Let’s let tp = TabularPandas(...). I don’t quite remember the exact attribute names off the top of my head, but it should be similar. To access the training or validation splits, we can write either tp.train or tp.valid respectively. To access the data and the targets, we can write tp.xs, tp.y, tp.train.xs, tp.train.y, tp.valid.xs, or tp.valid.y. It’s after accessing these data or targets that you can access specific rows.