Have three questions about model_devi section

[9Jaden]

1. Knowing that the model_devi job is completed by LAMMPS. I want to know what is the pair_style adopted in it? Is it the deep potential gained from the init_data?
2. I have been studying in surface system and need to fix the bottom layer of the configuration. The Selective Dynamics in POSCAR didn't work, so I was wondering how to fix the bottom layer during the model_devi section?
3. Besides, I have got a frozen model from DeePMD-kit, but when I want to do some further simulation job by LAMMPS with it. I always received the error below:

ERROR on proc 0: Neighbor list overflow, boost neigh_modify one (src/npair_full_bin_atomonly.cpp:89)
Last command: run 100000
application called MPI_Abort(MPI_COMM_WORLD, 1) - process 0
[unset]: write_line error; fd=-1 buf=:cmd=abort exitcode=1
:
system msg for write_line failure : Bad file descriptor

And in.lammps:

units metal
boundary p p p
atom_style atomic

neighbor 1.0 bin
neigh_modify every 10 delay 0 check no page 50000 one 5000

read_data C24.data
mass 1 64
mass 2 12

pair_style deepmd graph.pb
pair_coeff * *

velocity all create 1300.0 23456789
fix 1 all nvt temp 1300.0 1300.0 0.5
timestep 0.001

thermo_style custom step pe ke etotal temp press vol
thermo 100
dump 1 all custom 100 c18.dump id type x y z

run 100000

Also data file attached:
C24data.md
Want to know how to fix this error.

Great thanks for any help~

[HuangJiameng]

I want to know what is the pair_style adopted in it

pair_style is set as deepmd by default.

Is it the deep potential gained from the init_data

I do not know what you mean

2. There're many discussions about how to fix layers in LAMMPS's community, such as https://sourceforge.net/p/lammps/mailman/message/32396695/ . Maybe you can search for help there.

3. The reason for the failure is that there're too many atoms in one atom's neighbor (just like the log demonstrates). Maybe you should check your settings for LAMMPS, or check the quality of your Deep Potential Model by doing some validation tests.

[9Jaden]

Thanks. I think i know the answer of question 1

[taipinghu]

for your question 2, there are two ways to fix this issue.
(1) change the code of the dpgen (your-dpgen-path/dpgen/generator/lib/lammps.py) and add the fix butter layer code by setting velocity and force as 0.
(2) you can prepare a lammps input and set template in your param.json file.

[taipinghu]

you mean you want to fix some in POSCAR file? in dpgen iteration, VASP single point calculation will be performed. thus, fix atoms is not meaningful.

[9Jaden]

I want the configurations after model_devi have the bottom layer fixed. And the configurations are derived from the POSCAR I set. Now the question is how to change the code of lammps.py of my dpgen to only fix the bottom layer, since the POSCAR type of file can not lable out the bottom layer atoms.

[taipinghu]

I want the configurations after model_devi have the bottom layer fixed. And the configurations are derived from the POSCAR I set. Now the question is how to change the code of lammps.py of my dpgen to only fix the bottom layer, since the POSCAR type of file can not lable out the bottom layer atoms.

fix atom according to the atomic index since the atomic ids cannot be changed during the whole process.

[9Jaden]

but for different POSCARs I used, the number of total atoms changes so that the bottom layers' indexes change too.

[taipinghu]

but for different POSCARs I used, the number of total atoms changes so that the bottom layers' indexes change too.

OK, I see. You should change the code of the dpgen. you can insert some code in dpgenpath/generator/run.py. when the make_model_devi function finish, replace fix group atomic ids in in.lammps file.