Gru_Model_77.6%_Accuracy.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "oHAQuhSd0DaE"
      },
      "source": [
        "## Data collection"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "id": "xx2TjAex0DaF"
      },
      "outputs": [],
      "source": [
        "!wget -q https://git.io/J0fjL -O IAM_Words.zip\n",
        "!unzip -qq IAM_Words.zip\n",
        "!\n",
        "!mkdir data\n",
        "!mkdir data/words\n",
        "!tar -xf IAM_Words/words.tgz -C data/words\n",
        "!mv IAM_Words/words.txt data"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "82hW0Tf5ieoT",
        "outputId": "6fc92e18-1b92-464f-aeb9-9acb84de8f9a"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Mounted at /content/drive\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "raoUeTs40DaH"
      },
      "source": [
        "Preview how the dataset is organized. Lines prepended by \"#\" are just metadata information."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "RlKCqr9C0DaH",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "839680da-cf6a-4e15-b248-c06446c8879f"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "#--- words.txt ---------------------------------------------------------------#\n",
            "#\n",
            "# iam database word information\n",
            "#\n",
            "# format: a01-000u-00-00 ok 154 1 408 768 27 51 AT A\n",
            "#\n",
            "#     a01-000u-00-00  -> word id for line 00 in form a01-000u\n",
            "#     ok              -> result of word segmentation\n",
            "#                            ok: word was correctly\n",
            "#                            er: segmentation of word can be bad\n",
            "#\n",
            "#     154             -> graylevel to binarize the line containing this word\n",
            "#     1               -> number of components for this word\n",
            "#     408 768 27 51   -> bounding box around this word in x,y,w,h format\n",
            "#     AT              -> the grammatical tag for this word, see the\n",
            "#                        file tagset.txt for an explanation\n",
            "#     A               -> the transcription for this word\n",
            "#\n",
            "a01-000u-00-00 ok 154 408 768 27 51 AT A\n",
            "a01-000u-00-01 ok 154 507 766 213 48 NN MOVE\n",
            "a01-000u-00-02 ok 154 796 764 70 50 TO to\n",
            "a01-000u-00-03 ok 154 919 757 166 78 VB stop\n",
            "a01-000u-00-04 ok 154 1185 754 126 61 NPT Mr.\n",
            "a01-000u-00-05 ok 154 1438 746 382 73 NP Gaitskell\n",
            "a01-000u-00-06 ok 154 1896 757 173 72 IN from\n",
            "a01-000u-01-00 ok 156 395 932 441 100 VBG nominating\n",
            "a01-000u-01-01 ok 156 901 958 147 79 DTI any\n",
            "a01-000u-01-02 ok 156 1112 958 208 42 AP more\n",
            "a01-000u-01-03 ok 156 1400 937 294 59 NN Labour\n",
            "a01-000u-01-04 ok 156 1779 932 174 63 NN life\n",
            "a01-000u-01-05 ok 156 2008 933 237 70 NNS Peers\n",
            "a01-000u-02-00 ok 157 408 1106 65 70 BEZ is\n",
            "a01-000u-02-01 ok 157 541 1118 72 54 TO to\n",
            "a01-000u-02-02 ok 157 720 1114 113 63 BE be\n",
            "a01-000u-02-03 ok 157 916 1136 281 46 VBN made\n",
            "a01-000u-02-04 ok 157 1281 1117 80 59 IN at\n",
            "a01-000u-02-05 ok 157 1405 1140 64 35 AT a\n",
            "a01-000u-02-06 ok 157 1544 1115 339 96 NN meeting\n",
            "a01-000u-02-07 ok 157 1936 1106 91 74 INO of\n",
            "a01-000u-02-08 ok 157 2092 1121 302 65 NN Labour\n",
            "a01-000u-03-00 err 156 430 1290 177 59 NPTS M Ps\n",
            "a01-000u-03-01 err 156 705 1296 431 54 NR tomorrow\n",
            "a01-000u-03-02 err 156 1154 1346 9 10 . .\n",
            "a01-000u-03-03 err 156 1269 1292 320 68 NPT Mr.\n",
            "a01-000u-03-04 err 156 1570 1297 252 56 NP Michael\n",
            "a01-000u-03-05 err 156 1914 1292 190 63 NP Foot\n",
            "a01-000u-03-06 err 156 2179 1300 134 57 HVZ has\n",
            "a01-000u-04-00 ok 157 395 1475 153 71 VBN put\n",
            "a01-000u-04-01 ok 157 616 1482 221 48 RP down\n",
            "a01-000u-04-02 ok 157 891 1501 58 29 AT a\n"
          ]
        }
      ],
      "source": [
        "!head -50 data/words.txt"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "sqR-fSou0DaI"
      },
      "source": [
        "## Imports"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 2,
      "metadata": {
        "id": "GPCKRm9O0DaJ"
      },
      "outputs": [],
      "source": [
        "from tensorflow.keras.layers.experimental.preprocessing import StringLookup\n",
        "from tensorflow import keras\n",
        "\n",
        "import matplotlib.pyplot as plt\n",
        "import tensorflow as tf\n",
        "import numpy as np\n",
        "import os\n",
        "\n",
        "np.random.seed(42)\n",
        "tf.random.set_seed(42)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "CkUECYg40DaK"
      },
      "source": [
        "## Dataset splitting"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 3,
      "metadata": {
        "id": "5s9INtI60DaK"
      },
      "outputs": [],
      "source": [
        "base_path = \"data\"\n",
        "words_list = []\n",
        "\n",
        "words = open(f\"{base_path}/words.txt\", \"r\").readlines()\n",
        "for line in words:\n",
        "    if line[0] == \"#\":\n",
        "        continue\n",
        "    if line.split(\" \")[1] != \"err\":  # We don't need to deal with errored entries.\n",
        "        words_list.append(line)\n",
        "\n",
        "len(words_list)\n",
        "\n",
        "np.random.shuffle(words_list)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-bAZ0fuu0DaL"
      },
      "source": [
        "We will split the dataset into three subsets with a 90:5:5 ratio (train:validation:test)."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 4,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "PDJqmUZK0DaL",
        "outputId": "9611b6d1-68e2-4294-c5b5-3b98836200c0"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Total training samples: 86810\n",
            "Total validation samples: 4823\n",
            "Total test samples: 4823\n"
          ]
        }
      ],
      "source": [
        "split_idx = int(0.9 * len(words_list))\n",
        "train_samples = words_list[:split_idx]\n",
        "test_samples = words_list[split_idx:]\n",
        "\n",
        "val_split_idx = int(0.5 * len(test_samples))\n",
        "validation_samples = test_samples[:val_split_idx]\n",
        "test_samples = test_samples[val_split_idx:]\n",
        "\n",
        "assert len(words_list) == len(train_samples) + len(validation_samples) + len(\n",
        "    test_samples\n",
        "\n",
        ")\n",
        "\n",
        "print(f\"Total training samples: {len(train_samples)}\")\n",
        "print(f\"Total validation samples: {len(validation_samples)}\")\n",
        "print(f\"Total test samples: {len(test_samples)}\")"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "iITHuIyb0DaM"
      },
      "source": [
        "## Data input pipeline\n",
        "\n",
        "We start building our data input pipeline by first preparing the image paths."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 5,
      "metadata": {
        "id": "PhvX_QSz0DaM"
      },
      "outputs": [],
      "source": [
        "base_image_path = os.path.join(base_path, \"words\")\n",
        "\n",
        "\n",
        "def get_image_paths_and_labels(samples):\n",
        "    paths = []\n",
        "    corrected_samples = []\n",
        "    for (i, file_line) in enumerate(samples):\n",
        "        line_split = file_line.strip()\n",
        "        line_split = line_split.split(\" \")\n",
        "\n",
        "        # Each line split will have this format for the corresponding image:\n",
        "        # part1/part1-part2/part1-part2-part3.png\n",
        "        image_name = line_split[0]\n",
        "        partI = image_name.split(\"-\")[0]\n",
        "        partII = image_name.split(\"-\")[1]\n",
        "        img_path = os.path.join(\n",
        "            base_image_path, partI, partI + \"-\" + partII, image_name + \".png\"\n",
        "        )\n",
        "        if os.path.getsize(img_path):\n",
        "            paths.append(img_path)\n",
        "            corrected_samples.append(file_line.split(\"\\n\")[0])\n",
        "\n",
        "    return paths, corrected_samples\n",
        "\n",
        "\n",
        "train_img_paths, train_labels = get_image_paths_and_labels(train_samples)\n",
        "validation_img_paths, validation_labels = get_image_paths_and_labels(validation_samples)\n",
        "test_img_paths, test_labels = get_image_paths_and_labels(test_samples)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "dGr39Yui0DaM"
      },
      "source": [
        "Then we prepare the ground-truth labels."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 6,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "fi1gtbOo0DaN",
        "outputId": "488544e2-d088-4fbc-fae9-0b13cc228ba0"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Maximum length:  21\n",
            "Vocab size:  78\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "['sure',\n",
              " 'he',\n",
              " 'during',\n",
              " 'of',\n",
              " 'booty',\n",
              " 'gastronomy',\n",
              " 'boy',\n",
              " 'The',\n",
              " 'and',\n",
              " 'in']"
            ]
          },
          "metadata": {},
          "execution_count": 6
        }
      ],
      "source": [
        "# Find maximum length and the size of the vocabulary in the training data.\n",
        "train_labels_cleaned = []\n",
        "characters = set()\n",
        "max_len = 0\n",
        "\n",
        "for label in train_labels:\n",
        "    label = label.split(\" \")[-1].strip()\n",
        "    for char in label:\n",
        "        characters.add(char)\n",
        "\n",
        "    max_len = max(max_len, len(label))\n",
        "    train_labels_cleaned.append(label)\n",
        "\n",
        "characters = sorted(list(characters))\n",
        "\n",
        "print(\"Maximum length: \", max_len)\n",
        "print(\"Vocab size: \", len(characters))\n",
        "\n",
        "# Check some label samples.\n",
        "train_labels_cleaned[:10]"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "lyQBZnqz0DaN"
      },
      "source": [
        "Now we clean the validation and the test labels as well."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 7,
      "metadata": {
        "id": "csqr2q850DaN"
      },
      "outputs": [],
      "source": [
        "\n",
        "def clean_labels(labels):\n",
        "    cleaned_labels = []\n",
        "    for label in labels:\n",
        "        label = label.split(\" \")[-1].strip()\n",
        "        cleaned_labels.append(label)\n",
        "    return cleaned_labels\n",
        "\n",
        "\n",
        "validation_labels_cleaned = clean_labels(validation_labels)\n",
        "test_labels_cleaned = clean_labels(test_labels)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "97Wl_ggL0DaN"
      },
      "source": [
        "### Building the character vocabulary\n",
        "\n",
        "Keras provides different preprocessing layers to deal with different modalities of data.\n",
        "[This guide](https://keras.io/guides/preprocessing_layers/) provides a comprehensive introduction.\n",
        "Our example involves preprocessing labels at the character\n",
        "level. This means that if there are two labels, e.g. \"cat\" and \"dog\", then our character\n",
        "vocabulary should be {a, c, d, g, o, t} (without any special tokens). We use the\n",
        "[`StringLookup`](https://keras.io/api/layers/preprocessing_layers/categorical/string_lookup/)\n",
        "layer for this purpose."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 8,
      "metadata": {
        "id": "SuEDbdPW0DaO"
      },
      "outputs": [],
      "source": [
        "\n",
        "AUTOTUNE = tf.data.AUTOTUNE\n",
        "\n",
        "# Mapping characters to integers.\n",
        "char_to_num = StringLookup(vocabulary=list(characters), mask_token=None)\n",
        "\n",
        "# Mapping integers back to original characters.\n",
        "num_to_char = StringLookup(\n",
        "    vocabulary=char_to_num.get_vocabulary(), mask_token=None, invert=True\n",
        ")"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "T9lc2wM60DaO"
      },
      "source": [
        "### Resizing images without distortion\n",
        "\n",
        "Instead of square images, many OCR models work with rectangular images. This will become\n",
        "clearer in a moment when we will visualize a few samples from the dataset. While\n",
        "aspect-unaware resizing square images does not introduce a significant amount of\n",
        "distortion this is not the case for rectangular images. But resizing images to a uniform\n",
        "size is a requirement for mini-batching. So we need to perform our resizing such that\n",
        "the following criteria are met:\n",
        "\n",
        "* Aspect ratio is preserved.\n",
        "* Content of the images is not affected."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 9,
      "metadata": {
        "id": "vTAVv8QJ0DaO"
      },
      "outputs": [],
      "source": [
        "\n",
        "def distortion_free_resize(image, img_size):\n",
        "    w, h = img_size\n",
        "    image = tf.image.resize(image, size=(h, w), preserve_aspect_ratio=True)\n",
        "\n",
        "    # Check tha amount of padding needed to be done.\n",
        "    pad_height = h - tf.shape(image)[0]\n",
        "    pad_width = w - tf.shape(image)[1]\n",
        "\n",
        "    # Only necessary if you want to do same amount of padding on both sides.\n",
        "    if pad_height % 2 != 0:\n",
        "        height = pad_height // 2\n",
        "        pad_height_top = height + 1\n",
        "        pad_height_bottom = height\n",
        "    else:\n",
        "        pad_height_top = pad_height_bottom = pad_height // 2\n",
        "\n",
        "    if pad_width % 2 != 0:\n",
        "        width = pad_width // 2\n",
        "        pad_width_left = width + 1\n",
        "        pad_width_right = width\n",
        "    else:\n",
        "        pad_width_left = pad_width_right = pad_width // 2\n",
        "\n",
        "    image = tf.pad(\n",
        "        image,\n",
        "        paddings=[\n",
        "            [pad_height_top, pad_height_bottom],\n",
        "            [pad_width_left, pad_width_right],\n",
        "            [0, 0],\n",
        "        ],\n",
        "    )\n",
        "\n",
        "    image = tf.transpose(image, perm=[1, 0, 2])\n",
        "    image = tf.image.flip_left_right(image)\n",
        "    return image\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "kzEJO8kO0DaP"
      },
      "source": [
        "If we just go with the plain resizing then the images would look like so:\n",
        "\n",
        "![](https://i.imgur.com/eqq3s4N.png)\n",
        "\n",
        "Notice how this resizing would have introduced unnecessary stretching."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ODFegPTJ0DaP"
      },
      "source": [
        "### Putting the utilities together"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 10,
      "metadata": {
        "id": "aerzWmQk0DaP"
      },
      "outputs": [],
      "source": [
        "batch_size = 64\n",
        "padding_token = 99\n",
        "image_width = 128\n",
        "image_height = 32\n",
        "\n",
        "\n",
        "def preprocess_image(image_path, img_size=(image_width, image_height)):\n",
        "    image = tf.io.read_file(image_path)\n",
        "    image = tf.image.decode_png(image, 1)\n",
        "    image = distortion_free_resize(image, img_size)\n",
        "    image = tf.cast(image, tf.float32) / 255.0\n",
        "    return image\n",
        "\n",
        "\n",
        "def vectorize_label(label):\n",
        "    label = char_to_num(tf.strings.unicode_split(label, input_encoding=\"UTF-8\"))\n",
        "    length = tf.shape(label)[0]\n",
        "    pad_amount = max_len - length\n",
        "    label = tf.pad(label, paddings=[[0, pad_amount]], constant_values=padding_token)\n",
        "    return label\n",
        "\n",
        "\n",
        "def process_images_labels(image_path, label):\n",
        "    image = preprocess_image(image_path)\n",
        "    label = vectorize_label(label)\n",
        "    return {\"image\": image, \"label\": label}\n",
        "\n",
        "\n",
        "def prepare_dataset(image_paths, labels):\n",
        "    dataset = tf.data.Dataset.from_tensor_slices((image_paths, labels)).map(\n",
        "        process_images_labels, num_parallel_calls=AUTOTUNE\n",
        "    )\n",
        "    return dataset.batch(batch_size).cache().prefetch(AUTOTUNE)\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "H-SrSXS10DaP"
      },
      "source": [
        "## Prepare `tf.data.Dataset` objects"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 11,
      "metadata": {
        "id": "AYJWZfMy0DaP"
      },
      "outputs": [],
      "source": [
        "train_ds = prepare_dataset(train_img_paths, train_labels_cleaned)\n",
        "validation_ds = prepare_dataset(validation_img_paths, validation_labels_cleaned)\n",
        "test_ds = prepare_dataset(test_img_paths, test_labels_cleaned)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "aEqLhzkp0DaQ"
      },
      "source": [
        "## Visualize a few samples"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 12,
      "metadata": {
        "id": "U0CST5a10DaQ",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 433
        },
        "outputId": "d34d9c7e-b024-40bc-e301-00a614e74278"
      },
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 1080x576 with 16 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ],
      "source": [
        "for data in train_ds.take(1):\n",
        "    images, labels = data[\"image\"], data[\"label\"]\n",
        "\n",
        "    _, ax = plt.subplots(4, 4, figsize=(15, 8))\n",
        "\n",
        "    for i in range(16):\n",
        "        img = images[i]\n",
        "        img = tf.image.flip_left_right(img)\n",
        "        img = tf.transpose(img, perm=[1, 0, 2])\n",
        "        img = (img * 255.0).numpy().clip(0, 255).astype(np.uint8)\n",
        "        img = img[:, :, 0]\n",
        "\n",
        "        # Gather indices where label!= padding_token.\n",
        "        label = labels[i]\n",
        "        indices = tf.gather(label, tf.where(tf.math.not_equal(label, padding_token)))\n",
        "        # Convert to string.\n",
        "        label = tf.strings.reduce_join(num_to_char(indices))\n",
        "        label = label.numpy().decode(\"utf-8\")\n",
        "\n",
        "        ax[i // 4, i % 4].imshow(img, cmap=\"gray\")\n",
        "        ax[i // 4, i % 4].set_title(label)\n",
        "        ax[i // 4, i % 4].axis(\"off\")\n",
        "\n",
        "\n",
        "plt.show()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "yJuGrGG80DaQ"
      },
      "source": [
        "You will notice that the content of original image is kept as faithful as possible and has\n",
        "been padded accordingly."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "gkHZOWCh0DaQ"
      },
      "source": [
        "## Model\n",
        "\n",
        "Our model will use the CTC loss as an endpoint layer. For a detailed understanding of the\n",
        "CTC loss, refer to [this post](https://distill.pub/2017/ctc/)."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 13,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "plzcR0sS0DaQ",
        "outputId": "8645405e-f47b-4312-8f86-2c81b6e90ca8"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"handwriting_recognizer\"\n",
