conda pytorch gpu

I noticed that on my pc, the installation of the GPU version of pytorch (using setup.sh) silently fails. This means that everything works but I am not using my GPU for the experiments.

The output of the last successful CI/CD pipeline also shows Using cpu device. This made me wonder, does the computer on which we run the pipelines have a GPU? Does the installation work on your local computer @czs4581 ?

In order to get pytorch to use the GPU, I had to install it with pip. While officially, it is recommended to use https://pytorch.org/get-started/locally/ to install pytorch, I will try to write a script that first checks for the cuda version and then generates the correct pip install command.