Method __no_running_warning

lmdeploy/pytorch/engine/engine_loop.py:415–421

()

        next_running = None

        async def __no_running_warning():
            # TODO (JimyMa): add watermark check event instead of async sleep.
            # self.perfill_watermark_event.wait()
            logger.warning(f'no next prefill running request, Maybe cache is full, '
                           f'free gpu cache blocks: {scheduler.block_manager.get_num_free_gpu_blocks()}, '
                           f'total gpu cache blocks: {scheduler.block_manager.num_gpu_blocks}')
            await asyncio.sleep(0.1)

        while not self.stop_event.is_set():
            if self._sleep_requested: