C++ and Python version of the migraphx_test_runner to run models with inputs and outputs to verify correctness. The Python version was already merged to AMDMIGraphX.

You can use the following steps to run the test runner:

You can following the steps to build AMDMIGraphX from source using the rbuild.

The command line is:

rbuild build -d deps -B build --cxx=/opt/rocm/llvm/bin/clang++

Note: The names deps and build should be used in the above command since the C++ version of test runner hard-coded these names in the file CMakeLists.txt.

The commands to run the Python version test runner are:

1. Set the PYTHONPATH1

export PYTHONPATH=$PYTHONPATH:$path/AMDMIGraphX/build/lib

2. Run the test runner

python3 test_runner.py path/to/example

The test example here uses the standard test examples with inputs and inputs in protobuf format as in the onnx repo. You can download some examples using the script.

The C++ version of test runner can be built as:

cmake .. -DCMAKE_PREFIX_PATH=${path}/AMDMIGraphX

where ${path}/AMDMIGraphX is the MIGraphX repo folder.

Then the test runner can be executed as:

./migraphx_test_runner path/to/example

By default, the gpu target is used, you can set the target to ref with the option -t ref.