SAGEEncoder

class graphstorm.model.SAGEEncoder(h_dim, out_dim, num_hidden_layers=1, dropout=0, aggregator_type='mean', activation=<function relu>, num_ffn_layers_in_gnn=0, norm=None)

Bases: GraphConvEncoder

GraphSage Conv Encoder.

The SAGEEncoder employs several SAGEConv Layers as its encoding mechanism. The SAGEEncoder should be designated as the model’s encoder within Graphstorm.

Parameters

h_dim: int

Hidden dimension size.

out_dim: int

Output dimension size.

num_hidden_layers: int

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

dropout: float

Dropout rate. Default: 0.

aggregator_type: str

Message aggregation type. Options: mean, gcn, pool, lstm.

activation: torch.nn.functional

Activation function. Default: relu.

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 SAGEEncoder
from graphstorm import get_node_feat_size
from graphstorm.model import SAGEEncoder
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 = SAGEEncoder(4, 4,
                          num_hidden_layers=1,
                          dropout=0,
                          aggregator_type="mean",
                          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)

GraphSage 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

Node features for the default node type in the format of {dgl.DEFAULT_NTYPE: tensor}. The definition of dgl.DEFAULT_NTYPE can be found at DGL official Github site.

Returns

h: dict of Tensor

New node embeddings for the default node type in the format of {dgl.DEFAULT_NTYPE: tensor}. The definition of dgl.DEFAULT_NTYPE can be found at DGL official Github site.