ML MB - Revision history

Singraber at 07:56, 24 October 2025

2025-10-24T07:56:46Z

Singraber at 17:11, 15 February 2024

2024-02-15T17:11:19Z

← Older revision		Revision as of 17:11, 15 February 2024
Line 5:		Line 5:
	----		----
	The defaults for {{TAG\|ML_MB}} are different for each different {{TAG\|ML_MODE}} setting. Here are the defaults for each mode:		The defaults for {{TAG\|ML_MB}} are different for each different {{TAG\|ML_MODE}} setting. Here are the defaults for each mode:
	*{{~~TAG~~\|ML_MODE}}~~='TRAIN'~~:		*{{TAGO\|ML_MODE\|train}}:
	*No {{TAG\|ML_AB}} present (learning from scratch): min(1500 , max({{TAG\|NSW}} , 2{{TAG\|ML_MCONF_NEW}} * MAXAT_SP)		*No {{TAG\|ML_AB}} present (learning from scratch): <code>min(1500 , max({{TAG\|NSW}} , 2{{TAG\|ML_MCONF_NEW}} * MAXAT_SP)</code>
	*{{TAG\|ML_AB}} present (continuation of learning): MB_AB + min(1500, max({{TAG\|NSW}} , 2{{TAG\|ML_MCONF_NEW}} * MAXAT_SP)		*{{TAG\|ML_AB}} present (continuation of learning): <code>MB_AB + min(1500, max({{TAG\|NSW}} , 2{{TAG\|ML_MCONF_NEW}} * MAXAT_SP)</code>
	{{~~TAG~~\|ML_MODE}}~~='SELECT'~~: MB_AB + {{TAG\|ML_MCONF_NEW}} MAXAT_SP		{{TAGO\|ML_MODE\|select}}: <code>MB_AB + {{TAG\|ML_MCONF_NEW}} MAXAT_SP</code>
	*{{~~TAG~~\|ML_MODE}}~~='REFIT'~~: MB_AB + MAXAT_SP		*{{TAGO\|ML_MODE\|refit}}: <code>MB_AB + MAXAT_SP</code>
	*{{~~TAG~~\|ML_MODE}}~~='REFITBAYESIAN'~~: MB_AB + MAXAT_SP		*{{TAGO\|ML_MODE\|refitbayesian}}: <code>MB_AB + MAXAT_SP</code>
	*{{~~TAG~~\|ML_MODE}}~~='RUN'~~: MB_AB		*{{TAGO\|ML_MODE\|run}}: <code>MB_AB</code>
	using the following definitions:\|		using the following definitions:
	*MAXAT_SP = greatest number of atoms within all species among the current structures and the structures in the {{TAG\|ML_AB}} file (if present).		*<code>MAXAT_SP</code> = greatest number of atoms within all species among the current structures and the structures in the {{TAG\|ML_AB}} file (if present).
	*MB_AB = greatest number of local reference configurations within all species in the {{TAG\|ML_AB}} file.		*<code>MB_AB</code> = greatest number of local reference configurations within all species in the {{TAG\|ML_AB}} file.

	The default value for {{~~TAG~~\|ML_MODE}} ~~= ''TRAIN''~~ and {{~~TAG~~\|ML_MODE}} ~~= ''SELECT''~~ is a relatively safe value for most materials. However one might need to increase it to a greater value for liquids, polymers and amorphous systems, or when a MLFF for many different polytypes is trained. By default ({{~~TAG~~\|ML_LBASIS_DISCARD~~}}=''~~.TRUE.''), if the number of local reference configurations would exceed {{TAG\|ML_MB}}, VASP would continue the calculation and disposes of these, but one should make extensive tests whether the generated MLFF is sufficiently accurate.		The default value for {{TAGO\|ML_MODE\|train}} and {{TAGO\|ML_MODE\|select}} is a relatively safe value for most materials. However one might need to increase it to a greater value for liquids, polymers and amorphous systems, or when a MLFF for many different polytypes is trained. By default ({{TAGO\|ML_LBASIS_DISCARD\|.TRUE.}}), if the number of local reference configurations would exceed {{TAG\|ML_MB}}, VASP would continue the calculation and disposes of these, but one should make extensive tests whether the generated MLFF is sufficiently accurate.

	If VASP stops, subsequent training can be restarted from the existing ML_AB file. This avoids loss of already acquired training data.		If VASP stops, subsequent training can be restarted from the existing ML_AB file. This avoids loss of already acquired training data.

	Depending on the calculation mode VASP adds a little overhead in the allocation of {{TAG\|ML_MB}} arrays to buffer for new candidate structures. Here are the buffer sizes for each mode:		Depending on the calculation mode VASP adds a little overhead in the allocation of {{TAG\|ML_MB}} arrays to buffer for new candidate structures. Here are the buffer sizes for each mode:
	*{{~~TAG~~\|ML_MODE}}~~='TRAIN'~~:		*{{TAGO\|ML_MODE\|train}}:
	*No {{TAG\|ML_AB}} present (learning from scratch): min({{TAG\|NSW}} , {{TAG\|ML_MCONF_NEW}}) MAXAT_SP		*No {{TAG\|ML_AB}} present (learning from scratch): <code>min({{TAG\|NSW}} , {{TAG\|ML_MCONF_NEW}}) MAXAT_SP</code>
	*{{TAG\|ML_AB}} present (continuation of learning): min({{TAG\|NSW}} , {{TAG\|ML_MCONF_NEW}}) MAXAT_SP		*{{TAG\|ML_AB}} present (continuation of learning): <code>min({{TAG\|NSW}} , {{TAG\|ML_MCONF_NEW}}) MAXAT_SP</code>
	{{~~TAG~~\|ML_MODE}}~~='SELECT'~~: {{TAG\|ML_MCONF_NEW}} MAXAT_SP		{{TAGO\|ML_MODE\|select}}: <code>{{TAG\|ML_MCONF_NEW}} MAXAT_SP</code>
	*{{~~TAG~~\|ML_MODE}}~~='REFIT'~~: MAXAT_SP		*{{TAGO\|ML_MODE\|refit}}: <code>MAXAT_SP</code>
	*{{~~TAG~~\|ML_MODE}}~~='REFITBAYESIAN'~~: MAXAT_SP		*{{TAGO\|ML_MODE\|refitbayesian}}: <code>MAXAT_SP</code>
	*{{~~TAG~~\|ML_MODE}}~~='RUN'~~: 0		*{{TAGO\|ML_MODE\|run}}: <code>0</code>

	The flag {{TAG\|ML_MB}} also determines the maximum number of rows of the design matrix. This is usually a huge matrix. The design matrix is allocated statically at the beginning of the run, since several parts of the code use MPI shared memory and dynamic reallocation of these arrays can cause issues on many operating systems. An estimate of the size of the design matrix and all other large arrays is printed out to the {{TAG\|ML_LOGFILE}} before allocation. The design matrix is fully distributed in a block-cyclic fashion for scaLAPACK, and thus the memory requirement for each core scales inversely proportional to the number of used cores.

			The flag {{TAG\|ML_MB}} also determines the maximum number of rows of the design matrix. This is usually a huge matrix. The design matrix is allocated statically at the beginning of the run, since several parts of the code use MPI shared memory and dynamic reallocation of these arrays can cause issues on many operating systems. An estimate of the size of the design matrix and all other large arrays is printed out to the {{TAG\|ML_LOGFILE}} before allocation. The design matrix is fully distributed in a block-cyclic fashion for scaLAPACK, and thus the memory requirement for each core scales inversely proportional to the number of used cores.

	== Related tags and articles ==		== Related tags and articles ==

Karsai at 09:11, 14 April 2023

2023-04-14T09:11:22Z

← Older revision		Revision as of 09:11, 14 April 2023
Line 1:		Line 1:
	{{DISPLAYTITLE:ML_MB}}		{{DISPLAYTITLE:ML_MB}}
	{{TAGDEF\|ML_MB\|[integer]\|~~1500~~}}		{{TAGDEF\|ML_MB\|[integer]\|see below}}

	Description: This tag sets the maximum number of local reference configurations (i.e. basis functions in the kernel) in the machine learning force field method.		Description: This tag sets the maximum number of local reference configurations (i.e. basis functions in the kernel) in the machine learning force field method.
	----		----
	The default value is a relatively safe value for most materials. However one might need to increase it to a greater value for liquids, polymers and amorphous systems, or when a MLFF for many different polytypes is trained. ~~When the number of local reference configurations exceeds this value, VASP stops and prompts an error instruction to increase~~ {{TAG\|~~ML_MB~~}}. ~~The user can also force VASP to dispose~~ of local reference configurations ~~(basis functions in the kernel) by specifying the tag~~ {{TAG\| ~~ML_LBASIS_DISCARD~~}}~~. In this case~~, VASP ~~will not stop~~, but one should make extensive tests whether the generated MLFF is sufficiently accurate.		The defaults for {{TAG\|ML_MB}} are different for each different {{TAG\|ML_MODE}} setting. Here are the defaults for each mode:
			*{{TAG\|ML_MODE}}='TRAIN':
			*No {{TAG\|ML_AB}} present (learning from scratch): min(1500 , max({{TAG\|NSW}} , 2{{TAG\|ML_MCONF_NEW}} * MAXAT_SP)
			*{{TAG\|ML_AB}} present (continuation of learning): MB_AB + min(1500, max({{TAG\|NSW}} , 2{{TAG\|ML_MCONF_NEW}} * MAXAT_SP)
			{{TAG\|ML_MODE}}='SELECT': MB_AB + {{TAG\|ML_MCONF_NEW}} MAXAT_SP
			*{{TAG\|ML_MODE}}='REFIT': MB_AB + MAXAT_SP
			*{{TAG\|ML_MODE}}='REFITBAYESIAN': MB_AB + MAXAT_SP
			*{{TAG\|ML_MODE}}='RUN': MB_AB
			using the following definitions:\|
			*MAXAT_SP = greatest number of atoms within all species among the current structures and the structures in the {{TAG\|ML_AB}} file (if present).
			*MB_AB = greatest number of local reference configurations within all species in the {{TAG\|ML_AB}} file.

			The default value for {{TAG\|ML_MODE}} = ''TRAIN'' and {{TAG\|ML_MODE}} = ''SELECT'' is a relatively safe value for most materials. However one might need to increase it to a greater value for liquids, polymers and amorphous systems, or when a MLFF for many different polytypes is trained. By default ({{TAG\|ML_LBASIS_DISCARD}}=''.TRUE.''), if the number of local reference configurations would exceed {{TAG\|ML_MB}}, VASP would continue the calculation and disposes of these, but one should make extensive tests whether the generated MLFF is sufficiently accurate.

	If VASP stops, subsequent training can be restarted from the existing ML_AB file. This avoids loss of already acquired training data.		If VASP stops, subsequent training can be restarted from the existing ML_AB file. This avoids loss of already acquired training data.

	The flag {{TAG\|ML_MB}} also determines the maximum number of rows of the design matrix. This is usually a huge matrix. The design matrix is allocated statically at the beginning of the run, since several parts of the code use MPI shared memory and dynamic reallocation of these arrays can cause issues on many operating systems. An estimate of the size of the design matrix and all other large arrays is printed out to the {{TAG\|ML_LOGFILE}} before allocation. The design matrix is fully distributed in a block cyclic fashion for scaLAPACK, and thus the memory ~~requirements scale inverse proportionally~~ to the number of used ~~processors~~.		Depending on the calculation mode VASP adds a little overhead in the allocation of {{TAG\|ML_MB}} arrays to buffer for new candidate structures. Here are the buffer sizes for each mode:
			*{{TAG\|ML_MODE}}='TRAIN':
			*No {{TAG\|ML_AB}} present (learning from scratch): min({{TAG\|NSW}} , {{TAG\|ML_MCONF_NEW}}) MAXAT_SP
			*{{TAG\|ML_AB}} present (continuation of learning): min({{TAG\|NSW}} , {{TAG\|ML_MCONF_NEW}}) MAXAT_SP
			{{TAG\|ML_MODE}}='SELECT': {{TAG\|ML_MCONF_NEW}} MAXAT_SP
			*{{TAG\|ML_MODE}}='REFIT': MAXAT_SP
			*{{TAG\|ML_MODE}}='REFITBAYESIAN': MAXAT_SP
			*{{TAG\|ML_MODE}}='RUN': 0

			The flag {{TAG\|ML_MB}} also determines the maximum number of rows of the design matrix. This is usually a huge matrix. The design matrix is allocated statically at the beginning of the run, since several parts of the code use MPI shared memory and dynamic reallocation of these arrays can cause issues on many operating systems. An estimate of the size of the design matrix and all other large arrays is printed out to the {{TAG\|ML_LOGFILE}} before allocation. The design matrix is fully distributed in a block-cyclic fashion for scaLAPACK, and thus the memory requirement for each core scales inversely proportional to the number of used cores.


	== Related tags and articles ==		== Related tags and articles ==
	{{TAG\|ML_LMLFF}}, {{TAG\|ML_MCONF}}, {{TAG\| ML_LBASIS_DISCARD}}		{{TAG\|ML_LMLFF}}, {{TAG\|ML_MODE}}, {{TAG\|ML_MCONF_NEW}}, {{TAG\|ML_MCONF}}, {{TAG\| ML_LBASIS_DISCARD}}

	{{sc\|ML_MB\|Examples\|Examples that use this tag}}		{{sc\|ML_MB\|Examples\|Examples that use this tag}}

Kresse at 08:57, 19 December 2022

2022-12-19T08:57:24Z

← Older revision		Revision as of 08:57, 19 December 2022
Line 4:		Line 4:
	Description: This tag sets the maximum number of local reference configurations (i.e. basis functions in the kernel) in the machine learning force field method.		Description: This tag sets the maximum number of local reference configurations (i.e. basis functions in the kernel) in the machine learning force field method.
	----		----
	The default value is a relatively safe value for most materials. However one might need to increase it to a greater value for liquids, polymers and amorphous systems, or when a MLFF for many different polytypes is trained. When the number of local reference configurations exceeds this value, VASP stops and prompts an error instruction to increase {{TAG\|ML_MB}}. The user can also force VASP to dispose of local reference configurations (basis functions in the kernel) by specifying the tag {{TAG\| ML_LBASIS_DISCARD}}.		The default value is a relatively safe value for most materials. However one might need to increase it to a greater value for liquids, polymers and amorphous systems, or when a MLFF for many different polytypes is trained. When the number of local reference configurations exceeds this value, VASP stops and prompts an error instruction to increase {{TAG\|ML_MB}}. The user can also force VASP to dispose of local reference configurations (basis functions in the kernel) by specifying the tag {{TAG\| ML_LBASIS_DISCARD}}. In this case, VASP will not stop, but one should make extensive tests whether the generated MLFF is sufficiently accurate.

	If VASP stops, subsequent training can restarted from the existing ML_AB file. This avoids loss of already acquired training data.		If VASP stops, subsequent training can be restarted from the existing ML_AB file. This avoids loss of already acquired training data.

	The flag {{TAG\|ML_MB}} also determines the maximum number of rows of the design matrix. This is usually a huge matrix. The design matrix is allocated statically at the beginning of the run, since several parts of the code use MPI shared memory and dynamic reallocation of these arrays can cause issues on many operating systems. An estimate of the size of the design matrix and all other large arrays is printed out to the {{TAG\|ML_LOGFILE}} before allocation. The design matrix is fully distributed in a block cyclic fashion for scaLAPACK and thus the memory requirements scale inverse proportionally to the number of used processors.		The flag {{TAG\|ML_MB}} also determines the maximum number of rows of the design matrix. This is usually a huge matrix. The design matrix is allocated statically at the beginning of the run, since several parts of the code use MPI shared memory and dynamic reallocation of these arrays can cause issues on many operating systems. An estimate of the size of the design matrix and all other large arrays is printed out to the {{TAG\|ML_LOGFILE}} before allocation. The design matrix is fully distributed in a block cyclic fashion for scaLAPACK, and thus the memory requirements scale inverse proportionally to the number of used processors.

Kresse at 08:42, 19 December 2022

2022-12-19T08:42:39Z

← Older revision		Revision as of 08:42, 19 December 2022
Line 2:		Line 2:
	{{TAGDEF\|ML_MB\|[integer]\|1500}}		{{TAGDEF\|ML_MB\|[integer]\|1500}}

	Description: This tag sets the maximum number of local reference configurations in the machine learning force field method.		Description: This tag sets the maximum number of local reference configurations (i.e. basis functions in the kernel) in the machine learning force field method.
	----		----
	The default value is ~~usually~~ a safe value ~~but should be set~~ to a ~~higher~~ value ~~as soon as it~~ is ~~reached~~. When this ~~happens the code~~ stops and ~~gives~~ an error ~~instructing~~ to increase {{TAG\|ML_MB}}.		The default value is a relatively safe value for most materials. However one might need to increase it to a greater value for liquids, polymers and amorphous systems, or when a MLFF for many different polytypes is trained. When the number of local reference configurations exceeds this value, VASP stops and prompts an error instruction to increase {{TAG\|ML_MB}}. The user can also force VASP to dispose of local reference configurations (basis functions in the kernel) by specifying the tag {{TAG\| ML_LBASIS_DISCARD}}.
	This flag ~~sets~~ also the maximum number of rows ~~for~~ the design matrix~~, which~~ is usually a huge matrix. The design matrix is ~~to be~~ allocated statically at the beginning of the ~~program~~, since several parts of the code use MPI shared memory and dynamic reallocation of these arrays can cause ~~severe problems~~ on ~~some~~ systems. ~~So most~~ of the ~~main arrays are statically allocated in the code. An estimate~~ of the design matrix and all other large arrays is printed out to the {{TAG\|ML_LOGFILE}} before allocation. The design matrix is fully distributed in a block cyclic fashion for scaLAPACK and ~~should almost perfectly linearly~~ scale ~~with~~ the number of used processors.
			If VASP stops, subsequent training can restarted from the existing ML_AB file. This avoids loss of already acquired training data.

