Closed mkanai closed 1 month ago
@KangchengHou , I think you implemented this part. Should it be changed to np.float32
instead of np.float64
to keep the low-precision efficiency?
I believe the speed is not a main concern here - np.float64
should do fine
This PR fixes the following error caused by the deprecation of
np.float_
in the NumPy 2.0 release