Remove unused mypy ignores (#794)

* Remove unused mypy ignores

* Rerun with numpy-1.23.5

alibi_detect/ad/model_distillation.py

@@ -176,7 +176,7 @@ def score(self, X: np.ndarray, batch_size: int = int(1e10), return_predictions:
 
         # scale predictions
         if self.temperature != 1.:
-            y = y ** (1 / self.temperature)  # type: ignore
+            y = y ** (1 / self.temperature)
             y = (y / tf.reshape(tf.reduce_sum(y, axis=-1), (-1, 1))).numpy()
 
         if self.loss_type == 'kld':

alibi_detect/base.py

@@ -193,8 +193,8 @@ def _nested_detector(self):
         """
         The low-level nested detector.
         """
-        detector = self._detector if hasattr(self, '_detector') else self  # type: ignore[attr-defined]
-        detector = detector._detector if hasattr(detector, '_detector') else detector  # type: ignore[attr-defined]
+        detector = self._detector if hasattr(self, '_detector') else self
+        detector = detector._detector if hasattr(detector, '_detector') else detector
         return detector
 
 

alibi_detect/cd/base.py

@@ -151,9 +151,9 @@ def preprocess(self, x: Union[np.ndarray, list]) -> Tuple[Union[np.ndarray, list
                 x_ref = self.preprocess_fn(self.x_ref)
             else:
                 x_ref = self.x_ref
-            return x_ref, x  # type: ignore[return-value]
+            return x_ref, x
         else:
-            return self.x_ref, x  # type: ignore[return-value]
+            return self.x_ref, x
 
     def get_splits(
             self,
@@ -412,9 +412,9 @@ def preprocess(self, x: Union[np.ndarray, list]) -> Tuple[Union[np.ndarray, list
                 x_ref = self.preprocess_fn(self.x_ref)
             else:
                 x_ref = self.x_ref
-            return x_ref, x  # type: ignore[return-value]
+            return x_ref, x
         else:
-            return self.x_ref, x  # type: ignore[return-value]
+            return self.x_ref, x
 
     def get_splits(self, x_ref: Union[np.ndarray, list], x: Union[np.ndarray, list]) \
             -> Tuple[Tuple[Union[np.ndarray, list], Union[np.ndarray, list]],
@@ -441,7 +441,7 @@ def get_splits(self, x_ref: Union[np.ndarray, list], x: Union[np.ndarray, list])
 
         if isinstance(x_ref, np.ndarray):
             x_ref_tr, x_ref_te = x_ref[idx_ref_tr], x_ref[idx_ref_te]
-            x_cur_tr, x_cur_te = x[idx_cur_tr], x[idx_cur_te]  # type: ignore[call-overload]
+            x_cur_tr, x_cur_te = x[idx_cur_tr], x[idx_cur_te]
         elif isinstance(x, list):
             x_ref_tr, x_ref_te = [x_ref[_] for _ in idx_ref_tr], [x_ref[_] for _ in idx_ref_te]
             x_cur_tr, x_cur_te = [x[_] for _ in idx_cur_tr], [x[_] for _ in idx_cur_te]
@@ -814,7 +814,7 @@ def predict(self, x: Union[np.ndarray, list], return_p_val: bool = True, return_
                 pass
         self.x_ref = update_reference(self.x_ref, x, self.n, self.update_x_ref)  # type: ignore[arg-type]
         # used for reservoir sampling
-        self.n += len(x)  # type: ignore
+        self.n += len(x)
 
