Basic newbie question crossentropy loss

Newbie question :
I am working on a binary image classification problem.
The output of my model gives 2 features (n_out = 2).
For the labels in my dataset : Do I need to put only the label tensor or a tensor of bools ?? :
tensor[0] / tensor[1]
tensor[[0,1]] / tensor [[1,0]].
I read in pytorch it is the second solution but in fast ai it seems that we just need to put the label 0 or 1.