Slurm job script for shmem

This commit is contained in:
Martin Lilleeng Sætra 2021-08-06 17:13:11 +02:00
parent 9f88d611dd
commit 5f114fe7a5

View File

@ -3,14 +3,13 @@
#SBATCH -N 1 # number of nodes #SBATCH -N 1 # number of nodes
#SBATCH -n 1 # number of cores #SBATCH -n 1 # number of cores
#SBATCH -w g001 # DGX-2 node #SBATCH -w g001 # DGX-2 node
#SBATCH --gres=gpu:2 # number of V100's #SBATCH --gres=gpu:1 # number of V100's
#SBATCH -t 0-00:10 # time (D-HH:MM) #SBATCH -t 0-00:10 # time (D-HH:MM)
#SBATCH -o slurm.%N.%j.out # STDOUT #SBATCH -o slurm.%N.%j.out # STDOUT
#SBATCH -e slurm.%N.%j.err # STDERR #SBATCH -e slurm.%N.%j.err # STDERR
ulimit -s 10240 ulimit -s 10240
module load slurm module load slurm
module load openmpi/4.0.1
module load cuda10.1/toolkit/10.1.243 module load cuda10.1/toolkit/10.1.243
# Check how many gpu's your job got # Check how many gpu's your job got
@ -24,11 +23,11 @@ cp -r . /work/$USER/ShallowWaterGPU
# Run job # Run job
# (Assumes Miniconda is installed in user root dir.) # (Assumes Miniconda is installed in user root dir.)
cd /work/$USER/ShallowWaterGPU cd /work/$USER/ShallowWaterGPU
mpirun --mca btl_openib_if_include mlx5_0 --mca btl_openib_warn_no_device_params_found 0 $HOME/miniconda3/envs/ShallowWaterGPU_HPC/bin/python3 shmemTesting.py nvprof -o profiler_output $HOME/miniconda3/envs/ShallowWaterGPU_HPC/bin/python3 shmemTesting.py
cd $HOME/src/ShallowWaterGPU cd $HOME/src/ShallowWaterGPU
## Copy files from work directory: ## Copy files from work directory:
# (NOTE: Copying is not performed if job fails!) # (NOTE: Copying is not performed if job fails!)
cp /work/$USER/ShallowWaterGPU/*.log . cp /work/$USER/ShallowWaterGPU/*.log .
cp /work/$USER/ShallowWaterGPU/*.nc . cp /work/$USER/ShallowWaterGPU/*.nc .
cp /work/$USER/ShallowWaterGPU/profiler_output .