At UC Analysis Facility we provide multiple docker images. Two of them, AB-stable and AB-dev, come from a different repo. The ML platform images come from different branches of this repository. All of them are based on ML base, which is in turn based on nvidia cuda image. Github actions build and upload images to dockerhub/ivukotic and Harbor.
The definitions of the four images can be seen here: legacy, current, conda, julia.
Provides ancient python3.6 and python3.8 versions and come with full set of python data analysis packages (tensorflow, keras, visualizations, dask, uproot, jupyter, etc.)
Provides the same packages as the Legacy image but comes with python 3.12.
Unlike Legacy and Current images Conda image is based on a newer Nvidia Cuda image. Despite its name, it does not come with Conda but with Micromamba, which is just a CLI tool to manage Conda enviroments and works in exactly the same way.
This image is maintained by Jerry Ling. It comes with Python 3.8 and Julia 1.11.6.