Make NNCF common tensor statistics code pass mypy checks (#2865)

### Changes

General: 
- Added the target code path `nncf/common/tensor_statistics` to
`.mypy.ini.`
- Fixed mypy errors caused by type inconsistencies.

Used `# type:ignore` in the following cases:

- has no attribute
- incompatible types that couldn't be resolved without substantial code
changes.

### Related tickets

Closes Issue #2493 

### Tests

To validate that the changes did not affect the codebase, some pytest
tests were run.

---------

Co-authored-by: Daniil Lyakhov <[email protected]>

.mypy.ini

@@ -1,5 +1,5 @@
 [mypy]
-files = nncf/common/sparsity, nncf/common/graph, nncf/common/accuracy_aware_training/, nncf/common/utils/
+files = nncf/common/sparsity, nncf/common/graph, nncf/common/accuracy_aware_training/, nncf/common/utils/, nncf/common/tensor_statistics
 follow_imports = silent
 strict = True
 

nncf/common/tensor_statistics/aggregator.py

@@ -20,8 +20,9 @@
 from nncf.common.logging.track_progress import track
 from nncf.common.tensor import NNCFTensor
 from nncf.common.tensor_statistics.statistic_point import StatisticPointsContainer
+from nncf.data.dataset import DataItem
 from nncf.data.dataset import Dataset
-from nncf.tensor import Tensor
+from nncf.data.dataset import ModelInput
 
 TensorType = TypeVar("TensorType")
 TModel = TypeVar("TModel")
@@ -36,7 +37,7 @@ class StatisticsAggregator(ABC):
     Base class for statistics collection.
     """
 
-    def __init__(self, dataset: Dataset):
+    def __init__(self, dataset: Dataset[DataItem, ModelInput]):
         self.dataset = dataset
         self.stat_subset_size = None
         self.statistic_points = StatisticPointsContainer()
@@ -65,12 +66,12 @@ def collect_statistics(self, model: TModel, graph: NNCFGraph) -> None:
         model_transformer = factory.ModelTransformerFactory.create(model)
         merged_statistics = self._get_merged_statistic_points(self.statistic_points, model, graph)
         transformation_layout = self._get_transformation_layout_extra_outputs(merged_statistics)
-        model_with_outputs = model_transformer.transform(transformation_layout)
+        model_with_outputs: TModel = model_transformer.transform(transformation_layout)
         engine = factory.EngineFactory.create(model_with_outputs)
 
         iterations_number = self._get_iterations_number()
         empty_statistics = True
-        for input_data in track(
+        for input_data in track(  # type: ignore
             islice(self.dataset.get_inference_data(), iterations_number),
             total=iterations_number,
             description="Statistics collection",
@@ -141,7 +142,7 @@ def _get_merged_statistic_points(
 
     @staticmethod
     @abstractmethod
-    def _process_outputs(outputs: Any) -> Dict[str, Tensor]:
+    def _process_outputs(outputs: Any) -> Dict[str, NNCFTensor]:
         """
         Post-process model outputs for the further statistics collection.
 

nncf/common/tensor_statistics/collectors.py

@@ -12,9 +12,10 @@
 from abc import ABC
 from abc import abstractmethod
 from collections import deque
-from typing import List, Optional, Tuple
+from typing import Any, Deque, Dict, List, Optional, Tuple, Union, cast
 
 import numpy as np
+from numpy.typing import NDArray
 
 from nncf.common.tensor import NNCFTensor
 from nncf.common.tensor import TensorType
@@ -39,7 +40,7 @@ def __init__(self, reduction_shape: Optional[ReductionAxes] = None, num_samples:
         self._num_samples = num_samples
 
