Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

4G-HDNNP Prediction and LAMMPS interface #77

Draft
wants to merge 74 commits into
base: master
Choose a base branch
from
Draft

4G-HDNNP Prediction and LAMMPS interface #77

wants to merge 74 commits into from

Conversation

singraber
Copy link
Member

Implements the predictive capabilities for 4G-HDNNPs (https://arxiv.org/abs/2009.06484)

@singraber
Save NN setup in separate map to structs (WIP)
25a5cb7
@singraber
Working on per-element adjustable NN setup
d64acd2
@singraber
Changed setupNeuralNetwork() (WIP)
6f4ae5c
@singraber
Changed NN setup once again (WIP)
70cfb23
@singraber
NN setup work continued (WIP)
23bdf04
@singraber
Per-element NN setup should work, needs testing
d904f47
@singraber
More changes in Mode class (WIP)
b5fc486
@singraber
Allow to override default weights file name format
7cf84c3
@singraber
Fixed bug, nnp-predict works now again
ed06ade
@singraber
LAMMPS interface works again
42ce95e
@singraber
Added pair_style nnp/external and examples
7775054 
* Works for nnp-predict
* Not yet for RuNNer (different energy.out, nnforces.out format)
@singraber
Pair_style nnp/external now also works with RuNNer
3f29113
@singraber
Fixed Training class, ready for 4G-HDNNP code!
a2242b6
@singraber
Added hardness and sigma read-in
bbaabfb
@singraber
Merge branch '4G-HDNNP-prediction' of github.com:singraber/n2p2 into …
3974406 
…4G-HDNNP-prediction
@singraber singraber added enhancement New feature or request interface Concerns the interface to third-party software core Touches the libnnp core library labels Jan 19, 2021
@singraber singraber self-assigned this Jan 19, 2021
@singraber singraber changed the title 4G-HDNNP Prediction and LAMMPS interface [WIP] 4G-HDNNP Prediction and LAMMPS interface Jan 19, 2021
@philippmisof
Computation of dA/dr * Q has been implemented.
6d8a16e 
Result agrees with RuNNer for the AlAuMgO and the carbon chain examples.
@philippmisof
Implemented dChi/dG, agrees with RuNNer.
7bf8aaa
@singraber singraber mentioned this pull request Apr 2, 2021
Open
philippmisof and others added 3 commits April 2, 2021 21:30
@philippmisof
Added forces (total and elec.), agrees with RuNNer
b99866c 
Forces can now be predicted for periodic and non-periodic case.
@singraber
Ready to work on nnp-train
67d6a05 
Minor change: NL sorting only for 4G-HDNNPs, because it
breaks some tests (numerical accuracy).
@singraber
Excluded some broken pynnp CI tests
02f1adb
@singraber singraber changed the title [WIP] 4G-HDNNP Prediction and LAMMPS interface 4G-HDNNP Prediction and LAMMPS interface Apr 14, 2021
@singraber singraber marked this pull request as draft April 14, 2021 14:49
philippmisof and others added 17 commits April 21, 2021 15:28
@philippmisof
Quick fix for slow calculateForces
29c594f 
Added break condition to avoid loop over unnecessary neighbors. Will be
fixed more elegantely later.
@philippmisof
Merge branch '4G-HDNNP-prediction' of github.com:CompPhysVienna/n2p2 …
d4e0c65 
…into 4G-HDNNP-prediction
@emirkocer
QEq minimization for non-periodic case
2177c69
@emirkocer
Merge branch 'IF-nonperiodic' of github.com:emirkocer/n2p2 into 4G-HD…
4143ecb 
…NNP-prediction
@emirkocer
force member calls are replaced by utils
4879eaf
@emirkocer
non-periodic 4G-HDNNP (serial) is implemented
fa3bd4e
@emirkocer
periodic case implemented (serial)
8827e5b
@emirkocer
bugs fixed & cleaned up
a041104
@emirkocer
serial version is updated & bugs are fixed
2a23f3c
@emirkocer
clean-up of examples
8dd1bba
@singraber
Fixed allocation with uninitialized size
89cd893 
- Fixed invalid read in NN input.
@singraber
Fixed AlAuMgO LAMMPS example input
850e39b
@singraber
Cleanup of water example, fixed 8.0B real cutoff
5a6f194 
- see "water-test/correct-input-data" for consistent input for
nnp-predict/lammps-nnp.
@singraber
Added numeric vs. analytic check for dEQeq/dQ
ac9a358 
- Does NOT match yet, further investigation required...
@singraber
Fixed some bugs (type[i], cflength) in QEq_f
d96fab0 
- Reciprocal part still incorrect, k-vectors seem wrong
- Dev version, has unresolved segfaults!
@singraber
Found problem with different k-vectors
4ac5c16 
- LAMMPS uses scheme with different cutoff
- With current fix EQeq is now correct
- dEQeq needs to be fixed!
@singraber
Charges now match up to ~10^-9
1ab1c11 
- Maybe that is the limit for double precision?
@singraber singraber mentioned this pull request Sep 24, 2021
Closed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers
No reviews
Assignees

@singraber singraber

Labels
core Touches the libnnp core library enhancement New feature or request interface Concerns the interface to third-party software
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.

Implement and test sorted neighbor list
4 participants
@singraber @codecov-io @emirkocer @philippmisof