# How to check corresponding labels for predictions/softmax values?

Suppose that I have a tensor produced by `F.softmax(nn.Linear())` of shape [64, 2], where 64 is the batch size and 2 is the number of classes;

``````tensor([[0.5382, 0.4618],
[0.5180, 0.4820],
[0.5294, 0.4706],
[0.5272, 0.4728],
[0.5307, 0.4693],
[0.5291, 0.4709],
[0.5215, 0.4785],
[0.5300, 0.4700],
[0.5330, 0.4670],
``````

And that my label/class can either be 0 or 1.
How can I know which probability corresponds to which label? I suppose that at training time, I need to pick the probability value of the true tag and feed that to the loss function.

For the first sample, you have 0.5382 probability for class 0 and 0.4618 probability for class 1.