diff --git a/vllm/v1/worker/ubatching.py b/vllm/v1/worker/ubatching.py index 7449f196636f0..2dc84dc579092 100644 --- a/vllm/v1/worker/ubatching.py +++ b/vllm/v1/worker/ubatching.py @@ -64,9 +64,9 @@ class UBatchContext: self.cpu_signal_event.set() def _wait(self): - self.stream.wait_event(self.gpu_wait_event) self.cpu_wait_event.wait() self.cpu_wait_event.clear() + self.stream.wait_event(self.gpu_wait_event) self._restore_context() _CURRENT_CONTEXT: dict = {}