20210904_en_NLG.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1 Physical GPUs, 1 Logical GPUs\n",
      "Version:  2.5.0\n",
      "Eager mode:  True\n",
      "GPU is available\n"
     ]
    }
   ],
   "source": [
    "import os\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import pickle\n",
    "import json\n",
    "\n",
    "import tensorflow as tf\n",
    "from tensorflow.keras.models import Model\n",
    "from tensorflow.keras.layers import Input, Dense, LSTM, Embedding, Activation, MultiHeadAttention\n",
    "from tensorflow.keras.layers import LayerNormalization\n",
    "from tensorflow.keras.optimizers import Adam\n",
    "from tensorflow.keras.callbacks import ModelCheckpoint, EarlyStopping, CSVLogger\n",
    "from tensorflow.keras.datasets import imdb\n",
    "from tensorflow.keras.preprocessing import sequence\n",
    "\n",
    "os.environ['CUDA_VISIBLE_DEVICES']='1'\n",
    "tf.config.set_soft_device_placement(True)\n",
    "gpus = tf.config.experimental.list_physical_devices('GPU')\n",
    "if gpus:\n",
    "    try:\n",
    "        # Restrict TensorFlow to only use the fourth GPU\n",
    "        tf.config.experimental.set_visible_devices(gpus[0], 'GPU')\n",
    "        # Currently, memory growth needs to be the same across GPUs\n",
    "        for gpu in gpus:\n",
    "            tf.config.experimental.set_memory_growth(gpu, True)\n",
    "        logical_gpus = tf.config.experimental.list_logical_devices('GPU')\n",
    "        print(len(gpus), \"Physical GPUs,\", len(logical_gpus), \"Logical GPUs\")\n",
    "    except RuntimeError as e:\n",
    "        # Memory growth must be set before GPUs have been initialized\n",
    "        print(e)\n",
    "\n",
    "print(\"Version: \", tf.__version__)\n",
    "print(\"Eager mode: \", tf.executing_eagerly())\n",
    "print(\"GPU is\", \"available\" if tf.config.list_physical_devices(\"GPU\") else \"NOT AVAILABLE\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Constants"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "folder_name = '20210904_en_NLG'\n",
    "wv_dim = 32\n",
    "que_pad = 300\n",
    "num_words = 10000"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Load data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "<string>:6: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "25000\n",
      "2500\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/ai-lab/anaconda3/lib/python3.7/site-packages/tensorflow/python/keras/datasets/imdb.py:155: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
      "  x_train, y_train = np.array(xs[:idx]), np.array(labels[:idx])\n",
      "/home/ai-lab/anaconda3/lib/python3.7/site-packages/tensorflow/python/keras/datasets/imdb.py:156: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
      "  x_test, y_test = np.array(xs[idx:]), np.array(labels[idx:])\n"
     ]
    }
   ],
   "source": [
    "(text_train, class_train), (text_vali, class_vali) = imdb.load_data(\n",
    "    num_words = num_words-2,\n",
    ")\n",
    "text_vali = text_vali[:2500]\n",
    "print(len(text_train))\n",
    "print(len(text_vali))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Add BOS and EOS"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Index num_words -2 = BOS\n",
    "# Index num_words -1 = EOS\n",
    "null = [s.insert(0,num_words-2) for s in text_train]\n",
    "null = [s.append(num_words-1) for s in text_train]\n",
    "null = [s.insert(0,num_words-2) for s in text_vali]\n",
    "null = [s.append(num_words-1) for s in text_vali]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Padding"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "decoder_train shape:(25000, 300)\n",
      "teacher_train shape:(25000, 300)\n",
      "decoder_vali shape:(2500, 300)\n",
      "teacher_vali shape:(2500, 300)\n"
     ]
    }
   ],
   "source": [
    "# Data padding for training\n",
    "decoder_train = sequence.pad_sequences(\n",
    "    text_train,\n",
    "    maxlen = que_pad,\n",
    "    padding = 'post',\n",
    ")\n",
    "teacher_train = decoder_train[:,1:]\n",
    "teacher_train = sequence.pad_sequences(\n",
    "    teacher_train,\n",
    "    maxlen = que_pad,\n",
    "    padding = 'post',\n",
    ")\n",
    "# Data padding for valiation\n",
    "decoder_vali = sequence.pad_sequences(\n",
    "    text_vali,\n",
    "    maxlen = que_pad,\n",
    "    padding = 'post',\n",
    ")\n",
    "teacher_vali = decoder_vali[:,1:]\n",
    "teacher_vali = sequence.pad_sequences(\n",
    "    teacher_vali,\n",
    "    maxlen = que_pad,\n",
    "    padding = 'post',\n",
    ")\n",
    "print(f'decoder_train shape:{decoder_train.shape}')\n",
    "print(f'teacher_train shape:{teacher_train.shape}')\n",
    "print(f'decoder_vali shape:{decoder_vali.shape}')\n",
    "print(f'teacher_vali shape:{teacher_vali.shape}')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[9998    1   14   22   16   43  530  973 1622 1385   65  458 4468   66\n",
      "  3941    4  173   36  256    5   25  100   43  838  112   50  670    2\n",
      "     9   35  480  284    5  150    4  172  112  167    2  336  385   39\n",
      "     4  172 4536 1111   17  546   38   13  447    4  192   50   16    6\n",
      "   147 2025   19   14   22    4 1920 4613  469    4   22   71   87   12\n",
      "    16   43  530   38   76   15   13 1247    4   22   17  515   17   12\n",
      "    16  626   18    2    5   62  386   12    8  316    8  106    5    4\n",
      "  2223 5244   16  480   66 3785   33    4  130   12   16   38  619    5\n",
      "    25  124   51   36  135   48   25 1415   33    6   22   12  215   28\n",
      "    77   52    5   14  407   16   82    2    8    4  107  117 5952   15\n",
      "   256    4    2    7 3766    5  723   36   71   43  530  476   26  400\n",
      "   317   46    7    4    2 1029   13  104   88    4  381   15  297   98\n",
      "    32 2071   56   26  141    6  194 7486   18    4  226   22   21  134\n",
      "   476   26  480    5  144   30 5535   18   51   36   28  224   92   25\n",
      "   104    4  226   65   16   38 1334   88   12   16  283    5   16 4472\n",
      "   113  103   32   15   16 5345   19  178   32 9999    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]]\n",
      "[[   1   14   22   16   43  530  973 1622 1385   65  458 4468   66 3941\n",
      "     4  173   36  256    5   25  100   43  838  112   50  670    2    9\n",
      "    35  480  284    5  150    4  172  112  167    2  336  385   39    4\n",
      "   172 4536 1111   17  546   38   13  447    4  192   50   16    6  147\n",
      "  2025   19   14   22    4 1920 4613  469    4   22   71   87   12   16\n",
      "    43  530   38   76   15   13 1247    4   22   17  515   17   12   16\n",
      "   626   18    2    5   62  386   12    8  316    8  106    5    4 2223\n",
      "  5244   16  480   66 3785   33    4  130   12   16   38  619    5   25\n",
      "   124   51   36  135   48   25 1415   33    6   22   12  215   28   77\n",
      "    52    5   14  407   16   82    2    8    4  107  117 5952   15  256\n",
      "     4    2    7 3766    5  723   36   71   43  530  476   26  400  317\n",
      "    46    7    4    2 1029   13  104   88    4  381   15  297   98   32\n",
      "  2071   56   26  141    6  194 7486   18    4  226   22   21  134  476\n",
      "    26  480    5  144   30 5535   18   51   36   28  224   92   25  104\n",
      "     4  226   65   16   38 1334   88   12   16  283    5   16 4472  113\n",
      "   103   32   15   16 5345   19  178   32 9999    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]]\n"
     ]
    }
   ],
   "source": [
    "print(decoder_train[:1])\n",
    "print(teacher_train[:1])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Positional Encoding\n",
    "[Positional Encoding](https://www.tensorflow.org/text/tutorials/transformer)\n",
    "based on [Vaswani+17, Attention is All You Need](https://arxiv.org/abs/1706.03762)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_angles(pos, i, d_model):\n",
    "    angle_rates = 1 / np.power(10000, (2 * (i//2)) / np.float32(d_model))\n",
    "    return pos * angle_rates\n",
    "\n",
    "def positional_encoding(position, d_model):\n",
    "    angle_rads = get_angles(\n",
    "        np.arange(position)[:, np.newaxis],\n",
    "        np.arange(d_model)[np.newaxis, :],\n",
    "        d_model\n",
    "    )\n",
    "\n",
    "    # apply sin to even indices in the array; 2i\n",
    "    angle_rads[:, 0::2] = np.sin(angle_rads[:, 0::2])\n",
    "\n",
    "    # apply cos to odd indices in the array; 2i+1\n",
    "    angle_rads[:, 1::2] = np.cos(angle_rads[:, 1::2])\n",
    "\n",
    "    pos_encoding = angle_rads[np.newaxis, ...]\n",
    "\n",
    "    return tf.cast(pos_encoding, dtype=tf.float32)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Test positional encoding"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(1, 10, 