n",
      "X_embed_layer (Embedding)       (None, 20, 300)      6465900     X[0][0]                          \n",
      "__________________________________________________________________________________________________\n",
      "y_embed_layer (Embedding)       (None, 20, 300)      3763200     in_dec[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "lstm_enc (LSTM)                 [(None, 20, 256), (N 570368      X_embed_layer[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "lstm_dec_wo_attn (LSTM)         [(None, 20, 256), (N 570368      y_embed_layer[0][0]              \n",
      "                                                                 lstm_enc[0][1]                   \n",
      "                                                                 lstm_enc[0][2]                   \n",
      "__________________________________________________________________________________________________\n",
      "densor_dec (Dense)              (None, 20, 12544)    3223808     lstm_dec_wo_attn[0][0]           \n",
      "==================================================================================================\n",
      "Total params: 14,593,644\n",
      "Trainable params: 14,593,644\n",
      "Non-trainable params: 0\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# Put together\n",
    "model_wo_attn = Model(inputs=[X, in_dec], outputs=out_dec)\n",
    "model_wo_attn.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "model_wo_attn.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['sparse_categorical_accuracy'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Decoder with attention"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Attention layer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Define layers\n",
    "attn_repeator = layers.RepeatVector(max_len, name='attn_repeator')\n",
    "attn_concatenator = layers.Concatenate(name='attn_concatenator')\n",
    "attn_weights_computer = layers.Dense(1, activation='tanh', name='attn_weights_computer')\n",
    "\n",
    "def softmax_axis_1(x):\n",
    "    return softmax(x, axis=1)\n",
    "attn_weights_normalizer = layers.Activation(softmax_axis_1, name='attn_weights_normalizer')\n",
    "\n",
    "attn_context_computer = layers.Dot(axes=1, name='attn_context_computer')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def one_step_attn(a_enc, h_dec_prev):\n",
    "    # Repeat `h_dec_prev`\n",
    "    h_dec_prev_rep = attn_repeator(h_dec_prev)  # (None, timesteps, hidden_size_dec)\n",
    "    \n",
    "    # Compute the attention weights\n",
    "    h_dec_prev_w_a_enc = attn_concatenator([h_dec_prev_rep, a_enc])  # (None, timesteps, hidden_size_dec + hidden_size_enc)\n",
    "    attn_weights = attn_weights_computer(h_dec_prev_w_a_enc)  # (None, timesteps, 1)\n",
    "    \n",
    "    # Normalize all the weights using a softmax (the normalization is applied in the timestep's axis)\n",
    "    attn_weights_norm = attn_weights_normalizer(attn_weights)  # (None, timesteps, 1)\n",
    "    \n",
    "    # Use the normalized weights to compute the weighted-average encoder outputs (i.e., the context)\n",
    "    attn_context = attn_context_computer([attn_weights_norm, a_enc])  # (None, 1, hidden_size_dec)\n",
    "    \n",
    "    return attn_context, attn_weights_norm"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Decoder"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "y_embed_layer = layers.Embedding(input_dim=len(y_itos), output_dim=txt_embed_size, weights=[y_ft_embed], name='y_embed_layer')\n",
    "context_input_concatenator = layers.Concatenate(name='context_input_concatenator')\n",
    "densor_dec_0 = layers.Dense(hidden_size_dec, name='densor_dec_0')\n",
    "lstm_dec_w_attn = layers.LSTM(hidden_size_dec, return_sequences=False, return_state=True, name='lstm_dec_w_attn')\n",
    "densor_dec_1 = layers.Dense(len(y_itos), activation='softmax', name='densor_dec_1')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# One-step decoder\n",
    "def decode_w_attn(in_dec, a_enc, h_dec_prev, c_dec_prev):\n",
    "    # Compute context vector using the attention layer\n",
    "    attn_context, attn_weights_norm = one_step_attn(a_enc, h_dec_prev)\n",
    "    \n",
    "    # Concatenate with the decoder input\n",
    "    in_dec_embed = y_embed_layer(in_dec)\n",
    "    attn_context_w_input = context_input_concatenator([attn_context, in_dec_embed])\n",
    "    \n",
    "    # Reduce dimension to the original hidden size\n",
    "    attn_context_w_input = densor_dec_0(attn_context_w_input)\n",
    "    \n",
    "    # Feed into decoder LSTM\n",
    "    a_dec, h_dec, c_dec = lstm_dec_w_attn(attn_context_w_input, initial_state=[h_dec_prev, c_dec_prev])\n",
    "    \n",
    "    # Feed the output into a dense layer with softmax\n",
    "    out_dec = densor_dec_1(a_dec)\n",
    "    return out_dec, h_dec, c_dec, attn_weights_norm"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###  Put together"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Encode\n",
    "a_enc, h_enc, c_enc = encode(X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Decode\n",
    "in_dec_0 = Input(shape=(1, ), name='in_dec_0')  # Decoder starting token (\"_go_\")\n",
    "y = Input(shape=(max_len, ), name='y')\n",
    "y_slicer = layers.Lambda(lambda x: x[:, i:i+1], name='y_slicer')\n",
    "\n",
    "h_dec_prev, c_dec_prev = h_enc, c_enc\n",
    "in_dec = in_dec_0\n",
    "outs_dec = []\n",
    "\n",
    "for i in range(max_len):\n",
    "    out_dec, h_dec, c_dec, _ = decode_w_attn(in_dec, a_enc, h_dec_prev, c_dec_prev)\n",
    "    outs_dec.append(out_dec)\n",
    "    h_dec_prev, c_dec_prev = h_dec, c_dec\n",
    "    in_dec = y_slicer(y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "X (InputLayer)                  (None, 20)           0                                            \n",
      "__________________________________________________________________________________________________\n",
      "X_embed_layer (Embedding)       (None, 20, 300)      6465900     X[0][0]                          \n",
      "__________________________________________________________________________________________________\n",
      "lstm_enc (LSTM)                 [(None, 20, 256), (N 570368      X_embed_layer[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "attn_repeator (RepeatVector)    (None, 20, 256)      0           lstm_enc[0][1]                   \n",
      "                                                                 lstm_dec_w_attn[0][1]            \n",
      "                                                                 lstm_dec_w_attn[1][1]            \n",
      "                                                                 lstm_dec_w_attn[2][1]            \n",
      "                                                                 lstm_dec_w_attn[3][1]            \n",
      "                                                                 lstm_dec_w_attn[4][1]            \n",
      "                                                                 lstm_dec_w_attn[5][1]            \n",
      "                                                                 lstm_dec_w_attn[6][1]            \n",
      "                                                                 lstm_dec_w_attn[7][1]            \n",
      "                                                                 lstm_dec_w_attn[8][1]            \n",
      "                                                                 lstm_dec_w_attn[9][1]            \n",
      "                                                                 lstm_dec_w_attn[10][1]           \n",
      "                                                                 lstm_dec_w_attn[11][1]           \n",
      "                                                                 lstm_dec_w_attn[12][1]           \n",
      "                                                                 lstm_dec_w_attn[13][1]           \n",
      "                                                                 lstm_dec_w_attn[14][1]           \n",
      "                                                                 lstm_dec_w_attn[15][1]           \n",
      "                                                                 lstm_dec_w_attn[16][1]           \n",
      "                                                                 lstm_dec_w_attn[17][1]           \n",
      "                                                                 lstm_dec_w_attn[18][1]           \n",
      "__________________________________________________________________________________________________\n",
