Difference of RPC Connection on PYNQ-Z1, VTA

Hello, I am working on tutorial - Auto-tuning convolutional network on VTA. Until now, I used to connect on RPC server using TVM_HOME/apps/vta_rpc/start_rpc_server.sh but on this tutorial, I had to use different way - using RPC tracker.

Can anyone explain the difference between those two ways? Is it only that with the tracker, it tracks the time of networking? Also, if there are differences, how can I pick which way I should use while running my own work?

I think the tracker is designed for tuning with autotvm.

Thanks. Maybe that’s right :slight_smile: