During prediction in the evaluation mode, we put the predicted output as the input in the next time step. This is clear in LSTMCell
, however, for LSTM
, how should we do it as it wraps everything together ?
During prediction in the evaluation mode, we put the predicted output as the input in the next time step. This is clear in LSTMCell
, however, for LSTM
, how should we do it as it wraps everything together ?