tfr.utils.sort_by_scores

Sorts list of features according to per-example scores.

tfr.utils.sort_by_scores(
    scores, features_list, topn=None, shuffle_ties=True, seed=None, mask=None
)

scores A Tensor of shape [batch_size, list_size] representing the per-example scores.
features_list A list of Tensors to be sorted. The shape of the Tensor can be [batch_size, list_size] or [batch_size, list_size, feature_dims]. The latter is applicable for example features.
topn An integer as the cutoff of examples in the sorted list.
shuffle_ties A boolean. If True, randomly shuffle before the sorting.
seed The ops-level random seed used when shuffle_ties is True.
mask An optional Tensor of shape [batch_size, list_size] representing which entries are valid for sorting. Invalid entries will be pushed to the end.

A list of Tensors as the list of sorted features by scores.