{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "Igc5itf-xMGj"
   },
   "source": [
    "# Masakhane - Machine Translation for African Languages (Using JoeyNMT)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "x4fXCKCf36IK"
   },
   "source": [
    "## Note before beginning:\n",
    "### - The idea is that you should be able to make minimal changes to this in order to get SOME result for your own translation corpus. \n",
    "\n",
    "### - The tl;dr: Go to the **\"TODO\"** comments which will tell you what to update to get up and running\n",
    "\n",
    "### - If you actually want to have a clue what you're doing, read the text and peek at the links\n",
    "\n",
    "### - With 100 epochs, it should take around 7 hours to run in Google Colab\n",
    "\n",
    "### - Once you've gotten a result for your language, please attach and email your notebook that generated it to masakhanetranslation@gmail.com\n",
    "\n",
    "### - If you care enough and get a chance, doing a brief background on your language would be amazing. See examples in  [(Martinus, 2019)](https://arxiv.org/abs/1906.05685)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "l929HimrxS0a"
   },
   "source": [
    "## Retrieve your data & make a parallel corpus\n",
    "\n",
    "If you are wanting to use the JW300 data referenced on the Masakhane website or in our GitHub repo, you can use `opus-tools` to convert the data into a convenient format. `opus_read` from that package provides a convenient tool for reading the native aligned XML files and to convert them to TMX format. The tool can also be used to fetch relevant files from OPUS on the fly and to filter the data as necessary. [Read the documentation](https://pypi.org/project/opustools-pkg/) for more details.\n",
    "\n",
    "Once you have your corpus files in TMX format (an xml structure which will include the sentences in your target language and your source language in a single file), we recommend reading them into a pandas dataframe. Thankfully, Jade wrote a silly `tmx2dataframe` package which converts your tmx file to a pandas dataframe. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "oGRmDELn7Az0"
   },
   "outputs": [],
   "source": [
    "# from google.colab import drive\n",
    "# drive.mount('/content/drive')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "Cn3tgQLzUxwn"
   },
   "outputs": [],
   "source": [
    "# TODO: Set your source and target languages. Keep in mind, these traditionally use language codes as found here:\n",
    "# These will also become the suffix's of all vocab and corpus files used throughout\n",
    "import os\n",
    "source_language = \"en\"\n",
    "target_language = \"ki\" \n",
    "lc = False  # If True, lowercase the data.\n",
    "seed = 42  # Random seed for shuffling.\n",
    "tag = \"baseline\" # Give a unique name to your folder - this is to ensure you don't rewrite any models you've already submitted\n",
    "\n",
    "os.environ[\"src\"] = source_language # Sets them in bash as well, since we often use bash scripts\n",
    "os.environ[\"tgt\"] = target_language\n",
    "os.environ[\"tag\"] = tag\n",
    "\n",
    "# This will save it to a folder in our gdrive instead!\n",
    "# !mkdir -p \"/content/drive/My Drive/masakhane/$src-$tgt-$tag\"\n",
    "# os.environ[\"gdrive_path\"] = \"/content/drive/My Drive/masakhane/%s-%s-%s\" % (source_language, target_language, tag)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "kBSgJHEw7Nvx"
   },
   "outputs": [],
   "source": [
    "# !echo $gdrive_path"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "gA75Fs9ys8Y9"
   },
   "outputs": [],
   "source": [
    "# Install opus-tools\n",
    "#! pip install opustools-pkg"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Uncomment cell below if notebook is being run for the first time and you need to download the data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "xq-tDZVks7ZD"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Alignment file /proj/nlpl/data/OPUS/JW300/latest/xml/en-ki.xml.gz not found. The following files are available for downloading:\n",
      "\n",
      " 964 KB https://object.pouta.csc.fi/OPUS-JW300/v1/xml/en-ki.xml.gz\n",
      " 263 MB https://object.pouta.csc.fi/OPUS-JW300/v1/xml/en.zip\n",
      "  10 MB https://object.pouta.csc.fi/OPUS-JW300/v1/xml/ki.zip\n",
      "\n",
      " 274 MB Total size\n",
      "./JW300_latest_xml_en-ki.xml.gz ... 100% of 964 KB\n",
      "./JW300_latest_xml_en.zip ... 100% of 263 MBn.zip ... 96% of 263 MB\n",
      "./JW300_latest_xml_ki.zip ... 100% of 10 MB\n"
     ]
    }
   ],
   "source": [
    "# Downloading our corpus\n",
    "! opus_read -d JW300 -s $src -t $tgt -wm moses -w jw300.$src jw300.$tgt -q\n",
    "\n",
    "# extract the corpus file\n",
    "! gunzip JW300_latest_xml_$src-$tgt.xml.gz"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "n48GDRnP8y2G"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "--2020-02-19 05:55:36--  https://raw.githubusercontent.com/juliakreutzer/masakhane/master/jw300_utils/test/test.en-any.en\n",
      "Resolving raw.githubusercontent.com (raw.githubusercontent.com)... 151.101.192.133, 151.101.128.133, 151.101.64.133, ...\n",
      "Connecting to raw.githubusercontent.com (raw.githubusercontent.com)|151.101.192.133|:443... connected.\n",
      "HTTP request sent, awaiting response... 200 OK\n",
      "Length: 277791 (271K) [text/plain]\n",
      "Saving to: ‘test.en-any.en’\n",
      "\n",
      "test.en-any.en      100%[===================>] 271.28K  --.-KB/s    in 0.04s   \n",
      "\n",
      "2020-02-19 05:55:36 (6.56 MB/s) - ‘test.en-any.en’ saved [277791/277791]\n",
      "\n",
      "--2020-02-19 05:55:36--  https://raw.githubusercontent.com/juliakreutzer/masakhane/master/jw300_utils/test/test.en-ki.en\n",
      "Resolving raw.githubusercontent.com (raw.githubusercontent.com)... 151.101.192.133, 151.101.128.133, 151.101.64.133, ...\n",
      "Connecting to raw.githubusercontent.com (raw.githubusercontent.com)|151.101.192.133|:443... connected.\n",
      "HTTP request sent, awaiting response... 200 OK\n",
      "Length: 204462 (200K) [text/plain]\n",
      "Saving to: ‘test.en-ki.en’\n",
      "\n",
      "test.en-ki.en       100%[===================>] 199.67K  --.-KB/s    in 0.04s   \n",
      "\n",
      "2020-02-19 05:55:36 (4.69 MB/s) - ‘test.en-ki.en’ saved [204462/204462]\n",
      "\n",
      "--2020-02-19 05:55:37--  https://raw.githubusercontent.com/juliakreutzer/masakhane/master/jw300_utils/test/test.en-ki.ki\n",
      "Resolving raw.githubusercontent.com (raw.githubusercontent.com)... 151.101.192.133, 151.101.128.133, 151.101.64.133, ...\n",
      "Connecting to raw.githubusercontent.com (raw.githubusercontent.com)|151.101.192.133|:443... connected.\n",
      "HTTP request sent, awaiting response... 200 OK\n",
      "Length: 245872 (240K) [text/plain]\n",
      "Saving to: ‘test.en-ki.ki’\n",
      "\n",
      "test.en-ki.ki       100%[===================>] 240.11K  --.-KB/s    in 0.04s   \n",
      "\n",
      "2020-02-19 05:55:37 (5.23 MB/s) - ‘test.en-ki.ki’ saved [245872/245872]\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Download the global test set.\n",
    "! wget https://raw.githubusercontent.com/juliakreutzer/masakhane/master/jw300_utils/test/test.en-any.en\n",
    "  \n",
    "# And the specific test set for this language pair.\n",
    "os.environ[\"trg\"] = target_language \n",
    "os.environ[\"src\"] = source_language \n",
    "\n",
    "! wget https://raw.githubusercontent.com/juliakreutzer/masakhane/master/jw300_utils/test/test.en-$trg.en \n",
    "! mv test.en-$trg.en test.en\n",
    "! wget https://raw.githubusercontent.com/juliakreutzer/masakhane/master/jw300_utils/test/test.en-$trg.$trg \n",
    "! mv test.en-$trg.$trg test.$trg"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "NqDG-CI28y2L"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loaded 3571 global test sentences to filter from the training/dev data.\n"
     ]
    }
   ],
   "source": [
    "# Read the test data to filter from train and dev splits.\n",
    "# Store english portion in set for quick filtering checks.\n",
    "en_test_sents = set()\n",
    "filter_test_sents = \"test.en-any.en\"\n",
    "j = 0\n",
    "with open(filter_test_sents) as f:\n",
    "  for line in f:\n",
    "    en_test_sents.add(line.strip())\n",
    "    j += 1\n",
    "print('Loaded {} global test sentences to filter from the training/dev data.'.format(j))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "3CNdwLBCfSIl"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loaded data and skipped 4277/106699 lines since contained in test set.\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>source_sentence</th>\n",
       "      <th>target_sentence</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>© 2015 Watch Tower Bible and Tract Society of ...</td>\n",
       "      <td>© 2015 Watch Tower Bible and Tract Society of ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Two Questions Worth Asking</td>\n",
       "      <td>Ciũria Igĩrĩ cia Bata Tũngĩĩyũria</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>An Answer Worth Considering</td>\n",
       "      <td>Macokio ma Bata</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                     source_sentence  \\\n",
       "0  © 2015 Watch Tower Bible and Tract Society of ...   \n",
       "1                         Two Questions Worth Asking   \n",
       "2                        An Answer Worth Considering   \n",
       "\n",
       "                                     target_sentence  \n",
       "0  © 2015 Watch Tower Bible and Tract Society of ...  \n",
       "1                  Ciũria Igĩrĩ cia Bata Tũngĩĩyũria  \n",
       "2                                    Macokio ma Bata  "
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "\n",
    "# TMX file to dataframe\n",
    "source_file = 'jw300.' + source_language\n",
    "target_file = 'jw300.' + target_language\n",
    "\n",
    "source = []\n",
    "target = []\n",
    "skip_lines = []  # Collect the line numbers of the source portion to skip the same lines for the target portion.\n",
    "with open(source_file) as f:\n",
    "    for i, line in enumerate(f):\n",
    "        # Skip sentences that are contained in the test set.\n",
    "        if line.strip() not in en_test_sents:\n",
    "            source.append(line.strip())\n",
    "        else:\n",
    "            skip_lines.append(i)             \n",
    "with open(target_file) as f:\n",
    "    for j, line in enumerate(f):\n",
    "        # Only add to corpus if corresponding source was not skipped.\n",
    "        if j not in skip_lines:\n",
    "            target.append(line.strip())\n",
    "    \n",
    "print('Loaded data and skipped {}/{} lines since contained in test set.'.format(len(skip_lines), i))\n",
    "    \n",
    "df = pd.DataFrame(zip(source, target), columns=['source_sentence', 'target_sentence'])\n",
    "df.head(3)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "YkuK3B4p2AkN"
   },
   "source": [
    "## Pre-processing and export\n",
    "\n",
    "It is generally a good idea to remove duplicate translations and conflicting translations from the corpus. In practice, these public corpora include some number of these that need to be cleaned.\n",
    "\n",
    "In addition we will split our data into dev/test/train and export to the filesystem."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "M_2ouEOH1_1q"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.6/dist-packages/pandas/core/generic.py:6786: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "  self._update_inplace(new_data)\n",
      "/usr/local/lib/python3.6/dist-packages/ipykernel_launcher.py:8: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "  \n",
      "/usr/local/lib/python3.6/dist-packages/ipykernel_launcher.py:9: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "  if __name__ == '__main__':\n",
      "/usr/local/lib/python3.6/dist-packages/ipykernel_launcher.py:12: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "  if sys.path[0] == '':\n",
      "/usr/local/lib/python3.6/dist-packages/ipykernel_launcher.py:13: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "  del sys.path[0]\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "# drop duplicate translations\n",
    "df_pp = df.drop_duplicates()\n",
    "\n",
    "#drop empty lines (alp)\n",
    "df_pp['source_sentence'].replace('', np.nan, inplace=True)\n",
    "df_pp['target_sentence'].replace('', np.nan, inplace=True)\n",
    "df_pp.dropna(subset=['source_sentence'], inplace=True)\n",
    "df_pp.dropna(subset=['target_sentence'], inplace=True)\n",
    "\n",
    "# drop conflicting translations\n",
    "df_pp.drop_duplicates(subset='source_sentence', inplace=True)\n",
    "df_pp.drop_duplicates(subset='target_sentence', inplace=True)\n",
    "\n",
    "# Shuffle the data to remove bias in dev set selection.\n",
    "df_pp = df_pp.sample(frac=1, random_state=seed).reset_index(drop=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "hxxBOCA-xXhy"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "==> train.bpe.en <==\n",
      "These are often d@@ ue to the bo@@ d@@ y’s dir@@ ecting blood to the cor@@ e to support v@@ ital org@@ ans .\n",
      "The power to per@@ form fe@@ at@@ s like that could come only from God’s enem@@ ies in the spirit real@@ m . ​ — Ex@@ od@@ us 7 : 8 - 12 .\n",
      "In fact , that is what he promised to do for anointed ones in the new covenant : “ I will put my law with@@ in them , and in their heart I sha@@ ll wr@@ ite it . And I will become their God , and they themselves will become my people . ”\n",
      "Instead , Sa@@ ul ref@@ used to be mo@@ ld@@ ed .\n",
      "In@@ clud@@ ed among these were sever@@ al of Jesus ’ family members , who at first had not taken ser@@ i@@ ously the pos@@ sib@@ ility that one of their relati@@ ves could be the Mess@@ iah .\n",
      "B@@ enj@@ am@@ in , ag@@ ed 33 , is An@@ n@@ e - R@@ ak@@ el@@ ’s brother .\n",
      "Th@@ erefore , let us consider how we can over@@ come ob@@ st@@ ac@@ les that make it challeng@@ ing to listen to God’s vo@@ ice .\n",
      "You will have increas@@ ed satis@@ fac@@ tion that comes from p@@ ut@@ ting Kingdom inter@@ ests ahe@@ ad of your own .\n",
      "L@@ ater he est@@ ab@@ l@@ ished the Israelites in the Pro@@ mis@@ ed L@@ and .\n",
      "But there is one kind of love that we need more than any other ​ — Jehovah’s love .\n",
      "\n",
      "==> train.bpe.ki <==\n",
      "Ũndũ ũcio wĩk@@ ĩkaga nĩ ũndũ wa mwĩrĩ gũ@@ tw@@ ara thakame harĩ ciĩ@@ ga iria cia bata nĩguo ith@@ iĩ na mbere kũruta wĩra .\n",
      "Ũ@@ hot@@ i wa gwĩka maũndũ ta macio ũngĩ@@ oim@@ ire tu harĩ thũ cia Ngai iria ci@@ k@@ araga ci@@ ik@@ aro - inĩ cia roho . — Tha@@ ma 7 : 8 - 12 .\n",
      "Hatarĩ nganja , ũndũ ũcio nĩguo er@@ ĩire aitĩrĩrie maguta arĩa marĩ kĩr@@ ĩkanĩro - inĩ kĩrĩa kĩ@@ erũ : “ Nĩ gwĩkĩra ng@@ ek@@ ĩra watho wakwa ngoro - inĩ ciao thĩinĩ , o na ndĩ@@ w@@ andĩk@@ e meciria - inĩ mao ; na nd@@ u@@ ĩke Ngai wao , nao matu@@ ĩke andũ akwa . ”\n",
      "Ithenya rĩa ũguo , Sa@@ ul@@ u nde@@ etĩkĩrire .\n",
      "A@@ mwe ao maarĩ a famĩlĩ ya Jesu , arĩa mbere - inĩ mat@@ o@@ onaga ta mũndũ ũmwe met@@ ain@@ wo angĩ@@ atuĩkire Mesia .\n",
      "B@@ enj@@ am@@ in ũrĩa ũrĩ na mĩaka 33 , nĩ mũrũ wa nyina na An@@ n@@ e - R@@ ak@@ el .\n",
      "( Je@@ r . 17 : 9 ) Nĩ ũndũ ũcio , rekei twarĩrĩrie ũrĩa tũngĩ@@ hiũrania na mĩh@@ ĩ@@ ng@@ a ĩrĩa ĩ@@ ngĩt@@ ũma tw@@ age gũth@@ ikĩrĩria mũg@@ amb@@ o wa Ngai .\n",
      "Nĩ ũ@@ kũgĩa na kũ@@ igan@@ ĩra kũ@@ nene kũrĩa kuu@@ managa na kũiga maũndũ ma Ũthamaki mbere .\n",
      "( Tha@@ m . 1 : 15 - 20 ; 14 : 13 ) Thutha ũcio agĩt@@ ũma Aisiraeli maga@@ ac@@ ĩre B@@ ũrũri - inĩ wa Kĩ@@ ĩranĩro .\n",
      "Ĩndĩ wendo ũrĩa tũ@@ bataraga makĩria nĩ wa Jehova .\n",
      "\n",
      "==> train.en <==\n",
      "These are often due to the body’s directing blood to the core to support vital organs .\n",
      "The power to perform feats like that could come only from God’s enemies in the spirit realm . ​ — Exodus 7 : 8 - 12 .\n",
      "In fact , that is what he promised to do for anointed ones in the new covenant : “ I will put my law within them , and in their heart I shall write it . And I will become their God , and they themselves will become my people . ”\n",
      "Instead , Saul refused to be molded .\n",
      "Included among these were several of Jesus ’ family members , who at first had not taken seriously the possibility that one of their relatives could be the Messiah .\n",
      "Benjamin , aged 33 , is Anne - Rakel’s brother .\n",
      "Therefore , let us consider how we can overcome obstacles that make it challenging to listen to God’s voice .\n",
      "You will have increased satisfaction that comes from putting Kingdom interests ahead of your own .\n",
      "Later he established the Israelites in the Promised Land .\n",
      "But there is one kind of love that we need more than any other ​ — Jehovah’s love .\n",
      "\n",
      "==> train.ki <==\n",
      "Ũndũ ũcio wĩkĩkaga nĩ ũndũ wa mwĩrĩ gũtwara thakame harĩ ciĩga iria cia bata nĩguo ithiĩ na mbere kũruta wĩra .\n",
      "Ũhoti wa gwĩka maũndũ ta macio ũngĩoimire tu harĩ thũ cia Ngai iria cikaraga ciikaro - inĩ cia roho . — Thama 7 : 8 - 12 .\n",
      "Hatarĩ nganja , ũndũ ũcio nĩguo erĩire aitĩrĩrie maguta arĩa marĩ kĩrĩkanĩro - inĩ kĩrĩa kĩerũ : “ Nĩ gwĩkĩra ngekĩra watho wakwa ngoro - inĩ ciao thĩinĩ , o na ndĩwandĩke meciria - inĩ mao ; na nduĩke Ngai wao , nao matuĩke andũ akwa . ”\n",
      "Ithenya rĩa ũguo , Saulu ndeetĩkĩrire .\n",
      "Amwe ao maarĩ a famĩlĩ ya Jesu , arĩa mbere - inĩ matoonaga ta mũndũ ũmwe metainwo angĩatuĩkire Mesia .\n",
      "Benjamin ũrĩa ũrĩ na mĩaka 33 , nĩ mũrũ wa nyina na Anne - Rakel .\n",
      "( Jer . 17 : 9 ) Nĩ ũndũ ũcio , rekei twarĩrĩrie ũrĩa tũngĩhiũrania na mĩhĩnga ĩrĩa ĩngĩtũma twage gũthikĩrĩria mũgambo wa Ngai .\n",
      "Nĩ ũkũgĩa na kũiganĩra kũnene kũrĩa kuumanaga na kũiga maũndũ ma Ũthamaki mbere .\n",
      "( Tham . 1 : 15 - 20 ; 14 : 13 ) Thutha ũcio agĩtũma Aisiraeli magaacĩre Bũrũri - inĩ wa Kĩĩranĩro .\n",
      "Ĩndĩ wendo ũrĩa tũbataraga makĩria nĩ wa Jehova .\n",
      "==> dev.bpe.en <==\n",
      "R@@ es@@ ear@@ ch@@ ers at the Un@@ iver@@ s@@ ity of C@@ op@@ en@@ h@@ ag@@ en tr@@ ac@@ ked al@@ most 10@@ ,000 m@@ id@@ d@@ le - ag@@ ed people over an 11 - year peri@@ od and found that par@@ ti@@ ci@@ p@@ ants who f@@ requ@@ ently ar@@ gu@@ ed with someone close to them were far more likely to die pre@@ matu@@ rely than those who sel@@ dom had con@@ fl@@ ic@@ ts .\n",
      "Yes .\n",
      "He enjo@@ yed special in@@ ti@@ ma@@ cy with Jehovah , who en@@ ab@@ led him to dis@@ play “ aw@@ es@@ ome power ” as he led the Israelites to the Pro@@ mis@@ ed L@@ and . ​ — De@@ ut .\n",
      "HI@@ S@@ T@@ OR@@ Y : A@@ TH@@ E@@ I@@ S@@ T\n",
      "H@@ ere is another val@@ ue that seem@@ s to be on the wan@@ e today ​ — and another reason to imitate the faith of this kind@@ hear@@ ted young woman .\n",
      "They knew that sheep need care and attention in order to th@@ ri@@ ve .\n",
      "What four hor@@ se@@ men have been on the move since 1914 ?\n",
      "J@@ on : No , I d@@ on@@ ’t rec@@ all .\n",
      "s@@ hu@@ n pri@@ de ?\n",
      "L@@ ater , Joshua had the da@@ un@@ ting t@@ ask of s@@ et@@ t@@ ling God’s people in the Pro@@ mis@@ ed L@@ and .\n",
      "\n",
      "==> dev.bpe.ki <==\n",
      "A@@ thuth@@ uria y@@ un@@ iv@@ as@@ ĩt@@ ĩ - inĩ ya C@@ op@@ en@@ h@@ ag@@ en , nĩ maath@@ uth@@ ur@@ irie andũ hakuhĩ 10@@ ,000 a mĩaka ĩyo kwa ihinda rĩa mĩaka 11 na makĩ@@ ona atĩ arĩa ma@@ akar@@ ar@@ anagia na mũndũ wa famĩlĩ kaingĩ , ha@@ akoragwo na ũhot@@ ek@@ ek@@ u mũnene atĩ me@@ gũkua tene gũkĩra arĩa mat@@ a@@ akoragwo na ng@@ ar@@ ari kaingĩ .\n",
      "Ĩ@@ ĩ .\n",
      "Aarĩ na ũkuruhanu wa hakuhĩ na Jehova , na nĩ aamũ@@ he@@ ire ũhoti wa gwĩka “ maũndũ manene ma kũ@@ gu@@ o@@ yo@@ h@@ ania ” rĩrĩa aat@@ ongoragia Aisiraeli ma@@ thiĩ B@@ ũrũri wa Kĩ@@ ĩranĩro . — Gũ@@ c@@ ok .\n",
      "HI@@ S@@ T@@ OR@@ Ĩ : E@@ E@@ T@@ Ĩ@@ K@@ Ĩ@@ TI@@ E G@@ Ũ@@ TI@@ R@@ Ĩ NG@@ A@@ I\n",
      "Ĩ@@ yo nĩ ngumo ĩngĩ y@@ ag@@ ĩte mũno ũmũthĩ , na kĩu nĩ gĩtũmi kĩngĩ gĩa kwĩg@@ erekania na wĩtĩkio wa mũ@@ irĩtu ũcio warĩ mũtu@@ gi .\n",
      "Nĩ ma@@ amenyaga atĩ , nĩguo ng’ondu ith@@ er@@ em@@ e nĩ ib@@ at@@ araga kũr@@ ũmbũ@@ i@@ yo mũno .\n",
      "Nĩ atwar@@ ithia ar@@ ĩkũ a mb@@ ar@@ athi mak@@ oretwo makĩ@@ oneka kuuma 1914 ?\n",
      "John : Ac@@ a , nd@@ ir@@ ar@@ irikana .\n",
      "mwĩ@@ g@@ atho ?\n",
      "Thutha ũcio , Joshua nĩ aarĩ na wĩra mũritũ wa kũ@@ ga@@ ĩra andũ B@@ ũrũri wa Kĩ@@ ĩranĩro .\n",
      "\n",
      "==> dev.en <==\n",
      "Researchers at the University of Copenhagen tracked almost 10,000 middle - aged people over an 11 - year period and found that participants who frequently argued with someone close to them were far more likely to die prematurely than those who seldom had conflicts .\n",
      "Yes .\n",
      "He enjoyed special intimacy with Jehovah , who enabled him to display “ awesome power ” as he led the Israelites to the Promised Land . ​ — Deut .\n",
      "HISTORY : ATHEIST\n",
      "Here is another value that seems to be on the wane today ​ — and another reason to imitate the faith of this kindhearted young woman .\n",
      "They knew that sheep need care and attention in order to thrive .\n",
      "What four horsemen have been on the move since 1914 ?\n",
      "Jon : No , I don’t recall .\n",
      "shun pride ?\n",
      "Later , Joshua had the daunting task of settling God’s people in the Promised Land .\n",
      "\n",
      "==> dev.ki <==\n",
      "Athuthuria yunivasĩtĩ - inĩ ya Copenhagen , nĩ maathuthuririe andũ hakuhĩ 10,000 a mĩaka ĩyo kwa ihinda rĩa mĩaka 11 na makĩona atĩ arĩa maakararanagia na mũndũ wa famĩlĩ kaingĩ , haakoragwo na ũhotekeku mũnene atĩ megũkua tene gũkĩra arĩa mataakoragwo na ngarari kaingĩ .\n",
      "Ĩĩ .\n",
      "Aarĩ na ũkuruhanu wa hakuhĩ na Jehova , na nĩ aamũheire ũhoti wa gwĩka “ maũndũ manene ma kũguoyohania ” rĩrĩa aatongoragia Aisiraeli mathiĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "HISTORĨ : EETĨKĨTIE GŨTIRĨ NGAI\n",
      "Ĩyo nĩ ngumo ĩngĩ yagĩte mũno ũmũthĩ , na kĩu nĩ gĩtũmi kĩngĩ gĩa kwĩgerekania na wĩtĩkio wa mũirĩtu ũcio warĩ mũtugi .\n",
      "Nĩ maamenyaga atĩ , nĩguo ng’ondu ithereme nĩ ibataraga kũrũmbũiyo mũno .\n",
      "Nĩ atwarithia arĩkũ a mbarathi makoretwo makĩoneka kuuma 1914 ?\n",
      "John : Aca , ndiraririkana .\n",
      "mwĩgatho ?\n",
      "Thutha ũcio , Joshua nĩ aarĩ na wĩra mũritũ wa kũgaĩra andũ Bũrũri wa Kĩĩranĩro .\n"
     ]
    }
   ],
   "source": [
    "# This section does the split between train/dev for the parallel corpora then saves them as separate files\n",
    "# We use 1000 dev test and the given test set.\n",
    "import csv\n",
    "\n",
    "# Do the split between dev/train and create parallel corpora\n",
    "num_dev_patterns = 1000\n",
    "\n",
    "# Optional: lower case the corpora - this will make it easier to generalize, but without proper casing.\n",
    "if lc:  # Julia: making lowercasing optional\n",
    "    df_pp[\"source_sentence\"] = df_pp[\"source_sentence\"].str.lower()\n",
    "    df_pp[\"target_sentence\"] = df_pp[\"target_sentence\"].str.lower()\n",
    "\n",
    "# Julia: test sets are already generated\n",
    "dev = df_pp.tail(num_dev_patterns) # Herman: Error in original\n",
    "stripped = df_pp.drop(df_pp.tail(num_dev_patterns).index)\n",
    "\n",
    "with open(\"train.\"+source_language, \"w\") as src_file, open(\"train.\"+target_language, \"w\") as trg_file:\n",
    "  for index, row in stripped.iterrows():\n",
    "    src_file.write(row[\"source_sentence\"]+\"\\n\")\n",
    "    trg_file.write(row[\"target_sentence\"]+\"\\n\")\n",
    "    \n",
    "with open(\"dev.\"+source_language, \"w\") as src_file, open(\"dev.\"+target_language, \"w\") as trg_file:\n",
    "  for index, row in dev.iterrows():\n",
    "    src_file.write(row[\"source_sentence\"]+\"\\n\")\n",
    "    trg_file.write(row[\"target_sentence\"]+\"\\n\")\n",
    "\n",
    "#stripped[[\"source_sentence\"]].to_csv(\"train.\"+source_language, header=False, index=False)  # Herman: Added `header=False` everywhere\n",
    "#stripped[[\"target_sentence\"]].to_csv(\"train.\"+target_language, header=False, index=False)  # Julia: Problematic handling of quotation marks.\n",
    "\n",
    "#dev[[\"source_sentence\"]].to_csv(\"dev.\"+source_language, header=False, index=False)\n",
    "#dev[[\"target_sentence\"]].to_csv(\"dev.\"+target_language, header=False, index=False)\n",
    "\n",
    "\n",
    "# Doublecheck the format below. There should be no extra quotation marks or weird characters.\n",
    "! head train.*\n",
    "! head dev.*"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "epeCydmCyS8X"
   },
   "source": [
    "\n",
    "\n",
    "---\n",
    "\n",
    "\n",
    "## Installation of JoeyNMT\n",
    "\n",
    "JoeyNMT is a simple, minimalist NMT package which is useful for learning and teaching. Check out the documentation for JoeyNMT [here](https://joeynmt.readthedocs.io)  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "iBRMm4kMxZ8L"
   },
   "outputs": [],
   "source": [
    "# Install JoeyNMT\n",
    "#! git clone https://github.com/joeynmt/joeynmt.git\n",
    "#! cd joeynmt; pip3 install ."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "AaE77Tcppex9"
   },
   "source": [
    "# Preprocessing the Data into Subword BPE Tokens\n",
    "\n",
    "- One of the most powerful improvements for agglutinative languages (a feature of most Bantu languages) is using BPE tokenization [ (Sennrich, 2015) ](https://arxiv.org/abs/1508.07909).\n",
    "\n",
    "- It was also shown that by optimizing the umber of BPE codes we significantly improve results for low-resourced languages [(Sennrich, 2019)](https://www.aclweb.org/anthology/P19-1021) [(Martinus, 2019)](https://arxiv.org/abs/1906.05685)\n",
    "\n",
    "- Below we have the scripts for doing BPE tokenization of our data. We use 4000 tokens as recommended by [(Sennrich, 2019)](https://www.aclweb.org/anthology/P19-1021). You do not need to change anything. Simply running the below will be suitable. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "# One of the huge boosts in NMT performance was to use a different method of tokenizing. \n",
    "# Usually, NMT would tokenize by words. However, using a method called BPE gave amazing boosts to performance\n",
    "\n",
    "# Do subword NMT\n",
    "from os import path\n",
    "os.environ[\"src\"] = source_language # Sets them in bash as well, since we often use bash scripts\n",
    "os.environ[\"tgt\"] = target_language\n",
    "\n",
    "# Learn BPEs on the training data.\n",
    "os.environ[\"data_path\"] = path.join(\"../../joeynmt\", \"data\", source_language + target_language) # Herman! \n",
    "! subword-nmt learn-joint-bpe-and-vocab --input train.$src train.$tgt -s 4000 -o bpe.codes.4000 --write-vocabulary vocab.$src vocab.$tgt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Apply BPE splits to the development and test data.\n",
    "! subword-nmt apply-bpe -c bpe.codes.4000 --vocabulary vocab.$src < train.$src > train.bpe.$src\n",
    "! subword-nmt apply-bpe -c bpe.codes.4000 --vocabulary vocab.$tgt < train.$tgt > train.bpe.$tgt\n",
    "\n",
    "! subword-nmt apply-bpe -c bpe.codes.4000 --vocabulary vocab.$src < dev.$src > dev.bpe.$src\n",
    "! subword-nmt apply-bpe -c bpe.codes.4000 --vocabulary vocab.$tgt < dev.$tgt > dev.bpe.$tgt\n",
    "! subword-nmt apply-bpe -c bpe.codes.4000 --vocabulary vocab.$src < test.$src > test.bpe.$src\n",
    "! subword-nmt apply-bpe -c bpe.codes.4000 --vocabulary vocab.$tgt < test.$tgt > test.bpe.$tgt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "! sudo chmod 777 ../../joeynmt/data/"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "bpe.codes.4000\tdev.en\t     test.bpe.ki     test.ki\t   train.en\r\n",
      "dev.bpe.en\tdev.ki\t     test.en\t     train.bpe.en  train.ki\r\n",
      "dev.bpe.ki\ttest.bpe.en  test.en-any.en  train.bpe.ki  vocab.txt\r\n"
     ]
    }
   ],
   "source": [
    "# Create directory, move everyone we care about to the correct location\n",
    "! mkdir -p $data_path\n",
    "! cp train.* $data_path\n",
    "! cp test.* $data_path\n",
    "! cp dev.* $data_path\n",
    "! cp bpe.codes.4000 $data_path\n",
    "! ls $data_path"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Also move everything we care about to a mounted location in google drive (relevant if running in colab) at gdrive_path\n",
    "# ! cp train.* \"$gdrive_path\"\n",
    "# ! cp test.* \"$gdrive_path\"\n",
    "# ! cp dev.* \"$gdrive_path\"\n",
    "# ! cp bpe.codes.4000 \"$gdrive_path\"\n",
    "# ! ls \"$gdrive_path\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Create that vocab using build_vocab\n",
    "! sudo chmod 777 ../../joeynmt/scripts/build_vocab.py\n",
    "! ../../joeynmt/scripts/build_vocab.py ../../joeynmt/data/$src$tgt/train.bpe.$src ../../joeynmt/data/$src$tgt/train.bpe.$tgt --output_path ../../joeynmt/data/$src$tgt/vocab.txt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "H-TyjtmXB1mL"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "BPE Kikuyu Sentences\n",
      "N@@ go ĩrĩa nene ya wĩtĩkio ( Rora kĩbungo gĩa 12 - 14 )\n",
      "N@@ gũ@@ b@@ ia ya ũh@@ onokio ( Rora kĩbungo gĩa 15 - 18 )\n",
      "Nĩ ny@@ on@@ ete atĩ andũ nĩ math@@ ik@@ agĩrĩria wega mangĩ@@ ona atĩ nĩ wendete Bibilia mũno na nĩ ũr@@ er@@ utanĩria kũm@@ ateith@@ ia . ”\n",
      "R@@ ũ@@ hi@@ ũ rwa roho ( Rora kĩbungo gĩa 19 - 20 )\n",
      "Kũg@@ erera ũteithio wa Jehova , no tũhote kwĩ@@ hand@@ a wega tw@@ ĩt@@ irie Shaitani , thũ itũ .\n",
      "Combined BPE Vocab\n",
      "Hih@@\n",
      "ʺ\n",
      "KARAT@@\n",
      "ē@@\n",
      "Ó@@\n",
      "greg@@\n",
      "artic@@\n",
      "✔\n",
      "ā@@\n",
      "iraeli\n"
     ]
    }
   ],
   "source": [
    "# Some output\n",
    "! echo \"BPE Kikuyu Sentences\"\n",
    "! tail -n 5 test.bpe.$tgt\n",
    "! echo \"Combined BPE Vocab\"\n",
    "! tail -n 10 ../../joeynmt/data/$src$tgt/vocab.txt  # Herman"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "IlMitUHR8Qy-"
   },
   "outputs": [],
   "source": [
    "# Also move everything we care about to a mounted location in google drive (relevant if running in colab) at gdrive_path\n",
    "#! cp train.* \"$gdrive_path\"\n",
    "#! cp test.* \"$gdrive_path\"\n",
    "#! cp dev.* \"$gdrive_path\"\n",
    "#! cp bpe.codes.4000 \"$gdrive_path\"\n",
    "#! ls \"$gdrive_path\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "Ixmzi60WsUZ8"
   },
   "source": [
    "# Creating the JoeyNMT Config\n",
    "\n",
    "JoeyNMT requires a yaml config. We provide a template below. We've also set a number of defaults with it, that you may play with!\n",
    "\n",
    "- We used Transformer architecture \n",
    "- We set our dropout to reasonably high: 0.3 (recommended in  [(Sennrich, 2019)](https://www.aclweb.org/anthology/P19-1021))\n",
    "\n",
    "Things worth playing with:\n",
    "- The batch size (also recommended to change for low-resourced languages)\n",
    "- The number of epochs (we've set it at 30 just so it runs in about an hour, for testing purposes)\n",
    "- The decoder options (beam_size, alpha)\n",
    "- Evaluation metrics (BLEU versus Crhf4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "PIs1lY2hxMsl"
   },
   "outputs": [],
   "source": [
    "# This creates the config file for our JoeyNMT system. It might seem overwhelming so we've provided a couple of useful parameters you'll need to update\n",
    "# (You can of course play with all the parameters if you'd like!)\n",
    "\n",
    "name = '%s%s' % (source_language, target_language)\n",
    "# gdrive_path = os.environ[\"gdrive_path\"]\n",
    "\n",
    "# Create the config\n",
    "config = \"\"\"\n",
    "name: \"{name}_transformer\"\n",
    "\n",
    "data:\n",
    "    src: \"{source_language}\"\n",
    "    trg: \"{target_language}\"\n",
    "    train: \"data/{name}/train.bpe\"\n",
    "    dev:   \"data/{name}/dev.bpe\"\n",
    "    test:  \"data/{name}/test.bpe\"\n",
    "    level: \"bpe\"\n",
    "    lowercase: False\n",
    "    max_sent_length: 100\n",
    "    src_vocab: \"data/{name}/vocab.txt\"\n",
    "    trg_vocab: \"data/{name}/vocab.txt\"\n",
    "\n",
    "testing:\n",
    "    beam_size: 5\n",
    "    alpha: 1.0\n",
    "\n",
    "training:\n",
    "    #load_model: \"models/{name}_transformer/1000.ckpt\" # if uncommented, load a pre-trained model from this checkpoint\n",
    "    random_seed: 42\n",
    "    optimizer: \"adam\"\n",
    "    normalization: \"tokens\"\n",
    "    adam_betas: [0.9, 0.999] \n",
    "    scheduling: \"plateau\"           # TODO: try switching from plateau to Noam scheduling\n",
    "    patience: 5                     # For plateau: decrease learning rate by decrease_factor if validation score has not improved for this many validation rounds.\n",
    "    learning_rate_factor: 0.5       # factor for Noam scheduler (used with Transformer)\n",
    "    learning_rate_warmup: 1000      # warmup steps for Noam scheduler (used with Transformer)\n",
    "    decrease_factor: 0.7\n",
    "    loss: \"crossentropy\"\n",
    "    learning_rate: 0.0003\n",
    "    learning_rate_min: 0.00000001\n",
    "    weight_decay: 0.0\n",
    "    label_smoothing: 0.1\n",
    "    batch_size: 4096\n",
    "    batch_type: \"token\"\n",
    "    eval_batch_size: 3600\n",
    "    eval_batch_type: \"token\"\n",
    "    batch_multiplier: 1\n",
    "    early_stopping_metric: \"ppl\"\n",
    "    epochs: 30                     # TODO: Decrease for when playing around and checking of working. Around 30 is sufficient to check if its working at all\n",
    "    validation_freq: 1000          # TODO: Set to at least once per epoch.\n",
    "    logging_freq: 100\n",
    "    eval_metric: \"bleu\"\n",
    "    model_dir: \"models/{name}_transformer\"\n",
    "    overwrite: False               # TODO: Set to True if you want to overwrite possibly existing models. \n",
    "    shuffle: True\n",
    "    use_cuda: True\n",
    "    max_output_length: 100\n",
    "    print_valid_sents: [0, 1, 2, 3]\n",
    "    keep_last_ckpts: 3\n",
    "\n",
    "model:\n",
    "    initializer: \"xavier\"\n",
    "    bias_initializer: \"zeros\"\n",
    "    init_gain: 1.0\n",
    "    embed_initializer: \"xavier\"\n",
    "    embed_init_gain: 1.0\n",
    "    tied_embeddings: True\n",
    "    tied_softmax: True\n",
    "    encoder:\n",
    "        type: \"transformer\"\n",
    "        num_layers: 6\n",
    "        num_heads: 4             # TODO: Increase to 8 for larger data.\n",
    "        embeddings:\n",
    "            embedding_dim: 256   # TODO: Increase to 512 for larger data.\n",
    "            scale: True\n",
    "            dropout: 0.2\n",
    "        # typically ff_size = 4 x hidden_size\n",
    "        hidden_size: 256         # TODO: Increase to 512 for larger data.\n",
    "        ff_size: 1024            # TODO: Increase to 2048 for larger data.\n",
    "        dropout: 0.3\n",
    "    decoder:\n",
    "        type: \"transformer\"\n",
    "        num_layers: 6\n",
    "        num_heads: 4              # TODO: Increase to 8 for larger data.\n",
    "        embeddings:\n",
    "            embedding_dim: 256    # TODO: Increase to 512 for larger data.\n",
    "            scale: True\n",
    "            dropout: 0.2\n",
    "        # typically ff_size = 4 x hidden_size\n",
    "        hidden_size: 256         # TODO: Increase to 512 for larger data.\n",
    "        ff_size: 1024            # TODO: Increase to 2048 for larger data.\n",
    "        dropout: 0.3\n",
    "\"\"\".format(name=name, gdrive_path=\"n/a\", source_language=source_language, target_language=target_language)\n",
    "with open(\"../../joeynmt/configs/transformer_{name}.yaml\".format(name=name),'w') as f:\n",
    "    f.write(config)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "pIifxE3Qzuvs"
   },
   "source": [
    "# Train the Model\n",
    "\n",
    "This single line of joeynmt runs the training using the config we made above"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "total 136\r\n",
      "drwxr-xr-x 11 root root  4096 Oct 25 12:22 .\r\n",
      "drwxrwxrwx 11 root root  4096 Feb 19 06:15 ..\r\n",
      "drwxr-xr-x  8 root root  4096 Oct 24 15:14 .git\r\n",
      "-rw-r--r--  1 root root    49 Oct 24 15:14 .gitattributes\r\n",
      "drwxr-xr-x  3 root root  4096 Oct 24 15:14 .github\r\n",
      "-rw-r--r--  1 root root    71 Oct 24 15:14 .gitignore\r\n",
      "-rw-r--r--  1 root root 13514 Oct 24 15:14 .pylintrc\r\n",
      "-rw-r--r--  1 root root   159 Oct 24 15:14 .readthedocs.yml\r\n",
      "-rw-r--r--  1 root root   542 Oct 24 15:14 .travis.yml\r\n",
      "-rwxrw-rwx  1 root root  3354 Oct 24 15:14 CODE_OF_CONDUCT.md\r\n",
      "-rwxrw-rwx  1 root root  1071 Oct 24 15:14 LICENSE\r\n",
      "-rwxrw-rwx  1 root root 13286 Oct 24 15:14 README.md\r\n",
      "-rwxrw-rwx  1 root root  8229 Oct 24 15:14 benchmarks.md\r\n",
      "drwxrw-rwx  3 root root  4096 Feb 19 06:30 configs\r\n",
      "drwxrwxrwx  8 root root  4096 Feb 19 06:35 data\r\n",
      "drwxrw-rwx  4 root root  4096 Oct 24 15:14 docs\r\n",
      "-rwxrw-rwx  1 root root 14373 Oct 24 15:14 joey-small.png\r\n",
      "drwxrw-rwx  3 root root  4096 Oct 24 16:35 joeynmt\r\n",
      "drwxrwxrwx  8 root root  4096 Feb 19 07:55 models\r\n",
      "-rwxrw-rwx  1 root root   167 Oct 24 15:14 requirements.txt\r\n",
      "drwxrw-rwx  2 root root  4096 Oct 24 15:14 scripts\r\n",
      "-rwxrw-rwx  1 root root   810 Oct 24 15:14 setup.py\r\n",
      "drwxrw-rwx  4 root root  4096 Oct 24 15:14 test\r\n"
     ]
    }
   ],
   "source": [
    "! ls -la ../../joeynmt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "! sudo chmod 777 ../../joeynmt/models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "6ZBPFwT94WpI"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "/home/espoir_mur_gmail_com/.local/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:541: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_qint8 = np.dtype([(\"qint8\", np.int8, 1)])\n",
      "/home/espoir_mur_gmail_com/.local/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:542: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_quint8 = np.dtype([(\"quint8\", np.uint8, 1)])\n",
      "/home/espoir_mur_gmail_com/.local/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:543: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_qint16 = np.dtype([(\"qint16\", np.int16, 1)])\n",
      "/home/espoir_mur_gmail_com/.local/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:544: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_quint16 = np.dtype([(\"quint16\", np.uint16, 1)])\n",
      "/home/espoir_mur_gmail_com/.local/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:545: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_qint32 = np.dtype([(\"qint32\", np.int32, 1)])\n",
      "/home/espoir_mur_gmail_com/.local/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:550: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  np_resource = np.dtype([(\"resource\", np.ubyte, 1)])\n",
      "2020-02-20 07:37:55,793 Hello! This is Joey-NMT.\n",
      "2020-02-20 07:37:55,800 Total params: 12140800\n",
      "2020-02-20 07:37:55,801 Trainable parameters: ['decoder.layer_norm.bias', 'decoder.layer_norm.weight', 'decoder.layers.0.dec_layer_norm.bias', 'decoder.layers.0.dec_layer_norm.weight', 'decoder.layers.0.feed_forward.layer_norm.bias', 'decoder.layers.0.feed_forward.layer_norm.weight', 'decoder.layers.0.feed_forward.pwff_layer.0.bias', 'decoder.layers.0.feed_forward.pwff_layer.0.weight', 'decoder.layers.0.feed_forward.pwff_layer.3.bias', 'decoder.layers.0.feed_forward.pwff_layer.3.weight', 'decoder.layers.0.src_trg_att.k_layer.bias', 'decoder.layers.0.src_trg_att.k_layer.weight', 'decoder.layers.0.src_trg_att.output_layer.bias', 'decoder.layers.0.src_trg_att.output_layer.weight', 'decoder.layers.0.src_trg_att.q_layer.bias', 'decoder.layers.0.src_trg_att.q_layer.weight', 'decoder.layers.0.src_trg_att.v_layer.bias', 'decoder.layers.0.src_trg_att.v_layer.weight', 'decoder.layers.0.trg_trg_att.k_layer.bias', 'decoder.layers.0.trg_trg_att.k_layer.weight', 'decoder.layers.0.trg_trg_att.output_layer.bias', 'decoder.layers.0.trg_trg_att.output_layer.weight', 'decoder.layers.0.trg_trg_att.q_layer.bias', 'decoder.layers.0.trg_trg_att.q_layer.weight', 'decoder.layers.0.trg_trg_att.v_layer.bias', 'decoder.layers.0.trg_trg_att.v_layer.weight', 'decoder.layers.0.x_layer_norm.bias', 'decoder.layers.0.x_layer_norm.weight', 'decoder.layers.1.dec_layer_norm.bias', 'decoder.layers.1.dec_layer_norm.weight', 'decoder.layers.1.feed_forward.layer_norm.bias', 'decoder.layers.1.feed_forward.layer_norm.weight', 'decoder.layers.1.feed_forward.pwff_layer.0.bias', 'decoder.layers.1.feed_forward.pwff_layer.0.weight', 'decoder.layers.1.feed_forward.pwff_layer.3.bias', 'decoder.layers.1.feed_forward.pwff_layer.3.weight', 'decoder.layers.1.src_trg_att.k_layer.bias', 'decoder.layers.1.src_trg_att.k_layer.weight', 'decoder.layers.1.src_trg_att.output_layer.bias', 'decoder.layers.1.src_trg_att.output_layer.weight', 'decoder.layers.1.src_trg_att.q_layer.bias', 'decoder.layers.1.src_trg_att.q_layer.weight', 'decoder.layers.1.src_trg_att.v_layer.bias', 'decoder.layers.1.src_trg_att.v_layer.weight', 'decoder.layers.1.trg_trg_att.k_layer.bias', 'decoder.layers.1.trg_trg_att.k_layer.weight', 'decoder.layers.1.trg_trg_att.output_layer.bias', 'decoder.layers.1.trg_trg_att.output_layer.weight', 'decoder.layers.1.trg_trg_att.q_layer.bias', 'decoder.layers.1.trg_trg_att.q_layer.weight', 'decoder.layers.1.trg_trg_att.v_layer.bias', 'decoder.layers.1.trg_trg_att.v_layer.weight', 'decoder.layers.1.x_layer_norm.bias', 'decoder.layers.1.x_layer_norm.weight', 'decoder.layers.2.dec_layer_norm.bias', 'decoder.layers.2.dec_layer_norm.weight', 'decoder.layers.2.feed_forward.layer_norm.bias', 'decoder.layers.2.feed_forward.layer_norm.weight', 'decoder.layers.2.feed_forward.pwff_layer.0.bias', 'decoder.layers.2.feed_forward.pwff_layer.0.weight', 'decoder.layers.2.feed_forward.pwff_layer.3.bias', 'decoder.layers.2.feed_forward.pwff_layer.3.weight', 'decoder.layers.2.src_trg_att.k_layer.bias', 'decoder.layers.2.src_trg_att.k_layer.weight', 'decoder.layers.2.src_trg_att.output_layer.bias', 'decoder.layers.2.src_trg_att.output_layer.weight', 'decoder.layers.2.src_trg_att.q_layer.bias', 'decoder.layers.2.src_trg_att.q_layer.weight', 'decoder.layers.2.src_trg_att.v_layer.bias', 'decoder.layers.2.src_trg_att.v_layer.weight', 'decoder.layers.2.trg_trg_att.k_layer.bias', 'decoder.layers.2.trg_trg_att.k_layer.weight', 'decoder.layers.2.trg_trg_att.output_layer.bias', 'decoder.layers.2.trg_trg_att.output_layer.weight', 'decoder.layers.2.trg_trg_att.q_layer.bias', 'decoder.layers.2.trg_trg_att.q_layer.weight', 'decoder.layers.2.trg_trg_att.v_layer.bias', 'decoder.layers.2.trg_trg_att.v_layer.weight', 'decoder.layers.2.x_layer_norm.bias', 'decoder.layers.2.x_layer_norm.weight', 'decoder.layers.3.dec_layer_norm.bias', 'decoder.layers.3.dec_layer_norm.weight', 'decoder.layers.3.feed_forward.layer_norm.bias', 'decoder.layers.3.feed_forward.layer_norm.weight', 'decoder.layers.3.feed_forward.pwff_layer.0.bias', 'decoder.layers.3.feed_forward.pwff_layer.0.weight', 'decoder.layers.3.feed_forward.pwff_layer.3.bias', 'decoder.layers.3.feed_forward.pwff_layer.3.weight', 'decoder.layers.3.src_trg_att.k_layer.bias', 'decoder.layers.3.src_trg_att.k_layer.weight', 'decoder.layers.3.src_trg_att.output_layer.bias', 'decoder.layers.3.src_trg_att.output_layer.weight', 'decoder.layers.3.src_trg_att.q_layer.bias', 'decoder.layers.3.src_trg_att.q_layer.weight', 'decoder.layers.3.src_trg_att.v_layer.bias', 'decoder.layers.3.src_trg_att.v_layer.weight', 'decoder.layers.3.trg_trg_att.k_layer.bias', 'decoder.layers.3.trg_trg_att.k_layer.weight', 'decoder.layers.3.trg_trg_att.output_layer.bias', 'decoder.layers.3.trg_trg_att.output_layer.weight', 'decoder.layers.3.trg_trg_att.q_layer.bias', 'decoder.layers.3.trg_trg_att.q_layer.weight', 'decoder.layers.3.trg_trg_att.v_layer.bias', 'decoder.layers.3.trg_trg_att.v_layer.weight', 'decoder.layers.3.x_layer_norm.bias', 'decoder.layers.3.x_layer_norm.weight', 'decoder.layers.4.dec_layer_norm.bias', 'decoder.layers.4.dec_layer_norm.weight', 'decoder.layers.4.feed_forward.layer_norm.bias', 'decoder.layers.4.feed_forward.layer_norm.weight', 'decoder.layers.4.feed_forward.pwff_layer.0.bias', 'decoder.layers.4.feed_forward.pwff_layer.0.weight', 'decoder.layers.4.feed_forward.pwff_layer.3.bias', 'decoder.layers.4.feed_forward.pwff_layer.3.weight', 'decoder.layers.4.src_trg_att.k_layer.bias', 'decoder.layers.4.src_trg_att.k_layer.weight', 'decoder.layers.4.src_trg_att.output_layer.bias', 'decoder.layers.4.src_trg_att.output_layer.weight', 'decoder.layers.4.src_trg_att.q_layer.bias', 'decoder.layers.4.src_trg_att.q_layer.weight', 'decoder.layers.4.src_trg_att.v_layer.bias', 'decoder.layers.4.src_trg_att.v_layer.weight', 'decoder.layers.4.trg_trg_att.k_layer.bias', 'decoder.layers.4.trg_trg_att.k_layer.weight', 'decoder.layers.4.trg_trg_att.output_layer.bias', 'decoder.layers.4.trg_trg_att.output_layer.weight', 'decoder.layers.4.trg_trg_att.q_layer.bias', 'decoder.layers.4.trg_trg_att.q_layer.weight', 'decoder.layers.4.trg_trg_att.v_layer.bias', 'decoder.layers.4.trg_trg_att.v_layer.weight', 'decoder.layers.4.x_layer_norm.bias', 'decoder.layers.4.x_layer_norm.weight', 'decoder.layers.5.dec_layer_norm.bias', 'decoder.layers.5.dec_layer_norm.weight', 'decoder.layers.5.feed_forward.layer_norm.bias', 'decoder.layers.5.feed_forward.layer_norm.weight', 'decoder.layers.5.feed_forward.pwff_layer.0.bias', 'decoder.layers.5.feed_forward.pwff_layer.0.weight', 'decoder.layers.5.feed_forward.pwff_layer.3.bias', 'decoder.layers.5.feed_forward.pwff_layer.3.weight', 'decoder.layers.5.src_trg_att.k_layer.bias', 'decoder.layers.5.src_trg_att.k_layer.weight', 'decoder.layers.5.src_trg_att.output_layer.bias', 