How to solve "Could not load dynamic library 'cudart64_~.dll'"

With Tensorflow 2.1, tensorflow pip package contains both CPU and GPU versions of TF.

If the module can not search for the correct CUDA version, it emits the warning and falls back to CPU-only mode.

(cf. W in the beginning stands for warnings, errors have an E, and fatal errors have an F)

 

Such warnings:

Could not load dynamic library 'cudart64_100.dll'; dlerror: cudart64_100.dll not found
or
Could not load dynamic library 'cudart64_101.dll'; dlerror: cudart64_101.dll not found
or
Could not load dynamic library 'cudart64_102.dll'; dlerror: cudart64_102.dll not found
Ignore above cudart dlerror if you do not have a GPU set up on your machine.

** Solution **

 

1) Add Directories to PATH Variable

   C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\<CUDA Version>\bin\cudart64_<CUDA Version>.dll

   ex) C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1\bin\cudart64_101.dll 

   cf) https://helpdeskgeek.com/windows-10/add-windows-path-environment-variable/

 

2) Reboot!!! (After rebooting, it works)