         # populate drift dict
         cd = concept_drift_dict()
@@ -1205,7 +1205,7 @@ def predict(self,  # type: ignore[override]
         if isinstance(self.update_ref, dict) and self.preprocess_fn is not None and self.preprocess_at_init:
             x = self.preprocess_fn(x)
         self.x_ref = update_reference(self.x_ref, x, self.n, self.update_ref)  # type: ignore[arg-type]
-        self.c_ref = update_reference(self.c_ref, c, self.n, self.update_ref)  # type: ignore[arg-type]
+        self.c_ref = update_reference(self.c_ref, c, self.n, self.update_ref)
         # used for reservoir sampling
         self.n += len(x)
 

alibi_detect/cd/base_online.py

@@ -152,8 +152,8 @@ def _initialise_state(self) -> None:
         an example).
         """
         self.t = 0  # corresponds to a test set of ref data
-        self.test_stats = np.array([])  # type: ignore[var-annotated]
-        self.drift_preds = np.array([])  # type: ignore[var-annotated]
+        self.test_stats = np.array([])
+        self.drift_preds = np.array([])
 
     def reset(self) -> None:
         """
@@ -404,9 +404,9 @@ def _initialise_state(self) -> None:
         an example).
         """
         self.t = 0
-        self.xs = np.array([])  # type: ignore[var-annotated]
+        self.xs = np.array([])
         self.test_stats = np.empty([0, len(self.window_sizes), self.n_features])
-        self.drift_preds = np.array([])  # type: ignore[var-annotated]
+        self.drift_preds = np.array([])
 
     @abstractmethod
     def _check_drift(self, test_stats: np.ndarray, thresholds: np.ndarray) -> int:

alibi_detect/cd/chisquare.py

@@ -94,7 +94,7 @@ def __init__(
                                  'Dict[int, int], Dict[int, List[int]]')
         else:  # infer number of possible categories for each feature from reference data
             x_flat = self.x_ref.reshape(self.x_ref.shape[0], -1)
-            categories_per_feature = {f: list(np.unique(x_flat[:, f]))  # type: ignore
+            categories_per_feature = {f: list(np.unique(x_flat[:, f]))
                                       for f in range(self.n_features)}
         self.x_ref_categories = categories_per_feature
 

alibi_detect/cd/classifier.py

@@ -167,20 +167,20 @@ def __init__(
             [kwargs.pop(k, None) for k in pop_kwargs]
             if dataset is None:
                 kwargs.update({'dataset': TFDataset})
-            self._detector = ClassifierDriftTF(*args, **kwargs)  # type: ignore
+            self._detector = ClassifierDriftTF(*args, **kwargs)
         elif backend == Framework.PYTORCH:
             pop_kwargs = ['use_calibration', 'calibration_kwargs', 'use_oob']
             [kwargs.pop(k, None) for k in pop_kwargs]
             if dataset is None:
                 kwargs.update({'dataset': TorchDataset})
             if dataloader is None:
                 kwargs.update({'dataloader': DataLoader})
-            self._detector = ClassifierDriftTorch(*args, **kwargs)  # type: ignore
+            self._detector = ClassifierDriftTorch(*args, **kwargs)
         else:
             pop_kwargs = ['reg_loss_fn', 'optimizer', 'learning_rate', 'batch_size', 'preprocess_batch_fn',
                           'epochs', 'train_kwargs', 'device',  'dataset', 'dataloader', 'verbose']
             [kwargs.pop(k, None) for k in pop_kwargs]
-            self._detector = ClassifierDriftSklearn(*args, **kwargs)  # type: ignore
+            self._detector = ClassifierDriftSklearn(*args, **kwargs)
         self.meta = self._detector.meta
 
     def predict(self, x: Union[np.ndarray, list],  return_p_val: bool = True,

alibi_detect/cd/context_aware.py

@@ -115,9 +115,9 @@ def __init__(
 
         if backend == Framework.TENSORFLOW:
             kwargs.pop('device', None)
-            self._detector = ContextMMDDriftTF(*args, **kwargs)  # type: ignore
+            self._detector = ContextMMDDriftTF(*args, **kwargs)
         else:
-            self._detector = ContextMMDDriftTorch(*args, **kwargs)  # type: ignore
+            self._detector = ContextMMDDriftTorch(*args, **kwargs)
         self.meta = self._detector.meta
 
