search

deepmodeling / deepmd-kit

A deep learning package for many-body potential energy representation and molecular dynamics
https://docs.deepmodeling.com/projects/deepmd/
GNU Lesser General Public License v3.0
1.45k stars 499 forks source link

[BUG] _Unreasonable water structure when the lammps run time is over 1 ns #1099

Closed wangzhe86 closed 3 years ago

wangzhe86 commented 3 years ago

Unreasonable water structure when the lammps run time is over 1 ns

Deepmd-kit 2.0, .sh files, input file

I trained the example (/example/water/se_e2_a), and got the graph.pb Then, I copied the graph.pd to the lmp files (/water/lmp) to test the potential parameter. The lammps script can be run (NVT ensemble). I got a water.dump file, but the structure is unreasonable when the run time is over 1 ns. I found that the temperature suddenly rose from 330K to 4000K. The O and H are aggregated. Why does this problem occur?

image image
njzjz commented 3 years ago

The size of training data in examples directory is very small, so it's not surprising to find that the model is not good.