mindspore 2.4.0__cp311-cp311-macosx_11_0_arm64.whl

mindspore/ops/operations/_grad_ops.py

@@ -0,0 +1,2978 @@
+# Copyright 2020-2024 Huawei Technologies Co., Ltd
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ============================================================================
+
+"""Operators for gradients."""
+# pylint: disable=unused-import
+from __future__ import absolute_import
+
+from __future__ import division
+from mindspore._checkparam import _check_3d_int_or_tuple
+from mindspore.ops.operations.nn_ops import _check_positive_int_or_tuple
+from mindspore.ops import signature as sig
+from mindspore.ops._utils import get_concat_offset
+from mindspore.ops.primitive import Primitive, PrimitiveWithInfer, prim_attr_register
+import mindspore.context as context
+from mindspore import _checkparam as validator
+from mindspore.common import dtype as mstype
+from mindspore.communication.management import GlobalComm
+from mindspore.common._utils import is_shape_unknown, is_dim_unknown
+from ..auto_generate import (AbsGrad, ACosGrad, LogitGrad, AcoshGrad, AsinGrad, AsinhGrad, ReciprocalGrad, RsqrtGrad,
+                             SqrtGrad, BatchNormGrad, BatchNormGradGrad, BiasAddGrad, GeLUGrad, FastGeLUGrad,
+                             AvgPoolGrad, MinimumGrad, LogSoftmaxGrad, PReLUGrad, ReluGrad, ReLU6Grad, EluGrad,
+                             GatherDGradV2, ResizeBilinearGrad, ResizeLinear1DGrad, ResizeNearestNeighborV2Grad,
+                             SigmoidGrad, HSwishGrad, NLLLossGrad, AtanGrad, GridSampler3DGrad, GridSampler2DGrad,
+                             ResizeBicubicGrad, HSigmoidGrad, CholeskyGrad, ResizeNearestNeighborGrad, LayerNormGrad,
+                             HShrinkGrad, LayerNormGradGrad, SiLUGrad, MaximumGrad, MaximumGradGrad, RmsNormGrad,
+                             FlashAttentionScoreGrad, UpsampleTrilinear3DGrad, UpsampleNearest3DGrad, MaskedSelectGrad,
+                             BinaryCrossEntropyGrad, SoftShrinkGrad, SeluGrad)
+
+
+class SparseFillEmptyRowsGrad(Primitive):
+    """Performs grad of SparseFillEmptyRows operation."""
+
+    @prim_attr_register
+    def __init__(self):
+        """Initialize SparseFillEmptyRowsGrad."""
+        self.init_prim_io_names(inputs=['reverse_index_map', 'grad_values'],
+                                outputs=['y_values', 'y_default_value'])
+
+
+class ScaleAndTranslateGrad(Primitive):
+    """Performs grad of ScaleAndTranslate operation."""
+
+    @prim_attr_register
+    def __init__(self, kernel_type="lanczos3", antialias=True):
+        """Initialize ScaleAndTranslateGrad"""
+        validator.check_value_type("kernel_type", kernel_type, [str], self.name)
+        validator.check_string(kernel_type, ["lanczos1", "lanczos3", "lanczos5", "gaussian", "box", "triangle",
+                                             "keyscubic", "mitchellcubic"], "kernel_type", self.name)
+        validator.check_value_type("antialias", antialias, [bool], self.name)
+
+
+class SoftmaxGrad(Primitive):
+    """Performs grad of Softmax operation."""
+
+    @prim_attr_register
+    def __init__(self):
+        """Initialize SoftmaxGrad"""
+        self.init_prim_io_names(inputs=['y', 'dy'], outputs=['z'])
+
+
+class SyncBatchNormGrad(Primitive):
+    """Performs grad of SyncBatchNorm operation."""
+
+    @prim_attr_register
+    def __init__(self, epsilon=1e-5, group="group0", device_num=2):
+        validator.check_float_range(epsilon, 0, 1, validator.INC_RIGHT, 'epsilon', self.name)
+        if not isinstance(group, str):
+            raise TypeError("The group attr of SyncBatchNormGrad must be str.")
+        validator.check_int(device_num, 2, validator.GE, "device_num", self.name)
+
+
+class KLDivLossGrad(Primitive):
+    """Computes gradients for `KLDivLoss` operation."""
