Spelling_Collab.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Spelling.ipynb",
      "version": "0.3.2",
      "provenance": [],
      "collapsed_sections": [],
      "toc_visible": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "metadata": {
        "id": "mQcWw3-jgaeO",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 415
        },
        "outputId": "68338616-b7b6-4049-a0d9-e4352e990737"
      },
      "cell_type": "code",
      "source": [
        "# ignore some Keras warnings regarding deprecations and model saving \n",
        "import warnings\n",
        "warnings.filterwarnings('ignore')\n",
        "\n",
        "from keras.models import Model, load_model, Sequential\n",
        "from keras.layers import Input, LSTM, Dense, RepeatVector, \\\n",
        "                         TimeDistributed, Activation, GRU, Dropout, \\\n",
        "                         Bidirectional\n",
        "from keras.callbacks import EarlyStopping\n",
        "import pickle\n",
        "\n",
        "!rm helpers.py\n",
        "!wget https://github.com/cmantas/keras_seq2seq/raw/master/helpers.py 2>&1 /dev/null\n",
        "from helpers import *\n"
      ],
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Using TensorFlow backend.\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "--2018-10-31 16:18:07--  https://github.com/cmantas/keras_seq2seq/raw/master/helpers.py\n",
            "Resolving github.com (github.com)... 192.30.253.113, 192.30.253.112\n",
            "Connecting to github.com (github.com)|192.30.253.113|:443... connected.\n",
            "HTTP request sent, awaiting response... 302 Found\n",
            "Location: https://raw.githubusercontent.com/cmantas/keras_seq2seq/master/helpers.py [following]\n",
            "--2018-10-31 16:18:08--  https://raw.githubusercontent.com/cmantas/keras_seq2seq/master/helpers.py\n",
            "Resolving raw.githubusercontent.com (raw.githubusercontent.com)... 151.101.0.133, 151.101.64.133, 151.101.128.133, ...\n",
            "Connecting to raw.githubusercontent.com (raw.githubusercontent.com)|151.101.0.133|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 10541 (10K) [text/plain]\n",
            "Saving to: ‘helpers.py’\n",
            "\n",
            "\rhelpers.py            0%[                    ]       0  --.-KB/s               \rhelpers.py          100%[===================>]  10.29K  --.-KB/s    in 0s      \n",
            "\n",
            "2018-10-31 16:18:08 (91.5 MB/s) - ‘helpers.py’ saved [10541/10541]\n",
            "\n",
            "/dev/null: Scheme missing.\n",
            "FINISHED --2018-10-31 16:18:08--\n",
            "Total wall clock time: 1.4s\n",
            "Downloaded: 1 files, 10K in 0s (91.5 MB/s)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "sB9EGj_6oFzE",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 55
        },
        "outputId": "d3f5bce1-c9ef-48aa-cf18-99ce188b60a4"
      },
      "cell_type": "code",
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/drive')"
      ],
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Drive already mounted at /drive; to attempt to forcibly remount, call drive.mount(\"/drive\", force_remount=True).\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "ssnwOSf8oTLU",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "wy1nlSWwgaeb",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "Sentences from the [tatoeba dataset](https://tatoeba.org/eng/downloads)"
      ]
    },
    {
      "metadata": {
        "id": "beNmQf7Ggaee",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "epochs = 50  # Number of epochs to train for.\n",
        "train_size =200000 # Number of samples to train on.\n",
        "val_size = 2000\n",
        "# Path to the txt file on disk.\n",
        "data_path = '/drive/My Drive/ML/sentences.txt'\n",
        "\n",
        "noise = .05\n",
        "misspellings_count = 3\n",
        "batch_size = 1000  # Batch size for training.\n",
        "\n",
        "optimizer= 'adam'\n",
        "loss_fn='categorical_crossentropy'\n",
        "# Hand-pick maximum sequence lengths\n",
        "max_string_length = 25 # max (allowed) input string length\n",
        "max_seq_length = 30 # accomodate for the delimiters + spelling correction\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "y1OV04FfWo-d",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "YUU8yMQGgaeu",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 269
        },
        "outputId": "5d132c55-c253-425d-8f90-16293772318a"
      },
      "cell_type": "code",
      "source": [
        "all_phrases = load_preprocessed(data_path, max_string_length)\n",
        "assert len(all_phrases) > train_size + val_size\n",
        "train_phrases = all_phrases[:train_size]\n",
        "test_phrases = all_phrases[train_size: train_size + val_size]\n",
        "\n",
        "print('All phrases in dataset: ', len(all_phrases))\n",
        "print('Training phrases: ', len(train_phrases))\n",
        "print('Test phrases: ', len(test_phrases))\n",
        "\n",
        "print(\"\\n * \".join(['Examples:'] + all_phrases[:10]))"
      ],
      "execution_count": 9,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "All phrases in dataset:  239820\n",
            "Training phrases:  200000\n",
            "Test phrases:  2000\n",
            "Examples:\n",
            " * tom is reading a book now\n",
            " * tom is quite strong\n",
            " * what did you just do?\n",
            " * tom unloaded the car\n",
            " * you aren't hurt\n",
            " * i had a hard time\n",
            " * tom actually likes me\n",
            " * i'm loaded\n",
            " * i began working\n",
            " * how was the interview?\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "LGCbVmq9gae9",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        },
        "outputId": "ae09aed6-dae6-4a31-ac5d-1dd2624bf88c"
      },
      "cell_type": "code",
      "source": [
        "# create doken indices out of all phrases\n",
        "token_idx = token_index(all_phrases + ['\\t', '\\n'])\n",
        "# ^^ \\t and \\n are our [START] and [END] delimiters. With this trick\n",
        "# we are adding them to the token index\n",
        "\n",
        "num_encoder_tokens = len(token_idx)\n",
        "\n",
        "print('Number of unique tokens:', num_encoder_tokens)"
      ],
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Number of unique tokens: 57\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "tJjlQ1x_MVHH",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def simple_lstm(output_len, token_count):\n",
        "    \"\"\"Generate the model\"\"\"\n",
        "    latent_dim = 1024  # Latent dimensionality of the encoding space.\n",
        "    initializer = 'he_normal'\n",
        "\n",
        "    # \"Encode\" the input sequence using an RNN, producing an output of HIDDEN_SIZE\n",
        "    # note: in a situation where your input sequences have a variable length,\n",
        "    # use input_shape=(None, nb_feature).\n",
        "    encoder = LSTM(latent_dim, input_shape=(None, token_count), \n",
        "                   kernel_initializer=initializer)\n",
        "\n",
        "    # For the decoder's input, we repeat the encoded input for each time step\n",
        "    repeater = RepeatVector(output_len)\n",
        "\n",
        "    decoder = LSTM(latent_dim, return_sequences=True, kernel_initializer=initializer)\n",
        "\n",
        "    # For each of step of the output sequence, decide which character should be chosen\n",
        "    time_dist = TimeDistributed(Dense(token_count, kernel_initializer=initializer))\n",
        "    activation = Activation('softmax')\n",
        "    \n",
        "    model = Sequential([encoder, repeater, decoder, time_dist, activation])\n",
        "\n",
        "    model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
        "    return model"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "gaJaJF-muQ0X",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def deep_lstm(output_len, token_count):\n",
        "    \"\"\"Generate the model\"\"\"\n",
        "    latent_dim = 256  # Latent dimensionality of the encoding space.\n",
        "    initializer = 'he_normal'\n",
        "\n",
        "    # \"Encode\" the input sequence using an RNN, producing an output of HIDDEN_SIZE\n",
        "    # note: in a situation where your input sequences have a variable length,\n",
        "    # use input_shape=(None, nb_feature).\n",
        "    encoder_1 = LSTM(latent_dim, input_shape=(None, token_count),\n",
        "                     return_sequences=True,\n",
        "                     kernel_initializer=initializer)\n",
        "    \n",
        "    dropout1 = Dropout(.05)\n",
        "\n",
        "    encoder_2 = LSTM(latent_dim, input_shape=(None, token_count),\n",
        "                 kernel_initializer=initializer)\n",
        "\n",
        "    # For the decoder's input, we repeat the encoded input for each time step\n",
        "    repeater = RepeatVector(output_len)\n",
        "\n",
        "    decoder_1 = LSTM(latent_dim, return_sequences=True, kernel_initializer=initializer)\n",
        "    dropout2 = Dropout(.05)\n",
        "    decoder_2 = LSTM(latent_dim, return_sequences=True, kernel_initializer=initializer)\n",
        "\n",
        "    # For each of step of the output sequence, decide which character should be chosen\n",
        "    time_dist = TimeDistributed(Dense(token_count, kernel_initializer=initializer))\n",
        "    activation = Activation('softmax')\n",
        "    \n",
        "    model = Sequential([encoder_1, dropout1, encoder_2,\n",
        "                        repeater, \n",
        "                        decoder_1, dropout2, decoder_2, \n",
        "                        time_dist, activation])\n",
        "\n",
        "    model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
        "    return model\n",
        "  \n",
        "def simple_lstm_2(output_len, token_count):\n",
        "  # 50 epoch: loss: 0.0928 - acc: 0.7668 - val_loss: 0.1042 - val_acc: 0.7663\n",
        "  \"\"\"Generate the model\"\"\"\n",
        "  latent_dim = 512  # Latent dimensionality of the encoding space.\n",
        "\n",
        "  # \"Encode\" the input sequence using an RNN, producing an output of HIDDEN_SIZE\n",
        "  # note: in a situation where your input sequences have a variable length,\n",
        "  # use input_shape=(None, nb_feature).\n",
        "  encoder = LSTM(latent_dim, return_sequences=True,\n",
        "                 input_shape=(output_len, token_count))\n",
        "\n",
        "  # For the decoder's input, we repeat the encoded input for each time step\n",
        "  #repeater = RepeatVector(output_len)\n",
        "\n",
        "  decoder = LSTM(latent_dim, return_sequences=True)\n",
        "\n",
        "  # For each of step of the output sequence, decide which character should be chosen\n",
        "  time_dist = TimeDistributed(Dense(token_count,  activation='softmax'))\n",
        "  #activation = Activation('softmax')\n",
        "\n",
        "  model = Sequential([encoder, decoder, time_dist])\n",
        "\n",
        "  model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
        "  return model\n",
        "\n",
        "\n",
        "def bidrectional_lstm(output_len, token_count):\n",
        "  \"\"\"Generate the model\"\"\"\n",
        "  latent_dim = 512  # Latent dimensionality of the encoding space.\n",
        "\n",
        "  layers = [\n",
        "      # something like an encoder\n",
        "      Bidirectional(LSTM(latent_dim, return_sequences=True, dropout=.2, \n",
        "                         recurrent_dropout=0.2),\n",
        "                    input_shape=(output_len, token_count)),\n",
        "      \n",
        "      # ~decoder\n",
        "      LSTM(latent_dim, return_sequences=True, dropout=.2),\n",
        "      \n",
        "      # time-distr activation\n",
        "      TimeDistributed(Dense(token_count)),\n",
        "      TimeDistributed(Dense(token_count,  activation='softmax')),\n",
        "  ]\n",
        "\n",
        "  model = Sequential(layers)\n",
        "\n",
        "  model.compile(loss='categorical_crossentropy', optimizer='adam', \n",
        "                metrics=['accuracy'])\n",
        "  return model"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "mlLT8CCQgafU",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "gen_model = bidrectional_lstm"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "WQvM3NLMv0K0",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "rgmeSgicgafg",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "val_misspellings, val_correct = create_misspellings(test_phrases[:2000],\n",
        "                                            noise, misspellings_count,\n",
        "                                            max_seq_length)\n",
        "\n",
        "val_X = vectorize_batch(val_misspellings, token_idx,\n",
        "                  max_seq_length, dtype=np.bool)\n",
        "val_Y = vectorize_batch(wrap_with_delims(val_correct), token_idx,\n",
        "                  max_seq_length, dtype=np.bool)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "ppH0bcGn1q4E",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "GVN4uBM5gaft",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# early_stopping = EarlyStopping(monitor='val_loss', min_delta=.005, \n",
        "#                                patience=3, verbose=0, mode='auto')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "IQlrrEnDkNSJ",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "YKDRTwMZwzNl",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "v7Y4JCYfgag3",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def batcher(phrases, batch_size):\n",
        "  for i in range(0, len(phrases), batch_size):\n",
        "    frrom = i\n",
        "    to = i+batch_size\n",
        "    yield phrases[frrom:to]\n",
        "\n",
        "def chunkify(lst,n):\n",
        "  return [lst[i::n] for i in range(n)]\n",
        "    \n",
        "def misspelled_gen(phrases, batch_size, noise, misspellings_count,\n",
        "                  max_seq_length):\n",
        "  gen = batcher(phrases, batch_size)\n",
        "  for batch in gen:\n",
        "    misspelled, correct = create_misspellings(batch,\n",
        "                                      noise, misspellings_count,\n",
        "                                      max_seq_length)\n",
        "    mis_chunks = chunkify(misspelled, misspellings_count + 1)\n",
