Pytorch can't run on 5060 (original) (raw)

1. nvidia 5060， lowest driver is 576.52, latest is 577.00, cuda 12.9 (can’t not down to 12.8 cause has no the driver)

PS C:\Users\Administrator> nvidia-smi.exe
Wed Jul 30 12:11:18 2025
±----------------------------------------------------------------------------------------+
| NVIDIA-SMI 576.52 Driver Version: 576.52 CUDA Version: 12.9 |

2. latest torch version is 2.7.1 cuda 12.8

import torch
torch.version
‘2.7.1+cu128’

run failed and get the error：

RuntimeError: CUDA error: device kernel image is invalid

CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.

For debugging consider passing CUDA_LAUNCH_BLOCKING=1

Compile with TORCH_USE_CUDA_DSA to enable device-side assertions.

Hi @qizhao_2001 ,

  1. Update your NVIDIA driver to the latest version.
  2. Ensure that your CUDA version is compatible with your application or library.
  3. Pass CUDA_LAUNCH_BLOCKING=1 as an environment variable to enable device-side assertions.
  4. Verify that your PyTorch version is compatible with your CUDA version.
  5. If none of the above steps resolve the issue, please provide more details about your application, code, and environment to help further diagnose the problem.