See https://docs.computecanada.ca/wiki/Using_GPUs_with_Slurm. Essentially if we request more memory than 125 GB and don't specify a specific GPU model, it will attempt to assign us to a P100 node without enough memory and the job will not be scheduled.
See https://docs.computecanada.ca/wiki/Using_GPUs_with_Slurm. Essentially if we request more memory than 125 GB and don't specify a specific GPU model, it will attempt to assign us to a P100 node without enough memory and the job will not be scheduled.