https://beta.vasp.at/wiki/index.php?action=history&feed=atom&title=ML_ISTART 2026-04-23T01:04:25Z Revision history for this page on the wiki MediaWiki 1.43.8 https://beta.vasp.at/wiki/index.php?title=ML_ISTART&diff=22174&oldid=prev 2023-10-19T14:40:56Z

<p></p> <table style="background-color: #fff; color: #202122;" data-mw="interface"> <col class="diff-marker" /> <col class="diff-content" /> <col class="diff-marker" /> <col class="diff-content" /> <tr class="diff-title" lang="en"> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">← Older revision</td> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">Revision as of 14:40, 19 October 2023</td> </tr><tr><td colspan="2" class="diff-lineno" id="mw-diff-left-l1">Line 1:</td> <td colspan="2" class="diff-lineno">Line 1:</td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>{{DISPLAYTITLE:ML_ISTART}}</div></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>{{DISPLAYTITLE:ML_ISTART}}</div></td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>{{TAGDEF|ML_ISTART|[integer]|0}}</div></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>{{TAGDEF|ML_ISTART|[integer]|0}}</div></td></tr> <tr><td class="diff-marker" data-marker="−"></td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #ffe49c; vertical-align: top; white-space: pre-wrap;"><div> </div></td><td class="diff-marker" data-marker="+"></td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;"><div>{{NB|<ins style="font-weight: bold; text-decoration: none;">warning</ins>|This tag is deprecated and we advise to use {{TAG|ML_MODE}} instead.}}</div></td></tr> <tr><td class="diff-marker" data-marker="−"></td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #ffe49c; vertical-align: top; white-space: pre-wrap;"><div>{{NB|<del style="font-weight: bold; text-decoration: none;">mind</del>|This tag is deprecated and we advise to use {{TAG|ML_MODE}} instead.}}</div></td><td colspan="2" class="diff-side-added"></td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>Description: This tag selects the mode of operation (e.g. start from scratch, prediction-only,...) of the machine learning force fields method.</div></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>Description: This tag selects the mode of operation (e.g. start from scratch, prediction-only,...) of the machine learning force fields method.</div></td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>----</div></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>----</div></td></tr> </table>

Singraber https://beta.vasp.at/wiki/index.php?title=ML_ISTART&diff=19903&oldid=prev 2023-03-29T08:09:21Z

<p></p> <table style="background-color: #fff; color: #202122;" data-mw="interface"> <col class="diff-marker" /> <col class="diff-content" /> <col class="diff-marker" /> <col class="diff-content" /> <tr class="diff-title" lang="en"> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">← Older revision</td> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">Revision as of 08:09, 29 March 2023</td> </tr><tr><td colspan="2" class="diff-lineno" id="mw-diff-left-l11">Line 11:</td> <td colspan="2" class="diff-lineno">Line 11:</td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>{{NB|tip|Setting {{TAG|ML_ISTART}} {{=}} 1 together with {{TAG|NSW}} {{=}} 0 allows to repeat learning on the given training data and create a new force field in {{FILE|ML_FFN}} without actually performing additional MD steps. In this way force field parameters (e.g. cutoff radii, number of radial basis functions, etc.) can be varied without recalculating the entire trajectory. Moreover, because Bayesian error estimation is not required when no MD is run it is possible to switch the regression algorithm via the tag {{TAG|ML_IALGO_LINREG}} and check whether in this way better fitting results can be achieved. In order to avoid that the starting structure in the {{FILE|POSCAR}} file is processed and eventually added to the training data just set {{TAG|ML_CTIFOR}} to a large value (e.g. 1000).|:}}</div></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>{{NB|tip|Setting {{TAG|ML_ISTART}} {{=}} 1 together with {{TAG|NSW}} {{=}} 0 allows to repeat learning on the given training data and create a new force field in {{FILE|ML_FFN}} without actually performing additional MD steps. In this way force field parameters (e.g. cutoff radii, number of radial basis functions, etc.) can be varied without recalculating the entire trajectory. Moreover, because Bayesian error estimation is not required when no MD is run it is possible to switch the regression algorithm via the tag {{TAG|ML_IALGO_LINREG}} and check whether in this way better fitting results can be achieved. In order to avoid that the starting structure in the {{FILE|POSCAR}} file is processed and eventually added to the training data just set {{TAG|ML_CTIFOR}} to a large value (e.g. 1000).|:}}</div></td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>*{{TAG|ML_ISTART}} = 2: Prediction only. In this mode the previously trained machine learning force field is read from the {{FILE|ML_FF}} file. The MD simulation is driven with predictions from the force field only, no ab initio calculations are performed and no learning is executed. However, in order to monitor the quality of predictions the Bayesian error estimate of forces is still computed and logged in the {{FILE|ML_LOGFILE}}. This setting is typically used when the machine learning force field is considered mature and ready for production runs.</div></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>*{{TAG|ML_ISTART}} = 2: Prediction only. In this mode the previously trained machine learning force field is read from the {{FILE|ML_FF}} file. The MD simulation is driven with predictions from the force field only, no ab initio calculations are performed and no learning is executed. However, in order to monitor the quality of predictions the Bayesian error estimate of forces is still computed and logged in the {{FILE|ML_LOGFILE}}. This setting is typically used when the machine learning force field is considered mature and ready for production runs.</div></td></tr> <tr><td class="diff-marker" data-marker="−"></td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #ffe49c; vertical-align: top; white-space: pre-wrap;"><div>*{{TAG|ML_ISTART}} = 3: Learning from given ab initio data only, no MD time steps. In this operation mode a new machine learning force field is generated from ab initio data provided in the {{FILE|ML_AB}} file. The structures are read in and processed one by one as if harvested via an MD simulation. In other words, the same steps are performed as in on-the-fly training but the source of data is not an MD run but the series of structures available in {{FILE|ML_AB}}. This operation mode can be used to generate {{VASP}} machine learning force fields from pre-computed or external ab initio data sets. At first glance {{TAG|ML_ISTART}} = 3 looks very similar to the combination of {{TAG|ML_ISTART}} = 1 and {{TAG|NSW}} = 0 described above. However, there is an important difference: Setting {{TAG|ML_ISTART}} = 3 will ignore the list of local reference configurations in the {{FILE|ML_AB}} file and instead will determine a new collection which is written to the resulting {{FILE|ML_ABN}} file. {{NB|tip|If calculations for {{TAG|ML_ISTART}} {{=}} 3 are too time-consuming using the default settings, it is useful to increase {{TAG|ML_MCONF_NEW}} to values around 10-16 and set  {{TAG|ML_CDOUB}} {{=}} 4. This often accelerates the calculations by a factor  </div></td><td class="diff-marker" data-marker="+"></td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;"><div>*{{TAG|ML_ISTART}} = 3: Learning from given ab initio data only, no MD time steps. In this operation mode a new machine learning force field is generated from ab initio data provided in the {{FILE|ML_AB}} file. The structures are read in and processed one by one as if harvested via an MD simulation. In other words, the same steps are performed as in on-the-fly training but the source of data is not an MD run but the series of structures available in {{FILE|ML_AB}}. This operation mode can be used to generate {{VASP}} machine learning force fields from pre-computed or external ab initio data sets. At first glance {{TAG|ML_ISTART}} = 3 looks very similar to the combination of {{TAG|ML_ISTART}} = 1 and {{TAG|NSW}} = 0 described above. However, there is an important difference: Setting {{TAG|ML_ISTART}} = 3 will ignore the list of local reference configurations in the {{FILE|ML_AB}} file and instead will determine a new collection which is written to the resulting {{FILE|ML_ABN}} file. {{NB|tip|If calculations for {{TAG|ML_ISTART}} {{=}} 3 are too time-consuming using the default settings, it is useful to increase {{TAG|ML_MCONF_NEW}} to values around 10-16 and set  {{TAG|ML_CDOUB}} {{=}} 4. This often accelerates the calculations by a factor of 2-4.|:}} The {{TAG|ML_AB}} file may contain values for ''CTIFOR'' for each training structure. These are the thresholds used to sample that structure from the previous training. If a value for {{TAG|ML_CTIFOR}} is specified in the {{TAG|INCAR}} file, that value is then used and the thresholds from the {{TAG|ML_AB}} are ignored. Otherwise: 1) If thresholds exist in the {{TAG|ML_AB}} they are used. 2) If no thresholds are specified the default value for {{TAG|ML_CTIFOR}} is used.  </div></td></tr> <tr><td class="diff-marker" data-marker="−"></td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #ffe49c; vertical-align: top; white-space: pre-wrap;"><div><del style="font-weight: bold; text-decoration: none;"> </del>of 2-4.|:}} The {{TAG|ML_AB}} file may contain values for ''CTIFOR'' for each training structure. These are the thresholds used to sample that structure from the previous training. If a value for {{TAG|ML_CTIFOR}} is specified in the {{TAG|INCAR}} file, that value is then used and the thresholds from the {{TAG|ML_AB}} are ignored. Otherwise: 1) If thresholds exist in the {{TAG|ML_AB}} they are used. 2) If no thresholds are specified the default value for {{TAG|ML_CTIFOR}} is used.  </div></td><td colspan="2" class="diff-side-added"></td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>*{{TAG|ML_ISTART}} = 4: Refitting of the force field is done based on an existing {{TAG|ML_AB}} file, but the number of local reference configurations is taken from the {{TAG|ML_AB}} file. {{TAG|NSW}} on the input is ignored and only a single step is executed. No ab-initio calculation is carried out.  </div></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>*{{TAG|ML_ISTART}} = 4: Refitting of the force field is done based on an existing {{TAG|ML_AB}} file, but the number of local reference configurations is taken from the {{TAG|ML_AB}} file. {{TAG|NSW}} on the input is ignored and only a single step is executed. No ab-initio calculation is carried out.  </div></td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><br></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><br></td></tr> </table>

