v0.16.0

What's Changed

New Features 🎉

• Fit BayesianETS through its innovations recursion instead of the Kalman filter by @jessegrabowski in #777
• Implement conditional recovery for Laplace-marginalized variables by @fonnesbeck in #790
• MLX MCLMC sampler by @jessegrabowski in #743

Bugfixes 🪲

• Fix GeneralizedPoisson scalar draw by @ricardoV94 in #785
• Fix BetaNegativeBinomial generator parameterization by @fonnesbeck in #787
• Fix GEV log-CDF at the finite upper endpoint by @fonnesbeck in #788
• Preserve Prior core_dims during serialization by @fonnesbeck in #789
• Build the full-rank ADVI guide's Cholesky factor with constant indices by @jessegrabowski in #795

Maintenance 🔧

• Fix forecast with exogenous data on custom statespace models by @jessegrabowski in #794
• Run pathfinder on float32 models by @jessegrabowski in #813
• Bump pymc to 6.4 and pytensor to 3.3.6 by @jessegrabowski in #815

New Contributors

• @G26karthik made their first contribution in #802

Full Changelog: v0.15.1...v0.16.0