Do we need pytorch package? about ros2-vicon

[skim0119]

I think it works either way. I've seen docker images including pytorch/tensorflow natively, because they are big packages to install.

Within the image, we run some cache-cleanup to reduce the image size, but those steps will be neglected if we let users to install packages in each container. Downside is that we might need to provide cpu-image and gpu-image separately.

At this point, we can just specify in readme that python/pytorch are natively provided, and indicate their version. We let users to re-install if they need other version.

[hanson-hschang]

If we need to provide both the cpu-image and gpu-image, it might be better to remove the pytorch and let users to install there own package... It doesn't take too much time to install I believe. (At least for me today when I install the FastRodReconstruction repository)

Also, the python version is needed to be specified as well, I think.