matlab - 通过 MATLAB 调用 Tensorflow：CUDNN_STATUS_NOT_INITIALIZED。检索驱动程序版本时出错

Question

我有一个用 Tensorflow 实现的 CNN，它在 PyCharm IDE 和通过 Windows 调用它时都能完美运行cmd。现在我正在尝试system通过调用 python 脚本通过 MATLAB 命令运行网络。我收到以下错误：

...
2019-02-05 12:37:45.681913: E tensorflow/stream_executor/cuda/cuda_dnn.cc:373] Could not create cudnn handle: CUDNN_STATUS_NOT_INITIALIZED
2019-02-05 12:37:45.682038: E tensorflow/stream_executor/cuda/cuda_dnn.cc:377] Error retrieving driver version: Unimplemented: kernel reported driver version not implemented on Windows
2019-02-05 12:37:45.682192: E tensorflow/stream_executor/cuda/cuda_dnn.cc:373] Could not create cudnn handle: CUDNN_STATUS_NOT_INITIALIZED
2019-02-05 12:37:45.682306: E tensorflow/stream_executor/cuda/cuda_dnn.cc:377] Error retrieving driver version: Unimplemented: kernel reported driver version not implemented on Windows
...
tensorflow.python.framework.errors_impl.UnknownError: Failed to get convolution algorithm. This is probably because cuDNN failed to initialize, so try looking to see if a warning log message was printed above.

cmd这太奇怪了，因为我在 Windows或 Pycharm IDE中启动了完全相同的命令并且运行良好，但使用 MATLABsystem命令失败。

有什么想法？这可能与 MATLAB 中的 CUDA/CUDNN 冲突有关吗？

非常感谢

哈维尔。

标签： matlabtensorflow