Correctly measure inference in C++

Questions
2

It seems that module.run is non blocking call. So your time measurement should be after module.get_output(0) call.

More Info : Tvm Inference peformance almost 10x better than pytorch - #2 by vinx13