mindspore 2.6.0rc1__cp39-none-any.whl → 2.7.0__cp39-none-any.whl

mindspore/ops/auto_generate/gen_extend_func.py

@@ -17,294 +17,280 @@ from mindspore.common import dtype as mstype
 from mindspore.ops.auto_generate.pyboost_inner_prim import *
 
 
-def prod(input, dim=None, keepdim=False, dtype=None):
+def argmin(input, dim=None, keepdim=False):
     r"""
-    Reduces a dimension of a tensor by multiplying all elements in the dimension, by default. And also can
-    reduce a dimension of `input` along the `dim`. Determine whether the dimensions of the output and input are the
-    same by controlling `keepdim`.
+    Return the indices of the minimum values of a tensor across a dimension.
     
     Args:
-        input (Tensor[Number]): The input tensor. The dtype of the tensor to be reduced is number.
-            :math:`(N, *)` where :math:`*` means, any number of additional dimensions.
-        dim (int): The dimensions to reduce. Default: ``None`` , reduce all dimensions.
-            Only constant value is allowed. Assume the rank of `input` is r, and the value range is [-r,r).
-        keepdim (bool): If ``True`` , keep these reduced dimensions and the length is 1.
-            If ``False`` , don't keep these dimensions. Default: ``False`` .
-        dtype (:class:`mindspore.dtype`): The desired data type of returned Tensor. Default: ``None`` .
+        input (Tensor): Input tensor.
+        dim (Union[int, None], optional): Specify the axis for calculation. If `dim` is ``None`` , the indices of the minimum
+            value within the flattened input will be returned. Default: ``None`` .
+        keepdim (bool, optional): Whether the output tensor retains the specified
+            dimension. Ignored if `dim` is None. Default: ``False`` .
     
     Returns:
-        Tensor, has the same data type as input tensor.
-    
-        - If `dim` is ``None`` , and `keepdim` is  ``False`` ,
-          the output is a 0-D tensor representing the product of all elements in the input tensor.
-        - If `dim` is int, set as 1, and `keepdim` is  ``False`` ,
-          the shape of output is :math:`(input_0, input_2, ..., input_R)`.
+        Tensor, indices of the minimum values of the input tensor across a dimension.
     
     Raises:
-        TypeError: If `input` is not a Tensor.
-        TypeError: If `dim` is not one of the following: int or None.
-        TypeError: If `keepdim` is not a bool.
+        TypeError: If `keepdim` is not bool.
         ValueError: If `dim` is out of range.
     
     Supported Platforms:
-        ``Ascend`` ``GPU`` ``CPU``
+        ``Ascend``
     
     Examples:
-        >>> import mindspore
         >>> import numpy as np
-        >>> from mindspore import Tensor, ops
-        >>> x = Tensor(np.random.randn(3, 4, 5, 6).astype(np.float32))
-        >>> output = ops.prod_ext(x, 1, keepdim=True)
-        >>> result = output.shape
-        >>> print(result)
-        (3, 1, 5, 6)
-        >>> # case 1: Reduces a dimension by multiplying all elements in the dimension.
-        >>> x = Tensor(np.array([[[1, 1, 1, 1, 1, 1], [2, 2, 2, 2, 2, 2], [3, 3, 3, 3, 3, 3]],
-        ...                      [[4, 4, 4, 4, 4, 4], [5, 5, 5, 5, 5, 5], [6, 6, 6, 6, 6, 6]],
-        ...                      [[7, 7, 7, 7, 7, 7], [8, 8, 8, 8, 8, 8], [9, 9, 9, 9, 9, 9]]]), mindspore.float32)
-        >>> output = ops.prod_ext(x)
-        >>> print(output)
-        2.2833798e+33
-        >>> print(output.shape)
-        ()
-        >>> # case 2: Reduces a dimension along dim 0.
-        >>> output = ops.prod_ext(x, 0, True)
-        >>> print(output)
-        [[[ 28.  28.  28.  28.  28.  28.]
-        [ 80.  80.  80.  80.  80.  80.]
-        [162. 162. 162. 162. 162. 162.]]]
-        >>> # case 3: Reduces a dimension along dim 1.
-        >>> output = ops.prod_ext(x, 1, True)
-        >>> print(output)
-        [[[  6.   6.   6.   6.   6.   6.]]
-        [[120. 120. 120. 120. 120. 120.]]
-        [[504. 504. 504. 504. 504. 504.]]]
-        >>> # case 4: Reduces a dimension along dim 2.
-        >>> output = ops.prod_ext(x, 2, True)
+        >>> from mindspore import Tensor
+        >>> from mindspore import ops
+        >>> x = Tensor(np.array([[1, 20, 5], [67, 8, 9], [130, 24, 15]]).astype(np.float32))
+        >>> output = ops.auto_generate.argmin_ext(x, dim=-1)
         >>> print(output)
-        [[[1.00000e+00]
-        [6.40000e+01]
-        [7.29000e+02]]
-        [[4.09600e+03]
-        [1.56250e+04]
-        [4.66560e+04]]
-        [[1.17649e+05]
-        [2.62144e+05]
-        [5.31441e+05]]]
-    """
-    return prod_impl(input, dim, keepdim, dtype)
-
-
-def inplace_adds(input, other, alpha=1):
-    r"""
-    None
+        [0 1 2]
     """
-    return inplace_adds_impl(input, other, alpha)
+    return argmin_impl(input, dim, keepdim)
 
 
-def mm(input, mat2):
+def l1_loss(input, target, reduction='mean'):
     r"""
-    Returns the matrix product of two arrays.
-    If `input` is a :math:`(n \times m)` Tensor, `mat2` is a
-    :math:`(m \times p)` Tensor, `out` will be a :math:`(n \times p)` Tensor.
+    Calculate the mean absolute error between the `input` value and the `target` value.
     
-    Note:
-        This function cannot support broadcasting.
-        Refer to :func:`mindspore.ops.matmul` instead if you need a broadcastable function.
+    Assuming that the :math:`x` and :math:`y` are the predicted value and target value,
+    both are one-dimensional tensors of length :math:`N`, length :math:`N`, `reduction` is set to ``'none'`` ,
+    then calculate the loss of :math:`x` and :math:`y` without dimensionality reduction.
     
-    .. warning::
-        This is an experimental API that is subject to change or deletion.
+    The formula is as follows:
+    
+    .. math::
+        \ell(x, y) = L = \{l_1,\dots,l_N\}^\top, \quad \text{with } l_n = \left| x_n - y_n \right|,
+    
+    where :math:`N` is the batch size.
+    
+    If `reduction` is ``'mean'`` or ``'sum'`` , then:
+    
+    .. math::
+        \ell(x, y) =
+        \begin{cases}
+            \operatorname{mean}(L), & \text{if reduction} = \text{'mean';}\\
+            \operatorname{sum}(L),  & \text{if reduction} = \text{'sum'.}
+        \end{cases}
     
     Args:
-        input (Tensor): The first matrix of matrix multiplication.
-            The last dimension of `input` must be the same size as the first dimension of `mat2`.
-        mat2 (Tensor): The second matrix of matrix multiplication.
-            The last dimension of `input` must be the same size as the first dimension of `mat2`.
+        input (Tensor): Predicted value, Tensor of any dimension.
+        target (Tensor): Target value, usually has the same shape as the `input`.
+            If `input` and `target` have different shapes, make sure they can broadcast to each other.
+        reduction (str, optional): Apply specific reduction method to the output: ``'none'`` , ``'mean'`` ,
+            ``'sum'`` . Default: ``'mean'`` .
+    
+            - ``'none'``: no reduction will be applied.
+            - ``'mean'``: compute and return the mean of elements in the output. Notice: At least one of the input and target is float type when the reduction is ``'mean'`` .
+            - ``'sum'``: the output elements will be summed.
     
     Returns:
-        Tensor, the matrix product of the inputs.
+        Tensor or Scalar, if `reduction` is ``'none'`` , return a Tensor with same shape and dtype as `input`.
+        Otherwise, a scalar value will be returned.
     
     Raises:
-        ValueError: If the last dimension of `input` is not the same size as the
-            second-to-last dimension of `mat2`.
-        TypeError: If `input` or `mat2` is not a Tensor.
-        TypeError: If dtype of `input` or `mat2` is not float16, float32 or bfloat16.
+        TypeError: If `input` is not a Tensor.
+        TypeError: If `target` is not a Tensor.
+        ValueError: If `reduction` is not one of ``'none'`` , ``'mean'`` or ``'sum'`` .
     
     Supported Platforms:
         ``Ascend``
     
     Examples:
-        >>> import mindspore as ms
-        >>> from mindspore import ops
-        >>> import numpy as np
-        >>> x1 = ms.Tensor(np.random.rand(2, 3), ms.float32)
-        >>> x2 = ms.Tensor(np.random.rand(3, 4), ms.float32)
-        >>> out = ops.mm_ext(x1, x2)
-        >>> print(out.shape)
-        (2, 4)
+        >>> from mindspore import Tensor, ops
+        >>> from mindspore import dtype as mstype
+        >>> x = Tensor([[1, 2, 3], [4, 5, 6]], mstype.float32)
+        >>> target = Tensor([[6, 5, 4], [3, 2, 1]], mstype.float32)
+        >>> output = ops.l1_loss_ext(x, target, reduction="mean")
+        >>> print(output)
+        3.0
     """
-    return mm_impl(input, mat2)
+    return l1_loss_impl(input, target, converted_reduction)
 
 
-def isneginf(input):
+def inplace_add(input, other, alpha=1):
     r"""
-    Determines which elements are -inf for each position.
+    None
+    """
+    return inplace_add_impl(input, other, alpha)
+
+
+def sum(input, dim=None, keepdim=False, dtype=None):
+    r"""
+    Calculate sum of Tensor elements over a given dim.
     
-    .. warning::
-        - This API can be used only on the Atlas A2 training series.
+    Note:
+        The `dim` with tensor type is only used for compatibility with older versions and is not recommended.
     
     Args:
-        input (Tensor): Input Tensor.
+        input (Tensor): The input tensor.
+        dim (Union[None, int, tuple(int), list(int), Tensor]): Dimensions along which a sum is performed.
+            If ``None`` , sum all the elements of the input tensor.
+            If the `dim` is a tuple or list of ints, a sum is performed on all the dimensions specified in the tuple.
+            Must be in the range :math:`[-input.ndim, input.ndim)` . Default: ``None`` .
+        keepdim (bool): Whether the output tensor has `dim` retained or not.
+            If ``True`` , keep these reduced dimensions and the length is 1.
+            If ``False`` , don't keep these dimensions. Default: ``False`` .
+        dtype (:class:`mindspore.dtype`): The desired data type of returned Tensor. Default: ``None`` .
     
     Returns:
-        Tensor with the same shape as the input, where elements are `True` if the corresponding element in the `input` is negative infinity, and `False` otherwise.
+        A Tensor, sum of elements over a given `dim` in `input`.
     
     Raises:
-        TypeError: If the input is not a tensor.
+        TypeError: If `input` is not a Tensor.
+        TypeError: If `dim` is not an int, tulpe(int), list(int), Tensor or None.
+        ValueError: If `dim` is not in the range :math:`[-input.ndim, input.ndim)` .
+        TypeError: If `keepdim` is not a bool.
     
     Supported Platforms:
         ``Ascend`` ``GPU`` ``CPU``
     
     Examples:
-        >>> from mindspore import ops, Tensor
+        >>> import mindspore
+        >>> import numpy as np
+        >>> from mindspore import Tensor, ops
         >>> from mindspore import dtype as mstype
-        >>> output = ops.isneginf(Tensor([[-float("inf"), float("inf")], [1, -float("inf")]], mstype.float32))
-        >>> print(output)
-        [[ True False]
-         [False  True]]
+        >>> x = Tensor(np.array([[[1, 1, 1, 1, 1, 1], [2, 2, 2, 2, 2, 2], [3, 3, 3, 3, 3, 3]],
+        ...                      [[4, 4, 4, 4, 4, 4], [5, 5, 5, 5, 5, 5], [6, 6, 6, 6, 6, 6]],
+        ...                      [[7, 7, 7, 7, 7, 7], [8, 8, 8, 8, 8, 8], [9, 9, 9, 9, 9, 9]]]), mstype.float32)
+        >>> out = ops.sum_ext(x)
+        >>> print(out)
+        270.0
+        >>> out = ops.sum_ext(x, dim=2)
+        >>> print(out)
+        [[ 6. 12. 18.]
+        [24. 30. 36.]
+        [42. 48. 54.]]
+        >>> out = ops.sum_ext(x, dim=2, keepdim=True)
+        >>> print(out)
+        [[[ 6.]
+        [12.]
+        [18.]]
+        [[24.]
+        [30.]
+        [36.]]
+        [[42.]
+        [48.]
+        [54.]]]
     """
-    return isneginf_impl(input)
+    return sum_impl(input, dim, keepdim, dtype)
 
 
-def sub(input, other, alpha=1):
+def inplace_adds(input, other, alpha=1):
     r"""
-    Subtracts scaled other value from input Tensor.
+    None
+    """
+    return inplace_adds_impl(input, other, alpha)
+
+
+def acos(input):
+    r"""
+    Computes arccosine of input tensors element-wise.
     
     .. math::
     
-        out_{i} = input_{i} - alpha \times other_{i}
-    
-    Note:
-        - When the two inputs have different shapes,
-          they must be able to broadcast to a common shape.
-        - The two inputs and alpha comply with the implicit type conversion rules to make the data types
-          consistent.
+        out_i = \cos^{-1}(input_i)
     
     Args:
-        input (Union[Tensor, number.Number, bool]): The first input is a number.Number or
-            a bool or a tensor whose data type is
-            `number <https://www.mindspore.cn/docs/en/master/api_python/mindspore/mindspore.dtype.html>`_ or
-            `bool_ <https://www.mindspore.cn/docs/en/master/api_python/mindspore/mindspore.dtype.html>`_.
-        other (Union[Tensor, number.Number, bool]): The second input, is a number.Number or
-            a bool or a tensor whose data type is
-            `number <https://www.mindspore.cn/docs/en/master/api_python/mindspore/mindspore.dtype.html>`_ or
-            `bool_ <https://www.mindspore.cn/docs/en/master/api_python/mindspore/mindspore.dtype.html>`_.
-        alpha (number.Number): A scaling factor applied to `other`, default 1.
+        input (Tensor): The shape of tensor is
+            :math:`(N,*)`, where :math:`*` means any number of additional dimensions.
     
     Returns:
-        Tensor with a shape that is the same as the broadcasted shape of the input `input` and `other`,
-        and the data type is the one with higher precision or higher digits among the two inputs and alpha.
+        Tensor, has the same shape as `input`. The dtype of output is float32 when dtype of `input` is in [bool, int8, uint8, int16, int32, int64]. Otherwise output has the same dtype as `input`.
     
     Raises:
-        TypeError: If the type of `input`, `other`, or `alpha` is not one of the following: Tensor, number.Number, bool.
-        TypeError: If `alpha` is of type float but `input` and `other` are not of type float.
-        TypeError: If `alpha` is of type bool but `input` and `other` are not of type bool.
+        TypeError: If `input` is not a Tensor.
     
     Supported Platforms:
         ``Ascend`` ``GPU`` ``CPU``
     
     Examples:
-        >>> import numpy as np
         >>> import mindspore
-        >>> from mindspore import Tensor
-        >>> from mindspore import ops
-        >>> x = Tensor(np.array([4, 5, 6]).astype(np.float32))
-        >>> y = Tensor(1, mindspore.int32)
-        >>> alpha = 0.5
-        >>> output = ops.auto_generate.sub_ext(x, y, alpha)
+        >>> import numpy as np
+        >>> from mindspore import Tensor, ops
+        >>> input = Tensor(np.array([0.74, 0.04, 0.30, 0.56]), mindspore.float32)
+        >>> output = ops.acos_ext(input)
         >>> print(output)
-        [3.5 4.5 5.5]
-        >>> # the data type of x is float32, the data type of y is int32,
-        >>> # alpha is a float, and the output is the data format of higher precision float32.
-        >>> print(output.dtype)
-        Float32
+        [0.7377037  1.5307857 1.2661037 0.9764114]
     """
-    return sub_impl(input, other, alpha)
+    return acos_impl(input)
 
 
-def argmin(input, dim=None, keepdim=False):
+def diag(input, diagonal=0):
     r"""
-    Return the indices of the minimum values of a tensor across a dimension.
+    If input is a vector (1-D tensor), then returns a 2-D square tensor with the elements of input as the diagonal.
+    
+    If input is a matrix (2-D tensor), then returns a 1-D tensor with the diagonal elements of input.
+    
+    The argument diagonal controls which diagonal to consider:
+    
+    - If `diagonal` = 0, it is the main diagonal.
+    
+    - If `diagonal` > 0, it is above the main diagonal.
+    
+    - If `diagonal` < 0, it is below the main diagonal.
+    
+    .. warning::
+        This is an experimental API that is subject to change or deletion.
     
     Args:
-        input (Tensor): Input tensor.
-        dim (Union[int, None], optional): Specify the axis for calculation. If `dim` is ``None`` , the indices of the minimum
-            value within the flattened input will be returned. Default: ``None`` .
-        keepdim (bool, optional): Whether the output tensor retains the specified
-            dimension. Ignored if `dim` is None. Default: ``False`` .
+        input (Tensor): The input tensor.
+        diagonal (int, optional): the diagonal to consider. Defaults: ``0``.
     
     Returns:
-        Tensor, indices of the minimum values of the input tensor across a dimension.
+        Tensor, has the same dtype as the `input`, its shape is up to `diagonal`.
+    
+        - If `input` shape is :math:`(x_0)` : then output shape is :math:`(x_0 + \left | diagonal \right | , x_0 + \left | diagonal \right | )` 2-D Tensor.
+    
+        - If `input` shape is :math:`(x_0, x_1)` : then output shape is main diagonal to move :math:`(\left | diagonal \right |)` elements remains elements' length 1-D Tensor.
     
     Raises:
-        TypeError: If `keepdim` is not bool.
-        ValueError: If `dim` is out of range.
+        TypeError: If `input` is not a Tensor.
+        ValueError: If shape of `input` is not 1-D and 2-D.
     
     Supported Platforms:
         ``Ascend``
     
     Examples:
-        >>> import numpy as np
-        >>> from mindspore import Tensor
-        >>> from mindspore import mint
-        >>> x = Tensor(np.array([[1, 20, 5], [67, 8, 9], [130, 24, 15]]).astype(np.float32))
-        >>> output = mint.argmin(x, dim=-1)
+        >>> from mindspore import Tensor, ops
+        >>> input = Tensor([1, 2, 3, 4]).astype('int32')
+        >>> output = ops.auto_generate.diag_ext(input)
         >>> print(output)
-        [0 1 2]
-    """
-    return argmin_impl(input, dim, keepdim)
-
-
-def ffn(x, weight1, weight2, expertTokens=None, bias1=None, bias2=None, scale=None, offset=None, deqScale1=None, deqScale2=None, antiquant_scale1=None, antiquant_scale2=None, antiquant_offset1=None, antiquant_offset2=None, activation='fastgelu', inner_precise=0):
-    r"""
-    None
+        [[1 0 0 0]
+         [0 2 0 0]
+         [0 0 3 0]
+         [0 0 0 4]]
     """
-    return ffn_impl(x, weight1, weight2, expertTokens, bias1, bias2, scale, offset, deqScale1, deqScale2, antiquant_scale1, antiquant_scale2, antiquant_offset1, antiquant_offset2, converted_activation, inner_precise)
+    return diag_impl(input, diagonal)
 
 
-def elu(input, alpha=1.0):
+def log_softmax(input, dim=None, dtype=None):
     r"""
-    Exponential Linear Unit activation function.
-    
-    Applies the exponential linear unit function element-wise.
-    The activation function is defined as:
+    Applies the Log Softmax function to the input tensor on the specified axis.
+    Supposes a slice in the given axis, :math:`x` for each element :math:`x_i`,
+    the Log Softmax function is shown as follows:
     
     .. math::
+        \text{output}(x_i) = \log \left(\frac{\exp(x_i)} {\sum_{j = 0}^{N-1}\exp(x_j)}\right),
     
-        \text{ELU}(x)= \left\{
-        \begin{array}{align}
-            \alpha(e^{x}  - 1) & \text{if } x \le 0\\
-            x & \text{if } x \gt 0\\
-        \end{array}\right.
-    
-    Where :math:`x` is the element of input Tensor `input`, :math:`\alpha` is param `alpha`,
-    it determines the smoothness of ELU.
-    
-    ELU function graph:
-    
-    .. image:: ../images/ELU.png
-        :align: center
+    where :math:`N` is the length of the Tensor.
     
