Anaconda 中的 Tensorflow-Gpu 和 Cuda 驱动程序存在问题

Question

I have the following problem with Tensorflow-GPU. While trying to setup the gpus (in Jupyter) for a deep learning task, I get the following error:

InternalError                             Traceback (most recent call last)
<ipython-input-3-a08c39e19f9e> in <module>
     20     for gpu in gpus:
     21       tf.config.experimental.set_memory_growth(gpu, True)
---> 22     logical_gpus = tf.config.experimental.list_logical_devices('GPU')
     23     print(len(gpus), "Physical GPUs,", len(logical_gpus), "Logical GPUs")
     24   except RuntimeError as e:

~\Anaconda3\envs\tens-gpu\lib\site-packages\tensorflow_core\python\framework\config.py in list_logical_devices(device_type)
    345     List of LogicalDevice objects
    346   """
--> 347   return context.context().list_logical_devices(device_type=device_type)
    348 
    349 

~\Anaconda3\envs\tens-gpu\lib\site-packages\tensorflow_core\python\eager\context.py in list_logical_devices(self, device_type)
   1150   def list_logical_devices(self, device_type=None):
   1151     """Return logical devices."""
-> 1152     self.ensure_initialized()
   1153 
   1154     devices = []

~\Anaconda3\envs\tens-gpu\lib\site-packages\tensorflow_core\python\eager\context.py in ensure_initialized(self)
    490         if self._default_is_async == ASYNC:
    491           pywrap_tensorflow.TFE_ContextOptionsSetAsync(opts, True)
--> 492         self._context_handle = pywrap_tensorflow.TFE_NewContext(opts)
    493       finally:
    494         pywrap_tensorflow.TFE_DeleteContextOptions(opts)

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

My Cuda version is Cuda 10.0.130 and cudnn is 7.6.4, while I have an Intel UHD Graphics 620 and a NVIDIA GeForce MX150 gpus. The drivers are updated to the 'optimised' release and I really don't get what is not working. Could you please help me, as I've already tried everything I found online? Thank you in advance

Answer 1

The reason for this error is the mismatch of your installed CUDA Toolkit version and the version of the python package CUDA toolkit, which is usually installed as dependency of Tensorflow GPU

Running a CUDA application requires the system with at least one CUDA capable GPU and a driver that is compatible with the CUDA Toolkit. Please refer details in below

Could not load dynamic library 'cudart64_100.dll'; dlerror: cudart64_100.dll not found

The simplest way to fix is to install the latest NVIDIA GPU Computing Toolkit , because if it's not there, you will be missing the 'cudart64_100.dll' library.