pylibcugraphops.operators.agg_hg_basis_n2n_post_fwd#

pylibcugraphops.operators.agg_hg_basis_n2n_post_fwd = <nanobind.nb_func object>#

Computes the forward pass for node-to-node full-graph RGCN-like basis regularized: aggregation, with features being transformed after (post) this aggregation.

agg_hg_basis_n2n_post_fwd(
    output_embedding: device array, input_embedding: device array,
    weights_combination: Optional[device array], graph: pylibcugraphops.csc_hg_int[32|64],
    concat_own: bool = False, norm_by_out_degree: bool = False, stream_id: int = 0
)

Parameters:

output_embeddingdevice array type: Device array containing the output embeddings. Shape for concat_own=False: (graph.n_dst_nodes, dim_out); Shape for concat_own=True: (graph.n_dst_nodes, dim_out + dim_in), with dim_out = dim_in * n_bases if weights_combination is set, dim_out = dim_in * n_edge_types otherwise.
input_embeddingdevice array type: Device array containing the input embeddings. Shape: (graph.n_dst_nodes, dim_in).
weights_combinationdevice array type | None: Device array containing the combination weights. If None, no weights are used and features of different edge types are aggregated into separate output dimensions. Shape: (n_edge_types, n_bases) if set.
graphopaque graph type: The graph used for the operation.
concat_ownbool, default=False: Concatenate output node embeddings in the aggregation.
norm_by_out_degreebool, default=False: If set, output embeddings are normed by the degree of the output node.
stream_idint, default=0: CUDA stream pointer as a python int.