+
+    @prim_attr_register
+    def __init__(self, reduction='mean'):
+        device_target = context.get_context("device_target")
+        if device_target == "CPU":
+            support_mode = ['none', 'mean', 'batchmean', 'sum']
+        elif device_target == "GPU":
+            support_mode = ['none', 'mean', 'sum']
+        elif device_target == "Ascend":
+            support_mode = ['none', 'mean', 'batchmean', 'sum']
+        else:
+            raise ValueError(f"'{self.name}' unknown device target: '{device_target}'")
+        self.reduction = validator.check_string(reduction, support_mode, 'reduction', self.name)
+
+
+class LuUnpackGrad(Primitive):
+    """Computes gradients for `LuUnpack` operation."""
+
+    @prim_attr_register
+    def __init__(self, L_grad_flag, U_grad_flag):
+        validator.check_value_type("L_grad_flag", L_grad_flag, [bool], self.name)
+        validator.check_value_type("U_grad_flag", U_grad_flag, [bool], self.name)
+        self.add_prim_attr("cust_aicpu", self.name)
+
+
+class ConcatOffset(PrimitiveWithInfer):
+    """primitive for computing Concat's gradient."""
+
+    @prim_attr_register
+    def __init__(self, N=2, axis=0):
+        """Initialize ConcatOffset"""
+
+    def __infer__(self, input_x):
+        axis = self.axis
+        x_shp = input_x['shape']
+        x_type = input_x['dtype']
+        self.add_prim_attr('T', x_type[0].element_type())
+
+        # input_x is dynamic rank
+        rank = -1
+        is_dyn_rank = False
+        for _, sh in enumerate(x_shp):
+            if is_dim_unknown(sh):
+                is_dyn_rank = True
+            else:
+                rank = len(sh)
+        if is_dyn_rank:
+            return {
+                'shape': [len(x_shp), rank],
+                'dtype': mstype.int64,
+                'value': None
+            }
+
+        # if the dimension of input_x on the axis is dynamic
+        if axis < -rank or axis >= rank:
+            raise ValueError("For 'ConcatOffset', 'axis' must be in range [{}, {}), but got {}"
+                             .format(-rank, rank, axis))
+        if axis < 0:
+            axis = axis + rank
+        for each in x_shp:
+            if each[axis] == -1:
+                return {
+                    'shape': [len(x_shp), len(x_shp[0])],
+                    'dtype': mstype.int64,
+                    'value': None
+                }
+
+        offset, _, axis = get_concat_offset(x_shp, x_type, axis, self.name)
+        offset_values = []
+        for i in range(len(x_shp)):
+            values = []
+            for j in range(len(x_shp[0])):
+                value = 0
+                if j == axis:
+                    value = offset[i]
+                values.append(value)
+            offset_values.append(tuple(values))
+        out = {'shape': None,
+               'dtype': None,
+               'value': tuple(offset_values)}
+        return out
+
+
+class Conv3DBackpropFilter(Primitive):
+    """
+    Computes the gradients of convolution 3D with respect to the filter.
+
+    Args:
+        out_channel (int): The dimension of the output.
+        kernel_size (Union[int, tuple[int]]): The kernel size of the 3D convolution.
+        mode (int): Modes for different convolutions. Not currently used.
+        pad_mode (str): Modes to fill padding. It could be "valid", "same", or "pad". Default: "valid".
+        pad (Union(int, tuple[int])): The pad value to be filled. Default: 0. If `pad` is an integer, the paddings of
+                    head, tail, top, bottom, left and right are the same, equal to pad. If `pad` is a tuple of four
+                    integers, the padding of head, tail, top, bottom, left and right equal to pad[0], pad[1], pad[2],
+                    pad[3], pad[4] and pad[5] correspondingly.
+        stride (Union(int, tuple[int])): The stride to be applied to the convolution filter. Default: 1.
+        dilation (Union(int, tuple[int])): Specifies the space to use between kernel elements. Default: 1.
+        group (int): Splits input into groups. Default: 1.
+        data_format (str): The optional value for data format. Currently only support 'NCDHW'.
+
+    Inputs:
+        - **x** (Tensor) - The input of the convolution, then the shape is :math:`(C_{out}, C_{in}, D_{in}, K_1, K_2)`.
