{
 "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": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "oGRmDELn7Az0",
    "outputId": "5dfcaf0a-4e8d-4f84-a1fa-42940b6f2252"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Drive already mounted at /content/drive; to attempt to forcibly remount, call drive.mount(\"/content/drive\", force_remount=True).\n"
     ]
    }
   ],
   "source": [
    "from google.colab import drive\n",
    "drive.mount('/content/drive')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "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 = \"fon\" \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": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "kBSgJHEw7Nvx",
    "outputId": "f006df06-8d99-432e-d9f5-1b2227b2d1f6"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "/content/drive/My Drive/masakhane/en-fon-baseline\n"
     ]
    }
   ],
   "source": [
    "!echo $gdrive_path"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 102
    },
    "colab_type": "code",
    "id": "gA75Fs9ys8Y9",
    "outputId": "93dc8f60-2de8-45ba-9bb5-0afd9cf3b358"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Collecting opustools-pkg\n",
      "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/6c/9f/e829a0cceccc603450cd18e1ff80807b6237a88d9a8df2c0bb320796e900/opustools_pkg-0.0.52-py3-none-any.whl (80kB)\n",
      "\r",
      "\u001b[K     |████                            | 10kB 25.2MB/s eta 0:00:01\r",
      "\u001b[K     |████████                        | 20kB 4.5MB/s eta 0:00:01\r",
      "\u001b[K     |████████████▏                   | 30kB 6.3MB/s eta 0:00:01\r",
      "\u001b[K     |████████████████▏               | 40kB 8.1MB/s eta 0:00:01\r",
      "\u001b[K     |████████████████████▎           | 51kB 5.2MB/s eta 0:00:01\r",
      "\u001b[K     |████████████████████████▎       | 61kB 6.1MB/s eta 0:00:01\r",
      "\u001b[K     |████████████████████████████▎   | 71kB 6.9MB/s eta 0:00:01\r",
      "\u001b[K     |████████████████████████████████| 81kB 4.8MB/s \n",
      "\u001b[?25hInstalling collected packages: opustools-pkg\n",
      "Successfully installed opustools-pkg-0.0.52\n"
     ]
    }
   ],
   "source": [
    "# Install opus-tools\n",
    "! pip install opustools-pkg"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 187
    },
    "colab_type": "code",
    "id": "xq-tDZVks7ZD",
    "outputId": "e7343c5b-3bc2-4979-8fb3-b75b5a025164"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Alignment file /proj/nlpl/data/OPUS/JW300/latest/xml/en-fon.xml.gz not found. The following files are available for downloading:\n",
      "\n",
      "        ./JW300_latest_xml_fon.zip already exists\n",
      " 324 KB https://object.pouta.csc.fi/OPUS-JW300/v1/xml/en-fon.xml.gz\n",
      " 263 MB https://object.pouta.csc.fi/OPUS-JW300/v1/xml/en.zip\n",
      "\n",
      " 263 MB Total size\n",
      "./JW300_latest_xml_en-fon.xml.gz ... 100% of 324 KB\n",
      "./JW300_latest_xml_en.zip ... 100% of 263 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": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 578
    },
    "colab_type": "code",
    "id": "n48GDRnP8y2G",
    "outputId": "ed67a5b7-a1db-4d8b-bb37-58111706be8a"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "--2019-11-26 13:32:06--  https://raw.githubusercontent.com/juliakreutzer/masakhane/master/jw300_utils/test/test.en-any.en\n",
      "Resolving raw.githubusercontent.com (raw.githubusercontent.com)... 151.101.0.133, 151.101.64.133, 151.101.128.133, ...\n",
      "Connecting to raw.githubusercontent.com (raw.githubusercontent.com)|151.101.0.133|: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",
      "\r",
      "test.en-any.en        0%[                    ]       0  --.-KB/s               \r",
      "test.en-any.en      100%[===================>] 271.28K  --.-KB/s    in 0.02s   \n",
      "\n",
      "2019-11-26 13:32:06 (12.5 MB/s) - ‘test.en-any.en’ saved [277791/277791]\n",
      "\n",
      "--2019-11-26 13:32:09--  https://raw.githubusercontent.com/juliakreutzer/masakhane/master/jw300_utils/test/test.en-fon.en\n",
      "Resolving raw.githubusercontent.com (raw.githubusercontent.com)... 151.101.0.133, 151.101.64.133, 151.101.128.133, ...\n",
      "Connecting to raw.githubusercontent.com (raw.githubusercontent.com)|151.101.0.133|:443... connected.\n",
      "HTTP request sent, awaiting response... 200 OK\n",
      "Length: 206191 (201K) [text/plain]\n",
      "Saving to: ‘test.en-fon.en’\n",
      "\n",
      "test.en-fon.en      100%[===================>] 201.36K  --.-KB/s    in 0.02s   \n",
      "\n",
      "2019-11-26 13:32:10 (11.5 MB/s) - ‘test.en-fon.en’ saved [206191/206191]\n",
      "\n",
      "--2019-11-26 13:32:16--  https://raw.githubusercontent.com/juliakreutzer/masakhane/master/jw300_utils/test/test.en-fon.fon\n",
      "Resolving raw.githubusercontent.com (raw.githubusercontent.com)... 151.101.0.133, 151.101.64.133, 151.101.128.133, ...\n",
      "Connecting to raw.githubusercontent.com (raw.githubusercontent.com)|151.101.0.133|:443... connected.\n",
      "HTTP request sent, awaiting response... 200 OK\n",
      "Length: 266332 (260K) [text/plain]\n",
      "Saving to: ‘test.en-fon.fon’\n",
      "\n",
      "test.en-fon.fon     100%[===================>] 260.09K  --.-KB/s    in 0.02s   \n",
      "\n",
      "2019-11-26 13:32:17 (11.2 MB/s) - ‘test.en-fon.fon’ saved [266332/266332]\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": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "NqDG-CI28y2L",
    "outputId": "1cae02eb-e883-4be6-a4ae-a55de11f0b8a"
   },
   "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": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 159
    },
    "colab_type": "code",
    "id": "3CNdwLBCfSIl",
    "outputId": "a8139b78-0751-4893-91dc-306662889805"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loaded data and skipped 4199/34746 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>While reviewing a financial account with his s...</td>\n",
       "      <td>Hwenu e é ɖò akwɛzinzan ɖé lɛ́n xá gǎn tɔn wɛ ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>The interviewer responded , “ Your way of thin...</td>\n",
       "      <td>Mɛ e ɖò nǔ kanbyɔ ɛ wɛ é ɖɔ n’i ɖɔ : “ Linlin ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>One form of dishonesty that is particularly wi...</td>\n",
       "      <td>Nugbǒmaɖɔ sín alɔkpa e gbakpé tawun ɖò égbé lɛ...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                     source_sentence                                    target_sentence\n",
       "0  While reviewing a financial account with his s...  Hwenu e é ɖò akwɛzinzan ɖé lɛ́n xá gǎn tɔn wɛ ...\n",
       "1  The interviewer responded , “ Your way of thin...  Mɛ e ɖò nǔ kanbyɔ ɛ wɛ é ɖɔ n’i ɖɔ : “ Linlin ...\n",
       "2  One form of dishonesty that is particularly wi...  Nugbǒmaɖɔ sín alɔkpa e gbakpé tawun ɖò égbé lɛ..."
      ]
     },
     "execution_count": 11,
     "metadata": {
      "tags": []
     },
     "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",
    "# if you get TypeError: data argument can't be an iterator is because of your zip version run this below\n",
    "#df = pd.DataFrame(list(zip(source, target)), columns=['source_sentence', 'target_sentence'])\n",
    "df.head(3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "hvgFJhMy6178",
    "outputId": "c6239ec1-8975-462f-d70a-aeb3af66cc03"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'When two people trust each other , they feel safe and secure .'"
      ]
     },
     "execution_count": 15,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.source_sentence[10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "luzCebRY7FP0",
    "outputId": "1a013324-6412-4593-faf7-fff247d3bd86"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'Hwenu e mɛ wè ɖeji dó yeɖée lɛ wu é ɔ , ayi yetɔn nɔ j’ayǐ dó yeɖée lɛ wu .'"
      ]
     },
     "execution_count": 16,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.target_sentence[10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "TM6Uy1-W7Pxw",
    "outputId": "7c13e867-5e88-419e-bbd4-ab5224a29e99"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "30548"
      ]
     },
     "execution_count": 14,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(df)"
   ]
  },
  {
   "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": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 187
    },
    "colab_type": "code",
    "id": "M_2ouEOH1_1q",
    "outputId": "525b52d6-75d8-48ef-9d07-c169a071f614"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.6/dist-packages/ipykernel_launcher.py:6: 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:7: 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",
      "  import sys\n"
     ]
    }
   ],
   "source": [
    "# drop duplicate translations\n",
    "df_pp = df.drop_duplicates()\n",
    "\n",
    "# drop conflicting translations\n",
    "# (this is optional and something that you might want to comment out \n",
    "# depending on the size of your corpus)\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": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 819
    },
    "colab_type": "code",
    "id": "Z_1BwAApEtMk",
    "outputId": "7a7e4433-fe37-4424-d6e9-c4b0b237f648"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Collecting fuzzywuzzy\n",
      "  Downloading https://files.pythonhosted.org/packages/d8/f1/5a267addb30ab7eaa1beab2b9323073815da4551076554ecc890a3595ec9/fuzzywuzzy-0.17.0-py2.py3-none-any.whl\n",
      "Installing collected packages: fuzzywuzzy\n",
      "Successfully installed fuzzywuzzy-0.17.0\n",
      "Collecting python-Levenshtein\n",
      "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/42/a9/d1785c85ebf9b7dfacd08938dd028209c34a0ea3b1bcdb895208bd40a67d/python-Levenshtein-0.12.0.tar.gz (48kB)\n",
      "\u001b[K     |████████████████████████████████| 51kB 3.0MB/s \n",
      "\u001b[?25hRequirement already satisfied: setuptools in /usr/local/lib/python3.6/dist-packages (from python-Levenshtein) (41.6.0)\n",
      "Building wheels for collected packages: python-Levenshtein\n",
      "  Building wheel for python-Levenshtein (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
      "  Created wheel for python-Levenshtein: filename=python_Levenshtein-0.12.0-cp36-cp36m-linux_x86_64.whl size=144666 sha256=68318040f392110ed6edcb5d5e3370b72f980fb5d1e118d8b8fe41ad12d0eb72\n",
      "  Stored in directory: /root/.cache/pip/wheels/de/c2/93/660fd5f7559049268ad2dc6d81c4e39e9e36518766eaf7e342\n",
      "Successfully built python-Levenshtein\n",
      "Installing collected packages: python-Levenshtein\n",
      "Successfully installed python-Levenshtein-0.12.0\n",
      "00:00:00.02 0.00 percent complete\n",
      "00:00:24.63 3.49 percent complete\n",
      "00:00:49.66 6.99 percent complete\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:root:Applied processor reduces input query to empty string, all comparisons will have score 0. [Query: '']\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "00:01:16.43 10.48 percent complete\n",
      "00:01:41.96 13.97 percent complete\n",
      "00:02:06.36 17.47 percent complete\n",
      "00:02:30.61 20.96 percent complete\n",
      "00:02:55.65 24.45 percent complete\n",
      "00:03:19.86 27.95 percent complete\n",
      "00:03:45.11 31.44 percent complete\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:root:Applied processor reduces input query to empty string, all comparisons will have score 0. [Query: '*']\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "00:04:09.90 34.93 percent complete\n",
      "00:04:36.93 38.43 percent complete\n",
      "00:05:01.21 41.92 percent complete\n",
      "00:05:25.80 45.41 percent complete\n",
      "00:05:50.23 48.91 percent complete\n",
      "00:06:15.55 52.40 percent complete\n",
      "00:06:39.99 55.89 percent complete\n",
      "00:07:05.06 59.39 percent complete\n",
      "00:07:29.83 62.88 percent complete\n",
      "00:07:56.33 66.37 percent complete\n",
      "00:08:20.68 69.87 percent complete\n",
      "00:08:45.66 73.36 percent complete\n",
      "00:09:10.41 76.85 percent complete\n",
      "00:09:35.16 80.35 percent complete\n",
      "00:10:00.10 83.84 percent complete\n",
      "00:10:24.67 87.33 percent complete\n",
      "00:10:49.82 90.83 percent complete\n",
      "00:11:16.85 94.32 percent complete\n",
      "00:11:41.60 97.81 percent complete\n"
     ]
    }
   ],
   "source": [
    "# Install fuzzy wuzzy to remove \"almost duplicate\" sentences in the\n",
    "# test and training sets.\n",
    "! pip install fuzzywuzzy\n",
    "! pip install python-Levenshtein\n",
    "import time\n",
    "from fuzzywuzzy import process\n",
    "import numpy as np\n",
    "\n",
    "# reset the index of the training set after previous filtering\n",
    "df_pp.reset_index(drop=False, inplace=True)\n",
    "\n",
    "# Remove samples from the training data set if they \"almost overlap\" with the\n",
    "# samples in the test set.\n",
    "\n",
    "# Filtering function. Adjust pad to narrow down the candidate matches to\n",
    "# within a certain length of characters of the given sample.\n",
    "def fuzzfilter(sample, candidates, pad):\n",
    "  candidates = [x for x in candidates if len(x) <= len(sample)+pad and len(x) >= len(sample)-pad] \n",
    "  if len(candidates) > 0:\n",
    "    return process.extractOne(sample, candidates)[1]\n",
    "  else:\n",
    "    return np.nan\n",
    "\n",
    "# NOTE - This might run slow depending on the size of your training set. We are\n",
    "# printing some information to help you track how long it would take. \n",
    "scores = []\n",
    "start_time = time.time()\n",
    "for idx, row in df_pp.iterrows():\n",
    "  scores.append(fuzzfilter(row['source_sentence'], list(en_test_sents), 5))\n",
    "  if idx % 1000 == 0:\n",
    "    hours, rem = divmod(time.time() - start_time, 3600)\n",
    "    minutes, seconds = divmod(rem, 60)\n",
    "    print(\"{:0>2}:{:0>2}:{:05.2f}\".format(int(hours),int(minutes),seconds), \"%0.2f percent complete\" % (100.0*float(idx)/float(len(df_pp))))\n",
    "\n",
    "# Filter out \"almost overlapping samples\"\n",
    "df_pp['scores'] = scores\n",
    "df_pp = df_pp[df_pp['scores'] < 95]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 819
    },
    "colab_type": "code",
    "id": "hxxBOCA-xXhy",
    "outputId": "26b4e463-6d02-4a78-8002-2b331efe21f6"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "==> train.en <==\n",
      "If we realize that our love for material things is eclipsing our love for the Christ , we should reflect on Jesus ’ words : “ Guard against every sort of greed . ”\n",
      "The result ?\n",
      "Then my grandmother’s sister , who was over 80 years of age , began to study and was baptized before she died .\n",
      "If black Witnesses preached from door to door in a white neighborhood , they would be arrested ​ — and likely beaten up .\n",
      "Soon afterward Daniel died .\n",
      "At a diplomatic function , Herod clothed himself “ with royal raiment . ”\n",
      "( b ) How can we become effective in the ministry ?\n",
      "Giving refugees practical help requires , not a lot of money , but mainly our time and concern .\n",
      "Jesus gave an illustration about a king who forgave his slave a huge debt of 60,000,000 denarii .\n",
      "Jesus explained a principle that helps us to know whom we should obey .\n",
      "\n",
      "==> train.fon <==\n",
      "Nú mǐ ɖó ayi wu ɖɔ wanyiyi e mǐ ɖó nú agbaza sín nǔ lɛ é jɛ ta yí sín wanyiyi e mǐ ɖó nú Klisu é sí jí ɔ , mǐ ɖó na lin nǔ kpɔ́n dó xó e Jezu ɖɔ é jí : “ Mi cɔ́ miɖée dó nukúnkɛ́ndídó wu . ” ( Luk .\n",
      "Etɛ ka jɛ túnú tɔn ?\n",
      "Gɔ́ na ɔ , mamáa ce sín nagán , ee ko ɖó xwè 80 jɛji é lɔmɔ̌ jɛ Biblu kplɔ́n jí lobo bló baptɛm cobo kú .\n",
      "Enyi kúnnuɖetɔ́ mɛwi ɖé ɖò wɛn jla wɛ sín hɔn ɖé kɔn jɛ hɔn ɖé kɔn ɖò yovó lɛ sín xá mɛ ɔ , ye na wlí i bo sixu xò è .\n",
      "Enɛ gudo zaan ɔ , Daniel kú .\n",
      "Ðó Elodu ɖò tɛn acɛkpikpa tɔn mɛ wutu ɔ , é dó “ axɔsuwu tɔn . ”\n",
      "( b ) Nɛ̌ mǐ ka sixu bló gbɔn bɔ sinsɛnzɔ́ mǐtɔn na na sínsɛ́n hugǎn ?\n",
      "Alɔ tawun tawun lɛ didó mɛ ɖěɖee hɔn sín tò yetɔn mɛ lɛ é byɔ ɖɔ è ni zán akwɛ gègě ǎ , loɔ , nǔ taji ɔ wɛ nyí ɖɔ mǐ na zán hwenu mǐtɔn bo lɛ́ xlɛ́ ɖɔ nǔ yetɔn nɔ ɖu ayi mɛ nú mǐ .\n",
      "Jezu dó lǒ ɖé dó axɔsu e sɔ́ sikágankwɛ 10 000 , alǒ gankwɛ 60 000 000 kɛ mɛsɛntɔ́ tɔn lɛ ɖokpo é wu .\n",
      "Jezu tinmɛ nugbodòdó e sixu d’alɔ mǐ bɔ mǐ na tuùn mɛ e mǐ ɖó na setónú na é .\n",
      "==> dev.en <==\n",
      "I love pioneering in this territory .\n",
      "“ Too numerous to recount ” are the “ wonderful works ” we can thank and praise Jehovah for daily !\n",
      "Some Christian parents serving in a foreign - language field have come to realize that their children’s interest in the truth has waned .\n",
      "You certainly recognize that Jewish man as the one who came to be known as the apostle Paul .\n",
      "Before accepting a new assignment , a modest person will first find out what will be required of him .\n",
      "Others who had admirable qualities failed to win God’s approval .\n",
      "His momentary exuberance was not true joy . ​ — Prov .\n",
      "We can see that from the resurrections Jesus performed when outside of Nain and when in the home of Jairus .\n",
      "According to Revelation 14 : 6 , 7 , how do angels assist God’s people today ?\n",
      "Why do I obey God’s commands rather than imitate the world’s moral standards ?\n",
      "\n",
      "==> dev.fon <==\n",
      "Un yí wǎn nú gbexosin - alijitɔ́zɔ́ wiwa ɖò fí enɛ .\n",
      "“ Nùjiwǔ ” e Jehovah bló bɔ mǐ sixu dokú tɔn n’i bo lɛ́ kpa susu n’i ayihɔngbe ayihɔngbe é “ sukpɔ́ ” tawun .\n",
      "Mɛjitɔ́ Klisanwun e ɖò sinsɛnzɔ́ wà wɛ ɖò ayǐ e jí è nɔ dó gbè ɖevo ɖè lɛ é ɖé lɛ wá ɖ’ayi wu ɖɔ jlǒ e vǐ emitɔn lɛ ɖó nú nugbǒ ɔ é ɖò ɖiɖekpo wɛ .\n",
      "1 : 14 ) É ɖò wɛn ɖɔ a tuùn nya Jwifu enɛ e è wá ylɔ ɖɔ mɛsɛ́dó Pɔlu é .\n",
      "Cobonu mɛ e nɔ nɔ jlɛ̌ jí é ɖé na yí gbè nú azɔ̌ yɔyɔ̌ ɖé ɔ , é nɔ gbéjé nǔ e azɔ̌ ɔ na byɔ ɖò así tɔn lɛ é kpɔ́n hwɛ̌ .\n",
      "Mɛ ɖevo lɛ ka tíìn bo ɖó jijɔ ɖagbe lɛ có , nǔ yetɔn ka nyɔ́ nukún tɔn mɛ ǎ .\n",
      "Jo e é ki nú táan kpɛɖé é nyí awǎjijɛ jɔ awǎjijɛ ǎ . ​ — Nùx .\n",
      "Mǐ mɔ ɖɔ Jezu fɔ́n mɛ sín kú hwenu e é tɔ́n sín Nayinu é kpo hwenu e é ɖò Jayilɔsi xwé é kpo .\n",
      "Sɔgbe xá nǔ e Nǔɖexlɛ́mɛ 14 : 6 , 7 ɖɔ é ɔ , nɛ̌ wɛnsagun lɛ ka nɔ d’alɔ togun Mawu tɔn gbɔn ɖò égbé ?\n",
      "Etɛwu un ka nɔ nyì sɛ́n Mawu tɔn lɛ , bo ma nɔ wà nǔ gbɛ̀ ɔ ɖɔhun ǎ ?\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",
    "# 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": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1000
    },
    "colab_type": "code",
    "id": "iBRMm4kMxZ8L",
    "outputId": "71b591b8-0d99-43aa-f08a-4f3f6cd87998"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Cloning into 'joeynmt'...\n",
      "remote: Enumerating objects: 15, done.\u001b[K\n",
      "remote: Counting objects:   6% (1/15)\u001b[K\r",
      "remote: Counting objects:  13% (2/15)\u001b[K\r",
      "remote: Counting objects:  20% (3/15)\u001b[K\r",
      "remote: Counting objects:  26% (4/15)\u001b[K\r",
      "remote: Counting objects:  33% (5/15)\u001b[K\r",
      "remote: Counting objects:  40% (6/15)\u001b[K\r",
      "remote: Counting objects:  46% (7/15)\u001b[K\r",
      "remote: Counting objects:  53% (8/15)\u001b[K\r",
      "remote: Counting objects:  60% (9/15)\u001b[K\r",
      "remote: Counting objects:  66% (10/15)\u001b[K\r",
      "remote: Counting objects:  73% (11/15)\u001b[K\r",
      "remote: Counting objects:  80% (12/15)\u001b[K\r",
      "remote: Counting objects:  86% (13/15)\u001b[K\r",
      "remote: Counting objects:  93% (14/15)\u001b[K\r",
      "remote: Counting objects: 100% (15/15)\u001b[K\r",
      "remote: Counting objects: 100% (15/15), done.\u001b[K\n",
      "remote: Compressing objects: 100% (12/12), done.\u001b[K\n",
      "remote: Total 2199 (delta 4), reused 5 (delta 3), pack-reused 2184\u001b[K\n",
      "Receiving objects: 100% (2199/2199), 2.60 MiB | 4.29 MiB/s, done.\n",
      "Resolving deltas: 100% (1525/1525), done.\n",
      "Processing /content/joeynmt\n",
      "Requirement already satisfied: future in /usr/local/lib/python3.6/dist-packages (from joeynmt==0.0.1) (0.16.0)\n",
      "Requirement already satisfied: pillow in /usr/local/lib/python3.6/dist-packages (from joeynmt==0.0.1) (4.3.0)\n",
      "Requirement already satisfied: numpy<2.0,>=1.14.5 in /usr/local/lib/python3.6/dist-packages (from joeynmt==0.0.1) (1.17.4)\n",
      "Requirement already satisfied: setuptools>=41.0.0 in /usr/local/lib/python3.6/dist-packages (from joeynmt==0.0.1) (41.6.0)\n",
      "Requirement already satisfied: torch>=1.1 in /usr/local/lib/python3.6/dist-packages (from joeynmt==0.0.1) (1.3.1)\n",
      "Requirement already satisfied: tensorflow>=1.14 in /usr/local/lib/python3.6/dist-packages (from joeynmt==0.0.1) (1.15.0)\n",
      "Requirement already satisfied: torchtext in /usr/local/lib/python3.6/dist-packages (from joeynmt==0.0.1) (0.3.1)\n",
      "Collecting sacrebleu>=1.3.6\n",
      "  Downloading https://files.pythonhosted.org/packages/0e/e5/93d252182f7cbd4b59bb3ec5797e2ce33cfd6f5aadaf327db170cf4b7887/sacrebleu-1.4.2-py3-none-any.whl\n",
      "Collecting subword-nmt\n",
      "  Downloading https://files.pythonhosted.org/packages/74/60/6600a7bc09e7ab38bc53a48a20d8cae49b837f93f5842a41fe513a694912/subword_nmt-0.3.7-py2.py3-none-any.whl\n",
      "Requirement already satisfied: matplotlib in /usr/local/lib/python3.6/dist-packages (from joeynmt==0.0.1) (3.1.1)\n",
      "Requirement already satisfied: seaborn in /usr/local/lib/python3.6/dist-packages (from joeynmt==0.0.1) (0.9.0)\n",
      "Collecting pyyaml>=5.1\n",
      "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/e3/e8/b3212641ee2718d556df0f23f78de8303f068fe29cdaa7a91018849582fe/PyYAML-5.1.2.tar.gz (265kB)\n",
      "\u001b[K     |████████████████████████████████| 266kB 16.9MB/s \n",
      "\u001b[?25hCollecting pylint\n",
      "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/e9/59/43fc36c5ee316bb9aeb7cf5329cdbdca89e5749c34d5602753827c0aa2dc/pylint-2.4.4-py3-none-any.whl (302kB)\n",
      "\u001b[K     |████████████████████████████████| 307kB 49.9MB/s \n",
      "\u001b[?25hRequirement already satisfied: six==1.12 in /usr/local/lib/python3.6/dist-packages (from joeynmt==0.0.1) (1.12.0)\n",
      "Requirement already satisfied: olefile in /usr/local/lib/python3.6/dist-packages (from pillow->joeynmt==0.0.1) (0.46)\n",
      "Requirement already satisfied: absl-py>=0.7.0 in /usr/local/lib/python3.6/dist-packages (from tensorflow>=1.14->joeynmt==0.0.1) (0.8.1)\n",
      "Requirement already satisfied: keras-preprocessing>=1.0.5 in /usr/local/lib/python3.6/dist-packages (from tensorflow>=1.14->joeynmt==0.0.1) (1.1.0)\n",
      "Requirement already satisfied: termcolor>=1.1.0 in /usr/local/lib/python3.6/dist-packages (from tensorflow>=1.14->joeynmt==0.0.1) (1.1.0)\n",
      "Requirement already satisfied: grpcio>=1.8.6 in /usr/local/lib/python3.6/dist-packages (from tensorflow>=1.14->joeynmt==0.0.1) (1.15.0)\n",
      "Requirement already satisfied: keras-applications>=1.0.8 in /usr/local/lib/python3.6/dist-packages (from tensorflow>=1.14->joeynmt==0.0.1) (1.0.8)\n",
      "Requirement already satisfied: tensorboard<1.16.0,>=1.15.0 in /usr/local/lib/python3.6/dist-packages (from tensorflow>=1.14->joeynmt==0.0.1) (1.15.0)\n",
      "Requirement already satisfied: opt-einsum>=2.3.2 in /usr/local/lib/python3.6/dist-packages (from tensorflow>=1.14->joeynmt==0.0.1) (3.1.0)\n",
      "Requirement already satisfied: astor>=0.6.0 in /usr/local/lib/python3.6/dist-packages (from tensorflow>=1.14->joeynmt==0.0.1) (0.8.0)\n",
      "Requirement already satisfied: google-pasta>=0.1.6 in /usr/local/lib/python3.6/dist-packages (from tensorflow>=1.14->joeynmt==0.0.1) (0.1.8)\n",
      "Requirement already satisfied: gast==0.2.2 in /usr/local/lib/python3.6/dist-packages (from tensorflow>=1.14->joeynmt==0.0.1) (0.2.2)\n",
      "Requirement already satisfied: wheel>=0.26 in /usr/local/lib/python3.6/dist-packages (from tensorflow>=1.14->joeynmt==0.0.1) (0.33.6)\n",
      "Requirement already satisfied: wrapt>=1.11.1 in /usr/local/lib/python3.6/dist-packages (from tensorflow>=1.14->joeynmt==0.0.1) (1.11.2)\n",
      "Requirement already satisfied: tensorflow-estimator==1.15.1 in /usr/local/lib/python3.6/dist-packages (from tensorflow>=1.14->joeynmt==0.0.1) (1.15.1)\n",
      "Requirement already satisfied: protobuf>=3.6.1 in /usr/local/lib/python3.6/dist-packages (from tensorflow>=1.14->joeynmt==0.0.1) (3.10.0)\n",
      "Requirement already satisfied: tqdm in /usr/local/lib/python3.6/dist-packages (from torchtext->joeynmt==0.0.1) (4.28.1)\n",
      "Requirement already satisfied: requests in /usr/local/lib/python3.6/dist-packages (from torchtext->joeynmt==0.0.1) (2.21.0)\n",
      "Collecting portalocker\n",
      "  Downloading https://files.pythonhosted.org/packages/91/db/7bc703c0760df726839e0699b7f78a4d8217fdc9c7fcb1b51b39c5a22a4e/portalocker-1.5.2-py2.py3-none-any.whl\n",
      "Requirement already satisfied: typing in /usr/local/lib/python3.6/dist-packages (from sacrebleu>=1.3.6->joeynmt==0.0.1) (3.6.6)\n",
      "Requirement already satisfied: kiwisolver>=1.0.1 in /usr/local/lib/python3.6/dist-packages (from matplotlib->joeynmt==0.0.1) (1.1.0)\n",
      "Requirement already satisfied: python-dateutil>=2.1 in /usr/local/lib/python3.6/dist-packages (from matplotlib->joeynmt==0.0.1) (2.6.1)\n",
      "Requirement already satisfied: cycler>=0.10 in /usr/local/lib/python3.6/dist-packages (from matplotlib->joeynmt==0.0.1) (0.10.0)\n",
      "Requirement already satisfied: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in /usr/local/lib/python3.6/dist-packages (from matplotlib->joeynmt==0.0.1) (2.4.5)\n",
      "Requirement already satisfied: scipy>=0.14.0 in /usr/local/lib/python3.6/dist-packages (from seaborn->joeynmt==0.0.1) (1.3.2)\n",
      "Requirement already satisfied: pandas>=0.15.2 in /usr/local/lib/python3.6/dist-packages (from seaborn->joeynmt==0.0.1) (0.25.3)\n",
      "Collecting astroid<2.4,>=2.3.0\n",
      "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/ad/ae/86734823047962e7b8c8529186a1ac4a7ca19aaf1aa0c7713c022ef593fd/astroid-2.3.3-py3-none-any.whl (205kB)\n",
      "\u001b[K     |████████████████████████████████| 215kB 51.0MB/s \n",
      "\u001b[?25hCollecting isort<5,>=4.2.5\n",
      "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/e5/b0/c121fd1fa3419ea9bfd55c7f9c4fedfec5143208d8c7ad3ce3db6c623c21/isort-4.3.21-py2.py3-none-any.whl (42kB)\n",
      "\u001b[K     |████████████████████████████████| 51kB 6.5MB/s \n",
      "\u001b[?25hCollecting mccabe<0.7,>=0.6\n",
      "  Downloading https://files.pythonhosted.org/packages/87/89/479dc97e18549e21354893e4ee4ef36db1d237534982482c3681ee6e7b57/mccabe-0.6.1-py2.py3-none-any.whl\n",
      "Requirement already satisfied: h5py in /usr/local/lib/python3.6/dist-packages (from keras-applications>=1.0.8->tensorflow>=1.14->joeynmt==0.0.1) (2.8.0)\n",
      "Requirement already satisfied: markdown>=2.6.8 in /usr/local/lib/python3.6/dist-packages (from tensorboard<1.16.0,>=1.15.0->tensorflow>=1.14->joeynmt==0.0.1) (3.1.1)\n",
      "Requirement already satisfied: werkzeug>=0.11.15 in /usr/local/lib/python3.6/dist-packages (from tensorboard<1.16.0,>=1.15.0->tensorflow>=1.14->joeynmt==0.0.1) (0.16.0)\n",
      "Requirement already satisfied: idna<2.9,>=2.5 in /usr/local/lib/python3.6/dist-packages (from requests->torchtext->joeynmt==0.0.1) (2.8)\n",
      "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.6/dist-packages (from requests->torchtext->joeynmt==0.0.1) (2019.9.11)\n",
      "Requirement already satisfied: urllib3<1.25,>=1.21.1 in /usr/local/lib/python3.6/dist-packages (from requests->torchtext->joeynmt==0.0.1) (1.24.3)\n",
      "Requirement already satisfied: chardet<3.1.0,>=3.0.2 in /usr/local/lib/python3.6/dist-packages (from requests->torchtext->joeynmt==0.0.1) (3.0.4)\n",
      "Requirement already satisfied: pytz>=2017.2 in /usr/local/lib/python3.6/dist-packages (from pandas>=0.15.2->seaborn->joeynmt==0.0.1) (2018.9)\n",
      "Collecting typed-ast<1.5,>=1.4.0; implementation_name == \"cpython\" and python_version < \"3.8\"\n",
      "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/31/d3/9d1802c161626d0278bafb1ffb32f76b9d01e123881bbf9d91e8ccf28e18/typed_ast-1.4.0-cp36-cp36m-manylinux1_x86_64.whl (736kB)\n",
      "\u001b[K     |████████████████████████████████| 737kB 45.8MB/s \n",
      "\u001b[?25hCollecting lazy-object-proxy==1.4.*\n",
      "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/0b/dd/b1e3407e9e6913cf178e506cd0dee818e58694d9a5cd1984e3f6a8b9a10f/lazy_object_proxy-1.4.3-cp36-cp36m-manylinux1_x86_64.whl (55kB)\n",
      "\u001b[K     |████████████████████████████████| 61kB 7.8MB/s \n",
      "\u001b[?25hBuilding wheels for collected packages: joeynmt, pyyaml\n",
      "  Building wheel for joeynmt (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
      "  Created wheel for joeynmt: filename=joeynmt-0.0.1-cp36-none-any.whl size=72136 sha256=d117593e05da9e532a469ebc2f269231fc80415f40bc9d316cda47919044b9b5\n",
      "  Stored in directory: /tmp/pip-ephem-wheel-cache-4hxgsvsh/wheels/db/01/db/751cc9f3e7f6faec127c43644ba250a3ea7ad200594aeda70a\n",
      "  Building wheel for pyyaml (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
      "  Created wheel for pyyaml: filename=PyYAML-5.1.2-cp36-cp36m-linux_x86_64.whl size=44104 sha256=fb52239eedec1192734b9c05b0e67ddb748b25e86c2a5c09a444c72e1b0bf831\n",
      "  Stored in directory: /root/.cache/pip/wheels/d9/45/dd/65f0b38450c47cf7e5312883deb97d065e030c5cca0a365030\n",
      "Successfully built joeynmt pyyaml\n",
      "Installing collected packages: portalocker, sacrebleu, subword-nmt, pyyaml, typed-ast, lazy-object-proxy, astroid, isort, mccabe, pylint, joeynmt\n",
      "  Found existing installation: PyYAML 3.13\n",
      "    Uninstalling PyYAML-3.13:\n",
      "      Successfully uninstalled PyYAML-3.13\n",
      "Successfully installed astroid-2.3.3 isort-4.3.21 joeynmt-0.0.1 lazy-object-proxy-1.4.3 mccabe-0.6.1 portalocker-1.5.2 pylint-2.4.4 pyyaml-5.1.2 sacrebleu-1.4.2 subword-nmt-0.3.7 typed-ast-1.4.0\n"
     ]
    }
   ],
   "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": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 408
    },
    "colab_type": "code",
    "id": "H-TyjtmXB1mL",
    "outputId": "4f85916b-5a94-419d-94c7-296f3e65c729"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "bpe.codes.4000\tdev.en\t     test.bpe.fon    test.fon\t    train.en\n",
      "dev.bpe.en\tdev.fon      test.en\t     train.bpe.en   train.fon\n",
      "dev.bpe.fon\ttest.bpe.en  test.en-any.en  train.bpe.fon\n",
      "bpe.codes.4000\tdev.en\t     test.bpe.fon    test.fon\t    train.en\n",
      "dev.bpe.en\tdev.fon      test.en\t     train.bpe.en   train.fon\n",
      "dev.bpe.fon\ttest.bpe.en  test.en-any.en  train.bpe.fon\n",
      "BPE Xhosa Sentences\n",
      "Ðɛhan 6@@ 4 : 2 - 5 d’alɔ mì bɔ un mɔ ɖɔ un kún jló na nɔ mɛ ɖěɖee mɛ ɖevo lɛ nɔ xoɖɛ ɖɔ Mawu ni hwlɛn emi ɖò ye sí é mɛ ó !\n",
      "Un lɛ́vɔ wá mɔ nukúnnú jɛ wu ɖɔ enyi un kpò ɖò mɛ nú ɖɔ wɛ ɔ , un kún nyí kpɔ́ndéwú ɖagbe ɖé ó , b’ɛ na lɛ́ hɛn nyikɔ Jehovah tɔn k@@ wi@@ ji . ”\n",
      "Lin@@ da : “ Un xà t@@ l@@ ati mǐtɔn lɛ ganji bo na dó sixu sɔnǔ bá xwlé ye mɛ .\n",
      "Gbɛ@@ ̌@@ dido xá mɛ ɖěɖee nɔ ɖ’alɔ ɖò sinsɛnzɔ́ ɔ sín akpáxwé vovo lɛ mɛ é ko d’alɔ mì tawun .\n",
      "Un lɛ́ kpó ɖò ɖɛ xò sɛ́dó Jehovah wɛ bo ganjɛ wǔ tɔn . ”\n",
      "Combined BPE Vocab\n",
      "Tod@@\n",
      "righte@@\n",
      "xamu\n",
      "usal@@\n",
      "Mɔyiz@@\n",
      "ɔ́nd@@\n",
      "fín@@\n",
      "wɛli\n",
      "ɔlu\n",
      "mpli\n"
     ]
    }
   ],
   "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\n",
    "\n",
    "# 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\n",
    "\n",
    "# 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 $data_path\n",
    "\n",
    "# Also move everything we care about to a mounted location in google drive (relevant if running in colab) at gdrive_path\n",
    "! cp train.* \"$gdrive_path\"\n",
    "! cp test.* \"$gdrive_path\"\n",
    "! cp dev.* \"$gdrive_path\"\n",
    "! cp bpe.codes.4000 \"$gdrive_path\"\n",
    "! ls \"$gdrive_path\"\n",
    "\n",
    "# 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\n",
    "\n",
    "# Some output\n",
    "! echo \"BPE Xhosa Sentences\"\n",
    "! tail -n 5 test.bpe.$tgt\n",
    "! echo \"Combined BPE Vocab\"\n",
    "! tail -n 10 joeynmt/data/$src$tgt/vocab.txt  # Herman"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 68
    },
    "colab_type": "code",
    "id": "IlMitUHR8Qy-",
    "outputId": "9aa9345d-ad89-48ba-e9c6-03822b4d45e4"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "bpe.codes.4000\tdev.en\t     test.bpe.fon    test.fon\t    train.en\n",
      "dev.bpe.en\tdev.fon      test.en\t     train.bpe.en   train.fon\n",
      "dev.bpe.fon\ttest.bpe.en  test.en-any.en  train.bpe.fon\n"
     ]
    }
   ],
   "source": [
    "# Also move everything we care about to a mounted location in google drive (relevant if running in colab) at gdrive_path\n",
    "! cp train.* \"$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": 0,
   "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: \"{gdrive_path}/models/{name}_transformer/1.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=os.environ[\"gdrive_path\"], 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": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1000
    },
    "colab_type": "code",
    "id": "6ZBPFwT94WpI",
    "outputId": "9d1ffcd3-3a70-45da-f734-0f532d14cb30"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2019-11-26 14:00:30,601 Hello! This is Joey-NMT.\n",
      "2019-11-26 14:00:32,457 Total params: 12131328\n",
      "2019-11-26 14:00:32,459 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",
      "2019-11-26 14:00:38,324 cfg.name                           : enfon_transformer\n",
      "2019-11-26 14:00:38,324 cfg.data.src                       : en\n",
      "2019-11-26 14:00:38,324 cfg.data.trg                       : fon\n",
      "2019-11-26 14:00:38,324 cfg.data.train                     : data/enfon/train.bpe\n",
      "2019-11-26 14:00:38,325 cfg.data.dev                       : data/enfon/dev.bpe\n",
      "2019-11-26 14:00:38,325 cfg.data.test                      : data/enfon/test.bpe\n",
      "2019-11-26 14:00:38,325 cfg.data.level                     : bpe\n",
      "2019-11-26 14:00:38,325 cfg.data.lowercase                 : False\n",
      "2019-11-26 14:00:38,325 cfg.data.max_sent_length           : 100\n",
      "2019-11-26 14:00:38,325 cfg.data.src_vocab                 : data/enfon/vocab.txt\n",
      "2019-11-26 14:00:38,325 cfg.data.trg_vocab                 : data/enfon/vocab.txt\n",
      "2019-11-26 14:00:38,325 cfg.testing.beam_size              : 5\n",
      "2019-11-26 14:00:38,325 cfg.testing.alpha                  : 1.0\n",
      "2019-11-26 14:00:38,325 cfg.training.random_seed           : 42\n",
      "2019-11-26 14:00:38,326 cfg.training.optimizer             : adam\n",
      "2019-11-26 14:00:38,326 cfg.training.normalization         : tokens\n",
      "2019-11-26 14:00:38,326 cfg.training.adam_betas            : [0.9, 0.999]\n",
      "2019-11-26 14:00:38,326 cfg.training.scheduling            : plateau\n",
      "2019-11-26 14:00:38,326 cfg.training.patience              : 5\n",
      "2019-11-26 14:00:38,326 cfg.training.learning_rate_factor  : 0.5\n",
      "2019-11-26 14:00:38,326 cfg.training.learning_rate_warmup  : 1000\n",
      "2019-11-26 14:00:38,326 cfg.training.decrease_factor       : 0.7\n",
      "2019-11-26 14:00:38,326 cfg.training.loss                  : crossentropy\n",
      "2019-11-26 14:00:38,326 cfg.training.learning_rate         : 0.0003\n",
      "2019-11-26 14:00:38,327 cfg.training.learning_rate_min     : 1e-08\n",
      "2019-11-26 14:00:38,327 cfg.training.weight_decay          : 0.0\n",
      "2019-11-26 14:00:38,327 cfg.training.label_smoothing       : 0.1\n",
      "2019-11-26 14:00:38,327 cfg.training.batch_size            : 4096\n",
      "2019-11-26 14:00:38,327 cfg.training.batch_type            : token\n",
      "2019-11-26 14:00:38,327 cfg.training.eval_batch_size       : 3600\n",
      "2019-11-26 14:00:38,327 cfg.training.eval_batch_type       : token\n",
      "2019-11-26 14:00:38,327 cfg.training.batch_multiplier      : 1\n",
      "2019-11-26 14:00:38,327 cfg.training.early_stopping_metric : ppl\n",
      "2019-11-26 14:00:38,327 cfg.training.epochs                : 30\n",
      "2019-11-26 14:00:38,328 cfg.training.validation_freq       : 1000\n",
      "2019-11-26 14:00:38,328 cfg.training.logging_freq          : 100\n",
      "2019-11-26 14:00:38,328 cfg.training.eval_metric           : bleu\n",
      "2019-11-26 14:00:38,328 cfg.training.model_dir             : models/enfon_transformer\n",
      "2019-11-26 14:00:38,328 cfg.training.overwrite             : False\n",
      "2019-11-26 14:00:38,328 cfg.training.shuffle               : True\n",
      "2019-11-26 14:00:38,328 cfg.training.use_cuda              : True\n",
      "2019-11-26 14:00:38,328 cfg.training.max_output_length     : 100\n",
      "2019-11-26 14:00:38,328 cfg.training.print_valid_sents     : [0, 1, 2, 3]\n",
      "2019-11-26 14:00:38,328 cfg.training.keep_last_ckpts       : 3\n",
      "2019-11-26 14:00:38,329 cfg.model.initializer              : xavier\n",
      "2019-11-26 14:00:38,329 cfg.model.bias_initializer         : zeros\n",
      "2019-11-26 14:00:38,329 cfg.model.init_gain                : 1.0\n",
      "2019-11-26 14:00:38,329 cfg.model.embed_initializer        : xavier\n",
      "2019-11-26 14:00:38,329 cfg.model.embed_init_gain          : 1.0\n",
      "2019-11-26 14:00:38,329 cfg.model.tied_embeddings          : True\n",
      "2019-11-26 14:00:38,329 cfg.model.tied_softmax             : True\n",
      "2019-11-26 14:00:38,329 cfg.model.encoder.type             : transformer\n",
      "2019-11-26 14:00:38,329 cfg.model.encoder.num_layers       : 6\n",
      "2019-11-26 14:00:38,329 cfg.model.encoder.num_heads        : 4\n",
      "2019-11-26 14:00:38,330 cfg.model.encoder.embeddings.embedding_dim : 256\n",
      "2019-11-26 14:00:38,330 cfg.model.encoder.embeddings.scale : True\n",
      "2019-11-26 14:00:38,330 cfg.model.encoder.embeddings.dropout : 0.2\n",
      "2019-11-26 14:00:38,330 cfg.model.encoder.hidden_size      : 256\n",
      "2019-11-26 14:00:38,330 cfg.model.encoder.ff_size          : 1024\n",
      "2019-11-26 14:00:38,330 cfg.model.encoder.dropout          : 0.3\n",
      "2019-11-26 14:00:38,330 cfg.model.decoder.type             : transformer\n",
      "2019-11-26 14:00:38,330 cfg.model.decoder.num_layers       : 6\n",
      "2019-11-26 14:00:38,330 cfg.model.decoder.num_heads        : 4\n",
      "2019-11-26 14:00:38,330 cfg.model.decoder.embeddings.embedding_dim : 256\n",
      "2019-11-26 14:00:38,331 cfg.model.decoder.embeddings.scale : True\n",
      "2019-11-26 14:00:38,331 cfg.model.decoder.embeddings.dropout : 0.2\n",
      "2019-11-26 14:00:38,331 cfg.model.decoder.hidden_size      : 256\n",
      "2019-11-26 14:00:38,331 cfg.model.decoder.ff_size          : 1024\n",
      "2019-11-26 14:00:38,331 cfg.model.decoder.dropout          : 0.3\n",
      "2019-11-26 14:00:38,331 Data set sizes: \n",
      "\ttrain 27510,\n",
      "\tvalid 1000,\n",
      "\ttest 2713\n",
      "2019-11-26 14:00:38,331 First training example:\n",
      "\t[SRC] If we reali@@ ze that our love for material things is ec@@ li@@ p@@ sing our love for the Christ , we should reflect on Jesus ’ words : “ Gu@@ ard against every sor@@ t of gre@@ ed . ”\n",
      "\t[TRG] Nú mǐ ɖó ayi wu ɖɔ wanyiyi e mǐ ɖó nú agbaza sín nǔ lɛ é jɛ ta yí sín wanyiyi e mǐ ɖó nú Klisu é sí jí ɔ , mǐ ɖó na lin nǔ kpɔ́n dó xó e Jezu ɖɔ é jí : “ Mi cɔ́ miɖée dó nukún@@ k@@ ɛ́n@@ dí@@ dó wu . ” ( Luk .\n",
      "2019-11-26 14:00:38,331 First 10 words (src): (0) <unk> (1) <pad> (2) <s> (3) </s> (4) . (5) , (6) é (7) lɛ (8) ɔ (9) the\n",
      "2019-11-26 14:00:38,332 First 10 words (trg): (0) <unk> (1) <pad> (2) <s> (3) </s> (4) . (5) , (6) é (7) lɛ (8) ɔ (9) the\n",
      "2019-11-26 14:00:38,332 Number of Src words (types): 4184\n",
      "2019-11-26 14:00:38,333 Number of Trg words (types): 4184\n",
      "2019-11-26 14:00:38,333 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=4184),\n",
      "\ttrg_embed=Embeddings(embedding_dim=256, vocab_size=4184))\n",
      "2019-11-26 14:00:38,337 EPOCH 1\n",
      "2019-11-26 14:01:08,765 Epoch   1 Step:      100 Batch Loss:     5.094642 Tokens per Sec:     7341, Lr: 0.000300\n",
      "2019-11-26 14:01:39,090 Epoch   1 Step:      200 Batch Loss:     4.815321 Tokens per Sec:     7525, Lr: 0.000300\n",
      "2019-11-26 14:02:09,438 Epoch   1 Step:      300 Batch Loss:     4.436522 Tokens per Sec:     7544, Lr: 0.000300\n",
      "2019-11-26 14:02:26,611 Epoch   1: total training loss 1742.23\n",
      "2019-11-26 14:02:26,612 EPOCH 2\n",
      "2019-11-26 14:02:39,620 Epoch   2 Step:      400 Batch Loss:     4.331379 Tokens per Sec:     7479, Lr: 0.000300\n",
      "2019-11-26 14:03:09,957 Epoch   2 Step:      500 Batch Loss:     4.236403 Tokens per Sec:     7448, Lr: 0.000300\n",
      "2019-11-26 14:03:40,305 Epoch   2 Step:      600 Batch Loss:     3.822441 Tokens per Sec:     7380, Lr: 0.000300\n",
      "2019-11-26 14:04:10,836 Epoch   2 Step:      700 Batch Loss:     3.890893 Tokens per Sec:     7423, Lr: 0.000300\n",
      "2019-11-26 14:04:14,880 Epoch   2: total training loss 1396.25\n",
      "2019-11-26 14:04:14,880 EPOCH 3\n",
      "2019-11-26 14:04:41,409 Epoch   3 Step:      800 Batch Loss:     3.255218 Tokens per Sec:     7376, Lr: 0.000300\n",
      "2019-11-26 14:05:11,843 Epoch   3 Step:      900 Batch Loss:     3.479604 Tokens per Sec:     7369, Lr: 0.000300\n",
      "2019-11-26 14:05:42,564 Epoch   3 Step:     1000 Batch Loss:     3.530126 Tokens per Sec:     7480, Lr: 0.000300\n",
      "2019-11-26 14:07:15,329 Hooray! New best validation result [ppl]!\n",
      "2019-11-26 14:07:15,329 Saving new checkpoint.\n",
      "2019-11-26 14:07:15,584 Example #0\n",
      "2019-11-26 14:07:15,585 \tSource:     I love pioneering in this territory .\n",
      "2019-11-26 14:07:15,585 \tReference:  Un yí wǎn nú gbexosin - alijitɔ́zɔ́ wiwa ɖò fí enɛ .\n",
      "2019-11-26 14:07:15,585 \tHypothesis: É ɖò wɛn ɖɔ é ni mɔ nǔ e ɖò sinsɛnzɔ́ ɔ mɛ é .\n",
      "2019-11-26 14:07:15,585 Example #1\n",
      "2019-11-26 14:07:15,586 \tSource:     “ Too numerous to recount ” are the “ wonderful works ” we can thank and praise Jehovah for daily !\n",
      "2019-11-26 14:07:15,586 \tReference:  “ Nùjiwǔ ” e Jehovah bló bɔ mǐ sixu dokú tɔn n’i bo lɛ́ kpa susu n’i ayihɔngbe ayihɔngbe é “ sukpɔ́ ” tawun .\n",
      "2019-11-26 14:07:15,586 \tHypothesis: “ Un nɔ wà nǔ e ɖò sinsɛnzɔ́ ɔ mɛ lɛ é , bo nɔ nɔ wà nǔ e ɖò gbɛ̀ ɔ mɛ lɛ é .\n",
      "2019-11-26 14:07:15,586 Example #2\n",
      "2019-11-26 14:07:15,586 \tSource:     Some Christian parents serving in a foreign - language field have come to realize that their children’s interest in the truth has waned .\n",
      "2019-11-26 14:07:15,586 \tReference:  Mɛjitɔ́ Klisanwun e ɖò sinsɛnzɔ́ wà wɛ ɖò ayǐ e jí è nɔ dó gbè ɖevo ɖè lɛ é ɖé lɛ wá ɖ’ayi wu ɖɔ jlǒ e vǐ emitɔn lɛ ɖó nú nugbǒ ɔ é ɖò ɖiɖekpo wɛ .\n",
      "2019-11-26 14:07:15,587 \tHypothesis: É ɖò wɛn ɖɔ xó dó wusyɛn lanmɛ nú mɛ e ɖò agun tɔn mɛ lɛ é jí , bɔ ye na nɔ wà nǔ e ɖò sinsɛnzɔ́ ɔ mɛ lɛ é .\n",
      "2019-11-26 14:07:15,587 Example #3\n",
      "2019-11-26 14:07:15,587 \tSource:     You certainly recognize that Jewish man as the one who came to be known as the apostle Paul .\n",
      "2019-11-26 14:07:15,587 \tReference:  1 : 14 ) É ɖò wɛn ɖɔ a tuùn nya Jwifu enɛ e è wá ylɔ ɖɔ mɛsɛ́dó Pɔlu é .\n",
      "2019-11-26 14:07:15,587 \tHypothesis: É ɖò wɛn ɖɔ xó dó wusyɛn lanmɛ nú mɛ e ɖò sinsɛnzɔ́ ɔ mɛ é .\n",
      "2019-11-26 14:07:15,587 Validation result (greedy) at epoch   3, step     1000: bleu:   2.72, loss: 99911.8359, ppl:  29.0535, duration: 93.0228s\n",
      "2019-11-26 14:07:36,428 Epoch   3: total training loss 1254.81\n",
      "2019-11-26 14:07:36,429 EPOCH 4\n",
      "2019-11-26 14:07:45,842 Epoch   4 Step:     1100 Batch Loss:     3.662204 Tokens per Sec:     7252, Lr: 0.000300\n",
      "2019-11-26 14:08:16,145 Epoch   4 Step:     1200 Batch Loss:     3.202804 Tokens per Sec:     7434, Lr: 0.000300\n",
      "2019-11-26 14:08:46,301 Epoch   4 Step:     1300 Batch Loss:     3.270840 Tokens per Sec:     7407, Lr: 0.000300\n",
      "2019-11-26 14:09:16,542 Epoch   4 Step:     1400 Batch Loss:     3.876655 Tokens per Sec:     7399, Lr: 0.000300\n",
      "2019-11-26 14:09:25,135 Epoch   4: total training loss 1181.73\n",
      "2019-11-26 14:09:25,135 EPOCH 5\n",
      "2019-11-26 14:09:47,196 Epoch   5 Step:     1500 Batch Loss:     3.134502 Tokens per Sec:     7454, Lr: 0.000300\n",
      "2019-11-26 14:10:17,600 Epoch   5 Step:     1600 Batch Loss:     3.284228 Tokens per Sec:     7403, Lr: 0.000300\n",
      "2019-11-26 14:10:48,127 Epoch   5 Step:     1700 Batch Loss:     2.819642 Tokens per Sec:     7404, Lr: 0.000300\n",
      "2019-11-26 14:11:13,530 Epoch   5: total training loss 1100.60\n",
      "2019-11-26 14:11:13,530 EPOCH 6\n",
      "2019-11-26 14:11:18,464 Epoch   6 Step:     1800 Batch Loss:     3.228624 Tokens per Sec:     7396, Lr: 0.000300\n",
      "2019-11-26 14:11:48,768 Epoch   6 Step:     1900 Batch Loss:     2.696800 Tokens per Sec:     7410, Lr: 0.000300\n",
      "2019-11-26 14:12:19,085 Epoch   6 Step:     2000 Batch Loss:     2.260545 Tokens per Sec:     7352, Lr: 0.000300\n",
      "2019-11-26 14:13:52,290 Hooray! New best validation result [ppl]!\n",
      "2019-11-26 14:13:52,291 Saving new checkpoint.\n",
      "2019-11-26 14:13:52,551 Example #0\n",
      "2019-11-26 14:13:52,551 \tSource:     I love pioneering in this territory .\n",
      "2019-11-26 14:13:52,551 \tReference:  Un yí wǎn nú gbexosin - alijitɔ́zɔ́ wiwa ɖò fí enɛ .\n",
      "2019-11-26 14:13:52,552 \tHypothesis: Un nɔ mɔ ɖɔ emi kún nɔ wà mɔ̌ ó .\n",
      "2019-11-26 14:13:52,552 Example #1\n",
      "2019-11-26 14:13:52,552 \tSource:     “ Too numerous to recount ” are the “ wonderful works ” we can thank and praise Jehovah for daily !\n",
      "2019-11-26 14:13:52,552 \tReference:  “ Nùjiwǔ ” e Jehovah bló bɔ mǐ sixu dokú tɔn n’i bo lɛ́ kpa susu n’i ayihɔngbe ayihɔngbe é “ sukpɔ́ ” tawun .\n",
      "2019-11-26 14:13:52,552 \tHypothesis: “ Mi nú mǐ ni kpɔ́n nǔ e mǐ nɔ wà nú mɛ e ɖò nǔ e ɖò nukún wɛ lɛ é jí , bo ɖɔ : “ Mi nú mǐ ni nɔ gbeji nú Jehovah , bo nɔ lɛ́ mɔ nǔ e mǐ nɔ wà é . ”\n",
      "2019-11-26 14:13:52,552 Example #2\n",
      "2019-11-26 14:13:52,553 \tSource:     Some Christian parents serving in a foreign - language field have come to realize that their children’s interest in the truth has waned .\n",
      "2019-11-26 14:13:52,553 \tReference:  Mɛjitɔ́ Klisanwun e ɖò sinsɛnzɔ́ wà wɛ ɖò ayǐ e jí è nɔ dó gbè ɖevo ɖè lɛ é ɖé lɛ wá ɖ’ayi wu ɖɔ jlǒ e vǐ emitɔn lɛ ɖó nú nugbǒ ɔ é ɖò ɖiɖekpo wɛ .\n",
      "2019-11-26 14:13:52,553 \tHypothesis: Hwenu e è nɔ dó wusyɛn lanmɛ nú mɛ ɖevo lɛ é ɔ , ye nɔ mɔ ɖɔ emi kún nɔ mɔ ɖɔ emi kún nɔ wà nǔ xá emi ó .\n",
      "2019-11-26 14:13:52,553 Example #3\n",
      "2019-11-26 14:13:52,553 \tSource:     You certainly recognize that Jewish man as the one who came to be known as the apostle Paul .\n",
      "2019-11-26 14:13:52,553 \tReference:  1 : 14 ) É ɖò wɛn ɖɔ a tuùn nya Jwifu enɛ e è wá ylɔ ɖɔ mɛsɛ́dó Pɔlu é .\n",
      "2019-11-26 14:13:52,553 \tHypothesis: A na ɖɔ xó dó nǔ e wu è na dó é wu ɖɔ xó dó é wu .\n",
      "2019-11-26 14:13:52,554 Validation result (greedy) at epoch   6, step     2000: bleu:   5.03, loss: 83836.4688, ppl:  16.8957, duration: 93.4679s\n",
      "2019-11-26 14:14:23,173 Epoch   6 Step:     2100 Batch Loss:     2.636636 Tokens per Sec:     7366, Lr: 0.000300\n",
      "2019-11-26 14:14:36,074 Epoch   6: total training loss 1045.85\n",
      "2019-11-26 14:14:36,074 EPOCH 7\n",
      "2019-11-26 14:14:53,904 Epoch   7 Step:     2200 Batch Loss:     3.124637 Tokens per Sec:     7297, Lr: 0.000300\n",
      "2019-11-26 14:15:24,789 Epoch   7 Step:     2300 Batch Loss:     2.172596 Tokens per Sec:     7446, Lr: 0.000300\n",
      "2019-11-26 14:15:55,324 Epoch   7 Step:     2400 Batch Loss:     2.914614 Tokens per Sec:     7257, Lr: 0.000300\n",
      "2019-11-26 14:16:25,749 Epoch   7: total training loss 990.90\n",
      "2019-11-26 14:16:25,749 EPOCH 8\n",
      "2019-11-26 14:16:26,109 Epoch   8 Step:     2500 Batch Loss:     2.662452 Tokens per Sec:     6874, Lr: 0.000300\n",
      "2019-11-26 14:16:56,653 Epoch   8 Step:     2600 Batch Loss:     2.515854 Tokens per Sec:     7351, Lr: 0.000300\n",
      "2019-11-26 14:17:27,406 Epoch   8 Step:     2700 Batch Loss:     2.303658 Tokens per Sec:     7268, Lr: 0.000300\n",
      "2019-11-26 14:17:58,354 Epoch   8 Step:     2800 Batch Loss:     2.409138 Tokens per Sec:     7174, Lr: 0.000300\n",
      "2019-11-26 14:18:16,437 Epoch   8: total training loss 958.06\n",
      "2019-11-26 14:18:16,438 EPOCH 9\n",
      "2019-11-26 14:18:29,356 Epoch   9 Step:     2900 Batch Loss:     2.426154 Tokens per Sec:     7355, Lr: 0.000300\n",
      "2019-11-26 14:18:59,974 Epoch   9 Step:     3000 Batch Loss:     2.670738 Tokens per Sec:     7273, Lr: 0.000300\n",
      "2019-11-26 14:20:33,350 Hooray! New best validation result [ppl]!\n",
      "2019-11-26 14:20:33,351 Saving new checkpoint.\n",
      "2019-11-26 14:20:33,598 Example #0\n",
      "2019-11-26 14:20:33,598 \tSource:     I love pioneering in this territory .\n",
      "2019-11-26 14:20:33,599 \tReference:  Un yí wǎn nú gbexosin - alijitɔ́zɔ́ wiwa ɖò fí enɛ .\n",
      "2019-11-26 14:20:33,599 \tHypothesis: Un nɔ wà sinsɛnzɔ́ ɖò sinsɛnzɔ́ ɔ mɛ ɖò sinsɛnzɔ́ ɔ mɛ .\n",
      "2019-11-26 14:20:33,599 Example #1\n",
      "2019-11-26 14:20:33,600 \tSource:     “ Too numerous to recount ” are the “ wonderful works ” we can thank and praise Jehovah for daily !\n",
      "2019-11-26 14:20:33,600 \tReference:  “ Nùjiwǔ ” e Jehovah bló bɔ mǐ sixu dokú tɔn n’i bo lɛ́ kpa susu n’i ayihɔngbe ayihɔngbe é “ sukpɔ́ ” tawun .\n",
      "2019-11-26 14:20:33,600 \tHypothesis: “ Ahwanvu ” wɛ nyí ɖɔ è ni “ nɔ “ nɔ ” bo nɔ “ nɔ dó gbɔ nú Jehovah , bo nɔ lɛ́ nɔ gbeji nú Jehovah .\n",
      "2019-11-26 14:20:33,600 Example #2\n",
      "2019-11-26 14:20:33,600 \tSource:     Some Christian parents serving in a foreign - language field have come to realize that their children’s interest in the truth has waned .\n",
      "2019-11-26 14:20:33,601 \tReference:  Mɛjitɔ́ Klisanwun e ɖò sinsɛnzɔ́ wà wɛ ɖò ayǐ e jí è nɔ dó gbè ɖevo ɖè lɛ é ɖé lɛ wá ɖ’ayi wu ɖɔ jlǒ e vǐ emitɔn lɛ ɖó nú nugbǒ ɔ é ɖò ɖiɖekpo wɛ .\n",
      "2019-11-26 14:20:33,601 \tHypothesis: Mɛjitɔ́ Klisanwun lɛ nɔ ɖó nukún ɖɔ mɛ winnyawinnya lɛ ni nɔ mɔ ɖɔ emi kún nɔ mɔ ɖɔ emi kún na wà nǔ e ye ɖó é ó .\n",
      "2019-11-26 14:20:33,601 Example #3\n",
      "2019-11-26 14:20:33,602 \tSource:     You certainly recognize that Jewish man as the one who came to be known as the apostle Paul .\n",
      "2019-11-26 14:20:33,602 \tReference:  1 : 14 ) É ɖò wɛn ɖɔ a tuùn nya Jwifu enɛ e è wá ylɔ ɖɔ mɛsɛ́dó Pɔlu é .\n",
      "2019-11-26 14:20:33,602 \tHypothesis: ( Jaan 3 : 1 - 14 ) É ɖò wɛn ɖɔ è na ko ɖɔ xó dó mɛsɛ́dó Pɔlu wu ɖɔ : “ Mɛ e ɖò kú wɛ lɛ é .\n",
      "2019-11-26 14:20:33,602 Validation result (greedy) at epoch   9, step     3000: bleu:   8.59, loss: 74603.6562, ppl:  12.3755, duration: 93.6282s\n",
      "2019-11-26 14:21:04,261 Epoch   9 Step:     3100 Batch Loss:     2.652828 Tokens per Sec:     7283, Lr: 0.000300\n",
      "2019-11-26 14:21:34,968 Epoch   9 Step:     3200 Batch Loss:     2.133620 Tokens per Sec:     7516, Lr: 0.000300\n",
      "2019-11-26 14:21:39,535 Epoch   9: total training loss 911.46\n",
      "2019-11-26 14:21:39,536 EPOCH 10\n",
      "2019-11-26 14:22:05,404 Epoch  10 Step:     3300 Batch Loss:     2.298213 Tokens per Sec:     7238, Lr: 0.000300\n",
      "2019-11-26 14:22:35,769 Epoch  10 Step:     3400 Batch Loss:     2.885549 Tokens per Sec:     7437, Lr: 0.000300\n",
      "2019-11-26 14:23:06,231 Epoch  10 Step:     3500 Batch Loss:     2.850529 Tokens per Sec:     7403, Lr: 0.000300\n",
      "2019-11-26 14:23:28,448 Epoch  10: total training loss 885.28\n",
      "2019-11-26 14:23:28,448 EPOCH 11\n",
      "2019-11-26 14:23:36,374 Epoch  11 Step:     3600 Batch Loss:     2.306839 Tokens per Sec:     7311, Lr: 0.000300\n",
      "2019-11-26 14:24:06,560 Epoch  11 Step:     3700 Batch Loss:     2.788274 Tokens per Sec:     7385, Lr: 0.000300\n",
      "2019-11-26 14:24:37,391 Epoch  11 Step:     3800 Batch Loss:     2.138618 Tokens per Sec:     7420, Lr: 0.000300\n",
      "2019-11-26 14:25:07,753 Epoch  11 Step:     3900 Batch Loss:     2.272465 Tokens per Sec:     7262, Lr: 0.000300\n",
      "2019-11-26 14:25:17,967 Epoch  11: total training loss 853.94\n",
      "2019-11-26 14:25:17,968 EPOCH 12\n",
      "2019-11-26 14:25:38,922 Epoch  12 Step:     4000 Batch Loss:     2.089157 Tokens per Sec:     7291, Lr: 0.000300\n",
      "2019-11-26 14:27:12,572 Hooray! New best validation result [ppl]!\n",
      "2019-11-26 14:27:12,572 Saving new checkpoint.\n",
      "2019-11-26 14:27:12,857 Example #0\n",
      "2019-11-26 14:27:12,857 \tSource:     I love pioneering in this territory .\n",
      "2019-11-26 14:27:12,858 \tReference:  Un yí wǎn nú gbexosin - alijitɔ́zɔ́ wiwa ɖò fí enɛ .\n",
      "2019-11-26 14:27:12,858 \tHypothesis: Un yí wǎn nú sinsɛnzɔ́ ɔ ɖò fí e un nɔ wà ɖò tò enɛ mɛ é .\n",
      "2019-11-26 14:27:12,858 Example #1\n",
      "2019-11-26 14:27:12,858 \tSource:     “ Too numerous to recount ” are the “ wonderful works ” we can thank and praise Jehovah for daily !\n",
      "2019-11-26 14:27:12,858 \tReference:  “ Nùjiwǔ ” e Jehovah bló bɔ mǐ sixu dokú tɔn n’i bo lɛ́ kpa susu n’i ayihɔngbe ayihɔngbe é “ sukpɔ́ ” tawun .\n",
      "2019-11-26 14:27:12,858 \tHypothesis: “ Akpá e è wlan dó “ mɛ e ɖò nǔ kplɔ́n wɛ ” é ” ɔ , mǐ sixu mɔ ɖɔ Jehovah na “ nɔ gbeji nú Jehovah .\n",
      "2019-11-26 14:27:12,858 Example #2\n",
      "2019-11-26 14:27:12,859 \tSource:     Some Christian parents serving in a foreign - language field have come to realize that their children’s interest in the truth has waned .\n",
      "2019-11-26 14:27:12,859 \tReference:  Mɛjitɔ́ Klisanwun e ɖò sinsɛnzɔ́ wà wɛ ɖò ayǐ e jí è nɔ dó gbè ɖevo ɖè lɛ é ɖé lɛ wá ɖ’ayi wu ɖɔ jlǒ e vǐ emitɔn lɛ ɖó nú nugbǒ ɔ é ɖò ɖiɖekpo wɛ .\n",
      "2019-11-26 14:27:12,859 \tHypothesis: Mɛjitɔ́ Klisanwun ɖé lɛ nɔ ɖó nukún ɖɔ vǐ yetɔn lɛ nɔ mɔ nukúnnú jɛ nugbǒ ɔ mɛ , bo nɔ mɔ ɖɔ emi ɖó na mɔ nugbǒ nugbǒ ɔ .\n",
      "2019-11-26 14:27:12,859 Example #3\n",
      "2019-11-26 14:27:12,860 \tSource:     You certainly recognize that Jewish man as the one who came to be known as the apostle Paul .\n",
      "2019-11-26 14:27:12,860 \tReference:  1 : 14 ) É ɖò wɛn ɖɔ a tuùn nya Jwifu enɛ e è wá ylɔ ɖɔ mɛsɛ́dó Pɔlu é .\n",
      "2019-11-26 14:27:12,860 \tHypothesis: É ɖò wɛn ɖɔ a tuùn Jwifu Jwifu lɛ tɔn e ko tuùn Timɔtée é ɖé .\n",
      "2019-11-26 14:27:12,860 Validation result (greedy) at epoch  12, step     4000: bleu:  12.24, loss: 68261.9844, ppl:   9.9929, duration: 93.9380s\n",
      "2019-11-26 14:27:43,810 Epoch  12 Step:     4100 Batch Loss:     1.972761 Tokens per Sec:     7340, Lr: 0.000300\n",
      "2019-11-26 14:28:14,582 Epoch  12 Step:     4200 Batch Loss:     2.161382 Tokens per Sec:     7339, Lr: 0.000300\n",
      "2019-11-26 14:28:41,368 Epoch  12: total training loss 826.62\n",
      "2019-11-26 14:28:41,368 EPOCH 13\n",
      "2019-11-26 14:28:45,448 Epoch  13 Step:     4300 Batch Loss:     2.056029 Tokens per Sec:     6982, Lr: 0.000300\n",
      "2019-11-26 14:29:16,288 Epoch  13 Step:     4400 Batch Loss:     2.623316 Tokens per Sec:     7358, Lr: 0.000300\n",
      "2019-11-26 14:29:47,095 Epoch  13 Step:     4500 Batch Loss:     2.337763 Tokens per Sec:     7360, Lr: 0.000300\n",
      "2019-11-26 14:30:18,009 Epoch  13 Step:     4600 Batch Loss:     2.540540 Tokens per Sec:     7315, Lr: 0.000300\n",
      "2019-11-26 14:30:30,997 Epoch  13: total training loss 803.65\n",
      "2019-11-26 14:30:30,997 EPOCH 14\n",
      "2019-11-26 14:30:48,852 Epoch  14 Step:     4700 Batch Loss:     2.418147 Tokens per Sec:     7226, Lr: 0.000300\n",
      "2019-11-26 14:31:19,645 Epoch  14 Step:     4800 Batch Loss:     2.683759 Tokens per Sec:     7344, Lr: 0.000300\n",
      "2019-11-26 14:31:50,235 Epoch  14 Step:     4900 Batch Loss:     1.871324 Tokens per Sec:     7352, Lr: 0.000300\n",
      "2019-11-26 14:32:20,471 Epoch  14: total training loss 781.40\n",
      "2019-11-26 14:32:20,471 EPOCH 15\n",
      "2019-11-26 14:32:21,448 Epoch  15 Step:     5000 Batch Loss:     2.396113 Tokens per Sec:     7021, Lr: 0.000300\n",
      "2019-11-26 14:33:54,721 Hooray! New best validation result [ppl]!\n",
      "2019-11-26 14:33:54,721 Saving new checkpoint.\n",
      "2019-11-26 14:33:55,037 Example #0\n",
      "2019-11-26 14:33:55,037 \tSource:     I love pioneering in this territory .\n",
      "2019-11-26 14:33:55,037 \tReference:  Un yí wǎn nú gbexosin - alijitɔ́zɔ́ wiwa ɖò fí enɛ .\n",
      "2019-11-26 14:33:55,037 \tHypothesis: Un yí wǎn nú sinsɛnzɔ́ ɔ ɖò fí e un nɔ wà ɖò fí é .\n",
      "2019-11-26 14:33:55,037 Example #1\n",
      "2019-11-26 14:33:55,037 \tSource:     “ Too numerous to recount ” are the “ wonderful works ” we can thank and praise Jehovah for daily !\n",
      "2019-11-26 14:33:55,038 \tReference:  “ Nùjiwǔ ” e Jehovah bló bɔ mǐ sixu dokú tɔn n’i bo lɛ́ kpa susu n’i ayihɔngbe ayihɔngbe é “ sukpɔ́ ” tawun .\n",
      "2019-11-26 14:33:55,038 \tHypothesis: “ Mi kpɔ́n lee nǔ cí nú mǐ é , bo na dó wusyɛn lanmɛ nú mǐ ɖɔ mǐ ni nɔ gbeji nú Jehovah , bo na dó wusyɛn lanmɛ nú mǐ , bo na dó wusyɛn lanmɛ nú mǐ ɖɔ mǐ ni nɔ gbeji nú Jehovah !\n",
      "2019-11-26 14:33:55,038 Example #2\n",
      "2019-11-26 14:33:55,038 \tSource:     Some Christian parents serving in a foreign - language field have come to realize that their children’s interest in the truth has waned .\n",
      "2019-11-26 14:33:55,038 \tReference:  Mɛjitɔ́ Klisanwun e ɖò sinsɛnzɔ́ wà wɛ ɖò ayǐ e jí è nɔ dó gbè ɖevo ɖè lɛ é ɖé lɛ wá ɖ’ayi wu ɖɔ jlǒ e vǐ emitɔn lɛ ɖó nú nugbǒ ɔ é ɖò ɖiɖekpo wɛ .\n",
      "2019-11-26 14:33:55,038 \tHypothesis: Mɛjitɔ́ Klisanwun ɖé lɛ nɔ sɛ̀n ɖò xá e mɛ winnyawinnya lɛ nɔ mɔ ɖɔ emi kún ɖó na mɔ nugbǒ ɔ ó é ɖé lɛ ó .\n",
      "2019-11-26 14:33:55,038 Example #3\n",
      "2019-11-26 14:33:55,038 \tSource:     You certainly recognize that Jewish man as the one who came to be known as the apostle Paul .\n",
      "2019-11-26 14:33:55,038 \tReference:  1 : 14 ) É ɖò wɛn ɖɔ a tuùn nya Jwifu enɛ e è wá ylɔ ɖɔ mɛsɛ́dó Pɔlu é .\n",
      "2019-11-26 14:33:55,039 \tHypothesis: É ɖò wɛn ɖɔ a tuùn ɖɔ Jwifu lɛ ko tuùn ye .\n",
      "2019-11-26 14:33:55,039 Validation result (greedy) at epoch  15, step     5000: bleu:  14.34, loss: 64531.0781, ppl:   8.8115, duration: 93.5903s\n",
      "2019-11-26 14:34:25,789 Epoch  15 Step:     5100 Batch Loss:     2.222859 Tokens per Sec:     7383, Lr: 0.000300\n",
      "2019-11-26 14:34:56,436 Epoch  15 Step:     5200 Batch Loss:     2.189746 Tokens per Sec:     7351, Lr: 0.000300\n",
      "2019-11-26 14:35:27,218 Epoch  15 Step:     5300 Batch Loss:     2.335963 Tokens per Sec:     7297, Lr: 0.000300\n",
      "2019-11-26 14:35:43,859 Epoch  15: total training loss 767.66\n",
      "2019-11-26 14:35:43,860 EPOCH 16\n",
      "2019-11-26 14:35:57,639 Epoch  16 Step:     5400 Batch Loss:     2.270057 Tokens per Sec:     7111, Lr: 0.000300\n",
      "2019-11-26 14:36:28,032 Epoch  16 Step:     5500 Batch Loss:     2.056870 Tokens per Sec:     7154, Lr: 0.000300\n",
      "2019-11-26 14:36:58,671 Epoch  16 Step:     5600 Batch Loss:     1.063942 Tokens per Sec:     7178, Lr: 0.000300\n",
      "2019-11-26 14:37:29,562 Epoch  16 Step:     5700 Batch Loss:     2.415873 Tokens per Sec:     7487, Lr: 0.000300\n",
      "2019-11-26 14:37:34,438 Epoch  16: total training loss 759.58\n",
      "2019-11-26 14:37:34,439 EPOCH 17\n",
      "2019-11-26 14:38:00,049 Epoch  17 Step:     5800 Batch Loss:     2.065362 Tokens per Sec:     7273, Lr: 0.000300\n",
      "2019-11-26 14:38:30,954 Epoch  17 Step:     5900 Batch Loss:     2.120649 Tokens per Sec:     7279, Lr: 0.000300\n",
      "2019-11-26 14:39:01,390 Epoch  17 Step:     6000 Batch Loss:     2.329014 Tokens per Sec:     7347, Lr: 0.000300\n",
      "2019-11-26 14:40:34,786 Hooray! New best validation result [ppl]!\n",
      "2019-11-26 14:40:34,787 Saving new checkpoint.\n",
      "2019-11-26 14:40:35,087 Example #0\n",
      "2019-11-26 14:40:35,088 \tSource:     I love pioneering in this territory .\n",
      "2019-11-26 14:40:35,088 \tReference:  Un yí wǎn nú gbexosin - alijitɔ́zɔ́ wiwa ɖò fí enɛ .\n",
      "2019-11-26 14:40:35,088 \tHypothesis: Un yí wǎn nú sinsɛnzɔ́ ɔ ɖò fí enɛ .\n",
      "2019-11-26 14:40:35,088 Example #1\n",
      "2019-11-26 14:40:35,088 \tSource:     “ Too numerous to recount ” are the “ wonderful works ” we can thank and praise Jehovah for daily !\n",
      "2019-11-26 14:40:35,088 \tReference:  “ Nùjiwǔ ” e Jehovah bló bɔ mǐ sixu dokú tɔn n’i bo lɛ́ kpa susu n’i ayihɔngbe ayihɔngbe é “ sukpɔ́ ” tawun .\n",
      "2019-11-26 14:40:35,088 \tHypothesis: “ È flín nǔ e mǐ sixu mɔ ɖò “ Aklunɔ sín akpá ” é , bo na dó wusyɛn lanmɛ nú mǐ , bo na dó wusyɛn lanmɛ nú Jehovah , bo lɛ́ dokú n’i .\n",
      "2019-11-26 14:40:35,090 Example #2\n",
      "2019-11-26 14:40:35,090 \tSource:     Some Christian parents serving in a foreign - language field have come to realize that their children’s interest in the truth has waned .\n",
      "2019-11-26 14:40:35,090 \tReference:  Mɛjitɔ́ Klisanwun e ɖò sinsɛnzɔ́ wà wɛ ɖò ayǐ e jí è nɔ dó gbè ɖevo ɖè lɛ é ɖé lɛ wá ɖ’ayi wu ɖɔ jlǒ e vǐ emitɔn lɛ ɖó nú nugbǒ ɔ é ɖò ɖiɖekpo wɛ .\n",
      "2019-11-26 14:40:35,090 \tHypothesis: Mɛjitɔ́ Klisanwun ɖé lɛ nɔ sɛ̀n ɖò xá e mɛ winnyawinnya lɛ nɔ mɔ kpɔ́n ɖò wemaxɔmɛ é mɛ , bo nɔ mɔ ɖɔ nugbǒ ɔ ɖó na ɖó awǎjijɛ ɖò nugbǒ ɔ mɛ .\n",
