[CP2K:2077] File numbering for CELL_OPT runs restarts

Teodoro Laino teodor... at gmail.com
Mon May 18 11:52:24 CEST 2009


This message is post only to keep track of the test suite that shows 
that the cell_opt numbering scheme (+ restarting) is working.
If you have an additional case, where you can see that it does not work, 
please post it here..
Keep in mind that it has to run fast enough in order we don't waste too 
much human/computer time.

The test attached (the same posted in another related thread regarding 
the WALLTIME keyword) is using FIST.
Since the cell optimization is not related to the energy/forces 
evaluation this should be considered a fairly good test.
As I said, I'm opened to counterexamples.

Regards,
Teo



Maricarme... at cemes.fr wrote:
> Hi Teo,
>
> I actually downloaded the latest CVS version and recompiled the code,
> but I'm still getting the same issue (GEO_OPT files from previous
> CELL_OPT runs being appended because of the numbering scheme not being
> continued). I've tried with the RESTART_COUNTERS flag on (T) and off
> (F). I copy the input file below.
> Is there something else I'm missing?
>
> Thanks,
>
> Maricarmen
>
>
> P.S.: After the input file I include a copy of our previous messages
> regarding this subjet, made before I could subscribe to the group.
>
>
>
> &GLOBAL
>   PROJECT     Ag2V4O11_1-CellOpt
>   RUN_TYPE    CELL_OPT
>   PRINT_LEVEL MEDIUM
>   WALLTIME 84000
> &END GLOBAL
> &FORCE_EVAL
>   METHOD Quickstep
>   &DFT
>     BASIS_SET_FILE_NAME /scratch/grisolia/cp2k/tests/QS/BASIS_MOLOPT
>     POTENTIAL_FILE_NAME /scratch/grisolia/cp2k/tests/QS/GTH_POTENTIALS
>     LSD
>     &MGRID
>       CUTOFF 280
>       NGRIDS 5
>     &END MGRID
>     &QS
>       EPS_DEFAULT   1.0E-10
>       EXTRAPOLATION PS
>       EXTRAPOLATION_ORDER 1
>     &END QS
>     &SCF
>       SCF_GUESS RESTART
>       EPS_SCF 2.0E-7
>       MAX_SCF 30
>       &OUTER_SCF
>          EPS_SCF 2.0E-7
>          MAX_SCF 15
>       &END
>       &OT
>         MINIMIZER CG
>         PRECONDITIONER FULL_SINGLE_INVERSE
>         ENERGY_GAP 0.05
>       &END
>       &PRINT
>          &RESTART
>             FILENAME = Ag2V4O11_1.wfn
>          &END
>       &END
>     &END SCF
>     &XC
>       &XC_FUNCTIONAL PBE
>       &END XC_FUNCTIONAL
>     &END XC
>      &PRINT
>        &MO_CUBES
>           WRITE_CUBE F
>           NLUMO      20
>           NHOMO      20
>        &END
>      &END
>   &END DFT
>   &SUBSYS
>     &CELL
>       @INCLUDE Ag2V4O11_1.cell
>     &END CELL
>     &COORD
>       @INCLUDE Ag2V4O11_1.xyz
>     &END COORD
>     &KIND Ag
>       BASIS_SET DZVP-MOLOPT-SR-GTH
>       POTENTIAL GTH-BLYP-q11
>     &END KIND
>     &KIND O
>       BASIS_SET DZVP-MOLOPT-SR-GTH
>       POTENTIAL GTH-PBE-q6
>     &END KIND
>     &KIND V
>       BASIS_SET DZVP-MOLOPT-SR-GTH
>       POTENTIAL GTH-PBE-q13
>     &END KIND
>   &END SUBSYS
>   STRESS_TENSOR ANALYTICAL
> &END FORCE_EVAL
> &MOTION
>   &MD
>       TIMESTEP [fs] 0.5
>       STEPS         10000
>       TEMPERATURE   500
>       ENSEMBLE      NVE
>   &END
>   &CELL_OPT
>     TYPE GEO_OPT
>     OPTIMIZER CG
>     MAX_ITER 20
>     EXTERNAL_PRESSURE [bar] 0.0
>     MAX_DR 0.02
>     RMS_DR 0.01
>     MAX_FORCE 0.002
>     RMS_FORCE 0.001
>     KEEP_ANGLES F
>     &CG
>       &LINE_SEARCH
>         TYPE 2PNT
>         &2PNT
>         &END
>       &END
>     &END
>   &END
>   &GEO_OPT
>     MAX_ITER 300
>     MINIMIZER LBFGS
>   &END
> &END
> &EXT_RESTART
>    RESTART_FILE_NAME Ag2V4O11_1-CellOpt-1.restart
>    RESTART_DEFAULT  T
>    RESTART_COUNTERS F
> &END EXT_RESTART
>
>
>
>
> Laino Teodoro a écrit :
>   
>> Hi Maricarmen,
>>
>> this is a real bug and it is now fixed in the CVS.
>> Thanks for reporting this problem. I hope you solved as well the problem with the google list.
>> In fact, I prefer that the help request (next time) go directly to the mailing list, so we (all) can keep track of them.
>> Thanks and have a nice evening,
>> Teo
>>
>> On 4 May 2009, at 09:55, Maricarmen Grisolia wrote:
>>
>>     
>>> Ciao Teo,
>>>
>>> Thank you for your concern on the google group issue. I just realised it might be due to a difference in the email address on my google account (gris... at cemes.fr instead of maricarme... at cemes.fr). I'm trying to fix it.
>>>
>>> Meanwhile, I will ask you directly what I wanted to know (the reason why I asked for the group's subscription). I'm doing a DFT cell optimization on some metal-vanadium oxyde crystals (around 260 atoms). As this is a large system and I have a limited calculation time on my calculus center, I must do several restarts until cell optimization is completed. During this process, CP2K will write the outputs for every geometry optimization on each cell optimization step, and it will give them numbers, as MyProject-GEO_OPT-1.out, MyProject-GEO_OPT-2.out, and so forth. This is my problem: when I do the restart, CP2K will restart these numbers too, appending to the already existent files instead of writing new ones with the numbers starting where they have left. I would much rather have it continuing the numbering so as to preserve the sequence on the optimization process.
>>>
>>> I already tried with RESTART_COUNTERS F, but it's not working (this keyword is intended only for MD runs, not GEO_OPT, I guess). So, is there a way in which I can get CP2K to continue the numbering of the GEO-OPT output files during cell optimization? Otherwise I will have to move these files to another folder and/or renaming by hand each time I do a restart... That would be very annoying...
>>>
>>> I hope you can clarify this to me.
>>>
>>> I'm also having a segmentation fault issue (sigsegv error when starting a SCF cycle after the first one or two cell optimization steps), but I've read some threads on the group about this and I guess this is a machine configuration problem (the stack size limit thing), not CP2K's. I will ask my calculus center admins to see if they can solve it.
>>>
>>> Ciao,
>>> Maricarmen
>>>
>>>       
>
> >
>   

-------------- next part --------------
A non-text attachment was scrubbed...
Name: bug_report_cell.tgz
Type: application/octet-stream
Size: 2547831 bytes
Desc: not available
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20090518/517827dc/attachment.obj>


More information about the CP2K-user mailing list