labb5_NLP.ipynb

{
 "cells": [
  {
   "cell_type": "raw",
   "metadata": {},
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<b>Lab 5 - NLP</b>\n",
    "\n",
    "\n",
    "umu-id: anth2786\n",
    "\n",
    "<b>Part A</b> <br>\n",
    "is based on code from: https://colab.research.google.com/drive/1zuq1I_FudtB2W4OSOWff8ODqfqK8d9-G\n",
    "<br><br><b>Part B</b><br>\n",
    "is based on code from the book Deep Learning With Python, Chollet F. chapter 6.2. Found here: https://colab.research.google.com/github/fchollet/deep-learning-with-python-notebooks/blob/master/6.2-understanding-recurrent-neural-networks.ipynb\n",
    "<br>\n",
    "Part B also contains predictions with code instpired from:<br>\n",
    "https://towardsdatascience.com/simple-neural-network-sentiment-analyser-using-keras-179bc9f50fbc\n",
    "<br>\n",
    "http://drunkendatascience.com/simple-neural-network-sentiment-analyser-using-keras/\n",
    "<br>\n",
    "https://builtin.com/data-science/how-build-neural-network-keras\n",
    "\n",
    "<br>\n",
    "<b>VG Part</b><br>\n",
    "The BERT model is not included in this. Trying out 4 different way, 2 ways (gave file errors, one demanded more time on the data handling (as it was not described in the exemple model).  \n",
    "(Some code are still in this notebook, in the end)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "\n",
    "# Disable running on GPU\n",
    "os.environ[\"CUDA_VISIBLE_DEVICES\"] = \"-1\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Nicely formatted time string\n",
    "import time\n",
    "def hms_string(sec_elapsed):\n",
    "    h = int(sec_elapsed / (60 * 60))\n",
    "    m = int((sec_elapsed % (60 * 60)) / 60)\n",
    "    s = sec_elapsed % 60\n",
    "    return f\"{h}:{m:>02}:{s:>05.2f}\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Part A - Embedding and word2vec"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "from gensim.models import KeyedVectors"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Elapsed time: 0:00:40.40\n"
     ]
    }
   ],
   "source": [
    "start_time = time.time()\n",
    "\n",
    "EMBEDDING_FILE = 'GoogleNews-vectors-negative300.bin.gz' # from above\n",
    "word2vec = KeyedVectors.load_word2vec_format(EMBEDDING_FILE, binary=True)\n",
    "\n",
    "elapsed_time = time.time() - start_time\n",
    "print(\"Elapsed time: {}\".format(hms_string(elapsed_time)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(300,)"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "word2vec[\"cat\"].shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "\n",
    "# Make a function. cos(a,b) = (a dot b) / ( ||a|| * ||b|| )\n",
    "\n",
    "def cos(x1, x2):\n",
    "  return np.dot(x1, x2)/(np.linalg.norm(x1)*np.linalg.norm(x2))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.76094574"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Lets check how close cat and dog are in the vector space\n",
    "\n",
    "cos(word2vec[\"cat\"], word2vec[\"dog\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.11546826"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Lets check how close cat and egplant are in the vector space. Should be further\n",
    "# away (lower cos number)\n",
    "\n",
    "cos(word2vec[\"cat\"], word2vec[\"eggplant\"])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### Lets plot 100 words in a diagram"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Choose 100 random words\n",
    "\n",
    "import random\n",
    "vocab = random.sample(list(word2vec.vocab), 100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = np.array([word2vec[v] for v in vocab])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "from sklearn.manifold import TSNE\n",
    "tsne = TSNE(n_components=2, random_state=0)\n",
    "np.set_printoptions(suppress=True)\n",
    "\n",
    "\n",
    "Y = tsne.fit_transform(X)\n",
    "\n",
    "plt.scatter(Y[:, 0], Y[:, 1])\n",
    "for label, x, y in zip(vocab, Y[:, 0], Y[:, 1]):\n",
    "    plt.annotate(label, xy=(x, y), xytext=(0, 0), textcoords='offset points')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('queen', 0.7118192911148071)]"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Let's start by see where a movement takes us\n",
    "# If we go from man to King and make an equivalent move from woman, where will it  \n",
    "# take us\n",
    "\n",
    "word2vec.most_similar(positive=['woman', 'king'], negative=['man'], topn=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('pregnancy', 0.7064820528030396),\n",
       " ('pregant', 0.6582916378974915),\n",
       " ('Pregnant', 0.6250602006912231),\n",
       " ('miscarrying', 0.6189883947372437),\n",
       " ('collapse_Betsy_Sathers', 0.6159210205078125),\n",
       " ('Molly_Thornberg', 0.604171872138977),\n",
       " ('breastfeed', 0.5998175144195557),\n",
       " ('father_Moynahan_publicist', 0.5968271493911743),\n",
       " ('unborn_baby', 0.5936764478683472),\n",
       " ('preggers', 0.592155933380127)]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Let's check which words have best similarity to pregnant\n",
    "\n",
    "word2vec.most_similar('pregnant')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Looks good"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('uncle', 0.7730482220649719)]"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Let's try some other similar movement\n",
    "# If we go from pregnant to mother, where will similar move from father take us?\n",
    "\n",
    "word2vec.most_similar(positive=['father', 'mother'], negative=['pregnant'], topn=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('passively', 0.4318414330482483),\n",
       " ('concious', 0.4253477156162262),\n",
       " ('reactive', 0.4089198708534241),\n",
       " ('consciously', 0.38172537088394165),\n",
       " ('preconscious', 0.3749276399612427)]"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Let's try some other similar movement\n",
    "# If we go from good to passive,  where will similar move from conscious take us?\n",
    "\n",
    "word2vec.most_similar(positive=['conscious', 'passive'], negative=['good'], topn=5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('hamburger', 0.5281333923339844),\n",
       " ('hotdogs', 0.5276503562927246),\n",
       " ('burger', 0.4839725196361542),\n",
       " ('wiener', 0.4822864830493927),\n",
       " ('sandwich', 0.47812312841415405)]"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# If hotdog is a dish in USA, what is a similar dish in Sweden?\n",
    "\n",
    "word2vec.most_similar(positive=['sweden', 'hotdog'], negative=['usa'], topn=5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('son', 0.7779923677444458),\n",
       " ('uncle', 0.7626983523368835),\n",
       " ('younger_brother', 0.7591934204101562),\n",
       " ('nephew', 0.7300765514373779),\n",
       " ('eldest_son', 0.699535608291626)]"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# If we have a relationship between sister and brother. What is the equivalent\n",
    "# relationship for father\n",
    "\n",
    "word2vec.most_similar(positive=['father', 'brother'], negative=['sister'], topn=5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('walking', 0.6744711995124817),\n",
       " ('walk', 0.6334099769592285),\n",
       " ('walks', 0.5541554689407349),\n",
       " ('Walking', 0.4588717818260193),\n",
       " ('Kareem_Shuman', 0.4470715820789337)]"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Let's try some verbs\n",
    "# Swam - swimming gives Walked - ?\n",
    "\n",
    "word2vec.most_similar(positive=['walked', 'swimming'], negative=['swam'], topn=5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('weaker', 0.590004563331604),\n",
       " ('strengthened', 0.5677163004875183),\n",
       " ('strengthening', 0.5273597836494446),\n",
       " ('robust', 0.5244174003601074),\n",
       " ('strongest', 0.5159147381782532)]"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Dark -> stronger : strong -> ?\n",
    "\n",
    "word2vec.most_similar(positive=['strong', 'stronger'], negative=['dark'], topn=5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('bright', 0.4604077935218811),\n",
       " ('gray', 0.4363955855369568),\n",
       " ('inky_black', 0.43228602409362793),\n",
       " ('Dark', 0.41905760765075684),\n",
       " ('stuffy_cabins', 0.41465580463409424)]"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Stronger -> dark : strong -> ?\n",
    "\n",
    "word2vec.most_similar(positive=['strong', 'dark'], negative=['stronger'], topn=5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('darker', 0.7023787498474121),\n",
       " ('blacker', 0.5634409785270691),\n",
       " ('darkened', 0.5231126546859741),\n",
       " ('brighter', 0.5117559432983398),\n",
       " ('paler', 0.49426209926605225)]"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Let's try another verb\n",
    "# strong -> stronger : dark -> ?\n",
    "\n",
    "word2vec.most_similar(positive=['dark', 'stronger'], negative=['strong'], topn=5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Analysis"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We showed that similar words could be clustered. Eg similar words to pregnant is\n",
    "('pregnancy', 0.7064820528030396),\n",
    " ('pregant', 0.6582916378974915),\n",
    " ('Pregnant', 0.6250602006912231),\n",
    " ('miscarrying', 0.6189883947372437),"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Then when we \"calculating\" on relationships we get a rather good outcome. 3 out of 4 looking at common relationships<br>\n",
    "<br>\n",
    "<b>Good relationships</b><br>\n",
    "&nbsp;    usa -> hotdog, sweden - ?hamburger<br>\n",
    "&nbsp;    Swam - swimming gives Walked - ?walking<br>\n",
    "&nbsp;    strong -> stronger : dark -> ?darker<br>\n",
    "<br>\n",
    "<b>Not so good relationsships</b><br>\n",
    "&nbsp;    sister -> brother. father -> ?son<br>\n",
    "<br>\n",
    "\n",
    "But when we did input a difficult relationship, the model would make a <b>difficult estimation</b><br>\n",
    "\n",
    "\n",
    "&nbsp;    good -> passive,  conscious -> ?Passivley<br>\n",
    "&nbsp;    Dark -> stronger : strong -> ?weaker<br>\n",
    "&nbsp;    Stronger -> dark : strong -> ?bright<br>\n",
    "<br>\n",
    "So the model's estimations are interesting but a little bit on the weak side"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Part B. IMDB on RNN and LSTM"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Understanding recurrent neural networks\n",
    "\n",
    "This notebook contains the code samples found in Chapter 6, Section 2 of [Deep Learning with Python](https://www.manning.com/books/deep-learning-with-python?a_aid=keras&a_bid=76564dff). Note that the original text features far more content, in particular further explanations and figures: in this notebook, you will only find source code and related comments.\n",
    "\n",
    "---\n",
    "\n",
    "[...]\n",
    "\n",
    "## A first recurrent layer in Keras\n",
    "\n",
    "The process we just naively implemented in Numpy corresponds to an actual Keras layer: the `SimpleRNN` layer:\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'2.4.3'"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import keras\n",
    "keras.__version__"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.layers import SimpleRNN\n",
    "from keras.layers import Dense"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "There is just one minor difference: `SimpleRNN` processes batches of sequences, like all other Keras layers, not just a single sequence like \n",
    "in our Numpy example. This means that it takes inputs of shape `(batch_size, timesteps, input_features)`, rather than `(timesteps, \n",
    "input_features)`.\n",
    "\n",
    "Like all recurrent layers in Keras, `SimpleRNN` can be run in two different modes: it can return either the full sequences of successive \n",
    "outputs for each timestep (a 3D tensor of shape `(batch_size, timesteps, output_features)`), or it can return only the last output for each \n",
    "input sequence (a 2D tensor of shape `(batch_size, output_features)`). These two modes are controlled by the `return_sequences` constructor \n",
    "argument. Let's take a look at an example:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding (Embedding)        (None, None, 32)          320000    \n",
      "_________________________________________________________________\n",
      "simple_rnn (SimpleRNN)       (None, 32)                2080      \n",
      "=================================================================\n",
      "Total params: 322,080\n",
      "Trainable params: 322,080\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Embedding, SimpleRNN\n",
    "\n",
    "model = Sequential()\n",
    "model.add(Embedding(10000, 32))\n",
    "model.add(SimpleRNN(32))\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_1\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding_1 (Embedding)      (None, None, 32)          320000    \n",
