Machine Learning Engineering in Action

Appendix B. Setting up a development environment

There are many reasons to start with a fresh slate when working on a new project. The following list shows a few of the more relevant ones with respect to ML project work:

• Dependency management is easier with a clean environment.
• Isolation of temporary files, logs, and artifacts is simpler.
• Scripted environment creation makes porting to production easier.
• Installation of libraries is less complex with fewer dependency collisions.

While many options exist for creating isolable environments for development of new projects, this appendix provides guidance on using Docker along with Conda’s package management suite of tools, just as the companion repository to this book does.

B.1 The case for a clean ...