try:

    from sklearn.metrics import f1_score

    _has_sklearn = True

except (AttributeError, ImportError):

    _has_sklearn = False

def is_sklearn_available():

    return _has_sklearn

if _has_sklearn:

    def simple_accuracy(preds, labels):

        return (preds == labels).mean()

    def acc_and_f1(preds, labels):

        acc = simple_accuracy(preds, labels)

        f1 = f1_score(y_true=labels, y_pred=preds)

        return {

            "acc": acc,

            "f1": f1,

            "acc_and_f1": (acc + f1) / 2,

        }

    def glue_compute_metrics(preds, labels):

        assert len(preds) == len(labels)

        return acc_and_f1(preds, labels)