      "_________________________________________________________________\n",
      "simple_rnn_1 (SimpleRNN)     (None, None, 32)          2080      \n",
      "=================================================================\n",
      "Total params: 322,080\n",
      "Trainable params: 322,080\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model = Sequential()\n",
    "model.add(Embedding(10000, 32))\n",
    "model.add(SimpleRNN(32, return_sequences=True))\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "It is sometimes useful to stack several recurrent layers one after the other in order to increase the representational power of a network. In such a setup, you have to get all intermediate layers to return full sequences:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_2\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding_2 (Embedding)      (None, None, 32)          320000    \n",
      "_________________________________________________________________\n",
      "simple_rnn_2 (SimpleRNN)     (None, None, 32)          2080      \n",
      "_________________________________________________________________\n",
      "simple_rnn_3 (SimpleRNN)     (None, None, 32)          2080      \n",
      "_________________________________________________________________\n",
      "simple_rnn_4 (SimpleRNN)     (None, None, 32)          2080      \n",
      "_________________________________________________________________\n",
      "simple_rnn_5 (SimpleRNN)     (None, 32)                2080      \n",
      "=================================================================\n",
      "Total params: 328,320\n",
      "Trainable params: 328,320\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model = Sequential()\n",
    "model.add(Embedding(10000, 32))\n",
    "model.add(SimpleRNN(32, return_sequences=True))\n",
    "model.add(SimpleRNN(32, return_sequences=True))\n",
    "model.add(SimpleRNN(32, return_sequences=True))\n",
    "model.add(SimpleRNN(32))  # This last layer only returns the last outputs.\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now let's try to use such a model on the IMDB movie review classification problem. First, let's preprocess the data:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loading data...\n",
      "25000 train sequences\n",
      "25000 test sequences\n",
      "Pad sequences (samples x time)\n",
      "input_train shape: (25000, 500)\n",
      "input_test shape: (25000, 500)\n",
      "1\n",
      "[   0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    1   14   22   16   43  530  973 1622 1385   65  458 4468\n",
      "   66 3941    4  173   36  256    5   25  100   43  838  112   50  670\n",
      "    2    9   35  480  284    5  150    4  172  112  167    2  336  385\n",
      "   39    4  172 4536 1111   17  546   38   13  447    4  192   50   16\n",
      "    6  147 2025   19   14   22    4 1920 4613  469    4   22   71   87\n",
      "   12   16   43  530   38   76   15   13 1247    4   22   17  515   17\n",
      "   12   16  626   18    2    5   62  386   12    8  316    8  106    5\n",
      "    4 2223 5244   16  480   66 3785   33    4  130   12   16   38  619\n",
      "    5   25  124   51   36  135   48   25 1415   33    6   22   12  215\n",
      "   28   77   52    5   14  407   16   82    2    8    4  107  117 5952\n",
      "   15  256    4    2    7 3766    5  723   36   71   43  530  476   26\n",
      "  400  317   46    7    4    2 1029   13  104   88    4  381   15  297\n",
      "   98   32 2071   56   26  141    6  194 7486   18    4  226   22   21\n",
      "  134  476   26  480    5  144   30 5535   18   51   36   28  224   92\n",
      "   25  104    4  226   65   16   38 1334   88   12   16  283    5   16\n",
      " 4472  113  103   32   15   16 5345   19  178   32]\n",
      "0\n",
      "[   0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    1   13  244    6   87  337    7  628\n",
      " 2219    5   28  285   15  240   93   23  288  549   18 1455  673    4\n",
      "  241  534 3635 8448   20   38   54   13  258   46   44   14   13 1241\n",
      " 7258   12    5    5   51    9   14   45    6  762    7    2 1309  328\n",
      "    5  428 2473   15   26 1292    5 3939 6728    5 1960  279   13   92\n",
      "  124  803   52   21  279   14    9   43    6  762    7  595   15   16\n",
      "    2   23    4 1071  467    4  403    7  628 2219    8   97    6  171\n",
      " 3596   99  387   72   97   12  788   15   13  161  459   44    4 3939\n",
      " 1101  173   21   69    8  401    2    4  481   88   61 4731  238   28\n",
      "   32   11   32   14    9    6  545 1332  766    5  203   73   28   43\n",
      "   77  317   11    4    2  953  270   17    6 3616   13  545  386   25\n",
      "   92 1142  129  278   23   14  241   46    7  158]\n"
     ]
    }
   ],
   "source": [
    "from keras.datasets import imdb\n",
    "from keras.preprocessing import sequence\n",
    "\n",
    "max_features = 10000  # number of words to consider as features\n",
    "maxlen = 500  # cut texts after this number of words (among top max_features most common words)\n",
    "batch_size = 32\n",
    "\n",
    "print('Loading data...')\n",
    "(input_train, y_train), (input_test, y_test) = imdb.load_data(num_words=max_features)\n",
    "print(len(input_train), 'train sequences')\n",
    "print(len(input_test), 'test sequences')\n",
    "\n",
    "print('Pad sequences (samples x time)')\n",
    "input_train = sequence.pad_sequences(input_train, maxlen=maxlen)\n",
    "input_test = sequence.pad_sequences(input_test, maxlen=maxlen)\n",
    "print('input_train shape:', input_train.shape)\n",
    "print('input_test shape:', input_test.shape)\n",
    "\n",
    "print(y_train[0])\n",
    "print(input_train[0])\n",
    "\n",
    "print(y_train[100])\n",
    "print(input_train[100])\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's train a simple recurrent network using an Embedding layer and a SimpleRNN layer:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "157/157 [==============================] - 77s 489ms/step - loss: 0.6547 - acc: 0.6005 - val_loss: 0.5035 - val_acc: 0.7684\n",
      "Epoch 2/10\n",
      "157/157 [==============================] - 76s 485ms/step - loss: 0.3978 - acc: 0.8329 - val_loss: 0.4028 - val_acc: 0.8242\n",
      "Epoch 3/10\n",
      "157/157 [==============================] - 76s 486ms/step - loss: 0.2834 - acc: 0.8868 - val_loss: 0.4381 - val_acc: 0.8326\n",
      "Epoch 4/10\n",
      "157/157 [==============================] - 76s 485ms/step - loss: 0.2056 - acc: 0.9225 - val_loss: 0.4521 - val_acc: 0.8018\n",
      "Epoch 5/10\n",
      "157/157 [==============================] - 76s 483ms/step - loss: 0.1453 - acc: 0.9460 - val_loss: 0.4829 - val_acc: 0.8118\n",
      "Epoch 6/10\n",
      "157/157 [==============================] - 76s 485ms/step - loss: 0.1017 - acc: 0.9661 - val_loss: 0.5346 - val_acc: 0.8212\n",
      "Epoch 7/10\n",
      "157/157 [==============================] - 76s 485ms/step - loss: 0.0703 - acc: 0.9766 - val_loss: 0.6044 - val_acc: 0.8100\n",
      "Epoch 8/10\n",
      "157/157 [==============================] - 76s 485ms/step - loss: 0.0557 - acc: 0.9816 - val_loss: 0.5761 - val_acc: 0.8140\n",
      "Epoch 9/10\n",
      "157/157 [==============================] - 76s 485ms/step - loss: 0.0257 - acc: 0.9928 - val_loss: 0.6169 - val_acc: 0.8280\n",
      "Epoch 10/10\n",
      "157/157 [==============================] - 76s 485ms/step - loss: 0.0148 - acc: 0.9962 - val_loss: 0.7626 - val_acc: 0.7908\n",
      "Elapsed time: 0:12:48.56\n"
     ]
    }
   ],
   "source": [
    "start_time = time.time()\n",
    "\n",
    "model = Sequential()\n",
    "model.add(Embedding(max_features, 32))\n",
    "model.add(SimpleRNN(32))\n",
    "model.add(Dense(1, activation='sigmoid'))\n",
    "\n",
    "model.compile(optimizer='rmsprop', loss='binary_crossentropy', metrics=['acc'])\n",
    "history = model.fit(input_train, y_train,\n",
    "                    epochs=10,\n",
    "                    batch_size=128,\n",
    "                    validation_split=0.2)\n",
    "\n",
    "elapsed_time = time.time() - start_time\n",
    "print(\"Elapsed time: {}\".format(hms_string(elapsed_time)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's display the training and validation loss and accuracy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXoAAAEICAYAAABRSj9aAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8vihELAAAACXBIWXMAAAsTAAALEwEAmpwYAAArxElEQVR4nO3de5xVVf3/8deHQUAuIgimMcCggoghtxEVvGBewkuSpAWiQZqEpibf0i+mpaF8s6Q0v6mFeUsxNPNHVBLh7aulKQOiCSoiggxeGrlfhYHP74+1Z2bPYS5n4MycM3vez8fjPM45e6+9z+fsOfM5a6+1zl7m7oiISHI1y3YAIiJSv5ToRUQSToleRCThlOhFRBJOiV5EJOGU6EVEEk6Jvgkys9lmNjbTZbPJzJab2an1sF83s8Oix782sx+mU3YPXmeMmf19T+MUqYlpHH3jYGabYk9bA58BO6Pn33b36Q0fVe4ws+XAt9z96Qzv14Ge7r40U2XNrAB4H9jH3UszEqhIDZpnOwBJj7u3LXtcU1Izs+ZKHpIr9HnMDWq6aeTMbJiZFZvZf5vZx8ADZtbBzP5iZiVmtjZ6nB/b5nkz+1b0eJyZ/cPMpkZl3zezM/awbA8ze8HMNprZ02Z2l5k9Uk3c6cR4s5n9M9rf382sU2z9RWa2wsxWm9n1NRyfY8zsYzPLiy0718zeiB4PNrOXzWydmX1kZr8ysxbV7OtBM7sl9vyaaJsPzezilLJnmdlrZrbBzFaa2U2x1S9E9+vMbJOZHVd2bGPbDzGzeWa2Profku6xqeNx7mhmD0TvYa2ZzYytG2FmC6P38J6ZDY+WV2omM7Obyv7OZlYQNWFdYmYfAM9Gy/8Q/R3WR5+RI2Pb72tmP4/+nuujz9i+ZvZXM7sy5f28YWbnVvVepXpK9MlwENAR6A6MJ/xdH4iedwO2Ar+qYftjgHeATsDPgPvMzPag7KPAq8ABwE3ARTW8ZjoxXgB8EzgQaAF8H8DM+gD3RPv/fPR6+VTB3V8BNgNfTNnvo9HjncDE6P0cB5wCXF5D3EQxDI/iOQ3oCaT2D2wGvgHsD5wFXGZmX4nWnRjd7+/ubd395ZR9dwT+CtwZvbdfAH81swNS3sNux6YKtR3nhwlNgUdG+7o9imEw8Dvgmug9nAgsr+Y1qnIScATwpej5bMJxOhBYAMSbGqcCg4AhhM/xtcAu4CHgwrJCZtYP6EI4NlIX7q5bI7sR/uFOjR4PA7YDrWoo3x9YG3v+PKHpB2AcsDS2rjXgwEF1KUtIIqVA69j6R4BH0nxPVcV4Q+z55cDfosc/AmbE1rWJjsGp1ez7FuD+6HE7QhLuXk3Zq4H/F3vuwGHR4weBW6LH9wO3xsr1ipetYr93ALdHjwuiss1j68cB/4geXwS8mrL9y8C42o5NXY4zcDAhoXaootxvyuKt6fMXPb+p7O8ce2+H1BDD/lGZ9oQvoq1AvyrKtQLWEvo9IHwh3F0f/1NJv6lGnwwl7r6t7ImZtTaz30SnwhsITQX7x5svUnxc9sDdt0QP29ax7OeBNbFlACurCzjNGD+OPd4Si+nz8X27+2ZgdXWvRai9jzSzlsBIYIG7r4ji6BU1Z3wcxfE/hNp9bSrFAKxIeX/HmNlzUZPJemBCmvst2/eKlGUrCLXZMtUdm0pqOc5dCX+ztVVs2hV4L814q1J+bMwsz8xujZp/NlBxZtApurWq6rWiz/RjwIVm1gwYTTgDkTpSok+G1KFT3wMOB45x9/2oaCqorjkmEz4COppZ69iyrjWU35sYP4rvO3rNA6or7O6LCYnyDCo320BoAnqbUGvcD/jBnsRAOKOJexSYBXR19/bAr2P7rW2o24eEppa4bsCqNOJKVdNxXkn4m+1fxXYrgUOr2edmwtlcmYOqKBN/jxcAIwjNW+0Jtf6yGD4FttXwWg8BYwhNals8pZlL0qNEn0ztCKfD66L23hvr+wWjGnIRcJOZtTCz44Av11OMTwBnm9nxUcfpZGr/LD8KfJeQ6P6QEscGYJOZ9QYuSzOGx4FxZtYn+qJJjb8doba8LWrvviC2roTQZHJINft+CuhlZheYWXMz+zrQB/hLmrGlxlHlcXb3jwht53dHnbb7mFnZF8F9wDfN7BQza2ZmXaLjA7AQGBWVLwTOSyOGzwhnXa0JZ01lMewiNIP9wsw+H9X+j4vOvogS+y7g56g2v8eU6JPpDmBfQm3pX8DfGuh1xxA6NFcT2sUfI/yDV+UO9jBGd18EfIeQvD8itOMW17LZ7wkdhM+6+6ex5d8nJOGNwL1RzOnEMDt6D88CS6P7uMuByWa2kdCn8Hhs2y3AFOCfFkb7HJuy79XA2YTa+GpC5+TZKXGn6w5qPs4XATsIZzX/IfRR4O6vEjp7bwfWA/9HxVnGDwk18LXAj6l8hlSV3xHOqFYBi6M44r4P/BuYB6wBfkrl3PQ7oC+hz0f2gH4wJfXGzB4D3nb3ej+jkOQys28A4939+GzH0lipRi8ZY2ZHm9mh0an+cEK77MwshyWNWNQsdjkwLduxNGZK9JJJBxGG/m0ijAG/zN1fy2pE0miZ2ZcI/RmfUHvzkNRATTciIgmnGr2ISMLl3EXNOnXq5AUFBdkOQ0SkUZk/f/6n7t65qnU5l+gLCgooKirKdhgiIo2KmaX+mrqcmm5ERBJOiV5EJOGU6EVEEq7WNnozu5/wc+z/uPsXqlhvwC+BMwlX0Rvn7guidWOBG6Kit7j7Q3sS5I4dOyguLmbbtm21F5asaNWqFfn5+eyzzz7ZDkVEUqTTGfsgYaKC31Wz/gzChAI9CZNS3AMcE7uAUiHhSnbzzWxWNZdErVFxcTHt2rWjoKCA6ufDkGxxd1avXk1xcTE9evTIdjgikqLWpht3f4FwoaHqjAB+58G/CNe6Ppgws8xcdy+73vVcYPieBLlt2zYOOOAAJfkcZWYccMABOuMS2UPTp0NBATRrFu6nT69ti7rJxPDKLlSegKE4Wlbd8t2Y2XjCFHh065Z6We/yMhkIVeqL/j4ie2b6dBg/HrZEU/asWBGeA4wZk5nXyInOWHef5u6F7l7YuXOV4/1FRDKuvmvS6bj++ookX2bLlrA8UzKR6FdReaad/GhZdcsbndWrV9O/f3/69+/PQQcdRJcuXcqfb9++vcZti4qKuOqqq2p9jSFDhmQqXBFJQ1lNesUKcK+oSTd0sv/gg7ot3xOZSPSzgG9YcCywPpq5Zg5wejRzTQfg9GhZvcv0t/QBBxzAwoULWbhwIRMmTGDixInlz1u0aEFpaWm12xYWFnLnnXfW+hovvfTS3gUpInXSEDXpdFTTWl3t8j1Ra6I3s98TZqA/3MyKzewSM5tgZhOiIk8Bywiz7NxLuHY07r4GuJkwa8w8YHK0rF411Lf0uHHjmDBhAscccwzXXnstr776KscddxwDBgxgyJAhvPPOOwA8//zznH322QDcdNNNXHzxxQwbNoxDDjmk0hdA27Zty8sPGzaM8847j969ezNmzBjKrjD61FNP0bt3bwYNGsRVV11Vvt+45cuXc8IJJzBw4EAGDhxY6Qvkpz/9KX379qVfv35MmjQJgKVLl3LqqafSr18/Bg4cyHvv7c180CKNR0PUpNMxZQq0bl15WevWYXnGuHtO3QYNGuSpFi9evNuy6nTv7h5SfOVb9+5p76JGN954o992220+duxYP+uss7y0tNTd3devX+87duxwd/e5c+f6yJEj3d39ueee87POOqt82+OOO863bdvmJSUl3rFjR9++fbu7u7dp06a8/H777ecrV670nTt3+rHHHusvvviib9261fPz833ZsmXu7j5q1Kjy/cZt3rzZt27d6u7uS5Ys8bLj+dRTT/lxxx3nmzdvdnf31atXu7v74MGD/cknn3R3961bt5av3xN1+TuJZFt954q6eOSR8Lpm4f6RR+q+D6DIq8mrOXdRs73VkN/S559/Pnl5eQCsX7+esWPH8u6772Jm7Nixo8ptzjrrLFq2bEnLli058MAD+eSTT8jPz69UZvDgweXL+vfvz/Lly2nbti2HHHJI+Tj10aNHM23a7pPu7NixgyuuuIKFCxeSl5fHkiVLAHj66af55je/Seuo6tCxY0c2btzIqlWrOPfcc4HwoyeRpmLKlMqjXaAeatJpGjMmcyNsqpITo24yqSHau8q0adOm/PEPf/hDTj75ZN58803+/Oc/VzumvGXLluWP8/LyqmzfT6dMdW6//XY+97nP8frrr1NUVFRrZ7FINuTCaJcxY2DaNOjeHczC/bRp9ZtwsyVxib5B2ruqsH79erp0CT8TePDBBzO+/8MPP5xly5axfPlyAB577LFq4zj44INp1qwZDz/8MDt37gTgtNNO44EHHmBLVH1Zs2YN7dq1Iz8/n5kzZwLw2Wefla8XqS+5MtoFQlJfvhx27Qr3SUzykMBEn61v6WuvvZbrrruOAQMG1KkGnq59992Xu+++m+HDhzNo0CDatWtH+/btdyt3+eWX89BDD9GvXz/efvvt8rOO4cOHc84551BYWEj//v2ZOnUqAA8//DB33nknRx11FEOGDOHjjz/OeOySW7Jdm86V0S5NSc7NGVtYWOipE4+89dZbHHHEEVmKKHds2rSJtm3b4u585zvfoWfPnkycODHbYZXT3yn3pf4KE8IZb0M2WTRrFmryqcxCzVr2jJnNd/fCqtYlrkafZPfeey/9+/fnyCOPZP369Xz729/OdkjSyORCbboh+9EkSNyomySbOHFiTtXgpfHJhbHjuTTapalQjV6kCcmF2nRTGu2SK5ToRRpItjtBIXuj0lI1ldEuuUKJXqQB5MqQQtWmmyYlepEGkAudoGVUm256lOjTcPLJJzNnTuULb95xxx1cdtll1W4zbNgwyoaJnnnmmaxbt263MjfddFP5ePbqzJw5k8WLF5c//9GPfsTTTz9dh+glF+RCJ6g0XUr0aRg9ejQzZsyotGzGjBmMHj06re2feuop9t9//z167dREP3nyZE499dQ92pdkTy50gkrTpUSfhvPOO4+//vWv5deNWb58OR9++CEnnHACl112GYWFhRx55JHceOONVW5fUFDAp59+CsCUKVPo1asXxx9/fPmljCGMkT/66KPp168fX/3qV9myZQsvvfQSs2bN4pprrqF///689957jBs3jieeeAKAZ555hgEDBtC3b18uvvhiPvvss/LXu/HGGxk4cCB9+/bl7bff3i0mXc64YeVKJ6g0TY1uHP3VV8PChZndZ//+cMcd1a/v2LEjgwcPZvbs2YwYMYIZM2bwta99DTNjypQpdOzYkZ07d3LKKafwxhtvcNRRR1W5n/nz5zNjxgwWLlxIaWkpAwcOZNCgQQCMHDmSSy+9FIAbbriB++67jyuvvJJzzjmHs88+m/POO6/SvrZt28a4ceN45pln6NWrF9/4xje45557uPrqqwHo1KkTCxYs4O6772bq1Kn89re/rbT9gQceyNy5c2nVqhXvvvsuo0ePpqioiNmzZ/OnP/2JV155hdatW7NmTZhCYMyYMUyaNIlzzz2Xbdu2sUs/YayTsnbw668PzTXduoUkr/ZxaQiq0acp3nwTb7Z5/PHHGThwIAMGDGDRokWVmllSvfjii5x77rm0bt2a/fbbj3POOad83ZtvvskJJ5xA3759mT59OosWLaoxnnfeeYcePXrQq1cvAMaOHcsLL7xQvn7kyJEADBo0qPxCaHE7duzg0ksvpW/fvpx//vnlcad7OePWqdVTqZU6QSVb0qrRm9lw4JdAHvBbd781ZX134H6gM7AGuNDdi6N1O4F/R0U/cPdz2As11bzr04gRI5g4cSILFixgy5YtDBo0iPfff5+pU6cyb948OnTowLhx46q9PHFtxo0bx8yZM+nXrx8PPvggzz///F7FW3ap4+oucxy/nPGuXbt0LXqRBEtnKsE84C7gDKAPMNrM+qQUmwr8zt2PAiYDP4mt2+ru/aPbXiX5bGrbti0nn3wyF198cXltfsOGDbRp04b27dvzySefMHv27Br3ceKJJzJz5ky2bt3Kxo0b+fOf/1y+buPGjRx88MHs2LGD6bHB1e3atWPjxo277evwww9n+fLlLF26FAhXoTzppJPSfj+6nLFI05FO081gYKm7L3P37cAMYERKmT7As9Hj56pYnwijR4/m9ddfL0/0/fr1Y8CAAfTu3ZsLLriAoUOH1rj9wIED+frXv06/fv0444wzOProo8vX3XzzzRxzzDEMHTqU3r17ly8fNWoUt912GwMGDKjUAdqqVSseeOABzj//fPr27UuzZs2YMGEC6WpKlzPOhV+kimRTrZcpNrPzgOHu/q3o+UXAMe5+RazMo8Ar7v5LMxsJ/BHo5O6rzawUWAiUAre6+8wqXmM8MB6gW7dug1asWFFpvS5/2zjk4t8pFy7LK9IQGuIyxd8HTjKz14CTgFXAzmhd9+jFLwDuMLNDUzd292nuXujuhZ07d85QSCK59YtUkWxJpzN2FdA19jw/WlbO3T8ERgKYWVvgq+6+Llq3KrpfZmbPAwMADcKWBqFfpIqkV6OfB/Q0sx5m1gIYBcyKFzCzTmZWtq/rCCNwMLMOZtayrAwwFKh+/GENcm0mLKksV/8++kWqSBqJ3t1LgSuAOcBbwOPuvsjMJptZ2SiaYcA7ZrYE+BxQ9nu/I4AiM3ud0El7q7vXOdG3atWK1atX52wyaercndWrV+fkEE39IlWkkcwZu2PHDoqLi/d4jLrUv1atWpGfn88+++yT7VB2M326fpEqyVdTZ2yjSPQiIlIzTQ4uItKEKdGLiCScEr2ISMIp0YuIJJwSvdQbXWNGJDc0uolHpHFIvcbMihXhOWhoo0hDU41e6oWuMSOSO5TopV7oGjMiuUOJXuqFrjEjkjuU6KVe6BozIrlDiV7qxZgxYXKP7t3BLNxrsg+R7NCoG6k3Y8YosYvkAtXoRUQSToleRCThlOhFRBIurURvZsPN7B0zW2pmk6pY393MnjGzN8zseTPLj60ba2bvRrexmQxeRERqV2uiN7M84C7gDKAPMNrM+qQUmwr8zt2PAiYDP4m27QjcCBwDDAZuNLMOmQtfRERqk06NfjCw1N2Xuft2YAYwIqVMH+DZ6PFzsfVfAua6+xp3XwvMBYbvfdgiIpKudBJ9F2Bl7HlxtCzudWBk9PhcoJ2ZHZDmtpjZeDMrMrOikpKSdGMXEZE0ZKoz9vvASWb2GnASsArYme7G7j7N3QvdvbBz584ZCklERCC9H0ytArrGnudHy8q5+4dENXozawt81d3XmdkqYFjKts/vRbwiIlJH6dTo5wE9zayHmbUARgGz4gXMrJOZle3rOuD+6PEc4HQz6xB1wp4eLRMRkQZSa6J391LgCkKCfgt43N0XmdlkMzsnKjYMeMfMlgCfA6ZE264BbiZ8WcwDJkfLpB5pZicRiTN3z3YMlRQWFnpRUVG2w2i0Umd2gnDVSF1QTCTZzGy+uxdWtU6/jE0YzewkIqmU6BNGMzuJSCol+oTRzE4ikkqJPmE0s5OIpFKiTxjN7CQiqTTDVAJpZicRiVONXkQk4ZToRUQSToleRCThlOhFRBJOiV5EJOGU6EVEEk6JXkQk4ZToRUQSToleRCThlOhFRBIurURvZsPN7B0zW2pmk6pY383MnjOz18zsDTM7M1peYGZbzWxhdPt1pt+AiIjUrNZr3ZhZHnAXcBpQDMwzs1nuvjhW7AbCFIP3mFkf4CmgIFr3nrv3z2jUIiKStnRq9IOBpe6+zN23AzOAESllHNgvetwe+DBzIYqIyN5IJ9F3AVbGnhdHy+JuAi40s2JCbf7K2LoeUZPO/5nZCVW9gJmNN7MiMysqKSlJP3oREalVpjpjRwMPuns+cCbwsJk1Az4Curn7AOC/gEfNbL/Ujd19mrsXunth586dMxSSiIhAeol+FdA19jw/WhZ3CfA4gLu/DLQCOrn7Z+6+Olo+H3gP6LW3QYuISPrSSfTzgJ5m1sPMWgCjgFkpZT4ATgEwsyMIib7EzDpHnbmY2SFAT2BZpoIXEZHa1Trqxt1LzewKYA6QB9zv7ovMbDJQ5O6zgO8B95rZRELH7Dh3dzM7EZhsZjuAXcAEd19Tb+9GRER2Y+6e7RgqKSws9KKiomyHISLSqJjZfHcvrGqdfhmbQdOnQ0EBNGsW7qdPz3ZEIiKaHDxjpk+H8eNhy5bwfMWK8Bw0UbeIZJeabjKkoCAk91Tdu8Py5fX72hs2hNcou61YEe5LS+Fb34KzzgpnGSKSXDU13ahGnyEffFC35XWxbl3Vibzstm5d5fKtW4cvng0b4JxzoHdv+N734MILoVWrvY9HRBoXJfoM6dat6hp9t241b+cOa9dWn8RXrID16ytv06ZNSOQFBTB0aMXj7t3DfadOYBZq9H/4A9x2G1x6KdxwA1x5JVx2GXTsuLfvWKRxc4dly2DePNi5E772Ndhnn2xHVT/UdJMhqW30EGrWv/kNDB9ecyLfuLHyvtq1q0jeqUm8oCAkabP0Y3OH556DqVNh9uwQ1yWXwMSJ0KPHXrxpkUZk1aqQ1IuKKu7XxAZ79+wJP/kJjBxZt/+vXFFT040SfQZNnw7//d/hA7XvviEhr1sHmzdXLte+ffVJvKAA9t+//j5ob74JP/95iHXnTjjvPPj+9+Hoo+vn9USyYc2akMzjt48+Cuvy8uALXwif+bLbqlXhf3fxYhgyJJwFDxmS3fdQV0r0DaSoCL7yFSgpgSOPrDqJd+8eEnm2rVoF//u/8Otfh6ahk04KCf/MM9VxW5+2bAn9JDrGmbNpEyxYUDmpL4v9/r5Xr8pJvX//cFabqrQUHngAfvQj+PjjULO/9dZQ028MlOgbwGOPwbhxcOCBMGsW9OuX7YjSs2ED3Hcf3H47rFwJRxwROm7HjFHH7Z5wD7XJpUvD7b33Kt//5z8hyRxxBPTpEyoEffqEW48e+gKozWefwRtvVE7qb70Fu3aF9V27Vk7qgwbVvWK1aRP84hfws5+F15swIST/XL/eohJ9Pdq1K3wIpkyB44+HP/4xJPvGZseOio7bhQvhc5+Dq64KHbcdOmQ7utyya1doBkhN4mX3qZ3nXbvCoYfCYYeFZP7pp7BoUWgmKC6uKLfvvmGEVPwL4MgjwzZ5eQ37HnPBzp0hiceT+uuvh88qhMQbT+qFheFzmykffww//jHce2/4cp40Ca6+uuqzgVygRF9PNm2Ciy6CmTPh4ovh7ruhZctsR7V33OHZZ0PCnzMnjPAp67gtKMh2dA2ntDQMjS1L3vFEvmwZbN1aUbZ583BsDjusIqHHE3tNZ0br14dktnhxRfJftCicXZVp2TJ8AcRr/0ceCYccEl47CdzD8Y0n9QULKgY3tGsXEnk8sXfr1jCdpm+9FZL8rFnQpQvcckv4v8+1L18l+nqwfHkYo75oUTjNu+qqxtlTX5M33ggdt48+Gmqx558P11wTToeTYNs2eP/9qmvlZT84K9Oq1e5JvOy+W7fMJ9wNG+DttyuSf9kXQHwIb4sWcPjhu38BHHpo7gwTLC0NX2bx27p1FY8//jj0bRUVhWHGEI51//6Vk3qvXtlv1nrhhfD5f/VVOOqo0LTzpS9lN6Y4JfoMe+EF+OpXwynkY4/l1h+7PhQXw513hqGiGzbAsGHhAz98ePb/+WqzdWtI3EuW7F4zLy4ONcky++0XEnc8iZc9Pvjg3HivmzZV/gIou3///Yoy++wTEmO8+adPn/BeWrRI/7VKS8PfOzVRV5Wwq1sXH25clbw86Nu3ounl6KPDiJhc+aJK5Q6PPw7XXReO+WmnhbPfXOiTU6LPoHvvhcsvD6fNs2aFGlVTsWFDeP933BGSZJ8+FR232WyyKi0NNfAlS3a/xZtAIPSfVFUrP/RQOOCAxntWtnlz+AKI1/4XLw7NTGX/4s2bhy+APn3CfVltu7qEvWlT7a/bqlUYLty+fej0LHtc3a2qMrma1Gvy2Wdwzz1w883hTOSii0KTTteutW9bX5ToM6C0FP7rv8KQxNNPhxkzmm4nZdmZzNSpoXPsoINC09WECfV3TNzDkNB33909mS9bVrmZZf/9QyJLvR16aKi1NyVbt1Z8AcSbgd57LyTYvUnQ7dvX7Qwhidatg//5n3DGaxY6aydNCsemoSnR76W1a8PPo59+OnRK/uxnyekE2xvu4ZhMnQp//3vouL300vBh7959z/a5enVFAo8n9XffrdwMsO++YXxzr14V92W3xlwzbyi7duVGU1RSrFgRLjHyyCPh8/ejH4WKT0N+Ee51ojez4cAvCTNM/dbdb01Z3w14CNg/KjPJ3Z+K1l1HmFN2J3CVu8+p6bVyLdG//TZ8+cvhD/nrX4fRNbK7118PHbe//334AijruB04cPeymzfvnsTLHsd/kp6XF5rIyhJ4PKF36aJEJblnwYLwuX/22dAk+JOfhP68hqh47FWij+Z8XQKcBhQT5pAd7e6LY2WmAa+5+z1m1gd4yt0Lose/BwYDnweeBnq5+87qXi+XEv3s2TBqVGh/fvLJME5earZyZUXH7caN8MUvhqauZcsqEvqqlKnl8/Mr18jLEnqPHo2z/VaaNnf429/g2mvDJUeOPTac9Q4dWr+vu7eXKR4MLHX3ZdHOZgAjgMWxMg6UtX62Bz6MHo8AZrj7Z8D7ZrY02t/LdX4XDcg9/FL0mmvCiIA//WnPmyKamq5dwyiEG26o6Lh99tlwOturF5x6auWkfthhufsDFJE9YQZnnBEqOA89BD/8YagknntuqOFnZQCHu9d4A84jNNeUPb8I+FVKmYOBfxNq/GuBQdHyXwEXxsrdB5xXxWuMB4qAom7dunk2bdvmPm6cO7iPHOm+cWNWw2n0duxwX7s221GIZM+mTe633OLetq17Xp775Ze7f/JJ5l8HKPJq8nimWjlHAw+6ez5wJvCwmaW9b3ef5u6F7l7YOYsXlPjkk9DU8OCDoTPlD3+Atm2zFk4iNG+eGxdxE8mWNm3g+uvDSKdvfzs0ax56aBiOWdvvDDIlnWS8CoiPDs2PlsVdAjwO4O4vA62ATmlumxNeey38WOO118IPIn78Y3X2iUjmHHgg3HVXGOZ62mmhSadnT7j//nBdn/qUTiqbB/Q0sx5m1gIYBcxKKfMBcAqAmR1BSPQlUblRZtbSzHoAPYFXMxV8pjzxRGhDc4d//COMGBERqQ+HHx4Gd7z4Yrh8xiWXhEs+zJ5d+ZfamVRronf3UuAKYA7wFvC4uy8ys8lmdk5U7HvApWb2OmGUzbio2WgRoaa/GPgb8B2vYcRNQ9u1C266KST2fv3ChZSqGg4oIpJpxx8PL70Umoi3bg1zQXz5y/WT7JvsD6Y2b4axY8NlhceODe1mjf3KkyLSOG3fXjEk+Qc/2LN97O3wysT54AMYMSJcnXHq1HBpA/2SUkSypUULuPLK+tt/k0v0//xnmCJs2zb4y1/CeFcRkSRrUuNKHngATj45XNjqX/9SkheRpqFJJPqyK09efHGYBPuVV8KcnSIiTUHim27WrQvXq5kzJ7SB/eIXuvKkiDQtiU55S5aE6f7eew+mTQuX0BURaWoSm+jnzg3XkG/eHJ55Bk48MdsRiYhkR+La6N3DZXLPOCNcSfHVV5XkRaRpS1Si3749NM9897tw9tlhKGWPHtmOSkQkuxKT6EtKwrXO77svXCnuySehXbtsRyUikn2JaaM3C9PQPfoojB6d7WhERHJHYhJ9p06wcKGGToqIpEpM0w0oyYuIVCVRiV5ERHanRC8iknBK9CIiCZdWojez4Wb2jpktNbNJVay/3cwWRrclZrYutm5nbF3qFIQiIlLPau2+NLM84C7gNKAYmGdms9x9cVkZd58YK38lMCC2i63u3j9jEYuISJ2kU6MfDCx192Xuvh2YAYyoofxowryxIiKSA9JJ9F2AlbHnxdGy3ZhZd6AH8GxscSszKzKzf5nZV6rZbnxUpqikpCS9yEVEJC2Z7owdBTzh7jtjy7pHE9ZeANxhZoembuTu09y90N0LO3funOGQRESatnQS/Sqga+x5frSsKqNIabZx91XR/TLgeSq334uISD1LJ9HPA3qaWQ8za0FI5ruNnjGz3kAH4OXYsg5m1jJ63AkYCixO3VZEROpPraNu3L3UzK4A5gB5wP3uvsjMJgNF7l6W9EcBM9zdY5sfAfzGzHYRvlRujY/WERGR+meV83L2FRYWelFRUbbDEBFpVMxsftQfuhv9MlZEJOGU6EVEEk6JXkQk4ZToRUQSToleRCThlOhFRBJOiV5EJOGU6EVEEk6JXkQk4ZToRUQSToleRCThlOhFRBJOiV5EJOGU6EVEEk6JXkQk4ZToRUQSLq1Eb2bDzewdM1tqZpOqWH+7mS2MbkvMbF1s3Vgzeze6jc1g7CIikoZapxI0szzgLuA0oBiYZ2az4lMCuvvEWPkriSYAN7OOwI1AIeDA/GjbtRl9FyIiUq10avSDgaXuvszdtwMzgBE1lB8N/D56/CVgrruviZL7XGD43gQsIiJ1k06i7wKsjD0vjpbtxsy6Az2AZ+uyrZmNN7MiMysqKSlJJ24REUlTpjtjRwFPuPvOumzk7tPcvdDdCzt37pzhkEREmrZ0Ev0qoGvseX60rCqjqGi2qeu2IiJSD9JJ9POAnmbWw8xaEJL5rNRCZtYb6AC8HFs8BzjdzDqYWQfg9GiZiIg0kFpH3bh7qZldQUjQecD97r7IzCYDRe5elvRHATPc3WPbrjGzmwlfFgCT3X1NZt+CiIjUxGJ5OScUFhZ6UVFRtsMQEWlUzGy+uxdWtU6/jBURSTglehGRhFOiFxFJOCV6EZGEU6IXEUk4JXoRkYRTohcRSTglehGRhFOiFxFJOCV6EZGEU6IXEUk4JXoRkYRTohcRSTglehGRhFOiFxFJOCV6EZGESyvRm9lwM3vHzJaa2aRqynzNzBab2SIzezS2fKeZLYxuu01BKCIi9avWqQTNLA+4CzgNKAbmmdksd18cK9MTuA4Y6u5rzezA2C62unv/zIYtIiLpSqdGPxhY6u7L3H07MAMYkVLmUuAud18L4O7/yWyYIiKyp9JJ9F2AlbHnxdGyuF5ALzP7p5n9y8yGx9a1MrOiaPlXqnoBMxsflSkqKSmpS/wiIlKLWptu6rCfnsAwIB94wcz6uvs6oLu7rzKzQ4Bnzezf7v5efGN3nwZMgzA5eIZiEhER0qvRrwK6xp7nR8viioFZ7r7D3d8HlhASP+6+KrpfBjwPDNjLmEVEpA7SSfTzgJ5m1sPMWgCjgNTRMzMJtXnMrBOhKWeZmXUws5ax5UOBxYiISIOptenG3UvN7ApgDpAH3O/ui8xsMlDk7rOidaeb2WJgJ3CNu682syHAb8xsF+FL5db4aB0REal/5p5bTeKFhYVeVFSU7TBERBoVM5vv7oVVrdMvY0VEEk6JXkQk4ZToRUQSToleRCThlOhFRBJOiV5EJOGU6EVEEk6JXkQk4ZToRUQSToleRCThlOhFRBJOiV5EJOGU6EVEEk6JXkQk4ZToRUQSToleRCTh0kr0ZjbczN4xs6VmNqmaMl8zs8VmtsjMHo0tH2tm70a3sZkKPNX06VBQAM2ahfvp0+vrlUREGpdapxI0szzgLuA0wiTg88xsVnxKQDPrCVwHDHX3tWZ2YLS8I3AjUAg4MD/adm0m38T06TB+PGzZEp6vWBGeA4wZk8lXEhFpfNKp0Q8Glrr7MnffDswARqSUuRS4qyyBu/t/ouVfAua6+5po3VxgeGZCr3D99RVJvsyWLWG5iEhTl06i7wKsjD0vjpbF9QJ6mdk/zexfZja8DttiZuPNrMjMikpKStKPPvLBB3VbLiLSlGSqM7Y50BMYBowG7jWz/dPd2N2nuXuhuxd27ty5zi/erVvdlouINCXpJPpVQNfY8/xoWVwxMMvdd7j7+8ASQuJPZ9u9NmUKtG5deVnr1mG5iEhTl06inwf0NLMeZtYCGAXMSikzk1Cbx8w6EZpylgFzgNPNrIOZdQBOj5Zl1JgxMG0adO8OZuF+2jR1xIqIQBqjbty91MyuICToPOB+d19kZpOBInefRUVCXwzsBK5x99UAZnYz4csCYLK7r6mPNzJmjBK7iEhVzN2zHUMlhYWFXlRUlO0wREQaFTOb7+6FVa3TL2NFRBJOiV5EJOGU6EVEEk6JXkQk4XKuM9bMSoAVe7GLTsCnGQqnsdOxqEzHozIdjwpJOBbd3b3KX5zmXKLfW2ZWVF3Pc1OjY1GZjkdlOh4Vkn4s1HQjIpJwSvQiIgmXxEQ/LdsB5BAdi8p0PCrT8aiQ6GORuDZ6ERGpLIk1ehERiVGiFxFJuMQk+nQmMG8qzKyrmT0Xm6z9u9mOKdvMLM/MXjOzv2Q7lmwzs/3N7Akze9vM3jKz47IdUzaZ2cTo/+RNM/u9mbXKdkyZlohEH5vA/AygDzDazPpkN6qsKgW+5+59gGOB7zTx4wHwXeCtbAeRI34J/M3dewP9aMLHxcy6AFcBhe7+BcKl2EdlN6rMS0SiJ70JzJsMd//I3RdEjzcS/pF3m6u3qTCzfOAs4LfZjiXbzKw9cCJwH4C7b3f3dVkNKvuaA/uaWXOgNfBhluPJuKQk+rQmIW+KzKwAGAC8kuVQsukO4FpgV5bjyAU9gBLggagp67dm1ibbQWWLu68CpgIfAB8B693979mNKvOSkuilCmbWFvgjcLW7b8h2PNlgZmcD/3H3+dmOJUc0BwYC97j7AGAz0GT7tKIpTkcQvgA/D7QxswuzG1XmJSXRN8gk5I2Jme1DSPLT3f3JbMeTRUOBc8xsOaFJ74tm9kh2Q8qqYqDY3cvO8J4gJP6m6lTgfXcvcfcdwJPAkCzHlHFJSfTpTGDeZJiZEdpg33L3X2Q7nmxy9+vcPd/dCwifi2fdPXE1tnS5+8fASjM7PFp0CrA4iyFl2wfAsWbWOvq/OYUEdk7XOjl4Y1DdBOZZDiubhgIXAf82s4XRsh+4+1PZC0lyyJXA9KhStAz4ZpbjyRp3f8XMngAWEEarvUYCL4egSyCIiCRcUppuRESkGkr0IiIJp0QvIpJwSvQiIgmnRC8iknBK9CIiCadELyKScP8fCgTlz4NZpZ4AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "acc = history.history['acc']\n",
    "val_acc = history.history['val_acc']\n",
    "loss = history.history['loss']\n",
    "val_loss = history.history['val_loss']\n",
    "\n",
    "epochs = range(len(acc))\n",
    "\n",
    "plt.plot(epochs, acc, 'bo', label='Training acc')\n",
    "plt.plot(epochs, val_acc, 'b', label='Validation acc')\n",
    "plt.title('Training and validation accuracy')\n",
    "plt.legend()\n",
    "\n",
    "plt.figure()\n",
    "\n",
    "plt.plot(epochs, loss, 'bo', label='Training loss')\n",
    "plt.plot(epochs, val_loss, 'b', label='Validation loss')\n",
    "plt.title('Training and validation loss')\n",
    "plt.legend()\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "As a reminder, in chapter 3, our very first naive approach to this very dataset got us to 88% test accuracy. Unfortunately, our small recurrent network doesn't perform very well at all compared to this baseline (only up to 85% validation accuracy). Part of the problem is that our inputs only consider the first 500 words rather the full sequences -- hence our RNN has access to less information than our earlier baseline model. The remainder of the problem is simply that SimpleRNN isn't very good at processing long sequences, like text. Other types of recurrent layers perform much better. Let's take a look at some more advanced layers."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "pd.DataFrame(history.history).plot(figsize=(8,5))\n",
    "plt.grid(True)\n",
    "plt.gca().set_ylim(0,1)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test loss: 0.7379931807518005\n",
      "Test accuracy: 0.7904000282287598\n",
      "Elapsed time: 0:00:39.72\n"
     ]
    }
   ],
   "source": [
    "start_time = time.time()\n",
    "\n",
    "score = model.evaluate(input_test, y_test, verbose=0)\n",
    "print('Test loss: {}'.format(score[0]))\n",
    "print('Test accuracy: {}'.format(score[1]))\n",
    "\n",
    "elapsed_time = time.time() - start_time\n",
    "print(\"Elapsed time: {}\".format(hms_string(elapsed_time)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## A concrete LSTM example in Keras\n",
    "\n",
    "Now let's switch to more practical concerns: we will set up a model using a LSTM layer and train it on the IMDB data. Here's the network, \n",
    "similar to the one with `SimpleRNN` that we just presented. We only specify the output dimensionality of the LSTM layer, and leave every \n",
    "other argument (there are lots) to the Keras defaults. Keras has good defaults, and things will almost always \"just work\" without you \n",
