search

planck-npipe / hillipop

A cobaya high-ell likelihood polarized for planck 2020 data (NPIPE release)
GNU General Public License v3.0
3 stars 2 forks source link

Using hillipop with CLASS mod stops at first hillipop eval #17

Closed juan-s-cruz closed 11 months ago

juan-s-cruz commented 11 months ago

We are trying to use the hillipop likelihood in Cobaya however with a modified CLASS version as opposed to CAMB as in the example in this repository. The computation simply stops with

[planck_2020_hillipop.ttteee] Computing new state [planck_2020_hillipop.ttteee] *ERROR* The 'theory' attribute of likelihoods has been deprecated. Please use the equivalent 'provider' attribute instead.

We believe the theory module is running without a problem since it is outputting the correct things and the sampling just crashes at the first evaluation of hillipop. However we are not able to get past the above issue. I am unsure it has to do with the deprecation above or something else. Is there a way to solve this or at least get some more information via a debugging option?

xgarrido commented 11 months ago

Try to upgrade cobaya with pip install -U cobaya.

juan-s-cruz commented 11 months ago

We are already on the latest Cobaya.

xgarrido commented 11 months ago

I've checked, we have fixed the issue in the master branch on git but the current version v4.2.1 on PyPI still have the theory field. I will try to push a patched version on PyPI asap.

juan-s-cruz commented 11 months ago

I tested with the repository version and indeed the sampling runs without the error now. Thanks!

mtristram commented 11 months ago

New PyPI version v4.2.2 compliant with COBAYA v3.4