primeqa.calibration.train_confidence_calibrator.TaskArguments#

class primeqa.calibration.train_confidence_calibrator.TaskArguments(confidence_model_dir: str, scorer_type: str = 'weighted_sum_target_type_and_score_diff', task_heads: typing.Optional[primeqa.calibration.train_confidence_calibrator.object_reference] = None, preprocessor: primeqa.calibration.train_confidence_calibrator.object_reference = <class 'primeqa.mrc.processors.preprocessors.tydiqa.TyDiQAPreprocessor'>, postprocessor: primeqa.calibration.train_confidence_calibrator.object_reference = <class 'primeqa.mrc.processors.postprocessors.extractive.ExtractivePostProcessor'>, eval_metrics: primeqa.calibration.train_confidence_calibrator.object_reference = <class 'primeqa.mrc.metrics.tydi_f1.tydi_f1.TyDiF1'>, output_dropout_rate: float = 0.25, decoding_times_with_dropout: int = 5, max_iter_of_confidence_model_training: int = 200, prediction_reference_overlap_threshold: float = 0.5)#

Bases: object

Task specific arguments.

Methods

Attributes

`decoding_times_with_dropout`
`max_iter_of_confidence_model_training`
`output_dropout_rate`
`prediction_reference_overlap_threshold`
`scorer_type`
`task_heads`
`confidence_model_dir`

eval_metrics#: alias of primeqa.mrc.metrics.tydi_f1.tydi_f1.TyDiF1

postprocessor#: alias of primeqa.mrc.processors.postprocessors.extractive.ExtractivePostProcessor

preprocessor#: alias of primeqa.mrc.processors.preprocessors.tydiqa.TyDiQAPreprocessor