Hello,
I’m having difficulty understanding the CUDA installation process, particularly with the various versions displayed by different commands. In my conda environment, running the command conda list cudatoolkit
shows version 11.8.0. However, the command nvidia-smi
indicates that the CUDA version is 12.3, and nvcc --version
shows 12.2. Moreover, when I execute print(torch.version.cuda)
in Python, it displays 12.1. Could you please explain the differences between these versions? I realize this might be a basic question, but any clarification would be greatly appreciated!