Closed seberg closed 1 year ago
Yes, you are correct that we should split training and serving data. I kind of intentionally "forgot" about that to not overcomplicate things in that first iteration, but we can quickly add that. I'll take care of that.
ooops, sorry wrong repo, had this randomly open!
I just realized that my old output folder with data has a
serving
andtraining
folder. I suspect we need to generate data for both serving and training and I think we currently do not have a step for creating it (or use it).CC @csadorf I suspect this is simple and relatively low prio, because it is easy to fix things up if we want it.