Provide transform and inverse method implementations on AbstractBijection #193
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Provide an implementation for
transform
andinverse
inAbstractBijection
, by takingself.transform_and_log_det(x, condition)[0]
,self.inverse_and_log_det(y, condition)[0]
. The original motivation for separately defining the methods was to avoid computing the log_det duringtransform
andinverse
. However, under JIT, the code seems to be removed by the compilers dead code elimination, and I have not seen any performance degradation from switching to this simpler approach. There is likely a performance penalty of this approach outside of JIT, but this is not something I am majorly concerned about.We now only wrap the
*_log_det
methods with shape checking, as thetransform
andinverse
methods will be checked by calling the*_log_det
methods anyway. This does provide some risk for users who have defined custom bijections, as they will now be overwriting thetransform
andinverse
methods, with a non-shape checked version. In general, I'd recommend no longer defining theinverse
andtransform
methods.