How to use batchnorm1d and pack_padded_sequence at the same time?

hello everyone, I have a question about batchnorm1d and pack_padded_sequence.
I have an encoder-decoder model, and for the encoder, I use pack_padded_sequence to construct a batch input for it. Meanwhile, I also want to apply batchnorm1d to the encoder to avoid overfitting, so how should I do it.
Thanks in advance.