+          Currently dout data type only support float16 and float32.
+        - **dout** (Tensor) - The gradients w.r.t the output of the convolution. The shape conforms to the default
+          data_format :math:`(N, C_{out}, D_{out}, H_{out}, W_{out})`. Currently dout data type only support float16
+          and float32.
+        - **w_size** (tuple(int)) - A tuple describes the shape of the weight which conforms to the format
+          :math:`(N, C_{in}, D_{in}, H_{in}, W_{in})`.
+
+    Outputs:
+        Tensor, the gradients w.r.t the weight of convolution 3D. It has the same shape as the weight.
+
+    Supported Platforms:
+        ``Ascend``
+
+    Examples:
+        >>> x = Tensor(np.ones([16, 32, 13, 37, 33]), mindspore.float16)
+        >>> dout = Tensor(np.ones([16, 32, 10, 32, 32]), mindspore.float16)
+        >>> w = Tensor(np.ones([32, 32, 4, 6, 2]), mindspore.float16)
+        >>> conv3d_backprop_input = P.Conv3DBackpropInput(out_channel=4, kernel_size=(4, 6, 2))
+        >>> output = conv3d_backprop_input(x, dout, F.shape(w))
+        >>> print(output.shape)
+        (32, 32, 4, 6, 2)
+    """
+
+    @prim_attr_register
+    def __init__(self,
+                 out_channel,
+                 kernel_size,
+                 mode=1,
+                 pad_mode="valid",
+                 pad=0,
+                 stride=(1, 1, 1, 1, 1),
+                 dilation=(1, 1, 1, 1, 1),
+                 group=1,
+                 data_format="NCDHW"):
+        """Initialize Convolution"""
+        self.init_prim_io_names(inputs=['x', 'out_backprop', 'filter_size'], outputs=['y'])
+        self.out_channel = validator.check_positive_int(out_channel, 'out_channel', self.name)
+        self.kernel_size = _check_3d_int_or_tuple('kernel_size', kernel_size, self.name)
+        self.stride = _check_3d_int_or_tuple('stride', stride, self.name, allow_five=True, ret_five=True)
+        self.add_prim_attr('strides', self.stride)
+        self.dilation = _check_3d_int_or_tuple('dilation', dilation, self.name, allow_five=True, ret_five=True)
+        self.add_prim_attr('dilations', self.dilation)
+        validator.check_value_type('pad', pad, (int, tuple), self.name)
+        if isinstance(pad, int):
+            pad = (pad,) * 6
+        validator.check_equal_int(len(pad), 6, 'pad size', self.name)
+        self.add_prim_attr('pad', pad)
+        self.pad_list = pad
+        self.add_prim_attr('pad_list', self.pad_list)
+
+        validator.check_value_type('pad_mode', pad_mode, [str], self.name)
+        self.pad_mode = validator.check_string(pad_mode.lower(), ['valid', 'same', 'pad'], 'pad_mode', self.name)
+        if self.pad_mode != 'pad' and self.pad_list != (0, 0, 0, 0, 0, 0):
+            raise ValueError(f"For '{self.name}', when pad is not 0, pad_mode must be set as 'pad'.")
+        if self.pad_mode == 'pad':
+            for item in pad:
+                validator.check_non_negative_int(item, 'pad item', self.name)
+        self.add_prim_attr('pad_mode', self.pad_mode)
+
+        self.mode = validator.check_equal_int(mode, 1, 'mode', self.name)
+        self.add_prim_attr('mode', self.mode)
+        self.group = validator.check_positive_int(group, 'group', self.name)
+        self.add_prim_attr('groups', self.group)
+        self.format = validator.check_string(data_format, ['NCDHW'], 'format', self.name)
+        self.add_prim_attr('data_format', self.format)
+
+
+class Conv2DBackpropFilter(Primitive):
+    """
+    Computes the gradients of convolution with respect to the filter.
+
+    Args:
+        out_channel (int): The dimensionality of the output space.
+        kernel_size (Union[int, tuple[int]]): The size of the convolution window.
+        pad_mode (str): Modes to fill padding. It could be "valid", "same", or "pad". Default: "valid".
