[CI] Upgrade macos runner image

[mshr-h]

• macOS runner 12->14

[mshr-h]

Segmentation fault on macos-13 :(

https://github.com/microsoft/hummingbird/actions/runs/10092260187/job/27905556470

Fatal Python error: Segmentation fault

Thread 0x000070000b868000 (most recent call first):
tests/test_backends.py .........
File "/Users/runner/hostedtoolcache/Python/3.10.14/x64/lib/python3.10/selectors.py", line 416 in select
File "/Users/runner/hostedtoolcache/Python/3.10.14/x64/lib/python3.10/socketserver.py", line 232 in serve_forever
File "/Users/runner/hostedtoolcache/Python/3.10.14/x64/lib/python3.10/threading.py", line 953 in run
File "/Users/runner/hostedtoolcache/Python/3.10.14/x64/lib/python3.10/threading.py", line 1016 in _bootstrap_inner
File "/Users/runner/hostedtoolcache/Python/3.10.14/x64/lib/python3.10/threading.py", line 973 in _bootstrap

Thread 0x000070000a865000 (most recent call first):
File "/Users/runner/hostedtoolcache/Python/3.10.14/x64/lib/python3.10/site-packages/py4j/clientserver.py", line 58 in run
File "/Users/runner/hostedtoolcache/Python/3.10.14/x64/lib/python3.10/threading.py", line 1016 in _bootstrap_inner
File "/Users/runner/hostedtoolcache/Python/3.10.14/x64/lib/python3.10/threading.py", line 973 in _bootstrap

Thread 0x00007ff850e457c0 (most recent call first):
File "/Users/runner/hostedtoolcache/Python/3.10.14/x64/lib/python3.10/functools.py", line 63 in update_wrapper
File "/Users/runner/hostedtoolcache/Python/3.10.14/x64/lib/python3.10/site-packages/torch/onnx/symbolic_helper.py", line 308 in decorator
File "/Users/runner/hostedtoolcache/Python/3.10.14/x64/lib/python3.10/site-packages/torch/onnx/symbolic_opset13.py", line 339 in
File "/Users/runner/work/_temp/5491f745-7b5e-4065-a0ea-547c39b8115a.sh: line 1: 15671 Segmentation fault: 11 pytest

[mshr-h]

With macos-14, we hit another error.
https://github.com/microsoft/hummingbird/actions/runs/10093420039/job/27909084340

Do we have to be more tolerant? :(

_______ TestSklearnMatrixDecomposition.test_kernel_pca_converter_linear ________

self = <test_sklearn_decomposition.TestSklearnMatrixDecomposition testMethod=test_kernel_pca_converter_linear>

def test_kernel_pca_converter_linear(self):

  self._fit_model_pca(KernelPCA(n_components=5, kernel="linear"))

tests/test_sklearn_decomposition.py:79:

---

tests/test_sklearn_decomposition.py:35: in _fit_model_pca
np.testing.assert_allclose(model.transform(X_test), torch_model.transform(X_test), rtol=1e-6, atol=2 * 1e-5)

---

args = (<function assert_allclose..compare at 0x13e9743a0>, array([[ 17.46886258, 8.0934041 , 12.31343715, -5.393... -11.795859 ],
[-14.634164 , 15.9322605 , 0.75495166, 2.1367702 ,
12.430414 ]], dtype=float32))
kwds = {'equal_nan': True, 'err_msg': '', 'header': 'Not equal to tolerance rtol=1e-06, atol=2e-05', 'verbose': True}

@wraps(func)
def inner(*args, **kwds):
    with self._recreate_cm():

      return func(*args, **kwds)

E AssertionError:
E Not equal to tolerance rtol=1e-06, atol=2e-05
E
E Mismatched elements: 1 / 1800 (0.0556%)
E Max absolute difference: 4.51823837e-05
E Max relative difference: 1.33559948e-05
E x: array([[ 17.468863, 8.093404, 12.313437, -5.393746, -0.939549],
E [-13.784834, 7.6482 , -7.446306, 3.170859, 7.215178],
E [-20.562256, 2.393334, 9.65356 , -7.913929, 4.56979 ],...
E y: array([[ 17.468843, 8.093403, 12.313433, -5.393748, -0.939549],
E [-13.784828, 7.648199, -7.446309, 3.170859, 7.215181],
E [-20.562273, 2.393334, 9.653566, -7.913927, 4.569791],...

[mshr-h]

All tests have passed. But I changed atol and rtol in tests/test_sklearn_decomposition.py. I don't know if it's ok.
@ksaur @interesaaat

[ksaur]

All tests have passed. But I changed atol and rtol in tests/test_sklearn_decomposition.py. I don't know if it's ok. @ksaur @interesaaat

I think this rtol change is ok for this one test. confirming with @interesaaat ?

[ksaur]

In the meantime, i actually updated the runner checks this time!! 😆

[ksaur]

Thanks so much for all your help!!