Fix model copying for QDQ stripping #784
Open
+52
−21
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
This was referenced Aug 21, 2025
This comment was marked as outdated.
This comment was marked as outdated.
Sorry, something went wrong.
There was a problem hiding this comment.
Pull Request Overview
This PR fixes model copying functionality required for QDQ (Quantize/Dequantize) stripping when processing GPU models with bfloat16->float16 conversions and 16-bit integer quantization. The changes update the model copying mechanism to handle OrtValue initializers and refine the conditions for QDQ optimization.
- Updated model copying logic to use OrtValue-based initializer handling instead of TensorProto copying
- Added support for INT16/UINT16 data types in type checking and moved the logic from experimental GPU-only to general support
- Enhanced QDQ graph detection to specifically identify graphs with 16-bit quantization for targeted GPU optimization
Reviewed Changes
Copilot reviewed 3 out of 3 changed files in this pull request and generated 3 comments.
| File | Description |
|---|---|
| qdq_scales_fix.cpp | Simplified model copying to use OrtValue initializers and removed redundant tensor proto copying logic |
| data_ops.cc | Moved INT16/UINT16 type support from experimental GPU-only section to general type support |
| backend_manager.cc | Added detection for QDQ graphs with 16-bit quantization and refined GPU optimization conditions |
Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.
onnxruntime/core/providers/openvino/qdq_transformations/qdq_scales_fix.cpp
Outdated
Show resolved
Hide resolved
ankitm3k
approved these changes
Aug 22, 2025
MayureshV1
approved these changes
Aug 22, 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 the model copying portion required by QDQ stripping for GPU and bfloat16->float16 conversions. The copying was broken by the converting the initializers to OrtValues in the upstream repo.
To make the changes in a single place, a PR #768 with a small fix was embedded in this PR.
https://jira.devtools.intel.com/browse/CVS-171536