generate heatmap of attention weight when use simple_attention in seq2seq
Created by: gitforhf
as described in this issue https://github.com/PaddlePaddle/Paddle/issues/1269
attention weight的heap map如下图。该功能有现成的evaluator可以直接调用。
This requirement has been supported, but no solution has been given。 is there demo or tools to implement this feature。