Graph partitioning and Heterogeneous Execution

Thanks. But we will fail to lower if we don’t register tvm compute and schedule, right?