Hi, I am confused that you do not use the context in you code which is different from your paper

[bradley-code-again]

Hi, thanks for your wonderful sharing.
In a3c_gcn_seq2seq/net.py, method forward(...) of class Decoder, "context, attention = self.att(hidden_state, encoder_outputs, mask)", but the context is not used anywhere in the code.
From the paper, it should be used as part of the input to generate action, but in the code it is not.
I am confused on this, if you can give a bit more explanation, many thanks.

[GeminiLight]

Due to code refactoring, there is indeed a bug here. Thanks for your feedback, I'll fix it.