diff --git a/vllm/executor/ray_hpu_executor.py b/vllm/executor/ray_hpu_executor.py index ebfaafd29f92c..7e2085673323f 100644 --- a/vllm/executor/ray_hpu_executor.py +++ b/vllm/executor/ray_hpu_executor.py @@ -75,7 +75,7 @@ def shutdown(self) -> None: self.forward_dag.teardown() import ray for worker in self.workers: - ray.kill(worker) + worker.__ray_terminate__.remote() self.forward_dag = None def finish_measurements(self):