Maintenance: Oct 15, 2024

Miniforge and Anaconda are both popular tools for managing Python environments and packages, but they differ in a few key ways:

1. Size and Preinstalled Packages:

• Anaconda base environment came with a large number of preinstalled data science libraries. However, miniforge only includes the essential Conda and Mamba package managers along with commonly used packages such as numpy, pandas, matplotlib, etc., from the conda-forge channel.

2. Default Package Channels:

• Anaconda: Uses Anaconda’s proprietary channels (Anaconda repository) for package installations by default. These packages may have specific licensing restrictions.
• Miniforge: Uses conda-forge as its default channel, an open-source community-driven repository, ensuring more transparency and flexibility without proprietary limitations.

3. Licensing:

• Anaconda: The default Anaconda distribution has licensing restrictions for commercial and research use, requiring a paid license for certain types of usage.
• Miniforge: Has no such restrictions since it uses conda-forge, which provides fully open-source packages.

The use of your existing environments should not be affected by this change. For instructional use you may continue to install python packages from the licensed Anaconda default channels as shown in the example below.

conda install -n path-to-my-conda-enf seaborn -c anaconda

However, any use of such environment for research purposes is a violation of the Anaconda license unless you obtained your own license. Therefore, for research use, it is expected to replace packages installed through the anaconda restricted channels with packages from non-proprietary channels such as conda-forge.

The process and commands for creating conda environments using Miniforge are exactly the same. The only difference is that you need to load the Miniforge module instead of the Anaconda module on our system.Basically,

module load miniforge

conda create -n your_env_name_goes_here (default Python version: use conda info to find out)

Miniforge avoids violating Anaconda’s Terms of Service because it pulls packages from the conda-forge channel by default. Conda-forge is a community led collection of recipes, build infrastructure and distributions for the conda package manager and is free to use.

Existing environments will not be removed, however, any packages installed in your conda environments via the Anaconda default (proprietary) channel will need to be reinstalled through a different channel or tool such as conda-forge or pip.

The Miniforge module includes the conda package management system. You can use conda install <package-name> as you may have done previously using the Anaconda module. Miniforge uses the conda-forge channel by default.

If you are using Anaconda for instructional use or have your own license, you can install packages from the anaconda channel using conda install -c anaconda <package-name>

Yes. Miniforge supports installation of non-Conda packages with pip and uses the same syntax.

Yes. Miniforge supports installation of non-Conda packages with pip and uses the same syntax.

Once the environment is activated you an update packages and environments in the same fashion as you would with Anaconda. To activate all packages run:

conda update --all

To activate a single package you would run:

conda update <name>

Scripts using Anaconda packages that are also available within Miniforge may not break. However, any Anaconda packages that are not used soley for instructional purposes would be violating the Anaconda license terms, so they would need to be reinstalled through a different channel or tool such as conda-forge or pip.

The conda-forge channel is set as the default (and only) channel for Miniforge.

The Anaconda repository can only be used with miniforge provided that the packages are used strictly for instructional purposes.

First, you’ll need to load the Miniforge module and activate the environment

module load miniforge
source activate <env_name>

Next, you’ll need to export the existing environment to a yaml which will be used for rebuilding, then remove the existing environment. You’ll need to deactivate the environment prior to removing:

conda env export -f <env_name>.yml
conda deactivate
conda env remove --name <env_name>

You might encounter issues with resolving dependencies in which case you might need to leave out the versions of the packages in the yaml file and automatically install the most recent compatible versions. You can use the command conda env export | cut -f 1 -d "=" | grep -v "prefix" > <env_name>.yml in place of the above command to remove the versioning information of packages while exporting the environment into a yaml file.

You can then recreate the environment with Miniforge using the following:

conda env create -f <env_name>.yml

You can either submit a support request on our website or you can attend one of our office hour sessions. We meet virtually over Zoom on Tuesdays (3-5PM) and Thursdays (10AM-12PM) via Zoom. Links to the sessions can be found at the bottom of this page on our website.

Dear PI,

This message is important if you intend to use the anaconda/2023.07-py3.11 for your fall classes. Due to the recent licensing restrictions by Anaconda on research usage, we will be removing the licensed Anaconda distribution from our system on October 15, 2024.

As an alternative we have installed the miniforge/24.3.0-py3.11 module, which includes the essential Conda and Mamba package managers along with commonly used packages such as numpy, pandas, matplotlib, etc., from the conda-forge channel.

How should I prepare?

Between now and October 15, you may load the existing anaconda module or the newly installed miniforge/24.3.0-py3.11 module. Both modules provide the same conda commands to manage and use your conda environments.

Important: By default, the miniforge distribution will only provide packages from the conda-forge channel. Therefore, if you require packages from channels that are covered by the Anaconda repository Terms of Service (main/anaconda, r, msys2) you may specify this in your installation command but only for environments that are restricted to educational use, i.e., instructional work in your classes. For example, to install the seaborn package from the Anaconda default channels, you would use:

conda install –n path-to-my-conda-env seaborn –c anaconda

What to expect on October 15, 2024?

The licensed Anaconda distribution and base environment provided by the anaconda module will be removed from our systems on Oct 15, 2024. The anaconda module will redirect to the new miniforge/24.3.0-py3.11 module, effectively switching to conda-forge as the only default package installation channel and with a reduced number of preinstalled packages in the base environment.

The use of your existing environments should not be affected by this change. For instructional use you may continue to install python packages from the licensed Anaconda default channels as shown in the example above. However, any use of such environment for research purposes is a violation of the Anaconda license unless you obtained your own license.

We understand that these changes may cause inconvenience, but these changes are mandated by the Anaconda licensing condition which we cannot control. If you have any questions or concerns, please feel free to reach out.

At your service, RC staff

Research Computing

E hpc-support@virginia.edu
P 434.243.1107

University of Virginia
P.O. Box 400231
Charlottesville 22902