Thanks for reporting. This PR (https://github.com/apache/incubator-tvm/pull/6683) fixed this bug.
However, typically for large graphs like this. We don’t treat them as a single search task. Treating it as a single large graph makes the search very inefficient.
For large graphs, we use Relay to partition the large graphs into several small subgraphs and treat a small subgraph as a search task. We then use the auto-scheduler to optimize every subgraph and use a task scheduler to schedule search tasks.
The upstream of relay integration and task scheduler (PR) is still work in progress. Hopefully, we can finish the relay integration with tutorials in one or two weeks.