As a User, I want OOTB workbench image(s) with AMD libraries & packages

Description

Create the setup for building the AMD-based container image, which supports the jupyter packages and
AMD ROCm. This is similar to existing CUDA (NVidia) and HabanaAI images.

Blockers:

• We need some documentation/github repo to pull the information needed to build the images.
• Which are the libraries and packages

Dev:

• Create the base ROCm Dockerfile with ROCm binaries.
• Check if there is a separate Tensorflow or PyTorch for AMD.
• Build Notebook images (Datascience, Tensorflow, PyTorch) with base ROCm image.
• Test it on an AMD node
• Find ways to verify AMD binaries are set.
  • `rocminfo` 

• Reference:
  • https://rocm.docs.amd.com/projects/install-on-linux/en/latest/reference/docker-image-support-matrix.html
  • https://rocm.docs.amd.com/projects/rocSPARSE/en/latest/Linux_Install_Guide.html
  • https://github.com/ROCm/ROCm
  • https://github.com/ROCm/ROCm-docker 

 QA:

• Test it on the AMD node(might require AMD operator)

Docs:

• Document the AMD workbench in the RHOAI.

Questions 

• Where will be the material we need to fetch to build the images.
  • For NVidia we go to their github and get that information.
• Security Constraints