ML LFAST: Difference between revisions

From VASP Wiki
No edit summary
No edit summary
 
(8 intermediate revisions by 2 users not shown)
Line 1: Line 1:
{{DISPLAYTITLE:ML_LFAST}}
{{DISPLAYTITLE:ML_LFAST}}
{{TAGDEF|ML_LFAST|[logical]|.FALSE.}}
{{TAGDEF|ML_LFAST|[logical]}}
{{DEF|ML_LFAST|.TRUE.|for {{TAGO|ML_MODE|refit}}|.FALSE.|else}}


Description: This tag switches on the descriptors for refitting in the fast execution mode within machine learning force fields.
Description: This tag switches on the descriptors for refitting in the fast execution mode within machine learning force fields. This tag is usually switched on by {{TAGO|ML_MODE|refit}} and doesn't need to be set explicitely.
{{NB|mind|This tag is only available as of VASP.6.4.0}}
{{NB|mind|This tag is only available as of VASP.6.4.0}}
----
----
This tag switches on the descriptors for the fast execution mode. To use the fast execution mode the force field first has to be refit with the fast descriptors. As usual, the resulting {{TAG|ML_FFN}} has to be copied to {{TAG|ML_FF}} and the fast code will be automatically run in production mode runs by setting ({{TAG|ML_MODE}}=RUN or {{TAG|ML_ISTART}}=2).
This tag switches on the descriptors for the fast execution mode. To use the fast execution mode the force field first has to be refit with the fast descriptors. This is done by setting {{TAGO|ML_MODE|refit}}, which will automatically set {{TAG|ML_LFAST|.TRUE.}}. As usual, the resulting {{TAG|ML_FFN}} has to be copied to {{TAG|ML_FF}} and the fast code will be automatically run in production mode runs by setting {{TAG|ML_MODE|run}}.
 
To retrain the force field set the following:
*{{TAG|ML_MODE}}=REFIT
or
*{{TAG|ML_LFAST}}=.TRUE.
*{{TAG|ML_ISTART}}=4


The speedup of the fast method compared to the regular method is increasing with increasing number of local reference configurations.
The speedup of the fast method compared to the regular method is increasing with increasing number of local reference configurations.
Line 17: Line 12:
It should be noted that in the fast version no Bayesian error estimation is available.  
It should be noted that in the fast version no Bayesian error estimation is available.  


{{TAG|ML_LFAST}} sets the defaults {{TAG|ML_IALGO_LINREG}}=4, which is fitting with singular value decomposition and regularization ({{TAG|ML_SIGW0}} and {{TAG|ML_SIGV0}}).
Since the calculation time of the fast version is of the same order of magnitude as the timing for the output of the molecular-dynamics results, we advise decreasing the output frequency for molecular dynamics by the tag {{TAG|ML_OUTBLOCK}}. Additionally, the calculation and output of the pair-correlation function can be suppressed by {{TAGO|ML_OUTPUT_MODE|0}} as default. {{NB|deprecated|If {{TAGO|ML_LFAST|.TRUE.}} is explicitely used without the {{TAG|ML_MODE}} tag then it can only be used with {{TAGO|ML_ISTART|4}} and {{TAGO|ML_IALGO_LINREG|1|op=>}}.}}
 
{{TAG|ML_LFAST}} cannot be used with {{TAG|ML_IALGO_LINREG}}<3.
 
Since the calculation time of the fast version is of the same order of magnitude as the timing for the output of the molecular-dynamics results, we advise decreasing the output frequency for molecular dynamics. This is controlled by the tag {{TAG|ML_OUTBLOCK}}. By default, it writes out all molecular-dynamics results at every molecular-dynamics step. Additionally, the calculation and output of the pair-correlation function is suppressed by {{TAG|ML_OUTPUT_MODE}}=0 as default.


== Related tags and articles ==
== Related tags and articles ==
{{TAG|ML_LMLFF}}, {{TAG|ML_ISTART}}, {{TAG|ML_IERR}}, {{TAG|ML_OUTBLOCK}}, {{TAG|ML_OUTPUT_MODE}}
{{TAG|ML_LMLFF}}, {{TAG|ML_MODE}}, {{TAG|ML_IERR}}, {{TAG|ML_OUTBLOCK}}, {{TAG|ML_OUTPUT_MODE}}
----
----


[[Category:INCAR tag]][[Category:Machine-learned force fields]]
[[Category:INCAR tag]][[Category:Machine-learned force fields]]

Latest revision as of 15:30, 19 October 2023

ML_LFAST = [logical] 

Default: ML_LFAST = .TRUE. for ML_MODE = refit
= .FALSE. else

Description: This tag switches on the descriptors for refitting in the fast execution mode within machine learning force fields. This tag is usually switched on by ML_MODE = refit and doesn't need to be set explicitely.

Mind: This tag is only available as of VASP.6.4.0

This tag switches on the descriptors for the fast execution mode. To use the fast execution mode the force field first has to be refit with the fast descriptors. This is done by setting ML_MODE = refit, which will automatically set ML_LFAST. As usual, the resulting ML_FFN has to be copied to ML_FF and the fast code will be automatically run in production mode runs by setting ML_MODE.

The speedup of the fast method compared to the regular method is increasing with increasing number of local reference configurations.

It should be noted that in the fast version no Bayesian error estimation is available.

Since the calculation time of the fast version is of the same order of magnitude as the timing for the output of the molecular-dynamics results, we advise decreasing the output frequency for molecular dynamics by the tag ML_OUTBLOCK. Additionally, the calculation and output of the pair-correlation function can be suppressed by ML_OUTPUT_MODE = 0 as default.

Deprecated: If ML_LFAST = .TRUE. is explicitely used without the ML_MODE tag then it can only be used with ML_ISTART = 4 and ML_IALGO_LINREG > 1.

Related tags and articles

ML_LMLFF, ML_MODE, ML_IERR, ML_OUTBLOCK, ML_OUTPUT_MODE