64)\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX0AAAEKCAYAAAD+XoUoAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8vihELAAAACXBIWXMAAAsTAAALEwEAmpwYAAAhzUlEQVR4nO3de5RcZZnv8e+vuxOBcDcQQxIFnAyKHECNoOJBEPCEDBL0iIKKiDgRDxnR5Tkj4iyd0TVjxhkvODJgBoGoyEUUzGjkYhQBFU1A5I7EcAuJuQEhgUAu/Zw/9m4sK9VdVV273rr9Pmvt1bWvz7sD6+m3n3r3uxURmJlZb+hrdQPMzCwdJ30zsx7ipG9m1kOc9M3MeoiTvplZD3HSNzPrIU76ZmZNJOkiSask3T3Mfkn6mqQlku6U9JqSfdMlPZDvO7uI9jjpm5k11yXA9BH2HwtMzZdZwPkAkvqB8/L9+wMnS9q/0cY46ZuZNVFE3AQ8McIhM4FvReZWYFdJE4FDgCURsTQiNgGX58c2ZKDRC6Sgge1CY3dsdTPMrAPExrVrImKPRq7Rt/PkYMtztcS6Byg9cG5EzK0z3CTgsZL1Zfm2StsPrfPa2+iMpD92Rwb2O77VzTCzDrD5josfafgiW56rKedsvuPi5yJiWoPRVGFbjLC9IR2R9M3MkpJQX3+qaMuAKSXrk4HlwNhhtjfENX0zs22IvoGxVZeCzAfen4/ieT2wLiJWAIuAqZL2kTQWOCk/tiHu6ZuZlSuwpy/pMuAIYLykZcBngTEAEXEBsACYASwBngVOy/dtkTQbuA7oBy6KiHsabY+TvplZGQHqLybpR8TJVfYHcOYw+xaQ/VIojJO+mVk5ib50Nf2knPTNzCpI+EVuUk76Zmbl0o7eScpJ38ysjBB9A2Na3YymcNI3MyvXxT39po7Tl7SrpKsk3S/pPklvkLS7pBskPZj/3K2ZbTAzGw319VddOlGzH846F7g2Il4BHATcB5wNLIyIqcDCfN3MrH1IqL+/6tKJmpb0Je0MHA58EyAiNkXEU2SzxM3LD5sHnNCsNpiZjYbo3p5+M2v6+wKrgYslHQTcBpwFTMgfMSYiVkjas9LJkmaRzS0NY8Y1sZlmZmXUR39x0yy0lWaWdwaA1wDnR8SrgWeoo5QTEXMjYlpETNPAds1qo5nZttS9Pf1mJv1lwLKI+E2+fhXZL4GV+QsCyH+uamIbzMzqJuSkX6+I+BPwmKT98k1HAfeSzRJ3ar7tVOCHzWqDmdlodWvSb/Y4/b8DLs2nBV1KNntcH3ClpNOBR4ETm9wGM7P6dPE4/aYm/Yi4A6j0VpmjmhnXzKwxTvpmZj1DEn1junP0jpO+mVk5l3fMzHqLk76ZWQ/p61Orm9AUTvpmZmUkISd9M7Pe0d9fzGNMkqaTTT7ZD1wYEXPK9v8/4L356gDwSmCPiHhC0sPAemArsCUiKo2GrIuTvplZOVFIT19SP3AecAzZLAWLJM2PiHuHjomIfwP+LT/+bcDHI+KJksscGRFrGm5MrtlTK5uZdZxslk1VXWpwCLAkIpZGxCbgcrKZhodzMnBZ43cwPCd9M7NtiD5VX4DxkhaXLLPKLjQJeKxkfVm+bduI0g7AdOD7JZsDuF7SbRWuPSou75iZlau9vLOmSp290kVimGPfBvyyrLRzWEQsz6egv0HS/RFxUy0NG457+mZmFRRU3lkGTClZnwwsH+bYkygr7UTE8vznKuBqsnJRQ5z0zczKSNA/oKpLDRYBUyXtk088eRLZTMNl8bQL8GZKZh2WNE7STkOfgbcCdzd6by7vmJlVIDU+eicitkiaDVxHNmTzooi4R9IZ+f4L8kPfDlwfEc+UnD4BuDpvxwDw3Yi4ttE2OembmZWRVNgTuRGxAFhQtu2CsvVLgEvKti0FDiqkESWc9M3MKvATuWZmPcRJ38ysV4ihcfhdx0nfzKyMEH0D3Tm40UnfzKycPLWymVlPKWLIZjty0jczK5NNuNbqVjSHk76ZWTmXd8zMeonoK+glKu3GSd/MrIzc0zcz6y1+OGsUKr3fUdLuwBXA3sDDwLsi4slmtsPMrB4S9Hdp0k9RtDoyIg4uedHA2cDCiJgKLMzXzczaSn+fqi6dqBXfVMwE5uWf5wEntKANZmbDEtUTfqcm/WbX9Ife7xjANyJiLjAhIlYARMSK/DVg28jfB5m9E3LMuCY308zszyQY62kYRmWb9zvWemL+C2IuQN8O44d7p6SZWeEkGOjQnnw1TU36pe93lDT0fseVkibmvfyJwKpmtsHMrF7CX+TWbYT3O84HTs0PO5WSd0KambUFdW9Nv5lFqwnALZJ+D/wW+HH+fsc5wDGSHgSOydfNzNpG1tPvq7rUdC1puqQHJC2RtM1oRUlHSFon6Y58+Uyt545G08o7w73fMSLWAkc1K66ZWRGK6MlL6gfOI+vgLgMWSZofEfeWHXpzRBw3ynPr4idyzczK9ElFjd45BFiSd4KRdDnZsPVaEncj5w6rO8ckmZk1qF+qugDjJS0uWWaVXWYS8FjJ+rJ8W7k3SPq9pJ9IelWd59bFPX0zszJ1TMOwpmS2gYqXqrCtfAj67cDLImKDpBnANcDUGs+tm3v6ZmYVFDR6ZxkwpWR9MrC89ICIeDoiNuSfFwBjJI2v5dzRcE/fzKxMgQ9nLQKmStoHeBw4CXjPX8bSS4CVERGSDiHrjK8Fnqp27mg46Zv1KPX1t7oJbUsU80VuRGyRNBu4DugHLoqIeySdke+/AHgn8BFJW4CNwEkREUDFcxttk5O+mVmZIqdWzks2C8q2XVDy+evA12s9t1FO+mZmZbp5GgYnfbMqurUM0q33VYgufomKk76ZWZmh+fS7kZO+mVkFTvpmNejGkkHKe+raWP2d9f9Fn1+iYmbWQ1zTNzPrHeKFuXW6jpO+mVkFfU761qlcZ29M38DYZLFS1r77E95X38CYZLGKIKC/O3O+k76Z2TYEfa7pm5n1BgFjanwdYqdx0rdCpSqF9I1JV5ro1jJI/9jtk8Uas/2OyWKtK+AaLu+YmfUSyeUdM7NeITx6x8ysp7i8Y4Xq1iGHqWrtAwnr0QPbjUsWa+y4XZLFGpMw1ovGpavpryzgGhKM6fcXuWZmPcHlHTOzHuPyzihJ6gcWA49HxHGSdgeuAPYGHgbeFRFPNrsd7SZpeSfh8MaxO+ycJE7K0sR2u+yRLNYOu6T59wPYcdftujLWkgKuIVRYT1/SdOBcsvfcXhgRc8r2vxf4ZL66AfhIRPw+3/cwsB7YCmyJiGmNtidF0eos4L6S9bOBhRExFViYr5uZtY98ls1qS9XLZJ3e84Bjgf2BkyXtX3bYQ8CbI+JA4PPA3LL9R0bEwUUkfGhy0pc0Gfgb4MKSzTOBefnnecAJzWyDmVm9spp+9aUGhwBLImJpRGwCLifLgS+IiF+VVDtuBSYXeCvbaHZ556vA3wM7lWybEBErACJihaQ9K50oaRYwC4AxaUZPdOuImlQlF0hXChm3x15J4gDsuke60TsvnpBulMurJqUrkR00OV2sXxRwjTqmYRgvaXHJ+tyIKO2pTwIeK1lfBhw6wvVOB35Ssh7A9ZIC+EbZtUelaUlf0nHAqoi4TdIR9Z6f39xcgL4dxkexrTMzG4GgxhGba6qUXSr9PVAxn0k6kizpv6lk82ERsTzvHN8g6f6IuKmmlg2jmT39w4DjJc0AtgN2lvQdYKWkiXkvfyKwqoltMDOrW4FDNpcBU0rWJwPLt4knHUhWBj82ItYObY+I5fnPVZKuJisXNZT0m1bTj4hPRcTkiNgbOAn4WUS8D5gPnJofdirww2a1wcxsdLI3Z1VbarAImCppH0ljyXLh/L+IJL0U+AFwSkT8oWT7OEk7DX0G3grc3eidtWKc/hzgSkmnA48CJ7agDRWlrOmnnHUw5ZDDXSe/NEmcPSen+57itX81Plmsw/bdPVmsg16yU/WDCjJlu63JYn2ggGsU1dOPiC2SZgPXkQ3ZvCgi7pF0Rr7/AuAzwIuB/1QWc2ho5gTg6nzbAPDdiLi20TYlSfoRcSNwY/55LXBUirhmZqORTcNQzDj9iFgALCjbdkHJ5w8BH6pw3lLgoEIaUcJP5JqZVdClszA46ZdKWd4Zu9NuyWKlKrkA7LNfmlLIO14zKUkcgCP2SVdyedmfv8Nruq2/uzJZrOXX35gsVlH6Kg686XxO+mZmZYR7+mZmPaVLX5zlpG9mtg25p98TBl6U7sUc4/bovjo7wBmH75skzlv3TjfckFsuTxbqwe9ckyzWXdf+MVmsRU8+lyxWEUTN4/A7jpO+mVkFLu+YmfWQLs35TvqlxiScjXLivt1XcgGYsdv6JHFW/Munk8QB+NXcXyeLddOaZ5PFSlm+OHrPdDOVfulPjV+j51+XKOlFwP8me9vVC+dExOea0ywzs9bq0pxfc0//h8A64Dbg+eY1x8ysPaR4rWAr1Jr0J0fE9Ka2pA1sv9tLksWa+YZ0o3dSlVwA7p49O0mcy/77wSRxADZuTfc6h5MOTfek8RsvnlP9oIJcsXGfZLF4zZTqx1Sh/HWJ3ajWX2a/kvQ/mtoSM7M2IlVfOtGIPX1Jd5G95WUAOE3SUrLyjoDIX+RrZtZVRO+Wd45L0gozszajTu3KVzFi0o+IRwAkfTsiTindJ+nbwCkVT+xQu09JV089/bXpYt192tuTxbrgBw8kiXP4+B2SxAF4983fSBbrI3eme9J4+ofmJYu1/W4TksUqhPxw1qtKVyT1A68tvjlmZq0noKB3qLSdEctWkj4laT1woKSnJa3P11fhd9uaWReTVHXpRNXKO18AviDpCxHxqURtapmDD0z3J+jY7/1LslipSi4AHzr25UnirP3KZUniAOx8+heSxXrp645MFuvpyz6YLNac/d6RLNY/FHCN7IncAi4ESJoOnEv2jtwLI2JO2X7l+2cAzwIfiIjbazl3NGot75wj6R3Am8hG89wcEdc0GtzMrF0VkfPzUvh5wDHAMmCRpPkRcW/JYccCU/PlUOB84NAaz61braOSzgPOAO4C7gbOkHReI4HNzNqX6FP1pQaHAEsiYmlEbAIuB2aWHTMT+FZkbgV2lTSxxnPrVmtP/83AARERAJLmkf0CMDPrPsU9fDUJeKxkfRlZb77aMZNqPLdutSb9B4CXAo/k61OAOxsNXqtXv+Kl/PKX3faHxauqH1KQr74n3YyU3WjtDem+f0lpc8JYn1ibLF3wDzs0PpxXEWhway2Hjpe0uGR9bkTMLb1UhXPK5/UY7phazq1brUn/xcB9kn6br78O+LWk+QARcXyjDTEzayeKwVoOWxMR00bYv4yskzxkMrC8xmPG1nBu3WpN+p9pNJCZWecIqC3pV7MImCppH+Bx4CTgPWXHzAdmS7qcrHyzLiJWSFpdw7l1qynpR8QvJL0MmBoRP5W0PTAQEcNO3yhpO+Am4EV5nKsi4rOSdgeuIJub/2HgXRHxZGO3YWZWsGh8dtWI2CJpNnAd2bDLiyLiHkln5PsvABaQDddcQjZk87SRzm20TbW+ROVvgVnA7sDLyf7MuAA4aoTTngfeEhEbJI0BbpH0E+AdwMKImCPpbOBs4JMN3IOZWbGisJ4+EbGALLGXbrug5HMAZ9Z6bqNqHbJ5JnAY8HTekAeBPUc6IR9+tCFfHZMvQTbkaGjSj3nACfU12cys+RSDVZdOVGvSfz4fJwqApAFq+BZZUr+kO8imbbghIn4DTIiIFQD5z4q/PCTNkrRY0uLVa9bU2EwzsyIEDG6pvnSgWpP+LySdA2wv6Rjge8B/VzspIrZGxMFk5aBDJB1Qa8MiYm5ETIuIaXuMT/cScTMzgqy8U23pQLUm/bOB1WQPZH2YrMZU8xQXEfEUcCMwHViZP21G/nNV7c01M0shYHCw+tKBah29MyjpGuCaiFhdyzmS9gA2R8RT+Wifo4F/JRuedCowJ//p2TrNrO10as2+mmqvSxTwWWA22dNhkrQV+I+I+FyVa08E5uWTBvUBV0bEjyT9GrhS0unAo8CJjd6EmVnhejHpAx8jG7Xzuoh4CEDSvsD5kj4eEV8Z7sSIuBN4dYXtaxl5qKeZWWtFQG3TMHScajX99wMnDyV8gIhYCrwv32dm1pW6dchmtZ7+mIjYZrxkRKzOH7gyM+tCxT2c1W6qJf1No9xnZtbZejTpHyTp6QrbBWzXhPaYmbVegdMwtJtq78jtT9UQM7N2IXp0yKaZWRGioNdQpROwtTtH7zjpm5mVG5qGoQs56ZuZVeDyjpk1XeeVQbpVj36Ra2bWs5z0zcx6RBdPw+Ckb1aFSy69KIgtm1vdiKaodT59M7PeEWQ9/WpLgyTtLukGSQ/mP3ercMwUST+XdJ+keySdVbLvHyU9LumOfJlRLaaTvplZmSCIrVurLgU4G1gYEVOBhfl6uS3AJyLilcDrgTMl7V+y/ysRcXC+VH2JupO+mVm5INWbs2YC8/LP84ATtmlKxIqIuD3/vB64D5g02oCu6VuhXP+27lDzF7njJS0uWZ8bEXPrCDQhIlZAltwl7TnSwZL2JntPyW9KNs+W9H5gMdlfBE+OdA0nfTOzclHzF7lrImLaSAdI+inwkgq7Pl1PkyTtCHwf+FhEDE2EeT7webK/TT4PfAn44EjXcdI3M9tGEAUN2YyIo4fbJ2mlpIl5L38isGqY48aQJfxLI+IHJddeWXLMfwE/qtYeJ/0e4JKLVRLR6ha0saHRO803HzgVmJP//GH5Afm7yr8J3BcRXy7bN3GoPAS8Hbi7WkB/kWtmto1I9UXuHOAYSQ8Cx+TrSNpL0tBInMOAU4C3VBia+UVJd0m6EzgS+Hi1gO7pm5mVC4oakjlymIi1wFEVti8HZuSfbyGb4r/S+afUG9NJ38xsG56GwaxnufbduMFO+0esffROx3HSNzPbhnv6Zma9I93oneSalvQlTQG+RfZQwiDZk2rnStoduALYG3gYeFe1J8i6kYdRNqbTqgW16rgySI067a6CIIoZndN2mjlkc7hJgmqZYMjMrHUSzbLZCk3r6ecPDAzNKbFe0tAkQTOBI/LD5gE3Ap9sVjvMzOoWQWze1OpWNEWSmn7ZJEE1TTAkaRYwC2DKlCkpmmkFSFWdcBmkswx23I1FUQ9ftZ2mP5E7zCRBVUXE3IiYFhHT9hg/vnkNNDOrxOWd+g0zSVBNEwyZmbVMFDfhWrtpWk9/hEmChiYYgmEmGDIza7UYHKy6dKJm9vSHJgm6S9Id+bZzyCYUulLS6cCjwIlNbENdunUYZcrydzfW2lPeUcrad3Thf6vCRBBbOzOpV9PM0TvDThJEhQmGzMzaRUQwuHlLq5vRFH4i18ysXOCevhWrW0suqSJ1axkk6X2lC9WBQzad9M3MekZEMJhgPv1WcNI3M6ugU0fnVOOkb2ZWzqN3rJN1Y+02ZZ19a9LvX9LFSvpdRbJIxUg1eqfWWYclPQysB7YCWyJiWj3nl/KL0c3MKhjcOlh1KUA9sw4fGREHDyX8UZwPOOmbmW0rH7JZbSnATLLZhsl/ntDs813eKdGtwyhTlgy2JgqWsmSV6p4gbSkp5X11Wnmnjpr+eEmLS9bnRsTcOiLVNOsw2f/y10sK4BslMWo9/wVO+mZmZYKaR++sKSu3bEPST8neIFju03U06bCIWJ4n9Rsk3R8RN9Vx/guc9M3MykUwuKmYL3Ij4ujh9kmqadbhiFie/1wl6WrgEOAmRjFrsZN+i7g80WicdPe0JWHNJWl5J+UIqE6r7wQMphmnPzTr8ByGmXVY0jigL38D4TjgrcDnaj2/nL/INTMrE0SqL3LnAMdIehA4Jl9H0l6SFuTHTABukfR74LfAjyPi2pHOH4l7+mZm5QIiwTQMEbGWCrMO5+WcGfnnpcBB9Zw/Eid9M7NthKdh6AUph1F2Y50dYEui+9qc8N8v1T1B4vtKOJ/Y5k5LoJ5a2cysd0QEWwsavdNunPTNzLbh8k5PSDqMsgtLLgCbEt1YqjipYz23JV2ieXZzuvpOyliFcHnHzKyHBETKnllCTvpmZmWCKGoWzbbjpG9mVi4gOvHFvjVw0i+R8r9xyqF5SWvSiXpHGzen64Wtfz5dPXpDwhEj6zelu6+nn9ucLFYRImBrwn+flJz0zczKRbimb2bWSwad9Osj6SLgOGBVRByQb6v7fY4ppXxKNuXMjalKLgBPP5fmT+InNqYrF6x7Pl3J5cmE97Vmw/PJYq3dsClZrEJ08ZDNZs6yeQkwvWxb3e9zNDNLLYDBwai6dKKm9fQj4iZJe5dtngkckX+eB9wIfLJZbTAzG5UIf5FbkJrf5yhpFjALYMqUKUkal7KE91zCYKlKLgArn0nzZ/yqZ9KVJlasey5drKcSxlq3MVmstQn/DYsQXfxwVtu+RCUi5kbEtIiYtsf48a1ujpn1kjzpV1s6Ueqeft3vczQzS697n8hN3dMfep8j1Pg+RzOz5PIncqstnaiZQzYvI/vSdrykZcBnyd7feKWk04FHgRObFX80Uj65ujHhbIqp6uwAjyWqEz+4ckOSOACPrH0mWayVq59NFmtDwpr+hoTfVRQhSDNOv5Zh7JL2y48Zsi/wmYj4qqR/BP4WWJ3vOyciFjCCZo7eOXmYXXW9z9HMLLkIBtOM3hkaxj5H0tn5+l+MaIyIB4CDAST1A48DV5cc8pWI+PdaA7btF7lmZq0SkfX0qy0FmEk2fJ385wlVjj8K+GNEPDLagJ6GocSmhF/crO7CkgvAncvWJYlzf6I4AE+uSldKWrcmXXnnmdWPJov1/Lo1yWIVJdGbs2oexp47CbisbNtsSe8HFgOfqDbLgXv6ZmblonovP+/pj5e0uGSZVX4pST+VdHeFZWY9TZI0Fjge+F7J5vOBl5OVf1YAX6p2Hff0zczK1f5w1pqImDbipSKOHm6fpHqGsR8L3B4RK0uu/cJnSf8F/Khag93TNzMrE2QTrlVbClDPMPaTKSvt5L8ohrwduLtaQPf0S2zckm7I5uNPp5tGIFWdHeDOJWuTxFn1WLp7Wrf8oWSxNq5dnizWpmfS/RvGYIfNYxPB1k1JavoVh7FL2gu4MCJm5Os7AMcAHy47/4uSDib7PfVwhf3bcNI3MysTAYPR/E5gRKylwjD2iFgOzChZfxZ4cYXjTqk3ppO+mVkFWxMk/VZw0i/xdMKXZTywcn2yWKlKLgCPL1ld/aACPPXwXUniAGx8cmX1gwqSsgzyop12TxZr50l/nSzW8jsubvgaQdpZd1Ny0jczq8A9fTOzHjEYsKlDJ1Srxkm/xIr16UbULHroiWSxUpVcAFbff2uSOJsTjjxJWQYZ/4rXJ4t1wLRJyWL9n8P3TRZrxuVVB7DUxOUdM7MeEYTLO2ZmvcJf5JqZ9Rgn/R7w0JPpZjhctiRdTT9VnR3S1drH//XrksQBeO1RByeLNedtr0oWa79V6f6/uP/zH0wWqwgRHr1jZtYzAo/eMTPrGa7p94g7E07i9af770wWK+Xwxpe98W1J4vzzhw9NEgfgHVF14sLC3PK2/5Us1n/eviJZrF3G9CeLVRSXd8zMekRW0291K5rDSd/MrAL39M3MekQA6d6YnZaTfol7Hkz38ub1K/6YLFaqOjvAdf807JvhCrX5nFOrH1SQsy68PVmsA3beLlmsr9z8xWSx/nnDgcliMf2VDV8iCI/eMTPrFdnoHSd9M7Pe4C9yiyVpOnAu0E/2Hsg5rWhHuRV/eCxZrB0n7J0sVqqSC8AfDj08SZzrHnkqSRxIWwY58ubxyWJ97KOXVT+oIF/46muTxSqCe/oFktQPnEf2kt9lwCJJ8yPi3tRtMTMbjnv6xTkEWBIRSwEkXQ7MBJz0zawtDNK90zAoEv8JI+mdwPSI+FC+fgpwaETMLjtuFjArXz0ASPdYZDrjgXRDhtLoxnuC7ryvbrwngP0iYqdGLiDpWrJ/n2rWRMT0RmKl1oqevips2+Y3T0TMBeYCSFocEdOa3bDUuvG+uvGeoDvvqxvvCbL7avQanZbI69HXgpjLgCkl65OB5S1oh5lZz2lF0l8ETJW0j6SxwEnA/Ba0w8ys5yQv70TEFkmzgevIhmxeFBH3VDltbvNb1hLdeF/deE/QnffVjfcE3XtfhUj+Ra6ZmbVOK8o7ZmbWIk76ZmY9pK2TvqTpkh6QtETS2a1uTxEkTZH0c0n3SbpH0lmtblNRJPVL+p2kH7W6LUWRtKukqyTdn/83e0Or21QESR/P//+7W9JlktJN71kgSRdJWiXp7pJtu0u6QdKD+c/dWtnGdtO2Sb9kuoZjgf2BkyXt39pWFWIL8ImIeCXweuDMLrkvgLOA+1rdiIKdC1wbEa8ADqIL7k/SJOCjwLSIOIBsQMVJrW3VqF0ClI+pPxtYGBFTgYX5uuXaNulTMl1DRGwChqZr6GgRsSIibs8/rydLIpNa26rGSZoM/A1wYavbUhRJOwOHA98EiIhNEfFUSxtVnAFge0kDwA506LMyEXET8ETZ5pnAvPzzPOCElG1qd+2c9CcBpdNeLqMLkmMpSXsDrwZ+0+KmFOGrwN/TXS8c2hdYDVycl60ulDSu1Y1qVEQ8Dvw78CiwAlgXEde3tlWFmhARKyDrZAF7trg9baWdk35N0zV0Kkk7At8HPhYRT7e6PY2QdBywKiJua3VbCjYAvAY4PyJeDTxDF5QK8hr3TGAfYC9gnKT3tbZVlko7J/2una5B0hiyhH9pRPyg1e0pwGHA8ZIeJivDvUXSd1rbpEIsA5ZFxNBfYleR/RLodEcDD0XE6ojYDPwAeGOL21SklZImAuQ/V7W4PW2lnZN+V07XIElkNeL7IuLLrW5PESLiUxExOSL2Jvvv9LOI6PieY0T8CXhM0n75pqPojinAHwVeL2mH/P/Ho+iCL6hLzAeGXqJ8KvDDFral7bTt6xJHOV1DJzgMOAW4S9Id+bZzImJB65pkI/g74NK847EUOK3F7WlYRPxG0lXA7WSjyX5Hh05dIOky4AhgvKRlwGeBOcCVkk4n+wV3Yuta2H48DYOZWQ9p5/KOmZkVzEnfzKyHOOmbmfUQJ30zsx7ipG9m1kOc9C0pSVsl3ZHP7vg9STvUef5e+XBDJB0saUbJvuO7ZTZWs2bxkE1LStKGiNgx/3wpcNtoH1KT9AGymSJnF9hEs67mnr610s3AX+Xzn18j6U5Jt0o6EEDSm/O/Cu7IJzzbSdLe+V8JY4HPAe/O979b0gckfT0/92WSFubXXCjppfn2SyR9TdKvJC2V9M6W3b1ZCzjpW0vkU/oeC9wF/BPwu4g4EDgH+FZ+2P8FzoyIg4H/CWwcOj+fbvszwBURcXBEXFEW4uvAt/JrXgp8rWTfROBNwHFkT2+a9QwnfUtt+3z6icVkj8h/kywBfxsgIn4GvFjSLsAvgS9L+iiwa0RsqSPOG4Dv5p+/nccYck1EDEbEvcCERm7GrNO07dw71rU25j33F+STfpWLiJgj6cfADOBWSUcDz40ybumXV8+Xhh/l9cw6knv61g5uAt4LIOkIYE1EPC3p5RFxV0T8K9lfBq8oO289sNMw1/wVf34F4HuBW4putFknctK3dvCPwDRJd5LV2Iemxf1Y/qXt78nq+T8pO+/nwP5DX+SW7fsocFp+zVPI3t9r1vM8ZNPMrIe4p29m1kOc9M3MeoiTvplZD3HSNzPrIU76ZmY9xEnfzKyHOOmbmfWQ/w958Kjs2GzfdAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "from matplotlib import pyplot as plt\n",
    "n, d = 10, 64\n",
    "pos_encoding = positional_encoding(n, d)\n",
    "print(pos_encoding.shape)\n",
    "pos_encoding = pos_encoding[0]\n",
    "\n",
    "# Juggle the dimensions for the plot\n",
    "pos_encoding = tf.reshape(pos_encoding, (n, d//2, 2))\n",
    "pos_encoding = tf.transpose(pos_encoding, (2, 1, 0))\n",
    "pos_encoding = tf.reshape(pos_encoding, (d, n))\n",
    "\n",
    "plt.pcolormesh(pos_encoding, cmap='RdBu')\n",
    "plt.ylabel('Depth')\n",
    "plt.xlabel('Position')\n",
    "plt.colorbar()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Embedding"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "def getEmb():\n",
    "    initializer = tf.keras.initializers.GlorotNormal()\n",
    "    mask_embedding = initializer(shape=(1, wv_dim)).numpy()\n",
    "    emb = Embedding(\n",
    "        num_words, \n",
    "        wv_dim, \n",
    "        mask_zero = False,\n",
    "        input_length=(int(que_pad)),\n",
    "        trainable = True,\n",
    "    )\n",
    "    return emb"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Generator"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "[OneHot relaxation](https://www.tensorflow.org/probability/api_docs/python/tfp/distributions/RelaxedOneHotCategorical)\n",
    "based on [Jang+16](https://arxiv.org/abs/1611.01144) and [Maddison+16](https://arxiv.org/abs/1611.00712) \n",
    "<br>\n",
    "[Multi-Head Attention](https://www.tensorflow.org/api_docs/python/tf/keras/layers/MultiHeadAttention) \n",
    "and [Positional Encoding](https://www.tensorflow.org/text/tutorials/transformer)\n",
    "based on [Vaswani+17, Attention is All You Need](https://arxiv.org/abs/1706.03762)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "def Transformer(que_pad, wv_dim, mask = None, rate = 0.1):\n",
    "    # Inputs\n",
    "    mem  = Input((que_pad, wv_dim))\n",
    "    # Constants\n",
    "    ff_dim = wv_dim*16\n",
    "    # Multi-Head Attention\n",
    "    # TODO: concatenation\n",
    "    q = Dense(wv_dim)(mem)\n",
    "    k = Dense(wv_dim)(mem)\n",
    "    v = Dense(wv_dim)(mem)\n",
    "    #print(mask)\n",
    "    mem_new = MultiHeadAttention(\n",
    "        num_heads = 4,\n",
    "        key_dim = wv_dim, \n",
    "        value_dim = wv_dim\n",
    "    )(\n",
    "        q, k, v,\n",
    "        attention_mask = mask,\n",
    "    )\n",
    "    #mem_new = Dropout(rate)(mem_new)\n",
    "    mem_new = LayerNormalization(epsilon=1e-6)(mem_new+mem)\n",
    "    # Feed-Forward skip-connection\n",
    "    ffn = Dense(ff_dim, activation = 'relu')(mem_new)\n",
    "    ffn = Dense(wv_dim)(ffn)\n",
    "    #ffn = Dropout(rate)(ffn)\n",
    "    out = LayerNormalization(epsilon=1e-6)(ffn+mem_new)\n",
    "    model = Model(\n",
    "        mem,\n",
    "        [mem_new, out]\n",
    "    )\n",
    "    return model\n",
    "\n",
    "# GPT downstream\n",
    "def getGPT(emb):\n",
    "    _input = Input((que_pad,))\n",
    "    mem = emb(_input)\n",
    "    # position encoding\n",