'decoder.layers.5.src_trg_att.output_layer.weight', 'decoder.layers.5.src_trg_att.q_layer.bias', 'decoder.layers.5.src_trg_att.q_layer.weight', 'decoder.layers.5.src_trg_att.v_layer.bias', 'decoder.layers.5.src_trg_att.v_layer.weight', 'decoder.layers.5.trg_trg_att.k_layer.bias', 'decoder.layers.5.trg_trg_att.k_layer.weight', 'decoder.layers.5.trg_trg_att.output_layer.bias', 'decoder.layers.5.trg_trg_att.output_layer.weight', 'decoder.layers.5.trg_trg_att.q_layer.bias', 'decoder.layers.5.trg_trg_att.q_layer.weight', 'decoder.layers.5.trg_trg_att.v_layer.bias', 'decoder.layers.5.trg_trg_att.v_layer.weight', 'decoder.layers.5.x_layer_norm.bias', 'decoder.layers.5.x_layer_norm.weight', 'encoder.layer_norm.bias', 'encoder.layer_norm.weight', 'encoder.layers.0.feed_forward.layer_norm.bias', 'encoder.layers.0.feed_forward.layer_norm.weight', 'encoder.layers.0.feed_forward.pwff_layer.0.bias', 'encoder.layers.0.feed_forward.pwff_layer.0.weight', 'encoder.layers.0.feed_forward.pwff_layer.3.bias', 'encoder.layers.0.feed_forward.pwff_layer.3.weight', 'encoder.layers.0.layer_norm.bias', 'encoder.layers.0.layer_norm.weight', 'encoder.layers.0.src_src_att.k_layer.bias', 'encoder.layers.0.src_src_att.k_layer.weight', 'encoder.layers.0.src_src_att.output_layer.bias', 'encoder.layers.0.src_src_att.output_layer.weight', 'encoder.layers.0.src_src_att.q_layer.bias', 'encoder.layers.0.src_src_att.q_layer.weight', 'encoder.layers.0.src_src_att.v_layer.bias', 'encoder.layers.0.src_src_att.v_layer.weight', 'encoder.layers.1.feed_forward.layer_norm.bias', 'encoder.layers.1.feed_forward.layer_norm.weight', 'encoder.layers.1.feed_forward.pwff_layer.0.bias', 'encoder.layers.1.feed_forward.pwff_layer.0.weight', 'encoder.layers.1.feed_forward.pwff_layer.3.bias', 'encoder.layers.1.feed_forward.pwff_layer.3.weight', 'encoder.layers.1.layer_norm.bias', 'encoder.layers.1.layer_norm.weight', 'encoder.layers.1.src_src_att.k_layer.bias', 'encoder.layers.1.src_src_att.k_layer.weight', 'encoder.layers.1.src_src_att.output_layer.bias', 'encoder.layers.1.src_src_att.output_layer.weight', 'encoder.layers.1.src_src_att.q_layer.bias', 'encoder.layers.1.src_src_att.q_layer.weight', 'encoder.layers.1.src_src_att.v_layer.bias', 'encoder.layers.1.src_src_att.v_layer.weight', 'encoder.layers.2.feed_forward.layer_norm.bias', 'encoder.layers.2.feed_forward.layer_norm.weight', 'encoder.layers.2.feed_forward.pwff_layer.0.bias', 'encoder.layers.2.feed_forward.pwff_layer.0.weight', 'encoder.layers.2.feed_forward.pwff_layer.3.bias', 'encoder.layers.2.feed_forward.pwff_layer.3.weight', 'encoder.layers.2.layer_norm.bias', 'encoder.layers.2.layer_norm.weight', 'encoder.layers.2.src_src_att.k_layer.bias', 'encoder.layers.2.src_src_att.k_layer.weight', 'encoder.layers.2.src_src_att.output_layer.bias', 'encoder.layers.2.src_src_att.output_layer.weight', 'encoder.layers.2.src_src_att.q_layer.bias', 'encoder.layers.2.src_src_att.q_layer.weight', 'encoder.layers.2.src_src_att.v_layer.bias', 'encoder.layers.2.src_src_att.v_layer.weight', 'encoder.layers.3.feed_forward.layer_norm.bias', 'encoder.layers.3.feed_forward.layer_norm.weight', 'encoder.layers.3.feed_forward.pwff_layer.0.bias', 'encoder.layers.3.feed_forward.pwff_layer.0.weight', 'encoder.layers.3.feed_forward.pwff_layer.3.bias', 'encoder.layers.3.feed_forward.pwff_layer.3.weight', 'encoder.layers.3.layer_norm.bias', 'encoder.layers.3.layer_norm.weight', 'encoder.layers.3.src_src_att.k_layer.bias', 'encoder.layers.3.src_src_att.k_layer.weight', 'encoder.layers.3.src_src_att.output_layer.bias', 'encoder.layers.3.src_src_att.output_layer.weight', 'encoder.layers.3.src_src_att.q_layer.bias', 'encoder.layers.3.src_src_att.q_layer.weight', 'encoder.layers.3.src_src_att.v_layer.bias', 'encoder.layers.3.src_src_att.v_layer.weight', 'encoder.layers.4.feed_forward.layer_norm.bias', 'encoder.layers.4.feed_forward.layer_norm.weight', 'encoder.layers.4.feed_forward.pwff_layer.0.bias', 'encoder.layers.4.feed_forward.pwff_layer.0.weight', 'encoder.layers.4.feed_forward.pwff_layer.3.bias', 'encoder.layers.4.feed_forward.pwff_layer.3.weight', 'encoder.layers.4.layer_norm.bias', 'encoder.layers.4.layer_norm.weight', 'encoder.layers.4.src_src_att.k_layer.bias', 'encoder.layers.4.src_src_att.k_layer.weight', 'encoder.layers.4.src_src_att.output_layer.bias', 'encoder.layers.4.src_src_att.output_layer.weight', 'encoder.layers.4.src_src_att.q_layer.bias', 'encoder.layers.4.src_src_att.q_layer.weight', 'encoder.layers.4.src_src_att.v_layer.bias', 'encoder.layers.4.src_src_att.v_layer.weight', 'encoder.layers.5.feed_forward.layer_norm.bias', 'encoder.layers.5.feed_forward.layer_norm.weight', 'encoder.layers.5.feed_forward.pwff_layer.0.bias', 'encoder.layers.5.feed_forward.pwff_layer.0.weight', 'encoder.layers.5.feed_forward.pwff_layer.3.bias', 'encoder.layers.5.feed_forward.pwff_layer.3.weight', 'encoder.layers.5.layer_norm.bias', 'encoder.layers.5.layer_norm.weight', 'encoder.layers.5.src_src_att.k_layer.bias', 'encoder.layers.5.src_src_att.k_layer.weight', 'encoder.layers.5.src_src_att.output_layer.bias', 'encoder.layers.5.src_src_att.output_layer.weight', 'encoder.layers.5.src_src_att.q_layer.bias', 'encoder.layers.5.src_src_att.q_layer.weight', 'encoder.layers.5.src_src_att.v_layer.bias', 'encoder.layers.5.src_src_att.v_layer.weight', 'src_embed.lut.weight']\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2020-02-20 07:37:58,484 cfg.name                           : enki_transformer\n",
      "2020-02-20 07:37:58,484 cfg.data.src                       : en\n",
      "2020-02-20 07:37:58,484 cfg.data.trg                       : ki\n",
      "2020-02-20 07:37:58,484 cfg.data.train                     : data/enki/train.bpe\n",
      "2020-02-20 07:37:58,484 cfg.data.dev                       : data/enki/dev.bpe\n",
      "2020-02-20 07:37:58,484 cfg.data.test                      : data/enki/test.bpe\n",
      "2020-02-20 07:37:58,484 cfg.data.level                     : bpe\n",
      "2020-02-20 07:37:58,484 cfg.data.lowercase                 : False\n",
      "2020-02-20 07:37:58,484 cfg.data.max_sent_length           : 100\n",
      "2020-02-20 07:37:58,484 cfg.data.src_vocab                 : data/enki/vocab.txt\n",
      "2020-02-20 07:37:58,485 cfg.data.trg_vocab                 : data/enki/vocab.txt\n",
      "2020-02-20 07:37:58,485 cfg.testing.beam_size              : 5\n",
      "2020-02-20 07:37:58,485 cfg.testing.alpha                  : 1.0\n",
      "2020-02-20 07:37:58,485 cfg.training.random_seed           : 42\n",
      "2020-02-20 07:37:58,485 cfg.training.optimizer             : adam\n",
      "2020-02-20 07:37:58,485 cfg.training.normalization         : tokens\n",
      "2020-02-20 07:37:58,485 cfg.training.adam_betas            : [0.9, 0.999]\n",
      "2020-02-20 07:37:58,485 cfg.training.scheduling            : plateau\n",
      "2020-02-20 07:37:58,485 cfg.training.patience              : 5\n",
      "2020-02-20 07:37:58,485 cfg.training.learning_rate_factor  : 0.5\n",
      "2020-02-20 07:37:58,485 cfg.training.learning_rate_warmup  : 1000\n",
      "2020-02-20 07:37:58,485 cfg.training.decrease_factor       : 0.7\n",
      "2020-02-20 07:37:58,485 cfg.training.loss                  : crossentropy\n",
      "2020-02-20 07:37:58,485 cfg.training.learning_rate         : 0.0003\n",
      "2020-02-20 07:37:58,485 cfg.training.learning_rate_min     : 1e-08\n",
      "2020-02-20 07:37:58,486 cfg.training.weight_decay          : 0.0\n",
      "2020-02-20 07:37:58,486 cfg.training.label_smoothing       : 0.1\n",
      "2020-02-20 07:37:58,486 cfg.training.batch_size            : 4096\n",
      "2020-02-20 07:37:58,486 cfg.training.batch_type            : token\n",
      "2020-02-20 07:37:58,486 cfg.training.eval_batch_size       : 3600\n",
      "2020-02-20 07:37:58,486 cfg.training.eval_batch_type       : token\n",
      "2020-02-20 07:37:58,486 cfg.training.batch_multiplier      : 1\n",
      "2020-02-20 07:37:58,486 cfg.training.early_stopping_metric : ppl\n",
      "2020-02-20 07:37:58,486 cfg.training.epochs                : 30\n",
      "2020-02-20 07:37:58,486 cfg.training.validation_freq       : 1000\n",
      "2020-02-20 07:37:58,486 cfg.training.logging_freq          : 100\n",
      "2020-02-20 07:37:58,486 cfg.training.eval_metric           : bleu\n",
      "2020-02-20 07:37:58,486 cfg.training.model_dir             : models/enki_transformer\n",
      "2020-02-20 07:37:58,486 cfg.training.overwrite             : False\n",
      "2020-02-20 07:37:58,486 cfg.training.shuffle               : True\n",
      "2020-02-20 07:37:58,487 cfg.training.use_cuda              : True\n",
      "2020-02-20 07:37:58,487 cfg.training.max_output_length     : 100\n",
      "2020-02-20 07:37:58,487 cfg.training.print_valid_sents     : [0, 1, 2, 3]\n",
      "2020-02-20 07:37:58,487 cfg.training.keep_last_ckpts       : 3\n",
      "2020-02-20 07:37:58,487 cfg.model.initializer              : xavier\n",
      "2020-02-20 07:37:58,487 cfg.model.bias_initializer         : zeros\n",
      "2020-02-20 07:37:58,487 cfg.model.init_gain                : 1.0\n",
      "2020-02-20 07:37:58,487 cfg.model.embed_initializer        : xavier\n",
      "2020-02-20 07:37:58,487 cfg.model.embed_init_gain          : 1.0\n",
      "2020-02-20 07:37:58,487 cfg.model.tied_embeddings          : True\n",
      "2020-02-20 07:37:58,487 cfg.model.tied_softmax             : True\n",
      "2020-02-20 07:37:58,487 cfg.model.encoder.type             : transformer\n",
      "2020-02-20 07:37:58,487 cfg.model.encoder.num_layers       : 6\n",
      "2020-02-20 07:37:58,487 cfg.model.encoder.num_heads        : 4\n",
      "2020-02-20 07:37:58,487 cfg.model.encoder.embeddings.embedding_dim : 256\n",
      "2020-02-20 07:37:58,487 cfg.model.encoder.embeddings.scale : True\n",
      "2020-02-20 07:37:58,488 cfg.model.encoder.embeddings.dropout : 0.2\n",
      "2020-02-20 07:37:58,488 cfg.model.encoder.hidden_size      : 256\n",
      "2020-02-20 07:37:58,488 cfg.model.encoder.ff_size          : 1024\n",
      "2020-02-20 07:37:58,488 cfg.model.encoder.dropout          : 0.3\n",
      "2020-02-20 07:37:58,488 cfg.model.decoder.type             : transformer\n",
      "2020-02-20 07:37:58,488 cfg.model.decoder.num_layers       : 6\n",
      "2020-02-20 07:37:58,488 cfg.model.decoder.num_heads        : 4\n",
      "2020-02-20 07:37:58,488 cfg.model.decoder.embeddings.embedding_dim : 256\n",
      "2020-02-20 07:37:58,488 cfg.model.decoder.embeddings.scale : True\n",
      "2020-02-20 07:37:58,488 cfg.model.decoder.embeddings.dropout : 0.2\n",
      "2020-02-20 07:37:58,488 cfg.model.decoder.hidden_size      : 256\n",
      "2020-02-20 07:37:58,488 cfg.model.decoder.ff_size          : 1024\n",
      "2020-02-20 07:37:58,488 cfg.model.decoder.dropout          : 0.3\n",
      "2020-02-20 07:37:58,488 Data set sizes: \n",
      "\ttrain 92840,\n",
      "\tvalid 1000,\n",
      "\ttest 2691\n",
      "2020-02-20 07:37:58,489 First training example:\n",
      "\t[SRC] These are often d@@ ue to the bo@@ d@@ y’s dir@@ ecting blood to the cor@@ e to support v@@ ital org@@ ans .\n",
      "\t[TRG] Ũndũ ũcio wĩk@@ ĩkaga nĩ ũndũ wa mwĩrĩ gũ@@ tw@@ ara thakame harĩ ciĩ@@ ga iria cia bata nĩguo ith@@ iĩ na mbere kũruta wĩra .\n",
      "2020-02-20 07:37:58,489 First 10 words (src): (0) <unk> (1) <pad> (2) <s> (3) </s> (4) . (5) , (6) na (7) the (8) nĩ (9) to\n",
      "2020-02-20 07:37:58,489 First 10 words (trg): (0) <unk> (1) <pad> (2) <s> (3) </s> (4) . (5) , (6) na (7) the (8) nĩ (9) to\n",
      "2020-02-20 07:37:58,489 Number of Src words (types): 4221\n",
      "2020-02-20 07:37:58,489 Number of Trg words (types): 4221\n",
      "2020-02-20 07:37:58,490 Model(\n",
      "\tencoder=TransformerEncoder(num_layers=6, num_heads=4),\n",
      "\tdecoder=TransformerDecoder(num_layers=6, num_heads=4),\n",
      "\tsrc_embed=Embeddings(embedding_dim=256, vocab_size=4221),\n",
      "\ttrg_embed=Embeddings(embedding_dim=256, vocab_size=4221))\n",
      "2020-02-20 07:37:58,494 EPOCH 1\n",
      "2020-02-20 07:38:27,050 Epoch   1 Step:      100 Batch Loss:     5.094207 Tokens per Sec:     7556, Lr: 0.000300\n",
      "2020-02-20 07:38:55,736 Epoch   1 Step:      200 Batch Loss:     5.499734 Tokens per Sec:     7809, Lr: 0.000300\n",
      "2020-02-20 07:39:23,956 Epoch   1 Step:      300 Batch Loss:     4.892630 Tokens per Sec:     7675, Lr: 0.000300\n",
      "2020-02-20 07:39:52,754 Epoch   1 Step:      400 Batch Loss:     4.782716 Tokens per Sec:     7727, Lr: 0.000300\n",
      "2020-02-20 07:40:21,567 Epoch   1 Step:      500 Batch Loss:     4.550411 Tokens per Sec:     7685, Lr: 0.000300\n",
      "2020-02-20 07:40:50,307 Epoch   1 Step:      600 Batch Loss:     4.353094 Tokens per Sec:     7687, Lr: 0.000300\n",
      "2020-02-20 07:41:19,094 Epoch   1 Step:      700 Batch Loss:     4.602418 Tokens per Sec:     7669, Lr: 0.000300\n",
      "2020-02-20 07:41:47,585 Epoch   1 Step:      800 Batch Loss:     4.067693 Tokens per Sec:     7609, Lr: 0.000300\n",
      "2020-02-20 07:42:15,898 Epoch   1 Step:      900 Batch Loss:     3.944913 Tokens per Sec:     7505, Lr: 0.000300\n",
      "2020-02-20 07:42:44,747 Epoch   1 Step:     1000 Batch Loss:     4.178821 Tokens per Sec:     7611, Lr: 0.000300\n",
      "2020-02-20 07:44:15,452 Hooray! New best validation result [ppl]!\n",
      "2020-02-20 07:44:15,453 Saving new checkpoint.\n",
      "2020-02-20 07:44:15,661 Example #0\n",
      "2020-02-20 07:44:15,662 \tSource:     Researchers at the University of Copenhagen tracked almost 10,000 middle - aged people over an 11 - year period and found that participants who frequently argued with someone close to them were far more likely to die prematurely than those who seldom had conflicts .\n",
      "2020-02-20 07:44:15,662 \tReference:  Athuthuria yunivasĩtĩ - inĩ ya Copenhagen , nĩ maathuthuririe andũ hakuhĩ 10,000 a mĩaka ĩyo kwa ihinda rĩa mĩaka 11 na makĩona atĩ arĩa maakararanagia na mũndũ wa famĩlĩ kaingĩ , haakoragwo na ũhotekeku mũnene atĩ megũkua tene gũkĩra arĩa mataakoragwo na ngarari kaingĩ .\n",
      "2020-02-20 07:44:15,662 \tHypothesis: Maaĩ na mbere ya Maĩ ya Maĩ , nĩ maanyitire na mbere ya Maĩ na mbere ya Maĩ na mbere ya Marararararagagire na mbere ya mĩĩ na mbere ya mĩthĩ .\n",
      "2020-02-20 07:44:15,662 Example #1\n",
      "2020-02-20 07:44:15,662 \tSource:     Yes .\n",
      "2020-02-20 07:44:15,662 \tReference:  Ĩĩ .\n",
      "2020-02-20 07:44:15,662 \tHypothesis: Nĩ ũndũ ũcio .\n",
      "2020-02-20 07:44:15,662 Example #2\n",
      "2020-02-20 07:44:15,662 \tSource:     He enjoyed special intimacy with Jehovah , who enabled him to display “ awesome power ” as he led the Israelites to the Promised Land . ​ — Deut .\n",
      "2020-02-20 07:44:15,663 \tReference:  Aarĩ na ũkuruhanu wa hakuhĩ na Jehova , na nĩ aamũheire ũhoti wa gwĩka “ maũndũ manene ma kũguoyohania ” rĩrĩa aatongoragia Aisiraeli mathiĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 07:44:15,663 \tHypothesis: Nĩ ũndũ wa bata mũno mũno nĩ ũndũ wa Ngai , nĩ ũndũ wa Ngai nĩ ũndũ wa Ithe witũ wa Ngai nĩ ũndũ wa Ithe witũ wa Ithe witũ wa Ithe witũ wa Ngai . — 1 : 1 .\n",
      "2020-02-20 07:44:15,663 Example #3\n",
      "2020-02-20 07:44:15,663 \tSource:     HISTORY : ATHEIST\n",
      "2020-02-20 07:44:15,663 \tReference:  HISTORĨ : EETĨKĨTIE GŨTIRĨ NGAI\n",
      "2020-02-20 07:44:15,663 \tHypothesis: KA : “ ŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨŨA\n",
      "2020-02-20 07:44:15,663 Validation result at epoch   1, step     1000: bleu:   1.98, loss: 99303.9922, ppl:  50.3587, duration: 90.9150s\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2020-02-20 07:44:44,480 Epoch   1 Step:     1100 Batch Loss:     3.988148 Tokens per Sec:     7646, Lr: 0.000300\n",
      "2020-02-20 07:44:44,751 Epoch   1: total training loss 5155.24\n",
      "2020-02-20 07:44:44,752 EPOCH 2\n",
      "2020-02-20 07:45:13,301 Epoch   2 Step:     1200 Batch Loss:     3.747825 Tokens per Sec:     7729, Lr: 0.000300\n",
      "2020-02-20 07:45:42,119 Epoch   2 Step:     1300 Batch Loss:     3.764219 Tokens per Sec:     7797, Lr: 0.000300\n",
      "2020-02-20 07:46:10,766 Epoch   2 Step:     1400 Batch Loss:     3.736711 Tokens per Sec:     7631, Lr: 0.000300\n",
      "2020-02-20 07:46:39,691 Epoch   2 Step:     1500 Batch Loss:     3.946311 Tokens per Sec:     7654, Lr: 0.000300\n",
      "2020-02-20 07:47:08,246 Epoch   2 Step:     1600 Batch Loss:     3.595036 Tokens per Sec:     7729, Lr: 0.000300\n",
      "2020-02-20 07:47:36,814 Epoch   2 Step:     1700 Batch Loss:     3.649236 Tokens per Sec:     7372, Lr: 0.000300\n",
      "2020-02-20 07:48:05,665 Epoch   2 Step:     1800 Batch Loss:     3.596375 Tokens per Sec:     7804, Lr: 0.000300\n",
      "2020-02-20 07:48:34,279 Epoch   2 Step:     1900 Batch Loss:     3.455034 Tokens per Sec:     7611, Lr: 0.000300\n",
      "2020-02-20 07:49:02,976 Epoch   2 Step:     2000 Batch Loss:     3.500157 Tokens per Sec:     7663, Lr: 0.000300\n",
      "2020-02-20 07:50:33,745 Hooray! New best validation result [ppl]!\n",
      "2020-02-20 07:50:33,746 Saving new checkpoint.\n",
      "2020-02-20 07:50:33,947 Example #0\n",
      "2020-02-20 07:50:33,948 \tSource:     Researchers at the University of Copenhagen tracked almost 10,000 middle - aged people over an 11 - year period and found that participants who frequently argued with someone close to them were far more likely to die prematurely than those who seldom had conflicts .\n",
      "2020-02-20 07:50:33,948 \tReference:  Athuthuria yunivasĩtĩ - inĩ ya Copenhagen , nĩ maathuthuririe andũ hakuhĩ 10,000 a mĩaka ĩyo kwa ihinda rĩa mĩaka 11 na makĩona atĩ arĩa maakararanagia na mũndũ wa famĩlĩ kaingĩ , haakoragwo na ũhotekeku mũnene atĩ megũkua tene gũkĩra arĩa mataakoragwo na ngarari kaingĩ .\n",
      "2020-02-20 07:50:33,948 \tHypothesis: Kaingĩ nĩ maakenagio nĩ Kĩhibirania kĩa Kĩhibirania kĩa Kĩhibirania , nĩ ciathiaga na mbere ya Kĩhibirania , na nĩ ũndũ wa kũhunjĩria andũ aingĩ a famĩlĩ yake na njĩra ya kũhunjĩria andũ arĩa angĩ .\n",
      "2020-02-20 07:50:33,948 Example #1\n",
      "2020-02-20 07:50:33,948 \tSource:     Yes .\n",
      "2020-02-20 07:50:33,948 \tReference:  Ĩĩ .\n",
      "2020-02-20 07:50:33,948 \tHypothesis: Hihi nĩ kũrĩ ũguo .\n",
      "2020-02-20 07:50:33,949 Example #2\n",
      "2020-02-20 07:50:33,949 \tSource:     He enjoyed special intimacy with Jehovah , who enabled him to display “ awesome power ” as he led the Israelites to the Promised Land . ​ — Deut .\n",
      "2020-02-20 07:50:33,949 \tReference:  Aarĩ na ũkuruhanu wa hakuhĩ na Jehova , na nĩ aamũheire ũhoti wa gwĩka “ maũndũ manene ma kũguoyohania ” rĩrĩa aatongoragia Aisiraeli mathiĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 07:50:33,949 \tHypothesis: Nĩ ndaarutire atĩ Jehova nĩ aakenire na akĩmũteithagia Ngai , na nĩ aakenire “ Ngai . ” — Ath .\n",
      "2020-02-20 07:50:33,949 Example #3\n",
      "2020-02-20 07:50:33,949 \tSource:     HISTORY : ATHEIST\n",
      "2020-02-20 07:50:33,949 \tReference:  HISTORĨ : EETĨKĨTIE GŨTIRĨ NGAI\n",
      "2020-02-20 07:50:33,949 \tHypothesis: KARATATHI GA NYĨMBO :\n",
      "2020-02-20 07:50:33,950 Validation result at epoch   2, step     2000: bleu:   3.18, loss: 83300.2578, ppl:  26.7775, duration: 90.9731s\n",
      "2020-02-20 07:51:02,607 Epoch   2 Step:     2100 Batch Loss:     2.876269 Tokens per Sec:     7718, Lr: 0.000300\n",
      "2020-02-20 07:51:30,207 Epoch   2: total training loss 3904.64\n",
      "2020-02-20 07:51:30,207 EPOCH 3\n",
      "2020-02-20 07:51:31,470 Epoch   3 Step:     2200 Batch Loss:     3.112139 Tokens per Sec:     6915, Lr: 0.000300\n",
      "2020-02-20 07:52:00,021 Epoch   3 Step:     2300 Batch Loss:     3.386087 Tokens per Sec:     7602, Lr: 0.000300\n",
      "2020-02-20 07:52:28,940 Epoch   3 Step:     2400 Batch Loss:     3.473860 Tokens per Sec:     7726, Lr: 0.000300\n",
      "2020-02-20 07:52:57,706 Epoch   3 Step:     2500 Batch Loss:     3.362963 Tokens per Sec:     7690, Lr: 0.000300\n",
      "2020-02-20 07:53:26,189 Epoch   3 Step:     2600 Batch Loss:     3.410449 Tokens per Sec:     7566, Lr: 0.000300\n",
      "2020-02-20 07:53:55,039 Epoch   3 Step:     2700 Batch Loss:     3.013206 Tokens per Sec:     7670, Lr: 0.000300\n",
      "2020-02-20 07:54:24,005 Epoch   3 Step:     2800 Batch Loss:     3.246269 Tokens per Sec:     7617, Lr: 0.000300\n",
      "2020-02-20 07:54:52,704 Epoch   3 Step:     2900 Batch Loss:     3.079741 Tokens per Sec:     7532, Lr: 0.000300\n",
      "2020-02-20 07:55:21,631 Epoch   3 Step:     3000 Batch Loss:     3.617651 Tokens per Sec:     7677, Lr: 0.000300\n",
      "2020-02-20 07:56:52,426 Hooray! New best validation result [ppl]!\n",
      "2020-02-20 07:56:52,427 Saving new checkpoint.\n",
      "2020-02-20 07:56:52,626 Example #0\n",
      "2020-02-20 07:56:52,627 \tSource:     Researchers at the University of Copenhagen tracked almost 10,000 middle - aged people over an 11 - year period and found that participants who frequently argued with someone close to them were far more likely to die prematurely than those who seldom had conflicts .\n",
      "2020-02-20 07:56:52,627 \tReference:  Athuthuria yunivasĩtĩ - inĩ ya Copenhagen , nĩ maathuthuririe andũ hakuhĩ 10,000 a mĩaka ĩyo kwa ihinda rĩa mĩaka 11 na makĩona atĩ arĩa maakararanagia na mũndũ wa famĩlĩ kaingĩ , haakoragwo na ũhotekeku mũnene atĩ megũkua tene gũkĩra arĩa mataakoragwo na ngarari kaingĩ .\n",
      "2020-02-20 07:56:52,627 \tHypothesis: Ariũ a Ithe witũ arĩa mahikanĩtie na Kĩama Kĩrĩa Gĩtongoragia thĩinĩ wa thĩ , nĩ maambĩrĩirie kũmakinya ma mĩaka mĩingĩ na thutha wa mĩaka 20 na thutha wa ihinda rĩa mĩaka mĩingĩ , na thutha wa ihinda rĩa mĩaka mĩingĩ , na rĩrĩa rĩarĩ na andũ aingĩ nĩ maacokire makĩhũthĩra mĩciĩ - inĩ ya andũ aingĩ .\n",
      "2020-02-20 07:56:52,627 Example #1\n",
      "2020-02-20 07:56:52,627 \tSource:     Yes .\n",
      "2020-02-20 07:56:52,627 \tReference:  Ĩĩ .\n",
      "2020-02-20 07:56:52,628 \tHypothesis: Nĩ ũndũ wa bata .\n",
      "2020-02-20 07:56:52,628 Example #2\n",
      "2020-02-20 07:56:52,628 \tSource:     He enjoyed special intimacy with Jehovah , who enabled him to display “ awesome power ” as he led the Israelites to the Promised Land . ​ — Deut .\n",
      "2020-02-20 07:56:52,628 \tReference:  Aarĩ na ũkuruhanu wa hakuhĩ na Jehova , na nĩ aamũheire ũhoti wa gwĩka “ maũndũ manene ma kũguoyohania ” rĩrĩa aatongoragia Aisiraeli mathiĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 07:56:52,628 \tHypothesis: Nĩ aakenire mũno nĩ Jehova na njĩra ya kũheo wĩra wake , na nĩ ũndũ wa wendo wake wa andũ arĩa angĩ . — Ath .\n",
      "2020-02-20 07:56:52,628 Example #3\n",
      "2020-02-20 07:56:52,628 \tSource:     HISTORY : ATHEIST\n",
      "2020-02-20 07:56:52,629 \tReference:  HISTORĨ : EETĨKĨTIE GŨTIRĨ NGAI\n",
      "2020-02-20 07:56:52,629 \tHypothesis: NGĨKĨKĨKĨKĨKĨKĨKĨKĨKĨKĨKĨKĨKĨKĨKĨKĨKĨKĨKĨKĨKĨKA\n",
      "2020-02-20 07:56:52,629 Validation result at epoch   3, step     3000: bleu:   5.38, loss: 74742.7656, ppl:  19.1027, duration: 90.9969s\n",
