This repository was archived by the owner on Nov 29, 2022. It is now read-only.
This repository was archived by the owner on Nov 29, 2022. It is now read-only.
SparseCategoricalFocalLoss: ignore class? #24
Description
Hi,
TensorFlow's implementation of categorical cross-entropy accepts a parameter called ignore_class (see here) since v 2.10. In principle, it can be useful for NER tasks. It would be nice if such a parameter appeared in the calculation of focal loss, so an apple-to-apple comparison can be carried out.