[WebNN] Correct behavior of Softmax < opset 13 #26391
Open
+46
−10
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
ONNX Softmax operates with different semantics before and after opset 13.
Before opset 13, it normalizes over the flattened range of dimensions
starting from axis to the last dimension. Fix it by reshaping the input
to [M, N] where M = prod(d0..d{axis-1}) and N = prod(d{axis-1}..d{n-1})
and apply softmax along N, then reshaping back to the original shape.
guschmue
approved these changes
Oct 23, 2025
|
/azp run Linux QNN CI Pipeline,Win_TRT_Minimal_CUDA_Test_CI,Windows ARM64 QNN CI Pipeline,Windows GPU Doc Gen CI Pipeline |
|
Azure Pipelines successfully started running 4 pipeline(s). |
|
/azp run MacOS CI Pipeline / iphone_simulator (arm64) |
|
No pipelines are associated with this pull request. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 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Outdated suggestions cannot be applied.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.
ONNX Softmax operates with different semantics before and after opset 13. Before opset 13, it normalizes over the flattened range of dimensions starting from axis to the last dimension. Fix it by reshaping the input to [M, N] where M = prod(d0..d{axis-1}) and N = prod(d{axis-1}..d{n-1}) and applying softmax along N, then reshaping back to the original shape.