File 002 · Local Intelligence
Record 041 · Runtimes · 2018
18.41.UONNX Runtime
Microsoft
“A single runtime that runs ONNX graphs on CPU, CUDA, DirectML, CoreML, TensorRT.”
The interchange runtime.
ORT is how a model trained in PyTorch becomes a thing a Windows NPU or a Mac can execute. The plumbing of local.
The portability layer of on-device inference.
Filed notes
- ONNX execution
- EP plugins (DML, CoreML, TensorRT)
- Cross-OS
SIC
7372
Prepackaged Software
NAICS
511210
Software Publishers
