def test_parser_switch_layer_contain_diff_conv_same_parameter():

    class FinalConvNet(Cell, MetaFactory):
        def __init__(self):
            super().__init__()
            MetaFactory.__init__(self)
            self.conv = nn.Conv2d(1, 1, 3, weight_init='ones', pad_mode='pad')
            self.conv2 = nn.Conv2d(1, 1, 3, weight_init='normal', pad_mode='pad')
            self.weight = Parameter(Tensor(np.random.randn(1, 1, 3, 3).astype(np.float32)), name="aaa")
            self.conv.weight = self.weight
            self.conv2.weight = self.weight
            self.funcs = (self.conv, self.conv2)

        def construct(self, i, inputs):
            x = self.funcs[i](inputs)
            return x

    net = FinalConvNet()
    i = Tensor(0, mstype.int32)
    input = Tensor(np.ones([1, 1, 224, 224]).astype(np.float32))
    out = net(i, input)

    out_good = net.conv(input)
    allclose_nparray(out.asnumpy(), out_good.asnumpy(), 0.0001, 0.0001)

    back_net_good = GradOfAllInputsAndParams(net.conv, sens_param=False)
    back_net = GradOfAllInputsAndParams(net, sens_param=False)
    back_out = back_net(i, input)
    back_out_good = back_net_good(input)

    allclose_nparray(back_out[0][1].asnumpy(), back_out_good[0][0].asnumpy(), 0.0001, 0.0001)
    allclose_nparray(back_out[1][0].asnumpy(), back_out_good[1][0].asnumpy(), 0.0001, 0.0001)