Skip to content

First draft of a header for ONNXRunTime

tprocter46 requested to merge onnx_runtime_integration into release-3-1-x

Similar idea to RivetLWTNN.hh - provide the bare minimum to cleanly use an ONNX neural network inside a Rivet analysis. This is actually even more usefilf for Onnxruntime, as there is a LOT of boilerplate which looks terrible and is easy to mess up_. Like RivetLWTNN, we require the analysis is built separately against onnxruntime (i.e. -lonnxruntime -Lpath/to/onnxruntime/libs -Ipath/to/onnxruntime/cxx/headers). Unlike function-only RivetLWTNN, requires a class as ONNX requires that quite a few objects are preserved and dragged around with the network

Merge request reports