[QNN EP] MaxPool input rank-3 auto pad bug fix #24827
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
- Previously, padding for rank-3 MaxPool was only computed for auto_pad="NOTSET", using the final output shape. - Identified a broader issue during auto_pad="VALID" implementation: padding must be derived from the recalculated output shape. - Added unit tests to cover all use cases of auto_pad.
|
/azp run Linux QNN CI Pipeline,Win_TRT_Minimal_CUDA_Test_CI,Windows ARM64 QNN CI Pipeline,Windows GPU Doc Gen CI Pipeline,Windows x64 QNN CI Pipeline |
|
Azure Pipelines successfully started running 5 pipeline(s). |
HectorSVC
approved these changes
May 21, 2025
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Outdated suggestions cannot be applied.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.
Description
This PR fixes an issue in the padding calculation logic for rank-3 MaxPool operations when using auto_pad. The bug stemmed from using the final output shape (rank-3) to compute padding, rather than the correct intermediate shape (rank-4) that MaxPool actually operates on. The logic has been updated to use the reshaped rank-4 output for accurate padding computation. Unit tests have been added to validate behavior across all auto_pad modes.
Motivation and Context
While implementing support for auto_pad="VALID" in MaxPool, we discovered that the padding for MaxPool rank-3 was being calculated using the final output shape, which is rank-3. However, MaxPool internally operates on a reshaped rank-4 tensor (via pre- and post-processing reshapes). As a result, the padding logic was misaligned with the actual shape used during pooling, leading to test failures.