bring in two tricks from the cogview paper for reducing the chances of overflow, for attention and layernorm

dalle2_pytorch/dalle2_pytorch.py

@@ -496,7 +496,6 @@ class LayerNorm(nn.Module):
         self.g = nn.Parameter(torch.ones(dim))
 
     def forward(self, x):
-        x = x / x.amax(dim = -1, keepdim = True).detach()
         var = torch.var(x, dim = -1, unbiased = False, keepdim = True)
         mean = torch.mean(x, dim = -1, keepdim = True)
         return (x - mean) * (var + self.eps).rsqrt() * self.g
@@ -508,7 +507,6 @@ class ChanLayerNorm(nn.Module):
         self.g = nn.Parameter(torch.ones(1, dim, 1, 1))
 
     def forward(self, x):
-        x = x / x.amax(dim = 1, keepdim = True).detach()
         var = torch.var(x, dim = 1, unbiased = False, keepdim = True)
         mean = torch.mean(x, dim = 1, keepdim = True)
         return (x - mean) * (var + self.eps).rsqrt() * self.g

dalle2_pytorch/version.py

@@ -1 +1 @@
-__version__ = '0.16.2'
+__version__ = '0.16.1'