        "    cor_chunks = chunkify(correct, misspellings_count + 1)\n",
        "    \n",
        "    yield from zip(mis_chunks, cor_chunks)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "VEVe0e5dL21Q",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def training_gen(phrases, batch_size, noise, misspellings_count,\n",
        "                  max_seq_length, token_idx):\n",
        "\n",
        "  while True:\n",
        "    Random().shuffle(phrases)\n",
        "    mis_gen = misspelled_gen(phrases, batch_size, noise, misspellings_count,\n",
        "                             max_seq_length)\n",
        "    for mis, cor in mis_gen:\n",
        "      X = vectorize_batch(mis, token_idx,\n",
        "                          max_seq_length, dtype=np.bool)\n",
        "      Y = vectorize_batch(wrap_with_delims(cor), token_idx,\n",
        "                          max_seq_length, dtype=np.bool)\n",
        "      yield (X, Y)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "dBengM8jLTGL",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "steps_per_epoch = (len(train_phrases) *(misspellings_count + 1) / batch_size)#30 <-for test"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "naZSSbWHgag-",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 53
        },
        "outputId": "8d17f918-6008-4954-f2fb-40dc4f0ccaea"
      },
      "cell_type": "code",
      "source": [
        "gen = training_gen(train_phrases, batch_size, noise, misspellings_count, \n",
        "                   max_seq_length, token_idx)\n",
        "\n",
        "model = gen_model(max_seq_length, len(token_idx))\n",
        "# early_stopping = EarlyStopping(monitor='val_loss', min_delta=.005, \n",
        "#                                patience=3, verbose=0, mode='auto')\n",
        "\n",
        "hist = model.fit_generator(gen, validation_data=(val_X, val_Y),\n",
        "                    steps_per_epoch=steps_per_epoch,\n",
        "                    verbose=1, max_queue_size=1, epochs=epochs)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/50\n",
            "  7/800 [..............................] - ETA: 29:33 - loss: 2.8265 - acc: 0.0981"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "vJxLPPN9wIZl",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        },
        "outputId": "438291a5-7ce9-4edb-eb0d-5a2a66259296"
      },
      "cell_type": "code",
      "source": [
        "dual_evaluate_batch(val_correct, val_misspellings, model, token_idx, \n",
        "                    max_seq_length)"
      ],
      "execution_count": 67,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(0.0775, 0.515125)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 67
        }
      ]
    },
    {
      "metadata": {
        "id": "2aS0cxO5uv1J",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "model.save('/drive/My Drive/ML/n_to_n_bidirectional_dropouts_2xtimedist_latent_dim_512.h5')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "jjmLgnzkzb_8",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 377
        },
        "outputId": "cf0848ba-e475-4fa9-f53a-00ea7f7b1ad9"
      },
      "cell_type": "code",
      "source": [
        "plot_history(hist)"
      ],
      "execution_count": 15,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAeEAAAFnCAYAAACRo/HLAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nCCGy5P8HCvdG6Hay//UAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7fa735daa908>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "id": "kbvmmoH_utUS",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "vzp7-FHSgahB",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "LSRFCz-0gahF",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "Mrqwok5EgahM",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "sjAJ8MyLgahP",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "E4imo5GUgahV",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "MLe1Je4Rgahb",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 163
        },
        "outputId": "b8664eb2-69f9-4d48-844a-788b31032f55"
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "error",
          "ename": "NameError",
          "evalue": "ignored",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-1-8c1153337866>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mplot_history\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mhist\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
            "\u001b[0;31mNameError\u001b[0m: name 'plot_history' is not defined"
          ]
        }
      ]
    },
    {
      "metadata": {
        "id": "4UlT1J-Ygahf",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "mY9gbKCogahj",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "corrector = sequential_translator_fn(model, token_idx, max_seq_length)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "BC5T20RNgahs",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 377
