Hey guys! I currently have a project that is using torch==2.5.1+cu121 and it seems to be working just fine on my rtx3090. I have torch downloaded in a conda env, and have bundled this project into a exe file using pyinstaller. This also seems to be working fine; however when I try running this exe file on a different machine using rtx4080, im getting a “No kernel image is available for execution on this device” error. Was wondering why this was happening
The RTX 4080 reuses the same kernels as the 30XX series. Make sure the same PyTorch build is used and if so, try to narrow down which kernel fails to check if a 3rd party package is causing the issue.