Top Page > Browsing
Problems with NH MD and multiple continuation runs
Date: 2021/08/02 22:24
Name: Pavel Ondracka   <pavel.ondracka@email.cz>

So first of all the systemname.dat# continuation file is extremely useful, I have some long-time MD runs, and since we have quite short max jobtime limit, I just restart the jobs manually when killed by the scheduler. Also when I run the systemname.dat# file, I get systemname.dat##, and so on so multiple continuation runs are in theory possible. For this to work, the internal state like velocities and thermostat state needs to be included into the .dat# file, they are usually appended and it works fine. However now I noticed a problem with NH thermostat, i.e., some stuff is appended in a strange way.

So taking an example Gly_NH.dat and doing just few MD steps, the .dat# file contains:

----------------------------------------------------------------------
<MD.Init.Velocity
1 1418.12812465549632 -639.68997368704561 277.34239416514589
2 704.87749820695910 635.47023752775442 -1174.37669270961442
3 -361.98302554057739 321.98043746619283 -1257.66432420780416
4 807.54643862445289 -575.07608172579410 542.32637073093883
5 -1775.35613089288404 -997.18010779665963 3635.52572039333245
6 4198.84686479931588 1124.78170143343368 3117.26909190340666
7 -48.01066028523179 104.78752120527723 -1411.47957088142562
8 473.81763811770435 -3724.86851014468220 1527.65693444363637
9 -4607.07439031421836 -556.54485991111824 -3227.81205879084928
10 -810.79235737101715 4306.33963563264297 -2028.78786504676691
MD.Init.Velocity>


<MD.Init.Velocity.Prev
1 1460.29455134506861 -594.21525013588723 363.99454140159816
2 843.07225922432053 681.69722924080554 -1111.46484939476659
3 -258.34137962354851 412.87946463496331 -1170.28099380777189
4 868.08020354816290 -517.61641552431342 638.49144011490637
5 -1535.84685578350377 -755.09801746411995 3800.14858520716507
6 4295.66210803783997 1135.77781666036208 2822.62286860381892
7 -71.48297562808149 94.95472684514651 -1349.56668008313591
8 -71.38447339224349 -3619.86724193978262 1753.45958458325936
9 -4131.97205708078218 -669.68247513619303 -3312.98588828646507
10 -1125.70785159117281 4050.37705293710997 -2055.10251499413425
MD.Init.Velocity.Prev>

NH.R -0.000001979427717
NH.nzeta -0.000000239401917
NH.czeta -0.000000152041857
------------------------------------------------------------------------

so all looks OK, now I restart again from this file for few more steps. The relevant part of the .dat## file looks like this

-----------------------------------------------------------------------

<MD.Init.Velocity
1 1352.25874817298541 -728.59470627870394 99.88577893912608
2 256.41755087905369 540.45705617766782 -1266.13454445487400
3 -664.22554634330822 52.88992301133894 -1447.68880617887226
4 665.97061561088674 -703.28203877155886 326.15047919818772
5 -2584.56101426103896 -1829.77804737415295 3160.58005635603467
6 3918.72394791043735 1179.87720727330861 4370.91410885054393
7 155.56889297786475 239.57068418983405 -1496.93659562166590
8 2669.22704493872197 -4042.77189068729103 834.78218560044604
9 -6238.44086737751331 -63.90159903239377 -2795.74491640869155
10 469.06062749191170 5355.53341149195148 -1785.80774628023255
MD.Init.Velocity>
<MD.Init.Velocity.Prev
1 1399.99981891087145 -688.63751722073675 172.71843051112347
2 398.94153053952635 582.49002539145636 -1212.26369083794839
3 -560.40480518371294 137.55936062153032 -1369.05689341668267
4 733.02283465224957 -656.07656663707189 408.88811335170800
5 -2364.16518707952264 -1616.05745703155571 3303.75939351931402
6 4018.99613238830852 1182.27148737639618 4111.56162860108361
7 134.30820730277063 223.02925018561785 -1445.18925154363228
8 2181.30342929107974 -3950.38168644083817 1024.06632945433626
9 -5844.39301171041643 -155.50289567475406 -2855.31273676088631
10 204.38591621153648 5131.53773484659632 -1809.24075190004055
MD.Init.Velocity.Prev>
NH.R 0.000009511132209
NH.nzeta 0.000001383174729
NH.czeta 0.000002845255094



