Currie32 / Spell-Checker

A seq2seq model that can correct spelling mistakes.
213 stars 93 forks source link

multi-layer lstm implemention #12

Open turned2670 opened 5 years ago

turned2670 commented 5 years ago

Hi, thank you for your open-source program. I'm confused with how the multi-layer lstm was implemented. Without the MultiRNNCell, you leveraged the for loop and created a lstm for each layer. But I can't find out how the two layers are connected, because the inputs for the lstm of the second layer are still 'rnn_inputs', not the outputs from the previous layer.

Could you please point out how the two layers connected? Looking forward to your reply.

Thanks a lot.