cluster

HPC cluster

The files in this directory allow to run the code on a GPU-based High-Performance Computing (HPC) cluster. Although not mandatory, GPU acceleration is encouraged in order to speed up Sionna's ray-tracing and link-level simulations.

Requirements

• CUDA-capable HPC cluster.
• Singularity/Apptainer container support.
• Slurm scheduling.

Setup

Navigate to the project's root directory in the command line and run:

apptainer build --fakeroot --force cluster/sionna.sif cluster/sionna.def

Alternatively, you can pull the built container from GitHub Container Registry:

docker pull ghcr.io/rodrihgh/cissir/sionna.sif:v2

More information on Apptainer container creation

Usage

After building the container, you can submit a job to the HPC cluster with sbatch:

sbatch cluster/submit_job.sh

Once submitted, you can run the code on your browser via JupyterLab. JupyterLab's URL will be logged and written into the file jupyter.url in the project's root.

More information on Slurm commands