debug

torch_geometric/utils/rag/backend_utils.py

@@ -122,13 +122,13 @@ class RemoteGraphBackendLoader:
     graph_store_type: Type[ConvertableGraphStore]
     feature_store_type: Type[ConvertableFeatureStore]
 
-    def load(self, pid: Optional[int] = None,
-             is_sorted=False) -> RemoteGraphBackend:
+    def load(self, pid: Optional[int] = None) -> RemoteGraphBackend:
         if self.datatype == RemoteDataType.DATA:
             data_obj = torch.load(self.path, weights_only=False)
+            # is_sorted=true since assume nodes come sorted from indexer
             graph_store = self.graph_store_type.from_data(
                 edge_id=data_obj['edge_id'], edge_index=data_obj.edge_index,
-                num_nodes=data_obj.num_nodes, is_sorted=is_sorted)
+                num_nodes=data_obj.num_nodes, is_sorted=True)
             graph_store.edge_index = graph_store.edge_index.contiguous()
             feature_store = self.feature_store_type.from_data(
                 node_id=data_obj['node_id'], x=data_obj.x,
@@ -237,7 +237,7 @@ class to use. Defaults to LocalFeatureStore.
     if n_parts == 1:
         torch.save(data, path)
         return RemoteGraphBackendLoader(path, RemoteDataType.DATA, graph_db,
-                                        feature_db, is_sorted=True)
+                                        feature_db)
     else:
         partitioner = Partitioner(data=data, num_parts=n_parts, root=path)
         partitioner.generate_partition()