Weights & Biases - Intro to Pyenv for Machine Learning

If you jump between a lot of different machine learning projects, you probably find yourself running something like pip install -r requirements.txt By default, pip install puts libraries in your systemwide libraries folder. If one of your projects has requirements that conflicts with another, switching to that project and running pip install will effectively break your other project by modifying the systemwide python libraries it needs to run. Worse yet, many projects haven't fully moved to python3 yet! So you may find yourself juggling systemwide requirements across python2 and python3. A lot of folks use docker for this type of workflow issue.