AnyLabeling v0.4.39
Highlights
This release introduces deterministic, cross-platform ONNX Runtime accelerator selection.
Added
- Automatic CUDA selection for Linux and Windows GPU builds.
- CoreML selection on Apple Silicon.
- Opt-in DirectML, OpenVINO, TensorRT, ROCm, MIGraphX, WebGPU, CANN, QNN, and Vitis AI providers.
ANYLABELING_DEVICEoverrides, including Intel, Qualcomm, AMD, and Huawei NPU aliases.- CPU fallback when a requested accelerator is unavailable.
- Pip-managed CUDA 12 and cuDNN runtime discovery, including Windows DLL registration.
Availability
No binary artifacts were published for this tag. Installers with these changes and subsequent packaging fixes are available in v0.4.42.
Full changelog: v0.4.38...v0.4.39