wide_deep.py

# -*- coding: utf-8 -*-
"""tf.estimator API를 이용한 TensorFlow Wide & Deep Tutorial 예제"""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function

import os
import shutil

from absl import app as absl_app
from absl import flags
import tensorflow as tf  # pylint: disable=g-bad-import-order

from official.utils.flags import core as flags_core
from official.utils.logs import hooks_helper
from official.utils.logs import logger
from official.utils.misc import model_helpers

# 학습에 사용할 컬럼들을 정의합니다.
_CSV_COLUMNS = [
    'age', 'workclass', 'fnlwgt', 'education', 'education_num',
    'marital_status', 'occupation', 'relationship', 'race', 'gender',
    'capital_gain', 'capital_loss', 'hours_per_week', 'native_country',
    'income_bracket'
]

# 컬럼들의 기본값을 지정합니다.
_CSV_COLUMN_DEFAULTS = [[0], [''], [0], [''], [0], [''], [''], [''], [''], [''],
                        [0], [0], [0], [''], ['']]

# training과 validation에 사용할 train validation 데이터 개수를 지정합니다.
_NUM_EXAMPLES = {
    'train': 32561,
    'validation': 16281,
}


LOSS_PREFIX = {'wide': 'linear/', 'deep': 'dnn/'}


def define_wide_deep_flags():
  """model type과 학습을 위한 flag들을 지정합니다."""
  flags_core.define_base()
  flags_core.define_benchmark()

  flags.adopt_module_key_flags(flags_core)

  flags.DEFINE_enum(
      name="model_type", short_name="mt", default="wide_deep",
      enum_values=['wide', 'deep', 'wide_deep'],
      help="Select model topology.")

  flags_core.set_defaults(data_dir='/tmp/census_data',
                          model_dir='/tmp/census_model',
                          train_epochs=40,
                          epochs_between_evals=2,
                          batch_size=40)


def build_model_columns():
  """feature columns를 설정합니다."""
  # Continuous 컬럼들
  age = tf.feature_column.numeric_column('age')
  education_num = tf.feature_column.numeric_column('education_num')
  capital_gain = tf.feature_column.numeric_column('capital_gain')
  capital_loss = tf.feature_column.numeric_column('capital_loss')
  hours_per_week = tf.feature_column.numeric_column('hours_per_week')

  education = tf.feature_column.categorical_column_with_vocabulary_list(
      'education', [
          'Bachelors', 'HS-grad', '11th', 'Masters', '9th', 'Some-college',
          'Assoc-acdm', 'Assoc-voc', '7th-8th', 'Doctorate', 'Prof-school',
          '5th-6th', '10th', '1st-4th', 'Preschool', '12th'])

  marital_status = tf.feature_column.categorical_column_with_vocabulary_list(
      'marital_status', [
          'Married-civ-spouse', 'Divorced', 'Married-spouse-absent',
          'Never-married', 'Separated', 'Married-AF-spouse', 'Widowed'])

  relationship = tf.feature_column.categorical_column_with_vocabulary_list(
      'relationship', [
          'Husband', 'Not-in-family', 'Wife', 'Own-child', 'Unmarried',
          'Other-relative'])

  workclass = tf.feature_column.categorical_column_with_vocabulary_list(
      'workclass', [
          'Self-emp-not-inc', 'Private', 'State-gov', 'Federal-gov',
          'Local-gov', '?', 'Self-emp-inc', 'Without-pay', 'Never-worked'])

  # To show an example of hashing:
  occupation = tf.feature_column.categorical_column_with_hash_bucket(
      'occupation', hash_bucket_size=1000)

  # Transformations.
  age_buckets = tf.feature_column.bucketized_column(
      age, boundaries=[18, 25, 30, 35, 40, 45, 50, 55, 60, 65])

  # Wide columns and deep columns.
  base_columns = [
      education, marital_status, relationship, workclass, occupation,
      age_buckets,
  ]

  crossed_columns = [
      tf.feature_column.crossed_column(
          ['education', 'occupation'], hash_bucket_size=1000),
      tf.feature_column.crossed_column(
          [age_buckets, 'education', 'occupation'], hash_bucket_size=1000),
  ]

  wide_columns = base_columns + crossed_columns

  deep_columns = [
      age,
      education_num,
      capital_gain,
      capital_loss,
      hours_per_week,
      tf.feature_column.indicator_column(workclass),
      tf.feature_column.indicator_column(education),
      tf.feature_column.indicator_column(marital_status),
      tf.feature_column.indicator_column(relationship),
      # To show an example of embedding
      tf.feature_column.embedding_column(occupation, dimension=8),
  ]

  return wide_columns, deep_columns


def build_estimator(model_dir, model_type):
  """estimator model type에 따른 estimator를 설정합니다."""
  wide_columns, deep_columns = build_model_columns()
  hidden_units = [100, 75, 50, 25]

  # Create a tf.estimator.RunConfig to ensure the model is run on CPU, which
  # trains faster than GPU for this model.
  run_config = tf.estimator.RunConfig().replace(
      session_config=tf.ConfigProto(device_count={'GPU': 0}))

  if model_type == 'wide':
    return tf.estimator.LinearClassifier(
        model_dir=model_dir,
        feature_columns=wide_columns,
        config=run_config)
  elif model_type == 'deep':
    return tf.estimator.DNNClassifier(
        model_dir=model_dir,
        feature_columns=deep_columns,
        hidden_units=hidden_units,
        config=run_config)
  else:
    return tf.estimator.DNNLinearCombinedClassifier(
        model_dir=model_dir,
        linear_feature_columns=wide_columns,
        dnn_feature_columns=deep_columns,
        dnn_hidden_units=hidden_units,
        config=run_config)


def input_fn(data_file, num_epochs, shuffle, batch_size):
  """Estimator를 위한 input function을 정의합니다."""
  assert tf.gfile.Exists(data_file), (
      '%s not found. Please make sure you have run data_download.py and '
      'set the --data_dir argument to the correct path.' % data_file)
  # csv 파일을 parsing합니다.
  def parse_csv(value):
    print('Parsing', data_file)
    columns = tf.decode_csv(value, record_defaults=_CSV_COLUMN_DEFAULTS)
    features = dict(zip(_CSV_COLUMNS, columns))
    labels = features.pop('income_bracket')
    return features, tf.equal(labels, '>50K')

  # Extract lines from input files using the Dataset API.
  dataset = tf.data.TextLineDataset(data_file)

  if shuffle:
    dataset = dataset.shuffle(buffer_size=_NUM_EXAMPLES['train'])

  dataset = dataset.map(parse_csv, num_parallel_calls=5)

  # We call repeat after shuffling, rather than before, to prevent separate
  # epochs from blending together.
  dataset = dataset.repeat(num_epochs)
  dataset = dataset.batch(batch_size)
  return dataset


def export_model(model, model_type, export_dir):
  """SavedModel format으로 export합니다.

  Args:
    model: Estimator object
    model_type: model type을 나타내는 string. e.g. "wide", "deep" or "wide_deep"
    export_dir: model을 export할 폴더경로
  """
  wide_columns, deep_columns = build_model_columns()
  if model_type == 'wide':
    columns = wide_columns
  elif model_type == 'deep':
    columns = deep_columns
  else:
    columns = wide_columns + deep_columns
  feature_spec = tf.feature_column.make_parse_example_spec(columns)
  example_input_fn = (
      tf.estimator.export.build_parsing_serving_input_receiver_fn(feature_spec))
  model.export_savedmodel(export_dir, example_input_fn)


def run_wide_deep(flags_obj):
  """Wide-Deep training과 evaluation을 실행합니다.

  인자들(Args):
    flags_obj: parsed 플래그들(flag)
  """

  # model_dir 경로에 파일이 있으면 삭제합니다.
  shutil.rmtree(flags_obj.model_dir, ignore_errors=True)
  # tf.estimator API를 이용해서 학습모델을 생성합니다.
  model = build_estimator(flags_obj.model_dir, flags_obj.model_type)

  # 파일로부터 트레이닝 데이터와 테스트 데이터를 읽어옵니다.
  train_file = os.path.join(flags_obj.data_dir, 'adult.data')
  test_file = os.path.join(flags_obj.data_dir, 'adult.test')

  # train을 진행하고 flags.epochs_between_evals epoch 때마다 evaluation을 진행합니다.
  def train_input_fn():
    return input_fn(
        train_file, flags_obj.epochs_between_evals, True, flags_obj.batch_size)

  def eval_input_fn():
    return input_fn(test_file, 1, False, flags_obj.batch_size)

  run_params = {
      'batch_size': flags_obj.batch_size,
      'train_epochs': flags_obj.train_epochs,
      'model_type': flags_obj.model_type,
  }

  benchmark_logger = logger.get_benchmark_logger()
  benchmark_logger.log_run_info('wide_deep', 'Census Income', run_params,
                                test_id=flags_obj.benchmark_test_id)

  loss_prefix = LOSS_PREFIX.get(flags_obj.model_type, '')
  train_hooks = hooks_helper.get_train_hooks(
      flags_obj.hooks, batch_size=flags_obj.batch_size,
      tensors_to_log={'average_loss': loss_prefix + 'head/truediv',
                      'loss': loss_prefix + 'head/weighted_loss/Sum'})

   # train을 진행하고 flags.epochs_between_evals epoch 때마다 evaluation을 진행합니다.
  for n in range(flags_obj.train_epochs // flags_obj.epochs_between_evals):
    model.train(input_fn=train_input_fn, hooks=train_hooks)
    results = model.evaluate(input_fn=eval_input_fn)

    # evaluation 결과를 출력합니다.
    tf.logging.info('Results at epoch %d / %d',
                    (n + 1) * flags_obj.epochs_between_evals,
                    flags_obj.train_epochs)
    tf.logging.info('-' * 60)

    for key in sorted(results):
      tf.logging.info('%s: %s' % (key, results[key]))

    benchmark_logger.log_evaluation_result(results)

    if model_helpers.past_stop_threshold(
        flags_obj.stop_threshold, results['accuracy']):
      break

  # 모델을 Export합니다.
  if flags_obj.export_dir is not None:
    export_model(model, flags_obj.model_type, flags_obj.export_dir)


def main(_):
  with logger.benchmark_context(flags.FLAGS):
    run_wide_deep(flags.FLAGS)


if __name__ == '__main__':
  # 로깅 레벨을 설정합니다.
  tf.logging.set_verbosity(tf.logging.INFO)
  # 플래그들을 설정합니다.
  define_wide_deep_flags()
  # main 함수 호출합니다. 
  absl_app.run(main)