Fix target type (#133)

softwaremill · Oct 18, 2023 · 11c627e · 11c627e
1 parent 0fe28c6
commit 11c627e
Show file tree

Hide file tree

Showing 14 changed files with 36 additions and 205 deletions.
diff --git a/foxai/explainer/base_explainer.py b/foxai/explainer/base_explainer.py
@@ -4,7 +4,7 @@
 
 import torch
 
-from foxai.types import AttributionsType, ModelType, TargetType
+from foxai.types import AttributionsType, ModelType
 
 
 class Explainer(ABC):
@@ -16,7 +16,7 @@ def calculate_features(
         self,
         model: ModelType,
         input_data: torch.Tensor,  # TODO: add more generic way of passing model inputs # pylint: disable = (fixme)
-        pred_label_idx: Optional[TargetType] = None,
+        pred_label_idx: Optional[int] = None,
         **kwargs,
     ) -> AttributionsType:
         """Calculate features of given explainer.

diff --git a/foxai/explainer/computer_vision/algorithm/conductance.py b/foxai/explainer/computer_vision/algorithm/conductance.py
@@ -15,7 +15,7 @@
     get_last_conv_model_layer,
     preprocess_baselines,
 )
-from foxai.types import AttributionsType, LayerType, ModelType, TargetType
+from foxai.types import AttributionsType, LayerType, ModelType
 
 
 class LayerConductanceCVExplainer(Explainer):
@@ -49,7 +49,7 @@ def calculate_features(
         self,
         model: ModelType,
         input_data: torch.Tensor,
-        pred_label_idx: Optional[TargetType] = None,
+        pred_label_idx: Optional[int] = None,
         baselines: Optional[BaselineType] = None,
         additional_forward_args: Any = None,
         n_steps: int = 50,
@@ -72,24 +72,6 @@ def calculate_features(
                 this is usually the target class).
                 If the network returns a scalar value per example,
                 no target index is necessary.
-                For general 2D outputs, targets can be either:
-
-                - a single integer or a tensor containing a single
-                    integer, which is applied to all input examples
-
-                - a list of integers or a 1D tensor, with length matching
-                    the number of examples in inputs (dim 0). Each integer
-                    is applied as the target for the corresponding example.
-
-                For outputs with > 2 dimensions, targets can be either:
-
-                - A single tuple, which contains #output_dims - 1
-                    elements. This target index is applied to all examples.
-
-                - A list of tuples with length equal to the number of
-                    examples in inputs (dim 0), and each tuple containing
-                    #output_dims - 1 elements. Each tuple is applied as the
-                    target for the corresponding example.
                 Default: None
             baselines:
                 Baselines define the starting point from which integral

diff --git a/foxai/explainer/computer_vision/algorithm/deconv.py b/foxai/explainer/computer_vision/algorithm/deconv.py
@@ -12,7 +12,7 @@
 from foxai.array_utils import validate_result
 from foxai.explainer.base_explainer import Explainer
 from foxai.explainer.computer_vision.model_utils import modify_modules
-from foxai.types import AttributionsType, ModelType, TargetType
+from foxai.types import AttributionsType, ModelType
 
 
 class BaseDeconvolutionCVExplainer(Explainer):
@@ -37,7 +37,7 @@ def calculate_features(
         self,
         model: ModelType,
         input_data: torch.Tensor,
-        pred_label_idx: Optional[TargetType] = None,
+        pred_label_idx: Optional[int] = None,
         additional_forward_args: Any = None,
         **kwargs,
     ) -> AttributionsType:
@@ -54,24 +54,6 @@ def calculate_features(
                 this is usually the target class).
                 If the network returns a scalar value per example,
                 no target index is necessary.
-                For general 2D outputs, targets can be either:
-
-                - a single integer or a tensor containing a single
-                    integer, which is applied to all input examples
-
-                - a list of integers or a 1D tensor, with length matching
-                    the number of examples in inputs (dim 0). Each integer
-                    is applied as the target for the corresponding example.
-
-                For outputs with > 2 dimensions, targets can be either:
-
-                - A single tuple, which contains #output_dims - 1
-                    elements. This target index is applied to all examples.
-
-                - A list of tuples with length equal to the number of
-                    examples in inputs (dim 0), and each tuple containing
-                    #output_dims - 1 elements. Each tuple is applied as the
-                    target for the corresponding example.
                 Default: None
             additional_forward_args: If the forward function
                 requires additional arguments other than the inputs for

