Update TypeCheckAttr to infer what not to differentiate wrt for non-differentiable types. #22915
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
Contributor
Author
|
@swift-ci please test tensorflow |
Contributor
|
@dan-zheng is most familiar with this code path so I'll let him take a look first. My initial speculation is that the original attribute checking code was already checking |
Contributor
By "infer |
pschuh
changed the title
Contributor
Author
|
Perhaps? isDifferentiableParam is the only part that overlaps. I suppose I could make a helper lambda that does isDifferentiableParam but takes a bool whether or not to emit errors? |
dan-zheng
reviewed
Feb 26, 2019
This happens when inferring default 'wrt:' for @differentiable. This should resolve: TF-296.
Contributor
Author
|
@swift-ci please test tensorflow |
2 similar comments
Contributor
Author
|
@swift-ci please test tensorflow |
Contributor
Author
|
@swift-ci please test tensorflow |
rxwei
approved these changes
Feb 26, 2019
dan-zheng
approved these changes
Feb 26, 2019
rxwei
added a commit
to tensorflow/swift-apis
that referenced
this pull request
Feb 26, 2019
…ed! (#33) We've recently changed the type checker to improve the ergonomics of the `@differentiable` attribute. * [swiftlang/swift#22915](swiftlang/swift#22915): Explicit differentiation parameters are no longer required in a `@differentiable` attribute when the function has some arguments that do not conform to `Differentiable`. Those non-differentiable parameters will be skipped and the rest will be differentiated with respect to. * [swiftlang/swift#22877](swiftlang/swift#22877): On an instance method, when a `wrt:` is not specified, `self` is being implicitly included as a differentiation parameter. * [swiftlang/swift#22877](swiftlang/swift#22877): When a `@differentiable` requirement is not met, the `@differentiable` attribute fix-it will appear exactly as written in the original declaration instead of the most complex, canonical form. For instance, `@differentiable` instead of `@differentiable(wrt: (x))`. This greatly simplifies libraries and applications that use automatic differentiation. The protocol requirement `Layer.applied(to:in:)` becomes as simple as this: ```swift @differentiable func applied(to input: Input, in context: Context) -> Output ``` This PR updates deep learning APIs to use the simplest form of `@differentiable` possible. Hooray!
rxwei
pushed a commit
to rxwei/swift
that referenced
this pull request
May 11, 2019
…wiftlang#22915) This happens when inferring default 'wrt:' for @differentiable. This should resolve: TF-296.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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.
This happens when inferring default 'wrt:' for @differentiable.
This should resolve: TF-296.