Update formatting

emukit/bayesian_optimization/acquisitions/entropy_search.py

@@ -28,7 +28,6 @@ def __init__(
         proposal_function: Callable = None,
         burn_in_steps: int = 50,
     ) -> None:
-
         """
         Entropy Search acquisition function approximates the distribution of the global
         minimum and tries to decrease its entropy. See this paper for more details:
@@ -62,7 +61,6 @@ def __init__(
 
         # (unnormalized) density from which to sample the representer points to approximate pmin
         if proposal_function is None:
-
             ei = ExpectedImprovement(model)
 
             def prop_func(x):
@@ -286,7 +284,6 @@ def _sample_representer_points(self):
         return repr_points, repr_points_log
 
     def _get_proposal_function(self, model, space):
-
         # Define proposal function for multi-fidelity
         ei = ExpectedImprovement(model)
 

emukit/bayesian_optimization/acquisitions/expected_improvement.py

@@ -440,7 +440,6 @@ def _gradient_of_the_acquisition_second_term(
         else:
             B1, B2, B3 = np.zeros((q, d)), np.zeros((q, d)), np.zeros((q, d))
             for i in range(q):
-
                 # Assign helper variables needed by the gradients (See equation (6) for details)
                 ineq = [n for n in range(q) if n is not i]
                 Sigk_ik = Sigk[i, k]

emukit/bayesian_optimization/acquisitions/max_value_entropy_search.py

@@ -26,7 +26,6 @@ def __init__(
         num_samples: int = 10,
         grid_size: int = 5000,
     ) -> None:
-
         """
         MES acquisition function approximates the distribution of the value at the global
         minimum and tries to decrease its entropy. See this paper for more details:
@@ -159,7 +158,6 @@ def __init__(
         num_samples: int = 10,
         grid_size: int = 5000,
     ) -> None:
-
         """
         MUMBO acquisition function approximates the distribution of the value at the global
         minimum and tries to decrease its entropy.

emukit/bayesian_optimization/acquisitions/negative_lower_confidence_bound.py

@@ -12,7 +12,6 @@
 
 class NegativeLowerConfidenceBound(Acquisition):
     def __init__(self, model: Union[IModel, IDifferentiable], beta: float = 1.0) -> None:
-
         """
         This acquisition computes the negative lower confidence bound for a given input point. This is the same
         as optimizing the upper confidence bound if we would maximize instead of minimizing the objective function.

emukit/bayesian_optimization/epmgp.py

@@ -32,7 +32,6 @@ def joint_min(mu: np.ndarray, var: np.ndarray, with_derivatives: bool = False) -
         dlogPdSigma = np.zeros((D, int(0.5 * D * (D + 1))))
         dlogPdMudMu = np.zeros((D, D, D))
     for i in range(mu.shape[0]):
-
         # logP[k] ) self._min_factor(mu, var, 0)
         a = min_factor(mu, var, i)
 

emukit/bayesian_optimization/loops/bayesian_optimization_loop.py

@@ -29,7 +29,6 @@ def __init__(
         batch_size: int = 1,
         acquisition_optimizer: AcquisitionOptimizerBase = None,
     ):
-
         """
         Emukit class that implement a loop for building modular Bayesian optimization
 
@@ -74,7 +73,6 @@ def get_results(self):
 
 class BayesianOptimizationResults:
     def __init__(self, loop_state: LoopState):
-
         """
         Emukit class that takes as input the loop state and computes some results.
 

emukit/bayesian_optimization/loops/cost_sensitive_bayesian_optimization_loop.py

@@ -23,7 +23,6 @@ def __init__(
         update_interval: int = 1,
         acquisition_optimizer: AcquisitionOptimizerBase = None,
     ):
-
         """
         Emukit class that implements a loop for building modular cost sensitive Bayesian optimization.
 

emukit/bayesian_optimization/loops/unknown_constraint_bayesian_optimization_loop.py

@@ -27,7 +27,6 @@ def __init__(
         update_interval: int = 1,
         batch_size: int = 1,
     ):
-
         """
         Emukit class that implements a loop for building Bayesian optimization with an unknown constraint.
         For more information see:

emukit/benchmarking/loop_benchmarking/random_search.py

@@ -14,7 +14,6 @@ class RandomSearch(OuterLoop):
     def __init__(
         self, space: ParameterSpace, x_init: np.ndarray = None, y_init: np.ndarray = None, cost_init: np.ndarray = None
     ):
-
         """
         Simple loop to perform random search where in each iteration points are sampled uniformly at random
         over the input space.

emukit/core/loop/user_function_result.py

@@ -23,7 +23,7 @@ def __init__(self, X: np.ndarray, Y: np.ndarray, **kwargs) -> None:
             raise ValueError("y is expected to be 1-dimensional, actual dimensionality is {}".format(Y.ndim))
 
         self.extra_outputs = dict()
-        for (key, val) in kwargs.items():
+        for key, val in kwargs.items():
             if val.ndim != 1:
                 raise ValueError("Key word arguments must be 1-dimensional but {} is {}d".format(key, val.ndim))
             self.extra_outputs[key] = val

emukit/examples/fabolas/continuous_fidelity_entropy_search.py

@@ -63,7 +63,6 @@ def _sample_representer_points(self):
         return repr_points, repr_points_log
 
     def _get_proposal_function(self, model, space):
-
         # Define proposal function for multi-fidelity
         ei = ExpectedImprovement(model)
 

emukit/examples/fabolas/fabolas_model.py

@@ -9,7 +9,6 @@
 
 class FabolasKernel(GPy.kern.Kern):
     def __init__(self, input_dim, basis_func, a=1.0, b=1.0, active_dims=None):
-
         super(FabolasKernel, self).__init__(input_dim, active_dims, "fabolas_kernel")
 
         assert input_dim == 1

emukit/examples/gp_bayesian_optimization/single_objective_bayesian_optimization.py

@@ -37,7 +37,6 @@ def __init__(
         model_update_interval: int = int(1),
         batch_size: int = 1,
     ) -> None:
-
         """
         Generic class to run Bayesian optimization with GPyRegression model.
 

emukit/examples/gp_bayesian_optimization/unknown_constraint_bayesian_optimization.py

@@ -37,7 +37,6 @@ def __init__(
         batch_size: int = 1,
         model_update_interval: int = int(1),
     ) -> None:
-
         """
         Class to run Bayesian optimization with unknown contraints with GPyRegression model.
 

emukit/examples/multi_fidelity_dgp/multi_fidelity_deep_gp.py

@@ -233,7 +233,6 @@ def _build_likelihood(self):
         L = 0.0
         KL = 0.0
         for fidelity in range(self.num_layers):
-
             if (self._train_upto_fidelity != -1) and (fidelity > self._train_upto_fidelity):
                 continue
 

emukit/examples/preferential_batch_bayesian_optimization/pbbo/gp_models.py

@@ -129,7 +129,6 @@ def __init__(
         eta: float = 0.5,
         get_logger: Callable = None,
     ):
-
         super(EPComparisonGP, self).__init__(name=name)
 
         self.N, self.D = X.shape[0], X.shape[1]

emukit/examples/profet/meta_benchmarks/meta_forrester.py

@@ -65,7 +65,6 @@ def meta_forrester(fname_objective: str) -> Tuple[UserFunctionWrapper, Parameter
     objective.load_state_dict(data["state_dict"])
 
     def objective_function(config):
-
         Ht = np.repeat(task_feature_objective[None, :], config.shape[0], axis=0)
         x = np.concatenate((config, Ht), axis=1)
         x_norm = torch.from_numpy((x - x_mean_objective) / x_std_objective).float()

emukit/examples/profet/meta_benchmarks/meta_surrogates.py

@@ -19,7 +19,6 @@ def objective_function(
     log_objective=False,
     with_noise=True,
 ):
-
     Ht = np.repeat(task_feature_objective[None, :], config.shape[0], axis=0)
     x = np.concatenate((config, Ht), axis=1)
     x_norm = torch.from_numpy((x - x_mean_objective) / x_std_objective).float()

