v0.0.3

Fixed integration bug introduced in numpyro 0.20.1.

What's Changed

• Remove extraneous link by @DanWaxman in #176
• Detect use of typed keys in CD-Dynamax integration by @DanWaxman in #180
• bumping version to v0.0.3 deal with new numpyro release by @mattlevine22 in #181

Full Changelog: v0.0.2...v0.0.3

v0.0.2

The main change in this release is the addition of predict_times for simulation. This allows for combination of filters (for inference) and simulators (for prediction), among other patterns. For example,

with Simulator():
    with Filter():
         model(obs_times=obs_times, obs_values=obs_values, predict_times=predict_times)

A number of bugfixes are also included, including:

• Pinning the arviz version, due to breaking changes in the newest arviz
• Fixing a Blackjax bug that inadvertently added common random numbers by default (thanks @mlysy for spotting this)!

It also marks a move towards using standard distributions of our dependencies, which is necessary for PyPI integration.

Full Commit History

• Pin arviz version by @DanWaxman in #150
• Fix Blackjax Bug by @DanWaxman in #141
• Add badges and link to the docs by @DanWaxman in #152
• Use PyPI-based effectful by @DanWaxman in #161
• Upgrade cuthbert to current main by @DanWaxman in #159
• Predict Times by @DanWaxman in #135
• Migrate to CD-Dynamax public by @mattlevine22 in #165
• re-ran notebooks from main by @mattlevine22 in #171
• updating old error_if t0 statement by @mattlevine22 in #172
• Versioned docs page by @mattlevine22 in #163

Full Changelog: v0.0.1...v0.0.2

v0.0.1

Our initial release of dynestyx! Dynestyx is an extension of NumPyro built with first-class support for Bayesian inference in dynamical systems.

New Contributors

• @DanWaxman made their first contribution in #18
• @mattlevine22 made their first contribution in #12
• @eb8680 made their first contribution in #124

Full Changelog: https://github.com/BasisResearch/dynestyx/commits/v0.0.1