Karsai https://beta.vasp.at/wiki/index.php?title=ML_ISTART&diff=19902&oldid=prev 2023-03-29T08:09:00Z

<p></p> <table style="background-color: #fff; color: #202122;" data-mw="interface"> <col class="diff-marker" /> <col class="diff-content" /> <col class="diff-marker" /> <col class="diff-content" /> <tr class="diff-title" lang="en"> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">← Older revision</td> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">Revision as of 08:09, 29 March 2023</td> </tr><tr><td colspan="2" class="diff-lineno" id="mw-diff-left-l11">Line 11:</td> <td colspan="2" class="diff-lineno">Line 11:</td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>{{NB|tip|Setting {{TAG|ML_ISTART}} {{=}} 1 together with {{TAG|NSW}} {{=}} 0 allows to repeat learning on the given training data and create a new force field in {{FILE|ML_FFN}} without actually performing additional MD steps. In this way force field parameters (e.g. cutoff radii, number of radial basis functions, etc.) can be varied without recalculating the entire trajectory. Moreover, because Bayesian error estimation is not required when no MD is run it is possible to switch the regression algorithm via the tag {{TAG|ML_IALGO_LINREG}} and check whether in this way better fitting results can be achieved. In order to avoid that the starting structure in the {{FILE|POSCAR}} file is processed and eventually added to the training data just set {{TAG|ML_CTIFOR}} to a large value (e.g. 1000).|:}}</div></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>{{NB|tip|Setting {{TAG|ML_ISTART}} {{=}} 1 together with {{TAG|NSW}} {{=}} 0 allows to repeat learning on the given training data and create a new force field in {{FILE|ML_FFN}} without actually performing additional MD steps. In this way force field parameters (e.g. cutoff radii, number of radial basis functions, etc.) can be varied without recalculating the entire trajectory. Moreover, because Bayesian error estimation is not required when no MD is run it is possible to switch the regression algorithm via the tag {{TAG|ML_IALGO_LINREG}} and check whether in this way better fitting results can be achieved. In order to avoid that the starting structure in the {{FILE|POSCAR}} file is processed and eventually added to the training data just set {{TAG|ML_CTIFOR}} to a large value (e.g. 1000).|:}}</div></td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>*{{TAG|ML_ISTART}} = 2: Prediction only. In this mode the previously trained machine learning force field is read from the {{FILE|ML_FF}} file. The MD simulation is driven with predictions from the force field only, no ab initio calculations are performed and no learning is executed. However, in order to monitor the quality of predictions the Bayesian error estimate of forces is still computed and logged in the {{FILE|ML_LOGFILE}}. This setting is typically used when the machine learning force field is considered mature and ready for production runs.</div></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>*{{TAG|ML_ISTART}} = 2: Prediction only. In this mode the previously trained machine learning force field is read from the {{FILE|ML_FF}} file. The MD simulation is driven with predictions from the force field only, no ab initio calculations are performed and no learning is executed. However, in order to monitor the quality of predictions the Bayesian error estimate of forces is still computed and logged in the {{FILE|ML_LOGFILE}}. This setting is typically used when the machine learning force field is considered mature and ready for production runs.</div></td></tr> <tr><td class="diff-marker" data-marker="−"></td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #ffe49c; vertical-align: top; white-space: pre-wrap;"><div>*{{TAG|ML_ISTART}} = 3: Learning from given ab initio data only, no MD time steps. In this operation mode a new machine learning force field is generated from ab initio data provided in the {{FILE|ML_AB}} file. The structures are read in and processed one by one as if harvested via an MD simulation. In other words, the same steps are performed as in on-the-fly training but the source of data is not an MD run but the series of structures available in {{FILE|ML_AB}}. This operation mode can be used to generate {{VASP}} machine learning force fields from pre-computed or external ab initio data sets. At first glance {{TAG|ML_ISTART}} = 3 looks very similar to the combination of {{TAG|ML_ISTART}} = 1 and {{TAG|NSW}} = 0 described above. However, there is an important difference: Setting {{TAG|ML_ISTART}} = 3 will ignore the list of local reference configurations in the {{FILE|ML_AB}} file and instead will determine a new collection which is written to the resulting {{FILE|ML_ABN}} file. {{NB|tip|If calculations for {{TAG|ML_ISTART}} {{=}} 3 are too time-consuming using the default settings, it is useful to increase {{TAG|ML_MCONF_NEW}} to values around 10-16 and set  {{TAG|ML_CDOUB}} {{=}} 4. This often accelerates the calculations a factor 2-4.|:}} The {{TAG|ML_AB}} file may contain values for ''CTIFOR'' for each training structure. These are the thresholds used to sample that structure from the previous training. If a value for {{TAG|ML_CTIFOR}} is specified in the {{TAG|INCAR}} file, that value is then used and the thresholds from the {{TAG|ML_AB}} are ignored. Otherwise: 1) If thresholds exist in the {{TAG|ML_AB}} they are used. 2) If no thresholds are specified the default value for {{TAG|ML_CTIFOR}} is used.  </div></td><td class="diff-marker" data-marker="+"></td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;"><div>*{{TAG|ML_ISTART}} = 3: Learning from given ab initio data only, no MD time steps. In this operation mode a new machine learning force field is generated from ab initio data provided in the {{FILE|ML_AB}} file. The structures are read in and processed one by one as if harvested via an MD simulation. In other words, the same steps are performed as in on-the-fly training but the source of data is not an MD run but the series of structures available in {{FILE|ML_AB}}. This operation mode can be used to generate {{VASP}} machine learning force fields from pre-computed or external ab initio data sets. At first glance {{TAG|ML_ISTART}} = 3 looks very similar to the combination of {{TAG|ML_ISTART}} = 1 and {{TAG|NSW}} = 0 described above. However, there is an important difference: Setting {{TAG|ML_ISTART}} = 3 will ignore the list of local reference configurations in the {{FILE|ML_AB}} file and instead will determine a new collection which is written to the resulting {{FILE|ML_ABN}} file. {{NB|tip|If calculations for {{TAG|ML_ISTART}} {{=}} 3 are too time-consuming using the default settings, it is useful to increase {{TAG|ML_MCONF_NEW}} to values around 10-16 and set  {{TAG|ML_CDOUB}} {{=}} 4. This often accelerates the calculations <ins style="font-weight: bold; text-decoration: none;">by </ins>a factor  </div></td></tr> <tr><td colspan="2" class="diff-side-deleted"></td><td class="diff-marker" data-marker="+"></td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;"><div><ins style="font-weight: bold; text-decoration: none;"> of </ins>2-4.|:}} The {{TAG|ML_AB}} file may contain values for ''CTIFOR'' for each training structure. These are the thresholds used to sample that structure from the previous training. If a value for {{TAG|ML_CTIFOR}} is specified in the {{TAG|INCAR}} file, that value is then used and the thresholds from the {{TAG|ML_AB}} are ignored. Otherwise: 1) If thresholds exist in the {{TAG|ML_AB}} they are used. 2) If no thresholds are specified the default value for {{TAG|ML_CTIFOR}} is used.  </div></td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>*{{TAG|ML_ISTART}} = 4: Refitting of the force field is done based on an existing {{TAG|ML_AB}} file, but the number of local reference configurations is taken from the {{TAG|ML_AB}} file. {{TAG|NSW}} on the input is ignored and only a single step is executed. No ab-initio calculation is carried out.  </div></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>*{{TAG|ML_ISTART}} = 4: Refitting of the force field is done based on an existing {{TAG|ML_AB}} file, but the number of local reference configurations is taken from the {{TAG|ML_AB}} file. {{TAG|NSW}} on the input is ignored and only a single step is executed. No ab-initio calculation is carried out.  </div></td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><br></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><br></td></tr> </table>

Karsai https://beta.vasp.at/wiki/index.php?title=ML_ISTART&diff=19845&oldid=prev 2023-03-28T08:25:21Z

