diff --git a/vllm/compilation/backends.py b/vllm/compilation/backends.py
index f5fff344a1f4..c3c670422def 100644
--- a/vllm/compilation/backends.py
+++ b/vllm/compilation/backends.py
@@ -329,7 +329,8 @@ class PiecewiseCompileInterpreter(torch.fx.Interpreter):
             self.fake_mode.from_tensor(t) if isinstance(t, torch.Tensor) else t
             for t in args
         ]
-        return super().run(*fake_args)
+        with self.fake_mode:
+            return super().run(*fake_args)
 
     def call_module(self, target: torch.fx.node.Target,
                     args: Tuple[torch.fx.node.Argument,