Show HN: Dockerized GPU Deep Learning Solution (Code and Blog and TensorFlow Demo)

[viklas]

https://github.com/emergingstack/es-dev-stack

[flx42_]

I don't understand why you need to do that, tensorflow is already dockerized for GPUs, using the nvidia-docker images: https://github.com/tensorflow/tensorflow/tree/master/tensorf...

[sandGorgon]

Interesting - does anyone know if there is a dockerized install of Numpy with the right blas,etc libraries .

I am lost at figuring out the best way to configure all the dependencies for decent performance.

I don't have GPU, but I suppose that would make a difference?

[__mp]

Does docker support gpus or how does this work out?

[viklas → __mp]

The "--device" flag allows you to map devices through to a Docker container. It runs in 'privileged' mode though, so isn't suitable for a shared host.

Nvidia make it pretty straight-forward now but we had to branch from that approach a bit for the CoreOS deployment.

https://github.com/NVIDIA/nvidia-docker (Nice pictures)

https://docs.docker.com/engine/reference/run/ (Docker documentation, search for 'privileged')

The approach is a bit different depending on your host operating system. You'll also find there are constraints when you introduce a virtualisation layer, like virtualbox or parallels on your desktop - GPUs can be mapped through, but it's painful(ish).

[m-i-l → viklas]

There's also a blog entry about accessing GPU from Docker at http://marconijr.com/posts/docker-exposing-gpu/ .

[__mp → viklas]

Great! Thank you!

[zakk]

Three buzzwords in a single submission! (just joking, looks like a good project!)

[dharma1]

Nice. Does the host OS need to have CUDA installed?

[viklas → dharma1]

The first stage of the process is to take a vanilla CoreOS host and inject the CUDA drivers (one time process). After that, you can reboot the box and still retain the devices, for mapping into docker containers.