			The flag {{TAG\|ML_MB}} also determines the maximum number of rows of the design matrix. This is usually a huge matrix. The design matrix is allocated statically at the beginning of the run, since several parts of the code use MPI shared memory and dynamic reallocation of these arrays can cause issues on many operating systems. An estimate of the size of the design matrix and all other large arrays is printed out to the {{TAG\|ML_LOGFILE}} before allocation. The design matrix is fully distributed in a block cyclic fashion for scaLAPACK and thus the memory requirements scale inverse proportionally to the number of used processors.


	== Related tags and articles ==		== Related tags and articles ==
	{{TAG\|ML_LMLFF}}, {{TAG\|ML_MCONF}}		{{TAG\|ML_LMLFF}}, {{TAG\|ML_MCONF}}, {{TAG\| ML_LBASIS_DISCARD}}

	{{sc\|ML_MB\|Examples\|Examples that use this tag}}		{{sc\|ML_MB\|Examples\|Examples that use this tag}}

Karsai: /* Related tags and articles */

2022-04-08T13:27:27Z

Karsai at 07:34, 7 April 2022

2022-04-07T07:34:05Z

← Older revision		Revision as of 07:34, 7 April 2022
Line 1:		Line 1:
			{{DISPLAYTITLE:ML_MB}}
	{{TAGDEF\|ML_MB\|[integer]\|1500}}		{{TAGDEF\|ML_MB\|[integer]\|1500}}

Line 6:		Line 7:
	This flag sets also the maximum number of rows for the design matrix, which is usually a huge matrix. The design matrix is to be allocated statically at the beginning of the program, since several parts of the code use MPI shared memory and dynamic reallocation of these arrays can cause severe problems on some systems. So most of the main arrays are statically allocated in the code. An estimate of the design matrix and all other large arrays is printed out to the {{TAG\|ML_LOGFILE}} before allocation. The design matrix is fully distributed in a block cyclic fashion for scaLAPACK and should almost perfectly linearly scale with the number of used processors.		This flag sets also the maximum number of rows for the design matrix, which is usually a huge matrix. The design matrix is to be allocated statically at the beginning of the program, since several parts of the code use MPI shared memory and dynamic reallocation of these arrays can cause severe problems on some systems. So most of the main arrays are statically allocated in the code. An estimate of the design matrix and all other large arrays is printed out to the {{TAG\|ML_LOGFILE}} before allocation. The design matrix is fully distributed in a block cyclic fashion for scaLAPACK and should almost perfectly linearly scale with the number of used processors.

	== Related ~~Tags~~ and ~~Sections~~ ==		== Related tags and articles ==
	{{TAG\|ML_LMLFF}}, {{TAG\|ML_MCONF}}		{{TAG\|ML_LMLFF}}, {{TAG\|ML_MCONF}}

Line 12:		Line 13:
	----		----

	[[Category:INCAR]][[Category:Machine Learning]][[Category:Machine Learned Force Fields~~]][[Category: Alpha~~]]		[[Category:INCAR tag]][[Category:Machine Learning]][[Category:Machine Learned Force Fields]]

Singraber: TODO: Improve!

2021-10-11T21:24:04Z

TODO: Improve!

← Older revision		Revision as of 21:24, 11 October 2021
Line 1:		Line 1:
	{{TAGDEF\|ML_MB\|[integer]\|1500}}		{{TAGDEF\|ML_MB\|[integer]\|1500}}

	Description: This ~~flag~~ sets the maximum number of local configurations ~~(basis sets)~~ in the machine learning force field method.		Description: This tag sets the maximum number of local reference configurations in the machine learning force field method.
	----		----
	The default value is usually a safe value but should be set to a higher value as soon as it is reached. When this happens the code stops and gives an error instructing to increase {{TAG\|ML_MB}}.		The default value is usually a safe value but should be set to a higher value as soon as it is reached. When this happens the code stops and gives an error instructing to increase {{TAG\|ML_MB}}.
	This flag sets ~~alse~~ the maximum number of rows for the design matrix, which is usually a huge matrix. The design matrix is to be allocated statically at the beginning of the program, since several parts of the code use MPI shared memory and dynamic reallocation of these arrays can cause severe problems on some systems. So most of the main arrays are statically allocated in the code. An estimate of the design matrix and all other large arrays is printed out to the {{TAG\|ML_LOGFILE}} before allocation. The design matrix is fully distributed in a block cyclic fashion for scaLAPACK and should almost perfectly linearly scale with the number of used processors.		This flag sets also the maximum number of rows for the design matrix, which is usually a huge matrix. The design matrix is to be allocated statically at the beginning of the program, since several parts of the code use MPI shared memory and dynamic reallocation of these arrays can cause severe problems on some systems. So most of the main arrays are statically allocated in the code. An estimate of the design matrix and all other large arrays is printed out to the {{TAG\|ML_LOGFILE}} before allocation. The design matrix is fully distributed in a block cyclic fashion for scaLAPACK and should almost perfectly linearly scale with the number of used processors.

