Install / use

[kylebaron]

Do you have a short writeup for how to install and start using this? I haven't done Go before. But I'm ready to Go for it.

[vjd]

me too please!

[dpastoor]

yeah I can add a guide later today.

I can also push a binary if you just want to test it out, and not muck around with the code. That only involves sticking the binary in your path and you're off to the races.

My only question would be - what do you want to try it on?

Right now the only thing that is definitely stable is running models - but TBH if you're submitting jobs to a SGE queue it might not be worth it YET, since the design is first focused on assuming a single large machine (no SGE compute nodes).

These are on the short list of to-be-implemented soon:

General Run Related

• [IMPLEMENTED 90%] - running groups of models 'easily' by using subfolders/regex.
• using submission 'queue' to submit jobs not tested enough atm - not too much more work to implement. Please let me know if this is of interest and I can finish it up quickly
• piping stdout (gradient printing) of each run to various locations with more control. Right now redirected into a file called stdout.out in each run's folder. Will have more control over what is printed, so can stream to multiple places given different logging levels (eg only indicate 0 gradients on stdout but all to file, etc)
• dealing with debugging a model that isn't running properly where you have to dig around in the psn.lst file (objective is to re-parse the failed lst file and present the error and clean up behind you after acknowledging error)
• actually name the folder _est_ or _sim_ depending on run type - right now everything _est_

output related

• parser completely blows up for bayesian models and models that use $THETAI/$THETAR :-( the cause is I am doing a string match for like $THETA, so the priors and transformations cause some funkiness.
• only parsing theta names atm - omega blocks needed a little more thought due to how many ways you can write them, and how less often people use comments
• what gets pulled into the output tree for R

So, if you just want to use it to run jobs a little different than PsN I can push up a binary later today, else I can tackle (a subset) of these based on what you are most interested in using it for, and get that out.