en-kmb/jw300-baseline/kmb_baseline.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "kmb_baseline.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "toc_visible": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "dK0TQmS_OT_g",
        "colab_type": "text"
      },
      "source": [
        "# English to Kimbundu Baseline (Masakhane)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "GYLYy2KkOZD3",
        "colab_type": "text"
      },
      "source": [
        "## Dependencies"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "UiGBHYWtOSS9",
        "colab_type": "code",
        "outputId": "6579fb6c-fcb6-47a9-f02f-f075dafc3732",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 555
        }
      },
      "source": [
        "! apt-get install libgoogle-perftools-dev libsparsehash-dev"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Reading package lists... Done\n",
            "Building dependency tree       \n",
            "Reading state information... Done\n",
            "The following package was automatically installed and is no longer required:\n",
            "  libnvidia-common-430\n",
            "Use 'apt autoremove' to remove it.\n",
            "The following additional packages will be installed:\n",
            "  libunwind-dev\n",
            "The following NEW packages will be installed:\n",
            "  libgoogle-perftools-dev libsparsehash-dev libunwind-dev\n",
            "0 upgraded, 3 newly installed, 0 to remove and 16 not upgraded.\n",
            "Need to get 699 kB of archives.\n",
            "After this operation, 7,374 kB of additional disk space will be used.\n",
            "Get:1 http://archive.ubuntu.com/ubuntu bionic/main amd64 libunwind-dev amd64 1.2.1-8 [423 kB]\n",
            "Get:2 http://archive.ubuntu.com/ubuntu bionic/main amd64 libgoogle-perftools-dev amd64 2.5-2.2ubuntu3 [204 kB]\n",
            "Get:3 http://archive.ubuntu.com/ubuntu bionic/universe amd64 libsparsehash-dev all 2.0.2-1 [72.4 kB]\n",
            "Fetched 699 kB in 1s (750 kB/s)\n",
            "Selecting previously unselected package libunwind-dev:amd64.\n",
            "(Reading database ... 145155 files and directories currently installed.)\n",
            "Preparing to unpack .../libunwind-dev_1.2.1-8_amd64.deb ...\n",
            "Unpacking libunwind-dev:amd64 (1.2.1-8) ...\n",
            "Selecting previously unselected package libgoogle-perftools-dev.\n",
            "Preparing to unpack .../libgoogle-perftools-dev_2.5-2.2ubuntu3_amd64.deb ...\n",
            "Unpacking libgoogle-perftools-dev (2.5-2.2ubuntu3) ...\n",
            "Selecting previously unselected package libsparsehash-dev.\n",
            "Preparing to unpack .../libsparsehash-dev_2.0.2-1_all.deb ...\n",
            "Unpacking libsparsehash-dev (2.0.2-1) ...\n",
            "Setting up libsparsehash-dev (2.0.2-1) ...\n",
            "Setting up libunwind-dev:amd64 (1.2.1-8) ...\n",
            "Setting up libgoogle-perftools-dev (2.5-2.2ubuntu3) ...\n",
            "Processing triggers for man-db (2.8.3-2ubuntu0.1) ...\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "GyZDwcUgOlZo",
        "colab_type": "code",
        "outputId": "f5d67e2a-bb87-4afc-a164-6ee8b3d1afa5",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 139
        }
      },
      "source": [
        "! git clone https://github.com/clab/fast_align.git"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Cloning into 'fast_align'...\n",
            "remote: Enumerating objects: 9, done.\u001b[K\n",
            "remote: Counting objects: 100% (9/9), done.\u001b[K\n",
            "remote: Compressing objects: 100% (7/7), done.\u001b[K\n",
            "remote: Total 213 (delta 2), reused 4 (delta 2), pack-reused 204\u001b[K\n",
            "Receiving objects: 100% (213/213), 70.68 KiB | 3.07 MiB/s, done.\n",
            "Resolving deltas: 100% (110/110), done.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vrcmfy77Or0Z",
        "colab_type": "code",
        "outputId": "5d4d3ef5-c478-4caa-bc4e-69f23c8fad58",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 503
        }
      },
      "source": [
        "! cd fast_align && mkdir build && cd build && cmake .. && make"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "-- The C compiler identification is GNU 7.4.0\n",
            "-- The CXX compiler identification is GNU 7.4.0\n",
            "-- Check for working C compiler: /usr/bin/cc\n",
            "-- Check for working C compiler: /usr/bin/cc -- works\n",
            "-- Detecting C compiler ABI info\n",
            "-- Detecting C compiler ABI info - done\n",
            "-- Detecting C compile features\n",
            "-- Detecting C compile features - done\n",
            "-- Check for working CXX compiler: /usr/bin/c++\n",
            "-- Check for working CXX compiler: /usr/bin/c++ -- works\n",
            "-- Detecting CXX compiler ABI info\n",
            "-- Detecting CXX compiler ABI info - done\n",
            "-- Detecting CXX compile features\n",
            "-- Detecting CXX compile features - done\n",
            "-- Found SparseHash: /usr/include  \n",
            "-- Configuring done\n",
            "-- Generating done\n",
            "-- Build files have been written to: /content/fast_align/build\n",
            "\u001b[35m\u001b[1mScanning dependencies of target atools\u001b[0m\n",
            "[ 16%] \u001b[32mBuilding CXX object CMakeFiles/atools.dir/src/alignment_io.cc.o\u001b[0m\n",
            "[ 33%] \u001b[32mBuilding CXX object CMakeFiles/atools.dir/src/atools.cc.o\u001b[0m\n",
            "[ 50%] \u001b[32m\u001b[1mLinking CXX executable atools\u001b[0m\n",
            "[ 50%] Built target atools\n",
            "\u001b[35m\u001b[1mScanning dependencies of target fast_align\u001b[0m\n",
            "[ 66%] \u001b[32mBuilding CXX object CMakeFiles/fast_align.dir/src/fast_align.cc.o\u001b[0m\n",
            "[ 83%] \u001b[32mBuilding CXX object CMakeFiles/fast_align.dir/src/ttables.cc.o\u001b[0m\n",
            "[100%] \u001b[32m\u001b[1mLinking CXX executable fast_align\u001b[0m\n",
            "[100%] Built target fast_align\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "sJ2LEkCfO4Gx",
        "colab_type": "code",
        "outputId": "4552ff25-9464-4371-f416-f2f4ea60a2a4",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 124
        }
      },
      "source": [
        "! pip install opustools-pkg"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "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 29.7MB/s eta 0:00:01\r\u001b[K     |████████                        | 20kB 2.2MB/s eta 0:00:01\r\u001b[K     |████████████▏                   | 30kB 3.2MB/s eta 0:00:01\r\u001b[K     |████████████████▏               | 40kB 2.1MB/s eta 0:00:01\r\u001b[K     |████████████████████▎           | 51kB 2.6MB/s eta 0:00:01\r\u001b[K     |████████████████████████▎       | 61kB 3.1MB/s eta 0:00:01\r\u001b[K     |████████████████████████████▎   | 71kB 3.6MB/s eta 0:00:01\r\u001b[K     |████████████████████████████████| 81kB 3.2MB/s \n",
            "\u001b[?25hInstalling collected packages: opustools-pkg\n",
            "Successfully installed opustools-pkg-0.0.52\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_m9P3UU0PAXr",
        "colab_type": "code",
        "outputId": "89b37029-dee2-4ca7-e3c9-c08f45d9994d",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        }
      },
      "source": [
        "! git clone https://github.com/joeynmt/joeynmt.git\n",
        "! cd joeynmt; pip3 install ."
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Cloning into 'joeynmt'...\n",
            "remote: Enumerating objects: 84, done.\u001b[K\n",
            "remote: Counting objects: 100% (84/84), done.\u001b[K\n",
            "remote: Compressing objects: 100% (59/59), done.\u001b[K\n",
            "remote: Total 2268 (delta 50), reused 44 (delta 25), pack-reused 2184\u001b[K\n",
            "Receiving objects: 100% (2268/2268), 2.63 MiB | 17.18 MiB/s, done.\n",
            "Resolving deltas: 100% (1571/1571), 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) (6.2.2)\n",
            "Requirement already satisfied: numpy<2.0,>=1.14.5 in /usr/local/lib/python3.6/dist-packages (from joeynmt==0.0.1) (1.17.5)\n",
            "Requirement already satisfied: setuptools>=41.0.0 in /usr/local/lib/python3.6/dist-packages (from joeynmt==0.0.1) (45.1.0)\n",
            "Requirement already satisfied: torch>=1.1 in /usr/local/lib/python3.6/dist-packages (from joeynmt==0.0.1) (1.4.0)\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/45/31/1a135b964c169984b27fb2f7a50280fa7f8e6d9d404d8a9e596180487fd1/sacrebleu-1.4.3-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) (3.1.2)\n",
            "Requirement already satisfied: seaborn in /usr/local/lib/python3.6/dist-packages (from joeynmt==0.0.1) (0.9.1)\n",
            "Collecting pyyaml>=5.1\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/3d/d9/ea9816aea31beeadccd03f1f8b625ecf8f645bd66744484d162d84803ce5/PyYAML-5.3.tar.gz (268kB)\n",
            "\u001b[K     |████████████████████████████████| 276kB 8.1MB/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 14.8MB/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: 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: 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: 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: 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: absl-py>=0.7.0 in /usr/local/lib/python3.6/dist-packages (from tensorflow>=1.14->joeynmt==0.0.1) (0.9.0)\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: wheel>=0.26 in /usr/local/lib/python3.6/dist-packages (from tensorflow>=1.14->joeynmt==0.0.1) (0.34.2)\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: 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: 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) (0.8.1)\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: 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: requests in /usr/local/lib/python3.6/dist-packages (from torchtext->joeynmt==0.0.1) (2.21.0)\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: typing in /usr/local/lib/python3.6/dist-packages (from sacrebleu>=1.3.6->joeynmt==0.0.1) (3.6.6)\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: 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: 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: 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) (2.4.6)\n",
            "Requirement already satisfied: scipy>=0.17.1 in /usr/local/lib/python3.6/dist-packages (from seaborn->joeynmt==0.0.1) (1.4.1)\n",
            "Requirement already satisfied: pandas>=0.17.1 in /usr/local/lib/python3.6/dist-packages (from seaborn->joeynmt==0.0.1) (0.25.3)\n",
            "Collecting 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",
            "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 18.1MB/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 7.4MB/s \n",
            "\u001b[?25hRequirement 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: 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) (0.16.1)\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: 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: 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) (2019.11.28)\n",
            "Requirement already satisfied: pytz>=2017.2 in /usr/local/lib/python3.6/dist-packages (from pandas>=0.17.1->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/90/ed/5459080d95eb87a02fe860d447197be63b6e2b5e9ff73c2b0a85622994f4/typed_ast-1.4.1-cp36-cp36m-manylinux1_x86_64.whl (737kB)\n",
            "\u001b[K     |████████████████████████████████| 747kB 19.9MB/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 9.0MB/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=73017 sha256=21d2b5093d74cba0354895c618fae30fc41d9a1a2415f6889ab20ac1f2f0bad6\n",
            "  Stored in directory: /tmp/pip-ephem-wheel-cache-zyrii7fm/wheels/db/01/db/751cc9f3e7f6faec127c43644ba250a3ea7ad200594aeda70a\n",
            "  Building wheel for pyyaml (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "  Created wheel for pyyaml: filename=PyYAML-5.3-cp36-cp36m-linux_x86_64.whl size=44229 sha256=47a30aa84821b32517dc90994753e491af88e658d419f1b672ae53ccae2d7af2\n",
            "  Stored in directory: /root/.cache/pip/wheels/e4/76/4d/a95b8dd7b452b69e8ed4f68b69e1b55e12c9c9624dd962b191\n",
            "Successfully built joeynmt pyyaml\n",
            "Installing collected packages: portalocker, sacrebleu, subword-nmt, pyyaml, mccabe, typed-ast, lazy-object-proxy, astroid, isort, 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-5.3 sacrebleu-1.4.3 subword-nmt-0.3.7 typed-ast-1.4.1\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0Y1lkrt2peIt",
        "colab_type": "code",
        "outputId": "6a42b2d6-cabb-4d85-c6d8-0b9cb712d1a0",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 106
        }
      },
      "source": [
        "! pip install fuzzywuzzy"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "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"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Zt4JPGlNpmmG",
        "colab_type": "code",
        "outputId": "1e32e545-3c68-4154-ef57-94380bf1144a",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 228
        }
      },
      "source": [
        "! pip install python-Levenshtein"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Collecting python-Levenshtein\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/42/a9/d1785c85ebf9b7dfacd08938dd028209c34a0ea3b1bcdb895208bd40a67d/python-Levenshtein-0.12.0.tar.gz (48kB)\n",
            "\r\u001b[K     |██████▊                         | 10kB 31.1MB/s eta 0:00:01\r\u001b[K     |█████████████▌                  | 20kB 2.1MB/s eta 0:00:01\r\u001b[K     |████████████████████▏           | 30kB 2.7MB/s eta 0:00:01\r\u001b[K     |███████████████████████████     | 40kB 2.0MB/s eta 0:00:01\r\u001b[K     |████████████████████████████████| 51kB 2.1MB/s \n",
            "\u001b[?25hRequirement already satisfied: setuptools in /usr/local/lib/python3.6/dist-packages (from python-Levenshtein) (45.1.0)\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=144667 sha256=3c10c6cb5f031cdd3b567ad1ca99ded89d6e20b1928a0e7aca1ac93a8561b2af\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"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "akiy3TCiQgkP",
        "colab_type": "text"
      },
      "source": [
        "## Imports"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Y4YkB1RkQiAv",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from os import path\n",
        "import os\n",
        "import time\n",
        "\n",
        "import pandas as pd\n",
        "import numpy as np\n",
        "from nltk.tokenize import TreebankWordTokenizer\n",
        "from fuzzywuzzy import process"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "61_N3-mYPRY9",
        "colab_type": "text"
      },
      "source": [
        "## Data Gathering"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "EkW-mUdvQ1eY",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "source_language = 'en'\n",
        "target_language = 'kmb'\n",
        "os.environ[\"data_path\"] = path.join(\"joeynmt\", \"data\", source_language + target_language) \n",
        "os.environ[\"src\"] = source_language \n",
        "os.environ[\"tgt\"] = target_language"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "stPP3nXaQmK3",
        "colab_type": "code",
        "outputId": "3c6e86e4-faf3-4960-af3f-f7285d3983d7",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 355
        }
      },
      "source": [
        "# JW300 data\n",
        "! opus_read -d JW300 -s $tgt -t $src -wm moses -w jw300.$tgt jw300.$src -q\n",
        "\n",
        "source = []\n",
        "target = []\n",
        "with open('jw300.' + source_language) as f:\n",
        "  for _, line in enumerate(f):\n",
        "    source.append(line.strip())\n",
        "with open('jw300.' + target_language) as f:\n",
        "  for _, line in enumerate(f):\n",
        "    target.append(line.strip())\n",
        "\n",
        "jw300_raw = []\n",
        "for idx, line in enumerate(source):\n",
        "  if len(line) > 2:\n",
        "    if len(target[idx]) > 2:\n",
        "      jw300_raw.append([line, target[idx]])\n",
        "\n",
        "jw300 = pd.DataFrame(jw300_raw, columns=['source_sentence', 'target_sentence'])\n",
        "jw300.head(3)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "\n",
            "Alignment file /proj/nlpl/data/OPUS/JW300/latest/xml/en-kmb.xml.gz not found. The following files are available for downloading:\n",
            "\n",
            " 920 KB https://object.pouta.csc.fi/OPUS-JW300/v1/xml/en-kmb.xml.gz\n",
            " 263 MB https://object.pouta.csc.fi/OPUS-JW300/v1/xml/en.zip\n",
            "  10 MB https://object.pouta.csc.fi/OPUS-JW300/v1/xml/kmb.zip\n",
            "\n",
            " 274 MB Total size\n",
            "./JW300_latest_xml_en-kmb.xml.gz ... 100% of 920 KB\n",
            "./JW300_latest_xml_en.zip ... 100% of 263 MB\n",
            "./JW300_latest_xml_kmb.zip ... 100% of 10 MB\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "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>Table of Contents</td>\n",
              "      <td>Iala – mu</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>December 1 , 2010</td>\n",
              "      <td>1 Ua Katatu Ua 2011</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>Who Inhabit the Spirit Realm ?</td>\n",
              "      <td>O Kuiala ku Diulu Kuene Muene Athu mu Nzumbi</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "                  source_sentence                               target_sentence\n",
              "0               Table of Contents                                     Iala – mu\n",
              "1               December 1 , 2010                           1 Ua Katatu Ua 2011\n",
              "2  Who Inhabit the Spirit Realm ?  O Kuiala ku Diulu Kuene Muene Athu mu Nzumbi"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 10
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "IUJ6tYVWSXzd",
        "colab_type": "code",
        "outputId": "ca3bd0d1-0b2a-406f-ec92-ec802f8c888b",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 526
        }
      },
      "source": [
        "# Common test data\n",
        "source_test_file = 'test.en-' + target_language + '.en'\n",
        "target_test_file = 'test.en-' + target_language + '.' + target_language\n",
        "\n",
        "! wget https://raw.githubusercontent.com/jaderabbit/masakhane/master/jw300_utils/test/test.en-$tgt.en\n",
        "! wget https://raw.githubusercontent.com/jaderabbit/masakhane/master/jw300_utils/test/test.en-$tgt.$tgt\n",
        "\n",
        "source = []\n",
        "target = []\n",
        "with open(source_test_file) as f:\n",
        "  for _, line in enumerate(f):\n",
        "    source.append(line.strip())\n",
        "with open(target_test_file) as f:\n",
        "  for _, line in enumerate(f):\n",
        "    target.append(line.strip())\n",
        "\n",
        "! rm test.en-$tgt.en\n",
        "! rm test.en-$tgt.$tgt\n",
        "\n",
        "test_raw = []\n",
        "for idx, line in enumerate(source):\n",
        "  if len(line) > 2:\n",
        "    if len(target[idx]) > 2:\n",
        "      test_raw.append([line, target[idx]])\n",
        "\n",
        "df_test = pd.DataFrame(test_raw, columns=['source_sentence', 'target_sentence'])\n",
        "df_test.head(3)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--2020-02-04 19:04:17--  https://raw.githubusercontent.com/jaderabbit/masakhane/master/jw300_utils/test/test.en-kmb.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: 204945 (200K) [text/plain]\n",
            "Saving to: ‘test.en-kmb.en’\n",
            "\n",
            "\rtest.en-kmb.en        0%[                    ]       0  --.-KB/s               \rtest.en-kmb.en      100%[===================>] 200.14K  --.-KB/s    in 0.03s   \n",
            "\n",
            "2020-02-04 19:04:17 (6.53 MB/s) - ‘test.en-kmb.en’ saved [204945/204945]\n",
            "\n",
            "--2020-02-04 19:04:18--  https://raw.githubusercontent.com/jaderabbit/masakhane/master/jw300_utils/test/test.en-kmb.kmb\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: 230668 (225K) [text/plain]\n",
            "Saving to: ‘test.en-kmb.kmb’\n",
            "\n",
            "test.en-kmb.kmb     100%[===================>] 225.26K  --.-KB/s    in 0.04s   \n",
            "\n",
            "2020-02-04 19:04:18 (5.35 MB/s) - ‘test.en-kmb.kmb’ saved [230668/230668]\n",
            "\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "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>Dorcas “ abounded in good deeds and gifts of m...</td>\n",
              "      <td>Dorka , “ [ uavudile ] jimbote ni jimola [ ja ...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>What will be considered in this article , and ...</td>\n",
              "      <td>Ihi i tua - nda di longa ku mbandu íii , ni mu...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>Some names in this article have been changed .</td>\n",
              "      <td>Saí majina a a lungulula .</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "                                     source_sentence                                    target_sentence\n",
              "0  Dorcas “ abounded in good deeds and gifts of m...  Dorka , “ [ uavudile ] jimbote ni jimola [ ja ...\n",
              "1  What will be considered in this article , and ...  Ihi i tua - nda di longa ku mbandu íii , ni mu...\n",
              "2     Some names in this article have been changed .                         Saí majina a a lungulula ."