	== Related Tags and Sections ==		== Related Tags and Sections ==

Karsai at 08:59, 23 August 2021

2021-08-23T08:59:10Z

← Older revision		Revision as of 08:59, 23 August 2021
Line 1:		Line 1:
	{{TAGDEF\|~~ML_FF_MB_MB~~\|[integer]\|1500}}		{{TAGDEF\|ML_MB\|[integer]\|1500}}

	Description: This flag sets the maximum number of local configurations (basis sets) in the machine learning force field method.		Description: This flag sets the maximum number of local configurations (basis sets) in the machine learning force field method.
	----		----
	The default value is usually a safe value but should be set to a higher value as soon as it is reached. When this happens the code stops and gives an error instructing to increase {{TAG\|~~ML_FF_MB_MB~~}}.		The default value is usually a safe value but should be set to a higher value as soon as it is reached. When this happens the code stops and gives an error instructing to increase {{TAG\|ML_MB}}.
	This flag sets alse the maximum number of rows for the design matrix, which is usually a huge matrix. The design matrix is to be allocated statically at the beginning of the program, since several parts of the code use MPI shared memory and dynamic reallocation of these arrays can cause severe problems on some systems. So most of the main arrays are statically allocated in the code. An estimate of the design matrix and all other large arrays is printed out to the {{TAG\|ML_LOGFILE}} before allocation. The design matrix is fully distributed in a block cyclic fashion for scaLAPACK and should almost perfectly linearly scale with the number of used processors.		This flag sets alse the maximum number of rows for the design matrix, which is usually a huge matrix. The design matrix is to be allocated statically at the beginning of the program, since several parts of the code use MPI shared memory and dynamic reallocation of these arrays can cause severe problems on some systems. So most of the main arrays are statically allocated in the code. An estimate of the design matrix and all other large arrays is printed out to the {{TAG\|ML_LOGFILE}} before allocation. The design matrix is fully distributed in a block cyclic fashion for scaLAPACK and should almost perfectly linearly scale with the number of used processors.

	== Related Tags and Sections ==		== Related Tags and Sections ==
	{{TAG\|~~ML_FF_LMLFF~~}}, {{TAG\|~~ML_FF_MCONF~~}}		{{TAG\|ML_LMLFF}}, {{TAG\|ML_MCONF}}

	{{sc\|~~ML_FF_MB_MB~~\|Examples\|Examples that use this tag}}		{{sc\|ML_MB\|Examples\|Examples that use this tag}}
	----		----

	[[Category:INCAR]][[Category:Machine Learning]][[Category:Machine Learned Force Fields]][[Category: Alpha]]		[[Category:INCAR]][[Category:Machine Learning]][[Category:Machine Learned Force Fields]][[Category: Alpha]]

Karsai: Karsai moved page ML FF MB MB to ML MB

2021-08-23T06:44:59Z

Karsai moved page ML FF MB MB to ML MB

← Older revision	Revision as of 06:44, 23 August 2021
(No difference)

← Older revision		Revision as of 17:11, 15 February 2024
Line 5:		Line 5:
	----		----
	The defaults for {{TAG\|ML_MB}} are different for each different {{TAG\|ML_MODE}} setting. Here are the defaults for each mode:		The defaults for {{TAG\|ML_MB}} are different for each different {{TAG\|ML_MODE}} setting. Here are the defaults for each mode:
	*{{~~TAG~~\|ML_MODE}}~~='TRAIN'~~:		*{{TAGO\|ML_MODE\|train}}:
	*No {{TAG\|ML_AB}} present (learning from scratch): min(1500 , max({{TAG\|NSW}} , 2{{TAG\|ML_MCONF_NEW}} * MAXAT_SP)		*No {{TAG\|ML_AB}} present (learning from scratch): <code>min(1500 , max({{TAG\|NSW}} , 2{{TAG\|ML_MCONF_NEW}} * MAXAT_SP)</code>
	*{{TAG\|ML_AB}} present (continuation of learning): MB_AB + min(1500, max({{TAG\|NSW}} , 2{{TAG\|ML_MCONF_NEW}} * MAXAT_SP)		*{{TAG\|ML_AB}} present (continuation of learning): <code>MB_AB + min(1500, max({{TAG\|NSW}} , 2{{TAG\|ML_MCONF_NEW}} * MAXAT_SP)</code>
	{{~~TAG~~\|ML_MODE}}~~='SELECT'~~: MB_AB + {{TAG\|ML_MCONF_NEW}} MAXAT_SP		{{TAGO\|ML_MODE\|select}}: <code>MB_AB + {{TAG\|ML_MCONF_NEW}} MAXAT_SP</code>
	*{{~~TAG~~\|ML_MODE}}~~='REFIT'~~: MB_AB + MAXAT_SP		*{{TAGO\|ML_MODE\|refit}}: <code>MB_AB + MAXAT_SP</code>
	*{{~~TAG~~\|ML_MODE}}~~='REFITBAYESIAN'~~: MB_AB + MAXAT_SP		*{{TAGO\|ML_MODE\|refitbayesian}}: <code>MB_AB + MAXAT_SP</code>
	*{{~~TAG~~\|ML_MODE}}~~='RUN'~~: MB_AB		*{{TAGO\|ML_MODE\|run}}: <code>MB_AB</code>
	using the following definitions:\|		using the following definitions:
	*MAXAT_SP = greatest number of atoms within all species among the current structures and the structures in the {{TAG\|ML_AB}} file (if present).		*<code>MAXAT_SP</code> = greatest number of atoms within all species among the current structures and the structures in the {{TAG\|ML_AB}} file (if present).
	*MB_AB = greatest number of local reference configurations within all species in the {{TAG\|ML_AB}} file.		*<code>MB_AB</code> = greatest number of local reference configurations within all species in the {{TAG\|ML_AB}} file.

