Evaluating pytorch .pt model with TMVA::Reader

After following the PyTorch tutorial[1], I’m wondering about the recommended way to evaluate a pre-made .pt file using TMVA, e.g., that someone has trained independently of ROOT. I’d like to pick up the model.pt and use it my for own production based on a c++ script that runs TMVA. Normally I use the TMVA::Reader object, with method EvaluateMVA(). Is this possible when using a .pt file, or some other way using c++?

[1] new user, not able to link: “Tutorial: TMVA PyTorch Interface”


@moneta could you answer this one? Thanks!

Hi, thanks in advance!