How to use the fairlearn.metrics function in fairlearn

def test_negative_results(self):
        y_a = [0, 0, 1, 1, 0, 1, 1, 1]
        y_p = [0, 1, 1, 1, 1, 0, 0, 1]
        gid = [0, 0, 0, 0, 0, 1, 1, 1]

        def negative_results(y_true, y_pred):
            return -(len(y_true) + len(y_pred))

        result = metrics.metric_by_group(negative_results, y_a, y_p, gid)

        assert result.overall == -16
        assert result.by_group[0] == -10
        assert result.by_group[1] == -6
        assert result.minimum == -10
        assert result.maximum == -6
        assert result.range == 4
        assert np.isnan(result.range_ratio)

def test_smoke(self, transform_y_a, transform_y_p, transform_gid):
        y_a = transform_y_a([0, 0, 1, 1, 0, 1, 1, 1])
        y_p = transform_y_p([0, 1, 1, 1, 1, 0, 0, 1])
        gid = transform_gid([0, 0, 0, 0, 1, 1, 1, 1])

        result = metrics.metric_by_group(mock_func, y_a, y_p, gid)

        assert result.overall == 5
        assert len(result.by_group) == 2
        assert result.by_group[0] == 2
        assert result.by_group[1] == 3
        assert result.minimum == 2
        assert result.argmin_set == {0}
        assert result.maximum == 3
        assert result.argmax_set == {1}
        assert result.range == 1
        assert result.range_ratio == pytest.approx(0.6666666667)

supported_metrics_weighted = [(skm.accuracy_score, metrics.group_accuracy_score),
                              (skm.confusion_matrix, metrics.group_confusion_matrix),
                              (skm.zero_one_loss, metrics.group_zero_one_loss)]

# The following only work with binary data when called with their default arguments
supported_metrics_weighted_binary = [(skm.precision_score, metrics.group_precision_score),
                                     (skm.recall_score, metrics.group_recall_score),
                                     (skm.roc_auc_score, metrics.group_roc_auc_score),
                                     (skm.mean_squared_error, metrics.group_mean_squared_error)]
supported_metrics_weighted_binary = supported_metrics_weighted_binary + supported_metrics_weighted


metrics_no_sample_weights = [(skm.max_error, metrics.group_max_error),
                             (skm.mean_absolute_error, metrics.group_mean_absolute_error),
                             (skm.mean_squared_log_error, metrics.group_mean_squared_log_error),
                             (skm.median_absolute_error, metrics.group_median_absolute_error)]

supported_metrics_unweighted = metrics_no_sample_weights + supported_metrics_weighted_binary

# =======================================================


@pytest.mark.parametrize("func_tuple", supported_metrics_unweighted)
def test_metric_unweighted(func_tuple):
    metric_func = func_tuple[0]
    group_metric_func = func_tuple[1]

    result = group_metric_func(Y_true, Y_pred, groups)
    # We don't really care about the numbers (sklearn is responsible)
    # We just want to make sure we got a result
    assert len(result.by_group) == 5

Y_true = [0, 0, 0, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1]
Y_pred = [1, 0, 1, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1]

Y_true_ternary = [a, b, c, c, c, b, b, b, c, c, a, a, a, a, a, b, c, c]
Y_pred_ternary = [b, c, c, c, b, b, b, b, b, c, a, a, c, a, a, b, c, c]

groups = [3, 4, 1, 0, 0, 0, 3, 2, 0, 1, 2, 3, 4, 0, 1, 2, 3, 4]
weight = [1, 2, 3, 1, 2, 3, 4, 2, 3, 3, 2, 1, 2, 3, 1, 2, 3, 4]
group2 = [0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1]

# =======================================================

# Define as a dictionary so that the actual name can be seen
# when pytest builds the tests

supported_metrics_weighted = [(skm.accuracy_score, metrics.group_accuracy_score),
                              (skm.confusion_matrix, metrics.group_confusion_matrix),
                              (skm.zero_one_loss, metrics.group_zero_one_loss)]

# The following only work with binary data when called with their default arguments
supported_metrics_weighted_binary = [(skm.precision_score, metrics.group_precision_score),
                                     (skm.recall_score, metrics.group_recall_score),
                                     (skm.roc_auc_score, metrics.group_roc_auc_score),
                                     (skm.mean_squared_error, metrics.group_mean_squared_error)]
supported_metrics_weighted_binary = supported_metrics_weighted_binary + supported_metrics_weighted


metrics_no_sample_weights = [(skm.max_error, metrics.group_max_error),
                             (skm.mean_absolute_error, metrics.group_mean_absolute_error),
                             (skm.mean_squared_log_error, metrics.group_mean_squared_log_error),
                             (skm.median_absolute_error, metrics.group_median_absolute_error)]

def test_keys_and_weights(self, transform_y_a, transform_y_p, transform_gid, transform_s_w):
        a = "ABC"
        b = "DEF"
        c = "GHI"
        z = "something_longer"
        y_a = transform_y_a([0, 1, 1, 1, 0, 1, 1, 1])
        y_p = transform_y_p([0, 1, 1, 1, 1, 0, 0, 1])
        gid = transform_gid([a, z, a, b, b, c, c, c])
        s_w = transform_s_w([1, 1, 1, 5, 5, 7, 7, 7])

        grouped_metric_func = metrics.make_group_metric(mock_func_weight)
        result = grouped_metric_func(y_a, y_p, gid, s_w)
        assert result.overall == 28
        assert len(result.by_group) == 4
        assert result.by_group[a] == 1
        assert result.by_group[b] == 5
        assert result.by_group[c] == 21
        assert result.by_group[z] == 1
        assert result.minimum == 1
        assert result.maximum == 21
        assert result.argmin_set == {a, z}
        assert result.argmax_set == {c}
        assert result.range == 20
        assert result.range_ratio == pytest.approx(1.0/21.0)