After tensorflow 2.0.0 is installed, the following error occurs during the operation:

tensorflow.python.framework.errors_impl.InternalError: cudaGetDevice() failed. Status: CUDA driver version is insufficient for CUDA runtime version

This is because the CUDA driver version does not meet the CUDA running version. Previously, in order to install tensorflow 2.0.0, the CUDA running version was upgraded to CUDA 10.0, but the CUDA driver version was not upgraded, CUDA driver version and CUDA running version should meet the requirements of the following table（ https://docs.nvidia.com/cuda/cuda-toolkit-release-notes/index.html ):

After running NVIDIA SMI, the CUDA driver version is 385.54, which does not meet the above requirements

So we directly use the device manager of windows 8.1 to upgrade CUDA driver version with the driver software of automatic search update

After the upgrade, run NVIDIA SMI to show that the CUDA driver version has been upgraded to 416.34. At this time, if you run tensorflow 2.0.0, the above error will not appear