SweeperConfig Class

Holder for sweeper configurations.

Dummy initializer for feature sweeping configuration object.

No arguments are meant to be passed in directly. The below class methods are the intended functions for creating usable SweeperConfig objects.

as if force_text_dnn is specified by the user and we're testing BERT. :param _sampler: Property dict specifying which Sampler to use for this experiment, as needed by SamplerConfig.from_dict method. :param _estimator: Which estimator to use for this experiment, e.g. "logistic_regression". :param _scorer: The metric to use during the experiment, which is used alongside the task type to instantiate the Scorer object. :param _baseline: Dict specifying the baseline featurizers for this experiment. :param _experiment: Dict specifying the experimental featurizers for this experiment. :param _column_purposes: List of col purposes for which this experiment applies and other related settings. :param _epsilon: The minimum lift the experimental featurizer must provide in order to be chosen over the baseline, subject to scaling based on sample size. :param _scale_epsilon: Whether or not to scale epsilon inversely with sample size for this experiment.