English_To_isiZulu.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "name": "Copy_of_English_To_isiZulu.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "toc_visible": true,
      "machine_shape": "hm",
      "include_colab_link": true
    },
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/dinoanasta/English-To-isiZulu/blob/main/English_To_isiZulu.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "kbJeAod3d7PD"
      },
      "source": [
        "# English to isiZulu Machine Translation\n",
        "## COMS4054A - Natural Language Processing\n",
        "### Group Members:\n",
        "Jesse Bristow (1875955)<br>\n",
        "Chloë Smith (1877342)<br>\n",
        "Dino Anastasopoulos (1900661)<br>\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "VhLI87c7fMZ0"
      },
      "source": [
        "# Setup and Imports"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ndE20vfiT_5U",
        "outputId": "7467649c-ba90-4bdd-b332-694ff9e64de4"
      },
      "source": [
        "!pip install transformers"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Collecting transformers\n",
            "  Downloading transformers-4.12.5-py3-none-any.whl (3.1 MB)\n",
            "\u001b[K     |████████████████████████████████| 3.1 MB 5.2 MB/s \n",
            "\u001b[?25hRequirement already satisfied: numpy>=1.17 in /usr/local/lib/python3.7/dist-packages (from transformers) (1.19.5)\n",
            "Requirement already satisfied: requests in /usr/local/lib/python3.7/dist-packages (from transformers) (2.23.0)\n",
            "Requirement already satisfied: filelock in /usr/local/lib/python3.7/dist-packages (from transformers) (3.4.0)\n",
            "Collecting tokenizers<0.11,>=0.10.1\n",
            "  Downloading tokenizers-0.10.3-cp37-cp37m-manylinux_2_5_x86_64.manylinux1_x86_64.manylinux_2_12_x86_64.manylinux2010_x86_64.whl (3.3 MB)\n",
            "\u001b[K     |████████████████████████████████| 3.3 MB 47.4 MB/s \n",
            "\u001b[?25hCollecting pyyaml>=5.1\n",
            "  Downloading PyYAML-6.0-cp37-cp37m-manylinux_2_5_x86_64.manylinux1_x86_64.manylinux_2_12_x86_64.manylinux2010_x86_64.whl (596 kB)\n",
            "\u001b[K     |████████████████████████████████| 596 kB 54.1 MB/s \n",
            "\u001b[?25hRequirement already satisfied: importlib-metadata in /usr/local/lib/python3.7/dist-packages (from transformers) (4.8.2)\n",
            "Collecting huggingface-hub<1.0,>=0.1.0\n",
            "  Downloading huggingface_hub-0.1.2-py3-none-any.whl (59 kB)\n",
            "\u001b[K     |████████████████████████████████| 59 kB 6.8 MB/s \n",
            "\u001b[?25hRequirement already satisfied: packaging>=20.0 in /usr/local/lib/python3.7/dist-packages (from transformers) (21.3)\n",
            "Requirement already satisfied: regex!=2019.12.17 in /usr/local/lib/python3.7/dist-packages (from transformers) (2019.12.20)\n",
            "Collecting sacremoses\n",
            "  Downloading sacremoses-0.0.46-py3-none-any.whl (895 kB)\n",
            "\u001b[K     |████████████████████████████████| 895 kB 58.0 MB/s \n",
            "\u001b[?25hRequirement already satisfied: tqdm>=4.27 in /usr/local/lib/python3.7/dist-packages (from transformers) (4.62.3)\n",
            "Requirement already satisfied: typing-extensions>=3.7.4.3 in /usr/local/lib/python3.7/dist-packages (from huggingface-hub<1.0,>=0.1.0->transformers) (3.10.0.2)\n",
            "Requirement already satisfied: pyparsing!=3.0.5,>=2.0.2 in /usr/local/lib/python3.7/dist-packages (from packaging>=20.0->transformers) (3.0.6)\n",
            "Requirement already satisfied: zipp>=0.5 in /usr/local/lib/python3.7/dist-packages (from importlib-metadata->transformers) (3.6.0)\n",
            "Requirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.7/dist-packages (from requests->transformers) (3.0.4)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.7/dist-packages (from requests->transformers) (2021.10.8)\n",
            "Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.7/dist-packages (from requests->transformers) (1.24.3)\n",
            "Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.7/dist-packages (from requests->transformers) (2.10)\n",
            "Requirement already satisfied: click in /usr/local/lib/python3.7/dist-packages (from sacremoses->transformers) (7.1.2)\n",
            "Requirement already satisfied: joblib in /usr/local/lib/python3.7/dist-packages (from sacremoses->transformers) (1.1.0)\n",
            "Requirement already satisfied: six in /usr/local/lib/python3.7/dist-packages (from sacremoses->transformers) (1.15.0)\n",
            "Installing collected packages: pyyaml, tokenizers, sacremoses, huggingface-hub, transformers\n",
            "  Attempting uninstall: pyyaml\n",
            "    Found existing installation: PyYAML 3.13\n",
            "    Uninstalling PyYAML-3.13:\n",
            "      Successfully uninstalled PyYAML-3.13\n",
            "Successfully installed huggingface-hub-0.1.2 pyyaml-6.0 sacremoses-0.0.46 tokenizers-0.10.3 transformers-4.12.5\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 468
        },
        "id": "peeTbU1KZnRR",
        "outputId": "5d683ce5-b2d1-4663-d6a4-7ab20fea7b09"
      },
      "source": [
        "!pip install  matplotlib==3.4.2 --upgrade"
      ],
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Collecting matplotlib==3.4.2\n",
            "  Downloading matplotlib-3.4.2-cp37-cp37m-manylinux1_x86_64.whl (10.3 MB)\n",
            "\u001b[K     |████████████████████████████████| 10.3 MB 4.6 MB/s \n",
            "\u001b[?25hRequirement already satisfied: numpy>=1.16 in /usr/local/lib/python3.7/dist-packages (from matplotlib==3.4.2) (1.19.5)\n",
            "Requirement already satisfied: pyparsing>=2.2.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib==3.4.2) (3.0.6)\n",
            "Requirement already satisfied: pillow>=6.2.0 in /usr/local/lib/python3.7/dist-packages (from matplotlib==3.4.2) (7.1.2)\n",
            "Requirement already satisfied: kiwisolver>=1.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib==3.4.2) (1.3.2)\n",
            "Requirement already satisfied: cycler>=0.10 in /usr/local/lib/python3.7/dist-packages (from matplotlib==3.4.2) (0.11.0)\n",
            "Requirement already satisfied: python-dateutil>=2.7 in /usr/local/lib/python3.7/dist-packages (from matplotlib==3.4.2) (2.8.2)\n",
            "Requirement already satisfied: six>=1.5 in /usr/local/lib/python3.7/dist-packages (from python-dateutil>=2.7->matplotlib==3.4.2) (1.15.0)\n",
            "Installing collected packages: matplotlib\n",
            "  Attempting uninstall: matplotlib\n",
            "    Found existing installation: matplotlib 3.2.2\n",
            "    Uninstalling matplotlib-3.2.2:\n",
            "      Successfully uninstalled matplotlib-3.2.2\n",
            "\u001b[31mERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.\n",
            "albumentations 0.1.12 requires imgaug<0.2.7,>=0.2.5, but you have imgaug 0.2.9 which is incompatible.\u001b[0m\n",
            "Successfully installed matplotlib-3.4.2\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "application/vnd.colab-display-data+json": {
              "pip_warning": {
                "packages": [
                  "matplotlib",
                  "mpl_toolkits"
                ]
              }
            }
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "TaVKiz8eDwuZ"
      },
      "source": [
        "import collections\n",
        "from collections import defaultdict\n",
        "\n",
        "import helper\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "import math\n",
        "import os\n",
        "import cv2\n",
        "import random\n",
        "import skimage\n",
        "from skimage.color import rgb2gray\n",
        "from skimage import img_as_float\n",
        "import matplotlib.pyplot as plt\n",
        "import scipy.stats as ss\n",
        "from sklearn.model_selection import KFold\n",
        "\n",
        "from sklearn.model_selection import train_test_split\n",
        "\n",
        "from tensorflow import keras\n",
        "\n",
        "from keras.preprocessing.text import Tokenizer\n",
        "from keras.preprocessing.sequence import pad_sequences\n",
        "from keras.models import Model, Sequential\n",
        "from keras.layers import GRU, Input, Dense, TimeDistributed, Activation, RepeatVector, Bidirectional, Dropout, LSTM\n",
        "from keras.layers.embeddings import Embedding\n",
        "#from keras.optimizers import Adam\n",
        "from keras.losses import sparse_categorical_crossentropy\n",
        "\n",
        "from tokenizers import Tokenizer as Toke\n",
        "from tokenizers.models import BPE\n",
        "from tokenizers.trainers import BpeTrainer\n",
        "from tokenizers.pre_tokenizers import Whitespace\n",
        "\n",
        "from nltk.translate.bleu_score import corpus_bleu\n",
        "\n",
        "import warnings\n",
        "warnings.filterwarnings('ignore')"
      ],
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-zROixHqZqTB"
      },
      "source": [
        "NUM_EPOCHS = 200"
      ],
      "execution_count": 4,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "DBG79J0GfQRg"
      },
      "source": [
        "# Data Ingestion and Preprocessing"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Wgaz7hOXiE4w",
        "outputId": "027776ba-b2c9-41cb-aedc-78063097f84c"
      },
      "source": [
        "!wget https://zenodo.org/record/5035171/files/en-zu.eval.csv\n",
        "!wget https://zenodo.org/record/5035171/files/en-zu.training.csv"
      ],
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "--2021-11-25 18:03:13--  https://zenodo.org/record/5035171/files/en-zu.eval.csv\n",
            "Resolving zenodo.org (zenodo.org)... 137.138.76.77\n",
            "Connecting to zenodo.org (zenodo.org)|137.138.76.77|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 466677 (456K) [text/plain]\n",
            "Saving to: ‘en-zu.eval.csv’\n",
            "\n",
            "en-zu.eval.csv      100%[===================>] 455.74K  1.03MB/s    in 0.4s    \n",
            "\n",
            "2021-11-25 18:03:15 (1.03 MB/s) - ‘en-zu.eval.csv’ saved [466677/466677]\n",
            "\n",
            "--2021-11-25 18:03:15--  https://zenodo.org/record/5035171/files/en-zu.training.csv\n",
            "Resolving zenodo.org (zenodo.org)... 137.138.76.77\n",
            "Connecting to zenodo.org (zenodo.org)|137.138.76.77|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 1529050 (1.5M) [text/plain]\n",
            "Saving to: ‘en-zu.training.csv’\n",
            "\n",
            "en-zu.training.csv  100%[===================>]   1.46M  2.25MB/s    in 0.6s    \n",
            "\n",
            "2021-11-25 18:03:17 (2.25 MB/s) - ‘en-zu.training.csv’ saved [1529050/1529050]\n",
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "40VfR7poDvZP",
        "outputId": "3ddd9a24-52a6-47c6-c554-dd9ac2d8c821"
      },
      "source": [
        "df_train = pd.read_csv('en-zu.training.csv')\n",
        "print(df_train.head(3).columns)\n",
        "df_eval = pd.read_csv('en-zu.eval.csv')\n",
        "print(df_eval.head(3).columns)"
      ],
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Index(['en', 'zu', 'source'], dtype='object')\n",
            "Index(['en', 'zu', 'zu.1', 'Source'], dtype='object')\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "AOo6aSh9EMvh",
        "outputId": "db0309b9-0c41-4acc-abda-318ff0ad7daa"
      },
      "source": [
        "print(df_train['en'], df_train['zu'])\n",
        "english_sentences = df_train['en']\n",
        "zulu_sentences = df_train['zu']\n",
        "\n",
        "for i in range(len(english_sentences)):\n",
        "  english_sentences[i] = english_sentences[i].lower()\n",
        "  zulu_sentences[i] = zulu_sentences[i].lower()"
      ],
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "0       His solo albums in the 60s were some of the mo...\n",
            "1       From which it follows that if you want to save...\n",
            "2       Thirty years after St. John Paul II begged Moz...\n",
            "3       Wynford yw sylfaenydd Stafell Fyw yng Nghaerdy...\n",
            "4       Unai Emery's side have won just two of their l...\n",
            "                              ...                        \n",
            "4734    The picture shows the infant breastfeeding whi...\n",
            "4735    Blame the producers, who constantly switch and...\n",
            "4736                              (Javier Garcia/BPI/REX)\n",
            "4737    \"I have enormous respect for Laura, She does a...\n",
            "4738    But despite the chaotic circumstances of the s...\n",
            "Name: en, Length: 4739, dtype: object 0       Ama-albhamu akhe e-solo eminyaka yawo-60 angam...\n",
            "1       Ukusuka lapho kuzobe sekulandela ukuthi uma uf...\n",
            "2       Ngemuva kweminyaka engamashumi amathathu u-St....\n",
            "3       Wynford yw sylfaenydd Stafell Fyw yng Nghaerdy...\n",
            "4       Iqembu lika-Unai Emery lisanqobe imidlalo emib...\n",
            "                              ...                        \n",
            "4734    Isithombe sibonisa ingane encane encelayo nges...\n",
            "4735    Sola abakhiqizi, abaguqula njalo babeke endawe...\n",
            "4736                        (U-Javier Garcia/i-BPI/i-REX)\n",
            "4737    \"Ngimhlonipha kakhulu u- Laura, Wenza umsebenz...\n",
            "4738    Kodwa nanoma bekunezimo zokuvalwa ezibuxakalal...\n",
            "Name: zu, Length: 4739, dtype: object\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "UuOloxbRTxSr",
        "outputId": "0f76fa7b-58e2-4060-d5d1-b03ee7d0cbe9"
      },
      "source": [
        "# BPETokenizer training\n",
        "english_vocab_size_bpe = 10000\n",
        "zulu_vocab_size_bpe = 10000\n",
        "\n",
        "tokenizer_en = Toke(BPE(unk_token=\"[UNK]\"))\n",
        "tokenizer_zu = Toke(BPE(unk_token=\"[UNK]\"))\n",
        "\n",
        "trainer_en = BpeTrainer(vocab_size = english_vocab_size_bpe, special_tokens=[\"[UNK]\", \"[CLS]\", \"[SEP]\", \"[PAD]\", \"[MASK]\"])\n",
        "trainer_zu = BpeTrainer(vocab_size = zulu_vocab_size_bpe, special_tokens=[\"[UNK]\", \"[CLS]\", \"[SEP]\", \"[PAD]\", \"[MASK]\"])\n",
        "\n",
        "tokenizer_en.pre_tokenizer = Whitespace()\n",
        "tokenizer_zu.pre_tokenizer = Whitespace()\n",
        "\n",
        "#files = [f\"data/wikitext-103-raw/wiki.{split}.raw\" for split in [\"test\", \"train\", \"valid\"]]\n",
        "tokenizer_en.train_from_iterator(english_sentences, trainer_en)\n",
        "tokenizer_zu.train_from_iterator(zulu_sentences, trainer_zu)\n",
        "\n",
        "output = tokenizer_en.encode((\"Hello, y'all! How are you 😁 ?\").lower())\n",
        "output1 = tokenizer_en.encode((\"abukho ubuholi obungasusa abaphikisana nama-semites.\").lower())\n",
        "print(output.tokens)\n",
        "print(output1.tokens)\n",
        "outputz = tokenizer_zu.encode((\"Hello, y'all! How are you 😁 ?\").lower())\n",
        "outputz1 = tokenizer_zu.encode((\"abukho ubuholi obungasusa abaphikisana nama-semites.\").lower())\n",
        "print(outputz.tokens)\n",
        "print(outputz1.tokens)\n",
        "print(tokenizer_en.token_to_id(\"hel\"))\n",
        "print(tokenizer_en.id_to_token(1))"
      ],
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "['hello', ',', 'y', \"'\", 'all', '!', 'how', 'are', 'you', '[UNK]', '?']\n",
            "['ab', 'uk', 'ho', 'u', 'bu', 'hol', 'i', 'ob', 'un', 'gas', 'usa', 'ab', 'ap', 'hik', 'is', 'ana', 'nam', 'a', '-', 'semit', 'es', '.']\n",
            "['hel', 'lo', ',', 'y', \"'\", 'all', '!', 'ho', 'w', 'are', 'you', '[UNK]', '?']\n",
            "['abukho', 'ubuholi', 'obunga', 'susa', 'abaphikisana', 'nama', '-', 'semi', 'tes', '.']\n",
            "511\n",
            "[CLS]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "WLg8NjIE5fNL",
        "outputId": "7f1962d8-f847-4597-acde-b2ac538da821"
      },
      "source": [
        "# Remove empty rows and maybe rows that have sequences of length longer than x\n",
        "length_counts_en = defaultdict(lambda: 0)\n",
        "length_counts_zu = defaultdict(lambda: 0)\n",
        "inds_to_remove = []\n",
        "for i in range(len(english_sentences)):\n",
        "  length_counts_en[len(english_sentences[i].split())] += 1\n",
        "  length_counts_zu[len(zulu_sentences[i].split())] += 1\n",
        "  if len(english_sentences[i].split()) > 20 or len(zulu_sentences[i].split()) > 20:\n",
        "    inds_to_remove.append(i)\n",
        "\n",
        "old_eng_sent = english_sentences\n",
        "old_zu_sent = zulu_sentences\n",
        "english_sentences = english_sentences.drop(inds_to_remove, axis=0)\n",
        "zulu_sentences = zulu_sentences.drop(inds_to_remove, axis=0)\n",
        "\n",
        "english_words_counter = collections.Counter([word for sentence in english_sentences for word in sentence.split()])\n",
        "zulu_words_counter = collections.Counter([word for sentence in zulu_sentences for word in sentence.split()])\n",
        "\n",
        "print('{} English words.'.format(len([word for sentence in english_sentences for word in sentence.split()])))\n",
        "print('{} unique English words.'.format(len(english_words_counter)))\n",
        "print('10 Most common words in the English dataset:')\n",
        "print('\"' + '\" \"'.join(list(zip(*english_words_counter.most_common(10)))[0]) + '\"')\n",
        "print()\n",
        "print('{} Zulu words.'.format(len([word for sentence in zulu_sentences for word in sentence.split()])))\n",
        "print('{} unique Zulu words.'.format(len(zulu_words_counter)))\n",
        "print('10 Most common words in the Zulu dataset:')\n",
        "print('\"' + '\" \"'.join(list(zip(*zulu_words_counter.most_common(10)))[0]) + '\"')"
      ],
      "execution_count": 9,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "27271 English words.\n",
            "9538 unique English words.\n",
            "10 Most common words in the English dataset:\n",
            "\"the\" \"to\" \"a\" \"and\" \"in\" \"of\" \"is\" \"for\" \"was\" \"on\"\n",
            "\n",
            "21288 Zulu words.\n",
            "14241 unique Zulu words.\n",
            "10 Most common words in the Zulu dataset:\n",
            "\"ukuthi\" \"futhi\" \"kakhulu\" \"noma\" \"kodwa\" \"uma\" \"kusho\" \"kanye\" \"-\" \"kube\"\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "zdnvixB1cHOj"
      },
      "source": [
        "# Functions"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "h2-OoeLB5RWx"
      },
      "source": [
        "def tokenize(x):\n",
        "    \"\"\"\n",
        "    Tokenize x\n",
        "    :param x: List of sentences/strings to be tokenized\n",
        "    :return: Tuple of (tokenized x data, tokenizer used to tokenize x)\n",
        "    \"\"\"\n",
        "    # TODO: Implement\n",
        "    tokenizer = Tokenizer()\n",
        "    tokenizer.fit_on_texts(x)\n",
        "    return tokenizer.texts_to_sequences(x), tokenizer\n",
        "\n",
        "def bpe_tokenize(x, tokenizer_to_use):\n",
        "  # return the bpe tokenization of each sentence as a sequence\n",
        "  all_seq = []\n",
        "  for curr_sent in x:\n",
        "    otp = tokenizer_to_use.encode(curr_sent)\n",
        "    curr_tokenization = otp.tokens\n",
        "    curr_seq = []\n",
        "    for curr_token in curr_tokenization:\n",
        "      curr_seq.append(tokenizer_to_use.token_to_id(curr_token))\n",
        "\n",
        "    all_seq.append(curr_seq)\n",
        "  \n",
        "  return all_seq"
      ],
      "execution_count": 10,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "eKbuF9xa8rhD"
      },
      "source": [
        "def pad(x, length=None):\n",
        "    \"\"\"\n",
        "    Pad x\n",
        "    :param x: List of sequences.\n",
        "    :param length: Length to pad the sequence to.  If None, use length of longest sequence in x.\n",
        "    :return: Padded numpy array of sequences\n",
        "    \"\"\"\n",
        "    # TODO: Implement\n",
        "    return pad_sequences(x, maxlen=length, padding='post')"
      ],
      "execution_count": 11,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "JDwhpfrC8wAZ",
        "outputId": "8d7c3097-3bf5-4a5c-8d21-4e4b63b2b8ce"
      },
      "source": [
        "def preprocess(x, y):\n",
        "    \"\"\"\n",
        "    Preprocess x and y\n",
        "    :param x: Feature List of sentences\n",
        "    :param y: Label List of sentences\n",
        "    :return: Tuple of (Preprocessed x, Preprocessed y, x tokenizer, y tokenizer)\n",
        "    \"\"\"\n",
        "    preprocess_x, x_tk = tokenize(x)\n",
        "    preprocess_y, y_tk = tokenize(y)\n",
        "\n",
        "    preprocess_x_bpe = bpe_tokenize(x, tokenizer_en)\n",
        "    preprocess_y_bpe = bpe_tokenize(y, tokenizer_zu)\n",
        "\n",
        "    #Print min and max values of zulu sentences\n",
        "    min_zu = 2000\n",
        "    max_zu = 0\n",
        "    for sample_i, (sent, token_sent) in enumerate(zip(y, preprocess_y_bpe)):\n",
        "      if len(token_sent) != 0:\n",
        "        min_zu = min(min_zu, min(token_sent))\n",
        "        max_zu = max(max_zu, max(token_sent))\n",
        "    print('Min:', min_zu, ' Max:', max_zu)\n",
        "\n",
        "\n",
        "    preprocess_x = pad(preprocess_x)\n",
        "    preprocess_y = pad(preprocess_y)\n",
        "\n",
        "    preprocess_x_bpe = pad(preprocess_x_bpe)\n",
        "    preprocess_y_bpe = pad(preprocess_y_bpe)\n",
        "\n",
        "    # Keras's sparse_categorical_crossentropy function requires the labels to be in 3 dimensions\n",
        "    preprocess_y = preprocess_y.reshape(*preprocess_y.shape, 1)\n",
        "    preprocess_y_bpe = preprocess_y_bpe.reshape(*preprocess_y_bpe.shape, 1)\n",
        "\n",
        "    return preprocess_x, preprocess_y, x_tk, y_tk, preprocess_x_bpe, preprocess_y_bpe\n",
        "\n",
        "preproc_english_sentences, preproc_zulu_sentences, english_tokenizer, zulu_tokenizer, preproc_english_sentences_bpe, preproc_zulu_sentences_bpe =\\\n",
        "    preprocess(english_sentences, zulu_sentences)\n",
        "    \n",
        "preproc_english_sentences_test = preproc_english_sentences[int(0.9*len(preproc_english_sentences)):]\n",
        "preproc_zulu_sentences_test = preproc_zulu_sentences[int(0.9*len(preproc_zulu_sentences)):]\n",
        "preproc_english_sentences = preproc_english_sentences[:int(0.9*len(preproc_english_sentences))]\n",
        "preproc_zulu_sentences = preproc_zulu_sentences[:int(0.9*len(preproc_zulu_sentences))]\n",
        "\n",
        "preproc_english_sentences_test_bpe = preproc_english_sentences_bpe[int(0.9*len(preproc_english_sentences_bpe)):]\n",
        "preproc_zulu_sentences_test_bpe = preproc_zulu_sentences_bpe[int(0.9*len(preproc_zulu_sentences_bpe)):]\n",
        "preproc_english_sentences_bpe = preproc_english_sentences_bpe[:int(0.9*len(preproc_english_sentences_bpe))]\n",
        "preproc_zulu_sentences_bpe = preproc_zulu_sentences_bpe[:int(0.9*len(preproc_zulu_sentences_bpe))]\n",
        "\n",
        "max_english_sequence_length = preproc_english_sentences.shape[1]\n",
        "max_zulu_sequence_length = preproc_zulu_sentences.shape[1]\n",
        "english_vocab_size = len(english_tokenizer.word_index)\n",
        "zulu_vocab_size = len(zulu_tokenizer.word_index)\n",
        "\n",
        "max_english_sequence_length_bpe = preproc_english_sentences_bpe.shape[1]\n",
        "max_zulu_sequence_length_bpe = preproc_zulu_sentences_bpe.shape[1]\n",
        "\n",
        "print('Data Preprocessed')\n",
        "print(\"Max English sentence length:\", max_english_sequence_length)\n",
        "print(\"Max Zulu sentence length:\", max_zulu_sequence_length)\n",
        "print(\"English vocabulary size:\", english_vocab_size)\n",
        "print(\"Zulu vocabulary size:\", zulu_vocab_size)\n",
        "\n",
        "print()\n",
        "print(\"Max English sentence length(bpe):\", max_english_sequence_length_bpe)\n",
        "print(\"Max Zulu sentence length(bpe):\", max_zulu_sequence_length_bpe)\n",
        "print(\"English vocabulary size(bpe):\", english_vocab_size_bpe)\n",
        "print(\"Zulu vocabulary size(bpe):\", zulu_vocab_size_bpe)"
      ],
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Min: 5  Max: 9998\n",
            "Data Preprocessed\n",
            "Max English sentence length: 23\n",
            "Max Zulu sentence length: 25\n",
            "English vocabulary size: 7639\n",
            "Zulu vocabulary size: 12597\n",
            "\n",
            "Max English sentence length(bpe): 42\n",
            "Max Zulu sentence length(bpe): 58\n",
            "English vocabulary size(bpe): 10000\n",
            "Zulu vocabulary size(bpe): 10000\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "UyDsLcbe9Je2",
        "outputId": "41e39f72-2ad1-4aca-e5dc-5be8c1a59329"
      },
      "source": [
        "def logits_to_text(logits, tokenizer):\n",
        "    \"\"\"\n",
        "    Turn logits from a neural network into text using the tokenizer\n",
        "    :param logits: Logits from a neural network\n",
        "    :param tokenizer: Keras Tokenizer fit on the labels\n",
        "    :return: String that represents the text of the logits\n",
        "    \"\"\"\n",
        "    index_to_words = {id: word for word, id in tokenizer.word_index.items()}\n",
        "    index_to_words[0] = '<PAD>'\n",
        "\n",
        "    #print(\"ITW:\", index_to_words)\n",
        "    #print(\"ITW[42]:\", index_to_words[42])\n",
        "    # print(logits, np.shape(logits))\n",
        "    #print([index_to_words[prediction[0]] for prediction in logits])\n",
        "    # print([prediction for prediction in np.argmax(logits, 1)])\n",
        "\n",
        "    return ' '.join([index_to_words[prediction] for prediction in np.argmax(logits, 1)])\n",
        "    #return ' '.join([index_to_words[prediction[0]] for prediction in logits])\n",
        "\n",
        "\n",
        "print('`logits_to_text` function loaded.')"
      ],
      "execution_count": 13,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "`logits_to_text` function loaded.\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "eqGdZSySWjyg"
      },
      "source": [
        "def logits_to_seq(logits):\n",
        "    return [prediction for prediction in np.argmax(logits, 1)]"
      ],
      "execution_count": 14,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3xGA6cA6kFiT"
      },
      "source": [
        "def logits_to_text_bpe(logits, tokenizer):\n",
        "  #return ' '.join([tokenizer.id_to_token(prediction) for prediction in logits])\n",
        "  return ' '.join([tokenizer.id_to_token(prediction) for prediction in np.argmax(logits, 1)])\n",
        "\n",
        "#   print(preproc_english_sentences_bpe[0])\n",
        "# print(logits_to_text_bpe(preproc_english_sentences_bpe[0], tokenizer_en))"
      ],
      "execution_count": 15,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RjViswgo57_Y"
      },
      "source": [
        "def show_translation_comparison(input_sent, english_sent, actual_translation, model):\n",
        "  print('English Sentence: ', english_sent)\n",
        "  print('Model Prediction Sentence: ', logits_to_text(model.predict(input_sent)[0], zulu_tokenizer))\n",
        "  print('Actual Translation: ', actual_translation)"
      ],
      "execution_count": 16,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "AlFHjsGIWk30"
      },
      "source": [
        "def calc_bleu_score(references, candidates):\n",
        "  score = corpus_bleu(references, candidates) # References are the labels, and candidates are the predictions \n",
        "  return score"
      ],
      "execution_count": 17,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_pzgZfeeWoJq"
      },
      "source": [
        "def get_ref_cand(data_to_make_predictions, model, out_labels):\n",
        "  refs = []\n",
        "  cands = []\n",
        "  for i in range(len(data_to_make_predictions)):\n",
        "    refs.append(np.expand_dims(np.squeeze(out_labels[i], 1), axis=0))\n",
        "    cands.append(np.array(logits_to_seq(model.predict(data_to_make_predictions[i:i+1])[0])))\n",
        "\n",
        "  print(np.shape(refs), np.shape(cands)) #(157, 1, 22) (157, 22)\n",
        "  refs = np.array(refs)\n",
        "  cands = np.array(cands)\n",
        "  new_refs = []\n",
        "  new_cands = []\n",
        "  for i in range(len(refs)):\n",
        "    new_refs.append([refs[i][0][refs[i][0]!=0].tolist()])\n",
        "    # if len(new_refs[len(new_refs)-1][0]) == 0:\n",
        "    #   # print('herer')\n",
        "    #   new_cands[len(new_cands)-1][0].append(0)\n",
        "\n",
        "  for i in range(len(cands)):\n",
        "    new_cands.append(cands[i][cands[i]!=0].tolist())\n",
        "    # if len(new_cands[len(new_cands)-1]) == 0:\n",
        "    #   # print('herec')\n",
        "    #   new_cands[len(new_cands)-1].append(0)\n",
        "  return new_refs, new_cands"
      ],
      "execution_count": 18,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "e9edHocvgtCz"
      },
      "source": [
        "# Models"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "L9TnVkHYfVV6"
      },
      "source": [
        "## Model 1 - Simple RNN with Standard Tokenization"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "lekqixg39OOo"
      },
      "source": [
        "def simple_model(input_shape, output_sequence_length, english_vocab_size, zulu_vocab_size):\n",
        "    \"\"\"\n",
        "    Build and train a basic RNN on x and y\n",
        "    :param input_shape: Tuple of input shape\n",
        "    :param output_sequence_length: Length of output sequence\n",
        "    :param english_vocab_size: Number of unique English words in the dataset\n",
        "    :param zulu_vocab_size: Number of unique zulu words in the dataset\n",
        "    :return: Keras model built, but not trained\n",
        "    \"\"\"\n",
        "    # Hyperparameters\n",
        "    learning_rate = 0.005\n",
        "    \n",
        "    # TODO: Build the layers\n",
        "    model = Sequential()\n",
        "    model.add(GRU(256, input_shape=input_shape[1:], return_sequences=True))\n",
        "    model.add(TimeDistributed(Dense(1024, activation='relu')))\n",
        "    model.add(Dropout(0.5))\n",
        "    model.add(TimeDistributed(Dense(zulu_vocab_size+1, activation='softmax'))) \n",
        "    #model.add(TimeDistributed(Dense(2500, activation='softmax'))) \n",
        "\n",
        "    # Compile model\n",
        "    model.compile(loss=sparse_categorical_crossentropy,\n",
        "                  optimizer=keras.optimizers.Adam(learning_rate),\n",
        "                  metrics=['accuracy'])\n",
        "    return model\n",
        "\n"
      ],
      "execution_count": 19,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "oiWe-tJDuJC8",
        "outputId": "5471bb17-00f6-43e7-9ad7-90cc8606f319"
      },
      "source": [
        "#tests.test_simple_model(simple_model)\n",
        "\n",
        "# Reshaping the input to work with a basic RNN\n",
        "tmp_x = pad(preproc_english_sentences, max_zulu_sequence_length)\n",
        "tmp_x = tmp_x.reshape((-1, preproc_zulu_sentences.shape[-2], 1))\n",
        "tmp_x_test = pad(preproc_english_sentences_test, max_zulu_sequence_length)\n",
        "tmp_x_test = tmp_x_test.reshape((-1, preproc_zulu_sentences.shape[-2], 1))\n",
        "\n",
        "#print('Tmp_x', tmp_x[0], np.shape(tmp_x))\n",
        "#print(logits_to_text(tmp_x[0], english_tokenizer))\n",
        "\n",
        "# Train the neural network\n",
        "simple_rnn_model = simple_model(\n",
        "    tmp_x.shape,\n",
        "    max_zulu_sequence_length,\n",
        "    english_vocab_size,\n",
        "    zulu_vocab_size)\n",
        "\n",
        "print(simple_rnn_model.summary())\n",
        "\n",
        "model1 = simple_rnn_model.fit(tmp_x, preproc_zulu_sentences, batch_size=1024, epochs=NUM_EPOCHS, validation_split=0.2)\n",
        "\n",
        "# Print prediction(s)\n",
        "print(logits_to_text(simple_rnn_model.predict(tmp_x[:1])[0], zulu_tokenizer))"
      ],
      "execution_count": 20,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " gru (GRU)                   (None, 25, 256)           198912    \n",
            "                                                                 \n",
            " time_distributed (TimeDistr  (None, 25, 1024)         263168    \n",
            " ibuted)                                                         \n",
            "                                                                 \n",
            " dropout (Dropout)           (None, 25, 1024)          0         \n",
            "                                                                 \n",
            " time_distributed_1 (TimeDis  (None, 25, 12598)        12912950  \n",
            " tributed)                                                       \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 13,375,030\n",
            "Trainable params: 13,375,030\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n",
            "None\n",
            "Epoch 1/200\n",
            "2/2 [==============================] - 7s 901ms/step - loss: 9.1288 - accuracy: 0.2230 - val_loss: 7.5356 - val_accuracy: 0.5831\n",
            "Epoch 2/200\n",
            "2/2 [==============================] - 1s 377ms/step - loss: 6.1822 - accuracy: 0.5948 - val_loss: 4.9743 - val_accuracy: 0.5831\n",
            "Epoch 3/200\n",
            "2/2 [==============================] - 1s 378ms/step - loss: 4.1283 - accuracy: 0.5948 - val_loss: 4.7156 - val_accuracy: 0.5831\n",
            "Epoch 4/200\n",
            "2/2 [==============================] - 1s 407ms/step - loss: 3.9727 - accuracy: 0.5948 - val_loss: 4.8119 - val_accuracy: 0.5831\n",
            "Epoch 5/200\n",
            "2/2 [==============================] - 1s 376ms/step - loss: 3.9681 - accuracy: 0.5948 - val_loss: 4.6769 - val_accuracy: 0.5831\n",
            "Epoch 6/200\n",
            "2/2 [==============================] - 1s 376ms/step - loss: 3.8658 - accuracy: 0.5948 - val_loss: 4.8348 - val_accuracy: 0.5831\n",
            "Epoch 7/200\n",
            "2/2 [==============================] - 1s 381ms/step - loss: 3.7840 - accuracy: 0.5948 - val_loss: 4.8686 - val_accuracy: 0.5831\n",
            "Epoch 8/200\n",
            "2/2 [==============================] - 1s 401ms/step - loss: 3.7587 - accuracy: 0.5956 - val_loss: 4.9698 - val_accuracy: 0.5831\n",
            "Epoch 9/200\n",
            "2/2 [==============================] - 1s 379ms/step - loss: 3.7214 - accuracy: 0.5954 - val_loss: 5.0006 - val_accuracy: 0.5831\n",
            "Epoch 10/200\n",
            "2/2 [==============================] - 1s 380ms/step - loss: 3.6881 - accuracy: 0.5950 - val_loss: 5.0166 - val_accuracy: 0.5831\n",
            "Epoch 11/200\n",
            "2/2 [==============================] - 1s 378ms/step - loss: 3.6702 - accuracy: 0.5948 - val_loss: 5.0785 - val_accuracy: 0.5831\n",
            "Epoch 12/200\n",
            "2/2 [==============================] - 1s 394ms/step - loss: 3.6406 - accuracy: 0.5955 - val_loss: 5.0820 - val_accuracy: 0.5851\n",
            "Epoch 13/200\n",
            "2/2 [==============================] - 1s 380ms/step - loss: 3.6061 - accuracy: 0.5967 - val_loss: 5.1292 - val_accuracy: 0.5848\n",
            "Epoch 14/200\n",
            "2/2 [==============================] - 1s 378ms/step - loss: 3.5778 - accuracy: 0.5965 - val_loss: 5.1186 - val_accuracy: 0.5844\n",
            "Epoch 15/200\n",
            "2/2 [==============================] - 1s 381ms/step - loss: 3.5541 - accuracy: 0.5961 - val_loss: 5.1427 - val_accuracy: 0.5850\n",
            "Epoch 16/200\n",
            "2/2 [==============================] - 1s 395ms/step - loss: 3.5193 - accuracy: 0.5963 - val_loss: 5.1436 - val_accuracy: 0.5848\n",
            "Epoch 17/200\n",
            "2/2 [==============================] - 1s 378ms/step - loss: 3.4901 - accuracy: 0.5958 - val_loss: 5.1696 - val_accuracy: 0.5846\n",
            "Epoch 18/200\n",
            "2/2 [==============================] - 1s 377ms/step - loss: 3.4618 - accuracy: 0.5962 - val_loss: 5.1777 - val_accuracy: 0.5849\n",
            "Epoch 19/200\n",
            "2/2 [==============================] - 1s 376ms/step - loss: 3.4294 - accuracy: 0.5966 - val_loss: 5.1915 - val_accuracy: 0.5851\n",
            "Epoch 20/200\n",
            "2/2 [==============================] - 1s 392ms/step - loss: 3.3999 - accuracy: 0.5969 - val_loss: 5.2065 - val_accuracy: 0.5853\n",
            "Epoch 21/200\n",
            "2/2 [==============================] - 1s 379ms/step - loss: 3.3733 - accuracy: 0.5972 - val_loss: 5.2177 - val_accuracy: 0.5852\n",
            "Epoch 22/200\n",
            "2/2 [==============================] - 1s 377ms/step - loss: 3.3460 - accuracy: 0.5972 - val_loss: 5.2302 - val_accuracy: 0.5851\n",
            "Epoch 23/200\n",
            "2/2 [==============================] - 1s 377ms/step - loss: 3.3149 - accuracy: 0.5972 - val_loss: 5.2402 - val_accuracy: 0.5852\n",
            "Epoch 24/200\n",
            "2/2 [==============================] - 1s 399ms/step - loss: 3.2887 - accuracy: 0.5969 - val_loss: 5.2524 - val_accuracy: 0.5852\n",
            "Epoch 25/200\n",
            "2/2 [==============================] - 1s 388ms/step - loss: 3.2601 - accuracy: 0.5968 - val_loss: 5.2561 - val_accuracy: 0.5851\n",
            "Epoch 26/200\n",
            "2/2 [==============================] - 1s 376ms/step - loss: 3.2326 - accuracy: 0.5966 - val_loss: 5.2672 - val_accuracy: 0.5851\n",
            "Epoch 27/200\n",
            "2/2 [==============================] - 1s 379ms/step - loss: 3.2101 - accuracy: 0.5969 - val_loss: 5.2819 - val_accuracy: 0.5851\n",
            "Epoch 28/200\n",
            "2/2 [==============================] - 1s 399ms/step - loss: 3.1820 - accuracy: 0.5968 - val_loss: 5.2890 - val_accuracy: 0.5851\n",
            "Epoch 29/200\n",
            "2/2 [==============================] - 1s 378ms/step - loss: 3.1602 - accuracy: 0.5969 - val_loss: 5.3000 - val_accuracy: 0.5853\n",
            "Epoch 30/200\n",
            "2/2 [==============================] - 1s 380ms/step - loss: 3.1310 - accuracy: 0.5970 - val_loss: 5.3103 - val_accuracy: 0.5852\n",
            "Epoch 31/200\n",
            "2/2 [==============================] - 1s 381ms/step - loss: 3.1084 - accuracy: 0.5972 - val_loss: 5.3214 - val_accuracy: 0.5856\n",
            "Epoch 32/200\n",
            "2/2 [==============================] - 1s 395ms/step - loss: 3.0863 - accuracy: 0.5971 - val_loss: 5.3342 - val_accuracy: 0.5849\n",
            "Epoch 33/200\n",
            "2/2 [==============================] - 1s 377ms/step - loss: 3.0762 - accuracy: 0.5970 - val_loss: 5.3389 - val_accuracy: 0.5851\n",
            "Epoch 34/200\n",
            "2/2 [==============================] - 1s 375ms/step - loss: 3.0437 - accuracy: 0.5971 - val_loss: 5.3463 - val_accuracy: 0.5852\n",
            "Epoch 35/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 3.0133 - accuracy: 0.5970 - val_loss: 5.3535 - val_accuracy: 0.5854\n",
            "Epoch 36/200\n",
            "2/2 [==============================] - 1s 396ms/step - loss: 2.9996 - accuracy: 0.5970 - val_loss: 5.3631 - val_accuracy: 0.5857\n",
            "Epoch 37/200\n",
            "2/2 [==============================] - 1s 378ms/step - loss: 2.9741 - accuracy: 0.5967 - val_loss: 5.3680 - val_accuracy: 0.5857\n",
            "Epoch 38/200\n",
            "2/2 [==============================] - 1s 375ms/step - loss: 2.9519 - accuracy: 0.5973 - val_loss: 5.3779 - val_accuracy: 0.5863\n",
            "Epoch 39/200\n",
            "2/2 [==============================] - 1s 378ms/step - loss: 2.9291 - accuracy: 0.5975 - val_loss: 5.3816 - val_accuracy: 0.5860\n",
            "Epoch 40/200\n",
            "2/2 [==============================] - 1s 392ms/step - loss: 2.9086 - accuracy: 0.5975 - val_loss: 5.3892 - val_accuracy: 0.5866\n",
            "Epoch 41/200\n",
            "2/2 [==============================] - 1s 376ms/step - loss: 2.8910 - accuracy: 0.5972 - val_loss: 5.3928 - val_accuracy: 0.5865\n",
            "Epoch 42/200\n",
            "2/2 [==============================] - 1s 379ms/step - loss: 2.8645 - accuracy: 0.5978 - val_loss: 5.3974 - val_accuracy: 0.5866\n",
            "Epoch 43/200\n",
            "2/2 [==============================] - 1s 380ms/step - loss: 2.8442 - accuracy: 0.5980 - val_loss: 5.4095 - val_accuracy: 0.5868\n",
            "Epoch 44/200\n",
            "2/2 [==============================] - 1s 400ms/step - loss: 2.8172 - accuracy: 0.5977 - val_loss: 5.4162 - val_accuracy: 0.5868\n",
            "Epoch 45/200\n",
            "2/2 [==============================] - 1s 380ms/step - loss: 2.7917 - accuracy: 0.5978 - val_loss: 5.4185 - val_accuracy: 0.5863\n",
            "Epoch 46/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 2.7726 - accuracy: 0.5980 - val_loss: 5.4268 - val_accuracy: 0.5866\n",
            "Epoch 47/200\n",
            "2/2 [==============================] - 1s 380ms/step - loss: 2.7664 - accuracy: 0.5978 - val_loss: 5.4303 - val_accuracy: 0.5872\n",
            "Epoch 48/200\n",
            "2/2 [==============================] - 1s 390ms/step - loss: 2.7293 - accuracy: 0.5980 - val_loss: 5.4402 - val_accuracy: 0.5873\n",
            "Epoch 49/200\n",
            "2/2 [==============================] - 1s 378ms/step - loss: 2.7124 - accuracy: 0.5986 - val_loss: 5.4409 - val_accuracy: 0.5876\n",
            "Epoch 50/200\n",
            "2/2 [==============================] - 1s 378ms/step - loss: 2.6887 - accuracy: 0.5982 - val_loss: 5.4464 - val_accuracy: 0.5871\n",
            "Epoch 51/200\n",
            "2/2 [==============================] - 1s 379ms/step - loss: 2.6566 - accuracy: 0.5985 - val_loss: 5.4541 - val_accuracy: 0.5874\n",
            "Epoch 52/200\n",
            "2/2 [==============================] - 1s 392ms/step - loss: 2.6392 - accuracy: 0.5989 - val_loss: 5.4589 - val_accuracy: 0.5874\n",
            "Epoch 53/200\n",
            "2/2 [==============================] - 1s 377ms/step - loss: 2.6163 - accuracy: 0.5992 - val_loss: 5.4653 - val_accuracy: 0.5870\n",
            "Epoch 54/200\n",
            "2/2 [==============================] - 1s 382ms/step - loss: 2.5972 - accuracy: 0.5991 - val_loss: 5.4713 - val_accuracy: 0.5873\n",
            "Epoch 55/200\n",
            "2/2 [==============================] - 1s 381ms/step - loss: 2.5705 - accuracy: 0.6003 - val_loss: 5.4838 - val_accuracy: 0.5880\n",
            "Epoch 56/200\n",
            "2/2 [==============================] - 1s 393ms/step - loss: 2.5728 - accuracy: 0.5993 - val_loss: 5.4885 - val_accuracy: 0.5875\n",
            "Epoch 57/200\n",
            "2/2 [==============================] - 1s 378ms/step - loss: 2.5440 - accuracy: 0.6004 - val_loss: 5.4920 - val_accuracy: 0.5863\n",
            "Epoch 58/200\n",
            "2/2 [==============================] - 1s 376ms/step - loss: 2.5298 - accuracy: 0.6002 - val_loss: 5.4972 - val_accuracy: 0.5872\n",
            "Epoch 59/200\n",
            "2/2 [==============================] - 1s 380ms/step - loss: 2.4832 - accuracy: 0.6017 - val_loss: 5.4935 - val_accuracy: 0.5868\n",
            "Epoch 60/200\n",
            "2/2 [==============================] - 1s 400ms/step - loss: 2.4712 - accuracy: 0.6016 - val_loss: 5.5125 - val_accuracy: 0.5864\n",
            "Epoch 61/200\n",
            "2/2 [==============================] - 1s 377ms/step - loss: 2.4618 - accuracy: 0.6022 - val_loss: 5.5134 - val_accuracy: 0.5868\n",
            "Epoch 62/200\n",
            "2/2 [==============================] - 1s 379ms/step - loss: 2.4366 - accuracy: 0.6042 - val_loss: 5.5147 - val_accuracy: 0.5872\n",
            "Epoch 63/200\n",
            "2/2 [==============================] - 1s 383ms/step - loss: 2.4196 - accuracy: 0.6036 - val_loss: 5.5113 - val_accuracy: 0.5870\n",
            "Epoch 64/200\n",
            "2/2 [==============================] - 1s 391ms/step - loss: 2.4002 - accuracy: 0.6033 - val_loss: 5.5185 - val_accuracy: 0.5864\n",
            "Epoch 65/200\n",
            "2/2 [==============================] - 1s 376ms/step - loss: 2.3629 - accuracy: 0.6058 - val_loss: 5.5218 - val_accuracy: 0.5869\n",
            "Epoch 66/200\n",
            "2/2 [==============================] - 1s 381ms/step - loss: 2.3424 - accuracy: 0.6056 - val_loss: 5.5294 - val_accuracy: 0.5868\n",
            "Epoch 67/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 2.3230 - accuracy: 0.6084 - val_loss: 5.5386 - val_accuracy: 0.5862\n",
            "Epoch 68/200\n",
            "2/2 [==============================] - 1s 392ms/step - loss: 2.3039 - accuracy: 0.6075 - val_loss: 5.5411 - val_accuracy: 0.5861\n",
            "Epoch 69/200\n",
            "2/2 [==============================] - 1s 379ms/step - loss: 2.2882 - accuracy: 0.6084 - val_loss: 5.5504 - val_accuracy: 0.5864\n",
            "Epoch 70/200\n",
            "2/2 [==============================] - 1s 377ms/step - loss: 2.2651 - accuracy: 0.6088 - val_loss: 5.5568 - val_accuracy: 0.5851\n",
