Description
I followed the README instructions on compilation and at the end I faced the
UNAVAILABLE: Unsupported: Triton TRITONBACKEND API version: 1.16 does not support 'onnxruntime' TRITONBACKEND API version: 1.19
error when trying to run Triton Inference Server with the backend swapped. I think you forgot to mention in the README that it is absolutely necessary to specify correct -DTRITON_CORE_REPO_TAG, otherwise it fails on API version check in TRITONBACKEND_Initialize in onnxruntime.cc.
For example, you say to run
cmake -DCMAKE_INSTALL_PREFIX:PATH=`pwd`/install -DTRITON_BUILD_ONNXRUNTIME_VERSION=1.14.1 -DTRITON_BUILD_CONTAINER_VERSION=23.04 ..
"to build the ONNX Runtime backend for Triton 23.04" which is simply wrong and the correct command would be
cmake -DCMAKE_INSTALL_PREFIX:PATH=`pwd`/install -DTRITON_BUILD_ONNXRUNTIME_VERSION=1.14.1 -DTRITON_BUILD_CONTAINER_VERSION=23.04 -DTRITON_CORE_REPO_TAG=r23.04 ..