search

gabrielloye / Attention_Seq2seq-Translation

34 stars 23 forks source link

attn_combine is defined but not used in BahdanauDecoder forward process #2

Closed RHzhongzju closed 4 years ago

RHzhongzju commented 4 years ago

Hello, Gabriel! Thanks for your article on the Attention Mechanism! In the article, you provided the equation to calculate the alignment scores for Bahdanau Attention, but in the code, the attn_combine was defined but not used in the forward process.

gabrielloye commented 4 years ago

Hi @RHzhongzju, Thanks for pointing that out! Yes, defining attn_combine was unnecessary as it isn't used in the calculation of alignment scores and I've since removed it from the notebook.