Improve Lobster workflow preconverge step, kpoints, docs #277

JaGeo · 2023-03-21T20:04:43Z

Summary

I have now implemented some of the suggestions from #275 (same number of bands in preconverge static step and static step).

I also changed the number of kpoints. Again, I think some more tests are necessary to confirm that this reciprocal density of kpoints is emough.

I added a tutorial on how to access the lobsterpy plots from the database.

…preconverge

JaGeo · 2023-03-21T20:11:07Z

@utf, maybe, we could also check if auto_ispin is useful in this workflow as well.

codecov · 2023-03-21T20:28:58Z

Codecov Report

Merging #277 (33133e5) into main (6c1f611) will decrease coverage by 0.13%.
The diff coverage is 100.00%.

❗ Current head 33133e5 differs from pull request most recent head 6532893. Consider uploading reports for the commit 6532893 to get more accurate results

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #277      +/-   ##
==========================================
- Coverage   65.90%   65.78%   -0.13%     
==========================================
  Files          73       73              
  Lines        7048     7049       +1     
  Branches      901      900       -1     
==========================================
- Hits         4645     4637       -8     
- Misses       2137     2149      +12     
+ Partials      266      263       -3

Impacted Files	Coverage Δ
src/atomate2/lobster/schemas.py	`92.82% <ø> (ø)`
src/atomate2/vasp/flows/lobster.py	`96.36% <100.00%> (+1.91%)`	⬆️
src/atomate2/vasp/jobs/lobster.py	`95.08% <100.00%> (+1.63%)`	⬆️

... and 1 file with indirect coverage changes

JaGeo · 2023-03-21T22:07:46Z

Still not happy with the preconverge step. Will try to work on it further.

utf · 2023-03-21T22:09:12Z

Ok! I have a feeling using the NonSCF uniform maker will do the trick!

JaGeo · 2023-03-21T22:34:20Z

Yes, you are probably right. I will check it out tomorrow!

JaGeo · 2023-03-22T10:55:50Z

Test update (LobsterStaticMaker and LobsterUniformMaker)
check output schema for uuids and directories
check deletion of WAVECARs
remove WAVECAR copying from LobsterStaticMaker
Update incar_settings

utf · 2023-03-22T10:59:26Z

src/atomate2/vasp/flows/lobster.py

+            input_set_generator=StaticSetGenerator(
+                user_kpoints_settings={"reciprocal_density": 100},
+                user_incar_settings={
+                    "IBRION": 2,


I think you can remove IBRION, NSW, LCHARG, and ISIF as these are set by the static maker anyway (and IBRION, ISIF are not important with NSW = 0)

utf · 2023-03-22T10:59:40Z

src/atomate2/vasp/flows/lobster.py

+                user_kpoints_settings={"reciprocal_density": 400},
+                user_incar_settings={
+                    "LWAVE": True,
+                    "NSW": 0,


Similarly you can remove NSW and ICHARG here too.

Will ICHARG be set to 11 even if I give different incar settings?

EDIT: I will remove ICHARG as well. I somehow got lost with the settings.

utf · 2023-03-22T11:00:18Z

src/atomate2/vasp/jobs/lobster.py

@@ -61,14 +61,14 @@ class LobsterStaticMaker(BaseVaspMaker):
    name: str = "static_run"
    input_set_generator: VaspInputGenerator = field(
        default_factory=lambda: StaticSetGenerator(
-            user_kpoints_settings={"grid_density": 6000},
+            user_kpoints_settings={"reciprocal_density": 400},
            user_incar_settings={
                "IBRION": 2,


And IBRION, ISIF, NSW from here.

Short question: should I enable auto_ispin and things like this here as well?

update: I think I will just enable this as it will also save computational time.

Yeah, I think you can enable it. The default is for the lobster workflow to do a relaxation anyway, so I think it could be useful.

utf · 2023-03-22T11:00:49Z

src/atomate2/vasp/jobs/lobster.py

@@ -151,6 +159,8 @@ def update_user_incar_settings_maker(
        Structure object.
    prev_vasp_dir : Path or str
        Path or string to vasp files.
+    append_name: str or None


I guess you don't need append_name here since that was for the preconvergence stuff

Yes, you are right. I need to clean up.

JaGeo · 2023-03-22T12:35:41Z

I still need to pass the dir_names for the LobsterUniformMaker correctly. Deletion of WAVECAR.gz currently does not work fully. This is hopefully the last remaining issue with this new implementation.

Update: It seems to work fine! I just don't have the dir_name, uuid of the first static_maker of the LobsterUniformMaker in the final schema at the moment. I will try to find out how I can access this information as well.

JaGeo · 2023-03-22T12:53:46Z

src/atomate2/vasp/flows/lobster.py

+        )
+    )
+
+    def make(self, structure: Structure, prev_vasp_dir: str | Path | None = None):


Do you have a suggestion with regard to code reuse? I think I simply copied this part and just adapted the part above.

You could just subclass the UniformBandStructureMaker and remove the make function entirely?

In fact, probably even better is not even to subclass and just define the maker as a constant. E.g.

LOBSTER_UNIFORM_MAKER = UniformBandStructureMaker( name: str = "uniform lobster structure" static_maker: BaseVaspMaker = field( default_factory=lambda: StaticMaker( input_set_generator=StaticSetGenerator( user_kpoints_settings={"reciprocal_density": 100}, user_incar_settings={ "EDIFF": 1e-7, "LAECHG": False, "LVTOT": False, "LREAL": False, "ALGO": "Normal", "LWAVE": False, }, ) ) ) bs_maker: BaseVaspMaker = field( default_factory=lambda: NonSCFMaker( input_set_generator=NonSCFSetGenerator( user_kpoints_settings={"reciprocal_density": 400}, user_incar_settings={ "LWAVE": True, "ISYM": 0, }, ) ) ) )

Jimmy has done it this way in the phonon workflow and means we don't get too many makers. The line between adding a new maker vs reusing the old ones is quite ambitious though, so happy either way.

I like the constant in this case. I was thinking about subclassing but this would have added another layer of complexity as well.

(btw, I also like that we still get the VASP dos in this case. It also allows to compare Lobster and VASP dos if necessary. thus, I will keep the higher NEDOS setting)

JaGeo · 2023-03-22T13:24:15Z

@utf to me it looks like I currently can only access the output from the uniform_job and nothing else in the UniformBandStructureMaker. Shouldn't we add the dir_name and uuid from the static_job before as well?

atomate2/src/atomate2/vasp/flows/core.py

Line 224 in 6c1f611

return Flow(jobs, uniform_job.output, name=self.name)

JaGeo · 2023-03-22T13:28:09Z

Except for this issue, I am happy with the code for now. I very much hope that this version of the workflow is now fine for standard runs and also showcases the advantages of a workflow manager.

utf · 2023-03-22T13:52:58Z

I think its ok that you only store the dir_name and uuid of the non-scf step. If you really need it you could use the uuid of the non-scf to get the inputs of that job and get the static dir.

If you really want to include them we could add a all_outputs property to the Flow class in jobflow but this could be quite brittle if your flow maker has an job which isn't a vasp job.

JaGeo · 2023-03-22T13:55:00Z

Thanks! Then, I will leave it this way for now :).
I also found another tiny issue in the schema: when there are no cation-anions, one of the lobsterpy runs will fail and of course no text can be generated. It's fixed in this version.

utf · 2023-03-22T14:14:12Z

Great, thank you for these changes!

JaGeo and others added 8 commits March 21, 2023 15:50

Adapt number of bands in preconverge step

5feb790

Fix tests

1a5de31

Merge branch 'materialsproject:main' into Fix_preconverge

7614edc

Fix tests

77df27d

Merge branch 'Fix_preconverge' of github.com:JaGeo/atomate2 into Fix_…

e50fb86

…preconverge

Change kpoint schmema

8641507

Tutorial

28c6965

precommit

e760d7e

JaGeo changed the title ~~Fix preconverge~~ Improve Lobster workflow preconverge step, kpoints, docs Mar 21, 2023

inrease density

3c28a83

First pass at implementing LobsterUniformMaker

8c4141c

utf reviewed Mar 22, 2023

View reviewed changes

JaGeo added 6 commits March 22, 2023 12:03

Remove append_name

992e21d

Fix tests

eb5ce54

Fix precommit

79e56c3

Fix schema if describe is None

1a451e8

Add tests for uniform maker

4f0f65e

Fix linting

33133e5

JaGeo commented Mar 22, 2023

View reviewed changes

make constant

6532893

utf merged commit 161f685 into materialsproject:main Mar 22, 2023

utf mentioned this pull request Aug 14, 2023

Minting a new Atomate2 version #467

Closed

utf added the enhancement Improvements to existing features label Sep 1, 2023

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Improve Lobster workflow preconverge step, kpoints, docs #277

Improve Lobster workflow preconverge step, kpoints, docs #277

JaGeo commented Mar 21, 2023

JaGeo commented Mar 21, 2023

codecov bot commented Mar 21, 2023 •

edited

Loading

JaGeo commented Mar 21, 2023

utf commented Mar 21, 2023

JaGeo commented Mar 21, 2023

JaGeo commented Mar 22, 2023 •

edited

Loading

utf Mar 22, 2023

utf Mar 22, 2023

JaGeo Mar 22, 2023 •

edited

Loading

utf Mar 22, 2023

JaGeo Mar 22, 2023 •

edited

Loading

utf Mar 22, 2023

utf Mar 22, 2023

JaGeo Mar 22, 2023

JaGeo commented Mar 22, 2023 •

edited

Loading

JaGeo Mar 22, 2023

utf Mar 22, 2023

utf Mar 22, 2023

utf Mar 22, 2023

JaGeo Mar 22, 2023

JaGeo Mar 22, 2023

JaGeo commented Mar 22, 2023 •

edited

Loading

JaGeo commented Mar 22, 2023 •

edited

Loading

utf commented Mar 22, 2023

JaGeo commented Mar 22, 2023 •

edited

Loading

utf commented Mar 22, 2023

Improve Lobster workflow preconverge step, kpoints, docs #277

Improve Lobster workflow preconverge step, kpoints, docs #277

Conversation

JaGeo commented Mar 21, 2023

Summary

JaGeo commented Mar 21, 2023

codecov bot commented Mar 21, 2023 • edited Loading

Codecov Report

JaGeo commented Mar 21, 2023

utf commented Mar 21, 2023

JaGeo commented Mar 21, 2023

JaGeo commented Mar 22, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

JaGeo Mar 22, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

JaGeo Mar 22, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

JaGeo commented Mar 22, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

JaGeo commented Mar 22, 2023 • edited Loading

JaGeo commented Mar 22, 2023 • edited Loading

utf commented Mar 22, 2023

JaGeo commented Mar 22, 2023 • edited Loading

utf commented Mar 22, 2023

codecov bot commented Mar 21, 2023 •

edited

Loading

JaGeo commented Mar 22, 2023 •

edited

Loading

JaGeo Mar 22, 2023 •

edited

Loading

JaGeo Mar 22, 2023 •

edited

Loading

JaGeo commented Mar 22, 2023 •

edited

Loading

JaGeo commented Mar 22, 2023 •

edited

Loading

JaGeo commented Mar 22, 2023 •

edited

Loading

JaGeo commented Mar 22, 2023 •

edited

Loading