     def predict(self, x: Union[np.ndarray, list], c: np.ndarray,

alibi_detect/cd/keops/mmd.py

@@ -169,7 +169,7 @@ def score(self, x: Union[np.ndarray, list]) -> Tuple[float, float, float]:
         x_ref = torch.from_numpy(x_ref).float()  # type: ignore[assignment]
         x = torch.from_numpy(x).float()  # type: ignore[assignment]
         # compute kernel matrix, MMD^2 and apply permutation test
-        m, n = x_ref.shape[0], x.shape[0]  # type: ignore[union-attr]
+        m, n = x_ref.shape[0], x.shape[0]
         perms = [torch.randperm(m + n) for _ in range(self.n_permutations)]
         # TODO - Rethink typings (related to https://github.com/SeldonIO/alibi-detect/issues/540)
         x_all = torch.cat([x_ref, x], 0)  # type: ignore[list-item]

alibi_detect/cd/learned_kernel.py

@@ -166,7 +166,7 @@ def __init__(
             else:
                 detector = LearnedKernelDriftKeops
 
-        self._detector = detector(*args, **kwargs)  # type: ignore
+        self._detector = detector(*args, **kwargs)
         self.meta = self._detector.meta
 
     def predict(self, x: Union[np.ndarray, list],  return_p_val: bool = True,

alibi_detect/cd/lsdd.py

@@ -94,9 +94,9 @@ def __init__(
 
         if backend == Framework.TENSORFLOW:
             kwargs.pop('device', None)
-            self._detector = LSDDDriftTF(*args, **kwargs)  # type: ignore
+            self._detector = LSDDDriftTF(*args, **kwargs)
         else:
-            self._detector = LSDDDriftTorch(*args, **kwargs)  # type: ignore
+            self._detector = LSDDDriftTorch(*args, **kwargs)
         self.meta = self._detector.meta
 
     def predict(self, x: Union[np.ndarray, list], return_p_val: bool = True, return_distance: bool = True) \

alibi_detect/cd/lsdd_online.py

@@ -96,7 +96,7 @@ def __init__(
 
         if backend == Framework.TENSORFLOW:
             kwargs.pop('device', None)
-            self._detector = LSDDDriftOnlineTF(*args, **kwargs)  # type: ignore
+            self._detector = LSDDDriftOnlineTF(*args, **kwargs)
         else:
             self._detector = LSDDDriftOnlineTorch(*args, **kwargs)  # type: ignore
         self.meta = self._detector.meta

alibi_detect/cd/mmd.py

@@ -116,7 +116,7 @@ def __init__(
                 from alibi_detect.utils.keops.kernels import GaussianRBF  # type: ignore
             kwargs.update({'kernel': GaussianRBF})
 
-        self._detector = detector(*args, **kwargs)  # type: ignore
+        self._detector = detector(*args, **kwargs)
         self.meta = self._detector.meta
 
     def predict(self, x: Union[np.ndarray, list], return_p_val: bool = True, return_distance: bool = True) \

alibi_detect/cd/mmd_online.py

@@ -96,7 +96,7 @@ def __init__(
 
         if backend == Framework.TENSORFLOW:
             kwargs.pop('device', None)
-            self._detector = MMDDriftOnlineTF(*args, **kwargs)  # type: ignore
+            self._detector = MMDDriftOnlineTF(*args, **kwargs)
         else:
             self._detector = MMDDriftOnlineTorch(*args, **kwargs)  # type: ignore
         self.meta = self._detector.meta

alibi_detect/cd/pytorch/context_aware.py

@@ -135,7 +135,7 @@ def score(self,  # type: ignore[override]
         """
         x_ref, x = self.preprocess(x)
         x_ref = torch.from_numpy(x_ref).to(self.device)  # type: ignore[assignment]
-        c_ref = torch.from_numpy(self.c_ref).to(self.device)  # type: ignore[assignment]
+        c_ref = torch.from_numpy(self.c_ref).to(self.device)
 
