Computers

SLURM

Show all of the available partitions:

scontrol show partition

Show the information pertaining to a particular partition:
sinfo --partition=condo-slagergr
show partition condo-slagergr

Show the properties of a given node:

scontrol show node clrv1207

ISAAC

The user guide is here: ISAAC Documentation

The name of our condo is ISAAC-UTK0307. If you belong to this accont, you can use the partition condo-slagergr to run on the group condo.

The general campus open account is ACF-UTK0011.

In addition to the campus partitions, we can use condo-slagergr, but only to access the condo.

Load the necessary modules. The following is what I use by default in my .bashrc file:

# load necessary modules
module load gcc gsl openmpi lua

# for some reason the gsl module does not add an entry to the cpath
export CPATH=/spack/spack-0.17.2/apps/linux-rhel8-cascadelake/gcc-10.2.0/gsl-2.7-bf2eaja5t47limkvolsyaouai363gnqh/include

Example run script of job using a GPU and 64GB of memory for two hours

#!/bin/bash
#SBATCH --job-name=mlnn_neutrino
#SBATCH --account=isaac-utk0307
#SBATCH --nodes=1
#SBATCH --ntasks-per-node=1
#SBATCH --partition=condo-slagergr
#SBATCH --qos=condo
#SBATCH --cpus-per-task=16
#SBATCH --gpus=1
#SBATCH --mem=64G
#SBATCH --time=02:00:00

module load gcc/10.2.0
module load openmpi/4.0.3-gcc
module load hdf5/1.10.7-gcc
module load Python/3.9.10-gcc
module load cuda/11.4.2-gcc
source /nfs/home/jmcguig1/venv_Emu/bin/activate

srun python /nfs/home/jmcguig1/Rhea/stability_model/MLparent/mlnn_training/train_mlnn.py