      "2020-02-20 07:57:21,359 Epoch   3 Step:     3100 Batch Loss:     2.796993 Tokens per Sec:     7609, Lr: 0.000300\n",
      "2020-02-20 07:57:50,188 Epoch   3 Step:     3200 Batch Loss:     2.768728 Tokens per Sec:     7606, Lr: 0.000300\n",
      "2020-02-20 07:58:17,267 Epoch   3: total training loss 3413.38\n",
      "2020-02-20 07:58:17,267 EPOCH 4\n",
      "2020-02-20 07:58:19,424 Epoch   4 Step:     3300 Batch Loss:     2.948042 Tokens per Sec:     7130, Lr: 0.000300\n",
      "2020-02-20 07:58:48,122 Epoch   4 Step:     3400 Batch Loss:     2.958692 Tokens per Sec:     7780, Lr: 0.000300\n",
      "2020-02-20 07:59:16,755 Epoch   4 Step:     3500 Batch Loss:     2.594929 Tokens per Sec:     7705, Lr: 0.000300\n",
      "2020-02-20 07:59:45,458 Epoch   4 Step:     3600 Batch Loss:     3.204539 Tokens per Sec:     7688, Lr: 0.000300\n",
      "2020-02-20 08:00:14,042 Epoch   4 Step:     3700 Batch Loss:     2.986923 Tokens per Sec:     7522, Lr: 0.000300\n",
      "2020-02-20 08:00:42,804 Epoch   4 Step:     3800 Batch Loss:     2.783321 Tokens per Sec:     7706, Lr: 0.000300\n",
      "2020-02-20 08:01:11,215 Epoch   4 Step:     3900 Batch Loss:     2.713317 Tokens per Sec:     7489, Lr: 0.000300\n",
      "2020-02-20 08:01:40,145 Epoch   4 Step:     4000 Batch Loss:     2.642487 Tokens per Sec:     7759, Lr: 0.000300\n",
      "2020-02-20 08:03:11,204 Hooray! New best validation result [ppl]!\n",
      "2020-02-20 08:03:11,204 Saving new checkpoint.\n",
      "2020-02-20 08:03:11,417 Example #0\n",
      "2020-02-20 08:03:11,418 \tSource:     Researchers at the University of Copenhagen tracked almost 10,000 middle - aged people over an 11 - year period and found that participants who frequently argued with someone close to them were far more likely to die prematurely than those who seldom had conflicts .\n",
      "2020-02-20 08:03:11,418 \tReference:  Athuthuria yunivasĩtĩ - inĩ ya Copenhagen , nĩ maathuthuririe andũ hakuhĩ 10,000 a mĩaka ĩyo kwa ihinda rĩa mĩaka 11 na makĩona atĩ arĩa maakararanagia na mũndũ wa famĩlĩ kaingĩ , haakoragwo na ũhotekeku mũnene atĩ megũkua tene gũkĩra arĩa mataakoragwo na ngarari kaingĩ .\n",
      "2020-02-20 08:03:11,418 \tHypothesis: Roho wa Roma wa Kĩama Kĩrĩa Gĩtongoragia kĩa goro mũno nĩ kĩonainie andũ milioni nyingĩ thĩinĩ wa thĩ yothe , nĩ maanyitire na andũ arĩa maikaraga marĩ na gĩkeno .\n",
      "2020-02-20 08:03:11,418 Example #1\n",
      "2020-02-20 08:03:11,418 \tSource:     Yes .\n",
      "2020-02-20 08:03:11,418 \tReference:  Ĩĩ .\n",
      "2020-02-20 08:03:11,418 \tHypothesis: Hatarĩ nganja .\n",
      "2020-02-20 08:03:11,418 Example #2\n",
      "2020-02-20 08:03:11,418 \tSource:     He enjoyed special intimacy with Jehovah , who enabled him to display “ awesome power ” as he led the Israelites to the Promised Land . ​ — Deut .\n",
      "2020-02-20 08:03:11,419 \tReference:  Aarĩ na ũkuruhanu wa hakuhĩ na Jehova , na nĩ aamũheire ũhoti wa gwĩka “ maũndũ manene ma kũguoyohania ” rĩrĩa aatongoragia Aisiraeli mathiĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 08:03:11,419 \tHypothesis: Nĩ aakenire mũno nĩ ũndũ wa mwanya , Jehova nĩ aatuire itua rĩa ‘ akĩmũhe hinya ’ na akĩmũhe andũ arĩa angĩ . — Ath .\n",
      "2020-02-20 08:03:11,419 Example #3\n",
      "2020-02-20 08:03:11,419 \tSource:     HISTORY : ATHEIST\n",
      "2020-02-20 08:03:11,419 \tReference:  HISTORĨ : EETĨKĨTIE GŨTIRĨ NGAI\n",
      "2020-02-20 08:03:11,419 \tHypothesis: HIHI : Hihi ŨNDŨ\n",
      "2020-02-20 08:03:11,419 Validation result at epoch   4, step     4000: bleu:   8.08, loss: 69263.2109, ppl:  15.3877, duration: 91.2732s\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2020-02-20 08:03:40,183 Epoch   4 Step:     4100 Batch Loss:     3.001262 Tokens per Sec:     7681, Lr: 0.000300\n",
      "2020-02-20 08:04:09,053 Epoch   4 Step:     4200 Batch Loss:     2.729484 Tokens per Sec:     7615, Lr: 0.000300\n",
      "2020-02-20 08:04:37,737 Epoch   4 Step:     4300 Batch Loss:     2.872530 Tokens per Sec:     7649, Lr: 0.000300\n",
      "2020-02-20 08:05:03,603 Epoch   4: total training loss 3117.84\n",
      "2020-02-20 08:05:03,603 EPOCH 5\n",
      "2020-02-20 08:05:06,619 Epoch   5 Step:     4400 Batch Loss:     2.815613 Tokens per Sec:     7475, Lr: 0.000300\n",
      "2020-02-20 08:05:35,235 Epoch   5 Step:     4500 Batch Loss:     2.658053 Tokens per Sec:     7700, Lr: 0.000300\n",
      "2020-02-20 08:06:03,907 Epoch   5 Step:     4600 Batch Loss:     2.678884 Tokens per Sec:     7692, Lr: 0.000300\n",
      "2020-02-20 08:06:32,440 Epoch   5 Step:     4700 Batch Loss:     2.239846 Tokens per Sec:     7679, Lr: 0.000300\n",
      "2020-02-20 08:07:01,316 Epoch   5 Step:     4800 Batch Loss:     2.742702 Tokens per Sec:     7599, Lr: 0.000300\n",
      "2020-02-20 08:07:29,682 Epoch   5 Step:     4900 Batch Loss:     3.182590 Tokens per Sec:     7581, Lr: 0.000300\n",
      "2020-02-20 08:07:58,290 Epoch   5 Step:     5000 Batch Loss:     2.688263 Tokens per Sec:     7625, Lr: 0.000300\n",
      "2020-02-20 08:09:29,024 Hooray! New best validation result [ppl]!\n",
      "2020-02-20 08:09:29,024 Saving new checkpoint.\n",
      "2020-02-20 08:09:29,248 Example #0\n",
      "2020-02-20 08:09:29,248 \tSource:     Researchers at the University of Copenhagen tracked almost 10,000 middle - aged people over an 11 - year period and found that participants who frequently argued with someone close to them were far more likely to die prematurely than those who seldom had conflicts .\n",
      "2020-02-20 08:09:29,248 \tReference:  Athuthuria yunivasĩtĩ - inĩ ya Copenhagen , nĩ maathuthuririe andũ hakuhĩ 10,000 a mĩaka ĩyo kwa ihinda rĩa mĩaka 11 na makĩona atĩ arĩa maakararanagia na mũndũ wa famĩlĩ kaingĩ , haakoragwo na ũhotekeku mũnene atĩ megũkua tene gũkĩra arĩa mataakoragwo na ngarari kaingĩ .\n",
      "2020-02-20 08:09:29,248 \tHypothesis: Mabũrũri - inĩ ma Amerika ( U.S ) nĩ maahutirie andũ a ndũrĩrĩ nyingĩ mũno thĩinĩ wa thĩ yothe , andũ aingĩ nĩ maahũrũrũragia andũ arĩa maikaraga maikaraga makĩmũcara na andũ arĩa maikaraga makĩmũcara .\n",
      "2020-02-20 08:09:29,248 Example #1\n",
      "2020-02-20 08:09:29,249 \tSource:     Yes .\n",
      "2020-02-20 08:09:29,249 \tReference:  Ĩĩ .\n",
      "2020-02-20 08:09:29,249 \tHypothesis: Wĩtĩkio nĩ Ngai .\n",
      "2020-02-20 08:09:29,249 Example #2\n",
      "2020-02-20 08:09:29,249 \tSource:     He enjoyed special intimacy with Jehovah , who enabled him to display “ awesome power ” as he led the Israelites to the Promised Land . ​ — Deut .\n",
      "2020-02-20 08:09:29,249 \tReference:  Aarĩ na ũkuruhanu wa hakuhĩ na Jehova , na nĩ aamũheire ũhoti wa gwĩka “ maũndũ manene ma kũguoyohania ” rĩrĩa aatongoragia Aisiraeli mathiĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 08:09:29,249 \tHypothesis: Agĩcoka akĩgeria Jehova na akĩmũhe “ hinya , ” na nĩ aamũrathimire “ hinya wa kũheo gĩtĩo ” na akĩmũhe gĩtĩo . — Ahib .\n",
      "2020-02-20 08:09:29,249 Example #3\n",
      "2020-02-20 08:09:29,249 \tSource:     HISTORY : ATHEIST\n",
      "2020-02-20 08:09:29,250 \tReference:  HISTORĨ : EETĨKĨTIE GŨTIRĨ NGAI\n",
      "2020-02-20 08:09:29,250 \tHypothesis: HISSTI : MŨTŨTŨTŨTŨTA\n",
      "2020-02-20 08:09:29,250 Validation result at epoch   5, step     5000: bleu:   9.67, loss: 65603.0391, ppl:  13.3180, duration: 90.9585s\n",
      "2020-02-20 08:09:58,075 Epoch   5 Step:     5100 Batch Loss:     2.653653 Tokens per Sec:     7732, Lr: 0.000300\n",
      "2020-02-20 08:10:26,560 Epoch   5 Step:     5200 Batch Loss:     2.692245 Tokens per Sec:     7670, Lr: 0.000300\n",
      "2020-02-20 08:10:55,289 Epoch   5 Step:     5300 Batch Loss:     2.038270 Tokens per Sec:     7580, Lr: 0.000300\n",
      "2020-02-20 08:11:24,077 Epoch   5 Step:     5400 Batch Loss:     2.657159 Tokens per Sec:     7702, Lr: 0.000300\n",
      "2020-02-20 08:11:49,648 Epoch   5: total training loss 2930.16\n",
      "2020-02-20 08:11:49,649 EPOCH 6\n",
      "2020-02-20 08:11:52,907 Epoch   6 Step:     5500 Batch Loss:     2.449648 Tokens per Sec:     7027, Lr: 0.000300\n",
      "2020-02-20 08:12:21,699 Epoch   6 Step:     5600 Batch Loss:     2.688391 Tokens per Sec:     7641, Lr: 0.000300\n",
      "2020-02-20 08:12:50,114 Epoch   6 Step:     5700 Batch Loss:     2.303163 Tokens per Sec:     7634, Lr: 0.000300\n",
      "2020-02-20 08:13:18,903 Epoch   6 Step:     5800 Batch Loss:     1.819047 Tokens per Sec:     7710, Lr: 0.000300\n",
      "2020-02-20 08:13:47,711 Epoch   6 Step:     5900 Batch Loss:     1.995629 Tokens per Sec:     7695, Lr: 0.000300\n",
      "2020-02-20 08:14:16,270 Epoch   6 Step:     6000 Batch Loss:     2.383198 Tokens per Sec:     7528, Lr: 0.000300\n",
      "2020-02-20 08:15:46,992 Hooray! New best validation result [ppl]!\n",
      "2020-02-20 08:15:46,992 Saving new checkpoint.\n",
      "2020-02-20 08:15:47,211 Example #0\n",
      "2020-02-20 08:15:47,211 \tSource:     Researchers at the University of Copenhagen tracked almost 10,000 middle - aged people over an 11 - year period and found that participants who frequently argued with someone close to them were far more likely to die prematurely than those who seldom had conflicts .\n",
      "2020-02-20 08:15:47,211 \tReference:  Athuthuria yunivasĩtĩ - inĩ ya Copenhagen , nĩ maathuthuririe andũ hakuhĩ 10,000 a mĩaka ĩyo kwa ihinda rĩa mĩaka 11 na makĩona atĩ arĩa maakararanagia na mũndũ wa famĩlĩ kaingĩ , haakoragwo na ũhotekeku mũnene atĩ megũkua tene gũkĩra arĩa mataakoragwo na ngarari kaingĩ .\n",
      "2020-02-20 08:15:47,211 \tHypothesis: Ruthu ũrĩa warĩ na mĩaka 1,500 , andũ aingĩ nĩ maahutirie andũ milioni nyingĩ mũno thĩinĩ wa thĩ yothe , na nĩ maanyitirũo ũgeni andũ milioni nyingĩ mũno nĩ ũndũ wa gũkorũo makĩgeria kũmakania na andũ arĩa maahurĩtwo nĩ andũ arĩa maakuĩte .\n",
      "2020-02-20 08:15:47,211 Example #1\n",
      "2020-02-20 08:15:47,212 \tSource:     Yes .\n",
      "2020-02-20 08:15:47,212 \tReference:  Ĩĩ .\n",
      "2020-02-20 08:15:47,212 \tHypothesis: Hatarĩ nganja .\n",
      "2020-02-20 08:15:47,212 Example #2\n",
      "2020-02-20 08:15:47,212 \tSource:     He enjoyed special intimacy with Jehovah , who enabled him to display “ awesome power ” as he led the Israelites to the Promised Land . ​ — Deut .\n",
      "2020-02-20 08:15:47,212 \tReference:  Aarĩ na ũkuruhanu wa hakuhĩ na Jehova , na nĩ aamũheire ũhoti wa gwĩka “ maũndũ manene ma kũguoyohania ” rĩrĩa aatongoragia Aisiraeli mathiĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 08:15:47,212 \tHypothesis: Nĩ akenire mũno nĩ ũndũ wa mwanya mũno , ũrĩa Jehova aamwĩrire “ hinya wa ũthamaki ” ũrĩa waheire Aisiraeli , na akĩmwĩra Aisiraeli . — Gũcok .\n",
      "2020-02-20 08:15:47,212 Example #3\n",
      "2020-02-20 08:15:47,212 \tSource:     HISTORY : ATHEIST\n",
      "2020-02-20 08:15:47,212 \tReference:  HISTORĨ : EETĨKĨTIE GŨTIRĨ NGAI\n",
      "2020-02-20 08:15:47,212 \tHypothesis: HISU : MŨNDŨ\n",
      "2020-02-20 08:15:47,213 Validation result at epoch   6, step     6000: bleu:  11.02, loss: 62489.7344, ppl:  11.7781, duration: 90.9422s\n",
      "2020-02-20 08:16:15,655 Epoch   6 Step:     6100 Batch Loss:     1.549631 Tokens per Sec:     7505, Lr: 0.000300\n",
      "2020-02-20 08:16:44,322 Epoch   6 Step:     6200 Batch Loss:     1.894732 Tokens per Sec:     7690, Lr: 0.000300\n",
      "2020-02-20 08:17:13,342 Epoch   6 Step:     6300 Batch Loss:     2.294345 Tokens per Sec:     7703, Lr: 0.000300\n",
      "2020-02-20 08:17:41,809 Epoch   6 Step:     6400 Batch Loss:     2.516922 Tokens per Sec:     7504, Lr: 0.000300\n",
      "2020-02-20 08:18:10,715 Epoch   6 Step:     6500 Batch Loss:     2.287703 Tokens per Sec:     7821, Lr: 0.000300\n",
      "2020-02-20 08:18:36,050 Epoch   6: total training loss 2780.73\n",
      "2020-02-20 08:18:36,050 EPOCH 7\n",
      "2020-02-20 08:18:39,915 Epoch   7 Step:     6600 Batch Loss:     2.096579 Tokens per Sec:     7365, Lr: 0.000300\n",
      "2020-02-20 08:19:08,491 Epoch   7 Step:     6700 Batch Loss:     2.316535 Tokens per Sec:     7598, Lr: 0.000300\n",
      "2020-02-20 08:19:37,390 Epoch   7 Step:     6800 Batch Loss:     2.555669 Tokens per Sec:     7627, Lr: 0.000300\n",
      "2020-02-20 08:20:05,990 Epoch   7 Step:     6900 Batch Loss:     2.145194 Tokens per Sec:     7583, Lr: 0.000300\n",
      "2020-02-20 08:20:34,743 Epoch   7 Step:     7000 Batch Loss:     2.290301 Tokens per Sec:     7711, Lr: 0.000300\n",
      "2020-02-20 08:22:05,577 Hooray! New best validation result [ppl]!\n",
      "2020-02-20 08:22:05,578 Saving new checkpoint.\n",
      "2020-02-20 08:22:05,804 Example #0\n",
      "2020-02-20 08:22:05,804 \tSource:     Researchers at the University of Copenhagen tracked almost 10,000 middle - aged people over an 11 - year period and found that participants who frequently argued with someone close to them were far more likely to die prematurely than those who seldom had conflicts .\n",
      "2020-02-20 08:22:05,804 \tReference:  Athuthuria yunivasĩtĩ - inĩ ya Copenhagen , nĩ maathuthuririe andũ hakuhĩ 10,000 a mĩaka ĩyo kwa ihinda rĩa mĩaka 11 na makĩona atĩ arĩa maakararanagia na mũndũ wa famĩlĩ kaingĩ , haakoragwo na ũhotekeku mũnene atĩ megũkua tene gũkĩra arĩa mataakoragwo na ngarari kaingĩ .\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2020-02-20 08:22:05,804 \tHypothesis: Ariũ a Ithe witũ arĩa maikaraga Amerika ( U.S ) nĩ maahoririe andũ milioni nyingĩ mũno , na nĩ maakenire mũno nĩ ũndũ wa gũkorũo na andũ milioni nyingĩ mũno , na nĩ maanyitirũo ũgeni nĩ andũ arĩa maikaraga makuĩte ngoro mũno .\n",
      "2020-02-20 08:22:05,804 Example #1\n",
      "2020-02-20 08:22:05,805 \tSource:     Yes .\n",
      "2020-02-20 08:22:05,805 \tReference:  Ĩĩ .\n",
      "2020-02-20 08:22:05,805 \tHypothesis: Ĩĩ .\n",
      "2020-02-20 08:22:05,805 Example #2\n",
      "2020-02-20 08:22:05,805 \tSource:     He enjoyed special intimacy with Jehovah , who enabled him to display “ awesome power ” as he led the Israelites to the Promised Land . ​ — Deut .\n",
      "2020-02-20 08:22:05,805 \tReference:  Aarĩ na ũkuruhanu wa hakuhĩ na Jehova , na nĩ aamũheire ũhoti wa gwĩka “ maũndũ manene ma kũguoyohania ” rĩrĩa aatongoragia Aisiraeli mathiĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 08:22:05,805 \tHypothesis: Nĩ aakenire mũno nĩ ũndũ wa gũkorũo na wendo wa mwanya , Jehova nĩ aamwĩrire atĩ nĩ ‘ aakenagĩra ’ na akĩmũhe Aisiraeli . — Gũcok .\n",
      "2020-02-20 08:22:05,805 Example #3\n",
      "2020-02-20 08:22:05,805 \tSource:     HISTORY : ATHEIST\n",
      "2020-02-20 08:22:05,806 \tReference:  HISTORĨ : EETĨKĨTIE GŨTIRĨ NGAI\n",
      "2020-02-20 08:22:05,806 \tHypothesis: HISTORĨ : MŨNDŨ\n",
      "2020-02-20 08:22:05,806 Validation result at epoch   7, step     7000: bleu:  12.02, loss: 60542.1758, ppl:  10.9068, duration: 91.0617s\n",
      "2020-02-20 08:22:34,129 Epoch   7 Step:     7100 Batch Loss:     1.495246 Tokens per Sec:     7581, Lr: 0.000300\n",
      "2020-02-20 08:23:02,832 Epoch   7 Step:     7200 Batch Loss:     2.630145 Tokens per Sec:     7620, Lr: 0.000300\n",
      "2020-02-20 08:23:31,753 Epoch   7 Step:     7300 Batch Loss:     2.589347 Tokens per Sec:     7673, Lr: 0.000300\n",
      "2020-02-20 08:24:00,511 Epoch   7 Step:     7400 Batch Loss:     2.927073 Tokens per Sec:     7532, Lr: 0.000300\n",
      "2020-02-20 08:24:29,479 Epoch   7 Step:     7500 Batch Loss:     2.287795 Tokens per Sec:     7710, Lr: 0.000300\n",
      "2020-02-20 08:24:58,221 Epoch   7 Step:     7600 Batch Loss:     2.584190 Tokens per Sec:     7707, Lr: 0.000300\n",
      "2020-02-20 08:25:22,932 Epoch   7: total training loss 2664.20\n",
      "2020-02-20 08:25:22,933 EPOCH 8\n",
      "2020-02-20 08:25:27,446 Epoch   8 Step:     7700 Batch Loss:     2.356482 Tokens per Sec:     7481, Lr: 0.000300\n",
      "2020-02-20 08:25:56,100 Epoch   8 Step:     7800 Batch Loss:     1.703457 Tokens per Sec:     7641, Lr: 0.000300\n",
      "2020-02-20 08:26:24,929 Epoch   8 Step:     7900 Batch Loss:     1.696029 Tokens per Sec:     7637, Lr: 0.000300\n",
      "2020-02-20 08:26:53,887 Epoch   8 Step:     8000 Batch Loss:     2.438761 Tokens per Sec:     7708, Lr: 0.000300\n",
      "2020-02-20 08:28:24,734 Hooray! New best validation result [ppl]!\n",
      "2020-02-20 08:28:24,734 Saving new checkpoint.\n",
      "2020-02-20 08:28:24,953 Example #0\n",
      "2020-02-20 08:28:24,953 \tSource:     Researchers at the University of Copenhagen tracked almost 10,000 middle - aged people over an 11 - year period and found that participants who frequently argued with someone close to them were far more likely to die prematurely than those who seldom had conflicts .\n",
      "2020-02-20 08:28:24,953 \tReference:  Athuthuria yunivasĩtĩ - inĩ ya Copenhagen , nĩ maathuthuririe andũ hakuhĩ 10,000 a mĩaka ĩyo kwa ihinda rĩa mĩaka 11 na makĩona atĩ arĩa maakararanagia na mũndũ wa famĩlĩ kaingĩ , haakoragwo na ũhotekeku mũnene atĩ megũkua tene gũkĩra arĩa mataakoragwo na ngarari kaingĩ .\n",
      "2020-02-20 08:28:24,953 \tHypothesis: Athuthuria a ndini cia Kĩama Kĩrĩa Gĩtongoragia nĩ maacaragia andũ aingĩ mũno mwaka - inĩ wa 10,000 , na andũ aingĩ arĩa maarĩ na ũmenyeru , no kũhoteke nĩ maamenyire atĩ nĩ meekĩte ngoro mũno .\n",
      "2020-02-20 08:28:24,954 Example #1\n",
      "2020-02-20 08:28:24,954 \tSource:     Yes .\n",
      "2020-02-20 08:28:24,954 \tReference:  Ĩĩ .\n",
      "2020-02-20 08:28:24,954 \tHypothesis: Ĩĩ .\n",
      "2020-02-20 08:28:24,954 Example #2\n",
      "2020-02-20 08:28:24,954 \tSource:     He enjoyed special intimacy with Jehovah , who enabled him to display “ awesome power ” as he led the Israelites to the Promised Land . ​ — Deut .\n",
      "2020-02-20 08:28:24,954 \tReference:  Aarĩ na ũkuruhanu wa hakuhĩ na Jehova , na nĩ aamũheire ũhoti wa gwĩka “ maũndũ manene ma kũguoyohania ” rĩrĩa aatongoragia Aisiraeli mathiĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 08:28:24,954 \tHypothesis: Nĩ aakenire mũno nĩ ũndũ wa mwanya mũno , ũrĩa Jehova aamwĩrire “ hinya wa kũheo hinya ” nĩ ũndũ wa Aisiraeli , na nĩ aamwĩrire atĩ nĩ eetĩkĩrire . — Gũcok .\n",
      "2020-02-20 08:28:24,954 Example #3\n",
      "2020-02-20 08:28:24,955 \tSource:     HISTORY : ATHEIST\n",
      "2020-02-20 08:28:24,955 \tReference:  HISTORĨ : EETĨKĨTIE GŨTIRĨ NGAI\n",
      "2020-02-20 08:28:24,955 \tHypothesis: HIHI : MŨNDŨ\n",
      "2020-02-20 08:28:24,955 Validation result at epoch   8, step     8000: bleu:  13.29, loss: 58300.2930, ppl:   9.9832, duration: 91.0669s\n",
      "2020-02-20 08:28:53,679 Epoch   8 Step:     8100 Batch Loss:     2.217531 Tokens per Sec:     7678, Lr: 0.000300\n",
      "2020-02-20 08:29:21,960 Epoch   8 Step:     8200 Batch Loss:     2.531309 Tokens per Sec:     7547, Lr: 0.000300\n",
      "2020-02-20 08:29:50,620 Epoch   8 Step:     8300 Batch Loss:     2.193943 Tokens per Sec:     7643, Lr: 0.000300\n",
      "2020-02-20 08:30:19,274 Epoch   8 Step:     8400 Batch Loss:     2.727993 Tokens per Sec:     7610, Lr: 0.000300\n",
      "2020-02-20 08:30:48,181 Epoch   8 Step:     8500 Batch Loss:     2.947153 Tokens per Sec:     7682, Lr: 0.000300\n",
      "2020-02-20 08:31:17,010 Epoch   8 Step:     8600 Batch Loss:     2.230273 Tokens per Sec:     7612, Lr: 0.000300\n",
      "2020-02-20 08:31:45,839 Epoch   8 Step:     8700 Batch Loss:     2.263208 Tokens per Sec:     7545, Lr: 0.000300\n",
      "2020-02-20 08:32:09,655 Epoch   8: total training loss 2561.89\n",
      "2020-02-20 08:32:09,655 EPOCH 9\n",
      "2020-02-20 08:32:14,970 Epoch   9 Step:     8800 Batch Loss:     1.997649 Tokens per Sec:     7803, Lr: 0.000300\n",
      "2020-02-20 08:32:43,477 Epoch   9 Step:     8900 Batch Loss:     2.652678 Tokens per Sec:     7554, Lr: 0.000300\n",
      "2020-02-20 08:33:12,208 Epoch   9 Step:     9000 Batch Loss:     2.664539 Tokens per Sec:     7606, Lr: 0.000300\n",
      "2020-02-20 08:34:42,945 Hooray! New best validation result [ppl]!\n",
      "2020-02-20 08:34:42,945 Saving new checkpoint.\n",
      "2020-02-20 08:34:43,168 Example #0\n",
      "2020-02-20 08:34:43,168 \tSource:     Researchers at the University of Copenhagen tracked almost 10,000 middle - aged people over an 11 - year period and found that participants who frequently argued with someone close to them were far more likely to die prematurely than those who seldom had conflicts .\n",
      "2020-02-20 08:34:43,168 \tReference:  Athuthuria yunivasĩtĩ - inĩ ya Copenhagen , nĩ maathuthuririe andũ hakuhĩ 10,000 a mĩaka ĩyo kwa ihinda rĩa mĩaka 11 na makĩona atĩ arĩa maakararanagia na mũndũ wa famĩlĩ kaingĩ , haakoragwo na ũhotekeku mũnene atĩ megũkua tene gũkĩra arĩa mataakoragwo na ngarari kaingĩ .\n",
      "2020-02-20 08:34:43,168 \tHypothesis: Ahunjia a ndini ya U.S . nĩ maahutirie andũ aingĩ mũno , na makĩona atĩ andũ aingĩ nĩ maahutirie andũ aingĩ mũno , na nĩ maanyarirũo nĩ andũ arĩa maikaraga makĩenda kũmakania na andũ arĩa angĩ .\n",
      "2020-02-20 08:34:43,169 Example #1\n",
      "2020-02-20 08:34:43,169 \tSource:     Yes .\n",
      "2020-02-20 08:34:43,169 \tReference:  Ĩĩ .\n",
      "2020-02-20 08:34:43,169 \tHypothesis: Ĩĩ .\n",
      "2020-02-20 08:34:43,169 Example #2\n",
      "2020-02-20 08:34:43,169 \tSource:     He enjoyed special intimacy with Jehovah , who enabled him to display “ awesome power ” as he led the Israelites to the Promised Land . ​ — Deut .\n",
      "2020-02-20 08:34:43,169 \tReference:  Aarĩ na ũkuruhanu wa hakuhĩ na Jehova , na nĩ aamũheire ũhoti wa gwĩka “ maũndũ manene ma kũguoyohania ” rĩrĩa aatongoragia Aisiraeli mathiĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 08:34:43,169 \tHypothesis: Nĩ aakenire mũno nĩ ũndũ wa gũkorũo na ũcayanĩri , ũrĩa Jehova aamwĩrire “ hinya ” o ta ũrĩa Aisiraeli maamuonirie Aisiraeli na makĩmũhe . — Gũcok .\n",
      "2020-02-20 08:34:43,169 Example #3\n",
      "2020-02-20 08:34:43,170 \tSource:     HISTORY : ATHEIST\n",
      "2020-02-20 08:34:43,170 \tReference:  HISTORĨ : EETĨKĨTIE GŨTIRĨ NGAI\n",
      "2020-02-20 08:34:43,170 \tHypothesis: HISTI : NĨ WĨKĨKĨRĨRĨRĨRĨRĨRĨRĨRĨRĨRIA\n",
      "2020-02-20 08:34:43,170 Validation result at epoch   9, step     9000: bleu:  13.84, loss: 56454.3359, ppl:   9.2818, duration: 90.9613s\n",
      "2020-02-20 08:35:11,766 Epoch   9 Step:     9100 Batch Loss:     2.356254 Tokens per Sec:     7715, Lr: 0.000300\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2020-02-20 08:35:40,464 Epoch   9 Step:     9200 Batch Loss:     2.365337 Tokens per Sec:     7652, Lr: 0.000300\n",
      "2020-02-20 08:36:09,367 Epoch   9 Step:     9300 Batch Loss:     2.179333 Tokens per Sec:     7768, Lr: 0.000300\n",
      "2020-02-20 08:36:37,814 Epoch   9 Step:     9400 Batch Loss:     2.216345 Tokens per Sec:     7620, Lr: 0.000300\n",
      "2020-02-20 08:37:06,686 Epoch   9 Step:     9500 Batch Loss:     2.180648 Tokens per Sec:     7720, Lr: 0.000300\n",
      "2020-02-20 08:37:35,593 Epoch   9 Step:     9600 Batch Loss:     2.269558 Tokens per Sec:     7566, Lr: 0.000300\n",
      "2020-02-20 08:38:04,153 Epoch   9 Step:     9700 Batch Loss:     2.119459 Tokens per Sec:     7605, Lr: 0.000300\n",
