[CP2K-user] [CP2K:15560] stopping run without any error alert

Nicholas Winner nwi... at berkeley.edu
Mon Jun 14 17:20:54 UTC 2021


(L)BFGS is a second order minimizer, as a result it doesn't have a step 
size and relies on approximating the local region of your minimization as a 
second order equation. This means that if you are close to the minimum, it 
can be very fast at finding your minimum. It also means that the further 
the you are from the minimum, the more it will "jump around" in an attempt 
to fit a 2nd order equation where it shouldn't be. LBFGS is a limited 
memory variant of BFGS, which you can use when you want the speed of BFGS 
for very large systems (usually around ~1000 atoms is a good guess for when 
to use LBFGS). Because BFGS is faster, its often a good idea to use if your 
initial guess is at least reasonable. If its not, then CG will be much more 
stable. 

On Monday, June 14, 2021 at 9:23:53 AM UTC-7 tah... at gmail.com wrote:

> Dear  Nicholas
>
> Thank you for your helping!
>
> I will add this subsection to the CG section!
> OPTIMIZER CG
>     &CG
>       MAX_STEEP_STEPS  0
>       RESTART_LIMIT 9.0E-01
>       &LINE-SEARCH
>       TYPE 2PNT
>       & END LINE-SEARCH 
>     &END CG
> Is this true????
> Do you think it is possible to use other minimizers such as BFGS, LBFGS 
> instead of the CG Minimizer? Which one of them is better?
> How many maximum numbers of SCF iteration to be selected for inner and 
> outer SCF? I had previously set these numbers to 30 and 10, subsequently.
>
> Thanks a lot 
>
>
> On Sun, Jun 13, 2021 at 10:06 PM Nicholas Winner <n... at berkeley.edu> 
> wrote:
>
>> Your SCF procedure is near as fast as it can go without using very 
>> imprecise settings. You CELL_OPT, however, is slowing your calc down a lot. 
>> You have selected the CG optimizer, but you have not set the line_search 
>> type. For reasons I cannot fathom the *gold* line search is the default 
>> setting, which is an extremely expensive line search. Change the CG line 
>> search from TYPE GOLD to TYPE 2PNT and the number of optimization steps 
>> should decrease considerably. If you need more speed after that, you simply 
>> need more hardware in my opinion. 8 cores is not very much for this size of 
>> a system. 
>>
>> On Sunday, June 13, 2021 at 4:43:59 AM UTC-7 tah... at gmail.com wrote:
>>
>>> Dear Matthias
>>> Thanks for your reply. You helped me a lot. I will run my calculations 
>>> based on what you said. I hope it works and the result is good. 
>>>
>>> Sorry! I have got another question. How can I speed up my calculations? Very 
>>> slow!.... Is there any keyword to reduce the computational cost? 
>>>
>>> Thank you in advanced 
>>>
>>> On Sun, Jun 13, 2021 at 1:34 PM Krack Matthias (PSI) <mat... at psi.ch> 
>>> wrote:
>>>
>>>> Hi
>>>>
>>>>  
>>>>
>>>> Try to use only DZVP-SR-MOLOPT basis sets first and decrease 
>>>> EPS_DEFAULT to 1.0E-12 or less.
>>>>
>>>>  
>>>>
>>>> M.
>>>>
>>>>  
>>>>
>>>> *Von:* c... at googlegroups.com <c... at googlegroups.com> *Im Auftrag von 
>>>> *Tahere Kheshti
>>>> *Gesendet:* Samstag, 12. Juni 2021 20:35
>>>> *An:* c... at googlegroups.com
>>>> *Betreff:* [CP2K:15555] stopping run without any error alert
>>>>
>>>>  
>>>>
>>>> Dear cp2k users
>>>>
>>>>  
>>>>
>>>> I am a new user for cp2k and trying to run DFT  calculations on 
>>>> perovskite materials.
>>>>
>>>> I have started from the single crystal x-ray structure include H, C, N, 
>>>> Cl, Br, F, and In atoms. There are 114 atoms in my unit cell! 
>>>>
>>>> I want to optimize this structure using the* PBE XC functional* by *cell 
>>>> opt run type*. The DZVP-MOLOPT basis sets were used to describe 
>>>> non-metallic atoms and TZVP-SR-MOLOPT for In.
>>>>
>>>> After 10 days, *my running was stopped without any error alert! *I 
>>>> don't know what the problem is?*!*
>>>>
>>>> *It should be mentioned that the memory of my computer cluster is 16 GB 
>>>> with 8 cores! I am running cp2k v.6.1. in parallel 16 cores!*
>>>>
>>>>  
>>>>
>>>> *C*ould you please help me to fix it!
>>>>
>>>> IThe input and output files were attached.
>>>>
>>>> -- 
>>>> You received this message because you are subscribed to the Google 
>>>> Groups "cp2k" group.
>>>> To unsubscribe from this group and stop receiving emails from it, send 
>>>> an email to cp... at googlegroups.com.
>>>> To view this discussion on the web visit 
>>>> https://groups.google.com/d/msgid/cp2k/CAJwLP87zWBE2K_4mQc-HMvEQU6TtVgKp0BCcBki-pJyQnhy0BQ%40mail.gmail.com 
>>>> <https://groups.google.com/d/msgid/cp2k/CAJwLP87zWBE2K_4mQc-HMvEQU6TtVgKp0BCcBki-pJyQnhy0BQ%40mail.gmail.com?utm_medium=email&utm_source=footer>
>>>> .
>>>>
>>>> -- 
>>>> You received this message because you are subscribed to the Google 
>>>> Groups "cp2k" group.
>>>> To unsubscribe from this group and stop receiving emails from it, send 
>>>> an email to cp... at googlegroups.com.
>>>>
>>> To view this discussion on the web visit 
>>>> https://groups.google.com/d/msgid/cp2k/ad122d449afc450b8d1d67dc0ccc17ed%40psi.ch 
>>>> <https://groups.google.com/d/msgid/cp2k/ad122d449afc450b8d1d67dc0ccc17ed%40psi.ch?utm_medium=email&utm_source=footer>
>>>> .
>>>>
>>> -- 
>> You received this message because you are subscribed to the Google Groups 
>> "cp2k" group.
>> To unsubscribe from this group and stop receiving emails from it, send an 
>> email to cp... at googlegroups.com.
>>
> To view this discussion on the web visit 
>> https://groups.google.com/d/msgid/cp2k/536b8c37-9383-4939-99c5-6d36b03cba74n%40googlegroups.com 
>> <https://groups.google.com/d/msgid/cp2k/536b8c37-9383-4939-99c5-6d36b03cba74n%40googlegroups.com?utm_medium=email&utm_source=footer>
>> .
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20210614/b3fb6bdc/attachment.htm>


More information about the CP2K-user mailing list