    "    pe = positional_encoding(que_pad, wv_dim)\n",
    "    mem = LayerNormalization(epsilon=1e-6)(mem+pe)\n",
    "    mask = np.tri(que_pad, que_pad, 0) # (q_dim, k_dim)\n",
    "    # forward sentence\n",
    "    for i in range(1):\n",
    "        tLayer = Transformer(que_pad, wv_dim, mask)\n",
    "        mem, output = tLayer((mem))\n",
    "        output = Activation('relu')(output)\n",
    "        mem = Activation('relu')(mem)\n",
    "    # Concatenation and output\n",
    "    output = Dense(num_words)(output)\n",
    "    output = Activation('softmax')(output)\n",
    "    model = Model(\n",
    "        _input, \n",
    "        output,\n",
    "        name = 'GPT'\n",
    "    ) \n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"GPT\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_1 (InputLayer)         [(None, 300)]             0         \n",
      "_________________________________________________________________\n",
      "embedding (Embedding)        (None, 300, 32)           320000    \n",
      "_________________________________________________________________\n",
      "tf.__operators__.add (TFOpLa (None, 300, 32)           0         \n",
      "_________________________________________________________________\n",
      "layer_normalization (LayerNo (None, 300, 32)           64        \n",
      "_________________________________________________________________\n",
      "model (Functional)           [(None, 300, 32), (None,  53408     \n",
      "_________________________________________________________________\n",
      "activation (Activation)      (None, 300, 32)           0         \n",
      "_________________________________________________________________\n",
      "dense_5 (Dense)              (None, 300, 10000)        330000    \n",
      "_________________________________________________________________\n",
      "activation_2 (Activation)    (None, 300, 10000)        0         \n",
      "=================================================================\n",
      "Total params: 703,472\n",
      "Trainable params: 703,472\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "emb = getEmb()\n",
    "mleG=getGPT(emb)\n",
    "mleG.compile(\n",
    "    loss='sparse_categorical_crossentropy',\n",
    "    optimizer=Adam(),\n",
    "    metrics=['accuracy'],\n",
    ")\n",
    "mleG.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/300\n",
      "782/782 [==============================] - 39s 47ms/step - loss: 4.7933 - accuracy: 0.3820 - val_loss: 3.8223 - val_accuracy: 0.4282\n",
      "Epoch 2/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.7700 - accuracy: 0.4324 - val_loss: 3.5330 - val_accuracy: 0.4581\n",
      "Epoch 3/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.5552 - accuracy: 0.4497 - val_loss: 3.4251 - val_accuracy: 0.4653\n",
      "Epoch 4/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.4741 - accuracy: 0.4548 - val_loss: 3.3694 - val_accuracy: 0.4688\n",
      "Epoch 5/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.4126 - accuracy: 0.4590 - val_loss: 3.3181 - val_accuracy: 0.4731\n",
      "Epoch 6/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.3629 - accuracy: 0.4626 - val_loss: 3.2863 - val_accuracy: 0.4759\n",
      "Epoch 7/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.3275 - accuracy: 0.4652 - val_loss: 3.2622 - val_accuracy: 0.4775\n",
      "Epoch 8/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.2999 - accuracy: 0.4673 - val_loss: 3.2428 - val_accuracy: 0.4794\n",
      "Epoch 9/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.2777 - accuracy: 0.4691 - val_loss: 3.2276 - val_accuracy: 0.4806\n",
      "Epoch 10/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.2595 - accuracy: 0.4705 - val_loss: 3.2167 - val_accuracy: 0.4819\n",
      "Epoch 11/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.2444 - accuracy: 0.4717 - val_loss: 3.2070 - val_accuracy: 0.4829\n",
      "Epoch 12/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.2316 - accuracy: 0.4727 - val_loss: 3.1974 - val_accuracy: 0.4837\n",
      "Epoch 13/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.2209 - accuracy: 0.4735 - val_loss: 3.1926 - val_accuracy: 0.4842\n",
      "Epoch 14/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.2117 - accuracy: 0.4742 - val_loss: 3.1876 - val_accuracy: 0.4847\n",
      "Epoch 15/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.2037 - accuracy: 0.4749 - val_loss: 3.1831 - val_accuracy: 0.4849\n",
      "Epoch 16/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1966 - accuracy: 0.4754 - val_loss: 3.1787 - val_accuracy: 0.4857\n",
      "Epoch 17/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1903 - accuracy: 0.4759 - val_loss: 3.1765 - val_accuracy: 0.4856\n",