<p></p> <table style="background-color: #fff; color: #202122;" data-mw="interface"> <col class="diff-marker" /> <col class="diff-content" /> <col class="diff-marker" /> <col class="diff-content" /> <tr class="diff-title" lang="en"> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">← Older revision</td> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">Revision as of 08:25, 28 March 2023</td> </tr><tr><td colspan="2" class="diff-lineno" id="mw-diff-left-l11">Line 11:</td> <td colspan="2" class="diff-lineno">Line 11:</td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>{{NB|tip|Setting {{TAG|ML_ISTART}} {{=}} 1 together with {{TAG|NSW}} {{=}} 0 allows to repeat learning on the given training data and create a new force field in {{FILE|ML_FFN}} without actually performing additional MD steps. In this way force field parameters (e.g. cutoff radii, number of radial basis functions, etc.) can be varied without recalculating the entire trajectory. Moreover, because Bayesian error estimation is not required when no MD is run it is possible to switch the regression algorithm via the tag {{TAG|ML_IALGO_LINREG}} and check whether in this way better fitting results can be achieved. In order to avoid that the starting structure in the {{FILE|POSCAR}} file is processed and eventually added to the training data just set {{TAG|ML_CTIFOR}} to a large value (e.g. 1000).|:}}</div></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>{{NB|tip|Setting {{TAG|ML_ISTART}} {{=}} 1 together with {{TAG|NSW}} {{=}} 0 allows to repeat learning on the given training data and create a new force field in {{FILE|ML_FFN}} without actually performing additional MD steps. In this way force field parameters (e.g. cutoff radii, number of radial basis functions, etc.) can be varied without recalculating the entire trajectory. Moreover, because Bayesian error estimation is not required when no MD is run it is possible to switch the regression algorithm via the tag {{TAG|ML_IALGO_LINREG}} and check whether in this way better fitting results can be achieved. In order to avoid that the starting structure in the {{FILE|POSCAR}} file is processed and eventually added to the training data just set {{TAG|ML_CTIFOR}} to a large value (e.g. 1000).|:}}</div></td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>*{{TAG|ML_ISTART}} = 2: Prediction only. In this mode the previously trained machine learning force field is read from the {{FILE|ML_FF}} file. The MD simulation is driven with predictions from the force field only, no ab initio calculations are performed and no learning is executed. However, in order to monitor the quality of predictions the Bayesian error estimate of forces is still computed and logged in the {{FILE|ML_LOGFILE}}. This setting is typically used when the machine learning force field is considered mature and ready for production runs.</div></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>*{{TAG|ML_ISTART}} = 2: Prediction only. In this mode the previously trained machine learning force field is read from the {{FILE|ML_FF}} file. The MD simulation is driven with predictions from the force field only, no ab initio calculations are performed and no learning is executed. However, in order to monitor the quality of predictions the Bayesian error estimate of forces is still computed and logged in the {{FILE|ML_LOGFILE}}. This setting is typically used when the machine learning force field is considered mature and ready for production runs.</div></td></tr> <tr><td class="diff-marker" data-marker="−"></td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #ffe49c; vertical-align: top; white-space: pre-wrap;"><div>*{{TAG|ML_ISTART}} = 3: Learning from given ab initio data only, no MD time steps. In this operation mode a new machine learning force field is generated from ab initio data provided in the {{FILE|ML_AB}} file. The structures are read in and processed one by one as if harvested via an MD simulation. In other words, the same steps are performed as in on-the-fly training but the source of data is not an MD run but the series of structures available in {{FILE|ML_AB}}. This operation mode can be used to generate {{VASP}} machine learning force fields from pre-computed or external ab initio data sets. At first glance {{TAG|ML_ISTART}} = 3 looks very similar to the combination of {{TAG|ML_ISTART}} = 1 and {{TAG|NSW}} = 0 described above. However, there is an important difference: Setting {{TAG|ML_ISTART}} = 3 will ignore the list of local reference configurations in the {{FILE|ML_AB}} file and instead will determine a new collection which is written to the resulting {{FILE|ML_ABN}} file. {{NB|tip|If calculations for {{TAG|ML_ISTART}} {{=}} 3 are too time-consuming using the default settings, it is useful to increase {{TAG|ML_MCONF_NEW}} to values around 10-16 and set  {{TAG|ML_CDOUB}} {{=}} 4. This often accelerates the calculations a factor 2-4.|:}}</div></td><td class="diff-marker" data-marker="+"></td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;"><div>*{{TAG|ML_ISTART}} = 3: Learning from given ab initio data only, no MD time steps. In this operation mode a new machine learning force field is generated from ab initio data provided in the {{FILE|ML_AB}} file. The structures are read in and processed one by one as if harvested via an MD simulation. In other words, the same steps are performed as in on-the-fly training but the source of data is not an MD run but the series of structures available in {{FILE|ML_AB}}. This operation mode can be used to generate {{VASP}} machine learning force fields from pre-computed or external ab initio data sets. At first glance {{TAG|ML_ISTART}} = 3 looks very similar to the combination of {{TAG|ML_ISTART}} = 1 and {{TAG|NSW}} = 0 described above. However, there is an important difference: Setting {{TAG|ML_ISTART}} = 3 will ignore the list of local reference configurations in the {{FILE|ML_AB}} file and instead will determine a new collection which is written to the resulting {{FILE|ML_ABN}} file. {{NB|tip|If calculations for {{TAG|ML_ISTART}} {{=}} 3 are too time-consuming using the default settings, it is useful to increase {{TAG|ML_MCONF_NEW}} to values around 10-16 and set  {{TAG|ML_CDOUB}} {{=}} 4. This often accelerates the calculations a factor 2-4.|:}} <ins style="font-weight: bold; text-decoration: none;">The </ins>{{<ins style="font-weight: bold; text-decoration: none;">TAG</ins>|<ins style="font-weight: bold; text-decoration: none;">ML_AB}} file may contain values for ''CTIFOR'' for each training structure. These are the thresholds used to sample that structure from the previous training. If a value for {{TAG</ins>|<ins style="font-weight: bold; text-decoration: none;">ML_CTIFOR}} </ins>is <ins style="font-weight: bold; text-decoration: none;">specified in the {{TAG|INCAR}} file, that value is then used </ins>and <ins style="font-weight: bold; text-decoration: none;">the thresholds from the {{TAG|ML_AB}} are ignored. Otherwise: 1) If thresholds exist </ins>in the {{<ins style="font-weight: bold; text-decoration: none;">TAG</ins>|ML_AB}} <ins style="font-weight: bold; text-decoration: none;">they are used</ins>. <ins style="font-weight: bold; text-decoration: none;">2) If no thresholds are specified the default value for {{TAG</ins>|<ins style="font-weight: bold; text-decoration: none;">ML_CTIFOR</ins>}} <ins style="font-weight: bold; text-decoration: none;">is used. </ins></div></td></tr> <tr><td class="diff-marker" data-marker="−"></td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #ffe49c; vertical-align: top; white-space: pre-wrap;"><div>{{<del style="font-weight: bold; text-decoration: none;">NB</del>|<del style="font-weight: bold; text-decoration: none;">warning</del>|<del style="font-weight: bold; text-decoration: none;">This mode of operation </del>is <del style="font-weight: bold; text-decoration: none;">experimental </del>and <del style="font-weight: bold; text-decoration: none;">not well tested! Problems may arise, </del>in <del style="font-weight: bold; text-decoration: none;">particular if </del>the {{<del style="font-weight: bold; text-decoration: none;">FILE</del>|ML_AB}} <del style="font-weight: bold; text-decoration: none;">file contains structures with mixed number of elements and atom numbers</del>.|<del style="font-weight: bold; text-decoration: none;">:</del>}}</div></td><td colspan="2" class="diff-side-added"></td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>*{{TAG|ML_ISTART}} = 4: Refitting of the force field is done based on an existing {{TAG|ML_AB}} file, but the number of local reference configurations is taken from the {{TAG|ML_AB}} file. {{TAG|NSW}} on the input is ignored and only a single step is executed. No ab-initio calculation is carried out.  </div></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>*{{TAG|ML_ISTART}} = 4: Refitting of the force field is done based on an existing {{TAG|ML_AB}} file, but the number of local reference configurations is taken from the {{TAG|ML_AB}} file. {{TAG|NSW}} on the input is ignored and only a single step is executed. No ab-initio calculation is carried out.  </div></td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><br></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><br></td></tr> </table>

