Restrict yourself to low end GPU model

Dear users,

If you ask for a GPU without specifying any constraint, it may happens that the GPU allocated will be a very powerful one such as an A100 with 80GB ram. This is fine but if your code has only low GPUs requirement, this is a waste of resource for the other users.

If you submit a lot of jobs, it would be fair to restrict yourself to some GPUs models. This can be done for example like that:

#SBATCH --partition=shared-gpu
#SBATCH --gpus=1

In this case, you’ll let the high end GPU models to other users.

See here hpc:slurm [eResearch Doc]

And here hpc:hpc_clusters [eResearch Doc]