torch-rnn
torch-rnn copied to clipboard
Should I call model:clearState() before/after saving/loading a RNN model?
Hi,
I trained a LSTM model and got a decent validation accuracy after few epochs.
However, I am not very clear with the correct way to store/restore the states of LSTM. What is the correct sequence of steps to follow?
Should I call model:clearState()
before saving the model and also after loading the model for evaluation?