Karsai https://beta.vasp.at/wiki/index.php?title=ML_ISTART&diff=19825&oldid=prev 2023-03-27T13:45:59Z

<p></p> <table style="background-color: #fff; color: #202122;" data-mw="interface"> <col class="diff-marker" /> <col class="diff-content" /> <col class="diff-marker" /> <col class="diff-content" /> <tr class="diff-title" lang="en"> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">← Older revision</td> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">Revision as of 13:45, 27 March 2023</td> </tr><tr><td colspan="2" class="diff-lineno" id="mw-diff-left-l2">Line 2:</td> <td colspan="2" class="diff-lineno">Line 2:</td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>{{TAGDEF|ML_ISTART|[integer]|0}}</div></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>{{TAGDEF|ML_ISTART|[integer]|0}}</div></td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><br></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><br></td></tr> <tr><td class="diff-marker" data-marker="−"></td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #ffe49c; vertical-align: top; white-space: pre-wrap;"><div>{{NB|mind|This tag is deprecated and we advise to use {{<del style="font-weight: bold; text-decoration: none;">TAGML_MODE</del>}} instead.}}</div></td><td class="diff-marker" data-marker="+"></td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;"><div>{{NB|mind|This tag is deprecated and we advise to use {{<ins style="font-weight: bold; text-decoration: none;">TAG|ML_MODE</ins>}} instead.}}</div></td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>Description: This tag selects the mode of operation (e.g. start from scratch, prediction-only,...) of the machine learning force fields method.</div></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>Description: This tag selects the mode of operation (e.g. start from scratch, prediction-only,...) of the machine learning force fields method.</div></td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>----</div></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>----</div></td></tr> </table>

Karsai https://beta.vasp.at/wiki/index.php?title=ML_ISTART&diff=19824&oldid=prev 2023-03-27T13:45:49Z

<p></p> <table style="background-color: #fff; color: #202122;" data-mw="interface"> <col class="diff-marker" /> <col class="diff-content" /> <col class="diff-marker" /> <col class="diff-content" /> <tr class="diff-title" lang="en"> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">← Older revision</td> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">Revision as of 13:45, 27 March 2023</td> </tr><tr><td colspan="2" class="diff-lineno" id="mw-diff-left-l2">Line 2:</td> <td colspan="2" class="diff-lineno">Line 2:</td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>{{TAGDEF|ML_ISTART|[integer]|0}}</div></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>{{TAGDEF|ML_ISTART|[integer]|0}}</div></td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><br></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><br></td></tr> <tr><td colspan="2" class="diff-side-deleted"></td><td class="diff-marker" data-marker="+"></td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;"><div><ins style="font-weight: bold; text-decoration: none;">{{NB|mind|This tag is deprecated and we advise to use {{TAGML_MODE}} instead.}}</ins></div></td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>Description: This tag selects the mode of operation (e.g. start from scratch, prediction-only,...) of the machine learning force fields method.</div></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>Description: This tag selects the mode of operation (e.g. start from scratch, prediction-only,...) of the machine learning force fields method.</div></td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>----</div></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>----</div></td></tr> </table>

Karsai https://beta.vasp.at/wiki/index.php?title=ML_ISTART&diff=19340&oldid=prev 2023-01-25T11:33:08Z

