How to efficiently assign forbidden edges by nodes?

I currently use the following code to add forbidden edges, but it's taking so long when the adjacency matrix (in the example below named partial_order) is large for adding background knowledge, is there any way to make this more efficient in the causal-learn package?

import numpy as np
nodenames = final_df.columns
bk= BackgroundKnowledge()
position_matrix = np.argwhere(partial_order == 1)
for coordinate in position_matrix:
   x, y = coordinate 
   node1 = nodenames[x]
   node2 = nodenames[y]
   bk.add_forbidden_by_node(GraphNode(node2), GraphNode(node1))

Start by examining BackgroundKnowledge.add_forbidden_by_node and the loop over np.argwhere(partial_order == 1). Compare the current approach with the package's existing background-knowledge APIs, then verify that large matrices produce the same forbidden edges with substantially better runtime.