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)