emukit/examples/profet/train_meta_model.py

@@ -33,7 +33,6 @@
 
 
 def download_data(path, source="http://www.ml4aad.org/wp-content/uploads/2019/05/profet_data.tar.gz"):
-
     l = urlretrieve(source)[0]
 
     tar = tarfile.open(l)

emukit/model_wrappers/gpy_quadrature_wrappers.py

@@ -182,7 +182,6 @@ def __init__(
 
         # product kernel from parameters
         if gpy_matern is None:
-
             input_dim = len(lengthscales)
             if input_dim < 1:
                 raise ValueError("'lengthscales' must contain at least 1 value.")
@@ -295,7 +294,6 @@ def __init__(
 
         # product kernel from parameters
         if gpy_matern is None:
-
             input_dim = len(lengthscales)
             if input_dim < 1:
                 raise ValueError("'lengthscales' must contain at least 1 value.")
@@ -406,7 +404,6 @@ def __init__(
 
         # product kernel from parameters
         if gpy_matern is None:
-
             input_dim = len(lengthscales)
             if input_dim < 1:
                 raise ValueError("'lengthscales' must contain at least 1 value.")
@@ -533,7 +530,6 @@ def __init__(
         variance: Optional[float] = None,
         input_dim: Optional[int] = None,
     ):
-
         if gpy_brownian is not None:
             if input_dim is not None or variance is not None:
                 warnings.warn("gpy_brownian and variance is given. The variance will be ignore.")

emukit/quadrature/kernels/quadrature_brownian.py

@@ -44,7 +44,6 @@ def __init__(
         measure: IntegrationMeasure,
         variable_names: str,
     ) -> None:
-
         if measure.input_dim != 1:
             raise ValueError(
                 "Integration measure for Brownian motion kernel must be 1-dimensional. Current dimesnion is ({}).".format(

emukit/quadrature/kernels/quadrature_kernels.py

@@ -146,7 +146,6 @@ def __init__(
         measure: IntegrationMeasure,
         variable_names: str,
     ) -> None:
-
         super().__init__(kern=kern, measure=measure, variable_names=variable_names)
 
     def _qK_unscaled(self, x2: np.ndarray, skip: List[int] = None) -> np.ndarray:

emukit/quadrature/loop/wsabil_loop.py

@@ -30,7 +30,6 @@ class WSABILLoop(OuterLoop):
     def __init__(
         self, model: WSABIL, model_updater: ModelUpdater = None, acquisition_optimizer: AcquisitionOptimizerBase = None
     ):
-
         # WSABI-L is used with uncertainty sampling.
         acquisition = UncertaintySampling(model, measure_power=1)
 

emukit/test_functions/multi_fidelity/hartmann.py

@@ -62,7 +62,6 @@ def multi_fidelity_hartmann_3d() -> Tuple[MultiSourceFunctionWrapper, ParameterS
     delta = np.array([0.01, -0.01, -0.1, 0.1])
 
     def high(x):
-
         res = 0
         for i in range(4):
             temp = 0
@@ -73,7 +72,6 @@ def high(x):
         return res[:, None]
 
     def medium(x):
-
         alpha_m = alpha + delta
 
         res = 0
@@ -86,7 +84,6 @@ def medium(x):
         return res[:, None]
 
     def low(x):
-
         alpha_l = alpha + 2 * delta
 
         res = 0

integration_tests/emukit/bayesian_optimization/test_create_bayesian_optimization_loop.py

@@ -6,7 +6,6 @@
 
 
 def test_loop_state():
-
     fcn, cs = forrester_function()
     n_init = 5
 

tests/emukit/bayesian_optimization/test_epmgp.py

@@ -4,7 +4,6 @@
 
 
 def test_joint_min():
-
     # Uniform distribution
     n_points = 5
     m = np.ones([n_points])

tests/emukit/bayesian_optimization/test_local_penalization.py

@@ -8,7 +8,6 @@
 
 
 def test_penalization_function_shape():
-
     model = MockModel()
     lp = LocalPenalization(model)
     lp.update_batches(np.zeros((5, 1)), 1, -0.1)
@@ -18,7 +17,6 @@ def test_penalization_function_shape():
 
