Stop pip installing packages into your system (base) python: use virtual environments instead

Показать описание

As a lazy python user, I never built virtual environments to isolate my dependencies. Instead I would just pip/conda install until something broke. At which point I'd start over with a clean environment and repeat the cycle. I justified my laziness by saying I wasn't a software engineer, lying to myself that creating separate environments per project was too much of a hassle.

Don't be like me.

It's so easy to create clean, isolated environments per project. So if you want to put some cayenne on one meal but keep it separate from the others, you can do so.

This video shows one simple option, straight from the python docs (venv). There are lots of great tools in the Python world for managing environments and packages, but you can't go wrong getting started with venv.

At Continual, we're focused on bringing software engineering practices to ML teams. Lots of data scientists do cool quantitative stuff in notebooks within their research environment. We're trying to help get that code into production using the tried and true software delivery methods with some slight modifications for ML.