"init_data_sys" and "sys_configs" in dpgen (lammps-cp2k) example param.json file

[longkunxuluke]

Hi, I am trying to use dpgen working with LAMMPS and CP2K (without VASP and others). I follow this example but not sure how to properly set the two parameters in the title.

Instead of using dpgen init, my initial data was produced using dpdata to process CP2K AIMD output following this tutorial. Should I put the directory containing the generated deepmd/npy files in "init_data_sys"? Should I randomly pick 4 configurations from AIMD trajectory and convert them to POSCAR files and put their directory in "sys_configs"?

I am using a local workstation, without using VASP and dpgen init, I tried looking for example param.json and machine.json specifically for this scenario but have not found. Any help would be appreciated.

[HuangJiameng]

https://docs.deepmodeling.com/projects/dpgen/en/devel/user-guide/troubleshooting.html Hope it is helpful.

[njzjz]

Could you provide machine.json?

[longkunxuluke]

Please check the attached, thanks
json.zip

[njzjz]

The file is correct, but you do not use the latest version.

[longkunxuluke]

Sorry did you mean the latest version of dpgen, deepmd-kit, or the formats in my json files?

[njzjz]

decide_machine.py is different from that on the GitHub.

dpgen/dpgen/remote/decide_machine.py

Line 39 in 8dea29e

if isinstance(mdata[task_type], dict):

[njzjz]

sys_configs is the starting structures for MD simulations. You can set another sys_format if you don't want POSCAR.

Supported format can be found in the dpdata documentation.

[chenyuping6]

I also want to ask, how to select structures from AIMD trajectories and convert them into POSCAR files to place in "sys_configs"?