Roc_auc_score error

I have a CNN model for the classification task with 11 classes.
I want to use sklearn.metrics to print metrics like F1, precision, roc_auc, etc.
I’ve calculated y_real and y_pred as follow:

  with torch.no_grad():
        for inputs, targets in test_dl:
            targets = targets[0]

            inputs =
            outputs = model(inputs)
            _, preds = torch.max(outputs, 1)

            predict += [preds.cpu().numpy()]
            label += [targets.cpu().numpy()]

    y_pred = np.concatenate(predict)
    y_real = np.concatenate(label)
    return y_real, y_pred

y_real and y_pred work ok and without any error for precision_score, recall_score, f1_score, for example:
print(recall_score(y_real, y_pred))

but for roc_auc, it doesn’t accept y_real and y_pred:

print('ROC: {:.4f}\n'.format(roc_auc_score(y_real, y_pred)))
ValueError: multi_class must be in ('ovo', 'ovr')

I also tested with multi_class='ovo' and multi_class='ovr', I got this error:
AxisError: axis 1 is out of bounds for array of dimension 1

The type of y_real and y_pred and their shape for both of them are:

<class 'numpy.ndarray'>