microsoft / microsoft/winml-cli
SA: Dynamic input support — QNN Adreno GPU
Open
@fangyangci is already working on this.
Since Apr 9, 2026.
EP scale
GPU
P2
static-analyzer
triaged
- Dominant language
- Python
- Stars
- 40
- Forks
- 11
- Avg merge
- 1d 8h
- Merged PRs (30d)
- 50
Description
Overview
Extend static analyzer support for models with dynamic input shapes on the QNN Adreno GPU EP. Currently SA coverage collection (#148) targets static shape models. Dynamic input shapes are increasingly common (variable sequence length, batch size, resolution) and must be analyzed correctly for SA to give accurate compatibility predictions.
Scope
- Validate SA op classification correctness for dynamic-shape ONNX models on QNN Adreno GPU
- Ensure partial/unsupported op detection is not affected by dynamic dimensions
- Target: ~520k test cases with dynamic input shapes
Acceptance Criteria
- SA correctly classifies ops for dynamic-shape models on QNN Adreno GPU
- No false SUPPORTED results for ops that fail due to dynamic shapes
- Validated on 52w-scale dynamic input test cases
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Assessment
This issue has not been assessed yet.