IMDB Reviews with GRU (and optional LSTM and Conv1D).ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Course 3 - Week 3 - Lesson 2d.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "id": "P-AhVYeBWgQ3",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# NOTE: PLEASE MAKE SURE YOU ARE RUNNING THIS IN A PYTHON3 ENVIRONMENT\n",
        "\n",
        "import tensorflow as tf\n",
        "print(tf.__version__)\n",
        "\n",
        "# This is needed for the iterator over the data\n",
        "# But not necessary if you have TF 2.0 installed\n",
        "#!pip install tensorflow==2.0.0-beta0\n",
        "\n",
        "\n",
        "tf.enable_eager_execution()\n",
        "\n",
        "# !pip install -q tensorflow-datasets"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_IoM4VFxWpMR",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import tensorflow_datasets as tfds\n",
        "imdb, info = tfds.load(\"imdb_reviews\", with_info=True, as_supervised=True)\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "wHQ2Ko0zl7M4",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import numpy as np\n",
        "\n",
        "train_data, test_data = imdb['train'], imdb['test']\n",
        "\n",
        "training_sentences = []\n",
        "training_labels = []\n",
        "\n",
        "testing_sentences = []\n",
        "testing_labels = []\n",
        "\n",
        "# str(s.tonumpy()) is needed in Python3 instead of just s.numpy()\n",
        "for s,l in train_data:\n",
        "  training_sentences.append(str(s.numpy()))\n",
        "  training_labels.append(l.numpy())\n",
        "  \n",
        "for s,l in test_data:\n",
        "  testing_sentences.append(str(s.numpy()))\n",
        "  testing_labels.append(l.numpy())\n",
        "  \n",
        "training_labels_final = np.array(training_labels)\n",
        "testing_labels_final = np.array(testing_labels)\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7n15yyMdmoH1",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "vocab_size = 10000\n",
        "embedding_dim = 16\n",
        "max_length = 120\n",
        "trunc_type='post'\n",
        "oov_tok = \"<OOV>\"\n",
        "\n",
        "\n",
        "from tensorflow.keras.preprocessing.text import Tokenizer\n",
        "from tensorflow.keras.preprocessing.sequence import pad_sequences\n",
        "\n",
        "tokenizer = Tokenizer(num_words = vocab_size, oov_token=oov_tok)\n",
        "tokenizer.fit_on_texts(training_sentences)\n",
        "word_index = tokenizer.word_index\n",
        "sequences = tokenizer.texts_to_sequences(training_sentences)\n",
        "padded = pad_sequences(sequences,maxlen=max_length, truncating=trunc_type)\n",
        "\n",
        "testing_sequences = tokenizer.texts_to_sequences(testing_sentences)\n",
        "testing_padded = pad_sequences(testing_sequences,maxlen=max_length)\n",
        "\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "9axf0uIXVMhO",
        "colab": {}
      },
      "source": [
        "reverse_word_index = dict([(value, key) for (key, value) in word_index.items()])\n",
        "\n",
        "def decode_review(text):\n",
        "    return ' '.join([reverse_word_index.get(i, '?') for i in text])\n",
        "\n",
        "print(decode_review(padded[1]))\n",
        "print(training_sentences[1])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5NEpdhb8AxID",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "model = tf.keras.Sequential([\n",
        "    tf.keras.layers.Embedding(vocab_size, embedding_dim, input_length=max_length),\n",
        "    tf.keras.layers.Bidirectional(tf.keras.layers.GRU(32)),\n",
        "    tf.keras.layers.Dense(6, activation='relu'),\n",
        "    tf.keras.layers.Dense(1, activation='sigmoid')\n",
        "])\n",
        "model.compile(loss='binary_crossentropy',optimizer='adam',metrics=['accuracy'])\n",
        "model.summary()\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "V5LLrXC-uNX6",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "num_epochs = 50\n",
        "history = model.fit(padded, training_labels_final, epochs=num_epochs, validation_data=(testing_padded, testing_labels_final))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nHGYuU4jPYaj",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "\n",
        "\n",
        "def plot_graphs(history, string):\n",
        "  plt.plot(history.history[string])\n",
        "  plt.plot(history.history['val_'+string])\n",
        "  plt.xlabel(\"Epochs\")\n",
        "  plt.ylabel(string)\n",
        "  plt.legend([string, 'val_'+string])\n",
        "  plt.show()\n",
        "\n",
        "plot_graphs(history, 'accuracy')\n",
        "plot_graphs(history, 'loss')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "wSualgGPPK0S",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Model Definition with LSTM\n",
        "model = tf.keras.Sequential([\n",
        "    tf.keras.layers.Embedding(vocab_size, embedding_dim, input_length=max_length),\n",
        "    tf.keras.layers.Bidirectional(tf.keras.layers.LSTM(32)),\n",
        "    tf.keras.layers.Dense(6, activation='relu'),\n",
        "    tf.keras.layers.Dense(1, activation='sigmoid')\n",
        "])\n",
        "model.compile(loss='binary_crossentropy',optimizer='adam',metrics=['accuracy'])\n",
        "model.summary()\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "K_Jc7cY3Qxke",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Model Definition with Conv1D\n",
        "model = tf.keras.Sequential([\n",
        "    tf.keras.layers.Embedding(vocab_size, embedding_dim, input_length=max_length),\n",
        "    tf.keras.layers.Conv1D(128, 5, activation='relu'),\n",
        "    tf.keras.layers.GlobalAveragePooling1D(),\n",
        "    tf.keras.layers.Dense(6, activation='relu'),\n",
        "    tf.keras.layers.Dense(1, activation='sigmoid')\n",
        "])\n",
        "model.compile(loss='binary_crossentropy',optimizer='adam',metrics=['accuracy'])\n",
        "model.summary()\n"
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}