I have successfully installed the CUDA-Q environment on my university’s AI server, and CUDA-Q can be imported normally. However, when I try to run any CUDA-Q kernel (for example calling cudaq.set_target('tensornet') or running solvers.vqe()), the Jupyter kernel immediately crashes without showing any Python errors.
My GPU environment:
-
Tesla V100 32GB ×2
-
Driver Version: 570.133.07
-
CUDA Runtime Version: 12.8
nvidia-smi shows the GPUs are available, but CUDA-Q execution still causes the kernel to die instantly with this message:
[error] Disposing session as kernel process died. ExitCode: undefined.
I would like to ask what might cause this crash.
Is it due to driver/CUDA incompatibility, missing libraries, or V100 support issues?
Any guidance on how to debug or obtain detailed logs would be greatly appreciated.