<MD.Init.Velocity
1 1352.25874817298541 -728.59470627870394 99.88577893912608
2 256.41755087905369 540.45705617766782 -1266.13454445487400
3 -664.22554634330822 52.88992301133894 -1447.68880617887226
4 665.97061561088674 -703.28203877155886 326.15047919818772
5 -2584.56101426103896 -1829.77804737415295 3160.58005635603467
6 3918.72394791043735 1179.87720727330861 4370.91410885054393
7 155.56889297786475 239.57068418983405 -1496.93659562166590
8 2669.22704493872197 -4042.77189068729103 834.78218560044604
9 -6238.44086737751331 -63.90159903239377 -2795.74491640869155
10 469.06062749191170 5355.53341149195148 -1785.80774628023255
MD.Init.Velocity>
<MD.Init.Velocity.Prev
1 1399.99981891087145 -688.63751722073675 172.71843051112347
2 398.94153053952635 582.49002539145636 -1212.26369083794839
3 -560.40480518371294 137.55936062153032 -1369.05689341668267
4 733.02283465224957 -656.07656663707189 408.88811335170800
5 -2364.16518707952264 -1616.05745703155571 3303.75939351931402
6 4018.99613238830852 1182.27148737639618 4111.56162860108361
7 134.30820730277063 223.02925018561785 -1445.18925154363228
8 2181.30342929107974 -3950.38168644083817 1024.06632945433626
9 -5844.39301171041643 -155.50289567475406 -2855.31273676088631
10 204.38591621153648 5131.53773484659632 -1809.24075190004055
MD.Init.Velocity.Prev>
NH.R 0.000009511132209
NH.nzeta 0.000001383174729
NH.czeta 0.000002845255094

NH.R -0.000001979427717
NH.nzeta -0.000000239401917
NH.czeta -0.000000152041857

------------------------------------------------------------------------

The MD.Init.Velocity stuff looks OK (it is updated), but the NH.* block is strange, the last NH.* block is the same as from the first continuation file (I'm not 100% sure how the behavior is when some key is defined multiple times, but I expect the last one to take effect), so probably if I restart from this, the correct internal state of the thermostat won't be read properly. So I'm not sure if doing multiple continuation runs like this is supported or not, but it would be very nice to have it working.
メンテ
Page: [1]

Re: Problems with NH MD and multiple continuation runs ( No.1 )
Date: 2021/08/07 11:32
Name: T. Ozaki  <t-ozaki@issp.u-tokyo.ac.jp>

Hi,

Thank you very much for your report.
If a keyword is multiply provided in an input file, the first one is used for the calculation.
So, there is no effect by the current format of the dat## file.
However, the format should be corrected since it apparently makes users confused as you mentioned.
So, I am going to fix the problem in an upcoming patch.

Regards,

TO
メンテ
Re: Problems with NH MD and multiple continuation runs ( No.2 )
Date: 2021/08/08 03:49
Name: Pavel Ondracka  <pavel.ondracka@email.cz>

Thanks for the clarification, good to hear there is no issue. And yes, I believe that if it can be possible to not generate some keywords multiple times than it would definitely make the continuation .dat##, .dat### and so on files much easier to read (right now there gets really a lot of stuff appended eventually, so its tricky to read/understand).

Besides the NH.* stuff and duplicate MD.Init.Velocity and MD.Init.Velocity.Prev, what also shows a lot are duplicite "Atoms.UnitVectors", "NPT.WV.F0", and "Atoms.Unitvectors.Velocity" keys.

BTW the "NPT.WV.F0" and "Atoms.Unitvectors.Velocity" are not even documented in the manual and always contain just zeros in my dat## files, so if zero is the default maybe those keywords can be omitted completely as an another possible improvement/cleanup?
メンテ
Re: Problems with NH MD and multiple continuation runs ( No.3 )
Date: 2021/08/08 06:40
Name: T. Ozaki

Hi,

> Besides the NH.* stuff and duplicate MD.Init.Velocity and MD.Init.Velocity.Prev, what also shows
> a lot are duplicite "Atoms.UnitVectors", "NPT.WV.F0", and "Atoms.Unitvectors.Velocity" keys.

The duplication will also be corrected in an upcoming patch.

> BTW the "NPT.WV.F0" and "Atoms.Unitvectors.Velocity" are not even documented in the manual
> and always contain just zeros in my dat## files, so if zero is the default maybe those keywords
> can be omitted completely as an another possible improvement/cleanup?

They will be used in MD simulations with NPT ensemble which will be released in Ver. 4.0.

Regards,

TO
メンテ

Page: [1]

Thread Title (must) Move the thread to the top
Your Name (must)
E-Mail (must)
URL
Password (used in modification of the submitted text)
Comment (must)

   Save Cookie