diff --git a/predtuner/torchutil/common_qos.py b/predtuner/torchutil/common_qos.py
index f7c9ccdbcd2e9917c597e42ec4aae0181086f004..a6a19f6625453bcd381cdd392f1c8a0206c206f8 100644
--- a/predtuner/torchutil/common_qos.py
+++ b/predtuner/torchutil/common_qos.py
@@ -4,4 +4,4 @@ from torch import Tensor
 def accuracy(output: Tensor, target: Tensor) -> float:
     _, pred_labels = output.max(1)
     n_correct = (pred_labels == target).sum().item()
-    return n_correct / len(output)
+    return n_correct / len(output) * 100
diff --git a/test/test_torchapp.py b/test/test_torchapp.py
index 0184093fcad50987a7f1b252f4816fcbb081ff89..1621a08c094bdf27c931418bdf5f519e3984d849 100644
--- a/test/test_torchapp.py
+++ b/test/test_torchapp.py
@@ -44,7 +44,7 @@ class TestTorchApp(unittest.TestCase):
     def test_baseline_qos(self):
         app = self.get_app()
         qos, _ = app.measure_qos_perf({}, False)
-        self.assertAlmostEqual(qos, 0.88)
+        self.assertAlmostEqual(qos, 88.0)
 
     def test_tuning(self):
         app = TorchApp(