[WIP] Q-HDNNP: Simplified "charge-aware" HDNNP

[singraber]

No description provided.

[singraber]

Charge training seems to work (see examples/nnp-train/H2O_RPBE-D3):

List of extra keywords required:

nnp_type Q-HDNNP
global_hidden_layers_electrostatic      2
global_nodes_electrostatic              25 25
global_activation_electrostatic         t t l
charge_fraction                         1.0
task_batch_size_charge                  1
write_traincharges                      1

Things to do:

• Check energy prediction
• Force calculation
• Energy training: here (selection) and here (derivative computation)
• Force training: here (selection) and here (derivative computation)

[codecov-io]

Codecov Report

Merging #80 (cca139d) into master (c96ea3c) will decrease coverage by 2.53%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master      #80      +/-   ##
==========================================
- Coverage   72.52%   69.99%   -2.54%     
==========================================
  Files         126      131       +5     
  Lines       13311    14280     +969     
==========================================
+ Hits         9654     9995     +341     
- Misses       3657     4285     +628     

Flag	Coverage Δ
cpp	73.11% <ø> (-2.24%)	⬇️
python	69.99% <ø> (-2.54%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
src/application/nnp-comp2.cpp	0.00% <ø> (ø)
src/application/nnp-dataset.cpp	96.00% <ø> (+0.07%)	⬆️
src/application/nnp-predict.cpp	95.88% <ø> (+0.36%)	⬆️
src/application/nnp-scaling.cpp	96.34% <ø> (ø)
src/application/nnp-sfclust.cpp	92.18% <ø> (ø)
src/application/nnp-train.cpp	94.11% <ø> (-5.89%)	⬇️
...nterface/LAMMPS/src/USER-NNP/pair_nnp_external.cpp	0.00% <ø> (ø)
.../interface/LAMMPS/src/USER-NNP/pair_nnp_external.h	100.00% <ø> (ø)
src/libnnp/Atom.cpp	39.98% <ø> (+0.21%)	⬆️
src/libnnp/Atom.h	80.00% <ø> (+13.33%)	⬆️
... and 100 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update de4da77...cca139d. Read the comment docs.