nnsvs.train_util.compute_batch_pitch_regularization_weight

nnsvs.train_util.compute_batch_pitch_regularization_weight(lf0_score_denorm, decay_size)

Batch version of computing pitch regularization weight

Parameters:

lf0_score_denorm (Tensor) – (B, T)

Returns:

weights of shape (B, N, 1)

Return type:

Tensor