its-sandy opened this issue 5 years ago · 0 comments
How do we ensure that the padding of the reference mel spectogram is taken into account when the reference encoder is applied on a batch of mels?