Results
1
issues of
Kai Shu
It seems that the Attention layer is not properly computed. In the original paper, the vectors are computed as the weights sum of the weight and hidden state (h_i), but...