Using Stan/brms on Super Computer

At my work I can utilize our computing cluster to run my models. I’m curious if there is anything to be aware of or specific things I need to do to run these scripts. (Obviously, some of these questions can be answered by my HPC team)

For example brms takes an argument cores = <number>. On my local computer this is fine, but when launching jobs on the HPC, I specify cores in the queuing script.

I’ve also run into problems when using too many cores and getting inaccurate results or results that don’t converge with different software. Is this a problem for Stan?

Can anyone point me in the direction for resources about using Stan on a high performance machine?

I would think that your queuing script would run an R script that uses brms, and in that case, I’d just pass the number of cores as an argument to the R script and then the R script can use that argument as a value for the cores argument in brms.

1 Like