Software¶

Hydra provides software through environment modules. You list what is installed with module avail and put one package on your path with module load NAME/VERSION. Most other packages you install yourself, into a conda environment in your own space.

For a package that is installed, find and load its module.
For a Python package, or anything on conda-forge or Bioconda, set up conda and install it there.
For software you built yourself, write a module file so that your jobs can load it.

We have a page each for R, for Julia, MATLAB, IDL and Java, for GPUs and for containers.

Set the thread count from $NSLOTS

NumPy, R's linear-algebra libraries, IDL, Java and many other packages start one thread per CPU on the node unless told otherwise. A job that does this is oversubscribed and we can kill it. See Warning emails. Each page below says how to set the count for that software.

When you need to look something up, the reference pages cover the compilers, libraries and MPI implementations and the installed modules. For a handful of packages we provide a guide with a working job file. If you need software that requires a license, a build against the cluster's MPI or CUDA libraries, or a shared database, email SI-HPC@si.edu.

Last updated 2026-10-03