Is there a way to use PackedSequence in custom LSTM

nlp
1

Whats the difference of using a pad sequence from torch.nn.utils.rnn.pad_sequence() or a PackedSequence in the torch.nn.LSTM() Only computation? or also relates to the result.