      "2020-02-20 08:38:32,809 Epoch   9 Step:     9800 Batch Loss:     2.942793 Tokens per Sec:     7636, Lr: 0.000300\n",
      "2020-02-20 08:38:56,058 Epoch   9: total training loss 2479.96\n",
      "2020-02-20 08:38:56,059 EPOCH 10\n",
      "2020-02-20 08:39:01,856 Epoch  10 Step:     9900 Batch Loss:     2.136981 Tokens per Sec:     7099, Lr: 0.000300\n",
      "2020-02-20 08:39:30,763 Epoch  10 Step:    10000 Batch Loss:     1.932028 Tokens per Sec:     7686, Lr: 0.000300\n",
      "2020-02-20 08:41:01,582 Hooray! New best validation result [ppl]!\n",
      "2020-02-20 08:41:01,583 Saving new checkpoint.\n",
      "2020-02-20 08:41:01,800 Example #0\n",
      "2020-02-20 08:41:01,800 \tSource:     Researchers at the University of Copenhagen tracked almost 10,000 middle - aged people over an 11 - year period and found that participants who frequently argued with someone close to them were far more likely to die prematurely than those who seldom had conflicts .\n",
      "2020-02-20 08:41:01,800 \tReference:  Athuthuria yunivasĩtĩ - inĩ ya Copenhagen , nĩ maathuthuririe andũ hakuhĩ 10,000 a mĩaka ĩyo kwa ihinda rĩa mĩaka 11 na makĩona atĩ arĩa maakararanagia na mũndũ wa famĩlĩ kaingĩ , haakoragwo na ũhotekeku mũnene atĩ megũkua tene gũkĩra arĩa mataakoragwo na ngarari kaingĩ .\n",
      "2020-02-20 08:41:01,800 \tHypothesis: Ahunjia a ndini arĩa maarutaga wĩra thĩinĩ wa Amerika ya U.S . nĩ maahutirie andũ aingĩ mũno arĩa maarĩ na mĩaka 1,000 mĩhĩtũku , na andũ aingĩ arĩa maahutagia mũno nĩ maatuĩkĩte andũ aingĩ mũno gũkĩra arĩa maatuĩkĩte njera .\n",
      "2020-02-20 08:41:01,801 Example #1\n",
      "2020-02-20 08:41:01,801 \tSource:     Yes .\n",
      "2020-02-20 08:41:01,801 \tReference:  Ĩĩ .\n",
      "2020-02-20 08:41:01,801 \tHypothesis: Hatarĩ nganja , nĩ kũrĩ .\n",
      "2020-02-20 08:41:01,801 Example #2\n",
      "2020-02-20 08:41:01,801 \tSource:     He enjoyed special intimacy with Jehovah , who enabled him to display “ awesome power ” as he led the Israelites to the Promised Land . ​ — Deut .\n",
      "2020-02-20 08:41:01,801 \tReference:  Aarĩ na ũkuruhanu wa hakuhĩ na Jehova , na nĩ aamũheire ũhoti wa gwĩka “ maũndũ manene ma kũguoyohania ” rĩrĩa aatongoragia Aisiraeli mathiĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 08:41:01,801 \tHypothesis: Nĩ aakenire mũno nĩ ũndũ wa gũkorũo na tha cia mwanya , ũrĩa wamũteithirie ‘ ũhoti wa kũguna ’ ũrĩa aaheire Aisiraeli . — Gũcok .\n",
      "2020-02-20 08:41:01,801 Example #3\n",
      "2020-02-20 08:41:01,802 \tSource:     HISTORY : ATHEIST\n",
      "2020-02-20 08:41:01,802 \tReference:  HISTORĨ : EETĨKĨTIE GŨTIRĨ NGAI\n",
      "2020-02-20 08:41:01,802 \tHypothesis: HISTORĨ : MŨNDŨ WA MŨNDŨ\n",
      "2020-02-20 08:41:01,802 Validation result at epoch  10, step    10000: bleu:  14.94, loss: 54946.1016, ppl:   8.7454, duration: 91.0378s\n",
      "2020-02-20 08:41:30,337 Epoch  10 Step:    10100 Batch Loss:     2.203784 Tokens per Sec:     7552, Lr: 0.000300\n",
      "2020-02-20 08:41:59,020 Epoch  10 Step:    10200 Batch Loss:     2.008966 Tokens per Sec:     7606, Lr: 0.000300\n",
      "2020-02-20 08:42:27,877 Epoch  10 Step:    10300 Batch Loss:     2.561419 Tokens per Sec:     7678, Lr: 0.000300\n",
      "2020-02-20 08:42:56,899 Epoch  10 Step:    10400 Batch Loss:     1.821608 Tokens per Sec:     7584, Lr: 0.000300\n",
      "2020-02-20 08:43:25,665 Epoch  10 Step:    10500 Batch Loss:     2.253025 Tokens per Sec:     7725, Lr: 0.000300\n",
      "2020-02-20 08:43:54,527 Epoch  10 Step:    10600 Batch Loss:     2.005208 Tokens per Sec:     7653, Lr: 0.000300\n",
      "2020-02-20 08:44:23,453 Epoch  10 Step:    10700 Batch Loss:     2.419311 Tokens per Sec:     7659, Lr: 0.000300\n",
      "2020-02-20 08:44:52,086 Epoch  10 Step:    10800 Batch Loss:     2.266176 Tokens per Sec:     7521, Lr: 0.000300\n",
      "2020-02-20 08:45:20,943 Epoch  10 Step:    10900 Batch Loss:     1.995486 Tokens per Sec:     7627, Lr: 0.000300\n",
      "2020-02-20 08:45:44,100 Epoch  10: total training loss 2414.43\n",
      "2020-02-20 08:45:44,100 EPOCH 11\n",
      "2020-02-20 08:45:49,699 Epoch  11 Step:    11000 Batch Loss:     2.083320 Tokens per Sec:     7524, Lr: 0.000300\n",
      "2020-02-20 08:47:20,433 Hooray! New best validation result [ppl]!\n",
      "2020-02-20 08:47:20,433 Saving new checkpoint.\n",
      "2020-02-20 08:47:20,660 Example #0\n",
      "2020-02-20 08:47:20,660 \tSource:     Researchers at the University of Copenhagen tracked almost 10,000 middle - aged people over an 11 - year period and found that participants who frequently argued with someone close to them were far more likely to die prematurely than those who seldom had conflicts .\n",
      "2020-02-20 08:47:20,661 \tReference:  Athuthuria yunivasĩtĩ - inĩ ya Copenhagen , nĩ maathuthuririe andũ hakuhĩ 10,000 a mĩaka ĩyo kwa ihinda rĩa mĩaka 11 na makĩona atĩ arĩa maakararanagia na mũndũ wa famĩlĩ kaingĩ , haakoragwo na ũhotekeku mũnene atĩ megũkua tene gũkĩra arĩa mataakoragwo na ngarari kaingĩ .\n",
      "2020-02-20 08:47:20,661 \tHypothesis: Ahunjia a ndini arĩa maanyitanagĩra na andũ 10,000 arĩa maarĩ na ũmenyeru , nĩ maanyitire ũgeni andũ aingĩ mũno arĩa maanyitanagĩra nao , na nĩ maanyitire ũgeni o na gũtuĩka nĩ maanyaririrũo nĩ andũ aingĩ mũno .\n",
      "2020-02-20 08:47:20,661 Example #1\n",
      "2020-02-20 08:47:20,661 \tSource:     Yes .\n",
      "2020-02-20 08:47:20,661 \tReference:  Ĩĩ .\n",
      "2020-02-20 08:47:20,661 \tHypothesis: Ĩĩ .\n",
      "2020-02-20 08:47:20,661 Example #2\n",
      "2020-02-20 08:47:20,661 \tSource:     He enjoyed special intimacy with Jehovah , who enabled him to display “ awesome power ” as he led the Israelites to the Promised Land . ​ — Deut .\n",
      "2020-02-20 08:47:20,661 \tReference:  Aarĩ na ũkuruhanu wa hakuhĩ na Jehova , na nĩ aamũheire ũhoti wa gwĩka “ maũndũ manene ma kũguoyohania ” rĩrĩa aatongoragia Aisiraeli mathiĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 08:47:20,662 \tHypothesis: Nĩ aakenire mũno nĩ Jehova , ũrĩa wamũheire “ hinya ” wa kũmũtungatĩra Aisiraeli , o ta ũrĩa Aisiraeli maamenirie na agatwara . — Gũcok .\n",
      "2020-02-20 08:47:20,662 Example #3\n",
      "2020-02-20 08:47:20,662 \tSource:     HISTORY : ATHEIST\n",
      "2020-02-20 08:47:20,662 \tReference:  HISTORĨ : EETĨKĨTIE GŨTIRĨ NGAI\n",
      "2020-02-20 08:47:20,662 \tHypothesis: HISTORĨ : MŨTHAMWA\n",
      "2020-02-20 08:47:20,662 Validation result at epoch  11, step    11000: bleu:  15.73, loss: 53487.9922, ppl:   8.2563, duration: 90.9625s\n",
      "2020-02-20 08:47:49,665 Epoch  11 Step:    11100 Batch Loss:     2.348814 Tokens per Sec:     7665, Lr: 0.000300\n",
      "2020-02-20 08:48:18,330 Epoch  11 Step:    11200 Batch Loss:     2.741960 Tokens per Sec:     7695, Lr: 0.000300\n",
      "2020-02-20 08:48:47,204 Epoch  11 Step:    11300 Batch Loss:     2.578320 Tokens per Sec:     7649, Lr: 0.000300\n",
      "2020-02-20 08:49:15,891 Epoch  11 Step:    11400 Batch Loss:     2.703610 Tokens per Sec:     7607, Lr: 0.000300\n",
      "2020-02-20 08:49:44,426 Epoch  11 Step:    11500 Batch Loss:     1.739605 Tokens per Sec:     7697, Lr: 0.000300\n",
      "2020-02-20 08:50:13,202 Epoch  11 Step:    11600 Batch Loss:     1.945136 Tokens per Sec:     7618, Lr: 0.000300\n",
      "2020-02-20 08:50:42,062 Epoch  11 Step:    11700 Batch Loss:     2.214830 Tokens per Sec:     7605, Lr: 0.000300\n",
      "2020-02-20 08:51:10,705 Epoch  11 Step:    11800 Batch Loss:     2.239208 Tokens per Sec:     7513, Lr: 0.000300\n",
      "2020-02-20 08:51:39,166 Epoch  11 Step:    11900 Batch Loss:     2.213643 Tokens per Sec:     7572, Lr: 0.000300\n",
      "2020-02-20 08:52:07,861 Epoch  11 Step:    12000 Batch Loss:     2.215874 Tokens per Sec:     7743, Lr: 0.000300\n",
      "2020-02-20 08:53:38,618 Hooray! New best validation result [ppl]!\n",
      "2020-02-20 08:53:38,619 Saving new checkpoint.\n",
      "2020-02-20 08:53:38,840 Example #0\n",
      "2020-02-20 08:53:38,840 \tSource:     Researchers at the University of Copenhagen tracked almost 10,000 middle - aged people over an 11 - year period and found that participants who frequently argued with someone close to them were far more likely to die prematurely than those who seldom had conflicts .\n",
      "2020-02-20 08:53:38,840 \tReference:  Athuthuria yunivasĩtĩ - inĩ ya Copenhagen , nĩ maathuthuririe andũ hakuhĩ 10,000 a mĩaka ĩyo kwa ihinda rĩa mĩaka 11 na makĩona atĩ arĩa maakararanagia na mũndũ wa famĩlĩ kaingĩ , haakoragwo na ũhotekeku mũnene atĩ megũkua tene gũkĩra arĩa mataakoragwo na ngarari kaingĩ .\n",
      "2020-02-20 08:53:38,840 \tHypothesis: Athuthuria a sayansi ya U.S . nĩ maaharagire andũ 10,000 arĩa maarĩ na mĩaka 11 na andũ aingĩ arĩa maarĩ na ũmenyeru , na nĩ maamomĩrĩirie mũno nĩ ũndũ wa andũ ta acio angĩ ta acio maahurũkire mũno .\n",
      "2020-02-20 08:53:38,840 Example #1\n",
      "2020-02-20 08:53:38,840 \tSource:     Yes .\n",
      "2020-02-20 08:53:38,840 \tReference:  Ĩĩ .\n",
      "2020-02-20 08:53:38,841 \tHypothesis: Ĩĩ .\n",
      "2020-02-20 08:53:38,841 Example #2\n",
      "2020-02-20 08:53:38,841 \tSource:     He enjoyed special intimacy with Jehovah , who enabled him to display “ awesome power ” as he led the Israelites to the Promised Land . ​ — Deut .\n",
      "2020-02-20 08:53:38,841 \tReference:  Aarĩ na ũkuruhanu wa hakuhĩ na Jehova , na nĩ aamũheire ũhoti wa gwĩka “ maũndũ manene ma kũguoyohania ” rĩrĩa aatongoragia Aisiraeli mathiĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 08:53:38,841 \tHypothesis: Nĩ aakenagĩra mũno nĩ Jehova , ũrĩa wamũteithirie ‘ kuona hinya ’ ũrĩa aaheire Aisiraeli na akĩmũhe Aisiraeli . — Gũcok .\n",
      "2020-02-20 08:53:38,841 Example #3\n",
      "2020-02-20 08:53:38,841 \tSource:     HISTORY : ATHEIST\n",
      "2020-02-20 08:53:38,841 \tReference:  HISTORĨ : EETĨKĨTIE GŨTIRĨ NGAI\n",
      "2020-02-20 08:53:38,841 \tHypothesis: HISTI : MŨNDŨ WA\n",
      "2020-02-20 08:53:38,842 Validation result at epoch  11, step    12000: bleu:  16.23, loss: 52413.1172, ppl:   7.9134, duration: 90.9803s\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2020-02-20 08:54:02,102 Epoch  11: total training loss 2343.63\n",
      "2020-02-20 08:54:02,102 EPOCH 12\n",
      "2020-02-20 08:54:07,781 Epoch  12 Step:    12100 Batch Loss:     2.053180 Tokens per Sec:     7597, Lr: 0.000300\n",
      "2020-02-20 08:54:36,634 Epoch  12 Step:    12200 Batch Loss:     2.118457 Tokens per Sec:     7671, Lr: 0.000300\n",
      "2020-02-20 08:55:05,379 Epoch  12 Step:    12300 Batch Loss:     1.909680 Tokens per Sec:     7592, Lr: 0.000300\n",
      "2020-02-20 08:55:34,427 Epoch  12 Step:    12400 Batch Loss:     2.349315 Tokens per Sec:     7755, Lr: 0.000300\n",
      "2020-02-20 08:56:03,488 Epoch  12 Step:    12500 Batch Loss:     2.389056 Tokens per Sec:     7745, Lr: 0.000300\n",
      "2020-02-20 08:56:32,238 Epoch  12 Step:    12600 Batch Loss:     2.330261 Tokens per Sec:     7754, Lr: 0.000300\n",
      "2020-02-20 08:57:01,118 Epoch  12 Step:    12700 Batch Loss:     1.863147 Tokens per Sec:     7565, Lr: 0.000300\n",
      "2020-02-20 08:57:29,831 Epoch  12 Step:    12800 Batch Loss:     1.909142 Tokens per Sec:     7510, Lr: 0.000300\n",
      "2020-02-20 08:57:58,530 Epoch  12 Step:    12900 Batch Loss:     1.856022 Tokens per Sec:     7681, Lr: 0.000300\n",
      "2020-02-20 08:58:26,981 Epoch  12 Step:    13000 Batch Loss:     1.775300 Tokens per Sec:     7482, Lr: 0.000300\n",
      "2020-02-20 08:59:57,810 Hooray! New best validation result [ppl]!\n",
      "2020-02-20 08:59:57,811 Saving new checkpoint.\n",
      "2020-02-20 08:59:58,036 Example #0\n",
      "2020-02-20 08:59:58,036 \tSource:     Researchers at the University of Copenhagen tracked almost 10,000 middle - aged people over an 11 - year period and found that participants who frequently argued with someone close to them were far more likely to die prematurely than those who seldom had conflicts .\n",
      "2020-02-20 08:59:58,036 \tReference:  Athuthuria yunivasĩtĩ - inĩ ya Copenhagen , nĩ maathuthuririe andũ hakuhĩ 10,000 a mĩaka ĩyo kwa ihinda rĩa mĩaka 11 na makĩona atĩ arĩa maakararanagia na mũndũ wa famĩlĩ kaingĩ , haakoragwo na ũhotekeku mũnene atĩ megũkua tene gũkĩra arĩa mataakoragwo na ngarari kaingĩ .\n",
      "2020-02-20 08:59:58,036 \tHypothesis: Athuthuria a ndini ya Amerika ya U.S . nĩ maanyitirũo ũgeni andũ 10,000 arĩa maahinyanĩtio na andũ 10,000 arĩa maikaraga meiguĩte , na nĩ maanyaririrũo nĩ andũ arĩa maikaraga meendaga kũmakania , na nĩ ũndũ ũcio nĩ maametigagĩra kũmahe mũndũ ũcio .\n",
      "2020-02-20 08:59:58,036 Example #1\n",
      "2020-02-20 08:59:58,037 \tSource:     Yes .\n",
      "2020-02-20 08:59:58,037 \tReference:  Ĩĩ .\n",
      "2020-02-20 08:59:58,037 \tHypothesis: Ĩĩ nĩ ya ma .\n",
      "2020-02-20 08:59:58,037 Example #2\n",
      "2020-02-20 08:59:58,037 \tSource:     He enjoyed special intimacy with Jehovah , who enabled him to display “ awesome power ” as he led the Israelites to the Promised Land . ​ — Deut .\n",
      "2020-02-20 08:59:58,037 \tReference:  Aarĩ na ũkuruhanu wa hakuhĩ na Jehova , na nĩ aamũheire ũhoti wa gwĩka “ maũndũ manene ma kũguoyohania ” rĩrĩa aatongoragia Aisiraeli mathiĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 08:59:58,037 \tHypothesis: Nĩ aakenire mũno nĩ Jehova , ũrĩa wamũteithirie ‘ kũiguithania wega ’ arĩ we wamũteithirie Aisiraeli arĩa maameheire Aisiraeli . — Gũcok .\n",
      "2020-02-20 08:59:58,037 Example #3\n",
      "2020-02-20 08:59:58,037 \tSource:     HISTORY : ATHEIST\n",
      "2020-02-20 08:59:58,037 \tReference:  HISTORĨ : EETĨKĨTIE GŨTIRĨ NGAI\n",
      "2020-02-20 08:59:58,038 \tHypothesis: HISTORĨ : MŨTŨNDŨ WA MŨTŨKORO\n",
      "2020-02-20 08:59:58,038 Validation result at epoch  12, step    13000: bleu:  17.24, loss: 51187.1445, ppl:   7.5396, duration: 91.0555s\n",
      "2020-02-20 09:00:26,917 Epoch  12 Step:    13100 Batch Loss:     2.712152 Tokens per Sec:     7642, Lr: 0.000300\n",
      "2020-02-20 09:00:48,941 Epoch  12: total training loss 2270.78\n",
      "2020-02-20 09:00:48,941 EPOCH 13\n",
      "2020-02-20 09:00:55,663 Epoch  13 Step:    13200 Batch Loss:     2.075423 Tokens per Sec:     7437, Lr: 0.000300\n",
      "2020-02-20 09:01:24,646 Epoch  13 Step:    13300 Batch Loss:     1.580067 Tokens per Sec:     7692, Lr: 0.000300\n",
      "2020-02-20 09:01:53,104 Epoch  13 Step:    13400 Batch Loss:     1.696493 Tokens per Sec:     7555, Lr: 0.000300\n",
      "2020-02-20 09:02:21,739 Epoch  13 Step:    13500 Batch Loss:     1.939544 Tokens per Sec:     7557, Lr: 0.000300\n",
      "2020-02-20 09:02:50,234 Epoch  13 Step:    13600 Batch Loss:     1.964505 Tokens per Sec:     7495, Lr: 0.000300\n",
      "2020-02-20 09:03:19,017 Epoch  13 Step:    13700 Batch Loss:     1.995048 Tokens per Sec:     7612, Lr: 0.000300\n",
      "2020-02-20 09:03:47,836 Epoch  13 Step:    13800 Batch Loss:     2.006960 Tokens per Sec:     7682, Lr: 0.000300\n",
      "2020-02-20 09:04:16,439 Epoch  13 Step:    13900 Batch Loss:     1.876329 Tokens per Sec:     7609, Lr: 0.000300\n",
      "2020-02-20 09:04:45,047 Epoch  13 Step:    14000 Batch Loss:     1.815590 Tokens per Sec:     7685, Lr: 0.000300\n",
      "2020-02-20 09:06:15,814 Hooray! New best validation result [ppl]!\n",
      "2020-02-20 09:06:15,814 Saving new checkpoint.\n",
      "2020-02-20 09:06:16,032 Example #0\n",
      "2020-02-20 09:06:16,032 \tSource:     Researchers at the University of Copenhagen tracked almost 10,000 middle - aged people over an 11 - year period and found that participants who frequently argued with someone close to them were far more likely to die prematurely than those who seldom had conflicts .\n",
      "2020-02-20 09:06:16,032 \tReference:  Athuthuria yunivasĩtĩ - inĩ ya Copenhagen , nĩ maathuthuririe andũ hakuhĩ 10,000 a mĩaka ĩyo kwa ihinda rĩa mĩaka 11 na makĩona atĩ arĩa maakararanagia na mũndũ wa famĩlĩ kaingĩ , haakoragwo na ũhotekeku mũnene atĩ megũkua tene gũkĩra arĩa mataakoragwo na ngarari kaingĩ .\n",
      "2020-02-20 09:06:16,032 \tHypothesis: Athuthuria a bũrũri wa Ngeretha a bũrũri - inĩ wa Amerika ( US ) nĩ maanyitire ũgeni andũ 10,000 arĩa maanyitanagĩra nao mĩaka 11 na nuthu , na nĩ maanyitire ũgeni andũ arĩa maamatuĩkĩte a itua rĩa gũkũngũĩra andũ arĩa maikaraga marĩ na hinya gũkĩra arĩa maakuĩte .\n",
      "2020-02-20 09:06:16,032 Example #1\n",
      "2020-02-20 09:06:16,033 \tSource:     Yes .\n",
      "2020-02-20 09:06:16,033 \tReference:  Ĩĩ .\n",
      "2020-02-20 09:06:16,033 \tHypothesis: Ĩĩ nĩ ya bata .\n",
      "2020-02-20 09:06:16,033 Example #2\n",
      "2020-02-20 09:06:16,033 \tSource:     He enjoyed special intimacy with Jehovah , who enabled him to display “ awesome power ” as he led the Israelites to the Promised Land . ​ — Deut .\n",
      "2020-02-20 09:06:16,033 \tReference:  Aarĩ na ũkuruhanu wa hakuhĩ na Jehova , na nĩ aamũheire ũhoti wa gwĩka “ maũndũ manene ma kũguoyohania ” rĩrĩa aatongoragia Aisiraeli mathiĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 09:06:16,033 \tHypothesis: Nĩ aakenire mũno nĩ gũkorũo arĩ na tha cia mwanya cia Jehova , ũrĩa wamũteithirie kũmũtongoria “ hinya ” ũrĩa aaheire Aisiraeli kuuma kũrĩ Aisiraeli . — Gũcok .\n",
      "2020-02-20 09:06:16,033 Example #3\n",
      "2020-02-20 09:06:16,033 \tSource:     HISTORY : ATHEIST\n",
      "2020-02-20 09:06:16,034 \tReference:  HISTORĨ : EETĨKĨTIE GŨTIRĨ NGAI\n",
      "2020-02-20 09:06:16,034 \tHypothesis: HISTORĨ : MŨNDŨ WA MŨTHAMA\n",
      "2020-02-20 09:06:16,034 Validation result at epoch  13, step    14000: bleu:  17.69, loss: 50359.1641, ppl:   7.2972, duration: 90.9853s\n",
      "2020-02-20 09:06:44,749 Epoch  13 Step:    14100 Batch Loss:     2.142448 Tokens per Sec:     7711, Lr: 0.000300\n",
      "2020-02-20 09:07:13,131 Epoch  13 Step:    14200 Batch Loss:     2.089159 Tokens per Sec:     7544, Lr: 0.000300\n",
      "2020-02-20 09:07:36,342 Epoch  13: total training loss 2239.69\n",
      "2020-02-20 09:07:36,342 EPOCH 14\n",
      "2020-02-20 09:07:42,197 Epoch  14 Step:    14300 Batch Loss:     1.387570 Tokens per Sec:     7479, Lr: 0.000300\n",
      "2020-02-20 09:08:10,906 Epoch  14 Step:    14400 Batch Loss:     2.090689 Tokens per Sec:     7632, Lr: 0.000300\n",
      "2020-02-20 09:08:39,308 Epoch  14 Step:    14500 Batch Loss:     1.952181 Tokens per Sec:     7601, Lr: 0.000300\n",
      "2020-02-20 09:09:08,190 Epoch  14 Step:    14600 Batch Loss:     2.223345 Tokens per Sec:     7666, Lr: 0.000300\n",
      "2020-02-20 09:09:37,035 Epoch  14 Step:    14700 Batch Loss:     1.848605 Tokens per Sec:     7654, Lr: 0.000300\n",
      "2020-02-20 09:10:05,840 Epoch  14 Step:    14800 Batch Loss:     2.215981 Tokens per Sec:     7729, Lr: 0.000300\n",
      "2020-02-20 09:10:34,537 Epoch  14 Step:    14900 Batch Loss:     1.726101 Tokens per Sec:     7479, Lr: 0.000300\n",
      "2020-02-20 09:11:03,695 Epoch  14 Step:    15000 Batch Loss:     2.336810 Tokens per Sec:     7791, Lr: 0.000300\n",
      "2020-02-20 09:12:34,447 Hooray! New best validation result [ppl]!\n",
      "2020-02-20 09:12:34,447 Saving new checkpoint.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2020-02-20 09:12:34,674 Example #0\n",
      "2020-02-20 09:12:34,674 \tSource:     Researchers at the University of Copenhagen tracked almost 10,000 middle - aged people over an 11 - year period and found that participants who frequently argued with someone close to them were far more likely to die prematurely than those who seldom had conflicts .\n",
      "2020-02-20 09:12:34,675 \tReference:  Athuthuria yunivasĩtĩ - inĩ ya Copenhagen , nĩ maathuthuririe andũ hakuhĩ 10,000 a mĩaka ĩyo kwa ihinda rĩa mĩaka 11 na makĩona atĩ arĩa maakararanagia na mũndũ wa famĩlĩ kaingĩ , haakoragwo na ũhotekeku mũnene atĩ megũkua tene gũkĩra arĩa mataakoragwo na ngarari kaingĩ .\n",
      "2020-02-20 09:12:34,675 \tHypothesis: Athuthuria a sayansi a Amerika ( US ) nĩ maaregire mũno na andũ 10,000 arĩa maikaraga mĩaka 11 na makĩona atĩ andũ arĩa maikaraga meiguĩte nĩ maanyariragwo nĩ mũndũ ũcio warĩ na ngoro njega gũkĩra ũrĩa maamonaga arĩ na nganja .\n",
      "2020-02-20 09:12:34,675 Example #1\n",
      "2020-02-20 09:12:34,675 \tSource:     Yes .\n",
      "2020-02-20 09:12:34,675 \tReference:  Ĩĩ .\n",
      "2020-02-20 09:12:34,675 \tHypothesis: Ĩĩ .\n",
      "2020-02-20 09:12:34,676 Example #2\n",
      "2020-02-20 09:12:34,676 \tSource:     He enjoyed special intimacy with Jehovah , who enabled him to display “ awesome power ” as he led the Israelites to the Promised Land . ​ — Deut .\n",
      "2020-02-20 09:12:34,676 \tReference:  Aarĩ na ũkuruhanu wa hakuhĩ na Jehova , na nĩ aamũheire ũhoti wa gwĩka “ maũndũ manene ma kũguoyohania ” rĩrĩa aatongoragia Aisiraeli mathiĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 09:12:34,676 \tHypothesis: Nĩ aakenagĩra mũno nĩ Jehova , ũrĩa wamũteithirie kũmũmenyeria “ hinya ” ta ũrĩa aaheire Aisiraeli Aisiraeli maamenete na akĩmũhe . — Gũcok .\n",
      "2020-02-20 09:12:34,676 Example #3\n",
      "2020-02-20 09:12:34,676 \tSource:     HISTORY : ATHEIST\n",
      "2020-02-20 09:12:34,677 \tReference:  HISTORĨ : EETĨKĨTIE GŨTIRĨ NGAI\n",
      "2020-02-20 09:12:34,677 \tHypothesis: HISTORĨ : MŨTHAYA\n",
      "2020-02-20 09:12:34,677 Validation result at epoch  14, step    15000: bleu:  17.98, loss: 49436.1328, ppl:   7.0362, duration: 90.9811s\n",
      "2020-02-20 09:13:03,311 Epoch  14 Step:    15100 Batch Loss:     1.086067 Tokens per Sec:     7596, Lr: 0.000300\n",
      "2020-02-20 09:13:31,717 Epoch  14 Step:    15200 Batch Loss:     2.106038 Tokens per Sec:     7511, Lr: 0.000300\n",
      "2020-02-20 09:14:00,558 Epoch  14 Step:    15300 Batch Loss:     1.905053 Tokens per Sec:     7756, Lr: 0.000300\n",
      "2020-02-20 09:14:23,390 Epoch  14: total training loss 2181.83\n",
      "2020-02-20 09:14:23,390 EPOCH 15\n",
      "2020-02-20 09:14:29,402 Epoch  15 Step:    15400 Batch Loss:     1.916626 Tokens per Sec:     7707, Lr: 0.000300\n",
      "2020-02-20 09:14:58,177 Epoch  15 Step:    15500 Batch Loss:     2.086712 Tokens per Sec:     7739, Lr: 0.000300\n",
      "2020-02-20 09:15:26,412 Epoch  15 Step:    15600 Batch Loss:     2.135949 Tokens per Sec:     7551, Lr: 0.000300\n",
      "2020-02-20 09:15:55,243 Epoch  15 Step:    15700 Batch Loss:     2.031682 Tokens per Sec:     7695, Lr: 0.000300\n",
      "2020-02-20 09:16:23,787 Epoch  15 Step:    15800 Batch Loss:     2.066107 Tokens per Sec:     7684, Lr: 0.000300\n",
      "2020-02-20 09:16:52,324 Epoch  15 Step:    15900 Batch Loss:     1.943845 Tokens per Sec:     7621, Lr: 0.000300\n",
      "2020-02-20 09:17:21,231 Epoch  15 Step:    16000 Batch Loss:     2.085521 Tokens per Sec:     7758, Lr: 0.000300\n",
      "2020-02-20 09:18:51,973 Hooray! New best validation result [ppl]!\n",
      "2020-02-20 09:18:51,973 Saving new checkpoint.\n",
      "2020-02-20 09:18:52,190 Example #0\n",
      "2020-02-20 09:18:52,190 \tSource:     Researchers at the University of Copenhagen tracked almost 10,000 middle - aged people over an 11 - year period and found that participants who frequently argued with someone close to them were far more likely to die prematurely than those who seldom had conflicts .\n",
