How to make the system reach steady state as soon as possible when run Deep Potential MD using lammps?

[chunweizhu]

I have trained an alloy deep potential model using dpgen and evaluated the model's accuracy.
During the NPT molecular dynamics (MD) simulations using LAMMPS, the alloy system did not reach a steady stateas as soon as possible, despite running 10^6 steps with a duration of 2 fs per step.

Are there any suggestion to accelerate the system to reach a steady state?