Classify examples by different tasks, e.g., Link Perdiction, Node classification, Graph regression. Some examples do have the tag "Link Prediction", but actually, the code does not have any link prediction part.
Spilt examples as official codes and third-party codes. Among them, official examples must strictly obey certain coding rules.
Define some fixed training pipelines, and for each pipeline, DGL team should provide one official example. If you did these, researchers can just focus on their model implementation, not the other not important things.
🚀 Feature