            "__________________________________________________________________________________________________\n",
            " Layer (type)                   Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            " image (InputLayer)             [(None, 128, 32, 1)  0           []                               \n",
            "                                ]                                                                 \n",
            "                                                                                                  \n",
            " Conv1 (Conv2D)                 (None, 128, 32, 32)  320         ['image[0][0]']                  \n",
            "                                                                                                  \n",
            " Conv2 (Conv2D)                 (None, 128, 32, 64)  18496       ['Conv1[0][0]']                  \n",
            "                                                                                                  \n",
            " Conv3 (Conv2D)                 (None, 128, 32, 128  73856       ['Conv2[0][0]']                  \n",
            "                                )                                                                 \n",
            "                                                                                                  \n",
            " pool1 (MaxPooling2D)           (None, 64, 16, 128)  0           ['Conv3[0][0]']                  \n",
            "                                                                                                  \n",
            " Conv4 (Conv2D)                 (None, 64, 16, 512)  590336      ['pool1[0][0]']                  \n",
            "                                                                                                  \n",
            " pool2 (MaxPooling2D)           (None, 32, 8, 512)   0           ['Conv4[0][0]']                  \n",
            "                                                                                                  \n",
            " dropout (Dropout)              (None, 32, 8, 512)   0           ['pool2[0][0]']                  \n",
            "                                                                                                  \n",
            " Conv5 (Conv2D)                 (None, 32, 8, 1024)  4719616     ['dropout[0][0]']                \n",
            "                                                                                                  \n",
            " reshape (Reshape)              (None, 32, 8192)     0           ['Conv5[0][0]']                  \n",
            "                                                                                                  \n",
            " dense1 (Dense)                 (None, 32, 64)       524352      ['reshape[0][0]']                \n",
            "                                                                                                  \n",
            " dropout_1 (Dropout)            (None, 32, 64)       0           ['dense1[0][0]']                 \n",
            "                                                                                                  \n",
            " bidirectional (Bidirectional)  (None, 32, 256)      148992      ['dropout_1[0][0]']              \n",
            "                                                                                                  \n",
            " bidirectional_1 (Bidirectional  (None, 32, 512)     789504      ['bidirectional[0][0]']          \n",
            " )                                                                                                \n",
            "                                                                                                  \n",
            " label (InputLayer)             [(None, None)]       0           []                               \n",
            "                                                                                                  \n",
            " dense2 (Dense)                 (None, 32, 81)       41553       ['bidirectional_1[0][0]']        \n",
            "                                                                                                  \n",
            " ctc_loss (CTCLayer)            (None, 32, 81)       0           ['label[0][0]',                  \n",
            "                                                                  'dense2[0][0]']                 \n",
            "                                                                                                  \n",
            "==================================================================================================\n",
            "Total params: 6,907,025\n",
            "Trainable params: 6,907,025\n",
            "Non-trainable params: 0\n",
            "__________________________________________________________________________________________________\n"
          ]
        }
      ],
      "source": [
        "\n",
        "class CTCLayer(keras.layers.Layer):\n",
        "    def __init__(self, name=None):\n",
        "        super().__init__(name=name)\n",
        "        self.loss_fn = keras.backend.ctc_batch_cost\n",
        "\n",
        "    def call(self, y_true, y_pred):\n",
        "        batch_len = tf.cast(tf.shape(y_true)[0], dtype=\"int64\")\n",
        "        input_length = tf.cast(tf.shape(y_pred)[1], dtype=\"int64\")\n",
        "        label_length = tf.cast(tf.shape(y_true)[1], dtype=\"int64\")\n",
        "\n",
        "        input_length = input_length * tf.ones(shape=(batch_len, 1), dtype=\"int64\")\n",
        "        label_length = label_length * tf.ones(shape=(batch_len, 1), dtype=\"int64\")\n",
        "        loss = self.loss_fn(y_true, y_pred, input_length, label_length)\n",
        "        self.add_loss(loss)\n",
        "\n",
        "        # At test time, just return the computed predictions.\n",
        "        return y_pred\n",
        "\n",
        "\n",
        "def build_model():\n",
        "    # Inputs to the model\n",
        "    input_img = keras.Input(shape=(image_width, image_height, 1), name=\"image\")\n",
        "    labels = keras.layers.Input(name=\"label\", shape=(None,))\n",
        "\n",
        "    # First conv block.\n",
        "    x = keras.layers.Conv2D(32,(3, 3),activation=\"relu\",kernel_initializer=\"he_normal\",padding=\"same\",name=\"Conv1\",)(input_img)\n",
        "    x = keras.layers.Conv2D(64,(3, 3),activation=\"relu\",kernel_initializer=\"he_normal\",padding=\"same\",name=\"Conv2\",)(x)\n",
        "    x = keras.layers.Conv2D(128,(3, 3),activation=\"relu\",kernel_initializer=\"he_normal\",padding=\"same\",name=\"Conv3\",)(x)\n",
        "    x = keras.layers.MaxPooling2D((2, 2), name=\"pool1\")(x)\n",
        "    x = keras.layers.Conv2D(512,(3, 3),activation=\"relu\",kernel_initializer=\"he_normal\",padding=\"same\",name=\"Conv4\",)(x)\n",
        "    x = keras.layers.MaxPooling2D((2, 2), name=\"pool2\")(x)\n",
        "    x = keras.layers.Dropout(0.3)(x) #added this\n",
        "    x = keras.layers.Conv2D(1024,(3, 3),activation=\"relu\",kernel_initializer=\"he_normal\",padding=\"same\",name=\"Conv5\",)(x)\n",
        "    # x = keras.layers.AveragePooling2D((2, 2), name=\"pool2\")(x)\n",
        "    new_shape = ((image_width // 4), (image_height // 4) * 1024)\n",
        "    x = keras.layers.Reshape(target_shape=new_shape, name=\"reshape\")(x)\n",
        "    x = keras.layers.Dense(64, activation=\"relu\", name=\"dense1\")(x)\n",
        "    x = keras.layers.Dropout(0.25)(x)\n",
        "\n",
        "    # RNNs.\n",
        "    \n",
        "    # x = keras.layers.Bidirectional(\n",
        "    #     keras.layers.LSTM(256, return_sequences=True, dropout=0.25)\n",
        "    # )(x)\n",
        "    # # x = keras.layers.Bidirectional(\n",
        "    # #     keras.layers.LSTM(128, return_sequences=True, dropout=0.25)\n",
        "    # # )(x)\n",
        "    # # x = keras.layers.Bidirectional(\n",
        "    # #     keras.layers.LSTM(64, return_sequences=True, dropout=0.25)\n",
        "    # # )(x)\n",
        "    #GRU LAYER\n",
        "    x = keras.layers.Bidirectional(keras.layers.GRU(128, return_sequences=True, dropout=0.2))(x)\n",
        "    x = keras.layers.Bidirectional(keras.layers.GRU(256, return_sequences=True, dropout=0.2))(x)\n",
        "    \n",
        "    \n",
        "\n",
        "    # +2 is to account for the two special tokens introduced by the CTC loss.\n",
        "    # The recommendation comes here: https://git.io/J0eXP.\n",
        "    x = keras.layers.Dense(len(char_to_num.get_vocabulary()) + 2, activation=\"softmax\", name=\"dense2\")(x)\n",
        "    # Add CTC layer for calculating CTC loss at each step.\n",
        "    output = CTCLayer(name=\"ctc_loss\")(labels, x)\n",
        "\n",
        "    # Define the model.\n",
        "    model = keras.models.Model(inputs=[input_img, labels], outputs=output, name=\"handwriting_recognizer\")\n",
        "    # Optimizer.\n",
        "    opt = keras.optimizers.Adam()\n",
        "    # Compile the model and return.\n",
        "    model.compile(optimizer=opt)\n",
        "    return model\n",
        "\n",
        "\n",
        "# Get the model.\n",
        "model = build_model()\n",
        "model.summary()"
      ]
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "CzNxzgrSbiqa"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "4PWS-J1K0DaR"
      },
      "source": [
        "## Evaluation metric\n",
        "\n",
        "[Edit Distance](https://en.wikipedia.org/wiki/Edit_distance)\n",
        "is the most widely used metric for evaluating OCR models. In this section, we will\n",
        "implement it and use it as a callback to monitor our model."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "IqGiXq130DaR"
      },
      "source": [
        "We first segregate the validation images and their labels for convenience."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 14,
      "metadata": {
        "id": "ZJEDfRZC0DaR"
      },
      "outputs": [],
      "source": [
        "validation_images = []\n",
        "validation_labels = []\n",
        "\n",
        "for batch in validation_ds:\n",
        "    validation_images.append(batch[\"image\"])\n",
        "    validation_labels.append(batch[\"label\"])"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "d34x-sgz0DaR"
      },
      "source": [
        "Now, we create a callback to monitor the edit distances."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 15,
      "metadata": {
        "id": "2hJXa4Qd0DaR"
      },
      "outputs": [],
      "source": [
        "\n",
        "def calculate_edit_distance(labels, predictions):\n",
        "    # Get a single batch and convert its labels to sparse tensors.\n",
        "    saprse_labels = tf.cast(tf.sparse.from_dense(labels), dtype=tf.int64)\n",
        "\n",
        "    # Make predictions and convert them to sparse tensors.\n",
        "    input_len = np.ones(predictions.shape[0]) * predictions.shape[1]\n",
        "    predictions_decoded = keras.backend.ctc_decode(\n",
        "        predictions, input_length=input_len, greedy=True\n",
        "    )[0][0][:, :max_len]\n",
        "    sparse_predictions = tf.cast(\n",
        "        tf.sparse.from_dense(predictions_decoded), dtype=tf.int64\n",
        "    )\n",
        "\n",
        "    # Compute individual edit distances and average them out.\n",
        "    edit_distances = tf.edit_distance(\n",
        "        sparse_predictions, saprse_labels, normalize=False\n",
        "    )\n",
        "    return tf.reduce_mean(edit_distances)\n",
        "\n",
        "\n",
        "class EditDistanceCallback(keras.callbacks.Callback):\n",
        "    def __init__(self, pred_model):\n",
        "        super().__init__()\n",
        "        self.prediction_model = pred_model\n",
        "\n",
        "    def on_epoch_end(self, epoch, logs=None):\n",
        "        edit_distances = []\n",
        "\n",
        "        for i in range(len(validation_images)):\n",
        "            labels = validation_labels[i]\n",
        "            predictions = self.prediction_model.predict(validation_images[i])\n",
        "            edit_distances.append(calculate_edit_distance(labels, predictions).numpy())\n",
        "\n",
        "        print(\n",
        "            f\"Mean edit distance for epoch {epoch + 1}: {np.mean(edit_distances):.4f}\"\n",
        "        )\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "IuPLB0oc0DaT"
      },
      "source": [
        "## Training\n",
        "\n",
        "Now we are ready to kick off model training."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 16,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "FbNTX7WE0DaT",
        "outputId": "097dc8c1-5df0-4eae-b9f5-d46e985f7fe2"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/26\n",
            "2/2 [==============================] - 2s 10ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "1/1 [==============================] - 1s 1s/step\n",
            "Mean edit distance for epoch 1: 19.7982\n",
            "1357/1357 [==============================] - 226s 153ms/step - loss: 11.6067 - val_loss: 7.2544\n",
            "Epoch 2/26\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 18ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 16ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 18ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "1/1 [==============================] - 0s 30ms/step\n",
            "Mean edit distance for epoch 2: 17.8807\n",
            "1357/1357 [==============================] - 194s 143ms/step - loss: 4.9698 - val_loss: 3.0624\n",
            "Epoch 3/26\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 18ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 18ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 17ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "1/1 [==============================] - 0s 34ms/step\n",
            "Mean edit distance for epoch 3: 17.5720\n",
            "1357/1357 [==============================] - 197s 145ms/step - loss: 3.0535 - val_loss: 2.2728\n",
            "Epoch 4/26\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 17ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "1/1 [==============================] - 0s 32ms/step\n",
            "Mean edit distance for epoch 4: 17.4678\n",
            "1357/1357 [==============================] - 197s 145ms/step - loss: 2.4399 - val_loss: 1.9664\n",
            "Epoch 5/26\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "1/1 [==============================] - 0s 33ms/step\n",
            "Mean edit distance for epoch 5: 17.4334\n",
            "1357/1357 [==============================] - 197s 145ms/step - loss: 2.1207 - val_loss: 1.8804\n",
            "Epoch 6/26\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 17ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 22ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 23ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "1/1 [==============================] - 0s 33ms/step\n",
            "Mean edit distance for epoch 6: 17.3838\n",
            "1357/1357 [==============================] - 197s 145ms/step - loss: 1.8961 - val_loss: 1.7387\n",
            "Epoch 7/26\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 18ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 17ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "1/1 [==============================] - 0s 34ms/step\n",
            "Mean edit distance for epoch 7: 17.2908\n",
            "1357/1357 [==============================] - 198s 146ms/step - loss: 1.7393 - val_loss: 1.5933\n",
            "Epoch 8/26\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 22ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "1/1 [==============================] - 0s 31ms/step\n",
            "Mean edit distance for epoch 8: 17.2888\n",
            "1357/1357 [==============================] - 199s 147ms/step - loss: 1.6220 - val_loss: 1.5762\n",
            "Epoch 9/26\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 18ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 18ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 17ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 18ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "1/1 [==============================] - 0s 34ms/step\n",
            "Mean edit distance for epoch 9: 17.2976\n",
            "1357/1357 [==============================] - 199s 147ms/step - loss: 1.5179 - val_loss: 1.5390\n",
            "Epoch 10/26\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 25ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 22ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 24ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 18ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "1/1 [==============================] - 0s 37ms/step\n",
            "Mean edit distance for epoch 10: 17.2809\n",
            "1357/1357 [==============================] - 199s 147ms/step - loss: 1.4464 - val_loss: 1.5389\n",
            "Epoch 11/26\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 16ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 22ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 18ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 18ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "1/1 [==============================] - 0s 37ms/step\n",
            "Mean edit distance for epoch 11: 17.2736\n",
            "1357/1357 [==============================] - 199s 147ms/step - loss: 1.3909 - val_loss: 1.4959\n",
            "Epoch 12/26\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 16ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 18ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 22ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "1/1 [==============================] - 0s 31ms/step\n",
            "Mean edit distance for epoch 12: 17.2644\n",
            "1357/1357 [==============================] - 198s 146ms/step - loss: 1.3376 - val_loss: 1.4704\n",
            "Epoch 13/26\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 18ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 17ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 18ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 18ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "1/1 [==============================] - 0s 32ms/step\n",
            "Mean edit distance for epoch 13: 17.2539\n",
            "1357/1357 [==============================] - 199s 147ms/step - loss: 1.3014 - val_loss: 1.4504\n",
            "Epoch 14/26\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 18ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 26ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 17ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 22ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 22ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "1/1 [==============================] - 0s 33ms/step\n",
            "Mean edit distance for epoch 14: 17.2781\n",
            "1357/1357 [==============================] - 198s 146ms/step - loss: 1.2662 - val_loss: 1.5178\n",
            "Epoch 15/26\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 18ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 17ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 23ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 23ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "1/1 [==============================] - 0s 32ms/step\n",
            "Mean edit distance for epoch 15: 17.2302\n",
            "1357/1357 [==============================] - 199s 146ms/step - loss: 1.2380 - val_loss: 1.4448\n",
            "Epoch 16/26\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 22ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 16ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 23ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 23ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 18ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "1/1 [==============================] - 0s 32ms/step\n",