	The default value for {{~~TAG~~\|ML_MODE}} ~~= ''TRAIN''~~ and {{~~TAG~~\|ML_MODE}} ~~= ''SELECT''~~ is a relatively safe value for most materials. However one might need to increase it to a greater value for liquids, polymers and amorphous systems, or when a MLFF for many different polytypes is trained. By default ({{~~TAG~~\|ML_LBASIS_DISCARD~~}}=''~~.TRUE.''), if the number of local reference configurations would exceed {{TAG\|ML_MB}}, VASP would continue the calculation and disposes of these, but one should make extensive tests whether the generated MLFF is sufficiently accurate.		The default value for {{TAGO\|ML_MODE\|train}} and {{TAGO\|ML_MODE\|select}} is a relatively safe value for most materials. However one might need to increase it to a greater value for liquids, polymers and amorphous systems, or when a MLFF for many different polytypes is trained. By default ({{TAGO\|ML_LBASIS_DISCARD\|.TRUE.}}), if the number of local reference configurations would exceed {{TAG\|ML_MB}}, VASP would continue the calculation and disposes of these, but one should make extensive tests whether the generated MLFF is sufficiently accurate.

	If VASP stops, subsequent training can be restarted from the existing ML_AB file. This avoids loss of already acquired training data.		If VASP stops, subsequent training can be restarted from the existing ML_AB file. This avoids loss of already acquired training data.

	Depending on the calculation mode VASP adds a little overhead in the allocation of {{TAG\|ML_MB}} arrays to buffer for new candidate structures. Here are the buffer sizes for each mode:		Depending on the calculation mode VASP adds a little overhead in the allocation of {{TAG\|ML_MB}} arrays to buffer for new candidate structures. Here are the buffer sizes for each mode:
	*{{~~TAG~~\|ML_MODE}}~~='TRAIN'~~:		*{{TAGO\|ML_MODE\|train}}:
	*No {{TAG\|ML_AB}} present (learning from scratch): min({{TAG\|NSW}} , {{TAG\|ML_MCONF_NEW}}) MAXAT_SP		*No {{TAG\|ML_AB}} present (learning from scratch): <code>min({{TAG\|NSW}} , {{TAG\|ML_MCONF_NEW}}) MAXAT_SP</code>
	*{{TAG\|ML_AB}} present (continuation of learning): min({{TAG\|NSW}} , {{TAG\|ML_MCONF_NEW}}) MAXAT_SP		*{{TAG\|ML_AB}} present (continuation of learning): <code>min({{TAG\|NSW}} , {{TAG\|ML_MCONF_NEW}}) MAXAT_SP</code>
	{{~~TAG~~\|ML_MODE}}~~='SELECT'~~: {{TAG\|ML_MCONF_NEW}} MAXAT_SP		{{TAGO\|ML_MODE\|select}}: <code>{{TAG\|ML_MCONF_NEW}} MAXAT_SP</code>
	*{{~~TAG~~\|ML_MODE}}~~='REFIT'~~: MAXAT_SP		*{{TAGO\|ML_MODE\|refit}}: <code>MAXAT_SP</code>
	*{{~~TAG~~\|ML_MODE}}~~='REFITBAYESIAN'~~: MAXAT_SP		*{{TAGO\|ML_MODE\|refitbayesian}}: <code>MAXAT_SP</code>
	*{{~~TAG~~\|ML_MODE}}~~='RUN'~~: 0		*{{TAGO\|ML_MODE\|run}}: <code>0</code>

	The flag {{TAG\|ML_MB}} also determines the maximum number of rows of the design matrix. This is usually a huge matrix. The design matrix is allocated statically at the beginning of the run, since several parts of the code use MPI shared memory and dynamic reallocation of these arrays can cause issues on many operating systems. An estimate of the size of the design matrix and all other large arrays is printed out to the {{TAG\|ML_LOGFILE}} before allocation. The design matrix is fully distributed in a block-cyclic fashion for scaLAPACK, and thus the memory requirement for each core scales inversely proportional to the number of used cores.

			The flag {{TAG\|ML_MB}} also determines the maximum number of rows of the design matrix. This is usually a huge matrix. The design matrix is allocated statically at the beginning of the run, since several parts of the code use MPI shared memory and dynamic reallocation of these arrays can cause issues on many operating systems. An estimate of the size of the design matrix and all other large arrays is printed out to the {{TAG\|ML_LOGFILE}} before allocation. The design matrix is fully distributed in a block-cyclic fashion for scaLAPACK, and thus the memory requirement for each core scales inversely proportional to the number of used cores.

	== Related tags and articles ==		== Related tags and articles ==

← Older revision		Revision as of 09:11, 14 April 2023
Line 1:		Line 1:
	{{DISPLAYTITLE:ML_MB}}		{{DISPLAYTITLE:ML_MB}}
	{{TAGDEF\|ML_MB\|[integer]\|~~1500~~}}		{{TAGDEF\|ML_MB\|[integer]\|see below}}

	Description: This tag sets the maximum number of local reference configurations (i.e. basis functions in the kernel) in the machine learning force field method.		Description: This tag sets the maximum number of local reference configurations (i.e. basis functions in the kernel) in the machine learning force field method.
	----		----
	The default value is a relatively safe value for most materials. However one might need to increase it to a greater value for liquids, polymers and amorphous systems, or when a MLFF for many different polytypes is trained. ~~When the number of local reference configurations exceeds this value, VASP stops and prompts an error instruction to increase~~ {{TAG\|~~ML_MB~~}}. ~~The user can also force VASP to dispose~~ of local reference configurations ~~(basis functions in the kernel) by specifying the tag~~ {{TAG\| ~~ML_LBASIS_DISCARD~~}}~~. In this case~~, VASP ~~will not stop~~, but one should make extensive tests whether the generated MLFF is sufficiently accurate.		The defaults for {{TAG\|ML_MB}} are different for each different {{TAG\|ML_MODE}} setting. Here are the defaults for each mode:
			*{{TAG\|ML_MODE}}='TRAIN':
			*No {{TAG\|ML_AB}} present (learning from scratch): min(1500 , max({{TAG\|NSW}} , 2{{TAG\|ML_MCONF_NEW}} * MAXAT_SP)
			*{{TAG\|ML_AB}} present (continuation of learning): MB_AB + min(1500, max({{TAG\|NSW}} , 2{{TAG\|ML_MCONF_NEW}} * MAXAT_SP)
			{{TAG\|ML_MODE}}='SELECT': MB_AB + {{TAG\|ML_MCONF_NEW}} MAXAT_SP
			*{{TAG\|ML_MODE}}='REFIT': MB_AB + MAXAT_SP
			*{{TAG\|ML_MODE}}='REFITBAYESIAN': MB_AB + MAXAT_SP
			*{{TAG\|ML_MODE}}='RUN': MB_AB
			using the following definitions:\|
			*MAXAT_SP = greatest number of atoms within all species among the current structures and the structures in the {{TAG\|ML_AB}} file (if present).
			*MB_AB = greatest number of local reference configurations within all species in the {{TAG\|ML_AB}} file.

