Attempt to fix #1412. In principle we could replace the default dtype in backends to something more useful (e.g. type(x)) however this is not trivial because this function was designed for state vectors.
PS. I prefer to dtype=np.uint8 however torch does not like it.
Attempt to fix #1412. In principle we could replace the default dtype in backends to something more useful (e.g. type(x)) however this is not trivial because this function was designed for state vectors. PS. I prefer to
dtype=np.uint8
however torch does not like it.Checklist: