Brian Kroth

> > > > It seems like when we break the training of the HPO model for some trials (a low number), then we start to resume the training based...

Some things to note: - TunableGroups provide reusable parameter specifications for individual Environments (e.g., VM, OS, App) - They can be combined in an Experiment to do cross Environment co-tuning....

I haven't thought through the full implementation in mlos_bench yet, but it likely starts with something here: - https://github.com/microsoft/MLOS/tree/main/mlos_bench/mlos_bench/config/schemas/tunables - https://github.com/microsoft/MLOS/blob/main/mlos_bench/mlos_bench/tunables/tunable_groups.py - https://github.com/microsoft/MLOS/blob/main/mlos_bench/mlos_bench/optimizers/convert_configspace.py

Regarding your observation, I think it's mostly a documentation issue. This PR in ConfigSpace seems to indicate that Forbidden clauses between Hyperparameters, not just constants, is possible. https://github.com/automl/ConfigSpace/pull/245/files How I...

@motus any other thoughts?

Related: https://microsoft.github.io/FLAML/docs/Use-Cases/Tune-User-Defined-Function/#more-constraints-on-the-tuning

FWIW: https://github.com/automl/ConfigSpace/issues/352

Wrote some notes for this in https://github.com/microsoft/MLOS/issues/380#issuecomment-1832711391

#676 adds some initial support for this

Another related thing: - reallocate vs. deprovision This is especially important for VMs as dellocation (without deleting the original resources), is slightly more expensive than a reboot, since it involves...