            "Epoch 71/200\n",
            "2/2 [==============================] - 1s 378ms/step - loss: 2.2678 - accuracy: 0.6099 - val_loss: 5.5788 - val_accuracy: 0.5831\n",
            "Epoch 72/200\n",
            "2/2 [==============================] - 1s 402ms/step - loss: 2.2872 - accuracy: 0.6093 - val_loss: 5.5875 - val_accuracy: 0.5867\n",
            "Epoch 73/200\n",
            "2/2 [==============================] - 1s 381ms/step - loss: 2.2898 - accuracy: 0.6077 - val_loss: 5.5733 - val_accuracy: 0.5843\n",
            "Epoch 74/200\n",
            "2/2 [==============================] - 1s 376ms/step - loss: 2.2346 - accuracy: 0.6131 - val_loss: 5.5688 - val_accuracy: 0.5858\n",
            "Epoch 75/200\n",
            "2/2 [==============================] - 1s 379ms/step - loss: 2.2047 - accuracy: 0.6153 - val_loss: 5.5714 - val_accuracy: 0.5864\n",
            "Epoch 76/200\n",
            "2/2 [==============================] - 1s 392ms/step - loss: 2.1881 - accuracy: 0.6144 - val_loss: 5.5700 - val_accuracy: 0.5857\n",
            "Epoch 77/200\n",
            "2/2 [==============================] - 1s 383ms/step - loss: 2.1610 - accuracy: 0.6171 - val_loss: 5.5826 - val_accuracy: 0.5850\n",
            "Epoch 78/200\n",
            "2/2 [==============================] - 1s 377ms/step - loss: 2.1544 - accuracy: 0.6172 - val_loss: 5.5792 - val_accuracy: 0.5858\n",
            "Epoch 79/200\n",
            "2/2 [==============================] - 1s 379ms/step - loss: 2.1265 - accuracy: 0.6194 - val_loss: 5.5916 - val_accuracy: 0.5833\n",
            "Epoch 80/200\n",
            "2/2 [==============================] - 1s 394ms/step - loss: 2.1152 - accuracy: 0.6218 - val_loss: 5.5923 - val_accuracy: 0.5854\n",
            "Epoch 81/200\n",
            "2/2 [==============================] - 1s 378ms/step - loss: 2.1225 - accuracy: 0.6201 - val_loss: 5.5890 - val_accuracy: 0.5850\n",
            "Epoch 82/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 2.0890 - accuracy: 0.6222 - val_loss: 5.6062 - val_accuracy: 0.5835\n",
            "Epoch 83/200\n",
            "2/2 [==============================] - 1s 389ms/step - loss: 2.0641 - accuracy: 0.6285 - val_loss: 5.6006 - val_accuracy: 0.5851\n",
            "Epoch 84/200\n",
            "2/2 [==============================] - 1s 390ms/step - loss: 2.0676 - accuracy: 0.6238 - val_loss: 5.6003 - val_accuracy: 0.5849\n",
            "Epoch 85/200\n",
            "2/2 [==============================] - 1s 381ms/step - loss: 2.0338 - accuracy: 0.6274 - val_loss: 5.6120 - val_accuracy: 0.5838\n",
            "Epoch 86/200\n",
            "2/2 [==============================] - 1s 379ms/step - loss: 2.0218 - accuracy: 0.6299 - val_loss: 5.6143 - val_accuracy: 0.5841\n",
            "Epoch 87/200\n",
            "2/2 [==============================] - 1s 378ms/step - loss: 1.9968 - accuracy: 0.6310 - val_loss: 5.6100 - val_accuracy: 0.5844\n",
            "Epoch 88/200\n",
            "2/2 [==============================] - 1s 390ms/step - loss: 1.9788 - accuracy: 0.6345 - val_loss: 5.6143 - val_accuracy: 0.5844\n",
            "Epoch 89/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 1.9744 - accuracy: 0.6351 - val_loss: 5.6341 - val_accuracy: 0.5845\n",
            "Epoch 90/200\n",
            "2/2 [==============================] - 1s 380ms/step - loss: 1.9513 - accuracy: 0.6366 - val_loss: 5.6331 - val_accuracy: 0.5822\n",
            "Epoch 91/200\n",
            "2/2 [==============================] - 1s 378ms/step - loss: 1.9520 - accuracy: 0.6380 - val_loss: 5.6353 - val_accuracy: 0.5833\n",
            "Epoch 92/200\n",
            "2/2 [==============================] - 1s 397ms/step - loss: 1.9229 - accuracy: 0.6417 - val_loss: 5.6333 - val_accuracy: 0.5840\n",
            "Epoch 93/200\n",
            "2/2 [==============================] - 1s 378ms/step - loss: 1.9022 - accuracy: 0.6427 - val_loss: 5.6426 - val_accuracy: 0.5834\n",
            "Epoch 94/200\n",
            "2/2 [==============================] - 1s 382ms/step - loss: 1.8910 - accuracy: 0.6459 - val_loss: 5.6427 - val_accuracy: 0.5830\n",
            "Epoch 95/200\n",
            "2/2 [==============================] - 1s 378ms/step - loss: 1.8696 - accuracy: 0.6461 - val_loss: 5.6527 - val_accuracy: 0.5835\n",
            "Epoch 96/200\n",
            "2/2 [==============================] - 1s 396ms/step - loss: 1.8612 - accuracy: 0.6471 - val_loss: 5.6587 - val_accuracy: 0.5820\n",
            "Epoch 97/200\n",
            "2/2 [==============================] - 1s 382ms/step - loss: 1.8379 - accuracy: 0.6523 - val_loss: 5.6581 - val_accuracy: 0.5829\n",
            "Epoch 98/200\n",
            "2/2 [==============================] - 1s 377ms/step - loss: 1.8254 - accuracy: 0.6531 - val_loss: 5.6612 - val_accuracy: 0.5827\n",
            "Epoch 99/200\n",
            "2/2 [==============================] - 1s 377ms/step - loss: 1.8108 - accuracy: 0.6550 - val_loss: 5.6571 - val_accuracy: 0.5820\n",
            "Epoch 100/200\n",
            "2/2 [==============================] - 1s 400ms/step - loss: 1.7955 - accuracy: 0.6553 - val_loss: 5.6620 - val_accuracy: 0.5832\n",
            "Epoch 101/200\n",
            "2/2 [==============================] - 1s 379ms/step - loss: 1.7895 - accuracy: 0.6561 - val_loss: 5.6772 - val_accuracy: 0.5807\n",
            "Epoch 102/200\n",
            "2/2 [==============================] - 1s 379ms/step - loss: 1.7757 - accuracy: 0.6579 - val_loss: 5.6728 - val_accuracy: 0.5826\n",
            "Epoch 103/200\n",
            "2/2 [==============================] - 1s 381ms/step - loss: 1.7687 - accuracy: 0.6587 - val_loss: 5.6779 - val_accuracy: 0.5823\n",
            "Epoch 104/200\n",
            "2/2 [==============================] - 1s 395ms/step - loss: 1.7521 - accuracy: 0.6622 - val_loss: 5.6777 - val_accuracy: 0.5827\n",
            "Epoch 105/200\n",
            "2/2 [==============================] - 1s 381ms/step - loss: 1.7432 - accuracy: 0.6646 - val_loss: 5.6867 - val_accuracy: 0.5814\n",
            "Epoch 106/200\n",
            "2/2 [==============================] - 1s 382ms/step - loss: 1.7261 - accuracy: 0.6671 - val_loss: 5.6875 - val_accuracy: 0.5814\n",
            "Epoch 107/200\n",
            "2/2 [==============================] - 1s 381ms/step - loss: 1.7133 - accuracy: 0.6669 - val_loss: 5.6932 - val_accuracy: 0.5814\n",
            "Epoch 108/200\n",
            "2/2 [==============================] - 1s 393ms/step - loss: 1.6866 - accuracy: 0.6709 - val_loss: 5.6971 - val_accuracy: 0.5816\n",
            "Epoch 109/200\n",
            "2/2 [==============================] - 1s 381ms/step - loss: 1.6850 - accuracy: 0.6717 - val_loss: 5.6998 - val_accuracy: 0.5807\n",
            "Epoch 110/200\n",
            "2/2 [==============================] - 1s 381ms/step - loss: 1.6737 - accuracy: 0.6722 - val_loss: 5.7039 - val_accuracy: 0.5813\n",
            "Epoch 111/200\n",
            "2/2 [==============================] - 1s 380ms/step - loss: 1.6576 - accuracy: 0.6747 - val_loss: 5.7044 - val_accuracy: 0.5822\n",
            "Epoch 112/200\n",
            "2/2 [==============================] - 1s 392ms/step - loss: 1.6439 - accuracy: 0.6774 - val_loss: 5.7087 - val_accuracy: 0.5820\n",
            "Epoch 113/200\n",
            "2/2 [==============================] - 1s 380ms/step - loss: 1.6306 - accuracy: 0.6807 - val_loss: 5.7153 - val_accuracy: 0.5814\n",
            "Epoch 114/200\n",
            "2/2 [==============================] - 1s 382ms/step - loss: 1.6282 - accuracy: 0.6774 - val_loss: 5.7132 - val_accuracy: 0.5798\n",
            "Epoch 115/200\n",
            "2/2 [==============================] - 1s 385ms/step - loss: 1.6138 - accuracy: 0.6847 - val_loss: 5.7175 - val_accuracy: 0.5807\n",
            "Epoch 116/200\n",
            "2/2 [==============================] - 1s 394ms/step - loss: 1.6043 - accuracy: 0.6843 - val_loss: 5.7228 - val_accuracy: 0.5820\n",
            "Epoch 117/200\n",
            "2/2 [==============================] - 1s 381ms/step - loss: 1.5951 - accuracy: 0.6858 - val_loss: 5.7218 - val_accuracy: 0.5807\n",
            "Epoch 118/200\n",
            "2/2 [==============================] - 1s 383ms/step - loss: 1.5826 - accuracy: 0.6882 - val_loss: 5.7207 - val_accuracy: 0.5814\n",
            "Epoch 119/200\n",
            "2/2 [==============================] - 1s 385ms/step - loss: 1.5704 - accuracy: 0.6899 - val_loss: 5.7301 - val_accuracy: 0.5819\n",
            "Epoch 120/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 1.5609 - accuracy: 0.6895 - val_loss: 5.7284 - val_accuracy: 0.5814\n",
            "Epoch 121/200\n",
            "2/2 [==============================] - 1s 382ms/step - loss: 1.5530 - accuracy: 0.6920 - val_loss: 5.7313 - val_accuracy: 0.5813\n",
            "Epoch 122/200\n",
            "2/2 [==============================] - 1s 379ms/step - loss: 1.5405 - accuracy: 0.6944 - val_loss: 5.7414 - val_accuracy: 0.5825\n",
            "Epoch 123/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 1.5328 - accuracy: 0.6936 - val_loss: 5.7411 - val_accuracy: 0.5806\n",
            "Epoch 124/200\n",
            "2/2 [==============================] - 1s 391ms/step - loss: 1.5207 - accuracy: 0.6955 - val_loss: 5.7478 - val_accuracy: 0.5812\n",
            "Epoch 125/200\n",
            "2/2 [==============================] - 1s 378ms/step - loss: 1.5126 - accuracy: 0.6972 - val_loss: 5.7484 - val_accuracy: 0.5797\n",
            "Epoch 126/200\n",
            "2/2 [==============================] - 1s 381ms/step - loss: 1.5138 - accuracy: 0.6975 - val_loss: 5.7539 - val_accuracy: 0.5810\n",
            "Epoch 127/200\n",
            "2/2 [==============================] - 1s 377ms/step - loss: 1.5092 - accuracy: 0.6975 - val_loss: 5.7554 - val_accuracy: 0.5802\n",
            "Epoch 128/200\n",
            "2/2 [==============================] - 1s 394ms/step - loss: 1.4899 - accuracy: 0.7014 - val_loss: 5.7500 - val_accuracy: 0.5793\n",
            "Epoch 129/200\n",
            "2/2 [==============================] - 1s 380ms/step - loss: 1.4885 - accuracy: 0.7021 - val_loss: 5.7638 - val_accuracy: 0.5805\n",
            "Epoch 130/200\n",
            "2/2 [==============================] - 1s 379ms/step - loss: 1.4702 - accuracy: 0.7034 - val_loss: 5.7574 - val_accuracy: 0.5793\n",
            "Epoch 131/200\n",
            "2/2 [==============================] - 1s 380ms/step - loss: 1.4666 - accuracy: 0.7055 - val_loss: 5.7673 - val_accuracy: 0.5802\n",
            "Epoch 132/200\n",
            "2/2 [==============================] - 1s 402ms/step - loss: 1.4592 - accuracy: 0.7053 - val_loss: 5.7636 - val_accuracy: 0.5792\n",
            "Epoch 133/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 1.4494 - accuracy: 0.7075 - val_loss: 5.7664 - val_accuracy: 0.5800\n",
            "Epoch 134/200\n",
            "2/2 [==============================] - 1s 380ms/step - loss: 1.4364 - accuracy: 0.7111 - val_loss: 5.7737 - val_accuracy: 0.5803\n",
            "Epoch 135/200\n",
            "2/2 [==============================] - 1s 378ms/step - loss: 1.4801 - accuracy: 0.7032 - val_loss: 5.7706 - val_accuracy: 0.5787\n",
            "Epoch 136/200\n",
            "2/2 [==============================] - 1s 393ms/step - loss: 1.4428 - accuracy: 0.7084 - val_loss: 5.7762 - val_accuracy: 0.5813\n",
            "Epoch 137/200\n",
            "2/2 [==============================] - 1s 378ms/step - loss: 1.4269 - accuracy: 0.7120 - val_loss: 5.7714 - val_accuracy: 0.5802\n",
            "Epoch 138/200\n",
            "2/2 [==============================] - 1s 390ms/step - loss: 1.4057 - accuracy: 0.7161 - val_loss: 5.7776 - val_accuracy: 0.5802\n",
            "Epoch 139/200\n",
            "2/2 [==============================] - 1s 380ms/step - loss: 1.3990 - accuracy: 0.7145 - val_loss: 5.7775 - val_accuracy: 0.5798\n",
            "Epoch 140/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 1.3912 - accuracy: 0.7164 - val_loss: 5.7822 - val_accuracy: 0.5798\n",
            "Epoch 141/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 1.3778 - accuracy: 0.7185 - val_loss: 5.7878 - val_accuracy: 0.5801\n",
            "Epoch 142/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 1.3729 - accuracy: 0.7194 - val_loss: 5.7857 - val_accuracy: 0.5792\n",
            "Epoch 143/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 1.3640 - accuracy: 0.7231 - val_loss: 5.7952 - val_accuracy: 0.5806\n",
            "Epoch 144/200\n",
            "2/2 [==============================] - 1s 395ms/step - loss: 1.3610 - accuracy: 0.7228 - val_loss: 5.7940 - val_accuracy: 0.5809\n",
            "Epoch 145/200\n",
            "2/2 [==============================] - 1s 383ms/step - loss: 1.4005 - accuracy: 0.7126 - val_loss: 5.8041 - val_accuracy: 0.5790\n",
            "Epoch 146/200\n",
            "2/2 [==============================] - 1s 379ms/step - loss: 1.3763 - accuracy: 0.7190 - val_loss: 5.7980 - val_accuracy: 0.5784\n",
            "Epoch 147/200\n",
            "2/2 [==============================] - 1s 383ms/step - loss: 1.3812 - accuracy: 0.7182 - val_loss: 5.7971 - val_accuracy: 0.5813\n",
            "Epoch 148/200\n",
            "2/2 [==============================] - 1s 383ms/step - loss: 1.3703 - accuracy: 0.7183 - val_loss: 5.7997 - val_accuracy: 0.5787\n",
            "Epoch 149/200\n",
            "2/2 [==============================] - 1s 378ms/step - loss: 1.3559 - accuracy: 0.7208 - val_loss: 5.8007 - val_accuracy: 0.5782\n",
            "Epoch 150/200\n",
            "2/2 [==============================] - 1s 378ms/step - loss: 1.3556 - accuracy: 0.7228 - val_loss: 5.8080 - val_accuracy: 0.5797\n",
            "Epoch 151/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 1.3581 - accuracy: 0.7232 - val_loss: 5.8024 - val_accuracy: 0.5801\n",
            "Epoch 152/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 1.3390 - accuracy: 0.7253 - val_loss: 5.8053 - val_accuracy: 0.5794\n",
            "Epoch 153/200\n",
            "2/2 [==============================] - 1s 380ms/step - loss: 1.3256 - accuracy: 0.7276 - val_loss: 5.8009 - val_accuracy: 0.5801\n",
            "Epoch 154/200\n",
            "2/2 [==============================] - 1s 383ms/step - loss: 1.3085 - accuracy: 0.7295 - val_loss: 5.8112 - val_accuracy: 0.5800\n",
            "Epoch 155/200\n",
            "2/2 [==============================] - 1s 383ms/step - loss: 1.3077 - accuracy: 0.7304 - val_loss: 5.8123 - val_accuracy: 0.5789\n",
            "Epoch 156/200\n",
            "2/2 [==============================] - 1s 395ms/step - loss: 1.3089 - accuracy: 0.7312 - val_loss: 5.8039 - val_accuracy: 0.5816\n",
            "Epoch 157/200\n",
            "2/2 [==============================] - 1s 380ms/step - loss: 1.3854 - accuracy: 0.7175 - val_loss: 5.8104 - val_accuracy: 0.5773\n",
            "Epoch 158/200\n",
            "2/2 [==============================] - 1s 380ms/step - loss: 1.3354 - accuracy: 0.7230 - val_loss: 5.8114 - val_accuracy: 0.5802\n",
            "Epoch 159/200\n",
            "2/2 [==============================] - 1s 380ms/step - loss: 1.3311 - accuracy: 0.7244 - val_loss: 5.8191 - val_accuracy: 0.5783\n",
            "Epoch 160/200\n",
            "2/2 [==============================] - 1s 395ms/step - loss: 1.3108 - accuracy: 0.7309 - val_loss: 5.8138 - val_accuracy: 0.5803\n",
            "Epoch 161/200\n",
            "2/2 [==============================] - 1s 379ms/step - loss: 1.2917 - accuracy: 0.7319 - val_loss: 5.8117 - val_accuracy: 0.5804\n",
            "Epoch 162/200\n",
            "2/2 [==============================] - 1s 382ms/step - loss: 1.2919 - accuracy: 0.7323 - val_loss: 5.8248 - val_accuracy: 0.5789\n",
            "Epoch 163/200\n",
            "2/2 [==============================] - 1s 379ms/step - loss: 1.2900 - accuracy: 0.7336 - val_loss: 5.8154 - val_accuracy: 0.5798\n",
            "Epoch 164/200\n",
            "2/2 [==============================] - 1s 393ms/step - loss: 1.2755 - accuracy: 0.7352 - val_loss: 5.8245 - val_accuracy: 0.5799\n",
            "Epoch 165/200\n",
            "2/2 [==============================] - 1s 377ms/step - loss: 1.2679 - accuracy: 0.7373 - val_loss: 5.8233 - val_accuracy: 0.5796\n",
            "Epoch 166/200\n",
            "2/2 [==============================] - 1s 380ms/step - loss: 1.2562 - accuracy: 0.7386 - val_loss: 5.8312 - val_accuracy: 0.5804\n",
            "Epoch 167/200\n",
            "2/2 [==============================] - 1s 378ms/step - loss: 1.2475 - accuracy: 0.7405 - val_loss: 5.8301 - val_accuracy: 0.5798\n",
            "Epoch 168/200\n",
            "2/2 [==============================] - 1s 397ms/step - loss: 1.2486 - accuracy: 0.7413 - val_loss: 5.8328 - val_accuracy: 0.5792\n",
            "Epoch 169/200\n",
            "2/2 [==============================] - 1s 378ms/step - loss: 1.2358 - accuracy: 0.7420 - val_loss: 5.8385 - val_accuracy: 0.5815\n",
            "Epoch 170/200\n",
            "2/2 [==============================] - 1s 377ms/step - loss: 1.2329 - accuracy: 0.7430 - val_loss: 5.8393 - val_accuracy: 0.5788\n",
            "Epoch 171/200\n",
            "2/2 [==============================] - 1s 379ms/step - loss: 1.2318 - accuracy: 0.7429 - val_loss: 5.8459 - val_accuracy: 0.5799\n",
            "Epoch 172/200\n",
            "2/2 [==============================] - 1s 394ms/step - loss: 1.2314 - accuracy: 0.7429 - val_loss: 5.8309 - val_accuracy: 0.5802\n",
            "Epoch 173/200\n",
            "2/2 [==============================] - 1s 381ms/step - loss: 1.2355 - accuracy: 0.7438 - val_loss: 5.8483 - val_accuracy: 0.5805\n",
            "Epoch 174/200\n",
            "2/2 [==============================] - 1s 379ms/step - loss: 1.2259 - accuracy: 0.7435 - val_loss: 5.8382 - val_accuracy: 0.5786\n",
            "Epoch 175/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 1.2105 - accuracy: 0.7473 - val_loss: 5.8506 - val_accuracy: 0.5807\n",
            "Epoch 176/200\n",
            "2/2 [==============================] - 1s 398ms/step - loss: 1.2043 - accuracy: 0.7455 - val_loss: 5.8445 - val_accuracy: 0.5787\n",
            "Epoch 177/200\n",
            "2/2 [==============================] - 1s 383ms/step - loss: 1.1995 - accuracy: 0.7474 - val_loss: 5.8547 - val_accuracy: 0.5804\n",
            "Epoch 178/200\n",
            "2/2 [==============================] - 1s 382ms/step - loss: 1.1873 - accuracy: 0.7489 - val_loss: 5.8507 - val_accuracy: 0.5800\n",
            "Epoch 179/200\n",
            "2/2 [==============================] - 1s 380ms/step - loss: 1.1858 - accuracy: 0.7504 - val_loss: 5.8556 - val_accuracy: 0.5801\n",
            "Epoch 180/200\n",
            "2/2 [==============================] - 1s 391ms/step - loss: 1.1782 - accuracy: 0.7541 - val_loss: 5.8551 - val_accuracy: 0.5798\n",
            "Epoch 181/200\n",
            "2/2 [==============================] - 1s 381ms/step - loss: 1.1817 - accuracy: 0.7507 - val_loss: 5.8570 - val_accuracy: 0.5802\n",
            "Epoch 182/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 1.1748 - accuracy: 0.7518 - val_loss: 5.8571 - val_accuracy: 0.5799\n",
            "Epoch 183/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 1.1681 - accuracy: 0.7533 - val_loss: 5.8651 - val_accuracy: 0.5808\n",
            "Epoch 184/200\n",
            "2/2 [==============================] - 1s 389ms/step - loss: 1.1641 - accuracy: 0.7552 - val_loss: 5.8634 - val_accuracy: 0.5792\n",
            "Epoch 185/200\n",
            "2/2 [==============================] - 1s 383ms/step - loss: 1.1609 - accuracy: 0.7544 - val_loss: 5.8626 - val_accuracy: 0.5799\n",
            "Epoch 186/200\n",
            "2/2 [==============================] - 1s 380ms/step - loss: 1.1505 - accuracy: 0.7563 - val_loss: 5.8668 - val_accuracy: 0.5802\n",
            "Epoch 187/200\n",
            "2/2 [==============================] - 1s 380ms/step - loss: 1.1444 - accuracy: 0.7567 - val_loss: 5.8651 - val_accuracy: 0.5798\n",
            "Epoch 188/200\n",
            "2/2 [==============================] - 1s 397ms/step - loss: 1.1414 - accuracy: 0.7594 - val_loss: 5.8721 - val_accuracy: 0.5802\n",
            "Epoch 189/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 1.1340 - accuracy: 0.7601 - val_loss: 5.8702 - val_accuracy: 0.5794\n",
            "Epoch 190/200\n",
            "2/2 [==============================] - 1s 381ms/step - loss: 1.1346 - accuracy: 0.7586 - val_loss: 5.8802 - val_accuracy: 0.5799\n",
            "Epoch 191/200\n",
            "2/2 [==============================] - 1s 381ms/step - loss: 1.1247 - accuracy: 0.7628 - val_loss: 5.8745 - val_accuracy: 0.5785\n",
            "Epoch 192/200\n",
            "2/2 [==============================] - 1s 396ms/step - loss: 1.1220 - accuracy: 0.7612 - val_loss: 5.8792 - val_accuracy: 0.5795\n",
            "Epoch 193/200\n",
            "2/2 [==============================] - 1s 377ms/step - loss: 1.1110 - accuracy: 0.7630 - val_loss: 5.8764 - val_accuracy: 0.5801\n",
            "Epoch 194/200\n",
            "2/2 [==============================] - 1s 385ms/step - loss: 1.1156 - accuracy: 0.7620 - val_loss: 5.8811 - val_accuracy: 0.5799\n",
            "Epoch 195/200\n",
            "2/2 [==============================] - 1s 378ms/step - loss: 1.1116 - accuracy: 0.7636 - val_loss: 5.8799 - val_accuracy: 0.5811\n",
            "Epoch 196/200\n",
            "2/2 [==============================] - 1s 393ms/step - loss: 1.1093 - accuracy: 0.7628 - val_loss: 5.8821 - val_accuracy: 0.5817\n",
            "Epoch 197/200\n",
            "2/2 [==============================] - 1s 378ms/step - loss: 1.1069 - accuracy: 0.7617 - val_loss: 5.8784 - val_accuracy: 0.5792\n",
            "Epoch 198/200\n",
            "2/2 [==============================] - 1s 385ms/step - loss: 1.1033 - accuracy: 0.7641 - val_loss: 5.8795 - val_accuracy: 0.5795\n",
            "Epoch 199/200\n",
            "2/2 [==============================] - 1s 382ms/step - loss: 1.0976 - accuracy: 0.7662 - val_loss: 5.8793 - val_accuracy: 0.5802\n",
            "Epoch 200/200\n",
            "2/2 [==============================] - 1s 391ms/step - loss: 1.0857 - accuracy: 0.7683 - val_loss: 5.8849 - val_accuracy: 0.5810\n",
            "kodwa ngendawo bailey eziyinhloko solo eminyaka yawo 60 angamanye anezingoma ezimnandi <PAD> kwengake <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD>\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "3QpJFXJQfjzy"
      },
      "source": [
        "## Model 2 - Simple RNN with Byte-Pair Encoding"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "QQ2McZASuQQ1",
        "outputId": "7b9266e8-7769-47cc-e1a4-f859d59988ee"
      },
      "source": [
        "#tests.test_simple_model(simple_model)\n",
        "\n",
        "# Reshaping the input to work with a basic RNN\n",
        "tmp_x_bpe = pad(preproc_english_sentences_bpe, max_zulu_sequence_length_bpe)\n",
        "tmp_x_bpe = tmp_x_bpe.reshape((-1, preproc_zulu_sentences_bpe.shape[-2], 1))\n",
        "tmp_x_test_bpe = pad(preproc_english_sentences_test_bpe, max_zulu_sequence_length_bpe)\n",
        "tmp_x_test_bpe = tmp_x_test_bpe.reshape((-1, preproc_zulu_sentences_bpe.shape[-2], 1))\n",
        "\n",
        "#print('Tmp_x', tmp_x[0], np.shape(tmp_x))\n",
        "#print(logits_to_text(tmp_x[0], english_tokenizer))\n",
        "\n",
        "# Train the neural network\n",
        "simple_rnn_model_bpe = simple_model(\n",
        "    tmp_x_bpe.shape,\n",
        "    max_zulu_sequence_length_bpe,\n",
        "    english_vocab_size_bpe,\n",
        "    zulu_vocab_size_bpe)\n",
        "\n",
        "print(simple_rnn_model_bpe.summary())\n",
        "\n",
        "model2 = simple_rnn_model_bpe.fit(tmp_x_bpe, preproc_zulu_sentences_bpe, batch_size=1024, epochs=NUM_EPOCHS, validation_split=0.2)\n",
        "\n",
        "# Print prediction(s)\n",
        "print(logits_to_text_bpe(simple_rnn_model_bpe.predict(tmp_x_bpe[:1])[0], tokenizer_zu))"
      ],
      "execution_count": 21,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential_1\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " gru_1 (GRU)                 (None, 58, 256)           198912    \n",
            "                                                                 \n",
            " time_distributed_2 (TimeDis  (None, 58, 1024)         263168    \n",
            " tributed)                                                       \n",
            "                                                                 \n",
            " dropout_1 (Dropout)         (None, 58, 1024)          0         \n",
            "                                                                 \n",
            " time_distributed_3 (TimeDis  (None, 58, 10001)        10251025  \n",
            " tributed)                                                       \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 10,713,105\n",
            "Trainable params: 10,713,105\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n",
            "None\n",
            "Epoch 1/200\n",
            "2/2 [==============================] - 4s 1s/step - loss: 9.0434 - accuracy: 0.3393 - val_loss: 8.0487 - val_accuracy: 0.6792\n",
            "Epoch 2/200\n",
            "2/2 [==============================] - 1s 629ms/step - loss: 6.3244 - accuracy: 0.6756 - val_loss: 3.3791 - val_accuracy: 0.6792\n",
            "Epoch 3/200\n",
            "2/2 [==============================] - 1s 643ms/step - loss: 2.9621 - accuracy: 0.6805 - val_loss: 3.3374 - val_accuracy: 0.6794\n",
            "Epoch 4/200\n",
            "2/2 [==============================] - 1s 630ms/step - loss: 2.9112 - accuracy: 0.6768 - val_loss: 3.2427 - val_accuracy: 0.6658\n",
            "Epoch 5/200\n",
            "2/2 [==============================] - 1s 630ms/step - loss: 2.9314 - accuracy: 0.6649 - val_loss: 3.1695 - val_accuracy: 0.6660\n",
            "Epoch 6/200\n",
            "2/2 [==============================] - 1s 650ms/step - loss: 2.8229 - accuracy: 0.6668 - val_loss: 3.2328 - val_accuracy: 0.6733\n",
            "Epoch 7/200\n",
            "2/2 [==============================] - 1s 633ms/step - loss: 2.7550 - accuracy: 0.6728 - val_loss: 3.1727 - val_accuracy: 0.6740\n",
            "Epoch 8/200\n",
            "2/2 [==============================] - 1s 636ms/step - loss: 2.6717 - accuracy: 0.6749 - val_loss: 2.9683 - val_accuracy: 0.6758\n",
            "Epoch 9/200\n",
            "2/2 [==============================] - 1s 634ms/step - loss: 2.6247 - accuracy: 0.6766 - val_loss: 2.9432 - val_accuracy: 0.6792\n",
            "Epoch 10/200\n",
            "2/2 [==============================] - 1s 631ms/step - loss: 2.5912 - accuracy: 0.6803 - val_loss: 3.0126 - val_accuracy: 0.6810\n",
            "Epoch 11/200\n",
            "2/2 [==============================] - 1s 635ms/step - loss: 2.5833 - accuracy: 0.6804 - val_loss: 2.9819 - val_accuracy: 0.6794\n",
            "Epoch 12/200\n",
            "2/2 [==============================] - 1s 633ms/step - loss: 2.5621 - accuracy: 0.6796 - val_loss: 2.9566 - val_accuracy: 0.6787\n",
            "Epoch 13/200\n",
            "2/2 [==============================] - 1s 637ms/step - loss: 2.5538 - accuracy: 0.6794 - val_loss: 2.9831 - val_accuracy: 0.6798\n",
            "Epoch 14/200\n",
            "2/2 [==============================] - 1s 632ms/step - loss: 2.5367 - accuracy: 0.6805 - val_loss: 2.9994 - val_accuracy: 0.6807\n",
            "Epoch 15/200\n",
            "2/2 [==============================] - 1s 631ms/step - loss: 2.5286 - accuracy: 0.6817 - val_loss: 2.9832 - val_accuracy: 0.6814\n",
            "Epoch 16/200\n",
            "2/2 [==============================] - 1s 631ms/step - loss: 2.5202 - accuracy: 0.6820 - val_loss: 2.9796 - val_accuracy: 0.6816\n",
            "Epoch 17/200\n",
            "2/2 [==============================] - 1s 654ms/step - loss: 2.5136 - accuracy: 0.6822 - val_loss: 2.9937 - val_accuracy: 0.6816\n",
            "Epoch 18/200\n",
            "2/2 [==============================] - 1s 634ms/step - loss: 2.5073 - accuracy: 0.6822 - val_loss: 2.9844 - val_accuracy: 0.6811\n",
            "Epoch 19/200\n",
            "2/2 [==============================] - 1s 633ms/step - loss: 2.4992 - accuracy: 0.6823 - val_loss: 2.9822 - val_accuracy: 0.6812\n",
            "Epoch 20/200\n",
            "2/2 [==============================] - 1s 635ms/step - loss: 2.4942 - accuracy: 0.6820 - val_loss: 2.9854 - val_accuracy: 0.6813\n",
            "Epoch 21/200\n",
            "2/2 [==============================] - 1s 632ms/step - loss: 2.4869 - accuracy: 0.6822 - val_loss: 2.9869 - val_accuracy: 0.6812\n",
            "Epoch 22/200\n",
            "2/2 [==============================] - 1s 632ms/step - loss: 2.4797 - accuracy: 0.6821 - val_loss: 2.9874 - val_accuracy: 0.6812\n",
            "Epoch 23/200\n",
            "2/2 [==============================] - 1s 635ms/step - loss: 2.4729 - accuracy: 0.6825 - val_loss: 2.9960 - val_accuracy: 0.6815\n",
            "Epoch 24/200\n",
            "2/2 [==============================] - 1s 631ms/step - loss: 2.4646 - accuracy: 0.6828 - val_loss: 2.9991 - val_accuracy: 0.6817\n",
            "Epoch 25/200\n",
            "2/2 [==============================] - 1s 628ms/step - loss: 2.4593 - accuracy: 0.6831 - val_loss: 3.0028 - val_accuracy: 0.6820\n",
            "Epoch 26/200\n",
            "2/2 [==============================] - 1s 630ms/step - loss: 2.4527 - accuracy: 0.6827 - val_loss: 3.0018 - val_accuracy: 0.6820\n",
            "Epoch 27/200\n",
            "2/2 [==============================] - 1s 636ms/step - loss: 2.4461 - accuracy: 0.6829 - val_loss: 3.0162 - val_accuracy: 0.6823\n",
            "Epoch 28/200\n",
            "2/2 [==============================] - 1s 643ms/step - loss: 2.4393 - accuracy: 0.6832 - val_loss: 3.0171 - val_accuracy: 0.6823\n",
            "Epoch 29/200\n",
            "2/2 [==============================] - 1s 630ms/step - loss: 2.4351 - accuracy: 0.6832 - val_loss: 3.0242 - val_accuracy: 0.6822\n",
            "Epoch 30/200\n",
            "2/2 [==============================] - 1s 630ms/step - loss: 2.4279 - accuracy: 0.6833 - val_loss: 3.0236 - val_accuracy: 0.6822\n",
            "Epoch 31/200\n",
            "2/2 [==============================] - 1s 632ms/step - loss: 2.4215 - accuracy: 0.6835 - val_loss: 3.0349 - val_accuracy: 0.6822\n",
            "Epoch 32/200\n",
            "2/2 [==============================] - 1s 632ms/step - loss: 2.4169 - accuracy: 0.6836 - val_loss: 3.0395 - val_accuracy: 0.6823\n",
            "Epoch 33/200\n",
            "2/2 [==============================] - 1s 632ms/step - loss: 2.4119 - accuracy: 0.6836 - val_loss: 3.0397 - val_accuracy: 0.6822\n",
            "Epoch 34/200\n",
            "2/2 [==============================] - 1s 644ms/step - loss: 2.4072 - accuracy: 0.6835 - val_loss: 3.0512 - val_accuracy: 0.6823\n",
            "Epoch 35/200\n",
            "2/2 [==============================] - 1s 631ms/step - loss: 2.4030 - accuracy: 0.6837 - val_loss: 3.0541 - val_accuracy: 0.6824\n",
            "Epoch 36/200\n",
            "2/2 [==============================] - 1s 633ms/step - loss: 2.3962 - accuracy: 0.6836 - val_loss: 3.0559 - val_accuracy: 0.6824\n",
            "Epoch 37/200\n",
            "2/2 [==============================] - 1s 630ms/step - loss: 2.3922 - accuracy: 0.6837 - val_loss: 3.0664 - val_accuracy: 0.6825\n",
            "Epoch 38/200\n",
            "2/2 [==============================] - 1s 640ms/step - loss: 2.3854 - accuracy: 0.6837 - val_loss: 3.0663 - val_accuracy: 0.6825\n",
            "Epoch 39/200\n",
            "2/2 [==============================] - 1s 640ms/step - loss: 2.3860 - accuracy: 0.6839 - val_loss: 3.0647 - val_accuracy: 0.6825\n",
            "Epoch 40/200\n",
            "2/2 [==============================] - 1s 633ms/step - loss: 2.3794 - accuracy: 0.6838 - val_loss: 3.1004 - val_accuracy: 0.6825\n",
            "Epoch 41/200\n",
            "2/2 [==============================] - 1s 648ms/step - loss: 2.3789 - accuracy: 0.6840 - val_loss: 3.0748 - val_accuracy: 0.6827\n",
            "Epoch 42/200\n",
            "2/2 [==============================] - 1s 631ms/step - loss: 2.3725 - accuracy: 0.6840 - val_loss: 3.0951 - val_accuracy: 0.6825\n",
            "Epoch 43/200\n",
            "2/2 [==============================] - 1s 633ms/step - loss: 2.3694 - accuracy: 0.6842 - val_loss: 3.0888 - val_accuracy: 0.6827\n",
            "Epoch 44/200\n",
            "2/2 [==============================] - 1s 634ms/step - loss: 2.3597 - accuracy: 0.6840 - val_loss: 3.1041 - val_accuracy: 0.6827\n",
            "Epoch 45/200\n",
            "2/2 [==============================] - 1s 631ms/step - loss: 2.3563 - accuracy: 0.6843 - val_loss: 3.0967 - val_accuracy: 0.6826\n",
            "Epoch 46/200\n",
            "2/2 [==============================] - 1s 630ms/step - loss: 2.3485 - accuracy: 0.6844 - val_loss: 3.1229 - val_accuracy: 0.6831\n",
            "Epoch 47/200\n",
            "2/2 [==============================] - 1s 634ms/step - loss: 2.3472 - accuracy: 0.6845 - val_loss: 3.1072 - val_accuracy: 0.6835\n",
            "Epoch 48/200\n",
            "2/2 [==============================] - 1s 634ms/step - loss: 2.3428 - accuracy: 0.6850 - val_loss: 3.0984 - val_accuracy: 0.6839\n",
            "Epoch 49/200\n",
            "2/2 [==============================] - 1s 632ms/step - loss: 2.3359 - accuracy: 0.6845 - val_loss: 3.1349 - val_accuracy: 0.6835\n",
            "Epoch 50/200\n",
            "2/2 [==============================] - 1s 635ms/step - loss: 2.3296 - accuracy: 0.6852 - val_loss: 3.1064 - val_accuracy: 0.6837\n",
            "Epoch 51/200\n",
            "2/2 [==============================] - 1s 631ms/step - loss: 2.3246 - accuracy: 0.6849 - val_loss: 3.1394 - val_accuracy: 0.6839\n",
            "Epoch 52/200\n",
            "2/2 [==============================] - 1s 631ms/step - loss: 2.3172 - accuracy: 0.6854 - val_loss: 3.1201 - val_accuracy: 0.6847\n",
            "Epoch 53/200\n",
            "2/2 [==============================] - 1s 649ms/step - loss: 2.3125 - accuracy: 0.6851 - val_loss: 3.1207 - val_accuracy: 0.6849\n",
            "Epoch 54/200\n",
            "2/2 [==============================] - 1s 631ms/step - loss: 2.3062 - accuracy: 0.6856 - val_loss: 3.1490 - val_accuracy: 0.6850\n",
            "Epoch 55/200\n",
            "2/2 [==============================] - 1s 630ms/step - loss: 2.2984 - accuracy: 0.6859 - val_loss: 3.1357 - val_accuracy: 0.6851\n",
            "Epoch 56/200\n",
            "2/2 [==============================] - 1s 630ms/step - loss: 2.2916 - accuracy: 0.6861 - val_loss: 3.1488 - val_accuracy: 0.6856\n",
            "Epoch 57/200\n",
            "2/2 [==============================] - 1s 634ms/step - loss: 2.2828 - accuracy: 0.6864 - val_loss: 3.1618 - val_accuracy: 0.6849\n",
            "Epoch 58/200\n",
            "2/2 [==============================] - 1s 633ms/step - loss: 2.2807 - accuracy: 0.6866 - val_loss: 3.1567 - val_accuracy: 0.6855\n",
            "Epoch 59/200\n",
            "2/2 [==============================] - 1s 655ms/step - loss: 2.2749 - accuracy: 0.6865 - val_loss: 3.1441 - val_accuracy: 0.6857\n",
            "Epoch 60/200\n",
            "2/2 [==============================] - 1s 647ms/step - loss: 2.2694 - accuracy: 0.6865 - val_loss: 3.1853 - val_accuracy: 0.6854\n",
            "Epoch 61/200\n",
            "2/2 [==============================] - 1s 630ms/step - loss: 2.2614 - accuracy: 0.6866 - val_loss: 3.1578 - val_accuracy: 0.6857\n",
            "Epoch 62/200\n",
            "2/2 [==============================] - 1s 638ms/step - loss: 2.2522 - accuracy: 0.6867 - val_loss: 3.1769 - val_accuracy: 0.6858\n",
            "Epoch 63/200\n",
            "2/2 [==============================] - 1s 632ms/step - loss: 2.2459 - accuracy: 0.6867 - val_loss: 3.1743 - val_accuracy: 0.6858\n",
            "Epoch 64/200\n",
            "2/2 [==============================] - 1s 632ms/step - loss: 2.2372 - accuracy: 0.6870 - val_loss: 3.1956 - val_accuracy: 0.6857\n",
            "Epoch 65/200\n",
            "2/2 [==============================] - 1s 636ms/step - loss: 2.2320 - accuracy: 0.6872 - val_loss: 3.1895 - val_accuracy: 0.6857\n",
            "Epoch 66/200\n",
            "2/2 [==============================] - 1s 633ms/step - loss: 2.2227 - accuracy: 0.6872 - val_loss: 3.1889 - val_accuracy: 0.6859\n",
            "Epoch 67/200\n",
            "2/2 [==============================] - 1s 633ms/step - loss: 2.2201 - accuracy: 0.6874 - val_loss: 3.2295 - val_accuracy: 0.6854\n",
            "Epoch 68/200\n",
            "2/2 [==============================] - 1s 632ms/step - loss: 2.2195 - accuracy: 0.6870 - val_loss: 3.2040 - val_accuracy: 0.6859\n",
            "Epoch 69/200\n",
            "2/2 [==============================] - 1s 632ms/step - loss: 2.2033 - accuracy: 0.6871 - val_loss: 3.2072 - val_accuracy: 0.6860\n",
            "Epoch 70/200\n",
            "2/2 [==============================] - 1s 637ms/step - loss: 2.1950 - accuracy: 0.6877 - val_loss: 3.2168 - val_accuracy: 0.6859\n",
            "Epoch 71/200\n",
            "2/2 [==============================] - 1s 632ms/step - loss: 2.1883 - accuracy: 0.6877 - val_loss: 3.2183 - val_accuracy: 0.6861\n",
            "Epoch 72/200\n",
            "2/2 [==============================] - 1s 640ms/step - loss: 2.1830 - accuracy: 0.6880 - val_loss: 3.2249 - val_accuracy: 0.6861\n",
            "Epoch 73/200\n",
            "2/2 [==============================] - 1s 636ms/step - loss: 2.1742 - accuracy: 0.6878 - val_loss: 3.2311 - val_accuracy: 0.6862\n",
            "Epoch 74/200\n",
            "2/2 [==============================] - 1s 629ms/step - loss: 2.1672 - accuracy: 0.6883 - val_loss: 3.2396 - val_accuracy: 0.6864\n",
            "Epoch 75/200\n",
            "2/2 [==============================] - 1s 629ms/step - loss: 2.1600 - accuracy: 0.6883 - val_loss: 3.2347 - val_accuracy: 0.6862\n",
            "Epoch 76/200\n",
            "2/2 [==============================] - 1s 630ms/step - loss: 2.1546 - accuracy: 0.6882 - val_loss: 3.2387 - val_accuracy: 0.6856\n",
            "Epoch 77/200\n",
            "2/2 [==============================] - 1s 636ms/step - loss: 2.1515 - accuracy: 0.6881 - val_loss: 3.2545 - val_accuracy: 0.6862\n",
            "Epoch 78/200\n",
            "2/2 [==============================] - 1s 637ms/step - loss: 2.1390 - accuracy: 0.6883 - val_loss: 3.2541 - val_accuracy: 0.6862\n",
            "Epoch 79/200\n",
            "2/2 [==============================] - 1s 635ms/step - loss: 2.1335 - accuracy: 0.6881 - val_loss: 3.2607 - val_accuracy: 0.6862\n",
            "Epoch 80/200\n",
            "2/2 [==============================] - 1s 629ms/step - loss: 2.1262 - accuracy: 0.6881 - val_loss: 3.2625 - val_accuracy: 0.6862\n",
            "Epoch 81/200\n",
            "2/2 [==============================] - 1s 637ms/step - loss: 2.1184 - accuracy: 0.6886 - val_loss: 3.2723 - val_accuracy: 0.6862\n",
            "Epoch 82/200\n",
            "2/2 [==============================] - 1s 634ms/step - loss: 2.1099 - accuracy: 0.6887 - val_loss: 3.2700 - val_accuracy: 0.6862\n",
            "Epoch 83/200\n",
            "2/2 [==============================] - 1s 631ms/step - loss: 2.1013 - accuracy: 0.6890 - val_loss: 3.2789 - val_accuracy: 0.6861\n",
            "Epoch 84/200\n",
            "2/2 [==============================] - 1s 645ms/step - loss: 2.0995 - accuracy: 0.6885 - val_loss: 3.2741 - val_accuracy: 0.6861\n",
            "Epoch 85/200\n",
            "2/2 [==============================] - 1s 634ms/step - loss: 2.0921 - accuracy: 0.6888 - val_loss: 3.2908 - val_accuracy: 0.6861\n",
            "Epoch 86/200\n",
            "2/2 [==============================] - 1s 633ms/step - loss: 2.0826 - accuracy: 0.6890 - val_loss: 3.3052 - val_accuracy: 0.6861\n",
            "Epoch 87/200\n",
            "2/2 [==============================] - 1s 630ms/step - loss: 2.0946 - accuracy: 0.6886 - val_loss: 3.2937 - val_accuracy: 0.6864\n",
            "Epoch 88/200\n",
            "2/2 [==============================] - 1s 634ms/step - loss: 2.0793 - accuracy: 0.6889 - val_loss: 3.2943 - val_accuracy: 0.6862\n",
            "Epoch 89/200\n",
            "2/2 [==============================] - 1s 645ms/step - loss: 2.0820 - accuracy: 0.6885 - val_loss: 3.3195 - val_accuracy: 0.6865\n",
            "Epoch 90/200\n",
            "2/2 [==============================] - 1s 634ms/step - loss: 2.0669 - accuracy: 0.6887 - val_loss: 3.2951 - val_accuracy: 0.6863\n",
            "Epoch 91/200\n",
            "2/2 [==============================] - 1s 630ms/step - loss: 2.0602 - accuracy: 0.6893 - val_loss: 3.3321 - val_accuracy: 0.6863\n",
            "Epoch 92/200\n",
            "2/2 [==============================] - 1s 635ms/step - loss: 2.0495 - accuracy: 0.6892 - val_loss: 3.3043 - val_accuracy: 0.6866\n",
            "Epoch 93/200\n",
            "2/2 [==============================] - 1s 632ms/step - loss: 2.0467 - accuracy: 0.6893 - val_loss: 3.3210 - val_accuracy: 0.6860\n",
            "Epoch 94/200\n",
            "2/2 [==============================] - 1s 632ms/step - loss: 2.0438 - accuracy: 0.6893 - val_loss: 3.3547 - val_accuracy: 0.6859\n",
            "Epoch 95/200\n",
            "2/2 [==============================] - 1s 639ms/step - loss: 2.0531 - accuracy: 0.6888 - val_loss: 3.3309 - val_accuracy: 0.6862\n",
            "Epoch 96/200\n",
            "2/2 [==============================] - 1s 630ms/step - loss: 2.0333 - accuracy: 0.6889 - val_loss: 3.3355 - val_accuracy: 0.6863\n",
            "Epoch 97/200\n",
            "2/2 [==============================] - 1s 634ms/step - loss: 2.0178 - accuracy: 0.6894 - val_loss: 3.3357 - val_accuracy: 0.6864\n",
            "Epoch 98/200\n",
            "2/2 [==============================] - 1s 632ms/step - loss: 2.0123 - accuracy: 0.6889 - val_loss: 3.3388 - val_accuracy: 0.6865\n",
            "Epoch 99/200\n",
            "2/2 [==============================] - 1s 632ms/step - loss: 1.9995 - accuracy: 0.6899 - val_loss: 3.3467 - val_accuracy: 0.6866\n",
            "Epoch 100/200\n",
            "2/2 [==============================] - 1s 629ms/step - loss: 1.9946 - accuracy: 0.6899 - val_loss: 3.3480 - val_accuracy: 0.6862\n",
            "Epoch 101/200\n",
            "2/2 [==============================] - 1s 630ms/step - loss: 1.9849 - accuracy: 0.6901 - val_loss: 3.3489 - val_accuracy: 0.6862\n",
            "Epoch 102/200\n",
            "2/2 [==============================] - 1s 631ms/step - loss: 1.9773 - accuracy: 0.6904 - val_loss: 3.3635 - val_accuracy: 0.6863\n",
            "Epoch 103/200\n",
            "2/2 [==============================] - 1s 635ms/step - loss: 1.9723 - accuracy: 0.6902 - val_loss: 3.3534 - val_accuracy: 0.6864\n",
            "Epoch 104/200\n",
            "2/2 [==============================] - 1s 632ms/step - loss: 1.9621 - accuracy: 0.6907 - val_loss: 3.3616 - val_accuracy: 0.6865\n",
            "Epoch 105/200\n",
            "2/2 [==============================] - 1s 635ms/step - loss: 1.9573 - accuracy: 0.6907 - val_loss: 3.3652 - val_accuracy: 0.6862\n",
            "Epoch 106/200\n",
            "2/2 [==============================] - 1s 640ms/step - loss: 1.9484 - accuracy: 0.6905 - val_loss: 3.3765 - val_accuracy: 0.6862\n",
            "Epoch 107/200\n",
            "2/2 [==============================] - 1s 631ms/step - loss: 1.9443 - accuracy: 0.6912 - val_loss: 3.3695 - val_accuracy: 0.6864\n",
            "Epoch 108/200\n",
            "2/2 [==============================] - 1s 634ms/step - loss: 1.9386 - accuracy: 0.6910 - val_loss: 3.3852 - val_accuracy: 0.6864\n",
            "Epoch 109/200\n",
            "2/2 [==============================] - 1s 634ms/step - loss: 1.9293 - accuracy: 0.6912 - val_loss: 3.3853 - val_accuracy: 0.6863\n",
            "Epoch 110/200\n",
            "2/2 [==============================] - 1s 631ms/step - loss: 1.9218 - accuracy: 0.6915 - val_loss: 3.3877 - val_accuracy: 0.6865\n",
            "Epoch 111/200\n",
            "2/2 [==============================] - 1s 638ms/step - loss: 1.9172 - accuracy: 0.6920 - val_loss: 3.3926 - val_accuracy: 0.6863\n",
            "Epoch 112/200\n",
            "2/2 [==============================] - 1s 642ms/step - loss: 1.9106 - accuracy: 0.6920 - val_loss: 3.3998 - val_accuracy: 0.6868\n",
            "Epoch 113/200\n",
            "2/2 [==============================] - 1s 640ms/step - loss: 1.9069 - accuracy: 0.6917 - val_loss: 3.4087 - val_accuracy: 0.6864\n",
            "Epoch 114/200\n",
            "2/2 [==============================] - 1s 639ms/step - loss: 1.8965 - accuracy: 0.6923 - val_loss: 3.3976 - val_accuracy: 0.6866\n",
            "Epoch 115/200\n",
            "2/2 [==============================] - 1s 630ms/step - loss: 1.8939 - accuracy: 0.6924 - val_loss: 3.4037 - val_accuracy: 0.6864\n",
            "Epoch 116/200\n",
            "2/2 [==============================] - 1s 630ms/step - loss: 1.8867 - accuracy: 0.6923 - val_loss: 3.4336 - val_accuracy: 0.6864\n",
            "Epoch 117/200\n",
            "2/2 [==============================] - 1s 649ms/step - loss: 1.8839 - accuracy: 0.6924 - val_loss: 3.4109 - val_accuracy: 0.6866\n",
            "Epoch 118/200\n",
            "2/2 [==============================] - 1s 630ms/step - loss: 1.8771 - accuracy: 0.6927 - val_loss: 3.4144 - val_accuracy: 0.6864\n",
            "Epoch 119/200\n",
            "2/2 [==============================] - 1s 648ms/step - loss: 1.8648 - accuracy: 0.6932 - val_loss: 3.4405 - val_accuracy: 0.6867\n",
            "Epoch 120/200\n",
            "2/2 [==============================] - 1s 655ms/step - loss: 1.8742 - accuracy: 0.6928 - val_loss: 3.4335 - val_accuracy: 0.6865\n",
            "Epoch 121/200\n",
            "2/2 [==============================] - 1s 630ms/step - loss: 1.8629 - accuracy: 0.6930 - val_loss: 3.4290 - val_accuracy: 0.6863\n",
            "Epoch 122/200\n",
            "2/2 [==============================] - 1s 630ms/step - loss: 1.8509 - accuracy: 0.6938 - val_loss: 3.4339 - val_accuracy: 0.6864\n",
            "Epoch 123/200\n",
            "2/2 [==============================] - 1s 630ms/step - loss: 1.8458 - accuracy: 0.6939 - val_loss: 3.4546 - val_accuracy: 0.6866\n",
            "Epoch 124/200\n",
            "2/2 [==============================] - 1s 629ms/step - loss: 1.8438 - accuracy: 0.6938 - val_loss: 3.4444 - val_accuracy: 0.6865\n",
            "Epoch 125/200\n",
            "2/2 [==============================] - 1s 637ms/step - loss: 1.8370 - accuracy: 0.6941 - val_loss: 3.4385 - val_accuracy: 0.6868\n",
            "Epoch 126/200\n",
            "2/2 [==============================] - 1s 640ms/step - loss: 1.8344 - accuracy: 0.6948 - val_loss: 3.4416 - val_accuracy: 0.6865\n",
            "Epoch 127/200\n",
            "2/2 [==============================] - 1s 636ms/step - loss: 1.8282 - accuracy: 0.6949 - val_loss: 3.4367 - val_accuracy: 0.6863\n",
            "Epoch 128/200\n",
            "2/2 [==============================] - 1s 630ms/step - loss: 1.8285 - accuracy: 0.6949 - val_loss: 3.4424 - val_accuracy: 0.6867\n",
            "Epoch 129/200\n",
            "2/2 [==============================] - 1s 633ms/step - loss: 1.8182 - accuracy: 0.6956 - val_loss: 3.4690 - val_accuracy: 0.6866\n",
            "Epoch 130/200\n",
            "2/2 [==============================] - 1s 638ms/step - loss: 1.8204 - accuracy: 0.6948 - val_loss: 3.4571 - val_accuracy: 0.6865\n",
            "Epoch 131/200\n",
            "2/2 [==============================] - 1s 637ms/step - loss: 1.8084 - accuracy: 0.6960 - val_loss: 3.4433 - val_accuracy: 0.6864\n",
            "Epoch 132/200\n",
            "2/2 [==============================] - 1s 638ms/step - loss: 1.8129 - accuracy: 0.6958 - val_loss: 3.4839 - val_accuracy: 0.6866\n",
            "Epoch 133/200\n",
            "2/2 [==============================] - 1s 633ms/step - loss: 1.8033 - accuracy: 0.6961 - val_loss: 3.4574 - val_accuracy: 0.6863\n",
            "Epoch 134/200\n",
            "2/2 [==============================] - 1s 633ms/step - loss: 1.7918 - accuracy: 0.6965 - val_loss: 3.4610 - val_accuracy: 0.6865\n",
            "Epoch 135/200\n",
            "2/2 [==============================] - 1s 640ms/step - loss: 1.7788 - accuracy: 0.6975 - val_loss: 3.4718 - val_accuracy: 0.6868\n",
            "Epoch 136/200\n",
            "2/2 [==============================] - 1s 636ms/step - loss: 1.7737 - accuracy: 0.6982 - val_loss: 3.4569 - val_accuracy: 0.6863\n",
            "Epoch 137/200\n",
            "2/2 [==============================] - 1s 631ms/step - loss: 1.7732 - accuracy: 0.6984 - val_loss: 3.5073 - val_accuracy: 0.6868\n",
            "Epoch 138/200\n",
            "2/2 [==============================] - 1s 648ms/step - loss: 1.7723 - accuracy: 0.6973 - val_loss: 3.4698 - val_accuracy: 0.6864\n",
            "Epoch 139/200\n",
            "2/2 [==============================] - 1s 631ms/step - loss: 1.7631 - accuracy: 0.6988 - val_loss: 3.4874 - val_accuracy: 0.6863\n",
            "Epoch 140/200\n",
            "2/2 [==============================] - 1s 632ms/step - loss: 1.7557 - accuracy: 0.6986 - val_loss: 3.4929 - val_accuracy: 0.6865\n",
            "Epoch 141/200\n",
            "2/2 [==============================] - 1s 632ms/step - loss: 1.7462 - accuracy: 0.6989 - val_loss: 3.4838 - val_accuracy: 0.6859\n",
            "Epoch 142/200\n",
            "2/2 [==============================] - 1s 631ms/step - loss: 1.7443 - accuracy: 0.7000 - val_loss: 3.5028 - val_accuracy: 0.6864\n",
            "Epoch 143/200\n",
            "2/2 [==============================] - 1s 637ms/step - loss: 1.7339 - accuracy: 0.6995 - val_loss: 3.4871 - val_accuracy: 0.6861\n",
            "Epoch 144/200\n",
            "2/2 [==============================] - 1s 646ms/step - loss: 1.7280 - accuracy: 0.7013 - val_loss: 3.5107 - val_accuracy: 0.6863\n",
            "Epoch 145/200\n",
            "2/2 [==============================] - 1s 631ms/step - loss: 1.7204 - accuracy: 0.7014 - val_loss: 3.5042 - val_accuracy: 0.6865\n",
            "Epoch 146/200\n",
            "2/2 [==============================] - 1s 650ms/step - loss: 1.7141 - accuracy: 0.7012 - val_loss: 3.5118 - val_accuracy: 0.6860\n",
            "Epoch 147/200\n",
            "2/2 [==============================] - 1s 633ms/step - loss: 1.7063 - accuracy: 0.7024 - val_loss: 3.5209 - val_accuracy: 0.6865\n",
            "Epoch 148/200\n",
            "2/2 [==============================] - 1s 640ms/step - loss: 1.7032 - accuracy: 0.7014 - val_loss: 3.5112 - val_accuracy: 0.6861\n",
            "Epoch 149/200\n",
            "2/2 [==============================] - 1s 632ms/step - loss: 1.6960 - accuracy: 0.7025 - val_loss: 3.5163 - val_accuracy: 0.6864\n",
            "Epoch 150/200\n",
            "2/2 [==============================] - 1s 638ms/step - loss: 1.6872 - accuracy: 0.7037 - val_loss: 3.5259 - val_accuracy: 0.6862\n",
            "Epoch 151/200\n",
            "2/2 [==============================] - 1s 641ms/step - loss: 1.6880 - accuracy: 0.7030 - val_loss: 3.5171 - val_accuracy: 0.6860\n",
            "Epoch 152/200\n",
            "2/2 [==============================] - 1s 638ms/step - loss: 1.6815 - accuracy: 0.7036 - val_loss: 3.5281 - val_accuracy: 0.6854\n",
            "Epoch 153/200\n",
            "2/2 [==============================] - 1s 633ms/step - loss: 1.6754 - accuracy: 0.7049 - val_loss: 3.5408 - val_accuracy: 0.6862\n",
            "Epoch 154/200\n",
            "2/2 [==============================] - 1s 632ms/step - loss: 1.6751 - accuracy: 0.7037 - val_loss: 3.5352 - val_accuracy: 0.6861\n",
            "Epoch 155/200\n",
            "2/2 [==============================] - 1s 633ms/step - loss: 1.6671 - accuracy: 0.7049 - val_loss: 3.5284 - val_accuracy: 0.6852\n",
            "Epoch 156/200\n",
            "2/2 [==============================] - 1s 631ms/step - loss: 1.6644 - accuracy: 0.7051 - val_loss: 3.5376 - val_accuracy: 0.6861\n",
            "Epoch 157/200\n",
            "2/2 [==============================] - 1s 633ms/step - loss: 1.6524 - accuracy: 0.7062 - val_loss: 3.5433 - val_accuracy: 0.6856\n",
            "Epoch 158/200\n",
            "2/2 [==============================] - 1s 635ms/step - loss: 1.6474 - accuracy: 0.7066 - val_loss: 3.5421 - val_accuracy: 0.6860\n",
            "Epoch 159/200\n",
            "2/2 [==============================] - 1s 633ms/step - loss: 1.6398 - accuracy: 0.7070 - val_loss: 3.5487 - val_accuracy: 0.6860\n",
            "Epoch 160/200\n",
            "2/2 [==============================] - 1s 632ms/step - loss: 1.6324 - accuracy: 0.7080 - val_loss: 3.5562 - val_accuracy: 0.6855\n",
            "Epoch 161/200\n",
            "2/2 [==============================] - 1s 631ms/step - loss: 1.6304 - accuracy: 0.7087 - val_loss: 3.5596 - val_accuracy: 0.6856\n",
            "Epoch 162/200\n",
            "2/2 [==============================] - 1s 635ms/step - loss: 1.6229 - accuracy: 0.7079 - val_loss: 3.5734 - val_accuracy: 0.6862\n",
            "Epoch 163/200\n",
            "2/2 [==============================] - 1s 634ms/step - loss: 1.6526 - accuracy: 0.7054 - val_loss: 3.6082 - val_accuracy: 0.6857\n",
            "Epoch 164/200\n",
            "2/2 [==============================] - 1s 639ms/step - loss: 1.6785 - accuracy: 0.7056 - val_loss: 3.5439 - val_accuracy: 0.6844\n",
            "Epoch 165/200\n",
            "2/2 [==============================] - 1s 647ms/step - loss: 1.6585 - accuracy: 0.7057 - val_loss: 3.5815 - val_accuracy: 0.6858\n",
            "Epoch 166/200\n",
            "2/2 [==============================] - 1s 633ms/step - loss: 1.6424 - accuracy: 0.7067 - val_loss: 3.5425 - val_accuracy: 0.6852\n",
            "Epoch 167/200\n",
            "2/2 [==============================] - 1s 633ms/step - loss: 1.6246 - accuracy: 0.7092 - val_loss: 3.5801 - val_accuracy: 0.6850\n",
            "Epoch 168/200\n",
            "2/2 [==============================] - 1s 634ms/step - loss: 1.6172 - accuracy: 0.7084 - val_loss: 3.5585 - val_accuracy: 0.6860\n",
            "Epoch 169/200\n",
            "2/2 [==============================] - 1s 631ms/step - loss: 1.6120 - accuracy: 0.7094 - val_loss: 3.5685 - val_accuracy: 0.6853\n",
            "Epoch 170/200\n",
            "2/2 [==============================] - 1s 632ms/step - loss: 1.6009 - accuracy: 0.7102 - val_loss: 3.5619 - val_accuracy: 0.6853\n",
            "Epoch 171/200\n",
            "2/2 [==============================] - 1s 632ms/step - loss: 1.5893 - accuracy: 0.7117 - val_loss: 3.5765 - val_accuracy: 0.6854\n",
            "Epoch 172/200\n",
            "2/2 [==============================] - 1s 635ms/step - loss: 1.5829 - accuracy: 0.7118 - val_loss: 3.5686 - val_accuracy: 0.6852\n",
            "Epoch 173/200\n",
            "2/2 [==============================] - 1s 635ms/step - loss: 1.5750 - accuracy: 0.7129 - val_loss: 3.5810 - val_accuracy: 0.6853\n",
            "Epoch 174/200\n",
            "2/2 [==============================] - 1s 634ms/step - loss: 1.5705 - accuracy: 0.7131 - val_loss: 3.5820 - val_accuracy: 0.6856\n",
            "Epoch 175/200\n",
            "2/2 [==============================] - 1s 638ms/step - loss: 1.5701 - accuracy: 0.7129 - val_loss: 3.5849 - val_accuracy: 0.6862\n",
            "Epoch 176/200\n",
            "2/2 [==============================] - 1s 634ms/step - loss: 1.5594 - accuracy: 0.7136 - val_loss: 3.5794 - val_accuracy: 0.6852\n",
            "Epoch 177/200\n",
            "2/2 [==============================] - 1s 633ms/step - loss: 1.5540 - accuracy: 0.7149 - val_loss: 3.6032 - val_accuracy: 0.6857\n",
            "Epoch 178/200\n",
            "2/2 [==============================] - 1s 634ms/step - loss: 1.5523 - accuracy: 0.7139 - val_loss: 3.5912 - val_accuracy: 0.6851\n",
            "Epoch 179/200\n",
            "2/2 [==============================] - 1s 633ms/step - loss: 1.5449 - accuracy: 0.7164 - val_loss: 3.6021 - val_accuracy: 0.6854\n",
            "Epoch 180/200\n",
            "2/2 [==============================] - 1s 642ms/step - loss: 1.5357 - accuracy: 0.7160 - val_loss: 3.6062 - val_accuracy: 0.6856\n",
            "Epoch 181/200\n",
            "2/2 [==============================] - 1s 637ms/step - loss: 1.5318 - accuracy: 0.7172 - val_loss: 3.5987 - val_accuracy: 0.6850\n",
            "Epoch 182/200\n",
            "2/2 [==============================] - 1s 647ms/step - loss: 1.5296 - accuracy: 0.7174 - val_loss: 3.6086 - val_accuracy: 0.6851\n",
            "Epoch 183/200\n",
            "2/2 [==============================] - 1s 635ms/step - loss: 1.5181 - accuracy: 0.7184 - val_loss: 3.6058 - val_accuracy: 0.6849\n",
            "Epoch 184/200\n",
            "2/2 [==============================] - 1s 647ms/step - loss: 1.5145 - accuracy: 0.7192 - val_loss: 3.6104 - val_accuracy: 0.6850\n",
            "Epoch 185/200\n",
            "2/2 [==============================] - 1s 633ms/step - loss: 1.5115 - accuracy: 0.7192 - val_loss: 3.6121 - val_accuracy: 0.6850\n",
            "Epoch 186/200\n",
            "2/2 [==============================] - 1s 634ms/step - loss: 1.5055 - accuracy: 0.7192 - val_loss: 3.6178 - val_accuracy: 0.6850\n",
            "Epoch 187/200\n",
            "2/2 [==============================] - 1s 632ms/step - loss: 1.5005 - accuracy: 0.7198 - val_loss: 3.6243 - val_accuracy: 0.6854\n",
            "Epoch 188/200\n",
            "2/2 [==============================] - 1s 630ms/step - loss: 1.4986 - accuracy: 0.7206 - val_loss: 3.6180 - val_accuracy: 0.6849\n",
            "Epoch 189/200\n",
            "2/2 [==============================] - 1s 635ms/step - loss: 1.4936 - accuracy: 0.7209 - val_loss: 3.6190 - val_accuracy: 0.6847\n",
            "Epoch 190/200\n",
            "2/2 [==============================] - 1s 647ms/step - loss: 1.4914 - accuracy: 0.7215 - val_loss: 3.6404 - val_accuracy: 0.6852\n",
            "Epoch 191/200\n",
            "2/2 [==============================] - 1s 650ms/step - loss: 1.4893 - accuracy: 0.7205 - val_loss: 3.6388 - val_accuracy: 0.6856\n",
            "Epoch 192/200\n",
            "2/2 [==============================] - 1s 631ms/step - loss: 1.4779 - accuracy: 0.7218 - val_loss: 3.6183 - val_accuracy: 0.6843\n",
            "Epoch 193/200\n",
            "2/2 [==============================] - 1s 647ms/step - loss: 1.4863 - accuracy: 0.7227 - val_loss: 3.6500 - val_accuracy: 0.6854\n",
            "Epoch 194/200\n",
            "2/2 [==============================] - 1s 631ms/step - loss: 1.4782 - accuracy: 0.7227 - val_loss: 3.6409 - val_accuracy: 0.6852\n",
            "Epoch 195/200\n",
            "2/2 [==============================] - 1s 633ms/step - loss: 1.4717 - accuracy: 0.7231 - val_loss: 3.6307 - val_accuracy: 0.6849\n",
            "Epoch 196/200\n",
            "2/2 [==============================] - 1s 634ms/step - loss: 1.4677 - accuracy: 0.7239 - val_loss: 3.6520 - val_accuracy: 0.6847\n",
            "Epoch 197/200\n",
            "2/2 [==============================] - 1s 634ms/step - loss: 1.4711 - accuracy: 0.7233 - val_loss: 3.6386 - val_accuracy: 0.6847\n",
            "Epoch 198/200\n",
            "2/2 [==============================] - 1s 639ms/step - loss: 1.4591 - accuracy: 0.7240 - val_loss: 3.6357 - val_accuracy: 0.6847\n",
            "Epoch 199/200\n",
            "2/2 [==============================] - 1s 637ms/step - loss: 1.4530 - accuracy: 0.7257 - val_loss: 3.6556 - val_accuracy: 0.6853\n",
            "Epoch 200/200\n",
            "2/2 [==============================] - 1s 640ms/step - loss: 1.4478 - accuracy: 0.7256 - val_loss: 3.6450 - val_accuracy: 0.6849\n",
            "- - - vum ebusweni yaka solo [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] anezi [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] ngazi [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "8_4El-alfluj"
      },
      "source": [
        "## Model 3 - RNN with Word Embedding and Standard Tokenization"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "OdNIDpNrL6Oz"
      },
      "source": [
        "def embed_model(input_shape, output_sequence_length, english_vocab_size, zulu_vocab_size):\n",
        "    \"\"\"\n",
        "    Build and train a RNN model using word embedding on x and y\n",
        "    :param input_shape: Tuple of input shape\n",
        "    :param output_sequence_length: Length of output sequence\n",
        "    :param english_vocab_size: Number of unique English words in the dataset\n",
        "    :param zulu_vocab_size: Number of unique zulu words in the dataset\n",
        "    :return: Keras model built, but not trained\n",
        "    \"\"\"\n",
        "    # TODO: Implement\n",
        "\n",
        "    # Hyperparameters\n",
        "    learning_rate = 0.005\n",
        "    \n",
        "    # TODO: Build the layers\n",
        "    model = Sequential()\n",
        "    model.add(Embedding(english_vocab_size, 256, input_length=input_shape[1], input_shape=input_shape[1:]))\n",
        "    model.add(GRU(256, return_sequences=True))    \n",
        "    model.add(TimeDistributed(Dense(1024, activation='relu')))\n",
        "    model.add(Dropout(0.5))\n",
        "    model.add(TimeDistributed(Dense(zulu_vocab_size, activation='softmax'))) \n",
        "\n",
        "    # Compile model\n",
        "    model.compile(loss=sparse_categorical_crossentropy,\n",
        "                  optimizer=keras.optimizers.Adam(learning_rate),\n",
        "                  metrics=['accuracy'])\n",
        "    return model\n",
        "\n",
        "#tests.test_embed_model(embed_model)\n",
        "\n",
        "# TODO: Reshape the input\n"
      ],
      "execution_count": 22,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "zC3D81I3nRJp",
        "outputId": "f9e685e5-7f7d-4783-d66c-4bb27c1f733c"
      },
      "source": [
        "tmp_x = pad(preproc_english_sentences, preproc_zulu_sentences.shape[1])\n",
        "tmp_x = tmp_x.reshape((-1, preproc_zulu_sentences.shape[-2]))\n",
        "tmp_x_test = pad(preproc_english_sentences_test, max_zulu_sequence_length)\n",
        "tmp_x_test = tmp_x_test.reshape((-1, preproc_zulu_sentences.shape[-2], 1))\n",
        "\n",
        "# TODO: Train the neural network\n",
        "embed_rnn_model = embed_model(\n",
        "    tmp_x.shape,\n",
        "    preproc_zulu_sentences.shape[1],\n",
        "    len(english_tokenizer.word_index)+1,\n",
        "    len(zulu_tokenizer.word_index)+1)\n",
        "\n",
        "embed_rnn_model.summary()\n",
        "\n",
        "model3 = embed_rnn_model.fit(tmp_x, preproc_zulu_sentences, batch_size=1024, epochs=NUM_EPOCHS, validation_split=0.2)\n",
        "\n",
        "print(logits_to_text(embed_rnn_model.predict(tmp_x[:1])[0], zulu_tokenizer))"
      ],
      "execution_count": 23,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential_2\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " embedding (Embedding)       (None, 25, 256)           1955840   \n",
            "                                                                 \n",
            " gru_2 (GRU)                 (None, 25, 256)           394752    \n",
            "                                                                 \n",
            " time_distributed_4 (TimeDis  (None, 25, 1024)         263168    \n",
            " tributed)                                                       \n",
            "                                                                 \n",
            " dropout_2 (Dropout)         (None, 25, 1024)          0         \n",
            "                                                                 \n",
            " time_distributed_5 (TimeDis  (None, 25, 12598)        12912950  \n",
            " tributed)                                                       \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 15,526,710\n",
            "Trainable params: 15,526,710\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n",
            "Epoch 1/200\n",
            "2/2 [==============================] - 3s 890ms/step - loss: 9.3878 - accuracy: 0.2217 - val_loss: 8.2959 - val_accuracy: 0.5831\n",
            "Epoch 2/200\n",
            "2/2 [==============================] - 1s 379ms/step - loss: 6.9186 - accuracy: 0.5948 - val_loss: 5.3069 - val_accuracy: 0.5831\n",
            "Epoch 3/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 5.2533 - accuracy: 0.5948 - val_loss: 5.9016 - val_accuracy: 0.5831\n",
            "Epoch 4/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 5.6125 - accuracy: 0.5948 - val_loss: 5.8871 - val_accuracy: 0.5831\n",
            "Epoch 5/200\n",
            "2/2 [==============================] - 1s 399ms/step - loss: 5.2703 - accuracy: 0.5928 - val_loss: 6.2075 - val_accuracy: 0.0022\n",
            "Epoch 6/200\n",
            "2/2 [==============================] - 1s 382ms/step - loss: 5.1525 - accuracy: 0.2729 - val_loss: 5.0438 - val_accuracy: 0.5802\n",
            "Epoch 7/200\n",
            "2/2 [==============================] - 1s 396ms/step - loss: 4.2936 - accuracy: 0.5909 - val_loss: 4.8545 - val_accuracy: 0.5829\n",
            "Epoch 8/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 4.2197 - accuracy: 0.5947 - val_loss: 4.8685 - val_accuracy: 0.5828\n",
            "Epoch 9/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 4.0534 - accuracy: 0.5947 - val_loss: 4.8951 - val_accuracy: 0.5828\n",
            "Epoch 10/200\n",
            "2/2 [==============================] - 1s 383ms/step - loss: 3.9380 - accuracy: 0.5946 - val_loss: 4.9647 - val_accuracy: 0.5826\n",
            "Epoch 11/200\n",
            "2/2 [==============================] - 1s 391ms/step - loss: 3.8605 - accuracy: 0.5945 - val_loss: 4.9841 - val_accuracy: 0.5828\n",
            "Epoch 12/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 3.8117 - accuracy: 0.5951 - val_loss: 5.0198 - val_accuracy: 0.5841\n",
            "Epoch 13/200\n",
            "2/2 [==============================] - 1s 382ms/step - loss: 3.7699 - accuracy: 0.5958 - val_loss: 5.0737 - val_accuracy: 0.5828\n",
            "Epoch 14/200\n",
            "2/2 [==============================] - 1s 383ms/step - loss: 3.7150 - accuracy: 0.5953 - val_loss: 5.0866 - val_accuracy: 0.5829\n",
            "Epoch 15/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 3.6733 - accuracy: 0.5951 - val_loss: 5.1260 - val_accuracy: 0.5830\n",
            "Epoch 16/200\n",
            "2/2 [==============================] - 1s 382ms/step - loss: 3.6259 - accuracy: 0.5954 - val_loss: 5.1153 - val_accuracy: 0.5840\n",
            "Epoch 17/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 3.5796 - accuracy: 0.5962 - val_loss: 5.1421 - val_accuracy: 0.5848\n",
            "Epoch 18/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 3.5405 - accuracy: 0.5965 - val_loss: 5.1471 - val_accuracy: 0.5843\n",
            "Epoch 19/200\n",
            "2/2 [==============================] - 1s 394ms/step - loss: 3.4913 - accuracy: 0.5964 - val_loss: 5.1894 - val_accuracy: 0.5850\n",
            "Epoch 20/200\n",
            "2/2 [==============================] - 1s 382ms/step - loss: 3.4522 - accuracy: 0.5967 - val_loss: 5.2161 - val_accuracy: 0.5847\n",
            "Epoch 21/200\n",
            "2/2 [==============================] - 1s 408ms/step - loss: 3.4091 - accuracy: 0.5967 - val_loss: 5.2456 - val_accuracy: 0.5843\n",
            "Epoch 22/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 3.3640 - accuracy: 0.5969 - val_loss: 5.2718 - val_accuracy: 0.5837\n",
            "Epoch 23/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 3.3181 - accuracy: 0.5967 - val_loss: 5.2955 - val_accuracy: 0.5837\n",
            "Epoch 24/200\n",
            "2/2 [==============================] - 1s 383ms/step - loss: 3.2768 - accuracy: 0.5963 - val_loss: 5.3200 - val_accuracy: 0.5841\n",
            "Epoch 25/200\n",
            "2/2 [==============================] - 1s 391ms/step - loss: 3.2308 - accuracy: 0.5961 - val_loss: 5.3494 - val_accuracy: 0.5846\n",
            "Epoch 26/200\n",
            "2/2 [==============================] - 1s 405ms/step - loss: 3.1838 - accuracy: 0.5966 - val_loss: 5.3764 - val_accuracy: 0.5845\n",
            "Epoch 27/200\n",
            "2/2 [==============================] - 1s 385ms/step - loss: 3.1411 - accuracy: 0.5967 - val_loss: 5.4053 - val_accuracy: 0.5843\n",
            "Epoch 28/200\n",
            "2/2 [==============================] - 1s 400ms/step - loss: 3.0917 - accuracy: 0.5971 - val_loss: 5.4327 - val_accuracy: 0.5839\n",
            "Epoch 29/200\n",
            "2/2 [==============================] - 1s 390ms/step - loss: 3.0423 - accuracy: 0.5972 - val_loss: 5.4701 - val_accuracy: 0.5838\n",
            "Epoch 30/200\n",
            "2/2 [==============================] - 1s 383ms/step - loss: 2.9907 - accuracy: 0.5981 - val_loss: 5.5069 - val_accuracy: 0.5843\n",
            "Epoch 31/200\n",
            "2/2 [==============================] - 1s 382ms/step - loss: 2.9399 - accuracy: 0.5981 - val_loss: 5.5411 - val_accuracy: 0.5844\n",
            "Epoch 32/200\n",
            "2/2 [==============================] - 1s 388ms/step - loss: 2.8890 - accuracy: 0.5985 - val_loss: 5.5716 - val_accuracy: 0.5824\n",
            "Epoch 33/200\n",
            "2/2 [==============================] - 1s 389ms/step - loss: 2.8345 - accuracy: 0.5987 - val_loss: 5.5930 - val_accuracy: 0.5827\n",
            "Epoch 34/200\n",
            "2/2 [==============================] - 1s 385ms/step - loss: 2.7790 - accuracy: 0.5996 - val_loss: 5.6242 - val_accuracy: 0.5811\n",
            "Epoch 35/200\n",
            "2/2 [==============================] - 1s 399ms/step - loss: 2.7275 - accuracy: 0.5997 - val_loss: 5.6531 - val_accuracy: 0.5793\n",
            "Epoch 36/200\n",
            "2/2 [==============================] - 1s 383ms/step - loss: 2.6807 - accuracy: 0.6006 - val_loss: 5.6673 - val_accuracy: 0.5790\n",
            "Epoch 37/200\n",
            "2/2 [==============================] - 1s 383ms/step - loss: 2.6205 - accuracy: 0.6019 - val_loss: 5.6904 - val_accuracy: 0.5810\n",
            "Epoch 38/200\n",
            "2/2 [==============================] - 1s 383ms/step - loss: 2.5710 - accuracy: 0.6020 - val_loss: 5.7121 - val_accuracy: 0.5774\n",
            "Epoch 39/200\n",
            "2/2 [==============================] - 1s 391ms/step - loss: 2.5184 - accuracy: 0.6042 - val_loss: 5.7429 - val_accuracy: 0.5740\n",
            "Epoch 40/200\n",
            "2/2 [==============================] - 1s 398ms/step - loss: 2.4854 - accuracy: 0.6044 - val_loss: 5.7457 - val_accuracy: 0.5757\n",
            "Epoch 41/200\n",
            "2/2 [==============================] - 1s 389ms/step - loss: 2.4208 - accuracy: 0.6077 - val_loss: 5.7447 - val_accuracy: 0.5787\n",
            "Epoch 42/200\n",
            "2/2 [==============================] - 1s 394ms/step - loss: 2.3843 - accuracy: 0.6081 - val_loss: 5.7750 - val_accuracy: 0.5728\n",
            "Epoch 43/200\n",
            "2/2 [==============================] - 1s 383ms/step - loss: 2.3342 - accuracy: 0.6112 - val_loss: 5.7757 - val_accuracy: 0.5757\n",
            "Epoch 44/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 2.2814 - accuracy: 0.6153 - val_loss: 5.7903 - val_accuracy: 0.5734\n",
            "Epoch 45/200\n",
            "2/2 [==============================] - 1s 392ms/step - loss: 2.2252 - accuracy: 0.6185 - val_loss: 5.7963 - val_accuracy: 0.5720\n",
            "Epoch 46/200\n",
            "2/2 [==============================] - 1s 395ms/step - loss: 2.1780 - accuracy: 0.6201 - val_loss: 5.7972 - val_accuracy: 0.5741\n",
            "Epoch 47/200\n",
            "2/2 [==============================] - 1s 382ms/step - loss: 2.1353 - accuracy: 0.6251 - val_loss: 5.8299 - val_accuracy: 0.5708\n",
            "Epoch 48/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 2.0863 - accuracy: 0.6314 - val_loss: 5.8309 - val_accuracy: 0.5716\n",
            "Epoch 49/200\n",
            "2/2 [==============================] - 1s 406ms/step - loss: 2.0280 - accuracy: 0.6372 - val_loss: 5.8517 - val_accuracy: 0.5712\n",
            "Epoch 50/200\n",
            "2/2 [==============================] - 1s 383ms/step - loss: 1.9845 - accuracy: 0.6399 - val_loss: 5.8559 - val_accuracy: 0.5680\n",
            "Epoch 51/200\n",
            "2/2 [==============================] - 1s 388ms/step - loss: 1.9382 - accuracy: 0.6470 - val_loss: 5.8777 - val_accuracy: 0.5713\n",
            "Epoch 52/200\n",
            "2/2 [==============================] - 1s 385ms/step - loss: 1.8884 - accuracy: 0.6510 - val_loss: 5.8835 - val_accuracy: 0.5685\n",
            "Epoch 53/200\n",
            "2/2 [==============================] - 1s 398ms/step - loss: 1.8400 - accuracy: 0.6595 - val_loss: 5.9096 - val_accuracy: 0.5669\n",
            "Epoch 54/200\n",
            "2/2 [==============================] - 1s 397ms/step - loss: 1.7858 - accuracy: 0.6664 - val_loss: 5.9061 - val_accuracy: 0.5697\n",
            "Epoch 55/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 1.7336 - accuracy: 0.6744 - val_loss: 5.9308 - val_accuracy: 0.5634\n",
            "Epoch 56/200\n",
            "2/2 [==============================] - 1s 399ms/step - loss: 1.6953 - accuracy: 0.6783 - val_loss: 5.9434 - val_accuracy: 0.5646\n",
            "Epoch 57/200\n",
            "2/2 [==============================] - 1s 382ms/step - loss: 1.6448 - accuracy: 0.6858 - val_loss: 5.9458 - val_accuracy: 0.5670\n",
            "Epoch 58/200\n",
            "2/2 [==============================] - 1s 383ms/step - loss: 1.6040 - accuracy: 0.6922 - val_loss: 5.9691 - val_accuracy: 0.5636\n",
            "Epoch 59/200\n",
            "2/2 [==============================] - 1s 385ms/step - loss: 1.5616 - accuracy: 0.7007 - val_loss: 5.9756 - val_accuracy: 0.5619\n",
            "Epoch 60/200\n",
            "2/2 [==============================] - 1s 388ms/step - loss: 1.5139 - accuracy: 0.7075 - val_loss: 5.9822 - val_accuracy: 0.5653\n",
            "Epoch 61/200\n",
            "2/2 [==============================] - 1s 390ms/step - loss: 1.4639 - accuracy: 0.7150 - val_loss: 5.9925 - val_accuracy: 0.5635\n",
            "Epoch 62/200\n",
            "2/2 [==============================] - 1s 390ms/step - loss: 1.4266 - accuracy: 0.7222 - val_loss: 6.0105 - val_accuracy: 0.5620\n",
            "Epoch 63/200\n",
            "2/2 [==============================] - 1s 396ms/step - loss: 1.3837 - accuracy: 0.7288 - val_loss: 6.0174 - val_accuracy: 0.5628\n",
            "Epoch 64/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 1.3410 - accuracy: 0.7363 - val_loss: 6.0190 - val_accuracy: 0.5644\n",
            "Epoch 65/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 1.3158 - accuracy: 0.7416 - val_loss: 6.0356 - val_accuracy: 0.5629\n",
            "Epoch 66/200\n",
            "2/2 [==============================] - 1s 383ms/step - loss: 1.2751 - accuracy: 0.7479 - val_loss: 6.0501 - val_accuracy: 0.5587\n",
            "Epoch 67/200\n",
            "2/2 [==============================] - 1s 395ms/step - loss: 1.2500 - accuracy: 0.7528 - val_loss: 6.0660 - val_accuracy: 0.5609\n",
            "Epoch 68/200\n",
            "2/2 [==============================] - 1s 389ms/step - loss: 1.2097 - accuracy: 0.7604 - val_loss: 6.0494 - val_accuracy: 0.5648\n",
            "Epoch 69/200\n",
            "2/2 [==============================] - 1s 388ms/step - loss: 1.1755 - accuracy: 0.7661 - val_loss: 6.0785 - val_accuracy: 0.5587\n",
            "Epoch 70/200\n",
            "2/2 [==============================] - 1s 385ms/step - loss: 1.1395 - accuracy: 0.7721 - val_loss: 6.0821 - val_accuracy: 0.5601\n",
            "Epoch 71/200\n",
            "2/2 [==============================] - 1s 385ms/step - loss: 1.1093 - accuracy: 0.7783 - val_loss: 6.0821 - val_accuracy: 0.5634\n",
            "Epoch 72/200\n",
            "2/2 [==============================] - 1s 391ms/step - loss: 1.0779 - accuracy: 0.7834 - val_loss: 6.0989 - val_accuracy: 0.5596\n",
            "Epoch 73/200\n",
            "2/2 [==============================] - 1s 381ms/step - loss: 1.0532 - accuracy: 0.7887 - val_loss: 6.1075 - val_accuracy: 0.5590\n",
            "Epoch 74/200\n",
            "2/2 [==============================] - 1s 398ms/step - loss: 1.0172 - accuracy: 0.7963 - val_loss: 6.1110 - val_accuracy: 0.5628\n",
            "Epoch 75/200\n",
            "2/2 [==============================] - 1s 382ms/step - loss: 0.9948 - accuracy: 0.8015 - val_loss: 6.1298 - val_accuracy: 0.5585\n",
            "Epoch 76/200\n",
            "2/2 [==============================] - 1s 382ms/step - loss: 0.9653 - accuracy: 0.8062 - val_loss: 6.1280 - val_accuracy: 0.5618\n",
            "Epoch 77/200\n",
            "2/2 [==============================] - 1s 390ms/step - loss: 0.9405 - accuracy: 0.8122 - val_loss: 6.1315 - val_accuracy: 0.5612\n",
            "Epoch 78/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 0.9140 - accuracy: 0.8152 - val_loss: 6.1471 - val_accuracy: 0.5589\n",
            "Epoch 79/200\n",
            "2/2 [==============================] - 1s 385ms/step - loss: 0.8857 - accuracy: 0.8223 - val_loss: 6.1405 - val_accuracy: 0.5629\n",
            "Epoch 80/200\n",
            "2/2 [==============================] - 1s 389ms/step - loss: 0.8663 - accuracy: 0.8263 - val_loss: 6.1572 - val_accuracy: 0.5592\n",
            "Epoch 81/200\n",
            "2/2 [==============================] - 1s 395ms/step - loss: 0.8414 - accuracy: 0.8320 - val_loss: 6.1674 - val_accuracy: 0.5568\n",
            "Epoch 82/200\n",
            "2/2 [==============================] - 1s 395ms/step - loss: 0.8156 - accuracy: 0.8372 - val_loss: 6.1533 - val_accuracy: 0.5625\n",
            "Epoch 83/200\n",
            "2/2 [==============================] - 1s 385ms/step - loss: 0.8039 - accuracy: 0.8394 - val_loss: 6.1908 - val_accuracy: 0.5571\n",
            "Epoch 84/200\n",
            "2/2 [==============================] - 1s 383ms/step - loss: 0.7869 - accuracy: 0.8431 - val_loss: 6.1915 - val_accuracy: 0.5587\n",
            "Epoch 85/200\n",
            "2/2 [==============================] - 1s 382ms/step - loss: 0.7582 - accuracy: 0.8482 - val_loss: 6.1852 - val_accuracy: 0.5624\n",
            "Epoch 86/200\n",
            "2/2 [==============================] - 1s 397ms/step - loss: 0.7485 - accuracy: 0.8499 - val_loss: 6.2172 - val_accuracy: 0.5575\n",
            "Epoch 87/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 0.7385 - accuracy: 0.8521 - val_loss: 6.1955 - val_accuracy: 0.5610\n",
            "Epoch 88/200\n",
            "2/2 [==============================] - 1s 388ms/step - loss: 0.7074 - accuracy: 0.8584 - val_loss: 6.2118 - val_accuracy: 0.5582\n",
            "Epoch 89/200\n",
            "2/2 [==============================] - 1s 392ms/step - loss: 0.6993 - accuracy: 0.8611 - val_loss: 6.2194 - val_accuracy: 0.5587\n",
            "Epoch 90/200\n",
            "2/2 [==============================] - 1s 389ms/step - loss: 0.6792 - accuracy: 0.8658 - val_loss: 6.2132 - val_accuracy: 0.5617\n",
            "Epoch 91/200\n",
            "2/2 [==============================] - 1s 400ms/step - loss: 0.6617 - accuracy: 0.8683 - val_loss: 6.2352 - val_accuracy: 0.5580\n",
            "Epoch 92/200\n",
            "2/2 [==============================] - 1s 403ms/step - loss: 0.6461 - accuracy: 0.8739 - val_loss: 6.2240 - val_accuracy: 0.5624\n",
            "Epoch 93/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 0.6390 - accuracy: 0.8748 - val_loss: 6.2415 - val_accuracy: 0.5578\n",
            "Epoch 94/200\n",
            "2/2 [==============================] - 1s 401ms/step - loss: 0.6275 - accuracy: 0.8761 - val_loss: 6.2384 - val_accuracy: 0.5603\n",
            "Epoch 95/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 0.6095 - accuracy: 0.8806 - val_loss: 6.2344 - val_accuracy: 0.5615\n",
            "Epoch 96/200\n",
            "2/2 [==============================] - 1s 385ms/step - loss: 0.5993 - accuracy: 0.8836 - val_loss: 6.2622 - val_accuracy: 0.5577\n",
            "Epoch 97/200\n",
            "2/2 [==============================] - 1s 391ms/step - loss: 0.5838 - accuracy: 0.8864 - val_loss: 6.2517 - val_accuracy: 0.5628\n",
            "Epoch 98/200\n",
            "2/2 [==============================] - 1s 390ms/step - loss: 0.5798 - accuracy: 0.8873 - val_loss: 6.2651 - val_accuracy: 0.5577\n",
            "Epoch 99/200\n",
            "2/2 [==============================] - 1s 381ms/step - loss: 0.5682 - accuracy: 0.8885 - val_loss: 6.2630 - val_accuracy: 0.5591\n",
            "Epoch 100/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 0.5548 - accuracy: 0.8930 - val_loss: 6.2631 - val_accuracy: 0.5605\n",
            "Epoch 101/200\n",
            "2/2 [==============================] - 1s 388ms/step - loss: 0.5495 - accuracy: 0.8927 - val_loss: 6.2762 - val_accuracy: 0.5585\n",
            "Epoch 102/200\n",
            "2/2 [==============================] - 1s 398ms/step - loss: 0.5346 - accuracy: 0.8957 - val_loss: 6.2750 - val_accuracy: 0.5604\n",
            "Epoch 103/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 0.5234 - accuracy: 0.8997 - val_loss: 6.2906 - val_accuracy: 0.5580\n",
            "Epoch 104/200\n",
            "2/2 [==============================] - 1s 400ms/step - loss: 0.5140 - accuracy: 0.9025 - val_loss: 6.2855 - val_accuracy: 0.5604\n",
            "Epoch 105/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 0.5119 - accuracy: 0.9028 - val_loss: 6.2991 - val_accuracy: 0.5589\n",
            "Epoch 106/200\n",
            "2/2 [==============================] - 1s 390ms/step - loss: 0.4981 - accuracy: 0.9045 - val_loss: 6.3007 - val_accuracy: 0.5607\n",
            "Epoch 107/200\n",
            "2/2 [==============================] - 1s 389ms/step - loss: 0.4900 - accuracy: 0.9071 - val_loss: 6.3124 - val_accuracy: 0.5603\n",
            "Epoch 108/200\n",
            "2/2 [==============================] - 1s 388ms/step - loss: 0.4809 - accuracy: 0.9087 - val_loss: 6.3108 - val_accuracy: 0.5604\n",
            "Epoch 109/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 0.4796 - accuracy: 0.9090 - val_loss: 6.3118 - val_accuracy: 0.5605\n",
            "Epoch 110/200\n",
            "2/2 [==============================] - 1s 382ms/step - loss: 0.4717 - accuracy: 0.9124 - val_loss: 6.3217 - val_accuracy: 0.5596\n",
            "Epoch 111/200\n",
            "2/2 [==============================] - 1s 389ms/step - loss: 0.4628 - accuracy: 0.9128 - val_loss: 6.3178 - val_accuracy: 0.5605\n",
            "Epoch 112/200\n",
            "2/2 [==============================] - 1s 398ms/step - loss: 0.4516 - accuracy: 0.9161 - val_loss: 6.3196 - val_accuracy: 0.5606\n",
            "Epoch 113/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 0.4504 - accuracy: 0.9168 - val_loss: 6.3280 - val_accuracy: 0.5601\n",
            "Epoch 114/200\n",
            "2/2 [==============================] - 1s 400ms/step - loss: 0.4453 - accuracy: 0.9180 - val_loss: 6.3245 - val_accuracy: 0.5614\n",
            "Epoch 115/200\n",
            "2/2 [==============================] - 1s 385ms/step - loss: 0.4394 - accuracy: 0.9179 - val_loss: 6.3277 - val_accuracy: 0.5608\n",
            "Epoch 116/200\n",
            "2/2 [==============================] - 1s 388ms/step - loss: 0.4299 - accuracy: 0.9207 - val_loss: 6.3269 - val_accuracy: 0.5620\n",
            "Epoch 117/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 0.4244 - accuracy: 0.9221 - val_loss: 6.3359 - val_accuracy: 0.5598\n",
            "Epoch 118/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 0.4206 - accuracy: 0.9244 - val_loss: 6.3391 - val_accuracy: 0.5603\n",
            "Epoch 119/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 0.4126 - accuracy: 0.9245 - val_loss: 6.3447 - val_accuracy: 0.5593\n",
            "Epoch 120/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 0.4075 - accuracy: 0.9272 - val_loss: 6.3373 - val_accuracy: 0.5602\n",
            "Epoch 121/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 0.4006 - accuracy: 0.9287 - val_loss: 6.3512 - val_accuracy: 0.5599\n",
            "Epoch 122/200\n",
            "2/2 [==============================] - 1s 394ms/step - loss: 0.4014 - accuracy: 0.9275 - val_loss: 6.3541 - val_accuracy: 0.5604\n",
            "Epoch 123/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 0.3971 - accuracy: 0.9289 - val_loss: 6.3598 - val_accuracy: 0.5600\n",
            "Epoch 124/200\n",
            "2/2 [==============================] - 1s 398ms/step - loss: 0.3957 - accuracy: 0.9291 - val_loss: 6.3587 - val_accuracy: 0.5619\n",
            "Epoch 125/200\n",
            "2/2 [==============================] - 1s 400ms/step - loss: 0.3845 - accuracy: 0.9316 - val_loss: 6.3690 - val_accuracy: 0.5605\n",
            "Epoch 126/200\n",
            "2/2 [==============================] - 1s 392ms/step - loss: 0.3847 - accuracy: 0.9312 - val_loss: 6.3680 - val_accuracy: 0.5608\n",
            "Epoch 127/200\n",
            "2/2 [==============================] - 1s 400ms/step - loss: 0.3741 - accuracy: 0.9340 - val_loss: 6.3721 - val_accuracy: 0.5602\n",
            "Epoch 128/200\n",
            "2/2 [==============================] - 1s 398ms/step - loss: 0.3790 - accuracy: 0.9330 - val_loss: 6.3751 - val_accuracy: 0.5603\n",
            "Epoch 129/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 0.3719 - accuracy: 0.9348 - val_loss: 6.3725 - val_accuracy: 0.5610\n",
            "Epoch 130/200\n",
            "2/2 [==============================] - 1s 401ms/step - loss: 0.3656 - accuracy: 0.9368 - val_loss: 6.3865 - val_accuracy: 0.5591\n",
            "Epoch 131/200\n",
            "2/2 [==============================] - 1s 388ms/step - loss: 0.3616 - accuracy: 0.9376 - val_loss: 6.3798 - val_accuracy: 0.5611\n",
            "Epoch 132/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 0.3594 - accuracy: 0.9383 - val_loss: 6.3961 - val_accuracy: 0.5586\n",
            "Epoch 133/200\n",
            "2/2 [==============================] - 1s 388ms/step - loss: 0.3550 - accuracy: 0.9400 - val_loss: 6.3834 - val_accuracy: 0.5615\n",
            "Epoch 134/200\n",
            "2/2 [==============================] - 1s 389ms/step - loss: 0.3523 - accuracy: 0.9401 - val_loss: 6.4069 - val_accuracy: 0.5594\n",
            "Epoch 135/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 0.3493 - accuracy: 0.9407 - val_loss: 6.3944 - val_accuracy: 0.5611\n",
            "Epoch 136/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 0.3496 - accuracy: 0.9402 - val_loss: 6.4082 - val_accuracy: 0.5587\n",
            "Epoch 137/200\n",
            "2/2 [==============================] - 1s 395ms/step - loss: 0.3447 - accuracy: 0.9412 - val_loss: 6.3977 - val_accuracy: 0.5602\n",
            "Epoch 138/200\n",
            "2/2 [==============================] - 1s 394ms/step - loss: 0.3421 - accuracy: 0.9414 - val_loss: 6.4074 - val_accuracy: 0.5593\n",
            "Epoch 139/200\n",
            "2/2 [==============================] - 1s 396ms/step - loss: 0.3395 - accuracy: 0.9420 - val_loss: 6.3976 - val_accuracy: 0.5611\n",
            "Epoch 140/200\n",
            "2/2 [==============================] - 1s 396ms/step - loss: 0.3362 - accuracy: 0.9429 - val_loss: 6.4094 - val_accuracy: 0.5592\n",
            "Epoch 141/200\n",
            "2/2 [==============================] - 1s 402ms/step - loss: 0.3309 - accuracy: 0.9450 - val_loss: 6.4074 - val_accuracy: 0.5603\n",
            "Epoch 142/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 0.3337 - accuracy: 0.9437 - val_loss: 6.4121 - val_accuracy: 0.5594\n",
            "Epoch 143/200\n",
            "2/2 [==============================] - 1s 394ms/step - loss: 0.3260 - accuracy: 0.9459 - val_loss: 6.4039 - val_accuracy: 0.5614\n",
            "Epoch 144/200\n",
            "2/2 [==============================] - 1s 382ms/step - loss: 0.3245 - accuracy: 0.9468 - val_loss: 6.4221 - val_accuracy: 0.5597\n",
            "Epoch 145/200\n",
            "2/2 [==============================] - 1s 385ms/step - loss: 0.3234 - accuracy: 0.9464 - val_loss: 6.4146 - val_accuracy: 0.5615\n",
            "Epoch 146/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 0.3221 - accuracy: 0.9473 - val_loss: 6.4294 - val_accuracy: 0.5595\n",
            "Epoch 147/200\n",
            "2/2 [==============================] - 1s 383ms/step - loss: 0.3184 - accuracy: 0.9483 - val_loss: 6.4218 - val_accuracy: 0.5609\n",
            "Epoch 148/200\n",
            "2/2 [==============================] - 1s 394ms/step - loss: 0.3199 - accuracy: 0.9468 - val_loss: 6.4322 - val_accuracy: 0.5598\n",
            "Epoch 149/200\n",
            "2/2 [==============================] - 1s 385ms/step - loss: 0.3160 - accuracy: 0.9485 - val_loss: 6.4361 - val_accuracy: 0.5594\n",
            "Epoch 150/200\n",
            "2/2 [==============================] - 1s 396ms/step - loss: 0.3150 - accuracy: 0.9481 - val_loss: 6.4364 - val_accuracy: 0.5594\n",
            "Epoch 151/200\n",
            "2/2 [==============================] - 1s 403ms/step - loss: 0.3107 - accuracy: 0.9501 - val_loss: 6.4348 - val_accuracy: 0.5596\n",
            "Epoch 152/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 0.3110 - accuracy: 0.9498 - val_loss: 6.4410 - val_accuracy: 0.5592\n",
            "Epoch 153/200\n",
            "2/2 [==============================] - 1s 402ms/step - loss: 0.3092 - accuracy: 0.9498 - val_loss: 6.4409 - val_accuracy: 0.5596\n",
            "Epoch 154/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 0.3070 - accuracy: 0.9500 - val_loss: 6.4401 - val_accuracy: 0.5598\n",
            "Epoch 155/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 0.3056 - accuracy: 0.9517 - val_loss: 6.4352 - val_accuracy: 0.5607\n",
            "Epoch 156/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 0.3039 - accuracy: 0.9518 - val_loss: 6.4441 - val_accuracy: 0.5587\n",
            "Epoch 157/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 0.3005 - accuracy: 0.9530 - val_loss: 6.4460 - val_accuracy: 0.5586\n",
            "Epoch 158/200\n",
            "2/2 [==============================] - 1s 389ms/step - loss: 0.3010 - accuracy: 0.9521 - val_loss: 6.4476 - val_accuracy: 0.5588\n",
            "Epoch 159/200\n",
            "2/2 [==============================] - 1s 391ms/step - loss: 0.2967 - accuracy: 0.9531 - val_loss: 6.4498 - val_accuracy: 0.5591\n",
            "Epoch 160/200\n",
            "2/2 [==============================] - 1s 395ms/step - loss: 0.2976 - accuracy: 0.9532 - val_loss: 6.4441 - val_accuracy: 0.5605\n",
            "Epoch 161/200\n",
            "2/2 [==============================] - 1s 395ms/step - loss: 0.2957 - accuracy: 0.9540 - val_loss: 6.4534 - val_accuracy: 0.5586\n",
            "Epoch 162/200\n",
            "2/2 [==============================] - 1s 389ms/step - loss: 0.2961 - accuracy: 0.9535 - val_loss: 6.4488 - val_accuracy: 0.5603\n",
            "Epoch 163/200\n",
            "2/2 [==============================] - 1s 400ms/step - loss: 0.2960 - accuracy: 0.9522 - val_loss: 6.4598 - val_accuracy: 0.5591\n",
            "Epoch 164/200\n",
            "2/2 [==============================] - 1s 393ms/step - loss: 0.2922 - accuracy: 0.9541 - val_loss: 6.4579 - val_accuracy: 0.5597\n",
            "Epoch 165/200\n",
            "2/2 [==============================] - 1s 391ms/step - loss: 0.2928 - accuracy: 0.9542 - val_loss: 6.4509 - val_accuracy: 0.5604\n",
            "Epoch 166/200\n",
            "2/2 [==============================] - 1s 398ms/step - loss: 0.2908 - accuracy: 0.9544 - val_loss: 6.4543 - val_accuracy: 0.5594\n",
            "Epoch 167/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 0.2890 - accuracy: 0.9544 - val_loss: 6.4465 - val_accuracy: 0.5606\n",
            "Epoch 168/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 0.2899 - accuracy: 0.9549 - val_loss: 6.4663 - val_accuracy: 0.5585\n",
            "Epoch 169/200\n",
            "2/2 [==============================] - 1s 385ms/step - loss: 0.2845 - accuracy: 0.9559 - val_loss: 6.4517 - val_accuracy: 0.5611\n",
            "Epoch 170/200\n",
            "2/2 [==============================] - 1s 389ms/step - loss: 0.2873 - accuracy: 0.9548 - val_loss: 6.4658 - val_accuracy: 0.5586\n",
            "Epoch 171/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 0.2831 - accuracy: 0.9561 - val_loss: 6.4619 - val_accuracy: 0.5592\n",
            "Epoch 172/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 0.2838 - accuracy: 0.9566 - val_loss: 6.4607 - val_accuracy: 0.5595\n",
            "Epoch 173/200\n",
            "2/2 [==============================] - 1s 397ms/step - loss: 0.2835 - accuracy: 0.9572 - val_loss: 6.4580 - val_accuracy: 0.5604\n",
            "Epoch 174/200\n",
            "2/2 [==============================] - 1s 402ms/step - loss: 0.2807 - accuracy: 0.9573 - val_loss: 6.4665 - val_accuracy: 0.5592\n",
            "Epoch 175/200\n",
            "2/2 [==============================] - 1s 388ms/step - loss: 0.2797 - accuracy: 0.9580 - val_loss: 6.4651 - val_accuracy: 0.5595\n",
            "Epoch 176/200\n",
            "2/2 [==============================] - 1s 403ms/step - loss: 0.2820 - accuracy: 0.9568 - val_loss: 6.4737 - val_accuracy: 0.5586\n",
            "Epoch 177/200\n",
            "2/2 [==============================] - 1s 397ms/step - loss: 0.2787 - accuracy: 0.9577 - val_loss: 6.4681 - val_accuracy: 0.5597\n",
            "Epoch 178/200\n",
            "2/2 [==============================] - 1s 388ms/step - loss: 0.2797 - accuracy: 0.9572 - val_loss: 6.4691 - val_accuracy: 0.5597\n",
            "Epoch 179/200\n",
            "2/2 [==============================] - 1s 404ms/step - loss: 0.2774 - accuracy: 0.9578 - val_loss: 6.4687 - val_accuracy: 0.5599\n",
            "Epoch 180/200\n",
            "2/2 [==============================] - 1s 399ms/step - loss: 0.2754 - accuracy: 0.9579 - val_loss: 6.4776 - val_accuracy: 0.5587\n",
            "Epoch 181/200\n",
            "2/2 [==============================] - 1s 388ms/step - loss: 0.2728 - accuracy: 0.9594 - val_loss: 6.4719 - val_accuracy: 0.5598\n",
            "Epoch 182/200\n",
            "2/2 [==============================] - 1s 400ms/step - loss: 0.2719 - accuracy: 0.9591 - val_loss: 6.4780 - val_accuracy: 0.5596\n",
            "Epoch 183/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 0.2728 - accuracy: 0.9589 - val_loss: 6.4882 - val_accuracy: 0.5589\n",
            "Epoch 184/200\n",
            "2/2 [==============================] - 1s 393ms/step - loss: 0.2718 - accuracy: 0.9593 - val_loss: 6.4693 - val_accuracy: 0.5615\n",
            "Epoch 185/200\n",
            "2/2 [==============================] - 1s 412ms/step - loss: 0.2724 - accuracy: 0.9594 - val_loss: 6.4782 - val_accuracy: 0.5603\n",
            "Epoch 186/200\n",
            "2/2 [==============================] - 1s 403ms/step - loss: 0.2706 - accuracy: 0.9597 - val_loss: 6.4759 - val_accuracy: 0.5604\n",
            "Epoch 187/200\n",
            "2/2 [==============================] - 1s 385ms/step - loss: 0.2695 - accuracy: 0.9601 - val_loss: 6.4743 - val_accuracy: 0.5609\n",
            "Epoch 188/200\n",
            "2/2 [==============================] - 1s 400ms/step - loss: 0.2716 - accuracy: 0.9591 - val_loss: 6.4882 - val_accuracy: 0.5590\n",
            "Epoch 189/200\n",
            "2/2 [==============================] - 1s 383ms/step - loss: 0.2700 - accuracy: 0.9604 - val_loss: 6.4689 - val_accuracy: 0.5624\n",
            "Epoch 190/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 0.2693 - accuracy: 0.9599 - val_loss: 6.4804 - val_accuracy: 0.5609\n",
            "Epoch 191/200\n",
            "2/2 [==============================] - 1s 402ms/step - loss: 0.2690 - accuracy: 0.9601 - val_loss: 6.4883 - val_accuracy: 0.5604\n",
            "Epoch 192/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 0.2691 - accuracy: 0.9598 - val_loss: 6.4801 - val_accuracy: 0.5615\n",
            "Epoch 193/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 0.2655 - accuracy: 0.9605 - val_loss: 6.4964 - val_accuracy: 0.5595\n",
            "Epoch 194/200\n",
            "2/2 [==============================] - 1s 385ms/step - loss: 0.2658 - accuracy: 0.9610 - val_loss: 6.4937 - val_accuracy: 0.5602\n",
            "Epoch 195/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 0.2641 - accuracy: 0.9612 - val_loss: 6.4897 - val_accuracy: 0.5608\n",
            "Epoch 196/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 0.2635 - accuracy: 0.9611 - val_loss: 6.5076 - val_accuracy: 0.5583\n",
            "Epoch 197/200\n",
            "2/2 [==============================] - 1s 390ms/step - loss: 0.2659 - accuracy: 0.9605 - val_loss: 6.5085 - val_accuracy: 0.5589\n",
            "Epoch 198/200\n",
            "2/2 [==============================] - 1s 399ms/step - loss: 0.2633 - accuracy: 0.9603 - val_loss: 6.5031 - val_accuracy: 0.5598\n",
            "Epoch 199/200\n",
            "2/2 [==============================] - 1s 400ms/step - loss: 0.2633 - accuracy: 0.9605 - val_loss: 6.5136 - val_accuracy: 0.5583\n",
            "Epoch 200/200\n",
            "2/2 [==============================] - 1s 389ms/step - loss: 0.2637 - accuracy: 0.9602 - val_loss: 6.5081 - val_accuracy: 0.5589\n",
            "ukufaka albhamu akhe e solo eminyaka yawo 60 angamanye anezingoma ezimnandi kakhulu kwengake ngazizwa <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD>\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "m_T5_ysJftu1"
      },
      "source": [
        "## Model 4 - RNN with Word Embedding and Byte-Pair Encoding"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "bW3apkl6mYlH",
        "outputId": "810ef6e4-9cbe-4ade-ca44-4da8abac791d"
      },
      "source": [
        "tmp_x_bpe = pad(preproc_english_sentences_bpe, preproc_zulu_sentences_bpe.shape[1])\n",
        "tmp_x_bpe = tmp_x_bpe.reshape((-1, preproc_zulu_sentences_bpe.shape[-2]))\n",
        "tmp_x_test_bpe = pad(preproc_english_sentences_test_bpe, max_zulu_sequence_length_bpe)\n",
        "tmp_x_test_bpe = tmp_x_test_bpe.reshape((-1, preproc_zulu_sentences_bpe.shape[-2], 1))\n",
        "\n",
        "# TODO: Train the neural network\n",
        "embed_rnn_model_bpe = embed_model(\n",
        "    tmp_x_bpe.shape,\n",
        "    preproc_zulu_sentences_bpe.shape[1],\n",
        "    english_vocab_size_bpe+1,\n",
        "    zulu_vocab_size_bpe+1)\n",
        "\n",
        "embed_rnn_model_bpe.summary()\n",
        "\n",
        "model4 = embed_rnn_model_bpe.fit(tmp_x_bpe, preproc_zulu_sentences_bpe, batch_size=1024, epochs=NUM_EPOCHS, validation_split=0.2)\n",
        "\n",
        "# TODO: Print prediction(s)\n",
        "print(logits_to_text_bpe(embed_rnn_model_bpe.predict(tmp_x_bpe[:1])[0], tokenizer_zu))"
      ],
      "execution_count": 24,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential_3\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " embedding_1 (Embedding)     (None, 58, 256)           2560256   \n",
            "                                                                 \n",
            " gru_3 (GRU)                 (None, 58, 256)           394752    \n",
            "                                                                 \n",
            " time_distributed_6 (TimeDis  (None, 58, 1024)         263168    \n",
            " tributed)                                                       \n",
            "                                                                 \n",
            " dropout_3 (Dropout)         (None, 58, 1024)          0         \n",
            "                                                                 \n",
            " time_distributed_7 (TimeDis  (None, 58, 10001)        10251025  \n",
            " tributed)                                                       \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 13,469,201\n",
            "Trainable params: 13,469,201\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n",
            "Epoch 1/200\n",
            "2/2 [==============================] - 4s 1s/step - loss: 9.1503 - accuracy: 0.2568 - val_loss: 7.6734 - val_accuracy: 0.6792\n",
            "Epoch 2/200\n",
            "2/2 [==============================] - 1s 641ms/step - loss: 5.9423 - accuracy: 0.6803 - val_loss: 4.1665 - val_accuracy: 0.6792\n",
            "Epoch 3/200\n",
            "2/2 [==============================] - 1s 643ms/step - loss: 4.1673 - accuracy: 0.6803 - val_loss: 4.6396 - val_accuracy: 0.6792\n",
            "Epoch 4/200\n",
            "2/2 [==============================] - 1s 646ms/step - loss: 4.4889 - accuracy: 0.4850 - val_loss: 4.2269 - val_accuracy: 0.6794\n",
            "Epoch 5/200\n",
            "2/2 [==============================] - 1s 643ms/step - loss: 3.6784 - accuracy: 0.6806 - val_loss: 3.3639 - val_accuracy: 0.6698\n",
            "Epoch 6/200\n",
            "2/2 [==============================] - 1s 642ms/step - loss: 2.9871 - accuracy: 0.6713 - val_loss: 3.2345 - val_accuracy: 0.6794\n",
            "Epoch 7/200\n",
            "2/2 [==============================] - 1s 640ms/step - loss: 2.9021 - accuracy: 0.6801 - val_loss: 3.2323 - val_accuracy: 0.6792\n",
            "Epoch 8/200\n",
            "2/2 [==============================] - 1s 638ms/step - loss: 2.8280 - accuracy: 0.6803 - val_loss: 3.1813 - val_accuracy: 0.6789\n",
            "Epoch 9/200\n",
            "2/2 [==============================] - 1s 639ms/step - loss: 2.7541 - accuracy: 0.6751 - val_loss: 3.1479 - val_accuracy: 0.6735\n",
            "Epoch 10/200\n",
            "2/2 [==============================] - 1s 644ms/step - loss: 2.6718 - accuracy: 0.6737 - val_loss: 3.1084 - val_accuracy: 0.6743\n",
            "Epoch 11/200\n",
            "2/2 [==============================] - 1s 647ms/step - loss: 2.6270 - accuracy: 0.6703 - val_loss: 3.1429 - val_accuracy: 0.6804\n",
            "Epoch 12/200\n",
            "2/2 [==============================] - 1s 663ms/step - loss: 2.6046 - accuracy: 0.6808 - val_loss: 3.1117 - val_accuracy: 0.6804\n",
            "Epoch 13/200\n",
            "2/2 [==============================] - 1s 651ms/step - loss: 2.5769 - accuracy: 0.6816 - val_loss: 3.0744 - val_accuracy: 0.6807\n",
            "Epoch 14/200\n",
            "2/2 [==============================] - 1s 650ms/step - loss: 2.5451 - accuracy: 0.6817 - val_loss: 3.0700 - val_accuracy: 0.6805\n",
            "Epoch 15/200\n",
            "2/2 [==============================] - 1s 652ms/step - loss: 2.5179 - accuracy: 0.6825 - val_loss: 3.0536 - val_accuracy: 0.6803\n",
            "Epoch 16/200\n",
            "2/2 [==============================] - 1s 654ms/step - loss: 2.5019 - accuracy: 0.6834 - val_loss: 3.0894 - val_accuracy: 0.6798\n",
            "Epoch 17/200\n",
            "2/2 [==============================] - 1s 656ms/step - loss: 2.4768 - accuracy: 0.6832 - val_loss: 3.1053 - val_accuracy: 0.6805\n",
            "Epoch 18/200\n",
            "2/2 [==============================] - 2s 660ms/step - loss: 2.4604 - accuracy: 0.6839 - val_loss: 3.1221 - val_accuracy: 0.6815\n",
            "Epoch 19/200\n",
            "2/2 [==============================] - 1s 650ms/step - loss: 2.4420 - accuracy: 0.6849 - val_loss: 3.1491 - val_accuracy: 0.6811\n",
            "Epoch 20/200\n",
            "2/2 [==============================] - 1s 657ms/step - loss: 2.4266 - accuracy: 0.6854 - val_loss: 3.1424 - val_accuracy: 0.6814\n",
            "Epoch 21/200\n",
            "2/2 [==============================] - 1s 651ms/step - loss: 2.4114 - accuracy: 0.6855 - val_loss: 3.1644 - val_accuracy: 0.6816\n",
            "Epoch 22/200\n",
            "2/2 [==============================] - 1s 654ms/step - loss: 2.3949 - accuracy: 0.6862 - val_loss: 3.1613 - val_accuracy: 0.6820\n",
            "Epoch 23/200\n",
            "2/2 [==============================] - 1s 654ms/step - loss: 2.3797 - accuracy: 0.6864 - val_loss: 3.1831 - val_accuracy: 0.6822\n",
            "Epoch 24/200\n",
            "2/2 [==============================] - 1s 638ms/step - loss: 2.3639 - accuracy: 0.6873 - val_loss: 3.1757 - val_accuracy: 0.6825\n",
            "Epoch 25/200\n",
            "2/2 [==============================] - 1s 658ms/step - loss: 2.3472 - accuracy: 0.6866 - val_loss: 3.1935 - val_accuracy: 0.6824\n",
            "Epoch 26/200\n",
            "2/2 [==============================] - 1s 638ms/step - loss: 2.3330 - accuracy: 0.6871 - val_loss: 3.1994 - val_accuracy: 0.6821\n",
            "Epoch 27/200\n",
            "2/2 [==============================] - 1s 656ms/step - loss: 2.3162 - accuracy: 0.6875 - val_loss: 3.2212 - val_accuracy: 0.6822\n",
            "Epoch 28/200\n",
            "2/2 [==============================] - 1s 644ms/step - loss: 2.3009 - accuracy: 0.6881 - val_loss: 3.2281 - val_accuracy: 0.6822\n",
            "Epoch 29/200\n",
            "2/2 [==============================] - 1s 642ms/step - loss: 2.2861 - accuracy: 0.6880 - val_loss: 3.2499 - val_accuracy: 0.6823\n",
            "Epoch 30/200\n",
            "2/2 [==============================] - 1s 638ms/step - loss: 2.2725 - accuracy: 0.6887 - val_loss: 3.2697 - val_accuracy: 0.6820\n",
            "Epoch 31/200\n",
            "2/2 [==============================] - 1s 640ms/step - loss: 2.2551 - accuracy: 0.6884 - val_loss: 3.2937 - val_accuracy: 0.6815\n",
            "Epoch 32/200\n",
            "2/2 [==============================] - 1s 637ms/step - loss: 2.2406 - accuracy: 0.6891 - val_loss: 3.2990 - val_accuracy: 0.6799\n",
            "Epoch 33/200\n",
            "2/2 [==============================] - 1s 641ms/step - loss: 2.2296 - accuracy: 0.6895 - val_loss: 3.3079 - val_accuracy: 0.6768\n",
            "Epoch 34/200\n",
            "2/2 [==============================] - 1s 648ms/step - loss: 2.2242 - accuracy: 0.6893 - val_loss: 3.3371 - val_accuracy: 0.6788\n",
            "Epoch 35/200\n",
            "2/2 [==============================] - 1s 654ms/step - loss: 2.1994 - accuracy: 0.6898 - val_loss: 3.3764 - val_accuracy: 0.6813\n",
            "Epoch 36/200\n",
            "2/2 [==============================] - 1s 644ms/step - loss: 2.1878 - accuracy: 0.6903 - val_loss: 3.3530 - val_accuracy: 0.6761\n",
            "Epoch 37/200\n",
            "2/2 [==============================] - 1s 641ms/step - loss: 2.1766 - accuracy: 0.6906 - val_loss: 3.3892 - val_accuracy: 0.6779\n",
            "Epoch 38/200\n",
            "2/2 [==============================] - 1s 646ms/step - loss: 2.1536 - accuracy: 0.6914 - val_loss: 3.4188 - val_accuracy: 0.6790\n",
            "Epoch 39/200\n",
            "2/2 [==============================] - 1s 652ms/step - loss: 2.1406 - accuracy: 0.6910 - val_loss: 3.4108 - val_accuracy: 0.6759\n",
            "Epoch 40/200\n",
            "2/2 [==============================] - 1s 652ms/step - loss: 2.1204 - accuracy: 0.6924 - val_loss: 3.4366 - val_accuracy: 0.6767\n",
            "Epoch 41/200\n",
            "2/2 [==============================] - 1s 654ms/step - loss: 2.1020 - accuracy: 0.6926 - val_loss: 3.4673 - val_accuracy: 0.6773\n",
            "Epoch 42/200\n",
            "2/2 [==============================] - 2s 651ms/step - loss: 2.0913 - accuracy: 0.6932 - val_loss: 3.4754 - val_accuracy: 0.6770\n",
            "Epoch 43/200\n",
            "2/2 [==============================] - 1s 654ms/step - loss: 2.0698 - accuracy: 0.6930 - val_loss: 3.4708 - val_accuracy: 0.6725\n",
            "Epoch 44/200\n",
            "2/2 [==============================] - 1s 655ms/step - loss: 2.0540 - accuracy: 0.6943 - val_loss: 3.5222 - val_accuracy: 0.6787\n",
            "Epoch 45/200\n",
            "2/2 [==============================] - 1s 654ms/step - loss: 2.0402 - accuracy: 0.6941 - val_loss: 3.5035 - val_accuracy: 0.6731\n",
            "Epoch 46/200\n",
            "2/2 [==============================] - 1s 654ms/step - loss: 2.0189 - accuracy: 0.6949 - val_loss: 3.5324 - val_accuracy: 0.6746\n",
            "Epoch 47/200\n",
            "2/2 [==============================] - 1s 653ms/step - loss: 2.0104 - accuracy: 0.6958 - val_loss: 3.5530 - val_accuracy: 0.6759\n",
            "Epoch 48/200\n",
            "2/2 [==============================] - 2s 657ms/step - loss: 1.9964 - accuracy: 0.6960 - val_loss: 3.5613 - val_accuracy: 0.6744\n",
            "Epoch 49/200\n",
            "2/2 [==============================] - 2s 656ms/step - loss: 1.9709 - accuracy: 0.6961 - val_loss: 3.5707 - val_accuracy: 0.6713\n",
            "Epoch 50/200\n",
            "2/2 [==============================] - 1s 647ms/step - loss: 1.9458 - accuracy: 0.6973 - val_loss: 3.5994 - val_accuracy: 0.6753\n",
            "Epoch 51/200\n",
            "2/2 [==============================] - 2s 655ms/step - loss: 1.9215 - accuracy: 0.6978 - val_loss: 3.6093 - val_accuracy: 0.6722\n",
            "Epoch 52/200\n",
            "2/2 [==============================] - 1s 643ms/step - loss: 1.8987 - accuracy: 0.6994 - val_loss: 3.6194 - val_accuracy: 0.6707\n",
            "Epoch 53/200\n",
            "2/2 [==============================] - 1s 655ms/step - loss: 1.8758 - accuracy: 0.7006 - val_loss: 3.6505 - val_accuracy: 0.6746\n",
            "Epoch 54/200\n",
            "2/2 [==============================] - 1s 646ms/step - loss: 1.8497 - accuracy: 0.7011 - val_loss: 3.6764 - val_accuracy: 0.6735\n",
            "Epoch 55/200\n",
            "2/2 [==============================] - 1s 648ms/step - loss: 1.8301 - accuracy: 0.7020 - val_loss: 3.6833 - val_accuracy: 0.6727\n",
            "Epoch 56/200\n",
            "2/2 [==============================] - 1s 656ms/step - loss: 1.8058 - accuracy: 0.7035 - val_loss: 3.6952 - val_accuracy: 0.6672\n",
            "Epoch 57/200\n",
            "2/2 [==============================] - 1s 653ms/step - loss: 1.7906 - accuracy: 0.7036 - val_loss: 3.7122 - val_accuracy: 0.6708\n",
            "Epoch 58/200\n",
            "2/2 [==============================] - 1s 653ms/step - loss: 1.7589 - accuracy: 0.7062 - val_loss: 3.7378 - val_accuracy: 0.6735\n",
            "Epoch 59/200\n",
            "2/2 [==============================] - 1s 651ms/step - loss: 1.7376 - accuracy: 0.7082 - val_loss: 3.7402 - val_accuracy: 0.6712\n",
            "Epoch 60/200\n",
            "2/2 [==============================] - 2s 661ms/step - loss: 1.7075 - accuracy: 0.7091 - val_loss: 3.7469 - val_accuracy: 0.6666\n",
            "Epoch 61/200\n",
            "2/2 [==============================] - 1s 648ms/step - loss: 1.6891 - accuracy: 0.7114 - val_loss: 3.7670 - val_accuracy: 0.6694\n",
            "Epoch 62/200\n",
            "2/2 [==============================] - 1s 656ms/step - loss: 1.6575 - accuracy: 0.7136 - val_loss: 3.7823 - val_accuracy: 0.6700\n",
            "Epoch 63/200\n",
            "2/2 [==============================] - 1s 651ms/step - loss: 1.6318 - accuracy: 0.7156 - val_loss: 3.8028 - val_accuracy: 0.6707\n",
            "Epoch 64/200\n",
            "2/2 [==============================] - 1s 652ms/step - loss: 1.6058 - accuracy: 0.7177 - val_loss: 3.8053 - val_accuracy: 0.6683\n",
            "Epoch 65/200\n",
            "2/2 [==============================] - 1s 651ms/step - loss: 1.5779 - accuracy: 0.7214 - val_loss: 3.8172 - val_accuracy: 0.6663\n",
            "Epoch 66/200\n",
            "2/2 [==============================] - 1s 651ms/step - loss: 1.5513 - accuracy: 0.7235 - val_loss: 3.8351 - val_accuracy: 0.6692\n",
            "Epoch 67/200\n",
            "2/2 [==============================] - 1s 657ms/step - loss: 1.5254 - accuracy: 0.7255 - val_loss: 3.8625 - val_accuracy: 0.6712\n",
            "Epoch 68/200\n",
            "2/2 [==============================] - 1s 652ms/step - loss: 1.4995 - accuracy: 0.7279 - val_loss: 3.8677 - val_accuracy: 0.6682\n",
            "Epoch 69/200\n",
            "2/2 [==============================] - 1s 651ms/step - loss: 1.4704 - accuracy: 0.7321 - val_loss: 3.8803 - val_accuracy: 0.6648\n",
            "Epoch 70/200\n",
            "2/2 [==============================] - 1s 650ms/step - loss: 1.4577 - accuracy: 0.7331 - val_loss: 3.8859 - val_accuracy: 0.6639\n",
            "Epoch 71/200\n",
            "2/2 [==============================] - 2s 661ms/step - loss: 1.4354 - accuracy: 0.7360 - val_loss: 3.9036 - val_accuracy: 0.6669\n",
            "Epoch 72/200\n",
            "2/2 [==============================] - 1s 647ms/step - loss: 1.4009 - accuracy: 0.7407 - val_loss: 3.9251 - val_accuracy: 0.6707\n",
            "Epoch 73/200\n",
            "2/2 [==============================] - 1s 656ms/step - loss: 1.3887 - accuracy: 0.7399 - val_loss: 3.9208 - val_accuracy: 0.6676\n",
            "Epoch 74/200\n",
            "2/2 [==============================] - 1s 653ms/step - loss: 1.3566 - accuracy: 0.7447 - val_loss: 3.9230 - val_accuracy: 0.6640\n",
            "Epoch 75/200\n",
            "2/2 [==============================] - 1s 657ms/step - loss: 1.3388 - accuracy: 0.7473 - val_loss: 3.9459 - val_accuracy: 0.6683\n",
            "Epoch 76/200\n",
            "2/2 [==============================] - 1s 653ms/step - loss: 1.3112 - accuracy: 0.7491 - val_loss: 3.9563 - val_accuracy: 0.6675\n",
            "Epoch 77/200\n",
            "2/2 [==============================] - 2s 654ms/step - loss: 1.2824 - accuracy: 0.7532 - val_loss: 3.9595 - val_accuracy: 0.6642\n",
            "Epoch 78/200\n",
            "2/2 [==============================] - 1s 650ms/step - loss: 1.2606 - accuracy: 0.7566 - val_loss: 3.9772 - val_accuracy: 0.6673\n",
            "Epoch 79/200\n",
            "2/2 [==============================] - 1s 648ms/step - loss: 1.2358 - accuracy: 0.7610 - val_loss: 3.9788 - val_accuracy: 0.6675\n",
            "Epoch 80/200\n",
            "2/2 [==============================] - 1s 653ms/step - loss: 1.2079 - accuracy: 0.7638 - val_loss: 3.9799 - val_accuracy: 0.6635\n",
            "Epoch 81/200\n",
            "2/2 [==============================] - 1s 649ms/step - loss: 1.1896 - accuracy: 0.7669 - val_loss: 4.0030 - val_accuracy: 0.6682\n",
            "Epoch 82/200\n",
            "2/2 [==============================] - 1s 654ms/step - loss: 1.1621 - accuracy: 0.7711 - val_loss: 4.0083 - val_accuracy: 0.6675\n",
            "Epoch 83/200\n",
            "2/2 [==============================] - 1s 655ms/step - loss: 1.1346 - accuracy: 0.7743 - val_loss: 4.0163 - val_accuracy: 0.6625\n",
            "Epoch 84/200\n",
            "2/2 [==============================] - 1s 655ms/step - loss: 1.1178 - accuracy: 0.7771 - val_loss: 4.0330 - val_accuracy: 0.6669\n",
            "Epoch 85/200\n",
            "2/2 [==============================] - 1s 657ms/step - loss: 1.0883 - accuracy: 0.7819 - val_loss: 4.0520 - val_accuracy: 0.6657\n",
            "Epoch 86/200\n",
            "2/2 [==============================] - 1s 651ms/step - loss: 1.0750 - accuracy: 0.7835 - val_loss: 4.0495 - val_accuracy: 0.6632\n",
            "Epoch 87/200\n",
            "2/2 [==============================] - 1s 652ms/step - loss: 1.0544 - accuracy: 0.7874 - val_loss: 4.0625 - val_accuracy: 0.6659\n",
            "Epoch 88/200\n",
            "2/2 [==============================] - 1s 645ms/step - loss: 1.0280 - accuracy: 0.7900 - val_loss: 4.0753 - val_accuracy: 0.6681\n",
            "Epoch 89/200\n",
            "2/2 [==============================] - 1s 651ms/step - loss: 1.0130 - accuracy: 0.7921 - val_loss: 4.0722 - val_accuracy: 0.6637\n",
            "Epoch 90/200\n",
            "2/2 [==============================] - 1s 655ms/step - loss: 0.9978 - accuracy: 0.7950 - val_loss: 4.0797 - val_accuracy: 0.6625\n",
            "Epoch 91/200\n",
            "2/2 [==============================] - 1s 647ms/step - loss: 0.9811 - accuracy: 0.7975 - val_loss: 4.0934 - val_accuracy: 0.6688\n",
            "Epoch 92/200\n",
            "2/2 [==============================] - 2s 664ms/step - loss: 0.9769 - accuracy: 0.7968 - val_loss: 4.0910 - val_accuracy: 0.6636\n",
            "Epoch 93/200\n",
            "2/2 [==============================] - 1s 647ms/step - loss: 0.9490 - accuracy: 0.8018 - val_loss: 4.0989 - val_accuracy: 0.6635\n",
            "Epoch 94/200\n",
            "2/2 [==============================] - 2s 658ms/step - loss: 0.9278 - accuracy: 0.8064 - val_loss: 4.1058 - val_accuracy: 0.6656\n",
            "Epoch 95/200\n",
            "2/2 [==============================] - 1s 646ms/step - loss: 0.9125 - accuracy: 0.8082 - val_loss: 4.1192 - val_accuracy: 0.6626\n",
            "Epoch 96/200\n",
            "2/2 [==============================] - 1s 655ms/step - loss: 0.8918 - accuracy: 0.8120 - val_loss: 4.1172 - val_accuracy: 0.6642\n",
            "Epoch 97/200\n",
            "2/2 [==============================] - 1s 655ms/step - loss: 0.8716 - accuracy: 0.8161 - val_loss: 4.1354 - val_accuracy: 0.6659\n",
            "Epoch 98/200\n",
            "2/2 [==============================] - 2s 660ms/step - loss: 0.8567 - accuracy: 0.8168 - val_loss: 4.1381 - val_accuracy: 0.6633\n",
            "Epoch 99/200\n",
            "2/2 [==============================] - 1s 647ms/step - loss: 0.8373 - accuracy: 0.8204 - val_loss: 4.1456 - val_accuracy: 0.6649\n",
            "Epoch 100/200\n",
            "2/2 [==============================] - 1s 650ms/step - loss: 0.8156 - accuracy: 0.8251 - val_loss: 4.1591 - val_accuracy: 0.6654\n",
            "Epoch 101/200\n",
            "2/2 [==============================] - 1s 653ms/step - loss: 0.8080 - accuracy: 0.8252 - val_loss: 4.1633 - val_accuracy: 0.6651\n",
            "Epoch 102/200\n",
            "2/2 [==============================] - 1s 651ms/step - loss: 0.7892 - accuracy: 0.8295 - val_loss: 4.1721 - val_accuracy: 0.6632\n",
            "Epoch 103/200\n",
            "2/2 [==============================] - 1s 657ms/step - loss: 0.7736 - accuracy: 0.8326 - val_loss: 4.1820 - val_accuracy: 0.6656\n",
            "Epoch 104/200\n",
            "2/2 [==============================] - 1s 657ms/step - loss: 0.7568 - accuracy: 0.8344 - val_loss: 4.1869 - val_accuracy: 0.6642\n",
            "Epoch 105/200\n",
            "2/2 [==============================] - 1s 653ms/step - loss: 0.7398 - accuracy: 0.8367 - val_loss: 4.2005 - val_accuracy: 0.6625\n",
            "Epoch 106/200\n",
            "2/2 [==============================] - 2s 663ms/step - loss: 0.7339 - accuracy: 0.8394 - val_loss: 4.2071 - val_accuracy: 0.6660\n",
            "Epoch 107/200\n",
            "2/2 [==============================] - 1s 648ms/step - loss: 0.7163 - accuracy: 0.8418 - val_loss: 4.2125 - val_accuracy: 0.6663\n",
            "Epoch 108/200\n",
            "2/2 [==============================] - 1s 654ms/step - loss: 0.7023 - accuracy: 0.8443 - val_loss: 4.2206 - val_accuracy: 0.6634\n",
            "Epoch 109/200\n",
            "2/2 [==============================] - 1s 655ms/step - loss: 0.6931 - accuracy: 0.8453 - val_loss: 4.2284 - val_accuracy: 0.6631\n",
            "Epoch 110/200\n",
            "2/2 [==============================] - 1s 653ms/step - loss: 0.6796 - accuracy: 0.8489 - val_loss: 4.2326 - val_accuracy: 0.6666\n",
            "Epoch 111/200\n",
            "2/2 [==============================] - 2s 654ms/step - loss: 0.6728 - accuracy: 0.8491 - val_loss: 4.2390 - val_accuracy: 0.6653\n",
            "Epoch 112/200\n",
            "2/2 [==============================] - 1s 657ms/step - loss: 0.6581 - accuracy: 0.8517 - val_loss: 4.2522 - val_accuracy: 0.6597\n",
            "Epoch 113/200\n",
            "2/2 [==============================] - 2s 661ms/step - loss: 0.6579 - accuracy: 0.8500 - val_loss: 4.2455 - val_accuracy: 0.6659\n",
            "Epoch 114/200\n",
            "2/2 [==============================] - 1s 645ms/step - loss: 0.6479 - accuracy: 0.8514 - val_loss: 4.2568 - val_accuracy: 0.6654\n",
            "Epoch 115/200\n",
            "2/2 [==============================] - 1s 654ms/step - loss: 0.6405 - accuracy: 0.8546 - val_loss: 4.2592 - val_accuracy: 0.6591\n",
            "Epoch 116/200\n",
            "2/2 [==============================] - 1s 655ms/step - loss: 0.6352 - accuracy: 0.8551 - val_loss: 4.2622 - val_accuracy: 0.6675\n",
            "Epoch 117/200\n",
            "2/2 [==============================] - 1s 650ms/step - loss: 0.6294 - accuracy: 0.8559 - val_loss: 4.2660 - val_accuracy: 0.6615\n",
            "Epoch 118/200\n",
            "2/2 [==============================] - 1s 652ms/step - loss: 0.6148 - accuracy: 0.8591 - val_loss: 4.2602 - val_accuracy: 0.6638\n",
            "Epoch 119/200\n",
            "2/2 [==============================] - 1s 654ms/step - loss: 0.5993 - accuracy: 0.8614 - val_loss: 4.2651 - val_accuracy: 0.6630\n",
            "Epoch 120/200\n",
            "2/2 [==============================] - 1s 655ms/step - loss: 0.5872 - accuracy: 0.8638 - val_loss: 4.2711 - val_accuracy: 0.6621\n",
            "Epoch 121/200\n",
            "2/2 [==============================] - 2s 664ms/step - loss: 0.5743 - accuracy: 0.8669 - val_loss: 4.2765 - val_accuracy: 0.6650\n",
            "Epoch 122/200\n",
            "2/2 [==============================] - 1s 654ms/step - loss: 0.5657 - accuracy: 0.8683 - val_loss: 4.2918 - val_accuracy: 0.6604\n",
            "Epoch 123/200\n",
            "2/2 [==============================] - 1s 648ms/step - loss: 0.5570 - accuracy: 0.8695 - val_loss: 4.2859 - val_accuracy: 0.6661\n",
            "Epoch 124/200\n",
            "2/2 [==============================] - 1s 655ms/step - loss: 0.5478 - accuracy: 0.8716 - val_loss: 4.3018 - val_accuracy: 0.6616\n",
            "Epoch 125/200\n",
            "2/2 [==============================] - 1s 656ms/step - loss: 0.5388 - accuracy: 0.8729 - val_loss: 4.3078 - val_accuracy: 0.6658\n",
            "Epoch 126/200\n",
            "2/2 [==============================] - 1s 655ms/step - loss: 0.5246 - accuracy: 0.8768 - val_loss: 4.3069 - val_accuracy: 0.6635\n",
            "Epoch 127/200\n",
            "2/2 [==============================] - 1s 648ms/step - loss: 0.5170 - accuracy: 0.8776 - val_loss: 4.3170 - val_accuracy: 0.6640\n",
            "Epoch 128/200\n",
            "2/2 [==============================] - 1s 654ms/step - loss: 0.5085 - accuracy: 0.8800 - val_loss: 4.3202 - val_accuracy: 0.6653\n",
            "Epoch 129/200\n",
            "2/2 [==============================] - 2s 660ms/step - loss: 0.5005 - accuracy: 0.8804 - val_loss: 4.3281 - val_accuracy: 0.6617\n",
            "Epoch 130/200\n",
            "2/2 [==============================] - 1s 647ms/step - loss: 0.4900 - accuracy: 0.8845 - val_loss: 4.3282 - val_accuracy: 0.6661\n",
            "Epoch 131/200\n",
            "2/2 [==============================] - 2s 662ms/step - loss: 0.4799 - accuracy: 0.8857 - val_loss: 4.3434 - val_accuracy: 0.6622\n",
            "Epoch 132/200\n",
            "2/2 [==============================] - 1s 651ms/step - loss: 0.4794 - accuracy: 0.8850 - val_loss: 4.3437 - val_accuracy: 0.6657\n",
            "Epoch 133/200\n",
            "2/2 [==============================] - 2s 659ms/step - loss: 0.4722 - accuracy: 0.8864 - val_loss: 4.3464 - val_accuracy: 0.6634\n",
            "Epoch 134/200\n",
            "2/2 [==============================] - 1s 649ms/step - loss: 0.4616 - accuracy: 0.8893 - val_loss: 4.3560 - val_accuracy: 0.6633\n",
            "Epoch 135/200\n",
            "2/2 [==============================] - 1s 653ms/step - loss: 0.4547 - accuracy: 0.8902 - val_loss: 4.3520 - val_accuracy: 0.6653\n",
            "Epoch 136/200\n",
            "2/2 [==============================] - 1s 650ms/step - loss: 0.4490 - accuracy: 0.8915 - val_loss: 4.3659 - val_accuracy: 0.6632\n",
            "Epoch 137/200\n",
            "2/2 [==============================] - 1s 654ms/step - loss: 0.4390 - accuracy: 0.8933 - val_loss: 4.3670 - val_accuracy: 0.6639\n",
            "Epoch 138/200\n",
            "2/2 [==============================] - 1s 656ms/step - loss: 0.4388 - accuracy: 0.8925 - val_loss: 4.3710 - val_accuracy: 0.6646\n",
            "Epoch 139/200\n",
            "2/2 [==============================] - 2s 663ms/step - loss: 0.4311 - accuracy: 0.8944 - val_loss: 4.3768 - val_accuracy: 0.6621\n",
            "Epoch 140/200\n",
            "2/2 [==============================] - 1s 653ms/step - loss: 0.4235 - accuracy: 0.8964 - val_loss: 4.3764 - val_accuracy: 0.6663\n",
            "Epoch 141/200\n",
            "2/2 [==============================] - 1s 648ms/step - loss: 0.4228 - accuracy: 0.8962 - val_loss: 4.3850 - val_accuracy: 0.6622\n",
            "Epoch 142/200\n",
            "2/2 [==============================] - 1s 652ms/step - loss: 0.4140 - accuracy: 0.8987 - val_loss: 4.3925 - val_accuracy: 0.6646\n",
            "Epoch 143/200\n",
            "2/2 [==============================] - 1s 656ms/step - loss: 0.4048 - accuracy: 0.9000 - val_loss: 4.3947 - val_accuracy: 0.6637\n",
            "Epoch 144/200\n",
            "2/2 [==============================] - 1s 652ms/step - loss: 0.4003 - accuracy: 0.9009 - val_loss: 4.3996 - val_accuracy: 0.6638\n",
            "Epoch 145/200\n",
            "2/2 [==============================] - 1s 655ms/step - loss: 0.3945 - accuracy: 0.9024 - val_loss: 4.3987 - val_accuracy: 0.6652\n",
            "Epoch 146/200\n",
            "2/2 [==============================] - 1s 651ms/step - loss: 0.3869 - accuracy: 0.9047 - val_loss: 4.4050 - val_accuracy: 0.6629\n",
            "Epoch 147/200\n",
            "2/2 [==============================] - 1s 654ms/step - loss: 0.3847 - accuracy: 0.9038 - val_loss: 4.4090 - val_accuracy: 0.6648\n",
            "Epoch 148/200\n",
            "2/2 [==============================] - 1s 657ms/step - loss: 0.3787 - accuracy: 0.9057 - val_loss: 4.4107 - val_accuracy: 0.6633\n",
            "Epoch 149/200\n",
            "2/2 [==============================] - 1s 653ms/step - loss: 0.3739 - accuracy: 0.9064 - val_loss: 4.4247 - val_accuracy: 0.6611\n",
            "Epoch 150/200\n",
            "2/2 [==============================] - 1s 658ms/step - loss: 0.3707 - accuracy: 0.9070 - val_loss: 4.4145 - val_accuracy: 0.6660\n",
            "Epoch 151/200\n",
            "2/2 [==============================] - 1s 653ms/step - loss: 0.3662 - accuracy: 0.9084 - val_loss: 4.4336 - val_accuracy: 0.6608\n",
            "Epoch 152/200\n",
            "2/2 [==============================] - 1s 651ms/step - loss: 0.3575 - accuracy: 0.9099 - val_loss: 4.4360 - val_accuracy: 0.6641\n",
            "Epoch 153/200\n",
            "2/2 [==============================] - 1s 654ms/step - loss: 0.3524 - accuracy: 0.9104 - val_loss: 4.4303 - val_accuracy: 0.6635\n",
            "Epoch 154/200\n",
            "2/2 [==============================] - 1s 656ms/step - loss: 0.3449 - accuracy: 0.9135 - val_loss: 4.4407 - val_accuracy: 0.6630\n",
            "Epoch 155/200\n",
            "2/2 [==============================] - 1s 651ms/step - loss: 0.3427 - accuracy: 0.9136 - val_loss: 4.4346 - val_accuracy: 0.6640\n",
            "Epoch 156/200\n",
            "2/2 [==============================] - 1s 657ms/step - loss: 0.3349 - accuracy: 0.9156 - val_loss: 4.4426 - val_accuracy: 0.6619\n",
            "Epoch 157/200\n",
            "2/2 [==============================] - 1s 646ms/step - loss: 0.3313 - accuracy: 0.9158 - val_loss: 4.4462 - val_accuracy: 0.6642\n",
            "Epoch 158/200\n",
            "2/2 [==============================] - 1s 649ms/step - loss: 0.3253 - accuracy: 0.9177 - val_loss: 4.4542 - val_accuracy: 0.6636\n",
            "Epoch 159/200\n",
            "2/2 [==============================] - 1s 654ms/step - loss: 0.3218 - accuracy: 0.9178 - val_loss: 4.4656 - val_accuracy: 0.6611\n",
            "Epoch 160/200\n",
            "2/2 [==============================] - 1s 657ms/step - loss: 0.3134 - accuracy: 0.9205 - val_loss: 4.4623 - val_accuracy: 0.6637\n",
            "Epoch 161/200\n",
            "2/2 [==============================] - 1s 648ms/step - loss: 0.3142 - accuracy: 0.9202 - val_loss: 4.4725 - val_accuracy: 0.6623\n",
            "Epoch 162/200\n",
            "2/2 [==============================] - 1s 655ms/step - loss: 0.3122 - accuracy: 0.9203 - val_loss: 4.4727 - val_accuracy: 0.6627\n",
            "Epoch 163/200\n",
            "2/2 [==============================] - 1s 657ms/step - loss: 0.3025 - accuracy: 0.9235 - val_loss: 4.4705 - val_accuracy: 0.6653\n",
            "Epoch 164/200\n",
            "2/2 [==============================] - 1s 649ms/step - loss: 0.3023 - accuracy: 0.9226 - val_loss: 4.4874 - val_accuracy: 0.6613\n",
            "Epoch 165/200\n",
            "2/2 [==============================] - 1s 654ms/step - loss: 0.2966 - accuracy: 0.9233 - val_loss: 4.4817 - val_accuracy: 0.6642\n",
            "Epoch 166/200\n",
            "2/2 [==============================] - 1s 656ms/step - loss: 0.2912 - accuracy: 0.9243 - val_loss: 4.4860 - val_accuracy: 0.6630\n",
            "Epoch 167/200\n",
            "2/2 [==============================] - 2s 661ms/step - loss: 0.2871 - accuracy: 0.9267 - val_loss: 4.4888 - val_accuracy: 0.6620\n",
            "Epoch 168/200\n",
            "2/2 [==============================] - 1s 659ms/step - loss: 0.2827 - accuracy: 0.9269 - val_loss: 4.4876 - val_accuracy: 0.6639\n",
            "Epoch 169/200\n",
            "2/2 [==============================] - 1s 657ms/step - loss: 0.2826 - accuracy: 0.9270 - val_loss: 4.4973 - val_accuracy: 0.6619\n",
            "Epoch 170/200\n",
            "2/2 [==============================] - 1s 650ms/step - loss: 0.2748 - accuracy: 0.9281 - val_loss: 4.4996 - val_accuracy: 0.6627\n",
            "Epoch 171/200\n",
            "2/2 [==============================] - 1s 652ms/step - loss: 0.2713 - accuracy: 0.9303 - val_loss: 4.4990 - val_accuracy: 0.6641\n",
            "Epoch 172/200\n",
            "2/2 [==============================] - 1s 652ms/step - loss: 0.2652 - accuracy: 0.9310 - val_loss: 4.5099 - val_accuracy: 0.6622\n",
            "Epoch 173/200\n",
            "2/2 [==============================] - 2s 662ms/step - loss: 0.2646 - accuracy: 0.9310 - val_loss: 4.5081 - val_accuracy: 0.6646\n",
            "Epoch 174/200\n",
            "2/2 [==============================] - 1s 646ms/step - loss: 0.2632 - accuracy: 0.9314 - val_loss: 4.5097 - val_accuracy: 0.6636\n",
            "Epoch 175/200\n",
            "2/2 [==============================] - 1s 646ms/step - loss: 0.2600 - accuracy: 0.9323 - val_loss: 4.5206 - val_accuracy: 0.6628\n",
            "Epoch 176/200\n",
            "2/2 [==============================] - 2s 657ms/step - loss: 0.2586 - accuracy: 0.9332 - val_loss: 4.5194 - val_accuracy: 0.6633\n",
            "Epoch 177/200\n",
            "2/2 [==============================] - 1s 652ms/step - loss: 0.2536 - accuracy: 0.9338 - val_loss: 4.5226 - val_accuracy: 0.6639\n",
            "Epoch 178/200\n",
            "2/2 [==============================] - 1s 656ms/step - loss: 0.2499 - accuracy: 0.9348 - val_loss: 4.5285 - val_accuracy: 0.6619\n",
            "Epoch 179/200\n",
            "2/2 [==============================] - 1s 651ms/step - loss: 0.2479 - accuracy: 0.9353 - val_loss: 4.5221 - val_accuracy: 0.6645\n",
            "Epoch 180/200\n",
            "2/2 [==============================] - 1s 658ms/step - loss: 0.2452 - accuracy: 0.9347 - val_loss: 4.5330 - val_accuracy: 0.6629\n",
            "Epoch 181/200\n",
            "2/2 [==============================] - 1s 653ms/step - loss: 0.2407 - accuracy: 0.9368 - val_loss: 4.5362 - val_accuracy: 0.6620\n",
            "Epoch 182/200\n",
            "2/2 [==============================] - 1s 655ms/step - loss: 0.2387 - accuracy: 0.9377 - val_loss: 4.5340 - val_accuracy: 0.6647\n",
            "Epoch 183/200\n",
            "2/2 [==============================] - 2s 666ms/step - loss: 0.2370 - accuracy: 0.9379 - val_loss: 4.5488 - val_accuracy: 0.6610\n",
            "Epoch 184/200\n",
            "2/2 [==============================] - 1s 652ms/step - loss: 0.2351 - accuracy: 0.9377 - val_loss: 4.5461 - val_accuracy: 0.6638\n",
            "Epoch 185/200\n",
            "2/2 [==============================] - 1s 656ms/step - loss: 0.2321 - accuracy: 0.9388 - val_loss: 4.5521 - val_accuracy: 0.6624\n",
            "Epoch 186/200\n",
            "2/2 [==============================] - 1s 651ms/step - loss: 0.2286 - accuracy: 0.9397 - val_loss: 4.5572 - val_accuracy: 0.6623\n",
            "Epoch 187/200\n",
            "2/2 [==============================] - 1s 647ms/step - loss: 0.2262 - accuracy: 0.9395 - val_loss: 4.5571 - val_accuracy: 0.6627\n",
            "Epoch 188/200\n",
            "2/2 [==============================] - 1s 653ms/step - loss: 0.2222 - accuracy: 0.9418 - val_loss: 4.5654 - val_accuracy: 0.6624\n",
            "Epoch 189/200\n",
            "2/2 [==============================] - 1s 655ms/step - loss: 0.2187 - accuracy: 0.9412 - val_loss: 4.5621 - val_accuracy: 0.6622\n",
            "Epoch 190/200\n",
            "2/2 [==============================] - 1s 654ms/step - loss: 0.2156 - accuracy: 0.9435 - val_loss: 4.5660 - val_accuracy: 0.6635\n",
            "Epoch 191/200\n",
            "2/2 [==============================] - 1s 655ms/step - loss: 0.2118 - accuracy: 0.9435 - val_loss: 4.5715 - val_accuracy: 0.6621\n",
            "Epoch 192/200\n",
            "2/2 [==============================] - 2s 658ms/step - loss: 0.2086 - accuracy: 0.9453 - val_loss: 4.5719 - val_accuracy: 0.6640\n",
            "Epoch 193/200\n",
            "2/2 [==============================] - 1s 648ms/step - loss: 0.2074 - accuracy: 0.9438 - val_loss: 4.5738 - val_accuracy: 0.6626\n",
            "Epoch 194/200\n",
            "2/2 [==============================] - 1s 658ms/step - loss: 0.2045 - accuracy: 0.9460 - val_loss: 4.5811 - val_accuracy: 0.6627\n",
            "Epoch 195/200\n",
            "2/2 [==============================] - 1s 655ms/step - loss: 0.2023 - accuracy: 0.9458 - val_loss: 4.5754 - val_accuracy: 0.6630\n",
            "Epoch 196/200\n",
            "2/2 [==============================] - 2s 662ms/step - loss: 0.1977 - accuracy: 0.9476 - val_loss: 4.5801 - val_accuracy: 0.6632\n",
            "Epoch 197/200\n",
            "2/2 [==============================] - 1s 650ms/step - loss: 0.1961 - accuracy: 0.9481 - val_loss: 4.5907 - val_accuracy: 0.6621\n",
            "Epoch 198/200\n",
            "2/2 [==============================] - 2s 661ms/step - loss: 0.1946 - accuracy: 0.9479 - val_loss: 4.5968 - val_accuracy: 0.6622\n",
            "Epoch 199/200\n",
            "2/2 [==============================] - 1s 643ms/step - loss: 0.1930 - accuracy: 0.9490 - val_loss: 4.5921 - val_accuracy: 0.6639\n",
            "Epoch 200/200\n",
            "2/2 [==============================] - 1s 658ms/step - loss: 0.1943 - accuracy: 0.9480 - val_loss: 4.6068 - val_accuracy: 0.6611\n",
            "uphu - albhamu akhe e - solo eminyaka yawo - 60 anga manye anezi ngoma ezimnandi kakhulu kwe nga ke ngazi zwa . [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "sqJNnDGYfy6I"
      },
      "source": [
        "## Model 5 - Bidirectional RNN with Word Embedding and Standard Tokenization"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "yhdEiWgvNJhA"
      },
      "source": [
        "def bd_model(input_shape, output_sequence_length, english_vocab_size, zulu_vocab_size):\n",
        "    \"\"\"\n",
        "    Build and train a bidirectional RNN model on x and y\n",
        "    :param input_shape: Tuple of input shape\n",
