The distance result changes when I run several times?

[imerse]

As the title, I wonder does there have a random process in the flow ? Why the distance changes each time?
And does the forward process take more time on gpu than cpu? I test on v100 gpu with 402ms for the first pair and about 50ms on cpu. Looks strange! Any help will be appreciated! Thanks.

[richzhang]

Hmm, I'm a little confused wy it's loading a resnet18 model. Were you able to run the example command and get this?

[imerse]

Sorry, I misunderstood the location of the automatically downloaded model. Now the result of each inference is the same, but I still don't understand why the efficiency of running on GPU is not as good as on CPU.

[richzhang]

May be that communication dominates when running a single image. Try for a bigger batch of images and see if the trend still holds