        },
        "outputId": "5d0b8be1-22fa-470c-f292-5250b4596ddd"
      },
      "cell_type": "code",
      "source": [
        "[(corrector(p), p, corrector(p) == p) for p in train_phrases[:20]]"
      ],
      "execution_count": 29,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[('\\t                ', 'good to be here', False),\n",
              " ('\\t                   ', 'i need someone to hold me', False),\n",
              " ('\\t                  ', 'ann gave me this present', False),\n",
              " ('\\t                 ', 'tom has a lot of toys', False),\n",
              " ('\\t                 ', \"i'm furious with you\", False),\n",
              " ('\\t                 ', 'it was my first job', False),\n",
              " ('\\t                  ', \"here's a picture of her\", False),\n",
              " ('\\t                 ', 'mary loves australia', False),\n",
              " ('\\t                 ', 'is he married?', False),\n",
              " ('\\t                 ', 'this well is very deep', False),\n",
              " ('\\t                   ', \"we'd be foolish to do so\", False),\n",
              " ('\\t                 ', 'come meet everybody', False),\n",
              " ('\\t                 ', 'catch you later', False),\n",
              " ('\\t                 ', 'i wrote them a letter', False),\n",
              " ('\\t                 ', 'it kept me occupied', False),\n",
              " ('\\t                 ', 'it surprises me', False),\n",
              " ('\\t                 ', 'no one followed me', False),\n",
              " ('\\t                 ', \"that's not the point\", False),\n",
              " ('\\t                  ', 'he put on clean trousers', False),\n",
              " ('\\t                  ', \"we're doing all we can\", False)]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 29
        }
      ]
    },
    {
      "metadata": {
        "id": "CM8Tr9R8gahw",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 102
        },
        "outputId": "fd458bff-17fc-4ca6-8904-dbcedb2b2865"
      },
      "cell_type": "code",
      "source": [
        "print(evaluate_correct(train_phrases[:1000], corrector))\n",
        "print(evaluate_correct(train_phrases[-1000:], corrector))\n",
        "print(evaluate_misspelled(train_phrases[:1000], corrector))\n",
        "print(evaluate_correct(test_phrases[:1000], corrector))\n",
        "print(evaluate_misspelled(test_phrases[:1000], corrector))"
      ],
      "execution_count": 25,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "0.032\n",
            "0.029\n",
            "0.398\n",
            "0.044\n",
            "0.444\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "vEP2XCTkgah0",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "plot_history(model.history)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "ul5hf_MWgah4",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "s7dps9QGgakD",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "4BdJOyLbgakI",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# find max encoder seq legth\n",
        "#max_encoder_seq_length = encoder_model.get_layer('encoder_inputs').input_shape[-1]\n",
        "phrases = ['fire', 'stp', 'comein', 'get ot', 'i cant go','im sorry', \n",
        "           'h is busi', 'hes drunk', 'ill be lat', 'hold mi beer', 'pus the buton', \n",
        "          'coll me on my phone', 'helo boys and girls']\n",
        "\n",
        "[corrector(phrase) for phrase in phrases]"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "qBW7G7rogakR",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        },
        "outputId": "19c6fc09-081c-4d9b-abf9-c668a3d87479"
      },
      "cell_type": "code",
      "source": [
        "!hostname"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "f248df630aff\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "nXbTxbJBgakU",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "MPPoEfGDgakY",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "u-9i902Pgakd",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "xWU7NTj7gakf",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "AG3PcZDwgakw",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "C61OoBikgak0",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "4VR31Tk4gak8",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "U8lTgVL8galA",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "xlqzuqNBgalE",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "rpoE0BQ8galH",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "Kdbx5ZVVgalX",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "AnJm4WAHgalb",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "IDo_-IYbgalf",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "ic83KDn7gali",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}