added paddle.tensor.math.exp

ivy/data_classes/array/experimental/norms.py

@@ -51,6 +51,7 @@ def batch_norm(
         training: bool = False,
         eps: float = 1e-5,
         momentum: float = 1e-1,
+        data_format: str = "NSC",
         out: Optional[Tuple[ivy.Array, ivy.Array, ivy.Array]] = None,
     ) -> Tuple[ivy.Array, ivy.Array, ivy.Array]:
         """
@@ -61,39 +62,35 @@ def batch_norm(
         Parameters
         ----------
         self
-            Input array of shape (N, *S, C), where N is the batch dimension,
+            Input array of default shape (N, *S, C), where N is the batch dimension,
             *S corresponds to any number of spatial dimensions and
              C corresponds to the channel dimension.
+        training
+            If true, calculate and use the mean and variance of `x`. Otherwise, use the
+            provided `mean` and `variance`.
         mean
-            Array used for input's normalization. If ``training=True``
-            then it must be one dimensional with size equal to the size of
-            channel dimension C. If ``training=False`` then it can be of any
-            shape broadcastble to the input shape.
+            Mean array used for input's normalization. It can be of any shape
+            braodcastable to (N,*S,C).
         variance
-            Array for the input's normalization. If ``training=True``
-            then it must be one dimensional with size equal to the size of
-            channel dimension C. If ``training=False`` then it can be of any shape
-            broadcastble to the input shape.
+            Variance array used for input's normalization. It can be of any shape
+            braodcastable to (N,*S,C).
         offset
             An offset array. If present, will be added to the normalized input.
-            If ``training=True`` then it must be one dimensional with size equal
-            to the size of channel dimension C. If ``training=False`` then it can
-            be of any shape broadcastble to the input shape.
+            It can be of any shape broadcastable to (N,*S,C).
         scale
             A scale array. If present, the scale is applied to the normalized input.
-            If ``training=True`` then it must be one dimensional with size equal to
-            the size of channel dimension C. If ``training=False`` then it can be of
-            any shape broadcastble to the input shape.
-        training
-            If true, calculate and use the mean and variance of `x`. Otherwise, use the
-            provided `mean` and `variance`.
+            It can be of any shape broadcastable to (N,*S,C).
         eps
             A small float number to avoid dividing by 0.
         momentum
              the value used for the running_mean and running_var computation.
               Default value is 0.1.
+        data_format
+            The ordering of the dimensions in the input, one of "NSC" or "NCS",
+            where N is the batch dimension, S represents any number of spatial
+            dimensions and C is the channel dimension. Default is "NSC".
         out
-            optional output array, for writing the result to.
+            optional output arrays, for writing the result to.
 
         Returns
         -------
@@ -110,6 +107,7 @@ def batch_norm(
             training=training,
             eps=eps,
             momentum=momentum,
+            data_format=data_format,
             out=out,
         )
 
@@ -124,6 +122,7 @@ def instance_norm(
         training: bool = False,
         eps: float = 1e-5,
         momentum: float = 1e-1,
+        data_format: str = "NSC",
         out: Optional[Tuple[ivy.Array, ivy.Array, ivy.Array]] = None,
     ) -> Tuple[ivy.Array, ivy.Array, ivy.Array]:
         """
@@ -134,29 +133,19 @@ def instance_norm(
         Parameters
         ----------
         self
-            Input array of shape (N, *S, C), where N is the batch dimension,
+            Input array of shape default (N, *S, C), where N is the batch dimension,
             *S corresponds to any number of spatial dimensions and
              C corresponds to the channel dimension.
         mean
-            Mean array used for input's normalization. If ``training=True``
-            then it must be one dimensional with size equal to the size of
-            channel dimension C. If ``training=False`` then it can be of any
-            shape broadcastble to the input shape.
+            Mean array of size C used for input's normalization.
         variance
-            Variance array for the input's normalization. If ``training=True``
-            then it must be one dimensional with size equal to the size of
-            channel dimension C. If ``training=False`` then it can be of any shape
-            broadcastble to the input shape.
+            Variance array of size C used for input's normalization.
         offset
-            An offset array. If present, will be added to the normalized input.
-            If ``training=True`` then it must be one dimensional with size equal
-            to the size of channel dimension C. If ``training=False`` then it can
-            be of any shape broadcastble to the input shape.
+            An offset array of size C. If present, will be added
+            to the normalized input.
         scale
-            A scale array. If present, the scale is applied to the normalized input.
-            If ``training=True`` then it must be one dimensional with size equal to
-            the size of channel dimension C. If ``training=False`` then it can be of
-            any shape broadcastble to the input shape.
+            A scale array of size C. If present, the scale is
+            applied to the normalized input.
         training
             If true, calculate and use the mean and variance of `x`. Otherwise, use the
             provided `mean` and `variance`.
@@ -165,6 +154,10 @@ def instance_norm(
         momentum
              the value used for the running_mean and running_var computation.
               Default value is 0.1.
+        data_format
+            The ordering of the dimensions in the input, one of "NSC" or "NCS",
+            where N is the batch dimension, S represents any number of spatial
+            dimensions and C is the channel dimension. Default is "NSC".
         out
             optional output array, for writing the result to.
 
@@ -184,6 +177,7 @@ def instance_norm(
             eps=eps,
             momentum=momentum,
             out=out,
+            data_format=data_format,
         )
 
     def lp_normalize(

ivy/data_classes/array/experimental/statistical.py

@@ -357,7 +357,6 @@ def nanmedian(
 
         Examples
         --------
-
         With :class:`ivy.array` input and default backend set as `numpy`:
 
         >>> a = ivy.array([[10.0, ivy.nan, 4], [3, 2, 1]])