The Modules System

Research Computing uses a module system to load most software into a user’s environment. Most software is not accessible by default and must be loaded in. This allows Research Computing to provide multiple versions of the software concurrently and enables users to easily switch between different versions.

The module Command

Modules should be loaded in job scripts, interactive jobs, or on compile nodes only. They should not be loaded when on one of the login nodes. The login node will restrict the loading of modules, so you won’t be able to access software unless you do so through a job or a compile node.

To see what modules are available to load, ssh into a compile node by typing acompile from a login node, and type:

module avail

This will return a list of modules available to load into the environment. Please note if you run this command on a login node you will not receive a list of modules present on the system. It is also important to note that if a module has dependencies, you may not see the module listed until dependencies are loaded.

To load your chosen modules into the environment type:

module load some_module

# example: "module load anaconda"

You can specify the version of the software by appending a / with the version number:

module load some_module/version 

# example: "module load anaconda/2023.09"

The Lmod hierarchical module system provides five layers to support programs built with compiler and library consistency requirements. A module’s dependencies must be loaded before the module can be loaded.

The Layers include:

• Independent programs

• Compilers

• Compiler dependent programs

• MPI implementations

• MPI dependent programs

If you cannot load a module because of dependencies, you can use the module spider to find what dependencies you need to load the module.

module spider some_module

# example: "module spider openmpi"

Subcommands

The module command has a variety of subcommands, outlined in the table below. You may shorten the command to ml, but the shortened command may require specialized syntax.

Command	Shortened Command	Description	Example
module avail	ml av	List available software. Modules not listed here may have unmet dependencies which must be loaded for the module to be available.	module avail
module spider <module>	ml spider <module>	Searches for a particular software.	module spider openmpi
module load <module>	ml <module>	Load a module to use the software. In this example we are loading the GNU Compiler Collection. The default version will load because we have not specified a version.	module load gcc
module load <module>/<version>	ml <module>/<version>	Load GCC version 10.3.0	module load gcc/10.3.0
module unload <module>	ml -<module>	Remove or unload a module	module unload gcc
module swap <module> <new_module>	ml -<module> <new_module>	Swap a module. In this example we are unloading GCC and loading Intel. Any GCC-dependent modules will also be unloaded, and the intel-dependent versions (if available) will be loaded in their place.	module swap gcc intel
module purge	ml purge	Remove all modules. The slurm module will not be unloaded with this purge because it is sticky. Use the --force flag to unload a sticky module.	module purge
module save <name>	ml save <name>	Save the state of all loaded modules. In this example, we are saving all loaded modules as a collection called foo	module save foo
module restore <name>	ml restore <name>	Restore a state of saved modules. In this example, we are restoring all modules that were saved as the collection called foo	module restore foo
module help		Find information about additional module sub-commands.	module help

Loading Modules in a Job Script

Loading a module will enable access to the modules described in the software package. Additionally, modules will set or modify a user’s environment variables.

Modules in a job script can be loaded after your #SBATCH directives and before your actual executable is called. A sample job script that loads Anaconda into the environment is provided below:

#!bin/bash
#SBATCH --nodes=1
#SBATCH --time=00:01:00
#SBATCH --ntasks=1
#SBATCH --job-name=test-job
#SBATCH --output=test-job.%j.out

module purge
module load anaconda

python test-program.py