Make loading of pretrained gpt2 faster by avoiding initialization of Conv1D's weights (#21879)

apply normal_ after assigning weight as nn.Parameter to avoid unnecessary initialization computation

src/transformers/pytorch_utils.py

@@ -105,10 +105,9 @@ class Conv1D(nn.Module):
     def __init__(self, nf, nx):
         super().__init__()
         self.nf = nf
-        w = torch.empty(nx, nf)
-        nn.init.normal_(w, std=0.02)
-        self.weight = nn.Parameter(w)
+        self.weight = nn.Parameter(torch.empty(nx, nf))
         self.bias = nn.Parameter(torch.zeros(nf))
+        nn.init.normal_(self.weight, std=0.02)
 
     def forward(self, x):
         size_out = x.size()[:-1] + (self.nf,)