Introduce internal interface for ValuePropBackend
#201
+256
−107
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
Signed-off-by: Atanas Dimitrov <[email protected]>
Signed-off-by: Atanas Dimitrov <[email protected]>
neNasko1
requested review from
adityagoel4512,
cbourjau and
jbachurski
as code owners
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Outdated suggestions cannot be applied.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.
Since we currently "bless" the reference and onnxruntime implementations as the only way to propagate value internally, this PR refactors a bit the internals to allow a better interface for "custom" execution providers(i.e. @cbourjau rust runtime, onnx-mlir...).
Consult the tests for a POC which could also be used to register custom opsets as well.
An additional internal change is related to the
PropDicttype as previously it was too wide and didn't capture the correct semantics of the returned values from onnxruntime. As per the definition of from_ref_value we can see that the returned feed is of typedict[str, PropValue]which makes more sense to model around.This exposed a bug in the function value inference which was unnoticed and is now fixed.
Checklist
CHANGELOG.rstentry