+        pad (Union(int, tuple[int])): The pad value to be filled. Default: 0. If `pad` is an integer, the paddings of
+                    top, bottom, left and right are the same, equal to pad. If `pad` is a tuple of four integers, the
+                    padding of top, bottom, left and right equal to pad[0], pad[1], pad[2], and pad[3] correspondingly.
+        pad_list (tuple): The pad list like (top, bottom, left, right). Default: (0, 0, 0, 0).
+        mode (int): Modes for different convolutions. 0 Math convolution, 1 cross-correlation convolution ,
+                    2 deconvolution, 3 depthwise convolution. Default: 1.
+        stride (tuple): The stride to be applied to the convolution filter. Default: (1, 1).
+        dilation (tuple): Specifies the dilation rate to be used for the dilated convolution. Default: (1, 1, 1, 1).
+        group (int): Splits input into groups. Default: 1.
+        data_format (str) - The format of input and output data. It should be 'NHWC' or 'NCHW', \
+            default is 'NCHW'.
+
+    Returns:
+        Tensor, the gradients of convolution.
+    """
+
+    @prim_attr_register
+    def __init__(self,
+                 out_channel,
+                 kernel_size,
+                 pad_mode="valid",
+                 pad=0,
+                 pad_list=(0, 0, 0, 0),
+                 mode=1,
+                 stride=(1, 1),
+                 dilation=(1, 1, 1, 1),
+                 group=1,
+                 data_format="NCHW"):
+        """Initialize Convolution"""
+        self.init_prim_io_names(inputs=['out_backprop', 'input', 'filter_sizes'], outputs=['output'])
+        self.out_channel = out_channel
+        self.kernel_size = kernel_size
+        self.mode = mode
+        pad_mode = pad_mode.upper()
+        self.add_prim_attr('pad_mode', pad_mode)
+        if isinstance(pad, int):
+            pad = (pad,) * 4
+        else:
+            validator.check_equal_int(len(pad), 4, 'pad size', self.name)
+        self.add_prim_attr("pad", pad)
+        self.format = validator.check_string(data_format, ['NCHW', 'NHWC'], 'format', self.name)
+        if context.get_context("device_target") != "GPU" and self.format == "NHWC":
+            raise ValueError("NHWC format only support in GPU target.")
+        self.add_prim_attr('data_format', self.format)
+        self.stride = _check_positive_int_or_tuple('stride', stride, self.name, allow_four=True, ret_four=True)
+        self.add_prim_attr('stride', self.stride)
+        self.dilation = _check_positive_int_or_tuple('dilation', dilation, self.name, allow_four=True, ret_four=True)
+        self.add_prim_attr('dilation', self.dilation)
+        self.group = group
+        self.add_prim_attr('groups', group)
+        if pad_list:
+            for x in pad_list:
+                if x != -1:
+                    validator.check_non_negative_int(x, 'element of pad_list', self.name)
+        self.pad_list = pad_list
+
+
+class DepthwiseConv2dNativeBackpropFilter(PrimitiveWithInfer):
+    """
+    Returns the gradient of filter for DepthwiseConv2dNative.
+
+    Applies depthwise conv2d for the input, which will generate more channels with channel_multiplier.
+
+    Refer to class DepthwiseConv2dNative for more details.
+
+    Args:
+        channel_multiplier (int): The multiplier for the original output conv.
+        kernel_size (int or tuple): The size of the conv kernel.
+        mode (int): Modes for different convolutions. 0 Math convolution, 1 cross-correlation convolution,
+                       2 deconvolution,3 depthwise convolution. Default: 3.
+        pad_mode (str): The mode to fill padding which can be: "valid", "same" or "pad". Default: "valid".
+        pad (Union(int, tuple[int])): The pad value to be filled. Default: 0. If `pad` is an integer, the paddings of
+                    top, bottom, left and right are the same, equal to pad. If `pad` is a tuple of four integers, the
+                    padding of top, bottom, left and right equal to pad[0], pad[1], pad[2], and pad[3] correspondingly.
+        pad_list (tuple): The pad list like (top, bottom, left, right). Default: (0, 0, 0, 0).
+        stride (int): The stride to be applied to the convolution filter. Default: 1.
+        dilation (int): Specifies the space to use between kernel elements. Default: 1.
+        group (int): Splits input into groups. Default: 1.
+
+    Returns:
+        Tensor, the value is the gradient of filter for DepthwiseConv2dNative.
