values

pmrf.parameters.values(tree, where: str | Sequence[str] | Callable[[Any], Any] = '*', *, free_only: bool = False, space: Literal['raw', 'declared', 'physical'] = 'declared') → dict[str, Array]

Returns the values of a model’s parameters by name, in one space.

This is what update() accepts, so prf.update(m, prf.values(m, space=s), space=s) gives back m, with the same structure and jit cache key.

The raw value of a parameter under a joint prior (see prior()) is its slice of the prior’s whitened vector, in its own shape, so it depends on the other parameters under the prior, and is the same whichever parameters where selects.

Parameters:

  • tree (PyTree) – A model, or any collection of models and parameters.

  • where (str, Sequence[str] or Callable, default='*') – The parameters to read, as for params().

  • free_only (bool, default=False) – Only return free parameters.

  • space ({'declared', 'physical', 'raw'}, default='declared') – The space of the values. See Param.

Returns:

Names, as from params(), mapped to values.

Return type:

dict[str, jax.Array]

Examples

c = Capacitor(prf.Unconstrained(2.0, scale=1e-12))
prf.values(c)                      # {'C': 2.0}
prf.values(c, space='physical')    # {'C': 2e-12}