[torch.compile] fix deprecated code (#10948)

Signed-off-by: youkaichao <[email protected]>

vllm/compilation/backends.py

@@ -40,7 +40,7 @@ def wrap_inductor(graph,
     compilation_counter.num_inductor_compilations += 1
 
     from torch._inductor import config
-    current_config = config.shallow_copy_dict()
+    current_config = config.get_config_copy()
     from torch._inductor.compile_fx import compile_fx
 
     if additional_inductor_config is not None: