0.21.0

What's Changed

• Add pupport forward-mode differentiation with autoguide by @juanitorduz in #2163
• Fix TraceEnum_ELBO ignoring is_auxiliary option by @juanitorduz in #2164
• Fix: Tracer leaks during NUTS sampling on random_nnx_module when NNX module contains mutable state by @juanitorduz in #2162
• FIX: Running MCMC within JIT causes tracer leaks by @juanitorduz in #2165
• Example: Embarrassingly Parallel MCMC via Consensus Monte Carlo by @juanitorduz in #2167
• Remove vectorize_particles from RenyiELBO docstring (Issue #2166) by @juanitorduz in #2168
• fix repeated recompilation in no-progress-bar fori_collect path by @simeoncampos in #2171
• Update GPJax notebook to simplified API by @thomaspinder in #2177
• Effect handlers Tutorial by @juanitorduz in #2169
• Generate outputs by @thomaspinder in #2179
• improve type error message for 'fn' in sample function by @Qazalbash in #2175
• Fix xla_pmap_p import for JAX versions that removed pmap by @saitcakmak in #2173
• Change key type from prng_key to Array by @fehiepsi in #2178
• feat(gh-2142): add VSBC diagnostic by @michaelellis003 in #2170
• Allow zero rate for Poisson distribution. by @tillahoffmann in #2176
• Bump to 0.21.0 by @fehiepsi in #2180

New Contributors

• @simeoncampos made their first contribution in #2171
• @saitcakmak made their first contribution in #2173

Full Changelog: 0.20.1...0.21.0

0.20.1

What's Changed

• DOC: Minor wording changes to numpyro code comments by @star1327p in #2133
• chore: files formatted with ruff-v0.15.0 by @Qazalbash in #2136
• test: update test effected by the optax==0.2.7 release by @Qazalbash in #2137
• test: adjust assertion for my_fn_calls based on optax version in test/test_optimizers.py by @Qazalbash in #2140
• build: setup.py -> pyproject.toml by @Qazalbash in #2141
• chore: replace legacy jax.random.PRNGKey with modern jax.random.key by @Qazalbash in #2134
• Add PSIS k-hat diagnostic for variational inference by @michaelellis003 in #2139
• doc: Fix a few typos in numpyro code comments by @star1327p in #2144
• try fixing downloads datasets by @juanitorduz in #2151
• chore: remove dm-haiku support from numpyro.contrib.module by @theorashid in #2150
• Add GPJax notebook by @thomaspinder in #2149
• GPJax Notebook is correctly display in docs by @juanitorduz in #2152
• expose progress bar print rate option in MCMC by @mdmould in #2145
• Always call jnp.finfo on array dtypes, not arrays themselves by @dylanhmorris in #2147
• fix test because new jax 0.9.2 release by @juanitorduz in #2155
• CI(gh-2126): transition to uv for dependency management and testing by @Qazalbash in #2129
• fix test because new jax 0.9.2 release [v2] by @juanitorduz in #2156
• Use PyThon 3.14 in the Test Suite (CI) by @juanitorduz in #2157
• fix: remove frozen project build during doc generation by @Qazalbash in #2160
• Bump NumPyro 0.20.1 by @juanitorduz in #2159

New Contributors

• @michaelellis003 made their first contribution in #2139
• @thomaspinder made their first contribution in #2149
• @mdmould made their first contribution in #2145

