seq2seq.py 文件源码

def encode_final_state(self):
        context_vector       = self.input[-1, ]
        tiled_context_vector = T.tile(context_vector, (T.sum(self.dur_input), 1))

        return tiled_context_vector