     Args:
-        input (Tensor): The input of ELU is a Tensor of any dimension.
-        alpha (float, optional): The alpha value of ELU, the data type is float.
-            Default: ``1.0`` .
+        input (Tensor): The input Tensor.
+        dim (int, optional): The axis to perform the Log softmax operation. Default: ``None`` .
+    
+    Keyword Args:
+        dtype (:class:`mindspore.dtype`, optional): The desired dtype of returned Tensor. If not set to None, the input
+            Tensor will be cast to `dtype` before the operation is performed. This is useful for preventing overflows.
+            If set to None, stay the same as original Tensor. Default: ``None`` . Supported data type is {float16, float32, double, bfloat16}.
     
     Returns:
-        Tensor, has the same shape and data type as `input`.
+        Tensor, with the same shape as the input.
     
     Raises:
-        TypeError: If `alpha` is not a float.
+        TypeError: If `dim` is not an int.
+        ValueError: If `dim` is not in range [-len(input.shape), len(input.shape)).
     
     Supported Platforms:
         ``Ascend``
@@ -313,202 +299,128 @@ def elu(input, alpha=1.0):
         >>> import mindspore
         >>> import numpy as np
         >>> from mindspore import Tensor, ops
-        >>> x = Tensor(np.array([[-1.0, 4.0, -8.0], [2.0, -5.0, 9.0]]), mindspore.float32)
-        >>> output = ops.auto_generate.elu_ext(x)
+        >>> logits = Tensor(np.array([1, 2, 3, 4, 5]), mindspore.float32)
+        >>> output = ops.auto_generate.log_softmax(logits, dim=-1)
         >>> print(output)
-        [[-0.63212055  4.         -0.99966455]
-         [ 2.         -0.99326205  9.        ]]
+        [-4.4519143 -3.4519143 -2.4519143 -1.4519144 -0.4519144]
     """
-    return elu_impl(input, alpha)
+    return log_softmax_impl(input, dim, dtype)
 
 
-def mean(input, dim=None, keepdim=False, dtype=None):
+def cumsum(input, dim, dtype=None):
     r"""
-    Reduces all dimension of a tensor by averaging all elements in the dimension, by default.
-    And reduce a dimension of `input` along the specified `dim`. `keepdim`
-    determines whether the dimensions of the output and input are the same.
+    Computes the cumulative sum of input Tensor along `dim`.
     
-    Note:
-        The `dim` with tensor type is only used for compatibility with older versions and is not recommended.
+    .. math::
+    
+        y_i = x_1 + x_2 + x_3 + ... + x_i
     
     Args:
-        input (Tensor[Number]): The input tensor. The dtype of the tensor to be reduced is number.
-            :math:`(N, *)` where :math:`*` means, any number of additional dimensions.
-        dim (Union[int, tuple(int), list(int), Tensor]): The dimensions to reduce. Default: ``None`` ,
-            reduce all dimensions. Only constant value is allowed. Assume the rank of `input` is r,
-            and the value range is [-r,r).
-        keepdim (bool): If ``True`` , keep these reduced dimensions and the length is 1.
-            If ``False`` , don't keep these dimensions. Default: ``False`` .
-        dtype (:class:`mindspore.dtype`): The desired data type of returned Tensor. Default: ``None`` .
+        input (Tensor): The input Tensor.
+        dim (int): Dim along which the cumulative sum is computed.
+        dtype (:class:`mindspore.dtype`, optional): The desired dtype of returned Tensor. If specified,
+            the input Tensor will be cast to `dtype` before the computation. This is useful for preventing overflows.
+            If not specified, stay the same as original Tensor. Default: ``None`` .
     
     Returns:
-        Tensor, has the same data type as input tensor.
-    
-        - If `dim` is ``None`` , and `keepdim` is ``False`` ,
-          the output is a 0-D tensor representing the product of all elements in the input tensor.
-        - If `dim` is int, set as 1, and `keepdim` is ``False`` ,
-          the shape of output is :math:`(x_0, x_2, ..., x_R)`.
-        - If `dim` is tuple(int), set as (1, 2), and `keepdim` is ``False`` ,
-          the shape of output is :math:`(x_0, x_3, ..., x_R)`.
-        - If `dim` is 1-D Tensor, set as [1, 2], and `keepdim` is ``False`` ,
-          the shape of output is :math:`(x_0, x_3, ..., x_R)`.
+        Tensor, the shape of the output Tensor is consistent with the input Tensor's.
     
     Raises:
-        TypeError: If `x` is not a Tensor.
-        TypeError: If `dim` is not one of the following: int, tuple, list or Tensor.
-        TypeError: If `keepdim` is not a bool.
-        ValueError: If `dim` is out of range.
+        TypeError: If `input` is not a Tensor.
+        ValueError: If the `dim` is out of range.
     
     Supported Platforms:
-        ``Ascend`` ``GPU`` ``CPU``
+        ``Ascend``
     
     Examples:
-        >>> import mindspore
         >>> import numpy as np
-        >>> from mindspore import Tensor, ops
-        >>> x = Tensor(np.random.randn(3, 4, 5, 6).astype(np.float32))
-        >>> output = ops.mean_ext(x, 1, keepdim=True)
-        >>> result = output.shape
-        >>> print(result)
-        (3, 1, 5, 6)
-        >>> # case 1: Reduces a dimension by averaging all elements in the dimension.
-        >>> x = Tensor(np.array([[[2, 2, 2, 2, 2, 2], [2, 2, 2, 2, 2, 2], [2, 2, 2, 2, 2, 2]],
-        ... [[4, 4, 4, 4, 4, 4], [5, 5, 5, 5, 5, 5], [6, 6, 6, 6, 6, 6]],
-        ... [[6, 6, 6, 6, 6, 6], [8, 8, 8, 8, 8, 8], [10, 10, 10, 10, 10, 10]]]),
-        ... mindspore.float32)
-        >>> output = ops.mean_ext(x)
-        >>> print(output)
-        5.0
-        >>> print(output.shape)
-        ()
-        >>> # case 2: Reduces a dimension along the dim 0
-        >>> output = ops.mean_ext(x, 0, True)
-        >>> print(output)
-        [[[4. 4. 4. 4. 4. 4.]
-        [5. 5. 5. 5. 5. 5.]
-        [6. 6. 6. 6. 6. 6.]]]
-        >>> # case 3: Reduces a dimension along the dim 1
-        >>> output = ops.mean_ext(x, 1, True)
-        >>> print(output)
-        [[[2. 2. 2. 2. 2. 2.]]
-        [[5. 5. 5. 5. 5. 5.]]
-        [[8. 8. 8. 8. 8. 8.]]]
-        >>> # case 4: Reduces a dimension along the dim 2
-        >>> output = ops.mean_ext(x, 2, True)
-        >>> print(output)
-        [[[ 2.]
-        [ 2.]
-        [ 2.]]
-        [[ 4.]
-        [ 5.]
-        [ 6.]]
-        [[ 6.]
-        [ 8.]
-        [10.]]]
+        >>> from mindspore import Tensor
+        >>> import mindspore.ops as ops
+        >>> x = Tensor(np.array([[3, 4, 6, 10], [1, 6, 7, 9], [4, 3, 8, 7], [1, 3, 7, 9]]).astype(np.float32))
+        >>> # case 1: along the dim 0
+        >>> y = ops.auto_generate.cumsum_ext(x, 0)
+        >>> print(y)
+        [[ 3.  4.  6. 10.]
+        [ 4. 10. 13. 19.]
+        [ 8. 13. 21. 26.]
+        [ 9. 16. 28. 35.]]
+        >>> # case 2: along the dim 1
+        >>> y = ops.auto_generate.cumsum_ext(x, 1)
+        >>> print(y)
+        [[ 3.  7. 13. 23.]
+        [ 1.  7. 14. 23.]
+        [ 4.  7. 15. 22.]
+        [ 1.  4. 11. 20.]]
     """
-    return mean_impl(input, dim, keepdim, dtype)
+    return cumsum_impl(input, dim, dtype)
 
 
-def stack(tensors, dim=0):
+def outer(input, vec2):
     r"""
-    Stacks a list of tensors in specified dim.
+    Return outer product of `input` and `vec2`. If `input` is a vector of size :math:`n`
+    and `vec2` is a vector of size :math:`m` , then output must be a matrix of shape :math:`(n, m)` .
     
-    Stacks the list of input tensors with the same rank `R`, output is a tensor of rank `(R+1)`.
+    .. warning::
+        This is an experimental API that is subject to change or deletion.
     
-    Given input tensors of shape :math:`(x_1, x_2, ..., x_R)`. Set the number of input tensors as `N`.
-    If :math:`dim \ge 0`, the shape of the output tensor is
-    :math:`(x_1, x_2, ..., x_{dim}, N, x_{dim+1}, ..., x_R)`.
+    .. note::
+        This function does not broadcast.
     
     Args:
-        tensors (Union[tuple, list]): A Tuple or list of Tensor objects with the same shape and type.
-        dim (int, optional): Dimension to stack. The range is [-(R+1), R+1). Default: ``0`` .
+        input (Tensor): 1-D input vector.
+        vec2 (Tensor): 1-D input vector.
     
     Returns:
-        Tensor. A stacked Tensor with the same type as `tensors`.
+        out, 2-D matrix, the outer product of two vectors.
     
     Raises:
-        TypeError: If the data types of elements in `tensors` are not the same.
-        ValueError: If `dim` is out of the range [-(R+1), R+1);
-                    or if the shapes of elements in `tensors` are not the same.
+        TypeError: If `input` or `vec2` is not a Tensor.
+        TypeError: The implicitly converted data types of `input` and `vec2` are not one of float16, float32, float64, bool, uint8, int8, int16, int32, int64, complex64, complex128, bfloat16
+        ValueError: If the dimension of `input` or `vec2` is not equal to 1.
     
     Supported Platforms:
         ``Ascend``
     
     Examples:
         >>> import mindspore
-        >>> from mindspore import Tensor, ops
         >>> import numpy as np
-        >>> data1 = Tensor(np.array([0, 1]).astype(np.float32))
-        >>> data2 = Tensor(np.array([2, 3]).astype(np.float32))
-        >>> output = ops.auto_generate.stack_ext([data1, data2], 0)
-        >>> print(output)
-        [[0. 1.]
-         [2. 3.]]
-    """
-    return stack_impl(tensors, dim)
-
-
-def cummin(input, dim):
-    r"""
-    Returns a tuple (values, indices) where `values` is the cumulative minimum value of input Tensor `input`
-    along the dimension `dim`, and `indices` is the index location of each minimum value.
-    
-    .. math::
-        \begin{array}{ll} \\
-            y_{i} = \min(x_{1}, x_{2}, ... , x_{i})
-        \end{array}
-    
-    .. note::
-        O2 mode is not supported in Ascend.
-    
-    Args:
-        input (Tensor): The input Tensor, The dimension must be greater than 0.
-        dim (int): Operation dimension. The value of `dim` must be in the range `[-input.ndim, input.ndim - 1]`.
-    
-    Returns:
-        tuple [Tensor], tuple of 2 Tensors, containing the cumulative minimum of elements and the index.
-        The shape of each output tensor is the same as that of input `input`.
-    
-    Raises:
-        TypeError: If `input` is not a Tensor.
-        TypeError: If `input` is a Tensor, but the type is complex or bool.
-        TypeError: If `dim` is not an int.
-        ValueError: If `dim` is out the range of `[-input.ndim, input.ndim - 1]`.
-    
-    Supported Platforms:
-        ``Ascend``
-    
-    Examples:
-        >>> from mindspore import Tensor, ops
-        >>> import mindspore
-        >>> a = Tensor([-0.2284, -0.6628,  0.0975,  0.2680, -1.3298, -0.4220], mindspore.float32)
-        >>> output = ops.cummin_ext(a, dim=0)
-        >>> print(output[0])
-        [-0.2284 -0.6628 -0.6628 -0.6628 -1.3298 -1.3298]
-        >>> print(output[1])
-        [0 1 1 1 4 4]
+        >>> from mindspore import Tensor
+        >>> from mindspore import ops
+        >>> input = Tensor(np.array([7, 8, 9]), mindspore.int32)
+        >>> vec2 = Tensor(np.array([7, 10, 11]), mindspore.int32)
+        >>> out = ops.outer(input, vec2)
+        >>> print(out)
+        [[49 70 77]
+         [56 80 88]
+         [63 90 99]]
     """
-    return cummin_impl(input, dim)
+    return outer_impl(input, vec2)
 
 
-def flatten(input, start_dim=0, end_dim=-1):
+def matmul(input, other):
     r"""
-    Flatten a tensor along dimensions from `start_dim` to `end_dim`.
+    None
+    """
+    return matmul_impl(input, other)
+
+
+def asin(input):
+    r"""
+    Computes arcsine of input tensors element-wise.
+    
+    .. math::
+    
+        out_i = \sin^{-1}(input_i)
     
     Args:
-        input (Tensor): The input Tensor.
-        start_dim (int, optional): The first dimension to flatten. Default: ``0`` .
-        end_dim (int, optional): The last dimension to flatten. Default: ``-1`` .
+        input (Tensor): The shape of tensor is
+            :math:`(N,*)`, where :math:`*` means any number of additional dimensions.
     
     Returns:
-        Tensor. If no dimensions are flattened, returns the original `input`, otherwise return the flattened Tensor.
-        If `input` is a 0-dimensional Tensor, a 1-dimensional Tensor will be returned.
+        Tensor, has the same shape as `input`. The dtype of output is float32 when dtype of `input` is in [bool, int8, uint8, int16, int32, int64]. Otherwise output has the same dtype as `input`.
     
     Raises:
         TypeError: If `input` is not a Tensor.
-        TypeError: If `start_dim` or `end_dim` is not int.
-        ValueError: If `start_dim` is greater than `end_dim` after canonicalized.
-        ValueError: If `start_dim` or `end_dim` is not in range of [-input.dim, input.dim-1].
     
     Supported Platforms:
         ``Ascend`` ``GPU`` ``CPU``
@@ -517,151 +429,107 @@ def flatten(input, start_dim=0, end_dim=-1):
         >>> import mindspore
         >>> import numpy as np
         >>> from mindspore import Tensor, ops
-        >>> input_x = Tensor(np.ones(shape=[1, 2, 3, 4]), mindspore.float32)
-        >>> output = ops.auto_generate.flatten_ext(input_x)
-        >>> print(output.shape)
-        (24,)
+        >>> input = Tensor(np.array([0.74, 0.04, 0.30, 0.56]), mindspore.float32)
+        >>> output = ops.asin_ext(input)
+        >>> print(output)
+        [0.8330927  0.04001068  0.30469266  0.59438497 ]
     """
-    return flatten_impl(input, start_dim, end_dim)
+    return asin_impl(input)
 
 
-def atan2(input, other):
+def asinh(input):
     r"""
-    Returns arctangent of input/other element-wise.
+    Computes inverse hyperbolic sine of the input element-wise.
     
-    It returns :math:`\theta\ \in\ [-\pi, \pi]`
-    such that :math:`input = r*\sin(\theta), other = r*\cos(\theta)`, where :math:`r = \sqrt{input^2 + other^2}`.
+    .. math::
     
-    Note:
-        - Arg `input` and `other` comply with the implicit type conversion rules to make the data types consistent.
-          If they have different data types, the lower precision data type will be converted to relatively the
-          highest precision data type.
+        out_i = \sinh^{-1}(input_i)
     
     Args:
-        input (Tensor, Number.number): The input tensor or scalar.
-        other (Tensor, Number.number): The input tensor or scalar. It has the same shape with `input` or
-            its shape is able to broadcast with `input`.
+        input (Tensor): The input tensor of inverse hyperbolic sine function.
     
     Returns:
-        Tensor, the shape is the same as the one after broadcasting.
-        The dtype of output is float32 when dtype of `input` is in
-        [bool, int8, uint8, int16, int32, int64]. Otherwise output has the same dtype as `input`.
+        Tensor, has the same shape as `input`. The dtype of output is float32 when dtype of `input` is in [bool, int8, uint8, int16, int32, int64]. Otherwise output has the same dtype as `input`.
     
     Raises:
-        TypeError: If `input` or `other` is not a Tensor or scalar.
-        RuntimeError: If the data type of `input` and `other` conversion of Parameter is required
-                    when data type conversion of Parameter is not supported.
+        TypeError: If `input` is not a Tensor.
     
     Supported Platforms:
-        ``Ascend``
+        ``Ascend`` ``GPU`` ``CPU``
     
     Examples:
         >>> import mindspore
         >>> import numpy as np
         >>> from mindspore import Tensor, ops
-        >>> input = Tensor(np.array([0, 1]), mindspore.float32)
-        >>> other = Tensor(np.array([1, 1]), mindspore.float32)
-        >>> output = mint.atan2(input, other)
+        >>> input = Tensor(np.array([-5.0, 1.5, 3.0, 100.0]), mindspore.float32)
+        >>> output = ops.asinh_ext(input)
         >>> print(output)
-        [0.        0.7853982]
-    """
-    return atan2_impl(input, other)
-
-
-def sort(input, dim=-1, descending=False, stable=False):
-    r"""
-    None
+        [-2.3124385  1.1947632  1.8184465  5.298342 ]
     """
-    return sort_impl(input, dim, descending, stable)
+    return asinh_impl(input)
 
 
-def unfold(input, kernel_size, dilation=1, padding=0, stride=1):
+def logaddexp(input, other):
     r"""
-    Extracts sliding local blocks from a batched input tensor.
-    
-    Consider a batched input tensor of shape :math:`(N, C, *)`,
-    where :math:`N` is the batch dimension, :math:`C` is the channel dimension,
-    and :math:`*` represent arbitrary spatial dimensions. This operation flattens
-    each sliding `Kernel_size`- sized block within the spatial dimensions
-    of `input` into a column (i.e., last dimension) of a 3-D output
-    tensor of shape :math:`(N, C \times \prod(\text{kernel_size}), L)`, where
-    :math:`C \times \prod(\text{kernel_size})` is the total number of values
-    within each block (a block has :math:`\prod(\text{kernel_size})` spatial
-    locations each containing a `C`-channeled vector), and :math:`L` is
-    the total number of such blocks:
+    Computes the logarithm of the sum of exponentiations of the inputs.
+    This function is useful in statistics where the calculated probabilities of events may be
+    so small as to exceed the range of normal floating point numbers.
     
     .. math::
-        L = \prod_d \left\lfloor\frac{\text{spatial_size}[d] + 2 \times \text{padding}[d] %
-            - \text{dilation}[d] \times (\text{kernel_size}[d] - 1) - 1}{\text{stride}[d]} + 1\right\rfloor,
-    
-    where :math:`\text{spatial_size}` is formed by the spatial dimensions
-    of `input` (:math:`*` above), and :math:`d` is over all spatial
-    dimensions.
-    
-    Therefore, indexing `output` at the last dimension (column dimension)
-    gives all values within a certain block.
     
-    The `dilation`, `padding` and `stride` arguments specify
-    how the sliding blocks are retrieved.
+        out_i = \log(exp(input_i) + \exp(other_i))
     
     .. warning::
-        - Currently, batched(4D) image-like tensors are supported.
-        - For Ascend, it is only supported on platforms above Atlas A2.
+        This is an experimental API that is subject to change or deletion.
     
     Args:
-        input (Tensor): 4-D Tensor.
-        kernel_size (Union[int, tuple[int], list[int]]): The size of the kernel, should be two int
-            for height and width. If type is int, it means that height equal with width. Must be specified.
-        dilation (Union[int, tuple[int], list[int]], optional): The dilation of the window, should be two int
-            for height and width. If type is int, it means that height equal with width. Default: ``1`` .
-        padding (Union[int, tuple[int], list[int]], optional): The pad of the window, should be two int
-            for height and width. If type is int, it means that height equal with width. Default: ``0`` .
-        stride (Union[int, tuple[int], list[int]], optional): The stride of the window, should be two int
-            for height and width. If type is int, it means that height equal with width. Default: ``1`` .
+        input (Tensor): Input Tensor. The dtype of `input` must be float.
+        other (Tensor): Input Tensor. The dtype of `other` must be float.
+            If the shape of `input` is not equal to the shape of `other`,
+            they must be broadcastable to a common shape.
     
