# Stub for non-CUDA systems
def aimdo_to_tensor(v, device): return None