gigl.transforms.utils#

Attributes#

EdgeType

Functions#

add_edge_attr(data, values[, attr_name, edge_type_to_idx])

Helper function to add edge attributes to a HeteroData object.

add_node_attr(data, values[, attr_name, node_type_to_idx])

Helper function to add node attributes to a HeteroData object.

Module Contents#

gigl.transforms.utils.add_edge_attr(data, values, attr_name=None, edge_type_to_idx=None)[source]#

Helper function to add edge attributes to a HeteroData object.

Parameters:

  • data (torch_geometric.data.HeteroData) – The HeteroData object to modify.

  • values (Union[torch.Tensor, Dict[EdgeType, torch.Tensor]]) –

    Either: - A tensor of values in homogeneous edge order (requires edge_type_to_idx

    or will be computed from data.edge_types), OR

    • A dictionary mapping edge types to tensors of values for each type.

  • attr_name (Optional[str]) – The name of the attribute to add. If None, concatenates to existing edge_attr attribute for each edge type (or creates it).

  • edge_type_to_idx (Optional[Dict[EdgeType, Tuple[int, int]]]) – Optional mapping from edge type to (start, end) indices. Only used when values is a tensor. If None, it will be computed from data.edge_types.

Returns:

The modified HeteroData object.

Return type:

torch_geometric.data.HeteroData

gigl.transforms.utils.add_node_attr(data, values, attr_name=None, node_type_to_idx=None)[source]#

Helper function to add node attributes to a HeteroData object.

Parameters:

  • data (torch_geometric.data.HeteroData) – The HeteroData object to modify.

  • values (Union[torch.Tensor, Dict[str, torch.Tensor]]) –

    Either: - A tensor of values in homogeneous node order (requires node_type_to_idx

    or will be computed from data.node_types), OR

    • A dictionary mapping node types to tensors of values for each type.

  • attr_name (Optional[str]) – The name of the attribute to add. If None, concatenates to existing x attribute for each node type (or creates it).

  • node_type_to_idx (Optional[Dict[str, Tuple[int, int]]]) – Optional mapping from node type to (start, end) indices. Only used when values is a tensor. If None, it will be computed from data.node_types.

Returns:

The modified HeteroData object.

Return type:

torch_geometric.data.HeteroData

gigl.transforms.utils.EdgeType[source]#