peal.training.criterions.cross_entropy_loss¶
- peal.training.criterions.cross_entropy_loss(input, target, size_average=True, latent_code=None)[source]¶
Cross-entropy between logits and a soft (one-hot or mixed) target.
- Parameters:
input (
torch.Tensor) – Logits of shape[N, C]; log-softmax is applied here.target (
torch.Tensor) – Target distribution of the same shape (one-hot or a mixup of them).size_average (
bool, optional) – Divide the summed loss by the batch sizeN(default) or return the plain sum.latent_code (optional) – Ignored; present for the common criterion signature.
- Returns:
Scalar loss.
- Return type: