How we should convert an array containing np.complex128 datatype (say a complex 2d numpy array) into a torch .complex128 array of the same dimension?
torch.from_numpy should work:
a = np.random.randn(10, 10)
a = a.astype(np.complex128)
print(a.dtype)
# complex128
x = torch.from_numpy(a)
print(x.dtype)
# torch.complex128
Thank you ptrblck for solving the issue. The same command was not working a few days ago and now it is working well.