Simulation-based inference toolkit
sbi
to an new github organization: https://github.com/sbi-dev/sbi
sbi
docs: https://sbi-dev.github.io/sbi/
.sbi.analysis.pairplot
: upper
was replaced by offdiag
and will be deprecated in a future release.process_prior
(thanks to @musoke, #813)pairplot
:
upper
is now called offdiag
to match other kwargs.samples
and points
legend
and pass kwargs
for the legend.append_simulations
(thanks to @VictorSven, #803)simulate_for_sbi
(@jan-matthis, #876)3.9.13
(@michaeldeistler, #888, #900)simulate_for_sbi
on multiple workers (#762)enable_transforms
in sampler interface (#756)pyknos
with bugfix for APT with MDNs (#734)force_first_round_loss=True
as default (#729)ArviZ
integration (#727)arviz
for posterior plotting and MCMC diagnostics (#546, #607, thanks to @sethaxen)VIPosterior
with MultipleIndependent
prior, a51e93bpairplot
and MCMC kwargsDataset
to reduce memory load and add flexibility
with large data sets (#685, thanks to @tbmiller-astro)DirectPosterior
for more flexibility with custom priors (#714)sbi
v0.17.2
or older can not be loaded under sbi
v0.18.0
or newer.sample_with
can no longer be passed to .sample()
. Instead, the user has to rerun
.build_posterior(sample_with=...)
. (#573)posterior
no longer has the the method .sample_conditional()
. Using this
feature now requires using the sampler interface
(see tutorial
here) (#573)retrain_from_scratch_each_round
is now called retrain_from_scratch
(#598, thanks to @jnsbck)sbi v0.14.0
and v0.15.0
are not enforced. Using the interface prior to
those changes leads to an error (#645)sampler interface
(#573)Sequential Neural Variational Inference (SNVI)
(Glöckler et al. 2022) (#609, thanks to @manuelgloeckler)init_strategy
for MCMC (#646)ax
and fig
(#557)init_strategy
for MCMC (#605)mp_context
to allow for multi-chain pyro samplers (#608, thanks to @sethaxen)conditional_pairplot
and ActiveSubspace
(#613)MultipleIndependent
distribution as prior (#619)init_strategy
is now called proposal
instead of prior
(#602)pickle
(#617)posterior
(#631, thanks to @tomMoral)RestrictedPrior
as prior for SNPE
(#642)posterior = inference.build_posterior(sample_with_mcmc=True)
New syntax:
posterior = inference.build_posterior(sample_with="mcmc") # or "rejection"
kde=True
) (#525).DirectPosterior
trained
with MDNs (thanks @jnsbck #458).scatter
allowed for diagonal entries in pairplot (#510)SNPE_A
(thanks @famura, #496, #497)within_prior
checks (#506)