I'm encountering issues with linear dependencies in the basis set, leading to the following warning during the SCF procedure:

WARN: Singularity detected in overlap matrix (condition number = 2.15e+08).
SCF may be inaccurate and hard to converge.

def eig(h, s, cutoff=1e-7):
d, t = np.linalg.eigh(s)
mask = d > cutoff
x = t[:, mask] / np.sqrt(d[mask])
xhx = x.T @ h @ x
e, c = np.linalg.eigh(xhx)
c = x @ c
return e, c

mol = gto.Mole()
mol.atom = atom_block
mol.basis = '6-31++g**'
mol.spin = 0
mol.charge = 0
mol.build(verbose=0)

mf = scf.RHF(mol)
mf.verbose = 3
mf.max_cycle = 1000
mf.eig = eig
mf = addons.remove_linear_dep_(mf)
mf.scf()

Despite using addons.remove_linear_dep_, I still get the singular overlap matrix warning and unstable SCF behavior. Even after attempting to remove the smallest eigenvalues from the overlap matrix I get the warning.