excited module

Full Documentation for hippynn.graphs.nodes.excited module. Click here for a summary page.

Nodes for excited state modeling.

class HuberPhaseLoss(predicted, true)

Bases: _BaseCompareLoss

class LocalEnergyNode(name, parents, first_is_interacting=False, module='auto', **kwargs)

Bases: Energies, ExpandParents, HAtomRegressor, MultiNode

Predict a localized energy, with contributions from implicitly computed atoms.

Note

This node has parent expansion, following these procedures.

1. If matching (Network), then apply expansion0

2. If matching (Network, AtomIndexer), then apply expansion1

3. Asserts that the number of parents is 5

auto_module()

expansion0(net, *, purpose, **kwargs)

Used for creation from parents with signature (Network)

expansion1(net, pdindexer, **kwargs)

Used for creation from parents with signature (Network, AtomIndexer)

class MAEPhaseLoss(predicted, true)

Bases: _BaseCompareLoss

torch_module = LambdaModule(_mae_with_phases)

class MSEPhaseLoss(predicted, true)

Bases: _BaseCompareLoss

torch_module = LambdaModule(_mse_with_phases)

class NACRMultiStateNode(name, parents, module='auto', module_kwargs=None, **kwargs)

Bases: AutoKw, SingleNode

Compute the non-adiabatic coupling vector multiplied by the energy difference between all pairs of states.

class NACRNode(name: str, parents: Tuple, module='auto', module_kwargs=None, **kwargs)

Bases: AutoKw, SingleNode

Compute the non-adiabatic coupling vector multiplied by the energy difference between two states.