Claude seems to suggest that certain attributes are lost when we go from directed -> undirected graph (node/edge/graph level, not sure). See where this might happen.
pytorch-geometric looks at shapes to determine what is node- vs edge- level, which can get confusing. So we use a prefix to distinguish between the two.
Claude seems to suggest that certain attributes are lost when we go from directed -> undirected graph (node/edge/graph level, not sure). See where this might happen.
pytorch-geometric looks at shapes to determine what is node- vs edge- level, which can get confusing. So we use a prefix to distinguish between the two.