basically there are GPU servers spinning up and down in the background and we don't want to overload them.
billions-memory-82613
04/12/2023, 5:47 PM
Right now we stare at the deploy logs and update the % traffic to the experiment manually every 30 mins over the interval of T hours
(and stop increasing if the deploy goes bad)
f
fresh-football-47124
04/12/2023, 11:36 PM
Right now, we don’t have an automatic way to do ramping through the UI. We do have an API that would allow you to script it if you would like