High pressure crystals structure relaxations by JonathanSchmidt1 · Pull Request #340 · ddmms/ml-peg

JonathanSchmidt1 · 2026-02-04T12:06:55Z

Pre-review checklist for PR author

PR authors must check the boxes below when creating the PR.

I've confirmed the contribution guidelines.

Summary

Adds a new high-pressure relaxation benchmark that evaluates MLIP performance in predicting equilibrium structures under hydrostatic pressure.

3000 structures from the Alexandria database
Relaxed at 7 pressure conditions:
0, 25, 50, 75, 100, 125, 150 GPa
Results compared against PBE reference calculations

Metrics evaluated (at each pressure)

Volume per atom MAE (Å³/atom)
Enthalpy per atom MAE (eV/atom), calculated as:
H = E + PV
Convergence rate (%)

Methodology

Structures relaxed using janus-core's GeomOpt
ASE FixSymmetry constraint used to preserve crystallographic symmetry (matching DFT protocol)
FrechetCellFilter with specified scalar pressure
Force convergence threshold: 0.0002 eV/Å
Maximum of 500 steps
Up to 3 consecutive relaxation attempts if not converged
Obvious outliers based on the predicted energies are marked as non-converged and are part of the convergence percentage metric but not the MAEs

Linked issue

Resolves #338

Progress

Testing

Tested on the following models:

mace-mpa-0
orb-v3-consv-inf-mpa
Regression test comparing to results produced with scripts from the original paper for mace-mpa-0

New decorators / callbacks

No new decorators or callbacks required.

Note the pull request right now includes a new model

orb-v3-consv-inf-mpa
because I wanted to compare to the original paper but this could be removed.
I am looking for some feedback on the good/bad values for the metrics

…l-peg into high-pressure

Analysis: - Replace parity plots with density scatter via plot_density_scatter decorator - Add high_pressure_stats fixture pooling all pressures per model - Use build_density_inputs for volume/energy density plots - Rename output files to figure_volume_density.json / figure_energy_density.json Calc: - Write relaxed structures to relaxed_{pressure_label}.xyz per model/pressure - Annotate relaxed atoms with mat_id and pressure_gpa in atoms.info App: - Update plot filenames to match new density plot filenames Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

…l-peg into high-pressure

Co-authored-by: Elliott Kasoar <45317199+ElliottKasoar@users.noreply.github.com>

…utils/decorators.py Co-authored-by: Elliott Kasoar <45317199+ElliottKasoar@users.noreply.github.com>

joehart2001 · 2026-03-24T16:37:34Z

Hey @JonathanSchmidt1, thanks for the PR, its looking in great shape!

Some points:

It would be good to put a lot of the output information as part of the atoms.info like done in other benchmarks, as we're saving the xyz files for the structure visualisation anyway too. I know this could be slightly tricky with the convergence metric, and maybe only the csv will just contain this convergence info, whilst the atoms object can contain the rest. Would be good to know what you think about this
I've now implemented structure visualisation from density scatters, so it would be good to get that in. Take a look at e.g. NCIA_D1200 analysis and app. If you need any help with this then let me know and i can make a code suggestion. I've seen the suggestion you've made about the ordering of points, and i plan to look at that soon.

Sorry for the delay in getting back to you

Co-authored-by: ElliottKasoar <45317199+ElliottKasoar@users.noreply.github.com>

Co-authored-by: Elliott Kasoar <45317199+ElliottKasoar@users.noreply.github.com>

Co-authored-by: Domantas Kuryla <dk584@icepc16> Co-authored-by: ElliottKasoar <45317199+ElliottKasoar@users.noreply.github.com> Co-authored-by: Joseph Hart <92541539+joehart2001@users.noreply.github.com>

Co-authored-by: Elliott Kasoar <45317199+ElliottKasoar@users.noreply.github.com>

Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

Co-authored-by: ElliottKasoar <45317199+ElliottKasoar@users.noreply.github.com>

Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

Co-authored-by: ElliottKasoar <45317199+ElliottKasoar@users.noreply.github.com>

Co-authored-by: Elliott Kasoar <45317199+ElliottKasoar@users.noreply.github.com>

Co-authored-by: joehart2001 <jh2536@cam.ac.uk>

joehart2001 · 2026-04-13T15:18:57Z

+
+    return relaxed, converged, enthalpy / n_atoms
+
+


Suggested change

@pytest.mark.very_slow

joehart2001 · 2026-04-13T15:31:00Z

+
+        vol_trajs = collect_traj_assets(
+            data_path=DATA_PATH,
+            assets_prefix="assets/bulk_crystal/high_pressure_relaxation",


