You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@tvm.apache.org by GitBox <gi...@apache.org> on 2020/03/05 23:52:35 UTC

[GitHub] [incubator-tvm] comaniac commented on issue #4995: [AutoTVM] Avoid using RPC for LocalRunner

comaniac commented on issue #4995: [AutoTVM] Avoid using RPC for LocalRunner
URL: https://github.com/apache/incubator-tvm/pull/4995#issuecomment-595503657
 
 
   > The main reason to adopt the RPC runner is for the process isolation -- because we cannot guarantee that the local runner won't crash and as a result crash the AutoTVM process itself.
   > 
   > It might be useful though to have a persistent RPC server over the tuning process
   
   While I do agree with you that we should make sure the failure of a measurement won't crash the tuning process itself, I still don't think local RPC is the only and the best solution when considering its overhead and reliability. For example, we can use a separate thread to run `time_evaluator` locally and catch the exception if failed, just like what we did for task extraction.

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
users@infra.apache.org


With regards,
Apache Git Services