Full Changelog: 0.20.0...0.20.1

0.20.0

What's Changed

• numpyro.contrib.module: only collect mutables in nnx_module() when th… by @esennesh in #2061
• Use functional interface funsor.adjoint.adjoint to avoid tracer leak by @fehiepsi in #2002
• fix(tests): fix stderr calculation in test/test_distributions::test_entropy_samples to handle zero standard deviation by @Qazalbash in #2069
• Ensure compatibility NNX 0.11.2 by @juanitorduz in #2067
• Fix test for new NNX 0.12.0 by @juanitorduz in #2076
• Fix compatibility Predictive and new random keys by @juanitorduz in #2073
• Add cdf method for truncated distributions by @juanitorduz in #2074
• Fix get model relations when a param is a lambda function by @juanitorduz in #2078
• DOC: Correct typos in Hierarchical Forecasting by @star1327p in #2082
• DOC: Update the URL to Stochastic Volatility PyMC example by @star1327p in #2083
• DOC: Update MC Stan links in Examples by @star1327p in #2084
• Implement LeftCensoredDistribution and RightCensoredDistribution by @vanAmsterdam in #2081
• fix(gh-2036): MyPy Errors in numpyro.distributions.transforms Module by @Qazalbash in #2066
• DOC: Correct a typo: lets -> let's by @star1327p in #2087
• chore: update numpyro.util.set_platform for JAX_PLATFORM_NAME and JAX_PLATFORMS by @Qazalbash in #2091
• Implement IntervalCensoredDistribution by @vanAmsterdam in #2090
• DOC: Correct a few grammatical errors by @star1327p in #2097
• feat: add support for CUDA 12 and CUDA 13 by @Qazalbash in #2094
• Fix Beta with concentration1=1 gives nan log_prob at value=0 by @juanitorduz in #2089
• Add rational quadratic kernel to HSGP approximation by @juanitorduz in #2100
• Add uncondition effect handler by @juanitorduz in #2101
• Add Beta Negative Binomial distribution by @juanitorduz in #2102
• Add Inverse Wishart Distribution by @juanitorduz in #2103
• Fix uniform distribution log_prob support by @juanitorduz in #2095
• Bugfix/mutables leaves by @esennesh in #2105
• distributions.mixtures._MixtureBase: mixtures lack reparameterized sa… by @esennesh in #2106
• resolve deprecated warning from pre-commit hook by @Qazalbash in #2108
• feat(gh-2110): setup for astral-sh/ty by @Qazalbash in #2111
• Fix NumPy 2.4 deprecation warning when loading JSB Chorales dataset by @juanitorduz in #2113
• Refine ty config by @juanitorduz in #2114
• ci: pre-commit hook to auto-format yaml files by @Qazalbash in #2112
• doc(gh-2107): Improved documentation for Negative Binomial distributions by @Qazalbash in #2109
• Disable masking when running Predictive with infer_discrete=True by @fehiepsi in #2115
• DOC: Correct a few a/an usages by @star1327p in #2116
• FIX: JAX pxla is deprecated from 0.8.2 onwards by @juanitorduz in #2119
• Bump minimum python version to 3.11 and minimum jax version to 0.7 by @fehiepsi in #2120
• Add initial value to Gaussian state space distribution (fixes #2098). by @tillahoffmann in #2104
• fix(gh-2036): MyPy Errors in numpyro.distributions.constraints Module by @Qazalbash in #2085
• Add Types to numpyro.distributions.distribution by @juanitorduz in #2122
• Remove Typing Protocols by @juanitorduz in #2125
• ci: prek for pre-commit hooks check by @Qazalbash in #2121
• fix(gh-2123): Validate component distributions of MixtureSameFamily have parameter-independent support by @Qazalbash in #2127
• Bump numpyro to 0.20.0 by @fehiepsi in #2128
• Fix docs build for numpyro 0.20 by @fehiepsi in #2131

New Contributors

• @esennesh made their first contribution in #2061
• @star1327p made their first contribution in #2082

Full Changelog: 0.19.0...0.20.0

0.19.0

What's Changed

• Displaying CVAE figure example by @guspan-tanadi in #2011
• Equinox Integration by @kylejcaron in #2005
• fix(distributions.conjugate): support total_count_max in DirichletMultinomial distribution by @brendancooley in #2016
• [Example ] Bayesian CUPED by @juanitorduz in #2018
• [Typo] Add missing plus sign circular GP notebook by @juanitorduz in #2020
• Reenable haiku tests by @tare in #2021
• DOC: typo by @gdmcbain in #2023
• Fixed parameter updates for sequence layers in nn modules by @kylejcaron in #2024
• Avoid importing deprecated jax.util module by @jakevdp in #2025
• matrix normal doc: correlation -> covariance by @mochar in #2027
• refactor(infer.elbo): add type hints to elbo module by @brendancooley in #2028
• Display plate name on error by @emmanuel-ferdman in #2033
• fix(tests): fix for failing nested sampling unit tests by @Qazalbash in #2034
• feat(gh-299): Type hints in distributions modules by @Qazalbash in #2032
• feat: Dagum distribution by @Qazalbash in #2014
• Adds support for passing initial value to transform by @tingiskhan in #2038
• Allow scope divider in NNX and Equinox Random Modules by @juanitorduz in #2048
• update numpyro version in tbip notebook by @juanitorduz in #2049
• Handle renaming of pjit_p to jit_p by @tare in #2052
• Fix typing_extensions import issue by @tare in #2054
• Fix test to ensure compatibility with Flax 0.11 by @juanitorduz in #2057
• Test against Python 3.13 by @juanitorduz in #2058
• Update WishartCholesky.infer_shapes typing by @fehiepsi in #2059
• Bump to 0.19.0 by @fehiepsi in #2060