    "having to spend time tuning parameters by hand."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "157/157 [==============================] - 31s 199ms/step - loss: 0.5514 - acc: 0.7427 - val_loss: 0.3518 - val_acc: 0.8628\n",
      "Epoch 2/10\n",
      "157/157 [==============================] - 31s 195ms/step - loss: 0.3044 - acc: 0.8784 - val_loss: 0.3220 - val_acc: 0.8768\n",
      "Epoch 3/10\n",
      "157/157 [==============================] - 30s 194ms/step - loss: 0.2406 - acc: 0.9071 - val_loss: 0.2975 - val_acc: 0.8864\n",
      "Epoch 4/10\n",
      "157/157 [==============================] - 30s 194ms/step - loss: 0.2035 - acc: 0.9255 - val_loss: 0.2997 - val_acc: 0.8938\n",
      "Epoch 5/10\n",
      "157/157 [==============================] - 30s 194ms/step - loss: 0.1860 - acc: 0.9319 - val_loss: 0.3404 - val_acc: 0.8652\n",
      "Epoch 6/10\n",
      "157/157 [==============================] - 30s 192ms/step - loss: 0.1621 - acc: 0.9415 - val_loss: 0.4042 - val_acc: 0.8628\n",
      "Epoch 7/10\n",
      "157/157 [==============================] - 30s 193ms/step - loss: 0.1487 - acc: 0.9469 - val_loss: 0.3890 - val_acc: 0.8692\n",
      "Epoch 8/10\n",
      "157/157 [==============================] - 31s 196ms/step - loss: 0.1318 - acc: 0.9553 - val_loss: 0.3038 - val_acc: 0.8866\n",
      "Epoch 9/10\n",
      "157/157 [==============================] - 31s 196ms/step - loss: 0.1232 - acc: 0.9578 - val_loss: 0.4056 - val_acc: 0.8680\n",
      "Epoch 10/10\n",
      "157/157 [==============================] - 31s 196ms/step - loss: 0.1142 - acc: 0.9602 - val_loss: 0.3961 - val_acc: 0.8694\n",
      "Elapsed time: 0:05:09.78\n"
     ]
    }
   ],
   "source": [
    "from keras.layers import LSTM\n",
    "\n",
    "start_time = time.time()\n",
    "\n",
    "model2 = Sequential()\n",
    "model2.add(Embedding(max_features, 32))\n",
    "model2.add(LSTM(32))\n",
    "model2.add(Dense(1, activation='sigmoid'))\n",
    "\n",
    "model2.compile(optimizer='rmsprop',\n",
    "              loss='binary_crossentropy',\n",
    "              metrics=['acc'])\n",
    "history2 = model2.fit(input_train, y_train,\n",
    "                    epochs=10,\n",
    "                    batch_size=128,\n",
    "                    validation_split=0.2)\n",
    "\n",
    "elapsed_time = time.time() - start_time\n",
    "print(\"Elapsed time: {}\".format(hms_string(elapsed_time)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "acc = history2.history['acc']\n",
    "val_acc = history2.history['val_acc']\n",
    "loss = history2.history['loss']\n",
    "val_loss = history2.history['val_loss']\n",
    "\n",
    "epochs = range(len(acc))\n",
    "\n",
    "plt.plot(epochs, acc, 'bo', label='Training acc')\n",
    "plt.plot(epochs, val_acc, 'b', label='Validation acc')\n",
    "plt.title('Training and validation accuracy')\n",
    "plt.legend()\n",
    "\n",
    "plt.figure()\n",
    "\n",
    "plt.plot(epochs, loss, 'bo', label='Training loss')\n",
    "plt.plot(epochs, val_loss, 'b', label='Validation loss')\n",
    "plt.title('Training and validation loss')\n",
    "plt.legend()\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "pd.DataFrame(history2.history).plot(figsize=(8,5))\n",
    "plt.grid(True)\n",
    "plt.gca().set_ylim(0,1)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test loss: 0.43439844250679016\n",
      "Test accuracy: 0.8565199971199036\n",
      "Elapsed time: 0:00:18.60\n"
     ]
    }
   ],
   "source": [
    "start_time = time.time()\n",
    "\n",
    "score = model2.evaluate(input_test, y_test, verbose=0)\n",
    "print('Test loss: {}'.format(score[0]))\n",
    "print('Test accuracy: {}'.format(score[1]))\n",
    "\n",
    "elapsed_time = time.time() - start_time\n",
    "print(\"Elapsed time: {}\".format(hms_string(elapsed_time)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Let's implement some models based on the data above from Chollet"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### Implement Early stopping on the simple RNN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "from tensorflow.keras.callbacks import EarlyStopping\n",
    "from keras.layers import LSTM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "157/157 [==============================] - 14s 90ms/step - loss: 0.6114 - acc: 0.6708 - val_loss: 0.5297 - val_acc: 0.7802\n",
      "Epoch 2/10\n",
      "157/157 [==============================] - 15s 93ms/step - loss: 0.4047 - acc: 0.8417 - val_loss: 0.3780 - val_acc: 0.8436\n",
      "Epoch 3/10\n",
      "157/157 [==============================] - 15s 94ms/step - loss: 0.2886 - acc: 0.8840 - val_loss: 0.4927 - val_acc: 0.7578\n",
      "Elapsed time: 0:00:44.77\n"
     ]
    }
   ],
   "source": [
    "start_time = time.time()\n",
    "\n",
    "# Looking at the earlier graph, we are using 0 patiance\n",
    "early_stopping3 = EarlyStopping(\n",
    "    patience=0,\n",
    "    mode='auto')\n",
    "\n",
    "\n",
    "model3 = Sequential()\n",
    "model3.add(Embedding(max_features, 32))\n",
    "model3.add(SimpleRNN(32))\n",
    "model3.add(Dense(1, activation='sigmoid'))\n",
    "\n",
    "model3.compile(optimizer='rmsprop', loss='binary_crossentropy', metrics=['acc'])\n",
    "history3 = model3.fit(input_train, y_train,\n",
    "                    epochs=10,\n",
    "                    batch_size=128,\n",
    "                    validation_split=0.2,\n",
    "            callbacks=[early_stopping3])\n",
    "\n",
    "elapsed_time = time.time() - start_time\n",
    "print(\"Elapsed time: {}\".format(hms_string(elapsed_time)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "pd.DataFrame(history3.history).plot(figsize=(8,5))\n",
    "plt.grid(True)\n",
    "plt.gca().set_ylim(0,1)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test loss: 0.48372167348861694\n",
      "Test accuracy: 0.763480007648468\n",
      "Elapsed time: 0:00:09.08\n"
     ]
    }
   ],
   "source": [
    "start_time = time.time()\n",
    "\n",
    "score = model3.evaluate(input_test, y_test, verbose=0)\n",
    "print('Test loss: {}'.format(score[0]))\n",
    "print('Test accuracy: {}'.format(score[1]))\n",
    "\n",
    "elapsed_time = time.time() - start_time\n",
    "print(\"Elapsed time: {}\".format(hms_string(elapsed_time)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### Short analysis (more later)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "A rather low test accuracy with high loss. The model semms to perform better after only one epoch, looking at the graphs. It quickly became overfitted"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### Implement LSTM(32) with early stopping"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "157/157 [==============================] - 32s 204ms/step - loss: 0.5136 - acc: 0.7534 - val_loss: 0.3468 - val_acc: 0.8650\n",
      "Epoch 2/10\n",
      "157/157 [==============================] - 30s 193ms/step - loss: 0.2944 - acc: 0.8858 - val_loss: 0.3945 - val_acc: 0.8320\n",
      "Elapsed time: 0:01:04.77\n"
     ]
    }
   ],
   "source": [
    "start_time = time.time()\n",
    "\n",
    "# Looking at the earlier graph, we are using 1 patiance\n",
    "early_stopping4 = EarlyStopping(\n",
    "    patience=1,\n",
    "    mode='auto')\n",
    "\n",
    "model4 = Sequential()\n",
    "model4.add(Embedding(max_features, 32))\n",
    "model4.add(LSTM(32))\n",
    "model4.add(Dense(1, activation='sigmoid'))\n",
    "\n",
    "model4.compile(optimizer='rmsprop',\n",
    "              loss='binary_crossentropy',\n",
    "              metrics=['acc'])\n",
    "history4 = model4.fit(input_train, y_train,\n",
    "                    epochs=10,\n",
    "                    batch_size=128,\n",
    "                    validation_split=0.2,\n",
    "            callbacks=[early_stopping4])\n",
    "\n",
    "elapsed_time = time.time() - start_time\n",
    "print(\"Elapsed time: {}\".format(hms_string(elapsed_time)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "pd.DataFrame(history4.history).plot(figsize=(8,5))\n",
    "plt.grid(True)\n",
    "plt.gca().set_ylim(0,1)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test loss: 0.3930627405643463\n",
      "Test accuracy: 0.8315200209617615\n",
      "Elapsed time: 0:00:19.50\n"
     ]
    }
   ],
   "source": [
    "start_time = time.time()\n",
    "\n",
    "score = model4.evaluate(input_test, y_test, verbose=0)\n",
    "print('Test loss: {}'.format(score[0]))\n",
    "print('Test accuracy: {}'.format(score[1]))\n",
    "\n",
    "elapsed_time = time.time() - start_time\n",
    "print(\"Elapsed time: {}\".format(hms_string(elapsed_time)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### Short analysis (more later)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The basic LSTM outperformed the Simple RNN directly. Still high loss on the test set. The validation accuracy becomes worse the more epoch the model run"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### Implement LSTM(128) with early stopping"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "157/157 [==============================] - 124s 791ms/step - loss: 0.5364 - acc: 0.7397 - val_loss: 0.3344 - val_acc: 0.8616\n",
      "Epoch 2/10\n",
      "157/157 [==============================] - 134s 853ms/step - loss: 0.3563 - acc: 0.8530 - val_loss: 0.3159 - val_acc: 0.8766\n",
      "Epoch 3/10\n",
      "157/157 [==============================] - 136s 866ms/step - loss: 0.2891 - acc: 0.8898 - val_loss: 0.3067 - val_acc: 0.8750\n",
      "Epoch 4/10\n",
      "157/157 [==============================] - 136s 869ms/step - loss: 0.2363 - acc: 0.9125 - val_loss: 0.6423 - val_acc: 0.8276\n",
      "Elapsed time: 0:08:55.64\n"
     ]
    }
   ],
   "source": [
    "start_time = time.time()\n",
    "\n",
    "# Using the same 1 patiance, as LSTM(32)\n",
    "early_stopping5 = EarlyStopping(\n",
    "    patience=1,\n",
    "    mode='auto')\n",
    "\n",
    "model5 = Sequential()\n",
    "model5.add(Embedding(max_features, 32))\n",
    "model5.add(LSTM(128))\n",
    "model5.add(Dense(1, activation='sigmoid'))\n",
    "\n",
    "model5.compile(optimizer='rmsprop',\n",
    "              loss='binary_crossentropy',\n",
    "              metrics=['acc'])\n",
    "history5 = model5.fit(input_train, y_train,\n",
    "                    epochs=10,\n",
    "                    batch_size=128,\n",
    "                    validation_split=0.2,\n",
    "            callbacks=[early_stopping5])\n",
    "\n",
    "elapsed_time = time.time() - start_time\n",
    "print(\"Elapsed time: {}\".format(hms_string(elapsed_time)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "pd.DataFrame(history5.history).plot(figsize=(8,5))\n",
    "plt.grid(True)\n",
    "plt.gca().set_ylim(0,1)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test loss: 0.6742952466011047\n",
      "Test accuracy: 0.8130000233650208\n",
      "Elapsed time: 0:00:59.26\n"
     ]
    }
   ],
   "source": [
    "start_time = time.time()\n",
    "\n",
    "score = model5.evaluate(input_test, y_test, verbose=0)\n",
    "print('Test loss: {}'.format(score[0]))\n",
    "print('Test accuracy: {}'.format(score[1]))\n",
    "\n",
    "elapsed_time = time.time() - start_time\n",
    "print(\"Elapsed time: {}\".format(hms_string(elapsed_time)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### Short analysis (more later)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Easily overfitted with a high test loss."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### Implement BiDir LSTM(32) with early stopping"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "157/157 [==============================] - 69s 441ms/step - loss: 0.5773 - acc: 0.7084 - val_loss: 0.5011 - val_acc: 0.7652\n",
      "Epoch 2/10\n",
      "157/157 [==============================] - 67s 430ms/step - loss: 0.3561 - acc: 0.8543 - val_loss: 0.3111 - val_acc: 0.8730\n",
      "Epoch 3/10\n",
      "157/157 [==============================] - 65s 413ms/step - loss: 0.2618 - acc: 0.8996 - val_loss: 0.2962 - val_acc: 0.8772\n",
      "Epoch 4/10\n",
      "157/157 [==============================] - 64s 410ms/step - loss: 0.2197 - acc: 0.9211 - val_loss: 0.2990 - val_acc: 0.8764\n",
      "Elapsed time: 0:04:31.41\n"
     ]
    }
   ],
   "source": [
    "start_time = time.time()\n",
    "\n",
    "# Using the same 1 patiance, as LSTM(32)\n",
    "early_stopping8 = EarlyStopping(\n",
    "    patience=0,\n",
    "    mode='auto')\n",
    "\n",
    "model8 = Sequential()\n",
    "model8.add(Embedding(max_features, 32))\n",
    "model8.add(keras.layers.Bidirectional(LSTM(32)))\n",
    "model8.add(Dense(1, activation='sigmoid'))\n",
    "\n",
    "model8.compile(optimizer='rmsprop',\n",
    "              loss='binary_crossentropy',\n",
    "              metrics=['acc'])\n",
    "history8 = model8.fit(input_train, y_train,\n",
    "                    epochs=10,\n",
    "                    batch_size=128,\n",
    "                    validation_split=0.2,\n",
    "            callbacks=[early_stopping8])\n",
    "\n",