+    """
+
+    @prim_attr_register
+    def __init__(self,
+                 channel_multiplier,
+                 kernel_size,
+                 pad_mode="valid",
+                 pad=0,
+                 pad_list=(0, 0, 0, 0),
+                 mode=3,
+                 stride=1,
+                 dilation=1,
+                 group=1):
+        """Initialize Convolution"""
+        self.init_prim_io_names(inputs=['input', 'filter_size', 'dout'], outputs=['output'])
+        self.channel_multiplier = channel_multiplier
+        self.kernel_size = kernel_size
+        self.mode = mode
+        self.pad_mode = pad_mode
+        if isinstance(pad, int):
+            pad = (pad,) * 4
+        else:
+            validator.check_equal_int(len(pad), 4, 'pad size', self.name)
+        self.add_prim_attr("pad", pad)
+        self.pad_list = pad_list
+        self.stride = stride
+        self.dilation = dilation
+        self.group = group
+        self.add_prim_attr('data_format', "NCHW")
+
+    def __infer__(self, x, w_size, dout):
+        w_size_v = w_size['value']
+        args = {'x': x['dtype'], 'dout': dout['dtype']}
+        validator.check_tensors_dtypes_same_and_valid(args, mstype.number_type, self.name)
+        out = {
+            'value': None,
+            'shape': w_size_v,
+            'dtype': dout['dtype'],
+        }
+        return out
+
+
+class DepthwiseConv2dNativeBackpropInput(PrimitiveWithInfer):
+    """
+    Returns the gradient of input for DepthwiseConv2dNative.
+
+    Applies depthwise conv2d for the input, which will generate more channels with channel_multiplier.
+
+    Args:
+        channel_multiplier (int): The multiplier for the original output conv.
+        kernel_size (int or tuple): The size of the conv kernel.
+        mode (int): Modes for different convolutions. 0 Math convolution, 1 cross-correlation convolution ,
+                    2 deconvolution,3 depthwise convolution. Default: 3.
+        pad_mode (str):  Modes to fill padding. It could be "valid", "same", or "pad". Default: "valid".
+        pad (Union(int, tuple[int])): The pad value to be filled. Default: 0. If `pad` is an integer, the paddings of
+                    top, bottom, left and right are the same, equal to pad. If `pad` is a tuple of four integers, the
+                    padding of top, bottom, left and right equal to pad[0], pad[1], pad[2], and pad[3] correspondingly.
+        pad_list (tuple): The pad list like (top, bottom, left, right). Default: (0, 0, 0, 0).
+        stride (int): The stride to be applied to the convolution filter. Default: 1.
+        dilation (int): Specifies the space to use between kernel elements. Default: 1.
+        group (int): Splits input into groups. Default: 1.
+
+    Returns:
+        Tensor, the value is the gradient of input for DepthwiseConv2dNative.
+    """
+
+    @prim_attr_register
+    def __init__(self,
+                 channel_multiplier,
+                 kernel_size,
+                 pad_mode="valid",
+                 pad=0,
+                 pad_list=(0, 0, 0, 0),
+                 mode=3,
+                 stride=1,
+                 dilation=1,
+                 group=1):
+        """Initialize Convolution"""
+        self.init_prim_io_names(inputs=['input_size', 'filter', 'dout'], outputs=['output'])
+        self.channel_multiplier = channel_multiplier
+        self.kernel_size = kernel_size
+        self.mode = mode
+        self.pad_mode = pad_mode
+        if isinstance(pad, int):
+            pad = (pad,) * 4
+        else:
+            validator.check_equal_int(len(pad), 4, 'pad size', self.name)
+        self.add_prim_attr("pad", pad)
+        self.pad_list = pad_list
+        self.stride = stride
+        self.dilation = dilation
+        self.group = group
+        self.add_prim_attr('data_format', "NCHW")
+
+    def __infer__(self, x_size, w, dout):
+        args = {'w': w['dtype'], 'dout': dout['dtype']}
+        validator.check_tensors_dtypes_same_and_valid(args, mstype.number_type, self.name)
+        x_size_v = x_size['value']
+        out = {
+            'value': None,
+            'shape': x_size_v,
+            'dtype': dout['dtype'],
+        }
+        return out
+
+
+class DropoutGrad(Primitive):
+    """
+    The gradient of Dropout. During training, randomly zeroes some of the elements
+    of the input tensor with probability.
