be a bit more conservative and stick with layernorm (without bias) for now, given @borisdayma results https://twitter.com/borisdayma/status/1517227191477571585

dalle2_pytorch/dalle2_pytorch.py

@@ -137,23 +137,27 @@ def sigmoid_beta_schedule(timesteps):
 
 # diffusion prior
 
-class RMSNorm(nn.Module):
+class LayerNorm(nn.Module):
+    def __init__(self, dim):
+        super().__init__()
+        self.gamma = nn.Parameter(torch.ones(dim))
+        self.register_buffer("beta", torch.zeros(dim))
+
+    def forward(self, x):
+        return F.layer_norm(x, x.shape[-1:], self.gamma, self.beta)
+
+
+class ChanLayerNorm(nn.Module):
     def __init__(self, dim, eps = 1e-5):
         super().__init__()
         self.eps = eps
-        self.scale = dim ** 0.5
+        self.g = nn.Parameter(torch.ones(1, dim, 1, 1))
-        self.gamma = nn.Parameter(torch.ones(dim))
 
     def forward(self, x):
-        squared_sum = (x ** 2).sum(dim = -1, keepdim = True)
+        var = torch.var(x, dim = 1, unbiased = False, keepdim = True)
-        inv_norm = torch.rsqrt(squared_sum + self.eps)
+        mean = torch.mean(x, dim = 1, keepdim = True)
-        return x * inv_norm * self.gamma * self.scale
+        return (x - mean) / (var + self.eps).sqrt() * self.g
 
-class ChanRMSNorm(RMSNorm):
-    def forward(self, x):
-        squared_sum = (x ** 2).sum(dim = 1, keepdim = True)
-        inv_norm = torch.rsqrt(squared_sum + self.eps)
-        return x * inv_norm * rearrange(self.gamma, 'c -> 1 c 1 1') * self.scale
 
 class Residual(nn.Module):
     def __init__(self, fn):
@@ -249,10 +253,10 @@ def FeedForward(dim, mult = 4, dropout = 0., post_activation_norm = False):
 
     inner_dim = int(mult * dim)
     return nn.Sequential(
-        RMSNorm(dim),
+        LayerNorm(dim),
         nn.Linear(dim, inner_dim * 2, bias = False),
         SwiGLU(),
-        RMSNorm(inner_dim) if post_activation_norm else nn.Identity(),
+        LayerNorm(inner_dim) if post_activation_norm else nn.Identity(),
         nn.Dropout(dropout),
         nn.Linear(inner_dim, dim, bias = False)
     )
@@ -275,7 +279,8 @@ class Attention(nn.Module):
         inner_dim = dim_head * heads
 
         self.causal = causal
-        self.norm = RMSNorm(dim)
+        self.norm = LayerNorm(dim)
+        self.post_norm = LayerNorm(dim)     # sandwich norm from Coqview paper + Normformer
         self.dropout = nn.Dropout(dropout)
 
         self.null_kv = nn.Parameter(torch.randn(2, dim_head))
@@ -331,7 +336,8 @@ class Attention(nn.Module):
         out = einsum('b h i j, b j d -> b h i d', attn, v)
 
         out = rearrange(out, 'b h n d -> b n (h d)')
-        return self.to_out(out)
+        out = self.to_out(out)
+        return self.post_norm(out)
 
 class CausalTransformer(nn.Module):
     def __init__(
@@ -356,7 +362,7 @@ class CausalTransformer(nn.Module):
                 FeedForward(dim = dim, mult = ff_mult, dropout = ff_dropout)
             ]))
 
-        self.norm = RMSNorm(dim) if norm_out else nn.Identity()  # unclear in paper whether they projected after the classic layer norm for the final denoised image embedding, or just had the transformer output it directly: plan on offering both options
+        self.norm = LayerNorm(dim) if norm_out else nn.Identity()  # unclear in paper whether they projected after the classic layer norm for the final denoised image embedding, or just had the transformer output it directly: plan on offering both options
 
     def forward(
         self,
@@ -720,7 +726,7 @@ class ConvNextBlock(nn.Module):
 
         inner_dim = int(dim_out * mult)
         self.net = nn.Sequential(
-            ChanRMSNorm(dim) if norm else nn.Identity(),
+            ChanLayerNorm(dim) if norm else nn.Identity(),
             nn.Conv2d(dim, inner_dim, 3, padding = 1),
             nn.GELU(),
             nn.Conv2d(inner_dim, dim_out, 3, padding = 1)
@@ -756,8 +762,8 @@ class CrossAttention(nn.Module):
 
         context_dim = default(context_dim, dim)
 
-        self.norm = RMSNorm(dim)
+        self.norm = LayerNorm(dim)
-        self.norm_context = RMSNorm(context_dim)
+        self.norm_context = LayerNorm(context_dim)
         self.dropout = nn.Dropout(dropout)
 
         self.null_kv = nn.Parameter(torch.randn(2, dim_head))

setup.py

@@ -10,7 +10,7 @@ setup(
       'dream = dalle2_pytorch.cli:dream'
     ],
   },
-  version = '0.0.33',
+  version = '0.0.34',
   license='MIT',
   description = 'DALL-E 2',
   author = 'Phil Wang',