Which metric is lower at the end of training: the final training loss or the final test loss?

Which metric is lower at the end of training: the final training loss or the final test loss?: a task in SmolDataEnvs-harbor-train-sandbox: 📊 SmolDataEnvs: Harbor (train) (Harbor dataset). Files (in /home/user/input, no subfolders): - train.csv - test.csv