LSTM层在期望它们不接受任何输入形状时会接受

Question

我正在使用keras构建神经网络，我对LSTM层输入形状有些迷茫。下面是相关部分的图片。

两个塔都是相似的，唯一的区别是左侧接受任何长度的序列，而右侧仅接受长度5的序列。这导致它们的LSTM层分别接收不明确的序列长度和4的序列长度，两者都具有每步8个功能。因此，我希望两个LSTM层都应具有（1,8）的input_shape。

我现在感到困惑的是，两个LSTM层都可以毫无问题地接受任何输入形状，这就是为什么我认为这可能无法按我认为的方式工作的原因。我希望正确的LSTM层要求输入形状的第一维为1、2或4，因为只有这些尺寸才能划分输入序列4。此外，我希望两者都都需要第二维永远是8。

有人可以解释一下为什么LSTM层可以接受任何输入形状，以及它们是否使用input_shape =（1,8）正确处理序列？下面是相关代码。

 # Tower 1
 inp_sentence1 = Input(shape=(None, 300, 1)) 
 conv11 = Conv2D(32, (2, 300))(inp_sentence1) 
 reshape11 = K.squeeze(conv11, 2)
 maxpl11 = MaxPooling1D(4, data_format='channels_first')(reshape11) 
 lstm11 = LSTM(units=6, input_shape=(1,8))(maxpl11)

 # Tower 2
 inp_sentence2 = Input(shape=(5, 300, 1)) 
 conv21 = Conv2D(32, (2, 300))(inp_sentence2) 
 reshape21 = Reshape((4,32))(conv21)
 maxpl21 = MaxPooling1D(4, data_format='channels_first')(reshape21) 
 lstm21 = LSTM(units=6, input_shape=(1,8))(maxpl21) 

编辑：伪数据中问题的简短再现：

    # Tower 1
    inp_sentence1 = Input(shape=(None, 300, 1))
    conv11 = Conv2D(32, (2, 300))(inp_sentence1)
    reshape11 = K.squeeze(conv11, 2)
    maxpl11 = MaxPooling1D(4, data_format='channels_first')(reshape11)
    lstm11 = LSTM(units=6, input_shape=(1,8))(maxpl11)

    # Tower 2
    inp_sentence2 = Input(shape=(5, 300, 1))
    conv21 = Conv2D(32, (2, 300))(inp_sentence2)
    reshape21 = Reshape((4,32))(conv21)
    maxpl21 = MaxPooling1D(4, data_format='channels_first')(reshape21)
    lstm21 = LSTM(units=6, input_shape=(1,8))(maxpl21)

    # Combine towers
    substract = Subtract()([lstm11, lstm21])
    dense = Dense(16, activation='relu')(substract)
    final = Dense(1, activation='sigmoid')(dense)

    # Build model
    model = Model([inp_sentence1, inp_sentence2], final)
    model.compile(optimizer='adam',
                   loss='binary_crossentropy',
                   metrics=['accuracy'])

    # Create data
    random_length = random.randint(2, 10)
    x1 = numpy.random.random((100, random_length, 300))
    x2 = numpy.random.random((100, 5, 300))
    y = numpy.random.randint(2, size=100)

    # Train and predict on data
    model.fit([x1, x2], y, epochs=10, batch_size=5)
    prediction = model.predict([x1, x2])
    prediction = [round(x) for [x] in prediction]
    classification = prediction == y
    print("accuracy:", sum(classification)/len(prediction))