2019-02-21 18:36:53.481823: E C:\tf_jenkins\workspace\rel-win\M\windows-gpu\PY\3 6\tensorflow\stream_executor\cuda\cuda_dnn.cc:385] could not create cudnn handle : CUDNN_STATUS_NOT_INITIALIZED

UnknownError (see above for traceback): Failed to get convolution algorithm. This is probably because cuDNN failed to initialize, so try looking to see if a war
ning log message was printed above.

Could not open codec 'libopenh264': Unspecified error

2019-02-21 18:36:53.481823: E C:\tf_jenkins\workspace\rel-win\M\windows-gpu\PY\3
6\tensorflow\stream_executor\cuda\cuda_dnn.cc:385] could not create cudnn handle
: CUDNN_STATUS_NOT_INITIALIZED

 

2019-02-21 18:36:53.481823: E C:\tf_jenkins\workspace\rel-win\M\windows-gpu\PY\3
6\tensorflow\stream_executor\cuda\cuda_dnn.cc:385] could not create cudnn handle
: CUDNN_STATUS_NOT_INITIALIZED

 Tensorflow版本过高。安装的1.12.0，换成1.6.0，解决