      "attn_concatenator (Concatenate) (None, 20, 512)      0           attn_repeator[0][0]              \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_repeator[1][0]              \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_repeator[2][0]              \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_repeator[3][0]              \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_repeator[4][0]              \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_repeator[5][0]              \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_repeator[6][0]              \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_repeator[7][0]              \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_repeator[8][0]              \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_repeator[9][0]              \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_repeator[10][0]             \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_repeator[11][0]             \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_repeator[12][0]             \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_repeator[13][0]             \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_repeator[14][0]             \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_repeator[15][0]             \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_repeator[16][0]             \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_repeator[17][0]             \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_repeator[18][0]             \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_repeator[19][0]             \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "attn_weights_computer (Dense)   (None, 20, 1)        513         attn_concatenator[0][0]          \n",
      "                                                                 attn_concatenator[1][0]          \n",
      "                                                                 attn_concatenator[2][0]          \n",
      "                                                                 attn_concatenator[3][0]          \n",
      "                                                                 attn_concatenator[4][0]          \n",
      "                                                                 attn_concatenator[5][0]          \n",
      "                                                                 attn_concatenator[6][0]          \n",
      "                                                                 attn_concatenator[7][0]          \n",
      "                                                                 attn_concatenator[8][0]          \n",
      "                                                                 attn_concatenator[9][0]          \n",
      "                                                                 attn_concatenator[10][0]         \n",
      "                                                                 attn_concatenator[11][0]         \n",
      "                                                                 attn_concatenator[12][0]         \n",
      "                                                                 attn_concatenator[13][0]         \n",
      "                                                                 attn_concatenator[14][0]         \n",
      "                                                                 attn_concatenator[15][0]         \n",
      "                                                                 attn_concatenator[16][0]         \n",
      "                                                                 attn_concatenator[17][0]         \n",
      "                                                                 attn_concatenator[18][0]         \n",
      "                                                                 attn_concatenator[19][0]         \n",
      "__________________________________________________________________________________________________\n",
      "attn_weights_normalizer (Activa (None, 20, 1)        0           attn_weights_computer[0][0]      \n",
      "                                                                 attn_weights_computer[1][0]      \n",
      "                                                                 attn_weights_computer[2][0]      \n",
      "                                                                 attn_weights_computer[3][0]      \n",
      "                                                                 attn_weights_computer[4][0]      \n",
      "                                                                 attn_weights_computer[5][0]      \n",
      "                                                                 attn_weights_computer[6][0]      \n",
      "                                                                 attn_weights_computer[7][0]      \n",
      "                                                                 attn_weights_computer[8][0]      \n",
      "                                                                 attn_weights_computer[9][0]      \n",
      "                                                                 attn_weights_computer[10][0]     \n",
      "                                                                 attn_weights_computer[11][0]     \n",
      "                                                                 attn_weights_computer[12][0]     \n",
      "                                                                 attn_weights_computer[13][0]     \n",
      "                                                                 attn_weights_computer[14][0]     \n",
      "                                                                 attn_weights_computer[15][0]     \n",
      "                                                                 attn_weights_computer[16][0]     \n",
      "                                                                 attn_weights_computer[17][0]     \n",
      "                                                                 attn_weights_computer[18][0]     \n",
      "                                                                 attn_weights_computer[19][0]     \n",
      "__________________________________________________________________________________________________\n",
      "in_dec_0 (InputLayer)           (None, 1)            0                                            \n",
      "__________________________________________________________________________________________________\n",
      "attn_context_computer (Dot)     (None, 1, 256)       0           attn_weights_normalizer[0][0]    \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_weights_normalizer[1][0]    \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_weights_normalizer[2][0]    \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_weights_normalizer[3][0]    \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_weights_normalizer[4][0]    \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_weights_normalizer[5][0]    \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_weights_normalizer[6][0]    \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_weights_normalizer[7][0]    \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_weights_normalizer[8][0]    \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_weights_normalizer[9][0]    \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_weights_normalizer[10][0]   \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_weights_normalizer[11][0]   \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_weights_normalizer[12][0]   \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_weights_normalizer[13][0]   \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_weights_normalizer[14][0]   \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_weights_normalizer[15][0]   \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_weights_normalizer[16][0]   \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_weights_normalizer[17][0]   \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_weights_normalizer[18][0]   \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "                                                                 attn_weights_normalizer[19][0]   \n",
      "                                                                 lstm_enc[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "y_embed_layer (Embedding)       (None, 1, 300)       3763200     in_dec_0[0][0]                   \n",
      "                                                                 y_slicer[0][0]                   \n",
      "                                                                 y_slicer[1][0]                   \n",
      "                                                                 y_slicer[2][0]                   \n",
      "                                                                 y_slicer[3][0]                   \n",
      "                                                                 y_slicer[4][0]                   \n",
      "                                                                 y_slicer[5][0]                   \n",
