Script slurm-amber-20-cpu.sh to run sander.MPI on 2 nodes (16 tasks per node):

slurm-amber-20-cpu.sh

#!/bin/bash --login
#SBATCH --job-name=amber20cpu
#SBATCH --output=%x.o%j
#SBATCH --error=%x.e%j
#SBATCH --nodes=2
#SBATCH --ntasks-per-node=16
#SBATCH --time=0-01:00:00
#SBATCH --mem=16G
#SBATCH --partition=cpu
#SBATCH --constraint='broadwell|skylake'
#SBATCH --account=<account>
 
shopt -q login_shell || exit 1
 
module load gnu8 openmpi4
module load apptainer/1.0
 
container='/hpc/share/applications/amber/20/amber-20-cpu'
 
bind="${GROUP:+$GROUP,}${ARCHIVE:+$ARCHIVE,}${SCRATCH:+$SCRATCH,}$(test -d /node && echo /node)"
 
mpirun \
apptainer exec ${bind:+--bind $bind} "$container.sif" \
sander.MPI -p complex_azd.prmtop -c complex_azd.inpcrd -ref complex_azd.inpcrd -i min1.in -o min1.out -r min1.crd -O

Script slurm-amber-20-gpu.sh to run pmemd.cuda on 1 node with 1 GPU (4 tasks per node):

slurm-amber-20-gpu.sh

#!/bin/bash --login
#SBATCH --job-name=amber20gpu
#SBATCH --output=%x.o%j
#SBATCH --error=%x.e%j
#SBATCH --nodes=1
#SBATCH --ntasks-per-node=4
#SBATCH --gres=gpu:1
#SBATCH --time=0-01:00:00
#SBATCH --mem=16G
#SBATCH --partition=gpu
#SBATCH --account=<account>
 
shopt -q login_shell || exit 1
 
module load apptainer/1.0
 
container='/hpc/share/applications/amber/20/amber-20-gpu'
 
bind="${GROUP:+$GROUP,}${ARCHIVE:+$ARCHIVE,}${SCRATCH:+$SCRATCH,}$(test -d /node && echo /node)"
bind="${bind:+$bind,}/opt/hpc/system/nvidia/driver:/usr/local/nvidia/lib,/opt/hpc/system/nvidia/driver:/usr/local/nvidia/bin"
 
apptainer exec ${bind:+--bind $bind} "$container.sif" \
pmemd.cuda -i min.in -o test-min.out -p test.prmtop -c test.inpcrd -r test-min.rst -ref test.inpcrd -e test-min.en -inf test-min.mdinfo -O
 
apptainer exec ${bind:+--bind $bind} "$container.sif" \
pmemd.cuda -i test_amber_MD.in -o test-MD.out -p test.prmtop -c test-min.rst -r test-MD.rst -x test-MD.nc -e test-MD.en -inf test-MD.mdinfo -O

Run the following commands on the login node to open a shell on the compute node:

srun \
    --nodes=1 \
    --ntasks-per-node=4 \
    --time=0-01:00:00 \
    --mem=16G \
    --partition=cpu \
    --qos=cpu \
    --constraint='broadwell|skylake' \
    --account=<account> \
    --pty \
    bash

As usual the argument to the

--account=<account>

option must be the account you intend to use.

On the compute node run a shell within the amber/20/cpu container:

module load apptainer/1.0
module load amber/20/cpu

apptainer shell "$CONTAINER"

Inside the container run pdb4amber:

pdb4amber

To exit the container run exit.

To exit the shell on the compute node run the exit.