Question about selective read

[JuliRao]

I've read the code in models/copynet.py. I think in the third step 'get weighted attention to use for predicting next word', input_idx(y_t) should be used instead of input_idx(y_(t-1)), which is then used to generate y_(t+1).

[SilverLiningSL]

Have you figured it out? This is also where I get puzzled.