      "                                                                 y_slicer[6][0]                   \n",
      "                                                                 y_slicer[7][0]                   \n",
      "                                                                 y_slicer[8][0]                   \n",
      "                                                                 y_slicer[9][0]                   \n",
      "                                                                 y_slicer[10][0]                  \n",
      "                                                                 y_slicer[11][0]                  \n",
      "                                                                 y_slicer[12][0]                  \n",
      "                                                                 y_slicer[13][0]                  \n",
      "                                                                 y_slicer[14][0]                  \n",
      "                                                                 y_slicer[15][0]                  \n",
      "                                                                 y_slicer[16][0]                  \n",
      "                                                                 y_slicer[17][0]                  \n",
      "                                                                 y_slicer[18][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "context_input_concatenator (Con (None, 1, 556)       0           attn_context_computer[0][0]      \n",
      "                                                                 y_embed_layer[0][0]              \n",
      "                                                                 attn_context_computer[1][0]      \n",
      "                                                                 y_embed_layer[1][0]              \n",
      "                                                                 attn_context_computer[2][0]      \n",
      "                                                                 y_embed_layer[2][0]              \n",
      "                                                                 attn_context_computer[3][0]      \n",
      "                                                                 y_embed_layer[3][0]              \n",
      "                                                                 attn_context_computer[4][0]      \n",
      "                                                                 y_embed_layer[4][0]              \n",
      "                                                                 attn_context_computer[5][0]      \n",
      "                                                                 y_embed_layer[5][0]              \n",
      "                                                                 attn_context_computer[6][0]      \n",
      "                                                                 y_embed_layer[6][0]              \n",
      "                                                                 attn_context_computer[7][0]      \n",
      "                                                                 y_embed_layer[7][0]              \n",
      "                                                                 attn_context_computer[8][0]      \n",
      "                                                                 y_embed_layer[8][0]              \n",
      "                                                                 attn_context_computer[9][0]      \n",
      "                                                                 y_embed_layer[9][0]              \n",
      "                                                                 attn_context_computer[10][0]     \n",
      "                                                                 y_embed_layer[10][0]             \n",
      "                                                                 attn_context_computer[11][0]     \n",
      "                                                                 y_embed_layer[11][0]             \n",
      "                                                                 attn_context_computer[12][0]     \n",
      "                                                                 y_embed_layer[12][0]             \n",
      "                                                                 attn_context_computer[13][0]     \n",
      "                                                                 y_embed_layer[13][0]             \n",
      "                                                                 attn_context_computer[14][0]     \n",
      "                                                                 y_embed_layer[14][0]             \n",
      "                                                                 attn_context_computer[15][0]     \n",
      "                                                                 y_embed_layer[15][0]             \n",
      "                                                                 attn_context_computer[16][0]     \n",
      "                                                                 y_embed_layer[16][0]             \n",
      "                                                                 attn_context_computer[17][0]     \n",
      "                                                                 y_embed_layer[17][0]             \n",
      "                                                                 attn_context_computer[18][0]     \n",
      "                                                                 y_embed_layer[18][0]             \n",
      "                                                                 attn_context_computer[19][0]     \n",
      "                                                                 y_embed_layer[19][0]             \n",
      "__________________________________________________________________________________________________\n",
      "densor_dec_0 (Dense)            (None, 1, 256)       142592      context_input_concatenator[0][0] \n",
      "                                                                 context_input_concatenator[1][0] \n",
      "                                                                 context_input_concatenator[2][0] \n",
      "                                                                 context_input_concatenator[3][0] \n",
      "                                                                 context_input_concatenator[4][0] \n",
      "                                                                 context_input_concatenator[5][0] \n",
      "                                                                 context_input_concatenator[6][0] \n",
      "                                                                 context_input_concatenator[7][0] \n",
      "                                                                 context_input_concatenator[8][0] \n",
      "                                                                 context_input_concatenator[9][0] \n",
      "                                                                 context_input_concatenator[10][0]\n",
      "                                                                 context_input_concatenator[11][0]\n",
      "                                                                 context_input_concatenator[12][0]\n",
      "                                                                 context_input_concatenator[13][0]\n",
      "                                                                 context_input_concatenator[14][0]\n",
      "                                                                 context_input_concatenator[15][0]\n",
      "                                                                 context_input_concatenator[16][0]\n",
      "                                                                 context_input_concatenator[17][0]\n",
      "                                                                 context_input_concatenator[18][0]\n",
      "                                                                 context_input_concatenator[19][0]\n",
      "__________________________________________________________________________________________________\n",
      "lstm_dec_w_attn (LSTM)          [(None, 256), (None, 525312      densor_dec_0[0][0]               \n",
      "                                                                 lstm_enc[0][1]                   \n",
      "                                                                 lstm_enc[0][2]                   \n",
      "                                                                 densor_dec_0[1][0]               \n",
      "                                                                 lstm_dec_w_attn[0][1]            \n",
      "                                                                 lstm_dec_w_attn[0][2]            \n",
