microsoft / microsoft/winml-cli

SA: Coverage report investigation — OV NPU (2026-04-03)

Open
#309 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

P2 static-analyzer triaged
Dominant language
Python
Stars
40
Forks
11
Avg merge
1d 8h
Merged PRs (30d)
50

Description

Overview

Investigate regressions and unknown ops surfaced in the OpenVINO NPU SA coverage report (2026-04-03 run). Report: https://icy-moss-029643d00.6.azurestaticapps.net/sa_e2e_coverage/OpenVINOExecutionProvider_NPU/0403/sa_eval_report.html

Summary stats: Avg SUPPORTED (Pre): 95.5% → (Post): 95.5% | Avg delta: -0.1% | Avg UNKNOWN: 27 | All-SUPPORTED (Post): 99.1%

Regressions (Post < Pre)

Model Task Pre Post Delta
microsoft/swin-large-patch4-window7-224 image-classification 68.8% 66.7% -2.1%

Partial op patterns involved: Transpose

Unknown Ops

Model Unknown Count Op Types
microsoft/deberta-xlarge-mnli 1 Slice
microsoft/swin-large-patch4-window7-224 4 Add, Pad, Reshape, Slice
PekingU/rtdetr_r50vd_coco_o365 10 GridSample, Tile, TopK, Add, Concat, Div, Mul, Reshape, Slice, Sub
PekingU/rtdetr_r101vd_coco_o365 10 GridSample, Tile, TopK, Add, Concat, Div, Mul, Reshape, Slice, Sub
PekingU/rtdetr_v2_r18vd 8 GridSample, Tile, TopK, Add, Mul, Reshape, Slice, Sub

Acceptance Criteria

  • Root cause identified for swin-large regression (-2.1%) on OV NPU
  • Slice classified correctly for deberta-xlarge-mnli and swin on OV
  • GridSample, Tile, TopK added to OV coverage DB or flagged as unsupported
  • Next report shows 0 regressions and 0 unknown ops for the above models

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start with the linked OpenVINO NPU SA coverage report for the 2026-04-03 run and compare the pre/post results for swin-large-patch4-window7-224. Investigate the listed partial and unknown ops across the named models, then verify their classification or coverage status. Done means the regression root cause is identified, required ops are covered or explicitly unsupported, and a subsequent report has no listed regressions or unknown ops.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
machine-learning, testing-qa
Issue type
Bug
Difficulty
5/5
Estimated time
Over a week
Activity status
Quiet
Clarity
Mostly clear
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.