         # Hold out a portion of contexts for conditioning on
         n, n_held = len(c), int(len(c)*self.prop_c_held)

alibi_detect/cd/pytorch/learned_kernel.py

@@ -200,7 +200,7 @@ def score(self, x: Union[np.ndarray, list]) -> Tuple[float, float, float]:
         self.kernel = deepcopy(self.original_kernel) if self.retrain_from_scratch else self.kernel
         self.kernel = self.kernel.to(self.device)
         train_args = [self.j_hat, (dl_ref_tr, dl_cur_tr), self.device]
-        LearnedKernelDriftTorch.trainer(*train_args, **self.train_kwargs)  # type: ignore
+        LearnedKernelDriftTorch.trainer(*train_args, **self.train_kwargs)
 
         if isinstance(x_ref_te, np.ndarray) and isinstance(x_cur_te, np.ndarray):
             x_all = np.concatenate([x_ref_te, x_cur_te], axis=0)

alibi_detect/cd/pytorch/lsdd.py

@@ -115,8 +115,8 @@ def _initialize_kernel(self, x_ref: torch.Tensor):
     def _configure_normalization(self, x_ref: torch.Tensor, eps: float = 1e-12):
         x_ref_means = x_ref.mean(0)
         x_ref_stds = x_ref.std(0)
-        self._normalize = lambda x: (torch.as_tensor(x) - x_ref_means) / (x_ref_stds + eps)  # type: ignore[assignment]
-        self._unnormalize = lambda x: (torch.as_tensor(x) * (x_ref_stds + eps)  # type: ignore[assignment]
+        self._normalize = lambda x: (torch.as_tensor(x) - x_ref_means) / (x_ref_stds + eps)
+        self._unnormalize = lambda x: (torch.as_tensor(x) * (x_ref_stds + eps)
                                        + x_ref_means).cpu().numpy()
 
     def _configure_kernel_centers(self, x_ref: torch.Tensor):

alibi_detect/cd/pytorch/lsdd_online.py

@@ -103,7 +103,7 @@ def __init__(
         else:
             sigma = torch.from_numpy(sigma).to(self.device) if isinstance(sigma,  # type: ignore[assignment]
                                                                           np.ndarray) else None
-            self.kernel = GaussianRBF(sigma)  # type: ignore[arg-type]
+            self.kernel = GaussianRBF(sigma)
 
         if self.n_kernel_centers is None:
             self.n_kernel_centers = 2 * window_size

alibi_detect/cd/pytorch/mmd.py

@@ -128,7 +128,7 @@ def score(self, x: Union[np.ndarray, list]) -> Tuple[float, float, float]:
         x = torch.from_numpy(x).to(self.device)  # type: ignore[assignment]
         # compute kernel matrix, MMD^2 and apply permutation test using the kernel matrix
         # TODO: (See https://github.com/SeldonIO/alibi-detect/issues/540)
-        n = x.shape[0]  # type: ignore
+        n = x.shape[0]
         kernel_mat = self.kernel_matrix(x_ref, x)  # type: ignore[arg-type]
         kernel_mat = kernel_mat - torch.diag(kernel_mat.diag())  # zero diagonal
         mmd2 = mmd2_from_kernel_matrix(kernel_mat, n, permute=False, zero_diag=False)

alibi_detect/cd/pytorch/spot_the_diff.py

@@ -223,8 +223,8 @@ def predict(
         and the trained model.
         """
         preds = self._detector.predict(x, return_p_val, return_distance, return_probs, return_model=True)
-        preds['data']['diffs'] = preds['data']['model'].diffs.detach().cpu().numpy()  # type: ignore
-        preds['data']['diff_coeffs'] = preds['data']['model'].coeffs.detach().cpu().numpy()  # type: ignore
+        preds['data']['diffs'] = preds['data']['model'].diffs.detach().cpu().numpy()
+        preds['data']['diff_coeffs'] = preds['data']['model'].coeffs.detach().cpu().numpy()
         if not return_model:
             del preds['data']['model']
         return preds

