You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
demands:
the inner/outer steps of lammps can be correctly distinguished when recording the model_devi, if the exploration scheme involves simulation schemes requesting 【"M steps of A (inner steps) every N steps of B (outer steps)"】.
where to enhance:
interface with lammps 【USER-DEEPMD】
Detailed Description
current behavior :
All M inner-step configurations would be checked if the (nN+1) outer step is checked, but the "step" recorded in model_devi.out is always the outer step,
consequently, there could be more than M repeated “(nN+1)” step in model_devi.out, each could correspond to an A-operation-changed configuration that gives distinct model_devi values.
Then, masses pile up that, only the outer-step configuration with the unique outer-step number would be dumped by lammps, thus may lead to the mismatch between the model_devi value and the configuration.
e.g. 1 MC every 1 MD with the check frequency of 20:
After discussion, we decided to reject this request.
There is no "outer-step" in LAMMPS. For the same step, LAMMPS only dumps the first configuration. However, only processing the first one conflicts with reset_timestep when one repeatedly performs reset_timestep 0. LAMMPS throws errors when there is any active dump but does nothing with pair_style.
Summary
demands:
the inner/outer steps of lammps can be correctly distinguished when recording the model_devi, if the exploration scheme involves simulation schemes requesting 【"M steps of A (inner steps) every N steps of B (outer steps)"】.
where to enhance:
interface with lammps 【USER-DEEPMD】
Detailed Description
current behavior :
All M inner-step configurations would be checked if the (nN+1) outer step is checked, but the "step" recorded in model_devi.out is always the outer step,
consequently, there could be more than M repeated “(nN+1)” step in model_devi.out, each could correspond to an A-operation-changed configuration that gives distinct model_devi values.
Then, masses pile up that, only the outer-step configuration with the unique outer-step number would be dumped by lammps, thus may lead to the mismatch between the model_devi value and the configuration.
e.g. 1 MC every 1 MD with the check frequency of 20:
Further Information, Files, and Links
The text was updated successfully, but these errors were encountered: