CUDA error on RTX 5080 (Linux/WSL): no kernel image is available for execution on the device

[suha-glal]

I’m using an RTX 5080 on Linux/WSL, and I’m encountering the following error:

CUDA error: no kernel image is available for execution on the device

I’m not sure how to resolve this issue. Any guidance on what might be causing this or how to fix it would be greatly appreciated.

[yahskapar]

There's probably some kind of mismatch between your host drivers (the ones on Windows, which WSL should be using as far as I know) and PyTorch - I'd investigate that. Inside where you're trying to run the toolbox scripts (e.g., a. conda environment), try:

python3 -c "import torch; print(f'PyTorch: {torch.__version__}'); print(f'PyTorch CUDA: {torch.version.cuda}'); print(f'Is CUDA available: {torch.cuda.is_available()}')"

Perhaps some alternative set of commands would work - basically, try to see if there's any discrepancy in versions there (e.g., older CUDA version than expected, I think in your case with a RTX 5080 at least 12.8 may be required based on my quick online search).

You should be able to update appropriately using something like:

pip install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cu128

[ziyuanLi-alex]

I encountered similar issue when using Mamba related modules, and the same hardware & platform. It was likely caused by the custom Mamba module ( in ./tools/mamba ) lacking support for sm_120 compute ability. To fix this issue, add

    if bare_metal_version >= Version("12.2"):
        cc_flag.append("-gencode")
        cc_flag.append("arch=compute_120,code=sm_120")

in the cc_flag setup logic in setup.py.