alibi_detect/cd/spot_the_diff.py

@@ -143,7 +143,7 @@ def __init__(
             [kwargs.pop(k, None) for k in pop_kwargs]
             if dataset is None:
                 kwargs.update({'dataset': TFDataset})
-            self._detector = SpotTheDiffDriftTF(*args, **kwargs)  # type: ignore
+            self._detector = SpotTheDiffDriftTF(*args, **kwargs)
         else:
             if dataset is None:
                 kwargs.update({'dataset': TorchDataset})

alibi_detect/cd/tensorflow/classifier.py

@@ -185,7 +185,7 @@ def score(self, x: np.ndarray) -> Tuple[float, float, np.ndarray, np.ndarray,  #
                 self.train_kwargs['optimizer'] = optimizer.__class__.from_config(optimizer.get_config())
             train_args = [self.model, self.loss_fn, None]
             self.train_kwargs.update({'dataset': ds_tr})
-            trainer(*train_args, **self.train_kwargs)  # type: ignore
+            trainer(*train_args, **self.train_kwargs)
             preds = self.predict_fn(x_te, self.model)
             preds_oof_list.append(preds)
             idx_oof_list.append(idx_te)

alibi_detect/cd/tensorflow/learned_kernel.py

@@ -181,7 +181,7 @@ def score(self, x: Union[np.ndarray, list]) -> Tuple[float, float, float]:
 
         self.kernel = clone_model(self.original_kernel) if self.retrain_from_scratch else self.kernel
         train_args = [self.j_hat, (ds_ref_tr, ds_cur_tr)]
-        LearnedKernelDriftTF.trainer(*train_args, **self.train_kwargs)  # type: ignore
+        LearnedKernelDriftTF.trainer(*train_args, **self.train_kwargs)
 
         if isinstance(x_ref_te, np.ndarray) and isinstance(x_cur_te, np.ndarray):
             x_all = np.concatenate([x_ref_te, x_cur_te], axis=0)

alibi_detect/cd/tensorflow/lsdd.py

@@ -103,8 +103,8 @@ def _initialize_kernel(self, x_ref: tf.Tensor):
     def _configure_normalization(self, x_ref: tf.Tensor, eps: float = 1e-12):
         x_ref_means = tf.reduce_mean(x_ref, axis=0)
         x_ref_stds = tf.math.reduce_std(x_ref, axis=0)
-        self._normalize = lambda x: (x - x_ref_means) / (x_ref_stds + eps)  # type: ignore[assignment]
-        self._unnormalize = lambda x: (x * (x_ref_stds + eps) + x_ref_means).numpy()  # type: ignore[assignment]
+        self._normalize = lambda x: (x - x_ref_means) / (x_ref_stds + eps)
+        self._unnormalize = lambda x: (x * (x_ref_stds + eps) + x_ref_means).numpy()
 
     def _configure_kernel_centers(self, x_ref: tf.Tensor):
         "Set aside reference samples to act as kernel centers"

alibi_detect/cd/tensorflow/spot_the_diff.py

@@ -219,8 +219,8 @@ def predict(
         and the trained model.
         """
         preds = self._detector.predict(x, return_p_val, return_distance, return_probs, return_model=True)
-        preds['data']['diffs'] = preds['data']['model'].diffs.numpy()  # type: ignore
-        preds['data']['diff_coeffs'] = preds['data']['model'].coeffs.numpy()  # type: ignore
+        preds['data']['diffs'] = preds['data']['model'].diffs.numpy()
+        preds['data']['diff_coeffs'] = preds['data']['model'].coeffs.numpy()
         if not return_model:
             del preds['data']['model']
         return preds

alibi_detect/cd/utils.py

@@ -92,7 +92,7 @@ def new_model_fn(x: Union[np.ndarray, list]) -> np.ndarray:
         # shuffle
         n_x = len(x)
         perm = np.random.permutation(n_x)
-        x = x[perm] if is_np else [x[i] for i in perm]  # type: ignore[call-overload]
+        x = x[perm] if is_np else [x[i] for i in perm]
         # add extras if necessary
         final_batch_size = n_x % batch_size
         if final_batch_size != 0:

alibi_detect/od/llr.py

@@ -189,12 +189,12 @@ def fit(self,
             # train semantic model
             args = [self.dist_s, loss_fn, X]
             kwargs.update({'y_train': y, 'optimizer': optimizer_s})
-            trainer(*args, **kwargs)  # type: ignore[arg-type]
+            trainer(*args, **kwargs)
 
