48-text-generation-on-shahnameh-tensorflow.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<img src=\"http://akhavanpour.ir/notebook/images/srttu.gif\" alt=\"SRTTU\" style=\"width: 150px;\"/>\n",
    "\n",
    "[![Run in Google Colab](./lecture_images/colab.png)](https://colab.research.google.com/github/alireza-akhavan/class.vision/blob/master/48-text-generation-on-shahnameh-tensorflow.ipynb)\n",
    "[![Azure Notebooks](https://notebooks.azure.com/launch.png)](https://notebooks.azure.com/import/gh/Alireza-Akhavan/class.vision)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# <div style=\"direction:rtl;text-align:right;font-family:B Lotus, B Nazanin, Tahoma\"> تولید متن با استفاده از شبکه های بازگشتی در سطح کاراکتر  روی شاهنامه فروسی</div>\n",
    "\n",
    "<div style=\"direction:rtl;text-align:right;font-family:Tahoma\">\n",
    "کدها با تغییرات برگرفته از مستندات سایت تنسرفلو است.</div>\n",
    "\n",
    "[https://www.tensorflow.org/tutorials/sequences/text_generation](https://www.tensorflow.org/tutorials/sequences/text_generation)\n",
    "\n",
    "\n",
    "\n",
    "<div style=\"direction:rtl;text-align:right;font-family:Tahoma\">\n",
    "توصیه می‌شود قبل از این نوت‌بوک حتما نوت بوک\n",
    "</div>\n",
    "\n",
    "[42-text-generation-with-lstm.ipynb](http://nbviewer.jupyter.org/github/alireza-akhavan/class.vision/blob/master/42-text-generation-with-lstm.ipynb)\n",
    "\n",
    "\n",
    "<div style=\"direction:rtl;text-align:right;font-family:Tahoma\">\n",
    "را مرور کنید.\n",
    "<br>\n",
    "در این نوت بوک:\n",
    "\n",
    "    یک مدل character level بازگشتی خواهیم داشت\n",
    "    در این نوت بوک از لایه embedding برای کاراکترها و نه کلمات استفاده شده است\n",
    "    در این مثال از api تنسورفلو برای ورودی داده به نام tf.dataset استفاده شده\n",
    "    در این نوت بوک eager execution تنسورفلو فعال \n",
    "    این نوت بوک روی داده شاهنامه آموزش دیده است\n",
    "    از نسخه خاص GPU به نام tf.keras.layers.CuDNNGRU استفاده میکنیم.\n",
    "</div>\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "BwpJ5IffzRG6"
   },
   "source": [
    " <div style=\"direction:rtl;text-align:right;font-family:B Lotus, B Nazanin, Tahoma\">اگر در گوگل کولب اجرا میکنید حتما GPU را فعال کنید.</div>\n",
    "\n",
    "Note: Enable GPU acceleration to execute this notebook faster. In Colab: *Runtime > Change runtime type > Hardware acclerator > GPU*. If running locally make sure TensorFlow version >= 1.11.\n",
    "\n",
    "This tutorial includes runnable code implemented using [tf.keras](https://www.tensorflow.org/programmers_guide/keras) and [eager execution](https://www.tensorflow.org/programmers_guide/eager). \n",
    "\n",
    "While some of the sentences are grammatical, most do not make sense. The model has not learned the meaning of words, but consider:\n",
    "\n",
    "* The model is character-based. When training started, the model did not know how to spell an Persian word, or that words were even a unit of text.\n",
    "\n",
    "* As demonstrated below, the model is trained on small batches of text (100 characters each), and is still able to generate a longer sequence of text with coherent structure."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "srXC6pLGLwS6"
   },
   "source": [
    "## Setup"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "WGyKZj3bzf9p"
   },
   "source": [
    "### Import TensorFlow and other libraries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "yG_n40gFzf9s"
   },
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "tf.enable_eager_execution()\n",
    "\n",
    "import numpy as np\n",
    "import os\n",
    "import time"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "EHDoRoc5PKWz"
   },
   "source": [
    "### Download the Shahnameh dataset\n",
    "\n",
    "Change the following line to run this code on your own data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "pD_55cOxLkAb"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Downloading data from http://dataset.class.vision/NLP/shahnameh.txt\n",
      "4653056/4652876 [==============================] - 5s 1us/step\n"
     ]
    }
   ],
   "source": [
    "path_to_file = tf.keras.utils.get_file('shahnameh.txt', 'http://dataset.class.vision/NLP/shahnameh.txt')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "UHjdCjDuSvX_"
   },
   "source": [
    "### Read the data\n",
    "\n",
    "First, look in the text."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "aavnuByVymwK"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Length of text: 2653849 characters\n"
