将tf.nn.dropout（）应用于3DCNN时出现问题

Question

我想将tf.nn.dropout应用于3D-CNN层

但我面临着错误"ValueError: Can't convert non-rectangular Python sequence to Tensor."

代码是

inputs = layers.Input(shape=Xshape)

model1 = layers.Conv3D(64, input_shape=Xshape, kernel_size=(7, 7, 7), padding='same', activation='relu')

y=model1(inputs)

weights_as_numpy = model1.get_weights()

x=tf.nn.dropout(weights_as_numpy, rate = 0.5, seed = 1)

然后，我使用了data_tensor = tf.ragged.constant(weights_as_numpy)

但是我遇到了另一个错误：ValueError: all scalar values must have the same nesting depth

你能帮帮我吗

先谢谢你