解决tensorflow2.0提示cudnn版本不对 Failed to get convolution algorithm

tensorflow2.0报错：

Could not create cudnn handle: CUDNN_STATUS_NOT_INITIALIZED
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. [Op:Conv2D]
意思是 cudnn 版本过低，提示升级 cudnn 版本。但是我尝试了三台服务器十几个 cudnn 版本都不能用。最后是在某个集群上调试的时候发现无论使用哪个 cudnn 版本，报的错都是：

Loaded runtime CuDNN library: 7.5.1 but source was compiled with: 7.6.0.  CuDNN library major and minor version needs to match or have higher minor version in case of CuDNN 7.0 or later version. If using a binary install, upgrade your CuDNN library.  If building from sources, make sure the library loaded at runtime is compatible with the version specified during compile configuration.

也就是使用了 7.5.1 版本ÿ