     ]
    }
   ],
   "source": [
    "text = open(path_to_file , encoding=\"utf8\").read()\n",
    "# length of text is the number of characters in it\n",
    "print ('Length of text: {} characters'.format(len(text)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "Duhg9NrUymwO"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "|به نام خداوند جان و خرد\n",
      "|کزین برتر اندیشه برنگذرد\n",
      "|خداوند نام و خداوند جای\n",
      "|خداوند روزی ده رهنمای\n",
      "|خداوند کیوان و گردان سپهر\n",
      "|فروزنده ماه و ناهید و مهر\n",
      "|ز نام و نشان و گمان برترست\n",
      "|نگارندهٔ بر شده پیکرست\n",
      "|به بینندگان آفریننده را\n",
      "|نبینی مرنجان دو بین\n"
     ]
    }
   ],
   "source": [
    "# Take a look at the first 250 characters in text\n",
    "print(text[:250])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "IlCgQBRVymwR"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "48 unique characters\n"
     ]
    }
   ],
   "source": [
    "# The unique characters in the file\n",
    "vocab = sorted(set(text))\n",
    "print ('{} unique characters'.format(len(vocab)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "rNnrKn_lL-IJ"
   },
   "source": [
    "## Process the text"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "LFjSVAlWzf-N"
   },
   "source": [
    "### Vectorize the text\n",
    "\n",
    "Before training, we need to map strings to a numerical representation. Create two lookup tables: one mapping characters to numbers, and another for numbers to characters."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "IalZLbvOzf-F"
   },
   "outputs": [],
   "source": [
    "# Creating a mapping from unique characters to indices\n",
    "char2idx = {u:i for i, u in enumerate(vocab)}\n",
    "idx2char = np.array(vocab)\n",
    "\n",
    "text_as_int = np.array([char2idx[c] for c in text])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "tZfqhkYCymwX"
   },
   "source": [
    "Now we have an integer representation for each character. Notice that we mapped the character as indexes from 0 to `len(unique)`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "10\n",
      "0\n",
      "1\n",
      "17\n"
     ]
    }
   ],
   "source": [
    "print(char2idx['آ'])\n",
    "print(char2idx['\\n'])\n",
    "print(char2idx[' '])\n",
    "print(char2idx['ث'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "FYyNlCNXymwY"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{\n",
      "  '\\n':   0,\n",
      "  ' ' :   1,\n",
      "  '(' :   2,\n",
      "  ')' :   3,\n",
      "  '|' :   4,\n",
      "  '«' :   5,\n",
      "  '»' :   6,\n",
      "  '،' :   7,\n",
      "  '؟' :   8,\n",
      "  'ء' :   9,\n",
      "  'آ' :  10,\n",
      "  'أ' :  11,\n",
      "  'ؤ' :  12,\n",
      "  'ئ' :  13,\n",
      "  'ا' :  14,\n",
      "  'ب' :  15,\n",
      "  'ت' :  16,\n",
      "  'ث' :  17,\n",
      "  'ج' :  18,\n",
      "  'ح' :  19,\n",
      "  ...\n",
      "}\n"
     ]
    }
   ],
   "source": [
    "print('{')\n",
    "for char,_ in zip(char2idx, range(20)):\n",
    "    print('  {:4s}: {:3d},'.format(repr(char), char2idx[char]))\n",
    "print('  ...\\n}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "l1VKcQHcymwb"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "'|به نام خداون' ---- characters mapped to int ---- > [ 4 15 38  1 37 14 36  1 20 21 14 39 37]\n"
     ]
    }
   ],
   "source": [
    "# Show how the first 13 characters from the text are mapped to integers\n",
    "print ('{} ---- characters mapped to int ---- > {}'.format(repr(text[:13]), text_as_int[:13]))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "bbmsf23Bymwe"
   },
   "source": [
    "### The prediction task"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "wssHQ1oGymwe"
   },
   "source": [
    "Given a character, or a sequence of characters, what is the most probable next character? This is the task we're training the model to perform. The input to the model will be a sequence of characters, and we train the model to predict the output—the following character at each time step.\n",
    "\n",
    "Since RNNs maintain an internal state that depends on the previously seen elements, given all the characters computed until this moment, what is the next character?\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "hgsVvVxnymwf"
   },
   "source": [
    "### Create training examples and targets\n",
    "\n",
    "Next divide the text into example sequences. Each input sequence will contain `seq_length` characters from the text. \n",
    "\n",
