Hello,
I am having issues setting up ray correctly on Slurm.
I have a @ray.remote(num_cpus=0.5) function. I run things locally, and it seems to help. I am on a mac with 12cpus.
I then wanted to run on a higher number of cpus on Slurm. This is how I defined the sbatch job
#!/bin/bash
#SBATCH --mem 20G
#SBATCH --time=0-06:00:00
#SBATCH --tasks-per-node=1
#SBATCH --cpus-per-task=24
#SBATCH --nodes=1
this did not seem to work. When I ran top in the instance, there were a lot of ray::IDLE and then only one that seemed to be running the function. I am confused, I was expecting 48 workers to working at the same time. Any helps?
