SparseGAP implementation - #76

Closed

agoscinski wants to merge 239 commits into

Commits on Nov 20, 2023

Initial commit
frostedoyster
authored
3177680
Update README.md
frostedoyster
authored
3e73d1f
Make the package installable
frostedoyster
committed
9d83073
Set up CI, linters and tests (#1)
frostedoyster
authored
a0222c9

Commits on Nov 21, 2023

Add scaffold of all types of model (#2)
frostedoyster
authored
8937336

Commits on Nov 29, 2023

Remove setup.py-dependent build test
frostedoyster
committed
77c6378
Debug CI
frostedoyster
committed
81b8d07
Allow ls for debugging
frostedoyster
committed
23e92c3
Retrieve sdist to build wheels from it
frostedoyster
committed
6ec7658
Debug again?
frostedoyster
committed
d931eeb
This works
frostedoyster
committed
b132499
Little cleanup for new purpose of the repo
PicoCentauri
committed
f81dc7d
Replace matplotlib dependency
frostedoyster
committed
9d8e3cf
Merge pull request #3 from lab-cosmo/skeleton
PicoCentauri
authored
fbdffd5

Commits on Nov 30, 2023

Add basic cli interface to scripts (#5)
PicoCentauri
authored
aa3c55e
Fix ci job syntax error (#9)
PicoCentauri
authored
9a939d3
Add xyz structure and target reader (#8)
PicoCentauri
authored
b9f0502

Commits on Dec 1, 2023

Temporarily disable failing Windows tests (#12)
frostedoyster
authored
5b18376
Implement a temporary dataset class (#10)
frostedoyster
authored
bc8437f

Commits on Dec 5, 2023

Add documentation structure (#13)

frostedoyster
and
PicoCentauri
authored
56224a2

Commits on Dec 6, 2023

Add SOAP-BPNN (#7)
frostedoyster
authored
9c632bb
Rename to `metatensor-models` (#15)
frostedoyster
authored
99ca38b

Commits on Dec 7, 2023

Extract species list from dataset (#17)
frostedoyster
authored
35a5996

Commits on Dec 8, 2023

Add hydra parsing (#18)
PicoCentauri
authored
09c58cd

Commits on Dec 11, 2023

Implement saving and loading of models (#19)

frostedoyster
and
PicoCentauri
authored
44e4f11

Commits on Dec 13, 2023

Update cli interface (#21)

PicoCentauri
and
frostedoyster
authored
dc2299b

Commits on Jan 3, 2024

Change CLI API (#24)
PicoCentauri
authored
1a98445
Make train function self-contained (#25)
frostedoyster
authored
3882faa

Commits on Jan 11, 2024

Integrate with `metatensor.torch.atomistic` (#28)
frostedoyster
authored
4d7f847

Commits on Jan 13, 2024

Add gradient calculators (#26)
frostedoyster
authored
51df872

Commits on Jan 19, 2024

Add parser and update readers for new options.yaml (#30)
PicoCentauri
authored
08844c5

Commits on Jan 20, 2024

Add tutorial how to override arch params (#32)
PicoCentauri
authored
9e894b3

Commits on Jan 24, 2024

Finalize training procedure (#31)

frostedoyster
and
PicoCentauri
authored
50c5e36
Add length_unit + cleanup (#35)
PicoCentauri
authored
2689dc1
Compile slice and join (#36)
frostedoyster
authored
26a5297

Commits on Jan 25, 2024

Fix wrong shapes in ase stress parsers (#38)
PicoCentauri
authored
1e487a9

Commits on Jan 26, 2024

Provide better logging (#37)
frostedoyster
authored
e07e2a8
Reactivate MacOS tests (#41)
frostedoyster
authored
e54c82a
Display `metatensor-torch` torchscript classes correctly in docs (#40)
frostedoyster
authored
cfa37b2

Commits on Jan 31, 2024

Fix wrong name assignment in parser (#42)

frostedoyster
and
PicoCentauri
authored
872dd0d

Commits on Feb 1, 2024

Fix BPNN training issues (#44)
frostedoyster
authored
fcc7c1f

Commits on Feb 2, 2024

Export models (#34)

frostedoyster
and
PicoCentauri
authored
74ff6a4
Add ASE tutorial (#43)

frostedoyster
and
PicoCentauri
authored
df96bb0

Commits on Feb 5, 2024

Clean up logging (#45)
frostedoyster
authored
cd9a8ab
added check_units for dataset_options with test (#50)
DavideTisi
authored
091cf1f

Commits on Feb 6, 2024

Fix typo
frostedoyster
committed
c4720ad
Improvements to hydra handling (#54)
PicoCentauri
authored
8b39820
Add option to continue training from checkpoint (#49)

