graphstorm.model.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)

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_dimint

Hidden dimension

out_dimint

Output dimension

num_hidden_layersint

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

dropoutfloat

Dropout. Default 0.

num_ffn_layers_in_gnn: int

Number of ngnn gnn layers between GNN layers

normstr, optional

Normalization Method. Default: None

Examples:

# Build model and do full-graph inference on SAGEEncoder
from graphstorm import get_node_feat_size
from graphstorm.model.sage_encoder import SAGEEncoder
from graphstorm.model.node_decoder import EntityClassifier
from graphstorm.model import GSgnnNodeModel, GSNodeEncoderInputLayer
from graphstorm.dataloading import GSgnnNodeTrainData
from graphstorm.model import do_full_graph_inference

np_data = GSgnnNodeTrainData(...)

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=norm)
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)