            "Mean edit distance for epoch 16: 17.2279\n",
            "1357/1357 [==============================] - 200s 148ms/step - loss: 1.2058 - val_loss: 1.4209\n",
            "Epoch 17/26\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 18ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 18ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 22ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 18ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "1/1 [==============================] - 0s 34ms/step\n",
            "Mean edit distance for epoch 17: 17.2289\n",
            "1357/1357 [==============================] - 199s 146ms/step - loss: 1.1987 - val_loss: 1.4241\n",
            "Epoch 18/26\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "1/1 [==============================] - 0s 32ms/step\n",
            "Mean edit distance for epoch 18: 17.2292\n",
            "1357/1357 [==============================] - 200s 148ms/step - loss: 1.1780 - val_loss: 1.3935\n",
            "Epoch 19/26\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 23ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 23ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 23ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 24ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "1/1 [==============================] - 0s 35ms/step\n",
            "Mean edit distance for epoch 19: 17.2284\n",
            "1357/1357 [==============================] - 199s 146ms/step - loss: 1.1818 - val_loss: 1.4554\n",
            "Epoch 20/26\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 22ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 23ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 22ms/step\n",
            "2/2 [==============================] - 0s 18ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 22ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "1/1 [==============================] - 0s 31ms/step\n",
            "Mean edit distance for epoch 20: 17.2266\n",
            "1357/1357 [==============================] - 199s 147ms/step - loss: 1.1516 - val_loss: 1.4421\n",
            "Epoch 21/26\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 18ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 18ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 23ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 23ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 23ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 18ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 22ms/step\n",
            "2/2 [==============================] - 0s 22ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 23ms/step\n",
            "1/1 [==============================] - 0s 31ms/step\n",
            "Mean edit distance for epoch 21: 17.2114\n",
            "1357/1357 [==============================] - 201s 148ms/step - loss: 1.1564 - val_loss: 1.4101\n",
            "Epoch 22/26\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 22ms/step\n",
            "2/2 [==============================] - 0s 18ms/step\n",
            "2/2 [==============================] - 0s 22ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 25ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 22ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 25ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 23ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 23ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 23ms/step\n",
            "1/1 [==============================] - 0s 31ms/step\n",
            "Mean edit distance for epoch 22: 17.2182\n",
            "1357/1357 [==============================] - 202s 149ms/step - loss: 1.1512 - val_loss: 1.4027\n",
            "Epoch 23/26\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 27ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 27ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 22ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 22ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 23ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 23ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 23ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 23ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "1/1 [==============================] - 0s 32ms/step\n",
            "Mean edit distance for epoch 23: 17.2254\n",
            "1357/1357 [==============================] - 200s 148ms/step - loss: 1.1515 - val_loss: 1.4571\n",
            "Epoch 24/26\n",
            "2/2 [==============================] - 0s 22ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 22ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 24ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 25ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "1/1 [==============================] - 0s 32ms/step\n",
            "Mean edit distance for epoch 24: 17.2219\n",
            "1357/1357 [==============================] - 201s 148ms/step - loss: 1.1431 - val_loss: 1.4609\n",
            "Epoch 25/26\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 22ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 23ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 26ms/step\n",
            "2/2 [==============================] - 0s 24ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 23ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 22ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 22ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 18ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 22ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 22ms/step\n",
            "2/2 [==============================] - 0s 23ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "1/1 [==============================] - 0s 31ms/step\n",
            "Mean edit distance for epoch 25: 17.2324\n",
            "1357/1357 [==============================] - 201s 148ms/step - loss: 1.1499 - val_loss: 1.4792\n",
            "Epoch 26/26\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 22ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 23ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 22ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 22ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 18ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 22ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 22ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 22ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 21ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 22ms/step\n",
            "2/2 [==============================] - 0s 18ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 22ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 20ms/step\n",
            "2/2 [==============================] - 0s 22ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 18ms/step\n",
            "1/1 [==============================] - 0s 39ms/step\n",
            "Mean edit distance for epoch 26: 17.2078\n",
            "1357/1357 [==============================] - 199s 147ms/step - loss: 1.1454 - val_loss: 1.4336\n"
          ]
        }
      ],
      "source": [
        "epochs = 26\n",
        "\n",
        "model = build_model()\n",
        "prediction_model = keras.models.Model(\n",
        "    model.get_layer(name=\"image\").input, model.get_layer(name=\"dense2\").output\n",
        ")\n",
        "edit_distance_callback = EditDistanceCallback(prediction_model)\n",
        "\n",
        "# Train the model.\n",
        "history = model.fit(\n",
        "    train_ds,\n",
        "    validation_data=validation_ds,\n",
        "    epochs=epochs,\n",
        "    callbacks=[edit_distance_callback],\n",
        ")\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "MQKo1LVN0DaT"
      },
      "source": [
        "## Inference"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 17,
      "metadata": {
        "id": "ALW0259Z0DaT",
        "outputId": "67ccd508-a7b0-4a78-9e3b-9f5323493964",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 793
        }
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "len is :  2\n",
            "batch img shape:  (128, 32, 1)\n",
            "2/2 [==============================] - 0s 31ms/step\n",
            "(128, 32, 1)\n",
            "(128, 32, 1)\n",
            "(128, 32, 1)\n",
            "(128, 32, 1)\n",
            "(128, 32, 1)\n",
            "(128, 32, 1)\n",
            "(128, 32, 1)\n",
            "(128, 32, 1)\n",
            "(128, 32, 1)\n",
            "(128, 32, 1)\n",
            "(128, 32, 1)\n",
            "(128, 32, 1)\n",
            "(128, 32, 1)\n",
            "(128, 32, 1)\n",
            "(128, 32, 1)\n",
            "(128, 32, 1)\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 1080x576 with 16 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ],
      "source": [
        "\n",
        "# A utility function to decode the output of the network.\n",
        "def decode_batch_predictions(pred):\n",
        "    input_len = np.ones(pred.shape[0]) * pred.shape[1]\n",
        "    # Use greedy search. For complex tasks, you can use beam search.\n",
        "    results = keras.backend.ctc_decode(pred, input_length=input_len, greedy=True)[0][0][\n",
        "        :, :max_len\n",
        "    ]\n",
        "    # Iterate over the results and get back the text.\n",
        "    output_text = []\n",
        "    for res in results:\n",
        "        res = tf.gather(res, tf.where(tf.math.not_equal(res, -1)))\n",
        "        res = tf.strings.reduce_join(num_to_char(res)).numpy().decode(\"utf-8\")\n",
        "        output_text.append(res)\n",
        "    return output_text\n",
        "\n",
        "\n",
        "#  Let's check results on some test samples.\n",
        "for batch in test_ds.take(1):\n",
        "    # print(batch[\"label\"])\n",
        "    batch_images,batch_labels = batch[\"image\"],batch[\"label\"]\n",
        "    print(\"len is : \",len(batch))\n",
        "    print(\"batch img shape: \",batch_images[1].shape)\n",
        "    # print(batch)\n",
        "    _, ax = plt.subplots(4, 4, figsize=(15, 8))\n",
        "\n",
        "    preds = prediction_model.predict(batch_images)\n",
        "    pred_texts = decode_batch_predictions(preds)\n",
        "\n",
        "    for i in range(16):\n",
        "        img = batch_images[i]\n",
        "        print(img.shape)\n",
        "        img = tf.image.flip_left_right(img)\n",
        "        img = tf.transpose(img, perm=[1, 0, 2])\n",
        "        img = (img * 255.0).numpy().clip(0, 255).astype(np.uint8)\n",
        "        img = img[:, :, 0]\n",
        "\n",
        "        label = batch_labels[i]\n",
        "        indices = tf.gather(label, tf.where(tf.math.not_equal(label, padding_token)))\n",
        "        # Convert to string.\n",
        "        label = tf.strings.reduce_join(num_to_char(indices))\n",
        "        label = label.numpy().decode(\"utf-8\")\n",
        "\n",
        "        title = f\"Prediction: {pred_texts[i]} \\n\\n Original : {label}\"\n",
        "        ax[i // 4, i % 4].imshow(img, cmap=\"gray\")\n",
        "        ax[i // 4, i % 4].set_title(title)\n",
        "        ax[i // 4, i % 4].axis(\"off\")\n",
        "\n",
        "plt.show()\n",
        "\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Precision\n",
        "\n",
        "total_count = 0\n",
        "correct_count = 0\n",
        "\n",
        "for batch in test_ds.take(75):\n",
        "    batch_images,batch_labels = batch[\"image\"],batch[\"label\"]\n",
        "\n",
        "    preds = prediction_model.predict(batch_images)\n",
        "    pred_texts = decode_batch_predictions(preds)\n",
        "\n",
        "    for i in range(64):\n",
        "        label = batch_labels[i]\n",
        "        indices = tf.gather(label, tf.where(tf.math.not_equal(label, padding_token)))\n",
        "        # Convert to string.\n",
        "        label = tf.strings.reduce_join(num_to_char(indices))\n",
        "        label = label.numpy().decode(\"utf-8\")\n",
        "\n",
        "\n",
        "        title = f\"Prediction: {pred_texts[i]} \\n\\n Original : {label}\"\n",
        "        total_count = total_count+1\n",
        "        if(label == pred_texts[i]):\n",
        "          correct_count = correct_count+1 \n",
        "\n",
        "\n",
        "print(\"Total count : \",total_count)\n",
        "\n",
        "print(\"Correct count : \",correct_count)\n",
        "\n",
        "percentage = (correct_count/total_count)*100\n",
        "print(f\"Percentage : {percentage}% .\")\n",
        "        "
      ],
      "metadata": {
        "id": "OGtQI1IDyLiG",
        "outputId": "466469e1-8a89-4500-9879-34a13eec6ec4",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "execution_count": 18,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2/2 [==============================] - 0s 31ms/step\n",
            "2/2 [==============================] - 0s 30ms/step\n",
            "2/2 [==============================] - 0s 30ms/step\n",
            "2/2 [==============================] - 0s 29ms/step\n",
            "2/2 [==============================] - 0s 30ms/step\n",
            "2/2 [==============================] - 0s 34ms/step\n",
            "2/2 [==============================] - 0s 27ms/step\n",
            "2/2 [==============================] - 0s 26ms/step\n",
            "2/2 [==============================] - 0s 31ms/step\n",
            "2/2 [==============================] - 0s 31ms/step\n",
            "2/2 [==============================] - 0s 30ms/step\n",
            "2/2 [==============================] - 0s 19ms/step\n",
            "2/2 [==============================] - 0s 31ms/step\n",
            "2/2 [==============================] - 0s 28ms/step\n",
            "2/2 [==============================] - 0s 30ms/step\n",
            "2/2 [==============================] - 0s 26ms/step\n",
            "2/2 [==============================] - 0s 31ms/step\n",
            "2/2 [==============================] - 0s 30ms/step\n",
            "2/2 [==============================] - 0s 29ms/step\n",
            "2/2 [==============================] - 0s 28ms/step\n",
            "2/2 [==============================] - 0s 28ms/step\n",
            "2/2 [==============================] - 0s 30ms/step\n",
            "2/2 [==============================] - 0s 29ms/step\n",
            "2/2 [==============================] - 0s 29ms/step\n",
            "2/2 [==============================] - 0s 29ms/step\n",
            "2/2 [==============================] - 0s 29ms/step\n",
            "2/2 [==============================] - 0s 30ms/step\n",
            "2/2 [==============================] - 0s 30ms/step\n",
            "2/2 [==============================] - 0s 29ms/step\n",
            "2/2 [==============================] - 0s 29ms/step\n",
            "2/2 [==============================] - 0s 30ms/step\n",
            "2/2 [==============================] - 0s 31ms/step\n",
            "2/2 [==============================] - 0s 29ms/step\n",
            "2/2 [==============================] - 0s 29ms/step\n",
            "2/2 [==============================] - 0s 31ms/step\n",
            "2/2 [==============================] - 0s 30ms/step\n",
            "2/2 [==============================] - 0s 31ms/step\n",
            "2/2 [==============================] - 0s 30ms/step\n",
            "2/2 [==============================] - 0s 30ms/step\n",
            "2/2 [==============================] - 0s 29ms/step\n",
            "2/2 [==============================] - 0s 29ms/step\n",
            "2/2 [==============================] - 0s 31ms/step\n",
            "2/2 [==============================] - 0s 30ms/step\n",
            "2/2 [==============================] - 0s 30ms/step\n",
            "2/2 [==============================] - 0s 29ms/step\n",
            "2/2 [==============================] - 0s 29ms/step\n",
            "2/2 [==============================] - 0s 30ms/step\n",
            "2/2 [==============================] - 0s 30ms/step\n",
            "2/2 [==============================] - 0s 31ms/step\n",
            "2/2 [==============================] - 0s 28ms/step\n",
            "2/2 [==============================] - 0s 31ms/step\n",
            "2/2 [==============================] - 0s 28ms/step\n",
            "2/2 [==============================] - 0s 30ms/step\n",
            "2/2 [==============================] - 0s 30ms/step\n",
            "2/2 [==============================] - 0s 31ms/step\n",
            "2/2 [==============================] - 0s 29ms/step\n",
            "2/2 [==============================] - 0s 29ms/step\n",
            "2/2 [==============================] - 0s 30ms/step\n",
            "2/2 [==============================] - 0s 29ms/step\n",
            "2/2 [==============================] - 0s 30ms/step\n",
            "2/2 [==============================] - 0s 31ms/step\n",
            "2/2 [==============================] - 0s 28ms/step\n",
            "2/2 [==============================] - 0s 31ms/step\n",
            "2/2 [==============================] - 0s 31ms/step\n",
            "2/2 [==============================] - 0s 30ms/step\n",
            "2/2 [==============================] - 0s 31ms/step\n",
            "2/2 [==============================] - 0s 29ms/step\n",
            "2/2 [==============================] - 0s 29ms/step\n",
            "2/2 [==============================] - 0s 28ms/step\n",
            "2/2 [==============================] - 0s 30ms/step\n",
            "2/2 [==============================] - 0s 30ms/step\n",
            "2/2 [==============================] - 0s 30ms/step\n",
            "2/2 [==============================] - 0s 31ms/step\n",
            "2/2 [==============================] - 0s 31ms/step\n",
            "2/2 [==============================] - 0s 26ms/step\n",
            "Total count :  4800\n",
            "Correct count :  3726\n",
            "Percentage : 77.625% .\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "-DokrqTe5uaU",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 151
        },
        "outputId": "9472ac15-00eb-4ef4-cc02-a439aa3d00c3"
      },
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXAAAAB1CAYAAABXo7o4AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nO2dWYxkWVrffyf2PSL3vSqrsrqquqe7pwcBBo1lWWDkASPGD9ZoMMYgj9QvtgxjJDOYB2w/jWULg2UMagFmsBADHrAZIbAHj0GIB8bN0K2iFrqrqiurct9i37fjh4jv9I2oXCIyIiszss9PSmVmZGTcc+4993/P+c63KK01FovFYhk9XOfdAIvFYrGcDivgFovFMqJYAbdYLJYRxQq4xWKxjChWwC0Wi2VEsQJusVgsI8pAAq6U+pRS6j2l1COl1BeG1SiLxWKxnIw6rR+4UsoNvA98D7AOvA38oNb6/vCaZ7FYLJaj8Azwv98OPNJafwCglPoy8GngSAFXSl34qKFYLMbS0hKBQOC8m2KxWM6QcrnM2toa2Wz2vJvSC/ta66nuFwcR8AVgzfH7OvA3ut+klHoTeHOA47xQPvnJT/JzP/dz3Lx587ybYrFYzpD79+/z+c9/nq997Wvn3ZReeHrYi4MIeE9ord8C3oLRmIFbLJazp9lsorVGKWW+LP0zyCbmBrDk+H2x/ZrFYrEcidaaRqNBo9Gg2Wyed3NGmkEE/G3gJaXUNaWUD/gs8NXhNMtisVw2tNZGvMvlMvl8nnK5bF639M+pTSha67pS6p8B/xtwA7+qtb43tJZZLJZLRbPZpNFoUCgUePjwIalUivn5eW7cuIHP58PlcllTSp8MZAPXWv8B8AdDaovFYrmkaK1pNps0m01KpRKbm5tsbW3h8Xi4evUqXq/3vJs4kpz5JqbFYrForclms6RSKXK5HJVK5dD32Bl4f1gBt1gsZ06z2WRtbY379+9Tq9WeE3BrAz8dVsAtFsuZo7WmXC6TyWRoNBod9m4r3qfHJrOyWCxnykkCbX3BT48VcIvFcuYc5ypoxfv0WAG3WCznihXv02MF3GKxnCky85aZtsv1oeyIeFsRPx1WwC0Wy5njFPHu1+R1K+L9YwXcYrGcGc1mk1qtRq1Wo16vd/zNCvbgWDdCi8VyZjQaDSqVCtVq1SSucgp3t0nF0h9WwHtAa02tVqPZbOJyufB4WqfNDjyL5WgkfL5Wq1GtVmk0Gke+187GT4cV8B4oFAp88MEHpFIpxsbGmJ2dxev1EgqF8Pv95908i+XCUiwWefbsGYVCgVQqZYN2howV8B6oVqs8efKEtbU1lpaW8Pv9hEIhvF6vFXCL5RjK5TI7Oztks1lKpdJ5N+fSYQW8ByQIQWtNpVIhm81Sq9UIhUJEIpHzbp7FcmFpNBodJhRJWOXxeHC73Xg8Hms+GQAr4H2SyWR4+PAhwWCQYDDI+Pi4HYAWyxFI+thSqWQ2MT0eD7FYDL/fTzgctvfPAFgB7wOlFLVajVwuZzKqHWbTswPSYmkhFXicLoRKKWN+9Hq99n4ZACvgPeB2u4nFYkxMTFCtVimXyzQaDXK5HHt7ex3vdblcuN1ulFK43W5cLpfxXBGXKWfQgh28lsuOMw+KiPfk5CRjY2MkEgncbvc5t3B0OVHAlVK/Cnw/sKu1frX92jjwW8AysAp8RmudOrtmni8ej4exsTGTlL5SqdBoNEin02xubnaIsMfjwePx4HK58Pv9uN1ufD4fgUDACPlhbohWyC8mh0UQWnrHWYlHxrvP52Nubs54c1kBPz29zMB/DfjPwK87XvsC8HWt