Update utils.py (#410)

In some versions of scipy(e.g. 1.14), M.A might not be available. In such cases, you should use M.toarray() to convert the sparse matrix to a dense one.

Co-authored-by: Michaela Müller <51025211+mumichae@users.noreply.github.com>

scib/metrics/utils.py

@@ -138,7 +138,7 @@ def diffusion_nn(adata, k, max_iterations=26):
         )
 
     M.setdiag(0)
-    k_indices = np.argpartition(M.A, -k, axis=1)[:, -k:]
+    k_indices = np.argpartition(M.toarray(), -k, axis=1)[:, -k:]
 
     return k_indices