WIP: implement hopsy for init sampling

also: print statements for user info;

CADETProcess/optimization/axAdapater.py

@@ -134,7 +134,7 @@ class AxInterface(OptimizerBase):
     ftol = UnsignedFloat(default=0.0025)
     acquisition_fn = Typed(default=qNoisyExpectedImprovement)
     surrogate_model = Typed(default=FixedNoiseGP)
-    
+
     _specific_options = [
         'n_init_evals', 'n_max_evals', 'seed', 'xtol', 'cvtol', 'ftol',
         'acquisition_fn', 'surrogate_model'
@@ -174,7 +174,7 @@ def _setup_linear_constraints(optimizationProblem: OptimizationProblem):
             )
             parameter_constraints.append(constr)
 
-        return
+        return parameter_constraints
 
     @classmethod
     def _setup_searchspace(cls, optimizationProblem):
@@ -235,8 +235,12 @@ def run(self, optimization_problem, x0):
         search_space = self._setup_searchspace(self.optimization_problem)
         objectives = self._setup_objectives()
 
-
         if len(objectives) > 1:
+            is_moo = True
+        else:
+            is_moo = False
+
+        if is_moo:
             optimization_config = ax.MultiObjectiveOptimizationConfig(
                 objective=ax.MultiObjective(objectives)
             )
@@ -252,15 +256,15 @@ def run(self, optimization_problem, x0):
                 botorch_acqf_class=self.acquisition_fn,  # Optional, will use default if unspecified
             )
 
-        # Alternative: Use a model bridge directly with a botorch model. 
+        # Alternative: Use a model bridge directly with a botorch model.
         # This allows for more control and does not introduce another "magic"
         # middle layer in between. But currently I can't get this to work.
         # The above is as suggested in the PR
         # Model = BoTorchModel(
-        #   acquisition_class=UpperConfidenceBound, 
+        #   acquisition_class=UpperConfidenceBound,
         #   surrogate=Surrogate(FixedNoiseGP)
         # )
-        
+
         # model = TorchModelBridge(
         #     experiment=self.ax_experiment,
         #     search_space=search_space,
@@ -269,7 +273,7 @@ def run(self, optimization_problem, x0):
         #     transforms=Cont_X_trans + Y_trans
         # )
 
-        
+
         runner = CADETProcessRunner(
             optimization_problem=self.optimization_problem)
 
@@ -280,10 +284,22 @@ def run(self, optimization_problem, x0):
             runner=runner,
         )
 
+        # init_samples = self.optimization_problem.create_initial_values(
+        #     n_samples=self.n_init_evals,
+        #     method="chebyshev",
+        #     seed=self.seed + 5641,
+        # )
+
+        # self.optimization_problem.evaluate_objectives_population(
+        #     init_samples, n_cores=1
+        # )
+
+
+
 
         # TODO: termination criteria
-        print(f"Running Sobol initialization trials...")
         with manual_seed(seed=self.seed):
+            print(f"Running Sobol initialization trials...")
             sobol = Models.SOBOL(search_space=self.ax_experiment.search_space)
             for i in range(self.n_init_evals):
                 # Produce a GeneratorRun from the model, which contains proposed arm(s) and other metadata
@@ -307,17 +323,29 @@ def run(self, optimization_problem, x0):
 
                 trial_data = trial.fetch_data()
 
+
             i += 1
             while i < self.n_max_evals:
-                print(f"Running optimization trial {i+1}/{self.n_max_evals}...")
                 # Reinitialize GP+EI model at each step with updated data.
-
-                # this 
+                # this
                 model = Model(
                     experiment=self.ax_experiment,
                     data=self.ax_experiment.fetch_data(),
                 )
 
+                if i == self.n_init_evals:
+                    if is_moo:
+                        srgm = model.model.model._get_name()
+                        acqf = model.model.acqf_constructor.__name__.split("_")[1]
+                    else:
+                        srgm = model.model.surrogate.model._get_name()
+                        acqf = model.model.botorch_acqf_class.__name__
+                    print(f"Starting bayesian optimization loop...")
+                    print(f"Surrogate model: {srgm}")
+                    print(f"Acquisition function: {acqf}")
+
+                print(f"Running optimization trial {i+1}/{self.n_max_evals}...")
+
                 # generates samples
                 # samples can be accessed here by sample_generator.arms:
                 sample_generator = model.gen(n=1)
@@ -353,5 +381,5 @@ def run(self, optimization_problem, x0):
     test.test_single_objective()
     test.test_single_objective_linear_constraints()
     test.test_multi_objective()
-    
-    #x: [0.23783216 0.50267604 0.43198473], f: [4.81493504]
+
+    #x: [0.23783216 0.50267604 0.43198473], f: [4.81493504]