      "2020-02-20 09:18:52,190 \tReference:  Athuthuria yunivasĩtĩ - inĩ ya Copenhagen , nĩ maathuthuririe andũ hakuhĩ 10,000 a mĩaka ĩyo kwa ihinda rĩa mĩaka 11 na makĩona atĩ arĩa maakararanagia na mũndũ wa famĩlĩ kaingĩ , haakoragwo na ũhotekeku mũnene atĩ megũkua tene gũkĩra arĩa mataakoragwo na ngarari kaingĩ .\n",
      "2020-02-20 09:18:52,190 \tHypothesis: Athuthuria a sayansi a bũrũri wa Ngeretha , nĩ maanyitire ũgeni andũ 10,000 arĩa maaikaraga mĩaka 11 na makĩona atĩ andũ arĩa maikaraga meiguĩte nĩ maanyaririrũo mũno nĩ andũ arĩa maakuire ngoro mũno gũkĩra arĩa maakuire .\n",
      "2020-02-20 09:18:52,190 Example #1\n",
      "2020-02-20 09:18:52,191 \tSource:     Yes .\n",
      "2020-02-20 09:18:52,191 \tReference:  Ĩĩ .\n",
      "2020-02-20 09:18:52,191 \tHypothesis: Ĩĩ .\n",
      "2020-02-20 09:18:52,191 Example #2\n",
      "2020-02-20 09:18:52,191 \tSource:     He enjoyed special intimacy with Jehovah , who enabled him to display “ awesome power ” as he led the Israelites to the Promised Land . ​ — Deut .\n",
      "2020-02-20 09:18:52,191 \tReference:  Aarĩ na ũkuruhanu wa hakuhĩ na Jehova , na nĩ aamũheire ũhoti wa gwĩka “ maũndũ manene ma kũguoyohania ” rĩrĩa aatongoragia Aisiraeli mathiĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 09:18:52,191 \tHypothesis: Nĩ aakenire mũno nĩ Jehova , ũrĩa wamũteithirie kuonania “ hinya ” ũrĩa aaheire Aisiraeli wĩra - inĩ wa kũheenereria Aisiraeli . — Gũcok .\n",
      "2020-02-20 09:18:52,191 Example #3\n",
      "2020-02-20 09:18:52,191 \tSource:     HISTORY : ATHEIST\n",
      "2020-02-20 09:18:52,192 \tReference:  HISTORĨ : EETĨKĨTIE GŨTIRĨ NGAI\n",
      "2020-02-20 09:18:52,192 \tHypothesis: HISTORĨ : MŨNDŨ WA MŨNDŨ\n",
      "2020-02-20 09:18:52,192 Validation result at epoch  15, step    16000: bleu:  18.68, loss: 48922.7031, ppl:   6.8950, duration: 90.9596s\n",
      "2020-02-20 09:19:20,958 Epoch  15 Step:    16100 Batch Loss:     1.519474 Tokens per Sec:     7646, Lr: 0.000300\n",
      "2020-02-20 09:19:49,812 Epoch  15 Step:    16200 Batch Loss:     2.036353 Tokens per Sec:     7704, Lr: 0.000300\n",
      "2020-02-20 09:20:18,499 Epoch  15 Step:    16300 Batch Loss:     1.908288 Tokens per Sec:     7715, Lr: 0.000300\n",
      "2020-02-20 09:20:46,652 Epoch  15 Step:    16400 Batch Loss:     2.568640 Tokens per Sec:     7500, Lr: 0.000300\n",
      "2020-02-20 09:21:09,311 Epoch  15: total training loss 2141.58\n",
      "2020-02-20 09:21:09,311 EPOCH 16\n",
      "2020-02-20 09:21:15,430 Epoch  16 Step:    16500 Batch Loss:     1.885816 Tokens per Sec:     7525, Lr: 0.000300\n",
      "2020-02-20 09:21:44,326 Epoch  16 Step:    16600 Batch Loss:     1.957161 Tokens per Sec:     7570, Lr: 0.000300\n",
      "2020-02-20 09:22:12,939 Epoch  16 Step:    16700 Batch Loss:     1.120665 Tokens per Sec:     7681, Lr: 0.000300\n",
      "2020-02-20 09:22:41,896 Epoch  16 Step:    16800 Batch Loss:     1.895897 Tokens per Sec:     7707, Lr: 0.000300\n",
      "2020-02-20 09:23:10,833 Epoch  16 Step:    16900 Batch Loss:     2.079926 Tokens per Sec:     7752, Lr: 0.000300\n",
      "2020-02-20 09:23:39,284 Epoch  16 Step:    17000 Batch Loss:     2.081605 Tokens per Sec:     7553, Lr: 0.000300\n",
      "2020-02-20 09:25:10,031 Hooray! New best validation result [ppl]!\n",
      "2020-02-20 09:25:10,032 Saving new checkpoint.\n",
      "2020-02-20 09:25:10,253 Example #0\n",
      "2020-02-20 09:25:10,254 \tSource:     Researchers at the University of Copenhagen tracked almost 10,000 middle - aged people over an 11 - year period and found that participants who frequently argued with someone close to them were far more likely to die prematurely than those who seldom had conflicts .\n",
      "2020-02-20 09:25:10,254 \tReference:  Athuthuria yunivasĩtĩ - inĩ ya Copenhagen , nĩ maathuthuririe andũ hakuhĩ 10,000 a mĩaka ĩyo kwa ihinda rĩa mĩaka 11 na makĩona atĩ arĩa maakararanagia na mũndũ wa famĩlĩ kaingĩ , haakoragwo na ũhotekeku mũnene atĩ megũkua tene gũkĩra arĩa mataakoragwo na ngarari kaingĩ .\n",
      "2020-02-20 09:25:10,254 \tHypothesis: Athuthuria a kuuma Amerika ya Cenaken nĩ maanyitire ũgeni andũ 10,000 arĩa maakoragwo na mĩaka 11 na nĩ moonire atĩ andũ aingĩ arĩa maikaraga makĩhũthĩra njĩra ya kũmahunjĩria andũ arĩa maikaraga na ihenya gũkĩra arĩa maikaraga .\n",
      "2020-02-20 09:25:10,254 Example #1\n",
      "2020-02-20 09:25:10,254 \tSource:     Yes .\n",
      "2020-02-20 09:25:10,254 \tReference:  Ĩĩ .\n",
      "2020-02-20 09:25:10,254 \tHypothesis: Ĩĩ .\n",
      "2020-02-20 09:25:10,254 Example #2\n",
      "2020-02-20 09:25:10,254 \tSource:     He enjoyed special intimacy with Jehovah , who enabled him to display “ awesome power ” as he led the Israelites to the Promised Land . ​ — Deut .\n",
      "2020-02-20 09:25:10,254 \tReference:  Aarĩ na ũkuruhanu wa hakuhĩ na Jehova , na nĩ aamũheire ũhoti wa gwĩka “ maũndũ manene ma kũguoyohania ” rĩrĩa aatongoragia Aisiraeli mathiĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 09:25:10,255 \tHypothesis: Nĩ aakenagĩra mũno gũkorũo na ũrata wa mwanya na Jehova , ũrĩa wamũheire “ hinya ” ũrĩa aahetwo nĩ Aisiraeli kuuma Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 09:25:10,255 Example #3\n",
      "2020-02-20 09:25:10,255 \tSource:     HISTORY : ATHEIST\n",
      "2020-02-20 09:25:10,255 \tReference:  HISTORĨ : EETĨKĨTIE GŨTIRĨ NGAI\n",
      "2020-02-20 09:25:10,255 \tHypothesis: HISTORĨ : MŨTHAYA\n",
      "2020-02-20 09:25:10,255 Validation result at epoch  16, step    17000: bleu:  19.09, loss: 48392.4062, ppl:   6.7522, duration: 90.9697s\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2020-02-20 09:25:39,160 Epoch  16 Step:    17100 Batch Loss:     1.762879 Tokens per Sec:     7728, Lr: 0.000300\n",
      "2020-02-20 09:26:07,933 Epoch  16 Step:    17200 Batch Loss:     1.179826 Tokens per Sec:     7718, Lr: 0.000300\n",
      "2020-02-20 09:26:36,680 Epoch  16 Step:    17300 Batch Loss:     2.390864 Tokens per Sec:     7602, Lr: 0.000300\n",
      "2020-02-20 09:27:05,129 Epoch  16 Step:    17400 Batch Loss:     1.921284 Tokens per Sec:     7626, Lr: 0.000300\n",
      "2020-02-20 09:27:33,955 Epoch  16 Step:    17500 Batch Loss:     2.055123 Tokens per Sec:     7640, Lr: 0.000300\n",
      "2020-02-20 09:27:55,769 Epoch  16: total training loss 2104.33\n",
      "2020-02-20 09:27:55,769 EPOCH 17\n",
      "2020-02-20 09:28:02,825 Epoch  17 Step:    17600 Batch Loss:     1.867045 Tokens per Sec:     7402, Lr: 0.000300\n",
      "2020-02-20 09:28:31,292 Epoch  17 Step:    17700 Batch Loss:     2.136398 Tokens per Sec:     7434, Lr: 0.000300\n",
      "2020-02-20 09:28:59,751 Epoch  17 Step:    17800 Batch Loss:     2.207061 Tokens per Sec:     7606, Lr: 0.000300\n",
      "2020-02-20 09:29:28,678 Epoch  17 Step:    17900 Batch Loss:     1.921016 Tokens per Sec:     7704, Lr: 0.000300\n",
      "2020-02-20 09:29:57,532 Epoch  17 Step:    18000 Batch Loss:     1.766410 Tokens per Sec:     7772, Lr: 0.000300\n",
      "2020-02-20 09:31:28,293 Hooray! New best validation result [ppl]!\n",
      "2020-02-20 09:31:28,293 Saving new checkpoint.\n",
      "2020-02-20 09:31:28,507 Example #0\n",
      "2020-02-20 09:31:28,507 \tSource:     Researchers at the University of Copenhagen tracked almost 10,000 middle - aged people over an 11 - year period and found that participants who frequently argued with someone close to them were far more likely to die prematurely than those who seldom had conflicts .\n",
      "2020-02-20 09:31:28,507 \tReference:  Athuthuria yunivasĩtĩ - inĩ ya Copenhagen , nĩ maathuthuririe andũ hakuhĩ 10,000 a mĩaka ĩyo kwa ihinda rĩa mĩaka 11 na makĩona atĩ arĩa maakararanagia na mũndũ wa famĩlĩ kaingĩ , haakoragwo na ũhotekeku mũnene atĩ megũkua tene gũkĩra arĩa mataakoragwo na ngarari kaingĩ .\n",
      "2020-02-20 09:31:28,508 \tHypothesis: Andũ aingĩ arĩa maahanaga nĩ ũndũ wa kũingĩra thĩinĩ wa Amerika ( US ) nĩ maahũũrire andũ 10,000 arĩa maarĩ na mĩaka 11 na kĩndũ mĩaka 11 , na nĩ moonire atĩ andũ arĩa maikaraga meiguĩte nĩ maabataraga kũheo wĩra wa kũmakania na mũndũ ũcio .\n",
      "2020-02-20 09:31:28,508 Example #1\n",
      "2020-02-20 09:31:28,508 \tSource:     Yes .\n",
      "2020-02-20 09:31:28,508 \tReference:  Ĩĩ .\n",
      "2020-02-20 09:31:28,508 \tHypothesis: Ĩĩ .\n",
      "2020-02-20 09:31:28,508 Example #2\n",
      "2020-02-20 09:31:28,508 \tSource:     He enjoyed special intimacy with Jehovah , who enabled him to display “ awesome power ” as he led the Israelites to the Promised Land . ​ — Deut .\n",
      "2020-02-20 09:31:28,508 \tReference:  Aarĩ na ũkuruhanu wa hakuhĩ na Jehova , na nĩ aamũheire ũhoti wa gwĩka “ maũndũ manene ma kũguoyohania ” rĩrĩa aatongoragia Aisiraeli mathiĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 09:31:28,508 \tHypothesis: Nĩ aakenagĩra mũno nĩ Jehova , ũrĩa wamũteithirie kũmũtongoria ‘ ũhoti wa kũmenya ’ ũrĩa aaheire Aisiraeli maatũme Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 09:31:28,509 Example #3\n",
      "2020-02-20 09:31:28,509 \tSource:     HISTORY : ATHEIST\n",
      "2020-02-20 09:31:28,509 \tReference:  HISTORĨ : EETĨKĨTIE GŨTIRĨ NGAI\n",
      "2020-02-20 09:31:28,509 \tHypothesis: HISTORĨ : MŨTHAYA\n",
      "2020-02-20 09:31:28,509 Validation result at epoch  17, step    18000: bleu:  19.45, loss: 47702.1055, ppl:   6.5708, duration: 90.9765s\n",
      "2020-02-20 09:31:57,473 Epoch  17 Step:    18100 Batch Loss:     2.046036 Tokens per Sec:     7695, Lr: 0.000300\n",
      "2020-02-20 09:32:26,412 Epoch  17 Step:    18200 Batch Loss:     1.942086 Tokens per Sec:     7722, Lr: 0.000300\n",
      "2020-02-20 09:32:55,071 Epoch  17 Step:    18300 Batch Loss:     2.094379 Tokens per Sec:     7628, Lr: 0.000300\n",
      "2020-02-20 09:33:23,424 Epoch  17 Step:    18400 Batch Loss:     1.560269 Tokens per Sec:     7530, Lr: 0.000300\n",
      "2020-02-20 09:33:52,456 Epoch  17 Step:    18500 Batch Loss:     1.726429 Tokens per Sec:     7875, Lr: 0.000300\n",
      "2020-02-20 09:34:21,328 Epoch  17 Step:    18600 Batch Loss:     2.112112 Tokens per Sec:     7810, Lr: 0.000300\n",
      "2020-02-20 09:34:41,442 Epoch  17: total training loss 2058.65\n",
      "2020-02-20 09:34:41,442 EPOCH 18\n",
      "2020-02-20 09:34:50,328 Epoch  18 Step:    18700 Batch Loss:     1.858301 Tokens per Sec:     7633, Lr: 0.000300\n",
      "2020-02-20 09:35:18,944 Epoch  18 Step:    18800 Batch Loss:     2.224093 Tokens per Sec:     7693, Lr: 0.000300\n",
      "2020-02-20 09:35:47,899 Epoch  18 Step:    18900 Batch Loss:     2.186527 Tokens per Sec:     7647, Lr: 0.000300\n",
      "2020-02-20 09:36:16,711 Epoch  18 Step:    19000 Batch Loss:     1.819201 Tokens per Sec:     7533, Lr: 0.000300\n",
      "2020-02-20 09:37:47,406 Example #0\n",
      "2020-02-20 09:37:47,406 \tSource:     Researchers at the University of Copenhagen tracked almost 10,000 middle - aged people over an 11 - year period and found that participants who frequently argued with someone close to them were far more likely to die prematurely than those who seldom had conflicts .\n",
      "2020-02-20 09:37:47,406 \tReference:  Athuthuria yunivasĩtĩ - inĩ ya Copenhagen , nĩ maathuthuririe andũ hakuhĩ 10,000 a mĩaka ĩyo kwa ihinda rĩa mĩaka 11 na makĩona atĩ arĩa maakararanagia na mũndũ wa famĩlĩ kaingĩ , haakoragwo na ũhotekeku mũnene atĩ megũkua tene gũkĩra arĩa mataakoragwo na ngarari kaingĩ .\n",
      "2020-02-20 09:37:47,406 \tHypothesis: Athuthuria a sayanisi a sayanisi a sayanisi arĩa maarĩ na mĩaka 1,000 , nĩ maanyitire ũgeni andũ aingĩ arĩa maikaraga kuo ihinda iraya , na nĩ ũndũ ũcio makĩambĩrĩria kũhona mũndũ o wothe ũrĩa ũikaraga hakuhĩ na ũrĩa maabataraga .\n",
      "2020-02-20 09:37:47,407 Example #1\n",
      "2020-02-20 09:37:47,407 \tSource:     Yes .\n",
      "2020-02-20 09:37:47,407 \tReference:  Ĩĩ .\n",
      "2020-02-20 09:37:47,407 \tHypothesis: Ĩĩ , nĩ ma .\n",
      "2020-02-20 09:37:47,407 Example #2\n",
      "2020-02-20 09:37:47,407 \tSource:     He enjoyed special intimacy with Jehovah , who enabled him to display “ awesome power ” as he led the Israelites to the Promised Land . ​ — Deut .\n",
      "2020-02-20 09:37:47,407 \tReference:  Aarĩ na ũkuruhanu wa hakuhĩ na Jehova , na nĩ aamũheire ũhoti wa gwĩka “ maũndũ manene ma kũguoyohania ” rĩrĩa aatongoragia Aisiraeli mathiĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 09:37:47,407 \tHypothesis: Nĩ aakenagĩra mũno gũkorũo arĩ na ũrata wa mwanya na Jehova , ũrĩa wamũteithirie kuonania “ hinya ” ũrĩa aaheire Aisiraeli kuuma Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 09:37:47,407 Example #3\n",
      "2020-02-20 09:37:47,408 \tSource:     HISTORY : ATHEIST\n",
      "2020-02-20 09:37:47,408 \tReference:  HISTORĨ : EETĨKĨTIE GŨTIRĨ NGAI\n",
      "2020-02-20 09:37:47,408 \tHypothesis: HISTORĨ : MŨTHATHITHIO\n",
      "2020-02-20 09:37:47,408 Validation result at epoch  18, step    19000: bleu:  19.01, loss: 48030.9531, ppl:   6.6566, duration: 90.6960s\n",
      "2020-02-20 09:38:16,470 Epoch  18 Step:    19100 Batch Loss:     2.293683 Tokens per Sec:     7781, Lr: 0.000300\n",
      "2020-02-20 09:38:45,194 Epoch  18 Step:    19200 Batch Loss:     1.851934 Tokens per Sec:     7609, Lr: 0.000300\n",
      "2020-02-20 09:39:14,219 Epoch  18 Step:    19300 Batch Loss:     2.018710 Tokens per Sec:     7614, Lr: 0.000300\n",
      "2020-02-20 09:39:42,863 Epoch  18 Step:    19400 Batch Loss:     2.120527 Tokens per Sec:     7648, Lr: 0.000300\n",
      "2020-02-20 09:40:11,236 Epoch  18 Step:    19500 Batch Loss:     2.119176 Tokens per Sec:     7645, Lr: 0.000300\n",
      "2020-02-20 09:40:40,223 Epoch  18 Step:    19600 Batch Loss:     1.957167 Tokens per Sec:     7852, Lr: 0.000300\n",
      "2020-02-20 09:41:09,144 Epoch  18 Step:    19700 Batch Loss:     2.134189 Tokens per Sec:     7719, Lr: 0.000300\n",
      "2020-02-20 09:41:26,613 Epoch  18: total training loss 2028.71\n",
      "2020-02-20 09:41:26,613 EPOCH 19\n",
      "2020-02-20 09:41:37,628 Epoch  19 Step:    19800 Batch Loss:     1.751475 Tokens per Sec:     7251, Lr: 0.000300\n",
      "2020-02-20 09:42:06,506 Epoch  19 Step:    19900 Batch Loss:     2.067418 Tokens per Sec:     7728, Lr: 0.000300\n",
      "2020-02-20 09:42:35,108 Epoch  19 Step:    20000 Batch Loss:     2.300687 Tokens per Sec:     7609, Lr: 0.000300\n",
      "2020-02-20 09:44:05,850 Hooray! New best validation result [ppl]!\n",
      "2020-02-20 09:44:05,850 Saving new checkpoint.\n",
      "2020-02-20 09:44:06,071 Example #0\n",
      "2020-02-20 09:44:06,072 \tSource:     Researchers at the University of Copenhagen tracked almost 10,000 middle - aged people over an 11 - year period and found that participants who frequently argued with someone close to them were far more likely to die prematurely than those who seldom had conflicts .\n",
      "2020-02-20 09:44:06,072 \tReference:  Athuthuria yunivasĩtĩ - inĩ ya Copenhagen , nĩ maathuthuririe andũ hakuhĩ 10,000 a mĩaka ĩyo kwa ihinda rĩa mĩaka 11 na makĩona atĩ arĩa maakararanagia na mũndũ wa famĩlĩ kaingĩ , haakoragwo na ũhotekeku mũnene atĩ megũkua tene gũkĩra arĩa mataakoragwo na ngarari kaingĩ .\n",
      "2020-02-20 09:44:06,072 \tHypothesis: Athuthuria a sayansi a Carenagint nĩ maahũũrire andũ 10,000 arĩa maanyitĩte ũgeni kwa ihinda rĩa mĩaka 11 na nuthu , na nĩ moonire atĩ andũ arĩa mahoyaga Ngai nĩ maabataraga gũkorũo marĩ na ũũgĩ mũingĩ gũkĩra andũ arĩa maakuire .\n",
      "2020-02-20 09:44:06,072 Example #1\n",
      "2020-02-20 09:44:06,072 \tSource:     Yes .\n",
      "2020-02-20 09:44:06,072 \tReference:  Ĩĩ .\n",
      "2020-02-20 09:44:06,072 \tHypothesis: Ĩĩ .\n",
      "2020-02-20 09:44:06,073 Example #2\n",
      "2020-02-20 09:44:06,073 \tSource:     He enjoyed special intimacy with Jehovah , who enabled him to display “ awesome power ” as he led the Israelites to the Promised Land . ​ — Deut .\n",
      "2020-02-20 09:44:06,073 \tReference:  Aarĩ na ũkuruhanu wa hakuhĩ na Jehova , na nĩ aamũheire ũhoti wa gwĩka “ maũndũ manene ma kũguoyohania ” rĩrĩa aatongoragia Aisiraeli mathiĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 09:44:06,073 \tHypothesis: Nĩ aakenagĩra mũno gũkorũo arĩ na ũrata wa mwanya na Jehova , ũrĩa wamũteithirie kuona “ hinya ” ũrĩa aaheire Aisiraeli mathiĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 09:44:06,073 Example #3\n",
      "2020-02-20 09:44:06,073 \tSource:     HISTORY : ATHEIST\n",
      "2020-02-20 09:44:06,073 \tReference:  HISTORĨ : EETĨKĨTIE GŨTIRĨ NGAI\n",
      "2020-02-20 09:44:06,073 \tHypothesis: HISTORĨ : MŨTUNO\n",
      "2020-02-20 09:44:06,073 Validation result at epoch  19, step    20000: bleu:  20.04, loss: 46900.6914, ppl:   6.3662, duration: 90.9644s\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2020-02-20 09:44:34,643 Epoch  19 Step:    20100 Batch Loss:     1.911918 Tokens per Sec:     7558, Lr: 0.000300\n",
      "2020-02-20 09:45:03,459 Epoch  19 Step:    20200 Batch Loss:     2.009794 Tokens per Sec:     7638, Lr: 0.000300\n",
      "2020-02-20 09:45:32,382 Epoch  19 Step:    20300 Batch Loss:     2.239769 Tokens per Sec:     7841, Lr: 0.000300\n",
      "2020-02-20 09:46:01,129 Epoch  19 Step:    20400 Batch Loss:     1.981424 Tokens per Sec:     7631, Lr: 0.000300\n",
      "2020-02-20 09:46:30,005 Epoch  19 Step:    20500 Batch Loss:     1.979983 Tokens per Sec:     7786, Lr: 0.000300\n",
      "2020-02-20 09:46:58,752 Epoch  19 Step:    20600 Batch Loss:     1.382899 Tokens per Sec:     7669, Lr: 0.000300\n",
      "2020-02-20 09:47:27,195 Epoch  19 Step:    20700 Batch Loss:     1.610619 Tokens per Sec:     7357, Lr: 0.000300\n",
      "2020-02-20 09:47:56,029 Epoch  19 Step:    20800 Batch Loss:     1.765529 Tokens per Sec:     7841, Lr: 0.000300\n",
      "2020-02-20 09:48:12,791 Epoch  19: total training loss 2010.95\n",
      "2020-02-20 09:48:12,792 EPOCH 20\n",
      "2020-02-20 09:48:25,198 Epoch  20 Step:    20900 Batch Loss:     1.499889 Tokens per Sec:     7688, Lr: 0.000300\n",
      "2020-02-20 09:48:53,837 Epoch  20 Step:    21000 Batch Loss:     1.951232 Tokens per Sec:     7704, Lr: 0.000300\n",
      "2020-02-20 09:50:24,568 Hooray! New best validation result [ppl]!\n",
      "2020-02-20 09:50:24,569 Saving new checkpoint.\n",
      "2020-02-20 09:50:24,787 Example #0\n",
      "2020-02-20 09:50:24,788 \tSource:     Researchers at the University of Copenhagen tracked almost 10,000 middle - aged people over an 11 - year period and found that participants who frequently argued with someone close to them were far more likely to die prematurely than those who seldom had conflicts .\n",
      "2020-02-20 09:50:24,788 \tReference:  Athuthuria yunivasĩtĩ - inĩ ya Copenhagen , nĩ maathuthuririe andũ hakuhĩ 10,000 a mĩaka ĩyo kwa ihinda rĩa mĩaka 11 na makĩona atĩ arĩa maakararanagia na mũndũ wa famĩlĩ kaingĩ , haakoragwo na ũhotekeku mũnene atĩ megũkua tene gũkĩra arĩa mataakoragwo na ngarari kaingĩ .\n",
      "2020-02-20 09:50:24,788 \tHypothesis: Andũ aingĩ arĩa mathuthuragia Amerika ( US ) nĩ maahũũrire andũ 10,000 arĩa maarĩ na mĩaka ta 11 na nĩ moonire atĩ andũ arĩa maikaraga nao nĩ maanyariragwo mũno nĩ mũndũ ũmwe ũmwe ũmwe ũmwe ũmwe wao , no kũhoteke nĩ maakuĩte ngoro mũno .\n",
      "2020-02-20 09:50:24,788 Example #1\n",
      "2020-02-20 09:50:24,788 \tSource:     Yes .\n",
      "2020-02-20 09:50:24,788 \tReference:  Ĩĩ .\n",
      "2020-02-20 09:50:24,788 \tHypothesis: Ĩĩ .\n",
      "2020-02-20 09:50:24,788 Example #2\n",
      "2020-02-20 09:50:24,789 \tSource:     He enjoyed special intimacy with Jehovah , who enabled him to display “ awesome power ” as he led the Israelites to the Promised Land . ​ — Deut .\n",
      "2020-02-20 09:50:24,789 \tReference:  Aarĩ na ũkuruhanu wa hakuhĩ na Jehova , na nĩ aamũheire ũhoti wa gwĩka “ maũndũ manene ma kũguoyohania ” rĩrĩa aatongoragia Aisiraeli mathiĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 09:50:24,789 \tHypothesis: Nĩ aakenire mũno nĩ ũndũ wa gũkorũo na ũrata wa mwanya na Jehova , ũrĩa wamũteithirie kuonania “ hinya ” ũrĩa aaheire Aisiraeli kuuma Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 09:50:24,789 Example #3\n",
      "2020-02-20 09:50:24,789 \tSource:     HISTORY : ATHEIST\n",
      "2020-02-20 09:50:24,789 \tReference:  HISTORĨ : EETĨKĨTIE GŨTIRĨ NGAI\n",
      "2020-02-20 09:50:24,790 \tHypothesis: HISTORĨ : MŨNDŨ WA MŨNDŨ WA MŨNDŨ WA MŨTŨŨRO\n",
      "2020-02-20 09:50:24,790 Validation result at epoch  20, step    21000: bleu:  20.24, loss: 46317.4102, ppl:   6.2213, duration: 90.9519s\n",
      "2020-02-20 09:50:53,390 Epoch  20 Step:    21100 Batch Loss:     1.996995 Tokens per Sec:     7524, Lr: 0.000300\n",
      "2020-02-20 09:51:21,914 Epoch  20 Step:    21200 Batch Loss:     2.264500 Tokens per Sec:     7686, Lr: 0.000300\n",
      "2020-02-20 09:51:50,495 Epoch  20 Step:    21300 Batch Loss:     1.929861 Tokens per Sec:     7632, Lr: 0.000300\n",
      "2020-02-20 09:52:19,077 Epoch  20 Step:    21400 Batch Loss:     2.021441 Tokens per Sec:     7633, Lr: 0.000300\n",
      "2020-02-20 09:52:47,992 Epoch  20 Step:    21500 Batch Loss:     1.671376 Tokens per Sec:     7797, Lr: 0.000300\n",
      "2020-02-20 09:53:16,612 Epoch  20 Step:    21600 Batch Loss:     2.305511 Tokens per Sec:     7658, Lr: 0.000300\n",
      "2020-02-20 09:53:45,660 Epoch  20 Step:    21700 Batch Loss:     1.875114 Tokens per Sec:     7748, Lr: 0.000300\n",
      "2020-02-20 09:54:14,071 Epoch  20 Step:    21800 Batch Loss:     1.963596 Tokens per Sec:     7629, Lr: 0.000300\n",
      "2020-02-20 09:54:42,553 Epoch  20 Step:    21900 Batch Loss:     1.868513 Tokens per Sec:     7540, Lr: 0.000300\n",
      "2020-02-20 09:54:59,088 Epoch  20: total training loss 1987.05\n",
      "2020-02-20 09:54:59,088 EPOCH 21\n",
      "2020-02-20 09:55:11,340 Epoch  21 Step:    22000 Batch Loss:     1.768498 Tokens per Sec:     7455, Lr: 0.000300\n",
      "2020-02-20 09:56:42,091 Hooray! New best validation result [ppl]!\n",
      "2020-02-20 09:56:42,092 Saving new checkpoint.\n",
      "2020-02-20 09:56:42,308 Example #0\n",
      "2020-02-20 09:56:42,308 \tSource:     Researchers at the University of Copenhagen tracked almost 10,000 middle - aged people over an 11 - year period and found that participants who frequently argued with someone close to them were far more likely to die prematurely than those who seldom had conflicts .\n",
      "2020-02-20 09:56:42,308 \tReference:  Athuthuria yunivasĩtĩ - inĩ ya Copenhagen , nĩ maathuthuririe andũ hakuhĩ 10,000 a mĩaka ĩyo kwa ihinda rĩa mĩaka 11 na makĩona atĩ arĩa maakararanagia na mũndũ wa famĩlĩ kaingĩ , haakoragwo na ũhotekeku mũnene atĩ megũkua tene gũkĩra arĩa mataakoragwo na ngarari kaingĩ .\n",
      "2020-02-20 09:56:42,308 \tHypothesis: Athuthuria a ndini ya Ngeretha a bũrũri wa Amerika ( US ) nĩ maahũrire andũ 10,000 arĩa maarĩ na mĩaka ta 1011 na makĩona atĩ andũ arĩa maikaraga makĩenda gũkoma nĩ maakuĩte ngoro mũno gũkĩra arĩa maakuĩte .\n",
      "2020-02-20 09:56:42,308 Example #1\n",
      "2020-02-20 09:56:42,308 \tSource:     Yes .\n",