      "                                                                 densor_dec_0[2][0]               \n",
      "                                                                 lstm_dec_w_attn[1][1]            \n",
      "                                                                 lstm_dec_w_attn[1][2]            \n",
      "                                                                 densor_dec_0[3][0]               \n",
      "                                                                 lstm_dec_w_attn[2][1]            \n",
      "                                                                 lstm_dec_w_attn[2][2]            \n",
      "                                                                 densor_dec_0[4][0]               \n",
      "                                                                 lstm_dec_w_attn[3][1]            \n",
      "                                                                 lstm_dec_w_attn[3][2]            \n",
      "                                                                 densor_dec_0[5][0]               \n",
      "                                                                 lstm_dec_w_attn[4][1]            \n",
      "                                                                 lstm_dec_w_attn[4][2]            \n",
      "                                                                 densor_dec_0[6][0]               \n",
      "                                                                 lstm_dec_w_attn[5][1]            \n",
      "                                                                 lstm_dec_w_attn[5][2]            \n",
      "                                                                 densor_dec_0[7][0]               \n",
      "                                                                 lstm_dec_w_attn[6][1]            \n",
      "                                                                 lstm_dec_w_attn[6][2]            \n",
      "                                                                 densor_dec_0[8][0]               \n",
      "                                                                 lstm_dec_w_attn[7][1]            \n",
      "                                                                 lstm_dec_w_attn[7][2]            \n",
      "                                                                 densor_dec_0[9][0]               \n",
      "                                                                 lstm_dec_w_attn[8][1]            \n",
      "                                                                 lstm_dec_w_attn[8][2]            \n",
      "                                                                 densor_dec_0[10][0]              \n",
      "                                                                 lstm_dec_w_attn[9][1]            \n",
      "                                                                 lstm_dec_w_attn[9][2]            \n",
      "                                                                 densor_dec_0[11][0]              \n",
      "                                                                 lstm_dec_w_attn[10][1]           \n",
      "                                                                 lstm_dec_w_attn[10][2]           \n",
      "                                                                 densor_dec_0[12][0]              \n",
      "                                                                 lstm_dec_w_attn[11][1]           \n",
      "                                                                 lstm_dec_w_attn[11][2]           \n",
      "                                                                 densor_dec_0[13][0]              \n",
      "                                                                 lstm_dec_w_attn[12][1]           \n",
      "                                                                 lstm_dec_w_attn[12][2]           \n",
      "                                                                 densor_dec_0[14][0]              \n",
      "                                                                 lstm_dec_w_attn[13][1]           \n",
      "                                                                 lstm_dec_w_attn[13][2]           \n",
      "                                                                 densor_dec_0[15][0]              \n",
      "                                                                 lstm_dec_w_attn[14][1]           \n",
      "                                                                 lstm_dec_w_attn[14][2]           \n",
      "                                                                 densor_dec_0[16][0]              \n",
      "                                                                 lstm_dec_w_attn[15][1]           \n",
      "                                                                 lstm_dec_w_attn[15][2]           \n",
      "                                                                 densor_dec_0[17][0]              \n",
      "                                                                 lstm_dec_w_attn[16][1]           \n",
      "                                                                 lstm_dec_w_attn[16][2]           \n",
      "                                                                 densor_dec_0[18][0]              \n",
      "                                                                 lstm_dec_w_attn[17][1]           \n",
      "                                                                 lstm_dec_w_attn[17][2]           \n",
      "                                                                 densor_dec_0[19][0]              \n",
      "                                                                 lstm_dec_w_attn[18][1]           \n",
      "                                                                 lstm_dec_w_attn[18][2]           \n",
      "__________________________________________________________________________________________________\n",
      "y (InputLayer)                  (None, 20)           0                                            \n",
      "__________________________________________________________________________________________________\n",
      "y_slicer (Lambda)               (None, 1)            0           y[0][0]                          \n",
      "                                                                 y[0][0]                          \n",
      "                                                                 y[0][0]                          \n",
      "                                                                 y[0][0]                          \n",
      "                                                                 y[0][0]                          \n",
      "                                                                 y[0][0]                          \n",
      "                                                                 y[0][0]                          \n",
      "                                                                 y[0][0]                          \n",
      "                                                                 y[0][0]                          \n",
      "                                                                 y[0][0]                          \n",
      "                                                                 y[0][0]                          \n",
      "                                                                 y[0][0]                          \n",
      "                                                                 y[0][0]                          \n",
      "                                                                 y[0][0]                          \n",
      "                                                                 y[0][0]                          \n",
      "                                                                 y[0][0]                          \n",
      "                                                                 y[0][0]                          \n",
      "                                                                 y[0][0]                          \n",
      "                                                                 y[0][0]                          \n",
      "__________________________________________________________________________________________________\n",
      "densor_dec_1 (Dense)            (None, 12544)        3223808     lstm_dec_w_attn[0][0]            \n",
      "                                                                 lstm_dec_w_attn[1][0]            \n",
      "                                                                 lstm_dec_w_attn[2][0]            \n",
      "                                                                 lstm_dec_w_attn[3][0]            \n",
      "                                                                 lstm_dec_w_attn[4][0]            \n",