     @property
-    def num_samples(self) -> int:
+    def num_samples(self) -> Union[int, None]:
         return self._num_samples
 
     def register_input(self, x: TensorType) -> TensorType:
@@ -49,38 +50,38 @@ def register_input(self, x: TensorType) -> TensorType:
         if self._num_samples is not None and self._collected_samples >= self._num_samples:
             return x
         if self._reduction_shape is None:
-            self._reduction_shape = tuple(range(len(x.shape)))
+            self._reduction_shape = tuple(range(len(cast(NNCFTensor, x).shape)))
         self._register_input(x)
         self._collected_samples += 1
         return x
 
     @abstractmethod
-    def _register_input(self, x: TensorType):
+    def _register_input(self, x: TensorType) -> None:
         pass
 
-    def get_statistics(self):
+    def get_statistics(self) -> None:
         """Returns collected statistics, if present."""
         if self._collected_samples == 0:
             raise StatisticsNotCollectedError()
         return self._get_statistics()
 
     @abstractmethod
-    def _get_statistics(self):
+    def _get_statistics(self) -> None:
         pass
 
-    def enable(self):
+    def enable(self) -> None:
         self._enabled = True
 
-    def disable(self):
+    def disable(self) -> None:
         self._enabled = False
 
-    def reset(self):
+    def reset(self) -> None:
         """Resets all the statistics in the collector."""
         self._collected_samples = 0
         self._reset()
 
     @abstractmethod
-    def _reset(self):
+    def _reset(self) -> None:
         pass
 
     def collected_samples(self) -> int:
@@ -102,9 +103,9 @@ def __init__(
         self, reduction_shape: Optional[ReductionAxes] = None, num_samples: int = None, window_size: int = None
     ):
         super().__init__(reduction_shape, num_samples)
-        self._samples = deque(maxlen=window_size)
+        self._samples: Deque[int] = deque(maxlen=window_size)
 
-    def _reset(self):
+    def _reset(self) -> None:
         self._samples.clear()
 
 
@@ -121,10 +122,10 @@ def __init__(self, use_abs_max: bool, reduction_shape: ReductionAxes, num_sample
 
     @staticmethod
     @abstractmethod
-    def _get_processor():
+    def _get_processor() -> Any:
         pass
 
-    def _register_input_common(self, x: NNCFTensor):
+    def _register_input_common(self, x: NNCFTensor) -> None:
         min_reduced = self._tensor_processor.reduce_min(x, self._reduction_shape)
         if self._use_abs_max:
             x = self._tensor_processor.abs(x)
@@ -140,7 +141,7 @@ def _register_input_common(self, x: NNCFTensor):
         else:
             self._max_values = self._tensor_processor.max(max_reduced, self._max_values)
 
-    def _reset(self):
+    def _reset(self) -> None:
         self._min_values = None
         self._max_values = None
 
@@ -164,15 +165,15 @@ def __init__(
         self._use_abs_max = use_abs_max
         self._tensor_processor = self._get_processor()
 
-        self._all_min_values = deque(maxlen=window_size)
-        self._all_max_values = deque(maxlen=window_size)
+        self._all_min_values: Deque[int] = deque(maxlen=window_size)
+        self._all_max_values: Deque[int] = deque(maxlen=window_size)
 
     @staticmethod
     @abstractmethod
-    def _get_processor():
+    def _get_processor() -> Any:
         pass
 
-    def _register_input_common(self, x: NNCFTensor):
+    def _register_input_common(self, x: NNCFTensor) -> None:
         min_reduced = self._tensor_processor.reduce_min(x, self._reduction_shape)
         if self._use_abs_max:
             x = self._tensor_processor.abs(x)
@@ -186,14 +187,14 @@ def _register_input_common(self, x: NNCFTensor):
             self._all_max_values.append(max_reduced)
 
     @abstractmethod
-    def _min_aggregate(self):
+    def _min_aggregate(self) -> None:
         pass
 
     @abstractmethod
-    def _max_aggregate(self):
+    def _max_aggregate(self) -> None:
         pass
 
-    def _reset(self):
+    def _reset(self) -> None:
         self._all_min_values.clear()
         self._all_max_values.clear()
 
@@ -217,13 +218,13 @@ def __init__(
         self._use_means_of_mins = use_means_of_mins
         self._use_means_of_maxs = use_means_of_maxs
 
-    def _min_aggregate(self):
+    def _min_aggregate(self) -> Any:
         stacked_min = self._tensor_processor.stack(self._all_min_values)
         if self._use_means_of_mins:
             return self._tensor_processor.mean(stacked_min, axis=0)
         return self._tensor_processor.reduce_min(stacked_min, axis=0)
 
-    def _max_aggregate(self):
+    def _max_aggregate(self) -> Any:
         stacked_max = self._tensor_processor.stack(self._all_max_values)
         if self._use_means_of_maxs:
             return self._tensor_processor.mean(stacked_max, axis=0)
@@ -235,11 +236,11 @@ class MeanMinMaxStatisticCollector(MinMaxOfflineStatisticCollectorBase):
     Collector aggregates mean of minimum values and mean of maximum values.
     """
 
-    def _min_aggregate(self):
+    def _min_aggregate(self) -> Any:
         stacked_min = self._tensor_processor.stack(self._all_min_values)
         return self._tensor_processor.mean(stacked_min, axis=0)
 
-    def _max_aggregate(self):
+    def _max_aggregate(self) -> Any:
         stacked_max = self._tensor_processor.stack(self._all_max_values)
         return self._tensor_processor.mean(stacked_max, axis=0)
 
@@ -259,30 +260,30 @@ def __init__(self, channel_axis: int, num_samples: Optional[int] = None, window_
         super().__init__(num_samples=num_samples)
         self._channel_axis = channel_axis
         self._tensor_processor = self._get_processor()
-        self._all_values = deque(maxlen=window_size)
-        self._all_shapes = deque(maxlen=window_size)
+        self._all_values: Deque[int] = deque(maxlen=window_size)
+        self._all_shapes: Deque[int] = deque(maxlen=window_size)
 
     @staticmethod
     @abstractmethod
-    def _get_processor():
+    def _get_processor() -> Any:
         pass
 
-    def _register_input_common(self, x: NNCFTensor):
+    def _register_input_common(self, x: NNCFTensor) -> None:
         if self._channel_axis == 0:
             self._all_values.append(self._tensor_processor.batch_mean(x))
         else:
             self._all_values.append(self._tensor_processor.mean_per_channel(x, self._channel_axis))
-        self._all_shapes.append(x.shape)
+        self._all_shapes.append(cast(int, x.shape))
 
-    def _reset(self):
+    def _reset(self) -> None:
         self._all_values.clear()
         self._all_shapes.clear()
 
-    def _mean_aggregate(self):
+    def _mean_aggregate(self) -> Any:
         all_values_stack = self._tensor_processor.stack(self._all_values)
         return self._tensor_processor.mean(all_values_stack, 0)
 
-    def _shape(self):
+    def _shape(self) -> Any:
         return self._all_shapes[0]
 
 
@@ -298,17 +299,17 @@ def __init__(self, num_samples: Optional[int] = None) -> None:
             the number of samples that will be processed.
         """
         super().__init__(num_samples=num_samples)
-        self._all_values = []
+        self._all_values: List[int] = []
 
     @staticmethod
     @abstractmethod
-    def _get_processor():
+    def _get_processor() -> Any:
         pass
 
-    def _register_input_common(self, x: NNCFTensor):
-        self._all_values.append(x.tensor)
+    def _register_input_common(self, x: NNCFTensor) -> None:
+        self._all_values.append(cast(int, x.tensor))
 
-    def _reset(self):
+    def _reset(self) -> None:
         self._all_values.clear()
 
 
@@ -317,8 +318,10 @@ class MedianMADStatisticCollector(OfflineTensorStatisticCollector):
     Collector estimates median and median absolute deviation (MAD).
     """
 
