A problem about LSTM forward propagation result

1

When I set the sequence length to 1, Output vector and result [0] are the same.

image
image1911×1008 144 KB

If the sequence length is not equal to 1 , Output vector is not equal to hidden state vector.
image
image1117×143 14 KB

Who can tell me why. And how to calculate the output vector at this time. And what’s the meaning

2

Could you explain your use case a bit more and post a code snippet so that we can have a look, please?

PS: you can post code snippets by wrapping them into three backticks ```, which makes debugging easier. :wink:

3

Please look here pytorch/rnn.py at eae1b40400df2b3e6d777bcaaf60f338749cad0f · pytorch/pytorch · GitHub

The output tensor contain the hidden states for all the sequence lengths and result [1] tensor contain the hidden states just of the very last sequence length .