You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The cobaya-cosmo-generator does not warn the user and doesn't check the file for inconsistencies. Namely it still allows the user to output, even though no likelihood is specified or selected. Polychord suffers the most, because no default settings are provided as an example, and the end user might not know how to adjust the parameters, or what parameter choices are valid or not. I believe that @lukashergt 's experience may prove valuable here, in designing a set of sane defaults other than
sampler:
polychord: {}
The text was updated successfully, but these errors were encountered:
the details of the defaults are all listed in Cobaya's readthedocs. I think the defaults are all sensible.
Are you suggesting to copy them over into the cosmo generator? All of them or just a subset?
I think all of them would be too much, but I agree that for a first time user it might be helpful to have the two major tuning parameters nlive=25d and num_repeats=2d listed explicitly.
that is true. Cobaya-Cosmo-generator should at least want the end user if their config is not sensible, (e.g. because they haven’t chosen a likelihood). Some general sanity checks would be useful, and help guide new users.
All of them or just a subset?
the crucial subset. I think there should be a combo box that has “low/medium/high” for the number of live points, and the number of repeats sensible for the run. I defer to your judgement which if any should be included. With the #115 merged, some accommodations for PolyChord’s new features will be needed anyway, and an explicit use_supernest clause should be there at a minimum.
The
cobaya-cosmo-generator
does not warn the user and doesn't check the file for inconsistencies. Namely it still allows the user to output, even though no likelihood is specified or selected.Polychord
suffers the most, because no default settings are provided as an example, and the end user might not know how to adjust the parameters, or what parameter choices are valid or not. I believe that @lukashergt 's experience may prove valuable here, in designing a set of sane defaults other thanThe text was updated successfully, but these errors were encountered: