Skip to content

Set up an HPC system profile

hhemt describes each cluster in an hpc_system_config.yaml (one per cluster), threaded via --hpc-system-config or Toolkit.from_configs(..., hpc_system_config=...). Anonymized example profiles ship in-repo: test_data/norfolk_coastal_flooding/hpc_system_config_uva.yaml and hpc_system_config_frontier.yaml.

Fill in your allocation

Copy an example profile and fill exactly two fields:

  • default_account: "{your-allocation}" → your SLURM account (slurm_account).
  • container.sif_path → only if you run in container mode (execution_environment: container); the native default ignores the whole container: block.

Tip

Start with native (no SIF). Container mode is opt-in and needs a transferred, signed Apptainer SIF — see containers/README.md.

Choose a partition

Pick a partition name on your analysis config (hpc_ensemble_partition, hpc_setup_and_analysis_processing_partition) that keys into partitions: in the profile. GPU hardware and compilation backend DERIVE from the partition — you do not set gpu_hardware directly. (UVA: standard / gpu-a6000 / gpu-a100-80; Frontier: batch.)

See also