dask_submit_workers.sh

#!/bin/bash

#SBATCH -n 6 # Number of tasks 
#SBATCH -c 4 # Total number of core for one task
#SBATCH --mem-per-cpu=3G
# SBATCH -p clk
#SBATCH -t 04:00:00
# SBATCH -C clk  #For submit to clk
#SBATCH -C hwl  #For submit to hwl


MEMORY_PER_TASK=$(( $SLURM_CPUS_PER_TASK*$SLURM_MEM_PER_CPU ))
# Number of tasks 
echo SLURM_NTASKS: $SLURM_NTASKS  
echo SLURM_NTASKS_PER_NODE: $SLURM_NTASKS_PER_NODE
echo SLURM_CPUS_PER_TASK: $SLURM_CPUS_PER_TASK 
echo SLURM_NNODES: $SLURM_NNODES
echo SLURM_MEM_PER_CPU: $SLURM_MEM_PER_CPU

########## MODULE LOADING ############
#module load cesga/2020 gcc/system openmpi/4.0.5_ft3 dask/2021.3.0-python-3.6.12 #-> WORKS

# module load cesga/2020 gcc/system openmpi/4.0.5_ft3 dask/2021.6.0 #-> WORKS

# module load cesga/2022 gcc/system openmpi/4.1.4 dask/2022.2.0 myqlm/1.9.9

module load cesga/system miniconda3/22.11.1-1
conda activate fibratic_ml_online
#module load miniconda3
#conda activate qiskit_dask
######################################


#Cambia el fichero con la info del scheduler si fuera menester
SCHED_FILE="./scheduler_info.json"

#--mem-per-cpu $SLURM_MEM_PER_CPU \
srun -n $SLURM_NTASKS \
    -c $SLURM_CPUS_PER_TASK \
    --mem=$MEMORY_PER_TASK \
    python ./dask_cluster.py \
        -local $LUSTRE_SCRATCH \
        --worker \
        -scheduler_file $SCHED_FILE \
        -preload  ./PreLoad.py
        #--ib \

#--resv-ports=$SLURM_NTASKS -l \