[Python][CI] test-cuda-python fails, requires cuda-toolkit v11.2

[danepitkin]

Describe the bug, including details regarding any error messages, version, and platform.

test-cuda-python fails with

_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
usr/local/lib/python3.8/dist-packages/numba/cuda/dispatcher.py:539: in __call__
    return self.dispatcher.call(args, self.griddim, self.blockdim,
usr/local/lib/python3.8/dist-packages/numba/cuda/dispatcher.py:673: in call
    kernel = _dispatcher.Dispatcher._cuda_call(self, *args)
usr/local/lib/python3.8/dist-packages/numba/cuda/dispatcher.py:681: in _compile_for_args
    return self.compile(tuple(argtypes))
usr/local/lib/python3.8/dist-packages/numba/cuda/dispatcher.py:924: in compile
    kernel = _Kernel(self.py_func, argtypes, **self.targetoptions)
usr/local/lib/python3.8/dist-packages/numba/core/compiler_lock.py:35: in _acquire_compile_lock
    return func(*args, **kwargs)
usr/local/lib/python3.8/dist-packages/numba/cuda/dispatcher.py:103: in __init__
    self.cooperative = 'cudaCGGetIntrinsicHandle' in lib.get_asm_str()
usr/local/lib/python3.8/dist-packages/numba/cuda/codegen.py:119: in get_asm_str
    return self._join_ptxes(self._get_ptxes(cc=cc))
usr/local/lib/python3.8/dist-packages/numba/cuda/codegen.py:131: in _get_ptxes
    arch = nvvm.get_arch_option(*cc)
usr/local/lib/python3.8/dist-packages/numba/cuda/cudadrv/nvvm.py:439: in get_arch_option
    arch = find_closest_arch((major, minor))
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

mycc = (7, 5)

    def find_closest_arch(mycc):
        """
        Given a compute capability, return the closest compute capability supported
        by the CUDA toolkit.
    
        :param mycc: Compute capability as a tuple ``(MAJOR, MINOR)``
        :return: Closest supported CC as a tuple ``(MAJOR, MINOR)``
        """
        supported_ccs = NVVM().supported_ccs
    
        if not supported_ccs:
            msg = "No supported GPU compute capabilities found. " \
                  "Please check your cudatoolkit version matches your CUDA version."
>           raise NvvmSupportError(msg)
E           numba.cuda.cudadrv.error.NvvmSupportError: No supported GPU compute capabilities found. Please check your cudatoolkit version matches your CUDA version.

usr/local/lib/python3.8/dist-packages/numba/cuda/cudadrv/nvvm.py:412: NvvmSupportError
=============================== warnings summary ===============================
usr/local/lib/python3.8/dist-packages/pyarrow/tests/test_cuda_numba_interop.py: 19 warnings
  /usr/local/lib/python3.8/dist-packages/numba/cuda/cudadrv/nvvm.py:392: UserWarning: CUDA Toolkit 11.0 is unsupported by Numba - 11.2 is the minimum required version.

https://github.com/ursacomputing/crossbow/actions/runs/6319788893/job/17161299196

Component(s)

Python