Suggested change

assets_prefix="assets/bulk_crystal/high_pressure_relaxation",

assets_prefix="/assets/bulk_crystal/high_pressure_relaxation",

joehart2001 · 2026-04-13T15:31:13Z

+        )
+        energy_trajs = collect_traj_assets(
+            data_path=DATA_PATH,
+            assets_prefix="assets/bulk_crystal/high_pressure_relaxation",


Suggested change

assets_prefix="assets/bulk_crystal/high_pressure_relaxation",

assets_prefix="/assets/bulk_crystal/high_pressure_relaxation",

joehart2001 · 2026-04-13T15:41:53Z

+                fmax=fmax,
+                steps=max_steps,
+                filter_kwargs={"scalar_pressure": pressure_gpa},
+                calc_kwargs={"default_dtype": "float64"},


@ElliottKasoar does this work in our case and is it robust? I would reccomend using model.default_dtype = "float64" before calling get calculator, like done here

Yes, unfortunately, generally this needs to be defined before the calculator is instantiated, and default_dtype is also not guaranteed to be the correct kwarg for a given calculator, so yes that linked implementation is safer.

(model.default_dtype sets our custom attribute that knows how to set it for the models we currently support, whereas janus-core's calc_kwargs would be passed straight to the relevant ASE calculator at the time it's set up)

joehart2001 · 2026-04-13T15:46:37Z

Thanks for addressing the comments, the density plots look great! Just a few small things.

The suggestion to the assets path is due to a recent change we made
It seems your rebase has resulted in all the changes being commited to this branch and github is not realising these are not your changes. Maybe a fresh fetch and rebase would sort it out?

Also, the current density plot shows one density plot for all the pressures. Would you like to keep it this way or have one for each pressure?

JonathanSchmidt1 and others added 9 commits February 2, 2026 22:25

add high pressure benchmark

1883421

add random subsampling

cb5c052

remove outliers

8ef30cc

add orb-v3-cons-inf-mpa for comparison with original benchmark

5bbeba1

Merge branch 'ddmms:main' into high-pressure

b80781a

add docs for benchmark

1d2fb2e

Merge branch 'high-pressure' of https://github.com/JonathanSchmidt1/m…

de0ff5a

…l-peg into high-pressure

fix good/bad metrics, use internal mae function

3a06816

metrics.yml update

b0c5d56

JonathanSchmidt1 changed the title ~~High pressure crystals structure relaxations~~ [WIP] High pressure crystals structure relaxations Feb 4, 2026

ElliottKasoar added the new benchmark Proposals and suggestions for new benchmarks label Feb 6, 2026

JonathanSchmidt1 marked this pull request as draft February 9, 2026 08:40

JonathanSchmidt1 marked this pull request as ready for review February 9, 2026 08:41

JonathanSchmidt1 changed the title ~~[WIP] High pressure crystals structure relaxations~~ High pressure crystals structure relaxations Feb 9, 2026

JonathanSchmidt1 and others added 9 commits February 16, 2026 11:51

fix regression test to rerun instead of using output data

8d87f2a

Merge branch 'main' into high-pressure

62fe094

add density plots and xyz storage of relaxed structures

b6eb62a

Merge branch 'ddmms:main' into high-pressure

3f3b0aa

sort plots before density plot so high density points end up on top

1b871fa

fix xyz storage by removing calculator

6b61b60

Merge branch 'high-pressure' of https://github.com/JonathanSchmidt1/m…

6b4544e

…l-peg into high-pressure

fix callback removing active cell selection/plot

89993ba

ElliottKasoar reviewed Mar 4, 2026

View reviewed changes

Comment thread ml_peg/analysis/utils/decorators.py Outdated

Comment thread ml_peg/models/models.yml Outdated

JonathanSchmidt1 and others added 2 commits March 4, 2026 18:47

remove extra model ml_peg/models/models.yml

ede9d9c

Co-authored-by: Elliott Kasoar <45317199+ElliottKasoar@users.noreply.github.com>

remove sorting for better vis of high density points ml_peg/analysis/…

d066570

…utils/decorators.py Co-authored-by: Elliott Kasoar <45317199+ElliottKasoar@users.noreply.github.com>

kuryla and others added 3 commits April 8, 2026 11:20

Add QUID benchmark (ddmms#402)

343ba02

Co-authored-by: ElliottKasoar <45317199+ElliottKasoar@users.noreply.github.com>

Add silicon interstitial NEBs benchmark (ddmms#326)

7c485c3

Co-authored-by: Elliott Kasoar <45317199+ElliottKasoar@users.noreply.github.com>

Float64 precision for static tests (ddmms#408)

