The second to last one on my list for @loriab. What I found out only at the end was that Psi doesn't actually have a DF-LCCSD implementation. While Psi will not complain if you set cc_type to be DF and request lccsd, proc.py:run_cepa never actually checks cc_type before sending you to fnocc, so the computation proceeds using conventional integrals.
Accordingly, I test my reference implementation against the value I got when I constructed the DF integrals as four-index quantities and applied the tested conventional code. It's not an ideal test, but it'll do.
What are your new additions? Please provide a brief list.
Yikes, I’ll fix that in psi. I didn’t fully realize before yesterday that fnocc cepa was conv only either. My rule of thumb was that fnocc does everything, so long as it’s an rhf energy.
Description
The second to last one on my list for @loriab. What I found out only at the end was that Psi doesn't actually have a DF-LCCSD implementation. While Psi will not complain if you set
cc_type
to beDF
and requestlccsd
,proc.py:run_cepa
never actually checkscc_type
before sending you tofnocc
, so the computation proceeds using conventional integrals.Accordingly, I test my reference implementation against the value I got when I constructed the DF integrals as four-index quantities and applied the tested conventional code. It's not an ideal test, but it'll do.
What are your new additions? Please provide a brief list.
Status