Some changes to facilitate newly implemented CCSDt #148
Conversation
avoid considering auxiliary for orthogonality tests; Changes in hamiltonian.py to return a np array of (2, modim, modim) instead of a tuple, the tuple results in errors in ebcc calculations for UHF. PYSCF-wide default is the np array.
|
Is it better to fix
in https://github.com/BoothGroup/ebcc/blob/ff5240121b23e43c2e760968946f51900b13c805/ebcc/uebcc.py#L448 to |
That's what I thought at first. But I tested what pyscf returns for UHF mf.get_fock() and it is indeed (2, nmo, nmo ) and not a tuple. I recall this was put in as pad_to_match fix for active spaces with unequal numbers of electrons/orbitals in alpha and beta channels. |
|
cool, up to you |
typo in np.asarray
Codecov ReportAttention:
Additional details and impacted files@@ Coverage Diff @@
## master #148 +/- ##
==========================================
- Coverage 71.53% 0.00% -71.54%
==========================================
Files 152 152
Lines 20159 20164 +5
Branches 3343 0 -3343
==========================================
- Hits 14420 0 -14420
- Misses 4906 20164 +15258
+ Partials 833 0 -833 ☔ View full report in Codecov by Sentry. |
solver=EBCC and ansatz=CCSDt possibility added
| @@ -639,7 +640,7 @@ def get_fock(self, with_vext=True, use_seris=True, with_exxdiv=False): | |||
| + einsum("iipq->pq", gab[oa, oa]) # Coulomb | |||
| - einsum("ipqi->pq", gbb[ob, :, :, ob]) | |||
| ) # Exchange | |||
| fock = ((fock[0] + dfa), (fock[1] + dfb)) | |||
| fock = np.asarray(((fock[0] + dfa), (fock[1] + dfb))) | |||
There was a problem hiding this comment.
Can this can be packed into a single array, if fock[0] and fock[1] could have different shapes?
There was a problem hiding this comment.
It can not be if the shapes are different. Wouldn't Fock matrices for alpha and beta always have the same shape?
contributes implies active; removed the redundant active=True
update has_actspace flag
+ small bug fix, ebcc keyword is max_iter for ma_cycle
Added Active space support for EBCC CCSDt; Avoid considering auxiliary for orthogonality tests for EC-CC or TCC calculations. Changes in hamiltonian.py to return a np array of (2, modim, modim) instead of a tuple, the tuple results in errors in ebcc for UHF. PYSCF-wide default is the np array.