    "For each input sequence, the corresponding targets contain the same length of text, except shifted one character to the right. \n",
    "\n",
    "So break the text into chunks of `seq_length+1`. For example, say `seq_length` is 4 and our text is \"Hello\". The input sequence would be \"Hell\", and the target sequence \"ello\".\n",
    "\n",
    "To do this first use the `tf.data.Dataset.from_tensor_slices` function to convert the text vector into a stream of character indices."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "0UHJDA39zf-O"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "|\n",
      "ب\n",
      "ه\n",
      " \n",
      "ن\n",
      "ا\n",
      "م\n",
      " \n",
      "خ\n",
      "د\n",
      "ا\n",
      "و\n",
      "ن\n",
      "د\n",
      " \n"
     ]
    }
   ],
   "source": [
    "# The maximum length sentence we want for a single input in characters\n",
    "seq_length = 100\n",
    "examples_per_epoch = len(text)//seq_length\n",
    "\n",
    "# Create training examples / targets\n",
    "char_dataset = tf.data.Dataset.from_tensor_slices(text_as_int)\n",
    "\n",
    "for c in char_dataset.take(15):\n",
    "    print(idx2char[c.numpy()])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "-ZSYAcQV8OGP"
   },
   "source": [
    "The `batch` method lets us easily convert these individual characters to sequences of the desired size. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "l4hkDU3i7ozi"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "'|به نام خداوند جان و خرد\\n|کزین برتر اندیشه برنگذرد\\n|خداوند نام و خداوند جای\\n|خداوند روزی ده رهنمای\\n|خ'\n",
      "\n",
      "\n",
      "'داوند کیوان و گردان سپهر\\n|فروزنده ماه و ناهید و مهر\\n|ز نام و نشان و گمان برترست\\n|نگارندهٔ بر شده پیکر'\n",
      "\n",
      "\n",
      "'ست\\n|به بینندگان آفریننده را\\n|نبینی مرنجان دو بیننده را\\n|نیابد بدو نیز اندیشه راه\\n|که او برتر از نام و'\n",
      "\n",
      "\n",
      "' از جایگاه\\n|سخن هر چه زین گوهران بگذرد\\n|نیابد بدو راه جان و خرد\\n|خرد گر سخن برگزیند همی\\n|همان را گزین'\n",
      "\n",
      "\n",
      "'د که بیند همی\\n|ستودن نداند کس او را چو هست\\n|میان بندگی را ببایدت بست\\n|خرد را و جان را همی سنجد اوی\\n|د'\n",
      "\n",
      "\n"
     ]
    }
   ],
   "source": [
    "sequences = char_dataset.batch(seq_length+1, drop_remainder=True)\n",
    "\n",
    "for item in sequences.take(5):\n",
    "    print(repr(''.join(idx2char[item.numpy()])))\n",
    "    print(\"\\n\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "UbLcIPBj_mWZ"
   },
   "source": [
    "For each sequence, duplicate and shift it to form the input and target text by using the `map` method to apply a simple function to each batch:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "9NGu-FkO_kYU"
   },
   "outputs": [],
   "source": [
    "def split_input_target(chunk):\n",
    "    input_text = chunk[:-1]\n",
    "    target_text = chunk[1:]\n",
    "    return input_text, target_text\n",
    "\n",
    "dataset = sequences.map(split_input_target)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "hiCopyGZymwi"
   },
   "source": [
    "Print the first examples input and target values:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "GNbw-iR0ymwj"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Input data: \n",
      "'|به نام خداوند جان و خرد\\n|کزین برتر اندیشه برنگذرد\\n|خداوند نام و خداوند جای\\n|خداوند روزی ده رهنمای\\n|'\n",
      "Target data: \n",
      "'به نام خداوند جان و خرد\\n|کزین برتر اندیشه برنگذرد\\n|خداوند نام و خداوند جای\\n|خداوند روزی ده رهنمای\\n|خ'\n"
     ]
    }
   ],
   "source": [
    "for input_example, target_example in  dataset.take(1):\n",
    "    print ('Input data: ')\n",
    "    print(repr(''.join(idx2char[input_example.numpy()])))\n",
    "    print ('Target data: ')\n",
    "    print(repr(''.join(idx2char[target_example.numpy()])))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "_33OHL3b84i0"
   },
   "source": [
    "<div style=\"direction:rtl;text-align:right;font-family:B Lotus, B Nazanin, Tahoma\">\n",
    "input: <br>\n",
    "'|به نام خداوند جان و خرد<br>|کزین برتر اندیشه برنگذرد<br>|خداوند نام و خداوند جای<br>|خداوند روزی ده رهنمای<br>|'\n",
    "<br>output:<br>\n",
    "'به نام خداوند جان و خرد<br>|کزین برتر اندیشه برنگذرد<br>|خداوند نام و خداوند جای<br>|خداوند روزی ده رهنمای<br>|خ'\n",
    "<br><br>\n",
    "هر اندیسی از این وکتور به عنوان یک time step پردازش میشود.\n",
    "     برای ورودی در زمان 0، مدل ورودی اندیس \"|\" را دریافت کرده و سعی میکند اندیس \"ب\" را به عنوان کاراکتر بعدی پیشگویی کند.\n",
