add more labels for transformers training

[iulusoy]

At the moment, in transformers training, all labels are given only as begin moralization (2), and moralization (2). For example, if task=task2 is selected, the labels are still generated as 2 and 1, where in fact it should be distinguished between Fairness, Cheating, Harm, etc by labels 1, 2, 3, etc.