+
+    Args:
+        keep_prob (float): The keep rate, between 0 and 1, e.g. keep_prob = 0.9,
+          means dropping out 10% of input units. Default: 0.5.
+
+    Inputs:
+        - **shape** (tuple[int]) - The shape of target mask.
+
+    Outputs:
+        Tensor, the value of generated mask for input shape.
+
+    Examples:
+        >>> dropout_grad = ops.DropoutGrad(keep_prob=0.5)
+        >>> in = Tensor((20, 16, 50, 50))
+        >>> out = dropout_grad(in)
+    """
+
+    @prim_attr_register
+    def __init__(self, keep_prob=0.5):
+        self.keep_prob = validator.check_float_range(keep_prob, 0, 1, validator.INC_RIGHT, "keep_prob", self.name)
+
+
+class FlattenGrad(PrimitiveWithInfer):
+    """Performs gradients of Flatten."""
+
+    @prim_attr_register
+    def __init__(self):
+        self.init_prim_io_names(inputs=['x', 'shape'], outputs=['output'])
+
+    def __infer__(self, *args):
+        out = {
+            'value': None,
+            'shape': args[1]['value'],
+            'dtype': args[0]['dtype'],
+        }
+        return out
+
+
+class InstanceNormGrad(PrimitiveWithInfer):
+    """Gradients of InstanceNorm operation."""
+
+    @prim_attr_register
+    def __init__(self, epsilon=0.0, momentum=0.1):
+        self.init_prim_io_names(inputs=['dy', 'x', 'gamma', 'save_mean', 'save_variance'],
+                                outputs=['dx', 'bn_gamma', 'bn_beta'])
+
+
+class InstanceNormV2Grad(Primitive):
+    """Gradients of InstanceNormV2 operation."""
+
+    @prim_attr_register
+    def __init__(self, is_training=True, epsilon=1e-5):
+        self.init_prim_io_names(inputs=['dy', 'x', 'gamma', 'mean', 'variance', 'save_mean', 'save_variance'],
+                                outputs=['pd_x', 'pd_gamma', 'pd_beta'])
+        validator.check_is_float(epsilon, 'epsilon', self.name)
+        validator.check_float_range(epsilon, 0, 1, validator.INC_RIGHT, 'epsilon', self.name)
+        validator.check_bool(is_training, "is_training", self.name)
+
+
+class EinsumGrad(PrimitiveWithInfer):
+    """Gradients of Einsum."""
+
+    @prim_attr_register
+    def __init__(self, equation):
+        pass
+
+    def infer_shape(self, x_shapes, dout_shape):
+        out_shape = ()
+        for dim in x_shapes:
+            out_shape += (dim,)
+        return out_shape
+
+    def infer_dtype(self, x_types, dout_shape):
+        out_type = ()
+        for cur_type in x_types:
+            out_type += (cur_type,)
+        return out_type
+
+
+class UniqueGrad(Primitive):
+    """Gradients of Unique operation."""
+
+    @prim_attr_register
+    def __init__(self):
+        self.init_prim_io_names(inputs=['dy', 'y'], outputs=['dx'])
+
+
+class BNTrainingReduceGrad(Primitive):
+    """Gradients of FusedBatchNorm operation."""
+
+    @prim_attr_register
+    def __init__(self, epsilon=0.0001, data_format='NCHW'):
+        self.data_format = validator.check_string(data_format, ['NCHW', 'NHWC'], 'format', self.name)
+        _inputs = ['grads', 'x', 'diff_scale', 'diff_offset', 'scale', 'batch_mean', 'batch_variance']
+        self.init_prim_io_names(inputs=_inputs, outputs=['y'])
+
+
+class BNTrainingUpdateGrad(Primitive):
+    """Gradients of FusedBatchNorm operation."""
+
+    @prim_attr_register
+    def __init__(self, epsilon=0.0001, data_format='NCHW'):
+        self.data_format = validator.check_string(data_format, ['NCHW', 'NHWC'], 'format', self.name)
+        self.init_prim_io_names(inputs=['grads', 'x', 'batch_mean', 'batch_variance'],
+                                outputs=['diff_scale', 'diff_offset'])
+
+
+class NeighborExchangeV2Grad(PrimitiveWithInfer):
+    """"Gradients of NeighborExchangeV2 operation."""
