8000 IDUN · gpuocean/gpuocean Wiki · GitHub
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content

IDUN

Jump to bottom
FlorianBeiser edited this page Sep 15, 2023 · 3 revisions

For general information and acc 8934 ess see: https://www.hpc.ntnu.no/idun/

Preparation

  1. Connect to VPN

  2. Connect to login node, e.g. ssh <username>@idun-login1.hpc.ntnu.no

  3. Prepare environment, use module available to see list of available modules:

$ module load Python/3.7.4-GCCcore-8.3.0
$ module load CUDAcore/11.1.1
$ virtualenv gpuocean
$ source gpuocean/bin/activate
  1. Install all necessary packages in the environment, see the general README for a complete list.
$ pip install ...

Creating a job script

Lets create job.slurm as

#!/bin/sh
#SBATCH --partition=GPUQ
#SBATCH --account=<acount-name>
#SBATCH --time=00:01:00
#SBATCH --nodes=1
#SBATCH --ntasks-per-node=1
#SBATCH --gres=gpu:1 
#SBATCH --constraint=gpu32g
#SBATCH --job-name="IDUNtest"
#SBATCH --output=IDUNtest.out
module purge
module load Python/3.7.4-GCCcore-8.3.0
module load CUDAcore/11.1.1
source gpuocean/bin/activate 
export PYTHONPATH=${PYTHONPATH}:/cluster/home/floribei/havvarsel/gpuocean/src
python gpuocean/src/gpuocean/examples/dryrun.py

Schedule the job

$ sbatch job.slurm

Check the job

$ squeue -u <username>
$ scontrol show job <job-id>
Clone this wiki locally
0