Error-tf. function-decorated function tried to create variables on non-first call

ValueError: tf.function-decorated function tried to create variables on non-first call

stay tensorflow2 The following saved models always report errors , It's a headache :

model.save(weight_file)

The wrong position is fixed in :

out_data = tf.keras.layers.Conv2D(filters, [3,3])(concat_data)

Finally find out why , You need to be in __init__ Defined in the function body , Can be in call Call in the :

    def __init__(self,**kwargs):
        super(Up_sampling, self).__init__(**kwargs)
        self.Conv2D_256 = tf.keras.layers.Conv2D(256, [3, 3])
    def call(self, input,filters):
        if filters==256:
            out_data = self.Conv2D_256(input)

This will solve the problem well , The error disappears !