Failing GPU test that isn't run by the gpuCI

[m-albert]

I just ran local tests with cupy installed and got a failing test:

>       out = threshold_local(self.image, block_size, method='gaussian')

tests/test_dask_image/test_ndfilters/test_cupy_threshold.py:37:
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
dask_image/ndfilters/_threshold.py:87: in threshold_local
    thresh_image = _gaussian.gaussian_filter(image, sigma, mode=mode,
dask_image/ndfilters/_gaussian.py:65: in gaussian_filter
    sigma = _get_sigmas(image, sigma)
dask_image/ndfilters/_gaussian.py:25: in _get_sigmas
    nsigmas = np.array(sigma)
../../miniconda3/envs/test_dask_image/lib/python3.10/site-packages/dask/array/core.py:1703: in __array__
    x = np.array(x)
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

>   ???
E   TypeError: Implicit conversion to a NumPy array is not allowed. Please use `.get()` to construct a NumPy array explicitly.

cupy/_core/core.pyx:1480: TypeError

I think the error makes sense, as sigma is passed as a cupy backed dask array by dask_image.ndfilters.threshold_local:

dask-image/dask_image/ndfilters/_threshold.py

Line 83 in 91ea280

sigma = (da.asarray(block_size, like=image._meta) - 1) / 6.0

However, somehow the gpuCI doesn't pick it up.

tests.test_dask_image.test_ndfilters contains test_cupy_ndfilters.py and test_cupy_threshold.py. For some reason the tests in the former are run and the test defined in the latter is ignored... Why could this be?

[GenevieveBuckley]

I don't have good insight into the gpuCI. Perhaps @jakirkham or @charlesbluca might have a better idea about what might be going wrong here?

[jakirkham]

We seem to be using an old class-style setup in the threshold modules (with a decorator)

Maybe we should move away from this structure to something more similar to typical pytest usage (like functions maybe with a fixture)

dask-image/tests/test_dask_image/test_ndfilters/test_cupy_threshold.py

Line 20 in fba6645

class TestSimpleImage:

dask-image/tests/test_dask_image/test_ndfilters/test__threshold.py

Line 16 in fba6645

class TestSimpleImage:

[m-albert]

Think I found the problem. Trying to recreate the way the gpuci performs tests I realized that the ignored function is not marked with @pytest.mark.cupy. Marking it includes it in the tests, at least locally.