9ReVUl9o//6Tw2/e+SIRZOVymXq9TrPZNDe01ppqtUqhUOgQYtmYcblc1Ot1PB4P1WqVer1uBFxm6PJ/J4URO/9HvluGi1xrl8tl6jNaBuO4cep2u4142/F8ek4UcK31nyqllrte/jTwt9s/fwn4Ey6RgMusYWtri42NDcrlMslkklKpRK1WMwEJBwcH5HK5jv91mkjEfCJfzte7zSjd0WnOz5uYmGBiYgKfz0cikbCui0Om2WyyubnJxsYGgUCA69evk0gkADvzHpTu8ydeKGJqtA/KwTitDXxGa73V/nkbmBlSey4MzWaTVCrF06dPqdVqlEql53I55PN5s7PupNdghaMEXBDhr1QquN1ugsEg4XDYCvgQcF6jZrNJOp1mdXWVaDTK3NycEXDL2eCc1FhOz8CbmFprrZQ6UrGUUm8Cbw56nBeNUopoNMrs7CzVapVsNku1WjWbmABjY2OEQiHzfjhdeajDBFxCkOXvqVSKQqFAo9EwQm43gAZHVlviKSHXt1Qqmf0Ly9lgxXtwTivgO0qpOa31llJqDtg96o1a67eAtwCOE/qLhNia5+fnGR8fp1wus729TT6fJ5VKsb29jcvl4saNG6ysrAxlIDofABIwlMlkqFQqbG9v8+TJE7TWxmZ4/fp13njjDfMAsfSPuLhJwqVSqYRSinQ6jdfrJRwOm/0My3Cw5dOGy2kF/KvAjwBfbH//vaG16ALh9/vx+Xx4PB7C4TBaa4rFItBaAobDYcbGxoY6GGUGXyqVaDQaZqNHItnEjFMoFA4131h6QynVsSHtFBWZkUvkoGU4HCbeVsgHoxc3wt+ktWE5qZRaB36GlnD/tlLqc8BT4DNn2cjzQgaXz+cjFovh8/me8zo5i2NqrfF6vcRiMSPY8XicYrHI2toa6XT6zI7/UUI207xeLwsLC8bFc3JykkgkYr1RhojL5SIQCBAMBolGo3i93vNu0qWgFy+UHzziT9895LZcSCRvQywWIxwOk06nX4j/tsfjIRqNorUmFouxtLREOp02X5bBEVOZ2+1mYWGBubm5Y72DLKdHKUUwGCSRSBCLxUwshGUw7FnsAZmpyVL7RRzPibhbOd2unDZzKzKDI14RlrOj2WxSr9c7klpZBsMKeI/IzX0RbvKL0AZLf3zUIjqdGTxl8lMul01EZrdLruV0WAHvke6ltd3csvSCc5x8VETcKd7O/otoO8urWQbDCniPOCMrXyRaa1KpFMlkklwuZ7xP7APkYlKv1020bjabJZ/PAx9OACQp2mX236/X6+zt7ZHNZtnf36darZpJj3j3WIaDFfAeeNE2cCfNZpP19XXu3btHtVqlVCo9N7OxnD9yPSToq1KpmCpOgPEnX1lZIRaLvbAoxOPGyVkdv1ar8ejRI548eWIKOgBGvCWvkGVwrID3yYsScNnwqdfrlMtlE4UpWQ5lF/+yL8dHBREnieSUaE554MoGdLFYNJG8Z5GJr9FomBWaUyT7zVvvzJrZL81m04xZZwZOWZ3YycfwsALeB84B3x0EMmzy+TzPnj2jWCyyvb1No9EgEAiwvLxMIpFgbGxsJHKidOeCHta5GoY9eVhtq9VqbG1tkc1mKRaLpNNpqtUqtVqNRCJBo9GgWCya99XrdYLBINeuXWNmZqan4/YiwM52VKtVisViXzNd5x7P4uIiS0tLfT1g5Hw6jxmJRFhcXMTn87G9vc3u7q6ddAwRK+A90h25B5zpMrBQKPDs2TMymYzZvff5fFy9epWlpaWOdLQXle6NrLO4cQd9gA6jbfV6nf39fba3tykWi6RSKZrNpvF5FjGt1Wrs7e2xt7dHJBJhbGyM6enpE4/fbTI7yoVUPn9ra4tSqUQ6ne7b20N8430+HwsLCz0LuFO8nXs04XCYK1euEAwGqVQq7O3tWQEfIlbAT4lsyFQqFbNMHFRQG40GhUKBcrlMOp2mUqlQr9cJBAJEo1Gi0agpDDEqQSbFYpF8Pm/cySTfdiQS6Wt2J+YJsalKObvjzoFEs3ZH/Ym4SIHqRqNBJBIhEomc6pw6TSdKKVMnNRaLEYlETJ5xEfJcLke9XieTybC7u4vf7ycajR5qshD3u2w22zFh8Hq9JqKxWCxSLBYplUrkcjkqlQpKKeLxeMd5PwoR31qtZjJsFotFkskkPp+PcDh8ZFIv54NF+lYoFKhWqyZ2wePxdJiLznr1+lHCCngfOAdbs9mkUCiYQR6PxwcW8FqtxuPHj00O8kwmQ7PZ5MaNG9y4cQO/308sFhsZ8dZas729zf3796nX6yavzNTUFLdu3SISifT0GVprSqUS+/v7Zha3u7tLs9l8LgmYk/Hxcd544w0mJibMe5zvTSaT3Llzh1KpxEsvvcStW7dOlaO62WySy+XY399ndnaWl19+mXA4jNfrxev1orU2G3ePHz/m7t27VKtVHj9+zNbWFtPT07z22mtEo9FDP39vb487d+4Y27n07dVXXyUej7O+vs77779vEnJVq1UmJia4efMmwWAQn8+Hz+c7dMxI2xqNBvv7+9y/f598Ps/29ja1Wo1wOMzNmzeZnJw89hw4r7W0w+v14vP5CIVChEKhjgeUFfHhYAX8FDhnLHJTDWJOkc+r1+tks1n29vZMOlmXy0UoFGJqasrM8odtRz7JU2GQ4xWLRfb396nVambzNRAIHJlHvbst8lqj0aBUKhkbs5wjp4Af5p4mHhCHHaNSqZBMJsnn8ywsLJggk35FRcaCrMbGx8eNp0n39drb28PlctFsNslmsxQKBfx+/7GmDmlnoVAwx5PXG40G+Xye3d3djr663W7i8TiRSIRgMIjf7z9SwJ0rG2lbqVQilUp1eJGchDxkq9WqmXFLNanDHoxynq2Inx4r4H3idCk8ODig0WiQSCSIRqN9bSqKQDebTTKZjJldJpNJs8l15coVAoEAMzMzfYu3U6i6Z55SwECW5dKOwz4jGAyaakDHzeROaovH42F2dtYUspUludP9LplMUi6XjRhKm0RkMpmM+VsoFMLtdjM+Pk4oFCKTybC1tUW1WjXHPaydR2XCk/PQb7rTbhu/M+XBYdfLmRhNSu2d5Bcdj8e5deuWEchMJkO9Xufp06fs7u6aTW5Z3USjUcbGxohEIuaheVR/xOYNnRG+0WiU+fl5QqEQwWDwyLaJ2DcaDbNX4/F4TCrmeDz+3MPD6Q/uPL6lf6yA94lzwO3u7rKzs8P8/DxXrlw5cgl8GOJqVavV2NjY4MGDB8bm3Wg0mJqa4uWXXzYl1Hpx6equMuMUbOdMttFosLm5yerqKvV6nWKxaGaA3SaJiYkJbt26ZWzwp10FeDweFhcXuXr1Kh6Pp0PAxc67vr5OKpUin8+TyWQ6hM0plKFQiHA43LG8f/r0qXn4ST+629gdSdst4OKj3O/GnXyms3zeUZ8RCAQYHx83NmYRveNWQePj40SjUSqVCnfv3qVcLlOtVnn48GFH6lu/38+VK1dYWloy+cx72S+RB45zjCUSCZaXl/H7/cfmnK/X6xQKBWM2aTab+P1+lpeXuX79ujkXItZy3mRmLzN0y+mwAt4nHo+HYDBo7LK1Wo16vX7sDSjCIGYSwAhn94aczFZk5iOblifhFGg5jvjcOv2C5YaXm06O3y18clOL6cLtdg/stugM4hBxlnYVCgVKpZIpIO0UxO7sgFKRKBgMopTqEECllFkpBIPBnsRBZvhizuhHVA4z+5zkUeJ8oJ4UlCXXQh46IrLNZrMjJF3OjawkZDx1Jz/rbouzD+KzDq1xLvnwjxt/MhER8RZzjXwJzoexjKlCoUAgEDh2hWA5HivgfTI2NsbHPvYxSqUST548YWtr68QZjmwKVSoV8vm88c+tVqvmppyfn8fj8RCPx82mTz+eGkopY2aoVqscHBywt7d36CxWBFzC8rsDPtxuN9PT08TjcWq1Guvr62ituXHjhilgcdIN1y1M1WrVuEU6iyU4TSi5XI5qtUo8Hmd+fh6v10soFOpYgougeb1e6vU66XSara0tUqmU8XwQV0vJPX0cInjr6+vk83mi0Sgf+9jHTty0c/bxOAHufn+5XCaVSlEqlahUKj39nwixUop6vW5s3xK0I0LrdrtZW1szvtbOiu/d1+uwdheLRSqVilkhBYPBE4ONyuUyT58+JZVKEQ6HuX37Nn6//7maovIAkgfF+++/z9raGouLi9y8eXMkYhouIlbA+0BrTSQSYWlpydgjd3Z2ThQz8ZwQr5VMJtNxA83OzjIzM0MwGGR2dpZEInFqdzYR5p2dHWMiOSwA6ThcLhexWIyZmRkODg5YW1ujWCwyPT1Ns9nsO7hD2ua034rd9LDZ3fj4OLOzswQCARKJBOFw+FABKhQKrK+vs729bWbubrebyclJbty4caipx9l/5ww0lUqRSqWYmJjg+vXrffWvHxEXV71SqdTzMeD5GbbzwSuumbIvI/sAvawE5P+dx5HNR/EaOo5qtWq8glZWVpibmzMrpKOOW6lU2NzcNFHFN27c6OtcWD7ECngfOM0LR7mbiaeE029ZNgyr1aoxizhNBBLw4fP58Hq9fYl3o9Egk8kY/1/xHy8Wi2Z2PzY2duxGlCBmF0n3eXBwQC6XM94ZgBGO02z0iR1UNtvEPtr9QJiamjIz76OW1zKjk1QDXq+Xubk5fD4f0Wj02E07aVMgEGB6eppgMEg+nyeXy/XUn+5+9Sre0NocXFxcNKsv6e9JFWq6jyGzXDEXBQIBI469eo0chXiwHHUOtdak02lTI1ZWEiLIXq/3uXtDKUUikWBpaaljnDrNftaM0j9WwPvEaRs+zC3q4OCAZ8+eUalUSKVSHeYSt9vNtWvXzCaT2P8k0KF7I+k4RDiq1SqPHj3i0aNHxrbs9DuWZe3S0tKhn+GkUCiYG2tjY4ONjQ3zIBL3Mul3vxuZYsqp1WpMTU1x+/ZtwuGwKbPlROyvJ3koiAtdOp1maWmJ119/nUgkYjbvTiIej/Paa69RLpd59OgRjx8/7tkHvHuDGE52uVRKMTs7awJs5P/EVHTcsZzfAWKxGG+88Qbj4+PmehzWHmdbj/LAcX6u9MHv9x957uv1Ok+ePOHhw4cmOEn6EQ6H8fv9z51Hl8vFlStXmJqaIp1O8+6777Kzs2OOL5MEK+L90UtNzCXg14EZQANvaa1/Xik1DvwWsAysAp/RWqfOrqkXA+csvHtXXWaDMhvO5XJmqey8MSQ6MBQKPTfzOmkAy7JXZsKyJE8mkwAdM3hZBkcikedskt2I7Vvc0ZRSZobU/b5+ZpzdKKXwer1EIhGi0eiJbmrHtVdu/EajgdvtJhaLHRnReBgej8e42smqyPn50t5hIpt7p9n4dBZZlpJ7Y2NjJx6zn2Od9Dky5gqFAqlUyjwYZCV12IpKCAQC+P3+jjHmTNpmPVL6p5eRXgd+Qmv9l0qpKPBNpdQfAT8KfF1r/UWl1BeALwA/eXZNvXjIgM7lcjx48IBgMEgmkzF1MxcWFgiFQh2uZbJkl8Haz81Uq9VMbpRisUgmk6FSqbC/vw+0lufLy8sd9kefz3eieAPm4SI2Z60109PTZLNZ1tbWOvyre8XpU+33+5mZmSEWixmTkaxA+kUeluJt4Ty//ZxTsR07Z/q9PpyOMy8Mm1wux87Ojgmu6Scf/LDEGyCVSrG+vk6xWDSRsJLrJBqNMjU1hdvtPtIc0m2CdLvdpNNp7t69a+IeJDeMpTd6KWq8BWy1f84ppR4AC8CnaVWrB/gS8Cd8xAQcWjeICLhz0EYiERYWFlhcXDReE06792mo1+tmEzSZTLKxsWGCQbTWhMNhbt26xdTUlGnbccv67htNvBmgZV4Qf/G9vb2+BVxs1ILX62V+fp6FhYVjc3/0gph1xI7sDKDpZwYnAi7/e5ipohe6TRf9cpKo5nI5njx5Yh7avR5rmOINkE6nuXfvHvl83sQrBINBXnrpJWZnZ48da87XnSlmZSwHAgHC4TBTU1PWjNIHfd1BSqll4BPAN4CZtrgDbNMysRz2P28Cb56+iRcX54AV27DYdCORCIFAoMO2fdqBKf7R5XK5ww0xFAp1JBmSjdB+XA+Pek1mss4Zrdjy4cMCBf0cR2bOg+YyF1e6QXJLdx/beS2dtv6TbNqHPZQPszmfdPzjcLvdJv2AhNP3yyBmExFrOeeS2liShclKqtdr4Xa7jVlPkp1JXEQ2m8Xtdvfsw/9Rp2cBV0pFgN8Bflxrne2yF2ql1KFXT2v9FvBW+zMuVSZ3Z14P8bBw+rVKtrh+PDa6aTQarK2t8fjxY2q1mgkempyc5OMf/7gRbPEAEPPHoLMYpzDJZ0lek0AgYHKz9PI5gEkZIB4OsVjsVO3SWlMsFo1bpnhADGvWJhvDkozpqBwi8OFMsvshJzbd0yTGOox4PM7KygrlctmUauuFYZwTSZVbLBY5ODgwXkrXrl1jeXnZZMrs53jBYJDbt2+zvLzM+vq6SXa2trZGPp8nkUiwsrLSU7Kzjzo9CbhSyktLvH9Da/277Zd3lFJzWustpdQcsHtWjbyIiCh7PJ6OzaVYLMbCwoLZnBxUTLXWZLNZNjY2jKlEa83U1BQzMzNmI3TYqToP+99arWbcE/upayibVRLI5Pf7B0r+Je0olUpDr24uAix5UY4LMJEVSvdMXTaYh5Wv3e/3MzExQblc7tjw7WWmPyiS6ySbzZpoS6UUY2NjLC0tncpzxOv1mkCpQqGA2+2mVquZUnTNZpOrV6+eRXcuHb14oSjgV4AHWuufdfzpq8CPAF9sf/+9M2nhBcXlcjE5OWncqLqzwQmDzLylqouETHu9XuMjPTMzY2bfx5lCToOzKG0ymTS2ZgkQikQijI+PmwCbk0wF4oIo/vGBQODUhW2VaoXSj4+P4/P52NpqWfHkPEkkYT/Lb6cpRAKOarUaY2NjPW20ut1uwuEwiUSCZrPJ2tqa8TGXVLaD4Nz4k4eC1tr0WQJvholEekrk6MHBAVpr4wJ72mCzbqLRKNevXze5zGVVJSbDQUq7fRTo5cx8Evhh4K+UUu+2X/tXtIT7t5VSnwOeAp85myZeTLxeLwsLC0xMTBixcwZnDIqklpUkQVprs/SUsPvD8lR0bx6eBvEtf/z4cUc6UYlWjMfjzM3NEY/HjxUP5+aiLP0l9HuQyuTifpjJZPjggw/MrFlqMJ7GfiptrdVqrK6u4vF4uHbtGtPT0z0JeCKRMCuDe/fuARg/7UGFzuVy4fV6O2b1Iq7FYtH4bA9zJu50id3c3GRzc9P42ksemmGsMKampojFYpTLZd5991329/cpFArk83k8Hs9zecQtnfTihfJnwFEj47uH25zRQSllXNAOC1wYFNk0KpfLJqJS8lOcZBscZNYvOc5lc0nKaymljOeH2IjL5bIRcDmm01UMPozOk80w2QQb5CEjs1GJQpR9BmlXd1bAk85H94OmVquZ2W0v7ZSxEAqFzDWTJE9yjvqNsD3sGHJeZQNZqgDJOR5WIIzM7qX98rAVQR2mqMoM21msW66lzMCH2bfLhn20DYCIgwzmYfoAFwoFnjx5QqFQwOv1srKyYlKoDhtpt+RPkU1CaM12X3rpJUKhEFtbW2xtbeF2u1ldXTV5YOQ8yEwxFAqxsLBAMBgkHo9z+/Ztk/xLPncYSI3QaDRKvV7n8ePHps3hcNi047goR2iJYyQSYWZmpiPM2xk8c5x4iGlLXCPFZ/vZs2cUCgVisRgrKyun3rgVJO3C4uIiAM+ePWN9fZ2ZmRkWFxfxer0mAdVpkCAdSXH87NkztNYkEgkmJiaYnJw0JqVhTVjkvMqK1ufzmeOvra0xPz9v+nbauIHLjD0bAyA2SWe6z2GJeKlUYnNzk0wmw/Xr11lcXDw07HxQnJueyWSSBw8edCRaikQi3Lhxg0QigdvtNlGmGxsbpvai3ITiWiYVaTweD+FwmEgkYjxYDg4OhtZ2yX8yPj7O+vo6jx8/plQqEY/HTZTixMTEiQIOGLu61HUUG3AvuN1uExEpqXcbjQbb29tsb28zNTXF7OzsUAQ8FosxPT1NLpfj6dOnxqXU6c43iIDLvsve3h6PHj0iEAjw+uuvMzc3RyAQMGa7Yc+GPR6PyYC5u7vLN7/5TXK5nDm3kmPFCngn9mwMiAxkEULxQBjEC6H7ISDuisM01VQqFTPLlGAgKfALLdc1ETVZ/odCISYmJkxSLmdWPGdBhHK5TDKZ7KhwI8tx+DCDXblcNrlLTiMISilTczIcDjM5OWmOUalUKBQKZkPSmW9GxMD5OWICEXNVv66f8l6/329mqvl83mzKHRwcGPv8ccm2upEgITE/SbZJyeYoppp0Ok0wGOzpYdX9+eLf7Uzp22g0zBiQeAM5f2dlyvB4PDSbTVP0QsZdMpnE6/VSLpdN4q5+i2JfVqyAD4Bzw9CZr8IZ7DLoIBOvC8k8N4wZSLPZ5ODggN3dXcrlsvHzlYK4fr+fmzdvsry8bIRNKcXMzAzxeNz0U1wDi8Ui1WqVJ0+esLm5STKZ5O7dux1tlST+cuy3334bv9/PrVu3uHnz5qHueCchAS4SQTo2Nka5XObhw4esrq5SKBTIZrMmz/r4+Dh+v5+FhYWOHCJKKZOfRkqzndZ3f2pqim/7tm+jXC5z//593n//ffL5PHfu3MHv93Pt2jVeffXVI6u8H4a48cnKZ3Nzk3q9bsbZwcEBxWLRPAzj8XjPn91sNtnf3yedTlMoFNjc3KRSqTAxMcErr7yC3+9ndnaWSCQyUBTxSchDVMwzLpeLSqXC1tYWd+/epdlsGq+rxcVFXnvttb4fVpcRK+AD4pyBO2dL/YZ1H/f5Ug3HaarpZwbXjczacrmcMW3k83nzdxG8mZnO4NpgMPicH7J8TqlUMv7dErRzFLLR5/F4WFpaGsj0JA8JEXMpHCEz1nK5jFLKJFAKBoPPVbKRzwkGg1QqleeuWz9CLsmqqtUqq6urJjVvMpnE5XIxMTFx7CZu93iCDz1OZHO5UCiYv4nQSZ+cG6/d7T7smLJZLg87SQ87Pj5uyvkNYlfvBxHuQCDA2NgYtVqNnZ0dk0Nexn8kEjm2CtZR1+usEpSdJ1bAh4jMLt977z1CoRBLS0s9ZYs7CgmY2dnZ4e7du4RCIebm5vqqvSk3geT1lodMMpk0Sbfm5ubMzEe8O3ppt2zgSnKuq1evHrrJKl4N29vbpNNpE6UqmRJPk572sLZIZOT8/LwJHpIbvVqtkk6nyeVy1Ot1tre3zf/JKkfMHjKzHdRTZm5uzvi/b21tUSgUODg44N69e6YNNS8AAAx8SURBVB3FKuQculwustmsaaOU3CuXyxQKBVP3cnl5ucM7SFZTlUrF2MWj0Sizs7MdRahrtZoxkchn1mo1U4dUqhlJ0jWJJH7Rpgpnjpr5+XmgtQrZ3t4mm82STqd58OCBiZIVU5oUUBZvLcDUDxWzXrPZZHp6+tIkzbICPiRk5rm9vU0ymSQWixGLxQYScDFVbGxssLOzQyKRIBQKEY1Ge56Fi2nj2bNnHYIm+S3GxsZMAiyv12tunF5NNc4bPBqNsrKy8tx7tNbk83nefvttstmsqTwUCoWIxWJD8WGWG9nn87G8vMzS0pIRv3q9zgcffMCDBw/Mg0TeL8IvVZHkhh90M9rtdnP16lUWFhZMnEAul2Nra4uDgwP8fj8rKyum8pBs0KVSKdbW1iiXyyZdgIwDr9fLzZs3WVpaMtXtXS4X9+/fZ29vj1KpxKNHj1hdXWVpaYl4PN5hz69Wq+zu7pLL5Ugmk+zs7Bi3zmazyczMDK+//npHoYhB0kCcFpmFa625cuUK8/PzJhI0nU6zt7dHKpXquH6RSIRr1651pBTW7cITUgnr4cOHVCoVPvGJT/RUMm8UsAI+BMTbQjaWxFd6kHBxl8tFOBw2bl2yhC4UCuRyuZ5uLImelPqLMtjFLCPBQU47cj+2WWdbnd8Pa0etVjN5t2Xj0Xl+hikS4lssD5Z6vW7MP87ZpLjNKaU6bPndJo7ThqtLG2QjMBKJdJhGxHTRbDaNz72UW5MNZjHFeb1e40oXDAbx+XzGrBYKhQiHw2YzWoK/8vl8x4NKqkXJzB4wqy7AFCQOBAJDy+MyCLLCE/NYOBw2kxeneUk2YIvFojlfEnsg/ZX9HSl0clmwAj4EYrEYr732GsVikfX1dZ4+fTrwZyYSCd544w1KpRIffPABq6ur5PN57t27x+rq6nMCflQEZrVaJZlMUqlUmJyc5Pr16+YGlWX4xMQEgUDgTG9Yr9fL7OwsbrfbeGVA69zNzc2dyTGdWfwkZ7UzFWs6nebp06dG7OQBLJ4s3ellT+spE4lEeP311ymVSuZY9XqdXC7He++9Z4K03G43hUKBYrGIz+djaWmJSCRiglkkEZh4YIiXxsLCAoFAgEKhwPvvv8/Ozg4HBwe88847xnatlDLRqrVajUgkYkwx0WjUPODE/fOi2ImlHcFgkFdeeYUrV650OAysra2xvb1Ns9lka2uLnZ2djomIuEXKCsaZMvkyYAV8CEgyeinW++zZM+D56jX93BQSiFKv18lkMqyurlIul9nY2DCf1c9GptatGpBLS0vmJn1RN6rM+BOJBC6Xi/39fTY2Nkx2xbM8rpiCJiYmGB8fBz48H1tbW+zt7Znr5myLM+eIcNqZeCAQMMKTzWaN18j+/r7J5dL9uWJ+m5ubM1kmj9onGBsbI5FImLD37e3tY2t8KqUIh8NMTEyY7/24Nr5oxDQm9nARcJl1S5bEZDJpsoIKcv2kVqp4ulzUvvaLFfAh0R3Q02g0TH5jWfr2OmicniYSfXf16tWOwdn9/ThkEE9PTxsb94sOTZaUA+JBsLCwQL1eJxaLvbB2dHvwBINBpqeniUQixi/eWfXI6X89jOhDeZCFQiHcbjezs7NHltST/QFnqoDjzpM8rKampnoqajw5OWk+fxQFzTkpmJ+fN940EsfQHZ8h513K+V0WrIAPge5wcrH5HhwcoJQiHo8zOzvblwDIAJRCyHNzc8/NBk/6/+5BLA+S88grITUrJZH//Py8KUpxHrZWuS6vvPJKRzBSJpPhnXfeYXt728zMxUtlGEvvUChk7LPj4+NHpsMVO66Iay/nKBAIcPv2bW7cuHHi+JANaxm7o4TT8+TKlSvMzs6aVdVxroUy7g9LAjeqWAEfEs4dcRFzSa4UDAYHqhwTCAQIBAIdr/eaZOmiIDccfPgggaM3Ps+yHYIUNRZEAJwBJSLsw0qR4AyGGbYtVrwxPiocdW98lLACPkSUUkxOTvLyyy8b++7GxgZaa+bm5oZ6w14kce4X54zyovUjGAyysrLC5OQkiUTC+EKPopnBcvmxAj5ERMCDwSC5XI47d+6wvr5OIBAYeuWYUeY8fIt7JRAIcO3aNer1eodwX9T2Wj7aWAEfMuKa5yypJilKpYr3R3nJd5Fx7js4N3qtgFsuKlbAh4xkbHNWX9/f3zfJm15++WWuX79uBeGC0r2pd1k2uyyXk15qYgaAPwX87fd/RWv9M0qpa8CXgQngm8APa62rZ9nYUUBmbiLkgIkG8/l8LCwsnHMLLSdhRdsyKvQyA68A36W1zqtWdfo/U0r9IfAvgP+otf6yUuqXgM8Bv3iGbR0pvF6vqV0peDweE0xisVgsg9JLTUwNSK5Rb/tLA98F/MP2618C/jVWwA1On1wnkkHNYrFYBqUnG7hSyk3LTHID+AXgMZDWWotrxTpwqG1AKfUm8ObgTR0txEfVYrFYzoqejH1a64bW+g1gEfh24HavB9Bav6W1/lat9beeso0Wi8ViOYS+dmu01mngj4HvBBJKKZnBLwIbQ26bxWKxWI7hRAFXSk0ppRLtn4PA9wAPaAn5P2i/7UeA3zurRlosFovleVQPSZFep7VJ6aYl+L+ttf63SqnrtNwIx4F3gH+kta6c8Fl7QAHYH0LbLwqTXK7+wOXrk+3Pxeey9WnY/bmqtZ7qfvFEAR82Sqm/uEz28MvWH7h8fbL9ufhctj69qP7YiAWLxWIZUayAWywWy4hyHgL+1jkc8yy5bP2By9cn25+Lz2Xr0wvpzwu3gVssFotlOFgTisVisYwoVsAtFotlRHmhAq6U+pRS6j2l1COl1Bde5LGHgVJqSSn1x0qp+0qpe0qpH2u/Pq6U+iOl1MP297Hzbms/KKXcSql3lFK/3/79mlLqG+3r9FtKqeEWbzxjlFIJpdRXlFJ/rZR6oJT6zlG+Rkqpz7fH212l1G8qpQKjdI2UUr+qlNpVSt11vHbo9VAt/lO7X3eUUt9yfi0/miP69O/bY+6OUup/SABk+28/1e7Te0qpvzusdrwwAW8nxPoF4HuBV4AfVEq98qKOPyTqwE9orV8BvgP4p+0+fAH4utb6JeDr7d9HiR+jFV0r/DtaqYJvAClaqYJHiZ8H/pfW+jbwcVp9G8lrpJRaAP458K1a61dpBdR9ltG6Rr8GfKrrtaOux/cCL7W/3uTiZjj9NZ7v0x8Br2qtXwfeB34KoK0RnwU+1v6f/9LWw4F5kTPwbwceaa0/aBd++DLw6Rd4/IHRWm9prf+y/XOOljAs0OrHl9pv+xLw98+nhf2jlFoE/h7wy+3fFa1UwV9pv2XU+hMH/hbwKwBa62o7h8/IXiNaWUOD7dxDIWCLEbpGWus/BZJdLx91PT4N/Lpu8ee0ci7NvZiW9s5hfdJaf82RofXPaeWIglafvqy1rmitnwCPaOnhwLxIAV8A1hy/H5mCdhRQSi0DnwC+Acxorbfaf9oGZs6pWafh54B/CTTbv0/QY6rgC8o1YA/4r22z0C8rpcKM6DXSWm8A/wF4Rku4M7RSO4/yNYKjr8dl0Yl/Avxh++cz65PdxDwFSqkI8DvAj2uts86/tQtgjIRvplLq+4FdrfU3z7stQ8QDfAvwi1rrT9DKvdNhLhmxazRGawZ3DZgHwjy/dB9pRul69IJS6qdpmVt/46yP9SIFfANYcvw+kilo22Xlfgf4Da3177Zf3pFlXvv77nm1r08+CfyAUmqVlknru2jZj0c5VfA6sK61/kb796/QEvRRvUZ/B3iitd7TWteA36V13Ub5GsHR12OkdUIp9aPA9wM/pD8MsjmzPr1IAX8beKm9e+6jZdT/6gs8/sC07cO/AjzQWv+s409fpZVSF0Yota7W+qe01ota62Va1+P/aq1/iBFOFay13gbWlFK32i99N3CfEb1GtEwn36GUCrXHn/RnZK9Rm6Oux1eBf9z2RvkOIOMwtVxolFKfomWO/AGtddHxp68Cn1VK+VWrGPxLwP8bykG11i/sC/g+Wruzj4GffpHHHlL7/yatpd4d4N321/fRsht/HXgI/B9g/Lzbeoq+/W3g99s/X28PsEfAfwf8592+PvvyBvAX7ev0P4GxUb5GwL8B/hq4C/w3wD9K1wj4TVr2+xqtFdLnjroegOLDso1/Rcv75tz70GOfHtGydYs2/JLj/T/d7tN7wPcOqx02lN5isVhGFLuJabFYLCOKFXCLxWIZUayAWywWy4hiBdxisVhGFCvgFovFMqJYAbdYLJYRxQq4xWKxjCj/H9PSP6yCnofXAAAAAElFTkSuQmCC\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(32, 128)\n"
          ]
        }
      ],
      "source": [
        "## custom image\n",
        "\n",
        "img_path = \"/content/custom_images/medicine1.jpg\"\n",
        "image2 = preprocess_image(img_path)\n",
        "# label = vectorize_label(label)\n",
        "img2 = tf.image.flip_left_right(image2)\n",
        "img2 = tf.transpose(img2, perm=[1, 0, 2])\n",
        "img2 = (img2 * 255.0).numpy().clip(0, 255).astype(np.uint8)\n",
        "img2 = img2[:, :, 0]\n",
        "plt.imshow(img2,cmap=\"gray\")\n",
        "plt.show()\n",
        "img_list = [img2]\n",
        "print(img2.shape)\n",
        "# gathering text\n",
        "# preds2 = prediction_model.predict(img_list)\n",
        "# pred_texts2 = decode_batch_predictions(preds2)\n",
        "# print(pred_texts2)\n"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "\n",
        "def prepare_dataset_custom(img_paths_3):\n",
        "    dataset = tf.data.Dataset.from_tensor_slices((img_paths_3)).map(\n",
        "        preprocess_image, num_parallel_calls=AUTOTUNE\n",
        "    )\n",
        "    return dataset.batch(batch_size).cache().prefetch(AUTOTUNE)\n"
      ],
      "metadata": {
        "id": "9ujqdspc_cLF"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "img_paths_3 = [\"/content/data/words/r03/r03-053/r03-053-01-01.png\",\"/content/data/words/r03/r03-065/r03-065-01-03.png\",\"/content/data/words/r03/r03-090/r03-090-02-04.png\",\"/content/data/words/r03/r03-090/r03-090-06-01.png\"]\n",
        "custom_ds = prepare_dataset_custom(img_paths_3)\n",
        "custom_ds"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "DY-24gYcAS2A",
        "outputId": "4dfda63b-d684-47ac-dd80-75e602902417"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<PrefetchDataset element_spec=TensorSpec(shape=(None, None, None, 1), dtype=tf.float32, name=None)>"
            ]
          },
          "metadata": {},
          "execution_count": 33
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "for batch in custom_ds.take(1):\n",
        "    # print(batch[\"label\"])\n",
        "    # print(batch)\n",
        "    # batch_images = batch[\"image\"]\n",
        "    print(\"len is : \",len(batch))\n",
        "    print(\"batch img shape: \",batch[0].shape)\n",
        "    print(\"batch img shape: \",batch[1].shape)\n",
        "    \n",
        "    _, ax = plt.subplots(1, 4, figsize=(15, 8))\n",
        "\n",
        "    preds = prediction_model.predict(batch)\n",
        "    pred_texts = decode_batch_predictions(preds)\n",
        "    print(pred_texts)\n",
        "\n",
        "    for i in range(4):\n",
        "        img = batch[i]\n",
        "        img = tf.image.flip_left_right(img)\n",
        "        img = tf.transpose(img, perm=[1, 0, 2])\n",
        "        img = (img * 255.0).numpy().clip(0, 255).astype(np.uint8)\n",
        "        img = img[:, :, 0]\n",
        "\n",
        "        title = f\"Prediction: {pred_texts[i]}\"\n",
        "        ax[i].imshow(img, cmap=\"gray\")\n",
        "        ax[i].set_title(title)\n",
        "        ax[i].axis(\"off\")\n",
        "\n",
        "plt.show()\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 179
        },
        "id": "Z02CdFWzA0Gv",
        "outputId": "a1163673-4a4d-4535-b733-d5e8757bee23"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "len is :  4\n",
            "batch img shape:  (128, 32, 1)\n",
            "batch img shape:  (128, 32, 1)\n",
            "1/1 [==============================] - 0s 281ms/step\n",
            "['grey', 'what', 'earns', 'wite']\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 1080x576 with 4 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# !zip -r /content/content/models/my_model_epoch_100.tflite.zip /content/Folder_To_Zip\n",
        "\n",
        "from google.colab import files\n",
        "files.download(\"/content/content/models/my_model_epoch_100.tflite\")"
      ],
      "metadata": {
        "id": "LLhC4nidZ75p"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "model.save('content/models/my_model_epoch_100.h5')"
      ],
      "metadata": {
        "id": "PM0SKq-tsnPS"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "!zip -r /content/content/models/my_model_epoch_100.tflite.zip /content/content/models/my_model_epoch_100.tflite"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "m5M-CkUvbgxi",
        "outputId": "355fa50b-57b0-488b-b2fc-d068a15fb28d"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "  adding: content/content/models/my_model_epoch_100.tflite/ (stored 0%)\n",
            "  adding: content/content/models/my_model_epoch_100.tflite/saved_model.pb (deflated 90%)\n",
            "  adding: content/content/models/my_model_epoch_100.tflite/variables/ (stored 0%)\n",
            "  adding: content/content/models/my_model_epoch_100.tflite/variables/variables.data-00000-of-00001 (deflated 9%)\n",
            "  adding: content/content/models/my_model_epoch_100.tflite/variables/variables.index (deflated 71%)\n",
            "  adding: content/content/models/my_model_epoch_100.tflite/keras_metadata.pb (deflated 93%)\n",
            "  adding: content/content/models/my_model_epoch_100.tflite/assets/ (stored 0%)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "N56LRHK-0DaU"
      },
      "source": [
        "To get better results the model should be trained for at least 50 epochs."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "CyGe85Pm0DaU"
      },
      "source": [
        "## Final remarks\n",
        "\n",
        "* The `prediction_model` is fully compatible with TensorFlow Lite. If you are interested,\n",
        "you can use it inside a mobile application. You may find\n",
        "[this notebook](/~https://github.com/tulasiram58827/ocr_tflite/blob/main/colabs/captcha_ocr_tflite.ipynb)\n",
        "to be useful in this regard.\n",
        "* Not all the training examples are perfectly aligned as observed in this example. This\n",
        "can hurt model performance for complex sequences. To this end, we can leverage\n",
        "Spatial Transformer Networks ([Jaderberg et al.](https://arxiv.org/abs/1506.02025))\n",
        "that can help the model learn affine transformations that maximize its performance."
      ]
    }
  ],
  "metadata": {
    "colab": {
      "provenance": []
    },
    "gpuClass": "standard",
    "kernelspec": {
      "display_name": "Python 3",
      "language": "python",
      "name": "python3"
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.7.0"
    },
    "accelerator": "GPU"
  },
  "nbformat": 4,
  "nbformat_minor": 0
}