VEGA support pointed me to a site which they have used for GRIR443 benchmark. The slurm batch there is quite helpful if you are new to singularity myself.

#!/bin/bash

#SBATCH --nodes=2
#SBATCH --time 04:00:00
set -e; set -o pipefail

# Load required modules, if required
module load singularity

# Set cluster/experiment specific variables
readonly gpus_per_node=#
readonly benchmark_dir=#
readonly kpoints=2
readonly npool=2
readonly procs_per_gpu=2
readonly procs_per_node=$((gpus_per_node*procs_per_gpu))
readonly qe_sif="${benchmark_dir}/qe_6.8.sif"

# Build SIF, if it doesn't exist
if [[ ! -f "${qe_sif}" ]]; then
    singularity build ${qe_sif} docker://nvcr.io/hpc/quantum_espresso:v6.8
fi

# Attempt to start MPS server within container if needed
if (( procs_per_gpu > 1 )); then
    srun --ntasks-per-node=1 /bin/bash -c "nvidia-cuda-mps-control -d; sleep infinity" &
fi

echo "INFO: Running Quantum ESPRESSO with:"
echo "  ${SLURM_JOB_NUM_NODES:-$SLURM_NNODES} Nodes"
echo "  ${gpus_per_node} GPUs per node"
echo "  ${procs_per_node} MPI processes per node"
echo "  ${procs_per_gpu} MPI processes per GPU"
echo "  ${npool} Pools"
echo "  ${kpoints} KPoints"

# Launch parallel QE experiment
srun --mpi=pmi2 \
     --ntasks-per-node=${procs_per_node} \
     singularity run --nv "-B${benchmark_dir}:/host_pwd" --pwd /host_pwd \
     ${qe_sif} \
     pw.x \
     -input /host_pwd/ausurf.in \
     -npool ${npool} \
     2>&1 | tee qe_log.txt

# Attempt to exit MPS server
srun --ntasks-per-node=1 /bin/bash -c "echo quit | nvidia-cuda-mps-control || true"