This is unable to run because np.float_ is now replaced by np.float64 in the NumPy 2.0 release (see image for error details)

#2
by adi-0 - opened

image

Sign up or log in to comment