+
+    @prim_attr_register
+    def __init__(self, send_rank_ids, send_lens, recv_rank_ids, recv_lens, data_format,
+                 group=GlobalComm.WORLD_COMM_GROUP):
+        self.init_prim_io_names(inputs=['dy'], outputs=['dx'])
+        self.send_rank_ids = send_rank_ids
+        self.recv_rank_ids = recv_rank_ids
+        self.send_lens = send_lens
+        self.recv_lens = recv_lens
+        self.format = validator.check_string(data_format, ['NCHW'], 'format', self.name)
+        self.add_prim_attr('no_elimilate', True)
+
+    def __infer__(self, dy):
+        dy_shape = dy['shape']
+        validator.check(f'dy_shape.size()', len(dy_shape), f'4', 4, validator.EQ, self.name)
+        if self.send_rank_ids[5] != -1 or self.send_rank_ids[6] != -1 or self.send_rank_ids[7] != -1:
+            dy_shape[3] -= self.send_lens[2]
+
+        if self.send_rank_ids[1] != -1 or self.send_rank_ids[2] != -1 or self.send_rank_ids[3] != -1:
+            dy_shape[3] -= self.send_lens[3]
+
+        if self.send_rank_ids[0] != -1 or self.send_rank_ids[1] != -1 or self.send_rank_ids[7] != -1:
+            dy_shape[2] -= self.send_lens[0]
+
+        if self.send_rank_ids[3] != -1 or self.send_rank_ids[4] != -1 or self.send_rank_ids[5] != -1:
+            dy_shape[2] -= self.send_lens[1]
+
+        return {'shape': dy_shape,
+                'dtype': dy['dtype'],
+                'value': None}
+
+
+class _PoolGrad(PrimitiveWithInfer):
+    """Gradients of the max/avg pool operation."""
+
+    @prim_attr_register
+    def __init__(self, kernel_size, strides, pad_mode="VALID", data_format="NCHW"):
+        self.init_prim_io_names(inputs=['x_origin', 'out_origin', 'grad'], outputs=['output'])
+
+        validator.check_value_type('kernel_size', kernel_size, [int, tuple], self.name)
+        validator.check_value_type('strides', strides, [int, tuple], self.name)
+        validator.check_value_type('pad_mode', pad_mode, [str], self.name)
+        self.pad_mode = validator.check_string(pad_mode.upper(), ['VALID', 'SAME'], 'pad_mode', self.name)
+        self.add_prim_attr("pad_mode", self.pad_mode)
+        self.format = validator.check_string(data_format, ['NCHW', 'NHWC'], 'format', self.name)
+        if context.get_context("device_target") != "GPU" and self.format == "NHWC":
+            raise ValueError("NHWC format only support in GPU target.")
+        self.is_maxpoolgradwithargmax = (self.name == "MaxPoolGradWithArgmax")
+        if not self.is_maxpoolgradwithargmax:
+            self.add_prim_attr('data_format', self.format)
+
+        def _grad_check_int_or_tuple(arg_name, arg_val, is_argmax):
+            validator.check_value_type(arg_name, arg_val, (int, tuple), self.name)
+            error_msg = ValueError(f"For '{self.name}' the '{arg_name}' must be an positive int number "
+                                   f"or a tuple of two or four positive int numbers, but got {arg_val}")
+            if isinstance(arg_val, int):
+                ret = (1, arg_val, arg_val, 1) if is_argmax else (1, 1, arg_val, arg_val)
+            elif len(arg_val) == 2:
+                ret = (1, arg_val[0], arg_val[1], 1) if is_argmax else (1, 1, arg_val[0], arg_val[1])
+            elif len(arg_val) == 4:
+                ret = arg_val
+            else:
+                raise error_msg
+            # whether all elements of tuple are positive integers
+            for item in ret:
+                if not isinstance(item, int) or item <= 0:
+                    raise error_msg
+            return ret
+
+        kernel_size = _grad_check_int_or_tuple("kernel_size", kernel_size, self.is_maxpoolgradwithargmax)
+        strides = _grad_check_int_or_tuple("strides", strides, self.is_maxpoolgradwithargmax)
+        if self.format == "NCHW":
+            self.kernel_size = kernel_size
+            self.strides = strides
+        else:
+            self.kernel_size = [kernel_size[0], kernel_size[2], kernel_size[3], kernel_size[1]]
+            self.strides = [strides[0], strides[2], strides[3], strides[1]]
+        self.add_prim_attr("kernel_size", self.kernel_size)
+        self.add_prim_attr("strides", self.strides)
+
+
+class AvgPoolGradVm(_PoolGrad):
+    """Gradients of the avg pool operation for vm."""
