AcceleratorError: CUDA error: no kernel image is available for execution on the device
Search for `cudaErrorNoKernelImageForDevice’ in CUDA Runtime API :: CUDA Toolkit Documentation for more information.
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.
Install any of our binaries built with CUDA 12.8 as explained in other threads already for Blackwell support. E.g. on Linux x86 just run pip install torch and it will install the latest stable 2.10.0 release with CUDA 12.8 which supports your Blackwell GPU.
How about in windows?
The same requirements apply: install any of our binaries built with CUDA 12.8+.
From the install matrix just select the release you want:
Hey not working pls do something we need to train our models for our thesis
you might need to provide a bit more details than “not working” mate.
Are you certain you installed it with CUDA 12.8 as @ptrblck suggested? I had a similar issue on your hardware running 12.6
Maybe post the output of nvcc –version in here
