change up epsilon in layernorm the case of using fp16, thanks to @Veldrovive for figuring out this stabilizes training

dalle2_pytorch/dalle2_pytorch.py

@@ -580,7 +580,7 @@ class ChanLayerNorm(nn.Module):
 
         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
+        return (x - mean) * (var + eps).rsqrt() * self.g
 
 class Residual(nn.Module):
     def __init__(self, fn):

dalle2_pytorch/version.py

@@ -1 +1 @@
-__version__ = '1.4.1'
+__version__ = '1.4.2'