<p></p> <table style="background-color: #fff; color: #202122;" data-mw="interface"> <col class="diff-marker" /> <col class="diff-content" /> <col class="diff-marker" /> <col class="diff-content" /> <tr class="diff-title" lang="en"> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">← Older revision</td> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">Revision as of 11:33, 25 January 2023</td> </tr><tr><td colspan="2" class="diff-lineno" id="mw-diff-left-l12">Line 12:</td> <td colspan="2" class="diff-lineno">Line 12:</td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>*{{TAG|ML_ISTART}} = 3: Learning from given ab initio data only, no MD time steps. In this operation mode a new machine learning force field is generated from ab initio data provided in the {{FILE|ML_AB}} file. The structures are read in and processed one by one as if harvested via an MD simulation. In other words, the same steps are performed as in on-the-fly training but the source of data is not an MD run but the series of structures available in {{FILE|ML_AB}}. This operation mode can be used to generate {{VASP}} machine learning force fields from pre-computed or external ab initio data sets. At first glance {{TAG|ML_ISTART}} = 3 looks very similar to the combination of {{TAG|ML_ISTART}} = 1 and {{TAG|NSW}} = 0 described above. However, there is an important difference: Setting {{TAG|ML_ISTART}} = 3 will ignore the list of local reference configurations in the {{FILE|ML_AB}} file and instead will determine a new collection which is written to the resulting {{FILE|ML_ABN}} file. {{NB|tip|If calculations for {{TAG|ML_ISTART}} {{=}} 3 are too time-consuming using the default settings, it is useful to increase {{TAG|ML_MCONF_NEW}} to values around 10-16 and set  {{TAG|ML_CDOUB}} {{=}} 4. This often accelerates the calculations a factor 2-4.|:}}</div></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>*{{TAG|ML_ISTART}} = 3: Learning from given ab initio data only, no MD time steps. In this operation mode a new machine learning force field is generated from ab initio data provided in the {{FILE|ML_AB}} file. The structures are read in and processed one by one as if harvested via an MD simulation. In other words, the same steps are performed as in on-the-fly training but the source of data is not an MD run but the series of structures available in {{FILE|ML_AB}}. This operation mode can be used to generate {{VASP}} machine learning force fields from pre-computed or external ab initio data sets. At first glance {{TAG|ML_ISTART}} = 3 looks very similar to the combination of {{TAG|ML_ISTART}} = 1 and {{TAG|NSW}} = 0 described above. However, there is an important difference: Setting {{TAG|ML_ISTART}} = 3 will ignore the list of local reference configurations in the {{FILE|ML_AB}} file and instead will determine a new collection which is written to the resulting {{FILE|ML_ABN}} file. {{NB|tip|If calculations for {{TAG|ML_ISTART}} {{=}} 3 are too time-consuming using the default settings, it is useful to increase {{TAG|ML_MCONF_NEW}} to values around 10-16 and set  {{TAG|ML_CDOUB}} {{=}} 4. This often accelerates the calculations a factor 2-4.|:}}</div></td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>{{NB|warning|This mode of operation is experimental and not well tested! Problems may arise, in particular if the {{FILE|ML_AB}} file contains structures with mixed number of elements and atom numbers.|:}}</div></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>{{NB|warning|This mode of operation is experimental and not well tested! Problems may arise, in particular if the {{FILE|ML_AB}} file contains structures with mixed number of elements and atom numbers.|:}}</div></td></tr> <tr><td colspan="2" class="diff-side-deleted"></td><td class="diff-marker" data-marker="+"></td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;"><div><ins style="font-weight: bold; text-decoration: none;">*{{TAG|ML_ISTART}} = 4: Refitting of the force field is done based on an existing {{TAG|ML_AB}} file, but the number of local reference configurations is taken from the {{TAG|ML_AB}} file. {{TAG|NSW}} on the input is ignored and only a single step is executed. No ab-initio calculation is carried out. </ins></div></td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><br></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><br></td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>== Related tags and articles ==</div></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>== Related tags and articles ==</div></td></tr> </table>

Karsai https://beta.vasp.at/wiki/index.php?title=ML_ISTART&diff=19062&oldid=prev 2022-10-13T14:35:24Z

<p></p> <table style="background-color: #fff; color: #202122;" data-mw="interface"> <col class="diff-marker" /> <col class="diff-content" /> <col class="diff-marker" /> <col class="diff-content" /> <tr class="diff-title" lang="en"> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">← Older revision</td> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">Revision as of 14:35, 13 October 2022</td> </tr><tr><td colspan="2" class="diff-lineno" id="mw-diff-left-l10">Line 10:</td> <td colspan="2" class="diff-lineno">Line 10:</td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>{{NB|tip|Setting {{TAG|ML_ISTART}} {{=}} 1 together with {{TAG|NSW}} {{=}} 0 allows to repeat learning on the given training data and create a new force field in {{FILE|ML_FFN}} without actually performing additional MD steps. In this way force field parameters (e.g. cutoff radii, number of radial basis functions, etc.) can be varied without recalculating the entire trajectory. Moreover, because Bayesian error estimation is not required when no MD is run it is possible to switch the regression algorithm via the tag {{TAG|ML_IALGO_LINREG}} and check whether in this way better fitting results can be achieved. In order to avoid that the starting structure in the {{FILE|POSCAR}} file is processed and eventually added to the training data just set {{TAG|ML_CTIFOR}} to a large value (e.g. 1000).