        "    :param output_sequence_length: Length of output sequence\n",
        "    :param english_vocab_size: Number of unique English words in the dataset\n",
        "    :param zulu_vocab_size: Number of unique zulu words in the dataset\n",
        "    :return: Keras model built, but not trained\n",
        "    \"\"\"\n",
        "    # TODO: Implement\n",
        "\n",
        "    # Hyperparameters\n",
        "    learning_rate = 0.003\n",
        "    \n",
        "    # TODO: Build the layers\n",
        "    model = Sequential()\n",
        "    model.add(Embedding(english_vocab_size, 256, input_length=input_shape[1], input_shape=input_shape[1:]))\n",
        "    model.add(Bidirectional(GRU(128, return_sequences=True), input_shape=input_shape[1:]))\n",
        "    model.add(TimeDistributed(Dense(1024, activation='relu')))\n",
        "    model.add(Dropout(0.5))\n",
        "    model.add(TimeDistributed(Dense(zulu_vocab_size, activation='softmax'))) \n",
        "\n",
        "    # Compile model\n",
        "    model.compile(loss=sparse_categorical_crossentropy,\n",
        "                  optimizer=keras.optimizers.Adam(learning_rate),\n",
        "                  metrics=['accuracy'])\n",
        "    return model\n",
        "\n"
      ],
      "execution_count": 25,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "E1pv2ISeu05x",
        "outputId": "e198a9e3-1aec-4149-f01e-6b2ee39f6b07"
      },
      "source": [
        "#tests.test_bd_model(bd_model)\n",
        "\n",
        "# TODO: Reshape the input\n",
        "tmp_x = pad(preproc_english_sentences, preproc_zulu_sentences.shape[1])\n",
        "tmp_x = tmp_x.reshape((-1, preproc_zulu_sentences.shape[-2]))\n",
        "tmp_x_test = pad(preproc_english_sentences_test, max_zulu_sequence_length)\n",
        "tmp_x_test = tmp_x_test.reshape((-1, preproc_zulu_sentences.shape[-2], 1))\n",
        "\n",
        "\n",
        "# TODO: Train and Print prediction(s)\n",
        "embed_rnn_model1 = bd_model(\n",
        "    tmp_x.shape,\n",
        "    preproc_zulu_sentences.shape[1],\n",
        "    english_vocab_size+1,\n",
        "    zulu_vocab_size+1)\n",
        "\n",
        "embed_rnn_model1.summary()\n",
        "\n",
        "model5 = embed_rnn_model1.fit(tmp_x, preproc_zulu_sentences, batch_size=1024, epochs=NUM_EPOCHS, validation_split=0.2)\n",
        "\n",
        "print(logits_to_text(embed_rnn_model1.predict(tmp_x[:1])[0], zulu_tokenizer))"
      ],
      "execution_count": 26,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential_4\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " embedding_2 (Embedding)     (None, 25, 256)           1955840   \n",
            "                                                                 \n",
            " bidirectional (Bidirectiona  (None, 25, 256)          296448    \n",
            " l)                                                              \n",
            "                                                                 \n",
            " time_distributed_8 (TimeDis  (None, 25, 1024)         263168    \n",
            " tributed)                                                       \n",
            "                                                                 \n",
            " dropout_4 (Dropout)         (None, 25, 1024)          0         \n",
            "                                                                 \n",
            " time_distributed_9 (TimeDis  (None, 25, 12598)        12912950  \n",
            " tributed)                                                       \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 15,428,406\n",
            "Trainable params: 15,428,406\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n",
            "Epoch 1/200\n",
            "2/2 [==============================] - 6s 1s/step - loss: 9.4040 - accuracy: 0.2222 - val_loss: 9.0871 - val_accuracy: 0.5831\n",
            "Epoch 2/200\n",
            "2/2 [==============================] - 1s 383ms/step - loss: 8.8516 - accuracy: 0.5948 - val_loss: 6.8581 - val_accuracy: 0.5831\n",
            "Epoch 3/200\n",
            "2/2 [==============================] - 1s 399ms/step - loss: 5.9398 - accuracy: 0.5948 - val_loss: 5.5132 - val_accuracy: 0.5831\n",
            "Epoch 4/200\n",
            "2/2 [==============================] - 1s 392ms/step - loss: 5.0214 - accuracy: 0.5948 - val_loss: 4.6881 - val_accuracy: 0.5831\n",
            "Epoch 5/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 3.9455 - accuracy: 0.5948 - val_loss: 4.4626 - val_accuracy: 0.5831\n",
            "Epoch 6/200\n",
            "2/2 [==============================] - 1s 385ms/step - loss: 3.9250 - accuracy: 0.5948 - val_loss: 4.5973 - val_accuracy: 0.5832\n",
            "Epoch 7/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 4.0311 - accuracy: 0.5950 - val_loss: 4.5863 - val_accuracy: 0.5832\n",
            "Epoch 8/200\n",
            "2/2 [==============================] - 1s 397ms/step - loss: 3.8286 - accuracy: 0.5950 - val_loss: 4.7451 - val_accuracy: 0.5832\n",
            "Epoch 9/200\n",
            "2/2 [==============================] - 1s 382ms/step - loss: 3.7755 - accuracy: 0.5950 - val_loss: 4.9230 - val_accuracy: 0.5831\n",
            "Epoch 10/200\n",
            "2/2 [==============================] - 1s 397ms/step - loss: 3.7949 - accuracy: 0.5949 - val_loss: 4.8272 - val_accuracy: 0.5832\n",
            "Epoch 11/200\n",
            "2/2 [==============================] - 1s 381ms/step - loss: 3.6621 - accuracy: 0.5950 - val_loss: 4.8117 - val_accuracy: 0.5834\n",
            "Epoch 12/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 3.6836 - accuracy: 0.5952 - val_loss: 4.8520 - val_accuracy: 0.5833\n",
            "Epoch 13/200\n",
            "2/2 [==============================] - 1s 388ms/step - loss: 3.6715 - accuracy: 0.5952 - val_loss: 4.9078 - val_accuracy: 0.5832\n",
            "Epoch 14/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 3.6172 - accuracy: 0.5952 - val_loss: 4.9861 - val_accuracy: 0.5832\n",
            "Epoch 15/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 3.5969 - accuracy: 0.5951 - val_loss: 5.0510 - val_accuracy: 0.5833\n",
            "Epoch 16/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 3.5927 - accuracy: 0.5951 - val_loss: 5.0513 - val_accuracy: 0.5834\n",
            "Epoch 17/200\n",
            "2/2 [==============================] - 1s 392ms/step - loss: 3.5539 - accuracy: 0.5952 - val_loss: 5.0281 - val_accuracy: 0.5831\n",
            "Epoch 18/200\n",
            "2/2 [==============================] - 1s 395ms/step - loss: 3.5230 - accuracy: 0.5952 - val_loss: 5.0207 - val_accuracy: 0.5827\n",
            "Epoch 19/200\n",
            "2/2 [==============================] - 1s 385ms/step - loss: 3.5101 - accuracy: 0.5951 - val_loss: 5.0314 - val_accuracy: 0.5824\n",
            "Epoch 20/200\n",
            "2/2 [==============================] - 1s 399ms/step - loss: 3.4874 - accuracy: 0.5950 - val_loss: 5.0559 - val_accuracy: 0.5824\n",
            "Epoch 21/200\n",
            "2/2 [==============================] - 1s 411ms/step - loss: 3.4549 - accuracy: 0.5950 - val_loss: 5.0891 - val_accuracy: 0.5825\n",
            "Epoch 22/200\n",
            "2/2 [==============================] - 1s 385ms/step - loss: 3.4310 - accuracy: 0.5952 - val_loss: 5.1130 - val_accuracy: 0.5823\n",
            "Epoch 23/200\n",
            "2/2 [==============================] - 1s 401ms/step - loss: 3.4085 - accuracy: 0.5953 - val_loss: 5.1177 - val_accuracy: 0.5819\n",
            "Epoch 24/200\n",
            "2/2 [==============================] - 1s 399ms/step - loss: 3.3762 - accuracy: 0.5954 - val_loss: 5.1187 - val_accuracy: 0.5809\n",
            "Epoch 25/200\n",
            "2/2 [==============================] - 1s 388ms/step - loss: 3.3479 - accuracy: 0.5953 - val_loss: 5.1296 - val_accuracy: 0.5802\n",
            "Epoch 26/200\n",
            "2/2 [==============================] - 1s 402ms/step - loss: 3.3215 - accuracy: 0.5955 - val_loss: 5.1496 - val_accuracy: 0.5807\n",
            "Epoch 27/200\n",
            "2/2 [==============================] - 1s 390ms/step - loss: 3.2876 - accuracy: 0.5956 - val_loss: 5.1753 - val_accuracy: 0.5807\n",
            "Epoch 28/200\n",
            "2/2 [==============================] - 1s 382ms/step - loss: 3.2584 - accuracy: 0.5953 - val_loss: 5.2009 - val_accuracy: 0.5807\n",
            "Epoch 29/200\n",
            "2/2 [==============================] - 1s 381ms/step - loss: 3.2284 - accuracy: 0.5957 - val_loss: 5.2238 - val_accuracy: 0.5797\n",
            "Epoch 30/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 3.1945 - accuracy: 0.5958 - val_loss: 5.2471 - val_accuracy: 0.5795\n",
            "Epoch 31/200\n",
            "2/2 [==============================] - 1s 390ms/step - loss: 3.1660 - accuracy: 0.5963 - val_loss: 5.2702 - val_accuracy: 0.5789\n",
            "Epoch 32/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 3.1320 - accuracy: 0.5961 - val_loss: 5.2934 - val_accuracy: 0.5788\n",
            "Epoch 33/200\n",
            "2/2 [==============================] - 1s 400ms/step - loss: 3.1026 - accuracy: 0.5965 - val_loss: 5.3193 - val_accuracy: 0.5778\n",
            "Epoch 34/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 3.0721 - accuracy: 0.5966 - val_loss: 5.3477 - val_accuracy: 0.5780\n",
            "Epoch 35/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 3.0431 - accuracy: 0.5967 - val_loss: 5.3731 - val_accuracy: 0.5775\n",
            "Epoch 36/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 3.0087 - accuracy: 0.5968 - val_loss: 5.3955 - val_accuracy: 0.5772\n",
            "Epoch 37/200\n",
            "2/2 [==============================] - 1s 396ms/step - loss: 2.9799 - accuracy: 0.5972 - val_loss: 5.4191 - val_accuracy: 0.5757\n",
            "Epoch 38/200\n",
            "2/2 [==============================] - 1s 390ms/step - loss: 2.9481 - accuracy: 0.5980 - val_loss: 5.4433 - val_accuracy: 0.5767\n",
            "Epoch 39/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 2.9184 - accuracy: 0.5989 - val_loss: 5.4637 - val_accuracy: 0.5773\n",
            "Epoch 40/200\n",
            "2/2 [==============================] - 1s 388ms/step - loss: 2.8817 - accuracy: 0.5990 - val_loss: 5.4818 - val_accuracy: 0.5766\n",
            "Epoch 41/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 2.8499 - accuracy: 0.5998 - val_loss: 5.5014 - val_accuracy: 0.5765\n",
            "Epoch 42/200\n",
            "2/2 [==============================] - 1s 388ms/step - loss: 2.8204 - accuracy: 0.6005 - val_loss: 5.5300 - val_accuracy: 0.5762\n",
            "Epoch 43/200\n",
            "2/2 [==============================] - 1s 390ms/step - loss: 2.7974 - accuracy: 0.6009 - val_loss: 5.5564 - val_accuracy: 0.5755\n",
            "Epoch 44/200\n",
            "2/2 [==============================] - 1s 391ms/step - loss: 2.7587 - accuracy: 0.6020 - val_loss: 5.5701 - val_accuracy: 0.5758\n",
            "Epoch 45/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 2.7331 - accuracy: 0.6026 - val_loss: 5.5913 - val_accuracy: 0.5750\n",
            "Epoch 46/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 2.7024 - accuracy: 0.6032 - val_loss: 5.6066 - val_accuracy: 0.5760\n",
            "Epoch 47/200\n",
            "2/2 [==============================] - 1s 388ms/step - loss: 2.6722 - accuracy: 0.6032 - val_loss: 5.6212 - val_accuracy: 0.5750\n",
            "Epoch 48/200\n",
            "2/2 [==============================] - 1s 393ms/step - loss: 2.6446 - accuracy: 0.6038 - val_loss: 5.6329 - val_accuracy: 0.5751\n",
            "Epoch 49/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 2.6113 - accuracy: 0.6055 - val_loss: 5.6450 - val_accuracy: 0.5740\n",
            "Epoch 50/200\n",
            "2/2 [==============================] - 1s 400ms/step - loss: 2.5811 - accuracy: 0.6062 - val_loss: 5.6557 - val_accuracy: 0.5740\n",
            "Epoch 51/200\n",
            "2/2 [==============================] - 1s 399ms/step - loss: 2.5537 - accuracy: 0.6070 - val_loss: 5.6629 - val_accuracy: 0.5748\n",
            "Epoch 52/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 2.5249 - accuracy: 0.6081 - val_loss: 5.6788 - val_accuracy: 0.5735\n",
            "Epoch 53/200\n",
            "2/2 [==============================] - 1s 396ms/step - loss: 2.4913 - accuracy: 0.6090 - val_loss: 5.6875 - val_accuracy: 0.5730\n",
            "Epoch 54/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 2.4637 - accuracy: 0.6099 - val_loss: 5.7091 - val_accuracy: 0.5730\n",
            "Epoch 55/200\n",
            "2/2 [==============================] - 1s 382ms/step - loss: 2.4297 - accuracy: 0.6105 - val_loss: 5.7107 - val_accuracy: 0.5732\n",
            "Epoch 56/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 2.3959 - accuracy: 0.6140 - val_loss: 5.7352 - val_accuracy: 0.5715\n",
            "Epoch 57/200\n",
            "2/2 [==============================] - 1s 392ms/step - loss: 2.3660 - accuracy: 0.6141 - val_loss: 5.7313 - val_accuracy: 0.5742\n",
            "Epoch 58/200\n",
            "2/2 [==============================] - 1s 390ms/step - loss: 2.3374 - accuracy: 0.6161 - val_loss: 5.7693 - val_accuracy: 0.5708\n",
            "Epoch 59/200\n",
            "2/2 [==============================] - 1s 385ms/step - loss: 2.2960 - accuracy: 0.6174 - val_loss: 5.7491 - val_accuracy: 0.5736\n",
            "Epoch 60/200\n",
            "2/2 [==============================] - 1s 389ms/step - loss: 2.2695 - accuracy: 0.6213 - val_loss: 5.8071 - val_accuracy: 0.5677\n",
            "Epoch 61/200\n",
            "2/2 [==============================] - 1s 400ms/step - loss: 2.2479 - accuracy: 0.6210 - val_loss: 5.7653 - val_accuracy: 0.5749\n",
            "Epoch 62/200\n",
            "2/2 [==============================] - 1s 388ms/step - loss: 2.2203 - accuracy: 0.6239 - val_loss: 5.8258 - val_accuracy: 0.5670\n",
            "Epoch 63/200\n",
            "2/2 [==============================] - 1s 389ms/step - loss: 2.1844 - accuracy: 0.6264 - val_loss: 5.7872 - val_accuracy: 0.5722\n",
            "Epoch 64/200\n",
            "2/2 [==============================] - 1s 385ms/step - loss: 2.1361 - accuracy: 0.6300 - val_loss: 5.8089 - val_accuracy: 0.5702\n",
            "Epoch 65/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 2.0776 - accuracy: 0.6356 - val_loss: 5.8490 - val_accuracy: 0.5658\n",
            "Epoch 66/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 2.0418 - accuracy: 0.6393 - val_loss: 5.8080 - val_accuracy: 0.5732\n",
            "Epoch 67/200\n",
            "2/2 [==============================] - 1s 382ms/step - loss: 2.0113 - accuracy: 0.6435 - val_loss: 5.8694 - val_accuracy: 0.5658\n",
            "Epoch 68/200\n",
            "2/2 [==============================] - 1s 389ms/step - loss: 1.9607 - accuracy: 0.6451 - val_loss: 5.8591 - val_accuracy: 0.5683\n",
            "Epoch 69/200\n",
            "2/2 [==============================] - 1s 407ms/step - loss: 1.9040 - accuracy: 0.6542 - val_loss: 5.8571 - val_accuracy: 0.5708\n",
            "Epoch 70/200\n",
            "2/2 [==============================] - 1s 388ms/step - loss: 1.8515 - accuracy: 0.6600 - val_loss: 5.9134 - val_accuracy: 0.5649\n",
            "Epoch 71/200\n",
            "2/2 [==============================] - 1s 392ms/step - loss: 1.8031 - accuracy: 0.6664 - val_loss: 5.8890 - val_accuracy: 0.5689\n",
            "Epoch 72/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 1.7558 - accuracy: 0.6740 - val_loss: 5.9156 - val_accuracy: 0.5680\n",
            "Epoch 73/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 1.7070 - accuracy: 0.6797 - val_loss: 5.9423 - val_accuracy: 0.5654\n",
            "Epoch 74/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 1.6600 - accuracy: 0.6853 - val_loss: 5.9345 - val_accuracy: 0.5683\n",
            "Epoch 75/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 1.6051 - accuracy: 0.6960 - val_loss: 5.9767 - val_accuracy: 0.5645\n",
            "Epoch 76/200\n",
            "2/2 [==============================] - 1s 396ms/step - loss: 1.5694 - accuracy: 0.7011 - val_loss: 5.9691 - val_accuracy: 0.5655\n",
            "Epoch 77/200\n",
            "2/2 [==============================] - 1s 401ms/step - loss: 1.5070 - accuracy: 0.7108 - val_loss: 5.9899 - val_accuracy: 0.5654\n",
            "Epoch 78/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 1.4634 - accuracy: 0.7161 - val_loss: 6.0206 - val_accuracy: 0.5644\n",
            "Epoch 79/200\n",
            "2/2 [==============================] - 1s 403ms/step - loss: 1.4162 - accuracy: 0.7258 - val_loss: 6.0045 - val_accuracy: 0.5665\n",
            "Epoch 80/200\n",
            "2/2 [==============================] - 1s 388ms/step - loss: 1.3698 - accuracy: 0.7335 - val_loss: 6.0425 - val_accuracy: 0.5629\n",
            "Epoch 81/200\n",
            "2/2 [==============================] - 1s 392ms/step - loss: 1.3285 - accuracy: 0.7397 - val_loss: 6.0399 - val_accuracy: 0.5666\n",
            "Epoch 82/200\n",
            "2/2 [==============================] - 1s 401ms/step - loss: 1.2892 - accuracy: 0.7474 - val_loss: 6.0572 - val_accuracy: 0.5637\n",
            "Epoch 83/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 1.2422 - accuracy: 0.7550 - val_loss: 6.0735 - val_accuracy: 0.5636\n",
            "Epoch 84/200\n",
            "2/2 [==============================] - 1s 389ms/step - loss: 1.2049 - accuracy: 0.7600 - val_loss: 6.0704 - val_accuracy: 0.5652\n",
            "Epoch 85/200\n",
            "2/2 [==============================] - 1s 398ms/step - loss: 1.1613 - accuracy: 0.7689 - val_loss: 6.0980 - val_accuracy: 0.5631\n",
            "Epoch 86/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 1.1175 - accuracy: 0.7765 - val_loss: 6.0909 - val_accuracy: 0.5652\n",
            "Epoch 87/200\n",
            "2/2 [==============================] - 1s 394ms/step - loss: 1.0838 - accuracy: 0.7813 - val_loss: 6.1190 - val_accuracy: 0.5626\n",
            "Epoch 88/200\n",
            "2/2 [==============================] - 1s 385ms/step - loss: 1.0475 - accuracy: 0.7904 - val_loss: 6.1111 - val_accuracy: 0.5647\n",
            "Epoch 89/200\n",
            "2/2 [==============================] - 1s 388ms/step - loss: 1.0089 - accuracy: 0.7969 - val_loss: 6.1370 - val_accuracy: 0.5626\n",
            "Epoch 90/200\n",
            "2/2 [==============================] - 1s 389ms/step - loss: 0.9736 - accuracy: 0.8040 - val_loss: 6.1387 - val_accuracy: 0.5632\n",
            "Epoch 91/200\n",
            "2/2 [==============================] - 1s 385ms/step - loss: 0.9330 - accuracy: 0.8121 - val_loss: 6.1490 - val_accuracy: 0.5627\n",
            "Epoch 92/200\n",
            "2/2 [==============================] - 1s 388ms/step - loss: 0.9018 - accuracy: 0.8167 - val_loss: 6.1726 - val_accuracy: 0.5611\n",
            "Epoch 93/200\n",
            "2/2 [==============================] - 1s 383ms/step - loss: 0.8744 - accuracy: 0.8208 - val_loss: 6.1590 - val_accuracy: 0.5639\n",
            "Epoch 94/200\n",
            "2/2 [==============================] - 1s 385ms/step - loss: 0.8424 - accuracy: 0.8276 - val_loss: 6.1984 - val_accuracy: 0.5603\n",
            "Epoch 95/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 0.8133 - accuracy: 0.8340 - val_loss: 6.1803 - val_accuracy: 0.5629\n",
            "Epoch 96/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 0.7857 - accuracy: 0.8403 - val_loss: 6.2047 - val_accuracy: 0.5611\n",
            "Epoch 97/200\n",
            "2/2 [==============================] - 1s 390ms/step - loss: 0.7567 - accuracy: 0.8431 - val_loss: 6.2094 - val_accuracy: 0.5628\n",
            "Epoch 98/200\n",
            "2/2 [==============================] - 1s 397ms/step - loss: 0.7309 - accuracy: 0.8497 - val_loss: 6.2110 - val_accuracy: 0.5627\n",
            "Epoch 99/200\n",
            "2/2 [==============================] - 1s 397ms/step - loss: 0.7050 - accuracy: 0.8541 - val_loss: 6.2341 - val_accuracy: 0.5610\n",
            "Epoch 100/200\n",
            "2/2 [==============================] - 1s 396ms/step - loss: 0.6818 - accuracy: 0.8600 - val_loss: 6.2242 - val_accuracy: 0.5625\n",
            "Epoch 101/200\n",
            "2/2 [==============================] - 1s 396ms/step - loss: 0.6613 - accuracy: 0.8626 - val_loss: 6.2476 - val_accuracy: 0.5604\n",
            "Epoch 102/200\n",
            "2/2 [==============================] - 1s 395ms/step - loss: 0.6389 - accuracy: 0.8683 - val_loss: 6.2361 - val_accuracy: 0.5629\n",
            "Epoch 103/200\n",
            "2/2 [==============================] - 1s 394ms/step - loss: 0.6186 - accuracy: 0.8708 - val_loss: 6.2619 - val_accuracy: 0.5605\n",
            "Epoch 104/200\n",
            "2/2 [==============================] - 1s 395ms/step - loss: 0.6008 - accuracy: 0.8752 - val_loss: 6.2536 - val_accuracy: 0.5623\n",
            "Epoch 105/200\n",
            "2/2 [==============================] - 1s 389ms/step - loss: 0.5813 - accuracy: 0.8788 - val_loss: 6.2581 - val_accuracy: 0.5623\n",
            "Epoch 106/200\n",
            "2/2 [==============================] - 1s 391ms/step - loss: 0.5608 - accuracy: 0.8837 - val_loss: 6.2757 - val_accuracy: 0.5608\n",
            "Epoch 107/200\n",
            "2/2 [==============================] - 1s 409ms/step - loss: 0.5416 - accuracy: 0.8882 - val_loss: 6.2641 - val_accuracy: 0.5629\n",
            "Epoch 108/200\n",
            "2/2 [==============================] - 1s 400ms/step - loss: 0.5297 - accuracy: 0.8890 - val_loss: 6.2939 - val_accuracy: 0.5615\n",
            "Epoch 109/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 0.5101 - accuracy: 0.8935 - val_loss: 6.2873 - val_accuracy: 0.5624\n",
            "Epoch 110/200\n",
            "2/2 [==============================] - 1s 397ms/step - loss: 0.4973 - accuracy: 0.8974 - val_loss: 6.2981 - val_accuracy: 0.5611\n",
            "Epoch 111/200\n",
            "2/2 [==============================] - 1s 390ms/step - loss: 0.4800 - accuracy: 0.9014 - val_loss: 6.3070 - val_accuracy: 0.5610\n",
            "Epoch 112/200\n",
            "2/2 [==============================] - 1s 389ms/step - loss: 0.4684 - accuracy: 0.9030 - val_loss: 6.3085 - val_accuracy: 0.5623\n",
            "Epoch 113/200\n",
            "2/2 [==============================] - 1s 407ms/step - loss: 0.4524 - accuracy: 0.9059 - val_loss: 6.3266 - val_accuracy: 0.5600\n",
            "Epoch 114/200\n",
            "2/2 [==============================] - 1s 403ms/step - loss: 0.4400 - accuracy: 0.9085 - val_loss: 6.3216 - val_accuracy: 0.5610\n",
            "Epoch 115/200\n",
            "2/2 [==============================] - 1s 390ms/step - loss: 0.4276 - accuracy: 0.9101 - val_loss: 6.3363 - val_accuracy: 0.5605\n",
            "Epoch 116/200\n",
            "2/2 [==============================] - 1s 397ms/step - loss: 0.4149 - accuracy: 0.9129 - val_loss: 6.3453 - val_accuracy: 0.5604\n",
            "Epoch 117/200\n",
            "2/2 [==============================] - 1s 397ms/step - loss: 0.4026 - accuracy: 0.9175 - val_loss: 6.3434 - val_accuracy: 0.5613\n",
            "Epoch 118/200\n",
            "2/2 [==============================] - 1s 385ms/step - loss: 0.3935 - accuracy: 0.9189 - val_loss: 6.3548 - val_accuracy: 0.5604\n",
            "Epoch 119/200\n",
            "2/2 [==============================] - 1s 399ms/step - loss: 0.3816 - accuracy: 0.9217 - val_loss: 6.3560 - val_accuracy: 0.5614\n",
            "Epoch 120/200\n",
            "2/2 [==============================] - 1s 397ms/step - loss: 0.3713 - accuracy: 0.9232 - val_loss: 6.3620 - val_accuracy: 0.5606\n",
            "Epoch 121/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 0.3645 - accuracy: 0.9255 - val_loss: 6.3778 - val_accuracy: 0.5600\n",
            "Epoch 122/200\n",
            "2/2 [==============================] - 1s 393ms/step - loss: 0.3508 - accuracy: 0.9288 - val_loss: 6.3696 - val_accuracy: 0.5612\n",
            "Epoch 123/200\n",
            "2/2 [==============================] - 1s 392ms/step - loss: 0.3439 - accuracy: 0.9283 - val_loss: 6.3835 - val_accuracy: 0.5597\n",
            "Epoch 124/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 0.3353 - accuracy: 0.9318 - val_loss: 6.3858 - val_accuracy: 0.5607\n",
            "Epoch 125/200\n",
            "2/2 [==============================] - 1s 401ms/step - loss: 0.3275 - accuracy: 0.9333 - val_loss: 6.3899 - val_accuracy: 0.5620\n",
            "Epoch 126/200\n",
            "2/2 [==============================] - 1s 385ms/step - loss: 0.3242 - accuracy: 0.9346 - val_loss: 6.3971 - val_accuracy: 0.5608\n",
            "Epoch 127/200\n",
            "2/2 [==============================] - 1s 391ms/step - loss: 0.3128 - accuracy: 0.9372 - val_loss: 6.3982 - val_accuracy: 0.5609\n",
            "Epoch 128/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 0.3050 - accuracy: 0.9388 - val_loss: 6.4042 - val_accuracy: 0.5606\n",
            "Epoch 129/200\n",
            "2/2 [==============================] - 1s 393ms/step - loss: 0.3031 - accuracy: 0.9381 - val_loss: 6.4149 - val_accuracy: 0.5604\n",
            "Epoch 130/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 0.2936 - accuracy: 0.9413 - val_loss: 6.4208 - val_accuracy: 0.5610\n",
            "Epoch 131/200\n",
            "2/2 [==============================] - 1s 403ms/step - loss: 0.2873 - accuracy: 0.9424 - val_loss: 6.4296 - val_accuracy: 0.5600\n",
            "Epoch 132/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 0.2827 - accuracy: 0.9445 - val_loss: 6.4314 - val_accuracy: 0.5598\n",
            "Epoch 133/200\n",
            "2/2 [==============================] - 1s 391ms/step - loss: 0.2761 - accuracy: 0.9468 - val_loss: 6.4271 - val_accuracy: 0.5608\n",
            "Epoch 134/200\n",
            "2/2 [==============================] - 1s 388ms/step - loss: 0.2687 - accuracy: 0.9478 - val_loss: 6.4380 - val_accuracy: 0.5606\n",
            "Epoch 135/200\n",
            "2/2 [==============================] - 1s 395ms/step - loss: 0.2638 - accuracy: 0.9480 - val_loss: 6.4442 - val_accuracy: 0.5605\n",
            "Epoch 136/200\n",
            "2/2 [==============================] - 1s 389ms/step - loss: 0.2598 - accuracy: 0.9504 - val_loss: 6.4443 - val_accuracy: 0.5609\n",
            "Epoch 137/200\n",
            "2/2 [==============================] - 1s 389ms/step - loss: 0.2575 - accuracy: 0.9496 - val_loss: 6.4508 - val_accuracy: 0.5609\n",
            "Epoch 138/200\n",
            "2/2 [==============================] - 1s 388ms/step - loss: 0.2484 - accuracy: 0.9517 - val_loss: 6.4587 - val_accuracy: 0.5605\n",
            "Epoch 139/200\n",
            "2/2 [==============================] - 1s 385ms/step - loss: 0.2455 - accuracy: 0.9534 - val_loss: 6.4614 - val_accuracy: 0.5594\n",
            "Epoch 140/200\n",
            "2/2 [==============================] - 1s 393ms/step - loss: 0.2379 - accuracy: 0.9556 - val_loss: 6.4599 - val_accuracy: 0.5600\n",
            "Epoch 141/200\n",
            "2/2 [==============================] - 1s 393ms/step - loss: 0.2342 - accuracy: 0.9576 - val_loss: 6.4635 - val_accuracy: 0.5602\n",
            "Epoch 142/200\n",
            "2/2 [==============================] - 1s 389ms/step - loss: 0.2313 - accuracy: 0.9565 - val_loss: 6.4706 - val_accuracy: 0.5594\n",
            "Epoch 143/200\n",
            "2/2 [==============================] - 1s 388ms/step - loss: 0.2269 - accuracy: 0.9592 - val_loss: 6.4752 - val_accuracy: 0.5595\n",
            "Epoch 144/200\n",
            "2/2 [==============================] - 1s 395ms/step - loss: 0.2250 - accuracy: 0.9579 - val_loss: 6.4738 - val_accuracy: 0.5607\n",
            "Epoch 145/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 0.2187 - accuracy: 0.9601 - val_loss: 6.4855 - val_accuracy: 0.5590\n",
            "Epoch 146/200\n",
            "2/2 [==============================] - 1s 388ms/step - loss: 0.2146 - accuracy: 0.9611 - val_loss: 6.4892 - val_accuracy: 0.5597\n",
            "Epoch 147/200\n",
            "2/2 [==============================] - 1s 391ms/step - loss: 0.2093 - accuracy: 0.9628 - val_loss: 6.4860 - val_accuracy: 0.5606\n",
            "Epoch 148/200\n",
            "2/2 [==============================] - 1s 393ms/step - loss: 0.2067 - accuracy: 0.9624 - val_loss: 6.4852 - val_accuracy: 0.5609\n",
            "Epoch 149/200\n",
            "2/2 [==============================] - 1s 391ms/step - loss: 0.2036 - accuracy: 0.9636 - val_loss: 6.4952 - val_accuracy: 0.5603\n",
            "Epoch 150/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 0.2002 - accuracy: 0.9640 - val_loss: 6.4961 - val_accuracy: 0.5606\n",
            "Epoch 151/200\n",
            "2/2 [==============================] - 1s 393ms/step - loss: 0.1981 - accuracy: 0.9649 - val_loss: 6.4971 - val_accuracy: 0.5606\n",
            "Epoch 152/200\n",
            "2/2 [==============================] - 1s 384ms/step - loss: 0.1947 - accuracy: 0.9654 - val_loss: 6.5054 - val_accuracy: 0.5607\n",
            "Epoch 153/200\n",
            "2/2 [==============================] - 1s 390ms/step - loss: 0.1917 - accuracy: 0.9675 - val_loss: 6.5081 - val_accuracy: 0.5601\n",
            "Epoch 154/200\n",
            "2/2 [==============================] - 1s 400ms/step - loss: 0.1918 - accuracy: 0.9664 - val_loss: 6.5173 - val_accuracy: 0.5590\n",
            "Epoch 155/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 0.1836 - accuracy: 0.9697 - val_loss: 6.5094 - val_accuracy: 0.5595\n",
            "Epoch 156/200\n",
            "2/2 [==============================] - 1s 389ms/step - loss: 0.1843 - accuracy: 0.9691 - val_loss: 6.5130 - val_accuracy: 0.5606\n",
            "Epoch 157/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 0.1810 - accuracy: 0.9696 - val_loss: 6.5269 - val_accuracy: 0.5599\n",
            "Epoch 158/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 0.1782 - accuracy: 0.9703 - val_loss: 6.5268 - val_accuracy: 0.5600\n",
            "Epoch 159/200\n",
            "2/2 [==============================] - 1s 391ms/step - loss: 0.1730 - accuracy: 0.9712 - val_loss: 6.5278 - val_accuracy: 0.5601\n",
            "Epoch 160/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 0.1746 - accuracy: 0.9709 - val_loss: 6.5322 - val_accuracy: 0.5601\n",
            "Epoch 161/200\n",
            "2/2 [==============================] - 1s 388ms/step - loss: 0.1728 - accuracy: 0.9709 - val_loss: 6.5330 - val_accuracy: 0.5602\n",
            "Epoch 162/200\n",
            "2/2 [==============================] - 1s 391ms/step - loss: 0.1698 - accuracy: 0.9721 - val_loss: 6.5355 - val_accuracy: 0.5605\n",
            "Epoch 163/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 0.1687 - accuracy: 0.9731 - val_loss: 6.5406 - val_accuracy: 0.5606\n",
            "Epoch 164/200\n",
            "2/2 [==============================] - 1s 388ms/step - loss: 0.1640 - accuracy: 0.9740 - val_loss: 6.5409 - val_accuracy: 0.5609\n",
            "Epoch 165/200\n",
            "2/2 [==============================] - 1s 394ms/step - loss: 0.1643 - accuracy: 0.9734 - val_loss: 6.5418 - val_accuracy: 0.5609\n",
            "Epoch 166/200\n",
            "2/2 [==============================] - 1s 393ms/step - loss: 0.1608 - accuracy: 0.9741 - val_loss: 6.5523 - val_accuracy: 0.5606\n",
            "Epoch 167/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 0.1592 - accuracy: 0.9747 - val_loss: 6.5583 - val_accuracy: 0.5599\n",
            "Epoch 168/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 0.1605 - accuracy: 0.9743 - val_loss: 6.5508 - val_accuracy: 0.5606\n",
            "Epoch 169/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 0.1557 - accuracy: 0.9768 - val_loss: 6.5628 - val_accuracy: 0.5601\n",
            "Epoch 170/200\n",
            "2/2 [==============================] - 1s 391ms/step - loss: 0.1572 - accuracy: 0.9753 - val_loss: 6.5645 - val_accuracy: 0.5604\n",
            "Epoch 171/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 0.1555 - accuracy: 0.9760 - val_loss: 6.5544 - val_accuracy: 0.5610\n",
            "Epoch 172/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 0.1512 - accuracy: 0.9770 - val_loss: 6.5554 - val_accuracy: 0.5609\n",
            "Epoch 173/200\n",
            "2/2 [==============================] - 1s 393ms/step - loss: 0.1505 - accuracy: 0.9767 - val_loss: 6.5683 - val_accuracy: 0.5594\n",
            "Epoch 174/200\n",
            "2/2 [==============================] - 1s 391ms/step - loss: 0.1487 - accuracy: 0.9775 - val_loss: 6.5681 - val_accuracy: 0.5596\n",
            "Epoch 175/200\n",
            "2/2 [==============================] - 1s 395ms/step - loss: 0.1485 - accuracy: 0.9774 - val_loss: 6.5628 - val_accuracy: 0.5601\n",
            "Epoch 176/200\n",
            "2/2 [==============================] - 1s 388ms/step - loss: 0.1466 - accuracy: 0.9777 - val_loss: 6.5692 - val_accuracy: 0.5596\n",
            "Epoch 177/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 0.1455 - accuracy: 0.9780 - val_loss: 6.5755 - val_accuracy: 0.5595\n",
            "Epoch 178/200\n",
            "2/2 [==============================] - 1s 388ms/step - loss: 0.1444 - accuracy: 0.9790 - val_loss: 6.5800 - val_accuracy: 0.5600\n",
            "Epoch 179/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 0.1427 - accuracy: 0.9789 - val_loss: 6.5863 - val_accuracy: 0.5600\n",
            "Epoch 180/200\n",
            "2/2 [==============================] - 1s 388ms/step - loss: 0.1398 - accuracy: 0.9797 - val_loss: 6.5869 - val_accuracy: 0.5603\n",
            "Epoch 181/200\n",
            "2/2 [==============================] - 1s 390ms/step - loss: 0.1372 - accuracy: 0.9813 - val_loss: 6.5848 - val_accuracy: 0.5598\n",
            "Epoch 182/200\n",
            "2/2 [==============================] - 1s 387ms/step - loss: 0.1379 - accuracy: 0.9803 - val_loss: 6.5903 - val_accuracy: 0.5596\n",
            "Epoch 183/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 0.1391 - accuracy: 0.9806 - val_loss: 6.5911 - val_accuracy: 0.5602\n",
            "Epoch 184/200\n",
            "2/2 [==============================] - 1s 389ms/step - loss: 0.1362 - accuracy: 0.9806 - val_loss: 6.5871 - val_accuracy: 0.5602\n",
            "Epoch 185/200\n",
            "2/2 [==============================] - 1s 389ms/step - loss: 0.1349 - accuracy: 0.9811 - val_loss: 6.5889 - val_accuracy: 0.5601\n",
            "Epoch 186/200\n",
            "2/2 [==============================] - 1s 389ms/step - loss: 0.1324 - accuracy: 0.9824 - val_loss: 6.5976 - val_accuracy: 0.5600\n",
            "Epoch 187/200\n",
            "2/2 [==============================] - 1s 394ms/step - loss: 0.1312 - accuracy: 0.9827 - val_loss: 6.5989 - val_accuracy: 0.5607\n",
            "Epoch 188/200\n",
            "2/2 [==============================] - 1s 388ms/step - loss: 0.1309 - accuracy: 0.9829 - val_loss: 6.5983 - val_accuracy: 0.5609\n",
            "Epoch 189/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 0.1317 - accuracy: 0.9822 - val_loss: 6.6179 - val_accuracy: 0.5591\n",
            "Epoch 190/200\n",
            "2/2 [==============================] - 1s 392ms/step - loss: 0.1308 - accuracy: 0.9827 - val_loss: 6.6155 - val_accuracy: 0.5590\n",
            "Epoch 191/200\n",
            "2/2 [==============================] - 1s 395ms/step - loss: 0.1295 - accuracy: 0.9827 - val_loss: 6.6116 - val_accuracy: 0.5598\n",
            "Epoch 192/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 0.1281 - accuracy: 0.9832 - val_loss: 6.6142 - val_accuracy: 0.5601\n",
            "Epoch 193/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 0.1251 - accuracy: 0.9836 - val_loss: 6.6170 - val_accuracy: 0.5601\n",
            "Epoch 194/200\n",
            "2/2 [==============================] - 1s 389ms/step - loss: 0.1253 - accuracy: 0.9836 - val_loss: 6.6116 - val_accuracy: 0.5600\n",
            "Epoch 195/200\n",
            "2/2 [==============================] - 1s 389ms/step - loss: 0.1262 - accuracy: 0.9838 - val_loss: 6.6137 - val_accuracy: 0.5597\n",
            "Epoch 196/200\n",
            "2/2 [==============================] - 1s 394ms/step - loss: 0.1245 - accuracy: 0.9839 - val_loss: 6.6180 - val_accuracy: 0.5597\n",
            "Epoch 197/200\n",
            "2/2 [==============================] - 1s 390ms/step - loss: 0.1223 - accuracy: 0.9846 - val_loss: 6.6222 - val_accuracy: 0.5594\n",
            "Epoch 198/200\n",
            "2/2 [==============================] - 1s 398ms/step - loss: 0.1233 - accuracy: 0.9846 - val_loss: 6.6180 - val_accuracy: 0.5600\n",
            "Epoch 199/200\n",
            "2/2 [==============================] - 1s 397ms/step - loss: 0.1228 - accuracy: 0.9847 - val_loss: 6.6234 - val_accuracy: 0.5596\n",
            "Epoch 200/200\n",
            "2/2 [==============================] - 1s 386ms/step - loss: 0.1197 - accuracy: 0.9856 - val_loss: 6.6342 - val_accuracy: 0.5585\n",
            "WARNING:tensorflow:5 out of the last 5 calls to <function Model.make_predict_function.<locals>.predict_function at 0x7ffa5a0e8b90> triggered tf.function retracing. Tracing is expensive and the excessive number of tracings could be due to (1) creating @tf.function repeatedly in a loop, (2) passing tensors with different shapes, (3) passing Python objects instead of tensors. For (1), please define your @tf.function outside of the loop. For (2), @tf.function has experimental_relax_shapes=True option that relaxes argument shapes that can avoid unnecessary retracing. For (3), please refer to https://www.tensorflow.org/guide/function#controlling_retracing and https://www.tensorflow.org/api_docs/python/tf/function for  more details.\n",
            "ama albhamu akhe e solo eminyaka yawo 60 angamanye anezingoma ezimnandi kakhulu kwengake ngazizwa <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD>\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "HIRUUd1Yf45B"
      },
      "source": [
        "## Model 6 - Bidirectional RNN with Word Embedding and Byte-Pair Encoding"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "04xhuwvSu4Gl",
        "outputId": "52d6c9eb-0b57-4ebe-e45b-6fcfb7d5e2b4"
      },
      "source": [
        "tmp_x_bpe = pad(preproc_english_sentences_bpe, preproc_zulu_sentences_bpe.shape[1])\n",
        "tmp_x_bpe = tmp_x_bpe.reshape((-1, preproc_zulu_sentences_bpe.shape[-2]))\n",
        "tmp_x_test_bpe = pad(preproc_english_sentences_test_bpe, max_zulu_sequence_length_bpe)\n",
        "tmp_x_test_bpe = tmp_x_test_bpe.reshape((-1, preproc_zulu_sentences_bpe.shape[-2], 1))\n",
        "\n",
        "\n",
        "# TODO: Train and Print prediction(s)\n",
        "embed_rnn_model1_bpe = bd_model(\n",
        "    tmp_x_bpe.shape,\n",
        "    preproc_zulu_sentences_bpe.shape[1],\n",
        "    english_vocab_size_bpe+1,\n",
        "    zulu_vocab_size_bpe+1)\n",
        "\n",
        "embed_rnn_model1_bpe.summary()\n",
        "\n",
        "model6 = embed_rnn_model1_bpe.fit(tmp_x_bpe, preproc_zulu_sentences_bpe, batch_size=1024, epochs=NUM_EPOCHS, validation_split=0.2)\n",
        "\n",
        "print(logits_to_text_bpe(embed_rnn_model1_bpe.predict(tmp_x_bpe[:1])[0], tokenizer_zu))"
      ],
      "execution_count": 27,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential_5\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " embedding_3 (Embedding)     (None, 58, 256)           2560256   \n",
            "                                                                 \n",
            " bidirectional_1 (Bidirectio  (None, 58, 256)          296448    \n",
            " nal)                                                            \n",
            "                                                                 \n",
            " time_distributed_10 (TimeDi  (None, 58, 1024)         263168    \n",
            " stributed)                                                      \n",
            "                                                                 \n",
            " dropout_5 (Dropout)         (None, 58, 1024)          0         \n",
            "                                                                 \n",
            " time_distributed_11 (TimeDi  (None, 58, 10001)        10251025  \n",
            " stributed)                                                      \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 13,370,897\n",
            "Trainable params: 13,370,897\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n",
            "Epoch 1/200\n",
            "2/2 [==============================] - 6s 1s/step - loss: 9.1694 - accuracy: 0.2529 - val_loss: 8.7625 - val_accuracy: 0.6792\n",
            "Epoch 2/200\n",
            "2/2 [==============================] - 1s 639ms/step - loss: 8.4574 - accuracy: 0.6803 - val_loss: 5.9383 - val_accuracy: 0.6792\n",
            "Epoch 3/200\n",
            "2/2 [==============================] - 1s 652ms/step - loss: 4.8466 - accuracy: 0.6803 - val_loss: 3.6618 - val_accuracy: 0.6792\n",
            "Epoch 4/200\n",
            "2/2 [==============================] - 1s 642ms/step - loss: 3.5034 - accuracy: 0.6803 - val_loss: 3.5656 - val_accuracy: 0.6792\n",
            "Epoch 5/200\n",
            "2/2 [==============================] - 1s 644ms/step - loss: 3.1186 - accuracy: 0.6803 - val_loss: 3.1860 - val_accuracy: 0.6792\n",
            "Epoch 6/200\n",
            "2/2 [==============================] - 1s 643ms/step - loss: 2.8834 - accuracy: 0.6804 - val_loss: 3.2038 - val_accuracy: 0.6804\n",
            "Epoch 7/200\n",
            "2/2 [==============================] - 2s 659ms/step - loss: 2.9508 - accuracy: 0.6822 - val_loss: 3.1770 - val_accuracy: 0.6805\n",
            "Epoch 8/200\n",
            "2/2 [==============================] - 1s 655ms/step - loss: 2.8927 - accuracy: 0.6815 - val_loss: 3.1141 - val_accuracy: 0.6796\n",
            "Epoch 9/200\n",
            "2/2 [==============================] - 1s 658ms/step - loss: 2.7905 - accuracy: 0.6804 - val_loss: 3.1398 - val_accuracy: 0.6794\n",
            "Epoch 10/200\n",
            "2/2 [==============================] - 1s 656ms/step - loss: 2.7751 - accuracy: 0.6803 - val_loss: 3.1922 - val_accuracy: 0.6794\n",
            "Epoch 11/200\n",
            "2/2 [==============================] - 1s 657ms/step - loss: 2.7760 - accuracy: 0.6804 - val_loss: 3.1688 - val_accuracy: 0.6795\n",
            "Epoch 12/200\n",
            "2/2 [==============================] - 1s 656ms/step - loss: 2.7252 - accuracy: 0.6804 - val_loss: 3.1257 - val_accuracy: 0.6797\n",
            "Epoch 13/200\n",
            "2/2 [==============================] - 1s 656ms/step - loss: 2.6832 - accuracy: 0.6805 - val_loss: 3.1041 - val_accuracy: 0.6796\n",
            "Epoch 14/200\n",
            "2/2 [==============================] - 1s 655ms/step - loss: 2.6607 - accuracy: 0.6805 - val_loss: 3.0959 - val_accuracy: 0.6791\n",
            "Epoch 15/200\n",
            "2/2 [==============================] - 1s 654ms/step - loss: 2.6500 - accuracy: 0.6805 - val_loss: 3.0792 - val_accuracy: 0.6790\n",
            "Epoch 16/200\n",
            "2/2 [==============================] - 1s 654ms/step - loss: 2.6335 - accuracy: 0.6805 - val_loss: 3.0529 - val_accuracy: 0.6790\n",
            "Epoch 17/200\n",
            "2/2 [==============================] - 1s 658ms/step - loss: 2.6115 - accuracy: 0.6805 - val_loss: 3.0262 - val_accuracy: 0.6792\n",
            "Epoch 18/200\n",
            "2/2 [==============================] - 1s 656ms/step - loss: 2.5868 - accuracy: 0.6806 - val_loss: 3.0069 - val_accuracy: 0.6795\n",
            "Epoch 19/200\n",
            "2/2 [==============================] - 2s 654ms/step - loss: 2.5653 - accuracy: 0.6808 - val_loss: 2.9908 - val_accuracy: 0.6795\n",
            "Epoch 20/200\n",
            "2/2 [==============================] - 1s 652ms/step - loss: 2.5462 - accuracy: 0.6809 - val_loss: 2.9684 - val_accuracy: 0.6797\n",
            "Epoch 21/200\n",
            "2/2 [==============================] - 1s 656ms/step - loss: 2.5267 - accuracy: 0.6810 - val_loss: 2.9415 - val_accuracy: 0.6797\n",
            "Epoch 22/200\n",
            "2/2 [==============================] - 2s 664ms/step - loss: 2.5094 - accuracy: 0.6818 - val_loss: 2.9212 - val_accuracy: 0.6802\n",
            "Epoch 23/200\n",
            "2/2 [==============================] - 1s 649ms/step - loss: 2.4946 - accuracy: 0.6822 - val_loss: 2.9149 - val_accuracy: 0.6800\n",
            "Epoch 24/200\n",
            "2/2 [==============================] - 1s 659ms/step - loss: 2.4803 - accuracy: 0.6828 - val_loss: 2.9208 - val_accuracy: 0.6802\n",
            "Epoch 25/200\n",
            "2/2 [==============================] - 1s 654ms/step - loss: 2.4671 - accuracy: 0.6832 - val_loss: 2.9294 - val_accuracy: 0.6809\n",
            "Epoch 26/200\n",
            "2/2 [==============================] - 1s 656ms/step - loss: 2.4514 - accuracy: 0.6844 - val_loss: 2.9366 - val_accuracy: 0.6811\n",
            "Epoch 27/200\n",
            "2/2 [==============================] - 1s 659ms/step - loss: 2.4364 - accuracy: 0.6858 - val_loss: 2.9452 - val_accuracy: 0.6813\n",
            "Epoch 28/200\n",
            "2/2 [==============================] - 2s 659ms/step - loss: 2.4203 - accuracy: 0.6876 - val_loss: 2.9548 - val_accuracy: 0.6816\n",
            "Epoch 29/200\n",
            "2/2 [==============================] - 1s 658ms/step - loss: 2.4040 - accuracy: 0.6885 - val_loss: 2.9656 - val_accuracy: 0.6818\n",
            "Epoch 30/200\n",
            "2/2 [==============================] - 1s 652ms/step - loss: 2.3903 - accuracy: 0.6889 - val_loss: 2.9709 - val_accuracy: 0.6820\n",
            "Epoch 31/200\n",
            "2/2 [==============================] - 1s 652ms/step - loss: 2.3745 - accuracy: 0.6899 - val_loss: 2.9781 - val_accuracy: 0.6813\n",
            "Epoch 32/200\n",
            "2/2 [==============================] - 1s 657ms/step - loss: 2.3590 - accuracy: 0.6901 - val_loss: 2.9910 - val_accuracy: 0.6811\n",
            "Epoch 33/200\n",
            "2/2 [==============================] - 1s 659ms/step - loss: 2.3449 - accuracy: 0.6904 - val_loss: 3.0034 - val_accuracy: 0.6813\n",
            "Epoch 34/200\n",
            "2/2 [==============================] - 2s 661ms/step - loss: 2.3303 - accuracy: 0.6913 - val_loss: 3.0132 - val_accuracy: 0.6809\n",
            "Epoch 35/200\n",
            "2/2 [==============================] - 1s 656ms/step - loss: 2.3158 - accuracy: 0.6916 - val_loss: 3.0238 - val_accuracy: 0.6811\n",
            "Epoch 36/200\n",
            "2/2 [==============================] - 1s 652ms/step - loss: 2.2999 - accuracy: 0.6919 - val_loss: 3.0356 - val_accuracy: 0.6813\n",
            "Epoch 37/200\n",
            "2/2 [==============================] - 2s 654ms/step - loss: 2.2853 - accuracy: 0.6923 - val_loss: 3.0521 - val_accuracy: 0.6812\n",
            "Epoch 38/200\n",
            "2/2 [==============================] - 1s 662ms/step - loss: 2.2699 - accuracy: 0.6928 - val_loss: 3.0707 - val_accuracy: 0.6814\n",
            "Epoch 39/200\n",
            "2/2 [==============================] - 2s 660ms/step - loss: 2.2534 - accuracy: 0.6937 - val_loss: 3.0914 - val_accuracy: 0.6810\n",
            "Epoch 40/200\n",
            "2/2 [==============================] - 1s 650ms/step - loss: 2.2375 - accuracy: 0.6941 - val_loss: 3.1151 - val_accuracy: 0.6809\n",
            "Epoch 41/200\n",
            "2/2 [==============================] - 1s 655ms/step - loss: 2.2212 - accuracy: 0.6947 - val_loss: 3.1423 - val_accuracy: 0.6806\n",
            "Epoch 42/200\n",
            "2/2 [==============================] - 2s 658ms/step - loss: 2.2057 - accuracy: 0.6955 - val_loss: 3.1670 - val_accuracy: 0.6796\n",
            "Epoch 43/200\n",
            "2/2 [==============================] - 1s 649ms/step - loss: 2.1878 - accuracy: 0.6964 - val_loss: 3.2005 - val_accuracy: 0.6790\n",
            "Epoch 44/200\n",
            "2/2 [==============================] - 1s 655ms/step - loss: 2.1707 - accuracy: 0.6967 - val_loss: 3.2287 - val_accuracy: 0.6778\n",
            "Epoch 45/200\n",
            "2/2 [==============================] - 1s 653ms/step - loss: 2.1531 - accuracy: 0.6976 - val_loss: 3.2594 - val_accuracy: 0.6775\n",
            "Epoch 46/200\n",
            "2/2 [==============================] - 1s 654ms/step - loss: 2.1349 - accuracy: 0.6979 - val_loss: 3.2938 - val_accuracy: 0.6777\n",
            "Epoch 47/200\n",
            "2/2 [==============================] - 1s 652ms/step - loss: 2.1193 - accuracy: 0.6986 - val_loss: 3.3254 - val_accuracy: 0.6763\n",
            "Epoch 48/200\n",
            "2/2 [==============================] - 1s 656ms/step - loss: 2.0993 - accuracy: 0.6995 - val_loss: 3.3630 - val_accuracy: 0.6761\n",
            "Epoch 49/200\n",
            "2/2 [==============================] - 1s 651ms/step - loss: 2.0816 - accuracy: 0.6999 - val_loss: 3.3919 - val_accuracy: 0.6747\n",
            "Epoch 50/200\n",
            "2/2 [==============================] - 1s 652ms/step - loss: 2.0633 - accuracy: 0.7012 - val_loss: 3.4273 - val_accuracy: 0.6759\n",
            "Epoch 51/200\n",
            "2/2 [==============================] - 1s 657ms/step - loss: 2.0434 - accuracy: 0.7011 - val_loss: 3.4510 - val_accuracy: 0.6733\n",
            "Epoch 52/200\n",
            "2/2 [==============================] - 1s 655ms/step - loss: 2.0270 - accuracy: 0.7029 - val_loss: 3.4843 - val_accuracy: 0.6750\n",
            "Epoch 53/200\n",
            "2/2 [==============================] - 2s 653ms/step - loss: 2.0088 - accuracy: 0.7026 - val_loss: 3.5057 - val_accuracy: 0.6720\n",
            "Epoch 54/200\n",
            "2/2 [==============================] - 1s 654ms/step - loss: 1.9916 - accuracy: 0.7047 - val_loss: 3.5353 - val_accuracy: 0.6753\n",
            "Epoch 55/200\n",
            "2/2 [==============================] - 1s 653ms/step - loss: 1.9716 - accuracy: 0.7043 - val_loss: 3.5516 - val_accuracy: 0.6704\n",
            "Epoch 56/200\n",
            "2/2 [==============================] - 2s 658ms/step - loss: 1.9563 - accuracy: 0.7061 - val_loss: 3.5724 - val_accuracy: 0.6754\n",
            "Epoch 57/200\n",
            "2/2 [==============================] - 1s 656ms/step - loss: 1.9408 - accuracy: 0.7050 - val_loss: 3.5828 - val_accuracy: 0.6690\n",
            "Epoch 58/200\n",
            "2/2 [==============================] - 2s 657ms/step - loss: 1.9239 - accuracy: 0.7073 - val_loss: 3.6005 - val_accuracy: 0.6762\n",
            "Epoch 59/200\n",
            "2/2 [==============================] - 1s 645ms/step - loss: 1.9095 - accuracy: 0.7058 - val_loss: 3.6144 - val_accuracy: 0.6690\n",
            "Epoch 60/200\n",
            "2/2 [==============================] - 2s 646ms/step - loss: 1.8927 - accuracy: 0.7086 - val_loss: 3.6188 - val_accuracy: 0.6746\n",
            "Epoch 61/200\n",
            "2/2 [==============================] - 2s 662ms/step - loss: 1.8682 - accuracy: 0.7082 - val_loss: 3.6297 - val_accuracy: 0.6730\n",
            "Epoch 62/200\n",
            "2/2 [==============================] - 1s 648ms/step - loss: 1.8444 - accuracy: 0.7108 - val_loss: 3.6482 - val_accuracy: 0.6698\n",
            "Epoch 63/200\n",
            "2/2 [==============================] - 1s 651ms/step - loss: 1.8279 - accuracy: 0.7120 - val_loss: 3.6571 - val_accuracy: 0.6753\n",
            "Epoch 64/200\n",
            "2/2 [==============================] - 1s 649ms/step - loss: 1.8137 - accuracy: 0.7108 - val_loss: 3.6772 - val_accuracy: 0.6701\n",
            "Epoch 65/200\n",
            "2/2 [==============================] - 1s 653ms/step - loss: 1.7923 - accuracy: 0.7147 - val_loss: 3.6823 - val_accuracy: 0.6725\n",
            "Epoch 66/200\n",
            "2/2 [==============================] - 1s 652ms/step - loss: 1.7651 - accuracy: 0.7149 - val_loss: 3.6975 - val_accuracy: 0.6741\n",
            "Epoch 67/200\n",
            "2/2 [==============================] - 2s 657ms/step - loss: 1.7432 - accuracy: 0.7170 - val_loss: 3.7275 - val_accuracy: 0.6694\n",
            "Epoch 68/200\n",
            "2/2 [==============================] - 1s 648ms/step - loss: 1.7256 - accuracy: 0.7194 - val_loss: 3.7317 - val_accuracy: 0.6729\n",
            "Epoch 69/200\n",
            "2/2 [==============================] - 2s 658ms/step - loss: 1.7058 - accuracy: 0.7194 - val_loss: 3.7542 - val_accuracy: 0.6707\n",
            "Epoch 70/200\n",
            "2/2 [==============================] - 1s 647ms/step - loss: 1.6793 - accuracy: 0.7236 - val_loss: 3.7713 - val_accuracy: 0.6713\n",
            "Epoch 71/200\n",
            "2/2 [==============================] - 1s 649ms/step - loss: 1.6571 - accuracy: 0.7246 - val_loss: 3.7822 - val_accuracy: 0.6720\n",
            "Epoch 72/200\n",
            "2/2 [==============================] - 1s 650ms/step - loss: 1.6338 - accuracy: 0.7260 - val_loss: 3.8096 - val_accuracy: 0.6694\n",
            "Epoch 73/200\n",
            "2/2 [==============================] - 2s 659ms/step - loss: 1.6124 - accuracy: 0.7295 - val_loss: 3.8127 - val_accuracy: 0.6727\n",
            "Epoch 74/200\n",
            "2/2 [==============================] - 1s 649ms/step - loss: 1.5903 - accuracy: 0.7291 - val_loss: 3.8415 - val_accuracy: 0.6699\n",
            "Epoch 75/200\n",
            "2/2 [==============================] - 1s 648ms/step - loss: 1.5680 - accuracy: 0.7330 - val_loss: 3.8455 - val_accuracy: 0.6711\n",
            "Epoch 76/200\n",
            "2/2 [==============================] - 1s 652ms/step - loss: 1.5434 - accuracy: 0.7352 - val_loss: 3.8611 - val_accuracy: 0.6703\n",
            "Epoch 77/200\n",
            "2/2 [==============================] - 1s 652ms/step - loss: 1.5168 - accuracy: 0.7374 - val_loss: 3.8760 - val_accuracy: 0.6701\n",
            "Epoch 78/200\n",
            "2/2 [==============================] - 1s 650ms/step - loss: 1.4949 - accuracy: 0.7397 - val_loss: 3.8845 - val_accuracy: 0.6709\n",
            "Epoch 79/200\n",
            "2/2 [==============================] - 1s 650ms/step - loss: 1.4699 - accuracy: 0.7435 - val_loss: 3.9108 - val_accuracy: 0.6692\n",
            "Epoch 80/200\n",
            "2/2 [==============================] - 2s 666ms/step - loss: 1.4484 - accuracy: 0.7457 - val_loss: 3.9110 - val_accuracy: 0.6719\n",
            "Epoch 81/200\n",
            "2/2 [==============================] - 2s 701ms/step - loss: 1.4258 - accuracy: 0.7483 - val_loss: 3.9426 - val_accuracy: 0.6684\n",
            "Epoch 82/200\n",
            "2/2 [==============================] - 2s 659ms/step - loss: 1.4025 - accuracy: 0.7509 - val_loss: 3.9395 - val_accuracy: 0.6714\n",
            "Epoch 83/200\n",
            "2/2 [==============================] - 2s 649ms/step - loss: 1.3795 - accuracy: 0.7521 - val_loss: 3.9705 - val_accuracy: 0.6687\n",
            "Epoch 84/200\n",
            "2/2 [==============================] - 1s 647ms/step - loss: 1.3603 - accuracy: 0.7553 - val_loss: 3.9656 - val_accuracy: 0.6711\n",
            "Epoch 85/200\n",
            "2/2 [==============================] - 1s 648ms/step - loss: 1.3358 - accuracy: 0.7573 - val_loss: 4.0000 - val_accuracy: 0.6674\n",
            "Epoch 86/200\n",
            "2/2 [==============================] - 1s 648ms/step - loss: 1.3153 - accuracy: 0.7617 - val_loss: 3.9934 - val_accuracy: 0.6709\n",
            "Epoch 87/200\n",
            "2/2 [==============================] - 1s 652ms/step - loss: 1.2913 - accuracy: 0.7629 - val_loss: 4.0232 - val_accuracy: 0.6676\n",
            "Epoch 88/200\n",
            "2/2 [==============================] - 1s 656ms/step - loss: 1.2694 - accuracy: 0.7680 - val_loss: 4.0187 - val_accuracy: 0.6701\n",
            "Epoch 89/200\n",
            "2/2 [==============================] - 1s 647ms/step - loss: 1.2459 - accuracy: 0.7700 - val_loss: 4.0413 - val_accuracy: 0.6683\n",
            "Epoch 90/200\n",
            "2/2 [==============================] - 2s 656ms/step - loss: 1.2241 - accuracy: 0.7751 - val_loss: 4.0470 - val_accuracy: 0.6690\n",
            "Epoch 91/200\n",
            "2/2 [==============================] - 1s 650ms/step - loss: 1.2029 - accuracy: 0.7759 - val_loss: 4.0675 - val_accuracy: 0.6678\n",
            "Epoch 92/200\n",
            "2/2 [==============================] - 2s 652ms/step - loss: 1.1845 - accuracy: 0.7805 - val_loss: 4.0719 - val_accuracy: 0.6688\n",
            "Epoch 93/200\n",
            "2/2 [==============================] - 2s 656ms/step - loss: 1.1622 - accuracy: 0.7824 - val_loss: 4.0885 - val_accuracy: 0.6673\n",
            "Epoch 94/200\n",
            "2/2 [==============================] - 1s 651ms/step - loss: 1.1420 - accuracy: 0.7857 - val_loss: 4.0959 - val_accuracy: 0.6684\n",
            "Epoch 95/200\n",
            "2/2 [==============================] - 2s 658ms/step - loss: 1.1228 - accuracy: 0.7879 - val_loss: 4.1129 - val_accuracy: 0.6676\n",
            "Epoch 96/200\n",
            "2/2 [==============================] - 1s 645ms/step - loss: 1.1016 - accuracy: 0.7920 - val_loss: 4.1200 - val_accuracy: 0.6682\n",
            "Epoch 97/200\n",
            "2/2 [==============================] - 1s 651ms/step - loss: 1.0819 - accuracy: 0.7946 - val_loss: 4.1384 - val_accuracy: 0.6669\n",
            "Epoch 98/200\n",
            "2/2 [==============================] - 1s 655ms/step - loss: 1.0652 - accuracy: 0.7977 - val_loss: 4.1426 - val_accuracy: 0.6682\n",
            "Epoch 99/200\n",
            "2/2 [==============================] - 1s 649ms/step - loss: 1.0434 - accuracy: 0.7996 - val_loss: 4.1631 - val_accuracy: 0.6672\n",
            "Epoch 100/200\n",
            "2/2 [==============================] - 2s 658ms/step - loss: 1.0247 - accuracy: 0.8038 - val_loss: 4.1650 - val_accuracy: 0.6677\n",
            "Epoch 101/200\n",
            "2/2 [==============================] - 1s 656ms/step - loss: 1.0090 - accuracy: 0.8065 - val_loss: 4.1899 - val_accuracy: 0.6659\n",
            "Epoch 102/200\n",
            "2/2 [==============================] - 1s 654ms/step - loss: 0.9926 - accuracy: 0.8083 - val_loss: 4.1797 - val_accuracy: 0.6680\n",
            "Epoch 103/200\n",
            "2/2 [==============================] - 2s 653ms/step - loss: 0.9769 - accuracy: 0.8096 - val_loss: 4.2094 - val_accuracy: 0.6654\n",
            "Epoch 104/200\n",
            "2/2 [==============================] - 2s 669ms/step - loss: 0.9611 - accuracy: 0.8131 - val_loss: 4.1968 - val_accuracy: 0.6682\n",
            "Epoch 105/200\n",
            "2/2 [==============================] - 2s 660ms/step - loss: 0.9440 - accuracy: 0.8146 - val_loss: 4.2179 - val_accuracy: 0.6657\n",
            "Epoch 106/200\n",
            "2/2 [==============================] - 1s 648ms/step - loss: 0.9294 - accuracy: 0.8171 - val_loss: 4.2105 - val_accuracy: 0.6682\n",
            "Epoch 107/200\n",
            "2/2 [==============================] - 1s 650ms/step - loss: 0.9102 - accuracy: 0.8201 - val_loss: 4.2363 - val_accuracy: 0.6657\n",
            "Epoch 108/200\n",
            "2/2 [==============================] - 1s 647ms/step - loss: 0.8990 - accuracy: 0.8223 - val_loss: 4.2221 - val_accuracy: 0.6680\n",
            "Epoch 109/200\n",
            "2/2 [==============================] - 2s 656ms/step - loss: 0.8910 - accuracy: 0.8222 - val_loss: 4.2539 - val_accuracy: 0.6648\n",
            "Epoch 110/200\n",
            "2/2 [==============================] - 1s 649ms/step - loss: 0.8696 - accuracy: 0.8277 - val_loss: 4.2353 - val_accuracy: 0.6682\n",
            "Epoch 111/200\n",
            "2/2 [==============================] - 2s 669ms/step - loss: 0.8565 - accuracy: 0.8287 - val_loss: 4.2665 - val_accuracy: 0.6649\n",
            "Epoch 112/200\n",
            "2/2 [==============================] - 1s 647ms/step - loss: 0.8421 - accuracy: 0.8312 - val_loss: 4.2466 - val_accuracy: 0.6676\n",
            "Epoch 113/200\n",
            "2/2 [==============================] - 1s 649ms/step - loss: 0.8257 - accuracy: 0.8337 - val_loss: 4.2800 - val_accuracy: 0.6646\n",
            "Epoch 114/200\n",
            "2/2 [==============================] - 1s 647ms/step - loss: 0.8121 - accuracy: 0.8374 - val_loss: 4.2582 - val_accuracy: 0.6681\n",
            "Epoch 115/200\n",
            "2/2 [==============================] - 2s 670ms/step - loss: 0.7985 - accuracy: 0.8382 - val_loss: 4.2995 - val_accuracy: 0.6647\n",
            "Epoch 116/200\n",
            "2/2 [==============================] - 1s 651ms/step - loss: 0.7866 - accuracy: 0.8394 - val_loss: 4.2694 - val_accuracy: 0.6684\n",
            "Epoch 117/200\n",
            "2/2 [==============================] - 2s 668ms/step - loss: 0.7794 - accuracy: 0.8410 - val_loss: 4.3156 - val_accuracy: 0.6646\n",
            "Epoch 118/200\n",
            "2/2 [==============================] - 2s 661ms/step - loss: 0.7646 - accuracy: 0.8441 - val_loss: 4.2857 - val_accuracy: 0.6678\n",
            "Epoch 119/200\n",
            "2/2 [==============================] - 1s 651ms/step - loss: 0.7511 - accuracy: 0.8459 - val_loss: 4.3269 - val_accuracy: 0.6651\n",
            "Epoch 120/200\n",
            "2/2 [==============================] - 1s 649ms/step - loss: 0.7338 - accuracy: 0.8494 - val_loss: 4.3038 - val_accuracy: 0.6676\n",
            "Epoch 121/200\n",
            "2/2 [==============================] - 1s 650ms/step - loss: 0.7251 - accuracy: 0.8504 - val_loss: 4.3403 - val_accuracy: 0.6650\n",
            "Epoch 122/200\n",
            "2/2 [==============================] - 1s 650ms/step - loss: 0.7124 - accuracy: 0.8529 - val_loss: 4.3183 - val_accuracy: 0.6684\n",
            "Epoch 123/200\n",
            "2/2 [==============================] - 1s 653ms/step - loss: 0.7015 - accuracy: 0.8543 - val_loss: 4.3504 - val_accuracy: 0.6648\n",
            "Epoch 124/200\n",
            "2/2 [==============================] - 1s 655ms/step - loss: 0.6921 - accuracy: 0.8569 - val_loss: 4.3401 - val_accuracy: 0.6672\n",
            "Epoch 125/200\n",
            "2/2 [==============================] - 1s 647ms/step - loss: 0.6839 - accuracy: 0.8575 - val_loss: 4.3572 - val_accuracy: 0.6648\n",
            "Epoch 126/200\n",
            "2/2 [==============================] - 1s 650ms/step - loss: 0.6689 - accuracy: 0.8611 - val_loss: 4.3463 - val_accuracy: 0.6672\n",
            "Epoch 127/200\n",
            "2/2 [==============================] - 2s 657ms/step - loss: 0.6575 - accuracy: 0.8628 - val_loss: 4.3666 - val_accuracy: 0.6657\n",
            "Epoch 128/200\n",
            "2/2 [==============================] - 1s 658ms/step - loss: 0.6471 - accuracy: 0.8658 - val_loss: 4.3624 - val_accuracy: 0.6667\n",
            "Epoch 129/200\n",
            "2/2 [==============================] - 1s 651ms/step - loss: 0.6361 - accuracy: 0.8669 - val_loss: 4.3786 - val_accuracy: 0.6664\n",
            "Epoch 130/200\n",
            "2/2 [==============================] - 2s 661ms/step - loss: 0.6229 - accuracy: 0.8695 - val_loss: 4.3826 - val_accuracy: 0.6657\n",
            "Epoch 131/200\n",
            "2/2 [==============================] - 1s 653ms/step - loss: 0.6173 - accuracy: 0.8712 - val_loss: 4.3808 - val_accuracy: 0.6671\n",
            "Epoch 132/200\n",
            "2/2 [==============================] - 1s 648ms/step - loss: 0.6079 - accuracy: 0.8706 - val_loss: 4.3976 - val_accuracy: 0.6660\n",
            "Epoch 133/200\n",
            "2/2 [==============================] - 2s 656ms/step - loss: 0.5993 - accuracy: 0.8735 - val_loss: 4.3910 - val_accuracy: 0.6671\n",
            "Epoch 134/200\n",
            "2/2 [==============================] - 1s 651ms/step - loss: 0.5884 - accuracy: 0.8757 - val_loss: 4.4152 - val_accuracy: 0.6661\n",
            "Epoch 135/200\n",
            "2/2 [==============================] - 2s 660ms/step - loss: 0.5777 - accuracy: 0.8775 - val_loss: 4.4067 - val_accuracy: 0.6668\n",
            "Epoch 136/200\n",
            "2/2 [==============================] - 1s 646ms/step - loss: 0.5687 - accuracy: 0.8793 - val_loss: 4.4285 - val_accuracy: 0.6664\n",
            "Epoch 137/200\n",
            "2/2 [==============================] - 2s 659ms/step - loss: 0.5604 - accuracy: 0.8802 - val_loss: 4.4206 - val_accuracy: 0.6672\n",
            "Epoch 138/200\n",
            "2/2 [==============================] - 1s 647ms/step - loss: 0.5473 - accuracy: 0.8824 - val_loss: 4.4407 - val_accuracy: 0.6656\n",
            "Epoch 139/200\n",
            "2/2 [==============================] - 1s 650ms/step - loss: 0.5430 - accuracy: 0.8843 - val_loss: 4.4322 - val_accuracy: 0.6671\n",
            "Epoch 140/200\n",
            "2/2 [==============================] - 2s 660ms/step - loss: 0.5339 - accuracy: 0.8860 - val_loss: 4.4542 - val_accuracy: 0.6660\n",
            "Epoch 141/200\n",
            "2/2 [==============================] - 2s 658ms/step - loss: 0.5272 - accuracy: 0.8870 - val_loss: 4.4446 - val_accuracy: 0.6667\n",
            "Epoch 142/200\n",
            "2/2 [==============================] - 1s 650ms/step - loss: 0.5191 - accuracy: 0.8891 - val_loss: 4.4666 - val_accuracy: 0.6654\n",
            "Epoch 143/200\n",
            "2/2 [==============================] - 2s 657ms/step - loss: 0.5118 - accuracy: 0.8905 - val_loss: 4.4585 - val_accuracy: 0.6669\n",
            "Epoch 144/200\n",
            "2/2 [==============================] - 1s 652ms/step - loss: 0.5019 - accuracy: 0.8925 - val_loss: 4.4836 - val_accuracy: 0.6646\n",
            "Epoch 145/200\n",
            "2/2 [==============================] - 2s 656ms/step - loss: 0.4948 - accuracy: 0.8940 - val_loss: 4.4733 - val_accuracy: 0.6668\n",
            "Epoch 146/200\n",
            "2/2 [==============================] - 1s 649ms/step - loss: 0.4902 - accuracy: 0.8932 - val_loss: 4.4956 - val_accuracy: 0.6649\n",
            "Epoch 147/200\n",
            "2/2 [==============================] - 1s 650ms/step - loss: 0.4829 - accuracy: 0.8958 - val_loss: 4.4786 - val_accuracy: 0.6669\n",
            "Epoch 148/200\n",
            "2/2 [==============================] - 1s 649ms/step - loss: 0.4761 - accuracy: 0.8971 - val_loss: 4.5066 - val_accuracy: 0.6648\n",
            "Epoch 149/200\n",
            "2/2 [==============================] - 1s 650ms/step - loss: 0.4735 - accuracy: 0.8979 - val_loss: 4.4864 - val_accuracy: 0.6669\n",
            "Epoch 150/200\n",
            "2/2 [==============================] - 1s 652ms/step - loss: 0.4646 - accuracy: 0.8992 - val_loss: 4.5095 - val_accuracy: 0.6647\n",
            "Epoch 151/200\n",
            "2/2 [==============================] - 1s 647ms/step - loss: 0.4581 - accuracy: 0.9015 - val_loss: 4.4850 - val_accuracy: 0.6677\n",
            "Epoch 152/200\n",
            "2/2 [==============================] - 1s 651ms/step - loss: 0.4571 - accuracy: 0.8998 - val_loss: 4.5280 - val_accuracy: 0.6630\n",
            "Epoch 153/200\n",
            "2/2 [==============================] - 2s 657ms/step - loss: 0.4555 - accuracy: 0.9002 - val_loss: 4.4832 - val_accuracy: 0.6680\n",
            "Epoch 154/200\n",
            "2/2 [==============================] - 2s 654ms/step - loss: 0.4476 - accuracy: 0.9029 - val_loss: 4.5386 - val_accuracy: 0.6629\n",
            "Epoch 155/200\n",
            "2/2 [==============================] - 1s 649ms/step - loss: 0.4504 - accuracy: 0.9012 - val_loss: 4.4835 - val_accuracy: 0.6677\n",
            "Epoch 156/200\n",
            "2/2 [==============================] - 2s 653ms/step - loss: 0.4459 - accuracy: 0.9016 - val_loss: 4.5429 - val_accuracy: 0.6628\n",
            "Epoch 157/200\n",
            "2/2 [==============================] - 2s 653ms/step - loss: 0.4402 - accuracy: 0.9035 - val_loss: 4.4864 - val_accuracy: 0.6678\n",
            "Epoch 158/200\n",
            "2/2 [==============================] - 2s 656ms/step - loss: 0.4374 - accuracy: 0.9041 - val_loss: 4.5242 - val_accuracy: 0.6643\n",
            "Epoch 159/200\n",
            "2/2 [==============================] - 2s 656ms/step - loss: 0.4272 - accuracy: 0.9072 - val_loss: 4.5048 - val_accuracy: 0.6659\n",
            "Epoch 160/200\n",
            "2/2 [==============================] - 2s 649ms/step - loss: 0.4142 - accuracy: 0.9095 - val_loss: 4.5167 - val_accuracy: 0.6670\n",
            "Epoch 161/200\n",
            "2/2 [==============================] - 2s 657ms/step - loss: 0.4140 - accuracy: 0.9086 - val_loss: 4.5372 - val_accuracy: 0.6641\n",
            "Epoch 162/200\n",
            "2/2 [==============================] - 1s 649ms/step - loss: 0.4079 - accuracy: 0.9107 - val_loss: 4.5057 - val_accuracy: 0.6680\n",
            "Epoch 163/200\n",
            "2/2 [==============================] - 1s 648ms/step - loss: 0.4001 - accuracy: 0.9115 - val_loss: 4.5495 - val_accuracy: 0.6653\n",
            "Epoch 164/200\n",
            "2/2 [==============================] - 1s 649ms/step - loss: 0.3944 - accuracy: 0.9132 - val_loss: 4.5330 - val_accuracy: 0.6659\n",
            "Epoch 165/200\n",
            "2/2 [==============================] - 2s 661ms/step - loss: 0.3839 - accuracy: 0.9152 - val_loss: 4.5398 - val_accuracy: 0.6667\n",
            "Epoch 166/200\n",
            "2/2 [==============================] - 2s 654ms/step - loss: 0.3773 - accuracy: 0.9168 - val_loss: 4.5646 - val_accuracy: 0.6655\n",
            "Epoch 167/200\n",
            "2/2 [==============================] - 1s 649ms/step - loss: 0.3762 - accuracy: 0.9179 - val_loss: 4.5500 - val_accuracy: 0.6667\n",
            "Epoch 168/200\n",
            "2/2 [==============================] - 1s 649ms/step - loss: 0.3683 - accuracy: 0.9189 - val_loss: 4.5695 - val_accuracy: 0.6659\n",
            "Epoch 169/200\n",
            "2/2 [==============================] - 2s 659ms/step - loss: 0.3619 - accuracy: 0.9212 - val_loss: 4.5737 - val_accuracy: 0.6659\n",
            "Epoch 170/200\n",
            "2/2 [==============================] - 2s 657ms/step - loss: 0.3599 - accuracy: 0.9202 - val_loss: 4.5770 - val_accuracy: 0.6658\n",
            "Epoch 171/200\n",
            "2/2 [==============================] - 1s 649ms/step - loss: 0.3546 - accuracy: 0.9222 - val_loss: 4.5859 - val_accuracy: 0.6659\n",
            "Epoch 172/200\n",
            "2/2 [==============================] - 2s 660ms/step - loss: 0.3481 - accuracy: 0.9233 - val_loss: 4.5862 - val_accuracy: 0.6664\n",
            "Epoch 173/200\n",
            "2/2 [==============================] - 1s 650ms/step - loss: 0.3428 - accuracy: 0.9247 - val_loss: 4.5963 - val_accuracy: 0.6654\n",
            "Epoch 174/200\n",
            "2/2 [==============================] - 2s 652ms/step - loss: 0.3396 - accuracy: 0.9251 - val_loss: 4.5898 - val_accuracy: 0.6667\n",
            "Epoch 175/200\n",
            "2/2 [==============================] - 2s 661ms/step - loss: 0.3326 - accuracy: 0.9270 - val_loss: 4.6088 - val_accuracy: 0.6657\n",
            "Epoch 176/200\n",
            "2/2 [==============================] - 2s 649ms/step - loss: 0.3303 - accuracy: 0.9275 - val_loss: 4.5999 - val_accuracy: 0.6663\n",
            "Epoch 177/200\n",
            "2/2 [==============================] - 2s 659ms/step - loss: 0.3266 - accuracy: 0.9277 - val_loss: 4.6129 - val_accuracy: 0.6655\n",
            "Epoch 178/200\n",
            "2/2 [==============================] - 1s 652ms/step - loss: 0.3234 - accuracy: 0.9288 - val_loss: 4.6158 - val_accuracy: 0.6659\n",
            "Epoch 179/200\n",
            "2/2 [==============================] - 1s 649ms/step - loss: 0.3139 - accuracy: 0.9313 - val_loss: 4.6163 - val_accuracy: 0.6659\n",
            "Epoch 180/200\n",
            "2/2 [==============================] - 2s 660ms/step - loss: 0.3129 - accuracy: 0.9313 - val_loss: 4.6254 - val_accuracy: 0.6653\n",
            "Epoch 181/200\n",
            "2/2 [==============================] - 1s 650ms/step - loss: 0.3105 - accuracy: 0.9315 - val_loss: 4.6227 - val_accuracy: 0.6660\n",
            "Epoch 182/200\n",
            "2/2 [==============================] - 1s 649ms/step - loss: 0.3043 - accuracy: 0.9323 - val_loss: 4.6327 - val_accuracy: 0.6656\n",
            "Epoch 183/200\n",
            "2/2 [==============================] - 2s 655ms/step - loss: 0.3015 - accuracy: 0.9334 - val_loss: 4.6362 - val_accuracy: 0.6656\n",
            "Epoch 184/200\n",
            "2/2 [==============================] - 1s 651ms/step - loss: 0.2982 - accuracy: 0.9342 - val_loss: 4.6382 - val_accuracy: 0.6659\n",
            "Epoch 185/200\n",
            "2/2 [==============================] - 2s 660ms/step - loss: 0.2943 - accuracy: 0.9350 - val_loss: 4.6461 - val_accuracy: 0.6655\n",
            "Epoch 186/200\n",
            "2/2 [==============================] - 2s 651ms/step - loss: 0.2937 - accuracy: 0.9351 - val_loss: 4.6461 - val_accuracy: 0.6660\n",
            "Epoch 187/200\n",
            "2/2 [==============================] - 1s 649ms/step - loss: 0.2912 - accuracy: 0.9358 - val_loss: 4.6545 - val_accuracy: 0.6655\n",
            "Epoch 188/200\n",
            "2/2 [==============================] - 2s 662ms/step - loss: 0.2860 - accuracy: 0.9374 - val_loss: 4.6530 - val_accuracy: 0.6659\n",
            "Epoch 189/200\n",
            "2/2 [==============================] - 2s 656ms/step - loss: 0.2813 - accuracy: 0.9381 - val_loss: 4.6583 - val_accuracy: 0.6653\n",
            "Epoch 190/200\n",
            "2/2 [==============================] - 2s 660ms/step - loss: 0.2810 - accuracy: 0.9385 - val_loss: 4.6650 - val_accuracy: 0.6649\n",
            "Epoch 191/200\n",
            "2/2 [==============================] - 1s 650ms/step - loss: 0.2731 - accuracy: 0.9400 - val_loss: 4.6627 - val_accuracy: 0.6658\n",
            "Epoch 192/200\n",
            "2/2 [==============================] - 1s 650ms/step - loss: 0.2732 - accuracy: 0.9407 - val_loss: 4.6744 - val_accuracy: 0.6649\n",
            "Epoch 193/200\n",
            "2/2 [==============================] - 1s 651ms/step - loss: 0.2714 - accuracy: 0.9397 - val_loss: 4.6715 - val_accuracy: 0.6657\n",
            "Epoch 194/200\n",
            "2/2 [==============================] - 1s 648ms/step - loss: 0.2654 - accuracy: 0.9412 - val_loss: 4.6813 - val_accuracy: 0.6655\n",
            "Epoch 195/200\n",
            "2/2 [==============================] - 1s 652ms/step - loss: 0.2644 - accuracy: 0.9420 - val_loss: 4.6845 - val_accuracy: 0.6656\n",
            "Epoch 196/200\n",
            "2/2 [==============================] - 2s 655ms/step - loss: 0.2583 - accuracy: 0.9437 - val_loss: 4.6836 - val_accuracy: 0.6659\n",
            "Epoch 197/200\n",
            "2/2 [==============================] - 1s 650ms/step - loss: 0.2576 - accuracy: 0.9437 - val_loss: 4.6940 - val_accuracy: 0.6652\n",
            "Epoch 198/200\n",
            "2/2 [==============================] - 2s 653ms/step - loss: 0.2540 - accuracy: 0.9440 - val_loss: 4.6867 - val_accuracy: 0.6661\n",
            "Epoch 199/200\n",
            "2/2 [==============================] - 1s 655ms/step - loss: 0.2525 - accuracy: 0.9447 - val_loss: 4.7024 - val_accuracy: 0.6651\n",
            "Epoch 200/200\n",
            "2/2 [==============================] - 2s 657ms/step - loss: 0.2491 - accuracy: 0.9455 - val_loss: 4.6928 - val_accuracy: 0.6660\n",
            "WARNING:tensorflow:6 out of the last 6 calls to <function Model.make_predict_function.<locals>.predict_function at 0x7ff82a1264d0> triggered tf.function retracing. Tracing is expensive and the excessive number of tracings could be due to (1) creating @tf.function repeatedly in a loop, (2) passing tensors with different shapes, (3) passing Python objects instead of tensors. For (1), please define your @tf.function outside of the loop. For (2), @tf.function has experimental_relax_shapes=True option that relaxes argument shapes that can avoid unnecessary retracing. For (3), please refer to https://www.tensorflow.org/guide/function#controlling_retracing and https://www.tensorflow.org/api_docs/python/tf/function for  more details.\n",
            "i - albhamu akhe e - solo eminyaka yawo - 60 anga manye anezi ngoma ezimnandi kakhulu kwe nga ke ngazi zwa . [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "P4eDE0tZf-lk"
      },
      "source": [
        "## Model 7 - Bidirectional RNN with Word Embedding, Encoder-Decoder and Standard Tokenization"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "C5-CCBpaSj5Z"
      },
      "source": [
        "def encdec_model(input_shape, output_sequence_length, english_vocab_size, zulu_vocab_size):\n",
        "    \"\"\"\n",
        "    Build and train an encoder-decoder model on x and y\n",
        "    :param input_shape: Tuple of input shape\n",
        "    :param output_sequence_length: Length of output sequence\n",
        "    :param english_vocab_size: Number of unique English words in the dataset\n",
        "    :param zulu_vocab_size: Number of unique zulu words in the dataset\n",
        "    :return: Keras model built, but not trained\n",
        "    \"\"\"\n",
        "    # OPTIONAL: Implement\n",
        "    \n",
        "    # Hyperparameters\n",
        "    learning_rate = 0.001\n",
        "    \n",
        "    # Build the layers    \n",
        "    model = Sequential()\n",
        "    # Embedding\n",
        "    model.add(Embedding(english_vocab_size, 256, input_length=input_shape[1], input_shape=input_shape[1:]))\n",
        "    \n",
        "    # Encoder\n",
        "    model.add(Bidirectional(GRU(128), input_shape=input_shape[1:]))\n",
        "    model.add(RepeatVector(output_sequence_length))\n",
        "    # Decoder\n",
        "    model.add(Bidirectional(GRU(128, return_sequences=True), input_shape=input_shape[1:]))\n",
        "    model.add(TimeDistributed(Dense(1024, activation='relu')))\n",
        "    model.add(Dropout(0.5))\n",
        "    model.add(TimeDistributed(Dense(zulu_vocab_size, activation='softmax')))\n",
        "\n",
        "    # Compile model\n",
        "    model.compile(loss=sparse_categorical_crossentropy,\n",
        "                  optimizer=keras.optimizers.Adam(learning_rate),\n",
        "                  metrics=['accuracy'])\n",
        "    \n",
        "    return model"
      ],
      "execution_count": 28,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "xBPleeftSwS7",
        "outputId": "0840f753-c606-404b-ada3-12afcb9d9915"
      },
      "source": [
        "tmp_x = pad(preproc_english_sentences, preproc_zulu_sentences.shape[1])\n",
        "tmp_x = tmp_x.reshape((-1, preproc_zulu_sentences.shape[-2]))\n",
        "tmp_x_test = pad(preproc_english_sentences_test, max_zulu_sequence_length)\n",
        "tmp_x_test = tmp_x_test.reshape((-1, preproc_zulu_sentences.shape[-2], 1))\n",
        "\n",
        "\n",
        "# TODO: Train and Print prediction(s)\n",
        "encdec_rnn_model = encdec_model(\n",
        "    tmp_x.shape,\n",
        "    preproc_zulu_sentences.shape[1],\n",
        "    english_vocab_size+1,\n",
        "    zulu_vocab_size+1)\n",
        "\n",
        "encdec_rnn_model.summary()\n",
        "\n",
        "model7 = encdec_rnn_model.fit(tmp_x, preproc_zulu_sentences, batch_size=1024, epochs=NUM_EPOCHS, validation_split=0.2)\n",
        "\n",
        "\n",
        "print(logits_to_text(encdec_rnn_model.predict(tmp_x[:1])[0], zulu_tokenizer))"
      ],
      "execution_count": 29,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential_6\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " embedding_4 (Embedding)     (None, 25, 256)           1955840   \n",
            "                                                                 \n",
            " bidirectional_2 (Bidirectio  (None, 256)              296448    \n",
            " nal)                                                            \n",
            "                                                                 \n",
            " repeat_vector (RepeatVector  (None, 25, 256)          0         \n",
            " )                                                               \n",
            "                                                                 \n",
            " bidirectional_3 (Bidirectio  (None, 25, 256)          296448    \n",
            " nal)                                                            \n",
            "                                                                 \n",
            " time_distributed_12 (TimeDi  (None, 25, 1024)         263168    \n",
            " stributed)                                                      \n",
            "                                                                 \n",
            " dropout_6 (Dropout)         (None, 25, 1024)          0         \n",
            "                                                                 \n",
            " time_distributed_13 (TimeDi  (None, 25, 12598)        12912950  \n",
            " stributed)                                                      \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 15,724,854\n",
            "Trainable params: 15,724,854\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n",
            "Epoch 1/200\n",
            "2/2 [==============================] - 9s 2s/step - loss: 9.4296 - accuracy: 0.2237 - val_loss: 9.3608 - val_accuracy: 0.5831\n",
            "Epoch 2/200\n",
            "2/2 [==============================] - 1s 393ms/step - loss: 9.3232 - accuracy: 0.5948 - val_loss: 9.0985 - val_accuracy: 0.5831\n",
            "Epoch 3/200\n",
            "2/2 [==============================] - 1s 396ms/step - loss: 8.9764 - accuracy: 0.5948 - val_loss: 8.2979 - val_accuracy: 0.5831\n",
            "Epoch 4/200\n",
            "2/2 [==============================] - 1s 398ms/step - loss: 7.9376 - accuracy: 0.5948 - val_loss: 6.1717 - val_accuracy: 0.5831\n",
            "Epoch 5/200\n",
            "2/2 [==============================] - 1s 413ms/step - loss: 5.4562 - accuracy: 0.5948 - val_loss: 5.1421 - val_accuracy: 0.5831\n",
            "Epoch 6/200\n",
            "2/2 [==============================] - 1s 396ms/step - loss: 4.9741 - accuracy: 0.5948 - val_loss: 5.8074 - val_accuracy: 0.5831\n",
            "Epoch 7/200\n",
            "2/2 [==============================] - 1s 400ms/step - loss: 5.4629 - accuracy: 0.5948 - val_loss: 5.5468 - val_accuracy: 0.5831\n",
            "Epoch 8/200\n",
            "2/2 [==============================] - 1s 407ms/step - loss: 5.0309 - accuracy: 0.5948 - val_loss: 5.0974 - val_accuracy: 0.5831\n",
            "Epoch 9/200\n",
            "2/2 [==============================] - 1s 399ms/step - loss: 4.4289 - accuracy: 0.5948 - val_loss: 5.0290 - val_accuracy: 0.5831\n",
            "Epoch 10/200\n",
            "2/2 [==============================] - 1s 409ms/step - loss: 4.4631 - accuracy: 0.5948 - val_loss: 5.0297 - val_accuracy: 0.5831\n",
            "Epoch 11/200\n",
            "2/2 [==============================] - 1s 395ms/step - loss: 4.3687 - accuracy: 0.5948 - val_loss: 4.7327 - val_accuracy: 0.5831\n",
            "Epoch 12/200\n",
            "2/2 [==============================] - 1s 394ms/step - loss: 4.0051 - accuracy: 0.5948 - val_loss: 4.8177 - val_accuracy: 0.5831\n",
            "Epoch 13/200\n",
            "2/2 [==============================] - 1s 400ms/step - loss: 4.0212 - accuracy: 0.5948 - val_loss: 4.9186 - val_accuracy: 0.5831\n",
            "Epoch 14/200\n",
            "2/2 [==============================] - 1s 397ms/step - loss: 4.0540 - accuracy: 0.5948 - val_loss: 4.8134 - val_accuracy: 0.5831\n",
            "Epoch 15/200\n",
            "2/2 [==============================] - 1s 419ms/step - loss: 3.9252 - accuracy: 0.5948 - val_loss: 4.6932 - val_accuracy: 0.5831\n",
            "Epoch 16/200\n",
            "2/2 [==============================] - 1s 396ms/step - loss: 3.8947 - accuracy: 0.5948 - val_loss: 4.7046 - val_accuracy: 0.5831\n",
            "Epoch 17/200\n",
            "2/2 [==============================] - 1s 397ms/step - loss: 3.9406 - accuracy: 0.5948 - val_loss: 4.6810 - val_accuracy: 0.5831\n",
            "Epoch 18/200\n",
            "2/2 [==============================] - 1s 397ms/step - loss: 3.8667 - accuracy: 0.5948 - val_loss: 4.7048 - val_accuracy: 0.5831\n",
            "Epoch 19/200\n",
            "2/2 [==============================] - 1s 394ms/step - loss: 3.8231 - accuracy: 0.5948 - val_loss: 4.7594 - val_accuracy: 0.5831\n",
            "Epoch 20/200\n",
            "2/2 [==============================] - 1s 417ms/step - loss: 3.8303 - accuracy: 0.5948 - val_loss: 4.7413 - val_accuracy: 0.5831\n",
            "Epoch 21/200\n",
            "2/2 [==============================] - 1s 401ms/step - loss: 3.7840 - accuracy: 0.5948 - val_loss: 4.6677 - val_accuracy: 0.5831\n",
            "Epoch 22/200\n",
            "2/2 [==============================] - 1s 395ms/step - loss: 3.7357 - accuracy: 0.5948 - val_loss: 4.6234 - val_accuracy: 0.5831\n",
            "Epoch 23/200\n",
            "2/2 [==============================] - 1s 397ms/step - loss: 3.7322 - accuracy: 0.5948 - val_loss: 4.6351 - val_accuracy: 0.5831\n",
            "Epoch 24/200\n",
            "2/2 [==============================] - 1s 404ms/step - loss: 3.7091 - accuracy: 0.5948 - val_loss: 4.7233 - val_accuracy: 0.5831\n",
            "Epoch 25/200\n",
            "2/2 [==============================] - 1s 408ms/step - loss: 3.6980 - accuracy: 0.5948 - val_loss: 4.7435 - val_accuracy: 0.5831\n",
            "Epoch 26/200\n",
            "2/2 [==============================] - 1s 407ms/step - loss: 3.6816 - accuracy: 0.5948 - val_loss: 4.6755 - val_accuracy: 0.5831\n",
            "Epoch 27/200\n",
            "2/2 [==============================] - 1s 402ms/step - loss: 3.6546 - accuracy: 0.5948 - val_loss: 4.6305 - val_accuracy: 0.5831\n",
            "Epoch 28/200\n",
            "2/2 [==============================] - 1s 398ms/step - loss: 3.6478 - accuracy: 0.5948 - val_loss: 4.6292 - val_accuracy: 0.5831\n",
            "Epoch 29/200\n",
            "2/2 [==============================] - 1s 395ms/step - loss: 3.6331 - accuracy: 0.5948 - val_loss: 4.6554 - val_accuracy: 0.5831\n",
            "Epoch 30/200\n",
            "2/2 [==============================] - 1s 406ms/step - loss: 3.6206 - accuracy: 0.5950 - val_loss: 4.6709 - val_accuracy: 0.5831\n",
            "Epoch 31/200\n",
            "2/2 [==============================] - 1s 412ms/step - loss: 3.6019 - accuracy: 0.5951 - val_loss: 4.6719 - val_accuracy: 0.5831\n",
            "Epoch 32/200\n",
            "2/2 [==============================] - 1s 395ms/step - loss: 3.5939 - accuracy: 0.5955 - val_loss: 4.7153 - val_accuracy: 0.5831\n",
            "Epoch 33/200\n",
            "2/2 [==============================] - 1s 398ms/step - loss: 3.5783 - accuracy: 0.5957 - val_loss: 4.7268 - val_accuracy: 0.5846\n",
            "Epoch 34/200\n",
            "2/2 [==============================] - 1s 401ms/step - loss: 3.5625 - accuracy: 0.5961 - val_loss: 4.6960 - val_accuracy: 0.5856\n",
            "Epoch 35/200\n",
            "2/2 [==============================] - 1s 421ms/step - loss: 3.5547 - accuracy: 0.5963 - val_loss: 4.7313 - val_accuracy: 0.5857\n",
            "Epoch 36/200\n",
            "2/2 [==============================] - 1s 402ms/step - loss: 3.5431 - accuracy: 0.5969 - val_loss: 4.7338 - val_accuracy: 0.5860\n",
            "Epoch 37/200\n",
            "2/2 [==============================] - 1s 399ms/step - loss: 3.5300 - accuracy: 0.5973 - val_loss: 4.7512 - val_accuracy: 0.5864\n",
            "Epoch 38/200\n",
            "2/2 [==============================] - 1s 448ms/step - loss: 3.5181 - accuracy: 0.5974 - val_loss: 4.7972 - val_accuracy: 0.5864\n",
            "Epoch 39/200\n",
            "2/2 [==============================] - 1s 407ms/step - loss: 3.5080 - accuracy: 0.5976 - val_loss: 4.7981 - val_accuracy: 0.5864\n",
            "Epoch 40/200\n",
            "2/2 [==============================] - 1s 398ms/step - loss: 3.4965 - accuracy: 0.5979 - val_loss: 4.8315 - val_accuracy: 0.5864\n",
            "Epoch 41/200\n",
            "2/2 [==============================] - 1s 401ms/step - loss: 3.4848 - accuracy: 0.5978 - val_loss: 4.8203 - val_accuracy: 0.5863\n",
            "Epoch 42/200\n",
            "2/2 [==============================] - 1s 409ms/step - loss: 3.4749 - accuracy: 0.5981 - val_loss: 4.8645 - val_accuracy: 0.5863\n",
            "Epoch 43/200\n",
            "2/2 [==============================] - 1s 413ms/step - loss: 3.4676 - accuracy: 0.5984 - val_loss: 4.8559 - val_accuracy: 0.5861\n",
            "Epoch 44/200\n",
            "2/2 [==============================] - 1s 399ms/step - loss: 3.4632 - accuracy: 0.5982 - val_loss: 4.9051 - val_accuracy: 0.5863\n",
            "Epoch 45/200\n",
            "2/2 [==============================] - 1s 415ms/step - loss: 3.4576 - accuracy: 0.5986 - val_loss: 4.8896 - val_accuracy: 0.5860\n",
            "Epoch 46/200\n",
            "2/2 [==============================] - 1s 405ms/step - loss: 3.4450 - accuracy: 0.5985 - val_loss: 4.9112 - val_accuracy: 0.5861\n",
            "Epoch 47/200\n",
            "2/2 [==============================] - 1s 394ms/step - loss: 3.4363 - accuracy: 0.5988 - val_loss: 4.9221 - val_accuracy: 0.5863\n",
            "Epoch 48/200\n",
            "2/2 [==============================] - 1s 397ms/step - loss: 3.4276 - accuracy: 0.5984 - val_loss: 4.9380 - val_accuracy: 0.5866\n",
            "Epoch 49/200\n",
            "2/2 [==============================] - 1s 421ms/step - loss: 3.4166 - accuracy: 0.5990 - val_loss: 4.9634 - val_accuracy: 0.5870\n",
            "Epoch 50/200\n",
            "2/2 [==============================] - 1s 404ms/step - loss: 3.4095 - accuracy: 0.5987 - val_loss: 4.9571 - val_accuracy: 0.5866\n",
            "Epoch 51/200\n",
            "2/2 [==============================] - 1s 397ms/step - loss: 3.3974 - accuracy: 0.5995 - val_loss: 4.9796 - val_accuracy: 0.5871\n",
            "Epoch 52/200\n",
            "2/2 [==============================] - 1s 403ms/step - loss: 3.3876 - accuracy: 0.5994 - val_loss: 4.9784 - val_accuracy: 0.5869\n",
            "Epoch 53/200\n",
            "2/2 [==============================] - 1s 396ms/step - loss: 3.3776 - accuracy: 0.5993 - val_loss: 5.0105 - val_accuracy: 0.5870\n",
            "Epoch 54/200\n",
            "2/2 [==============================] - 1s 406ms/step - loss: 3.3700 - accuracy: 0.5994 - val_loss: 5.0025 - val_accuracy: 0.5867\n",
