Conversation
Upgrade 'nncf' to 3.0 Upgrade 'openvino-model-api' Upgrade 'datumaro'
github-actions
Bot
added
DEPENDENCY
There was a problem hiding this comment.
Pull request overview
Upgrades the OpenVINO/NNCF/Model API/Datumaro dependency stack and applies the required API/type adjustments across library and application code/tests to keep Geti training/inference and dataset flows working.
Changes:
- Bump
openvinoto~2026.0.0,nncfto~3.0, and pinopenvino-model-apito a specific commit. - Update tests and tooling for API/type changes (e.g., OpenVINO
Core, DatumaroExportMode, stricterPathusage). - Adjust benchmark recipe path resolution to correctly map string task identifiers to
OTXTaskType.
Reviewed changes
Copilot reviewed 9 out of 11 changed files in this pull request and generated 1 comment.
Show a summary per file
| File | Description |
|---|---|
| library/tests/unit/tools/test_converter.py | Switch config paths to Path and add pyrefly ignores for updated typing expectations. |
| library/tests/perf_v2/benchmark.py | Tighten sub_work_dir typing to Path and fix recipe folder lookup by casting task strings to OTXTaskType. |
| library/tests/integration/api/test_xai.py | Update OpenVINO import/usage to from openvino import Core. |
| library/tests/integration/api/test_geti_interaction.py | Adjust nested hyper-parameter mutation and add pyrefly ignore for updated typing. |
| library/pyproject.toml | Upgrade OpenVINO/NNCF versions and pin openvino-model-api to a commit. |
| application/backend/tests/integration/services/test_staged_dataset_service.py | Update Datumaro export to use ExportMode.SKIP instead of a boolean flag. |
| application/backend/pyproject.toml | Remove direct OpenVINO deps from backend (backend pulls OTX from ../../library). |
| application/backend/app/services/dataset_revision_service.py | Modify dataset revision export call (notably removes explicit export_images=True). |
| application/backend/app/execution/dataset_export/export.py | Update Datumaro imports to reflect new locations/types. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
| export_dataset( | ||
| dataset=dataset, | ||
| output_path=revision_path, | ||
| export_images=True, | ||
| as_zip=False, # Export as uncompressed directory, see #5070 for details | ||
| ) |
There was a problem hiding this comment.
save_revision() used to explicitly export images (export_images=True), but that argument is now removed. With Datumaro’s updated API (see usage of ExportMode elsewhere in this PR), this now relies on whatever the default export behavior is, which could unintentionally skip images or change how they’re materialized on disk. To preserve the previous behavior and avoid regressions in training/loading revisions, pass an explicit export_images mode (e.g., ExportMode.COPY/equivalent) and import ExportMode here as needed.
📊 Test coverage report
|
Docker Image SizesCPU
GPU
XPU
|
|
Codecov Report✅ All modified and coverable lines are covered by tests. 📢 Thoughts on this report? Let us know! |
Summary
Changes:
openvinoto 2026.0nncfto 3.0openvino-model-apidatumaroResolves #5608
How to test
Train and run inference with models in Geti
Checklist