      "                                                                 lstm_dec_w_attn[5][0]            \n",
      "                                                                 lstm_dec_w_attn[6][0]            \n",
      "                                                                 lstm_dec_w_attn[7][0]            \n",
      "                                                                 lstm_dec_w_attn[8][0]            \n",
      "                                                                 lstm_dec_w_attn[9][0]            \n",
      "                                                                 lstm_dec_w_attn[10][0]           \n",
      "                                                                 lstm_dec_w_attn[11][0]           \n",
      "                                                                 lstm_dec_w_attn[12][0]           \n",
      "                                                                 lstm_dec_w_attn[13][0]           \n",
      "                                                                 lstm_dec_w_attn[14][0]           \n",
      "                                                                 lstm_dec_w_attn[15][0]           \n",
      "                                                                 lstm_dec_w_attn[16][0]           \n",
      "                                                                 lstm_dec_w_attn[17][0]           \n",
      "                                                                 lstm_dec_w_attn[18][0]           \n",
      "                                                                 lstm_dec_w_attn[19][0]           \n",
      "==================================================================================================\n",
      "Total params: 14,691,693\n",
      "Trainable params: 14,691,693\n",
      "Non-trainable params: 0\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# Put together\n",
    "model_w_attn = Model(inputs=[X, in_dec_0, y], outputs=outs_dec)\n",
    "model_w_attn.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "model_w_attn.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['sparse_categorical_accuracy'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Train models"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Model without attention"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((123110, 20), (30871, 20))"
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Initialize decoder input\n",
    "in_dec_trn = np.hstack([np.array([y_stoi[GO]] * len(X_ix_trn)).reshape(-1, 1), y_ix_trn[:, :-1]])\n",
    "in_dec_val = np.hstack([np.array([y_stoi[GO]] * len(X_ix_val)).reshape(-1, 1), y_ix_val[:, :-1]])\n",
    "\n",
    "in_dec_trn.shape, in_dec_val.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((123110, 20, 1), (30871, 20, 1))"
      ]
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Reshape outputs in order to use `sparse_categorical_crossentropy`\n",
    "y_ix_trn_rs = y_ix_trn.reshape(y_ix_trn.shape[0], y_ix_trn.shape[1], 1)\n",
    "y_ix_val_rs = y_ix_val.reshape(y_ix_val.shape[0], y_ix_val.shape[1], 1)\n",
    "\n",
    "y_ix_trn_rs.shape, y_ix_val_rs.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Add callbacks\n",
    "callbacks = []\n",
    "\n",
    "reduce_lr = ReduceLROnPlateau(monitor='val_loss', factor=0.1, patience=5, verbose=1)\n",
    "callbacks.append(reduce_lr)\n",
    "\n",
    "stopper = EarlyStopping(monitor='val_loss', patience=5, verbose=1)\n",
    "callbacks.append(stopper)\n",
    "\n",
    "checkpoint = ModelCheckpoint(filepath=os.path.join(model_path, 'translate-keras-wo_attn.h5'), monitor='val_loss', save_best_only=True, verbose=1)\n",
    "callbacks.append(checkpoint)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# Train\n",
    "hist_wo_attn = model_wo_attn.fit(\n",
    "    x=[X_ix_trn, in_dec_trn],\n",
    "    y=y_ix_trn_rs,\n",
    "    batch_size=64,\n",
    "    epochs=20,\n",
    "    validation_data=([X_ix_val, in_dec_val], y_ix_val_rs),\n",
    "    shuffle=True,\n",
    "    callbacks=callbacks)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def plot_losses(model_hist):\n",
    "    loss = pd.DataFrame(zip(model_hist.history['loss'], model_hist.history['val_loss']), columns=['loss', 'val_loss'])\n",
    "    \n",
    "    # Find the epoch with the smallest loss\n",
    "    min_loss_ix = np.array(model_hist.history['val_loss']).argmin()\n",
    "\n",
    "    # Plot\n",
    "    ax = loss.plot()\n",
    "    ax.axvline(x=min_loss_ix, color='black', ls='--')\n",
    "\n",
    "    return loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>loss</th>\n",
       "      <th>val_loss</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1.836616</td>\n",
       "      <td>1.353840</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1.078551</td>\n",
       "      <td>0.937422</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.732719</td>\n",
       "      <td>0.761308</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.532355</td>\n",
       "      <td>0.674843</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.404407</td>\n",
       "      <td>0.637921</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>0.316250</td>\n",
       "      <td>0.623148</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>0.252256</td>\n",
       "      <td>0.624716</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>0.205485</td>\n",
       "      <td>0.630465</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>0.170112</td>\n",
       "      <td>0.639769</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>0.143448</td>\n",
       "      <td>0.652620</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>0.122098</td>\n",
       "      <td>0.668480</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        loss  val_loss\n",
       "0   1.836616  1.353840\n",
       "1   1.078551  0.937422\n",
       "2   0.732719  0.761308\n",
       "3   0.532355  0.674843\n",
       "4   0.404407  0.637921\n",
       "5   0.316250  0.623148\n",
       "6   0.252256  0.624716\n",
       "7   0.205485  0.630465\n",
       "8   0.170112  0.639769\n",
       "9   0.143448  0.652620\n",
       "10  0.122098  0.668480"
      ]
     },
     "execution_count": 47,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAvgAAAH0CAYAAABICFkFAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nayDb+PUAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "image/png": {
       "height": 250,
       "width": 380
      }
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_losses(hist_wo_attn)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Model with attention"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((123110, 1), (30871, 1))"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Initialize decoder input (\"_go_\")\n",
    "in_dec_0_trn = np.array([y_stoi[GO]] * len(X_ix_trn)).reshape(-1, 1)\n",
    "in_dec_0_val = np.array([y_stoi[GO]] * len(X_ix_val)).reshape(-1, 1)\n",
    "\n",
    "in_dec_0_trn.shape, in_dec_0_val.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((20, 123110, 1), (20, 30871, 1))"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Reshape outputs in order to use `sparse_categorical_crossentropy`\n",
    "y_ix_trn_rs = [y_ix_trn[:, i:i+1] for i in range(y_ix_trn.shape[1])]\n",
    "y_ix_val_rs = [y_ix_val[:, i:i+1] for i in range(y_ix_val.shape[1])]\n",
    "\n",
    "np.array(y_ix_trn_rs).shape, np.array(y_ix_val_rs).shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Add callbacks\n",
    "callbacks = []\n",
    "\n",
    "reduce_lr = ReduceLROnPlateau(monitor='val_loss', factor=0.1, patience=5, verbose=1)\n",
    "callbacks.append(reduce_lr)\n",
    "\n",
    "stopper = EarlyStopping(monitor='val_loss', patience=5, verbose=1)\n",
    "callbacks.append(stopper)\n",
    "\n",
    "checkpoint = ModelCheckpoint(filepath=os.path.join(model_path, 'translate-keras-w_attn.h5'), monitor='val_loss', save_best_only=True, verbose=1)\n",
    "callbacks.append(checkpoint)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Train\n",
    "hist_w_attn = model_w_attn.fit(\n",
    "    x=[X_ix_trn, in_dec_0_trn, y_ix_trn],\n",
    "    y=y_ix_trn_rs,\n",
    "    batch_size=64,\n",
    "    epochs=20,\n",
    "    validation_data=([X_ix_val, in_dec_0_val, y_ix_val], y_ix_val_rs),\n",
    "    shuffle=True,\n",
    "    callbacks=callbacks)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>loss</th>\n",
       "      <th>val_loss</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>37.668571</td>\n",
       "      <td>28.986769</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>22.782615</td>\n",
       "      <td>19.546392</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>15.469534</td>\n",
       "      <td>15.750059</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>11.320483</td>\n",
