Sklearn precision recall curve pos_label for unbalanced dataset which class probability to use

Question

i want to evaluate my model using the precision recall scores, because my data is unbalanced. Since I have a binary classification I am using a softmax at the end of my N