mindspore 2.6.0__cp311-none-any.whl → 2.7.0__cp311-none-any.whl

mindspore/ops/functional_overload.py

@@ -13,69 +13,161 @@
 # limitations under the License.
 # ============================================================================
 """Holding mint APIs"""
-from mindspore._c_expression import _greater_equal_instance
-from mindspore._c_expression import _lerp_instance
-from mindspore._c_expression import _xlogy_instance
-from mindspore._c_expression import _where_instance
-from mindspore._c_expression import _min_instance
-from mindspore._c_expression import _kthvalue_instance
-from mindspore._c_expression import _sub_instance
-from mindspore._c_expression import _gmm_instance
-from mindspore._c_expression import _div_instance
-from mindspore._c_expression import _matmul_reduce_scatter_instance
-from mindspore._c_expression import _add_instance
-from mindspore._c_expression import _fmod_instance
-from mindspore._c_expression import _pixel_shuffle_instance
-from mindspore._c_expression import _addcdiv_instance
-from mindspore._c_expression import _gelu_instance
-from mindspore._c_expression import _max_instance
+from mindspore._c_expression import _conv3d_instance
 from mindspore._c_expression import _floor_divide_instance
-from mindspore._c_expression import _nansum_instance
 from mindspore._c_expression import _empty_instance
-from mindspore._c_expression import _gmm_backward_instance
+from mindspore._c_expression import _addcdiv_instance
+from mindspore._c_expression import _gmm_instance
+from mindspore._c_expression import _rmod_instance
+from mindspore._c_expression import _nansum_instance
 from mindspore._c_expression import _remainder_instance
 from mindspore._c_expression import _repeat_interleave_instance
-from mindspore._c_expression import _bitwise_not_instance
-from mindspore._c_expression import _all_gather_matmul_instance
+from mindspore._c_expression import _kthvalue_instance
+from mindspore._c_expression import _greater_equal_instance
+from mindspore._c_expression import _gelu_instance
 from mindspore._c_expression import _clamp_instance
+from mindspore._c_expression import _empty_like_instance
+from mindspore._c_expression import _add_instance
+from mindspore._c_expression import _sub_instance
+from mindspore._c_expression import _bitwise_not_instance
 from mindspore._c_expression import _gmm_backward_fusion_instance
+from mindspore._c_expression import _xlogy_instance
+from mindspore._c_expression import _div_instance
+from mindspore._c_expression import _einsum_instance
+from mindspore._c_expression import _any_instance
+from mindspore._c_expression import _max_instance
+from mindspore._c_expression import _fmod_instance
+from mindspore._c_expression import _min_instance
+from mindspore._c_expression import _all_gather_matmul_instance
+from mindspore._c_expression import _gmm_backward_instance
+from mindspore._c_expression import _pixel_shuffle_instance
+from mindspore._c_expression import _quant_matmul_instance
+from mindspore._c_expression import _bernoulli__instance
+from mindspore._c_expression import _lerp_instance
+from mindspore._c_expression import _matmul_reduce_scatter_instance
+from mindspore._c_expression import _index_add_instance
+from mindspore._c_expression import _where_instance
 
-def greater_equal(*args, **kwargs):
+def conv3d(*args, **kwargs):
     r"""
-    greater_equal(input, other) -> Tensor
+    conv3d(input, weight, bias=None, stride=1, padding=0, dilation=1, groups=1) -> Tensor
 
-    Computes the boolean value of :math:`input >= other` element-wise.
+    Applies a 3D convolution over an input tensor. The input tensor is typically of
+    shape :math:`(N, C_{in}, D_{in}, H_{in}, W_{in})` or :math:`(C_{in}, D_{in}, H_{in}, W_{in})`,
+    where :math:`N` is batch size, :math:`C` is channel number, :math:`D, H, W` are the depth,
+    height and width of the feature graph, respectively.
+
+    The output is calculated based on formula:
 
     .. math::
 
-        out_{i} =\begin{cases}
-            & \text{True,    if } input_{i}>=other_{i} \\
-            & \text{False,   if } input_{i}<other_{i}
-            \end{cases}
+        \text{out}(N_i, C_{\text{out}_j}) = \text{bias}(C_{\text{out}_j}) +
+        \sum_{k = 0}^{C_{in} - 1} \text{ccor}({\text{weight}(C_{\text{out}_j}, k), \text{X}(N_i, k)})
+
+    where :math:`bias` is the output channel bias, :math:`ccor` is
+    the `cross-correlation <https://en.wikipedia.org/wiki/Cross-correlation>`_
+    , :math:`weight` is the convolution kernel value and :math:`X` represents the input feature map.
+
+    Here are the indices' meanings:
+
+    - :math:`i` corresponds to the batch number, the range is :math:`[0, N-1]`,
+      where :math:`N` is the batch size of the input.
+
+    - :math:`j` corresponds to the output channel, the range is :math:`[0, C_{out}-1]`,
+      where :math:`C_{out}` is the number of
+      output channels, which is also equal to the number of kernels.
+
+    - :math:`k` corresponds to the input channel, the range is :math:`[0, C_{in}-1]`,
+      where :math:`C_{in}` is the number of
+      input channels, which is also equal to the number of channels in the convolutional kernels.
+
+    Therefore, in the above formula, :math:`{bias}(C_{\text{out}_j})` represents the bias of the :math:`j`-th
+    output channel, :math:`{weight}(C_{\text{out}_j}, k)` represents the slice of the :math:`j`-th convolutional
+    kernel in the :math:`k`-th channel, and :math:`{X}(N_i, k)` represents the slice of the :math:`k`-th input
+    channel in the :math:`i`-th batch of the input feature map.
+
+    The shape of the convolutional kernel is given by :math:`(kd, kh, kw)` where :math:`kd` , :math:`kd` and\
+    :math:`kw` are the depth, height and width of the kernel, respectively.
+    If we consider the input and output channels as well as the `group` parameter, the complete kernel shape
+    will be :math:`(C_{out}, C_{in} / \text{group}, kd, kh, kw)`,
+    where `group` is the number of groups dividing `x`'s input channel when applying group convolution.
+
+    For more details about convolution layer, please refer to `Gradient Based Learning Applied to Document Recognition
+    <http://vision.stanford.edu/cs598_spring07/papers/Lecun98.pdf>`_.
+
+    The following lists some of the limitations of the parameters.
+
+    - input -- The input to the conv3d. The input must have each dimension size within the range [1, int32_max].
+    - weight -- Filters of shape :math:`(C_{out}, C_{in} / groups, kd, kh, kw)`. The value of :math:`kh`
+      and :math:`kw` is in the range [1, 511]. The remaining values are in the range [1, int32_max].
+      And :math:`kh*kw*k0` is less 65536 (k0 is 16. If data type is float32, k0 is 8).
+    - bias -- Bias Tensor with shape :math:`(C_{out})`. The shape must equal to the first dimension of the weight.
+    - stride -- The distance of kernel moving. It can be an int number or
+      tuple (noted by :math:`(stride_d, stride_h, stride_w)`). stride_h and stride_w are in the range [1, 63].
+      stride_d is in the range [1, 255].
+    - padding -- If padding is an int number, it is in the range [0, 255].
+    - dilation -- The value is in the range [1, 255].
+    - groups -- The value is in the range [1, 65535].
+    - :math:`C_{in} \% \text{groups} == 0 \quad \text{and} \quad C_{out} \% \text{groups} == 0` .
+    - :math:`weight[1] == C_{in} / groups` .
+    - :math:`H_{in} + PadUp + PadDown >= (kh - 1) * DilationH + 1` .
+    - :math:`W_{in} + PadLeft + PadRight >= (kw - 1) * DilationW + 1` .
+    - :math:`D_{in} + PadFront + PadBack >= (kd - 1) * DilationD + 1` .
+    - :math:`H_{out} = (H_{in} + PadUp + PadDown - ((kh - 1) * DilationH + 1)) / StrideH + 1` .
+    - :math:`W_{out} = (W_{in} + PadLeft + PadRight - ((kw - 1) * DilationW + 1)) / StrideW + 1` .
+    - :math:`D_{out} = (D_{in} + PadFront + PadBack - ((kd - 1) * DilationD + 1)) / StrideD + 1` .
+    - :math:`(D_{in}+PadFront+PadBack - ((kd-1)*DilationD+1)) \% StrideD <= PadBack` .
+    - :math:`(H_{in}+PadUp+PadDown - ((kh-1)*Dilationh+1)) \% StrideH <= PadDown` .
+    - :math:`stride_d <= kernel_d` .
+    - :math:`PadUp < kh` and :math:`PadDown < kh` . When `padding` = ``'valid'``, both PadUp and PadDown are zeros.
+      When `padding` = ``'same'``, pad can be calculated by
+      :math:`floor(((H_{out}-1) * strideH + (kh - 1) * DilationH + 1 - H_{in}) / 2)` for high dimension.
+      It is similar way to calculate the padding for depth and width dimension. And the depth and width
+      dimensions also have the same constraints.
+    - :math:`((kh - 1) * DilationH - PadUp)` should be in [0, 255]. It is the same constraint for depth
+      and width dimension.
+    - If `padding` is ``'same'``, `stride` must be 1.
 
-    Note:
-        - Inputs of `input` and `other` comply with the implicit type conversion rules to make the data types
-          consistent.
-        - The inputs must be two tensors or one tensor and one scalar.
-        - When the inputs are two tensors, dtypes of them cannot be bool at the same time,
-          and the shapes of them can be broadcast.
-        - When the inputs are one tensor and one scalar, the scalar could only be a constant.
-        - Broadcasting is supported.
-        - If the input Tensor can be broadcast, the low dimension will be extended to the corresponding high dimension
-          in another input by copying the value of the dimension.
+    .. warning::
+        It is only supported on Atlas A2 Training Series Products.
 
     Args:
-        input (Union[Tensor, Number]): The first input is a number
-            or a tensor whose data type is `number <https://www.mindspore.cn/docs/en/master/api_python/mindspore/mindspore.dtype.html#mindspore.dtype>`_ or `bool_ <https://www.mindspore.cn/docs/en/master/api_python/mindspore/mindspore.dtype.html#mindspore.dtype>`_.
-        other (Union[Tensor, Number]): Second input. When the first input is a Tensor, the second input should be a Number,
-            or a Tensor of the number or bool_ data type. When the first input is a Scalar,
-            the second input must be a Tensor of number or bool_ data type.
+        input (Tensor): Tensor of shape :math:`(N, C_{in}, D_{in}, H_{in}, W_{in})`.
+        weight (Tensor): Set size of kernel is :math:`(kd, kh,
+            kw)`, then the shape is :math:`(C_{out}, C_{in} / groups, kd, kh, kw)`.
+        bias (Tensor, optional): Bias Tensor with shape :math:`(C_{out})`.
+            When bias is ``None`` , zeros will be used. Default: ``None`` .
+        stride (Union(int, tuple[int], list[int]), optional): The distance of kernel moving, an int
+            number that represents the depth, the height and width of movement are both strides, or a
+            tuple of triple int numbers that
+            represent the depth, height and width of movement respectively. Default: ``1`` .
+        padding (Union(int, tuple[int], list[int], str), optional): Implicit paddings on both sides of the input `x`.
+            Can be a string, one integer or a tuple/list with 3 integers.
+            If `padding` is a string, the optional values are ``"same"`` , ``"valid"``.
+
+            - same: Adopts the way of completion. The height and width of the output will be equal to
+              the input `x` divided by stride. The padding will be evenly calculated in top and bottom,
+              left and right possiblily. Otherwise, the last extra padding will be calculated from the bottom
+              and the right side. If this mode is set, `stride` must be 1.
+
+            - valid: Adopts the way of discarding. The possible largest height and width of output will be returned
+              without padding. Extra pixels will be discarded.
+
+            If `padding` is one integer, the paddings of top, bottom, left and right are the same, equal to padding.
+            If `padding` is a tuple/list with 3 integers, the padding of head, tail, top, bottom,
+            left and right equal to pad[0], pad[0], pad[1], pad[1], pad[2] and pad[2] correspondingly. Default: ``0`` .
+        dilation (Union[int, tuple[int], list[int]], optional): Controlling the space between the kernel points.
+            Default: ``1`` .
+        groups (int, optional): Splits `input` into groups. Default: ``1`` .
 
     Returns:
-        Tensor, the shape is the same as the one after broadcasting, and the data type is bool.
+        Tensor, the same dtype as the `input`, with the shape :math:`(N, C_{out}, D_{out}, H_{out}, W_{out})`
+        or :math:`(C_{out}, D_{out}, H_{out}, W_{out})`.
 
     Raises:
-        TypeError: If neither `input` nor `other` is a Tensor.
+        TypeError: If `stride`, `padding` or `dilation` is neither an int nor a tuple.
+        TypeError: `groups` is not an int.
+        TypeError: If `bias` is not a Tensor.
 
     Supported Platforms:
         ``Ascend``
@@ -83,245 +175,394 @@ def greater_equal(*args, **kwargs):
     Examples:
         >>> import mindspore
         >>> import numpy as np
-        >>> from mindspore import Tensor, mint
-        >>> input = Tensor(np.array([1, 2, 3]), mindspore.int32)
-        >>> other = Tensor(np.array([1, 1, 4]), mindspore.int32)
-        >>> output = mint.greater_equal(input, other)
-        >>> print(output)
-        [True True False]
-        >>> y = 2.1
-        >>> output = mint.greater_equal(input, y)
-        >>> print(output)
-        [False False True]
+        >>> from mindspore import mint
+        >>> x = mindspore.Tensor(np.random.randn(12, 1, 60, 50, 8), mindspore.float16)
+        >>> w = mindspore.Tensor(np.random.randn(26, 1, 2, 4, 4), mindspore.float16)
+        >>> out = mint.nn.functional.conv3d(x, w)
+        >>> print(out.shape)
+        (12, 26, 59, 47, 5)
     """
-    return _greater_equal_instance(*args, **kwargs)
+    return _conv3d_instance(*args, **kwargs)
 
 
-def ge(*args, **kwargs):
+def floor_divide(*args, **kwargs):
     r"""
-    ge(input, other) -> Tensor
-
-    Alias for :func:`mindspore.mint.greater_equal`.
-    """
-    return _greater_equal_instance(*args, **kwargs)
+    floor_divide(input, other) -> Tensor
 
+    Divides the first input tensor by the second input tensor element-wise and round down to the closest integer.
 
-def lerp(*args, **kwargs):
-    r"""
-    lerp(input, end, weight) -> Tensor
+    Inputs of `input` and `other` comply with the implicit type conversion rules to make the data types consistent.
+    Inputs must be two tensors or one tensor and one scalar.
+    When the inputs are two tensors,
+    dtypes of them cannot be bool at the same time, and the shapes of them could be broadcast.
+    When the inputs are one tensor and one scalar,
+    the scalar could only be a constant.
 
-    Perform a linear interpolation of two tensors input and end based on a float or tensor weight.
+    .. math::
+        out_{i} = \text{floor}( \frac{input_i}{other_i})
 
-    If `weight` is a tensor, the shapes of three inputs need to be broadcast;
-    If `weight` is a float, the shapes of `input` and `end` need to be broadcast.
-    If `weight` is a float and platform is Ascend, the types of `input` and `end` need to be float32.
+    where the :math:`floor` indicates the Floor operator. For more details,
+    please refer to the :class:`mindspore.mint.floor` operator.
 
     .. warning::
         This is an experimental API that is subject to change or deletion.
 
-    .. math::
-        output_{i} = input_{i} + weight_{i} * (end_{i} - input_{i})
-
     Args:
-        input (Tensor): The tensor with the starting points. Data type must be float16 or float32.
-        end (Tensor): The tensor with the ending points. Data type must be the same as `input`.
-        weight (Union[float, Tensor]): The weight for the interpolation formula. Must be a float scalar
-            or a tensor with float16 or float32 data type.
+        input (Union[Tensor, Number, bool]): The first input is a number or
+            a bool or a tensor whose data type is number or bool.
+        other (Union[Tensor, Number, bool]): The second input is a number or
+            a bool or a tensor whose data type is number or bool.
 
     Returns:
-        Tensor, has the same type and shape as input `input`.
+        Tensor, the shape is the same as the one after broadcasting,
+        and the data type is the one with higher precision or higher digits among the two inputs.
 
     Raises:
-        TypeError: If `input` or `end` is not a tensor.
-        TypeError: If `weight` is neither scalar(float) nor tensor.
-        TypeError: If dtype of `input` or `end` is neither float16 nor float32.
-        TypeError: If dtype of `weight` is neither float16 nor float32 when it is a tensor.
-        TypeError: If `input` and `end` have different data types.
-        TypeError: If `input`, `end` and `weight` have different data types when `weight` is a tensor.
-        ValueError: If `end` could not be broadcast to a tensor with shape of `input`.
-        ValueError: If `weight` could not be broadcast to tensors with shapes of `input` and `end` when it is a tensor.
+        TypeError: If `input` and `other` are not the following: Tensor, number.Number or bool.
 
     Supported Platforms:
         ``Ascend`` ``GPU`` ``CPU``
 
     Examples:
         >>> import mindspore
-        >>> import numpy as np
         >>> from mindspore import Tensor, mint
-        >>> start = Tensor(np.array([1., 2., 3., 4.]), mindspore.float32)
-        >>> end = Tensor(np.array([10., 10., 10., 10.]), mindspore.float32)
-        >>> output = mint.lerp(start, end, 0.5)
+        >>> import numpy as np
+        >>> input = Tensor(np.array([2, 4, -1]), mindspore.int32)
+        >>> other = Tensor(np.array([3, 3, 3]), mindspore.int32)
+        >>> output = mint.floor_divide(input, other)
         >>> print(output)
-        [5.5 6. 6.5 7. ]
+        [ 0  1 -1]
+        >>> input = Tensor(2.0, mindspore.float32)
+        >>> other = Tensor(2.0, mindspore.float32)
+        >>> output = mint.floor_divide(input, other)
+        >>> print(output)
+        1.0
     """
-    return _lerp_instance(*args, **kwargs)
+    return _floor_divide_instance(*args, **kwargs)
 
 
-def xlogy(*args, **kwargs):
+def empty(*args, **kwargs):
     r"""
-    xlogy(input, other) -> Tensor
-
-    Computes the first input multiplied by the logarithm of second input element-wise.
-    Returns zero when `input` is zero.
+    empty(*size, *, dtype=None, device=None) -> Tensor
 
-    .. math::
-
-        out_i = input_{i}\log{other_{i}}
+    Creates a tensor with uninitialized data, whose shape, dtype and device are described by the argument `size`,
+    `dtype` and `device` respectively.
 
-    Inputs of `input` and `other` comply with the implicit type conversion rules to make the data types consistent.
-    The inputs must be two tensors or one tensor and one scalar.
-    When the inputs are two tensors, the shapes of them could be broadcast.
+    .. warning::
+        This is an experimental API that is subject to change or deletion.
 
     Args:
-        input (Union[Tensor, numbers.Number, bool]): The first input is a numbers.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, numbers.Number, bool]): The second input is a numbers.Number or
-            a bool or a tensor whose data type is number or bool when the first input is a tensor.
-            When the first input is Scalar, the second input must be a Tensor whose data type is number or bool.
+        size (Union[tuple[int], list[int], int]): The specified shape of output tensor. Can be variable numbers of
+            positive integers or tuple or list containing positive integers.
+
+    Keyword Args:
+        dtype (:class:`mindspore.dtype`, optional): The specified type of output tensor. If `dtype` is ``None`` ,
+            `mindspore.float32` will be used. Default: ``None`` .
+        device (string, optional): The specified device of the output tensor. In PyNative mode, ``"Ascend"``, ``"npu"``,
+            ``"cpu"`` and ``"CPU"`` are supported. In graph mode O0, ``"Ascend"`` and ``"npu"`` are supported. If `device = None`,
+            `mindspore.context.device_target` will be used. Default ``None``.
 
     Returns:
-        Tensor, the shape is the same as the one after broadcasting,
-        and the data type is the one with higher precision or higher digits among the two inputs.
+        Tensor, whose shape, dtype and device are defined by input.
 
     Raises:
-        TypeError: If `input` and `other` is not a numbers.Number or a bool or a Tensor.
-        ValueError: If `input` could not be broadcast to a tensor with shape of `other`.
+        TypeError:  If `size` is neither an int nor a tuple or list of int.
 
     Supported Platforms:
-        ``Ascend`` ``GPU`` ``CPU``
+        ``Ascend`` ``CPU``
 
     Examples:
         >>> import mindspore
-        >>> import numpy as np
-        >>> from mindspore import Tensor, ops
-        >>> input = Tensor(np.array([-5, 0, 4]), mindspore.float32)
-        >>> other = Tensor(np.array([2, 2, 2]), mindspore.float32)
-        >>> output = ops.xlogy(input, other)
+        >>> from mindspore import ops
+        >>> output = ops.empty((2, 3), dtype=mindspore.float32)
         >>> print(output)
-        [-3.465736   0.        2.7725887]
+        [[0. 0. 0.]
+         [0. 0. 0.]]
     """
-    return _xlogy_instance(*args, **kwargs)
+    return _empty_instance(*args, **kwargs)
 
 
-def where(*args, **kwargs):
+def addcdiv(*args, **kwargs):
     r"""
-    where(condition, input, other) -> Tensor
+    addcdiv_ext(input, tensor1, tensor2, *, value=1) -> Tensor
 
-    Selects elements from `input` or `other` based on `condition` and returns a tensor.
+    Performs the element-wise division of tensor tensor1 by tensor tensor2,
+    multiply the result by the scalar value and add it to input data.
 
     .. math::
-        output_i = \begin{cases} input_i,\quad &if\ condition_i \\ other_i,\quad &otherwise \end{cases}
+        y[i] = input[i] + value * (tensor1[i] / tensor2[i])
+
+    .. warning::
+        This is an experimental API that is subject to change or deletion.
 
     Args:
-        condition (Tensor[bool]): If true, yield `input`, otherwise yield `other`.
-        input (Union[Tensor, Scalar]): When `condition` is true, values to select from.
-        other (Union[Tensor, Scalar]): When `condition` is false, values to select from.
+        input (Tensor): The tensor to be added.
+        tensor1 (Tensor): The numerator tensor.
+        tensor2 (Tensor): The denominator tensor.
+
+    Keyword Args:
+        value (Number, optional): The multiplier for tensor1/tensor2. Default: ``1`` .
 
     Returns:
-        Tensor, elements are selected from `input` and `other`.
+        Tensor, has the same shape and dtype as tensor1/tensor2.
 
     Raises:
-        TypeError: If `condition` is not a tensor.
-        TypeError: If both `input` and `other` are scalars.
-        ValueError: If `condition`, `input` and `other` can not broadcast to each other.
+        TypeError: If dtype of `tensor1`, `tensor2`, or `input` is not tensor.
+        ValueError: If `tensor1` could not be broadcast to a tensor with shape of `tensor2`.
+        ValueError: If `value` could not be broadcast to tensors with shapes of `tensor1/tensor2`.
+        ValueError: If `input` could not be broadcast to tensors with shapes of `value*(tensor1/tensor2)`.
 
     Supported Platforms:
-        ``Ascend`` ``GPU`` ``CPU``
+        ``Ascend``
 
     Examples:
+        >>> import mindspore
         >>> import numpy as np
-        >>> from mindspore import tensor, ops
-        >>> from mindspore import dtype as mstype
-        >>> a = tensor(np.arange(4).reshape((2, 2)), mstype.float32)
-        >>> b = tensor(np.ones((2, 2)), mstype.float32)
-        >>> condition = a < 3
-        >>> output = ops.where(condition, a, b)
-        >>> print(output)
-        [[0. 1.]
-         [2. 1.]]
-    
-    .. function:: where(condition) -> Tensor
-        :noindex:
+        >>> from mindspore import Tensor, ops
+        >>> input_data = Tensor(np.array([1, 1, 1, 1]), mindspore.float32)
+        >>> x1 = Tensor(np.array([1, 2, 3, 4]), mindspore.float32)
+        >>> x2 = Tensor(np.array([4, 3, 2, 1]), mindspore.float32)
+        >>> y = ops.addcdiv_ext(input_data, x1, x2, value=1)
+        >>> print(y)
+        [1.25      1.6666667 2.5       5.       ]
+    """
+    return _addcdiv_instance(*args, **kwargs)
 
-    Identical to :func:`mindspore.ops.nonzero` with input `condition` and `as_tuple` being True.
+
+def gmm(*args, **kwargs):
+    r"""
+    gmm(x, weight, bias=None, group_list=None, group_type=0, group_list_type=0) -> tuple[Tensor]
+
+    Grouping matrix multiplication.
+
+    .. warning::
+        - This is an experimental API that is subject to change or deletion.
+        - `group_type` must be a constant.
+        - Only support on Atlas A2 training series.
+        - When the type of `group_list` is tuple[int] or list[int], it should a non-negative non-decreasing sequence,
+          indicating indexes of each group along the split axis. In this scenario, the arg `group_list_type` is useless.
+
+    .. note::
+        - When `group_type` is 2, the tensors in `x` must be non-continuous tensors which has
+          been transposed.
+        - Only when `group_type` is 0 and `bias` is None, the reverse derivative is supported,
+          which is implemented by ops.function.math_func.gmm_backward or through automatic differentiation.
+
+    Args:
+        x (tuple[Tensor]): The first tensors to be multiplied, whose num should be 1.
+        weight (tuple[Tensor]): The second tensors to be multiplied, whose num should be 1.
+        bias (tuple[Tensor], optional): Biases added to outputs, whose num should be 1.
+            The shape of each tensor in `bias` should be :math: `(group_list.shape[0], n)`
+            or :math: `(len(group_list), n)`. In the training scenario, the bias only supports None.
+            Default: ``None`` .
+        group_list (Union[Tensor, list[int], tuple[int]], optional): 1-D Tensor, list[int]
+            or tuple[int], indicating indexes or sizes of each group along the split axis.
+            When `group_list` is list[int] or tuple[int], it's length should be less than or equal to 128.
+            When `group_list` is a Tensor, it's size should be less than or equal to 1024.
+            Supported dtypes: int64.
+            Default: ``None`` .
+
+            - If `group_list_type` is 0, it must be  a non-negative non-decreasing sequence.
+              And when `group_type` is 0, the last element in `group_list` should be equal to
+              the first dimension of the tensor in `x` . When `group_type` is 2, the last element
+              in `group_list` should be equal to the second dimension of the tensor in `x` .
+
+            - If `group_list_type` is 1, the value in `group_list` are the sizes of each group.
+        group_type (int, optional): Represents the axes that need to be grouped. For example,
+            :math: `C[m,n] = A[m,k] \times B[k,n]`. Default: ``0`` .
+
+            - If `group_type` is 0, it means that the m-axis is grouped, meaning that the shape
+              of each tensor in `x` should be :math: `(m, k)` , the shape of each tensor in `weight`
+              should be :math: `(group_list.shape[0], k, n)` or :math: `(len(group_list), k, n)`,
+              and the shape of each tensor in result would be :math: `(m, n)` .
+
+            - If `group_type` is 2, it means that the k-axis is grouped, meaning that
+              the shape of each tensor in `x` should be :math: `(m, k)`, the shape of each
+              tensor in `weight` should be :math: `(k, n)`, and the shape of each tensor
+              in result would be :math: `(group_list.shape[0], m, n)` or :math: `(len(group_list), m, n)`.
+        group_list_type (int, optional): If it's 0, the value in `group_list` are the cumsum
+            result of the size of each group. If it's 1, the value in `group_list` are the size
+            of each group. Default: ``0`` .
+
+    `x` , `weight` and `bias` only support the following 3 type combinations:
+
+    - x: float16, weight: float16, bias: float16
+    - x: bfloat16, weight: bfloat16, bias: float32
+    - x: float32, weight: float32, bias: float32
+
+    Returns:
+        tuple[Tensor], the results of grouping matrix multiplication.
 
     Supported Platforms:
         ``Ascend``
+
+    Examples:
+        >>> import numpy as np
+        >>> from mindspore import Tensor, ops
+        >>> x = Tensor(np.random.uniform(0,1, (10, 20)).astype(np.float32))
+        >>> weight = Tensor(np.random.uniform(0,1, (4, 20, 8)).astype(np.float32))
+        >>> group_list = Tensor([2, 4, 2, 2])
+        >>> y = ops.function.math_func.gmm([x,], [weight,], group_list=group_list, group_list_type=1)
+        >>> print(y[0].shape)
+        >>> (10, 8)
+        >>> group_list = [2, 6, 8, 10]
+        >>> y = ops.function.math_func.gmm([x,], [weight,], group_list=group_list, group_list_type=0)
+        >>> print(y[0].shape)
+        >>> (10, 8)
     """
-    return _where_instance(*args, **kwargs)
+    return _gmm_instance(*args, **kwargs)
 
 
-def min(*args, **kwargs):
+def rmod(*args, **kwargs):
     r"""
-    min(input) -> Tensor
+    rmod(input, other) -> Tensor
+    """
+    return _rmod_instance(*args, **kwargs)
 
-    Returns the minimum value of the input tensor.
+
+def nansum(*args, **kwargs):
+    r"""
+    nansum(input, dim=None, keepdim=False, *, dtype=None) -> Tensor
+
+    Computes sum of `input` over a given dimension, treating NaNs as zero.
+
+    .. warning::
+        It is only supported on Atlas A2 Training Series Products.
+        This is an experimental API that is subject to change or deletion.
 