PicoCentauri
and
frostedoyster
authored
89f5d36

Commits on Feb 7, 2024

Setting global random seeds based on user input (#60)
DavideTisi
authored
7421e4d
Test training logging (#61)

frostedoyster
and
PicoCentauri
authored
d109020

Commits on Feb 8, 2024

Make SOAP-BPNN run on GPU (#58)
frostedoyster
authored
bd906ab
More Pythonic API for train and export (#67)
PicoCentauri
authored
bb7efb3
Add docs on what `metatensor-models` is (#64)

authored
14bc244

Commits on Feb 9, 2024

Change how eval function works (#65)

PicoCentauri
and
frostedoyster
authored
cdb74f9
Apply new naming conventions (#63)
frostedoyster
authored
9ecf690
Use `metatensor-learn`'s `Dataset`, `DataLoader`, and collate fns in `metatensor-models` (#52)

authored
ff9819f
Allow SOAP-BPNN to predict on a subset of atoms (#69)
frostedoyster
authored
5f2d6ab

Commits on Feb 10, 2024

Change reader output to follow new dataset design (#70)
PicoCentauri
authored
1028095
Remove `metatensor` hotfix (#71)
frostedoyster
authored
9b7d49e

Commits on Feb 12, 2024

Minor proofread to getting started docs
PicoCentauri
committed
ea4e750
CI infrastructure work
PicoCentauri
committed
a956320

Commits on Feb 13, 2024

first prototype tests passing
agoscinski
committed
666899f
add docs and pip install deps
agoscinski
committed
52486e7

Commits on Feb 14, 2024

fix gradients partially
agoscinski
committed
f7123d8
handle different atomic species
DavideTisi
committed
4ed0c74
fix torchscript errors
DavideTisi
committed
7786133
Fix metatensor-torch-related docs (#74)
PicoCentauri
authored
0f0260e

Commits on Feb 15, 2024

the model trains with energies
DavideTisi
committed
3e80448
works with the command line
DavideTisi
committed
dd5e105
still not work
DavideTisi
committed
8db3b61

Commits on Feb 16, 2024

Commits on Feb 19, 2024

Document hyperparameters for SOAP-BPNN (#72)

frostedoyster
and
PicoCentauri
authored
2304cb7
Fix Contributor images
PicoCentauri
committed
a5a5b7d
Add ".extxyz" file suffix
PicoCentauri
committed
aedbf98
Enforce new export rules (#83)
frostedoyster
authored
d1c1522

Commits on Feb 20, 2024

Implement the Alchemical Model from `torch_alchemical` repo (#66)

abmazitov
and
frostedoyster
authored
06ca9b4
Add debug flag and architecture error message (#80)
PicoCentauri
authored
f2c6899

Commits on Feb 21, 2024

Better comment for sphericart-torch import

frostedoyster
authored and
PicoCentauri
committed
98c799b
Test the `metatensor.models.utils.export` module

frostedoyster
authored and
PicoCentauri
committed
c9e2759
Added tests for model loading (#98)
PicoCentauri
authored
d06e67e
Rename symmetry labels (#81)
frostedoyster
authored
91a7dff
Update the alchemical model docs (#94)

abmazitov
and
Luthaf
authored
8c22054

Commits on Feb 22, 2024

Update docs (#104)
PicoCentauri
authored
174462c

Commits on Feb 23, 2024

Remove SOAP-BPNN bug

frostedoyster
authored and
PicoCentauri
committed
5dec6c1
Add test

frostedoyster
authored and
PicoCentauri
committed
51f06d3
Actually do the test

frostedoyster
authored and
PicoCentauri
committed
17a9303
Lint

frostedoyster
authored and
PicoCentauri
committed
b616b80

Commits on Feb 26, 2024

Commits on Feb 27, 2024

implement training on per atom energies by modifying loss modules (#105)

SanggyuChong
and
frostedoyster
authored
5726ef1

Commits on Feb 28, 2024

Setup codecov (#111)

PicoCentauri
and
Luthaf
authored
110bd0e
Cleanup of omegconf and export functions (#112)
PicoCentauri
authored
f416bb9

Commits on Feb 29, 2024

Fix codecov badge (#116)
PicoCentauri
authored
e18467b
Move `normalize.py` to utils of `alchemical_model` (#115)
frostedoyster
authored
ffaf7b8
Add an interface to PET (#68)

authored
67a4186

Commits on Mar 1, 2024

Improve docs for BPNN and alchemical model (#130)
PicoCentauri
authored
9895e98

Commits on Mar 5, 2024

Commits on Mar 7, 2024

Pin rascaline version, limit metatensor version (#133)

PicoCentauri
and
SanggyuChong
authored
5956653

Commits on Mar 11, 2024

Update the Alchemical Model for consistency with the torch-alchemical repo (#118)
abmazitov
authored
f4981b3

