找不到cuda gpu

[lanchongdashygo]

您好，可以请教一下，cuda版本是多少比较合适呢，错误提示如下，在跑demo_r.py时发现cuda不可用，但是可以成功跑通demo.py文件，目前环境是pytorch11.7 CUDA11.3 3090 24GB
Loading checkpoint shards: 0%| | 0/8 [00:01<?, ?it/s]
Traceback (most recent call last):
File "./demo_r.py", line 186, in
main()
File "./demo_r.py", line 111, in main
model = model_class.from_pretrained(args.model_path, device_map = device_map).half()
File "/root/miniconda3/lib/python3.8/site-packages/transformers/modeling_utils.py", line 2478, in from_pretrained
) = cls._load_pretrained_model(
File "/root/miniconda3/lib/python3.8/site-packages/transformers/modeling_utils.py", line 2794, in _load_pretrained_model
new_error_msgs, offload_index, state_dict_index = _load_state_dict_into_meta_model(
File "/root/miniconda3/lib/python3.8/site-packages/transformers/modeling_utils.py", line 663, in _load_state_dict_into_meta_model
set_module_tensor_to_device(model, param_name, param_device, **set_module_kwargs)
File "/root/miniconda3/lib/python3.8/site-packages/accelerate/utils/modeling.py", line 149, in set_module_tensor_to_device
new_value = value.to(device)
File "/root/miniconda3/lib/python3.8/site-packages/torch/cuda/init.py", line 229, in _lazy_init
torch._C._cuda_init()
RuntimeError: No CUDA GPUs are available

[LiuHC0428]

我们目前使用的是cuda11.2 3090 24GB

[BlakcPink]

CUDA_VISIBLE_DEVICES=0 python ./demo.py