我正在使用编解码器网络进行文本汇总
编码器是双向GRU
解码器是一个简单的GRU
这是我尝试过的
## ENCODER
encoder_inputs = Input(shape = (MAX_LEN,), dtype = 'int32')
encoder_embeds = embedding_layer(encoder_inputs)
x, forward_h, backward_h = Bidirectional(GRU(MAX_LEN, return_sequences = True, return_state = True))(encoder_embeds) #return state is by default True
encoder_state = Concatenate()([forward_h, backward_h])
## DECODER
decoder_inputs = Input(shape = (MAX_LEN,), dtype = 'float32')
decoder_embeds = embedding_layer(decoder_inputs)
decoder_outputs, state_s = GRU(MAX_LEN, return_sequences = True, return_state = True)(decoder_embeds, initial_state = encoder_state)
这是我得到的错误
An `initial_state` was passed that is not compatible with `cell.state_size`. Received `state_spec`=[InputSpec(shape=(None, 198), ndim=2)]; however `cell.state_size` is 99
我知道错误表明当期望99时，它收到的第二维为（99 + 99 =）198
那
由于 Concatenate（）操作
而发生错误当我要考虑两个方向上的最后一个隐藏状态时， Concatenate（）的替代方法是什么