AxClient.get_best_trial(...) does not account for optimization_config parameter

[vassilyl]

Steps:

• Complete a multi-objective experiment with Service API.
• To get best predicted value for a single objective, create a new OptimizationConfig with the objective and pass it to AxClient.get_best_trial.

Expected:

The call returns best parameterization and predictions.

Observed:

An exception is raised from the implementation method '_get_best_trial`.

Ax/ax/service/utils/best_point_mixin.py

Line 266 in c5fa11a

"Please use `get_pareto_optimal_parameters` for multi-objective "

since AxClient.get_best_trial(...) doesn't pass optimization_config parameter value to implementation

Ax/ax/service/ax_client.py

Line 1540 in c5fa11a

def get_best_trial(

I've noticed similar issue for

Ax/ax/service/ax_client.py

Line 1554 in c5fa11a

def get_pareto_optimal_parameters(

[mgarrard]

Hi @vassilyl could you provide the full repro for where you create a new OptimizationConfig? I'm curious if maybe the optimization config isn't being saved to the experiment.