Bug#1118519: opentsne: autopkgtest regression on some architectures with scikit 1.7.2

Timo Röhling <[email protected]>
Newsgroups gmane.linux.debian.ports.s390
Message-ID <176105271898.34868.6168174698171415329.reportbug__28729.2122420453$1761052890$gmane$org@roehling3.cms.fkie.fraunhofer.de>
Source: opentsne
Version: 1.0.2-4
Severity: serious
X-Debbugs-Cc: [email protected], [email protected], [email protected]
User: [email protected]
Usertags: arm64
User: [email protected]
Usertags: riscv64
User: [email protected]
Usertags: s390x
User: [email protected]
Usertags: scikit-learn-1.7

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA512

Dear maintainer,

your package has an autopkgtest regression with sckit-learn 1.7.2. I'm 
quoting the arm64 [1] error below, but it is the same failing test on 
risc64 [2] and s390x [3] as well:

[1] https://ci.debian.net/packages/o/opentsne/testing/arm64/65339152/
[2] https://ci.debian.net/packages/o/opentsne/testing/riscv64/65333862/
[3] https://ci.debian.net/packages/o/opentsne/testing/s390x/65330911/


220s =================================== FAILURES ===================================
220s _________ TestTSNECorrectnessUsingPrecomputedDistanceMatrix.test_iris __________
220s
220s self = <tests.test_correctness.TestTSNECorrectnessUsingPrecomputedDistanceMatrix testMethod=test_iris>
220s
220s     @unittest.skipIf(('i686' in machine) or machine == 'i386'
220s                        or machine == 'ppc64le',
220s                      reason=f"Skipping test on {machine} architecture")
220s     def test_iris(self):
220s         x = datasets.load_iris().data
220s         print(platform.machine())
220s         x += np.random.normal(0, 1e-3, x.shape)  # iris contains duplicate rows
220s
220s         distances = squareform(pdist(x))
220s         params = dict(initialization="random", random_state=0)
220s         embedding1 = TSNE(metric="precomputed", **params).fit(distances)
220s         embedding2 = TSNE(metric="euclidean", **params).fit(x)
220s
220s >       np.testing.assert_almost_equal(embedding1, embedding2)
220s E       AssertionError:
220s E       Arrays are not almost equal to 7 decimals
220s E
220s E       Mismatched elements: 300 / 300 (100%)
220s E       Max absolute difference among violations: 4.85427071
220s E       Max relative difference among violations: 47.91193852
220s E        ACTUAL: TSNEEmbedding([[ 11.855024 ,  15.4978719],
220s E                      [ 11.7985732,  12.7515209],
220s E                      [ 12.8316039,  13.2288263],...
220s E        DESIRED: TSNEEmbedding([[ 13.8717067,  11.1420085],
220s E                      [ 14.6164783,   8.511387 ],
220s E                      [ 15.4360712,   9.2738073],...
220s
220s tests/test_correctness.py:307: AssertionError


Cheers
Timo

-----BEGIN PGP SIGNATURE-----

iQIzBAEBCgAdFiEEmwPruYMA35fCsSO/zIxr3RQD9MoFAmj3iC4ACgkQzIxr3RQD
9MrQRA/+LR9VIiz9dvHEQ282uBSwRkocmdz+aJkA1OfL4YRv9s42sbo7kT8/nFOp
PrVc7Rjk1SXJHIzusvPlJxu+mVPqFwIiiuKNwh5Pae26WcAbjndSFnKKljuT0mLM
t2CGL5Ul4zes54xUYfV4DN3Ha3impJawv0V5w3x41DXt8VG18CI0jI8NXkye4pCp
4lTb8Vpb17qQgaDOTgd3Ai+wQinsGgHqAgQJXqA7xKIGsYQGmWxVW2ZYIuU3wWqV
DVUF8Pm0e3zRGUE7Up0MsanTIDTufkn5ZyJOBDD0fFTVoUvTXt3ZeSQuZlsfw+wk
etvT29qr4f8WE/GeIAVmkR5zEkSm9HN6UfBBDTVWCd6q942cD3Gyt2u4cibsv+eb
c5dtwRH9ceOJPGx7VzRaFrnzKMwlmYQT5+nWaJDU+Cx9W/9y+o2io1uOfmlpVsDV
txWU0fbngc2Sv52zXATt/AVLrcrdOtHmDmOLhIZgz9ges9NpCK8Jbl2fFD+8fB6j
pbIBovYJMmvX2TFOL8wOEYwY6mlFpeI8SflqAqZbzXlqwEhKOu/b6HhoPrVipSHB
qR5KUO7v/ySMWE2doyxtWXWzTEt1NtfdkGexAZkUcP/XLM0f6FCTuQqAMDfRfiBP
LSA1DhdsCeeje9mwvRadJf9FUyq/nkt5Q0QbtL8c5gndZ8IbJEI=
=tUDO
-----END PGP SIGNATURE-----
lmpx.com only provides a reader for public news (NNTP) servers. It is not affiliated with the servers or forums shown here and is not responsible for the content of articles, which is written by their respective authors.