trivial Baysian Optimization raises TypeError

[Permafacture]

I don't know if this is a sherpa issue or GPyOpt, but since sherpa is calling gPyOpt and determines what versions of libraries are installed, I think this is the place to start.

This minimal example raises an error

search_params = [ sherpa.Continuous('lr', [0.001, 0.1], 'log')]
algorithm = sherpa.algorithms.GPyOpt(max_num_trials=50)
study = sherpa.Study(parameters=search_params, algorithm=algorithm, lower_is_better=True, disable_dashboard=True)
for trial in study:
    loss = trial.parameters['lr']
    study.add_observation(trial=trial, iteration=i, objective=loss)
    study.finalize(trial=trial)
    study.save('.')

give this output

INFO:GP:initializing Y
INFO:GP:initializing inference method
INFO:GP:adding kernel and likelihood as parameters
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Input In [205], in <cell line: 14>()
      8 algorithm2 = sherpa.algorithms.GPyOpt(max_num_trials=50)
      9 study2 = sherpa.Study(parameters=search_params,
     10                       algorithm=algorithm2,
     11                       lower_is_better=True,
     12                       disable_dashboard=True)
---> 14 for trial2 in study2:
     15     loss = trial2.parameters['lr']
     16     study2.add_observation(trial=trial2, iteration=i, objective=loss)

File ~/code/trace2vec/.env-trace2vec/lib/python3.10/site-packages/sherpa/core.py:376, in Study.__next__(self)
    372 def __next__(self):
    373     """
    374     Allows to write `for trial in study:`.
    375     """
--> 376     t = self.get_suggestion()
    377     if isinstance(t, Trial):
    378         return t

File ~/code/trace2vec/.env-trace2vec/lib/python3.10/site-packages/sherpa/core.py:214, in Study.get_suggestion(self)
    211 if len(self._trial_queue) != 0:
    212     return self._trial_queue.popleft()
--> 214 p = self.algorithm.get_suggestion(self.parameters, self.results,
    215                                   self.lower_is_better)
    216 if isinstance(p, dict):
    217     self.num_trials += 1

File ~/code/trace2vec/.env-trace2vec/lib/python3.10/site-packages/sherpa/algorithms/bayesian_optimization.py:109, in GPyOpt.get_suggestion(self, parameters, results, lower_is_better)
    105 self.next_trials.clear()
    107 X, y, y_var = self._prepare_data_for_bayes_opt(parameters, results)
--> 109 batch = self._generate_bayesopt_batch(self.domain, X, y, y_var,
    110                                       lower_is_better)
    112 batch_list_of_dicts = self._reverse_to_sherpa_format(batch,
    113                                                      parameters)
    115 self.next_trials.extend(batch_list_of_dicts)

File ~/code/trace2vec/.env-trace2vec/lib/python3.10/site-packages/sherpa/algorithms/bayesian_optimization.py:149, in GPyOpt._generate_bayesopt_batch(self, domain, X, y, y_var, lower_is_better)
    137     kwargs = {'model_type': self.model_type}
    139 bo_step = gpyopt_package.methods.BayesianOptimization(f=None,
    140                                                       domain=domain,
    141                                                       X=X, Y=y_adjusted,
   (...)
    147                                                       exact_feval=False,
    148                                                       **kwargs)
--> 149 return bo_step.suggest_next_locations()

File ~/code/trace2vec/.env-trace2vec/lib/python3.10/site-packages/GPyOpt/core/bo.py:69, in BO.suggest_next_locations(self, context, pending_X, ignored_X)
     66 self.context = context
     67 self._update_model(self.normalization_type)
---> 69 suggested_locations = self._compute_next_evaluations(pending_zipped_X = pending_X, ignored_zipped_X = ignored_X)
     71 return suggested_locations

File ~/code/trace2vec/.env-trace2vec/lib/python3.10/site-packages/GPyOpt/core/bo.py:236, in BO._compute_next_evaluations(self, pending_zipped_X, ignored_zipped_X)
    233     duplicate_manager = None
    235 ### We zip the value in case there are categorical variables
--> 236 return self.space.zip_inputs(self.evaluator.compute_batch(duplicate_manager=duplicate_manager, context_manager= self.acquisition.optimizer.context_manager))

File ~/code/trace2vec/.env-trace2vec/lib/python3.10/site-packages/GPyOpt/core/evaluators/batch_local_penalization.py:37, in LocalPenalization.compute_batch(self, duplicate_manager, context_manager)
     33 k=1
     35 if self.batch_size >1:
     36     # ---------- Approximate the constants of the the method
---> 37     L = estimate_L(self.acquisition.model.model,self.acquisition.space.get_bounds())
     38     Min = self.acquisition.model.model.Y.min()
     40 # --- GET the remaining elements

File ~/code/trace2vec/.env-trace2vec/lib/python3.10/site-packages/GPyOpt/core/evaluators/batch_local_penalization.py:67, in estimate_L(model, bounds, storehistory)
     65 x0 = samples[np.argmin(pred_samples)]
     66 res = scipy.optimize.minimize(df,x0, method='L-BFGS-B',bounds=bounds, args = (model,x0), options = {'maxiter': 200})
---> 67 minusL = res.fun[0][0]
     68 L = -minusL
     69 if L<1e-7: L=10  ## to avoid problems in cases in which the model is flat.

TypeError: 'float' object is not subscriptable

And res.fun is a float

Versions:
python 3.10
Sherpa 1.0.6
GPyOpt 1.2.6
scipy 1.9.0