diff --git a/foxai/explainer/computer_vision/algorithm/deeplift.py b/foxai/explainer/computer_vision/algorithm/deeplift.py
@@ -23,7 +23,6 @@
     CustomAttributionFuncType,
     LayerType,
     ModelType,
-    TargetType,
 )
 
 
@@ -66,7 +65,7 @@ def calculate_features(
         self,
         model: ModelType,
         input_data: torch.Tensor,
-        pred_label_idx: Optional[TargetType] = None,
+        pred_label_idx: Optional[int] = None,
         baselines: Optional[BaselineType] = None,
         additional_forward_args: Any = None,
         custom_attribution_func: Optional[CustomAttributionFuncType] = None,

diff --git a/foxai/explainer/computer_vision/algorithm/deeplift_shap.py b/foxai/explainer/computer_vision/algorithm/deeplift_shap.py
@@ -23,7 +23,6 @@
     CustomAttributionFuncType,
     LayerType,
     ModelType,
-    TargetType,
 )
 
 
@@ -66,7 +65,7 @@ def calculate_features(
         self,
         model: ModelType,
         input_data: torch.Tensor,
-        pred_label_idx: Optional[TargetType] = None,
+        pred_label_idx: Optional[int] = None,
         baselines: Optional[BaselineType] = None,
         additional_forward_args: Any = None,
         custom_attribution_func: Optional[CustomAttributionFuncType] = None,

diff --git a/foxai/explainer/computer_vision/algorithm/gradcam.py b/foxai/explainer/computer_vision/algorithm/gradcam.py
@@ -21,7 +21,7 @@
     BaseObjectDetector,
 )
 from foxai.explainer.computer_vision.object_detection.types import ObjectDetectionOutput
-from foxai.types import AttributionsType, LayerType, ModelType, TargetType
+from foxai.types import AttributionsType, LayerType, ModelType
 
 
 class LayerBaseGradCAM:
@@ -168,7 +168,7 @@ def calculate_features(
         self,
         model: ModelType,
         input_data: torch.Tensor,
-        pred_label_idx: Optional[TargetType] = None,
+        pred_label_idx: Optional[int] = None,
         **kwargs,
     ) -> AttributionsType:
         """Generate features image with GradCAM algorithm explainer.
@@ -251,7 +251,7 @@ def calculate_features(
         self,
         model: ModelType,
         input_data: torch.Tensor,
-        pred_label_idx: Optional[TargetType] = None,
+        pred_label_idx: Optional[int] = None,
         additional_forward_args: Any = None,
         attribute_to_layer_input: bool = False,
         interpolate_mode: str = "nearest",
@@ -394,9 +394,7 @@ def calculate_features(
         self,
         model: ModelType,
         input_data: torch.Tensor,
-        pred_label_idx: Optional[
-            TargetType
-        ] = None,  # pylint: disable = (unused-argument)
+        pred_label_idx: Optional[int] = None,  # pylint: disable = (unused-argument)
         layer: Optional[LayerType] = None,
         **kwargs,
     ) -> AttributionsType:
@@ -418,24 +416,6 @@ def calculate_features(
                 this is usually the target class).
                 If the network returns a scalar value per example,
                 no target index is necessary.
-                For general 2D outputs, targets can be either:
-
-                - a single integer or a tensor containing a single
-                    integer, which is applied to all input examples
-
-                - a list of integers or a 1D tensor, with length matching
-                    the number of examples in inputs (dim 0). Each integer
-                    is applied as the target for the corresponding example.
-
-                For outputs with > 2 dimensions, targets can be either:
-
-                - A single tuple, which contains #output_dims - 1
-                    elements. This target index is applied to all examples.
-
-                - A list of tuples with length equal to the number of
-                    examples in inputs (dim 0), and each tuple containing
-                    #output_dims - 1 elements. Each tuple is applied as the
-                    target for the corresponding example.
                 Default: None
             layer: Layer for which attributions are computed.
                 If None provided, last convolutional layer from the model

