{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# import required libs\n",
    "import glob\n",
    "import os\n",
    "\n",
    "import tensorflow as tf\n",
    "import tensorflow_data_validation as tfdv\n",
    "print('TF version: {}'.format(tf.version.VERSION))\n",
    "print('TFDV version: {}'.format(tfdv.version.__version__))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Read artifact information from metadata store.\n",
    "import beam_dag_runner\n",
    "\n",
    "from tfx.orchestration import metadata\n",
    "from tfx.types import standard_artifacts\n",
    "\n",
    "metadata_connection_config = metadata.sqlite_metadata_connection_config(\n",
    "              beam_dag_runner.METADATA_PATH)\n",
    "with metadata.Metadata(metadata_connection_config) as store:\n",
    "    stats_artifacts = store.get_artifacts_by_type(standard_artifacts.ExampleStatistics.TYPE_NAME)\n",
    "    schema_artifacts = store.get_artifacts_by_type(standard_artifacts.Schema.TYPE_NAME)\n",
    "    anomalies_artifacts = store.get_artifacts_by_type(standard_artifacts.ExampleAnomalies.TYPE_NAME)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# configure output paths\n",
    "# Exact paths to output artifacts can also be found on KFP Web UI if you are using kubeflow.\n",
    "stats_path = stats_artifacts[-1].uri\n",
    "train_stats_file = os.path.join(stats_path, 'train', 'stats_tfrecord')\n",
    "eval_stats_file = os.path.join(stats_path, 'eval', 'stats_tfrecord')\n",
    "print(\"Train stats file:{}, Eval stats file:{}\".format(\n",
    "    train_stats_file, eval_stats_file))\n",
    "\n",
    "schema_file = os.path.join(schema_artifacts[-1].uri, 'schema.pbtxt')\n",
    "print(\"Generated schame file:{}\".format(schema_file))\n",
    "anomalies_file = os.path.join(anomalies_artifacts[-1].uri, 'anomalies.pbtxt')\n",
    "print(\"Generated anomalies file:{}\".format(anomalies_file))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# load generated statistics from StatisticsGen\n",
    "train_stats = tfdv.load_statistics(train_stats_file)\n",
    "eval_stats = tfdv.load_statistics(eval_stats_file)\n",
    "tfdv.visualize_statistics(lhs_statistics=eval_stats, rhs_statistics=train_stats,\n",
    "                          lhs_name='EVAL_DATASET', rhs_name='TRAIN_DATASET')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# load generated schema from SchemaGen\n",
    "schema = tfdv.load_schema_text(schema_file)\n",
    "tfdv.display_schema(schema=schema)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# load data vaildation result from ExampleValidator\n",
    "anomalies = tfdv.load_anomalies_text(anomalies_file)\n",
    "tfdv.display_anomalies(anomalies)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3.10.4 64-bit",
   "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": "3.10.4"
  },
  "pycharm": {
   "stem_cell": {
    "cell_type": "raw",
    "metadata": {
     "collapsed": false
    },
    "source": []
   }
  },
  "vscode": {
   "interpreter": {
    "hash": "4f946df053fbf2b937619d3c5458e7af74262f9a954d8797ba0b27400bcafe06"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}