cudaGetDevice() 导入张量流失败错误

Question

我安装了 Tensorflow gpu 以及它需要的所有 dll。但是当我尝试导入 keras 时，我得到了这个错误：

cudaGetDevice() failed. Status: CUDA driver version is insufficient for CUDA runtime version

这是完整的错误：

InternalError                             Traceback (most recent call last)
<ipython-input-1-88d96843a926> in <module>()
----> 1 import keras

C:\ProgramData\Anaconda3\envs\tensorflow\lib\site-packages\keras\__init__.py in <module>()
      1 from __future__ import absolute_import
      2 
----> 3 from . import utils
      4 from . import activations
      5 from . import applications

C:\ProgramData\Anaconda3\envs\tensorflow\lib\site-packages\keras\utils\__init__.py in <module>()
      4 from . import data_utils
      5 from . import io_utils
----> 6 from . import conv_utils
      7 
      8 # Globally-importable utils.

C:\ProgramData\Anaconda3\envs\tensorflow\lib\site-packages\keras\utils\conv_utils.py in <module>()
      1 from six.moves import range
      2 import numpy as np
----> 3 from .. import backend as K
      4 
      5 

C:\ProgramData\Anaconda3\envs\tensorflow\lib\site-packages\keras\backend\__init__.py in <module>()
     81 elif _BACKEND == 'tensorflow':
     82     sys.stderr.write('Using TensorFlow backend.\n')
---> 83     from .tensorflow_backend import *
     84 else:
     85     raise ValueError('Unknown backend: ' + str(_BACKEND))

C:\ProgramData\Anaconda3\envs\tensorflow\lib\site-packages\keras\backend\tensorflow_backend.py in <module>()
     48 # This list queries the devices.
     49 # We assume our devices don't change during our lifetime.
---> 50 _LOCAL_DEVICES = device_lib.list_local_devices()
     51 
     52 

C:\ProgramData\Anaconda3\envs\tensorflow\lib\site-packages\tensorflow\python\client\device_lib.py in list_local_devices()
     34     return m
     35 
---> 36   return [_convert(s) for s in pywrap_tensorflow.list_devices()]

C:\ProgramData\Anaconda3\envs\tensorflow\lib\site-packages\tensorflow\python\pywrap_tensorflow_internal.py in list_devices()
   1322 
   1323   with errors.raise_exception_on_not_ok_status() as status:
-> 1324     return ListDevices(status)
   1325 
   1326 

C:\ProgramData\Anaconda3\envs\tensorflow\lib\site-packages\tensorflow\python\framework\errors_impl.py in __exit__(self, type_arg, value_arg, traceback_arg)
    471             None, None,
    472             compat.as_text(c_api.TF_Message(self.status.status)),
--> 473             c_api.TF_GetCode(self.status.status))
    474     # Delete the underlying status object from memory otherwise it stays alive
    475     # as there is a reference to status from this from the traceback due to

InternalError: cudaGetDevice() failed. Status: CUDA driver version is insufficient for CUDA runtime version

谢谢

Answer 1

当我们使用conda 或pip 安装tensorflow-gpu 时，除非明确指定，否则会安装最新版本的tensorflow。大多数时候，我们的cuda 驱动程序没有更新为与已安装的tensorflow 版本兼容。 cuda驱动可以通过以下命令更新，更新后与TensorFlow兼容。

sudo apt-get purge nvidia-*

sudo apt-get install nvidia-375