How to execute pytorch 1.0 code on cuda 11


I am running a code that was developed by Python 3.6.8 and PyTorch 1.0 and CUDA 9.2. Here is the link to the code: GitHub - kensakurada/sscdnet: Semantic Scene Change Detection Network (CSCDNet + SSCDNet)

But my current workstation has the Nividia driver which is NVIDIA-SMI 460.106.00 Driver Version: 460.106.00 CUDA Version: 11.2. Just wonder how to execute this code in this environment? And I could not downgrade the Cuda version for this machine as I share this machine with other users. Any help will be appreciated.

Your NVIDIA driver should be backwards-compatible. You also won’t be able to build such an old PyTorch release with CUDA 11.x, so your best bet is to either update PyTorch or to use the 1.0 release with its original CUDA 9 runtime.