I suspect this is architecture related since the CI for this test is consistently passing. The problem was introduced during one of the recent numerical stability/optimization PRs. I'm just documenting this here and not fixing it because it seems like it might be a tough one to resolve.
research/berry/berrylib/test_berry.py:184: in test_fast_inla
np.testing.assert_allclose(out[0][0].sum(), 2111.808, rtol=1e-3)
E AssertionError:
E Not equal to tolerance rtol=0.001, atol=0
E
E Mismatched elements: 1 / 1 (100%)
E Max absolute difference: 50.70225391
E Max relative difference: 0.02400893
E x: array(2162.5103, dtype=float32)
E y: array(2111.808)
@constinit
I suspect this is architecture related since the CI for this test is consistently passing. The problem was introduced during one of the recent numerical stability/optimization PRs. I'm just documenting this here and not fixing it because it seems like it might be a tough one to resolve.