Open Johnmakuta opened 1 year ago
This is very dependent in the way you've been installing the libtorch
library. If you have a working install of PyTorch 2.0.0 on which cuda/cudnn work, I would recommend using it by activating the environment and setting LIBTORCH_USE_PYTORCH=1
(more details on how libtorch can be installed on the main tch readme.
I've been testing out the default test script and every time is says:
Even though I've testing in pytorch and tensorflow that my gpu is available and shown.