NVIDIA GeForce RTX 5070 Ti with CUDA capability sm_120

[harry_harry]

@ptrblck

Your environment is most likely using a PyTorch binary built with an older CUDA toolkit in your environment. E.g. we have seen these types of issues caused by 3rd party apps such as ComfyUI which create their own virtual environment and install the wrong binaries. Check your torch.__path__ to see which installation is used in which env.

i am sorry,I’m a beginner and don’t quite understand.
but i asked grok
he told me check

print(torch._path_)
print(torch.version.cuda)
print(torch.cuda.get_arch_list())

[‘C:\ProgramData\anaconda3\envs\yolov12\Lib\site-packages\torch’]
12.6
[‘sm_50’, ‘sm_60’, ‘sm_61’, ‘sm_70’, ‘sm_75’, ‘sm_80’, ‘sm_86’, ‘sm_90’]

[harry_harry]

i already fixed it,it is because i using vs code python terminal.then i go to use anaconda cli and fixed it,thank you somuch @ptrblck

[ptrblck]

That’s great to hear!