New Contributors

• @guspan-tanadi made their first contribution in #2011
• @gdmcbain made their first contribution in #2023
• @mochar made their first contribution in #2027
• @emmanuel-ferdman made their first contribution in #2033
• @tingiskhan made their first contribution in #2038

Full Changelog: 0.18.0...0.19.0

0.18.0

What's Changed

• Remove cached plates from AutoGuide. by @tillahoffmann in #1976
• Fix typo and tweak wording in get_dependencies() docs by @dylanhmorris in #1975
• Navigate to our forum for discussions by @fehiepsi in #1978
• Unify sample_posterior() signatures by @tare in #1979
• Add guidelines for bug report and feature request by @fehiepsi in #1980
• Make seed handler stateless by default by @fehiepsi in #1983
• Update paths to data files by @fehiepsi in #1996
• Add sum_sites option to sum loss over sites or return as dict. by @tillahoffmann in #1995
• Improve rng default in the contrib/model.py module by @juanitorduz in #1992
• Fix padding calculation by @tare in #1993
• Apply init_to_median in unconstrained space. by @tillahoffmann in #1994
• Fix tracer leak in hmc by @fehiepsi in #1985
• Fix trace leakage in SL-DAIS by @fehiepsi in #1984
• Mark xfail for leaked tracer tests by @fehiepsi in #1997
• Fix tracer leak in svi with mutable state by @fehiepsi in #2001
• Fix Memleak in ASVGD by @OlaRonning in #2003
• Flax NNX integration by @juanitorduz in #1990
• Pass value arg to optax, allowing use of reduce_on_plateau by @zmbc in #1974
• Add CirculantNormal distribution. by @tillahoffmann in #1988
• circulant gp notebook: minor improvements by @juanitorduz in #2004
• Fix behavior of logdiffexp(jnp.inf, jnp.inf), add test that would have caught it. by @dylanhmorris in #2007
• Port hierarchical multivariate forecasting model from Pyro to NumPyro by @juanitorduz in #2006
• Bump to 0.18.0 by @fehiepsi in #2009

New Contributors

• @zmbc made their first contribution in #1974

Full Changelog: 0.17.0...0.18.0

0.17.0

What's Changed

• Only include arg_constraints in pytree_data_fields if they are not lazy_propertys. by @tillahoffmann in #1929
• Ensure BlockNeuralAutoregressiveTransform maps real->real by @danielward27 in #1897
• Update coverall badge to point to the master branch by @fehiepsi in #1930
• Add openssf badge by @fehiepsi in #1931
• Add utility function to evaluate log density for individual sites. by @tillahoffmann in #1932
• Update link to jax repo by @fehiepsi in #1936
• Fix tracer leak in contrib.module by @fehiepsi in #1935
• Implement ICDF Methods for Truncated Distributions by @TheSkyentist in #1938
• Add type hints to the primitives module by @juanitorduz in #1940
• Add types types to handlers module by @juanitorduz in #1941
• Add types optim module by @juanitorduz in #1942
• Add batch and event shape to repr. by @tillahoffmann in #1946
• SteinVI: Recompute Score Function* For Each Particle Interaction in the Attractive Force by @OlaRonning in #1947
• Make Delta.log_prob jit-able on metal and raise explicit error for Delta sampling site during initialization. by @tillahoffmann in #1950
• Remove unnecessary device_put by @fehiepsi in #1952
• feat: Levy distribution by @Qazalbash in #1943
• SteinVI: New quadratic form kernel with Gauss Newton metric by @OlaRonning in #1953
• Make pip install calls compatible with zsh by @dylanhmorris in #1962
• Add log1mexp and logdiffexp functions by @dylanhmorris in #1960
• Fetch rng_key using prng_key message in block handler. by @tillahoffmann in #1957
• Add ComplexTransform and update biject_to registry. by @tillahoffmann in #1964
• Fix event dim of some arg constraints by @fehiepsi in #1967
• Addresses the deprecation messages of some jax.core functionalities by @fehiepsi in #1966
• Refactor test_gof into a separate file by @fehiepsi in #1965
• fix(tests): update tolerance levels and PRNGKey usage for improved test stability by @Qazalbash in #1959
• chore(ci): disable continue-on-error for all test jobs in CI workflow by @Qazalbash in #1968
• Do not create plates for observed sites in AutoGuide. by @tillahoffmann in #1972
• Fix promote_batch_shape logic to take batch shapes of all parameters by @fehiepsi in #1973
• Bump to 0.17.0 by @fehiepsi in #1969

New Contributors

• @TheSkyentist made their first contribution in #1938
• @dylanhmorris made their first contribution in #1962

Full Changelog: 0.16.1...0.17.0

0.16.1

What's Changed

• Support Python 3.9 by @juanitorduz in #1923
• added note that sbvm is only accurate for conc <= 10k by @OlaRonning in #1924
• Clarify the composability of handlers with jax primities by @fehiepsi in #1926
• Add coveralls to CI by @fehiepsi in #1925
• Add Bayesian VAR(2) example script #1658 by @aibit0111 in #1915
• Bump to 0.16.1 by @fehiepsi in #1927

New Contributors

• @aibit0111 made their first contribution in #1915

Full Changelog: 0.16.0...0.16.1

0.16.0

What's Changed

• Fix flaky test_change_point_x64 by @GaetanLepage in #1863
• Refactor validate_args method on distributions (fixes #1865). by @tillahoffmann in #1866
• gh-1870: Refactor log_prob method in _MixtureBase class to handle -jnp.inf by @Qazalbash in #1874
• use biased autocorrelation estimate by default (resolves #1785) by @jonny-so in #1877
• remove depreciated line from docs by @juanitorduz in #1882
• feat: add api compat arg to NS.get_samples by @nstarman in #1880
• Fixing nested sampling by @renecotyfanboy in #1871
• ci: fix pre-commit by @nstarman in #1892
• Bump CI to python 3.10 by @fehiepsi in #1890
• Refactoring SteinVI by @OlaRonning in #1883
• Fix handling of event dimensions in ComposeTransform (fixes #1893). by @tillahoffmann in #1894
• Correct event dimensions for ReshapeTransform. by @tillahoffmann in #1895
• Updated SteinVI docs by @OlaRonning in #1898
• Fixed norm const for SBVM by @OlaRonning in #1899
• CHANGED Stein BNN example to use prediction with noise (i.e. y_bnn -> y) by @OlaRonning in #1900
• [BUG] Latent exclude deterministic by @juanitorduz in #1901
• Allow callable for vectorization over particles on SVI by @juanitorduz in #1902
• Add Gaussian state space model distribution. by @tillahoffmann in #1904
• Proof of Concept: Types and MyPy by @juanitorduz in #1906
• Allow to collect nested dict keys in mcmc by @fehiepsi in #1905
• Add types for contrib.stochastic_support and improve the HSGP module by @juanitorduz in #1907
• [Types] Diagnostics and Utils by @juanitorduz in #1912
• Add method to retrieve distribution arguments. by @tillahoffmann in #1913
• Merge postprocess_fn into the fori_collect loop by @fehiepsi in #1910
• Hm/all deterministic by @hessammehr in #1914
• Bump to 0.16.0 by @fehiepsi in #1917

New Contributors

• @GaetanLepage made their first contribution in #1863
• @jonny-so made their first contribution in #1877
• @nstarman made their first contribution in #1880

Full Changelog: 0.15.3...0.16.0

0.15.3

What's Changed

• Computing number of gradient evaluations example by @andyElking in #1839
• Enable pmap progress bar with cpu backend / remove deprecated host_callback by @andrewdipper in #1841
• NumPyro and Other Samplers Tutorial by @juanitorduz in #1842
• avoids setting jax tracer as lazy property attribute by @sbidari in #1843
• enable progress bar for multi-gpu by @andrewdipper in #1849
• Resolve numerical instability in entropy of GeometricLogits. by @tillahoffmann in #1852
• Generic cdf and icdf implementation for scalar TransformedDistributions. by @tillahoffmann in #1853
• Raise error when using CircularReparam at observed site by @fehiepsi in #1856
• Update docs to make Predictive behavior more clear by @tomwallis in #1850
• Fix binominal distribution by @InfinityMod in #1860
• Fix the file path when rendering model by @fehiepsi in #1857
• Fix DiscreteUniform.enumerate_support with non-trivial batch shape by @fehiepsi in #1859
• Implementation of Doubly Truncated Power Law and Lower Truncated Power Law by @Qazalbash in #1807
• Bump to 0.15.3 by @fehiepsi in #1858

New Contributors

• @andyElking made their first contribution in #1839
• @sbidari made their first contribution in #1843
• @tomwallis made their first contribution in #1850
• @InfinityMod made their first contribution in #1860

Full Changelog: 0.15.2...0.15.3

0.15.2

What's Changed

• Simplify the provenance logic to prepare for the removal of jax named shape by @fehiepsi in #1837
• Bump to 0.15.2 by @fehiepsi in #1840

Full Changelog: 0.15.1...0.15.2