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 11
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "caJcH-I5PL19",
        "colab_type": "text"
      },
      "source": [
        "## Word Alignments for Corpus Filtering"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "outputId": "70061918-b183-417d-a1ca-4448e55277a0",
        "id": "ISn3UWvPPjkI",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 104
        }
      },
      "source": [
        "# combine Bible + JW300 data for fast_align\n",
        "common = pd.concat([jw300])\n",
        "common['combined'] = common['source_sentence'] + ' ||| ' + common['target_sentence']\n",
        "common['combined'].values.tolist()[0:5]"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "['Table of Contents ||| Iala – mu',\n",
              " 'December 1 , 2010 ||| 1 Ua Katatu Ua 2011',\n",
              " 'Who Inhabit the Spirit Realm ? ||| O Kuiala ku Diulu Kuene Muene Athu mu Nzumbi',\n",
              " 'FROM OUR COVER ||| TU SANGA - MU UÉ MILONGI ÍII',\n",
              " '3 Someone Is Out There \\u200b — But Who ? ||| 3 Kuene Athu mu Nzumbi \\u200b — a Nanhi ?']"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 12
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "gPkjmeHweCtd",
        "colab_type": "code",
        "outputId": "b4099aee-2db8-49bb-b763-b2ff8e23d839",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 211
        }
      },
      "source": [
        "# Output to a file\n",
        "with open(\"word_align_file.txt\", \"w\") as wa_file:\n",
        "  for sample in common['combined'].values.tolist():\n",
        "    wa_file.write(sample+\"\\n\")\n",
        "\n",
        "! head word_align_file.txt"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Table of Contents ||| Iala – mu\n",
            "December 1 , 2010 ||| 1 Ua Katatu Ua 2011\n",
            "Who Inhabit the Spirit Realm ? ||| O Kuiala ku Diulu Kuene Muene Athu mu Nzumbi\n",
            "FROM OUR COVER ||| TU SANGA - MU UÉ MILONGI ÍII\n",
            "3 Someone Is Out There ​ — But Who ? ||| 3 Kuene Athu mu Nzumbi ​ — a Nanhi ?\n",
            "4 Visions of the Spirit Realm ||| 4 Isuma ia Athu a Tungu mu Nzumbi\n",
            "7 Contact With the Spirit Realm ||| 7 Tu Tena Kuzuela ni Athu a Tungu mu Nzumbi ?\n",
            "REGULAR FEATURES ||| TUA - NDA DI LONGA UÉ\n",
            "10 Did You Know ? ||| 10 Atangi a Madivulu Metu Ebhula . . .\n",
            "11 Draw Close to God ​ — He Knows “ the Heart of the Sons of Mankind ” ||| MILONGI PHALA KU DI LONGA 28 ia Kauana katé ku 6 ia Katanu MBANDU IA 11 - MIMBU : 49 , 74\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "r7TtC6ojfogR",
        "colab_type": "code",
        "outputId": "cb1df94b-3d65-4a19-c3d8-b088a528b953",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        }
      },
      "source": [
        "# Generate word alignments\n",
        "! ./fast_align/build/fast_align -i word_align_file.txt -d -o -v -s > forward.align"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "ARG=i\n",
            "ARG=d\n",
            "ARG=o\n",
            "ARG=v\n",
            "ARG=s\n",
            "INITIAL PASS \n",
            ".................................................. [50000]\n",
            "...............................................\n",
            "expected target length = source length * 1.25735\n",
            "ITERATION 1\n",
            ".................................................. [50000]\n",
            "...............................................\n",
            "  log_e likelihood: -4.17171e+07\n",
            "  log_2 likelihood: -6.0185e+07\n",
            "     cross entropy: 29.8974\n",
            "        perplexity: 1e+09\n",
            "      posterior p0: 0.08\n",
            " posterior al-feat: -0.16963\n",
            "       size counts: 2676\n",
            "ITERATION 2\n",
            ".................................................. [50000]\n",
            "...............................................\n",
            "  log_e likelihood: -1.0431e+07\n",
            "  log_2 likelihood: -1.50487e+07\n",
            "     cross entropy: 7.47556\n",
            "        perplexity: 177.979\n",
            "      posterior p0: 0.0599725\n",
            " posterior al-feat: -0.138555\n",
            "       size counts: 2676\n",
            "  1  model al-feat: -0.139253 (tension=4)\n",
            "  2  model al-feat: -0.138985 (tension=4.01396)\n",
            "  3  model al-feat: -0.13882 (tension=4.02256)\n",
            "  4  model al-feat: -0.138719 (tension=4.02786)\n",
            "  5  model al-feat: -0.138656 (tension=4.03114)\n",
            "  6  model al-feat: -0.138618 (tension=4.03316)\n",
            "  7  model al-feat: -0.138594 (tension=4.03441)\n",
            "  8  model al-feat: -0.138579 (tension=4.03518)\n",
            "     final tension: 4.03566\n",
            "ITERATION 3\n",
            ".................................................. [50000]\n",
            "...............................................\n",
            "  log_e likelihood: -8.88493e+06\n",
            "  log_2 likelihood: -1.28182e+07\n",
            "     cross entropy: 6.36756\n",
            "        perplexity: 82.5707\n",
            "      posterior p0: 0.0555787\n",
            " posterior al-feat: -0.131422\n",
            "       size counts: 2676\n",
            "  1  model al-feat: -0.13857 (tension=4.03566)\n",
            "  2  model al-feat: -0.135876 (tension=4.17861)\n",
            "  3  model al-feat: -0.134234 (tension=4.26769)\n",
            "  4  model al-feat: -0.133212 (tension=4.32393)\n",
            "  5  model al-feat: -0.132567 (tension=4.35974)\n",
            "  6  model al-feat: -0.132157 (tension=4.38264)\n",
            "  7  model al-feat: -0.131895 (tension=4.39734)\n",
            "  8  model al-feat: -0.131727 (tension=4.4068)\n",
            "     final tension: 4.41288\n",
            "ITERATION 4\n",
            ".................................................. [50000]\n",
            "...............................................\n",
            "  log_e likelihood: -8.52964e+06\n",
            "  log_2 likelihood: -1.23057e+07\n",
            "     cross entropy: 6.11293\n",
            "        perplexity: 69.2111\n",
            "      posterior p0: 0.058605\n",
            " posterior al-feat: -0.12528\n",
            "       size counts: 2676\n",
            "  1  model al-feat: -0.131618 (tension=4.41288)\n",
            "  2  model al-feat: -0.129394 (tension=4.53964)\n",
            "  3  model al-feat: -0.12798 (tension=4.62192)\n",
            "  4  model al-feat: -0.127065 (tension=4.67592)\n",
            "  5  model al-feat: -0.126465 (tension=4.7116)\n",
            "  6  model al-feat: -0.126069 (tension=4.73529)\n",
            "  7  model al-feat: -0.125807 (tension=4.75106)\n",
            "  8  model al-feat: -0.125632 (tension=4.76158)\n",
            "     final tension: 4.76861\n",
            "ITERATION 5 (FINAL)\n",
            ".................................................. [50000]\n",
            "...............................................\n",
            "  log_e likelihood: -8.39616e+06\n",
            "  log_2 likelihood: -1.21131e+07\n",
            "     cross entropy: 6.01727\n",
            "        perplexity: 64.7709\n",
            "      posterior p0: 0\n",
            " posterior al-feat: 0\n",
            "       size counts: 2676\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "_YFtSFy2gHgn",
        "colab_type": "text"
      },
      "source": [
        "## Corpus Filtering"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4qshK1wnggKM",
        "colab_type": "code",
        "outputId": "e5279e1d-de8a-46e4-ee5e-c739258dbb25",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 258
        }
      },
      "source": [
        "# add word alignment scores into the common dataframe\n",
        "scores = []\n",
        "with open('forward.align') as f:\n",
        "  for _, line in enumerate(f):\n",
        "    scores.append(float(line.split(' ||| ')[-1]))\n",
        "\n",
        "common['scores'] = scores\n",
        "common.head()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "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",
              "      <th>combined</th>\n",
              "      <th>scores</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>Table of Contents</td>\n",
              "      <td>Iala – mu</td>\n",
              "      <td>Table of Contents ||| Iala – mu</td>\n",
              "      <td>-6.44585</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>December 1 , 2010</td>\n",
              "      <td>1 Ua Katatu Ua 2011</td>\n",
              "      <td>December 1 , 2010 ||| 1 Ua Katatu Ua 2011</td>\n",
              "      <td>-26.63540</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>Who Inhabit the Spirit Realm ?</td>\n",
              "      <td>O Kuiala ku Diulu Kuene Muene Athu mu Nzumbi</td>\n",
              "      <td>Who Inhabit the Spirit Realm ? ||| O Kuiala ku...</td>\n",
              "      <td>-32.64520</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>FROM OUR COVER</td>\n",
              "      <td>TU SANGA - MU UÉ MILONGI ÍII</td>\n",
              "      <td>FROM OUR COVER ||| TU SANGA - MU UÉ MILONGI ÍII</td>\n",
              "      <td>-17.60530</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>3 Someone Is Out There ​ — But Who ?</td>\n",
              "      <td>3 Kuene Athu mu Nzumbi ​ — a Nanhi ?</td>\n",
              "      <td>3 Someone Is Out There ​ — But Who ? ||| 3 Kue...</td>\n",
              "      <td>-34.61690</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "                        source_sentence  ...    scores\n",
              "0                     Table of Contents  ...  -6.44585\n",
              "1                     December 1 , 2010  ... -26.63540\n",
              "2        Who Inhabit the Spirit Realm ?  ... -32.64520\n",
              "3                        FROM OUR COVER  ... -17.60530\n",
              "4  3 Someone Is Out There ​ — But Who ?  ... -34.61690\n",
              "\n",
              "[5 rows x 4 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 15
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jhWPxbK0kyBP",
        "colab_type": "code",
        "outputId": "cfb2c28e-57b2-489c-ec1e-545f3ffe3edd",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 300
        }
      },
      "source": [
        "common.describe()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "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>scores</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>count</th>\n",
              "      <td>97218.000000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>mean</th>\n",
              "      <td>-89.522981</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>std</th>\n",
              "      <td>59.855089</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>min</th>\n",
              "      <td>-1106.860000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>25%</th>\n",
              "      <td>-119.555000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>50%</th>\n",
              "      <td>-79.433800</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>75%</th>\n",
              "      <td>-47.867900</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>max</th>\n",
              "      <td>-1.122730</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "             scores\n",
              "count  97218.000000\n",
              "mean     -89.522981\n",
              "std       59.855089\n",
              "min    -1106.860000\n",
              "25%     -119.555000\n",
              "50%      -79.433800\n",
              "75%      -47.867900\n",
              "max       -1.122730"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 16
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "PB9t9Zt7k9Ps",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# cut out anything below the 0.1 quantile (really bad)\n",
        "threshold = common.quantile(0.1, axis=0)['scores']\n",
        "common_clean = common[common['scores'] > threshold]"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_PF8Z7UylbsW",
        "colab_type": "code",
        "outputId": "64f7e6cf-74f5-4782-af2a-358981d9e24b",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "# how many did we lose?\n",
        "len(common_clean)/len(common)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.8999979427678002"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 18
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "WvH3ugvylv7v",
        "colab_type": "text"
      },
      "source": [
        "## Other pre-processing"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "LK1AiflclpdM",
        "colab_type": "code",
        "outputId": "d54bf3bb-3f7c-45d4-f2ef-947aae6d3188",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "# drop test data from common\n",
        "df_pp = common_clean[~common_clean['source_sentence'].isin(df_test['source_sentence'].values)]\n",
        "df_pp = df_pp[~df_pp['target_sentence'].isin(df_test['target_sentence'].values)]\n",
        "\n",
        "# remove duplicates\n",
        "df_pp.drop_duplicates(inplace=True)\n",
        "\n",
        "# remove conflicting translations\n",
        "df_pp.drop_duplicates(subset='source_sentence', inplace=True)\n",
        "df_pp.drop_duplicates(subset='target_sentence', inplace=True)\n",
        "\n",
        "# what's left in terms of number of samples?\n",
        "len(df_pp)/len(common)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.8135530457322718"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 19
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "iGYR610PxdGK",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# reset the index of the training set after filtering\n",
        "df_pp.reset_index(drop=False, inplace=True)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_IuSBECCoDsW",
        "colab_type": "code",
        "outputId": "beca0972-aeb1-484d-f5b7-dd4218fc7be2",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        }
      },
      "source": [
        "# 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",
        "  #candidates = [x for x in candidates if 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",
        "eng_test = df_test['source_sentence'].values.tolist()\n",
        "scores = []\n",
        "start_time = time.time()\n",
        "for idx, row in df_pp.iterrows():\n",
        "  scores.append(fuzzfilter(row['source_sentence'], eng_test, 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))))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "00:00:00.02 0.00 percent complete\n",
            "00:00:15.99 1.26 percent complete\n",
            "00:00:31.60 2.53 percent complete\n",
            "00:00:46.84 3.79 percent complete\n",
            "00:01:02.81 5.06 percent complete\n",
            "00:01:18.16 6.32 percent complete\n",
            "00:01:33.57 7.59 percent complete\n",
            "00:01:48.58 8.85 percent complete\n",
            "00:02:04.89 10.11 percent complete\n",
            "00:02:21.35 11.38 percent complete\n",
            "00:02:36.99 12.64 percent complete\n",
            "00:02:52.40 13.91 percent complete\n",
            "00:03:07.66 15.17 percent complete\n",
            "00:03:23.21 16.44 percent complete\n",
            "00:03:38.89 17.70 percent complete\n",
            "00:03:54.36 18.97 percent complete\n",
            "00:04:10.44 20.23 percent complete\n",
            "00:04:26.39 21.49 percent complete\n",
            "00:04:43.23 22.76 percent complete\n",
            "00:04:59.46 24.02 percent complete\n",
            "00:05:16.05 25.29 percent complete\n",
            "00:05:31.91 26.55 percent complete\n",
            "00:05:48.36 27.82 percent complete\n",
            "00:06:04.70 29.08 percent complete\n",
            "00:06:21.21 30.34 percent complete\n",
            "00:06:37.73 31.61 percent complete\n",
            "00:06:53.93 32.87 percent complete\n",
            "00:07:10.20 34.14 percent complete\n",
            "00:07:26.56 35.40 percent complete\n",
            "00:07:42.11 36.67 percent complete\n",
            "00:07:58.07 37.93 percent complete\n",
            "00:08:14.06 39.19 percent complete\n",
            "00:08:29.90 40.46 percent complete\n",
            "00:08:46.04 41.72 percent complete\n",
            "00:09:02.80 42.99 percent complete\n",
            "00:09:19.59 44.25 percent complete\n",
            "00:09:35.92 45.52 percent complete\n",
            "00:09:51.77 46.78 percent complete\n",
            "00:10:08.14 48.05 percent complete\n",
            "00:10:24.54 49.31 percent complete\n",
            "00:10:40.48 50.57 percent complete\n",
            "00:10:56.29 51.84 percent complete\n",
            "00:11:12.00 53.10 percent complete\n",
            "00:11:29.30 54.37 percent complete\n",
            "00:11:46.04 55.63 percent complete\n",
            "00:12:02.19 56.90 percent complete\n",
            "00:12:17.97 58.16 percent complete\n",
            "00:12:35.13 59.42 percent complete\n",
            "00:12:50.94 60.69 percent complete\n",
            "00:13:06.72 61.95 percent complete\n",
            "00:13:23.85 63.22 percent complete\n",
            "00:13:42.28 64.48 percent complete\n",
            "00:13:59.85 65.75 percent complete\n",
            "00:14:17.12 67.01 percent complete\n",
            "00:14:34.22 68.27 percent complete\n",
            "00:14:51.67 69.54 percent complete\n",
            "00:15:08.71 70.80 percent complete\n",
            "00:15:25.66 72.07 percent complete\n",
            "00:15:43.60 73.33 percent complete\n",
            "00:16:00.66 74.60 percent complete\n",
            "00:16:17.58 75.86 percent complete\n",
            "00:16:34.54 77.13 percent complete\n",
            "00:16:51.25 78.39 percent complete\n",
            "00:17:08.11 79.65 percent complete\n",
            "00:17:25.00 80.92 percent complete\n",
            "00:17:41.94 82.18 percent complete\n",
            "00:17:58.92 83.45 percent complete\n",
            "00:18:17.12 84.71 percent complete\n",
            "00:18:34.94 85.98 percent complete\n",
            "00:18:52.80 87.24 percent complete\n",
            "00:19:09.40 88.50 percent complete\n",
            "00:19:26.94 89.77 percent complete\n",
            "00:19:43.81 91.03 percent complete\n",
            "00:20:01.44 92.30 percent complete\n",
            "00:20:18.72 93.56 percent complete\n",
            "00:20:35.76 94.83 percent complete\n",
            "00:20:53.76 96.09 percent complete\n",
            "00:21:11.23 97.35 percent complete\n",
            "00:21:28.30 98.62 percent complete\n",
            "00:21:45.93 99.88 percent complete\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "uBgUuWVX8V13",
        "colab_type": "code",
        "outputId": "c65da7a6-f253-4389-b3fb-9c9bcae047c7",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "# Filter out \"almost overlapping samples\"\n",
        "df_pp['scores'] = scores\n",
        "df_pp = df_pp[df_pp['scores'] < 95]\n",
        "\n",
        "# what's left in terms of number of samples?\n",
        "len(df_pp)/len(common)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.7892982780966488"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 22
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RdHgU5xrmOKz",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "## Lower case the corpus\n",
        "df_pp[\"source_sentence\"] = df_pp[\"source_sentence\"].str.lower()\n",
        "df_pp[\"target_sentence\"] = df_pp[\"target_sentence\"].str.lower()\n",
        "df_test[\"source_sentence\"] = df_test[\"source_sentence\"].str.lower()\n",
        "df_test[\"target_sentence\"] = df_test[\"target_sentence\"].str.lower()\n",
        "\n",
        "# shuffle the training/dev data\n",
        "df_pp = df_pp.sample(frac=1).reset_index(drop=True)\n",
        "\n",
        "# Do the split between dev/train\n",
        "num_dev_patterns = 1000\n",
        "dev = df_pp.tail(num_dev_patterns)\n",
        "stripped = df_pp.drop(df_pp.tail(num_dev_patterns).index)\n",
        "\n",
        "# output the final parallel corpus files\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",
        "with open(\"test.\"+source_language, \"w\") as src_file, open(\"test.\"+target_language, \"w\") as trg_file:\n",
        "  for index, row in df_test.iterrows():\n",
        "    src_file.write(row[\"source_sentence\"]+\"\\n\")\n",
        "    trg_file.write(row[\"target_sentence\"]+\"\\n\")"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "WugPZnZ_mv5H",
        "colab_type": "code",
        "outputId": "d126a0ce-ec38-4751-c459-c2569fb052e0",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 211
        }
      },
      "source": [
        "! head train.en"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "some may feel that they do not need anyone to explain the bible to them .\n",
            "most of us have very busy lives , but we should not let anything , including our responsibilities , stop us from reading the bible .\n",
            "giving may also lower stress and blood pressure .\n",
            "jehovah created all things\n",
            "but what can be done if the marital bond is strained ?\n",
            "( b ) explain how it became clear that the governing body was different from the watch tower society .\n",
            "this issue of the watchtower examines the bible’s claim that it can guide us in every aspect of life .\n",
            "then you throw them out onto the ground all at once .\n",
            "having no arms , i can fully empathize with those who have limitations .\n",
            "but even in such a situation , a wife will do what she can to teach her children the truth .\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "N9lWcwN2m0O2",
        "colab_type": "code",
        "outputId": "1e9a1f52-de8f-4ce1-e00e-f731d1c1c121",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 211
        }
      },
      "source": [
        "! head train.kmb"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "nange sai athu a banza kuila , ka bhingi muthu phala ku a jimbuluila o milongi ia bibidia .\n",
            "( josuué 1 : 8 ) tuala ni ikalakalu iavulu , kidi muene , maji ki tua tokala kuehela kima ku tu fidisa kutanga o bibidia , né muene o salu ietu .\n",
            "o kubhana ku tu kuatekesa ue kusosolola o kuthandanganha , ni kulenga kua manhinga mu mixibha ietu .\n",
            "mukonda jihova , muéne ua bhange o ima ioso\n",
            "maji ihi ia tokala o ku bhanga se mu ukaza mu moneka maka ?\n",
            "( b ) jimbulula kiebhi o jiphange kiéza mu kuijiia kuila , o kibuka kia utuminu ki ki lungile ni sociedade torre de vigia .\n",
            "o kadivulu kaka o mulangidi , ka - nda zuela se kiebhi o milongi ia bibidia i tena ku tu kuatekesa ku muenhu uetu lelu .\n",
            "eie u ji lundulula joso bhoxi .\n",
            "mukonda dia kukamba o maku , ngi tena kuivua o ndolo ivua ió ala uá ni unema .\n",
            "né muene mu ithangana kála iii , o muhatu ua bhingi kubhanga ioso , phala kulonga o itumu ia jihova ku tuana tuê .\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "eNbaZnhznC8-",
        "colab_type": "text"
      },
      "source": [
        "## Subword BPE Tokens"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "PlSQH_bQm82c",
        "colab_type": "code",
        "outputId": "4dfde776-d0e2-4208-827b-75faac7f5740",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 384
        }
      },
      "source": [
        "# Do BPE\n",
        "! subword-nmt learn-joint-bpe-and-vocab --input train.$src train.$tgt -s 4000 -o bpe.codes.4000 --write-vocabulary vocab.$src vocab.$tgt\n",
        "\n",
        "! subword-nmt apply-bpe -c bpe.codes.4000 --vocabulary vocab.$src < train.$src > train.bpe.$src\n",
        "! subword-nmt apply-bpe -c bpe.codes.4000 --vocabulary vocab.$tgt < train.$tgt > train.bpe.$tgt\n",
        "! subword-nmt apply-bpe -c bpe.codes.4000 --vocabulary vocab.$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 $data_path\n",
        "\n",
        "# Create that vocab using build_vocab\n",
        "! sudo chmod 777 joeynmt/scripts/build_vocab.py\n",
        "! joeynmt/scripts/build_vocab.py joeynmt/data/$src$tgt/train.bpe.$src joeynmt/data/$src$tgt/train.bpe.$tgt --output_path joeynmt/data/$src$tgt/vocab.txt\n",
        "\n",
        "# Some output\n",
        "! echo \"BPE Target Sentences\"\n",
        "! tail -n 5 test.bpe.$tgt\n",
        "! echo \"Combined BPE Vocab\"\n",
        "! tail -n 10 joeynmt/data/en$tgt/vocab.txt"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "bpe.codes.4000\tdev.en\t     test.bpe.kmb  train.bpe.en   train.kmb\n",
            "dev.bpe.en\tdev.kmb      test.en\t   train.bpe.kmb\n",
            "dev.bpe.kmb\ttest.bpe.en  test.kmb\t   train.en\n",
            "BPE Target Sentences\n",
            "o ngu@@ b@@ u ya ku@@ xikana ( tala o kaxi 12 - 14 )\n",
            "o ka@@ pas@@ ete ka kubh@@ uluka ( tala o kaxi 15 - 18 )\n",
            "nga mono kwila o athu a xikina dingi se a mona kwila ey@@ e wa zolo mwene o bibidya , wa mu bhanga yoso i u tena phala ku a kwatekesa . ”\n",
            "o xi@@ bhata ya nzumbi ikôla ( tala o kaxi 19 - 20 )\n",
            "ni ki@@ kwat@@ ek@@ esu kya jihova tu tena kubh@@ ânga nê !\n",
            "Combined BPE Vocab\n",
            "urrec@@\n",
            "pher@@\n",
            "danii@@\n",
            "espe@@\n",
            "enhu\n",
            "ould\n",
            "beh@@\n",
            "paradi@@\n",
            "effor@@\n",
            "kobo\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "mox2nEPXnnOe",
        "colab_type": "text"
      },
      "source": [
        "## JoeyNMT Config"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ugHbYjQPnNo4",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# This creates the config file for our JoeyNMT system. \n",
        "name = '%s%s' % (source_language, target_language)\n",
        "\n",
        "config = \"\"\"\n",
        "name: \"{name}_transformer\"\n",
        "\n",
        "data:\n",
        "    src: \"{source_language}\"\n",
        "    trg: \"{target_language}\"\n",
        "    train: \"data/{name}/train.bpe\"\n",
        "    dev:   \"data/{name}/dev.bpe\"\n",
        "    test:  \"data/{name}/test.bpe\"\n",
        "    level: \"bpe\"\n",
        "    lowercase: False\n",
        "    max_sent_length: 100\n",
        "    src_vocab: \"data/{name}/vocab.txt\"\n",
        "    trg_vocab: \"data/{name}/vocab.txt\"\n",
        "\n",
        "testing:\n",
        "    beam_size: 5\n",
        "    alpha: 1.0\n",
        "\n",
        "training:\n",
        "    #load_model: \"models/{name}_transformer/12000.ckpt\" # if given, 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: \"noam\"            # Try switching from plateau to Noam scheduling\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",
        "    patience: 8\n",
        "    decrease_factor: 0.7\n",
        "    loss: \"crossentropy\"\n",
        "    learning_rate: 0.0002\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: \"eval_metric\" # \"ppl\"\n",
        "    epochs: 40\n",
        "    validation_freq: 2000\n",
        "    logging_freq: 200\n",
        "    eval_metric: \"bleu\"\n",
        "    model_dir: \"models/{name}_transformer\"\n",
        "    overwrite: True\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: 8\n",
        "        embeddings:\n",
        "            embedding_dim: 512\n",
        "            scale: True\n",
        "            dropout: 0.\n",
        "        # typically ff_size = 4 x hidden_size\n",
        "        hidden_size: 512\n",
        "        ff_size: 2048\n",
        "        dropout: 0.3\n",
        "    decoder:\n",
        "        type: \"transformer\"\n",
        "        num_layers: 6\n",
        "        num_heads: 8\n",
        "        embeddings:\n",
        "            embedding_dim: 512\n",
        "            scale: True\n",
        "            dropout: 0.\n",
        "        # typically ff_size = 4 x hidden_size\n",
        "        hidden_size: 512\n",
        "        ff_size: 2048\n",
        "        dropout: 0.3\n",
        "\"\"\".format(name=name, 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)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "eNQ_9LO4n04W",
        "colab_type": "text"
      },
      "source": [
        "## Train the model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hWwNavEHnxs8",
        "colab_type": "code",
        "outputId": "8e3ac93f-5914-49f1-afd9-5b452e19b4e8",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        }
      },
      "source": [
        "!cd joeynmt; python3 -m joeynmt train configs/transformer_$src$tgt.yaml"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "2020-02-04 19:34:52,715 Hello! This is Joey-NMT.\n",
            "2020-02-04 19:34:54,040 Total params: 46249984\n",
            "2020-02-04 19:34:54,042 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",
            "2020-02-04 19:35:03,901 cfg.name                           : enkmb_transformer\n",
            "2020-02-04 19:35:03,901 cfg.data.src                       : en\n",
            "2020-02-04 19:35:03,901 cfg.data.trg                       : kmb\n",
            "2020-02-04 19:35:03,901 cfg.data.train                     : data/enkmb/train.bpe\n",
            "2020-02-04 19:35:03,901 cfg.data.dev                       : data/enkmb/dev.bpe\n",
            "2020-02-04 19:35:03,901 cfg.data.test                      : data/enkmb/test.bpe\n",
            "2020-02-04 19:35:03,901 cfg.data.level                     : bpe\n",
            "2020-02-04 19:35:03,901 cfg.data.lowercase                 : False\n",
            "2020-02-04 19:35:03,901 cfg.data.max_sent_length           : 100\n",
            "2020-02-04 19:35:03,902 cfg.data.src_vocab                 : data/enkmb/vocab.txt\n",
            "2020-02-04 19:35:03,902 cfg.data.trg_vocab                 : data/enkmb/vocab.txt\n",
            "2020-02-04 19:35:03,902 cfg.testing.beam_size              : 5\n",
            "2020-02-04 19:35:03,902 cfg.testing.alpha                  : 1.0\n",
            "2020-02-04 19:35:03,902 cfg.training.random_seed           : 42\n",
            "2020-02-04 19:35:03,902 cfg.training.optimizer             : adam\n",
            "2020-02-04 19:35:03,902 cfg.training.normalization         : tokens\n",
            "2020-02-04 19:35:03,902 cfg.training.adam_betas            : [0.9, 0.999]\n",
            "2020-02-04 19:35:03,902 cfg.training.scheduling            : noam\n",
            "2020-02-04 19:35:03,902 cfg.training.learning_rate_factor  : 0.5\n",
            "2020-02-04 19:35:03,902 cfg.training.learning_rate_warmup  : 1000\n",
            "2020-02-04 19:35:03,902 cfg.training.patience              : 8\n",
            "2020-02-04 19:35:03,902 cfg.training.decrease_factor       : 0.7\n",
            "2020-02-04 19:35:03,902 cfg.training.loss                  : crossentropy\n",
            "2020-02-04 19:35:03,902 cfg.training.learning_rate         : 0.0002\n",
            "2020-02-04 19:35:03,902 cfg.training.learning_rate_min     : 1e-08\n",
            "2020-02-04 19:35:03,902 cfg.training.weight_decay          : 0.0\n",
            "2020-02-04 19:35:03,902 cfg.training.label_smoothing       : 0.1\n",
            "2020-02-04 19:35:03,902 cfg.training.batch_size            : 4096\n",
            "2020-02-04 19:35:03,902 cfg.training.batch_type            : token\n",
            "2020-02-04 19:35:03,902 cfg.training.eval_batch_size       : 3600\n",
            "2020-02-04 19:35:03,902 cfg.training.eval_batch_type       : token\n",
            "2020-02-04 19:35:03,902 cfg.training.batch_multiplier      : 1\n",
            "2020-02-04 19:35:03,902 cfg.training.early_stopping_metric : eval_metric\n",
            "2020-02-04 19:35:03,902 cfg.training.epochs                : 40\n",
            "2020-02-04 19:35:03,902 cfg.training.validation_freq       : 2000\n",
            "2020-02-04 19:35:03,903 cfg.training.logging_freq          : 200\n",
            "2020-02-04 19:35:03,903 cfg.training.eval_metric           : bleu\n",
            "2020-02-04 19:35:03,903 cfg.training.model_dir             : models/enkmb_transformer\n",
            "2020-02-04 19:35:03,903 cfg.training.overwrite             : True\n",
            "2020-02-04 19:35:03,903 cfg.training.shuffle               : True\n",
            "2020-02-04 19:35:03,903 cfg.training.use_cuda              : True\n",
            "2020-02-04 19:35:03,903 cfg.training.max_output_length     : 100\n",
            "2020-02-04 19:35:03,903 cfg.training.print_valid_sents     : [0, 1, 2, 3]\n",
            "2020-02-04 19:35:03,903 cfg.training.keep_last_ckpts       : 3\n",
            "2020-02-04 19:35:03,903 cfg.model.initializer              : xavier\n",
            "2020-02-04 19:35:03,903 cfg.model.bias_initializer         : zeros\n",
            "2020-02-04 19:35:03,903 cfg.model.init_gain                : 1.0\n",
            "2020-02-04 19:35:03,903 cfg.model.embed_initializer        : xavier\n",
            "2020-02-04 19:35:03,903 cfg.model.embed_init_gain          : 1.0\n",
            "2020-02-04 19:35:03,903 cfg.model.tied_embeddings          : True\n",
            "2020-02-04 19:35:03,903 cfg.model.tied_softmax             : True\n",
            "2020-02-04 19:35:03,903 cfg.model.encoder.type             : transformer\n",
            "2020-02-04 19:35:03,903 cfg.model.encoder.num_layers       : 6\n",
            "2020-02-04 19:35:03,903 cfg.model.encoder.num_heads        : 8\n",
            "2020-02-04 19:35:03,903 cfg.model.encoder.embeddings.embedding_dim : 512\n",
            "2020-02-04 19:35:03,903 cfg.model.encoder.embeddings.scale : True\n",
            "2020-02-04 19:35:03,903 cfg.model.encoder.embeddings.dropout : 0.0\n",
            "2020-02-04 19:35:03,903 cfg.model.encoder.hidden_size      : 512\n",
            "2020-02-04 19:35:03,903 cfg.model.encoder.ff_size          : 2048\n",
            "2020-02-04 19:35:03,903 cfg.model.encoder.dropout          : 0.3\n",
            "2020-02-04 19:35:03,903 cfg.model.decoder.type             : transformer\n",
            "2020-02-04 19:35:03,903 cfg.model.decoder.num_layers       : 6\n",
            "2020-02-04 19:35:03,903 cfg.model.decoder.num_heads        : 8\n",
            "2020-02-04 19:35:03,903 cfg.model.decoder.embeddings.embedding_dim : 512\n",
            "2020-02-04 19:35:03,904 cfg.model.decoder.embeddings.scale : True\n",
            "2020-02-04 19:35:03,904 cfg.model.decoder.embeddings.dropout : 0.0\n",
            "2020-02-04 19:35:03,904 cfg.model.decoder.hidden_size      : 512\n",
            "2020-02-04 19:35:03,904 cfg.model.decoder.ff_size          : 2048\n",
            "2020-02-04 19:35:03,904 cfg.model.decoder.dropout          : 0.3\n",
            "2020-02-04 19:35:03,904 Data set sizes: \n",
            "\ttrain 75734,\n",
            "\tvalid 1000,\n",
            "\ttest 2693\n",
            "2020-02-04 19:35:03,904 First training example:\n",
            "\t[SRC] some may feel that they do not need anyone to explain the bible to them .\n",
            "\t[TRG] nange sai athu a banza kuila , ka bhingi muthu phala ku a jimb@@ ulu@@ ila o milongi ia bibidia .\n",
            "2020-02-04 19:35:03,904 First 10 words (src): (0) <unk> (1) <pad> (2) <s> (3) </s> (4) , (5) . (6) o (7) ku (8) mu (9) a\n",
            "2020-02-04 19:35:03,904 First 10 words (trg): (0) <unk> (1) <pad> (2) <s> (3) </s> (4) , (5) . (6) o (7) ku (8) mu (9) a\n",
            "2020-02-04 19:35:03,904 Number of Src words (types): 4120\n",
            "2020-02-04 19:35:03,905 Number of Trg words (types): 4120\n",
            "2020-02-04 19:35:03,905 Model(\n",
            "\tencoder=TransformerEncoder(num_layers=6, num_heads=8),\n",
            "\tdecoder=TransformerDecoder(num_layers=6, num_heads=8),\n",
            "\tsrc_embed=Embeddings(embedding_dim=512, vocab_size=4120),\n",
            "\ttrg_embed=Embeddings(embedding_dim=512, vocab_size=4120))\n",
            "2020-02-04 19:35:03,908 EPOCH 1\n",
            "2020-02-04 19:36:07,411 Epoch   1 Step:      200 Batch Loss:     4.373751 Tokens per Sec:     7390, Lr: 0.000140\n",
            "2020-02-04 19:37:16,614 Epoch   1 Step:      400 Batch Loss:     3.446606 Tokens per Sec:     6708, Lr: 0.000280\n",
            "2020-02-04 19:38:25,797 Epoch   1 Step:      600 Batch Loss:     3.172154 Tokens per Sec:     6715, Lr: 0.000419\n",
            "2020-02-04 19:39:09,436 Epoch   1: total training loss 2923.78\n",
            "2020-02-04 19:39:09,436 EPOCH 2\n",
            "2020-02-04 19:39:35,740 Epoch   2 Step:      800 Batch Loss:     3.004483 Tokens per Sec:     6665, Lr: 0.000559\n",
            "2020-02-04 19:40:45,337 Epoch   2 Step:     1000 Batch Loss:     2.668867 Tokens per Sec:     6703, Lr: 0.000699\n",
            "2020-02-04 19:41:55,230 Epoch   2 Step:     1200 Batch Loss:     2.925653 Tokens per Sec:     6683, Lr: 0.000638\n",
            "2020-02-04 19:43:05,191 Epoch   2 Step:     1400 Batch Loss:     2.579067 Tokens per Sec:     6708, Lr: 0.000591\n",
            "2020-02-04 19:43:22,214 Epoch   2: total training loss 1923.77\n",
            "2020-02-04 19:43:22,214 EPOCH 3\n",
            "2020-02-04 19:44:14,693 Epoch   3 Step:     1600 Batch Loss:     2.535758 Tokens per Sec:     6675, Lr: 0.000552\n",
            "2020-02-04 19:45:24,087 Epoch   3 Step:     1800 Batch Loss:     2.087405 Tokens per Sec:     6751, Lr: 0.000521\n",
            "2020-02-04 19:46:33,332 Epoch   3 Step:     2000 Batch Loss:     2.259302 Tokens per Sec:     6709, Lr: 0.000494\n",
            "2020-02-04 19:47:00,454 Hooray! New best validation result [eval_metric]!\n",
            "2020-02-04 19:47:00,454 Saving new checkpoint.\n",
            "2020-02-04 19:47:02,115 Example #0\n",
            "2020-02-04 19:47:02,116 \tSource:     ( read philippians 2 : 5 - 8 . )\n",
            "2020-02-04 19:47:02,116 \tReference:  ( tanga filipe 2 : 5 - 8 . )\n",
            "2020-02-04 19:47:02,116 \tHypothesis: ( tanga filipe 2 : 5 - 8 . )\n",
            "2020-02-04 19:47:02,116 Example #1\n",
            "2020-02-04 19:47:02,116 \tSource:     today , more than ever before , there are so many things that can distract us .\n",
            "2020-02-04 19:47:02,116 \tReference:  lelu , kuene ima iavulu i tena ku tu landukisa , m’ukulu ndenge .\n",
            "2020-02-04 19:47:02,117 \tHypothesis: lelu , kuene ima iavulu i tena ku tu kuatekesa .\n",
            "2020-02-04 19:47:02,117 Example #2\n",
            "2020-02-04 19:47:02,117 \tSource:     the apostle paul warned about what can happen if we please ourselves first .\n",
            "2020-02-04 19:47:02,117 \tReference:  ( 1 nzuá 2 : 16 ) o poxolo phaulu ua tu dimuna ku ima i tena kubhita se tu suua ngó o ima i tua uabhela .\n",
            "2020-02-04 19:47:02,117 \tHypothesis: o poxolo phaulu ua dimuna o ima i tu bhanga se tu bhanga .\n",
            "2020-02-04 19:47:02,117 Example #3\n",
            "2020-02-04 19:47:02,117 \tSource:     what a privilege it is to live in these last days and to be part of jehovah’s incredible organization !\n",
            "2020-02-04 19:47:02,117 \tReference:  tua tokala ku lembalala izuua ioso kuila , ujitu ua dikota ku tokala mu kilunga kia jihova mu ixi , mu izuua isukidila - ku !\n",
            "2020-02-04 19:47:02,117 \tHypothesis: o ujitu iú , u tena ku tu kuatekesa ku kala ni ukamba uambote ni jihova !\n",
            "2020-02-04 19:47:02,117 Validation result (greedy) at epoch   3, step     2000: bleu:  14.86, loss: 47097.8398, ppl:   8.2633, duration: 28.7850s\n",
            "2020-02-04 19:48:03,094 Epoch   3: total training loss 1615.38\n",
            "2020-02-04 19:48:03,094 EPOCH 4\n",
            "2020-02-04 19:48:11,783 Epoch   4 Step:     2200 Batch Loss:     2.120787 Tokens per Sec:     6459, Lr: 0.000471\n",
