In KMeans intro, why do we use np.random.seed(1) before fitting the model with X target?

[HeyItWorked]

attempt: from StackoverFlow, i know that the function makes random numbers predictable, having np.random.rand() after calling np.random.seed() will generatate the same set of random number, is that not the case here or which function uses randomization from numpy that requires seed setting?

[HeyItWorked]

same issue arises when generating random color centroids, if i use colors = np.random.default_rng(1).uniform(0,1,(x,3)), value of x past 10 will give error mean of empty slice and same number of labels changed throughout all iterations (at 27k labels so entire data). However with np.random.seed(0) with colors = np.random.uniform(0,1,(x,3)), i was able to do iterations for 12 colors max. Please advise