      "2020-02-20 09:56:42,309 \tReference:  Ĩĩ .\n",
      "2020-02-20 09:56:42,309 \tHypothesis: Ĩĩ .\n",
      "2020-02-20 09:56:42,309 Example #2\n",
      "2020-02-20 09:56:42,309 \tSource:     He enjoyed special intimacy with Jehovah , who enabled him to display “ awesome power ” as he led the Israelites to the Promised Land . ​ — Deut .\n",
      "2020-02-20 09:56:42,309 \tReference:  Aarĩ na ũkuruhanu wa hakuhĩ na Jehova , na nĩ aamũheire ũhoti wa gwĩka “ maũndũ manene ma kũguoyohania ” rĩrĩa aatongoragia Aisiraeli mathiĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 09:56:42,309 \tHypothesis: Nĩ aakenagĩra mũno gũkorũo na ũrata wa mwanya na Jehova , ũrĩa wamũheire “ hinya ” ũrĩa aaheire Aisiraeli mathiĩ kũrĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 09:56:42,309 Example #3\n",
      "2020-02-20 09:56:42,309 \tSource:     HISTORY : ATHEIST\n",
      "2020-02-20 09:56:42,310 \tReference:  HISTORĨ : EETĨKĨTIE GŨTIRĨ NGAI\n",
      "2020-02-20 09:56:42,310 \tHypothesis: HISTORĨ : MŨNDŨ WA MŨNDŨ\n",
      "2020-02-20 09:56:42,310 Validation result at epoch  21, step    22000: bleu:  20.78, loss: 46154.4414, ppl:   6.1814, duration: 90.9692s\n",
      "2020-02-20 09:57:11,014 Epoch  21 Step:    22100 Batch Loss:     1.412232 Tokens per Sec:     7566, Lr: 0.000300\n",
      "2020-02-20 09:57:39,688 Epoch  21 Step:    22200 Batch Loss:     1.834065 Tokens per Sec:     7709, Lr: 0.000300\n",
      "2020-02-20 09:58:08,608 Epoch  21 Step:    22300 Batch Loss:     1.996953 Tokens per Sec:     7669, Lr: 0.000300\n",
      "2020-02-20 09:58:37,155 Epoch  21 Step:    22400 Batch Loss:     1.386650 Tokens per Sec:     7612, Lr: 0.000300\n",
      "2020-02-20 09:59:05,920 Epoch  21 Step:    22500 Batch Loss:     1.507768 Tokens per Sec:     7782, Lr: 0.000300\n",
      "2020-02-20 09:59:34,882 Epoch  21 Step:    22600 Batch Loss:     1.519434 Tokens per Sec:     7592, Lr: 0.000300\n",
      "2020-02-20 10:00:03,277 Epoch  21 Step:    22700 Batch Loss:     0.902683 Tokens per Sec:     7526, Lr: 0.000300\n",
      "2020-02-20 10:00:31,736 Epoch  21 Step:    22800 Batch Loss:     1.673352 Tokens per Sec:     7506, Lr: 0.000300\n",
      "2020-02-20 10:01:00,691 Epoch  21 Step:    22900 Batch Loss:     1.775687 Tokens per Sec:     7700, Lr: 0.000300\n",
      "2020-02-20 10:01:29,448 Epoch  21 Step:    23000 Batch Loss:     1.974908 Tokens per Sec:     7555, Lr: 0.000300\n",
      "2020-02-20 10:03:00,260 Hooray! New best validation result [ppl]!\n",
      "2020-02-20 10:03:00,260 Saving new checkpoint.\n",
      "2020-02-20 10:03:00,481 Example #0\n",
      "2020-02-20 10:03:00,482 \tSource:     Researchers at the University of Copenhagen tracked almost 10,000 middle - aged people over an 11 - year period and found that participants who frequently argued with someone close to them were far more likely to die prematurely than those who seldom had conflicts .\n",
      "2020-02-20 10:03:00,482 \tReference:  Athuthuria yunivasĩtĩ - inĩ ya Copenhagen , nĩ maathuthuririe andũ hakuhĩ 10,000 a mĩaka ĩyo kwa ihinda rĩa mĩaka 11 na makĩona atĩ arĩa maakararanagia na mũndũ wa famĩlĩ kaingĩ , haakoragwo na ũhotekeku mũnene atĩ megũkua tene gũkĩra arĩa mataakoragwo na ngarari kaingĩ .\n",
      "2020-02-20 10:03:00,482 \tHypothesis: Athuthuria a sayansi a bũrũri wa Ngeretha nĩ maahũrire andũ 10,000 arĩa maanyitĩte ũgeni nĩ maanyitaga ũgeni kwa ihinda rĩa mĩaka 11 na makĩona atĩ andũ arĩa maikaraga maikaraga makĩenda kũingĩra thĩinĩ wa thĩ nĩ maabataraga kũingĩra na hinya gũkĩra arĩa maakuĩte .\n",
      "2020-02-20 10:03:00,482 Example #1\n",
      "2020-02-20 10:03:00,482 \tSource:     Yes .\n",
      "2020-02-20 10:03:00,482 \tReference:  Ĩĩ .\n",
      "2020-02-20 10:03:00,482 \tHypothesis: Ĩĩ .\n",
      "2020-02-20 10:03:00,482 Example #2\n",
      "2020-02-20 10:03:00,483 \tSource:     He enjoyed special intimacy with Jehovah , who enabled him to display “ awesome power ” as he led the Israelites to the Promised Land . ​ — Deut .\n",
      "2020-02-20 10:03:00,483 \tReference:  Aarĩ na ũkuruhanu wa hakuhĩ na Jehova , na nĩ aamũheire ũhoti wa gwĩka “ maũndũ manene ma kũguoyohania ” rĩrĩa aatongoragia Aisiraeli mathiĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 10:03:00,483 \tHypothesis: Nĩ aakenire mũno nĩ gũkorũo na ũrata wa mwanya na Jehova , ũrĩa wamũteithirie kuonania “ hinya ” ũrĩa aaheire Aisiraeli kuuma Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 10:03:00,483 Example #3\n",
      "2020-02-20 10:03:00,483 \tSource:     HISTORY : ATHEIST\n",
      "2020-02-20 10:03:00,483 \tReference:  HISTORĨ : EETĨKĨTIE GŨTIRĨ NGAI\n",
      "2020-02-20 10:03:00,483 \tHypothesis: HISTORĨ : MŨTHAYA\n",
      "2020-02-20 10:03:00,483 Validation result at epoch  21, step    23000: bleu:  20.73, loss: 45510.0195, ppl:   6.0262, duration: 91.0343s\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2020-02-20 10:03:17,380 Epoch  21: total training loss 1963.73\n",
      "2020-02-20 10:03:17,380 EPOCH 22\n",
      "2020-02-20 10:03:29,098 Epoch  22 Step:    23100 Batch Loss:     1.923240 Tokens per Sec:     7282, Lr: 0.000300\n",
      "2020-02-20 10:03:58,047 Epoch  22 Step:    23200 Batch Loss:     1.682064 Tokens per Sec:     7787, Lr: 0.000300\n",
      "2020-02-20 10:04:26,859 Epoch  22 Step:    23300 Batch Loss:     1.587517 Tokens per Sec:     7528, Lr: 0.000300\n",
      "2020-02-20 10:04:55,573 Epoch  22 Step:    23400 Batch Loss:     1.631523 Tokens per Sec:     7699, Lr: 0.000300\n",
      "2020-02-20 10:05:24,618 Epoch  22 Step:    23500 Batch Loss:     1.568314 Tokens per Sec:     7826, Lr: 0.000300\n",
      "2020-02-20 10:05:53,717 Epoch  22 Step:    23600 Batch Loss:     1.872469 Tokens per Sec:     7666, Lr: 0.000300\n",
      "2020-02-20 10:06:22,576 Epoch  22 Step:    23700 Batch Loss:     1.770639 Tokens per Sec:     7607, Lr: 0.000300\n",
      "2020-02-20 10:06:51,321 Epoch  22 Step:    23800 Batch Loss:     1.423158 Tokens per Sec:     7790, Lr: 0.000300\n",
      "2020-02-20 10:07:19,891 Epoch  22 Step:    23900 Batch Loss:     1.713157 Tokens per Sec:     7610, Lr: 0.000300\n",
      "2020-02-20 10:07:48,726 Epoch  22 Step:    24000 Batch Loss:     1.856597 Tokens per Sec:     7662, Lr: 0.000300\n",
      "2020-02-20 10:09:19,569 Hooray! New best validation result [ppl]!\n",
      "2020-02-20 10:09:19,570 Saving new checkpoint.\n",
      "2020-02-20 10:09:19,790 Example #0\n",
      "2020-02-20 10:09:19,791 \tSource:     Researchers at the University of Copenhagen tracked almost 10,000 middle - aged people over an 11 - year period and found that participants who frequently argued with someone close to them were far more likely to die prematurely than those who seldom had conflicts .\n",
      "2020-02-20 10:09:19,791 \tReference:  Athuthuria yunivasĩtĩ - inĩ ya Copenhagen , nĩ maathuthuririe andũ hakuhĩ 10,000 a mĩaka ĩyo kwa ihinda rĩa mĩaka 11 na makĩona atĩ arĩa maakararanagia na mũndũ wa famĩlĩ kaingĩ , haakoragwo na ũhotekeku mũnene atĩ megũkua tene gũkĩra arĩa mataakoragwo na ngarari kaingĩ .\n",
      "2020-02-20 10:09:19,791 \tHypothesis: Athuthuria a sayansi a Cophagineen nĩ maahũrire andũ 10,000 arĩa maarĩ na ũmenyeru na nĩ maakoragwo na andũ ta 11 arĩa maikaraga nao nĩ maacemanagia na mũndũ ũmwe ũmwe ũmwe ũmwe ũmwe wa famĩlĩ ciao .\n",
      "2020-02-20 10:09:19,791 Example #1\n",
      "2020-02-20 10:09:19,791 \tSource:     Yes .\n",
      "2020-02-20 10:09:19,791 \tReference:  Ĩĩ .\n",
      "2020-02-20 10:09:19,791 \tHypothesis: Ĩĩ , nĩ ma .\n",
      "2020-02-20 10:09:19,791 Example #2\n",
      "2020-02-20 10:09:19,792 \tSource:     He enjoyed special intimacy with Jehovah , who enabled him to display “ awesome power ” as he led the Israelites to the Promised Land . ​ — Deut .\n",
      "2020-02-20 10:09:19,792 \tReference:  Aarĩ na ũkuruhanu wa hakuhĩ na Jehova , na nĩ aamũheire ũhoti wa gwĩka “ maũndũ manene ma kũguoyohania ” rĩrĩa aatongoragia Aisiraeli mathiĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 10:09:19,792 \tHypothesis: Nĩ aakenagĩra mũno gũkorũo arĩ na ũigananĩru harĩ Jehova , ũrĩa wamũteithirie kuonania “ hinya ” ũrĩa aaheire Aisiraeli kuuma Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 10:09:19,792 Example #3\n",
      "2020-02-20 10:09:19,792 \tSource:     HISTORY : ATHEIST\n",
      "2020-02-20 10:09:19,792 \tReference:  HISTORĨ : EETĨKĨTIE GŨTIRĨ NGAI\n",
      "2020-02-20 10:09:19,792 \tHypothesis: HISTORĨ : MŨTHENYA\n",
      "2020-02-20 10:09:19,792 Validation result at epoch  22, step    24000: bleu:  21.01, loss: 45274.8008, ppl:   5.9705, duration: 91.0652s\n",
      "2020-02-20 10:09:48,429 Epoch  22 Step:    24100 Batch Loss:     1.681103 Tokens per Sec:     7655, Lr: 0.000300\n",
      "2020-02-20 10:10:03,166 Epoch  22: total training loss 1927.07\n",
      "2020-02-20 10:10:03,167 EPOCH 23\n",
      "2020-02-20 10:10:17,127 Epoch  23 Step:    24200 Batch Loss:     1.616819 Tokens per Sec:     7528, Lr: 0.000300\n",
      "2020-02-20 10:10:45,879 Epoch  23 Step:    24300 Batch Loss:     1.644871 Tokens per Sec:     7648, Lr: 0.000300\n",
      "2020-02-20 10:11:14,341 Epoch  23 Step:    24400 Batch Loss:     1.687739 Tokens per Sec:     7555, Lr: 0.000300\n",
      "2020-02-20 10:11:43,520 Epoch  23 Step:    24500 Batch Loss:     1.709220 Tokens per Sec:     7809, Lr: 0.000300\n",
      "2020-02-20 10:12:12,283 Epoch  23 Step:    24600 Batch Loss:     1.626681 Tokens per Sec:     7661, Lr: 0.000300\n",
      "2020-02-20 10:12:41,135 Epoch  23 Step:    24700 Batch Loss:     1.770550 Tokens per Sec:     7669, Lr: 0.000300\n",
      "2020-02-20 10:13:09,848 Epoch  23 Step:    24800 Batch Loss:     2.103009 Tokens per Sec:     7645, Lr: 0.000300\n",
      "2020-02-20 10:13:38,773 Epoch  23 Step:    24900 Batch Loss:     1.181772 Tokens per Sec:     7754, Lr: 0.000300\n",
      "2020-02-20 10:14:07,824 Epoch  23 Step:    25000 Batch Loss:     1.290577 Tokens per Sec:     7717, Lr: 0.000300\n",
      "2020-02-20 10:15:38,594 Hooray! New best validation result [ppl]!\n",
      "2020-02-20 10:15:38,594 Saving new checkpoint.\n",
      "2020-02-20 10:15:38,809 Example #0\n",
      "2020-02-20 10:15:38,809 \tSource:     Researchers at the University of Copenhagen tracked almost 10,000 middle - aged people over an 11 - year period and found that participants who frequently argued with someone close to them were far more likely to die prematurely than those who seldom had conflicts .\n",
      "2020-02-20 10:15:38,809 \tReference:  Athuthuria yunivasĩtĩ - inĩ ya Copenhagen , nĩ maathuthuririe andũ hakuhĩ 10,000 a mĩaka ĩyo kwa ihinda rĩa mĩaka 11 na makĩona atĩ arĩa maakararanagia na mũndũ wa famĩlĩ kaingĩ , haakoragwo na ũhotekeku mũnene atĩ megũkua tene gũkĩra arĩa mataakoragwo na ngarari kaingĩ .\n",
      "2020-02-20 10:15:38,810 \tHypothesis: Athuthuria a sayansi thĩinĩ wa Amerika ya U.S . nĩ maahũrire andũ 10,000 arĩa maanyariragwo mũno nĩ andũ makĩria ma 10,000 , na nĩ moonire atĩ andũ arĩa maikaraga makĩenda mũno kũhikania nao nĩ maanyariragwo mũno nĩ mũndũ ũcio .\n",
      "2020-02-20 10:15:38,810 Example #1\n",
      "2020-02-20 10:15:38,810 \tSource:     Yes .\n",
      "2020-02-20 10:15:38,810 \tReference:  Ĩĩ .\n",
      "2020-02-20 10:15:38,810 \tHypothesis: Ĩĩ nĩ ya ma .\n",
      "2020-02-20 10:15:38,810 Example #2\n",
      "2020-02-20 10:15:38,810 \tSource:     He enjoyed special intimacy with Jehovah , who enabled him to display “ awesome power ” as he led the Israelites to the Promised Land . ​ — Deut .\n",
      "2020-02-20 10:15:38,810 \tReference:  Aarĩ na ũkuruhanu wa hakuhĩ na Jehova , na nĩ aamũheire ũhoti wa gwĩka “ maũndũ manene ma kũguoyohania ” rĩrĩa aatongoragia Aisiraeli mathiĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 10:15:38,810 \tHypothesis: Nĩ aakenagĩra mũno gũkorũo na ũrata wa mwanya hamwe na Jehova , ũrĩa wamũteithirie kuonania “ hinya ” ũrĩa aaheete Aisiraeli maatũme Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 10:15:38,810 Example #3\n",
      "2020-02-20 10:15:38,811 \tSource:     HISTORY : ATHEIST\n",
      "2020-02-20 10:15:38,811 \tReference:  HISTORĨ : EETĨKĨTIE GŨTIRĨ NGAI\n",
      "2020-02-20 10:15:38,811 \tHypothesis: HISTORĨ : MŨTHENYA\n",
      "2020-02-20 10:15:38,811 Validation result at epoch  23, step    25000: bleu:  20.84, loss: 44945.0039, ppl:   5.8933, duration: 90.9866s\n",
      "2020-02-20 10:16:06,934 Epoch  23 Step:    25100 Batch Loss:     1.948351 Tokens per Sec:     7459, Lr: 0.000300\n",
      "2020-02-20 10:16:35,615 Epoch  23 Step:    25200 Batch Loss:     1.705841 Tokens per Sec:     7677, Lr: 0.000300\n",
      "2020-02-20 10:16:48,981 Epoch  23: total training loss 1907.63\n",
      "2020-02-20 10:16:48,982 EPOCH 24\n",
      "2020-02-20 10:17:04,681 Epoch  24 Step:    25300 Batch Loss:     1.650472 Tokens per Sec:     7674, Lr: 0.000300\n",
      "2020-02-20 10:17:33,143 Epoch  24 Step:    25400 Batch Loss:     1.431659 Tokens per Sec:     7648, Lr: 0.000300\n",
      "2020-02-20 10:18:01,750 Epoch  24 Step:    25500 Batch Loss:     1.697678 Tokens per Sec:     7585, Lr: 0.000300\n",
      "2020-02-20 10:18:30,337 Epoch  24 Step:    25600 Batch Loss:     1.631979 Tokens per Sec:     7506, Lr: 0.000300\n",
      "2020-02-20 10:18:59,231 Epoch  24 Step:    25700 Batch Loss:     1.899202 Tokens per Sec:     7765, Lr: 0.000300\n",
      "2020-02-20 10:19:27,995 Epoch  24 Step:    25800 Batch Loss:     1.655041 Tokens per Sec:     7704, Lr: 0.000300\n",
      "2020-02-20 10:19:57,030 Epoch  24 Step:    25900 Batch Loss:     1.849452 Tokens per Sec:     7855, Lr: 0.000300\n",
      "2020-02-20 10:20:25,752 Epoch  24 Step:    26000 Batch Loss:     1.612436 Tokens per Sec:     7664, Lr: 0.000300\n",
      "2020-02-20 10:21:56,567 Hooray! New best validation result [ppl]!\n",
      "2020-02-20 10:21:56,567 Saving new checkpoint.\n",
      "2020-02-20 10:21:56,783 Example #0\n",
      "2020-02-20 10:21:56,783 \tSource:     Researchers at the University of Copenhagen tracked almost 10,000 middle - aged people over an 11 - year period and found that participants who frequently argued with someone close to them were far more likely to die prematurely than those who seldom had conflicts .\n",
      "2020-02-20 10:21:56,783 \tReference:  Athuthuria yunivasĩtĩ - inĩ ya Copenhagen , nĩ maathuthuririe andũ hakuhĩ 10,000 a mĩaka ĩyo kwa ihinda rĩa mĩaka 11 na makĩona atĩ arĩa maakararanagia na mũndũ wa famĩlĩ kaingĩ , haakoragwo na ũhotekeku mũnene atĩ megũkua tene gũkĩra arĩa mataakoragwo na ngarari kaingĩ .\n",
      "2020-02-20 10:21:56,784 \tHypothesis: Athuthuria a maũndũ ma ndini ya Gopenaken nĩ maahutirie andũ 10,000 arĩa maikaraga mĩaka 11 na makona atĩ andũ arĩa maikaraga hakuhĩ na andũ arĩa maikaraga hakuhĩ mũno nĩ maacemanagia na mũndũ ũmwe ũmwe ũmwe warĩ na wendi mũnene wa kũigua atĩ nĩ mangĩaiguire ta ũcio nĩ mangĩaiguire ta angĩamakuĩte .\n",
      "2020-02-20 10:21:56,784 Example #1\n",
      "2020-02-20 10:21:56,784 \tSource:     Yes .\n",
      "2020-02-20 10:21:56,784 \tReference:  Ĩĩ .\n",
      "2020-02-20 10:21:56,784 \tHypothesis: Ĩĩ .\n",
      "2020-02-20 10:21:56,784 Example #2\n",
      "2020-02-20 10:21:56,784 \tSource:     He enjoyed special intimacy with Jehovah , who enabled him to display “ awesome power ” as he led the Israelites to the Promised Land . ​ — Deut .\n",
      "2020-02-20 10:21:56,784 \tReference:  Aarĩ na ũkuruhanu wa hakuhĩ na Jehova , na nĩ aamũheire ũhoti wa gwĩka “ maũndũ manene ma kũguoyohania ” rĩrĩa aatongoragia Aisiraeli mathiĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 10:21:56,784 \tHypothesis: Nĩ aakenagĩra mũno gũkorũo arĩ na ũiguano wa mwanya wa Jehova , ũrĩa wamũteithirie kuonania “ hinya ” ũrĩa aaheire Aisiraeli Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 10:21:56,785 Example #3\n",
      "2020-02-20 10:21:56,785 \tSource:     HISTORY : ATHEIST\n",
      "2020-02-20 10:21:56,785 \tReference:  HISTORĨ : EETĨKĨTIE GŨTIRĨ NGAI\n",
      "2020-02-20 10:21:56,785 \tHypothesis: HISTORĨ : MŨNDŨ WA GŨCIO\n",
      "2020-02-20 10:21:56,785 Validation result at epoch  24, step    26000: bleu:  20.91, loss: 44849.4727, ppl:   5.8711, duration: 91.0321s\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2020-02-20 10:22:25,577 Epoch  24 Step:    26100 Batch Loss:     1.864329 Tokens per Sec:     7598, Lr: 0.000300\n",
      "2020-02-20 10:22:54,292 Epoch  24 Step:    26200 Batch Loss:     2.075580 Tokens per Sec:     7762, Lr: 0.000300\n",
      "2020-02-20 10:23:23,104 Epoch  24 Step:    26300 Batch Loss:     1.820727 Tokens per Sec:     7566, Lr: 0.000300\n",
      "2020-02-20 10:23:34,862 Epoch  24: total training loss 1887.59\n",
      "2020-02-20 10:23:34,863 EPOCH 25\n",
      "2020-02-20 10:23:52,007 Epoch  25 Step:    26400 Batch Loss:     1.723013 Tokens per Sec:     7676, Lr: 0.000300\n",
      "2020-02-20 10:24:20,607 Epoch  25 Step:    26500 Batch Loss:     1.900988 Tokens per Sec:     7592, Lr: 0.000300\n",
      "2020-02-20 10:24:49,418 Epoch  25 Step:    26600 Batch Loss:     1.875849 Tokens per Sec:     7739, Lr: 0.000300\n",
      "2020-02-20 10:25:18,264 Epoch  25 Step:    26700 Batch Loss:     1.768392 Tokens per Sec:     7637, Lr: 0.000300\n",
      "2020-02-20 10:25:46,990 Epoch  25 Step:    26800 Batch Loss:     1.530995 Tokens per Sec:     7613, Lr: 0.000300\n",
      "2020-02-20 10:26:15,628 Epoch  25 Step:    26900 Batch Loss:     1.746765 Tokens per Sec:     7693, Lr: 0.000300\n",
      "2020-02-20 10:26:44,316 Epoch  25 Step:    27000 Batch Loss:     1.983916 Tokens per Sec:     7604, Lr: 0.000300\n",
      "2020-02-20 10:28:15,102 Hooray! New best validation result [ppl]!\n",
      "2020-02-20 10:28:15,103 Saving new checkpoint.\n",
      "2020-02-20 10:28:15,320 Example #0\n",
      "2020-02-20 10:28:15,321 \tSource:     Researchers at the University of Copenhagen tracked almost 10,000 middle - aged people over an 11 - year period and found that participants who frequently argued with someone close to them were far more likely to die prematurely than those who seldom had conflicts .\n",
      "2020-02-20 10:28:15,321 \tReference:  Athuthuria yunivasĩtĩ - inĩ ya Copenhagen , nĩ maathuthuririe andũ hakuhĩ 10,000 a mĩaka ĩyo kwa ihinda rĩa mĩaka 11 na makĩona atĩ arĩa maakararanagia na mũndũ wa famĩlĩ kaingĩ , haakoragwo na ũhotekeku mũnene atĩ megũkua tene gũkĩra arĩa mataakoragwo na ngarari kaingĩ .\n",
      "2020-02-20 10:28:15,321 \tHypothesis: Athuthuria a sayansi a Cophen nĩ maageririe andũ 10,000 arĩa maarĩ na mĩaka 11 na makĩona atĩ arĩa maikaraga meiguĩte nĩ maabataraga gũkorũo na ũũgĩ mũingĩ gũkĩra arĩa maikaraga Amerika .\n",
      "2020-02-20 10:28:15,321 Example #1\n",
      "2020-02-20 10:28:15,321 \tSource:     Yes .\n",
      "2020-02-20 10:28:15,321 \tReference:  Ĩĩ .\n",
      "2020-02-20 10:28:15,321 \tHypothesis: Ĩĩ , nĩ ma .\n",
      "2020-02-20 10:28:15,321 Example #2\n",
      "2020-02-20 10:28:15,321 \tSource:     He enjoyed special intimacy with Jehovah , who enabled him to display “ awesome power ” as he led the Israelites to the Promised Land . ​ — Deut .\n",
      "2020-02-20 10:28:15,322 \tReference:  Aarĩ na ũkuruhanu wa hakuhĩ na Jehova , na nĩ aamũheire ũhoti wa gwĩka “ maũndũ manene ma kũguoyohania ” rĩrĩa aatongoragia Aisiraeli mathiĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 10:28:15,322 \tHypothesis: Nĩ aakenagĩra mũno gũkorũo na ũrata wa mwanya na Jehova , ũrĩa wamũteithirie kuonania “ hinya ” ũrĩa aaheire Aisiraeli kuuma Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 10:28:15,322 Example #3\n",
      "2020-02-20 10:28:15,322 \tSource:     HISTORY : ATHEIST\n",
      "2020-02-20 10:28:15,322 \tReference:  HISTORĨ : EETĨKĨTIE GŨTIRĨ NGAI\n",
      "2020-02-20 10:28:15,322 \tHypothesis: HISTORĨ : ATHITHIMO\n",
      "2020-02-20 10:28:15,322 Validation result at epoch  25, step    27000: bleu:  21.56, loss: 44519.7773, ppl:   5.7952, duration: 91.0057s\n",
      "2020-02-20 10:28:43,877 Epoch  25 Step:    27100 Batch Loss:     1.829659 Tokens per Sec:     7593, Lr: 0.000300\n",
      "2020-02-20 10:29:12,364 Epoch  25 Step:    27200 Batch Loss:     1.834180 Tokens per Sec:     7644, Lr: 0.000300\n",
      "2020-02-20 10:29:41,058 Epoch  25 Step:    27300 Batch Loss:     1.854759 Tokens per Sec:     7607, Lr: 0.000300\n",
      "2020-02-20 10:30:09,460 Epoch  25 Step:    27400 Batch Loss:     1.410918 Tokens per Sec:     7510, Lr: 0.000300\n",
      "2020-02-20 10:30:22,264 Epoch  25: total training loss 1885.63\n",
      "2020-02-20 10:30:22,265 EPOCH 26\n",
      "2020-02-20 10:30:38,420 Epoch  26 Step:    27500 Batch Loss:     1.413156 Tokens per Sec:     7708, Lr: 0.000300\n",
      "2020-02-20 10:31:07,344 Epoch  26 Step:    27600 Batch Loss:     1.523301 Tokens per Sec:     7695, Lr: 0.000300\n",
      "2020-02-20 10:31:35,985 Epoch  26 Step:    27700 Batch Loss:     1.359455 Tokens per Sec:     7702, Lr: 0.000300\n",
      "2020-02-20 10:32:04,571 Epoch  26 Step:    27800 Batch Loss:     1.613422 Tokens per Sec:     7691, Lr: 0.000300\n",
      "2020-02-20 10:32:33,531 Epoch  26 Step:    27900 Batch Loss:     1.787205 Tokens per Sec:     7723, Lr: 0.000300\n",
      "2020-02-20 10:33:01,963 Epoch  26 Step:    28000 Batch Loss:     1.592025 Tokens per Sec:     7654, Lr: 0.000300\n",
      "2020-02-20 10:34:32,736 Hooray! New best validation result [ppl]!\n",
      "2020-02-20 10:34:32,736 Saving new checkpoint.\n",
      "2020-02-20 10:34:32,958 Example #0\n",
      "2020-02-20 10:34:32,958 \tSource:     Researchers at the University of Copenhagen tracked almost 10,000 middle - aged people over an 11 - year period and found that participants who frequently argued with someone close to them were far more likely to die prematurely than those who seldom had conflicts .\n",
      "2020-02-20 10:34:32,958 \tReference:  Athuthuria yunivasĩtĩ - inĩ ya Copenhagen , nĩ maathuthuririe andũ hakuhĩ 10,000 a mĩaka ĩyo kwa ihinda rĩa mĩaka 11 na makĩona atĩ arĩa maakararanagia na mũndũ wa famĩlĩ kaingĩ , haakoragwo na ũhotekeku mũnene atĩ megũkua tene gũkĩra arĩa mataakoragwo na ngarari kaingĩ .\n",
      "2020-02-20 10:34:32,958 \tHypothesis: Athuthuria a ndini ya Amerika ya Ngeretha nĩ maaherithirie andũ 10,000 arĩa maikaraga mĩaka 11 na nĩ monire atĩ airĩtu arĩa maikaraga hakuhĩ na andũ arĩa maikaraga Amerika nĩ maakuĩte ngoro mũno gũkĩra arĩa maakuĩte .\n",
      "2020-02-20 10:34:32,958 Example #1\n",
      "2020-02-20 10:34:32,958 \tSource:     Yes .\n",
      "2020-02-20 10:34:32,958 \tReference:  Ĩĩ .\n",
      "2020-02-20 10:34:32,958 \tHypothesis: Ĩĩ .\n",
      "2020-02-20 10:34:32,959 Example #2\n",
      "2020-02-20 10:34:32,959 \tSource:     He enjoyed special intimacy with Jehovah , who enabled him to display “ awesome power ” as he led the Israelites to the Promised Land . ​ — Deut .\n",
      "2020-02-20 10:34:32,959 \tReference:  Aarĩ na ũkuruhanu wa hakuhĩ na Jehova , na nĩ aamũheire ũhoti wa gwĩka “ maũndũ manene ma kũguoyohania ” rĩrĩa aatongoragia Aisiraeli mathiĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 10:34:32,959 \tHypothesis: Nĩ aakenagĩra mũno gũkorũo na ũrata wa mwanya hamwe na Jehova , ũrĩa wamũteithirie kuonania “ hinya ” ũrĩa aaheire Aisiraeli mathiĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 10:34:32,959 Example #3\n",
