What's Changed
Major Changes 🛠
New Features 🎉
- Add Pareto to pymc.dims by @williambdean in #8422
- Add Neumann and mixed boundary conditions to HSGP by @juanitorduz in #8433
- Add more distributions to pymc.dims by @williambdean in #8487
Bugfixes 🪲
- Respect sample dims in
apply_function_over_datasetby @ricardoV94 in #8426 - Make ZeroSumNormal zero-sum check tolerance dtype-aware by @velochy in #8464
- Fix dims coercion to list in model_from_fgraph by @williambdean in #8465
- Fix DiscreteUniform.logcdf at the lower bound by @anishmehta24 in #8452
- Fix LKJ (corr and cov) forward sampling for n>2, and log density normalization by @tjburch in #8497
- Fix dataset-to-point conversion with no sample dimensions by @j45856021-dev in #8490
Documentation 📖
- Update architecture guide by @abhioverdue in #8475
Maintenance 🔧
- Slice sampler: cap expansion steps per dimension by @ricardoV94 in #8420
- Better support for pm.fit(..., backend="jax") by @hrideymarwah15 in #8472
- Improve VonMises logp overflow for large kappa by @spicysharkco in #8503
- Small cleanups and bug fixes by @ricardoV94 in #8522
- Update to pytensor 3.3.5 and better-optimize 0.5, and clear deprecations by @jessegrabowski in #8584
New Contributors
- @anurag-mds made their first contribution in #8430
- @ZainnQureshii made their first contribution in #8456
- @SatvikMishra08 made their first contribution in #8474
- @abhioverdue made their first contribution in #8475
- @anishmehta24 made their first contribution in #8452
- @hrideymarwah15 made their first contribution in #8472
- @spicysharkco made their first contribution in #8503
- @j45856021-dev made their first contribution in #8490
Full Changelog: v6.3.2...v6.4.0