-- AdminUser - 24 Jul 2024

Envio de Trabajos con SLURM usando las GPUS AMD MI210

1. Por cada trabajo a enviar cree un directorio; coloque alli los datos y scripts.

2. Verifique que los script usen rutas relativas("./", "../" )

script(job script) para el envio de un trabajo con SLURM usando GPUs AMD MI210

#!/bin/bash
#SBATCH --job-name=Test_MI210 #Nombre del Trabajo
#SBATCH --cluster=cecc
#SBATCH --partition=gpu.cecc  #Particion a usar
#SBATCH --nodes=1 #Number of nodes to use
#SBATCH --cpus-per-task=2 #CPU-cores per task (>1 if multi-threaded task)
#SBATCH --gres=gpu:1
#SBATCH --mem=32G #Total memory per node
#SBATCH --output=/homes/qteorica/juanpq/QE_Cacl_slurm_%j.out #Finished job notification
#SBATCH --error=/homes/qteorica/juanpq/QE_Cacl_slurm_%j.err #Error job notification
#SBATCH --mail-type=begin #Send email when job begins
#SBATCH --mail-type=end #Send email when job ends
#SBATCH --mail-user=mail_User@unal.edu.co
#SBATCH --export=SCRATCH_DIR=/scratch/$SLURM_JOB_ACCOUNT/$SLURM_JOB_USER/$SLURM_JOB_ID

export OMP_NUM_THREADS=12
export SINGULARITY_BINDPATH="/scratch:/scratch,/localapps:/localapps,/home:/home"

cd $SCRATCH_DIR

/usr/local/bin/singularity run --nv /localapps/qe7.2GPU.sif pw.x -inp K-points_6.in > K-points_6.out