We tried to build BIRNN encoder and decoder with anttention to get seq2seq2 translations.

*build encoder - decoder model :BiRNN + Attention

#BiRNN：The encoder will receive a sequence of words, a LongTensor of size (max_len x batch_size), and output an encoding for each word, a FloatTensor of size (max_len x batch_size x hidden_size).

Given Sentence :X = (x1,x2,..,xt) Set ht=f(xt,ht-1) and c=q(h1,

Each decoder output is conditioned on the previous outputs and some 𝐱 , where 𝐱 consists of the current hidden state and the attention "context", p(y_i |m{y_1,...,y_{i-1}, 𝐱) = g(y_{i-1}, s_i, c_i) s_i = f(s_{i-1}, y_{i-1}, c_i) where ci is context vector,where each weight a_{ij}is a normalized attention "energy"

Training: we use max_size = 25 and max_size= 40 to train the data

Reference

https://zhuanlan.zhihu.com/p/57155059