Commits on Mar 12, 2024

Commits on Mar 13, 2024

Rename outputs to targets
frostedoyster
committed
9773605
Rename "requested capabilities"
frostedoyster
committed
e9488b0
Explicit per_atom
frostedoyster
committed
9caf4ac
Fix substitution mistake
frostedoyster
committed
7ffdc8f
Fixing composition weights defined by the user (#131)
SanggyuChong
authored
8d75439
Merge branch 'main' into dataset-info
frostedoyster
authored
d5186c3
Changes from review
frostedoyster
committed
e444d6c
Rework model io utilities (#144)
PicoCentauri
authored
40ec363

Commits on Mar 14, 2024

Fix tests (#146)
frostedoyster
authored
555c01c
Merge branch 'main' into dataset-info
frostedoyster
authored
9b2cfbd
Fix tests
frostedoyster
committed
626f386

Commits on Mar 15, 2024

Merge branch 'main' into sparse-gap
frostedoyster
authored
9b77b01
Merge branch 'dataset-info' into sparse-gap
frostedoyster
committed
177f9b3
Finish renaming to `experimental.gap`
frostedoyster
committed
073371e
Make GAP only support torch.float64
frostedoyster
committed
162e417
Add rascaline dependency
frostedoyster
committed
c7f1617
Do not disable rascaline's logger (#147)
frostedoyster
authored
cc9fbc9
Record authors and maintainers for each architecture (#145)
Luthaf
authored
8d59de4

Commits on Mar 16, 2024

Fix regularizer bug (?)
frostedoyster
committed
60ca7d4
Merge branch 'main' into dataset-info
frostedoyster
authored
9767085
Set `per_atom` to `False` in capabilities
frostedoyster
committed
1ddce12

Commits on Mar 18, 2024

Commits on Mar 19, 2024

Commits on Mar 21, 2024

Implementing schedulers for SOAP-BPNN and alchemical models (#134)
SanggyuChong
authored
77ba5d8

Commits on Mar 26, 2024

Remove global dtypes (#155)
PicoCentauri
authored
dd4f0df

Commits on Mar 27, 2024

Improve `eval` (#150)
frostedoyster
authored
275f9b2

Commits on Apr 1, 2024

Add domain decomposition support for PET (#137)
abmazitov
authored
e398aa5

Commits on Apr 2, 2024

Fix SOAP feature count and sparsity in SOAP-BPNN (#164)
frostedoyster
authored
b8dfce7

Commits on Apr 3, 2024

Fix bug in RMSE accumulator (#162)
frostedoyster
authored
458fb30

Commits on Apr 4, 2024

Fix validation RMSE calculation in SOAP-BPNN (#160)
frostedoyster
authored
7badc61
Fix bug in scripted SOAP-BPNN without LayerNorm (#166)

frostedoyster
and
SanggyuChong
authored
0fb52b9
Fix combined dataloader bug (#163)
frostedoyster
authored
6f02805

Commits on Apr 11, 2024

Use default device obtained from architecture (#165)
PicoCentauri
authored
5548548

Commits on Apr 20, 2024

github username update (#170)
spozdn
authored
ff4d644

Commits on Apr 26, 2024

Temporarily suppress sphinx cache warning (#173)
PicoCentauri
authored
accbf35
Add GPU support for PET-LAMMPS interface (#171)
abmazitov
authored
fecec98
Implement faster ASE neighbor list (#172)

frostedoyster
and
Luthaf
authored
2fa2ac4

Commits on May 13, 2024

Correct typo in readme (#176)
computerscienceiscool
authored
77b6a08

Commits on May 14, 2024

Loss improvements (#177)
frostedoyster
authored
3fe08a1
Remove hydra dependency (#175)
PicoCentauri
authored
48361b1

Commits on May 15, 2024

Commits on May 16, 2024

Commits on May 17, 2024

Fix link to workflow page for test banner (#188)
PicoCentauri
authored
3fa7fbf
Merge branch 'main' into sparse-gap
frostedoyster
committed
0191479
Update to new infrastructure
frostedoyster
committed
42a9ab8
Fix TODOs
frostedoyster
committed
aeb5f8e
Fix torchscript test
frostedoyster
committed
159d723
Add authors and maintainers
frostedoyster
committed
5e365bd
Turn mean into composition weights
frostedoyster
committed
77789c1

Commits on May 18, 2024

Fix tests
frostedoyster
committed
5b71eee
Fix docs?
frostedoyster
committed
2b05af4

Commits on May 22, 2024

Commits on May 23, 2024

Evaluate per-atom properties (#191)

frostedoyster
and
PicoCentauri
authored
f29c2e7
Merge branch 'main' into sparse-gap
DavideTisi
authored
52a3ef6
removed a old TODO
DavideTisi
committed
0c45b08
Replace `trainer` function by `Trainer` class
PicoCentauri
committed
bd35efe
Allow users to set weights for different loss terms from outside (#193)
frostedoyster
authored
25113b9
Merge branch 'main' into sparse-gap
frostedoyster
committed
abc3ddb
Fix tests badge (#198)
frostedoyster
authored
1d12efb

Commits on May 24, 2024

Replace `merge_capabilities` with `merge_dataset_info`
frostedoyster
committed
cae7888
Allow extraction of targets from the current training datasets
frostedoyster
committed
8f5f8d4
Incorporate changes from main
frostedoyster
committed
75701df
Fix `--continue`
frostedoyster
committed
b17d7a8
Fix SOAP-BPNN tests
frostedoyster
committed
8948346
removed file wise definitions in tests
PicoCentauri
committed
d498409
start porting AlchemicalModel
PicoCentauri
committed
a76e4e4
Rework PET neighborlist converter (#200)

abmazitov
and
frostedoyster
authored
2bc152f
start porting AlchemicalModel
PicoCentauri
committed
9ca28fc
fix leftover merge conflict
PicoCentauri
committed
7ecfd45
Merge branch 'main' into trainer-class
frostedoyster
authored
34bb3f9
fix resolvers
PicoCentauri
committed
88fa680
lint
PicoCentauri
committed
26c250e
Change `load_checkpoint` to normal method, finished PET
frostedoyster
committed
35fe7fa
Merge branch 'trainer-class' of https://github.com/lab-cosmo/metatensor-models into trainer-class
frostedoyster
committed
7279bce
tests oh my tests
PicoCentauri
committed
d1fbb23
Fix PET tests
frostedoyster
committed
77795e6
Merge branch 'trainer-class' of https://github.com/lab-cosmo/metatensor-models into trainer-class
frostedoyster
committed
1b869e8
continue with tests
PicoCentauri
committed
47ae5e4
tests alchemical
PicoCentauri
committed
d1e5cc7
Fix PET test
frostedoyster
committed
84ca10b
Fix pet tests?!?!!?
frostedoyster
committed
74bb8e5
restore class method
PicoCentauri
committed
d4f16ab
Attempted alchemical fix
frostedoyster
committed
339a691
Merge branch 'trainer-class' of https://github.com/lab-cosmo/metatensor-models into trainer-class
frostedoyster
committed
fbeb765
Allow restarting alchemical models
frostedoyster
committed
12747cc
Merge branch 'main' into sparse-gap
frostedoyster
authored
f0e16b5
Add a `register_autograd_neighbors` and remove corresponding warning
frostedoyster
committed
0fd0b9b

Commits on May 25, 2024

Upgrade PET, make alchemical model work
frostedoyster
committed
a34ae7e
Collect extensions in export
frostedoyster
committed
3a721a3
Merge branch 'extensions' into sparse-gap
frostedoyster
committed
a74022c
Remove explicit extension imports
frostedoyster
committed
86660aa
Merge branch 'extensions' into sparse-gap
frostedoyster
authored
1244435
Use the same rascaline-torch for SOAP-BPNN and GAP
frostedoyster
committed
c54a530
Adapt GAP to new interface
frostedoyster
committed
ab5a70f
Make `num_sparse_points` a `krr` parameter
frostedoyster
committed
af3c915
Make GAP run on GPU
frostedoyster
committed
54caca9
Fix problem with `register_autograd_neighbors` at evaluation
frostedoyster
committed
db4d12e
Merge branch 'trainer-class' into extensions
frostedoyster
authored
5606c22
Merge branch 'extensions' into sparse-gap
frostedoyster
authored
1cd1576

Commits on May 26, 2024

Fix alchemical tests
frostedoyster
committed
5f580b7
Merge branch 'trainer-class' into extensions
frostedoyster
authored
b578d79
Merge branch 'extensions' into sparse-gap
frostedoyster
authored
9e918cd
Attempt to fix `register_autograd_neighbors`
frostedoyster
committed
36fcaa7
Fix register_autograd_neighbors?
frostedoyster
committed
386f22c
Merge branch 'trainer-class' into extensions
frostedoyster
authored
5a640c0
Merge branch 'extensions' into sparse-gap
frostedoyster
authored
2cbc19e

Commits on May 27, 2024

Try (and fail) to fix GAP tests
frostedoyster
committed
2837878
minor comment update
PicoCentauri
committed
6ba003a
proofread of docs page
PicoCentauri
committed
f430dd1
Added SelfContributionsWrapper while exporting PET (#204)

abmazitov
and
frostedoyster
authored
f073e1b
Merge branch 'trainer-class' into extensions
frostedoyster
authored
b2e15da
Merge branch 'extensions' into sparse-gap
frostedoyster
authored
317a9b7

Commits on May 29, 2024

fixed gap example and added a better error message
DavideTisi
committed
03732ce
added test to see the error
DavideTisi
committed
61ac413