Keras LSTM input_shape 的问题：预期 lstm_1_input 的形状为 (500, 2) 但得到的数组的形状为 (500, 5)

Question

x_train 和 y_train 是我的模型的输入和输出，形状分别为 (6508, 500, 5), (6508, 5)。

而模型是这样的：

model = Sequential()
model.add(LSTM(units=96, return_sequences=True, input_shape=x_train.shape[1:]))
model.add(Dropout(0.2))
model.add(LSTM(units=96, return_sequences=True))
model.add(Dropout(0.2))
model.add(LSTM(units=96))
model.add(Dropout(0.2))
model.add(Dense(units=5))

model.compile(loss='mean_squared_error', optimizer='adam', metrics=['mse'])

model.fit(x_train, y_train, epochs=epochs, batch_size=batch_size)

模型总结：

Model: "sequential_1"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
lstm_1 (LSTM)                (None, 500, 96)           39168     
_________________________________________________________________
dropout_1 (Dropout)          (None, 500, 96)           0         
_________________________________________________________________
lstm_2 (LSTM)                (None, 500, 96)           74112     
_________________________________________________________________
dropout_2 (Dropout)          (None, 500, 96)           0         
_________________________________________________________________
lstm_3 (LSTM)                (None, 96)                74112     
_________________________________________________________________
dropout_3 (Dropout)          (None, 96)                0         
_________________________________________________________________
dense_1 (Dense)              (None, 5)                 485       
=================================================================
Total params: 187,877
Trainable params: 187,877
Non-trainable params: 0

问题是lstm_1 需要 input_shape (500, 2) 而我的数据形状是 (500, 5)：

ValueError: Error when checking input: expected lstm_1_input to have shape (500, 2) but got array with shape (500, 5)

我打印图层的形状：

for layer in model.layers:
    print(layer.input_shape, end='\t')

# (None, 500, 5)  (None, 500, 96) (None, 500, 96) (None, 500, 96) (None, 500, 96) (None, 96)      (None, 96)

它为lstm_1 打印(None, 500, 5)，所以我无法找出问题所在。

Keras==2.3.0
tf==1.14.0

更新：

使用keras==2.2.5 或tf.keras 可以解决问题。

Answer 1

为了社区的利益，在回答部分提及解决方案。

使用tf.keras 代替keras 已解决问题。