CPU implementation

[maksad]

I have tried PreciseRoIPooling in my research and was satisfied with the result, now I would like to deploy my model using TorchServe and Amazon Elastic Inference (EI). One of the requirements on EI is "CPU compatibility" and having that variation of PreciseRoIPooling would greatly help. Is there any timeline about it?

[vacancy]

Currently, there is no plan for CPU forward/backward support.

[cybaol]

Is there any plan to achieve CPU version？

[evekeen]

Without a CPU implementation, there is little practical use in the library besides research.
I'd really like to try porting it to CPU however, I'm far from understanding how it works. Can anyone help?