Conda vs Anaconda

On RCAC community clusters we deploy modules for both conda and anaconda. These are both separate distributions of the "conda" package manager. The conda module points towards the Miniforge distribution of conda (https://github.com/conda-forge/miniforge), while the anaconda module loads the official Anaconda, Inc. distribution of conda (https://www.anaconda.com/download).

Although these two distributions behave similarly, the components within them vary slightly:

• Anaconda

  The anaconda distribution of conda provides a comprehensive suite of commonly used scientific and data analysis packages. When fetching packages, it will pull from the defaults channel, a package repository managed by Anaconda, Inc.

• defaults channel (via Anaconda, Inc.)

• Large number of base packages
• Licensed via Anaconda Inc.

• conda package manager

  • classic or libmamba solver

• Miniforge:

  The "Miniforge" distribution of conda (loaded with the conda module) provides only the package manager and a minimal suite of default packages. Further, when fetching new packages it will default to the open conda-forge channel.

• conda-forge channel

• Small number of base packages
• Open Source Licence
• conda package manager
  • classic or libmamba solver

	Anaconda ( anaconda module)	Miniforge (conda module)
Default channel	defaults (Anaconda, Inc.)	conda-forge (community-maintained)
Base packages	Large set of scientific & data analysis tools	Minimal
License	Anaconda Terms of Service	Open source
Package manager	conda (classic & libmamba solvers)	conda (classic & libmamba solvers)

Default Environment Locations

Specifically on RCAC, the default location for environments created via conda and anaconda also differ.

• Environments created with the conda module will all install to: ~/.conda/envs/
• Environments created with the anaconda module will all install to a specific subfolder for that version: ~/.conda/envs/$RCAC_ANACONDA_VERSION

Which One Should You Use?

In general, we reccomend using the conda module (Miniforge) for building environments, as it is lighter weight, open source, and defaults to the broad conda-forge channel with more up-to-date packages. The anaconda module can be useful if you need a one-off “general” Python environment with many essential packages (including Jupyter notebook support) preinstalled.

Side Note on conda init

Typically when you set up a conda distribution, you are prompted to run conda init. This command modifies your .bashrc to define a shell function for conda, such as:

type -a conda
conda is a function
conda () 
{ 
    \local cmd="${1-__missing__}";
    case "$cmd" in 
        activate | deactivate)
            __conda_activate "$@"
        ;;
        install | update | upgrade | remove | uninstall)
            __conda_exe "$@" || \return;
            __conda_reactivate
        ;;
        *)
            __conda_exe "$@"
        ;;
    esac
}

When you load conda or anaconda via RCAC’s modulefiles, this function is already provided by the module, so there is no need to run conda init on RCAC systems! In fact, having the conda init block in your .bashrc can cause ussues in seemingly unrealated areas.