Do you have any details on your experience training the models? Was this using vanilla RNNs, LSTM cells, did you use an attention mechanism, etc.? How long did you train for?