v0.19.0

[janfb]

v0.19.0

Major changes and bug fixes

• new option to sample posterior using importance sampling (Importance sampler #692)
• new option to use arviz for posterior plotting and MCMC diagnostics (Add features to support ArviZ integration #546, Add features to support ArviZ integration, Rebased #607, thanks to @sethaxen)
• fixes for using the VIPosterior with MultipleIndependent prior, a51e93b
• bug fix for sir (sequential importance reweighting) for MCMC initialization (Importance sampler #692)
• bug fix for SNPE-A 565082c
• bug fix for validation loader batch size (fix bug in clipped batch size #674, thanks to @bkmi)
• small bug fixes for pairplot and MCMC kwargs

Enhancements

• improved and new tutorials:
  • Tutorial for simulation-based calibration (SBC) (A tutorial on how to use SBC for SBI diagnosis #629, thanks to @psteinb)
  • Tutorial for sampling the conditional posterior (Tutorial for sampling the conditional #667)
• new option to use first-round loss in all rounds
• simulated data is now stored as Dataset to reduce memory load and add flexibility
  with large data sets (Changes to how data is stored #685, thanks to @tbmiller-astro)
• refactoring of summary write for better training logs with tensorboard (Refactor logging with summary writer #704)
• new option to find peaks of 1D posterior marginals without gradients (add posterior_peaks function to find peaks of posterior distribution. #707, Refactor "add posterior_peaks function to find peaks of posterior distribution." #708, thanks to @Ziaeemehr)
• new option to not use parameter transforms in DirectPosterior for more flexibility with custom priors (make transform optional for DirectPosterior. #714)

---

This discussion was created from the release v0.19.0.