How to use machine learning potentials to track the change in species numbers during a reaction process.

[jzh-w]

How can a machine learning potential trained with deepmd-kit be used to track the change in the number of intermediates during a reaction process, similar to how reactive force fields like ReaxFF do?

[njzjz]

The trajectory can be processed by ReacNetGenerator, https://github.com/deepmodeling/reacnetgenerator

[jzh-w]

I am using the server configured by my research group, not a supercomputer. Additionally, when using my own machine, I noticed that the CPU cores were not fully utilized.

[njzjz]

I suggest using a machine with multiple cores and enough memory and waiting for a night.

[jzh-w]

Okay, thank you! I will follow your suggestion and try again. Since I noticed during yesterday's run that the CPU cores were not fully utilized, even though the ReacNetGenerator documentation states that -np defaults to using all available processes, I want to set it manually. However, I am concerned that I might input the command incorrectly. So could you please check if the command I wrote can achieve the desired effect:

reacnetgenerator -i relax.lammpstrj -a B O -np 32 --type dump --nohmm

Once again, I would like to express my gratitude to you!

[njzjz]

reacnetgenerator -i relax.lammpstrj -a B O -np 32 --type dump --nohmm

This looks good to me.

[jzh-w]

Thank you once again! Recently, I read the content you published on "Zhihu", and it left a deep impression on me. I wish you an even brighter future!Thank you!