    "    در time step بعدی کاری مشابه انجام می‌دهد اما RNN علاوه بر ورودی در این لحظه زمانی contex مرحله قبل را هم دارد.\n",
    "\n",
    "</div>\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "0eBu9WZG84i0"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Step    0\n",
      "  input: 4 ('|')\n",
      "  expected output: 15 ('ب')\n",
      "Step    1\n",
      "  input: 15 ('ب')\n",
      "  expected output: 38 ('ه')\n",
      "Step    2\n",
      "  input: 38 ('ه')\n",
      "  expected output: 1 (' ')\n",
      "Step    3\n",
      "  input: 1 (' ')\n",
      "  expected output: 37 ('ن')\n",
      "Step    4\n",
      "  input: 37 ('ن')\n",
      "  expected output: 14 ('ا')\n"
     ]
    }
   ],
   "source": [
    "for i, (input_idx, target_idx) in enumerate(zip(input_example[:5], target_example[:5])):\n",
    "    print(\"Step {:4d}\".format(i))\n",
    "    print(\"  input: {} ({:s})\".format(input_idx, repr(idx2char[input_idx])))\n",
    "    print(\"  expected output: {} ({:s})\".format(target_idx, repr(idx2char[target_idx])))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "MJdfPmdqzf-R"
   },
   "source": [
    "### Create training batches\n",
    "\n",
    "We used `tf.data` to split the text into manageable sequences. But before feeding this data into the model, we need to shuffle the data and pack it into batches."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "p2pGotuNzf-S"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<BatchDataset shapes: ((64, 100), (64, 100)), types: (tf.int32, tf.int32)>"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Batch size \n",
    "BATCH_SIZE = 64\n",
    "steps_per_epoch = examples_per_epoch//BATCH_SIZE\n",
    "\n",
    "# Buffer size to shuffle the dataset\n",
    "# (TF data is designed to work with possibly infinite sequences, \n",
    "# so it doesn't attempt to shuffle the entire sequence in memory. Instead, \n",
    "# it maintains a buffer in which it shuffles elements).\n",
    "BUFFER_SIZE = 10000\n",
    "\n",
    "dataset = dataset.shuffle(BUFFER_SIZE).batch(BATCH_SIZE, drop_remainder=True)\n",
    "\n",
    "dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "r6oUuElIMgVx"
   },
   "source": [
    "## Build The Model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "m8gPwEjRzf-Z"
   },
   "source": [
    "Use `tf.keras.Sequential` to define the model. For this simple example three layers are used to define our model:\n",
    "\n",
    "* `tf.keras.layers.Embedding`: The input layer. A trainable lookup table that will map the numbers of each character to a vector with `embedding_dim` dimensions;\n",
    "* `tf.keras.layers.GRU`: A type of RNN with size `units=rnn_units` (You can also use a LSTM layer here.)\n",
    "* `tf.keras.layers.Dense`: The output layer, with `vocab_size` outputs."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "zHT8cLh7EAsg"
   },
   "outputs": [],
   "source": [
    "# Length of the vocabulary in chars\n",
    "vocab_size = len(vocab)\n",
    "\n",
    "# The embedding dimension \n",
    "embedding_dim = 256\n",
    "\n",
    "# Number of RNN units\n",
    "rnn_units = 1024"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "NNVB-jmMEOzP"
   },
   "source": [
    "Next define a function to build the model.\n",
    "\n",
    "Use `CuDNNGRU` if running on GPU.  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "wjKZrC39ELy0"
   },
   "outputs": [],
   "source": [
    "if tf.test.is_gpu_available():\n",
    "    rnn = tf.keras.layers.CuDNNGRU\n",
    "else:\n",
    "    import functools\n",
    "    rnn = functools.partial(\n",
    "        tf.keras.layers.GRU, recurrent_activation='sigmoid')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "https://www.tensorflow.org/api_docs/python/tf/keras/layers/GRU\n",
    "\n",
    "https://keras.io/layers/recurrent/#cudnngru"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "MtCrdfzEI2N0"
   },
   "outputs": [],
   "source": [
    "def build_model(vocab_size, embedding_dim, rnn_units, batch_size):\n",
    "    model = tf.keras.Sequential([\n",
    "    tf.keras.layers.Embedding(vocab_size, embedding_dim, \n",
    "                              batch_input_shape=[batch_size, None]),\n",
    "    rnn(rnn_units,\n",
    "        return_sequences=True, \n",
    "        recurrent_initializer='glorot_uniform',\n",
    "        stateful=True),\n",
    "    tf.keras.layers.Dense(vocab_size)\n",
    "    ])\n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "wwsrpOik5zhv"
   },
   "outputs": [],
   "source": [
    "model = build_model(\n",
    "  vocab_size = len(vocab), \n",
    "  embedding_dim=embedding_dim, \n",
    "  rnn_units=rnn_units, \n",
    "  batch_size=BATCH_SIZE)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "RkA5upJIJ7W7"
   },
   "source": [
    "For each character the model looks up the embedding, runs the GRU one timestep with the embedding as input, and applies the dense layer to generate logits predicting the log-liklihood of the next character:\n",
    "\n",
    "![A drawing of the data passing through the model](./lecture_images/text_generation_training.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "-ubPo0_9Prjb"
   },
   "source": [
    "## Try the model\n",
    "\n",
    "Now run the model to see that it behaves as expected.\n",
    "\n",
    "First check the shape of the output:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "C-_70kKAPrPU"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(64, 100, 48) # (batch_size, sequence_length, vocab_size)\n"
     ]
    }
   ],
   "source": [
    "for input_example_batch, target_example_batch in dataset.take(1): \n",
    "    example_batch_predictions = model(input_example_batch)\n",
    "    print(example_batch_predictions.shape, \"# (batch_size, sequence_length, vocab_size)\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "Q6NzLBi4VM4o"
   },
   "source": [
    "In the above example the sequence length of the input is `100` but the model can be run on inputs of any length: "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "vPGmAAXmVLGC"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding (Embedding)        (64, None, 256)           12288     \n",
      "_________________________________________________________________\n",
      "cu_dnngru (CuDNNGRU)         (64, None, 1024)          3938304   \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (64, None, 48)            49200     \n",
      "=================================================================\n",
      "Total params: 3,999,792\n",
      "Trainable params: 3,999,792\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "uwv0gEkURfx1"
   },
   "source": [
    "To get actual predictions from the model we need to sample from the output distribution, to get actual character indices. This distribution is defined by the logits over the character vocabulary. \n",
    "\n",
    "Note: It is important to _sample_ from this distribution as taking the _argmax_ of the distribution can easily get the model stuck in a loop.\n",
    "\n",
    "Try it for the first example in the batch:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "4V4MfFg0RQJg"
   },
   "outputs": [],
   "source": [
    "sampled_indices = tf.multinomial(example_batch_predictions[0], num_samples=1)\n",
    "sampled_indices = tf.squeeze(sampled_indices,axis=-1).numpy()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "QM1Vbxs_URw5"
   },
   "source": [
    "This gives us, at each timestep, a prediction of the next character index:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "YqFMUQc_UFgM"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ 7,  6, 18, 39, 33, 45, 22, 19, 18, 41, 40, 29, 33, 19, 26,  7, 30,\n",
       "       10,  3, 21, 36,  5, 32, 34,  7, 46, 23, 33,  0, 36,  7,  8, 39, 12,\n",
       "       30, 30, 32, 47, 35, 35, 35, 38, 31,  3, 39, 15, 25,  8, 38, 47,  1,\n",
       "        8, 39, 28, 19, 28, 14,  7, 10, 40, 26, 18,  4, 35, 46, 10, 46,  9,\n",
       "       41,  7, 24,  3, 22, 44, 45, 41, 33, 16, 16, 29, 30, 35, 36, 44, 13,\n",
       "       31, 47, 10, 30, 18,  8,  0, 23, 33, 27, 47, 19, 40, 26, 10],\n",
       "      dtype=int64)"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sampled_indices"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "LfLtsP3mUhCG"
   },
   "source": [
    "Decode these to see the text predicted by this untrained model:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "xWcFwPwLSo05"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Input: \n",
      " ' ناز و ز تو بپوشد سخن\\n|تهمتن بران گشت همداستان\\n|که فرخنده موبد زد این داستان\\n|چنین گفت خرم دل رهنمای'\n",
      "\n",
      "Next Char Predictions: \n",
      " '،»جوفگذحجپٔطفحش،ظآ)دم«غق،یرف\\nم،؟وؤظظغ\\u200cلللهع)وبس؟ه\\u200c ؟وضحضا،آٔشج|لیآیءپ،ز)ذکگپفتتطظلمکئع\\u200cآظج؟\\nرفص\\u200cحٔشآ'\n"
     ]
    }
   ],
   "source": [
    "print(\"Input: \\n\", repr(\"\".join(idx2char[input_example_batch[0]])))\n",
    "print()\n",
    "print(\"Next Char Predictions: \\n\", repr(\"\".join(idx2char[sampled_indices ])))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "LJL0Q0YPY6Ee"
   },
   "source": [
    "## Train the model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "YCbHQHiaa4Ic"
   },
   "source": [
    "At this point the problem can be treated as a standard classification problem. Given the previous RNN state, and the input this time step, predict the class of the next character."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "trpqTWyvk0nr"
   },
   "source": [
    "### Attach an optimizer, and a loss function"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "UAjbjY03eiQ4"
   },
   "source": [
    "The standard `tf.losses.sparse_softmax_cross_entropy` loss function works in this case because it is applied across the last dimension of the predictions. \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "4HrXTACTdzY-"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Prediction shape:  (64, 100, 48)  # (batch_size, sequence_length, vocab_size)\n",
      "scalar_loss:       3.8720946\n"
     ]
    }
   ],
   "source": [
    "example_batch_loss  = tf.losses.sparse_softmax_cross_entropy(target_example_batch, example_batch_predictions)\n",
    "print(\"Prediction shape: \", example_batch_predictions.shape, \" # (batch_size, sequence_length, vocab_size)\") \n",
    "print(\"scalar_loss:      \", example_batch_loss.numpy())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "jeOXriLcymww"
   },
   "source": [
    "Configure the training procedure using the `tf.keras.Model.compile` method. We'll use `tf.train.AdamOptimizer` with default arguments and `tf.losses.sparse_softmax_cross_entropy` as the loss function. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "DDl1_Een6rL0"
   },
   "outputs": [],
   "source": [
    "model.compile(\n",
    "    optimizer = tf.train.AdamOptimizer(),\n",
    "    loss = tf.losses.sparse_softmax_cross_entropy)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "ieSJdchZggUj"
   },
   "source": [
    "### Configure checkpoints"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "C6XBUUavgF56"
   },
   "source": [
    "Use a `tf.keras.callbacks.ModelCheckpoint` to ensure that checkpoints are saved during training:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'D:\\\\my cources\\\\deeplearning-part2\\\\notebook\\\\training_checkpoints'"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Directory where the checkpoints will be saved\n",
    "checkpoint_dir = os.path.join(os.getcwd(), 'training_checkpoints')\n",
    "checkpoint_dir"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "W6fWTriUZP-n"
   },
   "outputs": [],
   "source": [
    "\n",
    "# Name of the checkpoint files\n",
    "checkpoint_prefix = os.path.join(checkpoint_dir, \"ckpt_{epoch}\")\n",
    "\n",
    "checkpoint_callback=tf.keras.callbacks.ModelCheckpoint(\n",
    "    filepath=checkpoint_prefix,\n",
    "    save_weights_only=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "3Ky3F_BhgkTW"
   },
   "source": [
    "### Execute the training"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "IxdOA-rgyGvs"
   },
   "source": [
    "To keep this quick, train the model for just 3 epochs:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "7yGBE2zxMMHs"
   },
   "outputs": [],
   "source": [
    "EPOCHS=30"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "UK-hmKjYVoll"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/30\n",
      "414/414 [==============================] - 107s 259ms/step - loss: 2.1298\n",
      "Epoch 2/30\n",
      "414/414 [==============================] - 109s 264ms/step - loss: 1.4952\n",
      "Epoch 3/30\n",
      "414/414 [==============================] - 110s 266ms/step - loss: 1.3507\n",
      "Epoch 4/30\n",
      "414/414 [==============================] - 110s 266ms/step - loss: 1.2831\n",
      "Epoch 5/30\n",
      "414/414 [==============================] - 111s 267ms/step - loss: 1.2383\n",
      "Epoch 6/30\n",
      "414/414 [==============================] - 110s 267ms/step - loss: 1.2029\n",
      "Epoch 7/30\n",
      "414/414 [==============================] - 110s 267ms/step - loss: 1.1726\n",
      "Epoch 8/30\n",
      "414/414 [==============================] - 111s 267ms/step - loss: 1.1443\n",
      "Epoch 9/30\n",
      "414/414 [==============================] - 111s 269ms/step - loss: 1.1201\n",
      "Epoch 10/30\n",
      "414/414 [==============================] - 111s 267ms/step - loss: 1.0959\n",
      "Epoch 11/30\n",
      "414/414 [==============================] - 111s 268ms/step - loss: 1.0741\n",
      "Epoch 12/30\n",
      "414/414 [==============================] - 111s 269ms/step - loss: 1.0536\n",
      "Epoch 13/30\n",
      "414/414 [==============================] - 111s 267ms/step - loss: 1.0350\n",
      "Epoch 14/30\n",
      "414/414 [==============================] - 111s 268ms/step - loss: 1.0198\n",
      "Epoch 15/30\n",
      "414/414 [==============================] - 111s 267ms/step - loss: 1.0040\n",
      "Epoch 16/30\n",
      "414/414 [==============================] - 111s 267ms/step - loss: 0.9912\n",
      "Epoch 17/30\n",
      "414/414 [==============================] - 111s 267ms/step - loss: 0.9803\n",
      "Epoch 18/30\n",
      "414/414 [==============================] - 111s 268ms/step - loss: 0.9700\n",
      "Epoch 19/30\n",
      "414/414 [==============================] - 112s 272ms/step - loss: 0.9613\n",
      "Epoch 20/30\n",
      "414/414 [==============================] - 113s 272ms/step - loss: 0.9551\n",
      "Epoch 21/30\n",
      "414/414 [==============================] - 114s 275ms/step - loss: 0.9501\n",
      "Epoch 22/30\n",
      "414/414 [==============================] - 112s 270ms/step - loss: 0.9463\n",
      "Epoch 23/30\n",
      "414/414 [==============================] - 111s 269ms/step - loss: 0.9441\n",
      "Epoch 24/30\n",
      "414/414 [==============================] - 111s 269ms/step - loss: 0.9424\n",
      "Epoch 25/30\n",
      "414/414 [==============================] - 112s 270ms/step - loss: 0.9418\n",
      "Epoch 26/30\n",
      "414/414 [==============================] - 111s 269ms/step - loss: 0.9433\n",
      "Epoch 27/30\n",
      "414/414 [==============================] - 111s 269ms/step - loss: 0.9424\n",
      "Epoch 28/30\n",
      "414/414 [==============================] - 112s 270ms/step - loss: 0.9445\n",
      "Epoch 29/30\n",
      "414/414 [==============================] - 114s 276ms/step - loss: 0.9482\n",
      "Epoch 30/30\n",
      "414/414 [==============================] - 113s 272ms/step - loss: 0.9503\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(dataset.repeat(), epochs=EPOCHS, steps_per_epoch=steps_per_epoch, callbacks=[checkpoint_callback])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "kKkD5M6eoSiN"
   },
   "source": [
    "## Generate text"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "JIPcXllKjkdr"
   },
   "source": [
    "### Restore the latest checkpoint"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "LyeYRiuVjodY"
   },
   "source": [
    "To keep this prediction step simple, use a batch size of 1.\n",
    "\n",
    "Because of the way the RNN state is passed from timestep to timestep, the model only accepts a fixed batch size once built. \n",
    "\n",
    "To run the model with a different `batch_size`, we need to rebuild the model and restore the weights from the checkpoint.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "zk2WJ2-XjkGz"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'D:\\\\my cources\\\\deeplearning-part2\\\\notebook\\\\training_checkpoints\\\\ckpt_30'"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tf.train.latest_checkpoint(checkpoint_dir)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "LycQ-ot_jjyu"
   },
   "outputs": [],
   "source": [
    "model = build_model(vocab_size, embedding_dim, rnn_units, batch_size=1)\n",
    "\n",
    "model.load_weights(tf.train.latest_checkpoint(checkpoint_dir))\n",
    "\n",
    "model.build(tf.TensorShape([1, None]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "71xa6jnYVrAN"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding_1 (Embedding)      (1, None, 256)            12288     \n",
      "_________________________________________________________________\n",
      "cu_dnngru_1 (CuDNNGRU)       (1, None, 1024)           3938304   \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (1, None, 48)             49200     \n",
      "=================================================================\n",
      "Total params: 3,999,792\n",
      "Trainable params: 3,999,792\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "DjGz1tDkzf-u"
   },
   "source": [
    "### The prediction loop\n",
    "\n",
    "The following code block generates the text:\n",
    "\n",
    "* It Starts by choosing a start string, initializing the RNN state and setting the number of characters to generate.\n",
    "\n",
    "* Get the prediction distribution of the next character using the start string and the RNN state.\n",
    "\n",
    "* Then, use a multinomial distribution to calculate the index of the predicted character. Use this predicted character as our next input to the model.\n",
    "\n",
    "* The RNN state returned by the model is fed back into the model so that it now has more context, instead than only one word. After predicting the next word, the modified RNN states are again fed back into the model, which is how it learns as it gets more context from the previously predicted words.\n",
    "\n",
    "\n",
    "![To generate text the model's output is fed back to the input](./lecture_images/text_generation_sampling.png)\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "WvuwZBX5Ogfd"
   },
   "outputs": [],
   "source": [
    "def generate_text(model, start_string):\n",
    "    # Evaluation step (generating text using the learned model)\n",
    "\n",
    "    # Number of characters to generate\n",
    "    num_generate = 1000\n",
    "\n",
    "\n",
    "    # Converting our start string to numbers (vectorizing) \n",
    "    input_eval = [char2idx[s] for s in start_string]\n",
    "    input_eval = tf.expand_dims(input_eval, 0)\n",
    "\n",
    "    # Empty string to store our results\n",
    "    text_generated = []\n",
    "\n",
    "    # Low temperatures results in more predictable text.\n",
    "    # Higher temperatures results in more surprising text.\n",
    "    # Experiment to find the best setting.\n",
    "    temperature = 1.0\n",
    "\n",
    "    # Here batch size == 1\n",
    "    model.reset_states()\n",
    "    for i in range(num_generate):\n",
    "        predictions = model(input_eval)\n",
    "        # remove the batch dimension\n",
    "        predictions = tf.squeeze(predictions, 0)\n",
    "\n",
    "        # using a multinomial distribution to predict the word returned by the model\n",
    "        predictions = predictions / temperature\n",
    "        predicted_id = tf.multinomial(predictions, num_samples=1)[-1,0].numpy()\n",
    "\n",
    "        # We pass the predicted word as the next input to the model\n",
    "        # along with the previous hidden state\n",
    "        input_eval = tf.expand_dims([predicted_id], 0)\n",
    "\n",
    "        text_generated.append(idx2char[predicted_id])\n",
    "\n",
    "    return (start_string + ''.join(text_generated))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "ktovv0RFhrkn"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "به نام خداوند باز\n",
      "|همیشه جهان را کند کارزار\n",
      "|یکی آتشی دید یزدان‌پرست\n",
      "|چو بگذشت گردان به ایوان خویش\n",
      "|همیدون پیاده سر و تنش گفت\n",
      "|زمین شد پر از سر پر از خاک و ششت\n",
      "|چه آمد به گنجور گستهم گفت\n",
      "|که کاوس های تو پرخاش جوی\n",
      "|زمانه برآید ترا یار کس\n",
      "|رده\n",
      "|بسیچید روز و جگر برگرفت\n",
      "|به آواز گفتند یک شب به خنجر به رنج\n",
      "|تنش لشکری سوخت افراسیاب\n",
      "|که چون زخم کو را همه یاد کرد\n",
      "|زمان و زمان کشته شد زادشم\n",
      "|همی ز آهن آتش پر از بادگون\n",
      "|فرستاده گفت ای جهاندیده مرد\n",
      "|که دست نخستت ز اسفندیار\n",
      "|بدان تا بر و بیشه و بی‌شمار\n",
      "|همی خواب دارد هر آنک کس را ندیدیم کس\n",
      "|نباید که او را کسی زین سپس\n",
      "|نه پاداش با خوردنیها بخواند\n",
      "|شگفتی و از کوه دیدارشان\n",
      "|فروبارجویست مانم به چیست\n",
      "|جز از پهلوان جهان کرد راست\n",
      "|شهنشاه وهرکو پرا بگذرد\n",
      "|نکوهر بود\n",
      "|جز از نیکدل چاره گردند راه\n",
      "|همه خاک بیند به پیکار او\n",
      "|بدان نامداران جنگش دراز\n",
      "|به کش نیازی دهد\n",
      "|ز کیوان و باعواب قیدافه چون زرگرفت\n",
      "|به حقنا رسید از فزود\n",
      "|ولیکن یکی گفت شاه جگاه\n",
      "|بفرجام بس زنده گر دوستی\n",
      "|بغلتید و دل شاددل بود زور\n",
      "|شب آمد بخیره ز ما بنده‌ایم\n",
      "|هم استه بدین تیره خاک نژند\n",
      "|بیاورد گرسیوز آن \n"
     ]
    }
   ],
   "source": [
    "print(generate_text(model, start_string=\"به نام خد\"))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-block alert-info\">\n",
    "<div style=\"direction:rtl;text-align:right;font-family:B Lotus, B Nazanin, Tahoma\"> دانشگاه تربیت دبیر شهید رجایی<br>مباحث ویژه 2 - یادگیری عمیق پیشرفته<br>علیرضا اخوان پور<br>97-98<br>\n",
    "</div>\n",
    "<a href=\"https://www.srttu.edu/\">SRTTU.edu</a> - <a href=\"http://class.vision\">Class.Vision</a> - <a href=\"http://AkhavanPour.ir\">AkhavanPour.ir</a>\n",
    "</div>"
   ]
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "collapsed_sections": [],
   "name": "text_generation_sequential.ipynb",
   "private_outputs": true,
   "provenance": [],
   "toc_visible": true,
   "version": "0.3.2"
  },
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}