def __init__()

in tfx_addons/sampling/component.py [0:0]

• 25 lines of code
• 2 McCabe index (conditional complexity)

  def __init__(
      self,
      label: str,
      input_data: types.Channel = None,
      output_data: types.Channel = None,
      name: Optional[Text] = None,
      splits: Optional[List[Text]] = None,
      copy_others: Optional[bool] = True,
      shards: Optional[int] = 0,
      null_classes: Optional[List[Text]] = None,
      sampling_strategy: SamplingStrategy = SamplingStrategy.UNDERSAMPLE):
    """Construct a SamplerComponent.

    Args:
      input_data: A Channel of type `standard_artifacts.Examples`.
      output_data: A Channel of type `standard_artifacts.Examples`.
        By default, only the train split is sampled; all others are copied.
      name: Optional unique name. Necessary if multiple components are
        declared in the same pipeline.
      label: The name of the column containing class names to sample by.
      splits: A list containing splits to sample.
      copy_others: Determines whether we copy over the splits that aren't
        sampled, or just exclude them from the output artifact.
      shards: The number of files that each sampled split should
        contain. Default 0 is Beam's tfrecordio function's default.
      null_classes: A list determining which classes that we should not sample.
    """

    if not output_data:
      output_data = channel_utils.as_channel([standard_artifacts.Examples()])

    spec = SamplerSpec(
        input_data=input_data,
        output_data=output_data,
        label=label,
        name=name,
        splits=json_utils.dumps(splits),
        copy_others=int(copy_others),
        shards=shards,
        null_classes=json_utils.dumps(null_classes),
        sampling_strategy=sampling_strategy,
    )

    super().__init__(spec=spec)