Although the GNN plot (endpoint importance-fraction-plot) relies on potentially-cached fractions inputs, the large number of queries that are made to these high-level fractions means that the overall plot creation is rather slow, taking a few minutes. I think this could be due to a large number of database connections for trivial metadata.
To address this, make sure that the final result is also cached.
Although the GNN plot (endpoint
importance-fraction-plot) relies on potentially-cached fractions inputs, the large number of queries that are made to these high-level fractions means that the overall plot creation is rather slow, taking a few minutes. I think this could be due to a large number of database connections for trivial metadata.To address this, make sure that the final result is also cached.