在静态图中对variable矩阵使用fluid.layers.data_norm归一化，报如下问题

Error Message Summary:

PaddleCheckError: op data_norm does not have kernel for data_type[float]:data_layout[ANY_LAYOUT]:place[CUDAPlace(0)]:library_type[PLAIN] at [/paddle/paddle/fluid/framework/operator.cc:1010]
[operator < data_norm > error]

定义语句为：
weight = fluid.layers.create_parameter(shape=[self.in_dim, self.out_dim],
dtype='float32', default_initializer=fluid.initializer.Xavier(uniform=True))

    weight_norm = fluid.layers.data_norm(weight, name='weight')

去掉该归一化步骤可以正常运行，请帮忙看下该op的问题如何解决