Antony Lewis

https://arxiv.org/abs/2008.01074 https://arxiv.org/abs/2012.02726 (but would be a small breaking change for unit tests etc)

Use of correlation information

2

This it not really a bug - the code seems to work well (thank you, it looks useful)- but it feels as though there ought to be a way to...

It is quite easy to get "chain stuck" errors on non-dragging runs (or runs where dragging automatically turned off), at least when starting a new model for white a good...

We don't have a very clear way to get combined output, and currently some samplers produce the same result on all processes (minimize), and some produce per-process result on each...

Function to dump observables

2

For testing, it's useful to be able to easily get the observables for a specific model to check all settings are correct and that theory predictions agree with expectations. (equivalent...

for external_modules might be more logical to allow inclusion via a package-relative reference rather than a path (e.g. external likelihoods could re-use Planck calibration parameter).

Sampling data calibrated parameters and stuck chains

2

For example, for future CMB data might want to sample a high-L C_L amplitude parameter rather than A_s, to avoid issues with this parameter being extremely tightly constrained by the...

--upgrade fails

1

Copying last comment from https://github.com/CobayaSampler/cobaya/pull/237: Upgrading old versions I think can fail because this raises an error (old installs don't have .dat files): ``` elif opts.get("github_release"): try: with open(os.path.join(path, _version_filename),...

Because apparently max(1, np.inf//d)=1 for integer d.

https://arxiv.org/abs/2306.05460