Builds of ONNX runtime for various platforms.
This repo provides pre-built libraries of ONNX Runtime specifically for use with ort_dart.
This repo is based of the excellent work of alfatraining's ort-artifacts repo and pykeio's ort artifacts repo.
To make sure the binaries are available, the binaries are built and released in this repo. Otherwise, the binaries provided by the other repos might change, deleted, or not be compatible anymore with ort_dart. This helps ensures that the binaries that are needed/required are available. This repo also compiles a couple extra binaries for needed architectures/platforms.
Each platform has certain execution providers compiled by default.
🗒️ If you would like specific execution providers to be built-in by default feel free to make an issue so this can be discussed more!
Here is a matrix breakdown:
| Platform | Execution Providers |
|---|---|
| Android | NNAPI, XNNPACK |
| iOS | CoreML, XNNPACK |
| Linux | OpenVINO, XNNPACK |
| MacOS | CoreML, XNNPACK |
| WASM | WebGPU |
| Windows | DirectML, OpenVINO (x86_64 only), XNNPACK |
You can find more information about each platform below.
The minimum SDK is 28. While you can go lower, this is best for performance.
The minimum iOS version is 15. While you can go lower, this is best for performance.
glibc ≥ 2.35 & libstdc++ >= 12 (Ubuntu ≥ 22.04, Debian ≥ 12 ‘Bookworm’)
MacOS 13.3 or greater.
Comes bundled with the WebGPU execution provider.
⚠️ ARM Support is disabled for the time being until build can be fixed.
Supports Windows 10 and 11.
OpenVINO does not have builds for arm64. npm distribution states "Windows ARM is not supported".