graphstorm

The graphstorm package contains a function for environment setup and a set of utilization functions. Users can directly use the following code to use these functions.

>>> import graphstorm as gs
>>> gs.initialize()
>>> feat_size = gs.get_node_feat_size(g, node_feat_names={"author": "feat"})
>>> relation_names = gs.get_rel_names_for_reconstruct(
>>>                     g,
>>>                     reconstructed_embed_ntype=["paper"],
>>>                     feat_size=feat_size)

gsf.initialize	Initialize distributed training and inference context.
gsf.get_node_feat_size	Get the overall feature size of each node type with feature names specified in the node_feat_names.
gsf.get_rel_names_for_reconstruct	Get the edge type list for reconstructing node features.