(from https://www.rc.virginia.edu/userinfo/rivanna/queues/)

PartitionMax time / jobMax nodes / jobMax cores / jobMax cores / nodeMax memory / coreMax memory / node / jobSU Charge Rate
standard7 days140409GB375GB1.00
parallel3 days251000409GB375GB1.00
largemem4 days1161660GB975GB1.00
gpu3 days4101032GB375GB3.00 *
dev1 hour2846GB36GB0.00

Sample slurm scripts: https://www.rc.virginia.edu/userinfo/rivanna/slurm/

Numba Information/Notes:

  1. https://numba.pydata.org/numba-doc/latest/user/5minguide.html
  2. https://numba.pydata.org/numba-doc/dev/reference/numpysupported.html

Multiprocessing Example

  1.  Let's say we need to parallelize 10 parallel tasks in each job, and let's say we need to submit 20 jobs (so 200 tasks in total)

    Main.py >>>

    def run_replica(i):
        job_number = sys.argv[1]
        replica_number = 10*int(sys.argv[1]) + i

    if __name__ == '__main__':
        jobs = []
        for i in range(10):
            p = multiprocessing.Process(target=run_replica, args=(i,))

    job.slurm >>>

    #SBATCH --nodes=20
    #SBATCH --ntasks-per-node=10
    #SBATCH --time=10:00:00
    #SBATCH --output=slurm.out
    #SBATCH --error=slurm.err
    #SBATCH --partition=parallel
    #SBATCH -A spinquest_standard
    #SBATCH --array=0-20

    module load openmpi

    srun python3 Main.py $SLURM_ARRAY_TASK_ID


  • No labels