Installation & Set-Up

Installation & Set-Up#

The majority of the packages within CAF are written with Python and therefore require Python to be installed on the users system in order to use them. This section runs through the CAF recommendations for installing Python and our packages.

Python#

The recommended method for installing Python is using miniforge, which will install conda, mamba, and their dependencies, see Miniforge Installation for details. Python, and many other packages, can then be installed using the conda install command e.g. conda install python.

Tip

Most CAF packages are available on PyPI and conda-forge so if you’ve installed Python in a different way then pip install should work for all the released CAF packages.

Warning

CAF packages are not available on the default Anaconda channel, if you have installed Anaconda then the conda-forge channel should be added. A channel can be added as a default using conda config --add channels conda-forge, or can be used for a single install command with -c conda-forge e.g. conda install caf.toolkit -c conda-forge.

Miniforge Installation#

Miniforge is a self-contained installer which will install conda and setup conda-forge as the default channel. During the miniforge installation the installer will provides some options to customise the installation, these two are worth considering:

  • Register Miniforge3 as my default Python: this will allow other programs to detect Python in Miniforge3 on the system.

  • Add Miniforge3 to my PATH environment variable: this will allow conda commands to be called from the default Windows command-prompt, if this is disabled then conda commands should be ran from within the Miniforge3 prompt.

See also

Frequently Asked Questions for the reasons behind CAF’s miniforge recommendation.

Conda Environments#

Conda allows for different versions of Python, and dependencies, to be installed in separate environments. Environments can be setup using the conda create -n environment_name command and then need to be activated using conda activate environment_name.

When inside an environment you can install packages using the conda install command and they won’t affect packages installed in any other environments. Environments can be created, switched between and removed very easily so it is recommended to use separate environments for different tasks, where possible, to avoid having dependency conflicts.

For more information on working with conda and setting up separate environments for different tools and projects see Environments.

Contents