Nicolas

@andreamad8 @polakowo @yangjianxin1 @LHolten thank you for releasing your code! Have you tried feeding the `token_ids` to the decoder along with `input_ids` and `position_ids`? Seems like it breaks the model's...

Here are the inputs for a sample dialog (`"hi, man!", "oh, hi!", "what is your name, man?"`) that I pad from the left and feed to the decoder: ``` input_ids:...

@andreamad8 Thanks a lot for your response! @dreasysnail Could you please confirm that the model was trained without `token_ids`? Or are we using them in the wrong way?

Got it, thanks a lot for the clarification!

Guys, I got the similar lame results yesterday... My guess is that there are some foundational problems in this approach: - Since word2vec vectors are used for words representations and...

@farizrahman4u not yet, I'll set the experiment with Attention Seq2seq now. Meanwhile could you please post the link here to your dataset? And some results example.