Python⚓︎

Apparently, Python won't come bundled with macOS starting with v13. And only Python 2 seems to be bundled with versions as recent as Monterey.

pyenv⚓︎

While Anaconda is a good environment manager, pyenv is the simplest and most lightweight option for pure Python. Once pyenv has been installed with Homebrew, install a recent version of Python with

pyenv install <version>

Multiple installations are supported, so I can have separate Python 2 and Python 3 environments, or even separate Python 3.X ones. Each is bundled with pip for external package dependency management. The active environment can be found with

pyenv local

and can be switched to any other installation (applicable to the active directory) with

pyenv local <version>

Change the global default Python version with

pyenv global <version>

Anaconda⚓︎

For completeness, below are instructions to install Anaconda with Homebrew:

brew install --cask anaconda
conda init "$(basename "${SHELL}")"  # (1)
<exit and reopen shell>
conda activate
conda config --set auto_activate_base false
conda deactivate

If conda executable doesn't work, run $(brew --prefix)/anaconda3/bin/conda

Then, to install the Anaconda Navigator (a GUI for managing conda environments and installing related tools), do

conda activate
conda install -c anaconda anaconda-navigator -y
conda deactivate

Dependencies⚓︎

See https://github.com/eshwen/ds-python-boilerplate for a list of common Python dependencies for data science.

Tensorflow⚓︎

If pip install tensorflow doesn't work, it may because it does not yet run natively on Apple Silicon. Run the following commands to get a usable installation:

pyenv install miniforge3
pyenv local miniforge3
conda install -c apple tensorflow-deps -y

pip install -U pip
pip install tensorflow-macos
pip install tensorflow-metal  # Metal plugin for GPU acceleration

Then, before using Tensorflow, do either pyenv local miniforge3 to activate the environment it's installed in.