Simplify annotations as python>=3.9

Starting from python 3.9, we can
- replace Tuple, Dict, List, Type from typing by buitins tuple, dict, list, type
- replace typing.Callable by collections.abc.Callable
- replace typing.Sequence by collections.abc.Sequence

See https://peps.python.org/pep-0585/

docs/generate_nb_index.py

@@ -4,7 +4,6 @@
 import os
 import re
 import urllib.parse
-from typing import List, Tuple
 
 NOTEBOOKS_LIST_PLACEHOLDER = "[[notebooks-list]]"
 NOTEBOOKS_PAGE_TEMPLATE_RELATIVE_PATH = "tutorials.template.md"
@@ -25,15 +24,15 @@
 
 def extract_notebook_title_n_description(
     notebook_filepath: str,
-) -> Tuple[str, List[str]]:
+) -> tuple[str, list[str]]:
     # load notebook
     with open(notebook_filepath, "rt", encoding="utf-8") as f:
         notebook = json.load(f)
 
     # find title + description: from first cell,  h1 title + remaining text.
     # or title from filename else
     title = ""
-    description_lines: List[str] = []
+    description_lines: list[str] = []
     cell = notebook["cells"][0]
     if cell["cell_type"] == "markdown":
         firstline = cell["source"][0].strip()
@@ -51,7 +50,7 @@ def extract_notebook_title_n_description(
     return title, description_lines
 
 
-def filter_tags_from_description(description_lines: List[str], html_tag_to_remove: str) -> List[str]:
+def filter_tags_from_description(description_lines: list[str], html_tag_to_remove: str) -> list[str]:
     description = "".join(description_lines)
     # opening/closing tags
     opening_tag = html_tag_to_remove
@@ -147,7 +146,7 @@ def get_binder_link(
     return link
 
 
-def get_repo_n_branches_for_binder_n_github_links() -> Tuple[bool, str, str, str, str, str, bool]:
+def get_repo_n_branches_for_binder_n_github_links() -> tuple[bool, str, str, str, str, str, bool]:
     # repos + branches to use for binder environment and notebooks content.
     creating_links = True
     use_nbgitpuller = False

src/decomon/backward_layers/activations.py

@@ -1,5 +1,6 @@
 import warnings
-from typing import Any, Callable, Dict, List, Optional, Union
+from collections.abc import Callable
+from typing import Any, Optional, Union
 
 import keras.ops as K
 import numpy as np
@@ -39,7 +40,7 @@
 
 
 def backward_relu(
-    inputs: List[Tensor],
+    inputs: list[Tensor],
     dc_decomp: bool = False,
     perturbation_domain: Optional[PerturbationDomain] = None,
     alpha: float = 0.0,
@@ -48,7 +49,7 @@ def backward_relu(
     slope: Union[str, Slope] = Slope.V_SLOPE,
     mode: Union[str, ForwardMode] = ForwardMode.HYBRID,
     **kwargs: Any,
-) -> List[Tensor]:
+) -> list[Tensor]:
     """Backward  LiRPA of relu
 
     Args:
@@ -87,13 +88,13 @@ def backward_relu(
 
 
 def backward_sigmoid(
-    inputs: List[Tensor],
+    inputs: list[Tensor],
     dc_decomp: bool = False,
     perturbation_domain: Optional[PerturbationDomain] = None,
     slope: Union[str, Slope] = Slope.V_SLOPE,
     mode: Union[str, ForwardMode] = ForwardMode.HYBRID,
     **kwargs: Any,
-) -> List[Tensor]:
+) -> list[Tensor]:
     """Backward  LiRPA of sigmoid
 
     Args:
@@ -118,13 +119,13 @@ def backward_sigmoid(
 
 
 def backward_tanh(
-    inputs: List[Tensor],
+    inputs: list[Tensor],
     dc_decomp: bool = False,
     perturbation_domain: Optional[PerturbationDomain] = None,
     slope: Union[str, Slope] = Slope.V_SLOPE,
     mode: Union[str, ForwardMode] = ForwardMode.HYBRID,
     **kwargs: Any,
-) -> List[Tensor]:
+) -> list[Tensor]:
     """Backward  LiRPA of tanh
 
     Args:
@@ -149,13 +150,13 @@ def backward_tanh(
 
 
 def backward_hard_sigmoid(
-    inputs: List[Tensor],
+    inputs: list[Tensor],
     dc_decomp: bool = False,
     perturbation_domain: Optional[PerturbationDomain] = None,
     slope: Union[str, Slope] = Slope.V_SLOPE,
     mode: Union[str, ForwardMode] = ForwardMode.HYBRID,
     **kwargs: Any,
-) -> List[Tensor]:
+) -> list[Tensor]:
     """Backward  LiRPA of hard sigmoid
 
     Args:
@@ -178,13 +179,13 @@ def backward_hard_sigmoid(
 
 
 def backward_elu(
-    inputs: List[Tensor],
+    inputs: list[Tensor],
     dc_decomp: bool = False,
     perturbation_domain: Optional[PerturbationDomain] = None,
     slope: Union[str, Slope] = Slope.V_SLOPE,
     mode: Union[str, ForwardMode] = ForwardMode.HYBRID,
     **kwargs: Any,
-) -> List[Tensor]:
+) -> list[Tensor]:
     """Backward  LiRPA of Exponential Linear Unit
 
     Args:
@@ -208,13 +209,13 @@ def backward_elu(
 
 
 def backward_selu(
-    inputs: List[Tensor],
+    inputs: list[Tensor],
     dc_decomp: bool = False,
     perturbation_domain: Optional[PerturbationDomain] = None,
     slope: Union[str, Slope] = Slope.V_SLOPE,
     mode: Union[str, ForwardMode] = ForwardMode.HYBRID,
     **kwargs: Any,
-) -> List[Tensor]:
+) -> list[Tensor]:
     """Backward LiRPA of Scaled Exponential Linear Unit (SELU)
 
     Args:
@@ -238,13 +239,13 @@ def backward_selu(
 
 
 def backward_linear(
-    inputs: List[Tensor],
+    inputs: list[Tensor],
     dc_decomp: bool = False,
     perturbation_domain: Optional[PerturbationDomain] = None,
     slope: Union[str, Slope] = Slope.V_SLOPE,
     mode: Union[str, ForwardMode] = ForwardMode.HYBRID,
     **kwargs: Any,
-) -> List[Tensor]:
+) -> list[Tensor]:
     """Backward LiRPA of linear
 
     Args:
@@ -265,13 +266,13 @@ def backward_linear(
 
 
 def backward_exponential(
-    inputs: List[Tensor],
+    inputs: list[Tensor],
     dc_decomp: bool = False,
     perturbation_domain: Optional[PerturbationDomain] = None,
     slope: Union[str, Slope] = Slope.V_SLOPE,
     mode: Union[str, ForwardMode] = ForwardMode.HYBRID,
     **kwargs: Any,
-) -> List[Tensor]:
+) -> list[Tensor]:
     """Backward LiRPAof exponential
 
     Args:
@@ -294,13 +295,13 @@ def backward_exponential(
 
 
 def backward_softplus(
-    inputs: List[Tensor],
+    inputs: list[Tensor],
     dc_decomp: bool = False,
     perturbation_domain: Optional[PerturbationDomain] = None,
     slope: Union[str, Slope] = Slope.V_SLOPE,
     mode: Union[str, ForwardMode] = ForwardMode.HYBRID,
     **kwargs: Any,
-) -> List[Tensor]:
+) -> list[Tensor]:
     """Backward LiRPA of softplus
 
     Args:
@@ -325,13 +326,13 @@ def backward_softplus(
 
 
 def backward_softsign(
-    inputs: List[Tensor],
+    inputs: list[Tensor],
     dc_decomp: bool = False,
     perturbation_domain: Optional[PerturbationDomain] = None,
     slope: Union[str, Slope] = Slope.V_SLOPE,
     mode: Union[str, ForwardMode] = ForwardMode.HYBRID,
     **kwargs: Any,
-) -> List[Tensor]:
+) -> list[Tensor]:
     """Backward LiRPA of softsign
 
