train problem about FloatingType

[zhaoguangcheng]

When I run the code ,I meet a questions.
RuntimeError: Expected isFloatingType(grads[i].scalar_type()) to be true, but got false. (Could this error message be improved? If so, please report an enhancement request to PyTorch.) (validate_outputs at ..\torch\csrc\autograd\engine.cpp:476)
(no backtrace available)
I don't know how to solve it.I would appreciate it if you could help me.