            "2020-02-04 19:49:20,971 Epoch   4 Step:     2400 Batch Loss:     1.741877 Tokens per Sec:     6705, Lr: 0.000451\n",
            "2020-02-04 19:50:30,588 Epoch   4 Step:     2600 Batch Loss:     2.465448 Tokens per Sec:     6751, Lr: 0.000433\n",
            "2020-02-04 19:51:39,815 Epoch   4 Step:     2800 Batch Loss:     2.057468 Tokens per Sec:     6758, Lr: 0.000418\n",
            "2020-02-04 19:52:14,527 Epoch   4: total training loss 1468.87\n",
            "2020-02-04 19:52:14,527 EPOCH 5\n",
            "2020-02-04 19:52:48,681 Epoch   5 Step:     3000 Batch Loss:     1.418909 Tokens per Sec:     6716, Lr: 0.000403\n",
            "2020-02-04 19:53:57,999 Epoch   5 Step:     3200 Batch Loss:     1.938244 Tokens per Sec:     6788, Lr: 0.000391\n",
            "2020-02-04 19:55:07,478 Epoch   5 Step:     3400 Batch Loss:     1.893386 Tokens per Sec:     6756, Lr: 0.000379\n",
            "2020-02-04 19:56:16,738 Epoch   5 Step:     3600 Batch Loss:     1.922130 Tokens per Sec:     6749, Lr: 0.000368\n",
            "2020-02-04 19:56:24,670 Epoch   5: total training loss 1360.27\n",
            "2020-02-04 19:56:24,671 EPOCH 6\n",
            "2020-02-04 19:57:26,127 Epoch   6 Step:     3800 Batch Loss:     1.733316 Tokens per Sec:     6793, Lr: 0.000358\n",
            "2020-02-04 19:58:35,178 Epoch   6 Step:     4000 Batch Loss:     1.426156 Tokens per Sec:     6764, Lr: 0.000349\n",
            "2020-02-04 19:59:03,345 Hooray! New best validation result [eval_metric]!\n",
            "2020-02-04 19:59:03,345 Saving new checkpoint.\n",
            "2020-02-04 19:59:04,958 Example #0\n",
            "2020-02-04 19:59:04,959 \tSource:     ( read philippians 2 : 5 - 8 . )\n",
            "2020-02-04 19:59:04,959 \tReference:  ( tanga filipe 2 : 5 - 8 . )\n",
            "2020-02-04 19:59:04,959 \tHypothesis: ( tanga filipe 2 : 5 - 8 . )\n",
            "2020-02-04 19:59:04,959 Example #1\n",
            "2020-02-04 19:59:04,959 \tSource:     today , more than ever before , there are so many things that can distract us .\n",
            "2020-02-04 19:59:04,959 \tReference:  lelu , kuene ima iavulu i tena ku tu landukisa , m’ukulu ndenge .\n",
            "2020-02-04 19:59:04,960 \tHypothesis: lelu , kua bhingi dingi ima iavulu , phala ku tu kuatekesa ku dibhana ni maka enhá .\n",
            "2020-02-04 19:59:04,960 Example #2\n",
            "2020-02-04 19:59:04,960 \tSource:     the apostle paul warned about what can happen if we please ourselves first .\n",
            "2020-02-04 19:59:04,960 \tReference:  ( 1 nzuá 2 : 16 ) o poxolo phaulu ua tu dimuna ku ima i tena kubhita se tu suua ngó o ima i tua uabhela .\n",
            "2020-02-04 19:59:04,960 \tHypothesis: o poxolo phaulu ua tendelesa o ima ia - nda bhita se tu bhanga o ima ia iibha .\n",
            "2020-02-04 19:59:04,960 Example #3\n",
            "2020-02-04 19:59:04,960 \tSource:     what a privilege it is to live in these last days and to be part of jehovah’s incredible organization !\n",
            "2020-02-04 19:59:04,960 \tReference:  tua tokala ku lembalala izuua ioso kuila , ujitu ua dikota ku tokala mu kilunga kia jihova mu ixi , mu izuua isukidila - ku !\n",
            "2020-02-04 19:59:04,960 \tHypothesis: ujitu ua dikota ku kala ku muenhu mu izuua íii isukidila - ku , ni ku kala ni ukamba uambote ni jihova !\n",
            "2020-02-04 19:59:04,960 Validation result (greedy) at epoch   6, step     4000: bleu:  20.91, loss: 39856.2031, ppl:   5.9722, duration: 29.7820s\n",
            "2020-02-04 20:00:13,886 Epoch   6 Step:     4200 Batch Loss:     1.681094 Tokens per Sec:     6784, Lr: 0.000341\n",
            "2020-02-04 20:01:04,062 Epoch   6: total training loss 1288.14\n",
            "2020-02-04 20:01:04,062 EPOCH 7\n",
            "2020-02-04 20:01:22,592 Epoch   7 Step:     4400 Batch Loss:     1.869816 Tokens per Sec:     6580, Lr: 0.000333\n",
            "2020-02-04 20:02:32,112 Epoch   7 Step:     4600 Batch Loss:     1.868371 Tokens per Sec:     6776, Lr: 0.000326\n",
            "2020-02-04 20:03:40,888 Epoch   7 Step:     4800 Batch Loss:     1.494951 Tokens per Sec:     6731, Lr: 0.000319\n",
            "2020-02-04 20:04:49,964 Epoch   7 Step:     5000 Batch Loss:     1.940855 Tokens per Sec:     6738, Lr: 0.000313\n",
            "2020-02-04 20:05:14,992 Epoch   7: total training loss 1217.81\n",
            "2020-02-04 20:05:14,993 EPOCH 8\n",
            "2020-02-04 20:05:59,350 Epoch   8 Step:     5200 Batch Loss:     1.924535 Tokens per Sec:     6736, Lr: 0.000306\n",
            "2020-02-04 20:07:08,852 Epoch   8 Step:     5400 Batch Loss:     1.800949 Tokens per Sec:     6749, Lr: 0.000301\n",
            "2020-02-04 20:08:17,863 Epoch   8 Step:     5600 Batch Loss:     1.651664 Tokens per Sec:     6767, Lr: 0.000295\n",
            "2020-02-04 20:09:25,274 Epoch   8: total training loss 1154.32\n",
            "2020-02-04 20:09:25,274 EPOCH 9\n",
            "2020-02-04 20:09:26,981 Epoch   9 Step:     5800 Batch Loss:     1.459499 Tokens per Sec:     5603, Lr: 0.000290\n",
            "2020-02-04 20:10:36,310 Epoch   9 Step:     6000 Batch Loss:     1.670126 Tokens per Sec:     6785, Lr: 0.000285\n",
            "2020-02-04 20:11:02,199 Hooray! New best validation result [eval_metric]!\n",
            "2020-02-04 20:11:02,199 Saving new checkpoint.\n",
            "2020-02-04 20:11:03,924 Example #0\n",
            "2020-02-04 20:11:03,924 \tSource:     ( read philippians 2 : 5 - 8 . )\n",
            "2020-02-04 20:11:03,924 \tReference:  ( tanga filipe 2 : 5 - 8 . )\n",
            "2020-02-04 20:11:03,924 \tHypothesis: ( tanga filipe 2 : 5 - 8 . )\n",
            "2020-02-04 20:11:03,924 Example #1\n",
            "2020-02-04 20:11:03,925 \tSource:     today , more than ever before , there are so many things that can distract us .\n",
            "2020-02-04 20:11:03,925 \tReference:  lelu , kuene ima iavulu i tena ku tu landukisa , m’ukulu ndenge .\n",
            "2020-02-04 20:11:03,925 \tHypothesis: lelu , kua kambe ngó bhofele , ande dia ku bhita o ima i tena ku tu landukisa .\n",
            "2020-02-04 20:11:03,925 Example #2\n",
            "2020-02-04 20:11:03,925 \tSource:     the apostle paul warned about what can happen if we please ourselves first .\n",
            "2020-02-04 20:11:03,925 \tReference:  ( 1 nzuá 2 : 16 ) o poxolo phaulu ua tu dimuna ku ima i tena kubhita se tu suua ngó o ima i tua uabhela .\n",
            "2020-02-04 20:11:03,925 \tHypothesis: o poxolo phaulu ua dimuna ia lungu ni ima i tu bhita na - iu , se tu bhanga o ima i tua mesena .\n",
            "2020-02-04 20:11:03,925 Example #3\n",
            "2020-02-04 20:11:03,925 \tSource:     what a privilege it is to live in these last days and to be part of jehovah’s incredible organization !\n",
            "2020-02-04 20:11:03,925 \tReference:  tua tokala ku lembalala izuua ioso kuila , ujitu ua dikota ku tokala mu kilunga kia jihova mu ixi , mu izuua isukidila - ku !\n",
            "2020-02-04 20:11:03,925 \tHypothesis: ujitu ua dikota ku kala mu izuua isukidila - ku , ni ku bhanga mbandu ku kilunga kia jihova !\n",
            "2020-02-04 20:11:03,925 Validation result (greedy) at epoch   9, step     6000: bleu:  23.81, loss: 37121.8281, ppl:   5.2831, duration: 27.6148s\n",
            "2020-02-04 20:12:12,509 Epoch   9 Step:     6200 Batch Loss:     1.229483 Tokens per Sec:     6730, Lr: 0.000281\n",
            "2020-02-04 20:13:21,856 Epoch   9 Step:     6400 Batch Loss:     1.410562 Tokens per Sec:     6790, Lr: 0.000276\n",
            "2020-02-04 20:14:03,446 Epoch   9: total training loss 1114.11\n",
            "2020-02-04 20:14:03,446 EPOCH 10\n",
            "2020-02-04 20:14:31,209 Epoch  10 Step:     6600 Batch Loss:     1.540644 Tokens per Sec:     6716, Lr: 0.000272\n",
            "2020-02-04 20:15:39,776 Epoch  10 Step:     6800 Batch Loss:     1.627514 Tokens per Sec:     6744, Lr: 0.000268\n",
            "2020-02-04 20:16:49,088 Epoch  10 Step:     7000 Batch Loss:     1.325339 Tokens per Sec:     6780, Lr: 0.000264\n",
            "2020-02-04 20:17:58,791 Epoch  10 Step:     7200 Batch Loss:     1.359054 Tokens per Sec:     6726, Lr: 0.000260\n",
            "2020-02-04 20:18:13,983 Epoch  10: total training loss 1072.73\n",
            "2020-02-04 20:18:13,983 EPOCH 11\n",
            "2020-02-04 20:19:07,831 Epoch  11 Step:     7400 Batch Loss:     1.427467 Tokens per Sec:     6722, Lr: 0.000257\n",
            "2020-02-04 20:20:17,024 Epoch  11 Step:     7600 Batch Loss:     1.435952 Tokens per Sec:     6733, Lr: 0.000253\n",
            "2020-02-04 20:21:26,239 Epoch  11 Step:     7800 Batch Loss:     1.629339 Tokens per Sec:     6752, Lr: 0.000250\n",
            "2020-02-04 20:22:24,633 Epoch  11: total training loss 1036.26\n",
            "2020-02-04 20:22:24,633 EPOCH 12\n",
            "2020-02-04 20:22:35,411 Epoch  12 Step:     8000 Batch Loss:     1.378333 Tokens per Sec:     6642, Lr: 0.000247\n",
            "2020-02-04 20:23:01,670 Hooray! New best validation result [eval_metric]!\n",
            "2020-02-04 20:23:01,670 Saving new checkpoint.\n",
            "2020-02-04 20:23:03,279 Example #0\n",
            "2020-02-04 20:23:03,280 \tSource:     ( read philippians 2 : 5 - 8 . )\n",
            "2020-02-04 20:23:03,280 \tReference:  ( tanga filipe 2 : 5 - 8 . )\n",
            "2020-02-04 20:23:03,280 \tHypothesis: ( tanga filipe 2 : 5 - 8 . )\n",
            "2020-02-04 20:23:03,280 Example #1\n",
            "2020-02-04 20:23:03,280 \tSource:     today , more than ever before , there are so many things that can distract us .\n",
            "2020-02-04 20:23:03,280 \tReference:  lelu , kuene ima iavulu i tena ku tu landukisa , m’ukulu ndenge .\n",
            "2020-02-04 20:23:03,280 \tHypothesis: lelu , kua kambe ngó bhofele , ande dia ku bhita o ima ioso , saí ima iavulu i tena ku tu landukisa .\n",
            "2020-02-04 20:23:03,280 Example #2\n",
            "2020-02-04 20:23:03,280 \tSource:     the apostle paul warned about what can happen if we please ourselves first .\n",
            "2020-02-04 20:23:03,280 \tReference:  ( 1 nzuá 2 : 16 ) o poxolo phaulu ua tu dimuna ku ima i tena kubhita se tu suua ngó o ima i tua uabhela .\n",
            "2020-02-04 20:23:03,280 \tHypothesis: o poxolo phaulu ua dimuna o ima i tu bhita , se tu bhanga o ima i tua mesena .\n",
            "2020-02-04 20:23:03,280 Example #3\n",
            "2020-02-04 20:23:03,281 \tSource:     what a privilege it is to live in these last days and to be part of jehovah’s incredible organization !\n",
            "2020-02-04 20:23:03,281 \tReference:  tua tokala ku lembalala izuua ioso kuila , ujitu ua dikota ku tokala mu kilunga kia jihova mu ixi , mu izuua isukidila - ku !\n",
            "2020-02-04 20:23:03,281 \tHypothesis: ujitu ua dikota ku kala mu izuua isukidila - ku , ni ku bhanga mbandu ku kilunga kia jihova !\n",
            "2020-02-04 20:23:03,281 Validation result (greedy) at epoch  12, step     8000: bleu:  25.00, loss: 35674.4961, ppl:   4.9511, duration: 27.8696s\n",
            "2020-02-04 20:24:12,367 Epoch  12 Step:     8200 Batch Loss:     1.231285 Tokens per Sec:     6702, Lr: 0.000244\n",
            "2020-02-04 20:25:21,048 Epoch  12 Step:     8400 Batch Loss:     1.578112 Tokens per Sec:     6722, Lr: 0.000241\n",
            "2020-02-04 20:26:30,454 Epoch  12 Step:     8600 Batch Loss:     1.382758 Tokens per Sec:     6747, Lr: 0.000238\n",
            "2020-02-04 20:27:03,588 Epoch  12: total training loss 1010.16\n",
            "2020-02-04 20:27:03,589 EPOCH 13\n",
            "2020-02-04 20:27:39,888 Epoch  13 Step:     8800 Batch Loss:     1.088220 Tokens per Sec:     6760, Lr: 0.000236\n",
            "2020-02-04 20:28:49,663 Epoch  13 Step:     9000 Batch Loss:     1.332197 Tokens per Sec:     6763, Lr: 0.000233\n",
            "2020-02-04 20:29:58,891 Epoch  13 Step:     9200 Batch Loss:     1.640509 Tokens per Sec:     6770, Lr: 0.000230\n",
            "2020-02-04 20:31:08,391 Epoch  13 Step:     9400 Batch Loss:     1.451203 Tokens per Sec:     6703, Lr: 0.000228\n",
            "2020-02-04 20:31:14,434 Epoch  13: total training loss 970.57\n",
            "2020-02-04 20:31:14,434 EPOCH 14\n",
            "2020-02-04 20:32:17,273 Epoch  14 Step:     9600 Batch Loss:     1.308440 Tokens per Sec:     6706, Lr: 0.000226\n",
            "2020-02-04 20:33:25,976 Epoch  14 Step:     9800 Batch Loss:     1.202386 Tokens per Sec:     6760, Lr: 0.000223\n",
            "2020-02-04 20:34:35,241 Epoch  14 Step:    10000 Batch Loss:     1.115765 Tokens per Sec:     6750, Lr: 0.000221\n",
            "2020-02-04 20:34:57,532 Hooray! New best validation result [eval_metric]!\n",
            "2020-02-04 20:34:57,532 Saving new checkpoint.\n",
            "2020-02-04 20:34:59,217 Example #0\n",
            "2020-02-04 20:34:59,217 \tSource:     ( read philippians 2 : 5 - 8 . )\n",
            "2020-02-04 20:34:59,217 \tReference:  ( tanga filipe 2 : 5 - 8 . )\n",
            "2020-02-04 20:34:59,217 \tHypothesis: ( tanga filipe 2 : 5 - 8 . )\n",
            "2020-02-04 20:34:59,217 Example #1\n",
            "2020-02-04 20:34:59,218 \tSource:     today , more than ever before , there are so many things that can distract us .\n",
            "2020-02-04 20:34:59,218 \tReference:  lelu , kuene ima iavulu i tena ku tu landukisa , m’ukulu ndenge .\n",
            "2020-02-04 20:34:59,218 \tHypothesis: lelu , kua kambe ngó , ande dia ku bhita o ima ioso , saí ima iavulu i tena ku tu landukisa .\n",
            "2020-02-04 20:34:59,218 Example #2\n",
            "2020-02-04 20:34:59,218 \tSource:     the apostle paul warned about what can happen if we please ourselves first .\n",
            "2020-02-04 20:34:59,218 \tReference:  ( 1 nzuá 2 : 16 ) o poxolo phaulu ua tu dimuna ku ima i tena kubhita se tu suua ngó o ima i tua uabhela .\n",
            "2020-02-04 20:34:59,218 \tHypothesis: o poxolo phaulu ua dimuna o ima i tena kubhita , se tu i sangulukisa .\n",
            "2020-02-04 20:34:59,218 Example #3\n",
            "2020-02-04 20:34:59,218 \tSource:     what a privilege it is to live in these last days and to be part of jehovah’s incredible organization !\n",
            "2020-02-04 20:34:59,218 \tReference:  tua tokala ku lembalala izuua ioso kuila , ujitu ua dikota ku tokala mu kilunga kia jihova mu ixi , mu izuua isukidila - ku !\n",
            "2020-02-04 20:34:59,218 \tHypothesis: ujitu ua dikota ku kala mu izuua íii isukidila - ku , ni ku bhanga mbandu ku kilunga kia jihova !\n",
            "2020-02-04 20:34:59,219 Validation result (greedy) at epoch  14, step    10000: bleu:  26.10, loss: 34720.7383, ppl:   4.7438, duration: 23.9776s\n",
            "2020-02-04 20:35:48,955 Epoch  14: total training loss 946.58\n",
            "2020-02-04 20:35:48,955 EPOCH 15\n",
            "2020-02-04 20:36:08,972 Epoch  15 Step:    10200 Batch Loss:     1.149951 Tokens per Sec:     6741, Lr: 0.000219\n",
            "2020-02-04 20:37:18,232 Epoch  15 Step:    10400 Batch Loss:     1.112914 Tokens per Sec:     6765, Lr: 0.000217\n",
            "2020-02-04 20:38:27,355 Epoch  15 Step:    10600 Batch Loss:     1.485635 Tokens per Sec:     6714, Lr: 0.000215\n",
            "2020-02-04 20:39:36,381 Epoch  15 Step:    10800 Batch Loss:     1.262681 Tokens per Sec:     6753, Lr: 0.000213\n",
            "2020-02-04 20:39:59,681 Epoch  15: total training loss 917.70\n",
            "2020-02-04 20:39:59,681 EPOCH 16\n",
            "2020-02-04 20:40:45,804 Epoch  16 Step:    11000 Batch Loss:     1.141444 Tokens per Sec:     6700, Lr: 0.000211\n",
            "2020-02-04 20:41:54,926 Epoch  16 Step:    11200 Batch Loss:     1.221109 Tokens per Sec:     6773, Lr: 0.000209\n",
            "2020-02-04 20:43:03,746 Epoch  16 Step:    11400 Batch Loss:     1.410407 Tokens per Sec:     6695, Lr: 0.000207\n",
            "2020-02-04 20:44:10,887 Epoch  16: total training loss 897.40\n",
            "2020-02-04 20:44:10,887 EPOCH 17\n",
            "2020-02-04 20:44:12,739 Epoch  17 Step:    11600 Batch Loss:     1.301653 Tokens per Sec:     6571, Lr: 0.000205\n",
            "2020-02-04 20:45:22,036 Epoch  17 Step:    11800 Batch Loss:     1.178325 Tokens per Sec:     6740, Lr: 0.000203\n",
            "2020-02-04 20:46:31,137 Epoch  17 Step:    12000 Batch Loss:     1.176379 Tokens per Sec:     6731, Lr: 0.000202\n",
            "2020-02-04 20:47:03,313 Hooray! New best validation result [eval_metric]!\n",
            "2020-02-04 20:47:03,314 Saving new checkpoint.\n",
            "2020-02-04 20:47:05,095 Example #0\n",
            "2020-02-04 20:47:05,095 \tSource:     ( read philippians 2 : 5 - 8 . )\n",
            "2020-02-04 20:47:05,095 \tReference:  ( tanga filipe 2 : 5 - 8 . )\n",
            "2020-02-04 20:47:05,095 \tHypothesis: ( tanga filipe 2 : 5 - 8 . )\n",
            "2020-02-04 20:47:05,095 Example #1\n",
            "2020-02-04 20:47:05,095 \tSource:     today , more than ever before , there are so many things that can distract us .\n",
            "2020-02-04 20:47:05,095 \tReference:  lelu , kuene ima iavulu i tena ku tu landukisa , m’ukulu ndenge .\n",
            "2020-02-04 20:47:05,096 \tHypothesis: lelu , ande dia ku bhita o ima , saí ima iavulu i tena ku tu landukisa .\n",
            "2020-02-04 20:47:05,096 Example #2\n",
            "2020-02-04 20:47:05,096 \tSource:     the apostle paul warned about what can happen if we please ourselves first .\n",
            "2020-02-04 20:47:05,096 \tReference:  ( 1 nzuá 2 : 16 ) o poxolo phaulu ua tu dimuna ku ima i tena kubhita se tu suua ngó o ima i tua uabhela .\n",
            "2020-02-04 20:47:05,096 \tHypothesis: o poxolo phaulu ua dimuna o ima i tu tena o kubhanga , se tu i sangulukisa o muxima ua nzambi .\n",
            "2020-02-04 20:47:05,096 Example #3\n",
            "2020-02-04 20:47:05,096 \tSource:     what a privilege it is to live in these last days and to be part of jehovah’s incredible organization !\n",
            "2020-02-04 20:47:05,096 \tReference:  tua tokala ku lembalala izuua ioso kuila , ujitu ua dikota ku tokala mu kilunga kia jihova mu ixi , mu izuua isukidila - ku !\n",
            "2020-02-04 20:47:05,096 \tHypothesis: ujitu ua dikota ku kala mu izuua íii isukidila - ku , ni ku bhanga mbandu ku kilunga kia jihova !\n",
            "2020-02-04 20:47:05,096 Validation result (greedy) at epoch  17, step    12000: bleu:  26.16, loss: 34417.0859, ppl:   4.6797, duration: 33.9585s\n",
            "2020-02-04 20:48:14,113 Epoch  17 Step:    12200 Batch Loss:     1.303897 Tokens per Sec:     6802, Lr: 0.000200\n",
            "2020-02-04 20:48:55,324 Epoch  17: total training loss 872.81\n",
            "2020-02-04 20:48:55,324 EPOCH 18\n",
            "2020-02-04 20:49:23,275 Epoch  18 Step:    12400 Batch Loss:     1.313688 Tokens per Sec:     6647, Lr: 0.000198\n",
            "2020-02-04 20:50:32,802 Epoch  18 Step:    12600 Batch Loss:     1.247998 Tokens per Sec:     6754, Lr: 0.000197\n",
            "2020-02-04 20:51:42,179 Epoch  18 Step:    12800 Batch Loss:     1.266422 Tokens per Sec:     6754, Lr: 0.000195\n",
            "2020-02-04 20:52:51,013 Epoch  18 Step:    13000 Batch Loss:     1.123238 Tokens per Sec:     6736, Lr: 0.000194\n",
            "2020-02-04 20:53:06,337 Epoch  18: total training loss 851.96\n",
            "2020-02-04 20:53:06,337 EPOCH 19\n",
            "2020-02-04 20:53:59,248 Epoch  19 Step:    13200 Batch Loss:     0.979365 Tokens per Sec:     6745, Lr: 0.000192\n",
            "2020-02-04 20:55:08,508 Epoch  19 Step:    13400 Batch Loss:     1.147491 Tokens per Sec:     6792, Lr: 0.000191\n",
            "2020-02-04 20:56:17,426 Epoch  19 Step:    13600 Batch Loss:     1.091969 Tokens per Sec:     6755, Lr: 0.000189\n",
            "2020-02-04 20:57:16,512 Epoch  19: total training loss 832.52\n",
            "2020-02-04 20:57:16,512 EPOCH 20\n",
            "2020-02-04 20:57:26,043 Epoch  20 Step:    13800 Batch Loss:     1.193934 Tokens per Sec:     6556, Lr: 0.000188\n",
            "2020-02-04 20:58:34,991 Epoch  20 Step:    14000 Batch Loss:     1.229089 Tokens per Sec:     6751, Lr: 0.000187\n",
            "2020-02-04 20:59:00,784 Hooray! New best validation result [eval_metric]!\n",
            "2020-02-04 20:59:00,785 Saving new checkpoint.\n",
            "2020-02-04 20:59:02,527 Example #0\n",
            "2020-02-04 20:59:02,527 \tSource:     ( read philippians 2 : 5 - 8 . )\n",
            "2020-02-04 20:59:02,527 \tReference:  ( tanga filipe 2 : 5 - 8 . )\n",
            "2020-02-04 20:59:02,527 \tHypothesis: ( tanga filipe 2 : 5 - 8 . )\n",
            "2020-02-04 20:59:02,527 Example #1\n",
            "2020-02-04 20:59:02,527 \tSource:     today , more than ever before , there are so many things that can distract us .\n",
            "2020-02-04 20:59:02,527 \tReference:  lelu , kuene ima iavulu i tena ku tu landukisa , m’ukulu ndenge .\n",
            "2020-02-04 20:59:02,528 \tHypothesis: lelu , ande dia ku bhita o ima , saí ima iavulu i tena ku tu landukisa .\n",
            "2020-02-04 20:59:02,528 Example #2\n",
            "2020-02-04 20:59:02,528 \tSource:     the apostle paul warned about what can happen if we please ourselves first .\n",
            "2020-02-04 20:59:02,528 \tReference:  ( 1 nzuá 2 : 16 ) o poxolo phaulu ua tu dimuna ku ima i tena kubhita se tu suua ngó o ima i tua uabhela .\n",
            "2020-02-04 20:59:02,528 \tHypothesis: o poxolo phaulu ua dimuna o ima i tena kubhita , se tu i sangulukisa o muxima uetu .\n",
            "2020-02-04 20:59:02,528 Example #3\n",
            "2020-02-04 20:59:02,528 \tSource:     what a privilege it is to live in these last days and to be part of jehovah’s incredible organization !\n",
            "2020-02-04 20:59:02,528 \tReference:  tua tokala ku lembalala izuua ioso kuila , ujitu ua dikota ku tokala mu kilunga kia jihova mu ixi , mu izuua isukidila - ku !\n",
            "2020-02-04 20:59:02,528 \tHypothesis: ujitu ua dikota ku kala mu izuua íii isukidila - ku , ni ku bhanga mbandu ku kilunga kia jihova !\n",
            "2020-02-04 20:59:02,528 Validation result (greedy) at epoch  20, step    14000: bleu:  27.21, loss: 34284.7305, ppl:   4.6520, duration: 27.5373s\n",
            "2020-02-04 21:00:11,822 Epoch  20 Step:    14200 Batch Loss:     1.060636 Tokens per Sec:     6718, Lr: 0.000185\n",
            "2020-02-04 21:01:20,983 Epoch  20 Step:    14400 Batch Loss:     1.281659 Tokens per Sec:     6710, Lr: 0.000184\n",
            "2020-02-04 21:01:55,348 Epoch  20: total training loss 815.14\n",
            "2020-02-04 21:01:55,348 EPOCH 21\n",
            "2020-02-04 21:02:30,501 Epoch  21 Step:    14600 Batch Loss:     1.254682 Tokens per Sec:     6780, Lr: 0.000183\n",
            "2020-02-04 21:03:40,040 Epoch  21 Step:    14800 Batch Loss:     1.393568 Tokens per Sec:     6728, Lr: 0.000182\n",
            "2020-02-04 21:04:48,713 Epoch  21 Step:    15000 Batch Loss:     0.893659 Tokens per Sec:     6724, Lr: 0.000180\n",
            "2020-02-04 21:05:57,599 Epoch  21 Step:    15200 Batch Loss:     0.912125 Tokens per Sec:     6745, Lr: 0.000179\n",
            "2020-02-04 21:06:06,220 Epoch  21: total training loss 794.84\n",
            "2020-02-04 21:06:06,220 EPOCH 22\n",
            "2020-02-04 21:07:06,241 Epoch  22 Step:    15400 Batch Loss:     1.034763 Tokens per Sec:     6755, Lr: 0.000178\n",
            "2020-02-04 21:08:14,942 Epoch  22 Step:    15600 Batch Loss:     1.081931 Tokens per Sec:     6748, Lr: 0.000177\n",
            "2020-02-04 21:09:24,367 Epoch  22 Step:    15800 Batch Loss:     1.300140 Tokens per Sec:     6756, Lr: 0.000176\n",
            "2020-02-04 21:10:16,419 Epoch  22: total training loss 776.06\n",
            "2020-02-04 21:10:16,419 EPOCH 23\n",
            "2020-02-04 21:10:33,331 Epoch  23 Step:    16000 Batch Loss:     1.199994 Tokens per Sec:     6680, Lr: 0.000175\n",
            "2020-02-04 21:11:06,115 Hooray! New best validation result [eval_metric]!\n",
            "2020-02-04 21:11:06,115 Saving new checkpoint.\n",
            "2020-02-04 21:11:07,697 Example #0\n",
            "2020-02-04 21:11:07,698 \tSource:     ( read philippians 2 : 5 - 8 . )\n",
            "2020-02-04 21:11:07,698 \tReference:  ( tanga filipe 2 : 5 - 8 . )\n",
            "2020-02-04 21:11:07,698 \tHypothesis: ( tanga filipe 2 : 5 - 8 . )\n",
            "2020-02-04 21:11:07,698 Example #1\n",
            "2020-02-04 21:11:07,698 \tSource:     today , more than ever before , there are so many things that can distract us .\n",
            "2020-02-04 21:11:07,698 \tReference:  lelu , kuene ima iavulu i tena ku tu landukisa , m’ukulu ndenge .\n",
            "2020-02-04 21:11:07,698 \tHypothesis: lelu , kua kambe ngó bhofele , ande dia ima iavulu i tena ku tu landukisa .\n",
            "2020-02-04 21:11:07,698 Example #2\n",
            "2020-02-04 21:11:07,698 \tSource:     the apostle paul warned about what can happen if we please ourselves first .\n",
            "2020-02-04 21:11:07,698 \tReference:  ( 1 nzuá 2 : 16 ) o poxolo phaulu ua tu dimuna ku ima i tena kubhita se tu suua ngó o ima i tua uabhela .\n",
            "2020-02-04 21:11:07,698 \tHypothesis: o poxolo phaulu ua dimuna o ima i tena kubhita , se tu mesena ku ta o kituxi .\n",
            "2020-02-04 21:11:07,698 Example #3\n",
            "2020-02-04 21:11:07,699 \tSource:     what a privilege it is to live in these last days and to be part of jehovah’s incredible organization !\n",
            "2020-02-04 21:11:07,699 \tReference:  tua tokala ku lembalala izuua ioso kuila , ujitu ua dikota ku tokala mu kilunga kia jihova mu ixi , mu izuua isukidila - ku !\n",
            "2020-02-04 21:11:07,699 \tHypothesis: ujitu ua dikota ku kala mu izuua íii isukidila - ku , ni ku bhanga mbandu ku kilunga kia jihova !\n",
            "2020-02-04 21:11:07,699 Validation result (greedy) at epoch  23, step    16000: bleu:  27.29, loss: 34352.4297, ppl:   4.6661, duration: 34.3671s\n",
            "2020-02-04 21:12:16,709 Epoch  23 Step:    16200 Batch Loss:     0.907302 Tokens per Sec:     6697, Lr: 0.000174\n",
            "2020-02-04 21:13:26,235 Epoch  23 Step:    16400 Batch Loss:     1.082953 Tokens per Sec:     6728, Lr: 0.000173\n",
            "2020-02-04 21:14:35,507 Epoch  23 Step:    16600 Batch Loss:     0.971708 Tokens per Sec:     6791, Lr: 0.000172\n",
            "2020-02-04 21:15:01,924 Epoch  23: total training loss 760.18\n",
            "2020-02-04 21:15:01,924 EPOCH 24\n",
            "2020-02-04 21:15:44,333 Epoch  24 Step:    16800 Batch Loss:     0.790211 Tokens per Sec:     6703, Lr: 0.000170\n",
            "2020-02-04 21:16:53,763 Epoch  24 Step:    17000 Batch Loss:     1.323298 Tokens per Sec:     6747, Lr: 0.000169\n",