-    def _prepare_statistics(self):
-        per_channel_history = get_per_channel_history(self._samples, list(self._reduction_shape), discard_zeros=True)
+    def _prepare_statistics(self) -> Tuple[NDArray[Any], NDArray[Any]]:
+        per_channel_history = get_per_channel_history(
+            self._samples, cast(List[int], self._reduction_shape), discard_zeros=True
+        )
         per_channel_median = [np.median(channel_hist) for channel_hist in per_channel_history]
         per_channel_mad = []
         for idx, median in enumerate(per_channel_median):
@@ -343,8 +346,8 @@ def __init__(
         super().__init__(reduction_shape, num_samples, window_size)
         self._percentiles_to_collect = percentiles_to_collect
 
-    def _prepare_statistics(self):
-        per_channel_history = get_per_channel_history(self._samples, list(self._reduction_shape))
+    def _prepare_statistics(self) -> Dict[float, Any]:
+        per_channel_history = get_per_channel_history(self._samples, cast(List[int], self._reduction_shape))
         percentile_vs_values_dict = {}
         for pc in self._percentiles_to_collect:
             per_channel_percentiles = [np.percentile(channel_hist, pc) for channel_hist in per_channel_history]
@@ -366,10 +369,10 @@ def __init__(
         window_size: int = None,
     ):
         super().__init__(reduction_shape, num_samples, window_size)
-        self._all_pct_values = {}
+        self._all_pct_values: Dict[float, Any] = {}
         for pc in percentiles_to_collect:
             self._all_pct_values[pc] = deque(maxlen=window_size)
 
-    def _reset(self):
+    def _reset(self) -> None:
         for _, val in self._all_pct_values.items():
             val.clear()

nncf/common/tensor_statistics/reduction.py

@@ -9,10 +9,10 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from collections import deque
-from typing import List, Tuple
+from typing import Any, Deque, List, Tuple
 
 import numpy as np
+from numpy.typing import NDArray
 
 
 def get_channel_count_and_dim_idx(scale_shape: List[int]) -> Tuple[int, int]:
@@ -25,7 +25,7 @@ def get_channel_count_and_dim_idx(scale_shape: List[int]) -> Tuple[int, int]:
     return channel_count, channel_dim_idx
 
 
-def split_into_channels(input_: np.ndarray, scale_shape: List[int]) -> List[np.ndarray]:
+def split_into_channels(input_: NDArray[Any], scale_shape: List[int]) -> List[NDArray[Any]]:
     channel_count, channel_dim_idx = get_channel_count_and_dim_idx(scale_shape)
     channel_first_tensor = np.moveaxis(input_, channel_dim_idx, 0)
     if channel_count == 1:
@@ -37,9 +37,11 @@ def split_into_channels(input_: np.ndarray, scale_shape: List[int]) -> List[np.n
     return ret_list
 
 
-def get_per_channel_history(raw_input_history: deque, scale_shape: List[int], discard_zeros=False) -> List:
+def get_per_channel_history(
+    raw_input_history: Deque[Any], scale_shape: List[int], discard_zeros: bool = False
+) -> List[Any]:
     channel_count, _ = get_channel_count_and_dim_idx(scale_shape)
-    per_channel_history = [None for i in range(channel_count)]
+    per_channel_history: List[Any] = [None for i in range(channel_count)]
     for _ in range(len(raw_input_history)):
         entry = raw_input_history.popleft()
         split = split_into_channels(entry, scale_shape)
@@ -59,7 +61,7 @@ def get_per_channel_history(raw_input_history: deque, scale_shape: List[int], di
     return per_channel_history
 
 
-def np_percentile_reduce_like(input_: np.array, ref_tensor_shape: Tuple[int], q: float) -> np.array:
+def np_percentile_reduce_like(input_: NDArray[Any], ref_tensor_shape: Tuple[int], q: float) -> NDArray[Any]:
     numel = np.prod(ref_tensor_shape)
     if numel == 1:
         return np.array([np.percentile(input_, q)])