You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We currently are unable to handle isolated vertices for GNN sampling, since those are dropped during graph creation. This is causing issues for several users, who are unable to train their graphs with many isolated vertices. Adding self loops is not an option since those will cause a major accuracy drop for heterogeneous graphs and may also violate the schema of the users' upstream graph storage.
The text was updated successfully, but these errors were encountered:
We currently are unable to handle isolated vertices for GNN sampling, since those are dropped during graph creation. This is causing issues for several users, who are unable to train their graphs with many isolated vertices. Adding self loops is not an option since those will cause a major accuracy drop for heterogeneous graphs and may also violate the schema of the users' upstream graph storage.
The text was updated successfully, but these errors were encountered: