Skip to content

Using Lmod software modules

Lmod (Lua-based Modules) sets up your environment (paths and variables) for the software you choose. Most compilers, libraries and applications on Phoebe are installed centrally and loaded as modules. Phoebe and Koios have different module trees: see Phoebe and Koios below. Koios and the Phoebe GPU nodes use EESSI.

Where modules come from

Phoebe

Modules are grouped into stacks. Each numbered stack is built with one EasyBuild toolchain generation; foss/2024a, for example, is GCC 13.3 with OpenMPI, FlexiBLAS, FFTW and ScaLAPACK. Modules from different stacks don't mix: load one foss toolchain and modules built with it.

Stack Location Toolchain
2026a /cvmfs/2026a.phoebe.lan foss/2026.1 (GCC 15.2), lfoss/2026.1
2025a /cvmfs/2025a.phoebe.lan foss/2025a (GCC 14.2)
2024a /cvmfs/2024a.phoebe.lan foss/2024a (GCC 13.3)
2023a /cvmfs/2023a.phoebe.lan foss/2023a (GCC 12.3)
system /cvmfs/system.phoebe.lan none: CUDA, Miniforge3, Mathematica, MATLAB, Julia, VTune

On Phoebe, all stacks are available on the login node and on the Rocky Linux 8 compute nodes.

Phoebe GPU nodes (Rocky Linux 10)

The GPU nodes gpu1 and gpu2 (partition rocky10) have none of the stacks above: no module tree is loaded there by default. Use the EESSI software stack, which also has CUDA and cuDNN modules, or the NVIDIA HPC SDK modules described in GPU jobs.

For new work, use the newest stack that has what you need.

Koios

Koios has no module tree of its own: all modules come from the EESSI software stack, on the login node koios1.fzu.cz and on all compute and GPU nodes. No modules are available until you start EESSI, in each shell and at the start of each job script:

source /cvmfs/software.eessi.io/versions/2025.06/init/lmod/bash

After that, module avail, module spider and module load work as described below. EESSI has the foss toolchains (foss/2024a to foss/2025b in version 2025.06), Python with SciPy-bundle, HDF5, GSL, CFITSIO and much more.

The former Koios tree /cvmfs/c9.phoebe.lan has been retired together with Rocky Linux 9 on the login node. Programs and Python environments built with its modules need to be rebuilt with EESSI modules. A few of its modules have no EESSI counterpart (HEALPix, PFFT, VTune); build them yourself with an EESSI toolchain or ask the administrator.

Koios has no system stack: no Miniforge3 or Mathematica modules. For conda, install Miniforge3 into your home directory (see conda on Koios).

Search for available modules

module avail lists every module you can load, stack by stack. module -t avail prints one module per line, which is easier to search. On the login node:

[jose@login1 ~]$ module -t avail 2>&1 | grep -i '^python'
Python-bundle-PyPI/
Python-bundle-PyPI/2023.06-GCCcore-12.3.0
Python/
Python/3.11.3-GCCcore-12.3.0
Python-bundle-PyPI/
Python-bundle-PyPI/2024.06-GCCcore-13.3.0
Python/
Python/3.10.14-GCCcore-13.3.0-bare
Python/3.11.9-GCCcore-13.3.0-bare
Python/3.12.3-GCCcore-13.3.0
Python/
Python/3.13.1-GCCcore-14.2.0
Python/
Python/3.14.2-GCCcore-15.2.0

module spider finds a module in any stack and describes it:

[jose@login1 ~]$ module spider Mathematica

----------------------------------------------------------------------------
  Mathematica:
----------------------------------------------------------------------------
    Description:
      Mathematica is a computational software program used in many
      scientific, engineering, mathematical and computing fields.

     Versions:
        Mathematica/11.3.0
        Mathematica/12.0.0
        Mathematica/13.1.0
        Mathematica/14.1.0

module spider <name>/<version> shows how to load one version, including any modules it needs first.

Load a specific module

Use module load with the full name to get a predictable version:

[jose@login1 ~]$ module load IPython/8.14.0-GCCcore-12.3.0

Lmod also loads the modules it depends on.

Check loaded modules

module list, or its shorthand ml, shows the currently loaded modules, including the dependencies Lmod loaded for you.

Save the current session

You can save your loaded modules as a named collection, so that you can easily recreate the environment later. Choose a descriptive name:

[jose@login1 ~]$ module save my-ipython
Saved current collection of modules to: "my-ipython"

Collections are stored in ~/.config/lmod.

Restore a saved session

Use module restore to load a saved collection:

[jose@login1 ~]$ module restore my-ipython
Restoring modules from user's my-ipython

Unload all modules

Run module purge to unload all currently loaded modules. With a clean environment, load only the modules you need for your current task.

Clear the module cache

Lmod keeps a cache of the module trees in ~/.cache/lmod to make module avail and module spider faster. If a module that was just installed doesn't show up, or Lmod reports modules that no longer exist, delete the cache. Lmod rebuilds it the next time you use it:

rm -rf ~/.cache/lmod

Further reading