Small bug in CClassifierPytorch _backward
I found a small bug in CClassifierPytorch._backward
function for zeroing _cached_grad_s
variable.
Steps to reproduce
Entering the
if self._cached_s.grad is not None:
self._cached_s.grad.data.zero_()
section causes an Exception being thrown.
Possible fixes
if self._cached_s.grad is not None:
self._cached_s.grad.data.zero_()
Yeah, wrong low dash position :)