|:}}</div></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>{{NB|tip|Setting {{TAG|ML_ISTART}} {{=}} 1 together with {{TAG|NSW}} {{=}} 0 allows to repeat learning on the given training data and create a new force field in {{FILE|ML_FFN}} without actually performing additional MD steps. In this way force field parameters (e.g. cutoff radii, number of radial basis functions, etc.) can be varied without recalculating the entire trajectory. Moreover, because Bayesian error estimation is not required when no MD is run it is possible to switch the regression algorithm via the tag {{TAG|ML_IALGO_LINREG}} and check whether in this way better fitting results can be achieved. In order to avoid that the starting structure in the {{FILE|POSCAR}} file is processed and eventually added to the training data just set {{TAG|ML_CTIFOR}} to a large value (e.g. 1000).|:}}</div></td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>*{{TAG|ML_ISTART}} = 2: Prediction only. In this mode the previously trained machine learning force field is read from the {{FILE|ML_FF}} file. The MD simulation is driven with predictions from the force field only, no ab initio calculations are performed and no learning is executed. However, in order to monitor the quality of predictions the Bayesian error estimate of forces is still computed and logged in the {{FILE|ML_LOGFILE}}. This setting is typically used when the machine learning force field is considered mature and ready for production runs.</div></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>*{{TAG|ML_ISTART}} = 2: Prediction only. In this mode the previously trained machine learning force field is read from the {{FILE|ML_FF}} file. The MD simulation is driven with predictions from the force field only, no ab initio calculations are performed and no learning is executed. However, in order to monitor the quality of predictions the Bayesian error estimate of forces is still computed and logged in the {{FILE|ML_LOGFILE}}. This setting is typically used when the machine learning force field is considered mature and ready for production runs.</div></td></tr> <tr><td class="diff-marker" data-marker="−"></td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #ffe49c; vertical-align: top; white-space: pre-wrap;"><div>*{{TAG|ML_ISTART}} = 3: Learning from given ab initio data only, no MD time steps. In this operation mode a new machine learning force field is generated from ab initio data provided in the {{FILE|ML_AB}} file. The structures are read in and processed one by one as if harvested via an MD simulation. In other words, the same steps are performed as in on-the-fly training but the source of data is not an MD run but the series of structures available in {{FILE|ML_AB}}. This operation mode can be used to generate {{VASP}} machine learning force fields from pre-computed or external ab initio data sets. At first glance {{TAG|ML_ISTART}} = 3 looks very similar to the combination of {{TAG|ML_ISTART}} = 1 and {{TAG|NSW}} = 0 described above. However, there is an important difference: Setting {{TAG|ML_ISTART}} = 3 will ignore the list of local reference configurations in the {{FILE|ML_AB}} file and instead will determine a new collection which is written to the resulting {{FILE|ML_ABN}} file. If calculations for {{TAG|ML_ISTART}} = 3 are too <del style="font-weight: bold; text-decoration: none;">timeconsuming </del>using the default settings, it is useful to increase <del style="font-weight: bold; text-decoration: none;"> </del>{{TAG|ML_MCONF_NEW}} to values around 10-16 and set  {{TAG|ML_CDOUB}} = 4. This often accelerates the calculations a factor 2-4.</div></td><td class="diff-marker" data-marker="+"></td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;"><div>*{{TAG|ML_ISTART}} = 3: Learning from given ab initio data only, no MD time steps. In this operation mode a new machine learning force field is generated from ab initio data provided in the {{FILE|ML_AB}} file. The structures are read in and processed one by one as if harvested via an MD simulation. In other words, the same steps are performed as in on-the-fly training but the source of data is not an MD run but the series of structures available in {{FILE|ML_AB}}. This operation mode can be used to generate {{VASP}} machine learning force fields from pre-computed or external ab initio data sets. At first glance {{TAG|ML_ISTART}} = 3 looks very similar to the combination of {{TAG|ML_ISTART}} = 1 and {{TAG|NSW}} = 0 described above. However, there is an important difference: Setting {{TAG|ML_ISTART}} = 3 will ignore the list of local reference configurations in the {{FILE|ML_AB}} file and instead will determine a new collection which is written to the resulting {{FILE|ML_ABN}} file. <ins style="font-weight: bold; text-decoration: none;">{{NB|tip|</ins>If calculations for {{TAG|ML_ISTART}} <ins style="font-weight: bold; text-decoration: none;">{{</ins>=<ins style="font-weight: bold; text-decoration: none;">}} </ins>3 are too <ins style="font-weight: bold; text-decoration: none;">time-consuming </ins>using the default settings, it is useful to increase {{TAG|ML_MCONF_NEW}} to values around 10-16 and set  {{TAG|ML_CDOUB}} <ins style="font-weight: bold; text-decoration: none;">{{</ins>=<ins style="font-weight: bold; text-decoration: none;">}} </ins>4. This often accelerates the calculations a factor 2-4.<ins style="font-weight: bold; text-decoration: none;">|:}}</ins></div></td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>{{NB|warning|This mode of operation is experimental and not well tested! Problems may arise, in particular if the {{FILE|ML_AB}} file contains structures with mixed number of elements and atom numbers.|:}}</div></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>{{NB|warning|This mode of operation is experimental and not well tested! Problems may arise, in particular if the {{FILE|ML_AB}} file contains structures with mixed number of elements and atom numbers.|:}}</div></td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><br></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><br></td></tr> </table>

