gngdb
commented
9 years ago Dockerization is not possible on university machines outside of a virtualbox, so probably impossible. Maybe useful for Amazon EC2 instances, but probably not. Also, we'd rather not use those because they cost money.
Basically, we want to be able to put things in a virtual environments, so we can run tests and verify everyone has the correct packages installed; and that they're installed right. That's why we're using Python virtual environments. But, we can't using Python environments for things not written in Python (obviously), such as Caffe, OpenCV and anything else we might want to use.
A good option to solve this problem is docker, which containerizes applications (read not virtualization, ie much less overhead). Then we can run tests on these applications, quickly spin up virtual machines on Amazon EC2 and verify everyone is running the same thing.
For all of these things though, we need to make sure we can use the GPU from inside the docker container. It looks like this is possible, but we should make sure it can be done.
To close this issue should get a GPU working inside a docker container on Amazon EC2 or on stonesoup.
If you think this is unnecessary and we should just get on with the real work, you might be right. I think it could save enough time in the long run and might be very useful to learn in the even longer run(PhD).