shared_embedding.ipynb

shared_embedding.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Untitled86.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/gist/austinzh/ad4c4508bb3793ad1815e6f845b7deaa/untitled86.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "E1e4lKDUo1nA"
      },
      "source": [
        "import tensorflow as tf\n"
      ],
      "execution_count": 1,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-cAa9Yd_xSgd"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "i1tg7dfRUaKw",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "fab308a5-d543-4aa9-c6d3-de3cf660aa6d"
      },
      "source": [
        "import tensorflow as tf\n",
        "from tensorflow import feature_column\n",
        "import numpy as np\n",
        "tf.compat.v1.disable_eager_execution()\n",
        "def _initialized_session(config=None):\n",
        "    sess = tf.compat.v1.Session(config=config)\n",
        "    sess.run(tf.compat.v1.initializers.global_variables())\n",
        "    sess.run(tf.compat.v1.tables_initializer())\n",
        "    return sess\n",
        "\n",
        "with tf.Graph().as_default():\n",
        "    # Define categorical colunm for our text feature, which is preprocessed into sequence of tokens\n",
        "    text_column = feature_column.sequence_categorical_column_with_vocabulary_list(key='text', vocabulary_list=list(['asd', 'asdf']))\n",
        "\n",
        "    max_length = 6\n",
        "    sequence_feature_layer_inputs = {}\n",
        "    sequence_feature_layer_inputs['text'] = tf.keras.Input(\n",
        "        shape=(max_length,), name='text', dtype=tf.string)\n",
        "\n",
        "    text_embedding = feature_column.shared_embeddings([text_column], dimension=64)\n",
        "\n",
        "    # below is ok to save\n",
        "    # text_embedding = feature_column.embedding_column(text_column, dimension=8)\n",
        "\n",
        "    # Define SequenceFeatures layer to pass feature_columns into Keras model\n",
        "    sequence_feature_layer = tf.keras.experimental.SequenceFeatures(text_embedding)\n",
        "\n",
        "    # note here that SequenceFeatures layer produce tuple of two tensors as output. We need just first to pass next.\n",
        "    sequence_feature_layer_outputs, _ = sequence_feature_layer(\n",
        "        sequence_feature_layer_inputs)\n",
        "    x = tf.keras.layers.Conv1D(8, 4)(sequence_feature_layer_outputs)\n",
        "    x = tf.keras.layers.MaxPooling1D(2)(x)\n",
        "    x = tf.keras.layers.Dense(256, activation='relu')(x)\n",
        "    x = tf.keras.layers.Dropout(0.2)(x)\n",
        "    x = tf.keras.layers.GlobalAveragePooling1D()(x)\n",
        "    # This example supposes binary classification, as labels are 0 or 1\n",
        "    x = tf.keras.layers.Dense(1, activation='sigmoid')(x)\n",
        "\n",
        "    model = tf.keras.models.Model(\n",
        "        inputs=[v for v in sequence_feature_layer_inputs.values()], outputs=x)\n",
        "\n",
        "    model.summary()\n",
        "\n",
        "    # This example supposes binary classification, as labels are 0 or 1\n",
        "    model.compile(optimizer='adam',\n",
        "                  loss='binary_crossentropy',\n",
        "                  metrics=['accuracy']\n",
        "                  #run_eagerly=True\n",
        "                  )\n",
        "    model.build(input_shape={'text': (max_length,)})\n",
        "    with _initialized_session() as sess:\n",
        "      model.fit({'text': np.array([[\"hello\"]*6])})\n",
        "      model.save(\"models\")"
      ],
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"model\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "text (InputLayer)            [(None, 6)]               0         \n",
            "_________________________________________________________________\n",
            "sequence_features (SequenceF ((None, None, 64), (None, 0         \n",
            "_________________________________________________________________\n",
            "conv1d (Conv1D)              (None, None, 8)           2056      \n",
            "_________________________________________________________________\n",
            "max_pooling1d (MaxPooling1D) (None, None, 8)           0         \n",
            "_________________________________________________________________\n",
            "dense (Dense)                (None, None, 256)         2304      \n",
            "_________________________________________________________________\n",
            "dropout (Dropout)            (None, None, 256)         0         \n",
            "_________________________________________________________________\n",
            "global_average_pooling1d (Gl (None, 256)               0         \n",
            "_________________________________________________________________\n",
            "dense_1 (Dense)              (None, 1)                 257       \n",
            "=================================================================\n",
            "Total params: 4,617\n",
            "Trainable params: 4,617\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n",
            "Train on 1 samples\n",
            "1/1 [==============================] - 1s 627ms/sample - loss: 0.6931 - accuracy: 0.0000e+00\n",
            "INFO:tensorflow:Assets written to: models/assets\n"
          ],
          "name": "stdout"
        }
      ]
    }
  ]
}