Datasets:
|
Download README.md from Futuremark/winml-test-set: direct link, hf CLI and curl.
- Browser
- Download file 3.11 kB
-
https://huggingface.co/datasets/Futuremark/winml-test-set/resolve/main/README.md
- Command line
-
hf download hf://datasets/Futuremark/winml-test-set/README.md
-
curl -L -H "Authorization: Bearer $HF_TOKEN" -o README.md https://huggingface.co/datasets/Futuremark/winml-test-set/resolve/main/README.md
3.11 kB
| pretty_name: WinML Test Set | |
| dataset_type: computer_vision | |
| tags: | |
| - evaluation | |
| - accuracy-testing | |
| - windows | |
| - onnx | |
| - directml | |
| - regression | |
| license: | |
| - apache-2.0 # upstream subsets retain their own licenses | |
| - cc-by-4.0 | |
| - cc0-1.0 | |
| task_categories: | |
| - image-classification | |
| - object-detection | |
| - image-to-text | |
| - image-segmentation | |
| language: | |
| - en | |
| # WinML Test Set | |
| ## Dataset Summary | |
| **WinML Test Set** is an evaluation‑only collection for validating model accuracy and stability on Windows ML / DirectML / ONNX Runtime pipelines. It aggregates several permissively‑licensed sources and harmonizes schema for reproducible, regression‑grade testing across backends and versions. *Not intended for training.* | |
| --- | |
| ## Intended Use | |
| - **Accuracy** and **regression** benchmarking of **Windows ML / DirectML / ONNX Runtime** pipelines. | |
| - Cross‑version and cross‑backend **consistency checks** (CPU vs GPU / driver changes). | |
| - CI/CD gates for **non‑regression** evaluation. | |
| ### Out‑of‑Scope Uses | |
| - Training or fine‑tuning models. | |
| - Public leaderboard claims beyond this documented evaluation scope. | |
| - Any use involving personal or sensitive data. | |
| --- | |
| ### Dataset Creation | |
| Sources: See /SOURCES/ (one subfolder per upstream dataset). | |
| Curation: None — this distribution aggregates upstream subsets as‑is. | |
| Processing: None — no changes or label edits were applied in this release. | |
| ### Sources & Provenance | |
| Each upstream dataset has its own folder in /SOURCES/<name>/ with: | |
| SOURCE.md: provenance (URL, authors/publisher, license, retrieval date, included components). | |
| CHANGES.md: this release — no changes (aggregated as‑is). | |
| Maintaining per‑source provenance and change logs supports auditability and license forwarding. | |
| ### Licensing | |
| This is an aggregation. Each upstream subset retains its original license; you must comply with the license applicable to any subset you redistribute: | |
| Apache‑2.0 subsets — include a copy of the license; preserve attribution/patent/trademark notices; if upstream shipped a NOTICE, reproduce it; indicate changes when you modify files (none in this release). | |
| CC BY 4.0 subsets (e.g., Sama‑Coco) — provide TASL attribution (Title–Author–Source–License), link to the license, and indicate if changes were made (none in this release). Do not add legal/technological restrictions that limit the freedoms granted by CC BY. | |
| See /LICENSES for canonical texts and NOTICE for consolidated attribution (and any upstream NOTICE content). | |
| Our contributions in this repository (README, NOTICE, CHANGELOG, and any Futuremark‑authored split or metadata files) are released under CC0 1.0 (Public Domain Dedication). This applies only to our additions; upstream content remains under its original license. Documenting licensing and metadata in the card aids correct reuse on the Hub | |
| ### Changelog | |
| See CHANGELOG.md for versioned updates (counts, label changes, file moves). | |
| #### Contact | |
| For issues/takedowns: contact admin@futuremark.com. No endorsement is implied; trademarks belong to their owners. |