Cross-entropy rises while AUC rises (pytorch)

Question

I have a feed-forward neural network and a binary classification problem.

Defining the loss function as

def cross_entropy(ys,ts):
    cross_entropy = -torch.sum(ts * torch.log(ys+0.00001) + (1-ts)*torch.log(1-ys+0.00001))
    return cross_entropy

and the AUC as

def auc(ys, ts):
    ts = ts.detach().numpy()
    ys = ys.detach().numpy()
    return roc_auc_score(ts,ys)

where ts and ys is target/net-output (for class 1) respectively. For some reason, when I train, the cross-entropy rises and the AUC rises. I would think either should fall when the other one grows.