tf.data.experimental.make_batched_features_dataset

Возвращает Dataset словарей признаков из Example протоколов.

tf.data.experimental.make_batched_features_dataset(
    file_pattern, batch_size, features, reader=tf.data.TFRecordDataset,
    label_key=None, reader_args=None, num_epochs=None, shuffle=True,
    shuffle_buffer_size=10000, shuffle_seed=None, prefetch_buffer_size=None,
    reader_num_threads=None, parser_num_threads=None, sloppy_ordering=False,
    drop_final_batch=False
)

Если аргумент label_key указан, возвращает Dataset кортеж, содержащий словари признаков и метку.

Пример:

serialized_examples = [
  features {
    feature { key: "age" value { int64_list { value: [ 0 ] } } }
    feature { key: "gender" value { bytes_list { value: [ "f" ] } } }
    feature { key: "kws" value { bytes_list { value: [ "code", "art" ] } } }
  },
  features {
    feature { key: "age" value { int64_list { value: [] } } }
    feature { key: "gender" value { bytes_list { value: [ "f" ] } } }
    feature { key: "kws" value { bytes_list { value: [ "sports" ] } } }
  }
]

Мы можем использовать аргументы:

features: {
  "age": FixedLenFeature([], dtype=tf.int64, default_value=-1),
  "gender": FixedLenFeature([], dtype=tf.string),
  "kws": VarLenFeature(dtype=tf.string),
}

И ожидаемый результат:

{
  "age": [[0], [-1]],
  "gender": [["f"], ["f"]],
  "kws": SparseTensor(
    indices=[[0, 0], [0, 1], [1, 0]],
    values=["code", "art", "sports"]
    dense_shape=[2, 2]),
}