     Returns:
-        A Tensor, with same type as `input` .
-    
-    Shape:
-        - Input: :math:`(N, C, *)`
-        - Output: :math:`(N, C \times \prod(\text{kernel_size}), L)`
+        Tensor, with the same dtype as `input` and `other`.
     
     Raises:
-        TypeError: If any data type of `kernel_size`, `stride`, `dilation`, `padding` is not int, tuple or list.
-        ValueError: If `kernel_size`, `dilation`, `stride` value is not
-            greater than zero or elements number more than `2`.
-        ValueError: If `padding` value is less than zero.
+        TypeError: If `input` or `other` is not a Tensor.
+        TypeError: The dtype of `input` or `other` is not float.
     
     Supported Platforms:
         ``Ascend``
     
     Examples:
-        >>> import mindspore
         >>> import numpy as np
         >>> from mindspore import Tensor, ops
-        >>> x = Tensor(np.random.rand(4, 4, 32, 32), mindspore.float32)
-        >>> output = ops.auto_generate.unfold_ext(x, kernel_size=3, dilation=1, stride=1)
-        >>> print(output.shape)
-        (4, 36, 900)
+        >>> x1 = Tensor(np.array([1, 2, 3]).astype(np.float16))
+        >>> x2 = Tensor(np.array(2).astype(np.float16))
+        >>> output = ops.logaddexp_ext(x1, x2)
+        >>> print(output)
+        [2.312 2.693 3.312]
     """
-    return unfold_impl(input, converted_kernel_size, converted_dilation, converted_padding, converted_stride)
+    return logaddexp_impl(input, other)
 
 
-def frac(input):
+def log10(input):
     r"""
-    Calculates the fractional part of each element in the input.
+    Returns the logarithm to the base 10 of a tensor element-wise.
     
     .. math::
-        out_i = input_i - \lfloor |input_i| \rfloor * sgn(input_i)
+        y_i = \log_{10}(x_i)
     
     .. warning::
-        This is an experimental API that is subject to change or deletion.
+        - This is an experimental API that is subject to change or deletion.
+        - If the input value of operator Log10 is within the range (0, 0.01] or [0.95, 1.05], the output accuracy
+          may be affacted.
     
     Args:
-        input (Tensor): The input Tensor.
+        input (Tensor): Input Tensor of any dimension. The value must be greater than 0.
     
     Returns:
-        Tensor, has the same shape and type as input.
+        Tensor, has the same shape as the `input`, and the dtype changes according to the `input.dtype`.
+        
+        - if `input.dtype` is in [float16, float32, float64, bfloat16], the output dtype is the same as the `input.dtype`.
+        - if `input.dtype` is integer or boolean type, the output dtype is float32.
     
     Raises:
         TypeError: If `input` is not a Tensor.
@@ -673,277 +541,132 @@ def frac(input):
         >>> import mindspore
         >>> import numpy as np
         >>> from mindspore import Tensor, ops
-        >>> x = Tensor([2, 4.2, -2.5], mindspore.float16)
-        >>> output = ops.frac_ext(x)
+        >>> x = Tensor(np.array([3.0, 5.0, 7.0]), mindspore.float32)
+        >>> output = ops.auto_generate.log10_ext(x)
         >>> print(output)
-          [ 0.      0.1992 -0.5   ]
+        [0.47712136 0.69897    0.845098  ]
     """
-    return frac_impl(input)
+    return log10_impl(input)
 
 
-def select(input, dim, index):
+def flatten(input, start_dim=0, end_dim=-1):
     r"""
-    Slices the input tensor along the selected dimension at the given index.
-    
-    .. warning::
-        This is an experimental API that is subject to change or deletion.
+    Flatten a tensor along dimensions from `start_dim` to `end_dim`.
     
     Args:
-        input (Tensor): the input tensor.
-        dim (int): the dimension to slice.
-        index (int): the index to select with.
+        input (Tensor): The input Tensor.
+        start_dim (int, optional): The first dimension to flatten. Default: ``0`` .
+        end_dim (int, optional): The last dimension to flatten. Default: ``-1`` .
     
     Returns:
-        Tensor.
+        Tensor. If no dimensions are flattened, returns the original `input`, otherwise return the flattened Tensor.
+        If `input` is a 0-dimensional Tensor, a 1-dimensional Tensor will be returned.
     
     Raises:
-        TypeError: If input is not a Tensor.
+        TypeError: If `input` is not a Tensor.
+        TypeError: If `start_dim` or `end_dim` is not int.
+        ValueError: If `start_dim` is greater than `end_dim` after canonicalized.
+        ValueError: If `start_dim` or `end_dim` is not in range of [-input.dim, input.dim-1].
     
     Supported Platforms:
-        ``Ascend``
+        ``Ascend`` ``GPU`` ``CPU``
     
     Examples:
         >>> import mindspore
-        >>> from mindspore import Tensor, mint
-        >>> input = Tensor([[2, 3, 4, 5],[3, 2, 4, 5]])
-        >>> y = mint.select(input, 0, 0)
-        >>> print(y)
-        [2 3 4 5]
+        >>> import numpy as np
+        >>> from mindspore import Tensor, ops
+        >>> input_x = Tensor(np.ones(shape=[1, 2, 3, 4]), mindspore.float32)
+        >>> output = ops.auto_generate.flatten_ext(input_x)
+        >>> print(output.shape)
+        (24,)
     """
-    return select_impl(input, dim, index)
+    return flatten_impl(input, start_dim, end_dim)
 
 
-def mish(input):
+def elu(input, alpha=1.0):
     r"""
-    Computes MISH (A Self Regularized Non-Monotonic Neural Activation Function)
-    of input tensors element-wise.
+    Exponential Linear Unit activation function.
     
-    The formula is defined as follows:
+    Applies the exponential linear unit function element-wise.
+    The activation function is defined as:
     
     .. math::
-        \text{mish}(input) = input * \tanh(softplus(\text{input}))
     
-    See more details in `A Self Regularized Non-Monotonic Neural Activation Function 
-    <https://arxiv.org/abs/1908.08681>`_.
-    
-    Mish Activation Function Graph:
-    
-    .. image:: ../images/Mish.png
-        :align: center
-    
-    Args:
-        input (Tensor): The input of MISH. Supported dtypes: 
-    
-            - Ascend: float16, float32.
-    
-    Returns:
-        Tensor, has the same type and shape as the `input`.
-    
-    Raises:
-        TypeError: If `input` is not a Tensor.
-        TypeError: If dtype of `input` is not float16 or float32.
-    
-    Supported Platforms:
-        ``Ascend``
-    
-    Examples:
-        >>> import mindspore
-        >>> from mindspore import Tensor, ops
-        >>> import numpy as np
-        >>> x = Tensor(np.array([[-1.1, 4.0, -8.0], [2.0, -5.0, 9.0]]), mindspore.float32)
-        >>> output = ops.mish(x)
-        >>> print(output)
-        [[-3.0764845e-01 3.9974124e+00 -2.6832507e-03]
-         [ 1.9439589e+00 -3.3576239e-02 8.9999990e+00]]
-    """
-    return mish_impl(input)
-
-
-def logaddexp(input, other):
-    r"""
-    Computes the logarithm of the sum of exponentiations of the inputs.
-    This function is useful in statistics where the calculated probabilities of events may be
-    so small as to exceed the range of normal floating point numbers.
+        \text{ELU}(x)= \left\{
+        \begin{array}{align}
+            \alpha(e^{x}  - 1) & \text{if } x \le 0\\
+            x & \text{if } x \gt 0\\
+        \end{array}\right.
     
-    .. math::
+    Where :math:`x` is the element of input Tensor `input`, :math:`\alpha` is param `alpha`,
+    it determines the smoothness of ELU.
     
-        out_i = \log(exp(input_i) + \exp(other_i))
+    ELU function graph:
     
-    .. warning::
-        This is an experimental API that is subject to change or deletion.
+    .. image:: ../images/ELU.png
+        :align: center
     
     Args:
-        input (Tensor): Input Tensor. The dtype of `input` must be float.
-        other (Tensor): Input Tensor. The dtype of `other` must be float.
-            If the shape of `input` is not equal to the shape of `other`,
-            they must be broadcastable to a common shape.
+        input (Tensor): The input of ELU is a Tensor of any dimension.
+        alpha (float, optional): The alpha value of ELU, the data type is float.
+            Default: ``1.0`` .
     
     Returns:
-        Tensor, with the same dtype as `input` and `other`.
+        Tensor, has the same shape and data type as `input`.
     
     Raises:
-        TypeError: If `input` or `other` is not a Tensor.
-        TypeError: The dtype of `input` or `other` is not float.
+        TypeError: If `alpha` is not a float.
     
     Supported Platforms:
         ``Ascend``
     
-    Examples:
-        >>> import numpy as np
-        >>> from mindspore import Tensor, ops
-        >>> x1 = Tensor(np.array([1, 2, 3]).astype(np.float16))
-        >>> x2 = Tensor(np.array(2).astype(np.float16))
-        >>> output = ops.logaddexp_ext(x1, x2)
-        >>> print(output)
-        [2.312 2.693 3.312]
-    """
-    return logaddexp_impl(input, other)
-
-
-def sum(input, dim=None, keepdim=False, dtype=None):
-    r"""
-    Calculate sum of Tensor elements over a given dim.
-    
-    Note:
-        The `dim` with tensor type is only used for compatibility with older versions and is not recommended.
-    
-    Args:
-        input (Tensor): The input tensor.
-        dim (Union[None, int, tuple(int), list(int), Tensor]): Dimensions along which a sum is performed.
-            If ``None`` , sum all the elements of the input tensor.
-            If the `dim` is a tuple or list of ints, a sum is performed on all the dimensions specified in the tuple.
-            Must be in the range :math:`[-input.ndim, input.ndim)` . Default: ``None`` .
-        keepdim (bool): Whether the output tensor has `dim` retained or not.
-            If ``True`` , keep these reduced dimensions and the length is 1.
-            If ``False`` , don't keep these dimensions. Default: ``False`` .
-        dtype (:class:`mindspore.dtype`): The desired data type of returned Tensor. Default: ``None`` .
-    
-    Returns:
-        A Tensor, sum of elements over a given `dim` in `input`.
-    
-    Raises:
-        TypeError: If `input` is not a Tensor.
-        TypeError: If `dim` is not an int, tulpe(int), list(int), Tensor or None.
-        ValueError: If `dim` is not in the range :math:`[-input.ndim, input.ndim)` .
-        TypeError: If `keepdim` is not a bool.
-    
-    Supported Platforms:
-        ``Ascend`` ``GPU`` ``CPU``
-    
     Examples:
         >>> import mindspore
         >>> import numpy as np
         >>> from mindspore import Tensor, ops
-        >>> from mindspore import dtype as mstype
-        >>> x = Tensor(np.array([[[1, 1, 1, 1, 1, 1], [2, 2, 2, 2, 2, 2], [3, 3, 3, 3, 3, 3]],
-        ...                      [[4, 4, 4, 4, 4, 4], [5, 5, 5, 5, 5, 5], [6, 6, 6, 6, 6, 6]],
-        ...                      [[7, 7, 7, 7, 7, 7], [8, 8, 8, 8, 8, 8], [9, 9, 9, 9, 9, 9]]]), mstype.float32)
-        >>> out = ops.sum_ext(x)
-        >>> print(out)
-        270.0
-        >>> out = ops.sum_ext(x, dim=2)
-        >>> print(out)
-        [[ 6. 12. 18.]
-        [24. 30. 36.]
-        [42. 48. 54.]]
-        >>> out = ops.sum_ext(x, dim=2, keepdim=True)
-        >>> print(out)
-        [[[ 6.]
-        [12.]
-        [18.]]
-        [[24.]
-        [30.]
-        [36.]]
-        [[42.]
-        [48.]
-        [54.]]]
+        >>> x = Tensor(np.array([[-1.0, 4.0, -8.0], [2.0, -5.0, 9.0]]), mindspore.float32)
+        >>> output = ops.auto_generate.elu_ext(x)
+        >>> print(output)
+        [[-0.63212055  4.         -0.99966455]
+         [ 2.         -0.99326205  9.        ]]
     """
-    return sum_impl(input, dim, keepdim, dtype)
+    return elu_impl(input, alpha)
 
 
-def logsumexp(input, dim, keepdim=False):
+def acosh(input):
     r"""
-    Computes the logarithm of the sum of exponentiations of all elements along the specified `dim` dimension of the `input` (with numerical stabilization), and retains the dimension based on the `keepdim` parameter.
+    Computes inverse hyperbolic cosine of the inputs element-wise.
     
     .. math::
     
-        logsumexp(input) = \log(\sum(e^{input-input_{max}})) + input_{max}
-    
-    .. warning::
-        This is an experimental API that is subject to change or deletion.
-    
-    Args:
-        input (Tensor): Input Tensor.
-        dim (Union[int, tuple(int), list(int)], optional): The dimension to be reduced (the value should be within `[0, len(input.shape) - 1]`), when the `dim` is `()`, all dimensions are reduced.
-        keepdim (bool, optional): Whether the output tensor retains the dimension `dim`, default: `False`.
-    
-    Returns:
-        Tensor, the dtype changes according to the `input.dtype`, and the shape changes according to the values of `dim` and `keepdim`.
-    
-        - If `input.dtype` is in [float16, float32, bfloat16], the output dtype is the same as the `input.dtype`.
-        - If `input.dtype` is an integer or boolean type, the output dtype is float32.
-        - If `dim` is (), and `keepdim` is False, the output is a 0-D tensor representing the logarithm of the sum of exponentiations of all elements in the `input` tensor.
-        - If `dim` is `1`, and `keepdim` is False, the shape of output is :math:`(input.shape[0], input.shape[2], ..., input.shape[n])`.
-        - If `dim` is `(1, 2)`, and `keepdim` is False, the shape of output is :math:`(input.shape[0], input.shape[3], ..., input.shape[n])`.
-    
-    Raises:
-        TypeError: If `input` is not a Tensor.
-        TypeError: If dtype of `input` is not one of: bool, int8, int16, int32, int64, uint8, float16, float32, bfloat16.
-        TypeError: If `dim` is not an int or tuple(int) or list(list).
-        TypeError: If `keepdim` is not a bool.
-        ValueError: If the value of any elements of `dim` is not in the range `[0, len(input.shape) - 1]`.
-        RuntimeError: If any element of `dim` is repeated.
-    
-    Supported Platforms:
-        ``Ascend``
+        out_i = \cosh^{-1}(input_i)
     
-    Examples:
-        >>> import numpy as np
-        >>> from mindspore import Tensor, ops
-        >>> x = Tensor(np.random.randn(3, 4, 5, 6).astype(np.float32))
-        >>> output = ops.auto_generate.logsumexp_ext(x, 1, keepdim=True)
-        >>> print(output.shape)
-        (3, 1, 5, 6)
-    """
-    return logsumexp_impl(input, dim, keepdim)
-
-
-def trace(input):
-    r"""
-    Returns a new tensor that is the sum of the `input` main trace.
+    .. note::
+        Given an input tensor input, the function computes inverse hyperbolic cosine of every element.
+        Input range is [1, inf].
     
     Args:
-        input (Tensor): 2-D Tensor.
+        input (Tensor): The input tensor of inverse hyperbolic cosine function.
     
     Returns:
-        Tensor, when the data type of `input` is integer or bool, its data type is int64, otherwise it is the same as `input`, and size equals to 1.
+        Tensor, has the same shape as `input`. The dtype of output is float32 when dtype of `input` is in [bool, int8, uint8, int16, int32, int64]. Otherwise output has the same dtype as `input`.
     
     Raises:
         TypeError: If `input` is not a Tensor.
-        ValueError: If the dimension of `input` is not equal to 2.
-        TypeError:  If the dtype of `input` is not one of float16, float32, float64, bool, uint8, int8, int16, int32, int64, complex64, complex128, bfloat16.
     
     Supported Platforms:
-        ``Ascend``
+        ``Ascend`` ``GPU`` ``CPU``
     
     Examples:
         >>> import mindspore
         >>> import numpy as np
         >>> from mindspore import Tensor, ops
-        >>> input = Tensor(np.array([[10, 11, 12], [13, 14, 15], [16, 17, 18]]), mindspore.float32)
-        >>> output = ops.trace_ext(input)
-        >>> print(output)
-        42.0
-        >>> input = Tensor(np.arange(1, 13).reshape(3, 4), mindspore.float32)
-        >>> output = ops.trace_ext(input)
-        >>> print(output)
-        18.0
-        >>> input = Tensor(np.arange(12, 0, -1).reshape(4, 3), mindspore.float32)
-        >>> output = ops.trace_ext(input)
+        >>> input = Tensor(np.array([1.0, 1.5, 3.0, 100.0]), mindspore.float32)
+        >>> output = ops.acosh_ext(input)
         >>> print(output)
-        24.0
+        [0.        0.9624236 1.7627472 5.298292 ]
     """
-    return trace_impl(input)
+    return acosh_impl(input)
 
 
 def selu(input):
@@ -985,172 +708,207 @@ def selu(input):
     
     Examples:
         >>> import mindspore
-        >>> from mindspore import Tensor, mint
+        >>> from mindspore import Tensor, ops
         >>> import numpy as np
         >>> input = Tensor(np.array([[-1.0, 4.0, -8.0], [2.0, -5.0, 9.0]]), mindspore.float32)
-        >>> output = mint.nn.functional.selu(input)
+        >>> output = ops.auto_generate.selu_ext(input)
         >>> print(output)
         [[-1.1113307 4.202804 -1.7575096]
-        [ 2.101402 -1.7462534 9.456309 ]]
+         [ 2.101402 -1.7462534 9.456309 ]]
     """
     return selu_impl(input)
 
 
-def bmm(input, mat2):
+def mse_loss(input, target, reduction='mean'):
     r"""
-    Performs batch matrix-matrix multiplication of two three-dimensional tensors.
+    Calculates the mean squared error between the predicted value and the label value.
     
-    .. math::
-        \text{output}= \text{input} @ \text{mat2}
+    For detailed information, please refer to :class:`mindspore.nn.MSELoss`.
     
     Args:
-        input (Tensor): The first batch of matrices to be multiplied. Must be a three-dimensional tensor of shape `(b, n, m)`.
-        mat2 (Tensor): The second batch of matrices to be multiplied. Must be a three-dimensional tensor of shape `(b, m, p)`.
+        input (Tensor): Tensor of any dimension. The data type needs to be consistent with the `target`.
+            It should also be broadcastable with the `target`.
+        target (Tensor): The input label. Tensor of any dimension. The data type needs to be consistent with the `input`.
+            It should also be broadcastable with the `input`.
+        reduction (str, optional): Apply specific reduction method to the output: ``'mean'`` , ``'none'`` ,
+            ``'sum'`` . Default: ``'mean'`` .
+    
+            - ``'none'``: no reduction will be applied.
+            - ``'mean'``: compute and return the mean of elements in the output.
+            - ``'sum'``: the output elements will be summed.
     
     Returns:
-        Tensor, the output tensor of shape `(b, n, p)`, where each matrix is the product of the corresponding matrices in the input batches.
+        - Tensor. If `reduction` is ``'mean'`` or ``'sum'``, the shape of output is `Tensor Scalar`.
+        - If reduction is ``'none'``, the shape of output is the broadcasted shape of **input** and **target** .
     
     Raises:
-        ValueError: If `input` or `mat2` is not three-dimensional tensors.
-        ValueError: If the length of the third dimension of `input` is not equal to the length of the second dimension of `mat2`.
-        ValueError: If the batch size of the inputs is not equal to the batch size of the mat2.
+        ValueError: If `reduction` is not one of ``'mean'`` , ``'sum'`` or ``'none'``.
+        ValueError: If `input` and `target` are not broadcastable.
+        TypeError: If `input` and `target` are in different data type.
     