      "Epoch 18/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1848 - accuracy: 0.4763 - val_loss: 3.1729 - val_accuracy: 0.4863\n",
      "Epoch 19/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1797 - accuracy: 0.4767 - val_loss: 3.1695 - val_accuracy: 0.4863\n",
      "Epoch 20/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1750 - accuracy: 0.4771 - val_loss: 3.1682 - val_accuracy: 0.4865\n",
      "Epoch 21/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1710 - accuracy: 0.4775 - val_loss: 3.1640 - val_accuracy: 0.4873\n",
      "Epoch 22/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1671 - accuracy: 0.4778 - val_loss: 3.1641 - val_accuracy: 0.4870\n",
      "Epoch 23/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1635 - accuracy: 0.4780 - val_loss: 3.1616 - val_accuracy: 0.4873\n",
      "Epoch 24/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1604 - accuracy: 0.4784 - val_loss: 3.1596 - val_accuracy: 0.4874\n",
      "Epoch 25/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1574 - accuracy: 0.4786 - val_loss: 3.1608 - val_accuracy: 0.4874\n",
      "Epoch 26/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1547 - accuracy: 0.4788 - val_loss: 3.1581 - val_accuracy: 0.4877\n",
      "Epoch 27/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1518 - accuracy: 0.4791 - val_loss: 3.1563 - val_accuracy: 0.4877\n",
      "Epoch 28/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1496 - accuracy: 0.4792 - val_loss: 3.1557 - val_accuracy: 0.4880\n",
      "Epoch 29/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1475 - accuracy: 0.4794 - val_loss: 3.1571 - val_accuracy: 0.4881\n",
      "Epoch 30/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1451 - accuracy: 0.4796 - val_loss: 3.1537 - val_accuracy: 0.4882\n",
      "Epoch 31/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1431 - accuracy: 0.4798 - val_loss: 3.1527 - val_accuracy: 0.4881\n",
      "Epoch 32/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1412 - accuracy: 0.4799 - val_loss: 3.1525 - val_accuracy: 0.4884\n",
      "Epoch 33/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1394 - accuracy: 0.4801 - val_loss: 3.1519 - val_accuracy: 0.4884\n",
      "Epoch 34/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1378 - accuracy: 0.4802 - val_loss: 3.1488 - val_accuracy: 0.4887\n",
      "Epoch 35/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1363 - accuracy: 0.4803 - val_loss: 3.1504 - val_accuracy: 0.4888\n",
      "Epoch 36/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1346 - accuracy: 0.4804 - val_loss: 3.1484 - val_accuracy: 0.4888\n",
      "Epoch 37/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1333 - accuracy: 0.4805 - val_loss: 3.1477 - val_accuracy: 0.4886\n",
      "Epoch 38/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1317 - accuracy: 0.4806 - val_loss: 3.1493 - val_accuracy: 0.4886\n",
      "Epoch 39/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1303 - accuracy: 0.4808 - val_loss: 3.1476 - val_accuracy: 0.4887\n",
      "Epoch 40/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1291 - accuracy: 0.4808 - val_loss: 3.1489 - val_accuracy: 0.4888\n",
      "Epoch 41/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1278 - accuracy: 0.4810 - val_loss: 3.1501 - val_accuracy: 0.4889\n",
      "Epoch 42/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1267 - accuracy: 0.4810 - val_loss: 3.1465 - val_accuracy: 0.4889\n",
      "Epoch 43/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1255 - accuracy: 0.4812 - val_loss: 3.1443 - val_accuracy: 0.4889\n",
      "Epoch 44/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1243 - accuracy: 0.4812 - val_loss: 3.1465 - val_accuracy: 0.4894\n",
      "Epoch 45/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1234 - accuracy: 0.4813 - val_loss: 3.1458 - val_accuracy: 0.4892\n",
      "Epoch 46/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1221 - accuracy: 0.4815 - val_loss: 3.1429 - val_accuracy: 0.4888\n",
      "Epoch 47/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1213 - accuracy: 0.4815 - val_loss: 3.1448 - val_accuracy: 0.4894\n",
      "Epoch 48/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1204 - accuracy: 0.4816 - val_loss: 3.1431 - val_accuracy: 0.4893\n",
      "Epoch 49/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1195 - accuracy: 0.4816 - val_loss: 3.1418 - val_accuracy: 0.4892\n",
      "Epoch 50/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1185 - accuracy: 0.4817 - val_loss: 3.1436 - val_accuracy: 0.4893\n",
      "Epoch 51/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1174 - accuracy: 0.4819 - val_loss: 3.1430 - val_accuracy: 0.4896\n",
      "Epoch 52/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1167 - accuracy: 0.4819 - val_loss: 3.1439 - val_accuracy: 0.4893\n",
      "Epoch 53/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1160 - accuracy: 0.4820 - val_loss: 3.1418 - val_accuracy: 0.4895\n",
      "Epoch 54/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1153 - accuracy: 0.4820 - val_loss: 3.1407 - val_accuracy: 0.4897\n",
      "Epoch 55/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1144 - accuracy: 0.4820 - val_loss: 3.1396 - val_accuracy: 0.4896\n",
      "Epoch 56/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1137 - accuracy: 0.4821 - val_loss: 3.1442 - val_accuracy: 0.4894\n",
      "Epoch 57/300\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1131 - accuracy: 0.4821 - val_loss: 3.1406 - val_accuracy: 0.4897\n",
      "Epoch 58/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1123 - accuracy: 0.4822 - val_loss: 3.1411 - val_accuracy: 0.4897\n",
      "Epoch 59/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1116 - accuracy: 0.4823 - val_loss: 3.1404 - val_accuracy: 0.4896\n",
      "Epoch 60/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1109 - accuracy: 0.4823 - val_loss: 3.1393 - val_accuracy: 0.4896\n",
      "Epoch 61/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1103 - accuracy: 0.4824 - val_loss: 3.1387 - val_accuracy: 0.4894\n",
      "Epoch 62/300\n",
      "782/782 [==============================] - 37s 48ms/step - loss: 3.1097 - accuracy: 0.4824 - val_loss: 3.1400 - val_accuracy: 0.4897\n",
      "Epoch 63/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1091 - accuracy: 0.4825 - val_loss: 3.1419 - val_accuracy: 0.4897\n",
      "Epoch 64/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1085 - accuracy: 0.4825 - val_loss: 3.1382 - val_accuracy: 0.4899\n",
      "Epoch 65/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1079 - accuracy: 0.4825 - val_loss: 3.1381 - val_accuracy: 0.4898\n",
      "Epoch 66/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1074 - accuracy: 0.4826 - val_loss: 3.1403 - val_accuracy: 0.4900\n",
      "Epoch 67/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1068 - accuracy: 0.4827 - val_loss: 3.1402 - val_accuracy: 0.4899\n",
      "Epoch 68/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1064 - accuracy: 0.4827 - val_loss: 3.1387 - val_accuracy: 0.4899\n",
      "Epoch 69/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1056 - accuracy: 0.4828 - val_loss: 3.1388 - val_accuracy: 0.4900\n",
      "Epoch 70/300\n",
      "782/782 [==============================] - 37s 47ms/step - loss: 3.1053 - accuracy: 0.4828 - val_loss: 3.1384 - val_accuracy: 0.4903\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<tensorflow.python.keras.callbacks.History at 0x7f1e2c8c8510>"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mleG.fit(\n",
    "    decoder_train, \n",
    "    teacher_train, \n",
    "    batch_size=32, \n",
    "    epochs=300, \n",
    "    shuffle=True, \n",
    "    validation_data = (decoder_vali, teacher_vali), \n",
    "    callbacks=[\n",
    "        ModelCheckpoint(\n",
    "            f'./{folder_name}/mleG.h5', \n",
    "            save_best_only=True, \n",
    "            monitor = \"val_loss\"\n",
    "        ),\n",
    "        EarlyStopping(monitor='val_loss', patience=5),\n",
    "        CSVLogger(f'{folder_name}/mleG.csv'),\n",
    "    ]\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "79/79 [==============================] - 2s 32ms/step - loss: 3.1381 - accuracy: 0.4898\n",
      "[3.1381313800811768, 0.48980000615119934]\n"
     ]
    }
   ],
   "source": [
    "mleG.load_weights(f'./{folder_name}/mleG.h5')\n",
    "loss = mleG.evaluate(decoder_vali, teacher_vali)\n",
    "print(loss)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "No handles with labels found to put in legend.\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1800x1200 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import csv\n",
    "from matplotlib import pyplot as plt\n",
    "%matplotlib inline\n",
    "rows = csv.reader(open(f'{folder_name}/mleG.csv'))\n",
    "nll_mle = []\n",
    "for i, row in enumerate(rows):\n",
    "    if i == 0:\n",
    "        continue\n",
    "    nll_mle.append(float(row[-2]))\n",
    "plt.figure(dpi = 300)\n",
    "plt.plot(np.arange(1,len(nll_mle)+1), nll_mle)\n",
    "plt.grid(True)\n",
    "plt.ylabel(r'accuracy')\n",
    "plt.xlabel('Training Iterations (37s per iters)')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "np.savetxt(f'{folder_name}/mleG_nll_list.txt', nll_mle)"
   ]
  },
  {
   "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": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}