8ccab53

Co-authored-by: Domantas Kuryla <dk584@icepc16> Co-authored-by: ElliottKasoar <45317199+ElliottKasoar@users.noreply.github.com> Co-authored-by: Joseph Hart <92541539+joehart2001@users.noreply.github.com>

joehart2001 and others added 24 commits April 8, 2026 11:20

Add structure visualisation for density scatter plots (ddmms#416)

0a68cdd

Co-authored-by: Elliott Kasoar <45317199+ElliottKasoar@users.noreply.github.com>

Update and add bounds for dependencies (ddmms#419)

31b6f19

Replace trained_on_d3 with trained_on_dispersion (ddmms#413)

d68b7d0

Bump tornado from 6.5.4 to 6.5.5 (ddmms#422)

cc30a2a

Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

Add link to tutorial notebook in readme (ddmms#415)

c77f6d7

Co-authored-by: ElliottKasoar <45317199+ElliottKasoar@users.noreply.github.com>

Bump orjson from 3.11.5 to 3.11.6 (ddmms#424)

6fd8e5e

Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

Bump pyjwt from 2.10.1 to 2.12.0 (ddmms#428)

04843b6

Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

Bump pyasn1 from 0.6.2 to 0.6.3 (ddmms#430)

f8b88de

Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

apply configured weights (ddmms#432)

bdaa6b8

Bump ujson from 5.11.0 to 5.12.0 (ddmms#435)

a8b89a0

Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

Bump pygments from 2.19.2 to 2.20.0 (ddmms#440)

7d94e70

Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

Bump cryptography from 46.0.3 to 46.0.6 (ddmms#439)

a93c221

Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

Bump requests from 2.32.5 to 2.33.0 (ddmms#438)

baa9da9

Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

Add model selection dropdown (ddmms#378)

2a0f622

Bump aiohttp from 3.13.3 to 3.13.4 (ddmms#442)

4b01d1c

Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

Add liquid densities benchmark (ddmms#388)

5eb3e99

Co-authored-by: ElliottKasoar <45317199+ElliottKasoar@users.noreply.github.com>

Add SBH17 surface barriers benchmark (ddmms#375)

f60a50d

Co-authored-by: Elliott Kasoar <45317199+ElliottKasoar@users.noreply.github.com>

Add water density benchmark (ddmms#423)

5b59b5e

Co-authored-by: Elliott Kasoar <45317199+ElliottKasoar@users.noreply.github.com>

Update Orb (ddmms#443)

26cfe01

Add navigation sidebar and pages (ddmms#441)

da29ed0

Co-authored-by: joehart2001 <jh2536@cam.ac.uk>

Tighten glob pattern to exclude pyc files (ddmms#448)

f49753d

Fix initial metric weight propagation to summary tables (ddmms#447)

587300f

Fix loading assets (ddmms#450)

165054c

xyz result storage, structure plotting from scatter

0703ed3

joehart2001 reviewed Apr 13, 2026

View reviewed changes

joehart2001 mentioned this pull request Apr 15, 2026

Low dimensional benchmark #355

Open

5 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

High pressure crystals structure relaxations#340

High pressure crystals structure relaxations#340
JonathanSchmidt1 wants to merge 51 commits intoddmms:mainfrom
JonathanSchmidt1:high-pressure

JonathanSchmidt1 commented Feb 4, 2026 •

edited

Loading

Uh oh!

Uh oh!

Uh oh!

joehart2001 commented Mar 24, 2026

Uh oh!

joehart2001 Apr 13, 2026

Uh oh!

joehart2001 Apr 13, 2026

Uh oh!

joehart2001 Apr 13, 2026

Uh oh!

joehart2001 Apr 13, 2026 •

edited

Loading

Uh oh!

ElliottKasoar Apr 13, 2026 •

edited

Loading

Uh oh!

joehart2001 commented Apr 13, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

	assets_prefix="assets/bulk_crystal/high_pressure_relaxation",
	assets_prefix="/assets/bulk_crystal/high_pressure_relaxation",

Conversation

JonathanSchmidt1 commented Feb 4, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Pre-review checklist for PR author

Summary

Metrics evaluated (at each pressure)

Methodology

Linked issue

Progress

Testing

New decorators / callbacks

Uh oh!

Uh oh!

Uh oh!

joehart2001 commented Mar 24, 2026

Uh oh!

joehart2001 Apr 13, 2026

Choose a reason for hiding this comment

Uh oh!

joehart2001 Apr 13, 2026

Choose a reason for hiding this comment

Uh oh!

joehart2001 Apr 13, 2026

Choose a reason for hiding this comment

Uh oh!

joehart2001 Apr 13, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

ElliottKasoar Apr 13, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

joehart2001 commented Apr 13, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

JonathanSchmidt1 commented Feb 4, 2026 •

edited

Loading

joehart2001 Apr 13, 2026 •

edited

Loading

ElliottKasoar Apr 13, 2026 •

edited

Loading