Looks cool do python have à wrapper ? First thing to note is that tmva doesn’t seem to optimise gradient calculations only inference.
ROOT: TMVA tutorials seems to use it on a bunch of examples.
Ps : I think Lux can generate ONNX files so it would be duable maybe it can do Julia → mlir → ONNX (using jax)
However you would get similar perf as doing a Jax Vs tmva since Reactant uses the same cool features
1 Like