     Args:
-        input (Tensor): The input tensor.
+        input (Tensor): The input Tensor.
+        dim (Union[int, tuple(int)], optional): The dimensions to sum.
+            Dim must be in the range [-rank(input), rank(input)). Default: ``None``, which indicates the sum of all
+            elements in a tensor.
+        keepdim (bool, optional): Whether the output Tensor keeps dimensions or not. Default: ``False``, indicating that no dimension is kept.
+
+    Keyword Args:
+        dtype (:class:`mindspore.dtype`, optional): The dtype of output Tensor. Default: ``None``.
 
     Returns:
-        Scalar Tensor with the same dtype as `input`, the minimum value of the input.
+        Tensor, the sum of input `input` in the given dimension dim, treating NaNs as zero.
+
+        - If dim is None, keepdim is False,
+          the output is a 0-D Tensor representing the sum of all elements in the input Tensor.
+        - If dim is int, set as 2, and keepdim is False,
+          the shape of output is :math:`(input_1, input_3, ..., input_R)`.
+        - If dim is tuple(int) or list(int), set as (2, 3), and keepdim is False,
+          the shape of output is :math:`(input_1, input_4, ..., input_R)`.
+
+    Raises:
+        TypeError: If `input` is not Tensor.
+        TypeError: If `keepdim` is not a bool.
+        TypeError: If the dtype of `input` or `dtype` is complex type.
+        ValueError: If `dim` is not in [-rank(input), rank(input)).
 
     Supported Platforms:
-        ``Ascend`` ``GPU`` ``CPU``
+        ``Ascend``
 
     Examples:
         >>> import mindspore
         >>> import numpy as np
         >>> from mindspore import Tensor, mint
-        >>> x = Tensor(np.array([0.0, 0.4, 0.6, 0.7, 0.1]), mindspore.float32)
-        >>> output = mint.min(x)
+        >>> x = Tensor(np.array([[float("nan"), 2, 3], [1, 2, float("nan")]]), mindspore.float32)
+        >>> output1 = mint.nansum(x, dim=0, keepdim=False, dtype=mindspore.float32)
+        >>> output2 = mint.nansum(x, dim=0, keepdim=True, dtype=mindspore.float32)
+        >>> print(output1)
+        [1. 4. 3.]
+        >>> print(output2)
+        [[1. 4. 3.]]
+    """
+    return _nansum_instance(*args, **kwargs)
+
+
+def remainder(*args, **kwargs):
+    r"""
+    remainder(input, other) -> Tensor
+
+    Computes the remainder of `input` divided by `other` element-wise. The result has the same sign as the divisor and
+    its absolute value is less than that of `other`.
+
+    Supports broadcasting to a common shape and implicit type promotion.
+
+    .. code:: python
+
+        remainder(input, other) == input - input.div(other, rounding_mode="floor") * other
+
+    Note:
+        Complex inputs are not supported. At least one input need to be tensor, but not both are bool tensors.
+
+    Args:
+        input (Union[Tensor, numbers.Number, bool]): The dividend is a numbers.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, numbers.Number, bool]): The divisor is a numbers.Number or
+            a bool or a tensor whose data type is number or bool when the dividend is a tensor.
+            When the dividend is Scalar, the divisor must be a Tensor whose data type is number or bool.
+
+    Returns:
+        Tensor, with dtype promoted and shape broadcasted.
+
+    Raises:
+        TypeError: If `input` and `other` are not of types: (tensor, tensor), (tensor, number), (tensor, bool),
+            (number, tensor) or (bool, tensor).
+        ValueError: If `input` and `other` are not broadcastable.
+
+    Supported Platforms:
+        ``Ascend``
+
+    Examples:
+        >>> import numpy as np
+        >>> from mindspore import Tensor, mint
+        >>> x = Tensor(np.array([-4.0, 5.0, 6.0]).astype(np.float32))
+        >>> y = Tensor(np.array([3.0, 2.0, 3.0]).astype(np.float64))
+        >>> output = mint.remainder(x, y)
         >>> print(output)
-        0.0
+        [2.  1.  0.]
+    """
+    return _remainder_instance(*args, **kwargs)
 
-    .. function:: min(input, dim, keepdim=False) -> Tensor
-        :noindex:
 
-    Calculates the minimum value along with the given dim for the input tensor, and returns the minimum values and
-    indices.
+def repeat_interleave(*args, **kwargs):
+    r"""
+    repeat_interleave(input, repeats, dim=None, *, output_size=None) -> Tensor
+
+    Repeat elements of a tensor along an axis, like :func:`mindspore.numpy.repeat`.
 
-    Args:
-        input (Tensor) - The input tensor, can be any dimension. Set the shape of input tensor as
-            :math:`(input_1, input_2, ..., input_N)` , Complex tensor is not supported.
-        dim (int): The dimension to reduce.
-        keepdim (bool, optional): Whether to reduce dimension, if ``True`` the output will keep the same dimension as the
-            input, the output will reduce dimension if ``False``. Default: ``False``.
+    .. warning::
+        Only support on Atlas A2 training series.
 
-    Returns:
-        tuple (Tensor), tuple of 2 tensors, containing the minimum value of the self tensor along the given
-        dimension `dim` and the corresponding index.
+    Args:
+        input (Tensor): The tensor to repeat values for. Must be of types: float16,
+            float32, int8, uint8, int16, int32, or int64.
+        repeats (Union[int, tuple, list, Tensor]): The number of times to repeat, must be positive.
+        dim (int, optional): The dim along which to repeat, Default: ``None``. If dims is None,
+            the input Tensor will be flattened and the output will alse be flattened.
 
-        - **values** (Tensor) - The minimum value of input tensor, with the same shape as `index`, and same dtype as `input`.
-        - **index** (Tensor) - The index for the minimum value of the input tensor, with dtype int64. If `keepdim`
-          is ``True`` , the shape of output tensors is :math:`(input_1, input_2, ..., input_{dim-1}, 1, input_{dim+1}, ..., input_N)`.
-          Otherwise, the shape is :math:`(input_1, input_2, ..., input_{dim-1}, input_{dim+1}, ..., input_N)` .
+    Keyword Args:
+        output_size (int, optional): Total output size for the given axis (e.g. sum of repeats),
+            Default: ``None``.
 
-    Raises:
-        TypeError: If `input` is not Tensor.
-        TypeError: If `keepdim` is not a bool.
-        TypeError: If `dim` is not an int.
+    Returns:
+        One tensor with values repeated along the specified dim. If input has shape
+        :math:`(s1, s2, ..., sn)` and dim is i, the output will have shape :math:`(s1, s2, ...,
+        si * repeats, ..., sn)`. The output type will be the same as the type of `input`.
 
     Supported Platforms:
-        ``Ascend`` ``GPU`` ``CPU``
+        ``Ascend``
 
     Examples:
         >>> import mindspore
         >>> import numpy as np
         >>> from mindspore import Tensor, mint
-        >>> x = Tensor(np.array([0.0, 0.4, 0.6, 0.7, 0.1]), mindspore.float32)
-        >>> output, index = mint.min(x, 0, keepdim=True)
-        >>> print(output, index)
-        [0.0] [0]
-
-    .. function:: min(input, other) -> Tensor
-        :noindex:
-
-    For details, please refer to :func:`mindspore.mint.minimum`.
+        >>> input = Tensor(np.array([[0, 1, 2], [3, 4, 5]]), mindspore.int32)
+        >>> output = mint.repeat_interleave(input, repeats=2, dim=0)
+        >>> print(output)
+        [[0 1 2]
+         [0 1 2]
+         [3 4 5]
+         [3 4 5]]
     """
-    return _min_instance(*args, **kwargs)
+    return _repeat_interleave_instance(*args, **kwargs)
 
 
 def kthvalue(*args, **kwargs):
@@ -377,323 +618,260 @@ def kthvalue(*args, **kwargs):
     return _kthvalue_instance(*args, **kwargs)
 
 
-def sub(*args, **kwargs):
+def greater_equal(*args, **kwargs):
     r"""
-    sub(input, other, *, alpha=1) -> Tensor
+    greater_equal(input, other) -> Tensor
 
-    Subtracts scaled other value from self Tensor.
+    Computes the boolean value of :math:`input >= other` element-wise.
 
     .. math::
 
-        out_{i} = self_{i} - alpha \times other_{i}
+        out_{i} =\begin{cases}
+            & \text{True,    if } input_{i}>=other_{i} \\
+            & \text{False,   if } input_{i}<other_{i}
+            \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
+        - Inputs of `input` and `other` comply with the implicit type conversion rules to make the data types
           consistent.
+        - The inputs must be two tensors or one tensor and one scalar.
+        - When the inputs are two tensors, dtypes of them cannot be bool at the same time,
+          and the shapes of them can be broadcast.
+        - When the inputs are one tensor and one scalar, the scalar could only be a constant.
+        - Broadcasting is supported.
+        - If the input Tensor can be broadcast, the low dimension will be extended to the corresponding high dimension
+          in another input by copying the value of the dimension.
 
     Args:
-        input (Union[Tensor, number.Number, bool]): `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]): `other` 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>`_.
-
-    Keyword Args:
-        alpha (number.Number, optional): A scaling factor applied to `other`, default ``1``.
+        input (Union[Tensor, Number]): The first input is a number
+            or a tensor whose data type is `number <https://www.mindspore.cn/docs/en/master/api_python/mindspore/mindspore.dtype.html#mindspore.dtype>`_ or `bool <https://www.mindspore.cn/docs/en/master/api_python/mindspore/mindspore.dtype.html#mindspore.dtype>`_.
+        other (Union[Tensor, Number]): Second input. When the first input is a Tensor, the second input should be a Number,
+            or a Tensor of the number or bool data type. When the first input is a Scalar,
+            the second input must be a Tensor of number or bool data type.
 
     Returns:
-        Tensor with a shape that is the same as the broadcasted shape of the self `self` and `other`,
-        and the data type is the one with higher precision or higher digits among the two inputs and alpha.
+        Tensor, the shape is the same as the one after broadcasting, and the data type is bool.
 
     Raises:
-        TypeError: If the type of `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 neither `input` nor `other` is a Tensor.
 
     Supported Platforms:
-        ``Ascend`` ``GPU`` ``CPU``
+        ``Ascend``
 
     Examples:
-        >>> import numpy as np
         >>> import mindspore
+        >>> import numpy as np
         >>> from mindspore import Tensor, mint
-        >>> x = Tensor(np.array([4, 5, 6]).astype(np.float32))
-        >>> y = Tensor(1, mindspore.int32)
-        >>> alpha = 0.5
-        >>> output = mint.sub(x, y, alpha=alpha)
+        >>> input = Tensor(np.array([1, 2, 3]), mindspore.int32)
+        >>> other = Tensor(np.array([1, 1, 4]), mindspore.int32)
+        >>> output = mint.greater_equal(input, other)
         >>> 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
+        [True True False]
+        >>> y = 2.1
+        >>> output = mint.greater_equal(input, y)
+        >>> print(output)
+        [False False True]
     """
-    return _sub_instance(*args, **kwargs)
+    return _greater_equal_instance(*args, **kwargs)
 
 
-def __sub__(*args, **kwargs):
+def ge(*args, **kwargs):
     r"""
-    __sub__(input, other, *, alpha=1) -> Tensor
-
-    Alias for :func:`mindspore.mint.sub`.
-
-    .. method:: mint.__sub__(input, other, *, alpha=1) -> Tensor
-        :noindex:
+    ge(input, other) -> Tensor
 
-    Alias for overload function of :func:`mindspore.mint.sub`.
+    Alias for :func:`mindspore.mint.greater_equal`.
     """
-    return _sub_instance(*args, **kwargs)
+    return _greater_equal_instance(*args, **kwargs)
 
 
-def gmm(*args, **kwargs):
+def gelu(*args, **kwargs):
     r"""
-    gmm(x, weight, bias=None, group_list=None, group_type=0, group_list_type=0) -> tuple[Tensor]
+    gelu(input, *, approximate='none') -> Tensor
 
-    Grouping matrix multiplication.
+    Gaussian Error Linear Units activation function.
 
-    .. warning::
-        - This is an experimental API that is subject to change or deletion.
-        - `group_type` must be a constant.
-        - Only support on Atlas A2 training series.
-        - When the type of `group_list` is tuple[int] or list[int], it should a non-negative non-decreasing sequence,
-          indicating indexes of each group along the split axis. In this scenario, the arg `group_list_type` is useless.
+    GeLU is described in the paper `Gaussian Error Linear Units (GELUs) <https://arxiv.org/abs/1606.08415>`_.
+    And also please refer to `BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
+    <https://arxiv.org/abs/1810.04805>`_.
 
-    .. note::
-        - When `group_type` is 2, the tensors in `x` must be non-continuous tensors which has
-          been transposed.
-        - Only when `group_type` is 0 and `bias` is None, the reverse derivative is supported,
-          which is implemented by ops.function.math_func.gmm_backward or through automatic differentiation.
+    When `approximate` argument is `none`, GELU is defined as follows:
 
-    Args:
-        x (tuple[Tensor]): The first tensors to be multiplied, whose num should be 1.
-        weight (tuple[Tensor]): The second tensors to be multiplied, whose num should be 1.
-        bias (tuple[Tensor], optional): Biases added to outputs, whose num should be 1.
-            The shape of each tensor in `bias` should be :math: `(group_list.shape[0], n)`
-            or :math: `(len(group_list), n)`. In the training scenario, the bias only supports None.
-            Default: ``None`` .
-        group_list (Union[Tensor, list[int], tuple[int]], optional): 1-D Tensor, list[int]
-            or tuple[int], indicating indexes or sizes of each group along the split axis.
-            When `group_list` is list[int] or tuple[int], it's length should be less than or equal to 128.
-            When `group_list` is a Tensor, it's size should be less than or equal to 1024.
-            Supported dtypes: int64.
-            Default: ``None`` .
+    .. math::
+        GELU(x_i) = x_i*P(X < x_i),
 
-            - If `group_list_type` is 0, it must be  a non-negative non-decreasing sequence.
-              And when `group_type` is 0, the last element in `group_list` should be equal to
-              the first dimension of the tensor in `x` . When `group_type` is 2, the last element
-              in `group_list` should be equal to the second dimension of the tensor in `x` .
+    where :math:`P` is the cumulative distribution function of the standard Gaussian distribution,
+    :math:`x_i` is the input element.
 