Singraber https://beta.vasp.at/wiki/index.php?title=ML_ISTART&diff=18928&oldid=prev 2022-08-26T13:28:39Z

<p></p> <table style="background-color: #fff; color: #202122;" data-mw="interface"> <col class="diff-marker" /> <col class="diff-content" /> <col class="diff-marker" /> <col class="diff-content" /> <tr class="diff-title" lang="en"> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">← Older revision</td> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">Revision as of 13:28, 26 August 2022</td> </tr><tr><td colspan="2" class="diff-lineno" id="mw-diff-left-l11">Line 11:</td> <td colspan="2" class="diff-lineno">Line 11:</td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>*{{TAG|ML_ISTART}} = 2: Prediction only. In this mode the previously trained machine learning force field is read from the {{FILE|ML_FF}} file. The MD simulation is driven with predictions from the force field only, no ab initio calculations are performed and no learning is executed. However, in order to monitor the quality of predictions the Bayesian error estimate of forces is still computed and logged in the {{FILE|ML_LOGFILE}}. This setting is typically used when the machine learning force field is considered mature and ready for production runs.</div></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>*{{TAG|ML_ISTART}} = 2: Prediction only. In this mode the previously trained machine learning force field is read from the {{FILE|ML_FF}} file. The MD simulation is driven with predictions from the force field only, no ab initio calculations are performed and no learning is executed. However, in order to monitor the quality of predictions the Bayesian error estimate of forces is still computed and logged in the {{FILE|ML_LOGFILE}}. This setting is typically used when the machine learning force field is considered mature and ready for production runs.</div></td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>*{{TAG|ML_ISTART}} = 3: Learning from given ab initio data only, no MD time steps. In this operation mode a new machine learning force field is generated from ab initio data provided in the {{FILE|ML_AB}} file. The structures are read in and processed one by one as if harvested via an MD simulation. In other words, the same steps are performed as in on-the-fly training but the source of data is not an MD run but the series of structures available in {{FILE|ML_AB}}. This operation mode can be used to generate {{VASP}} machine learning force fields from pre-computed or external ab initio data sets. At first glance {{TAG|ML_ISTART}} = 3 looks very similar to the combination of {{TAG|ML_ISTART}} = 1 and {{TAG|NSW}} = 0 described above. However, there is an important difference: Setting {{TAG|ML_ISTART}} = 3 will ignore the list of local reference configurations in the {{FILE|ML_AB}} file and instead will determine a new collection which is written to the resulting {{FILE|ML_ABN}} file. If calculations for {{TAG|ML_ISTART}} = 3 are too timeconsuming using the default settings, it is useful to increase  {{TAG|ML_MCONF_NEW}} to values around 10-16 and set  {{TAG|ML_CDOUB}} = 4. This often accelerates the calculations a factor 2-4.</div></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>*{{TAG|ML_ISTART}} = 3: Learning from given ab initio data only, no MD time steps. In this operation mode a new machine learning force field is generated from ab initio data provided in the {{FILE|ML_AB}} file. The structures are read in and processed one by one as if harvested via an MD simulation. In other words, the same steps are performed as in on-the-fly training but the source of data is not an MD run but the series of structures available in {{FILE|ML_AB}}. This operation mode can be used to generate {{VASP}} machine learning force fields from pre-computed or external ab initio data sets. At first glance {{TAG|ML_ISTART}} = 3 looks very similar to the combination of {{TAG|ML_ISTART}} = 1 and {{TAG|NSW}} = 0 described above. However, there is an important difference: Setting {{TAG|ML_ISTART}} = 3 will ignore the list of local reference configurations in the {{FILE|ML_AB}} file and instead will determine a new collection which is written to the resulting {{FILE|ML_ABN}} file. If calculations for {{TAG|ML_ISTART}} = 3 are too timeconsuming using the default settings, it is useful to increase  {{TAG|ML_MCONF_NEW}} to values around 10-16 and set  {{TAG|ML_CDOUB}} = 4. This often accelerates the calculations a factor 2-4.</div></td></tr> <tr><td class="diff-marker" data-marker="−"></td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #ffe49c; vertical-align: top; white-space: pre-wrap;"><div> </div></td><td class="diff-marker" data-marker="+"></td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;"><div>{{NB|warning|This mode of operation is experimental and <ins style="font-weight: bold; text-decoration: none;">not well tested</ins>! Problems may arise, in particular if the {{FILE|ML_AB}} file contains structures with mixed number of elements and atom numbers.|:}}</div></td></tr> <tr><td class="diff-marker" data-marker="−"></td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #ffe49c; vertical-align: top; white-space: pre-wrap;"><div>{{NB|warning|This mode of operation is experimental and <del style="font-weight: bold; text-decoration: none;">still under development</del>! Problems may arise, in particular if the {{FILE|ML_AB}} file contains structures with mixed number of elements and atom numbers.|:}}</div></td><td colspan="2" class="diff-side-added"></td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><br></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><br></td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>== Related tags and articles ==</div></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>== Related tags and articles ==</div></td></tr> </table>

