In the previous LSTM tutorial, we used a single character at a time, now thebigram approach is to predict a character using two characters at a time.