Parallel MCMC

[AndreWaehlisch]

I like how easy it is to sample in parallel with the emcee python package (just set the threads keyword to >1). Do you have any plans to implement this in a similar fashion?

[montyvesselinov]

The parallel also works. You just need to add processors (addproc ()).

On Thu, Apr 25, 2019 at 06:04 André Wählisch [email protected] wrote:

I like how easy it is to sample in parallel with the emcee python package (just set the threads keyword to >1). Do you have any plans to implement this in a similar fashion?

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/madsjulia/AffineInvariantMCMC.jl/issues/1, or mute the thread https://github.com/notifications/unsubscribe-auth/ABK7C6AOEVIC62OLVVFUSJTPSGM6HANCNFSM4HIMPGTQ .

-- -monty monty.gitlab.io

[ericagol]

Thanks for developing this package, @montyvesselinov

I am also interested in running this in parallel. I was not able to use the addproc() command. I did find the setprocs() command in Mads setprocs() but I'm unclear on whether this requires loading the full Mads package in addition to AffineInvariantMCMC.

I am interested in running this both on my multi-core laptop, as well as in a slurm environment with 28 processors per node.

Could you point me to documentation or examples which would show me how to do this?

Thanks in advance!

-Eric