sorry for my poor English.
Thanks your code again.
could I ask another question? About the seq2seq algorithm.
In my opinion ,It use the last word's hidden state to represent a sentence(no matter forward or reversed lats ),then use another lstm to decode it. In your code it is :
does it mean your use every word(not the last word)to the next lstm to decode
sorry for my poor English. Thanks your code again. could I ask another question? About the seq2seq algorithm. In my opinion ,It use the last word's hidden state to represent a sentence(no matter forward or reversed lats ),then use another lstm to decode it. In your code it is :
does it mean your use every word(not the last word)to the next lstm to decode