Type-fix: make execute_model output type optional (#12020)

vllm/v1/executor/uniproc_executor.py

@@ -70,6 +70,7 @@ class UniprocExecutor(Executor):
         scheduler_output,
     ) -> ModelRunnerOutput:
         output = self.worker.execute_model(scheduler_output)
+        assert output is not None
         return output
 
     def profile(self, is_start: bool = True):

vllm/v1/worker/gpu_worker.py

@@ -200,7 +200,7 @@ class Worker:
     def execute_model(
         self,
         scheduler_output: "SchedulerOutput",
-    ) -> ModelRunnerOutput:
+    ) -> Optional[ModelRunnerOutput]:
         output = self.model_runner.execute_model(scheduler_output)
         return output if self.rank == 0 else None