Cross Entropy

Loss variant for multiclass classification. It interprets the model output as per-class logits and penalizes the negative log-probability of the correct class. Compatible with the multiclass policy.

Configuration

Selected as name: cross-entropy.

Option	Description	Type	Default
reduction	How per-sample losses are reduced.	mean | sum	mean
weight	Optional per-class weights for class imbalance.	list[float] | null	null
ignore_index	Target value that is ignored and contributes no gradient.	int	-100
label_smoothing	Amount of label smoothing in [0, 1].	float	0.0

components:
  loss:
    name: cross-entropy
    kwargs: {}