CODEWITHSUNDEEP
lstmpytorchmulti-layer
Amuoeba
Amuoeba

Reputation: 786

Is it possible to implement a multilayered LSTM with LSTMCells modules in PyTorch?

In PyTorch there is a LSTM module which in addition to input sequence, hidden states, and cell states accepts a num_layers argument which specifies how many layers will our LSTM have.

There is however another module LSTMCell which has just input size and number of hidden states as parameters, there is no num_layers since this is a single cell in a multi-layered LSTM.

My question is what is the proper way to connect together the LSTMCell modules to achieve a same effect as a multi layered LSTM with num_layers > 1

Upvotes: 4

Views: 4904

Answers (2)

Roger Trullo
Roger Trullo

Reputation: 1584

LSTMCell is the basic building block of an LSTM network. You should use the LSTM module (which uses LSTMCell internally). If you want to do this yourself, the best way is to read the source code (https://github.com/pytorch/pytorch/blob/master/torch/nn/modules/rnn.py).

Basically you want to use one LSTMCell for each layer, and you should be careful on how to go from input to output, layer by layer taking into account the hidden states. I also have basic implementation of a convolutional LSTM but the idea is the same. You can check it here: https://github.com/rogertrullo/pytorch_convlstm/

Upvotes: 2

Hillel
Hillel

Reputation: 331

If I understand correctly the hidden output of lstm cell at layer l-1 is the input to the lstm cell at layer l.

Upvotes: 1

Related Questions