Page 1 of 1

Cobaya: Issue with measuring speeds when derived parameters added

Posted: August 19 2020
by Agne Semenaite

I am having some trouble with cobaya's automatic determination of the oversampling power.

I have added a derived parameter by modifying the logp method of the likelihood I am using (sigma12, by adding the line

Code: Select all

params_values["_derived"]["sigma12"] = sigma12
), which seems to cause cobaya to get stuck at "Measuring speeds" stage of the run, regardless of the sampler used. However, manually passing the oversampling powers or running "evaluate" dummy sampler works fine.

When looking at the debug file, it seems like the issue is an error when computing a new state that results in zero likelihood value:

Code: Select all

[des_y1.clustering_copy] Ignored error at evaluation and assigned 0 likelihood (set 'stop_at_error: True' as an option for this component to stop here). Error message: TypeError("'NoneType' object does not support item assignment",)
I suppose the issue is with the params_values dictionary at this stage, but I am not sure how to solve this.

Of course, one can avoid the issue by setting the oversampling manually, but I was wondering if there is a way to make this work?