     Args:
@@ -359,7 +360,7 @@ def backward_softsign(
 
 
 def backward_softsign_(
-    inputs: List[Tensor],
+    inputs: list[Tensor],
     w_u_out: Tensor,
     b_u_out: Tensor,
     w_l_out: Tensor,
@@ -368,7 +369,7 @@ def backward_softsign_(
     mode: Union[str, ForwardMode] = ForwardMode.HYBRID,
     slope: Union[str, Slope] = Slope.V_SLOPE,
     **kwargs: Any,
-) -> List[Tensor]:
+) -> list[Tensor]:
     if perturbation_domain is None:
         perturbation_domain = BoxDomain()
     w_u_0, b_u_0, w_l_0, b_l_0 = get_linear_hull_s_shape(
@@ -391,14 +392,14 @@ def backward_softsign_(
 
 
 def backward_softmax(
-    inputs: List[Tensor],
+    inputs: list[Tensor],
     dc_decomp: bool = False,
     perturbation_domain: Optional[PerturbationDomain] = None,
     slope: Union[str, Slope] = Slope.V_SLOPE,
     mode: Union[str, ForwardMode] = ForwardMode.HYBRID,
     axis: int = -1,
     **kwargs: Any,
-) -> List[Tensor]:
+) -> list[Tensor]:
     """Backward LiRPA of softmax
 
     Args:
@@ -422,7 +423,7 @@ def backward_softmax(
     raise NotImplementedError()
 
 
-def deserialize(name: str) -> Callable[..., List[Tensor]]:
+def deserialize(name: str) -> Callable[..., list[Tensor]]:
     """Get the activation from name.
 
     Args:
@@ -457,7 +458,7 @@ def deserialize(name: str) -> Callable[..., List[Tensor]]:
     raise ValueError("Could not interpret " "activation function identifier:", name)
 
 
-def get(identifier: Any) -> Callable[..., List[Tensor]]:
+def get(identifier: Any) -> Callable[..., list[Tensor]]:
     """Get the `identifier` activation function.
 
     Args:

src/decomon/backward_layers/backward_layers.py

@@ -1,4 +1,4 @@
-from typing import Any, Dict, List, Optional, Tuple, Union
+from typing import Any, Optional, Union
 
 import keras
 import keras.ops as K
@@ -65,7 +65,7 @@ def __init__(
             )
         self.frozen_weights = False
 
-    def call(self, inputs: List[BackendTensor], **kwargs: Any) -> List[BackendTensor]:
+    def call(self, inputs: list[BackendTensor], **kwargs: Any) -> list[BackendTensor]:
         if len(inputs) == 0:
             inputs = self.layer.input
 
@@ -96,7 +96,7 @@ def call(self, inputs: List[BackendTensor], **kwargs: Any) -> List[BackendTensor
 
         return [w, b, w, b]
 
-    def build(self, input_shape: List[Tuple[Optional[int], ...]]) -> None:
+    def build(self, input_shape: list[tuple[Optional[int], ...]]) -> None:
         """
         Args:
             input_shape: list of input shape
@@ -166,7 +166,7 @@ def __init__(
             )
         self.frozen_weights = False
 
-    def get_affine_components(self, inputs: List[BackendTensor]) -> Tuple[BackendTensor, BackendTensor]:
+    def get_affine_components(self, inputs: list[BackendTensor]) -> tuple[BackendTensor, BackendTensor]:
         """Express the implicit affine matrix of the convolution layer.
 
