Keras说我的cuda驱动程序不兼容，但实际上它们是兼容版本

Question

我运行了一个简单的keras脚本，该脚本在MNIST数据库上训练了一个转换网络。此脚本可以在笔记本电脑上运行，但不能在使用GeForce RTX 2070显卡的PC上运行。

错误是这样：

  File "/home/squall/spencer/kaggle/understanding_cloud_organization/mnist_model.py", line 67, in <module>
    validation_data=(x_test, y_test))
  File "/home/squall/anaconda3/envs/thunder/lib/python3.6/site-packages/keras/engine/training.py", line 1239, in fit
    validation_freq=validation_freq)
  File "/home/squall/anaconda3/envs/thunder/lib/python3.6/site-packages/keras/engine/training_arrays.py", line 196, in fit_loop
    outs = fit_function(ins_batch)
  File "/home/squall/anaconda3/envs/thunder/lib/python3.6/site-packages/tensorflow/python/keras/backend.py", line 3292, in __call__
    run_metadata=self.run_metadata)
  File "/home/squall/anaconda3/envs/thunder/lib/python3.6/site-packages/tensorflow/python/client/session.py", line 1458, in __call__
    run_metadata_ptr)
tensorflow.python.framework.errors_impl.UnknownError: 2 root error(s) found.
  (0) Unknown: 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.
     [[{{node conv2d_1/convolution}}]]
     [[metrics/accuracy/Identity/_91]]
  (1) Unknown: 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.
     [[{{node conv2d_1/convolution}}]]
0 successful operations.
0 derived errors ignored.

Cuda是10.1。司机是418.56。 CuDNN是7.4.2。 Tensorflow是1.14。根据官方的Nvidia图表，这些都是兼容版本。

有什么想法吗？

Answer 1

尝试一下

PS：CUDA为10.0，cuDNN为7.6.3（对于CUDA10.0]