-            - If `group_list_type` is 1, the value in `group_list` are the sizes of each group.
-        group_type (int, optional): Represents the axes that need to be grouped. For example,
-            :math: `C[m,n] = A[m,k] \times B[k,n]`. Default: ``0`` .
+    When `approximate` argument is `tanh`, GELU is estimated with:
 
-            - If `group_type` is 0, it means that the m-axis is grouped, meaning that the shape
-              of each tensor in `x` should be :math: `(m, k)` , the shape of each tensor in `weight`
-              should be :math: `(group_list.shape[0], k, n)` or :math: `(len(group_list), k, n)`,
-              and the shape of each tensor in result would be :math: `(m, n)` .
+    .. math::
+        GELU(x_i) = 0.5 * x_i * (1 + \tanh(\sqrt(2 / \pi) * (x_i + 0.044715 * x_i^3)))
 
-            - If `group_type` is 2, it means that the k-axis is grouped, meaning that
-              the shape of each tensor in `x` should be :math: `(m, k)`, the shape of each
-              tensor in `weight` should be :math: `(k, n)`, and the shape of each tensor
-              in result would be :math: `(group_list.shape[0], m, n)` or :math: `(len(group_list), m, n)`.
-        group_list_type (int, optional): If it's 0, the value in `group_list` are the cumsum
-            result of the size of each group. If it's 1, the value in `group_list` are the size
-            of each group. Default: ``0`` .
+    GELU Activation Function Graph:
 
-    `x` , `weight` and `bias` only support the following 3 type combinations:
+    .. image:: ../images/GELU.png
+        :align: center
 
-    - x: float16, weight: float16, bias: float16
-    - x: bfloat16, weight: bfloat16, bias: float32
-    - x: float32, weight: float32, bias: float32
+    .. note::
+        On the Ascend platform, when `input` is -inf, its gradient is 0,
+        and when `input` is inf, its gradient is `dout`.
+
+    Args:
+        input (Tensor): The input of the activation function GeLU, the data type is float16, float32 or float64.
+
+    Keyword Args:
+        approximate (str, optional): the gelu approximation algorithm to use. Acceptable vaslues are ``'none'`` and ``'tanh'`` .
+            Default: ``'none'`` .
 
     Returns:
-        tuple[Tensor], the results of grouping matrix multiplication.
+        Tensor, with the same type and shape as `input`.
+
+    Raises:
+        TypeError: If `input` is not a Tensor.
+        TypeError: If dtype of `input` is not bfloat16, float16, float32 or float64.
+        ValueError: If `approximate` value is neither `none` nor `tanh`.
 
     Supported Platforms:
         ``Ascend``
 
     Examples:
+        >>> import mindspore
         >>> import numpy as np
-        >>> from mindspore import Tensor, ops
-        >>> x = Tensor(np.random.uniform(0,1, (10, 20)).astype(np.float32))
-        >>> weight = Tensor(np.random.uniform(0,1, (4, 20, 8)).astype(np.float32))
-        >>> group_list = Tensor([2, 4, 2, 2])
-        >>> y = ops.function.math_func.gmm([x,], [weight,], group_list=group_list, group_list_type=1)
-        >>> print(y[0].shape)
-        >>> (10, 8)
-        >>> group_list = [2, 6, 8, 10]
-        >>> y = ops.function.math_func.gmm([x,], [weight,], group_list=group_list, group_list_type=0)
-        >>> print(y[0].shape)
-        >>> (10, 8)
+        >>> from mindspore import Tensor, mint
+        >>> input = Tensor(np.array([[-1.0, 4.0, -8.0], [2.0, -5.0, 9.0]]), mindspore.float32)
+        >>> result = mint.nn.functional.gelu(input)
+        >>> print(result)
+        [[-1.58655241e-01  3.99987316e+00 -0.00000000e+00]
+         [ 1.95449972e+00 -1.41860323e-06  9.0000000e+00]]
+        >>> result = mint.nn.functional.gelu(input, approximate="tanh")
+        >>> print(result)
+        [[-1.58808023e-01  3.99992990e+00 -3.10779147e-21]
+         [ 1.95459759e+00 -2.29180174e-07  9.0000000e+00]]
     """
-    return _gmm_instance(*args, **kwargs)
+    return _gelu_instance(*args, **kwargs)
 
 
-def div(*args, **kwargs):
+def clamp(*args, **kwargs):
     r"""
-    div(input, other, *, rounding_mode=None) -> Tensor
-
-    Divides each element of the `input` by the corresponding element of the `other` .
-
-    .. math::
-
-        out_{i} = input_{i} / other_{i}
+    clamp(input, min=None, max=None) -> Tensor
 
-    .. note::
-        - When the two inputs have different shapes, they must be able to broadcast to a common shape.
-        - The two inputs can not be bool type at the same time,
-          [True, Tensor(True, bool\_), Tensor(np.array([True]), bool\_)] are all considered bool type.
-        - The two inputs comply with the implicit type conversion rules to make the data types
-          consistent.
+    Clamps tensor values between the specified minimum value and maximum value.
 
-    Args:
-        input (Union[Tensor, Number, bool]): The dividend.
-        other (Union[Tensor, Number, bool]): The divisor.
+    Limits the value of :math:`input` to a range, whose lower limit is `min` and upper limit is `max` .
 
-    Keyword Args:
-        rounding_mode (str, optional): Type of rounding applied to the result. Default: ``None`` .
-            Three types are defined as,
+    .. math::
 
-            - None: Default behavior, which is the same as true division in Python or `true_divide` in NumPy.
+        out_i= \left\{
+        \begin{array}{align}
+            max & \text{ if } input_i\ge max \\
+            input_i & \text{ if } min \lt input_i \lt max \\
+            min & \text{ if } input_i \le min \\
+        \end{array}\right.
 
-            - "floor": Rounds the division of the inputs down, which is the same as floor division in Python
-              or `floor_divide` in NumPy.
+    Note:
+        - `min` and `max` cannot be None at the same time;
+        - When `min` is None and `max` is not None, the elements in Tensor larger than `max` will become `max`;
+        - When `min` is not None and `max` is None, the elements in Tensor smaller than `min` will become `min`;
+        - If `min` is greater than `max`, the value of all elements in Tensor will be set to `max`;
+        - The data type of `input`, `min` and `max` should support implicit type conversion and cannot be bool type.
 
-            - "trunc": Rounds the division of the inputs towards zero, which is the same as C-style integer division.
+    Args:
+        input (Tensor): Input data, which type is Tensor. Tensors of arbitrary dimensions are supported.
+        min (Union(Tensor, float, int), optional): The minimum value. Default: ``None`` .
+        max (Union(Tensor, float, int), optional): The maximum value. Default: ``None`` .
 
     Returns:
-        Tensor, the shape is the same as the one after broadcasting,
-        and the data type is the one with higher precision or higher digits among the two inputs.
+        Tensor, a clipped Tensor.
+        The data type and shape are the same as input.
 
     Raises:
-        TypeError: If `input` and `other` is not one of the following: Tensor, Number, bool.
-        ValueError: If `rounding_mode` value is not None, "floor" or "trunc".
+        ValueError: If both `min` and `max` are None.
+        TypeError: If the type of `input` is not Tensor.
+        TypeError: If the type of `min` is not in None, Tensor, float or int.
+        TypeError: If the type of `max` is not in None, Tensor, float or int.
 
     Supported Platforms:
         ``Ascend``
 
     Examples:
+        >>> # case 1: the data type of input is Tensor
         >>> import mindspore
+        >>> from mindspore import Tensor, mint
         >>> import numpy as np
+        >>> min_value = Tensor(5, mindspore.float32)
+        >>> max_value = Tensor(20, mindspore.float32)
+        >>> input = Tensor(np.array([[1., 25., 5., 7.], [4., 11., 6., 21.]]), mindspore.float32)
+        >>> output = mint.clamp(input, min_value, max_value)
+        >>> print(output)
+        [[ 5. 20.  5.  7.]
+         [ 5. 11.  6. 20.]]
+        >>> # case 2: the data type of input is number
+        >>> import mindspore
         >>> from mindspore import Tensor, mint
-        >>> x = Tensor(np.array([1.0, 2.0, 3.0]), mindspore.float32)
-        >>> y = Tensor(np.array([4.0, 5.0, 6.0]), mindspore.float32)
-        >>> output = mint.div(x, y)
+        >>> import numpy as np
+        >>> min_value = 5
+        >>> max_value = 20
+        >>> input = Tensor(np.array([[1., 25., 5., 7.], [4., 11., 6., 21.]]), mindspore.float32)
+        >>> output = mint.clamp(input, min_value, max_value)
         >>> print(output)
-        [0.25 0.4 0.5]
+        [[ 5. 20.  5.  7.]
+         [ 5. 11.  6. 20.]]
     """
-    return _div_instance(*args, **kwargs)
+    return _clamp_instance(*args, **kwargs)
 
 
-def divide(*args, **kwargs):
+def clip(*args, **kwargs):
     r"""
-    divide(input, other, *, rounding_mode=None) -> Tensor
+    clip(input, min=None, max=None) -> Tensor
 
-    Alias for :func:`mindspore.mint.div`.
+    Alias for :func:`mindspore.mint.clamp`.
     """
-    return _div_instance(*args, **kwargs)
+    return _clamp_instance(*args, **kwargs)
 
 
-def matmul_reduce_scatter(*args, **kwargs):
+def empty_like(*args, **kwargs):
     r"""
-    matmul_reduce_scatter(input, x2, group, world_size, *, reduce_op='sum', bias=None, comm_turn=0, trans_input=False, trans_x2=False) -> Tensor
-
-    In the TP segmentation scenario, matmul and reducescatter are fused, and communication and computational
-    pipelines are parallelized within the fusion operator.
+    empty_like(input, *, dtype=None, device=None) -> Tensor
 
-    .. math::
-        output = reducescatter(input@x2)
+    Returns an uninitialized Tensor with the same shape as the `input`. Its dtype is specified by `dtype` and its
+    device is specified by `device`.
 
     .. warning::
         This is an experimental API that is subject to change or deletion.
 
     Args:
-        input (Tensor): The left matrix of matmul, the dtype supports float16 and bfloat16, the shape supports 2
-            dimensions, and the data format supports ND.
-        x2 (Tensor): The right matrix of matmul, the dtype needs to be consistent with ``input`` , the shape
-            supports 2 dimensions, and the data format supports ND.
-        group (str): Communication group name, can be created by ``create_group`` method, or use the default group
-            ``mindspore.communication.GlobalComm.WORLD_COMM_GROUP``.
-        world_size (int): The total number of ranks in the communication group, should be consistent with the number
-            of devices actually running, supporting ``2`` , ``4`` , and ``8`` .
+        input (Tensor): Tensor of any dimension.
 
     Keyword Args:
-        reduce_op (str, optional) The reduce operation type. Currently only ``'sum'`` is supported. Default:
-            ``'sum'`` .
-        bias (Tensor, optional): Currently only ``None`` is supported. Default: ``None`` .
-        comm_turn (int, optional): Indicates the granularity of communication between ranks. Currently only ``0``
-            is supported. Default: ``0`` .
-        trans_input (bool, optional): Indicates whether ``input`` is transposed. Currently only ``False`` is
-            supported. Default: ``False`` .
-        trans_x2 (bool, optional): Indicates whether ``x2`` is transposed. Default: ``False`` .
+        dtype (:class:`mindspore.dtype`, optional): The specified dtype of the output tensor. If `dtype = None`, the
+            tensor will have the same dtype as input `input`. Default ``None``.
+        device (string, optional): The specified device of the output tensor. In PyNative mode, ``"Ascend"``, ``"npu"``,
+            ``"cpu"`` and ``"CPU"`` are supported. In graph mode O0, ``"Ascend"`` and ``"npu"`` are supported. If `device = None`,
+            the value set by :func:`mindspore.set_device` will be used. Default ``None``.
 
     Returns:
-        - output (Tensor) - The result of allgather and matmul fusion calculations.
-
-    Note:
-        - When using this interface, please ensure that the driver firmware package and CANN package are both the
-          matching 8.0.RC2 version or a higher version, otherwise an error will be reported, such as BUS ERROR.
-        - The shape of ``input`` is (m, k), the shape of ``x2`` is (k, n), k is required to be equal, and the value
-          range of k is [256, 65535), and m is required to be an integer multiple of ``world_size`` . The shape of
-          ``output`` is (m * world_size, n).
-        - The common fusion operators in a model only support the same communication group.
+        Tensor, has the same shape, type and device as `input` but with uninitialized data (May be a random value).
 
     Raises:
-        TypeError: Any arg is of wrong type.
-        RuntimeError: The dtype of ``input`` or ``x2`` is neither float16 nor bfloat16.
-        RuntimeError: The dtypes of ``input`` and ``x2`` are different.
-        RuntimeError: The shape of ``input`` or ``x2`` is not two-dimensional.
-        RuntimeError: The k axis of ``input`` shape and ``x2`` shape are not equal.
-        RuntimeError: k is less than ``256`` or greater than or equal to ``65535`` .
-        RuntimeError: ``bias`` is not None.
-        RuntimeError: ``group`` does not exist.
-        RuntimeError: ``world_size`` is inconsistent with the actual number of running cards.
-        RuntimeError: ``world_size`` is not equal to ``2`` , ``4`` , or ``8`` .
-        RuntimeError: ``reduce_op`` is not ``'sum'`` .
-        RuntimeError: ``trans_input`` is ``True`` .
+        TypeError: If `input` is not a Tensor.
 
     Supported Platforms:
-        ``Ascend``
+        ``Ascend`` ``CPU``
 
     Examples:
-        .. note::
-            Before running the following examples, you need to configure the communication environment variables.
-
-            For Ascend/GPU/CPU devices, it is recommended to use the msrun startup method without any third-party or
-            configuration file dependencies. Please see the `msrun start up <https://www.mindspore.cn/tutorials/en/master/parallel/msrun_launcher.html>`_
-            for more details.
-
-            This example should be run with 2 devices.
-
-        >>> import mindspore as ms
-        >>> from mindspore import ops
-        >>> import numpy as np
-        >>> ms.communication.init()
-        >>> rank = ms.communication.get_rank()
-        >>> np.random.seed(rank)
-        >>> input = ms.Tensor(np.random.randn(1024, 256).astype(np.float32), dtype=ms.float16)
-        >>> x2 = ms.Tensor(np.random.randn(256, 512).astype(np.float32), dtype=ms.float16)
-        >>> group = ms.communication.GlobalComm.WORLD_COMM_GROUP
-        >>> world_size = ms.communication.get_group_size()
-        >>> reduce_op = ops.ReduceOp.SUM
-        >>> output = ops.matmul_reduce_scatter(
-        ...    input,
-        ...    x2,
-        ...    group,
-        ...    world_size,
-        ...    reduce_op=reduce_op,
-        ...    bias=None,
-        ...    comm_turn=0,
-        ...    trans_input=False,
-        ...    trans_x2=False,
-        ... )
-        >>> print(output.shape)
-        (512, 512)
+        >>> import mindspore
+        >>> from mindspore import ops, Tensor
+        >>> x = Tensor([[1, 2, 3], [4, 5, 6]])
+        >>> output1 = ops.empty_like(x)
+        >>> print(output1)
+        [[0 0 0]
+         [0 0 0]]
+        >>> output2 = ops.empty_like(x, dtype=mindspore.float64)
+        >>> print(output2)
+        [[0. 0. 0.]
+         [0. 0. 0.]]
     """
-    return _matmul_reduce_scatter_instance(*args, **kwargs)
+    return _empty_like_instance(*args, **kwargs)
 
 
 def add(*args, **kwargs):
@@ -715,10 +893,10 @@ def add(*args, **kwargs):
     Args:
         input (Union[Tensor, number.Number, bool]): `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>`_.
+            `bool <https://www.mindspore.cn/docs/en/master/api_python/mindspore/mindspore.dtype.html>`_.
         other (Union[Tensor, number.Number, bool]): `other` 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>`_.
+            `bool <https://www.mindspore.cn/docs/en/master/api_python/mindspore/mindspore.dtype.html>`_.
 
     Keyword Args:
         alpha (number.Number, optional): A scaling factor applied to `other`, default ``1``.
@@ -762,127 +940,272 @@ def __add__(*args, **kwargs):
     .. method:: mint.__add__(input, other, *, alpha=1) -> Tensor
         :noindex:
 
-    Alias for overload function of :func:`mindspore.mint.add`.
+    Alias for overload function of :func:`mindspore.mint.add`.
+    """
+    return _add_instance(*args, **kwargs)
+
+
+def sub(*args, **kwargs):
+    r"""
+    sub(input, other, *, alpha=1) -> Tensor
+
+    Subtracts scaled other value from self Tensor.
+
+    .. math::
+
+        out_{i} = self_{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 (Union[Tensor, number.Number, bool]): `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]): `other` 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>`_.
+
+    Keyword Args:
+        alpha (number.Number, optional): A scaling factor applied to `other`, default ``1``.
+
+    Returns:
+        Tensor with a shape that is the same as the broadcasted shape of the self `self` and `other`,
+        and the data type is the one with higher precision or higher digits among the two inputs and alpha.
+
+    Raises:
+        TypeError: If the type of `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 numpy as np
+        >>> import mindspore
+        >>> from mindspore import Tensor, mint
+        >>> x = Tensor(np.array([4, 5, 6]).astype(np.float32))
+        >>> y = Tensor(1, mindspore.int32)
+        >>> alpha = 0.5
+        >>> output = mint.sub(x, y, alpha=alpha)
+        >>> 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
+    """
+    return _sub_instance(*args, **kwargs)
+
+
+def __sub__(*args, **kwargs):
+    r"""
+    __sub__(input, other, *, alpha=1) -> Tensor
+
+    Alias for :func:`mindspore.mint.sub`.
+
+    .. method:: mint.__sub__(input, other, *, alpha=1) -> Tensor
+        :noindex:
+
+    Alias for overload function of :func:`mindspore.mint.sub`.
+    """
+    return _sub_instance(*args, **kwargs)
+
+
+def bitwise_not(*args, **kwargs):
+    r"""
+    bitwise_not(input) -> Tensor
+
+    Returns bitwise `not` of the input tensor.
+
+    .. warning::
+        This is an experimental API that is subject to change or deletion.
+
+    Args:
+        input (Tensor): The input tensor must be of integral or Boolean types.
+
+    Returns:
+        Tensor, has the same shape and type as `input`.
+
+    Raises:
+        TypeError: If `input` is not a Tensor.
+        RuntimeError: If dtype of `input` is not int or bool.
+
+    Supported Platforms:
+        ``Ascend``
+
+    Examples:
+        >>> import mindspore
+        >>> import numpy as np
+        >>> from mindspore import Tensor, mint
+        >>> x = Tensor(np.array([True, False, True, False]))
+        >>> y = mint.bitwise_not(x)
+        >>> print(y)
+        [False True False True]
+    """
+    return _bitwise_not_instance(*args, **kwargs)
+
+
+def gmm_backward_fusion(*args, **kwargs):
+    r"""
+    gmm_backward_fusion(grad, weight, *, group_list=None, group_list_type=0) -> tuple[tuple[Tensor]]
+
+    the grad of ops.function.math_func.gmm, only dx
     """
-    return _add_instance(*args, **kwargs)
+    return _gmm_backward_fusion_instance(*args, **kwargs)
 
 
-def fmod(*args, **kwargs):
+def xlogy(*args, **kwargs):
     r"""
-    fmod(input, other) -> Tensor
+    xlogy(input, other) -> Tensor
 
-    Computes the floating-point remainder of the division operation input/other.
+    Computes the first input multiplied by the logarithm of second input element-wise.
+    Returns zero when `input` is zero.
 
     .. math::
 
-        out = input - n * other
-
-    Where :math:`n` is :math:`input/other` with its fractional part truncated.
-    The returned value has the same sign as `input` and is less than `other` in magnitude.
+        out_i = input_{i}\log{other_{i}}
 
-    .. warning::
-        This is an experimental API that is subject to change or deletion.
+    Inputs of `input` and `other` comply with the implicit type conversion rules to make the data types consistent.
+    The inputs must be two tensors or one tensor and one scalar.
+    When the inputs are two tensors, the shapes of them could be broadcast.
 
     Args:
-        input (Tensor): the dividend.
-        other (Union[Tensor, Number]): the divisor.
+        input (Union[Tensor, numbers.Number, bool]): The first input is a numbers.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, numbers.Number, bool]): The second input is a numbers.Number or
+            a bool or a tensor whose data type is number or bool when the first input is a tensor.
+            When the first input is Scalar, the second input must be a Tensor whose data type is number or bool.
 
     Returns:
         Tensor, the shape is the same as the one after broadcasting,
         and the data type is the one with higher precision or higher digits among the two inputs.
 
     Raises:
-        TypeError: If `input` is not a Tensor.
+        TypeError: If `input` and `other` is not a numbers.Number or a bool or a Tensor.
+        ValueError: If `input` could not be broadcast to a tensor with shape of `other`.
 
     Supported Platforms:
-        ``Ascend``
+        ``Ascend`` ``GPU`` ``CPU``
 
     Examples:
         >>> import mindspore
         >>> import numpy as np
-        >>> from mindspore import Tensor, mint
-        >>> input = Tensor(np.array([-4., -3.5, 0, 3.5, 4]), mindspore.float32)
-        >>> output = mint.fmod(input, 2.5)
+        >>> from mindspore import Tensor, ops
+        >>> input = Tensor(np.array([-5, 0, 4]), mindspore.float32)
+        >>> other = Tensor(np.array([2, 2, 2]), mindspore.float32)
+        >>> output = ops.xlogy(input, other)
         >>> print(output)
-        [-1.5 -1.   0.   1.   1.5]
+        [-3.465736   0.        2.7725887]
     """
-    return _fmod_instance(*args, **kwargs)
+    return _xlogy_instance(*args, **kwargs)
 
 
-def pixel_shuffle(*args, **kwargs):
+def div(*args, **kwargs):
     r"""
-    pixel_shuffle(input, upscale_factor) -> Tensor
-
-    Rearrange elements in a tensor according to an upscaling factor.
+    div(input, other, *, rounding_mode=None) -> Tensor
 
-    Rearranges elements in a tensor of shape :math:`(*, C \times r^2, H, W)`
-    to a tensor of shape :math:`(*, C, H \times r, W \times r)`, where r is an upscale factor.
+    Divides each element of the `input` by the corresponding element of the `other` .
 
-    This is useful for implementing efficient sub-pixel convolution
-    with a stride of :math:`1/r`.
+    .. math::
 
-    For detailed introduction to the pixel_shuffle algorithm, refer to
-    `Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network <https://arxiv.org/abs/1609.05158>`_ .
+        out_{i} = input_{i} / other_{i}
 
-    .. warning::
-        This is an experimental API that is subject to change or deletion.
+    .. note::
+        - When the two inputs have different shapes, they must be able to broadcast to a common shape.
+        - The two inputs can not be bool type at the same time,
+          [True, Tensor(True), Tensor(np.array([True]))] are all considered bool type.
+        - The two inputs comply with the implicit type conversion rules to make the data types
+          consistent.
 
     Args:
-        input (Tensor): Tensor of shape :math:`(*, C \times r^2, H, W)` . The dimension of `input` is larger than 2,
-            and the length of third to last dimension can be divisible by the square of `upscale_factor`.
-        upscale_factor (int): factor to shuffle the input Tensor, and is a positive integer.
-            `upscale_factor` is the above-mentioned :math:`r`.
+        input (Union[Tensor, Number, bool]): The dividend.
+        other (Union[Tensor, Number, bool]): The divisor.
+
+    Keyword Args:
+        rounding_mode (str, optional): Type of rounding applied to the result. Default: ``None`` .
+            Three types are defined as,
+
+            - None: Default behavior, which is the same as true division in Python or `true_divide` in NumPy.
+
+            - "floor": Rounds the division of the inputs down, which is the same as floor division in Python
+              or `floor_divide` in NumPy.
+
+            - "trunc": Rounds the division of the inputs towards zero, which is the same as C-style integer division.
 
     Returns:
-        - **output** (Tensor) - Tensor of shape :math:`(*, C, H \times r, W \times r)` .
+        Tensor, the shape is the same as the one after broadcasting,
+        and the data type is the one with higher precision or higher digits among the two inputs.
 
     Raises:
-        ValueError: If `upscale_factor` is not a positive integer.
-        ValueError: If the length of third to last dimension is not divisible by the square of `upscale_factor`.
-        ValueError: If the dimension of `input` is less than 3.
+        TypeError: If `input` and `other` is not one of the following: Tensor, Number, bool.
+        ValueError: If `rounding_mode` value is not None, "floor" or "trunc".
 
     Supported Platforms:
         ``Ascend``
 
     Examples:
-        >>> from mindspore import mint
-        >>> input = mint.randn(1, 9, 4, 4)
-        >>> output = mint.nn.functional.pixel_shuffle(input, 3)
-        >>> print(output.shape)
-            (1, 1, 12, 12)
+        >>> import mindspore
+        >>> import numpy as np
+        >>> from mindspore import Tensor, mint
+        >>> x = Tensor(np.array([1.0, 2.0, 3.0]), mindspore.float32)
+        >>> y = Tensor(np.array([4.0, 5.0, 6.0]), mindspore.float32)
+        >>> output = mint.div(x, y)
+        >>> print(output)
+        [0.25 0.4 0.5]
     """
-    return _pixel_shuffle_instance(*args, **kwargs)
+    return _div_instance(*args, **kwargs)
 
 
-def addcdiv(*args, **kwargs):
+def divide(*args, **kwargs):
     r"""
-    addcdiv_ext(input, tensor1, tensor2, *, value=1) -> Tensor
+    divide(input, other, *, rounding_mode=None) -> Tensor
 
-    Performs the element-wise division of tensor tensor1 by tensor tensor2,
-    multiply the result by the scalar value and add it to input data.
+    Alias for :func:`mindspore.mint.div`.
+    """
+    return _div_instance(*args, **kwargs)
 
-    .. math::
-        y[i] = input[i] + value * (tensor1[i] / tensor2[i])
+
+def einsum(*args, **kwargs):
+    r"""
+    According to the Einstein summation Convention (Einsum),
+    the product of the input tensor elements is summed along the specified dimension.
+    You can use this operator to perform diagonal, reducesum, transpose, matmul, mul, inner product operations, etc.
+
+    Note:
+        The sublist format is also supported. For example, einsum_ext(op1, sublist1, op2, sublist2, ..., sublist_out).
+        In this format, equation can be derived by the sublists which are made up of Python's Ellipsis and list of
+        integers in [0, 52). Each operand is followed by a sublist and an output sublist is at the end.
+        Dynamic shape, dynamic rank input is not supported in `graph mode (mode=mindspore.GRAPH_MODE)
+        <https://www.mindspore.cn/tutorials/en/master/compile/static_graph.html>`_.
 
     .. warning::
         This is an experimental API that is subject to change or deletion.
 
     Args:
-        input (Tensor): The tensor to be added.
-        tensor1 (Tensor): The numerator tensor.
-        tensor2 (Tensor): The denominator tensor.
-
-    Keyword Args:
-        value (Number, optional): The multiplier for tensor1/tensor2. Default: ``1`` .
+        equation (str): Notation based on the Einstein summation convention, represent the operation you want to do.
+            the value can contain only letters, commas, ellipsis and arrow. The letters(must be in [a-zA-Z]) represent
+            input tensor dimension, commas(,) represent separate tensors, ellipsis indicates the tensor dimension that
+            you do not care about, the left of the arrow indicates the input tensors, and the right of it indicates the
+            desired output dimension. If there are no arrows in the equation, the letters that appear exactly once in
+            the equation will be part of the output, sorted in increasing alphabetical order. The output is computed by
+            multiplying the input operands element-wise, with their dimensions aligned based on the letters, and then
+            summing out the dimensions whose letters are not part of the output. If there is one arrow in the equation,
+            the output letters must appear at least once for some input operand and at most once for the output.
+        operands (Tensor): Input tensor used for calculation. The dtype of the tensor must be the same.
 
     Returns:
-        Tensor, has the same shape and dtype as tensor1/tensor2.
+        Tensor, the shape of it can be obtained from the `equation` , and the dtype is the same as input tensors.
 
     Raises:
-        TypeError: If dtype of `tensor1`, `tensor2`, or `input` is not tensor.
-        ValueError: If `tensor1` could not be broadcast to a tensor with shape of `tensor2`.
-        ValueError: If `value` could not be broadcast to tensors with shapes of `tensor1/tensor2`.
-        ValueError: If `input` could not be broadcast to tensors with shapes of `value*(tensor1/tensor2)`.
+        TypeError: If `equation` is invalid, or the `equation` does not match the input tensor.
+        ValueError: If the number in sublist is not in [0, 52) in sublist format.
 
