在如下配置，使用modelzoo中的mindspore镜像的基础上

代码：

real = np.random.rand(10, 512)

imag = np.random.rand(10, 512)

all_sig_mixture = real + 1j * imag

sig_mixture_comp = np.stack((real, imag), axis=1)

sig_mixture_comp = mindspore.Tensor(sig_mixture_comp, dtype=mindspore.float32)

net = UNet()

sample = sig_mixture_comp[1:1 + 1, :, :]

output = net(sample)就可以运行

代码：

real = np.random.rand(10, 512)

imag = np.random.rand(10, 512)

# 将实部和虚部组合成一个复数数组

all_sig_mixture = real + 1j * imag

real_part = ops.real(mindspore.Tensor(all_sig_mixture, dtype=mindspore.float32))

imag_part = ops.imag(mindspore.Tensor(all_sig_mixture, dtype=mindspore.float32))

sig_mixture_comp = ops.stack((real_part, imag_part), axis=1)

net = UNet()

sig1_out = mindspore.Tensor(shape=sig_mixture_comp.shape, dtype=mindspore.float32,init=One())

sample = sig_mixture_comp[1:1 + 1, :, :]

output = net(sample)

就会出现报错