pairs module

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

Nodes for finding and manipulating pairs and distances.

class DynamicPeriodicPairs(name, parents, dist_hard_max, **kwargs)

Bases: _DispatchNeighbors

Node for finding pairs in arbitrary periodic boundary conditions. Note: This will often be slower than PeriodicPairIndexer, but more general. If the speed is a concern, consider precomputing pairs with experiment.assembly.precompute_pairs

Note

This node has parent expansion, following these procedures.

1. If matching (PositionsNode, SpeciesNode, CellNode), then apply expand0

2. If matching (PositionsNode, Encoder, PaddingIndexer, CellNode), then apply expand1

3. Asserts that the number of parents is 8

4. Gets main_output of nodes: casts MultiNodes to their main output

5. Transforms the parents to have index states (<IdxType.SysAtom>, None, None, None, None, None, None, None)

auto_module_class

alias of TorchNeighbors

parent_expander: ParentExpander = <hippynn.graphs.nodes.base.definition_helpers.ParentExpander object>

class ExternalNeighborIndexer(*args, module='auto', module_kwargs=None, **kwargs)

Bases: AutoKw, ExpandParents, PairIndexer, MultiNode

auto_module_class

alias of ExternalNeighbors

auto_module_kwargs: Tuple[str] | Dict[str, str] | None = ('hard_dist_cutoff',)

input_names = ('coordinates', 'real_atoms', 'shifts', 'cell', 'ext_pair_first', 'ext_pair_second')

parent_expander: ParentExpander = <hippynn.graphs.nodes.base.definition_helpers.ParentExpander object>

class KDTreePairs(name, parents, dist_hard_max, **kwargs)

Bases: _DispatchNeighbors

Node for finding pairs under periodic boundary conditions using Scipy’s KD Tree algorithm. Cell must be orthorhombic.

Note

This node has parent expansion, following these procedures.

1. If matching (PositionsNode, SpeciesNode, CellNode), then apply expand0

2. If matching (PositionsNode, Encoder, PaddingIndexer, CellNode), then apply expand1

3. Asserts that the number of parents is 8

4. Gets main_output of nodes: casts MultiNodes to their main output

5. Transforms the parents to have index states (<IdxType.SysAtom>, None, None, None, None, None, None, None)

auto_module_class

alias of KDTreeNeighbors

parent_expander: ParentExpander = <hippynn.graphs.nodes.base.definition_helpers.ParentExpander object>

class KDTreePairsMemory(name, parents, dist_hard_max, skin, module='auto', module_kwargs=None, **kwargs)

Bases: _DispatchNeighbors, Memory

Implementation of KDTreePairs with an added memory component.

Stores current pair indices in memory and reuses them to compute the pair distances if no particle has moved more than skin/2 since last pair calculation. Otherwise uses the _pair_indexer_class to recompute the pairs.

Increasing the value of ‘skin’ will increase the number of pair distances computed at each step, but decrease the number of times new pairs must be computed. Skin should be set to zero while training for fastest results.

Note

This node has parent expansion, following these procedures.

1. If matching (PositionsNode, SpeciesNode, CellNode), then apply expand0

2. If matching (PositionsNode, Encoder, PaddingIndexer, CellNode), then apply expand1

3. Asserts that the number of parents is 8

4. Gets main_output of nodes: casts MultiNodes to their main output

5. Transforms the parents to have index states (<IdxType.SysAtom>, None, None, None, None, None, None, None)

auto_module_class

alias of KDTreePairsMemory

auto_module_kwargs: Tuple[str] | Dict[str, str] | None = ('dist_hard_max',)

parent_expander: ParentExpander = <hippynn.graphs.nodes.base.definition_helpers.ParentExpander object>

class Memory

Bases: object

reset_reuse_percentage()

property reuse_percentage

property skin

class MinDistNode(name, parents, module='auto', **kwargs)

Bases: ExpandParents, AutoNoKw, MultiNode

auto_module_class

alias of MinDistModule

expand0(pair_finder, **kwargs)

Used for creation from parents with signature (PairIndexer)

expand1(neigh_list, **kwargs)

Used for creation from parents with signature (PaddedNeighborNode)

expand2(neigh_list, pad_idxer, **kwargs)

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

input_names = ('rij_list', 'j_list', 'system_index', 'atom_index', 'inv_real_atoms', 'n_atoms_max', 'n_systems')

output_index_states = (<IdxType.Systems>, <IdxType.Systems>, <IdxType.Atoms>, <IdxType.Atoms>)

output_names = ('min_dist_mol', 'mol_locs', 'min_dist_atom', 'atom_pairlocs')

parent_expander: ParentExpander = <hippynn.graphs.nodes.base.definition_helpers.ParentExpander object>

class NumpyDynamicPairs(name, parents, dist_hard_max, **kwargs)

Bases: _DispatchNeighbors

auto_module_class

alias of NPNeighbors

parent_expander: ParentExpander = <hippynn.graphs.nodes.base.definition_helpers.ParentExpander object>

class OpenPairIndexer(name, parents, dist_hard_max, **kwargs)

Bases: AutoKw, ExpandParents, PairIndexer, MultiNode

auto_module_class

alias of OpenPairIndexer

expand0(pos, encode, indexer, **kwargs)

Used for creation from parents with signature (PositionsNode, Encoder, PaddingIndexer)

auto_module_kwargs: Tuple[str] | Dict[str, str] | None = {'hard_dist_cutoff': 'dist_hard_max'}

input_names = ('coordinates', 'nonblank', 'real_atoms', 'inv_real_atoms')

parent_expander: ParentExpander = <hippynn.graphs.nodes.base.definition_helpers.ParentExpander object>

class PaddedNeighborNode(name, parents, module='auto', **kwargs)

Bases: ExpandParents, AutoNoKw, MultiNode

auto_module_class

alias of PaddedNeighModule

expand0(pair_finder, **kwargs)

Used for creation from parents with signature (PairIndexer)

input_names = ('pair_first', 'pair_second', 'pair_coord')

output_index_states = (<IdxType.Atoms>, <IdxType.Atoms>)

output_names = ('j_list', 'rij_list')

parent_expander: ParentExpander = <hippynn.graphs.nodes.base.definition_helpers.ParentExpander object>

class PairCacher(*args, module='auto', module_kwargs=None, **kwargs)

Bases: AutoKw, ExpandParents, PairCache, SingleNode

auto_module_class

alias of PairCacher

expand0(pair_indexer, *args, purpose, **kwargs)

Used for creation from parents with signature (PairIndexer)

expand1(pair_indexer, atomidx, *args, purpose, **kwargs)

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

index_state: IdxType | None = 'Unlabeled'

input_names = ('pair_first', 'pair_second', 'cell_offsets', 'offset_index', 'real_atoms', 'system_index', 'n_atoms_max', 'n_systems')

parent_expander: ParentExpander = <hippynn.graphs.nodes.base.definition_helpers.ParentExpander object>

class PairDeIndexer(name, parents, module='auto', **kwargs)

Bases: ExpandParents, AutoNoKw, SingleNode

For converting from IdxType.Pairs to IdxType.SysAtomAtom (Padded form)

Note

This node has parent expansion, following these procedures.

1. If matching (Node), then apply expand0

2. If matching (Node, PaddingIndexer, PairIndexer), then apply expand1

auto_module_class

alias of PairDeIndexer

expand0(pair_features)

Used for creation from parents with signature (Node)

expand1(pair_features, pad_idx, pair_idx)

Used for creation from parents with signature (Node, PaddingIndexer, PairIndexer)

index_state: IdxType | None = 'SysAtomAtom'

input_names = ('pair_features', 'system_index', 'atom_index', 'n_systems', 'n_atoms_maxpair_first', 'pair_second')

parent_expander: ParentExpander = <hippynn.graphs.nodes.base.definition_helpers.ParentExpander object>

class PairFilter(name, parents, dist_hard_max, module='auto', **kwargs)

Bases: AutoKw, PairIndexer, ExpandParents, MultiNode

auto_module_class

alias of FilterDistance

expand0(pair_indexer, purpose)

Used for creation from parents with signature (PairIndexer)

parent_expander: ParentExpander = <hippynn.graphs.nodes.base.definition_helpers.ParentExpander object>

class PairReIndexer(name, parents, module='auto', **kwargs)

Bases: ExpandParents, AutoNoKw, SingleNode

For re-using index information to convert from IdxType.SysAtomAtom -> IdxType.Pairs

Note

This node has parent expansion, following these procedures.

1. If matching (Node), then apply expand0

2. If matching (Node, PaddingIndexer, PairIndexer), then apply expand1

3. Asserts that the number of parents is 5

4. Gets main_output of nodes: casts MultiNodes to their main output

auto_module_class

alias of PairReIndexer

expand0(pair_features)

Used for creation from parents with signature (Node)

expand1(pair_features, pad_idx, pair_idx)

Used for creation from parents with signature (Node, PaddingIndexer, PairIndexer)

index_state: IdxType | None = 'Pairs'

input_names = ('pair_features', 'system_index', 'atom_index', 'pair_first', 'pair_second')

parent_expander: ParentExpander = <hippynn.graphs.nodes.base.definition_helpers.ParentExpander object>

class PairUncacher(name, parents, dist_hard_max, **kwargs)

Bases: ExpandParents, AutoNoKw, PairIndexer, MultiNode

auto_module_class

alias of PairUncacher

expand0(sparse, *args, purpose, **kwargs)

