RelationalGCNEncoder

class graphstorm.model.RelationalGCNEncoder(g, h_dim, out_dim, num_bases=-1, num_hidden_layers=1, dropout=0, use_self_loop=True, last_layer_act=False, num_ffn_layers_in_gnn=0, norm=None)

Bases: GraphConvEncoder, GSgnnGNNEncoderInterface

Relational graph conv encoder.

The RelationalGCNEncoder employs several RelGraphConvLayer as its encoding mechanism. The RelationalGCNEncoder should be designated as the model’s encoder within Graphstorm.

Parameters

g: DistGraph

The distributed graph.

h_dim: int

Hidden dimension.

out_dim: int

Output dimension.

num_bases: int

Number of bases. If is None, use number of relation types. Default: None.

num_hidden_layers: int

Number of hidden layers. Total GNN layers is equal to num_hidden_layers + 1. Default: 1.

dropout: float

Dropout rate. Default 0.

use_self_loop: bool

Whether to add selfloop. Default: True.

last_layer_act: callable

Activation for the last layer. Default: None.

num_ffn_layers_in_gnn: int

Number of fnn layers between GNN layers. Default: 0.

norm: str

Normalization methods. Options:batch, layer, and None. Default: None, meaning no normalization.

Examples:

# Build model and do full-graph inference on RelationalGCNEncoder
from graphstorm import get_node_feat_size
from graphstorm.model import RelationalGCNEncoder
from graphstorm.model import EntityClassifier
from graphstorm.model import GSgnnNodeModel, GSNodeEncoderInputLayer
from graphstorm.dataloading import GSgnnData
from graphstorm.model import do_full_graph_inference

np_data = GSgnnData(...)

model = GSgnnNodeModel(alpha_l2norm=0)
feat_size = get_node_feat_size(np_data.g, "feat")
encoder = GSNodeEncoderInputLayer(g, feat_size, 4,
                                  dropout=0,
                                  use_node_embeddings=True)
model.set_node_input_encoder(encoder)

gnn_encoder = RelationalGCNEncoder(g, 4, 4,
                                   num_hidden_layers=1,
                                   dropout=0,
                                   use_self_loop=True,
                                   norm="batch")
model.set_gnn_encoder(gnn_encoder)
model.set_decoder(EntityClassifier(model.gnn_encoder.out_dims, 3, False))

h = do_full_graph_inference(model, np_data)

forward(blocks, h)

RGCN encoder forward computation.

Parameters

blocks: list of DGL MFGs

Sampled subgraph in the list of DGL message flow graphs (MFGs) format. More detailed information about DGL MFG can be found in DGL Neighbor Sampling Overview.

h: dict of Tensor

Input node features for each node type in the format of {ntype: tensor}.

Returns

h: dict of Tensor

New node embeddings for each node type in the format of {ntype: tensor}.