large difference in computing transformation matrix W on different device

[KarenMars]

Hi,

I obtained two different transformation matrix W on different machines, one is on my own desktop (Intel(R) Core(TM) i5-10210U CPU @ 1.60GHz 2.11 GHz), one is on the server machine ( Intel(R) Xeon(R) Gold 5118 CPU @ 2.30GHz), I ensure that the versions of depending packages are all the same.

        u, s, vh = np.linalg.svd(cov)
        self.W = np.dot(u, np.diag(1/np.sqrt(s)))

Does anyone have the same problem and have a solution of it, thank you very much!