Functional tensors for probabilistic programming
partial_sum_product
by @ordabayevy in https://github.com/pyro-ppl/funsor/pull/606
Full Changelog: https://github.com/pyro-ppl/funsor/compare/0.4.5...0.4.6
Minor release supporting numpyro 0.11.0
Minor release supporting jax 0.4.
Full Changelog: https://github.com/pyro-ppl/funsor/compare/0.4.2...0.4.3
ProvenanceTensor
for torch backendConstant
funsorfunsor.recipes.forward_fiter_backward_rsample()
Precondition
interpretation for Gaussians, plus funsor.recipes.forward_fiter_backward_precondition()
pprint.pprint()
and pdbsqrt(precision)
representation for Gaussian
Minor release updating to PyTorch 1.9
ExtendedDistribution
IndependentConstraint
.reduce(op, vars_not_in_inputs)
Tuple
Funsorsarkka_bilmes_product
interfacefunsor.domain.Domain
s: bint(n) -> Bint[n]
, reals(*shape) -> Reals[shape]
StatefulInterpretation
pattern to simplify context-dependent interpretationsnumpy
(default), torch
, and jax
. To change the backend, use funsor.set_backend("torch")
or funsor.set_backend("jax")
.pytest