Thermalisation period in training MD data #1997
-
|
Sorry for this very simple question! In MD simulations, there is usually a Thermalisation period (or annealing) which is discarded before one postprocesses the data (e.g., for pressure analysis, mean temperature, etc.) Depending on the system and pre-thermalisation, this can be a few hundred steps or 2-5 pico seconds. Should this data also be discarded for deepmd training purposes or is it safe to include it? |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 1 reply
-
|
The DP model should be able to thermalise the system, thus including the thermalising part of the AIMD trajectory would be helpful. The capacity of the DP model with default parameters is large enough so you can safely include the thermalising data without loss of accuracy. |
Beta Was this translation helpful? Give feedback.
-
|
Thank you very much! |
Beta Was this translation helpful? Give feedback.
The DP model should be able to thermalise the system, thus including the thermalising part of the AIMD trajectory would be helpful.
The capacity of the DP model with default parameters is large enough so you can safely include the thermalising data without loss of accuracy.