Slurm

LICCA Cluster Workflow

Agents NEVER have SSH access to LICCA. Agent writes the job script + binary; Jörn submits and retrieves results.

Steps

1. Write/update the experiment entrypoint in experiments/<topic>/<experiment>/main.rs
2. Copy the template from references/experiment.sh to experiments/<topic>/<experiment>/job.sh
3. Fill in the TODOs in the job script (binary name, resources, arguments)
4. Write resource justification table (mandatory):

Flag	Value	Why
--partition	...	Why this partition
--cpus-per-task	...	Why this many CPUs
--mem	...	Why this much memory