Spectral norm does not work with convolution layers?

I am not sure if I am doing something wrong but when I do the exact thing that the documentation does for linear layers when it comes to spectral normalization, the norm isn’t equal to 1.

snm = nn.utils.parametrizations.spectral_norm(nn.Conv2d(1, 3, 3))

print(torch.linalg.matrix_norm(snm.weight, 2))

Am I doing it wrong? if so what can I do to fix it?