Add data-science-image Docker image
requested to merge 13166-fix-the-ds-airflow-pipeline-to-use-the-data-science-image-rather-than-the-analyst-image into master
Relates #13166
This MR attempts to introduce a new image for the data science team, starting from the jupyter/tensorflow-notebook:python-3.9.6
as a base image (which already contains miniforge, with both conda and mamba package managers, details here.) and installing our Data Science dependencies, as well as PyEarth.
The image builds and installs all necessary dependencies, although PyEarth with warnings (as usual), but I have one main concern:
-
pipfile
&pipfile.lock
need to be copy/pasted in this repository. This is sub-optimal, because ideally we would have a single source of truth (that being the data-science repo) for the dependencies, otherwise we need to copy them over and update the image every time they change.
A simple idea would be to host this Dockerfile
on the data-science repository itself?
However, I am open to any other ideas / suggestions.
Edited by Rigerta Demiri