         Conv is a linear operator but its affine component is implicit
@@ -210,7 +210,7 @@ def get_affine_components(self, inputs: List[BackendTensor]) -> Tuple[BackendTen
 
         return w_out_, b_out_
 
-    def call(self, inputs: List[BackendTensor], **kwargs: Any) -> List[BackendTensor]:
+    def call(self, inputs: list[BackendTensor], **kwargs: Any) -> list[BackendTensor]:
         weight_, bias_ = self.get_affine_components(inputs)
         return [weight_, bias_] * 2
 
@@ -252,13 +252,13 @@ def __init__(
         self.activation_name = layer.get_config()["activation"]
         self.slope = Slope(slope)
         self.finetune = finetune
-        self.finetune_param: List[keras.Variable] = []
+        self.finetune_param: list[keras.Variable] = []
         if self.finetune:
             self.frozen_alpha = False
-        self.grid_finetune: List[keras.Variable] = []
+        self.grid_finetune: list[keras.Variable] = []
         self.frozen_grid = False
 
-    def get_config(self) -> Dict[str, Any]:
+    def get_config(self) -> dict[str, Any]:
         config = super().get_config()
         config.update(
             {
@@ -268,7 +268,7 @@ def get_config(self) -> Dict[str, Any]:
         )
         return config
 
-    def build(self, input_shape: List[Tuple[Optional[int], ...]]) -> None:
+    def build(self, input_shape: list[tuple[Optional[int], ...]]) -> None:
         """
         Args:
             input_shape: list of input shape
@@ -371,7 +371,7 @@ def build(self, input_shape: List[Tuple[Optional[int], ...]]) -> None:
 
         self.built = True
 
-    def call(self, inputs: List[BackendTensor], **kwargs: Any) -> List[BackendTensor]:
+    def call(self, inputs: list[BackendTensor], **kwargs: Any) -> list[BackendTensor]:
         # infer the output dimension
         if self.activation_name != "linear":
             if self.finetune:
@@ -452,7 +452,7 @@ def __init__(
             **kwargs,
         )
 
-    def call(self, inputs: List[BackendTensor], **kwargs: Any) -> List[BackendTensor]:
+    def call(self, inputs: list[BackendTensor], **kwargs: Any) -> list[BackendTensor]:
         return get_identity_lirpa(inputs)
 
 
@@ -477,7 +477,7 @@ def __init__(
             **kwargs,
         )
 
-    def call(self, inputs: List[BackendTensor], **kwargs: Any) -> List[BackendTensor]:
+    def call(self, inputs: list[BackendTensor], **kwargs: Any) -> list[BackendTensor]:
         return get_identity_lirpa(inputs)
 
 
@@ -504,7 +504,7 @@ def __init__(
         self.dims = layer.dims
         self.op = layer.call
 
-    def call(self, inputs: List[BackendTensor], **kwargs: Any) -> List[BackendTensor]:
+    def call(self, inputs: list[BackendTensor], **kwargs: Any) -> list[BackendTensor]:
         w_u_out, b_u_out, w_l_out, b_l_out = get_identity_lirpa(inputs)
 
         # w_u_out (None, n_in, n_out)
@@ -545,7 +545,7 @@ def __init__(
             **kwargs,
         )
 
-    def call(self, inputs: List[BackendTensor], **kwargs: Any) -> List[BackendTensor]:
+    def call(self, inputs: list[BackendTensor], **kwargs: Any) -> list[BackendTensor]:
         return get_identity_lirpa(inputs)
 
 
@@ -573,7 +573,7 @@ def __init__(
         self.axis = self.layer.axis
         self.op_flat = Flatten()
 
-    def call(self, inputs: List[BackendTensor], **kwargs: Any) -> List[BackendTensor]:
+    def call(self, inputs: list[BackendTensor], **kwargs: Any) -> list[BackendTensor]:
         y = inputs[-1]
         n_out = int(np.prod(y.shape[1:]))
 
@@ -625,5 +625,5 @@ def __init__(
             **kwargs,
         )
 
-    def call(self, inputs: List[BackendTensor], **kwargs: Any) -> List[BackendTensor]:
+    def call(self, inputs: list[BackendTensor], **kwargs: Any) -> list[BackendTensor]:
         return get_identity_lirpa(inputs)