Parallelizing PyMC+Nutpie on an HPC node (Slurm)

[bersavosh]

Hi there,

I was wondering if there are specific suggestions/recommendations on how to parallelize Nutpie sampling (with PyMC) on an HPC node that uses Slurm. I noticed PyMC forum has some discussion on BLAS level parallelization, I was wondering if this applies to Nutpie as well, or whether folks with more experience might have other suggestions (anything from tweaking something in the source code to specific instructions through sbatch or srun).

Thanks in advance!