I would like to know the CUDA Runtime version that is being used by tensorflow.
I found two cuda installations on my machine (Ubuntu 18.04), and I am not sure which version is currently being used. Weirdly, no environment variable like PATH
, LD_LIBRARY_PATH
or CUDA_HOME
is set, but I am still able to launch trainings on GPUs.
Is there a way to force tensorflow to tell which CUDA version it is using, like a cudaRuntimeGetVersion()
would do?