Jesús Torrado

Thanks! + Then I don't think there is anything to be done then on the *thinning* side, since the relevant auxiliary GetDist chains are already initialised with weights and logposterior...

@cmbant do you agree with the changes in the docs here, or have anything to add? https://github.com/CobayaSampler/cobaya/pull/180/commits/89b31887357ea132a09e458441869c781cceb065

Thanks a lot for the review! > Looks good to me thanks, though not very sure what you mean by "by weighting with samples with their probability". The first "with"...

Thanks! Could work. Since this is new, we can test it ourselves and see whether it feels convenient. I am going to be using it this coming week myself. To...

@williamjameshandley and @lukashergt, thanks a lot! I'll come back to this as soon as I am finished with https://github.com/CobayaSampler/cobaya/pull/222 (1-2 working days) As Antony said, ideally clik would do this....

In that case, it sounds like CAMB/CLASS (which one?) should detect a failed computation and not even run the likelihood, I think.

Hard to say where it should be tested. One possibility would be to add to the `Collector'`s (than handles how the products of CAMB/CLASS are gathered) an optional `test` property...

Maybe the simplest solution is a check just before the clik call, when the array containing Cl's and nuisance is constructed. Works for CAMB and CLASS.

Just after this: https://github.com/CobayaSampler/cobaya/blob/d9458386846eb782430d963821522bf1ab47a7ae/cobaya/likelihoods/base_classes/planck_clik.py#L126

There are actually 2 problems: - CAMB-specific: failing silently by producing unphysical results, instead of raising an error. - clik-specific: segafulting instead of raising a Python error when receiving nan's...