       "      <td>13.878397</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>8.695356</td>\n",
       "      <td>13.056421</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>6.884371</td>\n",
       "      <td>12.638755</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>5.564721</td>\n",
       "      <td>12.534421</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>4.592229</td>\n",
       "      <td>12.563303</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>3.865414</td>\n",
       "      <td>12.740409</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>3.297847</td>\n",
       "      <td>12.859746</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>2.851011</td>\n",
       "      <td>13.078931</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>2.488212</td>\n",
       "      <td>13.375171</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         loss   val_loss\n",
       "0   37.668571  28.986769\n",
       "1   22.782615  19.546392\n",
       "2   15.469534  15.750059\n",
       "3   11.320483  13.878397\n",
       "4    8.695356  13.056421\n",
       "5    6.884371  12.638755\n",
       "6    5.564721  12.534421\n",
       "7    4.592229  12.563303\n",
       "8    3.865414  12.740409\n",
       "9    3.297847  12.859746\n",
       "10   2.851011  13.078931\n",
       "11   2.488212  13.375171"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAuUAAAH0CAYAAACaWFNdAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nHvv/JvRK2vYnpr4AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "image/png": {
       "height": 250,
       "width": 370
      }
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_losses(hist_w_attn)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Inference"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Model without attention"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Reload the last checkpoint\n",
    "model_wo_attn = load_model(os.path.join(model_path, 'translate-keras-wo_attn.h5'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "X (InputLayer)               (None, 20)                0         \n",
      "_________________________________________________________________\n",
      "X_embed_layer (Embedding)    (None, 20, 300)           6465900   \n",
      "_________________________________________________________________\n",
      "lstm_enc (LSTM)              [(None, 20, 256), (None,  570368    \n",
      "=================================================================\n",
      "Total params: 7,036,268\n",
      "Trainable params: 7,036,268\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# Extract encoder\n",
    "X = Input(shape=(max_len, ), name='X')\n",
    "X_embed_layer = model_wo_attn.get_layer('X_embed_layer')\n",
    "lstm_enc = model_wo_attn.get_layer('lstm_enc')\n",
    "\n",
    "encoder = Model(X, encode(X))\n",
    "encoder.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "in_dec (InputLayer)             (None, 1)            0                                            \n",
      "__________________________________________________________________________________________________\n",
      "y_embed_layer (Embedding)       multiple             3763200     in_dec[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "h_dec_prev (InputLayer)         (None, 256)          0                                            \n",
      "__________________________________________________________________________________________________\n",
      "c_dec_prev (InputLayer)         (None, 256)          0                                            \n",
      "__________________________________________________________________________________________________\n",
      "lstm_dec_wo_attn (LSTM)         multiple             570368      y_embed_layer[1][0]              \n",
      "                                                                 h_dec_prev[0][0]                 \n",
      "                                                                 c_dec_prev[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "densor_dec (Dense)              multiple             3223808     lstm_dec_wo_attn[1][0]           \n",
      "==================================================================================================\n",
      "Total params: 7,557,376\n",
      "Trainable params: 7,557,376\n",
      "Non-trainable params: 0\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# Extract decoder\n",
    "in_dec = Input(shape=(1, ), name='in_dec')\n",
    "h_dec_prev = Input(shape=(hidden_size_dec, ), name='h_dec_prev')\n",
    "c_dec_prev = Input(shape=(hidden_size_dec, ), name='c_dec_prev')\n",
    "y_embed_layer = model_wo_attn.get_layer('y_embed_layer')\n",
    "lstm_dec_wo_attn = model_wo_attn.get_layer('lstm_dec_wo_attn')\n",
    "densor_dec = model_wo_attn.get_layer('densor_dec')\n",
    "\n",
    "decoder = Model([in_dec, h_dec_prev, c_dec_prev], decode_wo_attn(in_dec, h_dec_prev, c_dec_prev))\n",
    "decoder.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def decode_sequence(input_seqs, encoder, decoder, toks_stoi, toks_itos, max_len, w_attn=False, return_attn_weights=False):\n",
    "    # Generate encoder output\n",
    "    a_enc_pred, h_enc_pred, c_enc_pred = encoder.predict(input_seqs)\n",
    "\n",
    "    # Generate decoder output, one timestep at a time\n",
    "    outs_dec_pred = []\n",
    "    attn_weights = []\n",
    "    in_dec = np.array([toks_stoi[GO]] * len(input_seqs)).reshape(-1, 1)\n",
    "    h_dec_pred = h_enc_pred\n",
    "    c_dec_pred = c_enc_pred\n",
    "\n",
    "    for i in range(max_len):\n",
    "        if not w_attn:\n",
    "            out_dec_pred, h_dec_pred, c_dec_pred = decoder.predict([in_dec, h_dec_pred, c_dec_pred])\n",
    "        else:\n",
    "            out_dec_pred, h_dec_pred, c_dec_pred, attn_weight = decoder.predict([in_dec, a_enc_pred, h_dec_pred, c_dec_pred])\n",
    "\n",
    "        # Derive the predicted token\n",
    "        # Need to first reduce the dimension of `out_dec_pred` from 3D to 2D\n",
    "        # (It's 3D because there is a `timestep` of 1 in the middle)\n",
    "        out_dec_pred = out_dec_pred.reshape(out_dec_pred.shape[0], out_dec_pred.shape[-1])\n",
    "        out_dec_pred = out_dec_pred.argmax(axis=-1)\n",
    "        outs_dec_pred.append(out_dec_pred)\n",
    "\n",
    "        # Assign the predicted token as the next decoder input\n",
    "        in_dec = out_dec_pred\n",
    "        \n",
    "        # Save attention weights\n",
    "        if w_attn:\n",
    "            attn_weights.append(attn_weight.reshape(attn_weight.shape[0], attn_weight.shape[1]))\n",
    "\n",
    "    # Convert output index into tokens\n",
    "    outs_dec_pred = np.array(toks_itos)[np.array(outs_dec_pred).transpose()]\n",
    "    \n",
    "    # Clean up and output\n",
    "    outs_dec_pred = [detok(toks) for toks in outs_dec_pred]\n",
    "    if not return_attn_weights:\n",
    "        return outs_dec_pred\n",
    "    return outs_dec_pred, attn_weights\n",
    "\n",
    "# Convert to strings and clean up\n",
    "# detokenizer = Detok()\n",
    "\n",
    "def detok(toks):\n",
    "    # Cut before the first padding (if any)\n",
    "    toks = list(toks)\n",
    "    if PAD in toks:\n",
    "        first_pad = toks.index(PAD)\n",
    "        toks = toks[:first_pad]\n",
    "    \n",
    "    # return detokenizer.detokenize(toks)\n",
    "    return ' '.join(toks)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "30871"
      ]
     },
     "execution_count": 52,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Test inference with validation data\n",
    "y_val_pred_wo_attn = decode_sequence(X_ix_val, encoder, decoder, y_stoi, y_itos, max_len)\n",
    "len(y_val_pred_wo_attn)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "11659\n",
      "Input:\t\ttom passait une mauvaise journée .\n",
      "Predicted:\ttom was a bad day .\n",
      "True:\t\ttom was having a bad day .\n",
      "276\n",
      "Input:\t\tje suis saoul .\n",
      "Predicted:\ti 'm drunk .\n",
      "True:\t\ti 'm loaded .\n",
      "30038\n",
      "Input:\t\tje suis du genre qui aime réfléchir très attentivement aux choses .\n",
      "Predicted:\ti 'm kind of things to know who she has n't like to be busy .\n",
      "True:\t\ti 'm the type who likes to think things over very carefully .\n",
      "25795\n",
      "Input:\t\tje le fais presque toujours l'après-midi .\n",
      "Predicted:\ti always do this afternoon .\n",
      "True:\t\ti almost always do that in the afternoon .\n",
      "4874\n",
      "Input:\t\telles n' y arriveront pas .\n",
      "Predicted:\tthey wo n't do it .\n",
      "True:\t\tthey wo n't make it .\n"
     ]
    }
   ],
   "source": [
    "# Print out a few results\n",
    "np.random.seed(0)\n",
    "for i in np.random.choice(np.arange(len(y_val)), size=5, replace=False):\n",
    "    print(i)\n",