     Supported Platforms:
-        ``Ascend`` ``GPU`` ``CPU``
+        ``Ascend``
     
     Examples:
         >>> import mindspore
         >>> import numpy as np
-        >>> from mindspore import Tensor
-        >>> from mindspore import ops
-        >>> a = Tensor(np.ones(shape=[2, 3, 4]), mindspore.float32)
-        >>> b = Tensor(np.ones(shape=[2, 4, 5]), mindspore.float32)
-        >>> output = ops.auto_generate.bmm_ext(a, b)
+        >>> from mindspore import Tensor, ops
+        >>> logits = Tensor(np.array([1, 2, 3]), mindspore.float32)
+        >>> labels = Tensor(np.array([[1, 1, 1], [1, 2, 2]]), mindspore.float32)
+        >>> output = ops.mse_loss_ext(logits, labels, reduction='none')
         >>> print(output)
-        [[[4. 4. 4. 4. 4.]
-          [4. 4. 4. 4. 4.]
-          [4. 4. 4. 4. 4.]]
-         [[4. 4. 4. 4. 4.]
-          [4. 4. 4. 4. 4.]
-          [4. 4. 4. 4. 4.]]]
+        [[0. 1. 4.]
+         [0. 0. 1.]]
     """
-    return bmm_impl(input, mat2)
+    return mse_loss_impl(input, target, converted_reduction)
 
 
-def matrix_inverse(input):
+def t(input):
     r"""
-    Compute the inverse of the input matrix.
+    Transpose the input tensor.
+    
+    .. warning::
+        This is an experimental API that is subject to change or deletion.
     
     Args:
-        input (Tensor): A matrix to be calculated. Input `input` must be at least two dimensions, and the size of
-            the last two dimensions must be the same size.
+        input (Tensor): The input tensor.
     
     Returns:
-        Tensor, has the same type and shape as input`.
+        Tensor, transpose 2D tensor, return 1D tensor as it is.
     
     Raises:
-        TypeError: If `input` is not a Tensor.
-        ValueError: If the size of the last two dimensions of `input` is not the same.
-        ValueError: If the dimension of `input` is 1.
+        ValueError: If the dimension of `input` is greater than 2.
+        ValueError: If `input` is empty.
+        TypeError: If `input` is not a tensor.
     
     Supported Platforms:
         ``Ascend``
     
     Examples:
+        >>> import mindspore
+        >>> import numpy as np
         >>> from mindspore import Tensor, ops
+        >>> input = Tensor(np.array([[1, 2, 3], [4, 5, 6]]), mindspore.float32)
+        >>> output = ops.t_ext(input)
+        >>> print(output)
+        [[ 1. 4.]
+         [ 2. 5.]
+         [ 3. 6.]]
+    """
+    return t_impl(input)
+
+
+def isneginf(input):
+    r"""
+    Determines which elements are -inf for each position.
+    
+    .. warning::
+        - This API can be used only on the Atlas A2 training series.
+    
+    Args:
+        input (Tensor): Input Tensor.
+    
+    Returns:
+        Tensor with the same shape as the input, where elements are `True` if the corresponding element in the `input` is negative infinity, and `False` otherwise.
+    
+    Raises:
+        TypeError: If the input is not a tensor.
+    
+    Supported Platforms:
+        ``Ascend`` ``GPU`` ``CPU``
+    
+    Examples:
+        >>> from mindspore import ops, Tensor
         >>> from mindspore import dtype as mstype
-        >>> x = Tensor([[1., 2.], [3., 4.]], mstype.float32)
-        >>> print(ops.matrix_inverse_ext(x))
-        [[-2.   1. ]
-         [ 1.5 -0.5]]
+        >>> output = ops.isneginf(Tensor([[-float("inf"), float("inf")], [1, -float("inf")]], mstype.float32))
+        >>> print(output)
+        [[ True False]
+         [False  True]]
     """
-    return matrix_inverse_impl(input)
+    return isneginf_impl(input)
 
 
-def adaptive_avg_pool3d(input, output_size):
+def sort(input, dim=-1, descending=False, stable=False):
     r"""
     None
     """
-    return adaptive_avg_pool3d_impl(input, output_size)
+    return sort_impl(input, dim, descending, stable)
 
 
-def histc(input, bins=100, min=0, max=0):
+def softplus(input, beta=1, threshold=20):
     r"""
-    Computes the histogram of a tensor.
+    Applies softplus function to `input` element-wise.
     
-    The elements are sorted into equal width bins between `min` and `max`.
-    If `min` and `max` are both zero, the minimum and maximum values of the data are used.
+    The softplus function is shown as follows, x is the element of `input` :
     
-    Elements lower than min or higher than max are ignored.
+    .. math::
     
-    .. warning::
-        This is an experimental API that is subject to change or deletion.
-        If input is int64, valid values fit within int32; exceeding this may cause precision errors.
+        \text{output} = \frac{1}{beta}\log(1 + \exp(\text{beta * x}))
+    
+    where :math:`input * beta > threshold`, the implementation converts to the linear function to ensure numerical stability.
     
     Args:
-        input (Tensor): the input tensor.
-        bins (int, optional): Number of histogram bins, optional. If specified, must be positive. Default: ``100`` .
-        min (int, float, optional): the lower end of the range (inclusive), optional. Default: ``0`` .
-        max (int, float, optional): the upper end of the range (inclusive), optional. Default: ``0`` .
+        input (Tensor): Tensor of any dimension. Supported dtypes: 
+    
+            - Ascend: float16, float32, bfloat16.
+        beta (number.Number, optional): Scaling parameters in the softplus function. Default: ``1`` .
+        threshold (number.Number, optional): For numerical stability, the softplus function is converted 
+            to a threshold parameter of a linear function. Default: ``20`` .
     
     Returns:
-        A 1-D Tensor, has the same type as `input` with the shape :math:`(bins, )`.
+        Tensor, with the same type and shape as the input.
     
     Raises:
         TypeError: If `input` is not a Tensor.
-        TypeError: If `input` datatype is not in support list.
-        TypeError: If attr `min` or `max` is not float or int.
-        TypeError: If attr `bins` is not int.
-        ValueError: If attr value `min` > `max`.
-        ValueError: If attr `bins` <= 0.
+        TypeError: If dtype of `input` is not float16, float32, bfloat16.
     
     Supported Platforms:
-        ``Ascend``
+        ``Ascend`` 
     
     Examples:
+        >>> import mindspore
+        >>> import numpy as np
         >>> from mindspore import Tensor, ops
-        >>> x = Tensor([1., 2, 1])
-        >>> y = ops.histc_ext(x, bins=4, min=0, max=3)
-        >>> print(y)
-        [0 2 1 0]
+        >>> input = Tensor(np.array([0.1, 0.2, 30, 25]), mindspore.float32)
+        >>> output = ops.auto_generate.softplus_ext(input)
+        >>> print(output)
+        [0.74439657 0.7981388 30. 25.]
     """
-    return histc_impl(input, bins, min, max)
+    return softplus_impl(input, beta, threshold)
 
 
-def acosh(input):
+def frac(input):
     r"""
-    Computes inverse hyperbolic cosine of the inputs element-wise.
+    Calculates the fractional part of each element in the input.
     
     .. math::
+        out_i = input_i - \lfloor |input_i| \rfloor * sgn(input_i)
     
-        out_i = \cosh^{-1}(input_i)
-    
-    .. note::
-        Given an input tensor input, the function computes inverse hyperbolic cosine of every element.
-        Input range is [1, inf].
+    .. warning::
+        This is an experimental API that is subject to change or deletion.
     
     Args:
-        input (Tensor): The input tensor of inverse hyperbolic cosine function.
+        input (Tensor): The input Tensor.
     
     Returns:
-        Tensor, has the same shape as `input`. The dtype of output is float32 when dtype of `input` is in [bool, int8, uint8, int16, int32, int64]. Otherwise output has the same dtype as `input`.
+        Tensor, has the same shape and type as input.
     
     Raises:
         TypeError: If `input` is not a Tensor.
     
     Supported Platforms:
-        ``Ascend`` ``GPU`` ``CPU``
+        ``Ascend``
     
     Examples:
         >>> import mindspore
         >>> import numpy as np
         >>> from mindspore import Tensor, ops
-        >>> input = Tensor(np.array([1.0, 1.5, 3.0, 100.0]), mindspore.float32)
-        >>> output = ops.acosh_ext(input)
+        >>> x = Tensor([2, 4.2, -2.5], mindspore.float16)
+        >>> output = ops.frac_ext(x)
         >>> print(output)
-        [0.        0.9624236 1.7627472 5.298292 ]
+          [ 0.      0.1992 -0.5   ]
     """
-    return acosh_impl(input)
+    return frac_impl(input)
 
 
 def bincount(input, weights=None, minlength=0):
@@ -1186,35 +944,53 @@ def bincount(input, weights=None, minlength=0):
         ``Ascend``
     
     Examples:
-        >>> from mindspore import mint, Tensor
-        >>> print(mint.bincount(Tensor(np.arange(5))))
+        >>> from mindspore import ops, Tensor
+        >>> print(ops.auto_generate.bincount_ext(Tensor(np.arange(5))))
         [1 1 1 1 1]
-        >>> print(mint.bincount(Tensor(np.array([0, 1, 1, 3, 2, 1, 7]))))
+        >>> print(ops.auto_generate.bincount_ext(Tensor(np.array([0, 1, 1, 3, 2, 1, 7]))))
         [1 3 1 1 0 0 0 1]
         >>> w = Tensor(np.array([0.3, 0.5, 0.2, 0.7, 1., -0.6])) # weights
         >>> x = Tensor(np.array([0, 1, 1, 2, 2, 2]))
-        >>> print(mint.bincount(x,  weights=w, minlength=5))
+        >>> print(ops.auto_generate.bincount_ext(x,  weights=w, minlength=5))
         [0.3 0.7 1.1 0.  0. ]
     """
     return bincount_impl(input, weights, minlength)
 
 
-def asinh(input):
+def avg_pool1d(input, kernel_size, stride=None, padding=0, ceil_mode=False, count_include_pad=True):
     r"""
-    Computes inverse hyperbolic sine of the input element-wise.
+    Applies a 1D average pooling over an input Tensor which can be regarded as a composition of 1D input planes.
+    
+    Typically the input is of shape :math:`(N_{in}, C_{in}, L_{in})`, avg_pool1d outputs regional average in the
+    :math:`(L_{in})`-dimension. Given kernel size as :math:`ks = l_{ker}` and `stride` as :math:`s = s_0`, the
+    operation is as follows.
     
     .. math::
+        \text{output}(N_i, C_j, l) = \frac{1}{l_{ker}} \sum_{n=0}^{l_{ker}-1}
+        \text{input}(N_i, C_j, s_0 \times l + n)
     
-        out_i = \sinh^{-1}(input_i)
+    .. warning::
+        This is an experimental API that is subject to change or deletion.
     
     Args:
-        input (Tensor): The input tensor of inverse hyperbolic sine function.
+        input (Tensor): Tensor of shape :math:`(N, C_{in}, L_{in})`.
+        kernel_size (Union(int, tuple[int])): The size of kernel window used to take the average value.
+        stride (Union(int, tuple[int]), optional): The distance of kernel moving. `stride` can either be an int
+            number or a tuple of one int number. Default: ``None``, the same value as `kernel_size`.
+        padding (Union(int, tuple[int]), optional): The pad length to be filled. `padding` can either be an integer
+            or a tuple of one integer. Default: ``0`` .
+        ceil_mode (bool, optional): If True, apply ceil instead of floor to compute the output shape. Default: ``False``.
+        count_include_pad (bool, optional): If True, include the zero-padding in the averaging calculation. Default: ``True`` .
     
     Returns:
-        Tensor, has the same shape as `input`. The dtype of output is float32 when dtype of `input` is in [bool, int8, uint8, int16, int32, int64]. Otherwise output has the same dtype as `input`.
+        Tensor of shape :math:`(N, C_{in}, L_{out})`.
     
     Raises:
         TypeError: If `input` is not a Tensor.
+        TypeError: If `kernel_size` or `stride` is not an int.
+        TypeError: If `ceil_mode` or `count_include_pad` is not a bool.
+        ValueError: If `kernel_size` or `stride` is less than `1`.
+        ValueError: If `kernel_size` or `stride` or `padding` is not int nor a tuple whose length is greater than `1`.
     
     Supported Platforms:
         ``Ascend`` ``GPU`` ``CPU``
@@ -1223,69 +999,127 @@ def asinh(input):
         >>> import mindspore
         >>> import numpy as np
         >>> from mindspore import Tensor, ops
-        >>> input = Tensor(np.array([-5.0, 1.5, 3.0, 100.0]), mindspore.float32)
-        >>> output = ops.asinh_ext(input)
-        >>> print(output)
-        [-2.3124385  1.1947632  1.8184465  5.298342 ]
+        >>> input_x = Tensor(np.random.randint(0, 10, [1, 3, 6]), mindspore.float32)
+        >>> output = ops.auto_generate.avg_pool1d_ext(input_x, kernel_size=6, stride=1)
+        >>> print(output.shape)
+        (1, 3, 1)
     """
-    return asinh_impl(input)
+    return avg_pool1d_impl(input, kernel_size, stride, padding, ceil_mode, count_include_pad)
 
 
-def t(input):
+def mean(input, dim=None, keepdim=False, dtype=None):
     r"""
-    Transpose the input tensor.
+    Reduces all dimension of a tensor by averaging all elements in the dimension, by default.
+    And reduce a dimension of `input` along the specified `dim`. `keepdim`
+    determines whether the dimensions of the output and input are the same.
     
-    .. warning::
-        This is an experimental API that is subject to change or deletion.
+    Note:
+        The `dim` with tensor type is only used for compatibility with older versions and is not recommended.
     
     Args:
-        input (Tensor): The input tensor.
+        input (Tensor[Number]): The input tensor. The dtype of the tensor to be reduced is number.
+            :math:`(N, *)` where :math:`*` means, any number of additional dimensions.
+        dim (Union[int, tuple(int), list(int), Tensor]): The dimensions to reduce. Default: ``None`` ,
+            reduce all dimensions. Only constant value is allowed. Assume the rank of `input` is r,
+            and the value range is [-r,r).
+        keepdim (bool): If ``True`` , keep these reduced dimensions and the length is 1.
+            If ``False`` , don't keep these dimensions. Default: ``False`` .
+        dtype (:class:`mindspore.dtype`): The desired data type of returned Tensor. Default: ``None`` .
     
     Returns:
-        Tensor, transpose 2D tensor, return 1D tensor as it is.
+        Tensor, has the same data type as input tensor.
+    
+        - If `dim` is ``None`` , and `keepdim` is ``False`` ,
+          the output is a 0-D tensor representing the product of all elements in the input tensor.
+        - If `dim` is int, set as 1, and `keepdim` is ``False`` ,
+          the shape of output is :math:`(x_0, x_2, ..., x_R)`.
+        - If `dim` is tuple(int), set as (1, 2), and `keepdim` is ``False`` ,
+          the shape of output is :math:`(x_0, x_3, ..., x_R)`.
+        - If `dim` is 1-D Tensor, set as [1, 2], and `keepdim` is ``False`` ,
+          the shape of output is :math:`(x_0, x_3, ..., x_R)`.
     
     Raises:
-        ValueError: If the dimension of `input` is greater than 2.
-        ValueError: If `input` is empty.
-        TypeError: If `input` is not a tensor.
+        TypeError: If `x` is not a Tensor.
+        TypeError: If `dim` is not one of the following: int, tuple, list or Tensor.
+        TypeError: If `keepdim` is not a bool.
+        ValueError: If `dim` is out of range.
     
     Supported Platforms:
-        ``Ascend``
+        ``Ascend`` ``GPU`` ``CPU``
     
     Examples:
         >>> import mindspore
         >>> import numpy as np
         >>> from mindspore import Tensor, ops
-        >>> input = Tensor(np.array([[1, 2, 3], [4, 5, 6]]), mindspore.float32)
-        >>> output = ops.t_ext(input)
+        >>> x = Tensor(np.random.randn(3, 4, 5, 6).astype(np.float32))
+        >>> output = ops.mean_ext(x, 1, keepdim=True)
+        >>> result = output.shape
+        >>> print(result)
+        (3, 1, 5, 6)
+        >>> # case 1: Reduces a dimension by averaging all elements in the dimension.
+        >>> x = Tensor(np.array([[[2, 2, 2, 2, 2, 2], [2, 2, 2, 2, 2, 2], [2, 2, 2, 2, 2, 2]],
+        ... [[4, 4, 4, 4, 4, 4], [5, 5, 5, 5, 5, 5], [6, 6, 6, 6, 6, 6]],
+        ... [[6, 6, 6, 6, 6, 6], [8, 8, 8, 8, 8, 8], [10, 10, 10, 10, 10, 10]]]),
+        ... mindspore.float32)
+        >>> output = ops.mean_ext(x)
         >>> print(output)
-        [[ 1. 4.]
-         [ 2. 5.]
-         [ 3. 6.]]
+        5.0
+        >>> print(output.shape)
+        ()
+        >>> # case 2: Reduces a dimension along the dim 0
+        >>> output = ops.mean_ext(x, 0, True)
+        >>> print(output)
+        [[[4. 4. 4. 4. 4. 4.]
+        [5. 5. 5. 5. 5. 5.]
+        [6. 6. 6. 6. 6. 6.]]]
+        >>> # case 3: Reduces a dimension along the dim 1
+        >>> output = ops.mean_ext(x, 1, True)
+        >>> print(output)
+        [[[2. 2. 2. 2. 2. 2.]]
+        [[5. 5. 5. 5. 5. 5.]]
+        [[8. 8. 8. 8. 8. 8.]]]
+        >>> # case 4: Reduces a dimension along the dim 2
+        >>> output = ops.mean_ext(x, 2, True)
+        >>> print(output)
+        [[[ 2.]
+        [ 2.]
+        [ 2.]]
+        [[ 4.]
+        [ 5.]
+        [ 6.]]
+        [[ 6.]
+        [ 8.]
+        [10.]]]
     """
-    return t_impl(input)
+    return mean_impl(input, dim, keepdim, dtype)
 
 
-def log2(input):
+def atan2(input, other):
     r"""
-    Returns the logarithm to the base 2 of a tensor element-wise.
+    Returns arctangent of input/other element-wise.
     
-    .. math::
-        y_i = \log_2(x_i)
+    It returns :math:`\theta\ \in\ [-\pi, \pi]`
+    such that :math:`input = r*\sin(\theta), other = r*\cos(\theta)`, where :math:`r = \sqrt{input^2 + other^2}`.
     
-    .. warning::
-        - If the input value of operator Log2 is within the range (0, 0.01] or [0.95, 1.05], the output accuracy
-          may be affacted.
+    Note:
+        - Arg `input` and `other` comply with the implicit type conversion rules to make the data types consistent.
+          If they have different data types, the lower precision data type will be converted to relatively the
+          highest precision data type.
     
     Args:
-        input (Tensor): Input Tensor of any dimension. The value must be greater than 0.
+        input (Tensor, Number.number): The input tensor or scalar.
+        other (Tensor, Number.number): The input tensor or scalar. It has the same shape with `input` or
+            its shape is able to broadcast with `input`.
     
     Returns:
-        Tensor, has the same shape as the `input`. If `input.dtype` is of integer or boolean type, the output dtype
-        will be float32. Otherwise, the output dtype will be the same as `input.dtype`.
+        Tensor, the shape is the same as the one after broadcasting.
+        The dtype of output is float32 when dtype of `input` is in
+        [bool, int8, uint8, int16, int32, int64]. Otherwise output has the same dtype as `input`.
     
     Raises:
-        TypeError: If `input` is not a Tensor.
+        TypeError: If `input` or `other` is not a Tensor or scalar.
+        RuntimeError: If the data type of `input` and `other` conversion of Parameter is required
+                    when data type conversion of Parameter is not supported.
     
