具有BatchNormalization的Tenshorshape错误

时间:2019-10-21 05:17:27

标签: tensorflow keras

我很确定我的tenshorshapes很好,但是我通过batchNormalization不断收到此错误

 ('Input has undefined rank:', TensorShape(None))

这是我代码的相关部分

input_layer = Input(shape = (32,32,3), name = "Input_Layer")
initial_conv = Conv2D(16,
                  (3,3),
                  padding="same",
                  name = "first_conv")(input_layer)

##First Dense Layer

Batch_1 = BatchNormalization(axis = -1,
                         name = "Batch_1")(initial_conv)
activation_1 = relu(Batch_1)
Conv_1 = Conv2D(16,
            (3,3),
            padding = "same",
            use_bias = False,
            name = "Conv_1")(activation_1)

conc1 = Concatenate([initial_conv,Conv_1])

##Second Dense
Batch_2 = BatchNormalization(axis = -1,
                         name = "Batch_2")(conc1)

在第一批之后,每隔一批我都会遇到相同的错误。任何人都知道为什么会这样吗?

0 个答案:

没有答案