How to augment dataset?

I have a sequential dataset.
For example, x1, x2, x3 and a corresponding label y.

I want to expand data only in training session like

[x1, x2, x3] -> [(x1, x2, x3), (x2, x1, x3), (x2, x3, x1)] for label y

and in validate and test session. I don’t expand data just use [x1, x2, x3]
How can i augment data only in training time???

and I dont use pytorch dataloader.