Docker-stacks: GPU recipe?

[iyanmv]

I know that jupyter/tensorflow-notebook only supports CPU to avoid including NVIDIA propietary software in the project (see https://github.com/jupyter/docker-stacks/issues/516). However, would it be ok to include a recipe in docs/using/recipes.md?

I know that some people (including me) would like to work with an image based on jupyter/base-notebook but with GPU support because official docker images lacks some interesting features.

[parente]

If we're extra careful to point the reader to the NVIDIA EULA and explain that they have to accept the terms of the license to install CUDA and redistribute it in their Docker images, then I'm +0.001 on having it in the recipes.

It might be good to have others weigh in.

[jzf2101]

See jupyterhub/zero-to-jupyterhub-k8s#994 as well as a reference

[beniz]

Believing Jupyter is right to stay away from closed source binaries as provided by NVidia, though driven by immediate industry requirements, we have put a version of this repository with added GPU support, including recipes and an additional one for Tensorflow + Pytorch (see #745 ), here: https://github.com/jolibrain/docker-stacks

The Tensorflow + Pytorch recipe is here: https://github.com/jolibrain/docker-stacks/tree/master/jupyter-dd-notebook-gpu

Note: we're only planning to update the GPU recipes irregularly at the moment.

[parente]

@beniz Thanks for sharing your work.

This issue has been idle for some time now so I'm going to close it. If anyone would like to submit a PR to the recipes or community images doc pages linking to information about GPU-enabled images, I'll be happy to review it.