            "2020-02-04 21:18:03,061 Epoch  24 Step:    17200 Batch Loss:     1.141842 Tokens per Sec:     6748, Lr: 0.000168\n",
            "2020-02-04 21:19:12,428 Epoch  24 Step:    17400 Batch Loss:     1.137136 Tokens per Sec:     6745, Lr: 0.000168\n",
            "2020-02-04 21:19:12,779 Epoch  24: total training loss 742.43\n",
            "2020-02-04 21:19:12,779 EPOCH 25\n",
            "2020-02-04 21:20:22,121 Epoch  25 Step:    17600 Batch Loss:     1.002455 Tokens per Sec:     6736, Lr: 0.000167\n",
            "2020-02-04 21:21:31,568 Epoch  25 Step:    17800 Batch Loss:     1.016481 Tokens per Sec:     6740, Lr: 0.000166\n",
            "2020-02-04 21:22:40,598 Epoch  25 Step:    18000 Batch Loss:     0.992400 Tokens per Sec:     6755, Lr: 0.000165\n",
            "2020-02-04 21:23:06,908 Hooray! New best validation result [eval_metric]!\n",
            "2020-02-04 21:23:06,909 Saving new checkpoint.\n",
            "2020-02-04 21:23:08,660 Example #0\n",
            "2020-02-04 21:23:08,660 \tSource:     ( read philippians 2 : 5 - 8 . )\n",
            "2020-02-04 21:23:08,660 \tReference:  ( tanga filipe 2 : 5 - 8 . )\n",
            "2020-02-04 21:23:08,660 \tHypothesis: ( tanga filipe 2 : 5 - 8 . )\n",
            "2020-02-04 21:23:08,660 Example #1\n",
            "2020-02-04 21:23:08,661 \tSource:     today , more than ever before , there are so many things that can distract us .\n",
            "2020-02-04 21:23:08,661 \tReference:  lelu , kuene ima iavulu i tena ku tu landukisa , m’ukulu ndenge .\n",
            "2020-02-04 21:23:08,661 \tHypothesis: lelu , ande dia ku bhita o ima , saí ima iavulu i tena ku tu landukisa .\n",
            "2020-02-04 21:23:08,661 Example #2\n",
            "2020-02-04 21:23:08,661 \tSource:     the apostle paul warned about what can happen if we please ourselves first .\n",
            "2020-02-04 21:23:08,661 \tReference:  ( 1 nzuá 2 : 16 ) o poxolo phaulu ua tu dimuna ku ima i tena kubhita se tu suua ngó o ima i tua uabhela .\n",
            "2020-02-04 21:23:08,661 \tHypothesis: o poxolo phaulu ua dimuna o ima i tu tena o kubhanga , se tu i sangulukisa o muxima ua nzambi .\n",
            "2020-02-04 21:23:08,661 Example #3\n",
            "2020-02-04 21:23:08,661 \tSource:     what a privilege it is to live in these last days and to be part of jehovah’s incredible organization !\n",
            "2020-02-04 21:23:08,661 \tReference:  tua tokala ku lembalala izuua ioso kuila , ujitu ua dikota ku tokala mu kilunga kia jihova mu ixi , mu izuua isukidila - ku !\n",
            "2020-02-04 21:23:08,661 \tHypothesis: ujitu ua dikota ku kala mu izuua íii isukidila - ku , ni ku bhanga mbandu ku kilunga kia jihova !\n",
            "2020-02-04 21:23:08,661 Validation result (greedy) at epoch  25, step    18000: bleu:  27.56, loss: 34467.0391, ppl:   4.6902, duration: 28.0627s\n",
            "2020-02-04 21:23:51,963 Epoch  25: total training loss 728.06\n",
            "2020-02-04 21:23:51,963 EPOCH 26\n",
            "2020-02-04 21:24:17,140 Epoch  26 Step:    18200 Batch Loss:     1.026593 Tokens per Sec:     6702, Lr: 0.000164\n",
            "2020-02-04 21:25:26,300 Epoch  26 Step:    18400 Batch Loss:     0.830519 Tokens per Sec:     6723, Lr: 0.000163\n",
            "2020-02-04 21:26:35,436 Epoch  26 Step:    18600 Batch Loss:     0.875107 Tokens per Sec:     6723, Lr: 0.000162\n",
            "2020-02-04 21:27:45,162 Epoch  26 Step:    18800 Batch Loss:     1.146535 Tokens per Sec:     6752, Lr: 0.000161\n",
            "2020-02-04 21:28:03,264 Epoch  26: total training loss 715.50\n",
            "2020-02-04 21:28:03,265 EPOCH 27\n",
            "2020-02-04 21:28:54,111 Epoch  27 Step:    19000 Batch Loss:     1.102405 Tokens per Sec:     6734, Lr: 0.000160\n",
            "2020-02-04 21:30:03,059 Epoch  27 Step:    19200 Batch Loss:     0.899686 Tokens per Sec:     6778, Lr: 0.000159\n",
            "2020-02-04 21:31:12,257 Epoch  27 Step:    19400 Batch Loss:     1.039293 Tokens per Sec:     6745, Lr: 0.000159\n",
            "2020-02-04 21:32:13,601 Epoch  27: total training loss 699.15\n",
            "2020-02-04 21:32:13,601 EPOCH 28\n",
            "2020-02-04 21:32:21,225 Epoch  28 Step:    19600 Batch Loss:     0.930306 Tokens per Sec:     6531, Lr: 0.000158\n",
            "2020-02-04 21:33:30,177 Epoch  28 Step:    19800 Batch Loss:     0.697419 Tokens per Sec:     6700, Lr: 0.000157\n",
            "2020-02-04 21:34:39,432 Epoch  28 Step:    20000 Batch Loss:     1.109810 Tokens per Sec:     6770, Lr: 0.000156\n",
            "2020-02-04 21:35:04,403 Hooray! New best validation result [eval_metric]!\n",
            "2020-02-04 21:35:04,404 Saving new checkpoint.\n",
            "2020-02-04 21:35:06,124 Example #0\n",
            "2020-02-04 21:35:06,124 \tSource:     ( read philippians 2 : 5 - 8 . )\n",
            "2020-02-04 21:35:06,124 \tReference:  ( tanga filipe 2 : 5 - 8 . )\n",
            "2020-02-04 21:35:06,124 \tHypothesis: ( tanga filipe 2 : 5 - 8 . )\n",
            "2020-02-04 21:35:06,124 Example #1\n",
            "2020-02-04 21:35:06,125 \tSource:     today , more than ever before , there are so many things that can distract us .\n",
            "2020-02-04 21:35:06,125 \tReference:  lelu , kuene ima iavulu i tena ku tu landukisa , m’ukulu ndenge .\n",
            "2020-02-04 21:35:06,125 \tHypothesis: lelu , ande dia ku bhita o ima , saí ima i tena ku tu landukisa .\n",
            "2020-02-04 21:35:06,125 Example #2\n",
            "2020-02-04 21:35:06,125 \tSource:     the apostle paul warned about what can happen if we please ourselves first .\n",
            "2020-02-04 21:35:06,125 \tReference:  ( 1 nzuá 2 : 16 ) o poxolo phaulu ua tu dimuna ku ima i tena kubhita se tu suua ngó o ima i tua uabhela .\n",
            "2020-02-04 21:35:06,125 \tHypothesis: o poxolo phaulu ua dimuna o ima i tua tokala o kubhanga , se tu dióndo tua - nda dióndo .\n",
            "2020-02-04 21:35:06,125 Example #3\n",
            "2020-02-04 21:35:06,125 \tSource:     what a privilege it is to live in these last days and to be part of jehovah’s incredible organization !\n",
            "2020-02-04 21:35:06,125 \tReference:  tua tokala ku lembalala izuua ioso kuila , ujitu ua dikota ku tokala mu kilunga kia jihova mu ixi , mu izuua isukidila - ku !\n",
            "2020-02-04 21:35:06,125 \tHypothesis: ujitu ua dikota ku kala mu izuua íii isukidila - ku , ni ku bhanga mbandu ku kilunga kia jihova ku mundu uoso !\n",
            "2020-02-04 21:35:06,125 Validation result (greedy) at epoch  28, step    20000: bleu:  27.83, loss: 34512.3594, ppl:   4.6997, duration: 26.6926s\n",
            "2020-02-04 21:36:15,361 Epoch  28 Step:    20200 Batch Loss:     1.001297 Tokens per Sec:     6779, Lr: 0.000155\n",
            "2020-02-04 21:36:51,103 Epoch  28: total training loss 685.74\n",
            "2020-02-04 21:36:51,104 EPOCH 29\n",
            "2020-02-04 21:37:24,494 Epoch  29 Step:    20400 Batch Loss:     0.780829 Tokens per Sec:     6656, Lr: 0.000155\n",
            "2020-02-04 21:38:34,375 Epoch  29 Step:    20600 Batch Loss:     1.058090 Tokens per Sec:     6789, Lr: 0.000154\n",
            "2020-02-04 21:39:43,651 Epoch  29 Step:    20800 Batch Loss:     1.022833 Tokens per Sec:     6713, Lr: 0.000153\n",
            "2020-02-04 21:40:52,962 Epoch  29 Step:    21000 Batch Loss:     1.021863 Tokens per Sec:     6729, Lr: 0.000152\n",
            "2020-02-04 21:41:02,519 Epoch  29: total training loss 672.60\n",
            "2020-02-04 21:41:02,520 EPOCH 30\n",
            "2020-02-04 21:42:02,234 Epoch  30 Step:    21200 Batch Loss:     1.052732 Tokens per Sec:     6722, Lr: 0.000152\n",
            "2020-02-04 21:43:11,251 Epoch  30 Step:    21400 Batch Loss:     0.706357 Tokens per Sec:     6738, Lr: 0.000151\n",
            "2020-02-04 21:44:20,417 Epoch  30 Step:    21600 Batch Loss:     1.044521 Tokens per Sec:     6757, Lr: 0.000150\n",
            "2020-02-04 21:45:12,997 Epoch  30: total training loss 660.55\n",
            "2020-02-04 21:45:12,998 EPOCH 31\n",
            "2020-02-04 21:45:29,078 Epoch  31 Step:    21800 Batch Loss:     0.909594 Tokens per Sec:     6664, Lr: 0.000150\n",
            "2020-02-04 21:46:38,634 Epoch  31 Step:    22000 Batch Loss:     1.052121 Tokens per Sec:     6702, Lr: 0.000149\n",
            "2020-02-04 21:47:04,968 Example #0\n",
            "2020-02-04 21:47:04,968 \tSource:     ( read philippians 2 : 5 - 8 . )\n",
            "2020-02-04 21:47:04,968 \tReference:  ( tanga filipe 2 : 5 - 8 . )\n",
            "2020-02-04 21:47:04,968 \tHypothesis: ( tanga filipe 2 : 5 - 8 . )\n",
            "2020-02-04 21:47:04,968 Example #1\n",
            "2020-02-04 21:47:04,968 \tSource:     today , more than ever before , there are so many things that can distract us .\n",
            "2020-02-04 21:47:04,968 \tReference:  lelu , kuene ima iavulu i tena ku tu landukisa , m’ukulu ndenge .\n",
            "2020-02-04 21:47:04,968 \tHypothesis: lelu , ande dia kubhita o ima ioso , sai ima iavulu i tena ku tu landukisa .\n",
            "2020-02-04 21:47:04,969 Example #2\n",
            "2020-02-04 21:47:04,969 \tSource:     the apostle paul warned about what can happen if we please ourselves first .\n",
            "2020-02-04 21:47:04,969 \tReference:  ( 1 nzuá 2 : 16 ) o poxolo phaulu ua tu dimuna ku ima i tena kubhita se tu suua ngó o ima i tua uabhela .\n",
            "2020-02-04 21:47:04,969 \tHypothesis: o poxolo phaulu ua dimuna o ima i tena ku bhita , se tu bhanga o ima i tua mesena .\n",
            "2020-02-04 21:47:04,969 Example #3\n",
            "2020-02-04 21:47:04,969 \tSource:     what a privilege it is to live in these last days and to be part of jehovah’s incredible organization !\n",
            "2020-02-04 21:47:04,969 \tReference:  tua tokala ku lembalala izuua ioso kuila , ujitu ua dikota ku tokala mu kilunga kia jihova mu ixi , mu izuua isukidila - ku !\n",
            "2020-02-04 21:47:04,969 \tHypothesis: ujitu ua dikota ku kala mu izuua íii isukidila - ku , ni ku bhanga mbandu ku kilunga kia jihova !\n",
            "2020-02-04 21:47:04,969 Validation result (greedy) at epoch  31, step    22000: bleu:  27.66, loss: 34950.4023, ppl:   4.7929, duration: 26.3353s\n",
            "2020-02-04 21:48:13,895 Epoch  31 Step:    22200 Batch Loss:     0.803242 Tokens per Sec:     6752, Lr: 0.000148\n",
            "2020-02-04 21:49:23,200 Epoch  31 Step:    22400 Batch Loss:     1.006292 Tokens per Sec:     6735, Lr: 0.000148\n",
            "2020-02-04 21:49:50,363 Epoch  31: total training loss 649.05\n",
            "2020-02-04 21:49:50,363 EPOCH 32\n",
            "2020-02-04 21:50:32,894 Epoch  32 Step:    22600 Batch Loss:     1.001542 Tokens per Sec:     6767, Lr: 0.000147\n",
            "2020-02-04 21:51:41,897 Epoch  32 Step:    22800 Batch Loss:     1.021521 Tokens per Sec:     6709, Lr: 0.000146\n",
            "2020-02-04 21:52:51,202 Epoch  32 Step:    23000 Batch Loss:     1.034444 Tokens per Sec:     6747, Lr: 0.000146\n",
            "2020-02-04 21:54:00,508 Epoch  32 Step:    23200 Batch Loss:     0.662240 Tokens per Sec:     6760, Lr: 0.000145\n",
            "2020-02-04 21:54:01,151 Epoch  32: total training loss 636.39\n",
            "2020-02-04 21:54:01,151 EPOCH 33\n",
            "2020-02-04 21:55:09,802 Epoch  33 Step:    23400 Batch Loss:     0.973105 Tokens per Sec:     6725, Lr: 0.000144\n",
            "2020-02-04 21:56:19,202 Epoch  33 Step:    23600 Batch Loss:     0.733882 Tokens per Sec:     6752, Lr: 0.000144\n",
            "2020-02-04 21:57:27,851 Epoch  33 Step:    23800 Batch Loss:     0.902763 Tokens per Sec:     6733, Lr: 0.000143\n",
            "2020-02-04 21:58:11,830 Epoch  33: total training loss 626.07\n",
            "2020-02-04 21:58:11,830 EPOCH 34\n",
            "2020-02-04 21:58:37,167 Epoch  34 Step:    24000 Batch Loss:     0.886888 Tokens per Sec:     6783, Lr: 0.000143\n",
            "2020-02-04 21:59:03,626 Example #0\n",
            "2020-02-04 21:59:03,626 \tSource:     ( read philippians 2 : 5 - 8 . )\n",
            "2020-02-04 21:59:03,626 \tReference:  ( tanga filipe 2 : 5 - 8 . )\n",
            "2020-02-04 21:59:03,626 \tHypothesis: ( tanga filipe 2 : 5 - 8 . )\n",
            "2020-02-04 21:59:03,627 Example #1\n",
            "2020-02-04 21:59:03,627 \tSource:     today , more than ever before , there are so many things that can distract us .\n",
            "2020-02-04 21:59:03,627 \tReference:  lelu , kuene ima iavulu i tena ku tu landukisa , m’ukulu ndenge .\n",
            "2020-02-04 21:59:03,627 \tHypothesis: lelu , ande dia ima iavulu ku tu landukisa .\n",
            "2020-02-04 21:59:03,627 Example #2\n",
            "2020-02-04 21:59:03,627 \tSource:     the apostle paul warned about what can happen if we please ourselves first .\n",
            "2020-02-04 21:59:03,627 \tReference:  ( 1 nzuá 2 : 16 ) o poxolo phaulu ua tu dimuna ku ima i tena kubhita se tu suua ngó o ima i tua uabhela .\n",
            "2020-02-04 21:59:03,627 \tHypothesis: o poxolo phaulu ua dimuna o ima i tu tena o kubhanga , se tu bhanga o ima i tua mesena ia tu sangulukisa .\n",
            "2020-02-04 21:59:03,627 Example #3\n",
            "2020-02-04 21:59:03,627 \tSource:     what a privilege it is to live in these last days and to be part of jehovah’s incredible organization !\n",
