[Optimizers] Programmatic access to Optimizer's capabilities/requirements #119
Assignees
Milestone
Comments
Meganton
added a commit
that referenced
this issue
Apr 28, 2025
Adds `ignore_fidelity` parameter across optimizers for flexible fidelity management Introduces warnings and error handling for missing or unsupported fidelity/prior cases Updates grid, random, and Bayesian sampling to respect fidelity settings Improves test coverage for fidelity and prior requirements Fixes #118, #119, #120
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Right now, there's no way to programmatically know that an optimizer supports multi-fidelity, multi-objective, cost-awareness or requires priors to be set.
This makes parametrized testing quite difficult as a result. What would be ideal is that from
SearchMapping, we can traverse through it and get the capabilities of the optimizer, preferabbly fromcls.Supports.fidelity,cls.Supports.multi_objective, cls.Supports.priors`, etc...This would also be useful for CLI and documentation purposes, as right now this would all require manual specification and given then umber of optimizers, this would lead to a lot of inconsistencies building up over time if we were to do it manually.
One
partialissue is that we usepartialin theSearchMapping, meaning that we would have to gopartial._inner_function.Support.blah(or whatever its called). Not a huge issue unless the partial actually changes the capability of the optimizer.The text was updated successfully, but these errors were encountered: