v3.6.2

3.6.2 – 2026-03-27

General

• New grid scripts optional parameter --musthave-params (#477; thanks @louisl3grand)
• Fixed regression in grid script jobqueue (preventing correct job submissions; thanks @louisl3grand)
• Fixed inconsistency between shell invocation as cobaya-COMMAND and python -m cobaya COMMAND. Now the former also searched for components in the cwd (#486; thanks Ruchika for reporting).

MCMC

• Fixed bug: checkpoint file was not being written when convergence check failed (#483; thanks @eatdust for reporting)

Cosmology

• CAMB 1.6.6 support for use_non_linear_ratio: True to use external non-linear Theory model consistently

v3.6.1

3.6.1 – 2026-01-30

General

• Grid tools fix for more robustly managing running jobs (#464)
• Fixed type check error introduced by pandas 3.0 (#475; thanks @cgiovanetti)

Cosmology

• Add new DES-Dovekie SN release (#465; thanks @bap37)
• Fixed race condition caching inverse covariance in CamSpec likelihoods (#466)

v3.6

3.6 – 2025-10-28

General

• Minimum python 3.10; type hints updated accordingly (#419)
• Added standardized formatting with ruff (removing flake8) (#419, #442)
• Added pre-commit hooks for automated code quality checks (trailing whitespace, end-of-file-fixer, pyupgrade, ruff) (#419)
• Documentation can now be built as markdown for LLM context. Flat context file provided.
• Periodic sampled parameters implemented (for now fully supported by MCMC only), as well as tagging of derived parameters as periodic for analysis with GetDist (#440, #443; fixes #438)
• Added simple gaussian likelihood with optional normalization constant (#433)
• In Prior.reference(), now fixed values can be turned into a norm with given std (#429, #460; fixes #394)
• In SampleCollection, inconsistent sums of individual chi2's and prior's are allowed, for e.g. emulators (#456)
• sampler_type now an explicit parameter, instead of guessed from sampler name (#455)
• Fixed out-of-bounds bug in Pool2D (#457; fixes #454)
• use_renames option of theories deprecated (was not used)
• Empty dicts in component definitions now preserved (fixes #450; thanks yzhengsong)
• Improved truncnorm prior support (#431; thanks @xgarrido)

Cosmology

• Planck clik-based likelihoods now use the new clipy Python implementation (#444, #427)
• Planck 2015 likelihoods deprecated (#444)
• CLASS updated to 3.3 (#449; fixes #404, #405, #434, #448)
• Updated candl mentions in docs (#453, #447; thanks @Lbalkenhol)
• Updated DESI BAO DR2 references in docs (#422; thanks @armengau)

v3.5.7

3.5.7 – 2025-03-31

General

• Fixed issue with changing parameter order using gaussian_mixture with input_params
• cobaya-doc can be used to list components of a given type, e.g. cobaya-doc bao
• Small tidies for float output with numpy2

Cosmology

• BAO, DESI, and SN likelihood docs now automatically list all available internal likelihoods

v3.5.6

• Added DESI DR2 bao likelihoods (thanks @desihub)
• ACT DR6 added to list of external likelihoods (thanks @ACTCollaboration)

v3.5.5

3.5.5 – 2025-02-05

General

• Option for stricter run-time input type checking (#388, thanks @ggalloni)
• Detect and fix incomplete last lines when resuming or minimizing from existing runs (#306, #378)
• Added functions module and refactored some numerical functions into it
• Tidier numpy2 outputs
• Other minor fixes (thanks @mgerbino, @ark93-cosmo, @MariaLuisaGallavotti, @ewanchamberlain)

Cosmology

CLASS

• Regression: no multiple local installs after v3.3 (see #404)

v3.5.4

• Allow classes to have both yaml and class attributes as long as no duplicate keys
• Added get_modified_defaults() class method to cobaya components to dynamically set/modify defaults

3.5.3

• Added cobaya-run --allow-changes to resume or minimize after changing input settings (turns off checking)
• Updates for numpy 2 and other compatibility/deprecation fixes
• Fixes #357, #358, #360, #361, #362, #368
• Added _fast_chi_squared method to base class InstallableLikelihood

v3.5.1

• Added DESI 1yr BAO data and SN from Pantheon Plus, DESY5 and Union3 (thanks DESI team, @adematti, @rubind, @WillMatt4 and @rodri981)
• Use of vector parameters now documented (PR #191; inspired by @lukashergt, thanks!)

v3.5

General

• Updated UGE sample job submission template (for cobaya-job-run and cobaya-grid-run)
• Clarify log feedback when using oversample_thin
• Fixed #345, #346, #347, #348

Grid scripts

• Support for running grids of models, including grid getdist, PDF tables, importance sampling, minimization (almost all features of CosmoMC grid now available in Cobaya). See the new doc pages.

Minimization

• Support for iminuit minimizer and getting best-fits for all mpi runs (#332, thanks @ggalloni)
• Support for minimzation with an importance-sampled input yaml config