Help understanding what encInSeq, decInSeq, decOutSeq are in the encoder and decoder example. about rnn HOT 4 CLOSED

In your example picture, encInSeq = ABC, decInSeq = <EOS>WXYZ and decOutSeq = WXYZ<EOS>. The example is mini-batched, in this case with a batch size of 2 - so we have two training examples. The encoded input sequence can be of a different length from the decoded sequence (which have to be of the same length). Does that help?

from rnn.

Yes it does! Thank you so much for your help! It seem obvious now.

from rnn.

local encInSeq = torch.Tensor({{1,2,3},{3,2,1}})
local decInSeq = torch.Tensor({{1,2,3,4},{4,3,2,1}})
local decOutSeq = torch.Tensor({{2,3,4,1},{1,2,4,3}})

I think your data decInSeq and decOutSeq is not match? @douwekiela

from rnn.

@htw2012: The example does not use real data, mostly because I did not want to get into trouble with padding and masking. If you want to improve it, feel free to submit a PR.

from rnn.