Karsai https://beta.vasp.at/wiki/index.php?title=ML_ISTART&diff=18927&oldid=prev 2022-08-26T13:28:16Z

<p></p> <table style="background-color: #fff; color: #202122;" data-mw="interface"> <col class="diff-marker" /> <col class="diff-content" /> <col class="diff-marker" /> <col class="diff-content" /> <tr class="diff-title" lang="en"> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">← Older revision</td> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">Revision as of 13:28, 26 August 2022</td> </tr><tr><td colspan="2" class="diff-lineno" id="mw-diff-left-l12">Line 12:</td> <td colspan="2" class="diff-lineno">Line 12:</td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>*{{TAG|ML_ISTART}} = 3: Learning from given ab initio data only, no MD time steps. In this operation mode a new machine learning force field is generated from ab initio data provided in the {{FILE|ML_AB}} file. The structures are read in and processed one by one as if harvested via an MD simulation. In other words, the same steps are performed as in on-the-fly training but the source of data is not an MD run but the series of structures available in {{FILE|ML_AB}}. This operation mode can be used to generate {{VASP}} machine learning force fields from pre-computed or external ab initio data sets. At first glance {{TAG|ML_ISTART}} = 3 looks very similar to the combination of {{TAG|ML_ISTART}} = 1 and {{TAG|NSW}} = 0 described above. However, there is an important difference: Setting {{TAG|ML_ISTART}} = 3 will ignore the list of local reference configurations in the {{FILE|ML_AB}} file and instead will determine a new collection which is written to the resulting {{FILE|ML_ABN}} file. If calculations for {{TAG|ML_ISTART}} = 3 are too timeconsuming using the default settings, it is useful to increase  {{TAG|ML_MCONF_NEW}} to values around 10-16 and set  {{TAG|ML_CDOUB}} = 4. This often accelerates the calculations a factor 2-4.</div></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>*{{TAG|ML_ISTART}} = 3: Learning from given ab initio data only, no MD time steps. In this operation mode a new machine learning force field is generated from ab initio data provided in the {{FILE|ML_AB}} file. The structures are read in and processed one by one as if harvested via an MD simulation. In other words, the same steps are performed as in on-the-fly training but the source of data is not an MD run but the series of structures available in {{FILE|ML_AB}}. This operation mode can be used to generate {{VASP}} machine learning force fields from pre-computed or external ab initio data sets. At first glance {{TAG|ML_ISTART}} = 3 looks very similar to the combination of {{TAG|ML_ISTART}} = 1 and {{TAG|NSW}} = 0 described above. However, there is an important difference: Setting {{TAG|ML_ISTART}} = 3 will ignore the list of local reference configurations in the {{FILE|ML_AB}} file and instead will determine a new collection which is written to the resulting {{FILE|ML_ABN}} file. If calculations for {{TAG|ML_ISTART}} = 3 are too timeconsuming using the default settings, it is useful to increase  {{TAG|ML_MCONF_NEW}} to values around 10-16 and set  {{TAG|ML_CDOUB}} = 4. This often accelerates the calculations a factor 2-4.</div></td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><br></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><br></td></tr> <tr><td class="diff-marker" data-marker="−"></td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #ffe49c; vertical-align: top; white-space: pre-wrap;"><div>{{NB|warning|<del style="font-weight: bold; text-decoration: none;">{{TAG|ML_ISTART}}=3 </del>is experimental and still under development! Problems may arise, in particular if the {{FILE|ML_AB}} file contains structures with mixed number of elements and atom numbers.|:}}</div></td><td class="diff-marker" data-marker="+"></td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;"><div>{{NB|warning|<ins style="font-weight: bold; text-decoration: none;">This mode of operation </ins>is experimental and still under development! Problems may arise, in particular if the {{FILE|ML_AB}} file contains structures with mixed number of elements and atom numbers.|:}}</div></td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><br></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><br></td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>== Related tags and articles ==</div></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>== Related tags and articles ==</div></td></tr> <!-- diff cache key mediawiki-vw:diff:1.41:old-18926:rev-18927:php=table --> </table>

Karsai