search

ICSM / ampere

A tool to fit the SED and spectra of dusty objects to constrain, among other things, the dust properties and mineralogy
6 stars 2 forks source link

opt.sampler.lnprobability populates lnprob with only -inf #34

Closed ciskakemper closed 3 years ago

ciskakemper commented 3 years ago

In branch ysosilfit I am trying to generate an example that is modeled after pgQuasarSample.py. This example is called ysoSilFit.py The very short run with the initial guess announced in line 66 does not work properly. In particular, after succesfully calling opt.optimise in line 67, the procedure breaks on lnprob = opt.sampler.lnprobability, as this somehow gets populated entirely with -inf values. This has all sorts of weird consequences downstream of course. I tried varying nsamples in line 67, nwalkers in line 31 (30 is the minimum), and also the starting positions of the walkers in line 54, by varying the division factor (currently set to 100). None of it gave any different result. It was quite a lot of work to figure out where the problem was in the first place, as the error message was complaining about something else.

ciskakemper commented 3 years ago

This is/will be fixed, with a proper treatment of the mask issue.