     Supported Platforms:
         ``Ascend``
@@ -1293,127 +1127,53 @@ def log2(input):
     Examples:
         >>> import mindspore
         >>> import numpy as np
-        >>> from mindspore import Tensor, mint
-        >>> x = Tensor(np.array([3.0, 5.0, 7.0]), mindspore.float32)
-        >>> output = mint.log2(x)
+        >>> from mindspore import Tensor, ops
+        >>> input = Tensor(np.array([0, 1]), mindspore.float32)
+        >>> other = Tensor(np.array([1, 1]), mindspore.float32)
+        >>> output = ops.auto_generate.atan2_ext(input, other)
         >>> print(output)
-        [1.5849625 2.321928  2.807355 ]
-    """
-    return log2_impl(input)
-
-
-def cumsum(input, dim, dtype=None):
-    r"""
-    Computes the cumulative sum of input Tensor along `dim`.
-    
-    .. math::
-    
-        y_i = x_1 + x_2 + x_3 + ... + x_i
-    
-    Args:
-        input (Tensor): The input Tensor.
-        dim (int): Dim along which the cumulative sum is computed.
-        dtype (:class:`mindspore.dtype`, optional): The desired dtype of returned Tensor. If specified,
-            the input Tensor will be cast to `dtype` before the computation. This is useful for preventing overflows.
-            If not specified, stay the same as original Tensor. Default: ``None`` .
-    
-    Returns:
-        Tensor, the shape of the output Tensor is consistent with the input Tensor's.
-    
-    Raises:
-        TypeError: If `input` is not a Tensor.
-        ValueError: If the `dim` is out of range.
-    
-    Supported Platforms:
-        ``Ascend``
-    
-    Examples:
-        >>> import numpy as np
-        >>> from mindspore import Tensor
-        >>> import mindspore.ops as ops
-        >>> x = Tensor(np.array([[3, 4, 6, 10], [1, 6, 7, 9], [4, 3, 8, 7], [1, 3, 7, 9]]).astype(np.float32))
-        >>> # case 1: along the dim 0
-        >>> y = ops.auto_generate.cumsum_ext(x, 0)
-        >>> print(y)
-        [[ 3.  4.  6. 10.]
-        [ 4. 10. 13. 19.]
-        [ 8. 13. 21. 26.]
-        [ 9. 16. 28. 35.]]
-        >>> # case 2: along the dim 1
-        >>> y = ops.auto_generate.cumsum_ext(x, 1)
-        >>> print(y)
-        [[ 3.  7. 13. 23.]
-        [ 1.  7. 14. 23.]
-        [ 4.  7. 15. 22.]
-        [ 1.  4. 11. 20.]]
+        [0.        0.7853982]
     """
-    return cumsum_impl(input, dim, dtype)
+    return atan2_impl(input, other)
 
 
-def fold(input, output_size, kernel_size, dilation=1, padding=0, stride=1):
+def stack(tensors, dim=0):
     r"""
-    Combines an array of sliding local blocks into a large containing tensor.
-    
-    Consider a batched input tensor of shape :math:`(N, C \times \prod(\text{kernel_size}), L)` ,
-    where :math:`N` is the batch dimension, :math:`C \times \prod(\text{kernel_size})` is the
-    total number of values within each block (a block has :math:`\prod(\text{kernel_size})` spatial
-    locations each containing a `C`-channeled vector), and :math:`L` is the total number of such blocks:
-    
-    .. math::
-        L = \prod_d \left\lfloor\frac{\text{output_size}[d] + 2 \times \text{padding}[d] %
-            - \text{dilation}[d] \times (\text{kernel_size}[d] - 1) - 1}{\text{stride}[d]} + 1\right\rfloor,
-    
-    where :math:`d` is over all spatial dimensions.
-    
-    Therefore, `output_size` is the spatial shape of the large containing tensor of the sliding local blocks.
+    Stacks a list of tensors in specified dim.
     
-    The `dilation`, `padding` and `stride` arguments specify how the sliding blocks are retrieved.
+    Stacks the list of input tensors with the same rank `R`, output is a tensor of rank `(R+1)`.
     
-    .. warning::
-        Currently, only unbatched(3D) or batched(4D) image-like output tensors are supported.
+    Given input tensors of shape :math:`(x_1, x_2, ..., x_R)`. Set the number of input tensors as `N`.
+    If :math:`dim \ge 0`, the shape of the output tensor is
+    :math:`(x_1, x_2, ..., x_{dim}, N, x_{dim+1}, ..., x_R)`.
     
     Args:
-        input (Tensor): 2-D or 3-D Tensor.
-        output_size (Union[int, tuple[int], list[int]]): The shape of the spatial dimensions of
-            the output(i.e., output.shape[2:]).
-        kernel_size (Union[int, tuple[int], list[int]]): The size of the kernel, should be two int
-            for height and width. If type is int, it means that height equal with width. Must be specified.
-        dilation (Union[int, tuple[int], list[int]], optional): The size of the dilation, should be two int
-            for height and width. If type is int, it means that height equal with width. Default: ``1`` .
-        padding (Union[int, tuple[int], list[int]], optional): The size of the padding, should be two int
-            for height and width. If type is int, it means that height equal with width. Default: ``0`` .
-        stride (Union[int, tuple[int], list[int]], optional): The size of the stride, should be two int
-            for height and width. If type is int, it means that height equal with width. Default: ``1`` .
+        tensors (Union[tuple, list]): A Tuple or list of Tensor objects with the same shape and type.
+        dim (int, optional): Dimension to stack. The range is [-(R+1), R+1). Default: ``0`` .
     
     Returns:
-        A Tensor, with same type as `input` .
-    
-    Shape:
-        - Input: :math:`(N, C \times \prod(\text{kernel_size}), L)` or
-          :math:`(C \times \prod(\text{kernel_size}), L)`
-        - Output: :math:`(N, C, output\_size[0], output\_size[1], ...)` or
-          :math:`(C, output\_size[0], output\_size[1], ...)`
+        Tensor. A stacked Tensor with the same type as `tensors`.
     
     Raises:
-        TypeError: If `output_size`, `kernel_size`, `stride`, `dilation`, `padding` data type is not int, tuple or list.
-        ValueError: If `output_size`, `kernel_size`, `dilation`, `stride` value is not
-            greater than zero or elements number invalid.
-        ValueError: If `padding` value is less than zero or elements number invalid.
-        ValueError: If input.shape[-2] can't be divisible by the product of kernel_size.
-        ValueError: If `input.shape[-1]` is not equal to the calculated number of sliding blocks `L`.
+        TypeError: If the data types of elements in `tensors` are not the same.
+        ValueError: If `dim` is out of the range [-(R+1), R+1);
+                    or if the shapes of elements in `tensors` are not the same.
     
     Supported Platforms:
         ``Ascend``
     
     Examples:
-        >>> import numpy as np
+        >>> import mindspore
         >>> from mindspore import Tensor, ops
-        >>> x = Tensor(np.random.rand(16, 64, 25).astype(np.float32))
-        >>> output = ops.auto_generate.fold_ext(x, (8, 8), [2, 2], [2, 2], [2, 2], [2, 2])
-        >>> print(output.shape)
-        (16, 16, 8, 8)
+        >>> import numpy as np
+        >>> data1 = Tensor(np.array([0, 1]).astype(np.float32))
+        >>> data2 = Tensor(np.array([2, 3]).astype(np.float32))
+        >>> output = ops.auto_generate.stack_ext([data1, data2], 0)
+        >>> print(output)
+        [[0. 1.]
+         [2. 3.]]
     """
-    return fold_impl(input, converted_output_size, converted_kernel_size, converted_dilation, converted_padding, converted_stride)
+    return stack_impl(tensors, dim)
 
 
 def sub_tensor_(input, other, alpha=1):
@@ -1423,210 +1183,125 @@ def sub_tensor_(input, other, alpha=1):
     return sub_tensor_impl(input, other, alpha)
 
 
-def index_add(input, dim, index, source, alpha=1):
+def bmm(input, mat2):
     r"""
-    Accumulate the elements of `alpha` times `source` into the `input` by adding to the index in the order given in `index`. For example, if ``dim == 0`` , ``index[i] == j`` , and ``alpha = -1`` , then the `i` th row of `source` is subtracted from the `j` th row of `input` . The `dim` th dimension of `source` must have the same size as the length of `index` , and all other dimensions must match `input`, or an error will be raised. For a 3-D tensor, the output is defined as follows:
+    Performs batch matrix-matrix multiplication of two three-dimensional tensors.
     
     .. math::
-        \begin{array}{ll}
-        input[index[i],\ :,\ :]\ +=\ alpha * source[i,\ :,\ :]  \qquad \#if\ dim == 0 \\
-        input[:,\ \ index[i],\ :]\ +=\ alpha * source[:,\ \ i,\ :]  \qquad \#if\ dim == 1 \\
-        input[:,\ :,\ \ index[i]]\ +=\ alpha * source[:,\ :,\ \ i]  \qquad\#if\ dim == 2 \\
-        \end{array}
-        
-    .. warning::
-        This is an experimental API that is subject to change or deletion.
-        
-    Args:
-        input (Tensor): The input Tensor.
-        dim (int): The dimension along which to index.
-        index (Tensor): Add the value of "input Tensor" and `source` along the dimension of the `dim` according to the specified index value, with data type int32. The `index` must be 1D with the same size as the size of `source` in the `dim` dimension. The values of `index` should be in [0, b), where the b is the size of "input Tensor" in the `dim` dimension.
-        source (Tensor): The input tensor with the value to add. Must have same data type as "input Tensor". The shape must be the same as "input Tensor" except the `dim` th dimension.
-        alpha (number, optional): The scalar multiplier for source. Default: ``1``.
-        
-    Returns:
-        Tensor, has the same shape and dtype as `input`.
-    
-    Raises:
-        TypeError: If neither `index` nor `source` is a Tensor.
-        ValueError: If the value of `dim` is out of the dimension range of `source` shape.
-        ValueError: If `index` rank is not the same as `source` rank.
-        ValueError: If shape of `index` is not 1D or size of `index` is not equal to dimension of source[dim].
-        ValueError: If the shape of `source` is not the same as that of `input` except the `dim` axis.
-    
-    Supported Platforms:
-        ``Ascend``
-    
-    Examples:
-        >>> import numpy as np
-        >>> import mindspore
-        >>> from mindspore import Tensor, ops
-        >>> x = Tensor(np.array([[1, 2, 3], [4, 5, 6], [7, 8, 9]]), mindspore.float32)
-        >>> index = Tensor(np.array([0, 2]), mindspore.int32)
-        >>> y = Tensor(np.array([[0.5, 1.0], [1.0, 1.5], [2.0, 2.5]]), mindspore.float32)
-        >>> output = ops.auto_generate.index_add_ext(x, 1, index, y, alpha=1)
-        >>> print(output)
-        [[ 1.5  2.   4. ]
-         [ 5.   5.   7.5]
-         [ 9.   8.  11.5]]
-    """
-    return index_add_impl(input, dim, index, source, alpha)
-
-
-def argmax(input, dim=None, keepdim=False):
-    r"""
-    argmax(input) -> Tensor
-    
-    Return the indices of the maximum values of a tensor.
-    
-    Args:
-        input (Tensor): Input tensor.
-    
-    Returns:
-        Tensor.
-    
-    Supported Platforms:
-        ``Ascend``
-    
-    Examples:
-        >>> import numpy as np
-        >>> from mindspore import Tensor
-        >>> from mindspore import ops
-        >>> x = Tensor(np.array([[1, 20, 5], [67, 8, 9], [130, 24, 15]]).astype(np.float32))
-        >>> output = ops.auto_generate.argmax_ext(x)
-        >>> print(output)
-        6
-    
-    .. function:: argmax(input, dim, keepdim=False) -> Tensor
-        :noindex:
-    
-    Return the indices of the maximum values of a tensor across a dimension.
+        \text{output}= \text{input} @ \text{mat2}
     
     Args:
-        input (Tensor): Input tensor.
-        dim (int): The dimension to reduce. 
-        keepdim (bool, optional): Whether the output tensor retains the specified
-            dimension. Default: ``False`` .
+        input (Tensor): The first batch of matrices to be multiplied. Must be a three-dimensional tensor of shape `(b, n, m)`.
+        mat2 (Tensor): The second batch of matrices to be multiplied. Must be a three-dimensional tensor of shape `(b, m, p)`.
     
     Returns:
-        Tensor, indices of the maximum values across a dimension.
+        Tensor, the output tensor of shape `(b, n, p)`, where each matrix is the product of the corresponding matrices in the input batches.
     
     Raises:
-        TypeError: If `keepdim` is not bool.
-        ValueError: If `dim` is out of range.
+        ValueError: If `input` or `mat2` is not three-dimensional tensors.
+        ValueError: If the length of the third dimension of `input` is not equal to the length of the second dimension of `mat2`.
+        ValueError: If the batch size of the inputs is not equal to the batch size of the mat2.
     
     Supported Platforms:
-        ``Ascend``
+        ``Ascend`` ``GPU`` ``CPU``
     
     Examples:
+        >>> import mindspore
         >>> import numpy as np
         >>> from mindspore import Tensor
         >>> from mindspore import ops
-        >>> x = Tensor(np.array([[1, 20, 5], [67, 8, 9], [130, 24, 15]]).astype(np.float32))
-        >>> output = ops.auto_generate.argmax_ext(x, dim=-1)
+        >>> a = Tensor(np.ones(shape=[2, 3, 4]), mindspore.float32)
+        >>> b = Tensor(np.ones(shape=[2, 4, 5]), mindspore.float32)
+        >>> output = ops.auto_generate.bmm_ext(a, b)
         >>> print(output)
-        [1 0 0]
+        [[[4. 4. 4. 4. 4.]
+          [4. 4. 4. 4. 4.]
+          [4. 4. 4. 4. 4.]]
+         [[4. 4. 4. 4. 4.]
+          [4. 4. 4. 4. 4.]
+          [4. 4. 4. 4. 4.]]]
     """
-    return argmax_impl(input, dim, keepdim)
+    return bmm_impl(input, mat2)
 
 
-def add(input, other, alpha=1):
+def index_select(input, dim, index):
     r"""
-    Adds scaled other value to input Tensor.
-    
-    .. math::
-    
-        out_{i} = input_{i} + alpha \times other_{i}
+    Generates a new Tensor that accesses the values of `input` along the specified `dim` dimension
+    using the indices specified in `index`. The new Tensor has the same number of dimensions as `input`,
+    with the size of the `dim` dimension being equal to the length of `index`, and the size of all other
+    dimensions will be unchanged from the original `input` Tensor.
     
-    Note:
-        - When the two inputs have different shapes,
-          they must be able to broadcast to a common shape.
-        - The two inputs and alpha comply with the implicit type conversion rules to make the data types
-          consistent.
+    .. note::
+        The value of index must be in the range of `[0, input.shape[dim])`, the result is undefined out of range.
     
     Args:
-        input (Union[Tensor, number.Number, bool]): The first input is a number.Number or
-            a bool or a tensor whose data type is
-            `number <https://www.mindspore.cn/docs/en/master/api_python/mindspore/mindspore.dtype.html>`_ or
-            `bool_ <https://www.mindspore.cn/docs/en/master/api_python/mindspore/mindspore.dtype.html>`_.
-        other (Union[Tensor, number.Number, bool]): The second input, is a number.Number or
-            a bool or a tensor whose data type is
-            `number <https://www.mindspore.cn/docs/en/master/api_python/mindspore/mindspore.dtype.html>`_ or
-            `bool_ <https://www.mindspore.cn/docs/en/master/api_python/mindspore/mindspore.dtype.html>`_.
-        alpha (number.Number): A scaling factor applied to `other`, default 1.
+        input (Tensor): The input Tensor.
+        dim (int): The dimension to be indexed.
+        index (Tensor): A 1-D Tensor with the indices.
     
     Returns:
-        Tensor with a shape that is the same as the broadcasted shape of the input `input` and `other`,
-        and the data type is the one with higher precision or higher digits among the two inputs and alpha.
+        Tensor, has the same dtype as input Tensor.
     
     Raises:
-        TypeError: If the type of `input`, `other`, or `alpha` is not one of the following: Tensor, number.Number, bool.
-        TypeError: If `alpha` is of type float but `input` and `other` are not of type float.
-        TypeError: If `alpha` is of type bool but `input` and `other` are not of type bool.
+        TypeError: If `input` or `index` is not a Tensor.
+        TypeError: If `dim` is not int number.
+        ValueError: If the value of `dim` is out the range of `[-input.ndim, input.ndim - 1]`.
+        ValueError: If the dimension of `index` is not equal to 1.
     
     Supported Platforms:
-        ``Ascend`` ``GPU`` ``CPU``
+        ``Ascend``
     
     Examples:
-        >>> import numpy as np
         >>> import mindspore
-        >>> from mindspore import Tensor
-        >>> from mindspore import ops
-        >>> x = Tensor(1, mindspore.int32)
-        >>> y = Tensor(np.array([4, 5, 6]).astype(np.float32))
-        >>> alpha = 0.5
-        >>> output = ops.auto_generate.add_ext(x, y, alpha)
-        >>> print(output)
-        [3. 3.5 4.]
-        >>> # the data type of x is int32, the data type of y is float32,
-        >>> # alpha is a float, and the output is the data format of higher precision float32.
-        >>> print(output.dtype)
-        Float32
-    """
-    return add_impl(input, other, alpha)
-
-
-def tril(input, diagonal=0):
-    r"""
-    None
+        >>> from mindspore import Tensor, ops
+        >>> import numpy as np
+        >>> input = Tensor(np.arange(16).astype(np.float32).reshape(2, 2, 4))
+        >>> print(input)
+        [[[ 0.  1.  2.  3.]
+        [ 4.  5.  6.  7.]]
+        [[ 8.  9. 10. 11.]
+        [12. 13. 14. 15.]]]
+        >>> index = Tensor([0,], mindspore.int32)
+        >>> y = ops.auto_generate.index_select_ext(input, 1, index)
+        >>> print(y)
+        [[[ 0.  1.  2.  3.]]
+        [[ 8.  9. 10. 11.]]]
     """
-    return tril_impl(input, diagonal)
+    return index_select_impl(input, dim, index)
 
 
-def max_unpool2d(input, indices, kernel_size, stride=None, padding=0, output_size=None):
+def logsumexp(input, dim, keepdim=False):
     r"""
-    Computes the inverse of `max_pool2d`.
-    
-    `max_unpool2d` keeps the maximal value and set all position of non-maximal values to zero. Typically the input is of shape :math:`(N, C, H_{in}, W_{in})` or :math:`(C, H_{in}, W_{in})`, and the output is of shape :math:`(N, C, H_{out}, W_{out})` or :math:`(C, H_{out}, W_{out})`. The operation is as follows.
+    Computes the logarithm of the sum of exponentiations of all elements along the specified `dim` dimension of the `input` (with numerical stabilization), and retains the dimension based on the `keepdim` parameter.
     
     .. math::
-        \begin{array}{ll} \\
-        H_{out} = (H_{in} - 1) \times stride[0] - 2 \times padding[0] + kernel\_size[0] \\
-        W_{out} = (W_{in} - 1) \times stride[1] - 2 \times padding[1] + kernel\_size[1] \\
-        \end{array}
+    
+        logsumexp(input) = \log(\sum(e^{input-input_{max}})) + input_{max}
     
     .. warning::
         This is an experimental API that is subject to change or deletion.
     
     Args:
-        input (Tensor): The input Tensor to invert. Tensor of shape :math:`(N, C, H_{in}, W_{in})` or :math:`(C, H_{in}, W_{in})`.
-        indices (Tensor): Max values' index represented by the indices. Tensor of shape must be same with input 'input'. Values of indices must belong to :math:`[0, H_{in} \times W_{in} - 1]`. Data type must be in int32 or int64.
-        kernel_size (Union[int, tuple[int]]): The size of kernel used to take the maximum value, an int number that represents height and width of the kernel, or a tuple of two int numbers that represent height and width respectively.
-        stride (Union[int, tuple[int]], optional): The distance of kernel moving, an int number that represents the height and width of movement are both stride, or a tuple of two int numbers that represent height and width of movement respectively. Default: ``None`` , which indicates the moving step is `kernel_size` .
-        padding (Union[int, tuple[int]], optional): The pad value to be filled. Default: ``0`` . If `padding` is an integer, the paddings of height and width are the same, equal to padding. If `padding` is a tuple of two integers, the padding of height and width equal to padding[0] and padding[1] correspondingly.
-        output_size (tuple[int], optional): The target output size. Default: ``None`` . If output_size == (), then the shape of output computed by `kernel_size`, `stride` and `padding`. If output_size != (), then output_size must be :math:`(N, C, H, W)` , :math:`(C, H, W)` or :math:`(H, W)` and output_size must belong to :math:`[(N, C, H_{out} - stride[0], W_{out} - stride[1]), (N, C, H_{out} + stride[0], W_{out} + stride[1])]`.
+        input (Tensor): Input Tensor.
+        dim (Union[int, tuple(int), list(int)], optional): The dimension to be reduced (the value should be within `[0, len(input.shape) - 1]`), when the `dim` is `()`, all dimensions are reduced.
+        keepdim (bool, optional): Whether the output tensor retains the dimension `dim`, default: `False`.
     
