Hi,
So the Compute function is like the predict in other frameworks, like sklearn or I think also xgboost. You can either put in your data as a vector<float>, a RTensor<float> or a numpy.array. However, if you want to inject the model prediction into an RDataFrame, you have to add it in a Define node. Actually, we have a tutorial for this here.
Further, if you want to do this from Python, we have currently the issue that the Python bindings break if you want to make the TMVA::Experimental::Compute(model) work. We just found this out in another forum thread, see here. There you find also the solution to this issue.
Best
Stefan