diff --git a/tests/test_tf.py b/tests/test_tf.py
index 075093d..b24757a 100644
--- a/tests/test_tf.py
+++ b/tests/test_tf.py
@@ -22,7 +22,6 @@ def test_percentile():
 
     np.testing.assert_almost_equal(p_np, p_t)
 
-
 def test_normalize_tf():
     size = 1024
     curr_file_path = os.path.dirname(os.path.realpath(__file__))
diff --git a/tests/test_torch.py b/tests/test_torch.py
index 0d79ed1..168d2f1 100644
--- a/tests/test_torch.py
+++ b/tests/test_torch.py
@@ -8,7 +8,6 @@
 from torchvision import transforms
 from skimage.metrics import structural_similarity as ssim
 
-
 def test_cov():
     x = np.random.randn(10, 10)
     cov_np = np.cov(x)
diff --git a/torchstain/tf/utils/percentile.py b/torchstain/tf/utils/percentile.py
index 2db4aeb..de8c6ac 100644
--- a/torchstain/tf/utils/percentile.py
+++ b/torchstain/tf/utils/percentile.py
@@ -14,5 +14,5 @@ def percentile(t: tf.Tensor, q: float) -> Union[int, float]:
     :param q: Percentile to compute, which must be between 0 and 100 inclusive.
     :return: Resulting value (scalar).
     """
-    k = 1 + tf.math.round(.01 * tf.cast(q, tf.float32) * (tf.cast(tf.size(t), tf.float32) - 1))
-    return tf.sort(tf.reshape(t, [-1]))[tf.cast(k, tf.int32)]
+    k = 1 + tf.math.round(.01 * tf.cast(q, tf.float32) * (tf.cast(tf.math.reduce_prod(tf.size(t)), tf.float32) - 1))
+    return tf.sort(tf.reshape(t, [-1]))[tf.cast(k - 1, tf.int32)]
diff --git a/torchstain/torch/utils/percentile.py b/torchstain/torch/utils/percentile.py
index 188f1bb..08273d4 100644
--- a/torchstain/torch/utils/percentile.py
+++ b/torchstain/torch/utils/percentile.py
@@ -6,7 +6,7 @@
 """
 def percentile(t: torch.tensor, q: float) -> Union[int, float]:
     """
-    Return the ``q``-th percentile of the flattened input tensor's data.
+    Return the ``q``-th percentile of the flattenepip d input tensor's data.
 
     CAUTION:
      * Needs PyTorch >= 1.1.0, as ``torch.kthvalue()`` is used.