I want to do a task oriented dialog chatbot which is used to book restaurant.Because every dialog has different sequences(eg. some has 5 turns of dialogs which is 10 sentences while another may has 6 turns of dialogs which is 12 sentences totally),I don't know how to batch dataset.
Could you give me some tutorial or github example?
There are some related questions to this on Stackoverflow. I liked the explanation/answer provided here. The tldr version is to use Packed Sequence. The answer I linked to provides the following example (copied from the link):
a = [torch.tensor([1,2,3]), torch.tensor([3,4])]
b = torch.nn.utils.rnn.pad_sequence(a, batch_first=True)
>>>>
tensor([[ 1, 2, 3],
[ 3, 4, 0]])
torch.nn.utils.rnn.pack_padded_sequence(b, batch_first=True, lengths=[3,2])
>>>>PackedSequence(data=tensor([ 1, 3, 2, 4, 3]), batch_sizes=tensor([ 2, 2, 1]))