diff --git a/foxai/explainer/computer_vision/algorithm/gradient_shap.py b/foxai/explainer/computer_vision/algorithm/gradient_shap.py
@@ -17,7 +17,7 @@
     get_last_conv_model_layer,
     preprocess_baselines,
 )
-from foxai.types import AttributionsType, LayerType, ModelType, StdevsType, TargetType
+from foxai.types import AttributionsType, LayerType, ModelType, StdevsType
 
 
 class BaseGradientSHAPCVExplainer(Explainer):
@@ -57,7 +57,7 @@ def calculate_features(
         self,
         model: ModelType,
         input_data: torch.Tensor,
-        pred_label_idx: Optional[TargetType] = None,
+        pred_label_idx: Optional[int] = None,
         baselines: Optional[BaselineType] = None,
         n_samples: int = 5,
         stdevs: StdevsType = 0.0,
@@ -79,25 +79,6 @@ def calculate_features(
                 this is usually the target class).
                 If the network returns a scalar value per example,
                 no target index is necessary.
-                For general 2D outputs, targets can be either:
-
-                - a single integer or a tensor containing a single
-                    integer, which is applied to all input examples
-
-                - a list of integers or a 1D tensor, with length matching
-                    the number of examples in inputs (dim 0). Each integer
-                    is applied as the target for the corresponding example.
-
-                For outputs with > 2 dimensions, targets can be either:
-
-                - A single tuple, which contains #output_dims - 1
-                    elements. This target index is applied to all examples.
-
-                - A list of tuples with length equal to the number of
-                    examples in inputs (dim 0), and each tuple containing
-                    #output_dims - 1 elements. Each tuple is applied as the
-                    target for the corresponding example.
-
                 Default: None
             baselines:
                 Baselines define the starting point from which expectation

diff --git a/foxai/explainer/computer_vision/algorithm/guided_backprop.py b/foxai/explainer/computer_vision/algorithm/guided_backprop.py
@@ -12,7 +12,7 @@
 from foxai.array_utils import validate_result
 from foxai.explainer.base_explainer import Explainer
 from foxai.explainer.computer_vision.model_utils import modify_modules
-from foxai.types import AttributionsType, ModelType, TargetType
+from foxai.types import AttributionsType, ModelType
 
 
 class BaseGuidedBackpropCVExplainer(Explainer):
@@ -34,7 +34,7 @@ def calculate_features(
         self,
         model: ModelType,
         input_data: torch.Tensor,
-        pred_label_idx: Optional[TargetType] = None,
+        pred_label_idx: Optional[int] = None,
         additional_forward_args: Any = None,
         **kwargs,
     ) -> AttributionsType:
@@ -57,25 +57,6 @@ def calculate_features(
                 this is usually the target class).
                 If the network returns a scalar value per example,
                 no target index is necessary.
-                For general 2D outputs, targets can be either:
-
-                - a single integer or a tensor containing a single
-                    integer, which is applied to all input examples
-
-                - a list of integers or a 1D tensor, with length matching
-                    the number of examples in inputs (dim 0). Each integer
-                    is applied as the target for the corresponding example.
-
-                For outputs with > 2 dimensions, targets can be either:
-
-                - A single tuple, which contains #output_dims - 1
-                    elements. This target index is applied to all examples.
-
-                - A list of tuples with length equal to the number of
-                    examples in inputs (dim 0), and each tuple containing
-                    #output_dims - 1 elements. Each tuple is applied as the
-                    target for the corresponding example.
-
                 Default: None
             additional_forward_args: If the forward function
                 requires additional arguments other than the inputs for

