Merge pull request #57 from kaboroevich/typing_refactor

Typing refactor

Author: kaboroevich

stubs/pyDeepInsight/__init__.pyi renamed to pyDeepInsight/__init__.pyi

@@ -12,7 +12,7 @@
 import inspect
 import warnings
 
-from .utils import sparse_assignment
+from .utils._assignment import sparse_assignment
 
 
 class ImageTransformer:
@@ -25,13 +25,13 @@ class ImageTransformer:
     Attributes:
         _fe (ManifoldLearner): The feature extraction method used for
             dimensionality reduction. It must have a `fit_transform` method.
-        _dm (function): The discretization method used to assign data points to
+        _dm (Callable): The discretization method used to assign data points to
             pixel coordinates.
-        _pixels (tuple): The dimensions of the image matrix (height, width) to
+        _pixels (Tuple[int, int]): The dimensions of the image matrix (height, width) to
             which the data will be mapped.
-        _xrot (ndarray): The rotated coordinates of the data after
+        _xrot (np.ndarray): The rotated coordinates of the data after
             dimensionality reduction.
-        _coords (ndarray): The final pixel coordinates assigned to the data
+        _coords (np.ndarray): The final pixel coordinates assigned to the data
             points after discretization.
         DISCRETIZATION_OPTIONS (dict): A dictionary mapping discretization
             method names  to their corresponding class methods for pixel

stubs/pyDeepInsight/feature_selection.pyi renamed to pyDeepInsight/feature_selection.pyi

@@ -2,7 +2,7 @@ from typing import Any, Optional, Callable
 from typing_extensions import Protocol
 from numpy.typing import ArrayLike
 import numpy as np
-import torch
+from torch import Tensor
 
 
 class ManifoldLearner(Protocol):
@@ -85,10 +85,10 @@ class ImageTransformer:
     def _calculate_coords(self) -> None: ...
 
     def transform(self, X: np.ndarray, img_format: str = 'rgb',
-                  empty_value: int = 0) -> np.ndarray | torch.Tensor: ...
+                  empty_value: int = 0) -> np.ndarray | Tensor: ...
 
     def fit_transform(self, X: np.ndarray, **kwargs: Any
-                      ) -> np.ndarray | torch.Tensor: ...
+                      ) -> np.ndarray | Tensor: ...
 
     def inverse_transform(self, img: np.ndarray) -> np.ndarray: ...
 
@@ -104,7 +104,7 @@ class ImageTransformer:
     def _mat_to_rgb(mat: np.ndarray) -> np.ndarray: ...
 
     @staticmethod
-    def _mat_to_pytorch(mat: np.ndarray) -> torch.Tensor: ...
+    def _mat_to_pytorch(mat: np.ndarray) -> Tensor: ...
 
 
 class MRepImageTransformer:
@@ -134,15 +134,15 @@ class MRepImageTransformer:
                   empty_value: int = 0, collate: str = 'sample',
                   return_index: bool = True
                   ) -> (np.ndarray
-                        | torch.Tensor
+                        | Tensor
                         | tuple[np.ndarray, np.ndarray, np.ndarray]
-                        | tuple[torch.Tensor, np.ndarray, np.ndarray]): ...
+                        | tuple[Tensor, np.ndarray, np.ndarray]): ...
 
     def fit_transform(self, X: np.ndarray, **kwargs: Any
                       ) -> (np.ndarray
-                            | torch.Tensor
+                            | Tensor
                             | tuple[np.ndarray, np.ndarray, np.ndarray]
-                            | tuple[torch.Tensor, np.ndarray, np.ndarray]): ...
+                            | tuple[Tensor, np.ndarray, np.ndarray]): ...
 
     @staticmethod
     def prediction_reduction(input: np.ndarray, index: np.ndarray,

pyDeepInsight/image_transformer.py

@@ -9,7 +9,7 @@ def _sparsify_top_percentile(arr, p):
     smallest top-percentile values in each row.
 
     Args:
-        arr (ndarray): 2D array of shape (n_rows, n_cols) representing the
+        arr (np.ndarray): 2D array of shape (n_rows, n_cols) representing the
             input cost matrix.
         p (float): The fraction (0 < p <= 1) of the smallest values to retain
             per row.
@@ -34,7 +34,7 @@ def sparse_assignment(cost_matrix, p=0.1):
         smallest top-percentile values in each row.
 
         Args:
-            cost_matrix (ndarray): 2D array of shape (n_rows, n_cols)
+            cost_matrix (np.ndarray): 2D array of shape (n_rows, n_cols)
                 representing the cost matrix.
             p (float): The fraction (0 < p <= 1) of the smallest values to
                 retain per row during sparsification.

stubs/pyDeepInsight/image_transformer.pyi renamed to pyDeepInsight/image_transformer.pyi

@@ -0,0 +1,8 @@
+from typing import Tuple
+import numpy as np
+
+
+def _sparsify_top_percentile(arr: np.ndarray, p: float) -> np.ndarray: ...
+
+def sparse_assignment(cost_matrix: np.ndarray, p: float) -> Tuple[np.ndarray, np.ndarray]: ...
+

pyDeepInsight/py.typed

@@ -1,5 +1,5 @@
 import numpy as np
-import torch
+from torch import nn, Tensor
 
 
 def step_blur_kernel(kernel_size: int, amplification: float) -> np.ndarray: ...
@@ -9,14 +9,14 @@ def apply_blur_kernel(img: np.ndarray, kernel: np.ndarray) -> np.ndarray: ...
 def step_blur(img: np.ndarray, kernel_size: int,
               amplification: float) -> np.ndarray: ...
 
-class StepBlur2d(torch.nn.Module):
+class StepBlur2d(nn.Module):
 
     def_amp: float
-    kernel: torch.tensor
+    kernel: Tensor
 
-    def __init__(self, kernel_size: torch.tensor, amplification: float): ...
+    def __init__(self, kernel_size: Tensor, amplification: float): ...
 
-    def forward(self, input: torch.tensor) -> torch.tensor: ...
+    def forward(self, input: Tensor) -> Tensor: ...
 
     @staticmethod
     def step_kernel(kernel_size: int, amplification: float): ...
@@ -25,14 +25,14 @@ def imgaborfilt(image: np.ndarray, wavelength: float, orientation: float,
               SpatialFrequencyBandwidth: float,
               SpatialAspectRatio: float) -> np.ndarray: ...
 
-class GaborFilter2d(torch.nn.Module):
+class GaborFilter2d(nn.Module):
 
     def __init__(self, wavelength: float, orientation: float): ...
 
-    def forward(self, img: torch.tensor) -> torch.tensor: ...
+    def forward(self, img: Tensor) -> Tensor: ...
 
     @staticmethod
-    def pil_to_tensor(img: np.ndarray) -> torch.tensor: ...
+    def pil_to_tensor(img: np.ndarray) -> Tensor: ...
 
     @staticmethod
-    def tensor_to_pil(img: torch.tensor) -> np.ndarray: ...
+    def tensor_to_pil(img: Tensor) -> np.ndarray: ...