I want to load .onnx or .nn generated by mlagents in Keras or Pytorch, because I want to see some things in the shap library.
Has anyone had the experience of loading the weights generated by mlagents while training in either of these two frameworks?