When you install Pytorch via pip, I’m aware that it downloads a precompiled CUDA runtime that works with that version of Torch.
However, I cannot seem to find any information on how this works, or where the downloaded runtimes are placed. In a related issue, my version of Torch was clearly built with CUDA, yet doesn’t seem to be able to find any cuda runtimes - how is this possible?
Collecting environment information... PyTorch version: 1.5.1 Is debug build: No CUDA used to build PyTorch: 10.2 OS: Ubuntu 16.04.6 LTS GCC version: Could not collect CMake version: Could not collect Python version: 3.6 Is CUDA available: No CUDA runtime version: No CUDA GPU models and configuration: No CUDA Nvidia driver version: No CUDA cuDNN version: No CUDA Versions of relevant libraries: [pip3] numpy==1.19.0 [pip3] torch==1.5.1 [pip3] torchvision==0.2.0
Any information on how pip + pytorch manage the cuda runtime installation, and where this runtime is stored, would be helpful.