Unable to use `cpus-per-task`

[conceptofmind]

Describe the issue:

Hello,

Currently, I am unable to launch a job due to not being able to use cpus-per-task on our SLURM cluster. Is there any way to use a different configuration, or do I need to rewrite the SLURM job launcher in order to get around this? Currently, it looks like the command is hard-coded into the SLURM job class.

Minimal Complete Verifiable Example:

cluster = SLURMCluster(
    account="my-account",
    queue="my-gpu-cluster",
    cores=4,
    memory="4GB",
    processes=1,
    walltime="4:00:00",
)
cluster.scale(1)
client = Client(cluster)

Anything else we need to know?:

Environment:

    "dask[complete]>=2025.3.0"
    "dask-jobqueue>=0.9.0"
    "distributed>=2025.3.0",

Thank you,

Enrico

[guillaumeeb]

Hi @conceptofmind,

Slurm clusters often have some specific configuration like that.
In this case, you should be able to use job_directives_skip kwarg.

See https://jobqueue.dask.org/en/latest/clusters-advanced-tips-and-tricks.html#skipping-unrecognised-line-in-submission-script-with-job-directives-skip