When k is large enough in early iterations, there may be too few valid scores from which to calculate a top-k average. This results in a TypeError of trying to sum over an iterable of Optional[float]. This PR adds a simple change to convert these Nones to 0s in the sum() call. This depresses the top-k average, but I would argue that's logically correct when the top-k contains invalid scores.
Description
When
kis large enough in early iterations, there may be too few valid scores from which to calculate a top-kaverage. This results in aTypeErrorof trying to sum over an iterable ofOptional[float]. This PR adds a simple change to convert theseNones to0s in thesum()call. This depresses the top-kaverage, but I would argue that's logically correct when the top-kcontains invalid scores.Checklist