RuntimeError: No CUDA GPUs are available

I run the python3 offline_inference.py, then I meet this error.

INFO 05-19 10:39:22 base_llm_engine.py:190] Initializing workers with distributed init method: tcp://172.17.0.4:36106
len(self.workers):[Actor(RayWorker, b387e8cfcb6791f20607568201000000)]
rank:0 and worker:Actor(RayWorker, b387e8cfcb6791f20607568201000000) and local_rank:0
Traceback (most recent call last):
  File "/root/autodl-tmp/vattention/sarathi-lean/examples/offline_inference.py", line 26, in <module>
    llm_engine = LLMEngine.from_engine_args(
  File "/root/miniconda3/envs/vattn/lib/python3.10/site-packages/sarathi-0.1.7-py3.10-linux-x86_64.egg/sarathi/engine/llm_engine.py", line 17, in from_engine_args
    engine = BaseLLMEngine(*engine_configs)
  File "/root/miniconda3/envs/vattn/lib/python3.10/site-packages/sarathi-0.1.7-py3.10-linux-x86_64.egg/sarathi/engine/base_llm_engine.py", line 111, in __init__
    self._init_workers_ray()
  File "/root/miniconda3/envs/vattn/lib/python3.10/site-packages/sarathi-0.1.7-py3.10-linux-x86_64.egg/sarathi/engine/base_llm_engine.py", line 221, in _init_workers_ray
    self._run_workers(
  File "/root/miniconda3/envs/vattn/lib/python3.10/site-packages/sarathi-0.1.7-py3.10-linux-x86_64.egg/sarathi/engine/base_llm_engine.py", line 440, in _run_workers
    all_outputs = ray.get(all_outputs)
  File "/root/miniconda3/envs/vattn/lib/python3.10/site-packages/ray/_private/auto_init_hook.py", line 18, in auto_init_wrapper
    return fn(*args, **kwargs)
  File "/root/miniconda3/envs/vattn/lib/python3.10/site-packages/ray/_private/client_mode_hook.py", line 103, in wrapper
    return func(*args, **kwargs)
  File "/root/miniconda3/envs/vattn/lib/python3.10/site-packages/ray/_private/worker.py", line 2540, in get
    raise value.as_instanceof_cause()
ray.exceptions.RayTaskError(RuntimeError): ray::RayWorker.execute_method() (pid=25233, ip=172.17.0.4, actor_id=b387e8cfcb6791f20607568201000000, repr=<sarathi.engine.ray_utils.RayWorker object at 0x7fed2d109300>)
  File "/root/miniconda3/envs/vattn/lib/python3.10/site-packages/sarathi-0.1.7-py3.10-linux-x86_64.egg/sarathi/engine/ray_utils.py", line 32, in execute_method
    return executor(*args, **kwargs)
  File "/root/miniconda3/envs/vattn/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 116, in decorate_context
    return func(*args, **kwargs)
  File "/root/miniconda3/envs/vattn/lib/python3.10/site-packages/sarathi-0.1.7-py3.10-linux-x86_64.egg/sarathi/utils/threading_utils.py", line 17, in synced_method
    return method(self, *args, **kwargs)
  File "/root/miniconda3/envs/vattn/lib/python3.10/site-packages/sarathi-0.1.7-py3.10-linux-x86_64.egg/sarathi/worker/base_worker.py", line 99, in init_model
    torch.cuda.set_device(self.device)
  File "/root/miniconda3/envs/vattn/lib/python3.10/site-packages/torch/cuda/__init__.py", line 420, in set_device
    torch._C._cuda_setDevice(device)
  File "/root/miniconda3/envs/vattn/lib/python3.10/site-packages/torch/cuda/__init__.py", line 314, in _lazy_init
    torch._C._cuda_init()
RuntimeError: No CUDA GPUs are available

Start by running python3 offline_inference.py from examples/offline_inference.py and checking the initialization path through BaseLLMEngine._init_workers_ray and base_worker.py:init_model. Confirm whether the failure is reproducible and whether the issue requires a code change or a documented CUDA prerequisite; done means the example starts without the reported RuntimeError.