Fix using .A instead of .toarray()
#61
Conversation
tests/test_gpcca.py
Outdated
| @@ -524,6 +525,7 @@ def test_gpcca_brandts_sparse_is_not_densified(self, P: np.ndarray, sd: np.ndarr | |||
| with pytest.raises(ValueError, match=r"Sparse implementation is only available for `method='krylov'`."): | |||
| GPCCA(csr_matrix(P), eta=sd, method="brandts").optimize(3) | |||
|
|
|||
| @pytest.mark.skipif(Version(np.__version__) >= Version("2"), reason="Eigenvalue mismatch.") | |||
There was a problem hiding this comment.
So you skip this test if numpy version is >=2? Why is it ok for newer numpy versions to fail this test?
There was a problem hiding this comment.
I've been trying to figure out why new numpy version causes this test to fail, but was unable to figure it out. I will try later do a bisect on the commits, see exactly why, but was thinking in the meantime we could just skip it.
There was a problem hiding this comment.
I've reverted this, will leave this for a future PR to investigate.
| @@ -70,7 +70,7 @@ def bdc(q: np.ndarray, p: np.ndarray, sparse: bool = True) -> Tuple[np.ndarray, | |||
| mu[0] = 1.0 | |||
| mu[1:] = np.cumprod(p[:-1] / q[1:]) | |||
|
|
|||
| return (P if sparse else P.A), mu / np.sum(mu) | |||
| return (P if sparse else P.toarray()), mu / np.sum(mu) | |||
There was a problem hiding this comment.
What was the problem here, exactly?
There was a problem hiding this comment.
This is because scipy deprecated .A and .toarray() should be used in favor.
Regarding the changes in the docs, sphinx was throwing some warning, so I corrected the template and |
No description provided.