Templating logic
Using the templating engine
So far everything we’ve covered is theoretically possible with slurm
job arrays. However, in this section, we’re going to cover something that glurmo
is uniquely capable of: using templating logic.
To offer a motivating example, we have covered how to run a variety of simulations using parameter vectors. However, in real simulation studies, different parameter combinations may require different computational resources, in terms of memory, CPUs, and time. One solution to this problem is to simply request the maximum necessary resources for all simulations, but this is not ideal. If some jobs require radically more resources, then the less resource intensive settings will take longer than they have to to be scheduled, and will potentially prevent other jobs from being scheduled that could realistically be run at the same time. Ideally, we’d like to only request the minimum resources we need by setting. glurmo
can accomplish this using templating logic, we will discuss in the next section.
This part of the tutorial will focus on the conditional-lm
subdirectory, which can be found here.