Fix the bug due to deprecation of np.float

Created by: didi-hou

This PR fixed the bug due to the deprecation of np.float. It replaced the np.float with python built-in float and it's equivalent to np.float64