            "Epoch 55/200\n",
            "2/2 [==============================] - 1s 406ms/step - loss: 3.3601 - accuracy: 0.5996 - val_loss: 5.0373 - val_accuracy: 0.5872\n",
            "Epoch 56/200\n",
            "2/2 [==============================] - 1s 408ms/step - loss: 3.3490 - accuracy: 0.5993 - val_loss: 5.0251 - val_accuracy: 0.5869\n",
            "Epoch 57/200\n",
            "2/2 [==============================] - 1s 396ms/step - loss: 3.3402 - accuracy: 0.5996 - val_loss: 5.0619 - val_accuracy: 0.5872\n",
            "Epoch 58/200\n",
            "2/2 [==============================] - 1s 394ms/step - loss: 3.3271 - accuracy: 0.5998 - val_loss: 5.0560 - val_accuracy: 0.5871\n",
            "Epoch 59/200\n",
            "2/2 [==============================] - 1s 410ms/step - loss: 3.3151 - accuracy: 0.5996 - val_loss: 5.0866 - val_accuracy: 0.5873\n",
            "Epoch 60/200\n",
            "2/2 [==============================] - 1s 400ms/step - loss: 3.3021 - accuracy: 0.5991 - val_loss: 5.0873 - val_accuracy: 0.5872\n",
            "Epoch 61/200\n",
            "2/2 [==============================] - 1s 396ms/step - loss: 3.2902 - accuracy: 0.5995 - val_loss: 5.1100 - val_accuracy: 0.5870\n",
            "Epoch 62/200\n",
            "2/2 [==============================] - 1s 404ms/step - loss: 3.2756 - accuracy: 0.6001 - val_loss: 5.1207 - val_accuracy: 0.5869\n",
            "Epoch 63/200\n",
            "2/2 [==============================] - 1s 403ms/step - loss: 3.2636 - accuracy: 0.5997 - val_loss: 5.1363 - val_accuracy: 0.5865\n",
            "Epoch 64/200\n",
            "2/2 [==============================] - 1s 412ms/step - loss: 3.2527 - accuracy: 0.6002 - val_loss: 5.1641 - val_accuracy: 0.5870\n",
            "Epoch 65/200\n",
            "2/2 [==============================] - 1s 399ms/step - loss: 3.2368 - accuracy: 0.5998 - val_loss: 5.1708 - val_accuracy: 0.5865\n",
            "Epoch 66/200\n",
            "2/2 [==============================] - 1s 419ms/step - loss: 3.2273 - accuracy: 0.6000 - val_loss: 5.2078 - val_accuracy: 0.5869\n",
            "Epoch 67/200\n",
            "2/2 [==============================] - 1s 420ms/step - loss: 3.2138 - accuracy: 0.6002 - val_loss: 5.2191 - val_accuracy: 0.5870\n",
            "Epoch 68/200\n",
            "2/2 [==============================] - 1s 405ms/step - loss: 3.2015 - accuracy: 0.6003 - val_loss: 5.2359 - val_accuracy: 0.5867\n",
            "Epoch 69/200\n",
            "2/2 [==============================] - 1s 400ms/step - loss: 3.1873 - accuracy: 0.6005 - val_loss: 5.2600 - val_accuracy: 0.5866\n",
            "Epoch 70/200\n",
            "2/2 [==============================] - 1s 404ms/step - loss: 3.1725 - accuracy: 0.6006 - val_loss: 5.2724 - val_accuracy: 0.5861\n",
            "Epoch 71/200\n",
            "2/2 [==============================] - 1s 405ms/step - loss: 3.1580 - accuracy: 0.6008 - val_loss: 5.2840 - val_accuracy: 0.5856\n",
            "Epoch 72/200\n",
            "2/2 [==============================] - 1s 412ms/step - loss: 3.1460 - accuracy: 0.6009 - val_loss: 5.3074 - val_accuracy: 0.5866\n",
            "Epoch 73/200\n",
            "2/2 [==============================] - 1s 411ms/step - loss: 3.1315 - accuracy: 0.6007 - val_loss: 5.3132 - val_accuracy: 0.5854\n",
            "Epoch 74/200\n",
            "2/2 [==============================] - 1s 407ms/step - loss: 3.1182 - accuracy: 0.6006 - val_loss: 5.3187 - val_accuracy: 0.5845\n",
            "Epoch 75/200\n",
            "2/2 [==============================] - 1s 400ms/step - loss: 3.1078 - accuracy: 0.6008 - val_loss: 5.3321 - val_accuracy: 0.5848\n",
            "Epoch 76/200\n",
            "2/2 [==============================] - 1s 400ms/step - loss: 3.0874 - accuracy: 0.6009 - val_loss: 5.3516 - val_accuracy: 0.5858\n",
            "Epoch 77/200\n",
            "2/2 [==============================] - 1s 401ms/step - loss: 3.0789 - accuracy: 0.6010 - val_loss: 5.3662 - val_accuracy: 0.5858\n",
            "Epoch 78/200\n",
            "2/2 [==============================] - 1s 409ms/step - loss: 3.0662 - accuracy: 0.6002 - val_loss: 5.3659 - val_accuracy: 0.5846\n",
            "Epoch 79/200\n",
            "2/2 [==============================] - 1s 401ms/step - loss: 3.0488 - accuracy: 0.6007 - val_loss: 5.3768 - val_accuracy: 0.5838\n",
            "Epoch 80/200\n",
            "2/2 [==============================] - 1s 398ms/step - loss: 3.0407 - accuracy: 0.6007 - val_loss: 5.3888 - val_accuracy: 0.5841\n",
            "Epoch 81/200\n",
            "2/2 [==============================] - 1s 399ms/step - loss: 3.0208 - accuracy: 0.6016 - val_loss: 5.4009 - val_accuracy: 0.5852\n",
            "Epoch 82/200\n",
            "2/2 [==============================] - 1s 394ms/step - loss: 3.0106 - accuracy: 0.6013 - val_loss: 5.4166 - val_accuracy: 0.5860\n",
            "Epoch 83/200\n",
            "2/2 [==============================] - 1s 421ms/step - loss: 2.9975 - accuracy: 0.6010 - val_loss: 5.4183 - val_accuracy: 0.5846\n",
            "Epoch 84/200\n",
            "2/2 [==============================] - 1s 408ms/step - loss: 2.9810 - accuracy: 0.6008 - val_loss: 5.4175 - val_accuracy: 0.5818\n",
            "Epoch 85/200\n",
            "2/2 [==============================] - 1s 400ms/step - loss: 2.9823 - accuracy: 0.6014 - val_loss: 5.4341 - val_accuracy: 0.5840\n",
            "Epoch 86/200\n",
            "2/2 [==============================] - 1s 413ms/step - loss: 2.9565 - accuracy: 0.6009 - val_loss: 5.4436 - val_accuracy: 0.5849\n",
            "Epoch 87/200\n",
            "2/2 [==============================] - 1s 401ms/step - loss: 2.9532 - accuracy: 0.6009 - val_loss: 5.4440 - val_accuracy: 0.5847\n",
            "Epoch 88/200\n",
            "2/2 [==============================] - 1s 396ms/step - loss: 2.9376 - accuracy: 0.6004 - val_loss: 5.4609 - val_accuracy: 0.5825\n",
            "Epoch 89/200\n",
            "2/2 [==============================] - 1s 399ms/step - loss: 2.9390 - accuracy: 0.6007 - val_loss: 5.4508 - val_accuracy: 0.5817\n",
            "Epoch 90/200\n",
            "2/2 [==============================] - 1s 416ms/step - loss: 2.9214 - accuracy: 0.6013 - val_loss: 5.4868 - val_accuracy: 0.5860\n",
            "Epoch 91/200\n",
            "2/2 [==============================] - 1s 416ms/step - loss: 2.9194 - accuracy: 0.6010 - val_loss: 5.4629 - val_accuracy: 0.5832\n",
            "Epoch 92/200\n",
            "2/2 [==============================] - 1s 409ms/step - loss: 2.8991 - accuracy: 0.6012 - val_loss: 5.4820 - val_accuracy: 0.5833\n",
            "Epoch 93/200\n",
            "2/2 [==============================] - 1s 411ms/step - loss: 2.8913 - accuracy: 0.6007 - val_loss: 5.4775 - val_accuracy: 0.5845\n",
            "Epoch 94/200\n",
            "2/2 [==============================] - 1s 404ms/step - loss: 2.8838 - accuracy: 0.6008 - val_loss: 5.4979 - val_accuracy: 0.5823\n",
            "Epoch 95/200\n",
            "2/2 [==============================] - 1s 400ms/step - loss: 2.8965 - accuracy: 0.6008 - val_loss: 5.4793 - val_accuracy: 0.5830\n",
            "Epoch 96/200\n",
            "2/2 [==============================] - 1s 406ms/step - loss: 2.8812 - accuracy: 0.6007 - val_loss: 5.5076 - val_accuracy: 0.5842\n",
            "Epoch 97/200\n",
            "2/2 [==============================] - 1s 403ms/step - loss: 2.8624 - accuracy: 0.6009 - val_loss: 5.4882 - val_accuracy: 0.5817\n",
            "Epoch 98/200\n",
            "2/2 [==============================] - 1s 409ms/step - loss: 2.8457 - accuracy: 0.6014 - val_loss: 5.4987 - val_accuracy: 0.5853\n",
            "Epoch 99/200\n",
            "2/2 [==============================] - 1s 413ms/step - loss: 2.8280 - accuracy: 0.6017 - val_loss: 5.5091 - val_accuracy: 0.5808\n",
            "Epoch 100/200\n",
            "2/2 [==============================] - 1s 410ms/step - loss: 2.8316 - accuracy: 0.6019 - val_loss: 5.5015 - val_accuracy: 0.5852\n",
            "Epoch 101/200\n",
            "2/2 [==============================] - 1s 404ms/step - loss: 2.8198 - accuracy: 0.6015 - val_loss: 5.5156 - val_accuracy: 0.5827\n",
            "Epoch 102/200\n",
            "2/2 [==============================] - 1s 401ms/step - loss: 2.8119 - accuracy: 0.6007 - val_loss: 5.5102 - val_accuracy: 0.5839\n",
            "Epoch 103/200\n",
            "2/2 [==============================] - 1s 413ms/step - loss: 2.7989 - accuracy: 0.6010 - val_loss: 5.5126 - val_accuracy: 0.5826\n",
            "Epoch 104/200\n",
            "2/2 [==============================] - 1s 400ms/step - loss: 2.7810 - accuracy: 0.6013 - val_loss: 5.5239 - val_accuracy: 0.5835\n",
            "Epoch 105/200\n",
            "2/2 [==============================] - 1s 401ms/step - loss: 2.7789 - accuracy: 0.6017 - val_loss: 5.5158 - val_accuracy: 0.5832\n",
            "Epoch 106/200\n",
            "2/2 [==============================] - 1s 415ms/step - loss: 2.7692 - accuracy: 0.6013 - val_loss: 5.5298 - val_accuracy: 0.5823\n",
            "Epoch 107/200\n",
            "2/2 [==============================] - 1s 401ms/step - loss: 2.7563 - accuracy: 0.6015 - val_loss: 5.5310 - val_accuracy: 0.5845\n",
            "Epoch 108/200\n",
            "2/2 [==============================] - 1s 398ms/step - loss: 2.7468 - accuracy: 0.6008 - val_loss: 5.5370 - val_accuracy: 0.5809\n",
            "Epoch 109/200\n",
            "2/2 [==============================] - 1s 398ms/step - loss: 2.7395 - accuracy: 0.6009 - val_loss: 5.5459 - val_accuracy: 0.5843\n",
            "Epoch 110/200\n",
            "2/2 [==============================] - 1s 400ms/step - loss: 2.7264 - accuracy: 0.6014 - val_loss: 5.5481 - val_accuracy: 0.5805\n",
            "Epoch 111/200\n",
            "2/2 [==============================] - 1s 415ms/step - loss: 2.7181 - accuracy: 0.6019 - val_loss: 5.5567 - val_accuracy: 0.5829\n",
            "Epoch 112/200\n",
            "2/2 [==============================] - 1s 399ms/step - loss: 2.7100 - accuracy: 0.6014 - val_loss: 5.5553 - val_accuracy: 0.5825\n",
            "Epoch 113/200\n",
            "2/2 [==============================] - 1s 395ms/step - loss: 2.7029 - accuracy: 0.6013 - val_loss: 5.5653 - val_accuracy: 0.5818\n",
            "Epoch 114/200\n",
            "2/2 [==============================] - 1s 405ms/step - loss: 2.6923 - accuracy: 0.6011 - val_loss: 5.5679 - val_accuracy: 0.5831\n",
            "Epoch 115/200\n",
            "2/2 [==============================] - 1s 407ms/step - loss: 2.6835 - accuracy: 0.6015 - val_loss: 5.5770 - val_accuracy: 0.5794\n",
            "Epoch 116/200\n",
            "2/2 [==============================] - 1s 406ms/step - loss: 2.6773 - accuracy: 0.6016 - val_loss: 5.5787 - val_accuracy: 0.5819\n",
            "Epoch 117/200\n",
            "2/2 [==============================] - 1s 418ms/step - loss: 2.6665 - accuracy: 0.6017 - val_loss: 5.5801 - val_accuracy: 0.5814\n",
            "Epoch 118/200\n",
            "2/2 [==============================] - 1s 408ms/step - loss: 2.6613 - accuracy: 0.6015 - val_loss: 5.5940 - val_accuracy: 0.5807\n",
            "Epoch 119/200\n",
            "2/2 [==============================] - 1s 407ms/step - loss: 2.6506 - accuracy: 0.6017 - val_loss: 5.5926 - val_accuracy: 0.5817\n",
            "Epoch 120/200\n",
            "2/2 [==============================] - 1s 403ms/step - loss: 2.6405 - accuracy: 0.6020 - val_loss: 5.6015 - val_accuracy: 0.5820\n",
            "Epoch 121/200\n",
            "2/2 [==============================] - 1s 404ms/step - loss: 2.6291 - accuracy: 0.6015 - val_loss: 5.6083 - val_accuracy: 0.5786\n",
            "Epoch 122/200\n",
            "2/2 [==============================] - 1s 406ms/step - loss: 2.6240 - accuracy: 0.6017 - val_loss: 5.6118 - val_accuracy: 0.5819\n",
            "Epoch 123/200\n",
            "2/2 [==============================] - 1s 403ms/step - loss: 2.6168 - accuracy: 0.6019 - val_loss: 5.6113 - val_accuracy: 0.5800\n",
            "Epoch 124/200\n",
            "2/2 [==============================] - 1s 408ms/step - loss: 2.6015 - accuracy: 0.6022 - val_loss: 5.6290 - val_accuracy: 0.5791\n",
            "Epoch 125/200\n",
            "2/2 [==============================] - 1s 414ms/step - loss: 2.6073 - accuracy: 0.6018 - val_loss: 5.6281 - val_accuracy: 0.5780\n",
            "Epoch 126/200\n",
            "2/2 [==============================] - 1s 406ms/step - loss: 2.5928 - accuracy: 0.6018 - val_loss: 5.6322 - val_accuracy: 0.5832\n",
            "Epoch 127/200\n",
            "2/2 [==============================] - 1s 401ms/step - loss: 2.5969 - accuracy: 0.6019 - val_loss: 5.6360 - val_accuracy: 0.5771\n",
            "Epoch 128/200\n",
            "2/2 [==============================] - 1s 406ms/step - loss: 2.5805 - accuracy: 0.6022 - val_loss: 5.6481 - val_accuracy: 0.5791\n",
            "Epoch 129/200\n",
            "2/2 [==============================] - 1s 399ms/step - loss: 2.5654 - accuracy: 0.6022 - val_loss: 5.6347 - val_accuracy: 0.5797\n",
            "Epoch 130/200\n",
            "2/2 [==============================] - 1s 415ms/step - loss: 2.5532 - accuracy: 0.6020 - val_loss: 5.6479 - val_accuracy: 0.5807\n",
            "Epoch 131/200\n",
            "2/2 [==============================] - 1s 399ms/step - loss: 2.5367 - accuracy: 0.6019 - val_loss: 5.6534 - val_accuracy: 0.5768\n",
            "Epoch 132/200\n",
            "2/2 [==============================] - 1s 400ms/step - loss: 2.5231 - accuracy: 0.6030 - val_loss: 5.6519 - val_accuracy: 0.5811\n",
            "Epoch 133/200\n",
            "2/2 [==============================] - 1s 394ms/step - loss: 2.5095 - accuracy: 0.6025 - val_loss: 5.6481 - val_accuracy: 0.5781\n",
            "Epoch 134/200\n",
            "2/2 [==============================] - 1s 397ms/step - loss: 2.4955 - accuracy: 0.6034 - val_loss: 5.6681 - val_accuracy: 0.5788\n",
            "Epoch 135/200\n",
            "2/2 [==============================] - 1s 410ms/step - loss: 2.5045 - accuracy: 0.6027 - val_loss: 5.6710 - val_accuracy: 0.5758\n",
            "Epoch 136/200\n",
            "2/2 [==============================] - 1s 403ms/step - loss: 2.4889 - accuracy: 0.6036 - val_loss: 5.6699 - val_accuracy: 0.5831\n",
            "Epoch 137/200\n",
            "2/2 [==============================] - 1s 399ms/step - loss: 2.4959 - accuracy: 0.6030 - val_loss: 5.6739 - val_accuracy: 0.5755\n",
            "Epoch 138/200\n",
            "2/2 [==============================] - 1s 400ms/step - loss: 2.4728 - accuracy: 0.6035 - val_loss: 5.6771 - val_accuracy: 0.5790\n",
            "Epoch 139/200\n",
            "2/2 [==============================] - 1s 402ms/step - loss: 2.4381 - accuracy: 0.6029 - val_loss: 5.6634 - val_accuracy: 0.5808\n",
            "Epoch 140/200\n",
            "2/2 [==============================] - 1s 410ms/step - loss: 2.4391 - accuracy: 0.6033 - val_loss: 5.6852 - val_accuracy: 0.5768\n",
            "Epoch 141/200\n",
            "2/2 [==============================] - 1s 415ms/step - loss: 2.4345 - accuracy: 0.6034 - val_loss: 5.6825 - val_accuracy: 0.5788\n",
            "Epoch 142/200\n",
            "2/2 [==============================] - 1s 412ms/step - loss: 2.4097 - accuracy: 0.6038 - val_loss: 5.6686 - val_accuracy: 0.5817\n",
            "Epoch 143/200\n",
            "2/2 [==============================] - 1s 399ms/step - loss: 2.4207 - accuracy: 0.6040 - val_loss: 5.7306 - val_accuracy: 0.5725\n",
            "Epoch 144/200\n",
            "2/2 [==============================] - 1s 404ms/step - loss: 2.4559 - accuracy: 0.6004 - val_loss: 5.6679 - val_accuracy: 0.5818\n",
            "Epoch 145/200\n",
            "2/2 [==============================] - 1s 409ms/step - loss: 2.4055 - accuracy: 0.6040 - val_loss: 5.6934 - val_accuracy: 0.5771\n",
            "Epoch 146/200\n",
            "2/2 [==============================] - 1s 403ms/step - loss: 2.3705 - accuracy: 0.6045 - val_loss: 5.6778 - val_accuracy: 0.5798\n",
            "Epoch 147/200\n",
            "2/2 [==============================] - 1s 409ms/step - loss: 2.3625 - accuracy: 0.6046 - val_loss: 5.6831 - val_accuracy: 0.5787\n",
            "Epoch 148/200\n",
            "2/2 [==============================] - 1s 410ms/step - loss: 2.3461 - accuracy: 0.6057 - val_loss: 5.6976 - val_accuracy: 0.5785\n",
            "Epoch 149/200\n",
            "2/2 [==============================] - 1s 408ms/step - loss: 2.3282 - accuracy: 0.6061 - val_loss: 5.6887 - val_accuracy: 0.5795\n",
            "Epoch 150/200\n",
            "2/2 [==============================] - 1s 409ms/step - loss: 2.3126 - accuracy: 0.6064 - val_loss: 5.7011 - val_accuracy: 0.5782\n",
            "Epoch 151/200\n",
            "2/2 [==============================] - 1s 406ms/step - loss: 2.3008 - accuracy: 0.6071 - val_loss: 5.6976 - val_accuracy: 0.5794\n",
            "Epoch 152/200\n",
            "2/2 [==============================] - 1s 396ms/step - loss: 2.2839 - accuracy: 0.6070 - val_loss: 5.7164 - val_accuracy: 0.5760\n",
            "Epoch 153/200\n",
            "2/2 [==============================] - 1s 404ms/step - loss: 2.2727 - accuracy: 0.6072 - val_loss: 5.7076 - val_accuracy: 0.5792\n",
            "Epoch 154/200\n",
            "2/2 [==============================] - 1s 402ms/step - loss: 2.2598 - accuracy: 0.6082 - val_loss: 5.7258 - val_accuracy: 0.5748\n",
            "Epoch 155/200\n",
            "2/2 [==============================] - 1s 413ms/step - loss: 2.2524 - accuracy: 0.6090 - val_loss: 5.7147 - val_accuracy: 0.5796\n",
            "Epoch 156/200\n",
            "2/2 [==============================] - 1s 412ms/step - loss: 2.2403 - accuracy: 0.6080 - val_loss: 5.7386 - val_accuracy: 0.5758\n",
            "Epoch 157/200\n",
            "2/2 [==============================] - 1s 406ms/step - loss: 2.2347 - accuracy: 0.6083 - val_loss: 5.7210 - val_accuracy: 0.5792\n",
            "Epoch 158/200\n",
            "2/2 [==============================] - 1s 399ms/step - loss: 2.2229 - accuracy: 0.6095 - val_loss: 5.7445 - val_accuracy: 0.5764\n",
            "Epoch 159/200\n",
            "2/2 [==============================] - 1s 403ms/step - loss: 2.2075 - accuracy: 0.6095 - val_loss: 5.7368 - val_accuracy: 0.5765\n",
            "Epoch 160/200\n",
            "2/2 [==============================] - 1s 401ms/step - loss: 2.2026 - accuracy: 0.6100 - val_loss: 5.7482 - val_accuracy: 0.5776\n",
            "Epoch 161/200\n",
            "2/2 [==============================] - 1s 404ms/step - loss: 2.1912 - accuracy: 0.6108 - val_loss: 5.7563 - val_accuracy: 0.5739\n",
            "Epoch 162/200\n",
            "2/2 [==============================] - 1s 415ms/step - loss: 2.1809 - accuracy: 0.6106 - val_loss: 5.7490 - val_accuracy: 0.5782\n",
            "Epoch 163/200\n",
            "2/2 [==============================] - 1s 415ms/step - loss: 2.1663 - accuracy: 0.6108 - val_loss: 5.7702 - val_accuracy: 0.5737\n",
            "Epoch 164/200\n",
            "2/2 [==============================] - 1s 411ms/step - loss: 2.1640 - accuracy: 0.6104 - val_loss: 5.7530 - val_accuracy: 0.5783\n",
            "Epoch 165/200\n",
            "2/2 [==============================] - 1s 399ms/step - loss: 2.1489 - accuracy: 0.6122 - val_loss: 5.7728 - val_accuracy: 0.5747\n",
            "Epoch 166/200\n",
            "2/2 [==============================] - 1s 409ms/step - loss: 2.1409 - accuracy: 0.6112 - val_loss: 5.7713 - val_accuracy: 0.5760\n",
            "Epoch 167/200\n",
            "2/2 [==============================] - 1s 399ms/step - loss: 2.1349 - accuracy: 0.6119 - val_loss: 5.7759 - val_accuracy: 0.5757\n",
            "Epoch 168/200\n",
            "2/2 [==============================] - 1s 401ms/step - loss: 2.1197 - accuracy: 0.6138 - val_loss: 5.7856 - val_accuracy: 0.5751\n",
            "Epoch 169/200\n",
            "2/2 [==============================] - 1s 415ms/step - loss: 2.1108 - accuracy: 0.6126 - val_loss: 5.7869 - val_accuracy: 0.5754\n",
            "Epoch 170/200\n",
            "2/2 [==============================] - 1s 408ms/step - loss: 2.1004 - accuracy: 0.6148 - val_loss: 5.7901 - val_accuracy: 0.5751\n",
            "Epoch 171/200\n",
            "2/2 [==============================] - 1s 415ms/step - loss: 2.0898 - accuracy: 0.6145 - val_loss: 5.8008 - val_accuracy: 0.5732\n",
            "Epoch 172/200\n",
            "2/2 [==============================] - 1s 402ms/step - loss: 2.0911 - accuracy: 0.6148 - val_loss: 5.7959 - val_accuracy: 0.5748\n",
            "Epoch 173/200\n",
            "2/2 [==============================] - 1s 421ms/step - loss: 2.0796 - accuracy: 0.6156 - val_loss: 5.8043 - val_accuracy: 0.5742\n",
            "Epoch 174/200\n",
            "2/2 [==============================] - 1s 405ms/step - loss: 2.0672 - accuracy: 0.6156 - val_loss: 5.8109 - val_accuracy: 0.5742\n",
            "Epoch 175/200\n",
            "2/2 [==============================] - 1s 406ms/step - loss: 2.0625 - accuracy: 0.6165 - val_loss: 5.8085 - val_accuracy: 0.5744\n",
            "Epoch 176/200\n",
            "2/2 [==============================] - 1s 411ms/step - loss: 2.0564 - accuracy: 0.6167 - val_loss: 5.8171 - val_accuracy: 0.5747\n",
            "Epoch 177/200\n",
            "2/2 [==============================] - 1s 401ms/step - loss: 2.0487 - accuracy: 0.6157 - val_loss: 5.8278 - val_accuracy: 0.5730\n",
            "Epoch 178/200\n",
            "2/2 [==============================] - 1s 421ms/step - loss: 2.0372 - accuracy: 0.6177 - val_loss: 5.8263 - val_accuracy: 0.5739\n",
            "Epoch 179/200\n",
            "2/2 [==============================] - 1s 411ms/step - loss: 2.0365 - accuracy: 0.6173 - val_loss: 5.8287 - val_accuracy: 0.5735\n",
            "Epoch 180/200\n",
            "2/2 [==============================] - 1s 403ms/step - loss: 2.0315 - accuracy: 0.6177 - val_loss: 5.8310 - val_accuracy: 0.5744\n",
            "Epoch 181/200\n",
            "2/2 [==============================] - 1s 401ms/step - loss: 2.0261 - accuracy: 0.6180 - val_loss: 5.8496 - val_accuracy: 0.5719\n",
            "Epoch 182/200\n",
            "2/2 [==============================] - 1s 402ms/step - loss: 2.0210 - accuracy: 0.6180 - val_loss: 5.8408 - val_accuracy: 0.5739\n",
            "Epoch 183/200\n",
            "2/2 [==============================] - 1s 406ms/step - loss: 2.0053 - accuracy: 0.6186 - val_loss: 5.8416 - val_accuracy: 0.5733\n",
            "Epoch 184/200\n",
            "2/2 [==============================] - 1s 405ms/step - loss: 2.0001 - accuracy: 0.6199 - val_loss: 5.8415 - val_accuracy: 0.5738\n",
            "Epoch 185/200\n",
            "2/2 [==============================] - 1s 407ms/step - loss: 1.9964 - accuracy: 0.6184 - val_loss: 5.8488 - val_accuracy: 0.5732\n",
            "Epoch 186/200\n",
            "2/2 [==============================] - 1s 414ms/step - loss: 1.9928 - accuracy: 0.6202 - val_loss: 5.8527 - val_accuracy: 0.5736\n",
            "Epoch 187/200\n",
            "2/2 [==============================] - 1s 403ms/step - loss: 1.9844 - accuracy: 0.6206 - val_loss: 5.8596 - val_accuracy: 0.5742\n",
            "Epoch 188/200\n",
            "2/2 [==============================] - 1s 400ms/step - loss: 1.9787 - accuracy: 0.6207 - val_loss: 5.8531 - val_accuracy: 0.5741\n",
            "Epoch 189/200\n",
            "2/2 [==============================] - 1s 420ms/step - loss: 1.9732 - accuracy: 0.6207 - val_loss: 5.8610 - val_accuracy: 0.5735\n",
            "Epoch 190/200\n",
            "2/2 [==============================] - 1s 409ms/step - loss: 1.9702 - accuracy: 0.6211 - val_loss: 5.8645 - val_accuracy: 0.5723\n",
            "Epoch 191/200\n",
            "2/2 [==============================] - 1s 416ms/step - loss: 1.9626 - accuracy: 0.6215 - val_loss: 5.8621 - val_accuracy: 0.5736\n",
            "Epoch 192/200\n",
            "2/2 [==============================] - 1s 404ms/step - loss: 1.9533 - accuracy: 0.6223 - val_loss: 5.8616 - val_accuracy: 0.5739\n",
            "Epoch 193/200\n",
            "2/2 [==============================] - 1s 403ms/step - loss: 1.9522 - accuracy: 0.6207 - val_loss: 5.8591 - val_accuracy: 0.5746\n",
            "Epoch 194/200\n",
            "2/2 [==============================] - 1s 407ms/step - loss: 1.9594 - accuracy: 0.6206 - val_loss: 5.8521 - val_accuracy: 0.5766\n",
            "Epoch 195/200\n",
            "2/2 [==============================] - 1s 400ms/step - loss: 1.9983 - accuracy: 0.6191 - val_loss: 5.8487 - val_accuracy: 0.5769\n",
            "Epoch 196/200\n",
            "2/2 [==============================] - 1s 403ms/step - loss: 2.0086 - accuracy: 0.6186 - val_loss: 5.8961 - val_accuracy: 0.5700\n",
            "Epoch 197/200\n",
            "2/2 [==============================] - 1s 399ms/step - loss: 1.9912 - accuracy: 0.6187 - val_loss: 5.9165 - val_accuracy: 0.5665\n",
            "Epoch 198/200\n",
            "2/2 [==============================] - 1s 402ms/step - loss: 1.9860 - accuracy: 0.6198 - val_loss: 5.8505 - val_accuracy: 0.5764\n",
            "Epoch 199/200\n",
            "2/2 [==============================] - 1s 401ms/step - loss: 2.0028 - accuracy: 0.6192 - val_loss: 5.8641 - val_accuracy: 0.5734\n",
            "Epoch 200/200\n",
            "2/2 [==============================] - 1s 401ms/step - loss: 1.9596 - accuracy: 0.6205 - val_loss: 5.8934 - val_accuracy: 0.5701\n",
            "ukusakazwa utamatisi owomiswe elangeni solo wesithombe 70g emafutheni ohluzwe okusamanzana okusamanzana ezimnandi uqotshwa elifiselekayo <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD>\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "k8Eev7EgZ0SH"
      },
      "source": [
        "## Model 8 - Bidirectional RNN with Word Embedding, Encoder-Decoder and Byte-Pair Encoding"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "liaNFKxLZ0uL",
        "outputId": "42202590-e8dd-40df-f2f3-4542859cd7f0"
      },
      "source": [
        "tmp_x_bpe = pad(preproc_english_sentences_bpe, preproc_zulu_sentences_bpe.shape[1])\n",
        "tmp_x_bpe = tmp_x_bpe.reshape((-1, preproc_zulu_sentences_bpe.shape[-2]))\n",
        "tmp_x_test_bpe = pad(preproc_english_sentences_test_bpe, max_zulu_sequence_length_bpe)\n",
        "tmp_x_test_bpe = tmp_x_test_bpe.reshape((-1, preproc_zulu_sentences_bpe.shape[-2], 1))\n",
        "\n",
        "\n",
        "# TODO: Train and Print prediction(s)\n",
        "encdec_rnn_model_bpe = encdec_model(\n",
        "    tmp_x_bpe.shape,\n",
        "    preproc_zulu_sentences_bpe.shape[1],\n",
        "    english_vocab_size_bpe+1,\n",
        "    zulu_vocab_size_bpe+1)\n",
        "\n",
        "encdec_rnn_model_bpe.summary()\n",
        "\n",
        "model8 = encdec_rnn_model_bpe.fit(tmp_x_bpe, preproc_zulu_sentences_bpe, batch_size=1024, epochs=NUM_EPOCHS, validation_split=0.2)\n",
        "\n",
        "print(logits_to_text_bpe(encdec_rnn_model_bpe.predict(tmp_x_bpe[:1])[0], tokenizer_zu))"
      ],
      "execution_count": 30,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential_7\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " embedding_5 (Embedding)     (None, 58, 256)           2560256   \n",
            "                                                                 \n",
            " bidirectional_4 (Bidirectio  (None, 256)              296448    \n",
            " nal)                                                            \n",
            "                                                                 \n",
            " repeat_vector_1 (RepeatVect  (None, 58, 256)          0         \n",
            " or)                                                             \n",
            "                                                                 \n",
            " bidirectional_5 (Bidirectio  (None, 58, 256)          296448    \n",
            " nal)                                                            \n",
            "                                                                 \n",
            " time_distributed_14 (TimeDi  (None, 58, 1024)         263168    \n",
            " stributed)                                                      \n",
            "                                                                 \n",
            " dropout_7 (Dropout)         (None, 58, 1024)          0         \n",
            "                                                                 \n",
            " time_distributed_15 (TimeDi  (None, 58, 10001)        10251025  \n",
            " stributed)                                                      \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 13,667,345\n",
            "Trainable params: 13,667,345\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n",
            "Epoch 1/200\n",
            "2/2 [==============================] - 10s 2s/step - loss: 9.1953 - accuracy: 0.2547 - val_loss: 9.1041 - val_accuracy: 0.6792\n",
            "Epoch 2/200\n",
            "2/2 [==============================] - 2s 675ms/step - loss: 9.0569 - accuracy: 0.6803 - val_loss: 8.7559 - val_accuracy: 0.6792\n",
            "Epoch 3/200\n",
            "2/2 [==============================] - 2s 671ms/step - loss: 8.6031 - accuracy: 0.6803 - val_loss: 7.6799 - val_accuracy: 0.6792\n",
            "Epoch 4/200\n",
            "2/2 [==============================] - 2s 672ms/step - loss: 7.2374 - accuracy: 0.6803 - val_loss: 4.8081 - val_accuracy: 0.6792\n",
            "Epoch 5/200\n",
            "2/2 [==============================] - 2s 678ms/step - loss: 4.1632 - accuracy: 0.6803 - val_loss: 3.8506 - val_accuracy: 0.6792\n",
            "Epoch 6/200\n",
            "2/2 [==============================] - 2s 677ms/step - loss: 3.9498 - accuracy: 0.6803 - val_loss: 4.3919 - val_accuracy: 0.6792\n",
            "Epoch 7/200\n",
            "2/2 [==============================] - 2s 680ms/step - loss: 4.3471 - accuracy: 0.6803 - val_loss: 4.2702 - val_accuracy: 0.6792\n",
            "Epoch 8/200\n",
            "2/2 [==============================] - 2s 678ms/step - loss: 4.1668 - accuracy: 0.6803 - val_loss: 3.9698 - val_accuracy: 0.6792\n",
            "Epoch 9/200\n",
            "2/2 [==============================] - 2s 675ms/step - loss: 3.8970 - accuracy: 0.6803 - val_loss: 3.8340 - val_accuracy: 0.6792\n",
            "Epoch 10/200\n",
            "2/2 [==============================] - 2s 672ms/step - loss: 3.7506 - accuracy: 0.6802 - val_loss: 3.6826 - val_accuracy: 0.6792\n",
            "Epoch 11/200\n",
            "2/2 [==============================] - 2s 674ms/step - loss: 3.5695 - accuracy: 0.6803 - val_loss: 3.5430 - val_accuracy: 0.6792\n",
            "Epoch 12/200\n",
            "2/2 [==============================] - 2s 671ms/step - loss: 3.4067 - accuracy: 0.6803 - val_loss: 3.4408 - val_accuracy: 0.6792\n",
            "Epoch 13/200\n",
            "2/2 [==============================] - 2s 668ms/step - loss: 3.2972 - accuracy: 0.6803 - val_loss: 3.3736 - val_accuracy: 0.6792\n",
            "Epoch 14/200\n",
            "2/2 [==============================] - 2s 678ms/step - loss: 3.2138 - accuracy: 0.6803 - val_loss: 3.2969 - val_accuracy: 0.6792\n",
            "Epoch 15/200\n",
            "2/2 [==============================] - 2s 671ms/step - loss: 3.1282 - accuracy: 0.6803 - val_loss: 3.2155 - val_accuracy: 0.6792\n",
            "Epoch 16/200\n",
            "2/2 [==============================] - 2s 676ms/step - loss: 3.0507 - accuracy: 0.6803 - val_loss: 3.1436 - val_accuracy: 0.6792\n",
            "Epoch 17/200\n",
            "2/2 [==============================] - 2s 703ms/step - loss: 2.9885 - accuracy: 0.6803 - val_loss: 3.0721 - val_accuracy: 0.6792\n",
            "Epoch 18/200\n",
            "2/2 [==============================] - 2s 672ms/step - loss: 2.9229 - accuracy: 0.6803 - val_loss: 3.0141 - val_accuracy: 0.6792\n",
            "Epoch 19/200\n",
            "2/2 [==============================] - 2s 673ms/step - loss: 2.8723 - accuracy: 0.6803 - val_loss: 2.9779 - val_accuracy: 0.6792\n",
            "Epoch 20/200\n",
            "2/2 [==============================] - 2s 678ms/step - loss: 2.8382 - accuracy: 0.6803 - val_loss: 2.9467 - val_accuracy: 0.6792\n",
            "Epoch 21/200\n",
            "2/2 [==============================] - 2s 670ms/step - loss: 2.8121 - accuracy: 0.6803 - val_loss: 2.9187 - val_accuracy: 0.6792\n",
            "Epoch 22/200\n",
            "2/2 [==============================] - 2s 673ms/step - loss: 2.7929 - accuracy: 0.6803 - val_loss: 2.9009 - val_accuracy: 0.6792\n",
            "Epoch 23/200\n",
            "2/2 [==============================] - 2s 672ms/step - loss: 2.7824 - accuracy: 0.6803 - val_loss: 2.8887 - val_accuracy: 0.6792\n",
            "Epoch 24/200\n",
            "2/2 [==============================] - 2s 673ms/step - loss: 2.7672 - accuracy: 0.6803 - val_loss: 2.8818 - val_accuracy: 0.6792\n",
            "Epoch 25/200\n",
            "2/2 [==============================] - 2s 668ms/step - loss: 2.7574 - accuracy: 0.6803 - val_loss: 2.8784 - val_accuracy: 0.6792\n",
            "Epoch 26/200\n",
            "2/2 [==============================] - 2s 676ms/step - loss: 2.7487 - accuracy: 0.6803 - val_loss: 2.8710 - val_accuracy: 0.6792\n",
            "Epoch 27/200\n",
            "2/2 [==============================] - 2s 672ms/step - loss: 2.7376 - accuracy: 0.6803 - val_loss: 2.8620 - val_accuracy: 0.6792\n",
            "Epoch 28/200\n",
            "2/2 [==============================] - 2s 678ms/step - loss: 2.7243 - accuracy: 0.6803 - val_loss: 2.8547 - val_accuracy: 0.6792\n",
            "Epoch 29/200\n",
            "2/2 [==============================] - 2s 672ms/step - loss: 2.7110 - accuracy: 0.6803 - val_loss: 2.8492 - val_accuracy: 0.6792\n",
            "Epoch 30/200\n",
            "2/2 [==============================] - 2s 672ms/step - loss: 2.6942 - accuracy: 0.6803 - val_loss: 2.8467 - val_accuracy: 0.6792\n",
            "Epoch 31/200\n",
            "2/2 [==============================] - 2s 677ms/step - loss: 2.6787 - accuracy: 0.6803 - val_loss: 2.8460 - val_accuracy: 0.6792\n",
            "Epoch 32/200\n",
            "2/2 [==============================] - 2s 671ms/step - loss: 2.6663 - accuracy: 0.6803 - val_loss: 2.8469 - val_accuracy: 0.6792\n",
            "Epoch 33/200\n",
            "2/2 [==============================] - 2s 677ms/step - loss: 2.6547 - accuracy: 0.6803 - val_loss: 2.8511 - val_accuracy: 0.6792\n",
            "Epoch 34/200\n",
            "2/2 [==============================] - 2s 678ms/step - loss: 2.6394 - accuracy: 0.6803 - val_loss: 2.8581 - val_accuracy: 0.6792\n",
            "Epoch 35/200\n",
            "2/2 [==============================] - 2s 689ms/step - loss: 2.6230 - accuracy: 0.6804 - val_loss: 2.8581 - val_accuracy: 0.6792\n",
            "Epoch 36/200\n",
            "2/2 [==============================] - 2s 669ms/step - loss: 2.6069 - accuracy: 0.6806 - val_loss: 2.8569 - val_accuracy: 0.6792\n",
            "Epoch 37/200\n",
            "2/2 [==============================] - 2s 685ms/step - loss: 2.5931 - accuracy: 0.6810 - val_loss: 2.8712 - val_accuracy: 0.6795\n",
            "Epoch 38/200\n",
            "2/2 [==============================] - 2s 670ms/step - loss: 2.5818 - accuracy: 0.6818 - val_loss: 2.8818 - val_accuracy: 0.6806\n",
            "Epoch 39/200\n",
            "2/2 [==============================] - 2s 672ms/step - loss: 2.5680 - accuracy: 0.6824 - val_loss: 2.8759 - val_accuracy: 0.6818\n",
            "Epoch 40/200\n",
            "2/2 [==============================] - 2s 691ms/step - loss: 2.5513 - accuracy: 0.6836 - val_loss: 2.8876 - val_accuracy: 0.6828\n",
            "Epoch 41/200\n",
            "2/2 [==============================] - 2s 669ms/step - loss: 2.5340 - accuracy: 0.6846 - val_loss: 2.8965 - val_accuracy: 0.6840\n",
            "Epoch 42/200\n",
            "2/2 [==============================] - 2s 671ms/step - loss: 2.5140 - accuracy: 0.6860 - val_loss: 2.9188 - val_accuracy: 0.6849\n",
            "Epoch 43/200\n",
            "2/2 [==============================] - 2s 674ms/step - loss: 2.4963 - accuracy: 0.6871 - val_loss: 2.9258 - val_accuracy: 0.6854\n",
            "Epoch 44/200\n",
            "2/2 [==============================] - 2s 681ms/step - loss: 2.4791 - accuracy: 0.6885 - val_loss: 2.9457 - val_accuracy: 0.6847\n",
            "Epoch 45/200\n",
            "2/2 [==============================] - 2s 695ms/step - loss: 2.4643 - accuracy: 0.6895 - val_loss: 2.9734 - val_accuracy: 0.6836\n",
            "Epoch 46/200\n",
            "2/2 [==============================] - 2s 681ms/step - loss: 2.4501 - accuracy: 0.6902 - val_loss: 2.9937 - val_accuracy: 0.6821\n",
            "Epoch 47/200\n",
            "2/2 [==============================] - 2s 673ms/step - loss: 2.4443 - accuracy: 0.6908 - val_loss: 3.0049 - val_accuracy: 0.6794\n",
            "Epoch 48/200\n",
            "2/2 [==============================] - 2s 676ms/step - loss: 2.4494 - accuracy: 0.6912 - val_loss: 3.0242 - val_accuracy: 0.6818\n",
            "Epoch 49/200\n",
            "2/2 [==============================] - 2s 683ms/step - loss: 2.4626 - accuracy: 0.6887 - val_loss: 3.0968 - val_accuracy: 0.6847\n",
            "Epoch 50/200\n",
            "2/2 [==============================] - 2s 687ms/step - loss: 2.5297 - accuracy: 0.6891 - val_loss: 3.0074 - val_accuracy: 0.6825\n",
            "Epoch 51/200\n",
            "2/2 [==============================] - 2s 676ms/step - loss: 2.4687 - accuracy: 0.6902 - val_loss: 2.9667 - val_accuracy: 0.6754\n",
            "Epoch 52/200\n",
            "2/2 [==============================] - 2s 672ms/step - loss: 2.4683 - accuracy: 0.6896 - val_loss: 3.0802 - val_accuracy: 0.6854\n",
            "Epoch 53/200\n",
            "2/2 [==============================] - 2s 684ms/step - loss: 2.4825 - accuracy: 0.6888 - val_loss: 2.9544 - val_accuracy: 0.6834\n",
            "Epoch 54/200\n",
            "2/2 [==============================] - 2s 672ms/step - loss: 2.4371 - accuracy: 0.6911 - val_loss: 2.9342 - val_accuracy: 0.6786\n",
            "Epoch 55/200\n",
            "2/2 [==============================] - 2s 672ms/step - loss: 2.4403 - accuracy: 0.6914 - val_loss: 2.9863 - val_accuracy: 0.6849\n",
            "Epoch 56/200\n",
            "2/2 [==============================] - 2s 678ms/step - loss: 2.4367 - accuracy: 0.6897 - val_loss: 2.9948 - val_accuracy: 0.6852\n",
            "Epoch 57/200\n",
            "2/2 [==============================] - 2s 678ms/step - loss: 2.4183 - accuracy: 0.6905 - val_loss: 2.9620 - val_accuracy: 0.6813\n",
            "Epoch 58/200\n",
            "2/2 [==============================] - 2s 673ms/step - loss: 2.4239 - accuracy: 0.6919 - val_loss: 2.9816 - val_accuracy: 0.6843\n",
            "Epoch 59/200\n",
            "2/2 [==============================] - 2s 674ms/step - loss: 2.4046 - accuracy: 0.6906 - val_loss: 3.0045 - val_accuracy: 0.6849\n",
            "Epoch 60/200\n",
            "2/2 [==============================] - 2s 692ms/step - loss: 2.4057 - accuracy: 0.6902 - val_loss: 2.9735 - val_accuracy: 0.6820\n",
            "Epoch 61/200\n",
            "2/2 [==============================] - 2s 691ms/step - loss: 2.3987 - accuracy: 0.6916 - val_loss: 2.9900 - val_accuracy: 0.6826\n",
            "Epoch 62/200\n",
            "2/2 [==============================] - 2s 672ms/step - loss: 2.3889 - accuracy: 0.6913 - val_loss: 3.0405 - val_accuracy: 0.6847\n",
            "Epoch 63/200\n",
            "2/2 [==============================] - 2s 673ms/step - loss: 2.3900 - accuracy: 0.6906 - val_loss: 3.0200 - val_accuracy: 0.6822\n",
            "Epoch 64/200\n",
            "2/2 [==============================] - 2s 683ms/step - loss: 2.3804 - accuracy: 0.6919 - val_loss: 3.0247 - val_accuracy: 0.6814\n",
            "Epoch 65/200\n",
            "2/2 [==============================] - 2s 678ms/step - loss: 2.3744 - accuracy: 0.6916 - val_loss: 3.0567 - val_accuracy: 0.6831\n",
            "Epoch 66/200\n",
            "2/2 [==============================] - 2s 675ms/step - loss: 2.3713 - accuracy: 0.6911 - val_loss: 3.0532 - val_accuracy: 0.6812\n",
            "Epoch 67/200\n",
            "2/2 [==============================] - 2s 678ms/step - loss: 2.3656 - accuracy: 0.6922 - val_loss: 3.0673 - val_accuracy: 0.6813\n",
            "Epoch 68/200\n",
            "2/2 [==============================] - 2s 674ms/step - loss: 2.3581 - accuracy: 0.6915 - val_loss: 3.0844 - val_accuracy: 0.6816\n",
            "Epoch 69/200\n",
            "2/2 [==============================] - 2s 677ms/step - loss: 2.3537 - accuracy: 0.6914 - val_loss: 3.0793 - val_accuracy: 0.6777\n",
            "Epoch 70/200\n",
            "2/2 [==============================] - 2s 671ms/step - loss: 2.3501 - accuracy: 0.6921 - val_loss: 3.1057 - val_accuracy: 0.6808\n",
            "Epoch 71/200\n",
            "2/2 [==============================] - 2s 677ms/step - loss: 2.3442 - accuracy: 0.6915 - val_loss: 3.1091 - val_accuracy: 0.6787\n",
            "Epoch 72/200\n",
            "2/2 [==============================] - 2s 668ms/step - loss: 2.3392 - accuracy: 0.6924 - val_loss: 3.1210 - val_accuracy: 0.6790\n",
            "Epoch 73/200\n",
            "2/2 [==============================] - 2s 679ms/step - loss: 2.3334 - accuracy: 0.6918 - val_loss: 3.1308 - val_accuracy: 0.6782\n",
            "Epoch 74/200\n",
            "2/2 [==============================] - 2s 674ms/step - loss: 2.3298 - accuracy: 0.6921 - val_loss: 3.1415 - val_accuracy: 0.6775\n",
            "Epoch 75/200\n",
            "2/2 [==============================] - 2s 670ms/step - loss: 2.3246 - accuracy: 0.6920 - val_loss: 3.1573 - val_accuracy: 0.6781\n",
            "Epoch 76/200\n",
            "2/2 [==============================] - 2s 686ms/step - loss: 2.3199 - accuracy: 0.6920 - val_loss: 3.1635 - val_accuracy: 0.6762\n",
            "Epoch 77/200\n",
            "2/2 [==============================] - 2s 675ms/step - loss: 2.3154 - accuracy: 0.6920 - val_loss: 3.1780 - val_accuracy: 0.6770\n",
            "Epoch 78/200\n",
            "2/2 [==============================] - 2s 672ms/step - loss: 2.3117 - accuracy: 0.6924 - val_loss: 3.1878 - val_accuracy: 0.6762\n",
            "Epoch 79/200\n",
            "2/2 [==============================] - 2s 683ms/step - loss: 2.3075 - accuracy: 0.6922 - val_loss: 3.1967 - val_accuracy: 0.6757\n",
            "Epoch 80/200\n",
            "2/2 [==============================] - 2s 679ms/step - loss: 2.3028 - accuracy: 0.6929 - val_loss: 3.2100 - val_accuracy: 0.6764\n",
            "Epoch 81/200\n",
            "2/2 [==============================] - 2s 673ms/step - loss: 2.2989 - accuracy: 0.6920 - val_loss: 3.2179 - val_accuracy: 0.6741\n",
            "Epoch 82/200\n",
            "2/2 [==============================] - 2s 678ms/step - loss: 2.2961 - accuracy: 0.6931 - val_loss: 3.2304 - val_accuracy: 0.6758\n",
            "Epoch 83/200\n",
            "2/2 [==============================] - 2s 669ms/step - loss: 2.2906 - accuracy: 0.6927 - val_loss: 3.2407 - val_accuracy: 0.6750\n",
            "Epoch 84/200\n",
            "2/2 [==============================] - 2s 678ms/step - loss: 2.2857 - accuracy: 0.6924 - val_loss: 3.2496 - val_accuracy: 0.6734\n",
            "Epoch 85/200\n",
            "2/2 [==============================] - 2s 682ms/step - loss: 2.2823 - accuracy: 0.6937 - val_loss: 3.2602 - val_accuracy: 0.6744\n",
            "Epoch 86/200\n",
            "2/2 [==============================] - 2s 671ms/step - loss: 2.2781 - accuracy: 0.6936 - val_loss: 3.2761 - val_accuracy: 0.6755\n",
            "Epoch 87/200\n",
            "2/2 [==============================] - 2s 695ms/step - loss: 2.2739 - accuracy: 0.6933 - val_loss: 3.2790 - val_accuracy: 0.6731\n",
            "Epoch 88/200\n",
            "2/2 [==============================] - 2s 670ms/step - loss: 2.2700 - accuracy: 0.6938 - val_loss: 3.2909 - val_accuracy: 0.6726\n",
            "Epoch 89/200\n",
            "2/2 [==============================] - 2s 693ms/step - loss: 2.2665 - accuracy: 0.6945 - val_loss: 3.2995 - val_accuracy: 0.6724\n",
            "Epoch 90/200\n",
            "2/2 [==============================] - 2s 671ms/step - loss: 2.2622 - accuracy: 0.6945 - val_loss: 3.3087 - val_accuracy: 0.6722\n",
            "Epoch 91/200\n",
            "2/2 [==============================] - 2s 681ms/step - loss: 2.2567 - accuracy: 0.6943 - val_loss: 3.3184 - val_accuracy: 0.6717\n",
            "Epoch 92/200\n",
            "2/2 [==============================] - 2s 672ms/step - loss: 2.2558 - accuracy: 0.6944 - val_loss: 3.3270 - val_accuracy: 0.6693\n",
            "Epoch 93/200\n",
            "2/2 [==============================] - 2s 676ms/step - loss: 2.2663 - accuracy: 0.6952 - val_loss: 3.3402 - val_accuracy: 0.6657\n",
            "Epoch 94/200\n",
            "2/2 [==============================] - 2s 673ms/step - loss: 2.2957 - accuracy: 0.6938 - val_loss: 3.3281 - val_accuracy: 0.6690\n",
            "Epoch 95/200\n",
            "2/2 [==============================] - 2s 669ms/step - loss: 2.2708 - accuracy: 0.6937 - val_loss: 3.3936 - val_accuracy: 0.6820\n",
