When dpgen is running, why is there no virial file in the file converted from dpdata?

[12jscvb]

After calculating the single-point energy structure at the end of each cycle, dpdata converts it into npy format files, but only four files, box, energy,coord and force, and no virial file, while dpgen sets stress_tensor in cp2k input file. Besides, when I manually transfer the ouput file, virial.npy will appear. What is the reason for this? Thanks! Best wishes

[robinzyb]

You may not print stresses, although cp2k calculated the stresses.
https://manual.cp2k.org/trunk/CP2K_INPUT/FORCE_EVAL/PRINT/STRESS_TENSOR.html

[robinzyb]

the code works properly, this is exactly the same code used in dpgen.

import dpdata
#cp2kmd_dir = "./"
#cp2kmd_output_name = "new.out"
dp = dpdata.LabeledSystem("./output.txt",fmt="cp2kdata/e_f")
dp.to_deepmd_npy('test',fmt='deepmd-npy')
#print(cp2koutput)
print(dp)

--- You are parsing data using package Cp2kData ---
Atom names are fake chemical symbols as you set in cp2k input.
--- You are parsing data using package Cp2kData ---
Data Summary
Labeled System
-------------------
Frame Numbers      : 1
Atom Numbers       : 255
Including Virials  : Yes
Element List       :
-------------------
Al  N  H  Cl  O
55  20  80  20  80

have no idea where goes wrong. Could you try to make a new folder, for example test_dpgen

and copy the files of iteration.00000 to the new folder, as well as some dpgen inputs.
Then modify the dpgen.record to

0 0
0 1
0 2
0 3
0 4
0 5
0 6
0 7

and restart the dpgen.
See if the dpgen can correctly genreate virial.npy.

[12jscvb]

Ok, this is also the case when I used dpgen in the Bohrium platform before. The dpgen processing file does not contain virial. In addition, the initial data set contains virial information while the dpgen iteration process does not contain virial information. Does this affect the model?

[robinzyb]

It depends whether the npt simulations are used. If prediction of pressure is necessary, the answer is yes. If not, just ignore it. You wouldn't use it in the deepmd fitting.

[12jscvb]

thank Professor for reply. Benefit a lot

[robinzyb]

Has your problem been solved? Does the virial.npy file appear in your new DP-GEN run?

[12jscvb]

Unfortunately, Virial.npy still does not appear in new DP-GEN run 

…

---Original--- From: "Yongbin ***@***.***> Date: Tue, Feb 6, 2024 06:25 AM To: ***@***.***>; Cc: "Express ***@***.******@***.***>; Subject: Re: [deepmodeling/dpdata] When dpgen is running, why is there novirial file in the file converted from dpdata? (Discussion #600) Has your problem been solved? Does the virial.npy file appear in your new DP-GEN run? — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***>

[robinzyb]

It is unusual.
Are the dpgen and dpdata and cp2kdata installed in the same environment? If so, it should be ok.

[12jscvb]

I am using cp2k2023 version, will it be relevant. In addition, I found no cp2kdata installation package by pip list command, but dpgen automatically converted data without error (dpgen related packages were installed by pip install deepmd-kit command)

…

---Original--- From: "Yongbin ***@***.***> Date: Wed, Feb 21, 2024 15:49 PM To: ***@***.***>; Cc: "Express ***@***.******@***.***>; Subject: Re: [deepmodeling/dpdata] When dpgen is running, why is there novirial file in the file converted from dpdata? (Discussion #600) It is unusual. Are the dpgen and dpdata and cp2kdata installed in the same environment? If so, it should be ok. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***>

[12jscvb]

thank you relpy. Your meanings is that in the first few loops the temp parameter in the param.json file is set like "200 300 500", and as the potential function is improved, then the temp temperature can be increased, for example, set like "500 700 900". Also, do I need to re-run an aimd of cp2k at 200/300/500k temperature to get the initial data set?

[robinzyb]

"do I need to re-run an aimd of cp2k at 200/300/500k temperature to get the initial data set?" not necessary

[12jscvb]

ok,If the initial data set I provide is large enough,such as up to 1,000 or more, could I start with nvt of high temperature like 500/1000/1500k?

[robinzyb]

The size of initial data set doesn't matter but the quality of your potetial does

[12jscvb]

Thank you very much for your answer. Benefit a lot