diff --git a/denoising_diffusion_pytorch/denoising_diffusion_pytorch.py b/denoising_diffusion_pytorch/denoising_diffusion_pytorch.py
index aabe5b8..27d1774 100644
--- a/denoising_diffusion_pytorch/denoising_diffusion_pytorch.py
+++ b/denoising_diffusion_pytorch/denoising_diffusion_pytorch.py
@@ -61,9 +61,6 @@ def convert_image_to_fn(img_type, image):
         return image.convert(img_type)
     return image
 
-def l2norm(t):
-    return F.normalize(t, dim = -1)
-
 # normalization functions
 
 def normalize_to_neg_one_to_one(img):
@@ -239,9 +236,10 @@ class LinearAttention(nn.Module):
 class Attention(nn.Module):
     def __init__(self, dim, heads = 4, dim_head = 32, scale = 10):
         super().__init__()
-        self.scale = scale
+        self.scale = dim_head ** -0.5
         self.heads = heads
         hidden_dim = dim_head * heads
+
         self.to_qkv = nn.Conv2d(dim, hidden_dim * 3, 1, bias = False)
         self.to_out = nn.Conv2d(hidden_dim, dim, 1)
 
@@ -250,11 +248,12 @@ class Attention(nn.Module):
         qkv = self.to_qkv(x).chunk(3, dim = 1)
         q, k, v = map(lambda t: rearrange(t, 'b (h c) x y -> b h c (x y)', h = self.heads), qkv)
 
-        q, k = map(l2norm, (q, k))
+        q = q * self.scale
 
-        sim = einsum('b h d i, b h d j -> b h i j', q, k) * self.scale
+        sim = einsum('b h d i, b h d j -> b h i j', q, k)
         attn = sim.softmax(dim = -1)
         out = einsum('b h i j, b h d j -> b h i d', attn, v)
+
         out = rearrange(out, 'b h (x y) d -> b (h d) x y', x = h, y = w)
         return self.to_out(out)
 
diff --git a/setup.py b/setup.py
index 18f1e14..780cbf9 100644
--- a/setup.py
+++ b/setup.py
@@ -3,7 +3,7 @@ from setuptools import setup, find_packages
 setup(
   name = 'denoising-diffusion-pytorch',
   packages = find_packages(),
-  version = '0.27.12',
+  version = '0.28.0',
   license='MIT',
   description = 'Denoising Diffusion Probabilistic Models - Pytorch',
   author = 'Phil Wang',