Used for creation from parents with signature (PairCache)

expand1(sp, r, c, atomidx, *args, purpose, **kwargs)

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

Used for creation from parents with signature (PairCache, PositionsNode, CellNode, AtomIndexer)

auto_module_kwargs = ('dist_hard_max',)

index_state = 'Unlabeled'

input_names = ('sparsepairs', 'coordinates', 'cells', 'real_atoms', 'inv_real_atoms', 'n_atoms_max', 'n_systems')

output_index_states = (<IdxType.Pairs>, <IdxType.Pairs>, <IdxType.Pairs>, <IdxType.Pairs>, <IdxType.Pairs>, <IdxType.Pairs>)

output_names = ('pair_dist', 'pair_first', 'pair_second', 'pair_coord', 'cell_offsets', 'offset_index')

parent_expander: ParentExpander = <hippynn.graphs.nodes.base.definition_helpers.ParentExpander object>

class PeriodicPairIndexer(name, parents, dist_hard_max, **kwargs)

Bases: AutoKw, ExpandParents, PeriodicPairOutputs, PairIndexer, MultiNode

auto_module_class

alias of PeriodicPairIndexer

expand0(pos, spec, cell, *, purpose, **kwargs)

Used for creation from parents with signature (PositionsNode, SpeciesNode, CellNode)

expand1(pos, encode, indexer, cell, **kwargs)

Used for creation from parents with signature (PositionsNode, Encoder, PaddingIndexer, CellNode)

auto_module_kwargs: Tuple[str] | Dict[str, str] | None = {'hard_dist_cutoff': 'dist_hard_max'}

input_names = ('coordinates', 'nonblank', 'real_atoms', 'inv_real_atoms', 'cell')

parent_expander: ParentExpander = <hippynn.graphs.nodes.base.definition_helpers.ParentExpander object>

class PeriodicPairIndexerMemory(name, parents, dist_hard_max, skin, module='auto', module_kwargs=None, **kwargs)

Bases: PeriodicPairIndexer, Memory

Implementation of PeriodicPairIndexer with additional memory component.

Stores current pair indices in memory and reuses them to compute the pair distances if no particle has moved more than skin/2 since last pair calculation. Otherwise uses the _pair_indexer_class to recompute the pairs.

Increasing the value of ‘skin’ will increase the number of pair distances computed at each step, but decrease the number of times new pairs must be computed. Skin should be set to zero while training for fastest results.

Note

This node has parent expansion, following these procedures.

1. If matching (PositionsNode, SpeciesNode, CellNode), then apply expand0

2. If matching (PositionsNode, Encoder, PaddingIndexer, CellNode), then apply expand1

3. Asserts that the number of parents is 5

4. Gets main_output of nodes: casts MultiNodes to their main output

auto_module_class

alias of PeriodicPairIndexerMemory

children: Tuple[Node]

name: str

origin_node: Node | None

parent_expander: ParentExpander = <hippynn.graphs.nodes.base.definition_helpers.ParentExpander object>

parents: Tuple[Node]

torch_module: torch.nn.Module

class PeriodicPairOutputs

Bases: object

output_index_states = (<IdxType.Pairs>, <IdxType.Pairs>, <IdxType.Pairs>, <IdxType.Pairs>, <IdxType.Pairs>, <IdxType.Pairs>)

output_names = ('pair_dist', 'pair_first', 'pair_second', 'pair_coord', 'cell_offsets', 'offset_index')

class RDFBins(name, parents, bins=None, **kwargs)

Bases: AutoKw, ExpandParents, SingleNode

auto_module_class

alias of RDFBins

expand0(pos, spec, cell, *, purpose, dist_hard_max=None, **kwargs)

Build a default Periodic Pair indexer.

Used for creation from parents with signature (PositionsNode, SpeciesNode, CellNode)

expand1(pos, spec, *, purpose, dist_hard_max=None, **kwargs)

Builds an open pair indexer.

Used for creation from parents with signature (PositionsNode, SpeciesNode)

expand2(pairs, *, purpose, **kwargs)

Get the encoding and padding associated with a pair indexer.

Used for creation from parents with signature (PairIndexer)

expand3(pairs, one_hot, pad, *, purpose, **kwargs)

Expanded the needed children of pairs, encoder, and padding indexer.

Used for creation from parents with signature (PairIndexer, OneHotEncoder, PaddingIndexer)

auto_module_kwargs: Tuple[str] | Dict[str, str] | None = ('bins',)

index_state: IdxType | None = 'Scalar'

input_names = ('pair_dists', 'pair_first', 'pair_second', 'one_hot', 'n_systems')

parent_expander: ParentExpander = <hippynn.graphs.nodes.base.definition_helpers.ParentExpander object>

parent_expansion_kwargs: Tuple[str] | Dict[str, str] | None = ('dist_hard_max',)