Fix vulkan delegate partial test

Summary:
## Context

Previously the `test_vulkan_backend_partial` was written erroneously as a duplicate constant reference was present in the graph. This was not caught previously, but pytorch/pytorch#120664 added checks that now catch the error. This is a simple change to correct the badly written test.

Reviewed By: manuelcandales

Differential Revision: D54263601

fbshipit-source-id: fc915b0c918d0a158e2efa08c5e11cd3d5d77e57

Author: SS-JIA

backends/vulkan/test/test_vulkan_delegate.py

@@ -241,12 +241,8 @@ class SimpleModel(torch.nn.Module):
             def __init__(self):
                 super().__init__()
                 self.linear = torch.nn.Linear(10, 10)
-                self.offset_1 = self.weight = torch.rand(
-                    size=(2, 10), dtype=torch.float32
-                )
-                self.offset_2 = self.weight = torch.rand(
-                    size=(2, 10), dtype=torch.float32
-                )
+                self.offset_1 = torch.rand(size=(2, 10), dtype=torch.float32)
+                self.offset_2 = torch.rand(size=(2, 10), dtype=torch.float32)
 
             def forward(self, x):
                 return self.linear(x + self.offset_1) - self.offset_2