Large version of the OmniFold dataset in training and test splits

20M Simulated events based on the dataset of QCD jets used for studying unfolding in OmniFold: A Method to Simultaneously Unfold All Observables split into training and validation.

To read, simply open as an hdf5 file. Hard and reconstructed observables are stored into the datasets "hard" and "reco", respectively. These have dimensionality (n_events, 6). The six dimensions are ordered as: "jet mass", "jet width", "jet multiplicity", "jet soft drop mass", "jet groomed momentum fraction z_g", and "jet N-subjettiness ratio tau_21".