LASSI-PDFT updated the framework & example

[JangidBhavnesh]

I have changed the example file according to recent example file of lassi.

[MatthewRHermes]

It is unnecessary to check both lassi.LASSI and lassis.LASSIS: the latter is a child class of the former, so you only need to check the former.

[JangidBhavnesh]

Done.

[MatthewRHermes]

It would be good to have a test case involving more than one fragment. For instance, take a look at https://github.com/MatthewRHermes/mrh/blob/master/tests/lassi/test_22.py:

mrh/tests/lassi/test_22.py

Lines 29 to 59 in 3dab0fa

	xyz='''H 0 0 0
	H 1 0 0
	H 3 0 0
	H 4 0 0'''
	mol = gto.M (atom=xyz, basis='sto3g', symmetry=False, verbose=0, output='/dev/null')
	mf = scf.RHF (mol).run ()
	# Random Hamiltonian
	rng = np.random.default_rng (424)
	mf._eri = rng.random (mf._eri.shape)
	hcore = rng.random ((4,4))
	hcore = hcore + hcore.T
	mf.get_hcore = lambda *args: hcore
	# LASSCF with CASCI-limit model space
	las = LASSCF (mf, (2,2), (2,2), spin_sub=(1,1))
	las.lasci ()
	las1 = las
	for i in range (2): las1 = all_single_excitations (las1)
	charges, spins, smults, wfnsyms = get_space_info (las1)
	lroots = 4 - smults
	idx = (charges!=0) & (lroots==3)
	lroots[idx] = 1
	las1.conv_tol_grad = las.conv_tol_self = 9e99
	las1.lasci (lroots=lroots.T)
	las1.dump_spaces ()
	lsi = LASSI (las1)
	lsi.kernel (opt=0)
	# CASCI limit
	mc = mcscf.CASCI (mf, 4, 4).run ()

lsi.e_roots[0] equals mc.e_tot here. The relationship should also hold for PDFT.

[JangidBhavnesh]

Unit test updated.

LASSI-PDFT energy is matching the CASCI-PDFT up to 9-decimal points.

[JangidBhavnesh]

There are mcpdft, mspdft, their gradients, and property unit test are there at
https://github.com/MatthewRHermes/mrh/tree/master/tests/mcpdft

Should one remove them?, since they have been shifted to pyscf-forge.

[MatthewRHermes]

No do not remove them. I left them there for a reason. Those marked "dupe" do not run on github CI but I run them locally.