primeqa.mrc.metrics.nq_f1.nq_eval.extract_metrics_at_optimal_threshold

primeqa.mrc.metrics.nq_f1.nq_eval.extract_metrics_at_optimal_threshold(answer_stats: list) → Tuple[float, float, float, float]

Parameters

answer_stats – one of the dictionaries returned from score_answers

Returns

Tuple (f1, precision, recall, optimal_threshold)