     Returns:
-        Tensor, with shape :math:`(N, C, H_{out}, W_{out})` or :math:`(C, H_{out}, W_{out})`, with the same data type with `input`.
+        Tensor, the dtype changes according to the `input.dtype`, and the shape changes according to the values of `dim` and `keepdim`.
+    
+        - If `input.dtype` is in [float16, float32, bfloat16], the output dtype is the same as the `input.dtype`.
+        - If `input.dtype` is an integer or boolean type, the output dtype is float32.
+        - If `dim` is (), and `keepdim` is False, the output is a 0-D tensor representing the logarithm of the sum of exponentiations of all elements in the `input` tensor.
+        - If `dim` is `1`, and `keepdim` is False, the shape of output is :math:`(input.shape[0], input.shape[2], ..., input.shape[n])`.
+        - If `dim` is `(1, 2)`, and `keepdim` is False, the shape of output is :math:`(input.shape[0], input.shape[3], ..., input.shape[n])`.
     
     Raises:
-        TypeError: If data type of `input` or `indices` is not supported.
-        TypeError: If `kernel_size`, `stride` or `padding` is neither an int nor a tuple.
-        ValueError: If numbers in `stride`, `padding` or `kernel_size` are not positive.
-        ValueError: If the shapes of `input` and `indices` are different.
-        ValueError: If the length of `input` is not 3 or 4.
-        ValueError: If the type of `output_size` is not tuple.
-        ValueError: If `output_size` is not close to output size computed by attr `kernel_size`, `stride`, `padding`.
+        TypeError: If `input` is not a Tensor.
+        TypeError: If dtype of `input` is not one of: bool, int8, int16, int32, int64, uint8, float16, float32, bfloat16.
+        TypeError: If `dim` is not an int or tuple(int) or list(list).
+        TypeError: If `keepdim` is not a bool.
+        ValueError: If the value of any elements of `dim` is not in the range `[0, len(input.shape) - 1]`.
+        RuntimeError: If any element of `dim` is repeated.
     
     Supported Platforms:
         ``Ascend``
@@ -1634,78 +1309,72 @@ def max_unpool2d(input, indices, kernel_size, stride=None, padding=0, output_siz
     Examples:
         >>> import numpy as np
         >>> from mindspore import Tensor, ops
-        >>> input = Tensor(np.array([[[[0, 1], [8, 9]]]]).astype(np.float32))
-        >>> indices = Tensor(np.array([[[[0, 1], [2, 3]]]]).astype(np.int64))
-        >>> output = ops.max_unpool2d_ext(input, indices, 1, stride=1, padding=0)
-        >>> print(output.asnumpy())
-        [[[[0. 1.]
-           [8. 9.]]]]
+        >>> x = Tensor(np.random.randn(3, 4, 5, 6).astype(np.float32))
+        >>> output = ops.auto_generate.logsumexp_ext(x, 1, keepdim=True)
+        >>> print(output.shape)
+        (3, 1, 5, 6)
     """
-    return max_unpool2d_impl(input, indices, kernel_size, stride, padding, output_size)
+    return logsumexp_impl(input, dim, keepdim)
 
 
-def softplus(input, beta=1, threshold=20):
+def argsort(input, dim=-1, descending=False, stable=False):
     r"""
-    Applies softplus function to `input` element-wise.
-    
-    The softplus function is shown as follows, x is the element of `input` :
-    
-    .. math::
-    
-        \text{output} = \frac{1}{beta}\log(1 + \exp(\text{beta * x}))
+    Sorts the input tensor along the given dimension in specified order and return the sorted indices.
     
-    where :math:`input * beta > threshold`, the implementation converts to the linear function to ensure numerical stability.
+    .. warning::
+            This is an experimental optimizer API that is subject to change.
     
     Args:
-        input (Tensor): Tensor of any dimension. Supported dtypes: 
-    
-            - Ascend: float16, float32, bfloat16.
-        beta (number.Number, optional): Scaling parameters in the softplus function. Default: ``1`` .
-        threshold (number.Number, optional): For numerical stability, the softplus function is converted 
-            to a threshold parameter of a linear function. Default: ``20`` .
+        input(Tensor): The input tensor to sort.
+        dim (int, optional): The dim to sort along. Default: ``-1`` , means the last dimension.
+            The Ascend backend only supports sorting the last dimension.
+        descending (bool, optional): The sort order. If `descending` is ``True`` then the elements
+            are sorted in descending order by value. Otherwise sort in ascending order. Default: ``False`` .
+        stable (bool, optional): Whether to use stable sorting algorithm. Default: ``False``.
     
     Returns:
-        Tensor, with the same type and shape as the input.
+        Tensor, the indices of sorted input tensor. Data type is int64.
     
     Raises:
-        TypeError: If `input` is not a Tensor.
-        TypeError: If dtype of `input` is not float16, float32, bfloat16.
+        ValueError: If `dim` is out of range.
+        TypeError: If dtype of `dim` is not int32.
+        TypeError: If dtype of `descending` is not bool.
+        TypeError: If dtype of `stable` is not bool.
     
     Supported Platforms:
-        ``Ascend`` 
+        ``Ascend``
     
     Examples:
         >>> import mindspore
         >>> import numpy as np
         >>> from mindspore import Tensor, ops
-        >>> input = Tensor(np.array([0.1, 0.2, 30, 25]), mindspore.float32)
-        >>> output = ops.auto_generate.softplus_ext(input)
-        >>> print(output)
-        [0.74439657 0.7981388 30. 25.]
+        >>> x = Tensor(np.array([[8, 2, 1], [5, 9, 3], [4, 6, 7]]), mindspore.float16)
+        >>> sort = ops.auto_generate.argsort_ext(x)
+        >>> print(sort)
+        [[2 1 0]
+         [2 0 1]
+         [0 1 2]]
     """
-    return softplus_impl(input, beta, threshold)
+    return argsort_impl(input, dim, descending, stable)
 
 
-def log10(input):
+def log2(input):
     r"""
-    Returns the logarithm to the base 10 of a tensor element-wise.
+    Returns the logarithm to the base 2 of a tensor element-wise.
     
     .. math::
-        y_i = \log_{10}(x_i)
+        y_i = \log_2(x_i)
     
     .. warning::
-        - This is an experimental API that is subject to change or deletion.
-        - If the input value of operator Log10 is within the range (0, 0.01] or [0.95, 1.05], the output accuracy
+        - If the input value of operator Log2 is within the range (0, 0.01] or [0.95, 1.05], the output accuracy
           may be affacted.
     
     Args:
         input (Tensor): Input Tensor of any dimension. The value must be greater than 0.
     
     Returns:
-        Tensor, has the same shape as the `input`, and the dtype changes according to the `input.dtype`.
-        
-        - if `input.dtype` is in [float16, float32, float64, bfloat16], the output dtype is the same as the `input.dtype`.
-        - if `input.dtype` is integer or boolean type, the output dtype is float32.
+        Tensor, has the same shape as the `input`. If `input.dtype` is of integer or boolean type, the output dtype
+        will be float32. Otherwise, the output dtype will be the same as `input.dtype`.
     
     Raises:
         TypeError: If `input` is not a Tensor.
@@ -1716,216 +1385,260 @@ def log10(input):
     Examples:
         >>> import mindspore
         >>> import numpy as np
-        >>> from mindspore import Tensor, mint
+        >>> from mindspore import Tensor, ops
         >>> x = Tensor(np.array([3.0, 5.0, 7.0]), mindspore.float32)
-        >>> output = mint.log10(x)
+        >>> output = ops.auto_generate.log2_ext(x)
         >>> print(output)
-        [0.47712136 0.69897    0.845098  ]
+        [1.5849625 2.321928  2.807355 ]
     """
-    return log10_impl(input)
+    return log2_impl(input)
 
 
-def l1_loss(input, target, reduction='mean'):
+def add(input, other, alpha=1):
     r"""
-    Calculate the mean absolute error between the `input` value and the `target` value.
-    
-    Assuming that the :math:`x` and :math:`y` are the predicted value and target value,
-    both are one-dimensional tensors of length :math:`N`, length :math:`N`, `reduction` is set to ``'none'`` ,
-    then calculate the loss of :math:`x` and :math:`y` without dimensionality reduction.
-    
-    The formula is as follows:
+    Adds scaled other value to input Tensor.
     
     .. math::
-        \ell(x, y) = L = \{l_1,\dots,l_N\}^\top, \quad \text{with } l_n = \left| x_n - y_n \right|,
-    
-    where :math:`N` is the batch size.
     
-    If `reduction` is ``'mean'`` or ``'sum'`` , then:
+        out_{i} = input_{i} + alpha \times other_{i}
     
-    .. math::
-        \ell(x, y) =
-        \begin{cases}
-            \operatorname{mean}(L), & \text{if reduction} = \text{'mean';}\\
-            \operatorname{sum}(L),  & \text{if reduction} = \text{'sum'.}
-        \end{cases}
+    Note:
+        - When the two inputs have different shapes,
+          they must be able to broadcast to a common shape.
+        - The two inputs and alpha comply with the implicit type conversion rules to make the data types
+          consistent.
     
     Args:
-        input (Tensor): Predicted value, Tensor of any dimension.
-        target (Tensor): Target value, usually has the same shape as the `input`.
-            If `input` and `target` have different shapes, make sure they can broadcast to each other.
-        reduction (str, optional): Apply specific reduction method to the output: ``'none'`` , ``'mean'`` ,
-            ``'sum'`` . Default: ``'mean'`` .
-    
-            - ``'none'``: no reduction will be applied.
-            - ``'mean'``: compute and return the mean of elements in the output. Notice: At least one of the input and target is float type when the reduction is ``'mean'`` .
-            - ``'sum'``: the output elements will be summed.
+        input (Union[Tensor, number.Number, bool]): The first input is a number.Number or
+            a bool or a tensor whose data type is
+            `number <https://www.mindspore.cn/docs/en/master/api_python/mindspore/mindspore.dtype.html>`_ or
+            `bool <https://www.mindspore.cn/docs/en/master/api_python/mindspore/mindspore.dtype.html>`_.
+        other (Union[Tensor, number.Number, bool]): The second input, is a number.Number or
+            a bool or a tensor whose data type is
+            `number <https://www.mindspore.cn/docs/en/master/api_python/mindspore/mindspore.dtype.html>`_ or
+            `bool <https://www.mindspore.cn/docs/en/master/api_python/mindspore/mindspore.dtype.html>`_.
+        alpha (number.Number): A scaling factor applied to `other`, default 1.
     
     Returns:
-        Tensor or Scalar, if `reduction` is ``'none'`` , return a Tensor with same shape and dtype as `input`.
-        Otherwise, a scalar value will be returned.
+        Tensor with a shape that is the same as the broadcasted shape of the input `input` and `other`,
+        and the data type is the one with higher precision or higher digits among the two inputs and alpha.
     
     Raises:
-        TypeError: If `input` is not a Tensor.
-        TypeError: If `target` is not a Tensor.
-        ValueError: If `reduction` is not one of ``'none'`` , ``'mean'`` or ``'sum'`` .
+        TypeError: If the type of `input`, `other`, or `alpha` is not one of the following: Tensor, number.Number, bool.
+        TypeError: If `alpha` is of type float but `input` and `other` are not of type float.
+        TypeError: If `alpha` is of type bool but `input` and `other` are not of type bool.
     
     Supported Platforms:
-        ``Ascend``
+        ``Ascend`` ``GPU`` ``CPU``
     
     Examples:
-        >>> from mindspore import Tensor, ops
-        >>> from mindspore import dtype as mstype
-        >>> x = Tensor([[1, 2, 3], [4, 5, 6]], mstype.float32)
-        >>> target = Tensor([[6, 5, 4], [3, 2, 1]], mstype.float32)
-        >>> output = ops.l1_loss_ext(x, target, reduction="mean")
+        >>> import numpy as np
+        >>> import mindspore
+        >>> from mindspore import Tensor
+        >>> from mindspore import ops
+        >>> x = Tensor(1, mindspore.int32)
+        >>> y = Tensor(np.array([4, 5, 6]).astype(np.float32))
+        >>> alpha = 0.5
+        >>> output = ops.auto_generate.add_ext(x, y, alpha)
         >>> print(output)
-        3.0
+        [3. 3.5 4.]
+        >>> # the data type of x is int32, the data type of y is float32,
+        >>> # alpha is a float, and the output is the data format of higher precision float32.
+        >>> print(output.dtype)
+        Float32
     """
-    return l1_loss_impl(input, target, converted_reduction)
+    return add_impl(input, other, alpha)
 
 
-def asin(input):
+def adaptive_avg_pool3d(input, output_size):
     r"""
-    Computes arcsine of input tensors element-wise.
+    None
+    """
+    return adaptive_avg_pool3d_impl(input, output_size)
+
+
+def sub(input, other, alpha=1):
+    r"""
+    Subtracts scaled other value from input Tensor.
     
     .. math::
     
-        out_i = \sin^{-1}(input_i)
+        out_{i} = input_{i} - alpha \times other_{i}
+    
+    Note:
+        - When the two inputs have different shapes,
+          they must be able to broadcast to a common shape.
+        - The two inputs and alpha comply with the implicit type conversion rules to make the data types
+          consistent.
     
     Args:
-        input (Tensor): The shape of tensor is
-            :math:`(N,*)`, where :math:`*` means any number of additional dimensions.
+        input (Union[Tensor, number.Number, bool]): The first input is a number.Number or
+            a bool or a tensor whose data type is
+            `number <https://www.mindspore.cn/docs/en/master/api_python/mindspore/mindspore.dtype.html>`_ or
+            `bool <https://www.mindspore.cn/docs/en/master/api_python/mindspore/mindspore.dtype.html>`_.
+        other (Union[Tensor, number.Number, bool]): The second input, is a number.Number or
+            a bool or a tensor whose data type is
+            `number <https://www.mindspore.cn/docs/en/master/api_python/mindspore/mindspore.dtype.html>`_ or
+            `bool <https://www.mindspore.cn/docs/en/master/api_python/mindspore/mindspore.dtype.html>`_.
+        alpha (number.Number): A scaling factor applied to `other`, default 1.
     
     Returns:
-        Tensor, has the same shape as `input`. The dtype of output is float32 when dtype of `input` is in [bool, int8, uint8, int16, int32, int64]. Otherwise output has the same dtype as `input`.
+        Tensor with a shape that is the same as the broadcasted shape of the input `input` and `other`,
+        and the data type is the one with higher precision or higher digits among the two inputs and alpha.
     
     Raises:
-        TypeError: If `input` is not a Tensor.
+        TypeError: If the type of `input`, `other`, or `alpha` is not one of the following: Tensor, number.Number, bool.
+        TypeError: If `alpha` is of type float but `input` and `other` are not of type float.
+        TypeError: If `alpha` is of type bool but `input` and `other` are not of type bool.
     
     Supported Platforms:
         ``Ascend`` ``GPU`` ``CPU``
     
     Examples:
-        >>> import mindspore
         >>> import numpy as np
-        >>> from mindspore import Tensor, ops
-        >>> input = Tensor(np.array([0.74, 0.04, 0.30, 0.56]), mindspore.float32)
-        >>> output = ops.asin_ext(input)
+        >>> import mindspore
+        >>> from mindspore import Tensor
+        >>> from mindspore import ops
+        >>> x = Tensor(np.array([4, 5, 6]).astype(np.float32))
+        >>> y = Tensor(1, mindspore.int32)
+        >>> alpha = 0.5
+        >>> output = ops.auto_generate.sub_ext(x, y, alpha)
         >>> print(output)
-        [0.8330927  0.04001068  0.30469266  0.59438497 ]
-    """
-    return asin_impl(input)
-
-
-def inplace_add(input, other, alpha=1):
-    r"""
-    None
+        [3.5 4.5 5.5]
+        >>> # the data type of x is float32, the data type of y is int32,
+        >>> # alpha is a float, and the output is the data format of higher precision float32.
+        >>> print(output.dtype)
+        Float32
     """
-    return inplace_add_impl(input, other, alpha)
+    return sub_impl(input, other, alpha)
 
 
-def transpose(input, dim0, dim1):
+def leaky_relu(input, negative_slope=0.01):
     r"""
-    Interchange two axes of a tensor.
+    leaky_relu activation function. The element of `input` less than 0 times `negative_slope` .
     
-    .. warning::
-        This is an experimental API that is subject to change or deletion.
+    The activation function is defined as:
+    
+    .. math::
+        \text{leaky_relu}(input) = \begin{cases}input, &\text{if } input \geq 0; \cr
+        \text{negative_slope} * input, &\text{otherwise.}\end{cases}
+    
+    where :math:`negative\_slope` represents the `negative_slope` parameter.
+    
+    For more details, see `Rectifier Nonlinearities Improve Neural Network Acoustic Models
+    <https://ai.stanford.edu/~amaas/papers/relu_hybrid_icml2013_final.pdf>`_.
+    
+    LeakyReLU Activation Function Graph:
+    
+    .. image:: ../images/LeakyReLU.png
+        :align: center
     
     Args:
-        input(Tensor): Input tensor.
-        dim0 (int): First axis.
-        dim1 (int): Second axis.
+        input (Tensor): The input of leaky_relu is a Tensor of any dimension.
+        negative_slope (Union[int, float], optional): Slope of the activation function when the element of `input` is less than 0.
+          Default: ``0.01`` .
     
     Returns:
-        Transposed tensor, has the same data type as `input`.
+        Tensor, has the same type and shape as the `input`.
     
     Raises:
-        TypeError: If argument `input` is not Tensor.
-        TypeError: If `dim0` or `dim1` is not integer.
-        ValueError: If `dim0` or `dim1` is not in the range of :math:`[-ndim, ndim-1]`.
+        TypeError: If `input` is not a Tensor.
+        TypeError: If `negative_slope` is not a float or an int.
     
     Supported Platforms:
         ``Ascend``
     
     Examples:
+        >>> import mindspore
         >>> import numpy as np
-        >>> from mindspore import mint
-        >>> from mindspore import Tensor
-        >>> input = Tensor(np.ones((2,3,4), dtype=np.float32))
-        >>> output = mint.transpose(input, 0, 2)
-        >>> print(output.shape)
-        (4, 3, 2)
+        >>> from mindspore import Tensor, ops
+        >>> input = Tensor(np.array([[-1.0, 4.0, -8.0], [2.0, -5.0, 9.0]]), mindspore.float32)
+        >>> print(ops.extend.leaky_relu_ext(input, negative_slope=0.2))
+        [[-0.2  4.  -1.6]
+         [ 2.  -1.   9. ]]
     """
-    return transpose_impl(input, dim0, dim1)
+    return leaky_relu_impl(input, negative_slope)
 
 
-def avg_pool1d(input, kernel_size, stride=None, padding=0, ceil_mode=False, count_include_pad=True):
+def max_unpool2d(input, indices, kernel_size, stride=None, padding=0, output_size=None):
     r"""
-    Applies a 1D average pooling over an input Tensor which can be regarded as a composition of 1D input planes.
+    Computes the inverse of `max_pool2d`.
     
-    Typically the input is of shape :math:`(N_{in}, C_{in}, L_{in})`, avg_pool1d outputs regional average in the
-    :math:`(L_{in})`-dimension. Given kernel size as :math:`ks = l_{ker}` and `stride` as :math:`s = s_0`, the
-    operation is as follows.
+    `max_unpool2d` keeps the maximal value and set all position of non-maximal values to zero. Typically the input is of shape :math:`(N, C, H_{in}, W_{in})` or :math:`(C, H_{in}, W_{in})`, and the output is of shape :math:`(N, C, H_{out}, W_{out})` or :math:`(C, H_{out}, W_{out})`. The operation is as follows.
     
     .. math::
-        \text{output}(N_i, C_j, l) = \frac{1}{l_{ker}} \sum_{n=0}^{l_{ker}-1}
-        \text{input}(N_i, C_j, s_0 \times l + n)
+        \begin{array}{ll} \\
+        H_{out} = (H_{in} - 1) \times stride[0] - 2 \times padding[0] + kernel\_size[0] \\
+        W_{out} = (W_{in} - 1) \times stride[1] - 2 \times padding[1] + kernel\_size[1] \\
+        \end{array}
     
     .. warning::
         This is an experimental API that is subject to change or deletion.
     
     Args:
-        input (Tensor): Tensor of shape :math:`(N, C_{in}, L_{in})`.
-        kernel_size (Union(int, tuple[int])): The size of kernel window used to take the average value.
-        stride (Union(int, tuple[int]), optional): The distance of kernel moving. `stride` can either be an int
-            number or a tuple of one int number. Default: ``None``, the same value as `kernel_size`.
-        padding (Union(int, tuple[int]), optional): The pad length to be filled. `padding` can either be an integer
-            or a tuple of one integer. Default: ``0`` .
-        ceil_mode (bool, optional): If True, apply ceil instead of floor to compute the output shape. Default: ``False``.
-        count_include_pad (bool, optional): If True, include the zero-padding in the averaging calculation. Default: ``True`` .
+        input (Tensor): The input Tensor to invert. Tensor of shape :math:`(N, C, H_{in}, W_{in})` or :math:`(C, H_{in}, W_{in})`.
+        indices (Tensor): Max values' index represented by the indices. Tensor of shape must be same with input 'input'. Values of indices must belong to :math:`[0, H_{in} \times W_{in} - 1]`. Data type must be in int32 or int64.
+        kernel_size (Union[int, tuple[int]]): The size of kernel used to take the maximum value, an int number that represents height and width of the kernel, or a tuple of two int numbers that represent height and width respectively.
+        stride (Union[int, tuple[int]], optional): The distance of kernel moving, an int number that represents the height and width of movement are both stride, or a tuple of two int numbers that represent height and width of movement respectively. Default: ``None`` , which indicates the moving step is `kernel_size` .
+        padding (Union[int, tuple[int]], optional): The pad value to be filled. Default: ``0`` . If `padding` is an integer, the paddings of height and width are the same, equal to padding. If `padding` is a tuple of two integers, the padding of height and width equal to padding[0] and padding[1] correspondingly.
+        output_size (tuple[int], optional): The target output size. Default: ``None`` . If output_size == (), then the shape of output computed by `kernel_size`, `stride` and `padding`. If output_size != (), then output_size must be :math:`(N, C, H, W)` , :math:`(C, H, W)` or :math:`(H, W)` and output_size must belong to :math:`[(N, C, H_{out} - stride[0], W_{out} - stride[1]), (N, C, H_{out} + stride[0], W_{out} + stride[1])]`.
     
     Returns:
-        Tensor of shape :math:`(N, C_{in}, L_{out})`.
+        Tensor, with shape :math:`(N, C, H_{out}, W_{out})` or :math:`(C, H_{out}, W_{out})`, with the same data type with `input`.
     
     Raises:
-        TypeError: If `input` is not a Tensor.
-        TypeError: If `kernel_size` or `stride` is not an int.
-        TypeError: If `ceil_mode` or `count_include_pad` is not a bool.
-        ValueError: If `kernel_size` or `stride` is less than `1`.
-        ValueError: If `kernel_size` or `stride` or `padding` is not int nor a tuple whose length is greater than `1`.
+        TypeError: If data type of `input` or `indices` is not supported.
+        TypeError: If `kernel_size`, `stride` or `padding` is neither an int nor a tuple.
+        ValueError: If numbers in `stride`, `padding` or `kernel_size` are not positive.
+        ValueError: If the shapes of `input` and `indices` are different.
+        ValueError: If the length of `input` is not 3 or 4.
+        ValueError: If the type of `output_size` is not tuple.
+        ValueError: If `output_size` is not close to output size computed by attr `kernel_size`, `stride`, `padding`.
     
     Supported Platforms:
-        ``Ascend`` ``GPU`` ``CPU``
+        ``Ascend``
     
     Examples:
-        >>> import mindspore
         >>> import numpy as np
-        >>> from mindspore import Tensor, mint
-        >>> input_x = Tensor(np.random.randint(0, 10, [1, 3, 6]), mindspore.float32)
-        >>> output = mint.nn.functional.avg_pool1d(input_x, kernel_size=6, stride=1)
-        >>> print(output.shape)
-        (1, 3, 1)
+        >>> from mindspore import Tensor, ops
+        >>> input = Tensor(np.array([[[[0, 1], [8, 9]]]]).astype(np.float32))
+        >>> indices = Tensor(np.array([[[[0, 1], [2, 3]]]]).astype(np.int64))
+        >>> output = ops.max_unpool2d_ext(input, indices, 1, stride=1, padding=0)
+        >>> print(output.asnumpy())
+        [[[[0. 1.]
+           [8. 9.]]]]
     """
-    return avg_pool1d_impl(input, kernel_size, stride, padding, ceil_mode, count_include_pad)
+    return max_unpool2d_impl(input, indices, kernel_size, stride, padding, output_size)
 
 
-def atan(input):
+def prod(input, dim=None, keepdim=False, dtype=None):
     r"""
-    Computes the trigonometric inverse tangent of the input element-wise.
-    
-    .. math::
-    
-        out_i = \tan^{-1}(input_i)
+    Reduces a dimension of a tensor by multiplying all elements in the dimension, by default. And also can
+    reduce a dimension of `input` along the `dim`. Determine whether the dimensions of the output and input are the
+    same by controlling `keepdim`.
     
     Args:
-        input (Tensor): The shape of tensor is
-            :math:`(N,*)` where :math:`*` means, any number of additional dimensions.
+        input (Tensor[Number]): The input tensor. The dtype of the tensor to be reduced is number.
+            :math:`(N, *)` where :math:`*` means, any number of additional dimensions.
+        dim (int): The dimensions to reduce. Default: ``None`` , reduce all dimensions.
+            Only constant value is allowed. Assume the rank of `input` is r, and the value range is [-r,r).
+        keepdim (bool): If ``True`` , keep these reduced dimensions and the length is 1.
+            If ``False`` , don't keep these dimensions. Default: ``False`` .
+        dtype (:class:`mindspore.dtype`): The desired data type of returned Tensor. Default: ``None`` .
     
     Returns:
-        Tensor, has the same shape as `input`. The dtype of output is float32 when dtype of `input` is in [bool, int8, uint8, int16, int32, int64]. Otherwise output has the same dtype as `input`.
+        Tensor, has the same data type as input tensor.
+    
+        - If `dim` is ``None`` , and `keepdim` is  ``False`` ,
+          the output is a 0-D tensor representing the product of all elements in the input tensor.
+        - If `dim` is int, set as 1, and `keepdim` is  ``False`` ,
+          the shape of output is :math:`(input_0, input_2, ..., input_R)`.
     
     Raises:
         TypeError: If `input` is not a Tensor.
+        TypeError: If `dim` is not one of the following: int or None.
+        TypeError: If `keepdim` is not a bool.
+        ValueError: If `dim` is out of range.
     
     Supported Platforms:
         ``Ascend`` ``GPU`` ``CPU``
@@ -1934,12 +1647,60 @@ def atan(input):
         >>> import mindspore
         >>> import numpy as np
         >>> from mindspore import Tensor, ops
-        >>> input = Tensor(np.array([1.0, 0.0]), mindspore.float32)
-        >>> output = ops.atan_ext(input)
+        >>> x = Tensor(np.random.randn(3, 4, 5, 6).astype(np.float32))
+        >>> output = ops.prod_ext(x, 1, keepdim=True)
+        >>> result = output.shape
+        >>> print(result)
+        (3, 1, 5, 6)
+        >>> # case 1: Reduces a dimension by multiplying all elements in the dimension.
+        >>> x = Tensor(np.array([[[1, 1, 1, 1, 1, 1], [2, 2, 2, 2, 2, 2], [3, 3, 3, 3, 3, 3]],
+        ...                      [[4, 4, 4, 4, 4, 4], [5, 5, 5, 5, 5, 5], [6, 6, 6, 6, 6, 6]],
+        ...                      [[7, 7, 7, 7, 7, 7], [8, 8, 8, 8, 8, 8], [9, 9, 9, 9, 9, 9]]]), mindspore.float32)
+        >>> output = ops.prod_ext(x)
         >>> print(output)
-        [0.7853982 0.       ]
+        2.2833798e+33
+        >>> print(output.shape)
+        ()
+        >>> # case 2: Reduces a dimension along dim 0.
+        >>> output = ops.prod_ext(x, 0, True)
+        >>> print(output)
+        [[[ 28.  28.  28.  28.  28.  28.]
+        [ 80.  80.  80.  80.  80.  80.]
+        [162. 162. 162. 162. 162. 162.]]]
+        >>> # case 3: Reduces a dimension along dim 1.
+        >>> output = ops.prod_ext(x, 1, True)
+        >>> print(output)
+        [[[  6.   6.   6.   6.   6.   6.]]
+        [[120. 120. 120. 120. 120. 120.]]
+        [[504. 504. 504. 504. 504. 504.]]]
+        >>> # case 4: Reduces a dimension along dim 2.
+        >>> output = ops.prod_ext(x, 2, True)
+        >>> print(output)
+        [[[1.00000e+00]
+        [6.40000e+01]
+        [7.29000e+02]]
+        [[4.09600e+03]
+        [1.56250e+04]
+        [4.66560e+04]]
+        [[1.17649e+05]
+        [2.62144e+05]
+        [5.31441e+05]]]
     """
-    return atan_impl(input)
+    return prod_impl(input, dim, keepdim, dtype)
+
+
+def adaptive_avg_pool2d_grad(grad_output, x):
+    r"""
+    None
+    """
+    return adaptive_avg_pool2d_grad_impl(grad_output, x)
+
+
+def tril(input, diagonal=0):
+    r"""
+    None
+    """
+    return tril_impl(input, diagonal)
 
 
 def topk(input, k, dim=-1, largest=True, sorted=True):
@@ -2015,116 +1776,182 @@ def topk(input, k, dim=-1, largest=True, sorted=True):
     return topk_impl(input, k, dim, largest, sorted)
 
 
-def index_select(input, dim, index):
+def atan(input):
     r"""
-    Generates a new Tensor that accesses the values of `input` along the specified `dim` dimension
-    using the indices specified in `index`. The new Tensor has the same number of dimensions as `input`,
-    with the size of the `dim` dimension being equal to the length of `index`, and the size of all other
-    dimensions will be unchanged from the original `input` Tensor.
+    Computes the trigonometric inverse tangent of the input element-wise.
     
-    .. note::
-        The value of index must be in the range of `[0, input.shape[dim])`, the result is undefined out of range.
+    .. math::
+    
+        out_i = \tan^{-1}(input_i)
     
     Args:
-        input (Tensor): The input Tensor.
-        dim (int): The dimension to be indexed.
-        index (Tensor): A 1-D Tensor with the indices.
+        input (Tensor): The shape of tensor is
+            :math:`(N,*)` where :math:`*` means, any number of additional dimensions.
     
     Returns:
-        Tensor, has the same dtype as input Tensor.
+        Tensor, has the same shape as `input`. The dtype of output is float32 when dtype of `input` is in [bool, int8, uint8, int16, int32, int64]. Otherwise output has the same dtype as `input`.
     
     Raises:
-        TypeError: If `input` or `index` is not a Tensor.
-        TypeError: If `dim` is not int number.
-        ValueError: If the value of `dim` is out the range of `[-input.ndim, input.ndim - 1]`.
-        ValueError: If the dimension of `index` is not equal to 1.
+        TypeError: If `input` is not a Tensor.
     
     Supported Platforms:
-        ``Ascend``
+        ``Ascend`` ``GPU`` ``CPU``
     
     Examples:
         >>> import mindspore
-        >>> from mindspore import Tensor, ops
         >>> import numpy as np
-        >>> input = Tensor(np.arange(16).astype(np.float32).reshape(2, 2, 4))
-        >>> print(input)
-        [[[ 0.  1.  2.  3.]
-        [ 4.  5.  6.  7.]]
-        [[ 8.  9. 10. 11.]
-        [12. 13. 14. 15.]]]
-        >>> index = Tensor([0,], mindspore.int32)
-        >>> y = ops.auto_generate.index_select_ext(input, 1, index)
-        >>> print(y)
-        [[[ 0.  1.  2.  3.]]
-        [[ 8.  9. 10. 11.]]]
+        >>> from mindspore import Tensor, ops
+        >>> input = Tensor(np.array([1.0, 0.0]), mindspore.float32)
+        >>> output = ops.atan_ext(input)
+        >>> print(output)
+        [0.7853982 0.       ]
     """
-    return index_select_impl(input, dim, index)
+    return atan_impl(input)
 
 
-def argsort(input, dim=-1, descending=False, stable=False):
+def cummin(input, dim):
     r"""
-    Sorts the input tensor along the given dimension in specified order and return the sorted indices.
+    Returns a tuple (values, indices) where `values` is the cumulative minimum value of input Tensor `input`
+    along the dimension `dim`, and `indices` is the index location of each minimum value.
     
-    .. warning::
-            This is an experimental optimizer API that is subject to change.
+    .. math::
+        \begin{array}{ll} \\
+            y_{i} = \min(x_{1}, x_{2}, ... , x_{i})
+        \end{array}
+    
+    .. note::
+        GE backend is not supported in Ascend.
     
     Args:
-        input(Tensor): The input tensor to sort.
-        dim (int, optional): The dim to sort along. Default: ``-1`` , means the last dimension.
-            The Ascend backend only supports sorting the last dimension.
-        descending (bool, optional): The sort order. If `descending` is ``True`` then the elements
-            are sorted in descending order by value. Otherwise sort in ascending order. Default: ``False`` .
-        stable (bool, optional): Whether to use stable sorting algorithm. Default: ``False``.
+        input (Tensor): The input Tensor, The dimension must be greater than 0.
+        dim (int): Operation dimension. The value of `dim` must be in the range `[-input.ndim, input.ndim - 1]`.
     
     Returns:
-        Tensor, the indices of sorted input tensor. Data type is int64.
+        tuple [Tensor], tuple of 2 Tensors, containing the cumulative minimum of elements and the index.
+        The shape of each output tensor is the same as that of input `input`.
     
     Raises:
-        ValueError: If `dim` is out of range.
-        TypeError: If dtype of `dim` is not int32.
-        TypeError: If dtype of `descending` is not bool.
-        TypeError: If dtype of `stable` is not bool.
+        TypeError: If `input` is not a Tensor.
+        TypeError: If `input` is a Tensor, but the type is complex or bool.
+        TypeError: If `dim` is not an int.
+        ValueError: If `dim` is out the range of `[-input.ndim, input.ndim - 1]`.
     
     Supported Platforms:
         ``Ascend``
     
     Examples:
+        >>> from mindspore import Tensor, ops
         >>> import mindspore
-        >>> import numpy as np
-        >>> from mindspore import Tensor
-        >>> import mindspore.mint as mint
-        >>> x = Tensor(np.array([[8, 2, 1], [5, 9, 3], [4, 6, 7]]), mindspore.float16)
-        >>> sort = mint.argsort(x)
-        >>> print(sort)
-        [[2 1 0]
-        [2 0 1]
-        [0 1 2]]
+        >>> a = Tensor([-0.2284, -0.6628,  0.0975,  0.2680, -1.3298, -0.4220], mindspore.float32)
+        >>> output = ops.cummin_ext(a, dim=0)
+        >>> print(output[0])
+        [-0.2284 -0.6628 -0.6628 -0.6628 -1.3298 -1.3298]
+        >>> print(output[1])
+        [0 1 1 1 4 4]
     """
-    return argsort_impl(input, dim, descending, stable)
+    return cummin_impl(input, dim)
 
 
-def outer(input, vec2):
+def mm(input, mat2):
     r"""
-    Return outer product of `input` and `vec2`. If `input` is a vector of size :math:`n`
-    and `vec2` is a vector of size :math:`m` , then output must be a matrix of shape :math:`(n, m)` .
+    Returns the matrix product of two arrays.
+    If `input` is a :math:`(n \times m)` Tensor, `mat2` is a
+    :math:`(m \times p)` Tensor, `out` will be a :math:`(n \times p)` Tensor.
+    
+    Note:
+        This function cannot support broadcasting.
+        Refer to :func:`mindspore.ops.matmul` instead if you need a broadcastable function.
     
     .. warning::
         This is an experimental API that is subject to change or deletion.
     
-    .. note::
-        This function does not broadcast.
+    Args:
+        input (Tensor): The first matrix of matrix multiplication.
+            The last dimension of `input` must be the same size as the first dimension of `mat2`.
+        mat2 (Tensor): The second matrix of matrix multiplication.
+            The last dimension of `input` must be the same size as the first dimension of `mat2`.
+    
+    Returns:
+        Tensor, the matrix product of the inputs.
+    
+    Raises:
+        ValueError: If the last dimension of `input` is not the same size as the
+            second-to-last dimension of `mat2`.
+        TypeError: If `input` or `mat2` is not a Tensor.
+        TypeError: If dtype of `input` or `mat2` is not float16, float32 or bfloat16.
+    
+    Supported Platforms:
+        ``Ascend``
+    
+    Examples:
+        >>> import mindspore as ms
+        >>> from mindspore import ops
+        >>> import numpy as np
+        >>> x1 = ms.Tensor(np.random.rand(2, 3), ms.float32)
+        >>> x2 = ms.Tensor(np.random.rand(3, 4), ms.float32)
+        >>> out = ops.mm_ext(x1, x2)
+        >>> print(out.shape)
+        (2, 4)
+    """
+    return mm_impl(input, mat2)
+
+
+def unfold(input, kernel_size, dilation=1, padding=0, stride=1):
+    r"""
+    Extracts sliding local blocks from a batched input tensor.
+    
+    Consider a batched input tensor of shape :math:`(N, C, *)`,
+    where :math:`N` is the batch dimension, :math:`C` is the channel dimension,
+    and :math:`*` represent arbitrary spatial dimensions. This operation flattens
+    each sliding `Kernel_size`- sized block within the spatial dimensions
+    of `input` into a column (i.e., last dimension) of a 3-D output
+    tensor of shape :math:`(N, C \times \prod(\text{kernel_size}), L)`, where
+    :math:`C \times \prod(\text{kernel_size})` is the total number of values
+    within each block (a block has :math:`\prod(\text{kernel_size})` spatial
+    locations each containing a `C`-channeled vector), and :math:`L` is
+    the total number of such blocks:
+    
+    .. math::
+        L = \prod_d \left\lfloor\frac{\text{spatial_size}[d] + 2 \times \text{padding}[d] %
+            - \text{dilation}[d] \times (\text{kernel_size}[d] - 1) - 1}{\text{stride}[d]} + 1\right\rfloor,
+    
+    where :math:`\text{spatial_size}` is formed by the spatial dimensions
+    of `input` (:math:`*` above), and :math:`d` is over all spatial
+    dimensions.
+    
+    Therefore, indexing `output` at the last dimension (column dimension)
+    gives all values within a certain block.
+    
+    The `dilation`, `padding` and `stride` arguments specify
+    how the sliding blocks are retrieved.
+    
+    .. warning::
+        - Currently, batched(4D) image-like tensors are supported.
+        - For Ascend, it is only supported on platforms above Atlas A2.
     
     Args:
-        input (Tensor): 1-D input vector.
-        vec2 (Tensor): 1-D input vector.
+        input (Tensor): 4-D Tensor.
+        kernel_size (Union[int, tuple[int], list[int]]): The size of the kernel, should be two int
+            for height and width. If type is int, it means that height equal with width. Must be specified.
+        dilation (Union[int, tuple[int], list[int]], optional): The dilation of the window, should be two int
+            for height and width. If type is int, it means that height equal with width. Default: ``1`` .
+        padding (Union[int, tuple[int], list[int]], optional): The pad of the window, should be two int
+            for height and width. If type is int, it means that height equal with width. Default: ``0`` .
+        stride (Union[int, tuple[int], list[int]], optional): The stride of the window, should be two int
+            for height and width. If type is int, it means that height equal with width. Default: ``1`` .
     
