pylibcugraphops.operators.agg_concat_n2n_bwd

pylibcugraphops.operators.agg_concat_n2n_bwd = <nanobind.nb_func object>

Computes the backward pass for a simple aggregation using node features in an

node-to-node reduction (n2n) on a graph while concatenating the original features of output nodes at the end (agg_concat).

agg_concat_n2n_bwd(
    grad_input: device array, grad_output: device array,
    graph: pylibcugraphops.csc_int[32|64],
    aggregation_operation: pylibcugraphops.operators.AggOp = pylibcugraphops.operators.AggOp.Sum,
    output_extrema_location: Optional[device array] = None, stream_id: int = 0
) -> None

Parameters:

grad_inputdevice array type

Device array containing the output gradient on input embeddings of forward. Shape: (graph.n_dst_nodes, dim_in).

grad_outputdevice array type

Device array containing the input gradient on output embeddings of forward. Shape: (graph.n_dst_nodes, 2 * dim_in).

graphopaque graph type

The graph used for the operation.

aggregation_operationAggOp, default=AggOp.Sum

The kind of aggregation operation.

output_extrema_locationdevice array type | None

Device array containing the location of the min/max embeddings. This is required for min/max aggregation only, and can be None otherwise. Shape: (graph.n_dst_nodes, dim_in) if set.

stream_idint, default=0

CUDA stream pointer as a python int.