Fix issue with adjoint sensitivities for domains where rate coefficients vary #95
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
ReverseDiffVJP(true) assumed there was no programmatic branching in the function This is true for most simulations when the rate coefficients are constant, but the assumption breaks down when the rate coefficients aren't constant Originally ReverseDiffVJP(false) gave NaNs when run and TrackerVJP() would crash (as it does currently) and so ReverseDiffVJP(true) was set as the default This should resolve issues with inconsistency between forward sensitivities and adjoint sensitivities
…ate coefficients not constant)
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
So I'm no longer able to reproduce the issue I had originally where using ReverseDiffVJP(false) gave NaNs forcing us to use ReverseDiffVJP(true) (which assumes no programmatic branching in the code, which should be true for domains with constant rate coefficients, but often not true for other domains) see #84 . I'm not entirely sure if this is something we fixed internally while working on something separate or it's related to package versioning. If the issue does up later we should be able to resolve it by adjusting our compat requirements. This PR switches us to default ReverseDiffVJP(false) and adds a test for adjoint and forward sensitivities for the ConstantVDomain.