Glad that you enjoyed the article and thanks for the feedback! The primary reason why I developed this approach to injecting a Conda environment inside a Docker image is that I work with many users that are very comfortable using with Conda but need a little help transitioning to Docker. This article describes a general approach for users who want to develop projects inside a Conda environment to then inject that environment into a Docker image without having to write a Dockerfile from scratch.

FWIW I have found that almost everything available via pip is also available on Conda Forge. Most of my projects, and many of my users/clients’ projects, have non-Python dependencies that are available for installation via Conda but not via pip. If you work with GPUs, you should also be aware that a major advantage of Conda relative to pip is that you can manage all of your NVIDIA CUDA dependencies via Conda. By managing your CUDA dependencies via Conda you can use a standard Ubuntu base image for your Dockerfiles (instead of using NVIDIA images).