 
 def test_penalization_function_gradients_shape():
-
     model = MockModel()
     lp = LocalPenalization(model)
     lp.update_batches(np.zeros((5, 2)), 1, -0.1)

tests/emukit/core/test_constraints.py

@@ -76,15 +76,13 @@ def test_inequality_constraint_no_bounds():
 
 
 def test_inequality_constraint_all_inf_bound():
-
     lower = np.array([-np.inf, 0])
     upper = np.array([np.inf, 1])
     with pytest.raises(ValueError):
         InequalityConstraint(lower, upper)
 
 
 def test_inequality_constraint_unbounded():
-
     lower = np.array([-np.inf, 0])
     upper = None
     with pytest.raises(ValueError):

tests/emukit/experimental_design/test_batch_experimental_design.py

@@ -63,7 +63,6 @@ def optimize(self):
 
 
 def test_batch_point_calculator(mock_model):
-
     acquisition = mock.create_autospec(Acquisition)
     acquisition_optimizer = mock.create_autospec(GradientAcquisitionOptimizer)
     acquisition_optimizer.optimize.return_value = (np.zeros((1, 1)), 0)

tests/emukit/model_wrappers/test_gpy_wrappers_quadrature.py

@@ -215,7 +215,6 @@ def wrapper_matern52_2(dim1, gpy_matern52):
 
 @pytest.mark.parametrize("wrapper", gpy_test_list)
 def test_create_emukit_model_from_gpy_model_types(wrapper):
-
     gpy_model = GPy.models.GPRegression(kernel=wrapper["gpy_kernel"], X=wrapper["data"][0], Y=wrapper["data"][1])
     emukit_gp = create_emukit_model_from_gpy_model(gpy_model=gpy_model, measure=wrapper["measure"])
 

tests/emukit/quadrature/ground_truth_integrals_methods.py

@@ -96,7 +96,6 @@ def mc_integral_var_from_measure(num_samples: int, model: WarpedBayesianQuadratu
         vZ_SAMPLES = np.zeros(num_runs)
         _, vZ = model.integrate()
         for i in range(num_runs):
-
             vZ_samples = mc_integral_var_from_measure(num_samples, model)
             vZ_SAMPLES[i] = vZ_samples
             print(".", end="", flush=True)

tests/emukit/quadrature/test_quadrature_kernels.py

@@ -586,7 +586,6 @@ def test_qkernel_gradient_values(kernel_embedding):
 
 
 def test_brownian_qkernel_raises():
-
     # measure has wrong dimensionality
     wrong_bounds = [(1, 2), (1, 2)]
     measure = LebesgueMeasure.from_bounds(bounds=wrong_bounds)

tests/emukit/quadrature/test_quadrature_models.py

@@ -210,7 +210,6 @@ def test_warped_model_transforms(model):
 
 @pytest.mark.parametrize("model", all_models_test_list)
 def test_warped_model_gradient_values(model, data):
-
     # gradient of mean
     func = lambda z: model.predict(z)[0][:, 0]
     dfunc = lambda z: model.get_prediction_gradients(z)[0].T

tests/emukit/sensitivity/test_emukit_sensitivity.py

@@ -16,7 +16,6 @@ def space():
 
 
 def test_model_based_montecarlo_sensitivity(space):
-
     model = mock.create_autospec(IModel)
     model.predict.return_value = (0.1 * np.ones((3, 2)), np.zeros((3, 2)))
 
@@ -43,7 +42,6 @@ def test_model_based_montecarlo_sensitivity(space):
 
 
 def test_model_free_montecarlo_sensitivity(space):
-
     mock_function = lambda x: 0.1 * np.ones((3, 1))
 
     sensitivity = ModelFreeMonteCarloSensitivity(mock_function, space)

tests/emukit/test_functions/test_ishigami_function.py

@@ -4,7 +4,6 @@
 
 
 def test_ishigami_function():
-
     ishigami = Ishigami(a=5, b=0.1)
 
     assert ishigami.fidelity1(np.array([[0, 1, 0]])).shape == (1,)