     Supported Platforms:
         ``Ascend``
@@ -891,81 +1214,105 @@ def addcdiv(*args, **kwargs):
         >>> import mindspore
         >>> import numpy as np
         >>> from mindspore import Tensor, ops
-        >>> input_data = Tensor(np.array([1, 1, 1, 1]), mindspore.float32)
-        >>> x1 = Tensor(np.array([1, 2, 3, 4]), mindspore.float32)
-        >>> x2 = Tensor(np.array([4, 3, 2, 1]), mindspore.float32)
-        >>> y = ops.addcdiv_ext(input_data, x1, x2, value=1)
-        >>> print(y)
-        [1.25      1.6666667 2.5       5.       ]
+        >>> x = Tensor(np.array([1.0, 2.0, 4.0]), mindspore.float32)
+        >>> equation = "i->"
+        >>> output = ops.einsum_ext(equation, x)
+        >>> print(output)
+        7.0
+        >>> x = Tensor(np.array([1.0, 2.0, 4.0]), mindspore.float32)
+        >>> y = Tensor(np.array([2.0, 4.0, 3.0]), mindspore.float32)
+        >>> equation = "i,i->i"
+        >>> output = ops.einsum_ext(equation, x, y)
+        >>> print(output)
+        [ 2. 8. 12.]
+        >>> x = Tensor(np.array([[1.0, 2.0, 3.0], [4.0, 5.0, 6.0]]), mindspore.float32)
+        >>> y = Tensor(np.array([[2.0, 3.0], [1.0, 2.0], [4.0, 5.0]]), mindspore.float32)
+        >>> equation = "ij,jk->ik"
+        >>> output = ops.einsum_ext(equation, x, y)
+        >>> print(output)
+        [[16. 22.]
+         [37. 52.]]
+        >>> x = Tensor(np.array([[1.0, 2.0, 3.0], [4.0, 5.0, 6.0]]), mindspore.float32)
+        >>> equation = "ij->ji"
+        >>> output = ops.einsum_ext(equation, x)
+        >>> print(output)
+        [[1. 4.]
+         [2. 5.]
+         [3. 6.]]
+        >>> x = Tensor(np.array([[1.0, 2.0, 3.0], [4.0, 5.0, 6.0]]), mindspore.float32)
+        >>> equation = "ij->j"
+        >>> output = ops.einsum_ext(equation, x)
+        >>> print(output)
+        [5. 7. 9.]
+        >>> x = Tensor(np.array([[1.0, 2.0, 3.0], [4.0, 5.0, 6.0]]), mindspore.float32)
+        >>> equation = "...->"
+        >>> output = ops.einsum_ext(equation, x)
+        >>> print(output)
+        21.0
+        >>> x = Tensor(np.array([1.0, 2.0, 3.0]), mindspore.float32)
+        >>> y = Tensor(np.array([2.0, 4.0, 1.0]), mindspore.float32)
+        >>> equation = "j,i->ji"
+        >>> output = ops.einsum_ext(equation, x, y)
+        >>> print(output)
+        [[ 2. 4. 1.]
+         [ 4. 8. 2.]
+         [ 6. 12. 3.]]
+        >>> x = mindspore.Tensor([1, 2, 3, 4], mindspore.float32)
+        >>> y = mindspore.Tensor([1, 2], mindspore.float32)
+        >>> output = ops.einsum_ext(x, [..., 1], y, [..., 2], [..., 1, 2])
+        >>> print(output)
+        [[1. 2.]
+         [2. 4.]
+         [3. 6.]
+         [4. 8.]]
     """
-    return _addcdiv_instance(*args, **kwargs)
+    return _einsum_instance(*args, **kwargs)
 
 
-def gelu(*args, **kwargs):
+def any(*args, **kwargs):
     r"""
-    gelu(input, *, approximate='none') -> Tensor
-
-    Gaussian Error Linear Units activation function.
-
-    GeLU is described in the paper `Gaussian Error Linear Units (GELUs) <https://arxiv.org/abs/1606.08415>`_.
-    And also please refer to `BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
-    <https://arxiv.org/abs/1810.04805>`_.
-
-    When `approximate` argument is `none`, GELU is defined as follows:
+    any(input) -> Tensor
 
-    .. math::
-        GELU(x_i) = x_i*P(X < x_i),
+    Check if ``True`` is present in `input` .
 
-    where :math:`P` is the cumulative distribution function of the standard Gaussian distribution,
-    :math:`x_i` is the input element.
+    Args:
+        input (Tensor): The input tensor.
 
-    When `approximate` argument is `tanh`, GELU is estimated with:
+    Returns:
+        Tensor
 
-    .. math::
-        GELU(x_i) = 0.5 * x_i * (1 + \tanh(\sqrt(2 / \pi) * (x_i + 0.044715 * x_i^3)))
+    Supported Platforms:
+        ``Ascend`` ``GPU`` ``CPU``
 
-    GELU Activation Function Graph:
+    Examples:
+        >>> import mindspore
+        >>> input = mindspore.tensor([[True, False], [True, True]])
+        >>> mindspore.ops.functional_overload.any(input)
+        Tensor(shape=[], dtype=Bool, value= True)
 
-    .. image:: ../images/GELU.png
-        :align: center
+    .. function:: any(input, dim, keepdim=False) -> Tensor
+        :noindex:
 
-    .. note::
-        On the Ascend platform, when `input` is -inf, its gradient is 0,
-        and when `input` is inf, its gradient is `dout`.
+    Check if ``True`` is present in the specified dimension of `input` .
 
     Args:
-        input (Tensor): The input of the activation function GeLU, the data type is float16, float32 or float64.
-
-    Keyword Args:
-        approximate (str, optional): the gelu approximation algorithm to use. Acceptable vaslues are ``'none'`` and ``'tanh'`` .
-            Default: ``'none'`` .
+        input (Tensor): The input tensor.
+        dim (int): The dimensions to reduce.
+        keepdim (bool, optional): Whether the output tensor has dim retained or not. Default ``False`` .
 
     Returns:
-        Tensor, with the same type and shape as `input`.
-
-    Raises:
-        TypeError: If `input` is not a Tensor.
-        TypeError: If dtype of `input` is not bfloat16, float16, float32 or float64.
-        ValueError: If `approximate` value is neither `none` nor `tanh`.
+        Tensor
 
     Supported Platforms:
-        ``Ascend``
+        ``Ascend`` ``GPU`` ``CPU``
 
     Examples:
         >>> import mindspore
-        >>> import numpy as np
-        >>> from mindspore import Tensor, mint
-        >>> input = Tensor(np.array([[-1.0, 4.0, -8.0], [2.0, -5.0, 9.0]]), mindspore.float32)
-        >>> result = mint.nn.functional.gelu(input)
-        >>> print(result)
-        [[-1.58655241e-01  3.99987316e+00 -0.00000000e+00]
-         [ 1.95449972e+00 -1.41860323e-06  9.0000000e+00]]
-        >>> result = mint.nn.functional.gelu(input, approximate="tanh")
-        >>> print(result)
-        [[-1.58808023e-01  3.99992990e+00 -3.10779147e-21]
-         [ 1.95459759e+00 -2.29180174e-07  9.0000000e+00]]
+        >>> input = mindspore.tensor([[True, False], [True, True]])
+        >>> mindspore.ops.functional_overload.any(input, dim=1)
+        Tensor(shape=[2], dtype=Bool, value= [ True,  True])
     """
-    return _gelu_instance(*args, **kwargs)
+    return _any_instance(*args, **kwargs)
 
 
 def max(*args, **kwargs):
@@ -1039,152 +1386,223 @@ def max(*args, **kwargs):
     return _max_instance(*args, **kwargs)
 
 
-def floor_divide(*args, **kwargs):
+def fmod(*args, **kwargs):
     r"""
-    Divides the first input tensor by the second input tensor element-wise and round down to the closest integer.
+    fmod(input, other) -> Tensor
 
-    Inputs of `input` and `other` comply with the implicit type conversion rules to make the data types consistent.
-    Inputs must be two tensors or one tensor and one scalar.
-    When the inputs are two tensors,
-    dtypes of them cannot be bool at the same time, and the shapes of them could be broadcast.
-    When the inputs are one tensor and one scalar,
-    the scalar could only be a constant.
+    Computes the floating-point remainder of the division operation input/other.
+
+    .. math::
+
+        out = input - n * other
+
+    Where :math:`n` is :math:`input/other` with its fractional part truncated.
+    The returned value has the same sign as `input` and is less than `other` in magnitude.
+
+    .. warning::
+        This is an experimental API that is subject to change or deletion.
+
+    Args:
+        input (Tensor): the dividend.
+        other (Union[Tensor, Number]): the divisor.
+
+    Returns:
+        Tensor, the shape is the same as the one after broadcasting,
+        and the data type is the one with higher precision or higher digits among the two inputs.
+
+    Raises:
+        TypeError: If `input` is not a Tensor.
+
+    Supported Platforms:
+        ``Ascend``
+
+    Examples:
+        >>> import mindspore
+        >>> import numpy as np
+        >>> from mindspore import Tensor, mint
+        >>> input = Tensor(np.array([-4., -3.5, 0, 3.5, 4]), mindspore.float32)
+        >>> output = mint.fmod(input, 2.5)
+        >>> print(output)
+        [-1.5 -1.   0.   1.   1.5]
+    """
+    return _fmod_instance(*args, **kwargs)
 
-    .. math::
-        out_{i} = \text{floor}( \frac{input_i}{other_i})
 
-    where the :math:`floor` indicates the Floor operator. For more details,
-    please refer to the :class:`mindspore.mint.floor` operator.
+def min(*args, **kwargs):
+    r"""
+    min(input) -> Tensor
 
-    .. warning::
-        This is an experimental API that is subject to change or deletion.
+    Returns the minimum value of the input tensor.
 
     Args:
-        input (Union[Tensor, Number, bool]): The first input is a number or
-            a bool or a tensor whose data type is number or bool.
-        other (Union[Tensor, Number, bool]): The second input is a number or
-            a bool or a tensor whose data type is number or bool.
+        input (Tensor): The input tensor.
 
     Returns:
-        Tensor, the shape is the same as the one after broadcasting,
-        and the data type is the one with higher precision or higher digits among the two inputs.
-
-    Raises:
-        TypeError: If `input` and `other` are not the following: Tensor, number.Number or bool.
+        Scalar Tensor with the same dtype as `input`, the minimum value of the input.
 
     Supported Platforms:
         ``Ascend`` ``GPU`` ``CPU``
 
     Examples:
         >>> import mindspore
-        >>> from mindspore import Tensor, mint
         >>> import numpy as np
-        >>> input = Tensor(np.array([2, 4, -1]), mindspore.int32)
-        >>> other = Tensor(np.array([3, 3, 3]), mindspore.int32)
-        >>> output = mint.floor_divide(input, other)
-        >>> print(output)
-        [ 0  1 -1]
-        >>> input = Tensor(2.0, mindspore.float32)
-        >>> other = Tensor(2.0, mindspore.float32)
-        >>> output = mint.floor_divide(input, other)
+        >>> from mindspore import Tensor, mint
+        >>> x = Tensor(np.array([0.0, 0.4, 0.6, 0.7, 0.1]), mindspore.float32)
+        >>> output = mint.min(x)
         >>> print(output)
-        1.0
-    """
-    return _floor_divide_instance(*args, **kwargs)
-
-
-def nansum(*args, **kwargs):
-    r"""
-    nansum(input, dim=None, keepdim=False, *, dtype=None) -> Tensor
+        0.0
 
-    Computes sum of `input` over a given dimension, treating NaNs as zero.
+    .. function:: min(input, dim, keepdim=False) -> Tensor
+        :noindex:
 
-    .. warning::
-        It is only supported on Atlas A2 Training Series Products.
-        This is an experimental API that is subject to change or deletion.
+    Calculates the minimum value along with the given dim for the input tensor, and returns the minimum values and
+    indices.
 
     Args:
-        input (Tensor): The input Tensor.
-        dim (Union[int, tuple(int)], optional): The dimensions to sum.
-            Dim must be in the range [-rank(input), rank(input)). Default: ``None``, which indicates the sum of all
-            elements in a tensor.
-        keepdim (bool, optional): Whether the output Tensor keeps dimensions or not. Default: ``False``, indicating that no dimension is kept.
-
-    Keyword Args:
-        dtype (:class:`mindspore.dtype`, optional): The dtype of output Tensor. Default: ``None``.
+        input (Tensor) - The input tensor, can be any dimension. Set the shape of input tensor as
+            :math:`(input_1, input_2, ..., input_N)` , Complex tensor is not supported.
+        dim (int): The dimension to reduce.
+        keepdim (bool, optional): Whether to reduce dimension, if ``True`` the output will keep the same dimension as the
+            input, the output will reduce dimension if ``False``. Default: ``False``.
 
     Returns:
-        Tensor, the sum of input `input` in the given dimension dim, treating NaNs as zero.
+        tuple (Tensor), tuple of 2 tensors, containing the minimum value of the self tensor along the given
+        dimension `dim` and the corresponding index.
 
-        - If dim is None, keepdim is False,
-          the output is a 0-D Tensor representing the sum of all elements in the input Tensor.
-        - If dim is int, set as 2, and keepdim is False,
-          the shape of output is :math:`(input_1, input_3, ..., input_R)`.
-        - If dim is tuple(int) or list(int), set as (2, 3), and keepdim is False,
-          the shape of output is :math:`(input_1, input_4, ..., input_R)`.
+        - **values** (Tensor) - The minimum value of input tensor, with the same shape as `index`, and same dtype as `input`.
+        - **index** (Tensor) - The index for the minimum value of the input tensor, with dtype int64. If `keepdim`
+          is ``True`` , the shape of output tensors is :math:`(input_1, input_2, ..., input_{dim-1}, 1, input_{dim+1}, ..., input_N)`.
+          Otherwise, the shape is :math:`(input_1, input_2, ..., input_{dim-1}, input_{dim+1}, ..., input_N)` .
 
     Raises:
         TypeError: If `input` is not Tensor.
         TypeError: If `keepdim` is not a bool.
-        TypeError: If the dtype of `input` or `dtype` is complex type.
-        ValueError: If `dim` is not in [-rank(input), rank(input)).
+        TypeError: If `dim` is not an int.
 
     Supported Platforms:
-        ``Ascend``
+        ``Ascend`` ``GPU`` ``CPU``
 
     Examples:
         >>> import mindspore
         >>> import numpy as np
         >>> from mindspore import Tensor, mint
-        >>> x = Tensor(np.array([[float("nan"), 2, 3], [1, 2, float("nan")]]), mindspore.float32)
-        >>> output1 = mint.nansum(x, dim=0, keepdim=False, dtype=mindspore.float32)
-        >>> output2 = mint.nansum(x, dim=0, keepdim=True, dtype=mindspore.float32)
-        >>> print(output1)
-        [1. 4. 3.]
-        >>> print(output2)
-        [[1. 4. 3.]]
+        >>> x = Tensor(np.array([0.0, 0.4, 0.6, 0.7, 0.1]), mindspore.float32)
+        >>> output, index = mint.min(x, 0, keepdim=True)
+        >>> print(output, index)
+        [0.0] [0]
+
+    .. function:: min(input, other) -> Tensor
+        :noindex:
+
+    For details, please refer to :func:`mindspore.mint.minimum`.
     """
-    return _nansum_instance(*args, **kwargs)
+    return _min_instance(*args, **kwargs)
 
 
-def empty(*args, **kwargs):
+def all_gather_matmul(*args, **kwargs):
     r"""
-    empty(*size, dtype=None, device=None) -> Tensor
+    all_gather_matmul(input, x2, group, world_size, *, bias=None, gather_index=0, gather_output=True, comm_turn=0, trans_input=False, trans_x2=False) -> Tensor
 
-    Creates a tensor with uninitialized data, whose shape, dtype and device are described by the argument `size`,
-    `dtype` and `device` respectively.
+    In the TP segmentation scenario, allgather and matmul are fused, and communication and computational pipelines
+    are parallelized within the fusion operator.
+
+    .. math::
+        output = allgather(input)@x2
+
+        gather\_out = allgather(input)
 
     .. warning::
         This is an experimental API that is subject to change or deletion.
 
     Args:
-        size (Union[tuple[int], list[int], int]): The specified shape of output tensor. Can be variable numbers of
-            positive integers or tupled or list containing positive integers.
+        input (Tensor): The left matrix of matmul, the dtype supports float16 and bfloat16, the shape supports 2
+            dimensions, and the data format supports ND.
+        x2 (Tensor): The right matrix of matmul, the dtype needs to be consistent with ``input`` , the shape
+            supports 2 dimensions, and the data format supports ND.
+        group (str): Communication group name, can be created by ``create_group`` method, or use the default group
+            ``mindspore.communication.GlobalComm.WORLD_COMM_GROUP``.
+        world_size (int): The total number of ranks in the communication group, should be consistent with the number
+            of devices actually running, supporting ``2`` , ``4`` , and ``8`` .
 
     Keyword Args:
-        dtype (:class:`mindspore.dtype`, optional): The specified type of output tensor. If `dtype` is ``None`` ,
-            `mindspore.float32` will be used. Default: ``None`` .
-        device (string, optional): The specified device of the output tensor. Support ``CPU`` and ``Ascend``. If
-            `device = None`, `mindspore.context.device_target` will be used. Default ``None``.
+        bias (Tensor, optional): Currently only ``None`` is supported. Default: ``None`` .
+        gather_index (int, optional): Indicates the allgather operation object, ``0`` means gather ``input`` ,
+            ``1`` means gather ``x2`` . Currently only ``0`` is supported. Default: ``0`` .
+        gather_output (bool, optional): Indicates whether gather output is required. Default: ``True`` .
+        comm_turn (int, optional): Indicates the granularity of communication between ranks. Currently only ``0``
+            is supported. Default: ``0`` .
+        trans_input (bool, optional): Indicates whether ``input`` is transposed. Currently only ``False`` is
+            supported. Default: ``False`` .
+        trans_x2 (bool, optional): Indicates whether ``x2`` is transposed. Default: ``False`` .
 
     Returns:
-        Tensor, whose dtype and size are defined by input.
+        - output (Tensor) - The result of allgather and matmul fusion calculations.
+        - gather_out (Tensor) - The result of allgather. If gather_output is ``False`` , ``gather_out`` returns a
+          tensor with shape 0.
+
+    Note:
+        - When using this interface, please ensure that the driver firmware package and CANN package are both the
+          matching 8.0.RC2 version or a higher version, otherwise an error will be reported, such as BUS ERROR.
+        - The shape of ``input`` is (m, k), the shape of ``x2`` is (k, n), k is required to be equal, and the value
+          range of k is [256, 65535). The shape of ``output`` is (m * world_size, n), and the shape of
+          ``gather_out`` is (m * world_size, k).
+        - The common fusion operators in a model only support the same communication group.
 
     Raises:
-        TypeError:  If `size` is neither an int nor a tuple or list of int.
+        TypeError: Any arg is of wrong type.
+        RuntimeError: The dtype of ``input`` or ``x2`` is neither float16 nor bfloat16.
+        RuntimeError: The dtypes of ``input`` and ``x2`` are different.
+        RuntimeError: The shape of ``input`` or ``x2`` is not two-dimensional.
+        RuntimeError: The k axis of ``input`` shape and ``x2`` shape are not equal.
+        RuntimeError: k is less than ``256`` or greater than or equal to ``65535`` .
+        RuntimeError: ``bias`` is not None.
+        RuntimeError: ``group`` does not exist.
+        RuntimeError: ``world_size`` is inconsistent with the actual number of running cards.
+        RuntimeError: ``world_size`` is not equal to ``2`` , ``4`` , or ``8`` .
+        RuntimeError: ``gather_index`` is not ``0`` .
+        RuntimeError: ``trans_input`` is ``True`` .
 
     Supported Platforms:
         ``Ascend``
 
     Examples:
-        >>> import mindspore
+        .. note::
+            Before running the following examples, you need to configure the communication environment variables.
+
+            For Ascend/GPU/CPU devices, it is recommended to use the msrun startup method without any third-party or
+            configuration file dependencies. Please see the `msrun start up <https://www.mindspore.cn/tutorials/en/master/parallel/msrun_launcher.html>`_
+            for more details.
+
+            This example should be run with 2 devices.
+
+        >>> import mindspore as ms
+        >>> import numpy as np
         >>> from mindspore import ops
-        >>> output = ops.empty((2, 3), dtype=mindspore.float32)
-        >>> print(output)
-        [[0. 0. 0.]
-         [0. 0. 0.]]
+        >>> ms.communication.init()
+        >>> rank = ms.communication.get_rank()
+        >>> np.random.seed(rank)
+        >>> input = ms.Tensor(np.random.randn(128, 256).astype(np.float32), dtype=ms.float16)
+        >>> x2 = ms.Tensor(np.random.randn(256, 512).astype(np.float32), dtype=ms.float16)
+        >>> group = ms.communication.GlobalComm.WORLD_COMM_GROUP
+        >>> world_size = ms.communication.get_group_size()
+        >>> output, gather_out = ops.all_gather_matmul(
+        ...    input,
+        ...    x2,
+        ...    group,
+        ...    world_size,
+        ...    bias=None,
+        ...    gather_index=0,
+        ...    gather_output=True,
+        ...    comm_turn=0,
+        ...    trans_input=False,
+        ...    trans_x2=False,
+        ... )
+        >>> print(output.shape)
+        (256, 512)
+        >>> print(gather_out.shape)
+        (256, 256)
     """
-    return _empty_instance(*args, **kwargs)
+    return _all_gather_matmul_instance(*args, **kwargs)
 
 
 def gmm_backward(*args, **kwargs):
@@ -1196,142 +1614,174 @@ def gmm_backward(*args, **kwargs):
     return _gmm_backward_instance(*args, **kwargs)
 
 
-def remainder(*args, **kwargs):
+def pixel_shuffle(*args, **kwargs):
     r"""
-    remainder(input, other) -> Tensor
+    pixel_shuffle(input, upscale_factor) -> Tensor
 
-    Computes the remainder of `input` divided by `other` element-wise. The result has the same sign as the divisor and
-    its absolute value is less than that of `other`.
+    Rearrange elements in a tensor according to an upscaling factor.
 
-    Supports broadcasting to a common shape and implicit type promotion.
+    Rearranges elements in a tensor of shape :math:`(*, C \times r^2, H, W)`
+    to a tensor of shape :math:`(*, C, H \times r, W \times r)`, where r is an upscale factor.
 
-    .. code:: python
+    This is useful for implementing efficient sub-pixel convolution
+    with a stride of :math:`1/r`.
 
-        remainder(input, other) == input - input.div(other, rounding_mode="floor") * other
+    For detailed introduction to the pixel_shuffle algorithm, refer to
+    `Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network <https://arxiv.org/abs/1609.05158>`_ .
 
-    Note:
-        Complex inputs are not supported. At least one input need to be tensor, but not both are bool tensors.
+    .. warning::
+        This is an experimental API that is subject to change or deletion.
 
     Args:
-        input (Union[Tensor, numbers.Number, bool]): The dividend is a numbers.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, numbers.Number, bool]): The divisor is a numbers.Number or
-            a bool or a tensor whose data type is number or bool\_ when the dividend is a tensor.
-            When the dividend is Scalar, the divisor must be a Tensor whose data type is number or bool\_.
+        input (Tensor): Tensor of shape :math:`(*, C \times r^2, H, W)` . The dimension of `input` is larger than 2,
+            and the length of third to last dimension can be divisible by the square of `upscale_factor`.
+        upscale_factor (int): factor to shuffle the input Tensor, and is a positive integer.
+            `upscale_factor` is the above-mentioned :math:`r`.
 
     Returns:
-        Tensor, with dtype promoted and shape broadcasted.
+        - **output** (Tensor) - Tensor of shape :math:`(*, C, H \times r, W \times r)` .
 
     Raises:
-        TypeError: If `input` and `other` are not of types: (tensor, tensor), (tensor, number), (tensor, bool),
-            (number, tensor) or (bool, tensor).
-        ValueError: If `input` and `other` are not broadcastable.
+        ValueError: If `upscale_factor` is not a positive integer.
+        ValueError: If the length of third to last dimension is not divisible by the square of `upscale_factor`.
+        ValueError: If the dimension of `input` is less than 3.
 
     Supported Platforms:
         ``Ascend``
 
     Examples:
-        >>> import numpy as np
-        >>> from mindspore import Tensor, mint
-        >>> x = Tensor(np.array([-4.0, 5.0, 6.0]).astype(np.float32))
-        >>> y = Tensor(np.array([3.0, 2.0, 3.0]).astype(np.float64))
-        >>> output = mint.remainder(x, y)
-        >>> print(output)
-        [2.  1.  0.]
+        >>> from mindspore import mint
+        >>> input = mint.randn(1, 9, 4, 4)
+        >>> output = mint.nn.functional.pixel_shuffle(input, 3)
+        >>> print(output.shape)
+            (1, 1, 12, 12)
     """
-    return _remainder_instance(*args, **kwargs)
+    return _pixel_shuffle_instance(*args, **kwargs)
 
 
-def repeat_interleave(*args, **kwargs):
+def quant_matmul(*args, **kwargs):
     r"""
-    repeat_interleave(input, repeats, dim=None, *, output_size=None) -> Tensor
+    quant_matmul(x1, x2, scale, *, offset=None, pertoken_scale=None, bias=None, output_dtype=None, x1_dtype=None, x2_dtype=None, pertoken_scale_dtype=None, scale_dtype=None, group_sizes=None) -> Tensor
 
-    Repeat elements of a tensor along an axis, like :func:`mindspore.numpy.repeat`.
+    Used for quantized matrix multiplication.
 
     .. warning::
-        Only support on Atlas A2 training series.
+        This is an experimental API that is subject to change or deletion.
+        Only support on David training series.
 
     Args:
-        input (Tensor): The tensor to repeat values for. Must be of types: float16,
-            float32, int8, uint8, int16, int32, or int64.
-        repeats (Union[int, tuple, list, Tensor]): The number of times to repeat, must be positive.
-        dim (int, optional): The dim along which to repeat, Default: ``None``. If dims is None,
-            the input Tensor will be flattened and the output will alse be flattened.
+        x1 (Tensor): Tensor of shape :math:`(*, M, K)` . The dimension of `input` should be in [2, 6].
+        x2 (Tensor): Tensor of shape :math:`(*, K, N)` . The dimension of `input` should be in [2, 6].
+        scale (Tensor): Tensor of shape :math:`(T,)` . T should be equal to 1 or N, N is the last dimension of `x2`.
 
     Keyword Args:
-        output_size (int, optional): Total output size for the given axis (e.g. sum of repeats),
-            Default: ``None``.
+        offset (Tensor, optional): Tensor of shape :math:`(T,)` . T should be equal to 1 or N, N is the last dimension of `x2`. Default: ``None`` .
+        pertoken_scale (Tensor, optional): Tensor of shape :math:`(M,)` . M is second-to-last dimension of `x1`. Default: ``None`` .
+            A valid Tensor must deliver to `pertoken_scale` , ``None`` will cause unexpected error.
+        bias (Tensor, optional): Tensor of shape :math:`(N,)` or :math:`(B, 1, N)` , N is the last dimension of `x2`.
+            If dimension of `output` is 2, 4, 5 or 6, `bias` must has shape :math:`(N,)` . Default: ``None`` .
+        output_dtype (:class:`mindspore.dtype`, optional): the dtype of `output`. Default: ``None`` .
+        x1_dtype (:class:`mindspore.dtype`, optional): Cast `x1` to `x1_dtype` before calculation. Default: ``None`` .
+        x2_dtype (:class:`mindspore.dtype`, optional): Cast `x2` to `x2_dtype` before calculation. Default: ``None`` .
+        pertoken_scale_dtype (:class:`mindspore.dtype`, optional): Cast `pertoken_scale` to `pertoken_scale_dtype` before calculation. Default: ``None`` .
+        scale_dtype (:class:`mindspore.dtype`, optional): Cast `scale` to `scale_dtype` before calculation. Default: ``None`` .
+        group_sizes (Union[tuple(int), list(int)], optional): A sequence of int elements. Must have 3 elements. Default: ``None`` .
 
     Returns:
-        One tensor with values repeated along the specified dim. If input has shape
-        :math:`(s1, s2, ..., sn)` and dim is i, the output will have shape :math:`(s1, s2, ...,
-        si * repeats, ..., sn)`. The output type will be the same as the type of `input`.
+        Tensor of shape :math:`(*, M, N)` .
+
+    Raises:
+        ValueError: If dtype of `x1` is int8 or int32.
 
     Supported Platforms:
         ``Ascend``
 
     Examples:
-        >>> import mindspore
         >>> import numpy as np
-        >>> from mindspore import Tensor, mint
-        >>> input = Tensor(np.array([[0, 1, 2], [3, 4, 5]]), mindspore.int32)
-        >>> output = mint.repeat_interleave(input, repeats=2, dim=0)
-        >>> print(output)
-        [[0 1 2]
-         [0 1 2]
-         [3 4 5]
-         [3 4 5]]
+        >>> import mindspore as ms
+        >>> from mindspore import ops, Tensor
+        >>> x1 = Tensor(np.random.randn(2, 3, 4), ms.float8_e4m3)
+        >>> x2 = Tensor(np.random.randn(2, 4, 5), ms.float8_e4m3)
+        >>> scale = Tensor(np.random.randn(1,), ms.float32)
+        >>> pertoken_scale = Tensor(np.random.randn(3,), ms.float32)
+        >>> output = ops.auto_generate.quant_matmul(x1, x2, scale, pertoken_scale=pertoken_scale, output_dtype=ms.bfloat16)
+        >>> print(output.shape)
+        (2, 3, 5)
+        >>> print(output.dtype)
+        BFloat16
     """
-    return _repeat_interleave_instance(*args, **kwargs)
+    return _quant_matmul_instance(*args, **kwargs)
 
 
-def bitwise_not(*args, **kwargs):
+def bernoulli_(*args, **kwargs):
     r"""
-    bitwise_not(input) -> Tensor
+    bernoulli_(input, p, seed, offset) -> Tensor
 
-    Returns bitwise `not` of the input tensor.
+    Inner function, used for Tensor.bernoulli_.
+    """
+    return _bernoulli__instance(*args, **kwargs)
+
+
+def lerp(*args, **kwargs):
+    r"""
+    lerp(input, end, weight) -> Tensor
+
+    Perform a linear interpolation of two tensors input and end based on a float or tensor weight.
+
+    If `weight` is a tensor, the shapes of three inputs need to be broadcast;
+    If `weight` is a float, the shapes of `input` and `end` need to be broadcast.
+    If `weight` is a float and platform is Ascend, the types of `input` and `end` need to be float32.
 
     .. warning::
         This is an experimental API that is subject to change or deletion.
 
+    .. math::
+        output_{i} = input_{i} + weight_{i} * (end_{i} - input_{i})
+
     Args:
-        input (Tensor): The input tensor must be of integral or Boolean types.
+        input (Tensor): The tensor with the starting points. Data type must be float16 or float32.
+        end (Tensor): The tensor with the ending points. Data type must be the same as `input`.
+        weight (Union[float, Tensor]): The weight for the interpolation formula. Must be a float scalar
+            or a tensor with float16 or float32 data type.
 
     Returns:
-        Tensor, has the same shape and type as `input`.
+        Tensor, has the same type and shape as input `input`.
 
     Raises:
-        TypeError: If `input` is not a Tensor.
-        RuntimeError: If dtype of `input` is not int or bool.
+        TypeError: If `input` or `end` is not a tensor.
+        TypeError: If `weight` is neither scalar(float) nor tensor.
+        TypeError: If dtype of `input` or `end` is neither float16 nor float32.
+        TypeError: If dtype of `weight` is neither float16 nor float32 when it is a tensor.
+        TypeError: If `input` and `end` have different data types.
+        TypeError: If `input`, `end` and `weight` have different data types when `weight` is a tensor.
+        ValueError: If `end` could not be broadcast to a tensor with shape of `input`.
+        ValueError: If `weight` could not be broadcast to tensors with shapes of `input` and `end` when it is a tensor.
 
     Supported Platforms:
-        ``Ascend``
+        ``Ascend`` ``GPU`` ``CPU``
 
     Examples:
         >>> import mindspore
         >>> import numpy as np
         >>> from mindspore import Tensor, mint
-        >>> x = Tensor(np.array([True, False, True, False]))
-        >>> y = mint.bitwise_not(x)
-        >>> print(y)
-        [False True False True]
+        >>> start = Tensor(np.array([1., 2., 3., 4.]), mindspore.float32)
+        >>> end = Tensor(np.array([10., 10., 10., 10.]), mindspore.float32)
+        >>> output = mint.lerp(start, end, 0.5)
+        >>> print(output)
+        [5.5 6. 6.5 7. ]
     """
-    return _bitwise_not_instance(*args, **kwargs)
+    return _lerp_instance(*args, **kwargs)
 
 
-def all_gather_matmul(*args, **kwargs):
+def matmul_reduce_scatter(*args, **kwargs):
     r"""
-    all_gather_matmul(input, x2, group, world_size, *, bias=None, gather_index=0, gather_output=True, comm_turn=0, trans_input=False, trans_x2=False) -> Tensor
+    matmul_reduce_scatter(input, x2, group, world_size, *, reduce_op='sum', bias=None, comm_turn=0, trans_input=False, trans_x2=False) -> Tensor
 
-    In the TP segmentation scenario, allgather and matmul are fused, and communication and computational pipelines
-    are parallelized within the fusion operator.
+    In the TP segmentation scenario, matmul and reducescatter are fused, and communication and computational
+    pipelines are parallelized within the fusion operator.
 
     .. math::
-        output = allgather(input)@x2
-
-        gather\_out = allgather(input)
+        output = reducescatter(input@x2)
 
     .. warning::
         This is an experimental API that is subject to change or deletion.
@@ -1347,10 +1797,9 @@ def all_gather_matmul(*args, **kwargs):
             of devices actually running, supporting ``2`` , ``4`` , and ``8`` .
 
     Keyword Args:
+        reduce_op (str, optional) The reduce operation type. Currently only ``'sum'`` is supported. Default:
+            ``'sum'`` .
         bias (Tensor, optional): Currently only ``None`` is supported. Default: ``None`` .
-        gather_index (int, optional): Indicates the allgather operation object, ``0`` means gather ``input`` ,
-            ``1`` means gather ``x2`` . Currently only ``0`` is supported. Default: ``0`` .
-        gather_output (bool, optional): Indicates whether gather output is required. Default: ``True`` .
         comm_turn (int, optional): Indicates the granularity of communication between ranks. Currently only ``0``
             is supported. Default: ``0`` .
         trans_input (bool, optional): Indicates whether ``input`` is transposed. Currently only ``False`` is
@@ -1359,15 +1808,13 @@ def all_gather_matmul(*args, **kwargs):
 
     Returns:
         - output (Tensor) - The result of allgather and matmul fusion calculations.
-        - gather_out (Tensor) - The result of allgather. If gather_output is ``False`` , ``gather_out`` returns a
-          tensor with shape 0.
 
     Note:
         - When using this interface, please ensure that the driver firmware package and CANN package are both the
           matching 8.0.RC2 version or a higher version, otherwise an error will be reported, such as BUS ERROR.
         - The shape of ``input`` is (m, k), the shape of ``x2`` is (k, n), k is required to be equal, and the value
-          range of k is [256, 65535). The shape of ``output`` is (m * world_size, n), and the shape of
-          ``gather_out`` is (m * world_size, k).
+          range of k is [256, 65535), and m is required to be an integer multiple of ``world_size`` . The shape of
+          ``output`` is (m * world_size, n).
         - The common fusion operators in a model only support the same communication group.
 
     Raises:
@@ -1381,7 +1828,7 @@ def all_gather_matmul(*args, **kwargs):
         RuntimeError: ``group`` does not exist.
         RuntimeError: ``world_size`` is inconsistent with the actual number of running cards.
         RuntimeError: ``world_size`` is not equal to ``2`` , ``4`` , or ``8`` .
-        RuntimeError: ``gather_index`` is not ``0`` .
+        RuntimeError: ``reduce_op`` is not ``'sum'`` .
         RuntimeError: ``trans_input`` is ``True`` .
 
     Supported Platforms:
@@ -1398,151 +1845,172 @@ def all_gather_matmul(*args, **kwargs):
             This example should be run with 2 devices.
 
         >>> import mindspore as ms
-        >>> import numpy as np
         >>> from mindspore import ops
+        >>> import numpy as np
         >>> ms.communication.init()
         >>> rank = ms.communication.get_rank()
         >>> np.random.seed(rank)
-        >>> input = ms.Tensor(np.random.randn(128, 256).astype(np.float32), dtype=ms.float16)
+        >>> input = ms.Tensor(np.random.randn(1024, 256).astype(np.float32), dtype=ms.float16)
         >>> x2 = ms.Tensor(np.random.randn(256, 512).astype(np.float32), dtype=ms.float16)
         >>> group = ms.communication.GlobalComm.WORLD_COMM_GROUP
         >>> world_size = ms.communication.get_group_size()
-        >>> output, gather_out = ops.all_gather_matmul(
+        >>> reduce_op = ops.ReduceOp.SUM
+        >>> output = ops.matmul_reduce_scatter(
         ...    input,
         ...    x2,
         ...    group,
         ...    world_size,
+        ...    reduce_op=reduce_op,
         ...    bias=None,
-        ...    gather_index=0,
-        ...    gather_output=True,
         ...    comm_turn=0,
         ...    trans_input=False,
         ...    trans_x2=False,
         ... )
         >>> print(output.shape)
-        (256, 512)
-        >>> print(gather_out.shape)
-        (256, 256)
+        (512, 512)
     """
-    return _all_gather_matmul_instance(*args, **kwargs)
+    return _matmul_reduce_scatter_instance(*args, **kwargs)
 
 
-def clamp(*args, **kwargs):
+def index_add(*args, **kwargs):
     r"""
-    clamp(input, min=None, max=None) -> Tensor
-
-    Clamps tensor values between the specified minimum value and maximum value.
+    index_add(input, dim, index, source, *, alpha=1) -> Tensor
 
-    Limits the value of :math:`input` to a range, whose lower limit is `min` and upper limit is `max` .
+    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:
 
     .. 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}
 
-        out_i= \left\{
-        \begin{array}{align}
-            max & \text{ if } input_i\ge max \\
-            input_i & \text{ if } min \lt input_i \lt max \\
-            min & \text{ if } input_i \le min \\
-        \end{array}\right.
-
-    Note:
-        - `min` and `max` cannot be None at the same time;
-        - When `min` is None and `max` is not None, the elements in Tensor larger than `max` will become `max`;
-        - When `min` is not None and `max` is None, the elements in Tensor smaller than `min` will become `min`;
-        - If `min` is greater than `max`, the value of all elements in Tensor will be set to `max`;
-        - The data type of `input`, `min` and `max` should support implicit type conversion and cannot be bool type.
+    .. warning::
+        This is an experimental API that is subject to change or deletion.
 
     Args:
-        input (Tensor): Input data, which type is Tensor. Tensors of arbitrary dimensions are supported.
-        min (Union(Tensor, float, int), optional): The minimum value. Default: ``None`` .
-        max (Union(Tensor, float, int), optional): The maximum value. Default: ``None`` .
+        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.
+
+    Keyword Args:
+        alpha (number, optional): The scalar multiplier for source. Default: ``1``.
 
     Returns:
-        Tensor, a clipped Tensor.
-        The data type and shape are the same as input.
+        Tensor, has the same shape and dtype as `input`.
 
     Raises:
-        ValueError: If both `min` and `max` are None.
-        TypeError: If the type of `input` is not Tensor.
-        TypeError: If the type of `min` is not in None, Tensor, float or int.
-        TypeError: If the type of `max` is not in None, Tensor, float or int.
+        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:
-        >>> # case 1: the data type of input is Tensor
-        >>> import mindspore
-        >>> from mindspore import Tensor, mint
         >>> import numpy as np
-        >>> min_value = Tensor(5, mindspore.float32)
-        >>> max_value = Tensor(20, mindspore.float32)
-        >>> input = Tensor(np.array([[1., 25., 5., 7.], [4., 11., 6., 21.]]), mindspore.float32)
-        >>> output = mint.clamp(input, min_value, max_value)
-        >>> print(output)
-        [[ 5. 20.  5.  7.]
-         [ 5. 11.  6. 20.]]
-        >>> # case 2: the data type of input is number
         >>> import mindspore
         >>> from mindspore import Tensor, mint
-        >>> import numpy as np
-        >>> min_value = 5
-        >>> max_value = 20
-        >>> input = Tensor(np.array([[1., 25., 5., 7.], [4., 11., 6., 21.]]), mindspore.float32)
-        >>> output = mint.clamp(input, min_value, max_value)
+        >>> 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 = mint.index_add(x, 1, index, y, alpha=1)
         >>> print(output)
-        [[ 5. 20.  5.  7.]
-         [ 5. 11.  6. 20.]]
+        [[ 1.5  2.   4. ]
+         [ 5.   5.   7.5]
+         [ 9.   8.  11.5]]
     """
-    return _clamp_instance(*args, **kwargs)
+    return _index_add_instance(*args, **kwargs)
 
 
-def clip(*args, **kwargs):
+def where(*args, **kwargs):
     r"""
-    clip(input, min=None, max=None) -> Tensor
+    where(condition, input, other) -> Tensor
 
-    Alias for :func:`mindspore.mint.clamp`.
-    """
-    return _clamp_instance(*args, **kwargs)
+    Selects elements from `input` or `other` based on `condition` and returns a tensor.
 
+    .. math::
+        output_i = \begin{cases} input_i,\quad &if\ condition_i \\ other_i,\quad &otherwise \end{cases}
 
-def gmm_backward_fusion(*args, **kwargs):
-    r"""
-    gmm_backward_fusion(grad, weight, *, group_list=None, group_list_type=0) -> tuple[tuple[Tensor]]
+    Args:
+        condition (Tensor[bool]): If true, yield `input`, otherwise yield `other`.
+        input (Union[Tensor, Scalar]): When `condition` is true, values to select from.
+        other (Union[Tensor, Scalar]): When `condition` is false, values to select from.
 
-    the grad of ops.function.math_func.gmm, only dx
+    Returns:
+        Tensor, elements are selected from `input` and `other`.
+
+    Raises:
+        TypeError: If `condition` is not a tensor.
+        TypeError: If both `input` and `other` are scalars.
+        ValueError: If `condition`, `input` and `other` can not broadcast to each other.
+
+    Supported Platforms:
+        ``Ascend`` ``GPU`` ``CPU``
+
+    Examples:
+        >>> import numpy as np
+        >>> from mindspore import tensor, ops
+        >>> from mindspore import dtype as mstype
+        >>> a = tensor(np.arange(4).reshape((2, 2)), mstype.float32)
+        >>> b = tensor(np.ones((2, 2)), mstype.float32)
+        >>> condition = a < 3
+        >>> output = ops.where(condition, a, b)
+        >>> print(output)
+        [[0. 1.]
+         [2. 1.]]
+    
+    .. function:: where(condition) -> Tensor
+        :noindex:
+
+    Identical to :func:`mindspore.ops.nonzero` with input `condition` and `as_tuple` being True.
+
+    Supported Platforms:
+        ``Ascend``
     """
-    return _gmm_backward_fusion_instance(*args, **kwargs)
+    return _where_instance(*args, **kwargs)
 
 __all__ = [
+    "conv3d",
+    "floor_divide",
+    "empty",
+    "addcdiv",
+    "gmm",
+    "rmod",
+    "nansum",
+    "remainder",
+    "repeat_interleave",
+    "kthvalue",
     "greater_equal",
     "ge",
-    "lerp",
-    "xlogy",
-    "where",
-    "min",
-    "kthvalue",
+    "gelu",
+    "clamp",
+    "clip",
+    "empty_like",
+    "add",
+    "__add__",
     "sub",
     "__sub__",
-    "gmm",
+    "bitwise_not",
+    "gmm_backward_fusion",
+    "xlogy",
     "div",
     "divide",
-    "matmul_reduce_scatter",
-    "add",
-    "__add__",
-    "fmod",
-    "pixel_shuffle",
-    "addcdiv",
-    "gelu",
+    "einsum",
+    "any",
     "max",
-    "floor_divide",
-    "nansum",
-    "empty",
-    "gmm_backward",
-    "remainder",
-    "repeat_interleave",
-    "bitwise_not",
+    "fmod",
+    "min",
     "all_gather_matmul",
-    "clamp",
-    "clip",
-    "gmm_backward_fusion",
+    "gmm_backward",
+    "pixel_shuffle",
+    "quant_matmul",
+    "bernoulli_",
+    "lerp",
+    "matmul_reduce_scatter",
+    "index_add",
+    "where",
 ]