0

我真的希望有人可以帮助我。我正在尝试构建一个具有 2 个输出 [[1 个数字]、[3 个数字]] 的网络，而我想使用两个不同的损失函数。我也使用生成器。所以我的问题是将 keras 生成器与多输出、多损失网络结合起来。在下面，我将向您展示我的部分代码以及我收到的相应错误消息：

发电机输出

...    
while True:
            it1 = gen1.next()
            it2 = gen2.next()
            it3 = gen3.next()
            it4 = gen4.next()
    
            # If generator is used for training and validating
            if to_predict == False:
                
                yield [it1[0], it2[0], it3[0]], [it1[1], it2[1], it3[1], it4[1]]  #Yield both images and their mutual label

网络

...
        model = keras.Model(inputs=[Input1, Input2, Input3], outputs=[Output1, Output2])


        OPTIMIZER = keras.optimizers.Adam(HP_lr)
        
        ### Compile
        model.compile(loss = {'out1': Loss1(), 
                              'out2': Loss2()},
                      loss_weights={'out1': 1,
                                    'out2': 1}, 
                      metrics = {'out1': Loss1(),
                                 'out2': Loss1()},
                      optimizer=OPTIMIZER
                     )

调谐器

...
atuner.search(train_generator,    ...)

错误信息

ValueError: The two structures don't have the same sequence length. Input structure has length 4, while shallow structure has length 2.

4

0 回答 0