     Returns:
-        out, 2-D matrix, the outer product of two vectors.
+        A Tensor, with same type as `input` .
+    
+    Shape:
+        - Input: :math:`(N, C, *)`
+        - Output: :math:`(N, C \times \prod(\text{kernel_size}), L)`
     
     Raises:
-        TypeError: If `input` or `vec2` is not a Tensor.
-        TypeError: The implicitly converted data types of `input` and `vec2` are not one of float16, float32, float64, bool, uint8, int8, int16, int32, int64, complex64, complex128, bfloat16
-        ValueError: If the dimension of `input` or `vec2` is not equal to 1.
+        TypeError: If any data type of `kernel_size`, `stride`, `dilation`, `padding` is not int, tuple or list.
+        ValueError: If `kernel_size`, `dilation`, `stride` value is not
+            greater than zero or elements number more than `2`.
+        ValueError: If `padding` value is less than zero.
     
     Supported Platforms:
         ``Ascend``
@@ -2132,174 +1959,182 @@ def outer(input, vec2):
     Examples:
         >>> import mindspore
         >>> import numpy as np
-        >>> from mindspore import Tensor
-        >>> from mindspore import ops
-        >>> input = Tensor(np.array([7, 8, 9]), mindspore.int32)
-        >>> vec2 = Tensor(np.array([7, 10, 11]), mindspore.int32)
-        >>> out = ops.outer(input, vec2)
-        >>> print(out)
-        [[49 70 77]
-         [56 80 88]
-         [63 90 99]]
+        >>> from mindspore import Tensor, ops
+        >>> x = Tensor(np.random.rand(4, 4, 32, 32), mindspore.float32)
+        >>> output = ops.auto_generate.unfold_ext(x, kernel_size=3, dilation=1, stride=1)
+        >>> print(output.shape)
+        (4, 36, 900)
     """
-    return outer_impl(input, vec2)
+    return unfold_impl(input, converted_kernel_size, converted_dilation, converted_padding, converted_stride)
 
 
-def log_softmax(input, dim=None, dtype=None):
+def mish(input):
     r"""
-    Applies the Log Softmax function to the input tensor on the specified axis.
-    Supposes a slice in the given axis, :math:`x` for each element :math:`x_i`,
-    the Log Softmax function is shown as follows:
+    Computes MISH (A Self Regularized Non-Monotonic Neural Activation Function)
+    of input tensors element-wise.
+    
+    The formula is defined as follows:
     
     .. math::
-        \text{output}(x_i) = \log \left(\frac{\exp(x_i)} {\sum_{j = 0}^{N-1}\exp(x_j)}\right),
+        \text{mish}(input) = input * \tanh(softplus(\text{input}))
     
-    where :math:`N` is the length of the Tensor.
+    See more details in `A Self Regularized Non-Monotonic Neural Activation Function 
+    <https://arxiv.org/abs/1908.08681>`_.
+    
+    Mish Activation Function Graph:
+    
+    .. image:: ../images/Mish.png
+        :align: center
     
     Args:
-        input (Tensor): The input Tensor.
-        dim (int, optional): The axis to perform the Log softmax operation. Default: ``None`` .
+        input (Tensor): The input of MISH. Supported dtypes: 
     
-    Keyword Args:
-        dtype (:class:`mindspore.dtype`, optional): The desired dtype of returned Tensor. If not set to None, the input
-            Tensor will be cast to `dtype` before the operation is performed. This is useful for preventing overflows.
-            If set to None, stay the same as original Tensor. Default: ``None`` . Supported data type is {float16, float32, double, bfloat16}.
+            - Ascend: float16, float32.
     
     Returns:
-        Tensor, with the same shape as the input.
+        Tensor, has the same type and shape as the `input`.
     
     Raises:
-        TypeError: If `dim` is not an int.
-        ValueError: If `dim` is not in range [-len(input.shape), len(input.shape)).
+        TypeError: If `input` is not a Tensor.
+        TypeError: If dtype of `input` is not float16 or float32.
     
     Supported Platforms:
         ``Ascend``
     
     Examples:
         >>> import mindspore
-        >>> import numpy as np
         >>> from mindspore import Tensor, ops
-        >>> logits = Tensor(np.array([1, 2, 3, 4, 5]), mindspore.float32)
-        >>> output = ops.auto_generate.log_softmax(logits, dim=-1)
+        >>> import numpy as np
+        >>> x = Tensor(np.array([[-1.1, 4.0, -8.0], [2.0, -5.0, 9.0]]), mindspore.float32)
+        >>> output = ops.mish(x)
         >>> print(output)
-        [-4.4519143 -3.4519143 -2.4519143 -1.4519144 -0.4519144]
+        [[-3.0764845e-01 3.9974124e+00 -2.6832507e-03]
+         [ 1.9439589e+00 -3.3576239e-02 8.9999990e+00]]
     """
-    return log_softmax_impl(input, dim, dtype)
+    return mish_impl(input)
 
 
-def matmul(input, other):
+def ffn(x, weight1, weight2, expertTokens=None, bias1=None, bias2=None, scale=None, offset=None, deqScale1=None, deqScale2=None, antiquant_scale1=None, antiquant_scale2=None, antiquant_offset1=None, antiquant_offset2=None, activation='fastgelu', inner_precise=0):
     r"""
     None
     """
-    return matmul_impl(input, other)
+    return ffn_impl(x, weight1, weight2, expertTokens, bias1, bias2, scale, offset, deqScale1, deqScale2, antiquant_scale1, antiquant_scale2, antiquant_offset1, antiquant_offset2, converted_activation, inner_precise)
 
 
-def acos(input):
+def histc(input, bins=100, min=0, max=0):
     r"""
-    Computes arccosine of input tensors element-wise.
+    Computes the histogram of a tensor.
     
-    .. math::
+    The elements are sorted into equal width bins between `min` and `max`.
+    If `min` and `max` are both zero, the minimum and maximum values of the data are used.
     
-        out_i = \cos^{-1}(input_i)
+    Elements lower than min or higher than max are ignored.
+    
+    .. warning::
+        This is an experimental API that is subject to change or deletion.
+        If input is int64, valid values fit within int32; exceeding this may cause precision errors.
     
     Args:
-        input (Tensor): The shape of tensor is
-            :math:`(N,*)`, where :math:`*` means any number of additional dimensions.
+        input (Tensor): the input tensor.
+        bins (int, optional): Number of histogram bins, optional. If specified, must be positive. Default: ``100`` .
+        min (int, float, optional): the lower end of the range (inclusive), optional. Default: ``0`` .
+        max (int, float, optional): the upper end of the range (inclusive), optional. Default: ``0`` .
     
     Returns:
-        Tensor, has the same shape as `input`. The dtype of output is float32 when dtype of `input` is in [bool, int8, uint8, int16, int32, int64]. Otherwise output has the same dtype as `input`.
+        A 1-D Tensor, has the same type as `input` with the shape :math:`(bins, )`.
     
     Raises:
         TypeError: If `input` is not a Tensor.
+        TypeError: If `input` datatype is not in support list.
+        TypeError: If attr `min` or `max` is not float or int.
+        TypeError: If attr `bins` is not int.
+        ValueError: If attr value `min` > `max`.
+        ValueError: If attr `bins` <= 0.
     
     Supported Platforms:
-        ``Ascend`` ``GPU`` ``CPU``
+        ``Ascend``
     
     Examples:
-        >>> import mindspore
-        >>> import numpy as np
         >>> from mindspore import Tensor, ops
-        >>> input = Tensor(np.array([0.74, 0.04, 0.30, 0.56]), mindspore.float32)
-        >>> output = ops.acos_ext(input)
-        >>> print(output)
-        [0.7377037  1.5307857 1.2661037 0.9764114]
+        >>> x = Tensor([1., 2, 1])
+        >>> y = ops.histc_ext(x, bins=4, min=0, max=3)
+        >>> print(y)
+        [0 2 1 0]
     """
-    return acos_impl(input)
+    return histc_impl(input, bins, min, max)
 
 
-def leaky_relu(input, negative_slope=0.01):
+def argmax(input, dim=None, keepdim=False):
     r"""
-    leaky_relu activation function. The element of `input` less than 0 times `negative_slope` .
+    argmax(input) -> Tensor
     
-    The activation function is defined as:
+    Return the indices of the maximum values of a tensor.
     
-    .. math::
-        \text{leaky_relu}(input) = \begin{cases}input, &\text{if } input \geq 0; \cr
-        \text{negative_slope} * input, &\text{otherwise.}\end{cases}
+    Args:
+        input (Tensor): Input tensor.
     
-    where :math:`negative\_slope` represents the `negative_slope` parameter.
+    Returns:
+        Tensor.
     
-    For more details, see `Rectifier Nonlinearities Improve Neural Network Acoustic Models
-    <https://ai.stanford.edu/~amaas/papers/relu_hybrid_icml2013_final.pdf>`_.
+    Supported Platforms:
+        ``Ascend``
     
-    LeakyReLU Activation Function Graph:
+    Examples:
+        >>> import numpy as np
+        >>> from mindspore import Tensor
+        >>> from mindspore import ops
+        >>> x = Tensor(np.array([[1, 20, 5], [67, 8, 9], [130, 24, 15]]).astype(np.float32))
+        >>> output = ops.auto_generate.argmax_ext(x)
+        >>> print(output)
+        6
     
-    .. image:: ../images/LeakyReLU.png
-        :align: center
+    .. function:: argmax(input, dim, keepdim=False) -> Tensor
+        :noindex:
+    
+    Return the indices of the maximum values of a tensor across a dimension.
     
     Args:
-        input (Tensor): The input of leaky_relu is a Tensor of any dimension.
-        negative_slope (Union[int, float], optional): Slope of the activation function when the element of `input` is less than 0.
-          Default: ``0.01`` .
+        input (Tensor): Input tensor.
+        dim (int): The dimension to reduce. 
+        keepdim (bool, optional): Whether the output tensor retains the specified
+            dimension. Default: ``False`` .
     
     Returns:
-        Tensor, has the same type and shape as the `input`.
+        Tensor, indices of the maximum values across a dimension.
     
     Raises:
-        TypeError: If `input` is not a Tensor.
-        TypeError: If `negative_slope` is not a float or an int.
+        TypeError: If `keepdim` is not bool.
+        ValueError: If `dim` is out of range.
     
     Supported Platforms:
         ``Ascend``
     
     Examples:
-        >>> import mindspore
         >>> import numpy as np
-        >>> from mindspore import Tensor, ops
-        >>> input = Tensor(np.array([[-1.0, 4.0, -8.0], [2.0, -5.0, 9.0]]), mindspore.float32)
-        >>> print(ops.extend.leaky_relu_ext(input, negative_slope=0.2))
-        [[-0.2  4.  -1.6]
-         [ 2.  -1.   9. ]]
+        >>> from mindspore import Tensor
+        >>> from mindspore import ops
+        >>> x = Tensor(np.array([[1, 20, 5], [67, 8, 9], [130, 24, 15]]).astype(np.float32))
+        >>> output = ops.auto_generate.argmax_ext(x, dim=-1)
+        >>> print(output)
+        [1 0 0]
     """
-    return leaky_relu_impl(input, negative_slope)
+    return argmax_impl(input, dim, keepdim)
 
 
-def mse_loss(input, target, reduction='mean'):
+def trace(input):
     r"""
-    Calculates the mean squared error between the predicted value and the label value.
-    
-    For detailed information, please refer to :class:`mindspore.nn.MSELoss`.
+    Returns a new tensor that is the sum of the `input` main trace.
     
     Args:
-        input (Tensor): Tensor of any dimension. The data type needs to be consistent with the `target`.
-            It should also be broadcastable with the `target`.
-        target (Tensor): The input label. Tensor of any dimension. The data type needs to be consistent with the `input`.
-            It should also be broadcastable with the `input`.
-        reduction (str, optional): Apply specific reduction method to the output: ``'mean'`` , ``'none'`` ,
-            ``'sum'`` . Default: ``'mean'`` .
-    
-            - ``'none'``: no reduction will be applied.
-            - ``'mean'``: compute and return the mean of elements in the output.
-            - ``'sum'``: the output elements will be summed.
+        input (Tensor): 2-D Tensor.
     
     Returns:
-        - Tensor. If `reduction` is ``'mean'`` or ``'sum'``, the shape of output is `Tensor Scalar`.
-        - If reduction is ``'none'``, the shape of output is the broadcasted shape of **input** and **target** .
+        Tensor, when the data type of `input` is integer or bool, its data type is int64, otherwise it is the same as `input`, and size equals to 1.
     
     Raises:
-        ValueError: If `reduction` is not one of ``'mean'`` , ``'sum'`` or ``'none'``.
-        ValueError: If `input` and `target` are not broadcastable.
-        TypeError: If `input` and `target` are in different data type.
+        TypeError: If `input` is not a Tensor.
+        ValueError: If the dimension of `input` is not equal to 2.
+        TypeError:  If the dtype of `input` is not one of float16, float32, float64, bool, uint8, int8, int16, int32, int64, complex64, complex128, bfloat16.
     
     Supported Platforms:
         ``Ascend``
@@ -2308,67 +2143,114 @@ def mse_loss(input, target, reduction='mean'):
         >>> import mindspore
         >>> import numpy as np
         >>> from mindspore import Tensor, ops
-        >>> logits = Tensor(np.array([1, 2, 3]), mindspore.float32)
-        >>> labels = Tensor(np.array([[1, 1, 1], [1, 2, 2]]), mindspore.float32)
-        >>> output = ops.mse_loss_ext(logits, labels, reduction='none')
+        >>> input = Tensor(np.array([[10, 11, 12], [13, 14, 15], [16, 17, 18]]), mindspore.float32)
+        >>> output = ops.trace_ext(input)
         >>> print(output)
-        [[0. 1. 4.]
-         [0. 0. 1.]]
+        42.0
+        >>> input = Tensor(np.arange(1, 13).reshape(3, 4), mindspore.float32)
+        >>> output = ops.trace_ext(input)
+        >>> print(output)
+        18.0
+        >>> input = Tensor(np.arange(12, 0, -1).reshape(4, 3), mindspore.float32)
+        >>> output = ops.trace_ext(input)
+        >>> print(output)
+        24.0
     """
-    return mse_loss_impl(input, target, converted_reduction)
+    return trace_impl(input)
 
 
-def adaptive_avg_pool2d_grad(grad_output, x):
+def matrix_inverse(input):
     r"""
-    None
+    Compute the inverse of the input matrix.
+    
+    Args:
+        input (Tensor): A matrix to be calculated. Input `input` must be at least two dimensions, and the size of
+            the last two dimensions must be the same size.
+    
+    Returns:
+        Tensor, has the same type and shape as input`.
+    
+    Raises:
+        TypeError: If `input` is not a Tensor.
+        ValueError: If the size of the last two dimensions of `input` is not the same.
+        ValueError: If the dimension of `input` is 1.
+    
+    Supported Platforms:
+        ``Ascend``
+    
+    Examples:
+        >>> from mindspore import Tensor, ops
+        >>> from mindspore import dtype as mstype
+        >>> x = Tensor([[1., 2.], [3., 4.]], mstype.float32)
+        >>> print(ops.matrix_inverse_ext(x))
+        [[-2.   1. ]
+         [ 1.5 -0.5]]
     """
-    return adaptive_avg_pool2d_grad_impl(grad_output, x)
+    return matrix_inverse_impl(input)
 
 
-def diag(input, diagonal=0):
+def fold(input, output_size, kernel_size, dilation=1, padding=0, stride=1):
     r"""
-    If input is a vector (1-D tensor), then returns a 2-D square tensor with the elements of input as the diagonal.
+    Combines an array of sliding local blocks into a large containing tensor.
     
-    If input is a matrix (2-D tensor), then returns a 1-D tensor with the diagonal elements of input.
+    Consider a batched input tensor of shape :math:`(N, C \times \prod(\text{kernel_size}), L)` ,
+    where :math:`N` is the batch dimension, :math:`C \times \prod(\text{kernel_size})` is the
+    total number of values within each block (a block has :math:`\prod(\text{kernel_size})` spatial
+    locations each containing a `C`-channeled vector), and :math:`L` is the total number of such blocks:
     
-    The argument diagonal controls which diagonal to consider:
+    .. math::
+        L = \prod_d \left\lfloor\frac{\text{output_size}[d] + 2 \times \text{padding}[d] %
+            - \text{dilation}[d] \times (\text{kernel_size}[d] - 1) - 1}{\text{stride}[d]} + 1\right\rfloor,
     
-    - If `diagonal` = 0, it is the main diagonal.
+    where :math:`d` is over all spatial dimensions.
     
-    - If `diagonal` > 0, it is above the main diagonal.
+    Therefore, `output_size` is the spatial shape of the large containing tensor of the sliding local blocks.
     
-    - If `diagonal` < 0, it is below the main diagonal.
+    The `dilation`, `padding` and `stride` arguments specify how the sliding blocks are retrieved.
     
     .. warning::
-        This is an experimental API that is subject to change or deletion.
+        Currently, only unbatched(3D) or batched(4D) image-like output tensors are supported.
     
     Args:
-        input (Tensor): The input tensor.
-        diagonal (int, optional): the diagonal to consider. Defaults: ``0``.
+        input (Tensor): 2-D or 3-D Tensor.
+        output_size (Union[int, tuple[int], list[int]]): The shape of the spatial dimensions of
+            the output(i.e., output.shape[2:]).
+        kernel_size (Union[int, tuple[int], list[int]]): The size of the kernel, should be two int
+            for height and width. If type is int, it means that height equal with width. Must be specified.
+        dilation (Union[int, tuple[int], list[int]], optional): The size of the dilation, should be two int
+            for height and width. If type is int, it means that height equal with width. Default: ``1`` .
+        padding (Union[int, tuple[int], list[int]], optional): The size of the padding, should be two int
+            for height and width. If type is int, it means that height equal with width. Default: ``0`` .
+        stride (Union[int, tuple[int], list[int]], optional): The size of the stride, should be two int
+            for height and width. If type is int, it means that height equal with width. Default: ``1`` .
     
     Returns:
-        Tensor, has the same dtype as the `input`, its shape is up to `diagonal`.
-    
-        - If `input` shape is :math:`(x_0)` : then output shape is :math:`(x_0 + \left | diagonal \right | , x_0 + \left | diagonal \right | )` 2-D Tensor.
+        A Tensor, with same type as `input` .
     
-        - If `input` shape is :math:`(x_0, x_1)` : then output shape is main diagonal to move :math:`(\left | diagonal \right |)` elements remains elements' length 1-D Tensor.
+    Shape:
+        - Input: :math:`(N, C \times \prod(\text{kernel_size}), L)` or
+          :math:`(C \times \prod(\text{kernel_size}), L)`
+        - Output: :math:`(N, C, output\_size[0], output\_size[1], ...)` or
+          :math:`(C, output\_size[0], output\_size[1], ...)`
     
     Raises:
-        TypeError: If `input` is not a Tensor.
-        ValueError: If shape of `input` is not 1-D and 2-D.
+        TypeError: If `output_size`, `kernel_size`, `stride`, `dilation`, `padding` data type is not int, tuple or list.
+        ValueError: If `output_size`, `kernel_size`, `dilation`, `stride` value is not
+            greater than zero or elements number invalid.
+        ValueError: If `padding` value is less than zero or elements number invalid.
+        ValueError: If input.shape[-2] can't be divisible by the product of kernel_size.
+        ValueError: If `input.shape[-1]` is not equal to the calculated number of sliding blocks `L`.
     
     Supported Platforms:
         ``Ascend``
     
     Examples:
-        >>> from mindspore import Tensor, mint
-        >>> input = Tensor([1, 2, 3, 4]).astype('int32')
-        >>> output = mint.diag(input)
-        >>> print(output)
-        [[1 0 0 0]
-         [0 2 0 0]
-         [0 0 3 0]
-         [0 0 0 4]]
+        >>> import numpy as np
+        >>> from mindspore import Tensor, ops
+        >>> x = Tensor(np.random.rand(16, 64, 25).astype(np.float32))
+        >>> output = ops.auto_generate.fold_ext(x, (8, 8), [2, 2], [2, 2], [2, 2], [2, 2])
+        >>> print(output.shape)
+        (16, 16, 8, 8)
     """
-    return diag_impl(input, diagonal)
+    return fold_impl(input, converted_output_size, converted_kernel_size, converted_dilation, converted_padding, converted_stride)