🐛 Bug Report

guess_bounds output differs between numpy<2 and latest numpy when points have dtype np.float32 . Casting points to np.float64 appears to revert to the old output.