change call_delegate_autograd

Summary: Some changes angela told me to make 😂

Differential Revision: D56222503

Author: mcr229

exir/delegate.py

@@ -102,7 +102,7 @@ def fake_requires_grad(var):
                         var.requires_grad = True
                 return var
 
-            return pytree.tree_map(fake_requires_grad, res)
+            return pytree.tree_map_only(torch.Tensor, fake_requires_grad, res)
 
         return res