High Performance Computing (HPC) conf. suggestion - node/tasks

[zkhodzhaev]

Hello,

As I know Mumax runs in GPU, and it only uses 1 node/ script. So, I can't distribute 1 script into different nodes at the same time. So, I am trying to find out what configuration in Slurm gives the best performance in Mumax (my tests were inconsistent - errors).

Ex: (1. Can Mumax distribute script into tasks? or i should just use -N 1)

#SBATCH -n 4 # number of tasks per node #SBATCH -N 1 # total number of mpi tasks requested #SBATCH -p gpu-a100 # Queue name #SBATCH -t 4:00:00 # Run time

Thank you for your help.

[jedcheng]

Saw your question when looking for something else. Have you resolved the problem?

Although I don't understand your problem, my usual practice is to generate a bunch of scripts with a generator using a range of parameters (for example the length of an object).

Then I would launch 16 independent jobs simultaneously to quickly sweep through parameters on the HPC.

[zkhodzhaev]

Thank you for your feedback. I was looking for a more in-script solution.