      "2019-11-26 14:40:35,090 Example #3\n",
      "2019-11-26 14:40:35,091 \tSource:     You certainly recognize that Jewish man as the one who came to be known as the apostle Paul .\n",
      "2019-11-26 14:40:35,091 \tReference:  1 : 14 ) É ɖò wɛn ɖɔ a tuùn nya Jwifu enɛ e è wá ylɔ ɖɔ mɛsɛ́dó Pɔlu é .\n",
      "2019-11-26 14:40:35,091 \tHypothesis: É ɖò wɛn ɖɔ a tuùn ɖɔ Jwifu lɛ kún tuùn ɖɔ emi kún nyí mɛsɛ́dó ó .\n",
      "2019-11-26 14:40:35,091 Validation result (greedy) at epoch  17, step     6000: bleu:  16.15, loss: 61767.5156, ppl:   8.0275, duration: 93.7005s\n",
      "2019-11-26 14:40:58,239 Epoch  17: total training loss 738.82\n",
      "2019-11-26 14:40:58,239 EPOCH 18\n",
      "2019-11-26 14:41:05,759 Epoch  18 Step:     6100 Batch Loss:     2.057728 Tokens per Sec:     7376, Lr: 0.000300\n",
      "2019-11-26 14:41:36,034 Epoch  18 Step:     6200 Batch Loss:     1.926301 Tokens per Sec:     7317, Lr: 0.000300\n",
      "2019-11-26 14:42:07,031 Epoch  18 Step:     6300 Batch Loss:     2.024534 Tokens per Sec:     7422, Lr: 0.000300\n",
      "2019-11-26 14:42:37,848 Epoch  18 Step:     6400 Batch Loss:     2.170782 Tokens per Sec:     7388, Lr: 0.000300\n",
      "2019-11-26 14:42:47,674 Epoch  18: total training loss 723.88\n",
      "2019-11-26 14:42:47,674 EPOCH 19\n",
      "2019-11-26 14:43:08,864 Epoch  19 Step:     6500 Batch Loss:     2.022346 Tokens per Sec:     7366, Lr: 0.000300\n",
      "2019-11-26 14:43:39,532 Epoch  19 Step:     6600 Batch Loss:     1.915361 Tokens per Sec:     7285, Lr: 0.000300\n",
      "2019-11-26 14:44:10,213 Epoch  19 Step:     6700 Batch Loss:     1.691444 Tokens per Sec:     7363, Lr: 0.000300\n",
      "2019-11-26 14:44:37,292 Epoch  19: total training loss 708.88\n",
      "2019-11-26 14:44:37,292 EPOCH 20\n",
      "2019-11-26 14:44:40,694 Epoch  20 Step:     6800 Batch Loss:     1.902304 Tokens per Sec:     7373, Lr: 0.000300\n",
      "2019-11-26 14:45:11,502 Epoch  20 Step:     6900 Batch Loss:     2.074996 Tokens per Sec:     7406, Lr: 0.000300\n",
      "2019-11-26 14:45:42,154 Epoch  20 Step:     7000 Batch Loss:     2.250434 Tokens per Sec:     7260, Lr: 0.000300\n",
      "2019-11-26 14:47:15,549 Hooray! New best validation result [ppl]!\n",
      "2019-11-26 14:47:15,550 Saving new checkpoint.\n",
      "2019-11-26 14:47:15,863 Example #0\n",
      "2019-11-26 14:47:15,863 \tSource:     I love pioneering in this territory .\n",
      "2019-11-26 14:47:15,864 \tReference:  Un yí wǎn nú gbexosin - alijitɔ́zɔ́ wiwa ɖò fí enɛ .\n",
      "2019-11-26 14:47:15,864 \tHypothesis: Un yí wǎn nú gbexosin - alijitɔ́zɔ́ ɖò fí enɛ .\n",
      "2019-11-26 14:47:15,864 Example #1\n",
      "2019-11-26 14:47:15,864 \tSource:     “ Too numerous to recount ” are the “ wonderful works ” we can thank and praise Jehovah for daily !\n",
      "2019-11-26 14:47:15,864 \tReference:  “ Nùjiwǔ ” e Jehovah bló bɔ mǐ sixu dokú tɔn n’i bo lɛ́ kpa susu n’i ayihɔngbe ayihɔngbe é “ sukpɔ́ ” tawun .\n",
      "2019-11-26 14:47:15,864 \tHypothesis: “ Togun ɔ ” ɔ , mǐ sixu mɔ ɖɔ ‘ Jehovah na kpa susu nú Jehovah , bo na kpa susu n’i .\n",
      "2019-11-26 14:47:15,864 Example #2\n",
      "2019-11-26 14:47:15,865 \tSource:     Some Christian parents serving in a foreign - language field have come to realize that their children’s interest in the truth has waned .\n",
      "2019-11-26 14:47:15,865 \tReference:  Mɛjitɔ́ Klisanwun e ɖò sinsɛnzɔ́ wà wɛ ɖò ayǐ e jí è nɔ dó gbè ɖevo ɖè lɛ é ɖé lɛ wá ɖ’ayi wu ɖɔ jlǒ e vǐ emitɔn lɛ ɖó nú nugbǒ ɔ é ɖò ɖiɖekpo wɛ .\n",
      "2019-11-26 14:47:15,865 \tHypothesis: Mɛjitɔ́ Klisanwun ɖé lɛ nɔ sɛ̀n ɖò xá e mɛ vǐ yetɔn lɛ nɔ mɔ ɖɔ nugbǒ ɔ kún ɖó ayi wu ó lɛ é mɛ .\n",
      "2019-11-26 14:47:15,865 Example #3\n",
      "2019-11-26 14:47:15,866 \tSource:     You certainly recognize that Jewish man as the one who came to be known as the apostle Paul .\n",
      "2019-11-26 14:47:15,866 \tReference:  1 : 14 ) É ɖò wɛn ɖɔ a tuùn nya Jwifu enɛ e è wá ylɔ ɖɔ mɛsɛ́dó Pɔlu é .\n",
      "2019-11-26 14:47:15,866 \tHypothesis: É ɖò wɛn ɖɔ a tuùn ɖɔ Jwifu e ko tuùn ɖɔ Jwifu lɛ wɛ nyí mɛsɛ́dó Pɔlu é ɔ , é wá tuùn i .\n",
      "2019-11-26 14:47:15,866 Validation result (greedy) at epoch  20, step     7000: bleu:  17.57, loss: 59520.3320, ppl:   7.4417, duration: 93.7114s\n",
      "2019-11-26 14:47:46,653 Epoch  20 Step:     7100 Batch Loss:     1.648879 Tokens per Sec:     7420, Lr: 0.000300\n",
      "2019-11-26 14:48:00,386 Epoch  20: total training loss 693.25\n",
      "2019-11-26 14:48:00,386 EPOCH 21\n",
      "2019-11-26 14:48:17,234 Epoch  21 Step:     7200 Batch Loss:     1.636328 Tokens per Sec:     7419, Lr: 0.000300\n",
      "2019-11-26 14:48:48,219 Epoch  21 Step:     7300 Batch Loss:     1.877475 Tokens per Sec:     7234, Lr: 0.000300\n",
      "2019-11-26 14:49:18,934 Epoch  21 Step:     7400 Batch Loss:     1.953899 Tokens per Sec:     7246, Lr: 0.000300\n",
      "2019-11-26 14:49:49,633 Epoch  21 Step:     7500 Batch Loss:     2.067929 Tokens per Sec:     7261, Lr: 0.000300\n",
      "2019-11-26 14:49:50,886 Epoch  21: total training loss 690.18\n",
      "2019-11-26 14:49:50,886 EPOCH 22\n",
      "2019-11-26 14:50:20,324 Epoch  22 Step:     7600 Batch Loss:     2.043521 Tokens per Sec:     7264, Lr: 0.000300\n",
      "2019-11-26 14:50:50,987 Epoch  22 Step:     7700 Batch Loss:     2.141779 Tokens per Sec:     7378, Lr: 0.000300\n",
      "2019-11-26 14:51:21,543 Epoch  22 Step:     7800 Batch Loss:     2.099143 Tokens per Sec:     7436, Lr: 0.000300\n",
      "2019-11-26 14:51:40,395 Epoch  22: total training loss 674.03\n",
      "2019-11-26 14:51:40,395 EPOCH 23\n",
      "2019-11-26 14:51:52,543 Epoch  23 Step:     7900 Batch Loss:     2.173620 Tokens per Sec:     7561, Lr: 0.000300\n",
      "2019-11-26 14:52:23,387 Epoch  23 Step:     8000 Batch Loss:     2.227573 Tokens per Sec:     7290, Lr: 0.000300\n",
      "2019-11-26 14:53:56,766 Hooray! New best validation result [ppl]!\n",
      "2019-11-26 14:53:56,766 Saving new checkpoint.\n",
      "2019-11-26 14:53:57,064 Example #0\n",
      "2019-11-26 14:53:57,065 \tSource:     I love pioneering in this territory .\n",
      "2019-11-26 14:53:57,065 \tReference:  Un yí wǎn nú gbexosin - alijitɔ́zɔ́ wiwa ɖò fí enɛ .\n",
      "2019-11-26 14:53:57,065 \tHypothesis: Un yí wǎn nú gbexosin - alijitɔ́zɔ́ ɔ ɖò fí enɛ .\n",
      "2019-11-26 14:53:57,065 Example #1\n",
      "2019-11-26 14:53:57,065 \tSource:     “ Too numerous to recount ” are the “ wonderful works ” we can thank and praise Jehovah for daily !\n",
      "2019-11-26 14:53:57,066 \tReference:  “ Nùjiwǔ ” e Jehovah bló bɔ mǐ sixu dokú tɔn n’i bo lɛ́ kpa susu n’i ayihɔngbe ayihɔngbe é “ sukpɔ́ ” tawun .\n",
      "2019-11-26 14:53:57,066 \tHypothesis: “ Togun ɔ ” jiwǔ tawun ɖɔ mǐ ni kpa susu nú Jehovah , bo lɛ́ dokú n’i n’i .\n",
      "2019-11-26 14:53:57,066 Example #2\n",
      "2019-11-26 14:53:57,067 \tSource:     Some Christian parents serving in a foreign - language field have come to realize that their children’s interest in the truth has waned .\n",
      "2019-11-26 14:53:57,067 \tReference:  Mɛjitɔ́ Klisanwun e ɖò sinsɛnzɔ́ wà wɛ ɖò ayǐ e jí è nɔ dó gbè ɖevo ɖè lɛ é ɖé lɛ wá ɖ’ayi wu ɖɔ jlǒ e vǐ emitɔn lɛ ɖó nú nugbǒ ɔ é ɖò ɖiɖekpo wɛ .\n",
      "2019-11-26 14:53:57,067 \tHypothesis: Mɛjitɔ́ ɖé lɛ nɔ sɛ̀n ɖò xá e mɛ vǐ yetɔn lɛ nɔ dó gbè ɖevo mɛ é ɖé lɛ mɔ ɖɔ nugbǒ ɔ kún ɖó awǎjijɛ ó .\n",
      "2019-11-26 14:53:57,067 Example #3\n",
      "2019-11-26 14:53:57,067 \tSource:     You certainly recognize that Jewish man as the one who came to be known as the apostle Paul .\n",
      "2019-11-26 14:53:57,067 \tReference:  1 : 14 ) É ɖò wɛn ɖɔ a tuùn nya Jwifu enɛ e è wá ylɔ ɖɔ mɛsɛ́dó Pɔlu é .\n",
      "2019-11-26 14:53:57,067 \tHypothesis: É ɖò wɛn ɖɔ a tuùn ɖɔ Jwifu e ko tuùn nǔ e ye tuùn lɛ é wɛ nyí mɛsɛ́dó Pɔlu .\n",
      "2019-11-26 14:53:57,067 Validation result (greedy) at epoch  23, step     8000: bleu:  18.66, loss: 57724.9688, ppl:   7.0045, duration: 93.6799s\n",
      "2019-11-26 14:54:27,396 Epoch  23 Step:     8100 Batch Loss:     1.993855 Tokens per Sec:     7262, Lr: 0.000300\n",
      "2019-11-26 14:54:58,225 Epoch  23 Step:     8200 Batch Loss:     1.721530 Tokens per Sec:     7242, Lr: 0.000300\n",
      "2019-11-26 14:55:03,931 Epoch  23: total training loss 664.34\n",
      "2019-11-26 14:55:03,931 EPOCH 24\n",
      "2019-11-26 14:55:28,869 Epoch  24 Step:     8300 Batch Loss:     1.609951 Tokens per Sec:     7309, Lr: 0.000300\n",
      "2019-11-26 14:55:59,418 Epoch  24 Step:     8400 Batch Loss:     1.596233 Tokens per Sec:     7278, Lr: 0.000300\n",
      "2019-11-26 14:56:30,194 Epoch  24 Step:     8500 Batch Loss:     0.993275 Tokens per Sec:     7256, Lr: 0.000300\n",
      "2019-11-26 14:56:54,172 Epoch  24: total training loss 659.35\n",
      "2019-11-26 14:56:54,173 EPOCH 25\n",
      "2019-11-26 14:57:00,783 Epoch  25 Step:     8600 Batch Loss:     1.881945 Tokens per Sec:     7052, Lr: 0.000300\n",
      "2019-11-26 14:57:31,541 Epoch  25 Step:     8700 Batch Loss:     1.651695 Tokens per Sec:     7310, Lr: 0.000300\n",
      "2019-11-26 14:58:02,219 Epoch  25 Step:     8800 Batch Loss:     1.661671 Tokens per Sec:     7294, Lr: 0.000300\n",
      "2019-11-26 14:58:33,202 Epoch  25 Step:     8900 Batch Loss:     2.083404 Tokens per Sec:     7341, Lr: 0.000300\n",
      "2019-11-26 14:58:44,241 Epoch  25: total training loss 641.46\n",
      "2019-11-26 14:58:44,241 EPOCH 26\n",
      "2019-11-26 14:59:04,129 Epoch  26 Step:     9000 Batch Loss:     2.332623 Tokens per Sec:     7369, Lr: 0.000300\n",
      "2019-11-26 15:00:36,933 Hooray! New best validation result [ppl]!\n",
      "2019-11-26 15:00:36,933 Saving new checkpoint.\n",
      "2019-11-26 15:00:37,275 Example #0\n",
      "2019-11-26 15:00:37,276 \tSource:     I love pioneering in this territory .\n",
      "2019-11-26 15:00:37,276 \tReference:  Un yí wǎn nú gbexosin - alijitɔ́zɔ́ wiwa ɖò fí enɛ .\n",
      "2019-11-26 15:00:37,276 \tHypothesis: Un yí wǎn nú gbexosin - alijitɔ́zɔ́ wiwa ɖò fí enɛ .\n",
      "2019-11-26 15:00:37,276 Example #1\n",
      "2019-11-26 15:00:37,276 \tSource:     “ Too numerous to recount ” are the “ wonderful works ” we can thank and praise Jehovah for daily !\n",
      "2019-11-26 15:00:37,276 \tReference:  “ Nùjiwǔ ” e Jehovah bló bɔ mǐ sixu dokú tɔn n’i bo lɛ́ kpa susu n’i ayihɔngbe ayihɔngbe é “ sukpɔ́ ” tawun .\n",
      "2019-11-26 15:00:37,276 \tHypothesis: “ Togun ” ɔ , è nɔ mɔ nǔ jɛ wu ɖɔ mǐ sixu kpa susu nú Jehovah , bo dokú nú Jehovah .\n",
      "2019-11-26 15:00:37,276 Example #2\n",
      "2019-11-26 15:00:37,276 \tSource:     Some Christian parents serving in a foreign - language field have come to realize that their children’s interest in the truth has waned .\n",
      "2019-11-26 15:00:37,277 \tReference:  Mɛjitɔ́ Klisanwun e ɖò sinsɛnzɔ́ wà wɛ ɖò ayǐ e jí è nɔ dó gbè ɖevo ɖè lɛ é ɖé lɛ wá ɖ’ayi wu ɖɔ jlǒ e vǐ emitɔn lɛ ɖó nú nugbǒ ɔ é ɖò ɖiɖekpo wɛ .\n",
      "2019-11-26 15:00:37,277 \tHypothesis: Mɛjitɔ́ Klisanwun ɖé lɛ nɔ sɛ̀n sinsɛnzɔ́ ɖò xá e mɛ vǐ yetɔn lɛ nɔ dó gbè ɖevo ɖè é ɖé lɛ mɔ ɖɔ emi ɖó ayi wu ɖɔ nugbǒ ɔ tɔn lɛ ɖó ayi wu .\n",
      "2019-11-26 15:00:37,277 Example #3\n",
      "2019-11-26 15:00:37,277 \tSource:     You certainly recognize that Jewish man as the one who came to be known as the apostle Paul .\n",
      "2019-11-26 15:00:37,277 \tReference:  1 : 14 ) É ɖò wɛn ɖɔ a tuùn nya Jwifu enɛ e è wá ylɔ ɖɔ mɛsɛ́dó Pɔlu é .\n",
      "2019-11-26 15:00:37,277 \tHypothesis: É ɖò wɛn ɖɔ a tuùn ɖɔ Jwifu e nyí Jwifu lɛ é ɖokpo wɛ nyí mɛsɛ́dó Pɔlu .\n",
      "2019-11-26 15:00:37,277 Validation result (greedy) at epoch  26, step     9000: bleu:  19.42, loss: 56730.3633, ppl:   6.7735, duration: 93.1475s\n",
      "2019-11-26 15:01:07,938 Epoch  26 Step:     9100 Batch Loss:     1.694081 Tokens per Sec:     7233, Lr: 0.000300\n",
      "2019-11-26 15:01:38,622 Epoch  26 Step:     9200 Batch Loss:     1.467636 Tokens per Sec:     7367, Lr: 0.000300\n",
      "2019-11-26 15:02:07,152 Epoch  26: total training loss 629.64\n",
      "2019-11-26 15:02:07,152 EPOCH 27\n",
      "2019-11-26 15:02:09,702 Epoch  27 Step:     9300 Batch Loss:     1.865959 Tokens per Sec:     7843, Lr: 0.000300\n",
      "2019-11-26 15:02:40,799 Epoch  27 Step:     9400 Batch Loss:     1.647859 Tokens per Sec:     7249, Lr: 0.000300\n",
      "2019-11-26 15:03:11,592 Epoch  27 Step:     9500 Batch Loss:     1.956773 Tokens per Sec:     7326, Lr: 0.000300\n",
      "2019-11-26 15:03:42,463 Epoch  27 Step:     9600 Batch Loss:     1.818972 Tokens per Sec:     7349, Lr: 0.000300\n",
      "2019-11-26 15:03:56,874 Epoch  27: total training loss 617.07\n",
      "2019-11-26 15:03:56,875 EPOCH 28\n",
      "2019-11-26 15:04:13,630 Epoch  28 Step:     9700 Batch Loss:     1.014780 Tokens per Sec:     7280, Lr: 0.000300\n",
      "2019-11-26 15:04:44,556 Epoch  28 Step:     9800 Batch Loss:     1.767276 Tokens per Sec:     7404, Lr: 0.000300\n",
      "2019-11-26 15:05:15,211 Epoch  28 Step:     9900 Batch Loss:     2.007863 Tokens per Sec:     7223, Lr: 0.000300\n",
      "2019-11-26 15:05:45,999 Epoch  28 Step:    10000 Batch Loss:     1.755681 Tokens per Sec:     7373, Lr: 0.000300\n",
      "2019-11-26 15:07:19,565 Hooray! New best validation result [ppl]!\n",
      "2019-11-26 15:07:19,565 Saving new checkpoint.\n",
      "2019-11-26 15:07:19,864 Example #0\n",
      "2019-11-26 15:07:19,864 \tSource:     I love pioneering in this territory .\n",
      "2019-11-26 15:07:19,865 \tReference:  Un yí wǎn nú gbexosin - alijitɔ́zɔ́ wiwa ɖò fí enɛ .\n",
      "2019-11-26 15:07:19,865 \tHypothesis: Un yí wǎn nú gbexosin - alijitɔ́zɔ́ ɖò fí e un ɖè é .\n",
      "2019-11-26 15:07:19,865 Example #1\n",
      "2019-11-26 15:07:19,865 \tSource:     “ Too numerous to recount ” are the “ wonderful works ” we can thank and praise Jehovah for daily !\n",
      "2019-11-26 15:07:19,865 \tReference:  “ Nùjiwǔ ” e Jehovah bló bɔ mǐ sixu dokú tɔn n’i bo lɛ́ kpa susu n’i ayihɔngbe ayihɔngbe é “ sukpɔ́ ” tawun .\n",
      "2019-11-26 15:07:19,865 \tHypothesis: “ Togun Mawu tɔn ” jiwǔ tawun , bɔ mǐ sixu dokú nú Jehovah , bo dokú n’i .\n",
      "2019-11-26 15:07:19,866 Example #2\n",
      "2019-11-26 15:07:19,866 \tSource:     Some Christian parents serving in a foreign - language field have come to realize that their children’s interest in the truth has waned .\n",
      "2019-11-26 15:07:19,866 \tReference:  Mɛjitɔ́ Klisanwun e ɖò sinsɛnzɔ́ wà wɛ ɖò ayǐ e jí è nɔ dó gbè ɖevo ɖè lɛ é ɖé lɛ wá ɖ’ayi wu ɖɔ jlǒ e vǐ emitɔn lɛ ɖó nú nugbǒ ɔ é ɖò ɖiɖekpo wɛ .\n",
      "2019-11-26 15:07:19,866 \tHypothesis: Mɛjitɔ́ Klisanwun ɖé lɛ nɔ sɛ̀n sinsɛnzɔ́ ɖò xá e mɛ vǐ yetɔn lɛ nɔ dó gbè ɖevo mɛ é ɖé lɛ mɔ ɖɔ nugbǒ ɔ ɖó kpɔ́ .\n",
      "2019-11-26 15:07:19,866 Example #3\n",
      "2019-11-26 15:07:19,867 \tSource:     You certainly recognize that Jewish man as the one who came to be known as the apostle Paul .\n",
      "2019-11-26 15:07:19,867 \tReference:  1 : 14 ) É ɖò wɛn ɖɔ a tuùn nya Jwifu enɛ e è wá ylɔ ɖɔ mɛsɛ́dó Pɔlu é .\n",
      "2019-11-26 15:07:19,867 \tHypothesis: É ɖò wɛn ɖɔ a tuùn ɖɔ Jwifu e wá tuùn mɛsɛ́dó Pɔlu é wɛ nyí mɛsɛ́dó Pɔlu .\n",
      "2019-11-26 15:07:19,867 Validation result (greedy) at epoch  28, step    10000: bleu:  20.28, loss: 54994.6992, ppl:   6.3884, duration: 93.8673s\n",
      "2019-11-26 15:07:20,499 Epoch  28: total training loss 611.54\n",
      "2019-11-26 15:07:20,500 EPOCH 29\n",
      "2019-11-26 15:07:50,840 Epoch  29 Step:    10100 Batch Loss:     1.650072 Tokens per Sec:     7307, Lr: 0.000300\n",
      "2019-11-26 15:08:21,642 Epoch  29 Step:    10200 Batch Loss:     1.353416 Tokens per Sec:     7449, Lr: 0.000300\n",
      "2019-11-26 15:08:52,008 Epoch  29 Step:    10300 Batch Loss:     1.798775 Tokens per Sec:     7367, Lr: 0.000300\n",
      "2019-11-26 15:09:09,910 Epoch  29: total training loss 604.65\n",
      "2019-11-26 15:09:09,911 EPOCH 30\n",
      "2019-11-26 15:09:22,935 Epoch  30 Step:    10400 Batch Loss:     1.441740 Tokens per Sec:     7415, Lr: 0.000300\n",
      "2019-11-26 15:09:53,582 Epoch  30 Step:    10500 Batch Loss:     2.227500 Tokens per Sec:     7292, Lr: 0.000300\n",
      "2019-11-26 15:10:24,477 Epoch  30 Step:    10600 Batch Loss:     2.074497 Tokens per Sec:     7352, Lr: 0.000300\n",
      "2019-11-26 15:10:55,341 Epoch  30 Step:    10700 Batch Loss:     2.114685 Tokens per Sec:     7480, Lr: 0.000300\n",
      "2019-11-26 15:10:59,098 Epoch  30: total training loss 591.97\n",
      "2019-11-26 15:10:59,098 Training ended after  30 epochs.\n",
      "2019-11-26 15:10:59,098 Best validation result (greedy) at step    10000:   6.39 ppl.\n",
      "2019-11-26 15:11:51,463  dev bleu:  21.61 [Beam search decoding with beam size = 5 and alpha = 1.0]\n",
      "2019-11-26 15:11:51,464 Translations saved to: models/enfon_transformer/00010000.hyps.dev\n",
      "2019-11-26 15:13:36,964 test bleu:  31.07 [Beam search decoding with beam size = 5 and alpha = 1.0]\n",
      "2019-11-26 15:13:36,967 Translations saved to: models/enfon_transformer/00010000.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": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "MBoDS09JM807",
    "outputId": "6cf7a2ff-fba7-4dd8-beb8-127ab80a5fe3"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "cp: cannot create symbolic link '/content/drive/My Drive/masakhane/en-fon-baseline/models/enfon_transformer/best.ckpt': Function not implemented\n"
     ]
    }
   ],
   "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}${tgt}_transformer/\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "AYCOImAVVCFJ"
   },
   "outputs": [],
   "source": [
    "# Copy the created models from the notebook storage to google drive for persistant storage \n",
    "!cp joeynmt/models/${src}${tgt}_transformer/best.ckpt \"$gdrive_path/models/${src}${tgt}_transformer/\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 102
    },
    "colab_type": "code",
    "id": "cSoGZWeeUFob",
    "outputId": "0f33d9fc-7df4-4593-96b0-860099939aba"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "00010000.hyps.dev   2000.hyps  7000.hyps  best.ckpt\t trg_vocab.txt\n",
      "00010000.hyps.test  3000.hyps  8000.ckpt  config.yaml\t validations.txt\n",
      "10000.ckpt\t    4000.hyps  8000.hyps  src_vocab.txt\n",
      "10000.hyps\t    5000.hyps  9000.ckpt  tensorboard\n",
      "1000.hyps\t    6000.hyps  9000.hyps  train.log\n"
     ]
    }
   ],
   "source": [
    "!ls joeynmt/models/${src}${tgt}_transformer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 187
    },
    "colab_type": "code",
    "id": "n94wlrCjVc17",
    "outputId": "30c9c4b5-a0e3-4291-ae0c-92d4ad75765b"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Steps: 1000\tLoss: 99911.83594\tPPL: 29.05352\tbleu: 2.72236\tLR: 0.00030000\t*\n",
      "Steps: 2000\tLoss: 83836.46875\tPPL: 16.89572\tbleu: 5.02770\tLR: 0.00030000\t*\n",
      "Steps: 3000\tLoss: 74603.65625\tPPL: 12.37551\tbleu: 8.58942\tLR: 0.00030000\t*\n",
      "Steps: 4000\tLoss: 68261.98438\tPPL: 9.99286\tbleu: 12.24441\tLR: 0.00030000\t*\n",
      "Steps: 5000\tLoss: 64531.07812\tPPL: 8.81153\tbleu: 14.33851\tLR: 0.00030000\t*\n",
      "Steps: 6000\tLoss: 61767.51562\tPPL: 8.02748\tbleu: 16.15283\tLR: 0.00030000\t*\n",
      "Steps: 7000\tLoss: 59520.33203\tPPL: 7.44165\tbleu: 17.57435\tLR: 0.00030000\t*\n",
      "Steps: 8000\tLoss: 57724.96875\tPPL: 7.00449\tbleu: 18.66193\tLR: 0.00030000\t*\n",
      "Steps: 9000\tLoss: 56730.36328\tPPL: 6.77346\tbleu: 19.42256\tLR: 0.00030000\t*\n",
      "Steps: 10000\tLoss: 54994.69922\tPPL: 6.38840\tbleu: 20.28424\tLR: 0.00030000\t*\n"
     ]
    }
   ],
   "source": [
    "# Output our validation accuracy\n",
    "! cat \"$gdrive_path/models/${src}${tgt}_transformer/validations.txt\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 68
    },
    "colab_type": "code",
    "id": "66WhRE9lIhoD",
    "outputId": "61637247-fd6b-41f5-b4a0-5d0a024af6d1"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2019-11-26 15:25:20,452 Hello! This is Joey-NMT.\n",
      "2019-11-26 15:26:14,929  dev bleu:  21.61 [Beam search decoding with beam size = 5 and alpha = 1.0]\n",
      "2019-11-26 15:28:00,670 test bleu:  31.07 [Beam search decoding with beam size = 5 and alpha = 1.0]\n"
     ]
    }
   ],
   "source": [
    "# Test our model\n",
    "! cd joeynmt; python3 -m joeynmt test \"$gdrive_path/models/${src}${tgt}_transformer/config.yaml\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "r54BFMrWURq5"
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "collapsed_sections": [],
   "name": "Copie de 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": "3.7.1"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}