			The default value for {{TAG\|ML_MODE}} = ''TRAIN'' and {{TAG\|ML_MODE}} = ''SELECT'' is a relatively safe value for most materials. However one might need to increase it to a greater value for liquids, polymers and amorphous systems, or when a MLFF for many different polytypes is trained. By default ({{TAG\|ML_LBASIS_DISCARD}}=''.TRUE.''), if the number of local reference configurations would exceed {{TAG\|ML_MB}}, VASP would continue the calculation and disposes of these, but one should make extensive tests whether the generated MLFF is sufficiently accurate.

	If VASP stops, subsequent training can be restarted from the existing ML_AB file. This avoids loss of already acquired training data.		If VASP stops, subsequent training can be restarted from the existing ML_AB file. This avoids loss of already acquired training data.

	The flag {{TAG\|ML_MB}} also determines the maximum number of rows of the design matrix. This is usually a huge matrix. The design matrix is allocated statically at the beginning of the run, since several parts of the code use MPI shared memory and dynamic reallocation of these arrays can cause issues on many operating systems. An estimate of the size of the design matrix and all other large arrays is printed out to the {{TAG\|ML_LOGFILE}} before allocation. The design matrix is fully distributed in a block cyclic fashion for scaLAPACK, and thus the memory ~~requirements scale inverse proportionally~~ to the number of used ~~processors~~.		Depending on the calculation mode VASP adds a little overhead in the allocation of {{TAG\|ML_MB}} arrays to buffer for new candidate structures. Here are the buffer sizes for each mode:
			*{{TAG\|ML_MODE}}='TRAIN':
			*No {{TAG\|ML_AB}} present (learning from scratch): min({{TAG\|NSW}} , {{TAG\|ML_MCONF_NEW}}) MAXAT_SP
			*{{TAG\|ML_AB}} present (continuation of learning): min({{TAG\|NSW}} , {{TAG\|ML_MCONF_NEW}}) MAXAT_SP
			{{TAG\|ML_MODE}}='SELECT': {{TAG\|ML_MCONF_NEW}} MAXAT_SP
			*{{TAG\|ML_MODE}}='REFIT': MAXAT_SP
			*{{TAG\|ML_MODE}}='REFITBAYESIAN': MAXAT_SP
			*{{TAG\|ML_MODE}}='RUN': 0

			The flag {{TAG\|ML_MB}} also determines the maximum number of rows of the design matrix. This is usually a huge matrix. The design matrix is allocated statically at the beginning of the run, since several parts of the code use MPI shared memory and dynamic reallocation of these arrays can cause issues on many operating systems. An estimate of the size of the design matrix and all other large arrays is printed out to the {{TAG\|ML_LOGFILE}} before allocation. The design matrix is fully distributed in a block-cyclic fashion for scaLAPACK, and thus the memory requirement for each core scales inversely proportional to the number of used cores.


	== Related tags and articles ==		== Related tags and articles ==
	{{TAG\|ML_LMLFF}}, {{TAG\|ML_MCONF}}, {{TAG\| ML_LBASIS_DISCARD}}		{{TAG\|ML_LMLFF}}, {{TAG\|ML_MODE}}, {{TAG\|ML_MCONF_NEW}}, {{TAG\|ML_MCONF}}, {{TAG\| ML_LBASIS_DISCARD}}

	{{sc\|ML_MB\|Examples\|Examples that use this tag}}		{{sc\|ML_MB\|Examples\|Examples that use this tag}}

← Older revision		Revision as of 07:56, 24 October 2025
Line 5:		Line 5:
	----		----
	The defaults for {{TAG\|ML_MB}} are different for each different {{TAG\|ML_MODE}} setting. Here are the defaults for each mode:		The defaults for {{TAG\|ML_MB}} are different for each different {{TAG\|ML_MODE}} setting. Here are the defaults for each mode:
	*{{~~TAGO~~\|ML_MODE\|train}}:		*{{TAG\|ML_MODE\|train}}:
	*No {{TAG\|ML_AB}} present (learning from scratch): <code>min(1500 , max({{TAG\|NSW}} , 2{{TAG\|ML_MCONF_NEW}} * MAXAT_SP)</code>		*No {{TAG\|ML_AB}} present (learning from scratch): <code>min(1500 , max({{TAG\|NSW}} , 2{{TAG\|ML_MCONF_NEW}} * MAXAT_SP)</code>
	*{{TAG\|ML_AB}} present (continuation of learning): <code>MB_AB + min(1500, max({{TAG\|NSW}} , 2{{TAG\|ML_MCONF_NEW}} * MAXAT_SP)</code>		*{{TAG\|ML_AB}} present (continuation of learning): <code>MB_AB + min(1500, max({{TAG\|NSW}} , 2{{TAG\|ML_MCONF_NEW}} * MAXAT_SP)</code>
	{{~~TAGO~~\|ML_MODE\|select}}: <code>MB_AB + {{TAG\|ML_MCONF_NEW}} MAXAT_SP</code>		{{TAG\|ML_MODE\|select}}: <code>MB_AB + {{TAG\|ML_MCONF_NEW}} MAXAT_SP</code>
	*{{~~TAGO~~\|ML_MODE\|refit}}: <code>MB_AB + MAXAT_SP</code>		*{{TAG\|ML_MODE\|refit}}: <code>MB_AB + MAXAT_SP</code>
	*{{~~TAGO~~\|ML_MODE\|refitbayesian}}: <code>MB_AB + MAXAT_SP</code>		*{{TAG\|ML_MODE\|refitbayesian}}: <code>MB_AB + MAXAT_SP</code>
	*{{~~TAGO~~\|ML_MODE\|run}}: <code>MB_AB</code>		*{{TAG\|ML_MODE\|run}}: <code>MB_AB</code>
	using the following definitions:		using the following definitions:
	*<code>MAXAT_SP</code> = greatest number of atoms within all species among the current structures and the structures in the {{TAG\|ML_AB}} file (if present).		*<code>MAXAT_SP</code> = greatest number of atoms within all species among the current structures and the structures in the {{TAG\|ML_AB}} file (if present).
	*<code>MB_AB</code> = greatest number of local reference configurations within all species in the {{TAG\|ML_AB}} file.		*<code>MB_AB</code> = greatest number of local reference configurations within all species in the {{TAG\|ML_AB}} file.

	The default value for {{~~TAGO~~\|ML_MODE\|train}} and {{~~TAGO~~\|ML_MODE\|select}} is a relatively safe value for most materials. However one might need to increase it to a greater value for liquids, polymers and amorphous systems, or when a MLFF for many different polytypes is trained. By default ({{~~TAGO~~\|ML_LBASIS_DISCARD\|.TRUE.}}), if the number of local reference configurations would exceed {{TAG\|ML_MB}}, VASP would continue the calculation and disposes of these, but one should make extensive tests whether the generated MLFF is sufficiently accurate.		The default value for {{TAG\|ML_MODE\|train}} and {{TAG\|ML_MODE\|select}} is a relatively safe value for most materials. However one might need to increase it to a greater value for liquids, polymers and amorphous systems, or when a MLFF for many different polytypes is trained. By default ({{TAG\|ML_LBASIS_DISCARD\|.TRUE.}}), if the number of local reference configurations would exceed {{TAG\|ML_MB}}, VASP would continue the calculation and disposes of these, but one should make extensive tests whether the generated MLFF is sufficiently accurate.

	If VASP stops, subsequent training can be restarted from the existing ML_AB file. This avoids loss of already acquired training data.		If VASP stops, subsequent training can be restarted from the existing ML_AB file. This avoids loss of already acquired training data.

	Depending on the calculation mode VASP adds a little overhead in the allocation of {{TAG\|ML_MB}} arrays to buffer for new candidate structures. Here are the buffer sizes for each mode:		Depending on the calculation mode VASP adds a little overhead in the allocation of {{TAG\|ML_MB}} arrays to buffer for new candidate structures. Here are the buffer sizes for each mode:
	*{{~~TAGO~~\|ML_MODE\|train}}:		*{{TAG\|ML_MODE\|train}}:
	*No {{TAG\|ML_AB}} present (learning from scratch): <code>min({{TAG\|NSW}} , {{TAG\|ML_MCONF_NEW}}) MAXAT_SP</code>		*No {{TAG\|ML_AB}} present (learning from scratch): <code>min({{TAG\|NSW}} , {{TAG\|ML_MCONF_NEW}}) MAXAT_SP</code>
	*{{TAG\|ML_AB}} present (continuation of learning): <code>min({{TAG\|NSW}} , {{TAG\|ML_MCONF_NEW}}) MAXAT_SP</code>		*{{TAG\|ML_AB}} present (continuation of learning): <code>min({{TAG\|NSW}} , {{TAG\|ML_MCONF_NEW}}) MAXAT_SP</code>
	{{~~TAGO~~\|ML_MODE\|select}}: <code>{{TAG\|ML_MCONF_NEW}} MAXAT_SP</code>		{{TAG\|ML_MODE\|select}}: <code>{{TAG\|ML_MCONF_NEW}} MAXAT_SP</code>
	*{{~~TAGO~~\|ML_MODE\|refit}}: <code>MAXAT_SP</code>		*{{TAG\|ML_MODE\|refit}}: <code>MAXAT_SP</code>
	*{{~~TAGO~~\|ML_MODE\|refitbayesian}}: <code>MAXAT_SP</code>		*{{TAG\|ML_MODE\|refitbayesian}}: <code>MAXAT_SP</code>
	*{{~~TAGO~~\|ML_MODE\|run}}: <code>0</code>		*{{TAG\|ML_MODE\|run}}: <code>0</code>

	The flag {{TAG\|ML_MB}} also determines the maximum number of rows of the design matrix. This is usually a huge matrix. The design matrix is allocated statically at the beginning of the run, since several parts of the code use MPI shared memory and dynamic reallocation of these arrays can cause issues on many operating systems. An estimate of the size of the design matrix and all other large arrays is printed out to the {{TAG\|ML_LOGFILE}} before allocation. The design matrix is fully distributed in a block-cyclic fashion for scaLAPACK, and thus the memory requirement for each core scales inversely proportional to the number of used cores.		The flag {{TAG\|ML_MB}} also determines the maximum number of rows of the design matrix. This is usually a huge matrix. The design matrix is allocated statically at the beginning of the run, since several parts of the code use MPI shared memory and dynamic reallocation of these arrays can cause issues on many operating systems. An estimate of the size of the design matrix and all other large arrays is printed out to the {{TAG\|ML_LOGFILE}} before allocation. The design matrix is fully distributed in a block-cyclic fashion for scaLAPACK, and thus the memory requirement for each core scales inversely proportional to the number of used cores.

← Older revision		Revision as of 13:27, 8 April 2022
Line 13:		Line 13:
	----		----

	[[Category:INCAR tag]][[Category:Machine ~~Learning]][[Category:Machine Learned Force Fields~~]]		[[Category:INCAR tag]][[Category:Machine-learned force fields]]