def _annotate_dataset_statistics()

in model_card_toolkit/model_card_toolkit.py [0:0]

• 19 lines of code
• 8 McCabe index (conditional complexity)

  def _annotate_dataset_statistics(self, model_card: ModelCard) -> ModelCard:
    """Annotates a model card with info from TFDV dataset statistics.

    Graphics for the dataset statistics are generated and appended to the
    Dataset section.

    Dataset statistics are read from both TfdvSource or MlmdSource, whichever is
    provided. Using both may cause duplicates to be recorded. If neither is
    provided, this function will be a no-op.

    Args:
      model_card: The model card object to annotate with TFDV dataset
        statistics.

    Returns:
      The model_card with dataset statistics annotated.
    """
    if self._source and self._source.tfdv:
      for dataset_stats_path in self._source.tfdv.dataset_statistics_paths:
        if self._source.tfdv.features_include or self._source.tfdv.features_exclude:
          data_stats = tfx_util.read_stats_protos_and_filter_features(
              dataset_stats_path, self._source.tfdv.features_include,
              self._source.tfdv.features_exclude)
        else:
          data_stats = tfx_util.read_stats_protos(dataset_stats_path)
        graphics.annotate_dataset_feature_statistics_plots(
            model_card, data_stats)
    if self._store:
      stats_artifacts = tfx_util.get_stats_artifacts_for_model(
          self._store, self._artifact_with_model_uri.id)
      for stats_artifact in stats_artifacts:
        data_stats = tfx_util.read_stats_protos(stats_artifact.uri)
        graphics.annotate_dataset_feature_statistics_plots(
            model_card, data_stats)
    return model_card