    "elapsed_time = time.time() - start_time\n",
    "print(\"Elapsed time: {}\".format(hms_string(elapsed_time)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "pd.DataFrame(history8.history).plot(figsize=(8,5))\n",
    "plt.grid(True)\n",
    "plt.gca().set_ylim(0,1)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test loss: 0.3078640401363373\n",
      "Test accuracy: 0.8717600107192993\n",
      "Elapsed time: 0:00:22.65\n"
     ]
    }
   ],
   "source": [
    "start_time = time.time()\n",
    "\n",
    "score = model8.evaluate(input_test, y_test, verbose=0)\n",
    "print('Test loss: {}'.format(score[0]))\n",
    "print('Test accuracy: {}'.format(score[1]))\n",
    "\n",
    "elapsed_time = time.time() - start_time\n",
    "print(\"Elapsed time: {}\".format(hms_string(elapsed_time)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Analysis"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "SimpleRNN 32\n",
    "Test loss: 0.48372167348861694<br>\n",
    "Test accuracy: 0.763480007648468<br>\n",
    "<br>   \n",
    "LSTM 32<br>\n",
    "Test loss: 0.3930627405643463<br>\n",
    "Test accuracy: 0.8315200209617615<br>\n",
    "<br>    \n",
    "LSTM 128<br>\n",
    "Test loss: 0.3930627405643463<br>\n",
    "Test accuracy: 0.8315200209617615<br>\n",
    "<br>    \n",
    "Bidir LSTM 32<br>\n",
    "Test loss: 0.3078640401363373<br>\n",
    "Test accuracy: 0.8717600107192993<br>\n",
    "<br>\n",
    "\n",
    "The simpleRNN model was \"simple\" and performed worst. LSTM models outperformed the simpleRNN. Whether having 32 or 128 neurons/units in the model, did not make any difference, but running a Bidirectional LSTM got made a difference. \n",
    "<br>\n",
    "The Bidir LSTM 32 achieve the highest accuracy on the test data (0,87) as well as lowest test loss (0,31)\n",
    "<br>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### Let's make a test on new test senteces"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Code inspired from:<br>\n",
    "https://towardsdatascience.com/simple-neural-network-sentiment-analyser-using-keras-179bc9f50fbc\n",
    "<br>\n",
    "http://drunkendatascience.com/simple-neural-network-sentiment-analyser-using-keras/\n",
    "<br>\n",
    "https://builtin.com/data-science/how-build-neural-network-keras\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.preprocessing import sequence\n",
    "import string\n",
    "import numpy as np\n",
    "\n",
    "max_features = 10000\n",
    "maxlen = 500\n",
    "word_dict = imdb.get_word_index() \n",
    "\n",
    "def analyze(text, modelnumber, debug = False): \n",
    "    # Prepare the input by removing punctuation characters, converting characters \n",
    "    # to lower case, and removing words containing numbers \n",
    "    \n",
    "    translator = str.maketrans('', '', string.punctuation) \n",
    "    text = text.translate(translator) \n",
    "    text = text.lower().split(' ') \n",
    "    text = [word for word in text if word.isalpha()] \n",
    "    if debug: print(text)\n",
    "    \n",
    "    #padded_input = \"\"\n",
    "    \n",
    "    # Generate an input tensor \n",
    "    \n",
    "    input = [1] \n",
    "    for word in text: \n",
    "        if word in word_dict and word_dict[word] < max_features: \n",
    "            input.append(word_dict[word]) \n",
    "            if debug: print(word_dict.get(word))\n",
    "        else: \n",
    "            input.append(2)  \n",
    "        \n",
    "        \n",
    "    padded_input = sequence.pad_sequences([input], maxlen=maxlen) \n",
    "        \n",
    "    # Invoke the model and return the result \n",
    "    if debug: print(padded_input)\n",
    "    \n",
    "    if modelnumber == 1:\n",
    "        result = model.predict(np.array([padded_input][0]))[0][0]\n",
    "    if modelnumber == 2:\n",
    "        result = model2.predict(np.array([padded_input][0]))[0][0]\n",
    "    if modelnumber == 3:\n",
    "        result = model3.predict(np.array([padded_input][0]))[0][0]\n",
    "    if modelnumber == 4:\n",
    "        result = model4.predict(np.array([padded_input][0]))[0][0]\n",
    "    if modelnumber == 5:\n",
    "        result = model5.predict(np.array([padded_input][0]))[0][0]\n",
    "    if modelnumber == 6:\n",
    "        result = model8.predict(np.array([padded_input][0]))[0][0]\n",
    "\n",
    "    return result"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "def decode(text):\n",
    "    index = imdb.get_word_index()\n",
    "    reverse_index = dict([(value, key) for (key, value) in index.items()]) \n",
    "    decoded = \" \".join( [reverse_index.get(i - 3, \"#\") for i in text] )\n",
    "    return decoded"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "model  1   score: 0.00 : That movie was absolutely awful\n",
      "model  1   score: 0.23 : This is a bad review, terrible, I hate Huffington Post\n",
      "model  1   score: 0.90 : This movie is not worth the money\n",
      "model  1   score: 0.95 : The only positive thing with this movie is the music\n",
      "model  1   score: 0.03 : The acting was a bit lacking\n",
      "model  1   score: 0.03 : The film was creative and surprising\n",
      "model  1   score: 0.96 : This movie is worth the money\n",
      "model  1   score: 0.76 : I loved the movie. I just have to see it\n",
      "model  1   score: 0.98 : best ever I must say it is lovely\n",
      "model  1   score: 0.00 : This is a must see\n",
      "model  1   score: 0.02 : This is a great review, an amazing blog, I love Huffington Post\n",
      "model  1   score: 0.14 : Absolutely fantastic!\n",
      " \n",
      "model  2   score: 0.09 : That movie was absolutely awful\n",
      "model  2   score: 0.37 : This is a bad review, terrible, I hate Huffington Post\n",
      "model  2   score: 0.79 : This movie is not worth the money\n",
      "model  2   score: 0.35 : The only positive thing with this movie is the music\n",
      "model  2   score: 0.51 : The acting was a bit lacking\n",
      "model  2   score: 0.13 : The film was creative and surprising\n",
      "model  2   score: 0.79 : This movie is worth the money\n",
      "model  2   score: 0.74 : I loved the movie. I just have to see it\n",
      "model  2   score: 0.61 : best ever I must say it is lovely\n",
      "model  2   score: 0.46 : This is a must see\n",
      "model  2   score: 0.57 : This is a great review, an amazing blog, I love Huffington Post\n",
      "model  2   score: 0.18 : Absolutely fantastic!\n",
      " \n",
      "model  3   score: 0.60 : That movie was absolutely awful\n",
      "model  3   score: 0.51 : This is a bad review, terrible, I hate Huffington Post\n",
      "model  3   score: 0.91 : This movie is not worth the money\n",
      "model  3   score: 0.52 : The only positive thing with this movie is the music\n",
      "model  3   score: 0.85 : The acting was a bit lacking\n",
      "model  3   score: 0.75 : The film was creative and surprising\n",
      "model  3   score: 0.92 : This movie is worth the money\n",
      "model  3   score: 0.83 : I loved the movie. I just have to see it\n",
      "model  3   score: 0.84 : best ever I must say it is lovely\n",
      "model  3   score: 0.71 : This is a must see\n",
      "model  3   score: 0.66 : This is a great review, an amazing blog, I love Huffington Post\n",
      "model  3   score: 0.60 : Absolutely fantastic!\n",
      " \n",
      "model  4   score: 0.38 : That movie was absolutely awful\n",
      "model  4   score: 0.45 : This is a bad review, terrible, I hate Huffington Post\n",
      "model  4   score: 0.74 : This movie is not worth the money\n",
      "model  4   score: 0.52 : The only positive thing with this movie is the music\n",
      "model  4   score: 0.67 : The acting was a bit lacking\n",
      "model  4   score: 0.54 : The film was creative and surprising\n",
      "model  4   score: 0.74 : This movie is worth the money\n",
      "model  4   score: 0.78 : I loved the movie. I just have to see it\n",
      "model  4   score: 0.64 : best ever I must say it is lovely\n",
      "model  4   score: 0.54 : This is a must see\n",
      "model  4   score: 0.48 : This is a great review, an amazing blog, I love Huffington Post\n",
      "model  4   score: 0.44 : Absolutely fantastic!\n",
      " \n",
      "model  5   score: 0.14 : That movie was absolutely awful\n",
      "model  5   score: 0.34 : This is a bad review, terrible, I hate Huffington Post\n",
      "model  5   score: 0.84 : This movie is not worth the money\n",
      "model  5   score: 0.32 : The only positive thing with this movie is the music\n",
      "model  5   score: 0.60 : The acting was a bit lacking\n",
      "model  5   score: 0.40 : The film was creative and surprising\n",
      "model  5   score: 0.85 : This movie is worth the money\n",
      "model  5   score: 0.76 : I loved the movie. I just have to see it\n",
      "model  5   score: 0.61 : best ever I must say it is lovely\n",
      "model  5   score: 0.45 : This is a must see\n",
      "model  5   score: 0.52 : This is a great review, an amazing blog, I love Huffington Post\n",
      "model  5   score: 0.24 : Absolutely fantastic!\n",
      " \n",
      "model  6   score: 0.29 : That movie was absolutely awful\n",
      "model  6   score: 0.48 : This is a bad review, terrible, I hate Huffington Post\n",
      "model  6   score: 0.76 : This movie is not worth the money\n",
      "model  6   score: 0.27 : The only positive thing with this movie is the music\n",
      "model  6   score: 0.52 : The acting was a bit lacking\n",
      "model  6   score: 0.34 : The film was creative and surprising\n",
      "model  6   score: 0.75 : This movie is worth the money\n",
      "model  6   score: 0.72 : I loved the movie. I just have to see it\n",
      "model  6   score: 0.74 : best ever I must say it is lovely\n",
      "model  6   score: 0.61 : This is a must see\n",
      "model  6   score: 0.66 : This is a great review, an amazing blog, I love Huffington Post\n",
      "model  6   score: 0.43 : Absolutely fantastic!\n",
      " \n",
      " \n",
      " \n",
      "model 6  score: 0.29 : That movie was absolutely awful\n",
      "model 6  score: 0.76 : This movie is not worth the money\n",
      "model 6  score: 0.27 : The only positive thing with this movie is the music\n",
      "model 6  score: 0.52 : The acting was a bit lacking\n",
      "model 6  score: 0.34 : The film was creative and surprising\n",
      "model 6  score: 0.43 : Absolutely fantastic!\n"
     ]
    }
   ],
   "source": [
    "test_sentences = [\n",
    "  \"That movie was absolutely awful\",\n",
    "  \"This is a bad review, terrible, I hate Huffington Post\",\n",
    "  \"This movie is not worth the money\",\n",
    "  \"The only positive thing with this movie is the music\",\n",
    "  \"The acting was a bit lacking\",\n",
    "  \"The film was creative and surprising\",\n",
    "  \"This movie is worth the money\",\n",
    "  \"I loved the movie. I just have to see it\",\n",
    "  \"best ever I must say it is lovely\",\n",
    "  \"This is a must see\",\n",
    "  \"This is a great review, an amazing blog, I love Huffington Post\",\n",
    "  \"Absolutely fantastic!\"   \n",
    "]\n",
    "\n",
    "for i in range (1,7):\n",
    "    for s in test_sentences:\n",
    "        score = analyze(s,i)\n",
    "        print(\"model \",i, \"  score: {:.2f} :\".format(score) , s)\n",
    "    print(\" \")\n",
    "\n",
    "test_sentences = [\n",
    "  \"That movie was absolutely awful\",\n",
    "  \"This movie is not worth the money\",\n",
    "  \"The only positive thing with this movie is the music\",\n",
    "  \"The acting was a bit lacking\",\n",
    "  \"The film was creative and surprising\",\n",
    "  \"Absolutely fantastic!\"   \n",
    "]\n",
    "\n",
    "print(\" \")\n",
    "print(\" \")\n",
    "for s in test_sentences:\n",
    "    score = analyze(s,6)\n",
    "    print(\"model 6  score: {:.2f} :\".format(score) , s)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<b>The Model called 6 above and below is equivalent to model 8 (Bidir LSTM 32)</b>\n",
    "<br>\n",
    "The Bidir model seems to perform best. The Sentences are set from worst (top) to best (bottom). It is not a clear cut estimation of the model. Though the there seems to be higher values further down. But \"Absolute fantastic\" should have a better score.<br><br>\n",
    "The sentence \"This movie is not wirth the money\", get a much higher score than it should. Probably because the \"not\" is not ranked as it should in the sentence. It is also possible that sentences in the training data is not usually phrased that way.<br><br>\n",
    "Looking only at the last 6 sentences that should be evaluated, it is the \"not worth the money\" that stands out. Otherwise the 50% better sentences had better scores, then the 50% wors sentences. But the ranking among the better senteces was not that good.\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "######  NON IN THE TASK (for fun) Some movie review"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " \n",
      " \n",
      "model 6  score: 0.07\n",
      "model 6  score: 0.24\n",
      "model 6  score: 0.07\n"
     ]
    }
   ],
   "source": [
    "test_sentences = [\n",
    "  \"Longer than the Gotterdammerung… and not nearly as funny!” Noël Coward is reputed to have quipped on seeing the protracted (four and a half hour) world premiere of Lerner and Loewe’s Camelot in Toronto in 1960.Afflicted by director Moss Hart having a heart attack and Lerner getting a stomach ulcer, the pair’s arduously achieved follow-up to My Fair Lady was duly pruned and – graced by Richard Burton as King Arthur and Julie Andrews as Guenevere – did respectable business on Broadway, during which time it acquired the boosting fandom of JFK. Yet even at the time it looked much inferior to their earlier masterpiece. And even running now at just 80 minutes in an outdoor concert staging at the Watermill, it feels like it’s all moat and no castle. Caught twixt jest and earnest, pre-Python but aware of chivalry’s pomposity, song after song trots by, tilting between effortful comedy and chaste emoting. It’s soporific stuff; were it not for the fresh air I’d have slept-a-lot.No question, it’s a miracle that in a matter of weeks director Paul Hart has marshalled and rehearsed a company of 10 (the majority actor-musicians) and ensured they can take their place in a socially distanced way on a pavilion-style stage festooned with coats of arms and fairy-lights. I salute the collective effort and applaud individual contributions. As Guenevere, Caroline Sheen has a presence and voice as gleaming as a polished goblet. Her real-life husband Michael Jibson – an Olivier-winning King George in Hamilton – exudes an essential Arthurian decency even if, in his suited attire, he hardly looks battle-ready. Marc Antolin’s conceited Lancelot has his moments, too. It’s largely (understandably) a script in hand affair, but the problem is the by-the-book nature of the material. That holy grail of musicals, a vital raison d’etre, is missing.\",\n",
    "  \"With more than 40 live-action versions of A Christmas Carol already in the can, what the Dickens is left to be done with it? Sibling filmmakers Jacqui and David Morris, who directed 2018’s superb Nureyev documentary, have come up with a compelling new angle: a danced-through version of the fable, performed in a stage-like space by a specially convened troupe, and with the characters’ voices supplied by an all-star (though off-screen) cast. The result is a mostly beguiling, intermittently jarring hybrid – much closer to “normal” cinema than, say, a straightforward screening of a play, but with a certain frisson of live-ness that works in its favour more often than not. Aimed squarely at families, it’s also one for children (and adults) who are already familiar with the original Charles Dickens novella – or, failing that, perhaps the immortal Alastair Sim and/or Muppets-led adaptations thereof. The appeal of this new Christmas Carol is rooted in its own particular adaptational choices: the sets that resemble grisly, Gustave Doré-like etchings, or the intermingling of ballet and body-popping, all captured by a camera that weaves between the dancers like a ghostly understudy just as often as it withdraws to a more conventional vantage point in the stalls. Siân Phillips plays the narrator, reading aloud from the original Dickens – or at least a well-whittled digest. The scene is a Victorian parlour, where two children are re-enacting the tale in their model theatre for their parents and younger sister, the latter of whom watches the whole thing come to life before her eyes. It really takes off when Michael Nunn’s Scrooge, voiced by Simon Russell Beale with (intentional or otherwise) nods to Sim, returns to his gloomy suite of rooms and encounters Marley’s ghost – who is danced by Russell Maliphant and voiced by Andy Serkis with such harmonious, blood-curdling gusto that it almost justifies the film’s voice-and-body-splitting gambit in one fell swoop. Serkis is so good, in fact, that he makes you crave another live-action version in which he’s physically present – that is, in addition to last year’s Peaky Blinders-inflected BBC version, in which he played the Ghost of Christmas Past. A Christmas Carol Martin Freeman also makes a convincing voice of Bob Cratchit, because of course he does, though alas, many of the other pairings just don’t mesh smoothly enough to prevent the device itself from feeling awkwardly conspicuous. Daniel Kaluuya doesn’t provide the jovial boom implied by Mikey Boats’s strapping Ghost of Christmas Present, while Carey Mulligan barely registers as the voice of Scrooge’s great lost love Belle, who is danced with great delicacy and warmth by Grace Jabbari.In fact, the most effective sequences here tend to be the ones least driven by dialogue: Marley’s ghost wrestling with his chains; the foretold death of Tiny Tim; the energised ensemble bustle of the London streets. Immersive rather than absorbing, it’s a film that makes you pine for the theatre, where this kind of offbeat treatment tends to land with fewer bumps.\",\n",
    "  \"Before I point out the bleeding obvious – namely that Six is a marvellous show, dripping with invention and intelligence, and one which brings not just happiness in the moment but hope for the future of the British musical – I need to attack the fatuous. Its composers Toby Marlow and Lucy Moss completed their witty, magpie-minded sonic onslaught on behalf of Henry VIII’s wives while in their final year at Cambridge (2016-17); it has since become a theatrical phenomenon and they deserve the riches, opportunities and acclaim that have come their way.  Cambridge University, you will be aware, has recently been at the centre of a freedom of speech storm. The governing body has just voted – crucially – for “tolerance” of differing views rather than required “respect” for them, a means of upholding/ensuring academic freedom and fairness. That message should be instinctive to the university’s graduate diaspora and amplified by it. But in a lamentable episode last month, Marlow and Moss cancelled an interview with Graham Norton on his Radio 2 programme because they “didn’t feel comfortable being on the same billing as JK Rowling”; they were reluctant to “feature as guests on a show where she is promoting her work until she has made significant efforts to make amends with the trans community”.  A snub, then; a quasi-boycott. Given that Six has transferred from the Arts to the (beautifully refurbished) Lyric, you wonder how they sleep at night; after all, it’s only a short stroll along Shaftesbury Avenue to another Nimax theatre, the Palace, home to the (Covid-suspended) Harry Potter. Eek! I jest. But, seriously, theatre upholds diversity of opinion and the right to dissent, or it is nothing. There’s an onus on those who work in it to fight against, not uphold, cancel culture. Minor though it might seem to disdain contact with a fellow writer – and Rowling is insulated by her queenly wealth – it’s part of a trend of treating even reasoned opinions as heretical, worthy of excommunication. And there’s an added irony here. Laying aside the fact that Henry’s appalling record as marriage material stemmed from the perceived usefulness – or otherwise – of his spouses as reproductive vessels, the winning thrust of Six is about letting muted, marginalised (female) voices be heard. Amy Shaw in Six at the Lyric Theatre Credit: Pamela Raith The evening is an 80-minute irreverent blast from the past aimed at setting the record straight. The format is mock Tudor talent-show, with us in the imaginary position of Simon Cowell-like judges. The sextet (in combatively playful, period-allusive attire – think Vivienne Westwood meets Valkyries) vie with each other, via vocal prowess and potted autobiography, to assert their claim to have had it the worst, and gain victimhood’s crown. There’s ample room, then, for comic cat-fights – preening, posturing and backchat. But the superficial bling carries a deeper sting. Who, actually, did suffer the most? Is it no contest if you got beheaded? What about the broken heart, the dashed romantic dreams, the death caused by child-birth? Taking modish liberties to articulate the real experience of these corseted characters, the show demands we put ourselves in their restrictive place: would their mindsets – their outrage, their hurt – be so very far removed from that of their sisters today? History is flipped on its head to become herstory and you’re left wondering if Henry’s fame owes much, if not all, to them (his reign in fact the sideshow). The production (directed by Moss, and Jamie Armitage) is more polished than ever – the choreography and lighting so synchronised it’s like the deftest human embroidery. None of the cast (invidious to single them out) misses a beat – or impish wink – as they hurtle through Hamilton-like rap, heart-stopping soul, Hi-NRG beats and girl-power anthems. Covid alas prevented Broadway from seeing Six – the curtain fell on opening night. Albeit the guillotine of Tier 3 hangs over London, the West End can still be royally pleased with itself for pulling off this riotous, life-affirming act of restoration\"\n",
    "]\n",
    "\n",
    "# 2 star: https://www.telegraph.co.uk/theatre/what-to-see/camelot-watermill-theatre-newbury-review-medieval-musical-moat/\n",
    "# 3 star: https://www.telegraph.co.uk/films/0/christmas-carol-review-simon-russell-beale-raises-ghost-festive/\n",
    "# 5 star: https://www.telegraph.co.uk/theatre/what-to-see/six-lyric-theatre-review-beacon-hope-future-british-musicals/\n",
    "\n",
    "\n",
    "print(\" \")\n",
    "print(\" \")\n",
    "for s in test_sentences:\n",
    "    score = analyze(s,6)\n",
    "    print(\"model 6  score: {:.2f}\".format(score))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### Analysis\n",
    "The model could not be used on long reviews. The 2 star and 5 star review got the same score\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " \n",
      " \n",
      "model 6  score: 0.50\n",
      "model 6  score: 0.74\n",
      "model 6  score: 0.27\n"
     ]
    }
   ],
   "source": [
    "# Just headlines\n",
    "\n",
    "test_sentences = [\n",
    "  \"This outdoor concert staging is a valiant effort under difficult circumstances – but this 1960 musical isn't worth resurrecting\",\n",
    "  \"In this danced version of Dickens’s tale, Russell Beale, Martin Freeman and Carey Mulligan remind us on screen what we’re missing on stage\",\n",
    "  \"This marvellous mock-Tudor musical celebrates muted female voices – a lesson its composers should apply in their own lives\"\n",
    "]\n",
    "\n",
    "# 2 star: https://www.telegraph.co.uk/theatre/what-to-see/camelot-watermill-theatre-newbury-review-medieval-musical-moat/\n",
    "# 3 star: https://www.telegraph.co.uk/films/0/christmas-carol-review-simon-russell-beale-raises-ghost-festive/\n",
    "# 5 star: https://www.telegraph.co.uk/theatre/what-to-see/six-lyric-theatre-review-beacon-hope-future-british-musicals/\n",
    "\n",
    "\n",
    "print(\" \")\n",
    "print(\" \")\n",
    "for s in test_sentences:\n",
    "    score = analyze(s,6)\n",
    "    print(\"model 6  score: {:.2f}\".format(score))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### Analysis\n",
    "Looking only at intros, the scores where more differentiated, but this time the 5 star review got the lowest sentoment"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "######  NON IN THE TASK (for fun) Some Tweets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " \n",
      " \n",
      "model 6  score: 0.49 : I had a GREAT week, thanks to YOU! If you need anything, ple\n",
      "model 6  score: 0.65 : My heart goes out to the Malaysian people. This is such a tr\n",
      "model 6  score: 0.78 : I just attended a group wedding for people who have been wai\n",
      "model 6  score: 0.17 : The Supreme Court had ZERO interest in the merits of the gre\n",
      "model 6  score: 0.18 : Now that the Biden Administration will be a scandal plagued \n",
      "model 6  score: 0.58 : Here’s a great way to call voters in Georgia and help them g\n",
      "model 6  score: 0.21 : Let’s all do our part this Thanksgiving to keep people safe \n",
      "model 6  score: 0.07 : IT’S HERE! IT’S REALLY HERE! I finally got back my smile! Ho\n",
      "model 6  score: 0.58 : When you need a reminder that you have the power in you - ev\n",
      "model 6  score: 0.63 : Always great to return to Spain and to Catalunya, always har\n",
      "model 6  score: 0.33 : Felling good and healthy!\n"
     ]
    }
   ],
   "source": [
    "test_sentences = [\n",
    "  \"I had a GREAT week, thanks to YOU! If you need anything, please reach out. ❤️ ❤️ ❤️ #WorldSmileDay pic.twitter.com/ZpVmQPmcyc\", \n",
    "  \"My heart goes out to the Malaysian people. This is such a tragedy. Words can't express how sad it is. I wish we could just have peace. #MH17\",\n",
    "  \"I just attended a group wedding for people who have been waiting years to get married. Just wow. #equality #lovewins\",\n",
    "  \"The Supreme Court had ZERO interest in the merits of the greatest voter fraud ever perpetrated on the United States of America. All they were interested in is “standing”, which makes it very difficult for the President to present a case on the merits. 75,000,000 votes!\",\n",
    "  \"Now that the Biden Administration will be a scandal plagued mess for years to come, it is much easier for the Supreme Court of the United States to follow the Constitution and do what everybody knows has to be done. They must show great Courage & Wisdom. Save the USA!!!\",\n",
    "  \"Here’s a great way to call voters in Georgia and help them get ready to vote. A couple hours this weekend could help Democrats win the Senate.\",\n",
    "  \"Let’s all do our part this Thanksgiving to keep people safe and healthy. Celebrate virtually, if you can. Wear a mask. And as always, listen to the experts. The choices you make could save lives.\",\n",
    "  \"IT’S HERE! IT’S REALLY HERE! I finally got back my smile! Hope this record puts one on your face #SMILE IS OUT EVERYWHERE NOW! LOVE YOU GUYS SO MUCH ENJOY (sent from my hospital bed lol) \",\n",
    "  \"When you need a reminder that you have the power in you - even tomorrow at the polls - here’s a playlist http://katy.to/empowered\",\n",
    "  \"Always great to return to Spain and to Catalunya, always hard to play in Camp Nou against one of the best teams I ever faced. Today we were a team of Champions! A true, strong and united family! Playing like this, we have nothing to fear until the end of the season... Let’s go!\",\n",
    "  \"Felling good and healthy!\",\n",
    "    \n",
    "]\n",
    "\n",
    "# Random\n",
    "# Random\n",
    "# Random\n",
    "# Donald Trump\n",
    "# Donald Trump\n",
    "# Barack Obama\n",
    "# Barack Obama\n",
    "# Katey Perry\n",
    "# Katey Perry\n",
    "# Christiano Ronaldo\n",
    "# Christiano Ronaldo\n",
    "\n",
    "print(\" \")\n",
    "print(\" \")\n",
    "for s in test_sentences:\n",
    "    score = analyze(s,6)\n",
    "    s_short = s[:60]\n",
    "    print(\"model 6  score: {:.2f} :\".format(score), s_short )\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### Analysis\n",
    "The sentiment scores are here and ther, but the negative Donald Trump tweets stand out and are accuratley estimated on the sentiment.<br>\n",
    "Wedding seems to be scored high (that is a positive happening)<br>\n",
    "Katey Perry's weet from the hospital is a positive tweet, but got a really bad sentiment score. <br>Barak Obamas second tweet looks happy, but it is in a negative environment (covid), and the model estimated on the negative side<br><br>\n",
    "Otherwise it is not that bad predictions."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## VG Del"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Prep för windows:\n",
    "https://github.com/soskek/bert-chainer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import math\n",
    "import datetime\n",
    "\n",
    "from tqdm import tqdm\n",
    "\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "\n",
    "import tensorflow as tf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'2.3.1'"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tf.__version__"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "os.environ[\"CUDA_VISIBLE_DEVICES\"] = \"1\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "if tf.__version__.startswith(\"1.\"):\n",
    "  tf.enable_eager_execution()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "from bert import BertModelLayer\n",
    "\n",
    "l_bert = BertModelLayer(**BertModelLayer.Params(\n",
    "  vocab_size               = 16000,        # embedding params\n",
    "  use_token_type           = True,\n",
    "  use_position_embeddings  = True,\n",
    "  token_type_vocab_size    = 2,\n",
    "\n",
    "  num_layers               = 12,           # transformer encoder params\n",
    "  hidden_size              = 768,\n",
    "  hidden_dropout           = 0.1,\n",
    "  intermediate_size        = 4*768,\n",
    "  intermediate_activation  = \"gelu\",\n",
    "\n",
    "  adapter_size             = None,         # see arXiv:1902.00751 (adapter-BERT)\n",
    "\n",
    "  shared_layer             = False,        # True for ALBERT (arXiv:1909.11942)\n",
    "  embedding_size           = None,         # None for BERT, wordpiece embedding size for ALBERT\n",
    "\n",
    "#  name                     = \"bert\"        # any other Keras layer params\n",
    "))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "import bert\n",
    "from bert import BertModelLayer\n",
    "from bert.loader import StockBertConfig, map_stock_config_to_params, load_stock_weights\n",
    "#from bert.tokenization.bert_tokenization import FullTokenizer\n",
    "from bert import bert_tokenization\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "from tensorflow import keras\n",
    "import os\n",
    "import re\n",
    "\n",
    "# Load all files from a directory in a DataFrame.\n",
    "def load_directory_data(directory):\n",
    "  data = {}\n",
    "  data[\"sentence\"] = []\n",
    "  data[\"sentiment\"] = []\n",
    "  for file_path in tqdm(os.listdir(directory), desc=os.path.basename(directory)):\n",
    "    with tf.io.gfile.GFile(os.path.join(directory, file_path), \"r\") as f:\n",
    "      data[\"sentence\"].append(f.read())\n",
    "      data[\"sentiment\"].append(re.match(\"\\d+_(\\d+)\\.txt\", file_path).group(1))\n",
    "  return pd.DataFrame.from_dict(data)\n",
    "\n",
    "# Merge positive and negative examples, add a polarity column and shuffle.\n",
    "def load_dataset(directory):\n",
    "  pos_df = load_directory_data(os.path.join(directory, \"pos\"))\n",
    "  neg_df = load_directory_data(os.path.join(directory, \"neg\"))\n",
    "  pos_df[\"polarity\"] = 1\n",
    "  neg_df[\"polarity\"] = 0\n",
    "  return pd.concat([pos_df, neg_df]).sample(frac=1).reset_index(drop=True)\n",
    "\n",
    "# Download and process the dataset files.\n",
    "def download_and_load_datasets(force_download=False):\n",
    "  dataset = tf.keras.utils.get_file(\n",
    "      fname=\"aclImdb.tar.gz\", \n",
    "      origin=\"http://ai.stanford.edu/~amaas/data/sentiment/aclImdb_v1.tar.gz\", \n",
    "      extract=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "\n",
    "\n",
    "Let's use the MovieReviewData class below, to prepare/encode the data for feeding into our BERT model, by:\n",
    "\n",
    "    tokenizing the text\n",
    "    trim or pad it to a max_seq_len length\n",
    "    append the special tokens [CLS] and [SEP]\n",
    "    convert the string tokens to numerical IDs using the original model's token encoding from vocab.txt\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "import bert\n",
    "from bert import BertModelLayer\n",
    "from bert.loader import StockBertConfig, map_stock_config_to_params, load_stock_weights\n",
    "#from bert.tokenization import FullTokenizer\n",
    "\n",
    "\n",
    "class MovieReviewData:\n",
    "    DATA_COLUMN = \"sentence\"\n",
    "    LABEL_COLUMN = \"polarity\"\n",
    "\n",
    "    def __init__(self, tokenizer: bert_tokenization.FullTokenizer, sample_size=None, max_seq_len=1024):\n",
    "    #def __init__(self, tokenizer: FullTokenizer, sample_size=None, max_seq_len=1024):\n",
    "        self.tokenizer = tokenizer\n",
    "        self.sample_size = sample_size\n",
    "        self.max_seq_len = 0\n",
    "        train, test = download_and_load_datasets()\n",
    "        \n",
    "        train, test = map(lambda df: df.reindex(df[MovieReviewData.DATA_COLUMN].str.len().sort_values().index), \n",
    "                          [train, test])\n",
    "                \n",
    "        if sample_size is not None:\n",
    "            assert sample_size % 128 == 0\n",
    "            train, test = train.head(sample_size), test.head(sample_size)\n",
    "            # train, test = map(lambda df: df.sample(sample_size), [train, test])\n",
    "        \n",
    "        ((self.train_x, self.train_y),\n",
    "         (self.test_x, self.test_y)) = map(self._prepare, [train, test])\n",
    "\n",
    "        print(\"max seq_len\", self.max_seq_len)\n",
    "        self.max_seq_len = min(self.max_seq_len, max_seq_len)\n",
    "        ((self.train_x, self.train_x_token_types),\n",
    "         (self.test_x, self.test_x_token_types)) = map(self._pad, \n",
    "                                                       [self.train_x, self.test_x])\n",
    "\n",
    "    def _prepare(self, df):\n",
    "        x, y = [], []\n",
    "        with tqdm(total=df.shape[0], unit_scale=True) as pbar:\n",
    "            for ndx, row in df.iterrows():\n",
    "                text, label = row[MovieReviewData.DATA_COLUMN], row[MovieReviewData.LABEL_COLUMN]\n",
    "                tokens = self.tokenizer.tokenize(text)\n",
    "                tokens = [\"[CLS]\"] + tokens + [\"[SEP]\"]\n",
    "                token_ids = self.tokenizer.convert_tokens_to_ids(tokens)\n",
    "                self.max_seq_len = max(self.max_seq_len, len(token_ids))\n",
    "                x.append(token_ids)\n",
    "                y.append(int(label))\n",
    "                pbar.update()\n",
    "        return np.array(x), np.array(y)\n",
    "\n",
    "    def _pad(self, ids):\n",
    "        x, t = [], []\n",
    "        token_type_ids = [0] * self.max_seq_len\n",
    "        for input_ids in ids:\n",
    "            input_ids = input_ids[:min(len(input_ids), self.max_seq_len - 2)]\n",
    "            input_ids = input_ids + [0] * (self.max_seq_len - len(input_ids))\n",
    "            x.append(np.array(input_ids))\n",
    "            t.append(token_type_ids)\n",
    "        return np.array(x), np.array(t)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "\n",
    "A tweak\n",
    "\n",
    "Because of a tf.train.load_checkpoint limitation requiring list permissions on the google storage bucket, we need to copy the pre-trained BERT weights locally.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "uncased_L-12_H-768_A-12/\n"
     ]
    }
   ],
   "source": [
    "bert_ckpt_dir = \"uncased_L-12_H-768_A-12/\"\n",
    "bert_ckpt_file = bert_ckpt_dir + \"bert_model.ckpt\"\n",
    "bert_config_file = bert_ckpt_dir + \"bert_config.json\"\n",
    "\n",
    "print(bert_ckpt_dir)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Wall time: 76 ms\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "#tokenizer = FullTokenizer(vocab_file=os.path.join(bert_ckpt_dir, \"vocab.txt\"))\n",
    "tokenizer = bert_tokenization.FullTokenizer(vocab_file=os.path.join(bert_ckpt_dir, \"vocab.txt\"))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "ename": "TypeError",
     "evalue": "cannot unpack non-iterable NoneType object",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mTypeError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-22-03971bca32f2>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m\u001b[0m\n\u001b[1;32m----> 1\u001b[1;33m \u001b[0mdata\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mMovieReviewData\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mtokenizer\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0msample_size\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;36m10\u001b[0m\u001b[1;33m*\u001b[0m\u001b[1;36m128\u001b[0m\u001b[1;33m*\u001b[0m\u001b[1;36m2\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mmax_seq_len\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;36m128\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[1;32m<ipython-input-19-856b305d42b5>\u001b[0m in \u001b[0;36m__init__\u001b[1;34m(self, tokenizer, sample_size, max_seq_len)\u001b[0m\n\u001b[0;32m     14\u001b[0m         \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0msample_size\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0msample_size\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     15\u001b[0m         \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mmax_seq_len\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;36m0\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 16\u001b[1;33m         \u001b[0mtrain\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mtest\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mdownload_and_load_datasets\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     17\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     18\u001b[0m         train, test = map(lambda df: df.reindex(df[MovieReviewData.DATA_COLUMN].str.len().sort_values().index), \n",
      "\u001b[1;31mTypeError\u001b[0m: cannot unpack non-iterable NoneType object"
     ]
    }
   ],
   "source": [
    "data = MovieReviewData(tokenizer, sample_size=10*128*2, max_seq_len=128)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "# pip install pytorch-pretrained-bert\n",
    "# pip install torchnlp"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import sys\n",
    "import numpy as np\n",
    "import random as rn\n",
    "import pandas as pd\n",
    "import torch\n",
    "from pytorch_pretrained_bert import BertModel\n",
    "from torch import nn\n",
    "from torchnlp.datasets import imdb_dataset\n",
    "# --> We are using our own uploaded dataset.\n",
    "from pytorch_pretrained_bert import BertTokenizer\n",
    "from keras.preprocessing.sequence import pad_sequences\n",
    "from torch.utils.data import TensorDataset, DataLoader, RandomSampler, SequentialSampler\n",
    "from torch.optim import Adam\n",
    "from torch.nn.utils import clip_grad_norm_\n",
    "from IPython.display import clear_output\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor([[0.2140, 0.2482, 0.2680],\n",
      "        [0.1146, 0.2093, 0.0161],\n",
      "        [0.0426, 0.0290, 0.1481],\n",
      "        [0.9391, 0.6205, 0.0977],\n",
      "        [0.2925, 0.7449, 0.4561]])\n"
     ]
    }
   ],
   "source": [
    "# Check if Torch is installed\n",
    "print(torch.rand(5, 3))\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "True"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Check if it works with CUDA\n",
    "torch.cuda.is_available()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "rn.seed(321)\n",
    "np.random.seed(321)\n",
    "torch.manual_seed(321)\n",
    "torch.cuda.manual_seed(321)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "info = os.stat(dir_path + f)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Exception ignored in: <function tqdm.__del__ at 0x00000000203F4B80>\n",
      "Traceback (most recent call last):\n",
      "  File \"c:\\python38\\lib\\site-packages\\tqdm\\std.py\", line 1128, in __del__\n",
      "    self.close()\n",
      "  File \"c:\\python38\\lib\\site-packages\\tqdm\\notebook.py\", line 274, in close\n",
      "    self.sp(bar_style='success')\n",
      "AttributeError: 'tqdm' object has no attribute 'sp'\n"
     ]
    }
   ],
   "source": [
    "import pickle\n",
    "from tensorflow.keras.utils import get_file\n",
    "\n",
    "path = get_file('imdb_full.pkl',\n",
    "               origin='https://s3.amazonaws.com/text-datasets/imdb_full.pkl',\n",
    "                md5_hash='d091312047c43cf9e4e38fef92437263')\n",
    "f = open(path, 'rb')\n",
    "(training_data, training_labels), (test_data, test_labels) = pickle.load(f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "training_labels[0]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[1mDownloading and preparing dataset imdb_reviews/plain_text/1.0.0 (download: Unknown size, generated: Unknown size, total: Unknown size) to C:\\Users\\sun\\tensorflow_datasets\\imdb_reviews\\plain_text\\1.0.0...\u001b[0m\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "5718d81393b143fb9808d1f43d957d53",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(HTML(value='Dl Completed...'), FloatProgress(value=1.0, bar_style='info', layout=Layout(width='…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "ab7731e6a83c4bfbb4d77eb63c002aa2",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(HTML(value='Dl Size...'), FloatProgress(value=1.0, bar_style='info', layout=Layout(width='20px'…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(HTML(value=''), FloatProgress(value=0.0, max=3.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(HTML(value=''), FloatProgress(value=1.0, bar_style='info', layout=Layout(width='20px'), max=1.0…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Shuffling and writing examples to C:\\/Users/sun/tensorflow_datasets/imdb_reviews/plain_text/1.0.0.incompleteRZIX92/imdb_reviews-train.tfrecord\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "f3b9f732185941d1a8ebf775c3cb7a4d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(HTML(value=''), FloatProgress(value=0.0, max=25000.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(HTML(value=''), FloatProgress(value=1.0, bar_style='info', layout=Layout(width='20px'), max=1.0…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Shuffling and writing examples to C:\\/Users/sun/tensorflow_datasets/imdb_reviews/plain_text/1.0.0.incompleteRZIX92/imdb_reviews-test.tfrecord\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "53446541b9784fd295e2d90b1c29ad31",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(HTML(value=''), FloatProgress(value=0.0, max=25000.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(HTML(value=''), FloatProgress(value=1.0, bar_style='info', layout=Layout(width='20px'), max=1.0…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Shuffling and writing examples to C:\\/Users/sun/tensorflow_datasets/imdb_reviews/plain_text/1.0.0.incompleteRZIX92/imdb_reviews-unsupervised.tfrecord\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "d740fb619efa4c51ac0ebd0cd03c36ee",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(HTML(value=''), FloatProgress(value=0.0, max=50000.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[1mDataset imdb_reviews downloaded and prepared to C:\\Users\\sun\\tensorflow_datasets\\imdb_reviews\\plain_text\\1.0.0. Subsequent calls will reuse this data.\u001b[0m\n",
      "{'label': <tf.Tensor: shape=(), dtype=int64, numpy=0>, 'text': <tf.Tensor: shape=(), dtype=string, numpy=b\"This was an absolutely terrible movie. Don't be lured in by Christopher Walken or Michael Ironside. Both are great actors, but this must simply be their worst role in history. Even their great acting could not redeem this movie's ridiculous storyline. This movie is an early nineties US propaganda piece. The most pathetic scenes were those when the Columbian rebels were making their cases for revolutions. Maria Conchita Alonso appeared phony, and her pseudo-love affair with Walken was nothing but a pathetic emotional plug in a movie that was devoid of any real meaning. I am disappointed that there are movies like this, ruining actor's like Christopher Walken's good name. I could barely sit through it.\">}\n",
      "{'label': <tf.Tensor: shape=(), dtype=int64, numpy=0>, 'text': <tf.Tensor: shape=(), dtype=string, numpy=b'I have been known to fall asleep during films, but this is usually due to a combination of things including, really tired, being warm and comfortable on the sette and having just eaten a lot. However on this occasion I fell asleep because the film was rubbish. The plot development was constant. Constantly slow and boring. Things seemed to happen, but with no explanation of what was causing them or why. I admit, I may have missed part of the film, but i watched the majority of it and everything just seemed to happen of its own accord without any real concern for anything else. I cant recommend this film at all.'>}\n",
      "{'label': <tf.Tensor: shape=(), dtype=int64, numpy=0>, 'text': <tf.Tensor: shape=(), dtype=string, numpy=b'Mann photographs the Alberta Rocky Mountains in a superb fashion, and Jimmy Stewart and Walter Brennan give enjoyable performances as they always seem to do. <br /><br />But come on Hollywood - a Mountie telling the people of Dawson City, Yukon to elect themselves a marshal (yes a marshal!) and to enforce the law themselves, then gunfighters battling it out on the streets for control of the town? <br /><br />Nothing even remotely resembling that happened on the Canadian side of the border during the Klondike gold rush. Mr. Mann and company appear to have mistaken Dawson City for Deadwood, the Canadian North for the American Wild West.<br /><br />Canadian viewers be prepared for a Reefer Madness type of enjoyable howl with this ludicrous plot, or, to shake your head in disgust.'>}\n",
      "{'label': <tf.Tensor: shape=(), dtype=int64, numpy=1>, 'text': <tf.Tensor: shape=(), dtype=string, numpy=b'This is the kind of film for a snowy Sunday afternoon when the rest of the world can go ahead with its own business as you descend into a big arm-chair and mellow for a couple of hours. Wonderful performances from Cher and Nicolas Cage (as always) gently row the plot along. There are no rapids to cross, no dangerous waters, just a warm and witty paddle through New York life at its best. A family film in every sense and one that deserves the praise it received.'>}\n"
     ]
    }
   ],
   "source": [
    "\n",
    "\n",
    "import tensorflow_datasets as tfds\n",
    "ds = tfds.load('imdb_reviews', split='train')\n",
    "\n",
    "for ex in ds.take(4):\n",
    "  print(ex)\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loading data...\n",
      "25000 train sequences\n",
      "25000 test sequences\n",
      "Pad sequences (samples x time)\n",
      "input_train shape: (25000, 500)\n",
      "input_test shape: (25000, 500)\n",
      "1\n",
      "[   0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    1   14   22   16   43  530  973 1622 1385   65  458 4468\n",
      "   66 3941    4  173   36  256    5   25  100   43  838  112   50  670\n",
      "    2    9   35  480  284    5  150    4  172  112  167    2  336  385\n",
      "   39    4  172 4536 1111   17  546   38   13  447    4  192   50   16\n",
      "    6  147 2025   19   14   22    4 1920 4613  469    4   22   71   87\n",
      "   12   16   43  530   38   76   15   13 1247    4   22   17  515   17\n",
      "   12   16  626   18    2    5   62  386   12    8  316    8  106    5\n",
      "    4 2223 5244   16  480   66 3785   33    4  130   12   16   38  619\n",
      "    5   25  124   51   36  135   48   25 1415   33    6   22   12  215\n",
      "   28   77   52    5   14  407   16   82    2    8    4  107  117 5952\n",
      "   15  256    4    2    7 3766    5  723   36   71   43  530  476   26\n",
      "  400  317   46    7    4    2 1029   13  104   88    4  381   15  297\n",
      "   98   32 2071   56   26  141    6  194 7486   18    4  226   22   21\n",
      "  134  476   26  480    5  144   30 5535   18   51   36   28  224   92\n",
      "   25  104    4  226   65   16   38 1334   88   12   16  283    5   16\n",
      " 4472  113  103   32   15   16 5345   19  178   32]\n",
      "0\n",
      "[   0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
      "    0    0    0    0    0    0    1   13  244    6   87  337    7  628\n",
      " 2219    5   28  285   15  240   93   23  288  549   18 1455  673    4\n",
      "  241  534 3635 8448   20   38   54   13  258   46   44   14   13 1241\n",
      " 7258   12    5    5   51    9   14   45    6  762    7    2 1309  328\n",
      "    5  428 2473   15   26 1292    5 3939 6728    5 1960  279   13   92\n",
      "  124  803   52   21  279   14    9   43    6  762    7  595   15   16\n",
      "    2   23    4 1071  467    4  403    7  628 2219    8   97    6  171\n",
      " 3596   99  387   72   97   12  788   15   13  161  459   44    4 3939\n",
      " 1101  173   21   69    8  401    2    4  481   88   61 4731  238   28\n",
      "   32   11   32   14    9    6  545 1332  766    5  203   73   28   43\n",
      "   77  317   11    4    2  953  270   17    6 3616   13  545  386   25\n",
      "   92 1142  129  278   23   14  241   46    7  158]\n"
     ]
    }
   ],
   "source": [
    "from keras.datasets import imdb\n",
    "from keras.preprocessing import sequence\n",
    "\n",
    "max_features = 10000  # number of words to consider as features\n",
    "maxlen = 500  # cut texts after this number of words (among top max_features most common words)\n",
    "batch_size = 32\n",
    "\n",
    "print('Loading data...')\n",
    "(input_train, y_train), (input_test, y_test) = imdb.load_data(num_words=max_features)\n",
    "print(len(input_train), 'train sequences')\n",
    "print(len(input_test), 'test sequences')\n",
    "\n",
    "print('Pad sequences (samples x time)')\n",
    "input_train = sequence.pad_sequences(input_train, maxlen=maxlen)\n",
    "input_test = sequence.pad_sequences(input_test, maxlen=maxlen)\n",
    "print('input_train shape:', input_train.shape)\n",
    "print('input_test shape:', input_test.shape)\n",
    "\n",
    "print(y_train[0])\n",
    "print(input_train[0])\n",
    "\n",
    "print(y_train[100])\n",
    "print(input_train[100])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'train_texts' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-9-24ed9450ed31>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m\u001b[0m\n\u001b[1;32m----> 1\u001b[1;33m \u001b[0mlen\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mtrain_texts\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mlen\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mtrain_labels\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mlen\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mtest_texts\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mlen\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mtest_labels\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[1;31mNameError\u001b[0m: name 'train_texts' is not defined"
     ]
    }
   ],
   "source": [
    "len(train_texts), len(train_labels), len(test_texts), len(test_labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}