Slurmise
Predicting how much time, memory, and cpu cores a SLURM job will need is notoriously difficult, especially if the same job needs to be run multiple times with different parameters and on differently sized inputs.
Requesting too little risks out of time or out of memory failures for some jobs, while requesting too much results in inefficient cluster usage and unnecessarily long queue times.
slurmise attempts to address these issues by maintaining a database of previous jobs that is used to predict the requirements of the current job submission.
Getting Started