    "    print('Input:\\t\\t' + detok(X_val[i]))\n",
    "    print('Predicted:\\t' + y_val_pred_wo_attn[i])\n",
    "    print('True:\\t\\t' + detok(y_val[i]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Save predictions\n",
    "pickle.dump(y_val_pred_wo_attn, open(os.path.join(data_path, 'translate-pred_wo_attn.pkl'), 'wb'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Model with attention"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Reload the last checkpoint\n",
    "model_w_attn = load_model(os.path.join(model_path, 'translate-keras-w_attn.h5'), custom_objects={'softmax_axis_1': softmax_axis_1, 'i': 0})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "X (InputLayer)               (None, 20)                0         \n",
      "_________________________________________________________________\n",
      "X_embed_layer (Embedding)    (None, 20, 300)           6465900   \n",
      "_________________________________________________________________\n",
      "lstm_enc (LSTM)              [(None, 20, 256), (None,  570368    \n",
      "=================================================================\n",
      "Total params: 7,036,268\n",
      "Trainable params: 7,036,268\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# Extract encoder\n",
    "X = Input(shape=(max_len, ), name='X')\n",
    "X_embed_layer = model_w_attn.get_layer('X_embed_layer')\n",
    "lstm_enc = model_w_attn.get_layer('lstm_enc')\n",
    "\n",
    "encoder = Model(X, encode(X))\n",
    "encoder.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "h_dec_prev (InputLayer)         (None, 256)          0                                            \n",
      "__________________________________________________________________________________________________\n",
      "attn_repeator (RepeatVector)    (None, 20, 256)      0           h_dec_prev[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "a_enc (InputLayer)              (None, 20, 256)      0                                            \n",
      "__________________________________________________________________________________________________\n",
      "attn_concatenator (Concatenate) (None, 20, 512)      0           attn_repeator[20][0]             \n",
      "                                                                 a_enc[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "attn_weights_computer (Dense)   (None, 20, 1)        513         attn_concatenator[20][0]         \n",
      "__________________________________________________________________________________________________\n",
      "attn_weights_normalizer (Activa (None, 20, 1)        0           attn_weights_computer[20][0]     \n",
      "__________________________________________________________________________________________________\n",
      "in_dec (InputLayer)             (None, 1)            0                                            \n",
      "__________________________________________________________________________________________________\n",
      "attn_context_computer (Dot)     (None, 1, 256)       0           attn_weights_normalizer[20][0]   \n",
      "                                                                 a_enc[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "y_embed_layer (Embedding)       (None, 1, 300)       3763200     in_dec[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "context_input_concatenator (Con (None, 1, 556)       0           attn_context_computer[20][0]     \n",
      "                                                                 y_embed_layer[20][0]             \n",
      "__________________________________________________________________________________________________\n",
      "densor_dec_0 (Dense)            (None, 1, 256)       142592      context_input_concatenator[20][0]\n",
      "__________________________________________________________________________________________________\n",
      "c_dec_prev (InputLayer)         (None, 256)          0                                            \n",
      "__________________________________________________________________________________________________\n",
      "lstm_dec_w_attn (LSTM)          [(None, 256), (None, 525312      densor_dec_0[20][0]              \n",
      "                                                                 h_dec_prev[0][0]                 \n",
      "                                                                 c_dec_prev[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "densor_dec_1 (Dense)            (None, 12544)        3223808     lstm_dec_w_attn[20][0]           \n",
      "==================================================================================================\n",
      "Total params: 7,655,425\n",
      "Trainable params: 7,655,425\n",
      "Non-trainable params: 0\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# Extract decoder\n",
    "in_dec = Input(shape=(1, ), name='in_dec')\n",
    "a_enc = Input(shape=(max_len, hidden_size_enc), name='a_enc')\n",
    "h_dec_prev = Input(shape=(hidden_size_dec, ), name='h_dec_prev')\n",
    "c_dec_prev = Input(shape=(hidden_size_dec, ), name='c_dec_prev')\n",
    "\n",
    "y_embed_layer = model_w_attn.get_layer('y_embed_layer')\n",
    "attn_repeator = model_w_attn.get_layer('attn_repeator')\n",
    "attn_concatenator = model_w_attn.get_layer('attn_concatenator')\n",
    "attn_weights_computer = model_w_attn.get_layer('attn_weights_computer')\n",
    "attn_weights_normalizer = model_w_attn.get_layer('attn_weights_normalizer')\n",
    "attn_context_computer = model_w_attn.get_layer('attn_context_computer')\n",
    "context_input_concatenator = model_w_attn.get_layer('context_input_concatenator')\n",
    "densor_dec_0 = model_w_attn.get_layer('densor_dec_0')\n",
    "lstm_dec_w_attn = model_w_attn.get_layer('lstm_dec_w_attn')\n",
    "densor_dec_1 = model_w_attn.get_layer('densor_dec_1')\n",
    "\n",
    "decoder = Model([in_dec, a_enc, h_dec_prev, c_dec_prev], decode_w_attn(in_dec, a_enc, h_dec_prev, c_dec_prev))\n",
    "decoder.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "30871"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Test inference with validation data\n",
    "y_val_pred_w_attn, y_val_attn_weights = decode_sequence(X_ix_val, encoder, decoder, y_stoi, y_itos, max_len, w_attn=True, return_attn_weights=True)\n",
    "len(y_val_pred_w_attn)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "11659\n",
      "Input:\t\ttom passait une mauvaise journée .\n",
      "Predicted:\ttom was a bad day .\n",
      "True:\t\ttom was having a bad day .\n",
      "276\n",
      "Input:\t\tje suis saoul .\n",
      "Predicted:\ti 'm drunk .\n",
      "True:\t\ti 'm loaded .\n",
      "30038\n",
      "Input:\t\tje suis du genre qui aime réfléchir très attentivement aux choses .\n",
      "Predicted:\ti 'm kind of things that i 'm looking forward to hearing .\n",
      "True:\t\ti 'm the type who likes to think things over very carefully .\n",
      "25795\n",
      "Input:\t\tje le fais presque toujours l'après-midi .\n",
      "Predicted:\ti always do it almost the afternoon .\n",
      "True:\t\ti almost always do that in the afternoon .\n",
      "4874\n",
      "Input:\t\telles n' y arriveront pas .\n",
      "Predicted:\tthey wo n't be there .\n",
      "True:\t\tthey wo n't make it .\n"
     ]
    }
   ],
   "source": [
    "np.random.seed(0)\n",
    "for i in np.random.choice(np.arange(len(y_val)), size=5, replace=False):\n",
    "    print(i)\n",
    "    print('Input:\\t\\t' + detok(X_val[i]))\n",
    "    print('Predicted:\\t' + y_val_pred_w_attn[i])\n",
    "    print('True:\\t\\t' + detok(y_val[i]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Save predictions\n",
    "pickle.dump(y_val_pred_w_attn, open(os.path.join(data_path, 'translate-pred_w_attn.pkl'), 'wb'))\n",
    "pickle.dump(y_val_attn_weights, open(os.path.join(data_path, 'translate-attn_weights_w_attn.pkl'), 'wb'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(30871, 4)"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Combine the two models' predictions\n",
    "y_val_pred_wo_attn = pickle.load(open(os.path.join(data_path, 'translate-pred_wo_attn.pkl'), 'rb'))\n",
    "y_val_pred_w_attn = pickle.load(open(os.path.join(data_path, 'translate-pred_w_attn.pkl'), 'rb'))\n",
    "\n",
    "y_val_true_vs_pred = pd.DataFrame({\n",
    "    'source': [' '.join(s) for s in X_val],\n",
    "    'true_target': [' '.join(s) for s in y_val],\n",
    "    'pred_wo_attn': y_val_pred_wo_attn,\n",
    "    'pred_w_attn': y_val_pred_w_attn\n",
    "})\n",
    "y_val_true_vs_pred.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>source</th>\n",
       "      <th>true_target</th>\n",
       "      <th>pred_wo_attn</th>\n",
       "      <th>pred_w_attn</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>11659</th>\n",
       "      <td>tom passait une mauvaise journée .</td>\n",