diff --git a/foxai/explainer/computer_vision/algorithm/integrated_gradients.py b/foxai/explainer/computer_vision/algorithm/integrated_gradients.py
@@ -17,7 +17,7 @@
     get_last_conv_model_layer,
     preprocess_baselines,
 )
-from foxai.types import AttributionsType, LayerType, ModelType, TargetType
+from foxai.types import AttributionsType, LayerType, ModelType
 
 
 class BaseIntegratedGradientsCVExplainer(Explainer):
@@ -61,7 +61,7 @@ def calculate_features(
         self,
         model: ModelType,
         input_data: torch.Tensor,
-        pred_label_idx: Optional[TargetType] = None,
+        pred_label_idx: Optional[int] = None,
         baselines: Optional[BaselineType] = None,
         additional_forward_args: Any = None,
         n_steps: int = 50,
@@ -84,25 +84,6 @@ def calculate_features(
                 this is usually the target class).
                 If the network returns a scalar value per example,
                 no target index is necessary.
-                For general 2D outputs, targets can be either:
-
-                - a single integer or a tensor containing a single
-                    integer, which is applied to all input examples
-
-                - a list of integers or a 1D tensor, with length matching
-                    the number of examples in inputs (dim 0). Each integer
-                    is applied as the target for the corresponding example.
-
-                For outputs with > 2 dimensions, targets can be either:
-
-                - A single tuple, which contains #output_dims - 1
-                    elements. This target index is applied to all examples.
-
-                - A list of tuples with length equal to the number of
-                    examples in inputs (dim 0), and each tuple containing
-                    #output_dims - 1 elements. Each tuple is applied as the
-                    target for the corresponding example.
-
                 Default: None
             baselines:
                 Baselines define the starting point from which integral

diff --git a/foxai/explainer/computer_vision/algorithm/noise_tunnel.py b/foxai/explainer/computer_vision/algorithm/noise_tunnel.py
@@ -12,7 +12,7 @@
 from foxai.array_utils import validate_result
 from foxai.explainer.base_explainer import Explainer
 from foxai.explainer.computer_vision.model_utils import get_last_conv_model_layer
-from foxai.types import AttributionsType, LayerType, ModelType, StdevsType, TargetType
+from foxai.types import AttributionsType, LayerType, ModelType, StdevsType
 
 
 class BaseNoiseTunnelCVExplainer(Explainer):
@@ -34,7 +34,7 @@ def calculate_features(
         self,
         model: ModelType,
         input_data: torch.Tensor,
-        pred_label_idx: Optional[TargetType] = None,
+        pred_label_idx: Optional[int] = None,
         nt_type: str = "smoothgrad",
         nt_samples: int = 5,
         nt_samples_batch_size: Optional[int] = None,
@@ -56,25 +56,6 @@ def calculate_features(
                 this is usually the target class).
                 If the network returns a scalar value per example,
                 no target index is necessary.
-                For general 2D outputs, targets can be either:
-
-                - a single integer or a tensor containing a single
-                    integer, which is applied to all input examples
-
-                - a list of integers or a 1D tensor, with length matching
-                    the number of examples in inputs (dim 0). Each integer
-                    is applied as the target for the corresponding example.
-
-                For outputs with > 2 dimensions, targets can be either:
-
-                - A single tuple, which contains #output_dims - 1
-                    elements. This target index is applied to all examples.
-
-                - A list of tuples with length equal to the number of
-                    examples in inputs (dim 0), and each tuple containing
-                    #output_dims - 1 elements. Each tuple is applied as the
-                    target for the corresponding example.
-
                 Default: None
             nt_type: Smoothing type of the attributions.
                 `smoothgrad`, `smoothgrad_sq` or `vargrad`