Backpropagation in RNNs when using packed padded inputs

1

I’m trying to train an RNN using batches of sequences. Some of those sequences have to be padded and packed. My question is, should I also pack the target batch corresponding to a packed and padded input?

Thanks