diff --git a/setup.py b/setup.py
index 8b7a52bb..6f7d1dd5 100644
--- a/setup.py
+++ b/setup.py
@@ -3,7 +3,7 @@
 setup(
   name = 'x-transformers',
   packages = find_packages(exclude=['examples']),
-  version = '1.37.7',
+  version = '1.37.8',
   license='MIT',
   description = 'X-Transformers - Pytorch',
   author = 'Phil Wang',
diff --git a/x_transformers/attend.py b/x_transformers/attend.py
index 48828a7d..665695c7 100644
--- a/x_transformers/attend.py
+++ b/x_transformers/attend.py
@@ -211,12 +211,12 @@ def flash_attn(
 
         if self.l2_distance:
             k_norm_sq = k.norm(dim = -1, keepdim = True) ** 2
-            k = F.pad(k, (0, 1), value = 1.)
-            k = torch.cat((k, -k_norm_sq), dim = -1)
+            k = F.pad(k, (0, 1), value = -1.)
+            k = torch.cat((k, k_norm_sq), dim = -1)
 
             q_norm_sq = q.norm(dim = -1, keepdim = True) ** 2
-            q = torch.cat((2 * q, -q_norm_sq), dim = -1)
-            q = F.pad(q, (0, 1), value = 1.)
+            q = torch.cat((2 * q, q_norm_sq), dim = -1)
+            q = F.pad(q, (0, 1), value = -1.)
 
         # handle scale - by default they scale by dim_head ** -0.5, but need to take care if using cosine sim attention