      "2020-02-20 10:34:32,959 \tSource:     HISTORY : ATHEIST\n",
      "2020-02-20 10:34:32,959 \tReference:  HISTORĨ : EETĨKĨTIE GŨTIRĨ NGAI\n",
      "2020-02-20 10:34:32,959 \tHypothesis: HISTORĨ : ATHITHĨ\n",
      "2020-02-20 10:34:32,959 Validation result at epoch  26, step    28000: bleu:  21.84, loss: 44272.7500, ppl:   5.7390, duration: 90.9952s\n",
      "2020-02-20 10:35:01,356 Epoch  26 Step:    28100 Batch Loss:     1.574245 Tokens per Sec:     7478, Lr: 0.000300\n",
      "2020-02-20 10:35:30,182 Epoch  26 Step:    28200 Batch Loss:     2.014296 Tokens per Sec:     7727, Lr: 0.000300\n",
      "2020-02-20 10:35:58,789 Epoch  26 Step:    28300 Batch Loss:     1.785048 Tokens per Sec:     7831, Lr: 0.000300\n",
      "2020-02-20 10:36:27,460 Epoch  26 Step:    28400 Batch Loss:     1.633065 Tokens per Sec:     7737, Lr: 0.000300\n",
      "2020-02-20 10:36:56,208 Epoch  26 Step:    28500 Batch Loss:     1.182130 Tokens per Sec:     7667, Lr: 0.000300\n",
      "2020-02-20 10:37:06,673 Epoch  26: total training loss 1845.57\n",
      "2020-02-20 10:37:06,674 EPOCH 27\n",
      "2020-02-20 10:37:25,465 Epoch  27 Step:    28600 Batch Loss:     1.389211 Tokens per Sec:     7785, Lr: 0.000300\n",
      "2020-02-20 10:37:53,926 Epoch  27 Step:    28700 Batch Loss:     1.641958 Tokens per Sec:     7588, Lr: 0.000300\n",
      "2020-02-20 10:38:22,489 Epoch  27 Step:    28800 Batch Loss:     1.556203 Tokens per Sec:     7664, Lr: 0.000300\n",
      "2020-02-20 10:38:51,242 Epoch  27 Step:    28900 Batch Loss:     1.782225 Tokens per Sec:     7667, Lr: 0.000300\n",
      "2020-02-20 10:39:19,787 Epoch  27 Step:    29000 Batch Loss:     1.610288 Tokens per Sec:     7767, Lr: 0.000300\n",
      "2020-02-20 10:40:50,626 Hooray! New best validation result [ppl]!\n",
      "2020-02-20 10:40:50,626 Saving new checkpoint.\n",
      "2020-02-20 10:40:50,839 Example #0\n",
      "2020-02-20 10:40:50,839 \tSource:     Researchers at the University of Copenhagen tracked almost 10,000 middle - aged people over an 11 - year period and found that participants who frequently argued with someone close to them were far more likely to die prematurely than those who seldom had conflicts .\n",
      "2020-02-20 10:40:50,840 \tReference:  Athuthuria yunivasĩtĩ - inĩ ya Copenhagen , nĩ maathuthuririe andũ hakuhĩ 10,000 a mĩaka ĩyo kwa ihinda rĩa mĩaka 11 na makĩona atĩ arĩa maakararanagia na mũndũ wa famĩlĩ kaingĩ , haakoragwo na ũhotekeku mũnene atĩ megũkua tene gũkĩra arĩa mataakoragwo na ngarari kaingĩ .\n",
      "2020-02-20 10:40:50,840 \tHypothesis: Andũ aingĩ arĩa mathuthuragia ũhoro wa ma nĩ maacemanirie na andũ makĩria ma 10,000 arĩa maikaraga Amerika ( U.S . ) , nĩ maacemanirie na andũ makĩria ma 10,000 arĩa maikaraga meiguĩte atĩ nĩ maikaraga meiguĩte mũno , na nĩ maabataraga gũkorũo na ũũgĩ mũingĩ mũno gũkĩra ũrĩa maabataraga gwĩka .\n",
      "2020-02-20 10:40:50,840 Example #1\n",
      "2020-02-20 10:40:50,840 \tSource:     Yes .\n",
      "2020-02-20 10:40:50,840 \tReference:  Ĩĩ .\n",
      "2020-02-20 10:40:50,840 \tHypothesis: Ĩĩ , nĩ ma .\n",
      "2020-02-20 10:40:50,840 Example #2\n",
      "2020-02-20 10:40:50,840 \tSource:     He enjoyed special intimacy with Jehovah , who enabled him to display “ awesome power ” as he led the Israelites to the Promised Land . ​ — Deut .\n",
      "2020-02-20 10:40:50,840 \tReference:  Aarĩ na ũkuruhanu wa hakuhĩ na Jehova , na nĩ aamũheire ũhoti wa gwĩka “ maũndũ manene ma kũguoyohania ” rĩrĩa aatongoragia Aisiraeli mathiĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 10:40:50,841 \tHypothesis: Nĩ aakenagĩra gũkorũo arĩ na ũrata wa mwanya hamwe na Jehova , ũrĩa wamũhotithirie kuonania “ hinya ” ũrĩa aaheire Aisiraeli kuuma Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 10:40:50,841 Example #3\n",
      "2020-02-20 10:40:50,841 \tSource:     HISTORY : ATHEIST\n",
      "2020-02-20 10:40:50,841 \tReference:  HISTORĨ : EETĨKĨTIE GŨTIRĨ NGAI\n",
      "2020-02-20 10:40:50,841 \tHypothesis: HISTORĨ : MŨNDŨ WA MŨTŨŨRĨRĨRIA\n",
      "2020-02-20 10:40:50,841 Validation result at epoch  27, step    29000: bleu:  21.95, loss: 43975.0117, ppl:   5.6720, duration: 91.0531s\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2020-02-20 10:41:19,132 Epoch  27 Step:    29100 Batch Loss:     1.771853 Tokens per Sec:     7520, Lr: 0.000300\n",
      "2020-02-20 10:41:47,661 Epoch  27 Step:    29200 Batch Loss:     1.573860 Tokens per Sec:     7592, Lr: 0.000300\n",
      "2020-02-20 10:42:16,152 Epoch  27 Step:    29300 Batch Loss:     1.881424 Tokens per Sec:     7574, Lr: 0.000300\n",
      "2020-02-20 10:42:44,940 Epoch  27 Step:    29400 Batch Loss:     1.846929 Tokens per Sec:     7745, Lr: 0.000300\n",
      "2020-02-20 10:43:13,726 Epoch  27 Step:    29500 Batch Loss:     1.933888 Tokens per Sec:     7783, Lr: 0.000300\n",
      "2020-02-20 10:43:42,499 Epoch  27 Step:    29600 Batch Loss:     1.905533 Tokens per Sec:     7771, Lr: 0.000300\n",
      "2020-02-20 10:43:52,443 Epoch  27: total training loss 1841.99\n",
      "2020-02-20 10:43:52,443 EPOCH 28\n",
      "2020-02-20 10:44:11,122 Epoch  28 Step:    29700 Batch Loss:     1.790683 Tokens per Sec:     7548, Lr: 0.000300\n",
      "2020-02-20 10:44:40,124 Epoch  28 Step:    29800 Batch Loss:     1.714535 Tokens per Sec:     7697, Lr: 0.000300\n",
      "2020-02-20 10:45:08,739 Epoch  28 Step:    29900 Batch Loss:     1.724976 Tokens per Sec:     7549, Lr: 0.000300\n",
      "2020-02-20 10:45:37,637 Epoch  28 Step:    30000 Batch Loss:     1.460413 Tokens per Sec:     7724, Lr: 0.000300\n",
      "2020-02-20 10:47:08,362 Hooray! New best validation result [ppl]!\n",
      "2020-02-20 10:47:08,363 Saving new checkpoint.\n",
      "2020-02-20 10:47:08,581 Example #0\n",
      "2020-02-20 10:47:08,581 \tSource:     Researchers at the University of Copenhagen tracked almost 10,000 middle - aged people over an 11 - year period and found that participants who frequently argued with someone close to them were far more likely to die prematurely than those who seldom had conflicts .\n",
      "2020-02-20 10:47:08,581 \tReference:  Athuthuria yunivasĩtĩ - inĩ ya Copenhagen , nĩ maathuthuririe andũ hakuhĩ 10,000 a mĩaka ĩyo kwa ihinda rĩa mĩaka 11 na makĩona atĩ arĩa maakararanagia na mũndũ wa famĩlĩ kaingĩ , haakoragwo na ũhotekeku mũnene atĩ megũkua tene gũkĩra arĩa mataakoragwo na ngarari kaingĩ .\n",
      "2020-02-20 10:47:08,581 \tHypothesis: Athuthuria a ndini iria cieĩtaga cia Cophaginen nĩ maahũrire andũ 10,000 arĩa maikaraga mĩaka 11 na makona atĩ andũ arĩa maaikaraga nyũmba kwa nyũmba nĩ maabataraga kũraihu na andũ aingĩ mũno , na nĩ maabataraga kũrĩha mũno gũkĩra arĩa maakuire .\n",
      "2020-02-20 10:47:08,581 Example #1\n",
      "2020-02-20 10:47:08,582 \tSource:     Yes .\n",
      "2020-02-20 10:47:08,582 \tReference:  Ĩĩ .\n",
      "2020-02-20 10:47:08,582 \tHypothesis: Ĩĩ , nĩ ma .\n",
      "2020-02-20 10:47:08,582 Example #2\n",
      "2020-02-20 10:47:08,582 \tSource:     He enjoyed special intimacy with Jehovah , who enabled him to display “ awesome power ” as he led the Israelites to the Promised Land . ​ — Deut .\n",
      "2020-02-20 10:47:08,582 \tReference:  Aarĩ na ũkuruhanu wa hakuhĩ na Jehova , na nĩ aamũheire ũhoti wa gwĩka “ maũndũ manene ma kũguoyohania ” rĩrĩa aatongoragia Aisiraeli mathiĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 10:47:08,583 \tHypothesis: Nĩ aakenagĩra gũkorũo arĩ na ũrata wa mwanya na Jehova , ũrĩa wamũteithirie kuonania “ hinya ” ũrĩa aaheire Aisiraeli kuuma Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 10:47:08,583 Example #3\n",
      "2020-02-20 10:47:08,583 \tSource:     HISTORY : ATHEIST\n",
      "2020-02-20 10:47:08,583 \tReference:  HISTORĨ : EETĨKĨTIE GŨTIRĨ NGAI\n",
      "2020-02-20 10:47:08,583 \tHypothesis: HISTORĨ : AATHITHITHIA\n",
      "2020-02-20 10:47:08,583 Validation result at epoch  28, step    30000: bleu:  22.04, loss: 43765.9258, ppl:   5.6254, duration: 90.9454s\n",
      "2020-02-20 10:47:37,547 Epoch  28 Step:    30100 Batch Loss:     1.974211 Tokens per Sec:     7690, Lr: 0.000300\n",
      "2020-02-20 10:48:06,116 Epoch  28 Step:    30200 Batch Loss:     1.686540 Tokens per Sec:     7764, Lr: 0.000300\n",
      "2020-02-20 10:48:34,775 Epoch  28 Step:    30300 Batch Loss:     1.586944 Tokens per Sec:     7724, Lr: 0.000300\n",
      "2020-02-20 10:49:03,398 Epoch  28 Step:    30400 Batch Loss:     1.704639 Tokens per Sec:     7629, Lr: 0.000300\n",
      "2020-02-20 10:49:32,144 Epoch  28 Step:    30500 Batch Loss:     1.779434 Tokens per Sec:     7681, Lr: 0.000300\n",
      "2020-02-20 10:50:00,711 Epoch  28 Step:    30600 Batch Loss:     1.921632 Tokens per Sec:     7526, Lr: 0.000300\n",
      "2020-02-20 10:50:29,432 Epoch  28 Step:    30700 Batch Loss:     1.641440 Tokens per Sec:     7665, Lr: 0.000300\n",
      "2020-02-20 10:50:38,833 Epoch  28: total training loss 1824.33\n",
      "2020-02-20 10:50:38,834 EPOCH 29\n",
      "2020-02-20 10:50:58,256 Epoch  29 Step:    30800 Batch Loss:     1.743072 Tokens per Sec:     7640, Lr: 0.000300\n",
      "2020-02-20 10:51:27,084 Epoch  29 Step:    30900 Batch Loss:     1.748749 Tokens per Sec:     7730, Lr: 0.000300\n",
      "2020-02-20 10:51:55,714 Epoch  29 Step:    31000 Batch Loss:     1.765801 Tokens per Sec:     7622, Lr: 0.000300\n",
      "2020-02-20 10:53:26,572 Hooray! New best validation result [ppl]!\n",
      "2020-02-20 10:53:26,573 Saving new checkpoint.\n",
      "2020-02-20 10:53:26,791 Example #0\n",
      "2020-02-20 10:53:26,792 \tSource:     Researchers at the University of Copenhagen tracked almost 10,000 middle - aged people over an 11 - year period and found that participants who frequently argued with someone close to them were far more likely to die prematurely than those who seldom had conflicts .\n",
      "2020-02-20 10:53:26,792 \tReference:  Athuthuria yunivasĩtĩ - inĩ ya Copenhagen , nĩ maathuthuririe andũ hakuhĩ 10,000 a mĩaka ĩyo kwa ihinda rĩa mĩaka 11 na makĩona atĩ arĩa maakararanagia na mũndũ wa famĩlĩ kaingĩ , haakoragwo na ũhotekeku mũnene atĩ megũkua tene gũkĩra arĩa mataakoragwo na ngarari kaingĩ .\n",
      "2020-02-20 10:53:26,792 \tHypothesis: Athuthuria a ndini iria ciekĩirũo Amerika ( US ) nĩ maahũũrire andũ 10,000 arĩa maarĩ na mĩaka 1,000 na nĩ moonire atĩ andũ arĩa maikaraga maikaraga mateiguĩte nĩ maabataraga kũraihanĩrĩirie na mũndũ o wothe ũngĩamakuaga .\n",
      "2020-02-20 10:53:26,792 Example #1\n",
      "2020-02-20 10:53:26,792 \tSource:     Yes .\n",
      "2020-02-20 10:53:26,792 \tReference:  Ĩĩ .\n",
      "2020-02-20 10:53:26,792 \tHypothesis: Ĩĩ .\n",
      "2020-02-20 10:53:26,792 Example #2\n",
      "2020-02-20 10:53:26,792 \tSource:     He enjoyed special intimacy with Jehovah , who enabled him to display “ awesome power ” as he led the Israelites to the Promised Land . ​ — Deut .\n",
      "2020-02-20 10:53:26,792 \tReference:  Aarĩ na ũkuruhanu wa hakuhĩ na Jehova , na nĩ aamũheire ũhoti wa gwĩka “ maũndũ manene ma kũguoyohania ” rĩrĩa aatongoragia Aisiraeli mathiĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 10:53:26,793 \tHypothesis: Nĩ aakenagĩra gũkorũo na ũrata wa mwanya hamwe na Jehova , ũrĩa wamũteithirie kuonania “ hinya wa kũmenya ” rĩrĩa aatũmĩte Aisiraeli mathiĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 10:53:26,793 Example #3\n",
      "2020-02-20 10:53:26,793 \tSource:     HISTORY : ATHEIST\n",
      "2020-02-20 10:53:26,793 \tReference:  HISTORĨ : EETĨKĨTIE GŨTIRĨ NGAI\n",
      "2020-02-20 10:53:26,793 \tHypothesis: HISTORĨ : ATHITHIO\n",
      "2020-02-20 10:53:26,793 Validation result at epoch  29, step    31000: bleu:  21.95, loss: 43635.5312, ppl:   5.5965, duration: 91.0785s\n",
      "2020-02-20 10:53:55,150 Epoch  29 Step:    31100 Batch Loss:     1.534370 Tokens per Sec:     7511, Lr: 0.000300\n",
      "2020-02-20 10:54:23,846 Epoch  29 Step:    31200 Batch Loss:     1.992012 Tokens per Sec:     7703, Lr: 0.000300\n",
      "2020-02-20 10:54:52,553 Epoch  29 Step:    31300 Batch Loss:     1.876786 Tokens per Sec:     7530, Lr: 0.000300\n",
      "2020-02-20 10:55:21,365 Epoch  29 Step:    31400 Batch Loss:     1.750656 Tokens per Sec:     7723, Lr: 0.000300\n",
      "2020-02-20 10:55:50,094 Epoch  29 Step:    31500 Batch Loss:     1.856229 Tokens per Sec:     7601, Lr: 0.000300\n",
      "2020-02-20 10:56:18,738 Epoch  29 Step:    31600 Batch Loss:     1.482456 Tokens per Sec:     7722, Lr: 0.000300\n",
      "2020-02-20 10:56:47,468 Epoch  29 Step:    31700 Batch Loss:     2.071452 Tokens per Sec:     7725, Lr: 0.000300\n",
      "2020-02-20 10:57:16,108 Epoch  29 Step:    31800 Batch Loss:     1.545179 Tokens per Sec:     7619, Lr: 0.000300\n",
      "2020-02-20 10:57:25,315 Epoch  29: total training loss 1810.48\n",
      "2020-02-20 10:57:25,315 EPOCH 30\n",
      "2020-02-20 10:57:44,711 Epoch  30 Step:    31900 Batch Loss:     1.588805 Tokens per Sec:     7604, Lr: 0.000300\n",
      "2020-02-20 10:58:13,289 Epoch  30 Step:    32000 Batch Loss:     1.289834 Tokens per Sec:     7524, Lr: 0.000300\n",
      "2020-02-20 10:59:44,111 Hooray! New best validation result [ppl]!\n",
      "2020-02-20 10:59:44,111 Saving new checkpoint.\n",
      "2020-02-20 10:59:44,334 Example #0\n",
      "2020-02-20 10:59:44,335 \tSource:     Researchers at the University of Copenhagen tracked almost 10,000 middle - aged people over an 11 - year period and found that participants who frequently argued with someone close to them were far more likely to die prematurely than those who seldom had conflicts .\n",
      "2020-02-20 10:59:44,335 \tReference:  Athuthuria yunivasĩtĩ - inĩ ya Copenhagen , nĩ maathuthuririe andũ hakuhĩ 10,000 a mĩaka ĩyo kwa ihinda rĩa mĩaka 11 na makĩona atĩ arĩa maakararanagia na mũndũ wa famĩlĩ kaingĩ , haakoragwo na ũhotekeku mũnene atĩ megũkua tene gũkĩra arĩa mataakoragwo na ngarari kaingĩ .\n",
      "2020-02-20 10:59:44,335 \tHypothesis: Athuthuria a maũndũ marĩa meekirũo nĩ andũ 10,000 arĩa maikaraga Amerika ( US ) nĩ meethagathagathagĩte mũno nĩ ũndũ wa andũ 10,000 arĩa maikaraga mĩaka 11 na nĩ moonire atĩ andũ arĩa mahoyaga mahoyaga kaingĩ nĩ maakuĩte .\n",
      "2020-02-20 10:59:44,335 Example #1\n",
      "2020-02-20 10:59:44,335 \tSource:     Yes .\n",
      "2020-02-20 10:59:44,335 \tReference:  Ĩĩ .\n",
      "2020-02-20 10:59:44,335 \tHypothesis: Ĩĩ .\n",
      "2020-02-20 10:59:44,335 Example #2\n",
      "2020-02-20 10:59:44,336 \tSource:     He enjoyed special intimacy with Jehovah , who enabled him to display “ awesome power ” as he led the Israelites to the Promised Land . ​ — Deut .\n",
      "2020-02-20 10:59:44,336 \tReference:  Aarĩ na ũkuruhanu wa hakuhĩ na Jehova , na nĩ aamũheire ũhoti wa gwĩka “ maũndũ manene ma kũguoyohania ” rĩrĩa aatongoragia Aisiraeli mathiĩ Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 10:59:44,336 \tHypothesis: Nĩ aakenagĩra mũno gũkorũo arĩ mwĩhokeku harĩ Jehova , ũrĩa wamũteithirie kuonania “ hinya ” ũrĩa aaheete Aisiraeli nĩguo maatũme Bũrũri wa Kĩĩranĩro . — Gũcok .\n",
      "2020-02-20 10:59:44,336 Example #3\n",
      "2020-02-20 10:59:44,336 \tSource:     HISTORY : ATHEIST\n",
      "2020-02-20 10:59:44,336 \tReference:  HISTORĨ : EETĨKĨTIE GŨTIRĨ NGAI\n",
      "2020-02-20 10:59:44,336 \tHypothesis: HISTORĨ : AATHITHIE\n",
      "2020-02-20 10:59:44,336 Validation result at epoch  30, step    32000: bleu:  22.59, loss: 43360.7656, ppl:   5.5361, duration: 91.0465s\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2020-02-20 11:00:12,807 Epoch  30 Step:    32100 Batch Loss:     1.610383 Tokens per Sec:     7718, Lr: 0.000300\n",
      "2020-02-20 11:00:41,777 Epoch  30 Step:    32200 Batch Loss:     1.876286 Tokens per Sec:     7646, Lr: 0.000300\n",
      "2020-02-20 11:01:10,048 Epoch  30 Step:    32300 Batch Loss:     1.924929 Tokens per Sec:     7672, Lr: 0.000300\n",
      "2020-02-20 11:01:38,556 Epoch  30 Step:    32400 Batch Loss:     1.818684 Tokens per Sec:     7576, Lr: 0.000300\n",
      "2020-02-20 11:02:07,424 Epoch  30 Step:    32500 Batch Loss:     1.408155 Tokens per Sec:     7746, Lr: 0.000300\n",
      "2020-02-20 11:02:36,124 Epoch  30 Step:    32600 Batch Loss:     1.853153 Tokens per Sec:     7527, Lr: 0.000300\n",
      "2020-02-20 11:03:04,892 Epoch  30 Step:    32700 Batch Loss:     1.727104 Tokens per Sec:     7646, Lr: 0.000300\n",
      "2020-02-20 11:03:33,842 Epoch  30 Step:    32800 Batch Loss:     2.103340 Tokens per Sec:     7715, Lr: 0.000300\n",
      "2020-02-20 11:04:02,603 Epoch  30 Step:    32900 Batch Loss:     1.930896 Tokens per Sec:     7685, Lr: 0.000300\n",
      "2020-02-20 11:04:11,941 Epoch  30: total training loss 1796.41\n",
      "2020-02-20 11:04:11,941 Training ended after  30 epochs.\n",
      "2020-02-20 11:04:11,941 Best validation result at step    32000:   5.54 ppl.\n",
      "2020-02-20 11:05:01,199  dev bleu:  23.83 [Beam search decoding with beam size = 5 and alpha = 1.0]\n",
      "2020-02-20 11:05:01,201 Translations saved to: models/enki_transformer/00032000.hyps.dev\n",
      "2020-02-20 11:06:32,445 test bleu:  36.06 [Beam search decoding with beam size = 5 and alpha = 1.0]\n",
      "2020-02-20 11:06:32,447 Translations saved to: models/enki_transformer/00032000.hyps.test\n"
     ]
    }
   ],
   "source": [
    "# Train the model\n",
    "# You can press Ctrl-C to stop. And then run the next cell to save your checkpoints! \n",
    "! cd ../../joeynmt; python3 -m joeynmt train configs/transformer_$src$tgt.yaml"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "collapsed": true,
    "id": "MBoDS09JM807"
   },
   "outputs": [],
   "source": [
    "# Copy the created models from the notebook storage to google drive for persistant storage \n",
    "!cp -r joeynmt/models/${src}${tgt}_transformer/* \"$gdrive_path/models/${src}${tgt}_transformer/\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "n94wlrCjVc17"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Steps: 1000\tLoss: 99303.99219\tPPL: 50.35875\tbleu: 1.98324\tLR: 0.00030000\t*\r\n",
      "Steps: 2000\tLoss: 83300.25781\tPPL: 26.77751\tbleu: 3.17995\tLR: 0.00030000\t*\r\n",
      "Steps: 3000\tLoss: 74742.76562\tPPL: 19.10268\tbleu: 5.37697\tLR: 0.00030000\t*\r\n",
      "Steps: 4000\tLoss: 69263.21094\tPPL: 15.38773\tbleu: 8.07771\tLR: 0.00030000\t*\r\n",
      "Steps: 5000\tLoss: 65603.03906\tPPL: 13.31800\tbleu: 9.66573\tLR: 0.00030000\t*\r\n",
      "Steps: 6000\tLoss: 62489.73438\tPPL: 11.77814\tbleu: 11.01826\tLR: 0.00030000\t*\r\n",
      "Steps: 7000\tLoss: 60542.17578\tPPL: 10.90676\tbleu: 12.02475\tLR: 0.00030000\t*\r\n",
      "Steps: 8000\tLoss: 58300.29297\tPPL: 9.98320\tbleu: 13.29180\tLR: 0.00030000\t*\r\n",
      "Steps: 9000\tLoss: 56454.33594\tPPL: 9.28175\tbleu: 13.84273\tLR: 0.00030000\t*\r\n",
      "Steps: 10000\tLoss: 54946.10156\tPPL: 8.74538\tbleu: 14.93520\tLR: 0.00030000\t*\r\n",
      "Steps: 11000\tLoss: 53487.99219\tPPL: 8.25632\tbleu: 15.73370\tLR: 0.00030000\t*\r\n",
      "Steps: 12000\tLoss: 52413.11719\tPPL: 7.91340\tbleu: 16.22676\tLR: 0.00030000\t*\r\n",
      "Steps: 13000\tLoss: 51187.14453\tPPL: 7.53963\tbleu: 17.23973\tLR: 0.00030000\t*\r\n",
      "Steps: 14000\tLoss: 50359.16406\tPPL: 7.29724\tbleu: 17.69230\tLR: 0.00030000\t*\r\n",
      "Steps: 15000\tLoss: 49436.13281\tPPL: 7.03619\tbleu: 17.98120\tLR: 0.00030000\t*\r\n",
      "Steps: 16000\tLoss: 48922.70312\tPPL: 6.89505\tbleu: 18.67521\tLR: 0.00030000\t*\r\n",
      "Steps: 17000\tLoss: 48392.40625\tPPL: 6.75224\tbleu: 19.09471\tLR: 0.00030000\t*\r\n",
      "Steps: 18000\tLoss: 47702.10547\tPPL: 6.57077\tbleu: 19.45177\tLR: 0.00030000\t*\r\n",
      "Steps: 19000\tLoss: 48030.95312\tPPL: 6.65660\tbleu: 19.01308\tLR: 0.00030000\t\r\n",
      "Steps: 20000\tLoss: 46900.69141\tPPL: 6.36620\tbleu: 20.03544\tLR: 0.00030000\t*\r\n",
      "Steps: 21000\tLoss: 46317.41016\tPPL: 6.22132\tbleu: 20.23505\tLR: 0.00030000\t*\r\n",
      "Steps: 22000\tLoss: 46154.44141\tPPL: 6.18143\tbleu: 20.78092\tLR: 0.00030000\t*\r\n",
      "Steps: 23000\tLoss: 45510.01953\tPPL: 6.02620\tbleu: 20.72684\tLR: 0.00030000\t*\r\n",
      "Steps: 24000\tLoss: 45274.80078\tPPL: 5.97052\tbleu: 21.00866\tLR: 0.00030000\t*\r\n",
      "Steps: 25000\tLoss: 44945.00391\tPPL: 5.89331\tbleu: 20.83810\tLR: 0.00030000\t*\r\n",
      "Steps: 26000\tLoss: 44849.47266\tPPL: 5.87113\tbleu: 20.91362\tLR: 0.00030000\t*\r\n",
      "Steps: 27000\tLoss: 44519.77734\tPPL: 5.79523\tbleu: 21.55712\tLR: 0.00030000\t*\r\n",
      "Steps: 28000\tLoss: 44272.75000\tPPL: 5.73901\tbleu: 21.84418\tLR: 0.00030000\t*\r\n",
      "Steps: 29000\tLoss: 43975.01172\tPPL: 5.67197\tbleu: 21.95025\tLR: 0.00030000\t*\r\n",
      "Steps: 30000\tLoss: 43765.92578\tPPL: 5.62535\tbleu: 22.03771\tLR: 0.00030000\t*\r\n",
      "Steps: 31000\tLoss: 43635.53125\tPPL: 5.59648\tbleu: 21.95044\tLR: 0.00030000\t*\r\n",
      "Steps: 32000\tLoss: 43360.76562\tPPL: 5.53612\tbleu: 22.59278\tLR: 0.00030000\t*\r\n"
     ]
    }
   ],
   "source": [
    "# Output our validation accuracy\n",
    "! cat \"../../joeynmt/models/${src}${tgt}_transformer/validations.txt\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "66WhRE9lIhoD"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "/home/espoir_mur_gmail_com/.local/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:541: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_qint8 = np.dtype([(\"qint8\", np.int8, 1)])\n",
      "/home/espoir_mur_gmail_com/.local/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:542: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_quint8 = np.dtype([(\"quint8\", np.uint8, 1)])\n",
      "/home/espoir_mur_gmail_com/.local/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:543: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_qint16 = np.dtype([(\"qint16\", np.int16, 1)])\n",
      "/home/espoir_mur_gmail_com/.local/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:544: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_quint16 = np.dtype([(\"quint16\", np.uint16, 1)])\n",
      "/home/espoir_mur_gmail_com/.local/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:545: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_qint32 = np.dtype([(\"qint32\", np.int32, 1)])\n",
      "/home/espoir_mur_gmail_com/.local/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:550: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  np_resource = np.dtype([(\"resource\", np.ubyte, 1)])\n",
      "2020-02-20 12:00:28,924 -  dev bleu:  23.83 [Beam search decoding with beam size = 5 and alpha = 1.0]\n"
     ]
    }
   ],
   "source": [
    "# Test our model\n",
    "! cd ../../joeynmt; python3 -m joeynmt test \"models/${src}${tgt}_transformer/config.yaml\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": []
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "collapsed_sections": [],
   "name": "starter_notebook.ipynb",
   "provenance": [],
   "toc_visible": true
  },
  "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": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}