            "Epoch 96/200\n",
            "2/2 [==============================] - 2s 685ms/step - loss: 2.3403 - accuracy: 0.6920 - val_loss: 3.3907 - val_accuracy: 0.6474\n",
            "Epoch 97/200\n",
            "2/2 [==============================] - 2s 677ms/step - loss: 2.3899 - accuracy: 0.6730 - val_loss: 3.3751 - val_accuracy: 0.6808\n",
            "Epoch 98/200\n",
            "2/2 [==============================] - 2s 686ms/step - loss: 2.3522 - accuracy: 0.6916 - val_loss: 3.3646 - val_accuracy: 0.6525\n",
            "Epoch 99/200\n",
            "2/2 [==============================] - 2s 673ms/step - loss: 2.3594 - accuracy: 0.6765 - val_loss: 3.3564 - val_accuracy: 0.6810\n",
            "Epoch 100/200\n",
            "2/2 [==============================] - 2s 674ms/step - loss: 2.3426 - accuracy: 0.6911 - val_loss: 3.2828 - val_accuracy: 0.6711\n",
            "Epoch 101/200\n",
            "2/2 [==============================] - 2s 672ms/step - loss: 2.2818 - accuracy: 0.6914 - val_loss: 3.2919 - val_accuracy: 0.6640\n",
            "Epoch 102/200\n",
            "2/2 [==============================] - 2s 672ms/step - loss: 2.2703 - accuracy: 0.6923 - val_loss: 3.3421 - val_accuracy: 0.6821\n",
            "Epoch 103/200\n",
            "2/2 [==============================] - 2s 678ms/step - loss: 2.2944 - accuracy: 0.6912 - val_loss: 3.2810 - val_accuracy: 0.6714\n",
            "Epoch 104/200\n",
            "2/2 [==============================] - 2s 677ms/step - loss: 2.2578 - accuracy: 0.6944 - val_loss: 3.2783 - val_accuracy: 0.6680\n",
            "Epoch 105/200\n",
            "2/2 [==============================] - 2s 670ms/step - loss: 2.2470 - accuracy: 0.6945 - val_loss: 3.3038 - val_accuracy: 0.6794\n",
            "Epoch 106/200\n",
            "2/2 [==============================] - 2s 673ms/step - loss: 2.2550 - accuracy: 0.6918 - val_loss: 3.3014 - val_accuracy: 0.6743\n",
            "Epoch 107/200\n",
            "2/2 [==============================] - 2s 686ms/step - loss: 2.2361 - accuracy: 0.6936 - val_loss: 3.3033 - val_accuracy: 0.6704\n",
            "Epoch 108/200\n",
            "2/2 [==============================] - 2s 676ms/step - loss: 2.2306 - accuracy: 0.6936 - val_loss: 3.3061 - val_accuracy: 0.6767\n",
            "Epoch 109/200\n",
            "2/2 [==============================] - 2s 686ms/step - loss: 2.2335 - accuracy: 0.6920 - val_loss: 3.3034 - val_accuracy: 0.6730\n",
            "Epoch 110/200\n",
            "2/2 [==============================] - 2s 673ms/step - loss: 2.2178 - accuracy: 0.6946 - val_loss: 3.3172 - val_accuracy: 0.6705\n",
            "Epoch 111/200\n",
            "2/2 [==============================] - 2s 675ms/step - loss: 2.2157 - accuracy: 0.6943 - val_loss: 3.3300 - val_accuracy: 0.6775\n",
            "Epoch 112/200\n",
            "2/2 [==============================] - 2s 676ms/step - loss: 2.2143 - accuracy: 0.6926 - val_loss: 3.3274 - val_accuracy: 0.6733\n",
            "Epoch 113/200\n",
            "2/2 [==============================] - 2s 675ms/step - loss: 2.2043 - accuracy: 0.6943 - val_loss: 3.3336 - val_accuracy: 0.6701\n",
            "Epoch 114/200\n",
            "2/2 [==============================] - 2s 676ms/step - loss: 2.2012 - accuracy: 0.6949 - val_loss: 3.3452 - val_accuracy: 0.6759\n",
            "Epoch 115/200\n",
            "2/2 [==============================] - 2s 688ms/step - loss: 2.1984 - accuracy: 0.6928 - val_loss: 3.3537 - val_accuracy: 0.6731\n",
            "Epoch 116/200\n",
            "2/2 [==============================] - 2s 672ms/step - loss: 2.1930 - accuracy: 0.6943 - val_loss: 3.3597 - val_accuracy: 0.6716\n",
            "Epoch 117/200\n",
            "2/2 [==============================] - 2s 674ms/step - loss: 2.1880 - accuracy: 0.6939 - val_loss: 3.3663 - val_accuracy: 0.6739\n",
            "Epoch 118/200\n",
            "2/2 [==============================] - 2s 671ms/step - loss: 2.1855 - accuracy: 0.6934 - val_loss: 3.3771 - val_accuracy: 0.6709\n",
            "Epoch 119/200\n",
            "2/2 [==============================] - 2s 672ms/step - loss: 2.1813 - accuracy: 0.6953 - val_loss: 3.3846 - val_accuracy: 0.6716\n",
            "Epoch 120/200\n",
            "2/2 [==============================] - 2s 681ms/step - loss: 2.1760 - accuracy: 0.6939 - val_loss: 3.3906 - val_accuracy: 0.6721\n",
            "Epoch 121/200\n",
            "2/2 [==============================] - 2s 680ms/step - loss: 2.1733 - accuracy: 0.6944 - val_loss: 3.4007 - val_accuracy: 0.6694\n",
            "Epoch 122/200\n",
            "2/2 [==============================] - 2s 670ms/step - loss: 2.1690 - accuracy: 0.6954 - val_loss: 3.4081 - val_accuracy: 0.6722\n",
            "Epoch 123/200\n",
            "2/2 [==============================] - 2s 678ms/step - loss: 2.1659 - accuracy: 0.6936 - val_loss: 3.4162 - val_accuracy: 0.6701\n",
            "Epoch 124/200\n",
            "2/2 [==============================] - 2s 673ms/step - loss: 2.1618 - accuracy: 0.6955 - val_loss: 3.4229 - val_accuracy: 0.6701\n",
            "Epoch 125/200\n",
            "2/2 [==============================] - 2s 695ms/step - loss: 2.1596 - accuracy: 0.6949 - val_loss: 3.4318 - val_accuracy: 0.6709\n",
            "Epoch 126/200\n",
            "2/2 [==============================] - 2s 674ms/step - loss: 2.1551 - accuracy: 0.6948 - val_loss: 3.4406 - val_accuracy: 0.6691\n",
            "Epoch 127/200\n",
            "2/2 [==============================] - 2s 677ms/step - loss: 2.1526 - accuracy: 0.6957 - val_loss: 3.4445 - val_accuracy: 0.6706\n",
            "Epoch 128/200\n",
            "2/2 [==============================] - 2s 682ms/step - loss: 2.1478 - accuracy: 0.6952 - val_loss: 3.4551 - val_accuracy: 0.6686\n",
            "Epoch 129/200\n",
            "2/2 [==============================] - 2s 682ms/step - loss: 2.1463 - accuracy: 0.6956 - val_loss: 3.4610 - val_accuracy: 0.6698\n",
            "Epoch 130/200\n",
            "2/2 [==============================] - 2s 673ms/step - loss: 2.1446 - accuracy: 0.6952 - val_loss: 3.4692 - val_accuracy: 0.6684\n",
            "Epoch 131/200\n",
            "2/2 [==============================] - 2s 673ms/step - loss: 2.1385 - accuracy: 0.6967 - val_loss: 3.4751 - val_accuracy: 0.6694\n",
            "Epoch 132/200\n",
            "2/2 [==============================] - 2s 676ms/step - loss: 2.1367 - accuracy: 0.6954 - val_loss: 3.4849 - val_accuracy: 0.6686\n",
            "Epoch 133/200\n",
            "2/2 [==============================] - 2s 673ms/step - loss: 2.1331 - accuracy: 0.6966 - val_loss: 3.4897 - val_accuracy: 0.6691\n",
            "Epoch 134/200\n",
            "2/2 [==============================] - 2s 676ms/step - loss: 2.1288 - accuracy: 0.6959 - val_loss: 3.4982 - val_accuracy: 0.6685\n",
            "Epoch 135/200\n",
            "2/2 [==============================] - 2s 684ms/step - loss: 2.1266 - accuracy: 0.6963 - val_loss: 3.5041 - val_accuracy: 0.6689\n",
            "Epoch 136/200\n",
            "2/2 [==============================] - 2s 673ms/step - loss: 2.1241 - accuracy: 0.6962 - val_loss: 3.5120 - val_accuracy: 0.6679\n",
            "Epoch 137/200\n",
            "2/2 [==============================] - 2s 675ms/step - loss: 2.1215 - accuracy: 0.6973 - val_loss: 3.5172 - val_accuracy: 0.6685\n",
            "Epoch 138/200\n",
            "2/2 [==============================] - 2s 673ms/step - loss: 2.1175 - accuracy: 0.6965 - val_loss: 3.5258 - val_accuracy: 0.6678\n",
            "Epoch 139/200\n",
            "2/2 [==============================] - 2s 680ms/step - loss: 2.1156 - accuracy: 0.6975 - val_loss: 3.5288 - val_accuracy: 0.6685\n",
            "Epoch 140/200\n",
            "2/2 [==============================] - 2s 672ms/step - loss: 2.1124 - accuracy: 0.6968 - val_loss: 3.5391 - val_accuracy: 0.6673\n",
            "Epoch 141/200\n",
            "2/2 [==============================] - 2s 674ms/step - loss: 2.1101 - accuracy: 0.6978 - val_loss: 3.5420 - val_accuracy: 0.6683\n",
            "Epoch 142/200\n",
            "2/2 [==============================] - 2s 678ms/step - loss: 2.1053 - accuracy: 0.6970 - val_loss: 3.5516 - val_accuracy: 0.6676\n",
            "Epoch 143/200\n",
            "2/2 [==============================] - 2s 676ms/step - loss: 2.1009 - accuracy: 0.6975 - val_loss: 3.5562 - val_accuracy: 0.6678\n",
            "Epoch 144/200\n",
            "2/2 [==============================] - 2s 680ms/step - loss: 2.0992 - accuracy: 0.6980 - val_loss: 3.5610 - val_accuracy: 0.6677\n",
            "Epoch 145/200\n",
            "2/2 [==============================] - 2s 687ms/step - loss: 2.0962 - accuracy: 0.6973 - val_loss: 3.5701 - val_accuracy: 0.6669\n",
            "Epoch 146/200\n",
            "2/2 [==============================] - 2s 680ms/step - loss: 2.0948 - accuracy: 0.6988 - val_loss: 3.5719 - val_accuracy: 0.6679\n",
            "Epoch 147/200\n",
            "2/2 [==============================] - 2s 676ms/step - loss: 2.0915 - accuracy: 0.6983 - val_loss: 3.5811 - val_accuracy: 0.6671\n",
            "Epoch 148/200\n",
            "2/2 [==============================] - 2s 672ms/step - loss: 2.0883 - accuracy: 0.6977 - val_loss: 3.5896 - val_accuracy: 0.6664\n",
            "Epoch 149/200\n",
            "2/2 [==============================] - 2s 684ms/step - loss: 2.0860 - accuracy: 0.6988 - val_loss: 3.5906 - val_accuracy: 0.6675\n",
            "Epoch 150/200\n",
            "2/2 [==============================] - 2s 674ms/step - loss: 2.0830 - accuracy: 0.6977 - val_loss: 3.5976 - val_accuracy: 0.6667\n",
            "Epoch 151/200\n",
            "2/2 [==============================] - 2s 676ms/step - loss: 2.0799 - accuracy: 0.6981 - val_loss: 3.6078 - val_accuracy: 0.6656\n",
            "Epoch 152/200\n",
            "2/2 [==============================] - 2s 673ms/step - loss: 2.0789 - accuracy: 0.6991 - val_loss: 3.6082 - val_accuracy: 0.6668\n",
            "Epoch 153/200\n",
            "2/2 [==============================] - 2s 685ms/step - loss: 2.0739 - accuracy: 0.6994 - val_loss: 3.6090 - val_accuracy: 0.6681\n",
            "Epoch 154/200\n",
            "2/2 [==============================] - 2s 683ms/step - loss: 2.0750 - accuracy: 0.6980 - val_loss: 3.6202 - val_accuracy: 0.6667\n",
            "Epoch 155/200\n",
            "2/2 [==============================] - 2s 676ms/step - loss: 2.0683 - accuracy: 0.6986 - val_loss: 3.6295 - val_accuracy: 0.6653\n",
            "Epoch 156/200\n",
            "2/2 [==============================] - 2s 682ms/step - loss: 2.0666 - accuracy: 0.6997 - val_loss: 3.6315 - val_accuracy: 0.6662\n",
            "Epoch 157/200\n",
            "2/2 [==============================] - 2s 686ms/step - loss: 2.0641 - accuracy: 0.6988 - val_loss: 3.6356 - val_accuracy: 0.6663\n",
            "Epoch 158/200\n",
            "2/2 [==============================] - 2s 671ms/step - loss: 2.0598 - accuracy: 0.6995 - val_loss: 3.6460 - val_accuracy: 0.6657\n",
            "Epoch 159/200\n",
            "2/2 [==============================] - 2s 680ms/step - loss: 2.0588 - accuracy: 0.6995 - val_loss: 3.6499 - val_accuracy: 0.6648\n",
            "Epoch 160/200\n",
            "2/2 [==============================] - 2s 673ms/step - loss: 2.0559 - accuracy: 0.7005 - val_loss: 3.6488 - val_accuracy: 0.6659\n",
            "Epoch 161/200\n",
            "2/2 [==============================] - 2s 673ms/step - loss: 2.0546 - accuracy: 0.7000 - val_loss: 3.6504 - val_accuracy: 0.6662\n",
            "Epoch 162/200\n",
            "2/2 [==============================] - 2s 673ms/step - loss: 2.0507 - accuracy: 0.6996 - val_loss: 3.6593 - val_accuracy: 0.6662\n",
            "Epoch 163/200\n",
            "2/2 [==============================] - 2s 678ms/step - loss: 2.0464 - accuracy: 0.6998 - val_loss: 3.6686 - val_accuracy: 0.6646\n",
            "Epoch 164/200\n",
            "2/2 [==============================] - 2s 676ms/step - loss: 2.0439 - accuracy: 0.7005 - val_loss: 3.6750 - val_accuracy: 0.6649\n",
            "Epoch 165/200\n",
            "2/2 [==============================] - 2s 676ms/step - loss: 2.0416 - accuracy: 0.7009 - val_loss: 3.6767 - val_accuracy: 0.6648\n",
            "Epoch 166/200\n",
            "2/2 [==============================] - 2s 676ms/step - loss: 2.0360 - accuracy: 0.7009 - val_loss: 3.6792 - val_accuracy: 0.6646\n",
            "Epoch 167/200\n",
            "2/2 [==============================] - 2s 671ms/step - loss: 2.0326 - accuracy: 0.7013 - val_loss: 3.6858 - val_accuracy: 0.6650\n",
            "Epoch 168/200\n",
            "2/2 [==============================] - 2s 675ms/step - loss: 2.0289 - accuracy: 0.7001 - val_loss: 3.6919 - val_accuracy: 0.6630\n",
            "Epoch 169/200\n",
            "2/2 [==============================] - 2s 676ms/step - loss: 2.0288 - accuracy: 0.7004 - val_loss: 3.7019 - val_accuracy: 0.6626\n",
            "Epoch 170/200\n",
            "2/2 [==============================] - 2s 674ms/step - loss: 2.0302 - accuracy: 0.7003 - val_loss: 3.7200 - val_accuracy: 0.6599\n",
            "Epoch 171/200\n",
            "2/2 [==============================] - 2s 672ms/step - loss: 2.0614 - accuracy: 0.6972 - val_loss: 3.7338 - val_accuracy: 0.6578\n",
            "Epoch 172/200\n",
            "2/2 [==============================] - 2s 675ms/step - loss: 2.0615 - accuracy: 0.6942 - val_loss: 3.6923 - val_accuracy: 0.6654\n",
            "Epoch 173/200\n",
            "2/2 [==============================] - 2s 677ms/step - loss: 2.0277 - accuracy: 0.6990 - val_loss: 3.6721 - val_accuracy: 0.6701\n",
            "Epoch 174/200\n",
            "2/2 [==============================] - 2s 675ms/step - loss: 2.0531 - accuracy: 0.6970 - val_loss: 3.7001 - val_accuracy: 0.6630\n",
            "Epoch 175/200\n",
            "2/2 [==============================] - 2s 673ms/step - loss: 2.0335 - accuracy: 0.6981 - val_loss: 3.7056 - val_accuracy: 0.6606\n",
            "Epoch 176/200\n",
            "2/2 [==============================] - 2s 672ms/step - loss: 2.0238 - accuracy: 0.6993 - val_loss: 3.6795 - val_accuracy: 0.6700\n",
            "Epoch 177/200\n",
            "2/2 [==============================] - 2s 674ms/step - loss: 2.0367 - accuracy: 0.6961 - val_loss: 3.6870 - val_accuracy: 0.6638\n",
            "Epoch 178/200\n",
            "2/2 [==============================] - 2s 672ms/step - loss: 2.0177 - accuracy: 0.6987 - val_loss: 3.7065 - val_accuracy: 0.6616\n",
            "Epoch 179/200\n",
            "2/2 [==============================] - 2s 672ms/step - loss: 2.0131 - accuracy: 0.7003 - val_loss: 3.6722 - val_accuracy: 0.6693\n",
            "Epoch 180/200\n",
            "2/2 [==============================] - 2s 677ms/step - loss: 2.0202 - accuracy: 0.6978 - val_loss: 3.7079 - val_accuracy: 0.6623\n",
            "Epoch 181/200\n",
            "2/2 [==============================] - 2s 674ms/step - loss: 2.0171 - accuracy: 0.6981 - val_loss: 3.6852 - val_accuracy: 0.6645\n",
            "Epoch 182/200\n",
            "2/2 [==============================] - 2s 670ms/step - loss: 2.0068 - accuracy: 0.6993 - val_loss: 3.6869 - val_accuracy: 0.6700\n",
            "Epoch 183/200\n",
            "2/2 [==============================] - 2s 671ms/step - loss: 2.0043 - accuracy: 0.6966 - val_loss: 3.7127 - val_accuracy: 0.6573\n",
            "Epoch 184/200\n",
            "2/2 [==============================] - 2s 683ms/step - loss: 2.0061 - accuracy: 0.6973 - val_loss: 3.6882 - val_accuracy: 0.6701\n",
            "Epoch 185/200\n",
            "2/2 [==============================] - 2s 841ms/step - loss: 2.0006 - accuracy: 0.6974 - val_loss: 3.6943 - val_accuracy: 0.6623\n",
            "Epoch 186/200\n",
            "2/2 [==============================] - 2s 700ms/step - loss: 1.9875 - accuracy: 0.6989 - val_loss: 3.6979 - val_accuracy: 0.6659\n",
            "Epoch 187/200\n",
            "2/2 [==============================] - 2s 680ms/step - loss: 1.9774 - accuracy: 0.6996 - val_loss: 3.6880 - val_accuracy: 0.6659\n",
            "Epoch 188/200\n",
            "2/2 [==============================] - 2s 680ms/step - loss: 1.9720 - accuracy: 0.6991 - val_loss: 3.7143 - val_accuracy: 0.6623\n",
            "Epoch 189/200\n",
            "2/2 [==============================] - 2s 674ms/step - loss: 1.9680 - accuracy: 0.7011 - val_loss: 3.6945 - val_accuracy: 0.6670\n",
            "Epoch 190/200\n",
            "2/2 [==============================] - 2s 675ms/step - loss: 1.9638 - accuracy: 0.6997 - val_loss: 3.7185 - val_accuracy: 0.6633\n",
            "Epoch 191/200\n",
            "2/2 [==============================] - 2s 691ms/step - loss: 1.9564 - accuracy: 0.7005 - val_loss: 3.7131 - val_accuracy: 0.6647\n",
            "Epoch 192/200\n",
            "2/2 [==============================] - 2s 674ms/step - loss: 1.9497 - accuracy: 0.7012 - val_loss: 3.7197 - val_accuracy: 0.6655\n",
            "Epoch 193/200\n",
            "2/2 [==============================] - 2s 678ms/step - loss: 1.9439 - accuracy: 0.6988 - val_loss: 3.7354 - val_accuracy: 0.6620\n",
            "Epoch 194/200\n",
            "2/2 [==============================] - 2s 683ms/step - loss: 1.9388 - accuracy: 0.7019 - val_loss: 3.7274 - val_accuracy: 0.6656\n",
            "Epoch 195/200\n",
            "2/2 [==============================] - 2s 684ms/step - loss: 1.9345 - accuracy: 0.6999 - val_loss: 3.7414 - val_accuracy: 0.6640\n",
            "Epoch 196/200\n",
            "2/2 [==============================] - 2s 686ms/step - loss: 1.9299 - accuracy: 0.7005 - val_loss: 3.7474 - val_accuracy: 0.6625\n",
            "Epoch 197/200\n",
            "2/2 [==============================] - 2s 672ms/step - loss: 1.9274 - accuracy: 0.7019 - val_loss: 3.7455 - val_accuracy: 0.6661\n",
            "Epoch 198/200\n",
            "2/2 [==============================] - 2s 679ms/step - loss: 1.9210 - accuracy: 0.7000 - val_loss: 3.7531 - val_accuracy: 0.6635\n",
            "Epoch 199/200\n",
            "2/2 [==============================] - 2s 670ms/step - loss: 1.9186 - accuracy: 0.7008 - val_loss: 3.7648 - val_accuracy: 0.6637\n",
            "Epoch 200/200\n",
            "2/2 [==============================] - 2s 672ms/step - loss: 1.9145 - accuracy: 0.7016 - val_loss: 3.7572 - val_accuracy: 0.6647\n",
            "u - - - - - - - - - - - - - - - - - - - - . [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK] [UNK]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "_NOxalnr6bLg",
        "outputId": "ce61b5af-4769-42f1-c9f9-e1a6fd12f55b"
      },
      "source": [
        "model_for_comparison = embed_rnn_model1\n",
        "print(\"From Training Data:\")\n",
        "print(type(english_sentences))\n",
        "show_translation_comparison(tmp_x[:1], english_sentences.values[:1], zulu_sentences.values[:1], model_for_comparison)\n",
        "print()\n",
        "show_translation_comparison(tmp_x[1:2], english_sentences.values[1:2], zulu_sentences.values[1:2], model_for_comparison)\n",
        "print()\n",
        "show_translation_comparison(tmp_x[2:3], english_sentences.values[2:3], zulu_sentences.values[2:3], model_for_comparison)\n",
        "print()\n",
        "print(\"From Testing Data:\")\n",
        "show_translation_comparison(tmp_x_test[:1], english_sentences.values[len(preproc_english_sentences):len(preproc_english_sentences)+1], zulu_sentences.values[len(preproc_zulu_sentences):len(preproc_zulu_sentences)+1], model_for_comparison)\n",
        "print()\n",
        "show_translation_comparison(tmp_x_test[1:2], english_sentences.values[len(preproc_english_sentences)+1:len(preproc_english_sentences)+2], zulu_sentences.values[len(preproc_zulu_sentences)+1:len(preproc_zulu_sentences)+2], model_for_comparison)\n",
        "print()\n",
        "show_translation_comparison(tmp_x_test[4:5], english_sentences.values[len(preproc_english_sentences)+4:len(preproc_english_sentences)+5], zulu_sentences.values[len(preproc_zulu_sentences)+4:len(preproc_zulu_sentences)+5], model_for_comparison)\n",
        "print()\n",
        "\n",
        "# print(english_sentences.values[:30])"
      ],
      "execution_count": 57,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "From Training Data:\n",
            "<class 'pandas.core.series.Series'>\n",
            "English Sentence:  [\"his solo albums in the 60s were some of the most brilliant and breathtaking beautiful sounds i've ever heard.\"]\n",
            "Model Prediction Sentence:  ama albhamu akhe e solo eminyaka yawo 60 angamanye anezingoma ezimnandi kakhulu kwengake ngazizwa <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD>\n",
            "Actual Translation:  ['ama-albhamu akhe e-solo eminyaka yawo-60 angamanye anezingoma ezimnandi kakhulu kwengake ngazizwa.']\n",
            "\n",
            "English Sentence:  ['his father was a chauffeur.']\n",
            "Model Prediction Sentence:  uyise wayengushofa <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD>\n",
            "Actual Translation:  ['uyise wayengushofa.']\n",
            "\n",
            "English Sentence:  ['no leadership in removing the anti-semites.']\n",
            "Model Prediction Sentence:  abukho ubuholi obungasusa abaphikisana nama semites <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD>\n",
            "Actual Translation:  ['abukho ubuholi obungasusa abaphikisana nama-semites.']\n",
            "\n",
            "From Testing Data:\n",
            "English Sentence:  ['one user said: \"gordon is a typical negative vote leave curmudgeon.']\n",
            "Model Prediction Sentence:  i jose uthe uthe white elibheka ukunambitheka ukhona <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD>\n",
            "Actual Translation:  ['omunye umsebenzisi uthe: \"ugordon uyi-curmudgeon ejwayelekile yevote leave.\\n']\n",
            "\n",
            "English Sentence:  ['an inquest into the deaths of the london bridge and borough market attack victims begins']\n",
            "Model Prediction Sentence:  amaholide 215 saphakathi i broadway emzimbeni ongamazi <PAD> <PAD> kusho kusho lemakethe 2018 2018 lwe <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD>\n",
            "Actual Translation:  ['kuqala uphenyo ngokushona kwabantu abahlaselwa yilondon bridge naseborough market\\n']\n",
            "\n",
            "English Sentence:  [\"and that's terrifying for anyone in houston's path.\"]\n",
            "Model Prediction Sentence:  empeleni ama imali imnandi 500 <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD>\n",
            "Actual Translation:  ['futhi lokho kuyesabisa kunoma ngubani osendleleni kahouston.\\n']\n",
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "sBPM8itagRWA"
      },
      "source": [
        "# Testing and Analysis"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "lfPgkxM1gV6T"
      },
      "source": [
        "## Model 1 BLEU score"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "5Anqi_EucR6U",
        "outputId": "9c5b0742-2433-4762-e2fb-2a3b5f2b3965"
      },
      "source": [
        "ref, cand = get_ref_cand(tmp_x_test, simple_rnn_model, preproc_zulu_sentences_test)\n",
        "model1_bleu = calc_bleu_score(ref, cand)\n",
        "ref_t, cand_t = get_ref_cand(tmp_x, simple_rnn_model, preproc_zulu_sentences)\n",
        "model1_bleu_t = calc_bleu_score(ref_t, cand_t)\n",
        "print('Model 1 BLEU Score(Testing Data): ', model1_bleu)\n",
        "print('Model 1 BLEU Score(Training Data): ', model1_bleu_t)"
      ],
      "execution_count": 32,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(228, 1, 25) (228, 25)\n",
            "(2046, 1, 25) (2046, 25)\n",
            "Model 1 BLEU Score(Testing Data):  0.1953026214008288\n",
            "Model 1 BLEU Score(Training Data):  0.2141769267007398\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "uRblZNNIga8J"
      },
      "source": [
        "## Model 2 BLEU score"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "FVqHjjPwcWte",
        "outputId": "23f74a49-fd07-4f7b-fc81-a03dd6a28824"
      },
      "source": [
        "ref, cand = get_ref_cand(tmp_x_test_bpe, simple_rnn_model_bpe, preproc_zulu_sentences_test_bpe)\n",
        "model2_bleu = calc_bleu_score(ref, cand)\n",
        "ref_t, cand_t = get_ref_cand(tmp_x_bpe, simple_rnn_model_bpe, preproc_zulu_sentences_bpe)\n",
        "model2_bleu_t = calc_bleu_score(ref_t, cand_t)\n",
        "print('Model 2 BLEU Score(Testing Data): ', model2_bleu)\n",
        "print('Model 2 BLEU Score(Training Data): ', model2_bleu_t)"
      ],
      "execution_count": 33,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(228, 1, 58) (228, 58)\n",
            "(2046, 1, 58) (2046, 58)\n",
            "Model 2 BLEU Score(Testing Data):  0.0834943630469507\n",
            "Model 2 BLEU Score(Training Data):  0.024486134383787102\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "S1UKDjSegbBa"
      },
      "source": [
        "## Model 3 BLEU score"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "iaY4tRYhcZDC",
        "outputId": "5b75f9a9-7dbb-4eb5-e7a1-e37afbf24049"
      },
      "source": [
        "ref, cand = get_ref_cand(tmp_x_test, embed_rnn_model, preproc_zulu_sentences_test)\n",
        "model3_bleu = calc_bleu_score(ref, cand)\n",
        "ref_t, cand_t = get_ref_cand(tmp_x, embed_rnn_model, preproc_zulu_sentences)\n",
        "model3_bleu_t = calc_bleu_score(ref_t, cand_t)\n",
        "print('Model 3 BLEU Score(Testing Data): ', model3_bleu)\n",
        "print('Model 3 BLEU Score(Training Data): ', model3_bleu_t)"
      ],
      "execution_count": 34,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(228, 1, 25) (228, 25)\n",
            "(2046, 1, 25) (2046, 25)\n",
            "Model 3 BLEU Score(Testing Data):  0.049257100964287225\n",
            "Model 3 BLEU Score(Training Data):  0.7232061030994564\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "3vvwIyk0gbJL"
      },
      "source": [
        "## Model 4 BLEU score"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "FULZmQoZcbAM",
        "outputId": "af91d369-aab0-4a4f-d8cd-b5ce50fdcd4b"
      },
      "source": [
        "ref, cand = get_ref_cand(tmp_x_test_bpe, embed_rnn_model_bpe, preproc_zulu_sentences_test_bpe)\n",
        "model4_bleu = calc_bleu_score(ref, cand)\n",
        "ref_t, cand_t = get_ref_cand(tmp_x_bpe, embed_rnn_model_bpe, preproc_zulu_sentences_bpe)\n",
        "model4_bleu_t = calc_bleu_score(ref_t, cand_t)\n",
        "print('Model 4 BLEU Score(Testing Data): ', model4_bleu)\n",
        "print('Model 4 BLEU Score(Training Data): ', model4_bleu_t)"
      ],
      "execution_count": 35,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(228, 1, 58) (228, 58)\n",
            "(2046, 1, 58) (2046, 58)\n",
            "Model 4 BLEU Score(Testing Data):  0.02113412505386779\n",
            "Model 4 BLEU Score(Training Data):  0.7090198920641712\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "hWdY2UghgbOx"
      },
      "source": [
        "## Model 5 BLEU score"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "eQYykyzscd9K",
        "outputId": "f3354673-ddff-465c-da2d-cdb3a4425143"
      },
      "source": [
        "\n",
        "ref, cand = get_ref_cand(tmp_x_test, embed_rnn_model1, preproc_zulu_sentences_test)\n",
        "model5_bleu = calc_bleu_score(ref, cand)\n",
        "ref_t, cand_t = get_ref_cand(tmp_x, embed_rnn_model1, preproc_zulu_sentences)\n",
        "model5_bleu_t = calc_bleu_score(ref_t, cand_t)\n",
        "print('Model 5 BLEU Score(Testing Data): ', model5_bleu)\n",
        "print('Model 5 BLEU Score(Training Data): ', model5_bleu_t)"
      ],
      "execution_count": 36,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(228, 1, 25) (228, 25)\n",
            "(2046, 1, 25) (2046, 25)\n",
            "Model 5 BLEU Score(Testing Data):  0.05696413050399299\n",
            "Model 5 BLEU Score(Training Data):  0.7838435688220137\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Rh0VS2JCgbU9"
      },
      "source": [
        "## Model 6 BLEU score"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "6NOcjSM7cgfm",
        "outputId": "456592ae-3736-4583-ef93-32ae590bfbde"
      },
      "source": [
        "ref, cand = get_ref_cand(tmp_x_test_bpe, embed_rnn_model1_bpe, preproc_zulu_sentences_test_bpe)\n",
        "model6_bleu = calc_bleu_score(ref, cand)\n",
        "ref_t, cand_t = get_ref_cand(tmp_x_bpe, embed_rnn_model1_bpe, preproc_zulu_sentences_bpe)\n",
        "model6_bleu_t = calc_bleu_score(ref_t, cand_t)\n",
        "print('Model 6 BLEU Score(Testing Data): ', model6_bleu)\n",
        "print('Model 6 BLEU Score(Training Data): ', model6_bleu_t)"
      ],
      "execution_count": 37,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(228, 1, 58) (228, 58)\n",
            "(2046, 1, 58) (2046, 58)\n",
            "Model 6 BLEU Score(Testing Data):  0.020144643580463625\n",
            "Model 6 BLEU Score(Training Data):  0.685158325107435\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "4UrynaCJgiHY"
      },
      "source": [
        "## Model 7 BLEU score"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ABrXVU8NgTYq",
        "outputId": "f528b0b7-2541-4002-dcf8-478f73f28d79"
      },
      "source": [
        "ref, cand = get_ref_cand(tmp_x_test, encdec_rnn_model, preproc_zulu_sentences_test)\n",
        "model7_bleu = calc_bleu_score(ref, cand)\n",
        "ref_t, cand_t = get_ref_cand(tmp_x, encdec_rnn_model, preproc_zulu_sentences)\n",
        "model7_bleu_t = calc_bleu_score(ref_t, cand_t)\n",
        "print('Model 7 BLEU Score(Testing Data): ', model7_bleu)\n",
        "print('Model 7 BLEU Score(Training Data): ', model7_bleu_t)"
      ],
      "execution_count": 38,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(228, 1, 25) (228, 25)\n",
            "(2046, 1, 25) (2046, 25)\n",
            "Model 7 BLEU Score(Testing Data):  0.3412611317820572\n",
            "Model 7 BLEU Score(Training Data):  0.03149528004263639\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Behgn9AfcM0p"
      },
      "source": [
        "## Model 8 BLEU score"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "dY7n3fMycM0q",
        "outputId": "860ee130-915d-4a2b-f2ef-15b6371bd127"
      },
      "source": [
        "ref, cand = get_ref_cand(tmp_x_test_bpe, encdec_rnn_model_bpe, preproc_zulu_sentences_test_bpe)\n",
        "model8_bleu = calc_bleu_score(ref, cand)\n",
        "ref_t, cand_t = get_ref_cand(tmp_x_bpe, encdec_rnn_model_bpe, preproc_zulu_sentences_bpe)\n",
        "model8_bleu_t = calc_bleu_score(ref_t, cand_t)\n",
        "print('Model 8 BLEU Score(Testing Data): ', model8_bleu)\n",
        "print('Model 8 BLEU Score(Training Data): ', model8_bleu_t)"
      ],
      "execution_count": 39,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(228, 1, 58) (228, 58)\n",
            "(2046, 1, 58) (2046, 58)\n",
            "Model 8 BLEU Score(Testing Data):  0.15434788493435153\n",
            "Model 8 BLEU Score(Training Data):  0.15324777334042008\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "9qqNOoYiduRg"
      },
      "source": [
        "## BLEU Score Plots"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 279
        },
        "id": "fGnBhjROWiX9",
        "outputId": "a0d6f7e6-0e8f-468b-8566-96c5ec57f5db"
      },
      "source": [
        "models = ['1', '2', '3', '4', '5', '6', '7', '8', ]\n",
        "scores_t = [element * 100 for element in [model1_bleu_t, model2_bleu_t, model3_bleu_t, model4_bleu_t, model5_bleu_t, model6_bleu_t, model7_bleu_t, model8_bleu_t]]\n",
        "plt.bar(models, scores_t, color=[\"red\", \"blue\"])\n",
        "plt.xlabel(\"Model\")\n",
        "plt.ylabel(\"Training BLEU Score\")\n",
        "plt.savefig(\"BLEU_Scores_Training.png\", bbox_inches='tight', transparent=True)\n",
        "plt.show()"
      ],
      "execution_count": 58,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX4AAAEGCAYAAABiq/5QAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAWQElEQVR4nO3dfZBldX3n8fcHBhYYRAZoJyMIQyJLllg82SEoLKWMuBgpoLKWgYg7laV2dmuNQnQrortbSCq7qxWiuKmEzQTQ0RgIIi74UEZqwBh2DdgDKI8WShyYycA0KPKQjSh89497mmma7jt3ejj3dnPer6pb9zzce853BubTv/6d3/mdVBWSpO7YZdQFSJKGy+CXpI4x+CWpYwx+SeoYg1+SOmbJqAsYxAEHHFArV64cdRmStKhs2LDh0aoam7l9UQT/ypUrmZiYGHUZkrSoJNk423a7eiSpY1oN/iS/m+TuJHcluTLJHkkOTXJLku8n+asku7dZgyTphVoL/iQHAu8DxqvqdcCuwFnAx4BPVNVrgR8D57ZVgyTpxdru6lkC7JlkCbAXsAU4Gbim2b8OOLPlGiRJ07QW/FW1GbgYeJBe4P8E2AA8XlU/bz62CThwtu8nWZNkIsnE5ORkW2VKUue02dWzDDgDOBR4NbAUOHXQ71fV2qoar6rxsbEXjUaSJM1Tm109bwH+vqomq+pnwLXACcC+TdcPwEHA5hZrkCTN0GbwPwgcn2SvJAFWAfcANwHvaD6zGriuxRokSTO02cd/C72LuLcBdzbnWgt8EHh/ku8D+wOXt1WDJOnFWr1zt6ouBC6csfkB4Lg2zyuNTDK6c/tQJQ3IO3clqWMMfknqGINfkjrG4JekjjH4JaljDH5J6hiDX5I6xuCXpI4x+CWpYwx+SeoYg1+SOsbgl6SOMfglqWMMfknqGINfkjqm1fn4tYg5r7z0smWLX5I6xuCXpI5pLfiTHJ7kjmmvJ5Kcn2S/JDckub95X9ZWDXp5Skb3kl4O2nzY+veq6uiqOhp4PfCPwBeBC4D1VXUYsL5ZlyQNybC6elYBP6iqjcAZwLpm+zrgzCHVIElieMF/FnBls7y8qrY0yw8Dy2f7QpI1SSaSTExOTg6jRknqhNaDP8nuwOnA52fuq6oCZh27V1Vrq2q8qsbHxsZarlKSumMYLf63AbdV1SPN+iNJVgA071uHUIMkqTGM4D+bbd08ANcDq5vl1cB1Q6hBktRo9c7dJEuBU4B/P23zR4Grk5wLbATe2WYNknq8GVtTWg3+qnoa2H/GtsfojfKRJI2Ad+5KUscY/JLUMQa/JHWMwS9JHWPwS1LHGPyS1DEGvyR1jMEvSR1j8EtSxxj8ktQxBr8kdYzBL0kdY/BLUscY/JLUMQa/JHWMwS9JHWPwS1LHtBr8SfZNck2S+5Lcm+QNSfZLckOS+5v3ZW3WIEl6obZb/J8EvlZVvwwcBdwLXACsr6rDgPXNuiRpSFoL/iSvBE4CLgeoqmeq6nHgDGBd87F1wJlt1SBJerE2W/yHApPAp5LcnuSyJEuB5VW1pfnMw8Dy2b6cZE2SiSQTk5OTLZYpSd3SZvAvAY4FLq2qY4CnmdGtU1UF1Gxfrqq1VTVeVeNjY2MtlilJ3TJw8CfZawePvQnYVFW3NOvX0PtB8EiSFc0xVwBbd/C4kqSdsN3gT/LGJPcA9zXrRyX50+19r6oeBh5KcnizaRVwD3A9sLrZthq4bj6FS5LmZ8kAn/kE8K/oBTZV9Z0kJw14/PcCn0uyO/AA8Nv0fthcneRcYCPwzh2uWpI0b4MEP1X1UJLpm54d8Ht3AOOz7Fo1yPclSS+9QYL/oSRvBCrJbsB59MbjS5IWoUEu7v4H4D3AgcBm4OhmXZK0CPVt8SfZFfhkVb1rSPVIklrWt8VfVc8ChzQXZyVJLwOD9PE/APyfJNfTuwkLgKr6eGtVSZJaM0jw/6B57QK8ot1yJElt227wV9VFAEn2btafarsoSVJ7Brlz93VJbgfuBu5OsiHJr7RfmiSpDYMM51wLvL+qDqmqQ4APAH/eblmSpLYMEvxLq+qmqZWq+gawtLWKJEmtGmhUT5L/Cny2WT+H3kgfSdIiNEiL/98CY8C1wBeAA5ptkqRFaJBRPT8G3jeEWiRJQzDIqJ4bkuw7bX1Zkr9utyxJUlsG6eo5oHlIOvD8bwCvaq8kSVKbBgn+55IcPLWS5BDmeE6uJGnhG2RUz38Gbk7yN0CAfwmsabUqSVJrBrm4+7UkxwLH02vpn19Vj7ZemSSpFXN29SQ5JMkrAZqgfxp4K/BvBp2mOckPk9yZ5I4kE822/ZoLxvc378tegj+HJGlA/fr4r6a5QzfJ0cDngQeBo4A/3YFzvLmqjq6qqWfvXgCsr6rDgPXNuiRpSPp19exZVf/QLJ8DXFFVf5RkF+COnTjnGcCbmuV1wDeAD+7E8SRJO6Bfiz/Tlk+m1zqnqp7bgeMX8PVmRs+pC8LLq2pLs/wwsHzWkydrkkwkmZicnNyBU0qS+unX4r8xydXAFmAZcCNAkhXAMwMe/8Sq2pzkVcANSe6bvrOqKsmsQ0Orai29mUEZHx93+KgkvUT6tfjPpzc/zw/pBfjPmu2/QG+I53ZV1ebmfSvwReA44JHmh8fUD5Gt86pckjQvc7b4q6qAq2bZfvsgB06yFNilqp5slt8K/D5wPbAa+Gjzft086pYkzdMgN3DN13Lgi0mmzvOXzT0B3wauTnIusBF4Z4s1SJJmaC34q+oBekM/Z25/DFjV1nklSf0NMlePJOllZM4Wf5I7eeFkbAU8CtwEXFxV/9RybZKkFvTr6jltlm370bsg+8fAv2ulIklSq/qN6tk4y+aNwO1JBhrZI0laeObbx++1AUlapPr18R87y+Zl9Obt+WZrFUmSWtWvj/+PZqwX8Bi9SdXWtlWQJKld/fr43zzMQiRJw9HvQSyXTFs+b8a+T7dYkySpRf0u0p40bXn1jH1HtlCLJGkIBp2PP3N+SpK0qPS7uLtL8zzcXaYtT/0A2LX1yiRJregX/K8ENrAt7G+bts8Ho0jSItVvVM/KufYlObCVaiRJrZvvHbjfekmrkCQNzXyD34u9krRIzTf47eOXpEWq31w9f8zsAR9g30FPkGRXYALYXFWnJTmU3rN896d38fjdVfXMDlUtSZq3fqN6Jua5b6bzgHuBfZr1jwGfqKqrkvwv4Fzg0h04niRpJ/Qb1bNuZw+e5CDg7cB/A96f3pPXTwZ+q/nIOuAjGPySNDT95uo5IMmFSd6XZO8klya5K8l1SV474PEvAX4PeK5Z3x94vKp+3qxvAmYdGppkTZKJJBOTk5MDnk6StD39Lu7+JfDPgMOAW4EHgHcAXwYu296Bk5wGbK2qDfMprKrWVtV4VY2PjY3N5xCSpFn06+NfXlUfbrpnNlbVHzbb70vyngGOfQJwepJfB/ag18f/SWDfJEuaVv9BwOadqF+StIP6tfifBaiqAh6dse+5F3/8harqQ1V1UHMH8FnAjVX1LuAmer85QG/Wz+t2tGhJ0vz1a/H/YpLr6Q3fnFqmWT90J875QeCqJH8A3A5cvhPHkiTtoH7Bf8a05Ytn7Ju53ldVfYPeIxupqgeA43bk+5Kkl06/4Zx/M8xCJEnDMd8pGyRJi5TBL0kdY/BLUsf0u7gLQJIv8eLJ2n5Cb76eP6uqf2qjMElSOwZp8T8APAX8efN6AngS+OfNuiRpEdluix94Y1X96rT1LyX5dlX9apK72ypMktSOQVr8eyc5eGqlWd67WXUefUlaZAZp8X8AuDnJD9h21+5/TLKU3rTKkqRFZLvBX1VfTXIY8MvNpu9Nu6B7SWuVSZJaMUiLH+D1wMrm80cloao+01pVkqTWDDKc87PALwF30MzYSW94p8EvSYvQIC3+ceCIZnpmSdIiN8ionruAX2i7EEnScAzS4j8AuCfJrcBPpzZW1emtVSVJas0gwf+RtouQJA3PIMM5nZdfkl5G5uzjT3Jz8/5kkiemvZ5M8sT2DpxkjyS3JvlOkruTXNRsPzTJLUm+n+Svkuz+0v1xJEnbM2fwV9WJzfsrqmqfaa9XVNU+Axz7p8DJVXUUcDRwapLjgY8Bn6iq1wI/Bs7d+T+GJGlQA83Hn2TXJK9OcvDUa3vfqZ6nmtXdmlcBJwPXNNvXAWfOo25J0jwNcgPXe4ELgUeA55rNBRw5wHd3BTYArwX+BPgB8HhV/bz5yCbgwB0vW5I0X4OM6jkPOLyqHtvRg1fVs8DRSfYFvsi2+X62K8kaYA3AwQdv9xcMSdKABunqeYjeE7fmraoeB24C3gDsm2TqB85BwOY5vrO2qsaranxsbGxnTi9JmmaQFv8DwDeSfIUX3sD18X5fSjIG/KyqHk+yJ3AKvQu7NwHvAK4CVgPXzbN2SdI8DBL8Dzav3ZvXoFYA65p+/l2Aq6vqy0nuAa5K8gfA7cDlO1izJGknDHID10XzOXBVfRc4ZpbtDwDHzeeYkqSdN2fwJ7mkqs5P8iV6o3hewLl6JGlx6tfi/2zzfvEwCpEkDcecwV9VG5p35+qRpJeRQW7gOgz4H8ARwB5T26vqF1usS5LUkkHG8X8KuBT4OfBmeo9c/Is2i5IktWeQ4N+zqtYDqaqNVfUR4O3tliVJassg4/h/mmQX4P4kv0PvTtu92y1LktSWQVr85wF7Ae8DXg+cQ++OW0nSItS3xd/cdfubVfWfgKeA3x5KVZKk1vR7AteSZnbNE4dYjySpZf1a/LcCxwK3J7ke+Dzw9NTOqrq25dokSS0Y5OLuHsBj9J6cVUCad4NfkhahfsH/qiTvB+5iW+BPedHcPZKkxaFf8O9Kb9hmZtln8EvSItUv+LdU1e8PrRJJ0lD0G8c/W0tfkrTI9Qv+VUOrQpI0NP2mZf7RMAtpTUb4i0t5KUTSwjPIlA3zkuQ1SW5Kck+Su5Oc12zfL8kNSe5v3pe1VYMk6cVaC3560zh/oKqOAI4H3pPkCOACYH1VHQasb9YlSUPSWvBX1Zaquq1ZfhK4FzgQOANY13xsHXBmWzVIkl6szRb/85KsBI4BbgGWV9WWZtfDwPI5vrMmyUSSicnJyWGUKUmd0HrwJ9kb+AJwflU9MX1fVRVz3AxWVWuraryqxsfGxtouU5I6o9XgT7IbvdD/3LRJ3R5JsqLZvwLY2mYNkqQXanNUT4DLgXur6uPTdl3Ptge5rAaua6sGSdKLDTI753ydALwbuDPJHc22DwMfBa5Oci6wEXhnizVIkmZoLfir6mbmnvbBu4IlaUSGMqpHkrRwGPyS1DEGvyR1jMEvSR1j8EtSxxj8ktQxbY7jl6RF7+X4SA9b/JLUMQa/JHWMwS9JHWPwS1LHGPyS1DEGvyR1jMEvSR1j8EtSxxj8ktQxBr8kdUybz9y9IsnWJHdN27ZfkhuS3N+8L2vr/JKk2bXZ4v80cOqMbRcA66vqMGB9sy5JGqLWgr+qvgn8aMbmM4B1zfI64My2zi9Jmt2w+/iXV9WWZvlhYPlcH0yyJslEkonJycnhVCdJHTCyi7tVVcCck45W1dqqGq+q8bGxsSFWJkkvb8MO/keSrABo3rcO+fyS1HnDDv7rgdXN8mrguiGfX5I6r83hnFcC3wIOT7IpybnAR4FTktwPvKVZlyQNUWuPXqyqs+fYtaqtc0qSts87dyWpYwx+SeoYg1+SOsbgl6SOMfglqWMMfknqGINfkjrG4JekjjH4JaljDH5J6hiDX5I6xuCXpI5pbZI2SRpYMrpz15zPg3rZssUvSR1ji3+EbORIGgVb/JLUMQa/JHWMwS9JHTOS4E9yapLvJfl+kgtGUYMkddXQgz/JrsCfAG8DjgDOTnLEsOuQpK4aRYv/OOD7VfVAVT0DXAWcMYI6JKmTRjGc80DgoWnrm4Bfm/mhJGuANc3qU0m+N4TaZnMA8Oi8vtn+eM151zaEoaT+vc2Pf2/z49/b7A6ZbeOCHcdfVWuBtaOuI8lEVY2Puo7ZWNv8WNv8WNv8LMTaRtHVsxl4zbT1g5ptkqQhGEXwfxs4LMmhSXYHzgKuH0EdktRJQ+/qqaqfJ/kd4K+BXYErquruYdexA0be3dSHtc2Ptc2Ptc3Pgqst5aQtktQp3rkrSR1j8EtSxxj8c0hyRZKtSe4adS3TJXlNkpuS3JPk7iTnjbqmKUn2SHJrku80tV006ppmSrJrktuTfHnUtUyX5IdJ7kxyR5KJUdczXZJ9k1yT5L4k9yZ5w6hrAkhyePP3NfV6Isn5o65rSpLfbf4d3JXkyiR7jLqmKfbxzyHJScBTwGeq6nWjrmdKkhXAiqq6LckrgA3AmVV1z4hLI0mApVX1VJLdgJuB86rq70Zc2vOSvB8YB/apqtNGXc+UJD8ExqtqfjchtSjJOuBvq+qyZiTeXlX1+Kjrmq6ZCmYz8GtVtXEB1HMgvf//j6iq/5fkauCrVfXp0VbWY4t/DlX1TeBHo65jpqraUlW3NctPAvfSuxt65KrnqWZ1t+a1YFoWSQ4C3g5cNupaFoskrwROAi4HqKpnFlroN1YBP1gIoT/NEmDPJEuAvYB/GHE9zzP4F7EkK4FjgFtGW8k2TVfKHcBW4IaqWjC1AZcAvwc8N+pCZlHA15NsaKYrWSgOBSaBTzVdZJclWTrqomZxFnDlqIuYUlWbgYuBB4EtwE+q6uujrWobg3+RSrI38AXg/Kp6YtT1TKmqZ6vqaHp3ZB+XZEF0kyU5DdhaVRtGXcscTqyqY+nNWvuepqtxIVgCHAtcWlXHAE8DC2oq9ab76XTg86OuZUqSZfQmnzwUeDWwNMk5o61qG4N/EWr6z78AfK6qrh11PbNpugNuAk4ddS2NE4DTm770q4CTk/zFaEvapmkhUlVbgS/Sm8V2IdgEbJr2m9s19H4QLCRvA26rqkdGXcg0bwH+vqomq+pnwLXAG0dc0/MM/kWmuYB6OXBvVX181PVMl2Qsyb7N8p7AKcB9o62qp6o+VFUHVdVKet0CN1bVgmiBJVnaXKin6UZ5K7AgRpNV1cPAQ0kObzatAkY+kGCGs1lA3TyNB4Hjk+zV/JtdRe963IJg8M8hyZXAt4DDk2xKcu6oa2qcALybXot1ahjbr4+6qMYK4KYk36U3J9MNVbWghk0uUMuBm5N8B7gV+EpVfW3ENU33XuBzzX/Xo4H/PuJ6ntf8oDyFXot6wWh+Q7oGuA24k17WLpipGxzOKUkdY4tfkjrG4JekjjH4JaljDH5J6hiDX5I6xuCXgCQ1/YauJEuSTO7oLJ7NLJsH7OxnpDYZ/FLP08DrmhvPoDc2fPMI65FaY/BL23yV3uydMONu0CT7JfnfSb6b5O+SHNls3z/J15t51y8DMu075zTPJ7gjyZ81UwdLI2fwS9tcBZzVPDDjSF446+lFwO1VdSTwYeAzzfYLgZur6lfozbFzMECSfwH8JnBCM2nds8C7hvKnkLZjyagLkBaKqvpuM9X12fRa/9OdCPzr5nM3Ni39fejNVf8bzfavJPlx8/lVwOuBb/emamFPelNVSyNn8EsvdD29edTfBOy/E8cJsK6qPvRSFCW9lOzqkV7oCuCiqrpzxva/pemqSfIm4NHmOQjfBH6r2f42YFnz+fXAO5K8qtm3X5JD2i9f2j5b/NI0VbUJ+J+z7PoIcEUzQ+U/Aqub7RcBVya5G/i/9KbjparuSfJf6D1VaxfgZ8B7gIX0aEB1lLNzSlLH2NUjSR1j8EtSxxj8ktQxBr8kdYzBL0kdY/BLUscY/JLUMf8fJYf/PVOmzkQAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "k6Pn5GvSdw6d",
        "outputId": "43f951c1-b5bf-4148-b073-05ecde0cac45"
      },
      "source": [
        "models = ['1', '2', '3', '4', '5', '6', '7', '8', ]\n",
        "scores = [element * 100 for element in [model1_bleu, model2_bleu, model3_bleu, model4_bleu, model5_bleu, model6_bleu, model7_bleu, model8_bleu]]\n",
        "plt.bar(models,scores, color=[\"red\", \"blue\"])\n",
        "plt.title(\"Testing Data Performance\")\n",
        "plt.xlabel(\"Model\")\n",
        "plt.ylabel(\"BLEU Score\")\n",
        "plt.savefig(\"BLEU_Scores_Testing.png\", bbox_inches='tight', transparent=True)\n",
        "plt.show()"
      ],
      "execution_count": 59,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7ZlJQA3VSopF"
      },
      "source": [
        "# Plots"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "IitME49BWEcp"
      },
      "source": [
        "\n",
        "train_loss1 = model1.history['loss']\n",
        "val_loss1   = model1.history['val_loss']\n",
        "train_acc1  = model1.history['accuracy']\n",
        "val_acc1    = model1.history['val_accuracy']\n",
        "\n",
        "\n",
        "train_loss2 = model2.history['loss']\n",
        "val_loss2   = model2.history['val_loss']\n",
        "train_acc2  = model2.history['accuracy']\n",
        "val_acc2    = model2.history['val_accuracy']\n",
        "\n",
        "\n",
        "train_loss3 = model3.history['loss']\n",
        "val_loss3   = model3.history['val_loss']\n",
        "train_acc3  = model3.history['accuracy']\n",
        "val_acc3    = model3.history['val_accuracy']\n",
        "\n",
        "\n",
        "train_loss4 = model4.history['loss']\n",
        "val_loss4   = model4.history['val_loss']\n",
        "train_acc4  = model4.history['accuracy']\n",
        "val_acc4    = model4.history['val_accuracy']\n",
        "\n",
        "\n",
        "train_loss5 = model5.history['loss']\n",
        "val_loss5   = model5.history['val_loss']\n",
        "train_acc5  = model5.history['accuracy']\n",
        "val_acc5    = model5.history['val_accuracy']\n",
        "\n",
        "\n",
        "train_loss6 = model6.history['loss']\n",
        "val_loss6   = model6.history['val_loss']\n",
        "train_acc6  = model6.history['accuracy']\n",
        "val_acc6    = model6.history['val_accuracy']\n",
        "\n",
        "\n",
        "train_loss7 = model7.history['loss']\n",
        "val_loss7   = model7.history['val_loss']\n",
        "train_acc7  = model7.history['accuracy']\n",
        "val_acc7    = model7.history['val_accuracy']\n",
        "\n",
        "train_loss8 = model8.history['loss']\n",
        "val_loss8   = model8.history['val_loss']\n",
        "train_acc8  = model8.history['accuracy']\n",
        "val_acc8    = model8.history['val_accuracy']"
      ],
      "execution_count": 51,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "crzNWqHpUCY_",
        "outputId": "34e147ce-7544-44c1-870c-0d4e48bc85d7"
      },
      "source": [
        "metrics = plt.figure()\n",
        "metrics.set_figheight(30)\n",
        "metrics.set_figwidth(10)\n",
        "\n",
        "xc = range(NUM_EPOCHS)\n",
        "\n",
        "metrics.add_subplot(4, 1, 1)\n",
        "plt.title(\"Training Loss\")\n",
        "plt.plot(xc, train_loss1, label=\"Model 1\")\n",
        "plt.plot(xc, train_loss2, label=\"Model 2\")\n",
        "plt.plot(xc, train_loss3, label=\"Model 3\")\n",
        "plt.plot(xc, train_loss4, label=\"Model 4\")\n",
        "plt.plot(xc, train_loss5, label=\"Model 5\")\n",
        "plt.plot(xc, train_loss6, label=\"Model 6\")\n",
        "plt.plot(xc, train_loss7, label=\"Model 7\")\n",
        "plt.plot(xc, train_loss8, label=\"Model 8\")\n",
        "plt.legend()\n",
        "\n",
        "metrics.add_subplot(4, 1, 2)\n",
        "plt.title(\"Validation Loss\")\n",
        "plt.plot(xc, val_loss1, label=\"Model 1\")\n",
        "plt.plot(xc, val_loss2, label=\"Model 2\")\n",
        "plt.plot(xc, val_loss3, label=\"Model 3\")\n",
        "plt.plot(xc, val_loss4, label=\"Model 4\")\n",
        "plt.plot(xc, val_loss5, label=\"Model 5\")\n",
        "plt.plot(xc, val_loss6, label=\"Model 6\")\n",
        "plt.plot(xc, val_loss7, label=\"Model 7\")\n",
        "plt.plot(xc, val_loss8, label=\"Model 8\")\n",
        "plt.legend()\n",
        "\n",
        "metrics.add_subplot(4, 1, 3)\n",
        "plt.title(\"Training Accuracy\")\n",
        "plt.plot(xc, train_acc1, label=\"Model 1\")\n",
        "plt.plot(xc, train_acc2, label=\"Model 2\")\n",
        "plt.plot(xc, train_acc3, label=\"Model 3\")\n",
        "plt.plot(xc, train_acc4, label=\"Model 4\")\n",
        "plt.plot(xc, train_acc5, label=\"Model 5\")\n",
        "plt.plot(xc, train_acc6, label=\"Model 6\")\n",
        "plt.plot(xc, train_acc7, label=\"Model 7\")\n",
        "plt.plot(xc, train_acc8, label=\"Model 8\")\n",
        "plt.legend()\n",
        "\n",
        "metrics.add_subplot(4, 1, 4)\n",
        "plt.title(\"Validation Accuracy\")\n",
        "plt.plot(xc, val_acc1, label=\"Model 1\")\n",
        "plt.plot(xc, val_acc2, label=\"Model 2\")\n",
        "plt.plot(xc, val_acc3, label=\"Model 3\")\n",
        "plt.plot(xc, val_acc4, label=\"Model 4\")\n",
        "plt.plot(xc, val_acc5, label=\"Model 5\")\n",
        "plt.plot(xc, val_acc6, label=\"Model 6\")\n",
        "plt.plot(xc, val_acc7, label=\"Model 7\")\n",
        "plt.plot(xc, val_acc8, label=\"Model 8\")\n",
        "plt.legend()\n",
        "plt.savefig(\"LossesAndAccuracy.png\", bbox_inches='tight', transparent=True)\n",
        "\n",
        "plt.show()"
      ],
      "execution_count": 52,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x2160 with 4 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "IugVpGaUd_uc"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}