move ad before mandatory inlining #21560
Merged
Conversation
This file contains hidden or 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
rxwei
approved these changes
Dec 28, 2018
|
@swift-ci please test tensorflow |
rxwei
reviewed
Dec 29, 2018
| // FIXME: Propagate wrt indices correctly so that this actually takes the | ||
| // gradient wrt only the first parameter, as intended. | ||
| // SupersetAdjointTests.test("CrossModule") { | ||
| // expectEqual(1, gradient(at: 1, 2, in: (+) as @autodiff (Float, @nondiff Float) -> Float)) |
There was a problem hiding this comment.
Actually, this will not work semantically. The arity of the differential operator determines the number of results returned by the pullback. We can't expect to exhaustively define differential operators for arbitrary combinations of @nondiff and non-@nondiff parameters.
Therefore, gradient(at: 1, 2, in: ...), gives you a pullback that returns a tuple. To obtain VJPs only w.r.t. the first parameter, you need to pass the single argument in place of at: and pass non-diff arguments as constants, i.e. gradient(at: 1, in: { x in x + 2 }). The function being passed to Builtin.autodiffApply will not have a @nondiff on its parameter types.
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Outdated suggestions cannot be applied.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.
Moves AD before mandatory inlining. Necessary fixes were:
Unfortunately, this does not fix stored property differentiation because SILGen actually generates direct
struct_extracts instead of getter calls in some cases. Also, this does not fixtensor_autodiff_runtime.swift. I will look into these two problems next.