版本資訊Release notes
此頁面會在最新的 Windows 10 SDK 和 NuGet 套件組建中記錄 Windows ML 的更新。This page records updates to Windows ML in the latest builds of the Windows 10 SDK and NuGet Package.
Windows ML NuGet 套件 – 版本 1.4Windows ML NuGet Package – Version 1.4
- 在此下載 NuGetDownload NuGet here
- 建置於 ONNX Runtime 1.4 上Built on ONNX Runtime 1.4
- 支援 ONNX 1.6 和 opset 11Support for ONNX 1.6 and opset 11
- 一般使用性和效能提升General usability and performance improvements
Windows ML NuGet 套件 - 版本 1.3Windows ML NuGet Package - Version 1.3
- 在此下載 NuGetDownload NuGet here
- 建置於 ONNX Runtime 1.3 上Built on ONNX Runtime 1.3
- 對應至 MachineLearningContract v3Corresponds to MachineLearningContract v3
- 支援 ONNX 1.6 和 opset 11Support for ONNX 1.6 and opset 11
- CPU 執行向下支援到 Windows 8.1;GPU 執行向下支援到 Windows 10 版本 1709CPU execution supported down to Windows 8.1; GPU execution supported down to Windows 10 version 1709
- 經過認證的已知已測試路徑是使用 C++ 的傳統型應用程式。Certified known tested paths are Desktop Applications using C++. 尚不支援儲存應用程式和 Windows 應用程式認證套件。Store applications and the Windows Application Certification Kit are not yet supported.
組建 19041 (Windows 10 2004 版)Build 19041 (Windows 10, version 2004)
支援 ONNX 1.4 和 opset 9 (CPU 和 GPU)Support for ONNX 1.4 and opset 9 (CPU and GPU)
API 介面新增項目:API Surface additions:
- CloseModelOnSessionCreation:新增可設定的 LearningModelSessionOptions 參數,用以減少工作記憶體。CloseModelOnSessionCreation: new LearningModelSessionOptions parameter to configure to reduce working memory.
工具:Tooling:
- WinMLTools 轉換器支援新的 ONNX 版本和 opsetWinMLTools converters supports new ONNX versions and opset
- 公開新效能計量的 WinMLRunner 最佳化Optimizations to WinMLRunner exposing new performance metrics
組建 18362 (Windows 10 版本 1903)Build 18362 (Windows 10, version 1903)
先前小眾測試組建中的所有功能和更新:All features and updates from previous flighted builds:
- ONNX 1.3 支援ONNX 1.3 support
- 支援透過訓練後權數量化來縮減模型大小。Support for model size reduction via post-training weight quantization. 您可以使用最新的 WinMLTools 版本,將模型的權數向下封裝到 int8。You can use the latest version of WinMLTools to pack the weights of your model down to int8.
- 從 Windows 10 SDK 移除 mlgen—改用下列其中一個 Visual Studio 擴充功能:Removal of mlgen from the Windows 10 SDK—use one of the following Visual Studio extensions instead:
- Visual Studio 2017:Windows Machine Learning 程式碼產生器 VS 2017Visual Studio 2017: Windows Machine Learning Code Generator VS 2017
- Visual Studio 2019:Windows Machine Learning 程式碼產生器Visual Studio 2019: Windows Machine Learning Code Generator
組建 18829Build 18829
- 已從 Windows 10 SDK 移除 mlgen。mlgen has been removed from the Windows 10 SDK. 相反地,請根據您的版本安裝下列其中一個 Visual Studio 擴充功能:Instead, install one of the following Visual Studio extensions depending on your version:
- Visual Studio 2017:Windows Machine Learning 程式碼產生器 VS 2017Visual Studio 2017: Windows Machine Learning Code Generator VS 2017
- Visual Studio 2019:Windows Machine Learning 程式碼產生器Visual Studio 2019: Windows Machine Learning Code Generator
組建 18290Build 18290
- 最小支援的 ONNX 版本 = 1.2.2 (opset 7)Min supported ONNX version = 1.2.2 (opset 7)
- 最小支援的 ONNX 版本 = 1.3 (opset 8)Max supported ONNX version = 1.3 (opset 8)
- 支援透過訓練後權數量化來縮減模型大小。Supports model size reduction via post-training weight quantization. 您可以使用最新的 WinMLTools 版本,將模型的權數向下封裝到 int8。You can use the latest version of WinMLTools to pack the weights of your model down to int8.
組建 17763 (Windows 10 版本 1809)Build 17763 (Windows 10, version 1809)
- Windows Machine Learning 的第一個正式發行版本。First official release of Windows Machine Learning.
- 需要 ONNX v 1.2。Requires ONNX v1.2.
- Windows.AI.MachineLearning 命名空間已取代 Windows.AI.MachineLearning.Preview 命名空間。Windows.AI.MachineLearning.Preview namespace deprecated in favor of Windows.AI.MachineLearning namespace.
已知問題Known issues
- 對於包含序列的模型,MLGen 會產生 IList<Dictionary<key, value>> (而不是適當的 IList<IDictionary<key, value>> ),進而導致空的結果。For models containing sequences, MLGen generates an IList<Dictionary<key, value>> instead of the proper IList<IDictionary<key, value>>, leading to empty results. 若要修正此問題,只要將自動產生的程式碼取代為適當的 IList<IDictionary<key, value>> 。To fix this issue, simply replace the automatically generated code with the appropriate IList<IDictionary<key, value>> instead.
組建 17723Build 17723
- 需要 ONNX v 1.2。Requires ONNX v1.2.
- 支援採用 GPU 型模型推斷的 F16 資料類型,以獲得更佳效能及降低模型使用量。Supports F16 datatypes with GPU-based model inferences for better performance and reduced model footprint. 您可以使用 WinMLTools 將您的模型從 FP32 轉換為 FP16。You can use WinMLTools to convert your models from FP32 to FP16.
- 可讓桌面應用程式透過 WinRT/C++使用 Windows.AI.MachineLearning API。Allows desktop apps to consume Windows.AI.MachineLearning APIs with WinRT/C++.
注意
使用下列資源取得 Windows ML 的說明:Use the following resources for help with Windows ML:
- 如需詢問或回答有關 Windows ML 的技術問題,請使用 Stack Overflow 上的 windows-machine-learning 標籤。To ask or answer technical questions about Windows ML, please use the windows-machine-learning tag on Stack Overflow.
- 如需回報錯誤 (bug),請在 GitHub 上提出問題。To report a bug, please file an issue on our GitHub.
- 如需要求功能,請前往 Windows 開發人員意見反應。To request a feature, please head over to Windows Developer Feedback.