Cuda 12.0 pytorch

I have ubuntu server containing python 3.10.3
Cuda 12.0
Nvidia driver 525.105.17
Torch 2.2.1+cu121
Torch.version.cuda 12.1
Still getting runtime error: cuda error : opration not supported
If anyone know solution kindly share

Could you post a minimal and executable code snippet reproducing the issue, please?

[quote=“yashika1, post:1, topic:198803”]
Thank you for respone,i am really stuck,I am running a detectron2 script to get detection and there i m getting the error

A screenshot unfortunately doesn’t allow us to copy/paste the code and execute it, so let me know if you can post the minimal and executable code snippet.

Import torch
device = torch.device(“cuda”)

This points to a broken installation and you might want to reinstall PyTorch, the NVIDIA driver, etc.

Did reinstalled pytorch,but any other way than reinstalling nvidia driver

Are you sure you installed PyTorch compiled to use CUDA? Can you show us the output of “torch.cuda.is_available()” command?