       "      <td>tom was having a bad day .</td>\n",
       "      <td>tom was a bad day .</td>\n",
       "      <td>tom was a bad day .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>276</th>\n",
       "      <td>je suis saoul .</td>\n",
       "      <td>i 'm loaded .</td>\n",
       "      <td>i 'm drunk .</td>\n",
       "      <td>i 'm drunk .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30038</th>\n",
       "      <td>je suis du genre qui aime réfléchir très atten...</td>\n",
       "      <td>i 'm the type who likes to think things over v...</td>\n",
       "      <td>i 'm kind of things to know who she has n't li...</td>\n",
       "      <td>i 'm kind of things that i 'm looking forward ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25795</th>\n",
       "      <td>je le fais presque toujours l'après-midi .</td>\n",
       "      <td>i almost always do that in the afternoon .</td>\n",
       "      <td>i always do this afternoon .</td>\n",
       "      <td>i always do it almost the afternoon .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4874</th>\n",
       "      <td>elles n' y arriveront pas .</td>\n",
       "      <td>they wo n't make it .</td>\n",
       "      <td>they wo n't do it .</td>\n",
       "      <td>they wo n't be there .</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                  source  \\\n",
       "11659                 tom passait une mauvaise journée .   \n",
       "276                                      je suis saoul .   \n",
       "30038  je suis du genre qui aime réfléchir très atten...   \n",
       "25795         je le fais presque toujours l'après-midi .   \n",
       "4874                         elles n' y arriveront pas .   \n",
       "\n",
       "                                             true_target  \\\n",
       "11659                         tom was having a bad day .   \n",
       "276                                        i 'm loaded .   \n",
       "30038  i 'm the type who likes to think things over v...   \n",
       "25795         i almost always do that in the afternoon .   \n",
       "4874                               they wo n't make it .   \n",
       "\n",
       "                                            pred_wo_attn  \\\n",
       "11659                                tom was a bad day .   \n",
       "276                                         i 'm drunk .   \n",
       "30038  i 'm kind of things to know who she has n't li...   \n",
       "25795                       i always do this afternoon .   \n",
       "4874                                 they wo n't do it .   \n",
       "\n",
       "                                             pred_w_attn  \n",
       "11659                                tom was a bad day .  \n",
       "276                                         i 'm drunk .  \n",
       "30038  i 'm kind of things that i 'm looking forward ...  \n",
       "25795              i always do it almost the afternoon .  \n",
       "4874                              they wo n't be there .  "
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Sample a few to review\n",
    "y_val_true_vs_pred_samp = y_val_true_vs_pred.sample(100, random_state=0)\n",
    "y_val_true_vs_pred_samp = y_val_true_vs_pred_samp[['source', 'true_target', 'pred_wo_attn', 'pred_w_attn']]\n",
    "y_val_true_vs_pred_samp.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Output\n",
    "y_val_true_vs_pred_samp.to_csv(os.path.join(data_path, 'translate-true_vs_pred_samp.csv'), index=False, encoding='utf-8')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "## Compute BLEU scores"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# Compute corpus-level BLEU score\n",
    "def compute_corpus_bleu(references, candidates):\n",
    "    # BLEU-1\n",
    "    bleu_1 = corpus_bleu(references, candidates, weights=(1, 0, 0, 0))\n",
    "    \n",
    "    # BLEU-2\n",
    "    bleu_2 = corpus_bleu(references, candidates, weights=(0.5, 0.5, 0, 0))\n",
    "    \n",
    "    # BLEU-3\n",
    "    bleu_3 = corpus_bleu(references, candidates, weights=(0.33, 0.33, 0.33, 0))\n",
    "    \n",
    "    # BLEU-4\n",
    "    bleu_4 = corpus_bleu(references, candidates, weights=(0.25, 0.25, 0.25, 0.25))\n",
    "    \n",
    "    return bleu_1, bleu_2, bleu_3, bleu_4"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.8119917446035697,\n",
       " 0.7366436711095151,\n",
       " 0.6803116445969866,\n",
       " 0.6344698382568674)"
      ]
     },
     "execution_count": 56,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Without attention\n",
    "compute_corpus_bleu([ref_dict[' '.join(x)] for x in X_val], y_val_pred_wo_attn)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.80145851710511, 0.73102862866273, 0.6786159432986426, 0.6357094173241202)"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# With attention\n",
    "compute_corpus_bleu([ref_dict[' '.join(x)] for x in X_val], y_val_pred_w_attn)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1.999, 5.0]    0.160604\n",
       "(6.0, 7.0]      0.153445\n",
       "(7.0, 8.0]      0.142885\n",
       "(5.0, 6.0]      0.136989\n",
       "(8.0, 9.0]      0.119886\n",
       "(10.0, 12.0]    0.108905\n",
       "(12.0, 20.0]    0.089534\n",
       "(9.0, 10.0]     0.087752\n",
       "Name: source_len_grp, dtype: float64"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Compute BLEU by sentence length\n",
    "y_val_true_vs_pred['source_len'] = y_val_true_vs_pred['source'].str.split().str.len()\n",
    "y_val_true_vs_pred['source_len_grp'] = pd.qcut(y_val_true_vs_pred['source_len'], 10, duplicates='drop')\n",
    "y_val_true_vs_pred['source_len_grp'].value_counts(normalize=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7f3da2a12190>"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAusAAAKOCAYAAAAfwFfUAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\negAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "image/png": {
       "height": 327,
       "width": 373
      }
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "def compute_corpus_bleu_using_df(sources, candidates):\n",
    "    references = [ref_dict[source] for source in sources.tolist()]\n",
    "    return compute_corpus_bleu(references, candidates)\n",
    "\n",
    "bleu_by_len_wo_attn = y_val_true_vs_pred.groupby('source_len_grp').apply(lambda grp: compute_corpus_bleu_using_df(grp['source'], grp['pred_wo_attn'])[-1])\n",
    "bleu_by_len_w_attn = y_val_true_vs_pred.groupby('source_len_grp').apply(lambda grp: compute_corpus_bleu_using_df(grp['source'], grp['pred_w_attn'])[-1])\n",
    "\n",
    "# Combine and visualize\n",
    "pd.DataFrame(bleu_by_len_wo_attn, columns=['wo_attn']).join(pd.DataFrame(bleu_by_len_w_attn, columns=['w_attn'])).plot(kind='bar', title='BLEU score per source length')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Visualize attention weights"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def visualize_attn_weights(input_seq, output_seq, attn_weights, ax):\n",
    "    # Remove paddings from the attention weights\n",
    "    attn_weights = attn_weights[:len(output_seq), :len(input_seq)]\n",
    "    \n",
    "    # Convert into a dataframe\n",
    "    attn_weights = pd.DataFrame(attn_weights, index=output_seq, columns=input_seq)\n",
    "    \n",
    "    # Visualize as a heatmap\n",
    "    fig = sns.heatmap(attn_weights[::-1], cmap='bone', ax=ax)\n",
    "    fig.set_xticklabels(fig.get_xticklabels(), rotation=90)\n",
    "    fig.set_yticklabels(fig.get_yticklabels(), rotation=0)\n",
    "    return fig"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "5042"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# For visualization purposes, select the predictions that match the truths perfectly\n",
    "ix = np.where([true == pred.split() for true, pred in zip(y_val, y_val_pred_w_attn)])[0]\n",
    "len(ix)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Sample a few of them to visualize\n",
    "np.random.seed(0)\n",
    "ix_samp = np.random.choice(ix, 9, replace=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAB3cAAAZwCAYAAABu18+LAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<Figure size 1152x1008 with 18 Axes>"
      ]
     },
     "metadata": {
      "image/png": {
       "height": 824,
       "width": 955
      }
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Visualize\n",
    "fig, axes = plt.subplots(3, 3, figsize=(16, 14))\n",
    "\n",
    "for i, ax in enumerate(np.ravel(axes)):\n",
    "    input_seq = X_val[ix_samp[i]]\n",
    "    output_seq = y_val_pred_w_attn[ix_samp[i]].split()\n",
    "    attn_weights = np.array([a[ix_samp[i]] for a in y_val_attn_weights])\n",
    "    visualize_attn_weights(input_seq, output_seq, attn_weights, ax)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2 (General DS)",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}