Lammps： Warnning ：Radial neighbor list length of type 0 is not enough

[tiandijunhao]

When I use freeze.dp model in lammps, it always has a warning which is " Warnning ：Radial neighbor list length of type 0 is not enough" .

What's the reason, please

[njzjz]

sel in your model, which gives the maximum possible number of neighbors in the cut-off radius, is not enough.

[tiandijunhao]

sel in your model, which gives the maximum possible number of neighbors in the cut-off radius, is not enough.

Thank you very much. Does this mean that I only need to modify the value of sel, keep the rcut at 6A, and then retrain the model? In addition, if I want this model to be able to simulate a large cell in lammps (for example, 10 * 10 * 10, 2000 atoms in total, fcc structure). Is the higher the sel value, the better?

[tiandijunhao]

Thank you very much. Does this mean that I only need to modify the value of sel, keep the rcut at 6A, and then retrain the model? In addition, if I want this model to be able to simulate a large cell in lammps (for example, 10 * 10 * 10, 2000 atoms in total, fcc structure). Is the higher the sel value, the better?

[njzjz]

sel in your model, which gives the maximum possible number of neighbors in the cut-off radius, is not enough.

Thank you very much. Does this mean that I only need to modify the value of sel, keep the rcut at 6A, and then retrain the model? In addition, if I want this model to be able to simulate a large cell in lammps (for example, 10 * 10 * 10, 2000 atoms in total, fcc structure). Is the higher the sel value, the better?

No, the model will become slower if you set it too high. It’s recommended to estimate the maximum atom number in the cutoff radius (in your model is 6 angstroms).

[tiandijunhao]

thank you very much

…

---Original--- From: "Jinzhe Zeng"<notifications@github.com&gt; Date: Wed, Oct 23, 2019 21:49 PM To: "deepmodeling/deepmd-kit"<deepmd-kit@noreply.github.com&gt;; Cc: "tiandijunhao"<1048309667@qq.com&gt;;"Author"<author@noreply.github.com&gt;; Subject: Re: [deepmodeling/deepmd-kit] Lammps： Warnning ：Radial neighbor list length of type 0 is not enough (#108) sel in your model, which gives the maximum possible number of neighbors in the cut-off radius, is not enough. Thank you very much. Does this mean that I only need to modify the value of sel, keep the rcut at 6A, and then retrain the model? In addition, if I want this model to be able to simulate a large cell in lammps (for example, 10 * 10 * 10, 2000 atoms in total, fcc structure). Is the higher the sel value, the better? No, the model will become slower if you set it too high. It’s recommended to estimate the maximum atom number in the cutoff radius (in your model is 6 angstroms). — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.