[GitHub] ArmageddonKnight commented on issue #9010: get_next_state parameter in FusedRNN constructor

[GitBox <gi...@apache.org>]

ArmageddonKnight commented on issue #9010: get_next_state parameter in FusedRNN constructor
URL: https://github.com/apache/incubator-mxnet/issues/9010#issuecomment-350621786
 
 
   Hi Haibin,
   
   Thanks for your response. But what will be returned when `get_next_state = false`? I noticed that in the `charRNN` example located in the `cpp-package`:
   
   https://github.com/apache/incubator-mxnet/blob/20666c3186b3d7166cae625b48c81713c6811d67/cpp-package/example/charRNN.cpp#L168 
   
   We have the following code:
   
   ```c++
     auto rnn = RNN(embed, rnn_params, rnn_h_init, rnn_c_init, num_hidden, num_lstm_layer, 
   RNNMode::kLstm, false, dropout, !isTrain);
   ```
   
   It seems to me that the last parameter in the C++ API (i.e. `!isTrain`) corresponds to the `get_next_state` (because its documentation says `state_outputs Whether to have the states as symbol outputs`). But why should it be set to `true` while training and `false` while not training?

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
users@infra.apache.org


With regards,
Apache Git Services