Using Python Packages via Conda (New Cluster)

Conda is a Python package manager often used for the distribution of scientific software. It provides features over and above the Python PyPi package manager 'pip', including support for non-Python libraries (e.g. CUDA) that Python software may depend upon.

Software is distributed via 'channels', which are web based download locations, the default install locations will depend on where you got Conda from; Anaconda, Miniconda and Condaforge are the most likely sources and differ as follows:

• Anaconda - this is a 'batteries included' commercial package installation and may include many packages you do not need. This is free to use for Personal, Educational, Open-Source or Small Business use as set out in the Anaconda license agreement. 
• Miniconda - this is the basic 'conda' command/minimal install option using the same channel as Anaconda and subject to the same license agreement.
  
• Miniforge/Mambaforge - this is an alternative to Miniconda but with the default channel set to Conda Forge. This is completely open-source, with no conda specific restrictive licensing. We would recommend this option in general.

The SLURM cluster has both Miniconda and Miniforge available via environment modules, so there is no need to install either in your personal home folder (although any environments you create will be located there by default). To get started:

module add Miniforge3

You will now have the 'conda' command available to you.

To create an environment use something like:

conda create -n "myenv" python=3.12

Then activate it and install some Python libraries:

conda activate myenv
conda install numpy

Further details are available on our Python pages.