[DL]運行Keras時遇到CUDNN_STATUS_INTERNAL_ERROR問題

問題描述

最近在執行Keras程式碼時，遇到了以下的問題：

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_3/convolution}}]]
     [[loss_1/mul/_201]]
  (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_3/convolution}}]]
0 successful operations.
0 derived errors ignored.

環境描述

當前的tensorflow版本: 1.14.0
當前的keras版本: 2.2.4
當前的cuda版本: cuda 10

解決辦法

一開始以為是cuda版本或是gpu驅動不相容的問題，不過後來研究了一下找到了solution，不用重裝cuda和驅動就可以解決，解法是參照github的issue加入這段code：

from tensorflow.compat.v1 import ConfigProto
from tensorflow.compat.v1 import InteractiveSession
 
config = ConfigProto()
config.gpu_options.allow_growth = True
session = InteractiveSession(config=config)

之後就可以正常執行Keras。