            "2020-02-04 21:59:03,627 \tReference:  tua tokala ku lembalala izuua ioso kuila , ujitu ua dikota ku tokala mu kilunga kia jihova mu ixi , mu izuua isukidila - ku !\n",
            "2020-02-04 21:59:03,628 \tHypothesis: ujitu ua dikota ku kala mu izuua íii isukidila - ku , ni ku bhanga mbandu ku kilunga kia jihova mu ngongo ioso !\n",
            "2020-02-04 21:59:03,628 Validation result (greedy) at epoch  34, step    24000: bleu:  27.81, loss: 35239.2383, ppl:   4.8554, duration: 26.4609s\n",
            "2020-02-04 22:00:12,651 Epoch  34 Step:    24200 Batch Loss:     0.735904 Tokens per Sec:     6769, Lr: 0.000142\n",
            "2020-02-04 22:01:21,751 Epoch  34 Step:    24400 Batch Loss:     0.764096 Tokens per Sec:     6725, Lr: 0.000141\n",
            "2020-02-04 22:02:31,084 Epoch  34 Step:    24600 Batch Loss:     0.675966 Tokens per Sec:     6724, Lr: 0.000141\n",
            "2020-02-04 22:02:48,997 Epoch  34: total training loss 615.29\n",
            "2020-02-04 22:02:48,997 EPOCH 35\n",
            "2020-02-04 22:03:40,033 Epoch  35 Step:    24800 Batch Loss:     0.966231 Tokens per Sec:     6778, Lr: 0.000140\n",
            "2020-02-04 22:04:49,458 Epoch  35 Step:    25000 Batch Loss:     1.004890 Tokens per Sec:     6685, Lr: 0.000140\n",
            "2020-02-04 22:05:58,423 Epoch  35 Step:    25200 Batch Loss:     0.799023 Tokens per Sec:     6731, Lr: 0.000139\n",
            "2020-02-04 22:06:59,902 Epoch  35: total training loss 604.89\n",
            "2020-02-04 22:06:59,902 EPOCH 36\n",
            "2020-02-04 22:07:07,737 Epoch  36 Step:    25400 Batch Loss:     0.851759 Tokens per Sec:     6783, Lr: 0.000139\n",
            "2020-02-04 22:08:16,531 Epoch  36 Step:    25600 Batch Loss:     0.867299 Tokens per Sec:     6737, Lr: 0.000138\n",
            "2020-02-04 22:09:25,729 Epoch  36 Step:    25800 Batch Loss:     0.928555 Tokens per Sec:     6769, Lr: 0.000138\n",
            "2020-02-04 22:10:34,258 Epoch  36 Step:    26000 Batch Loss:     0.984406 Tokens per Sec:     6720, Lr: 0.000137\n",
            "2020-02-04 22:11:00,673 Hooray! New best validation result [eval_metric]!\n",
            "2020-02-04 22:11:00,673 Saving new checkpoint.\n",
            "2020-02-04 22:11:02,345 Example #0\n",
            "2020-02-04 22:11:02,345 \tSource:     ( read philippians 2 : 5 - 8 . )\n",
            "2020-02-04 22:11:02,345 \tReference:  ( tanga filipe 2 : 5 - 8 . )\n",
            "2020-02-04 22:11:02,346 \tHypothesis: ( tanga filipe 2 : 5 - 8 . )\n",
            "2020-02-04 22:11:02,346 Example #1\n",
            "2020-02-04 22:11:02,346 \tSource:     today , more than ever before , there are so many things that can distract us .\n",
            "2020-02-04 22:11:02,346 \tReference:  lelu , kuene ima iavulu i tena ku tu landukisa , m’ukulu ndenge .\n",
            "2020-02-04 22:11:02,346 \tHypothesis: lelu , m’ukulu , kuene ima iavulu i tena ku tu landukisa .\n",
            "2020-02-04 22:11:02,346 Example #2\n",
            "2020-02-04 22:11:02,346 \tSource:     the apostle paul warned about what can happen if we please ourselves first .\n",
            "2020-02-04 22:11:02,346 \tReference:  ( 1 nzuá 2 : 16 ) o poxolo phaulu ua tu dimuna ku ima i tena kubhita se tu suua ngó o ima i tua uabhela .\n",
            "2020-02-04 22:11:02,347 \tHypothesis: o poxolo phaulu ua dimuna o ima i tu tena kubhanga , se tu dióndo dianga kusota o kitambuijilu kia poxolo .\n",
            "2020-02-04 22:11:02,347 Example #3\n",
            "2020-02-04 22:11:02,347 \tSource:     what a privilege it is to live in these last days and to be part of jehovah’s incredible organization !\n",
            "2020-02-04 22:11:02,347 \tReference:  tua tokala ku lembalala izuua ioso kuila , ujitu ua dikota ku tokala mu kilunga kia jihova mu ixi , mu izuua isukidila - ku !\n",
            "2020-02-04 22:11:02,347 \tHypothesis: ujitu ua dikota ku kala mu izuua íii isukidila - ku , ni ku bhanga mbandu ku kilunga kia jihova mu kilunga kiê !\n",
            "2020-02-04 22:11:02,347 Validation result (greedy) at epoch  36, step    26000: bleu:  28.57, loss: 35323.6875, ppl:   4.8738, duration: 28.0886s\n",
            "2020-02-04 22:11:38,789 Epoch  36: total training loss 595.59\n",
            "2020-02-04 22:11:38,789 EPOCH 37\n",
            "2020-02-04 22:12:11,812 Epoch  37 Step:    26200 Batch Loss:     0.837748 Tokens per Sec:     6688, Lr: 0.000137\n",
            "2020-02-04 22:13:20,839 Epoch  37 Step:    26400 Batch Loss:     0.885409 Tokens per Sec:     6753, Lr: 0.000136\n",
            "2020-02-04 22:14:29,896 Epoch  37 Step:    26600 Batch Loss:     0.945491 Tokens per Sec:     6753, Lr: 0.000135\n",
            "2020-02-04 22:15:38,565 Epoch  37 Step:    26800 Batch Loss:     0.869143 Tokens per Sec:     6750, Lr: 0.000135\n",
            "2020-02-04 22:15:49,475 Epoch  37: total training loss 585.76\n",
            "2020-02-04 22:15:49,475 EPOCH 38\n",
            "2020-02-04 22:16:48,111 Epoch  38 Step:    27000 Batch Loss:     0.751874 Tokens per Sec:     6739, Lr: 0.000134\n",
            "2020-02-04 22:17:57,532 Epoch  38 Step:    27200 Batch Loss:     0.771263 Tokens per Sec:     6784, Lr: 0.000134\n",
            "2020-02-04 22:19:06,625 Epoch  38 Step:    27400 Batch Loss:     0.903140 Tokens per Sec:     6732, Lr: 0.000133\n",
            "2020-02-04 22:19:59,952 Epoch  38: total training loss 574.80\n",
            "2020-02-04 22:19:59,952 EPOCH 39\n",
            "2020-02-04 22:20:15,691 Epoch  39 Step:    27600 Batch Loss:     0.824561 Tokens per Sec:     6609, Lr: 0.000133\n",
            "2020-02-04 22:21:24,953 Epoch  39 Step:    27800 Batch Loss:     0.907379 Tokens per Sec:     6778, Lr: 0.000133\n",
            "2020-02-04 22:22:33,948 Epoch  39 Step:    28000 Batch Loss:     0.561887 Tokens per Sec:     6718, Lr: 0.000132\n",
            "2020-02-04 22:22:59,746 Example #0\n",
            "2020-02-04 22:22:59,747 \tSource:     ( read philippians 2 : 5 - 8 . )\n",
            "2020-02-04 22:22:59,747 \tReference:  ( tanga filipe 2 : 5 - 8 . )\n",
            "2020-02-04 22:22:59,747 \tHypothesis: ( tanga filipe 2 : 5 - 8 . )\n",
            "2020-02-04 22:22:59,747 Example #1\n",
            "2020-02-04 22:22:59,747 \tSource:     today , more than ever before , there are so many things that can distract us .\n",
            "2020-02-04 22:22:59,747 \tReference:  lelu , kuene ima iavulu i tena ku tu landukisa , m’ukulu ndenge .\n",
            "2020-02-04 22:22:59,747 \tHypothesis: lelu , ande dia ima iavulu ku bhita , saí ima iavulu i tena ku tu landukisa .\n",
            "2020-02-04 22:22:59,747 Example #2\n",
            "2020-02-04 22:22:59,747 \tSource:     the apostle paul warned about what can happen if we please ourselves first .\n",
            "2020-02-04 22:22:59,747 \tReference:  ( 1 nzuá 2 : 16 ) o poxolo phaulu ua tu dimuna ku ima i tena kubhita se tu suua ngó o ima i tua uabhela .\n",
            "2020-02-04 22:22:59,747 \tHypothesis: o poxolo phaulu ua dimuna o ima i tena kubhita , se tu bhanga o ima i tua mesena .\n",
            "2020-02-04 22:22:59,747 Example #3\n",
            "2020-02-04 22:22:59,748 \tSource:     what a privilege it is to live in these last days and to be part of jehovah’s incredible organization !\n",
            "2020-02-04 22:22:59,748 \tReference:  tua tokala ku lembalala izuua ioso kuila , ujitu ua dikota ku tokala mu kilunga kia jihova mu ixi , mu izuua isukidila - ku !\n",
            "2020-02-04 22:22:59,748 \tHypothesis: ujitu ua dikota ku kala mu izuua íii isukidila - ku , ni ku bhanga mbandu ku kilunga kia jihova !\n",
            "2020-02-04 22:22:59,748 Validation result (greedy) at epoch  39, step    28000: bleu:  27.94, loss: 35734.3711, ppl:   4.9644, duration: 25.7991s\n",
            "2020-02-04 22:24:09,053 Epoch  39 Step:    28200 Batch Loss:     0.667817 Tokens per Sec:     6729, Lr: 0.000132\n",
            "2020-02-04 22:24:36,539 Epoch  39: total training loss 566.02\n",
            "2020-02-04 22:24:36,539 EPOCH 40\n",
            "2020-02-04 22:25:18,384 Epoch  40 Step:    28400 Batch Loss:     0.907208 Tokens per Sec:     6721, Lr: 0.000131\n",
            "2020-02-04 22:26:27,322 Epoch  40 Step:    28600 Batch Loss:     0.836263 Tokens per Sec:     6749, Lr: 0.000131\n",
            "2020-02-04 22:27:36,618 Epoch  40 Step:    28800 Batch Loss:     0.577048 Tokens per Sec:     6768, Lr: 0.000130\n",
            "2020-02-04 22:28:45,588 Epoch  40 Step:    29000 Batch Loss:     0.725040 Tokens per Sec:     6770, Lr: 0.000130\n",
            "2020-02-04 22:28:46,906 Epoch  40: total training loss 557.25\n",
            "2020-02-04 22:28:46,907 Training ended after  40 epochs.\n",
            "2020-02-04 22:28:46,907 Best validation result (greedy) at step    26000:  28.57 eval_metric.\n",
            "2020-02-04 22:29:27,124  dev bleu:  28.81 [Beam search decoding with beam size = 5 and alpha = 1.0]\n",
            "2020-02-04 22:29:27,125 Translations saved to: models/enkmb_transformer/00026000.hyps.dev\n",
            "2020-02-04 22:31:02,627 test bleu:  32.76 [Beam search decoding with beam size = 5 and alpha = 1.0]\n",
            "2020-02-04 22:31:02,629 Translations saved to: models/enkmb_transformer/00026000.hyps.test\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "EgjqnSBFn2eh",
        "colab_type": "code",
        "outputId": "2d516f75-cac4-43ff-fbc8-27b40f508ea4",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 260
        }
      },
      "source": [
        "! cat joeynmt/models/enkmb_transformer/validations.txt"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Steps: 2000\tLoss: 47097.83984\tPPL: 8.26328\tbleu: 14.85684\tLR: 0.00049411\t*\n",
            "Steps: 4000\tLoss: 39856.20312\tPPL: 5.97219\tbleu: 20.91001\tLR: 0.00034939\t*\n",
            "Steps: 6000\tLoss: 37121.82812\tPPL: 5.28307\tbleu: 23.80892\tLR: 0.00028527\t*\n",
            "Steps: 8000\tLoss: 35674.49609\tPPL: 4.95110\tbleu: 25.00066\tLR: 0.00024705\t*\n",
            "Steps: 10000\tLoss: 34720.73828\tPPL: 4.74383\tbleu: 26.10309\tLR: 0.00022097\t*\n",
            "Steps: 12000\tLoss: 34417.08594\tPPL: 4.67967\tbleu: 26.15923\tLR: 0.00020172\t*\n",
            "Steps: 14000\tLoss: 34284.73047\tPPL: 4.65198\tbleu: 27.20639\tLR: 0.00018675\t*\n",
            "Steps: 16000\tLoss: 34352.42969\tPPL: 4.66613\tbleu: 27.29388\tLR: 0.00017469\t*\n",
            "Steps: 18000\tLoss: 34467.03906\tPPL: 4.69017\tbleu: 27.56135\tLR: 0.00016470\t*\n",
            "Steps: 20000\tLoss: 34512.35938\tPPL: 4.69971\tbleu: 27.82828\tLR: 0.00015625\t*\n",
            "Steps: 22000\tLoss: 34950.40234\tPPL: 4.79293\tbleu: 27.65960\tLR: 0.00014898\t\n",
            "Steps: 24000\tLoss: 35239.23828\tPPL: 4.85541\tbleu: 27.80829\tLR: 0.00014264\t\n",
            "Steps: 26000\tLoss: 35323.68750\tPPL: 4.87383\tbleu: 28.57189\tLR: 0.00013704\t*\n",
            "Steps: 28000\tLoss: 35734.37109\tPPL: 4.96441\tbleu: 27.94465\tLR: 0.00013206\t\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9NOP9uXVsMDi",
        "colab_type": "code",
        "outputId": "f649ef87-c285-4c4b-b7cb-f44c2251781d",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 69
        }
      },
      "source": [
        "! cd joeynmt; python3 -m joeynmt test models/enkmb_transformer/config.yaml "
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "2020-02-04 22:38:57,549 Hello! This is Joey-NMT.\n",
            "2020-02-04 22:39:38,351  dev bleu:  28.81 [Beam search decoding with beam size = 5 and alpha = 1.0]\n",
            "2020-02-04 22:41:09,183 test bleu:  32.76 [Beam search decoding with beam size = 5 and alpha = 1.0]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "NItstdy5XK8i",
        "colab_type": "code",
        "outputId": "fac74fb9-9561-44a6-8eed-8166ad3d7f42",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 124
        }
      },
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive')"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Go to this URL in a browser: https://accounts.google.com/o/oauth2/auth?client_id=947318989803-6bn6qk8qdgf4n4g3pfee6491hc0brc4i.apps.googleusercontent.com&redirect_uri=urn%3aietf%3awg%3aoauth%3a2.0%3aoob&response_type=code&scope=email%20https%3a%2f%2fwww.googleapis.com%2fauth%2fdocs.test%20https%3a%2f%2fwww.googleapis.com%2fauth%2fdrive%20https%3a%2f%2fwww.googleapis.com%2fauth%2fdrive.photos.readonly%20https%3a%2f%2fwww.googleapis.com%2fauth%2fpeopleapi.readonly\n",
            "\n",
            "Enter your authorization code:\n",
            "··········\n",
            "Mounted at /content/drive\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0LXxVAOVp9Y8",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "!mkdir -p /content/drive/My\\ Drive/masakhane/kmb"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8C1ZqQU3XqWM",
        "colab_type": "code",
        "outputId": "590d6509-f9da-406f-ee9b-66c8ac3a1105",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 54
        }
      },
      "source": [
        "!cp -r joeynmt/models/enkmb_transformer /content/drive/My\\ Drive/masakhane/kmb/"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "cp: cannot create symbolic link '/content/drive/My Drive/masakhane/kmb/enkmb_transformer/best.ckpt': Operation not supported\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nS2NiVSMX5bB",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}