Building a custom rstudio with cuda libraries

[sneumann]

What operating system related to this question?

Linux

Question

Hi,
we'd like to build a custom rstudio with cuda support installed, and need some nudging in which direction to explore options.
More precisely, we'd like to build on top of bioconductor/bioconductor_docker:RELEASE_3_* ,
which in turn is build on top of a rocker/rstudio which eventually builds on vanilla ubuntu:noble.
I also know about the rocker/ml images, which build directly on top of e.g. nvidia/cuda:11.8.0-cudnn8-devel-ubuntu22.04.
=> How to combine these for a local franken-rstudio ?

1. Is there any sane way to use multi-stage builds and combine from bioconductor_docker and rocker/ml ?
2. Is it more promising to grab the respective Dockerfiles, and edit them locally to build the FROM chain
   rocker/ml => rocker/myrstudio => bioconductor/mybioconductor_docker => my_ml_rstudio_bioc_frankenstein ?
3. Anything simple I missed ?

This is a bit related to #1 in this repo :-)

Thanks in advance,
yours,
Steffen

[sneumann]

Hi, after finding https://hub.docker.com/r/bioconductor/ml this question can be close. Yours, Steffen