You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
A user can add the TransformValuesOpt as an extra rewrite and have transforms applied during log-probability construction. This process starts with a user-owned and constructed instance of TransformValuesOpt, which takes a dict of values-to-transforms; however, the transforms in this dict aren't necessarily RVTransforms, since they can include the DEFAULT_TRANSFORM sentinel. In these cases, the user doesn't actually have a usable map between values and RVTransforms with which they can—say—apply their own "forward" and "backward" transforms in a manner consistent with the log-probabilities they produce with TransformValuesOpt.
To remedy this, we could perform the _default_transformed_rv checks upfront (e.g. in TransformValuesOpt.__init__) and make the results available to the user via TransformValuesOpt.values_to_transforms.
The text was updated successfully, but these errors were encountered:
A user can add the
TransformValuesOpt
as an extra rewrite and have transforms applied during log-probability construction. This process starts with a user-owned and constructed instance ofTransformValuesOpt
, which takes adict
of values-to-transforms; however, the transforms in thisdict
aren't necessarilyRVTransform
s, since they can include theDEFAULT_TRANSFORM
sentinel. In these cases, the user doesn't actually have a usable map between values andRVTransform
s with which they can—say—apply their own "forward" and "backward" transforms in a manner consistent with the log-probabilities they produce withTransformValuesOpt
.To remedy this, we could perform the
_default_transformed_rv
checks upfront (e.g. inTransformValuesOpt.__init__
) and make the results available to the user viaTransformValuesOpt.values_to_transforms
.The text was updated successfully, but these errors were encountered: