重塑Keras中的张量列表

Question

如何将输入传递给连接层，例如将其传递给TimeDistributed层时，输出的形状应该为（None，1，1536）而不是（None，32，1536）？

input_shape = (32, 32, 32, 4)
cnn_count = 5

def create_cnn_lstm_model():

    model = create_shared_weight_cnn()

    for i in range(cnn_count):
        temp_name_input = 'input_' + str(i)
        globals()[temp_name_input] = Input(shape=input_shape)

    combined_input = []
    for i in range(cnn_count):
        temp_name_input = 'input_' + str(i)
        combined_input.append(globals()[temp_name_input])

    seq = concatenate(combined_input)
    out = TimeDistributed(Lambda(lambda x: model(combined_input)))(seq)
    out = LSTM(512)(out)
    out = Dense(4, activation='softmax')(out)

    cnn_lstm = Model(inputs=combined_input, outputs=out)

    return cnn_lstm