gigl.src.mocking.lib.pyg_to_training_samples#

Attributes#

`DEFAULT_NUM_HOPS_FOR_DATASETS`
`DEFAULT_NUM_NEGATIVE_SAMPLES_PER_POS_EDGE`
`DEFAULT_NUM_NODES_PER_HOP`
`DEFAULT_PYG_NODE_ANCHOR_SPLIT_TRANSFORM`

Functions#

`build_k_hop_subgraphs_from_pyg_heterodata`(hetero_data, ...)	Given inputs, return a map of each root node of type root_node_type and index in `root_node_idxs'
`build_node_anchor_link_prediction_samples_from_pyg_heterodata`(...)
`build_pyg_heterodata_from_mocked_dataset_info`(...)	Given a MockedDatasetInfo object, build a HeteroData object to use PyG convenience functions.
`build_supervised_node_classification_samples_from_pyg_heterodata`(...)

Module Contents#

gigl.src.mocking.lib.pyg_to_training_samples.build_k_hop_subgraphs_from_pyg_heterodata(hetero_data, graph_metadata_pb_wrapper, root_node_type, root_node_idxs=None, num_hops=DEFAULT_NUM_HOPS_FOR_DATASETS, num_neighbors=DEFAULT_NUM_NODES_PER_HOP)[source]#

Given inputs, return a map of each root node of type root_node_type and index in root_node_idxs’ to GraphPbWrappers which describe the `num_hops surrounding subgraph.

Parameters:

hetero_data (torch_geometric.data.HeteroData)
graph_metadata_pb_wrapper (gigl.src.common.types.pb_wrappers.graph_metadata.GraphMetadataPbWrapper)
root_node_type (gigl.src.common.types.graph_data.NodeType)
root_node_idxs (Optional[torch.Tensor])
num_hops (int)
num_neighbors (int)

Return type:

dict[gigl.src.common.types.graph_data.NodeId, gigl.src.common.types.pb_wrappers.graph_data_types.GraphPbWrapper]

gigl.src.mocking.lib.pyg_to_training_samples.build_node_anchor_link_prediction_samples_from_pyg_heterodata(hetero_data, sample_edge_type, graph_metadata_pb_wrapper, mocked_dataset_info)[source]#

Parameters:

hetero_data (torch_geometric.data.HeteroData)
sample_edge_type (gigl.src.common.types.graph_data.EdgeType)
graph_metadata_pb_wrapper (gigl.src.common.types.pb_wrappers.graph_metadata.GraphMetadataPbWrapper)
mocked_dataset_info (gigl.src.mocking.lib.mocked_dataset_resources.MockedDatasetInfo)

Return type:

Tuple[list[snapchat.research.gbml.training_samples_schema_pb2.NodeAnchorBasedLinkPredictionSample], list[snapchat.research.gbml.training_samples_schema_pb2.RootedNodeNeighborhood], list[snapchat.research.gbml.training_samples_schema_pb2.RootedNodeNeighborhood]]

gigl.src.mocking.lib.pyg_to_training_samples.build_pyg_heterodata_from_mocked_dataset_info(mocked_dataset_info)[source]#

Given a MockedDatasetInfo object, build a HeteroData object to use PyG convenience functions.

Parameters:: mocked_dataset_info (gigl.src.mocking.lib.mocked_dataset_resources.MockedDatasetInfo)
Return type:: torch_geometric.data.HeteroData

gigl.src.mocking.lib.pyg_to_training_samples.build_supervised_node_classification_samples_from_pyg_heterodata(hetero_data, root_node_type, graph_metadata_pb_wrapper)[source]#

Parameters:

hetero_data (torch_geometric.data.HeteroData)
root_node_type (gigl.src.common.types.graph_data.NodeType)
graph_metadata_pb_wrapper (gigl.src.common.types.pb_wrappers.graph_metadata.GraphMetadataPbWrapper)

Return type:

list[snapchat.research.gbml.training_samples_schema_pb2.SupervisedNodeClassificationSample]

gigl.src.mocking.lib.pyg_to_training_samples.DEFAULT_NUM_HOPS_FOR_DATASETS = 1[source]#

gigl.src.mocking.lib.pyg_to_training_samples.DEFAULT_NUM_NEGATIVE_SAMPLES_PER_POS_EDGE = 1[source]#

gigl.src.mocking.lib.pyg_to_training_samples.DEFAULT_NUM_NODES_PER_HOP = 5[source]#

gigl.src.mocking.lib.pyg_to_training_samples.DEFAULT_PYG_NODE_ANCHOR_SPLIT_TRANSFORM[source]#