microsoft / microsoft/winml-cli
SA: QDQ dynamic input support — all EPs (low priority)
Open
Nobody has claimed this yet.
EP scale
P2
QDQ
static-analyzer
triaged
- Dominant language
- Python
- Stars
- 40
- Forks
- 11
- Avg merge
- 1d 8h
- Merged PRs (30d)
- 50
Description
Overview
Extend static analyzer to correctly classify ops in QDQ (quantized) models that also have dynamic input shapes. This is data-driven and low priority — only pursue if coverage data from static+dynamic phases reveals meaningful gaps with QDQ models.
Scope (all EPs)
- QNN Adreno GPU
- Intel NPU (OpenVINO)
- Intel GPU (OpenVINO)
- Intel CPU (OpenVINO)
- AMD NPU (VitisAI)
Acceptance Criteria
- SA correctly classifies ops for QDQ + dynamic-shape models on all EPs
- Validated on 52w-scale QDQ dynamic input test cases per EP
- Gated on findings from static (#148, #94, #142, #143, #145) and dynamic (#301, #302, #303, #304, #305) phases
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.
Research direction
Start by reviewing the static phases (#148, #94, #142, #143) and dynamic phases (#301–#305), then inspect coverage findings for QDQ models with dynamic input shapes. Validate any changes against the 52w-scale QDQ dynamic input test cases for each listed execution provider. Done means correct op classification across all five EPs, gated by meaningful coverage gaps.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- machine-learning
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Quiet
- Clarity
- Mostly clear
- Newbie friendliness
- 42/100