26hzhang / neuralnetworkslite Goto Github PK

Hi Isaac, I was reading your code ConstructGraph.java in encdeclstm. There are two questions:

1. In regular seq2seq, the information passed to decoder includes decoderInput and the hidden unit. The hidden unit value is the thoughtVector in your code. And it usually passed to the decoder as its first hidden unit value. However, I see in your code, it uses as input together with decoderInput. Do you know how to use it as the normal hidden unit value?
2. I wonder if you could add one more embedding layer to decoderInput also.

I am new to deeplearning4j. I have played with it for some time but still cannot figure it out. Thank you so much for your time!