[DLMED] update according to comments

monai/handlers/mean_dice.py

@@ -48,9 +48,8 @@ def __init__(
                 a combination of argmax and to_onehot. Defaults to False.
             sigmoid: whether to add sigmoid function to the output prediction before computing Dice.
                 Defaults to False.
-            other_act: if don't want to use `sigmoid`, use other callable function to execute
-                other activation layers, Defaults to ``None``. for example:
-                `other_act = lambda x: torch.tanh(x)`.
+            other_act: callable function to replace `sigmoid` as activation layer if needed, Defaults to ``None``.
+                for example: `other_act = torch.tanh`.
             logit_thresh: the threshold value to round value to 0.0 and 1.0. Defaults to None (no thresholding).
             output_transform: transform the ignite.engine.state.output into [y_pred, y] pair.
             device: device specification in case of distributed computation usage.

monai/handlers/roc_auc.py

@@ -27,9 +27,8 @@ class ROCAUC(Metric):  # type: ignore # incorrectly typed due to optional_import
     Args:
         to_onehot_y: whether to convert `y` into the one-hot format. Defaults to False.
         softmax: whether to add softmax function to `y_pred` before computation. Defaults to False.
-        other_act: if don't want to use `softmax`, use other callable function to execute
-            other activation layers, Defaults to ``None``. for example:
-            `other_act = lambda x: torch.log_softmax(x)`.
+        other_act: callable function to replace `softmax` as activation layer if needed, Defaults to ``None``.
+            for example: `other_act = lambda x: torch.log_softmax(x)`.
         average: {``"macro"``, ``"weighted"``, ``"micro"``, ``"none"``}
             Type of averaging performed if not binary classification. Defaults to ``"macro"``.
 
@@ -48,7 +47,8 @@ class ROCAUC(Metric):  # type: ignore # incorrectly typed due to optional_import
         device: device specification in case of distributed computation usage.
 
     Note:
-        ROCAUC expects y to be comprised of 0's and 1's.  y_pred must either be probability estimates or confidence values.
+        ROCAUC expects y to be comprised of 0's and 1's.
+        y_pred must either be probability estimates or confidence values.
 
     """
 

monai/losses/dice.py

@@ -56,7 +56,7 @@ def __init__(
             softmax: if True, apply a softmax function to the prediction.
             other_act: if don't want to use `sigmoid` or `softmax`, use other callable function to execute
                 other activation layers, Defaults to ``None``. for example:
-                `other_act = lambda x: torch.tanh(x)`.
+                `other_act = torch.tanh`.
             squared_pred: use squared versions of targets and predictions in the denominator or not.
             jaccard: compute Jaccard Index (soft IoU) instead of dice or not.
             reduction: {``"none"``, ``"mean"``, ``"sum"``}
@@ -69,10 +69,12 @@ def __init__(
         Raises:
             ValueError: reduction={reduction} is invalid. Valid options are: none, mean or sum.
             ValueError: can only enable 1 of sigmoid, softmax and other_act.
+            ValueError: other_act must be a Callable function.
 
         """
         super().__init__(reduction=LossReduction(reduction).value)
-
+        if other_act is not None and not callable(other_act):
+            raise ValueError("other_act must be a Callable function.")
         if int(sigmoid) + int(softmax) + int(other_act is not None) > 1:
             raise ValueError("can only enable 1 of sigmoid, softmax and other_act.")
         self.include_background = include_background
@@ -105,8 +107,6 @@ def forward(self, input: torch.Tensor, target: torch.Tensor, smooth: float = 1e-
                 input = torch.softmax(input, 1)
 
         if self.other_act is not None:
-            if not callable(self.other_act):
-                raise ValueError("other_act must be a Callable function.")
             input = self.other_act(input)
 
         if self.to_onehot_y:
@@ -220,7 +220,7 @@ def __init__(
             softmax: If True, apply a softmax function to the prediction.
             other_act: if don't want to use `sigmoid` or `softmax`, use other callable function to execute
                 other activation layers, Defaults to ``None``. for example:
-                `other_act = lambda x: torch.tanh(x)`.
+                `other_act = torch.tanh`.
             squared_pred: use squared versions of targets and predictions in the denominator or not.
             w_type: {``"square"``, ``"simple"``, ``"uniform"``}
                 Type of function to transform ground truth volume to a weight factor. Defaults to ``"square"``.
@@ -237,7 +237,8 @@ def __init__(
 
         """
         super().__init__(reduction=LossReduction(reduction).value)
-
+        if other_act is not None and not callable(other_act):
+            raise ValueError("other_act must be a Callable function.")
         if int(sigmoid) + int(softmax) + int(other_act is not None) > 1:
             raise ValueError("can only enable 1 of sigmoid, softmax and other_act.")
         self.include_background = include_background
@@ -274,8 +275,6 @@ def forward(self, input: torch.Tensor, target: torch.Tensor, smooth: float = 1e-
                 input = torch.softmax(input, 1)
 
         if self.other_act is not None:
-            if not callable(self.other_act):
-                raise ValueError("other_act must be a Callable function.")
             input = self.other_act(input)
 
         if self.to_onehot_y:

monai/losses/tversky.py

@@ -51,7 +51,7 @@ def __init__(
             softmax: If True, apply a softmax function to the prediction.
             other_act: if don't want to use `sigmoid` or `softmax`, use other callable function to execute
                 other activation layers, Defaults to ``None``. for example:
-                `other_act = lambda x: torch.tanh(x)`.
+                `other_act = torch.tanh`.
             alpha: weight of false positives
             beta: weight of false negatives
             reduction: {``"none"``, ``"mean"``, ``"sum"``}
@@ -67,7 +67,8 @@ def __init__(
         """
 
         super().__init__(reduction=LossReduction(reduction).value)
-
+        if other_act is not None and not callable(other_act):
+            raise ValueError("other_act must be a Callable function.")
         if int(sigmoid) + int(softmax) + int(other_act is not None) > 1:
             raise ValueError("can only enable 1 of sigmoid, softmax and other_act.")
         self.include_background = include_background
@@ -100,8 +101,6 @@ def forward(self, input: torch.Tensor, target: torch.Tensor, smooth: float = 1e-
                 input = torch.softmax(input, 1)
 
         if self.other_act is not None:
-            if not callable(self.other_act):
-                raise ValueError("other_act must be a Callable function.")
             input = self.other_act(input)
 
         if self.to_onehot_y:

monai/metrics/meandice.py

@@ -36,12 +36,12 @@ class DiceMetric:
         mutually_exclusive: if True, `y_pred` will be converted into a binary matrix using
             a combination of argmax and to_onehot.  Defaults to False.
         sigmoid: whether to add sigmoid function to y_pred before computation. Defaults to False.
-        other_act: if don't want to use `sigmoid`, use other callable function to execute
-            other activation layers, Defaults to ``None``. for example:
-            `other_act = lambda x: torch.tanh(x)`.
+        other_act: callable function to replace `sigmoid` as activation layer if needed, Defaults to ``None``.
+            for example: `other_act = torch.tanh`.
         logit_thresh: the threshold value used to convert (for example, after sigmoid if `sigmoid=True`)
             `y_pred` into a binary matrix. Defaults to 0.5.
-        reduction: {``"none"``, ``"mean"``, ``"sum"``, ``"mean_batch"``, ``"sum_batch"``, ``"mean_channel"``, ``"sum_channel"``}
+        reduction: {``"none"``, ``"mean"``, ``"sum"``, ``"mean_batch"``, ``"sum_batch"``,
+            ``"mean_channel"``, ``"sum_channel"``}
             Define the mode to reduce computation result of 1 batch data. Defaults to ``"mean"``.
 
     """
@@ -151,9 +151,8 @@ def compute_meandice(
         mutually_exclusive: if True, `y_pred` will be converted into a binary matrix using
             a combination of argmax and to_onehot.  Defaults to False.
         sigmoid: whether to add sigmoid function to y_pred before computation. Defaults to False.
-        other_act: if don't want to use `sigmoid`, use other callable function to execute
-            other activation layers, Defaults to ``None``. for example:
-            `other_act = lambda x: torch.tanh(x)`.
+        other_act: callable function to replace `sigmoid` as activation layer if needed, Defaults to ``None``.
+            for example: `other_act = torch.tanh`.
         logit_thresh: the threshold value used to convert (for example, after sigmoid if `sigmoid=True`)
             `y_pred` into a binary matrix. Defaults to 0.5.
 

monai/metrics/rocauc.py

@@ -71,9 +71,8 @@ def compute_roc_auc(
             example shape: [16, 1] will be converted into [16, 2] (where `2` is inferred from `y_pred`).
         to_onehot_y: whether to convert `y` into the one-hot format. Defaults to False.
         softmax: whether to add softmax function to `y_pred` before computation. Defaults to False.
-        other_act: if don't want to use `softmax`, use other callable function to execute
-            other activation layers, Defaults to ``None``. for example:
-            `other_act = lambda x: torch.log_softmax(x)`.
+        other_act: callable function to replace `softmax` as activation layer if needed, Defaults to ``None``.
+            for example: `other_act = lambda x: torch.log_softmax(x)`.
         average: {``"macro"``, ``"weighted"``, ``"micro"``, ``"none"``}
             Type of averaging performed if not binary classification.
             Defaults to ``"macro"``.

tests/test_compute_meandice.py

@@ -143,7 +143,7 @@
         "to_onehot_y": False,
         "mutually_exclusive": False,
         "logit_thresh": 0.0,
-        "other_act": lambda x: torch.tanh(x),
+        "other_act": torch.tanh,
     },
     [[0.8]],
 ]

tests/test_dice_loss.py

@@ -109,7 +109,7 @@
         0.470451,
     ],
     [  # shape: (2, 1, 2, 2), (2, 1, 2, 2)
-        {"include_background": True, "other_act": lambda x: torch.tanh(x)},
+        {"include_background": True, "other_act": torch.tanh},
         {
             "input": torch.tensor([[[[1.0, -1.0], [-1.0, 1.0]]], [[[1.0, -1.0], [-1.0, 1.0]]]]),
             "target": torch.tensor([[[[1.0, 1.0], [1.0, 1.0]]], [[[1.0, 0.0], [1.0, 0.0]]]]),

tests/test_generalized_dice_loss.py

@@ -109,7 +109,7 @@
         0.250023,
     ],
     [  # shape: (2, 1, 2, 2), (2, 1, 2, 2)
-        {"include_background": True, "other_act": lambda x: torch.tanh(x)},
+        {"include_background": True, "other_act": torch.tanh},
         {
             "input": torch.tensor([[[[1.0, -1.0], [-1.0, 1.0]]], [[[1.0, -1.0], [-1.0, 1.0]]]]),
             "target": torch.tensor([[[[1.0, 1.0], [1.0, 1.0]]], [[[1.0, 0.0], [1.0, 0.0]]]]),

tests/test_tversky_loss.py

@@ -100,7 +100,7 @@
         0.247366,
     ],
     [  # shape: (2, 1, 2, 2), (2, 1, 2, 2)
-        {"include_background": True, "other_act": lambda x: torch.tanh(x)},
+        {"include_background": True, "other_act": torch.tanh},
         {
             "input": torch.tensor([[[[1.0, -1.0], [-1.0, 1.0]]], [[[1.0, -1.0], [-1.0, 1.0]]]]),
             "target": torch.tensor([[[[1.0, 1.0], [1.0, 1.0]]], [[[1.0, 0.0], [1.0, 0.0]]]]),