How to train the encoder decoder from sequence of features

Open ducanh841988 opened this issue 8 years ago • 1 comments

Apr 27 '17 07:04 ducanh841988

Dear Eriguchi, I want to train the encoder decoder directly from sequence of features instead of features in dictionary. I try to change your code but it is impossible for me. Currently, your source code is trained from [sentence --> sentence]. Then word embedded is employed for getting features for input sentence and output sentence. In my case, I want to train from [sequence of features --> sentence]. So I just use word embedded for output sentence. Can you help me?

Apr 27 '17 07:04 ducanh841988