RuntimeError: CUDA unknown error - this may be due to an incorrectly set up environment, e.g. changi

最近在训练网络模型时，出现以下错误：

/home/xw/anaconda3/envs/openmmlab/lib/python3.7/site-packages/torch/cuda/__init__.py:52: UserWarning: CUDA initialization: CUDA unknown error - this may be due to an incorrectly set up environment, e.g. changing env variable CUDA_VISIBLE_DEVICES after program start. Setting the available devices to be zero. (Triggered internally at  /opt/conda/conda-bld/pytorch_1623448265233/work/c10/cuda/CUDAFunctions.cpp:115.)
  return torch._C._cuda_getDeviceCount() > 0
No CUDA runtime is found, using CUDA_HOME='/usr/local/cuda-11.1'
Traceback (most recent call last):
  File "configs/trainval/tinaface/test_widerface.py", line 132, in <module>
    main()
  File "configs/trainval/tinaface/test_widerface.py", line 125, in main
    engine, data_loader = prepare(cfg, args.checkpoint,args.device)
  File "configs/trainval/tinaface/test_widerface.py", line 77, in prepare
    device = torch.cuda.current_device()
  File "/home/xw/anaconda3/envs/openmmlab/lib/python3.7/site-packages/torch/cuda/__init__.py", line 432, in current_device
    _lazy_init()
  File "/home/xw/anaconda3/envs/openmmlab/lib/python3.7/site-packages/torch/cuda/__init__.py", line 172, in _lazy_init
    torch._C._cuda_init()
RuntimeError: CUDA unknown error - this may be due to an incorrectly set up environment, e.g. changing env variable CUDA_VISIBLE_DEVICES after program start. Setting the available devices to be zero.

检查显卡驱动nvcc -V

nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2020 NVIDIA Corporation
Built on Tue_Sep_15_19:10:02_PDT_2020
Cuda compilation tools, release 11.1, V11.1.74
Build cuda_11.1.TC455_06.29069683_0

查看显卡：watch nvidia-smi

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 470.103.01   Driver Version: 470.103.01   CUDA Version: 11.4     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  NVIDIA GeForce ...  Off  | 00000000:19:00.0 Off |                  N/A |
|  0%   57C    P8    26W / 370W |     11MiB / 10018MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
|   1  NVIDIA GeForce ...  Off  | 00000000:1A:00.0 Off |                  N/A |
|  0%   57C    P8    25W / 370W |     11MiB / 10018MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
|   2  NVIDIA GeForce ...  Off  | 00000000:67:00.0 Off |                  N/A |
|  0%   84C    P2   275W / 370W |   9381MiB / 10018MiB |     96%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
|   3  NVIDIA GeForce ...  Off  | 00000000:68:00.0 Off |                  N/A |
|  0%   83C    P2   282W / 370W |   7904MiB / 10015MiB |     97%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------

运行：print(torch.cuda.is_available())
报错如下：

Traceback (most recent call last):
              File "<stdin>", line 1, in <module>
              File "/home/kumar/anaconda3/envs/pytorch/lib/python3.8/site-packages/torch/cuda/__init__.py", line 430, in current_device
                _lazy_init()
              File "/home/kumar/anaconda3/envs/pytorch/lib/python3.8/site-packages/torch/cuda/__init__.py", line 170, in _lazy_init
                torch._C._cuda_init()
            RuntimeError: CUDA unknown error - this may be due to an incorrectly set up environment, e.g. changing env variable CUDA_VISIBLE_DEVICES after program start. Setting the available devices to be zero.

解决方案：

apt-get install nvidia-modprobe 

一分钟秒解决