Cuda kernel image error

i have torch ‘1.11.0+cu102’ with cuda version 10.2
and i have NVIDIA GeForce RTX 3060 Ti
nvidia driver is 470.74

here is the Nvidia-Smi result
NVIDIA-SMI 470.74 Driver Version: 470.74 CUDA Version: 11.4
GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
MIG M. |
|===============================+======================+===== |
| 0 NVIDIA GeForce … Off | 00000000:0B:00.0 Off | N/A |
| 0% 48C P0 1W / 240W | 0MiB / 7982MiB | 0% Default |
| | | N/A |

and i am trung to run the training of pix2pix model,
and i get this error message:

alternatively i have the same envirement at collab or Kaggle and it works verry well,

how can i make it work on my card please?
PS: im working on vertual machine connected using ssh so i don’t have access to the UI of systeme,
i have centos as OS.

Can you trying installing torch with cuda 11 and see if it solves your issue? If I remember correctly from one of @ptrblck’s answers, this could be the solution for the issue you are having.

1 Like

@ptrblck @InnovArul thnx it workt now