             # train background model
             args = [self.dist_b, loss_fn, X_back]
             kwargs.update({'y_train': y_back, 'optimizer': optimizer_b})
-            trainer(*args, **kwargs)  # type: ignore[arg-type]
+            trainer(*args, **kwargs)
 
     def infer_threshold(self,
                         X: np.ndarray,

alibi_detect/od/pytorch/base.py

@@ -232,7 +232,7 @@ def predict(self, x: torch.Tensor) -> TorchOutlierDetectorOutput:
         ValueError
             Raised if the detector is not fit on reference data.
         """
-        self.check_fitted()  # type: ignore
+        self.check_fitted()
         raw_scores = self.score(x)
         scores = self._ensembler(raw_scores)
 

alibi_detect/saving/_tensorflow/loading.py

@@ -275,7 +275,7 @@ def load_detector_legacy(filepath: Union[str, os.PathLike], suffix: str, **kwarg
         model = load_model(model_dir, custom_objects=custom_objects)
         detector = init_ad_md(state_dict, md, model)
     elif detector_name == 'OutlierProphet':
-        detector = init_od_prophet(state_dict)  # type: ignore[assignment]
+        detector = init_od_prophet(state_dict)
     elif detector_name == 'SpectralResidual':
         detector = init_od_sr(state_dict)  # type: ignore[assignment]
     elif detector_name == 'OutlierSeq2Seq':
@@ -291,7 +291,7 @@ def load_detector_legacy(filepath: Union[str, os.PathLike], suffix: str, **kwarg
             logger.warning('No model found in {}, setting `model` to `None`.'.format(model_dir))
             model = None
         if detector_name == 'KSDrift':
-            load_fn = init_cd_ksdrift  # type: ignore[assignment]
+            load_fn = init_cd_ksdrift
         elif detector_name == 'MMDDriftTF':
             load_fn = init_cd_mmddrift  # type: ignore[assignment]
         elif detector_name == 'ChiSquareDrift':

alibi_detect/saving/saving.py

@@ -138,7 +138,7 @@ def _save_detector_config(detector: ConfigurableDetector,
 
     # Get the detector config (with artefacts still within it)
     if hasattr(detector, 'get_config'):
-        cfg = detector.get_config()  # type: ignore[union-attr]  # TODO - remove once all detectors have get_config
+        cfg = detector.get_config()  # TODO - remove once all detectors have get_config
         cfg = validate_config(cfg, resolved=True)
     else:
         raise NotImplementedError(f'{detector_name} does not yet support config.toml based saving.')
@@ -492,7 +492,7 @@ def _save_kernel_config(kernel: Callable,
     # if a DeepKernel
     if hasattr(kernel, 'proj'):
         if hasattr(kernel, 'get_config'):
-            cfg_kernel = kernel.get_config()  # type: ignore[attr-defined]
+            cfg_kernel = kernel.get_config()
         else:
             raise AttributeError("The detector's `kernel` must have a .get_config() method for it to be saved.")
         # Serialize the kernels (if needed)
@@ -511,7 +511,7 @@ def _save_kernel_config(kernel: Callable,
         else:  # if an object
             kernel_class = kernel.__class__
             if hasattr(kernel, 'get_config'):
-                cfg_kernel = kernel.get_config()  # type: ignore[attr-defined]
+                cfg_kernel = kernel.get_config()
                 cfg_kernel['init_sigma_fn'], _ = _serialize_object(cfg_kernel['init_sigma_fn'], base_path,
                                                                    local_path.joinpath('init_sigma_fn'))
             else: