what's the steps to realize run model on qdsp or enpu?

[FMzzq]

firstly, we can train a f32 model by pytorch or tensorflow, as we all know.
then, we can use Aimet to get quantization table , it is onnx and json file?
and this is the point, how can we use thes onnx and json to run on chip, maybe runtime lib or else?
can some one give the rigth step to realize that?