RuntimeError: Backward is not reentrant

[FCInter]

I compile the project and run the test.py. I got the error

RuntimeError: Backward is not reentrant, i.e., running backward with same input and grad_output multiple times gives different values, although analytical gradient matches numerical gradient

It seems the gradient check is not passed. I'm using python 3.6 with pytorch 1.0.0.

What's wrong with it?

[chengdazhi]

Try pytorch 0.4.1?

[FCInter]

@chengdazhi Sorry limited by environment, I can only use 1.0.0. What if I ignore this error and directly use it in my network? Will this be a problem?

[chengdazhi]

I suggested you use the dcn operators in mmdetection repo, they support pytorch 1.0.0+

[Andyflying]

@FCInter I met this issue too, did you solve it?