Atoms in lammps don't move

[psn417]

Hello, I get a .pb file from dpgen, an use it to run a lammps simulation, but the atoms don't move in every frame. This is trained by smooth version.
When I don't use smooth version, lammps run correctly.
Is it a bug or I make some mistakes?

this is my input file:

dimension 3
boundary p p p
units metal
atom_style atomic

read_data pos
mass * 23

pair_style deepmd graph.pb
pair_coeff

timestep 0.001

dump 2 all atom 10 2000.atom

fix 1 all nvt temp 1 2000 0.1
run 2500
unfix 1

fix 2 all nvt temp 2000 2000 0.1
run 7500
unfix 2

S. Pan

[psn417]

dpgen gives me 4 .pb files, I should use them together or use a single one?

[njzjz]

Hi, Can you dump forces in lammps and see whether forces are zero? ———————— Jinzhe Zeng PhD student in Chemistry Rutgers University jinzhe.zeng@rutgers.edu CV: https://njzjz.win/cv

…

On Feb 20, 2020, at 08:37, psn417 ***@***.***> wrote:  Hello, I get a .pb file from dpgen, an use it to run a lammps simulation, but the atoms don't move in every frame. This is trained by smooth version. When I don't use smooth version, lammps run correctly. Is it a bug or I make some mistakes? this is my input file: dimension 3 boundary p p p units metal atom_style atomic read_data pos mass * 23 pair_style deepmd graph.pb pair_coeff timestep 0.001 dump 2 all atom 10 2000.atom fix 1 all nvt temp 1 2000 0.1 run 2500 unfix 1 fix 2 all nvt temp 2000 2000 0.1 run 7500 unfix 2 S. Pan — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or unsubscribe.

[jameswind]

Also, can you report the training performance? Say, the training and testing error, etc. On Thu, Feb 20, 2020 at 12:52 PM Jinzhe Zeng <notifications@github.com> wrote:

…

Hi, Can you dump forces in lammps and see whether forces are zero? ———————— Jinzhe Zeng PhD student in Chemistry Rutgers University ***@***.*** CV: https://njzjz.win/cv > On Feb 20, 2020, at 08:37, psn417 ***@***.***> wrote: > >  > Hello, I get a .pb file from dpgen, an use it to run a lammps simulation, but the atoms don't move in every frame. This is trained by smooth version. > When I don't use smooth version, lammps run correctly. > Is it a bug or I make some mistakes? > > this is my input file: > > dimension 3 > boundary p p p > units metal > atom_style atomic > > read_data pos > mass * 23 > > pair_style deepmd graph.pb > pair_coeff > > timestep 0.001 > > dump 2 all atom 10 2000.atom > > fix 1 all nvt temp 1 2000 0.1 > run 2500 > unfix 1 > > fix 2 all nvt temp 2000 2000 0.1 > run 7500 > unfix 2 > > S. Pan > > — > You are receiving this because you are subscribed to this thread. > Reply to this email directly, view it on GitHub, or unsubscribe. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <https://github.com/deepmodeling/deepmd-kit/issues/190?email_source=notifications&email_token=AEJ6DCYG6VOSOXTSURNL3LDRD27MTA5CNFSM4KYO7NLKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEMPNMJY#issuecomment-589223463>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AEJ6DC3LOZMWF35Z6UCROG3RD27MTANCNFSM4KYO7NLA> .

[psn417]

Thank you! I find that I must add "velocity create" command in lammps script to let the atoms move. It's strange that I can use lammps without this command when I don't use smooth version.
I also wonder which of the 4 .pb files should I use?

[njzjz]

This behavior is weird... Can you ensure forces are not zero? The velocity command only gives an initial velocity. The usage of 4 files is to print their standard deviations. You can see the article of dpgen for details: https://doi.org/10.1016/j.cpc.2020.107206 -- Jinzhe Zeng PhD student in Chemistry Rutgers University jinzhe.zeng@rutgers.edu CV: https://njzjz.win/cv

…

On Fri, Feb 21, 2020 at 7:54 AM psn417 ***@***.***> wrote: Thank you! I find that I must add "velocity create" command in lammps script to let the atoms move. It's strange that I can use lammps without this command when I don't use smooth version. I also wonder which of the 4 .pb files should I use? — You are receiving this because you commented. Reply to this email directly, view it on GitHub <https://github.com/deepmodeling/deepmd-kit/issues/190?email_source=notifications&email_token=ACIOQ7QIYOR5NCVQEENGLX3RD7FIZA5CNFSM4KYO7NLKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEMSTOBI#issuecomment-589641477>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACIOQ7VYBH3WSFMIBTTVWOTRD7FIZANCNFSM4KYO7NLA> .