+
+    @prim_attr_register
+    def __init__(self, kernel_size=1, strides=1, pad_mode="VALID"):
+        super(AvgPoolGradVm, self).__init__(kernel_size, strides, pad_mode)
+        self.init_prim_io_names(inputs=['x_origin', 'grad', 'mean_matrix', 'kernel_matrix'], outputs=['output'])
+
+    def __infer__(self, origin_input, dout, mean_matrix, kernel_matrix):
+        out = {
+            'value': None,
+            'shape': tuple(origin_input['value']),
+            'dtype': dout['dtype'],
+        }
+
+        return out
+
+
+class AvgPoolGradGe(_PoolGrad):
+    """Gradients of the avg pool operation for ge."""
+
+    @prim_attr_register
+    def __init__(self, kernel_size=1, strides=1, pad_mode="VALID", data_format="NCHW"):
+        super(AvgPoolGradGe, self).__init__(kernel_size, strides, pad_mode, data_format)
+
+    def __infer__(self, origin_input, dout):
+        out = {
+            'value': None,
+            'shape': tuple(origin_input['value']),
+            'dtype': dout['dtype'],
+        }
+
+        return out
+
+
+class AvgPoolGradV1(Primitive):
+    """Gradients of the AvgPoolV1 operation."""
+
+    @prim_attr_register
+    def __init__(self, kernel_size=1, strides=1, pad_mode="VALID", data_format="NCHW"):
+        validator.check_value_type('kernel_size', kernel_size, [int, tuple], self.name)
+        validator.check_value_type('strides', strides, [int, tuple], self.name)
+        self.pad_mode = validator.check_string(
+            pad_mode.upper(), ['VALID', 'SAME'], 'pad_mode', self.name)
+        self.add_prim_attr("pad_mode", self.pad_mode)
+        self.format = validator.check_string(
+            data_format, ['NCHW', 'NHWC'], 'format', self.name)
+        self.add_prim_attr('data_format', self.format)
+
+        def _avgpoolgrad_check_int_or_tuple(argname, argval):
+            validator.check_value_type(argname, argval, (int, tuple), self.name)
+            errormsg = ValueError(f"For '{self.name}' the '{argname}' should be an positive int number "
+                                  f"or a tuple of two or four positive int numbers, but got {argval}")
+            if isinstance(argval, int):
+                ret = (1, 1, argval, argval)
+            elif len(argval) == 2:
+                ret = (1, 1, argval[0], argval[1])
+            elif len(argval) == 4:
+                ret = argval
+            else:
+                raise errormsg
+            # whether all elements of tuple are positive integers?
+            for it in ret:
+                if not isinstance(it, int) or it <= 0:
+                    raise errormsg
+            return ret
+
+        self.kernel_size = _avgpoolgrad_check_int_or_tuple(
+            "kernel_size", kernel_size)
+        self.strides = _avgpoolgrad_check_int_or_tuple("strides", strides)
+
+        self.kernel_size_adapt = self.kernel_size if self.format == "NCHW" else (
+            self.kernel_size[0], self.kernel_size[2], self.kernel_size[3], self.kernel_size[1])
+        self.strides_adapt = self.strides if self.format == "NCHW" else (
+            self.strides[0], self.strides[2], self.strides[3], self.strides[1])
+
+        # If length of some attrs is 4 we regard it as legal, either by using the op directly,
+        # or passed from an instance of forward op AvgPoolV1.
+        if len(self.kernel_size) == 4:
+            self.kernel_size_adapt = self.kernel_size
+        if len(self.strides) == 4:
+            self.strides_adapt = self.strides
+
+        self.add_prim_attr("kernel_size", self.kernel_size_adapt)
+        self.add_prim_attr("strides", self.strides_adapt)
+
+
+class AdaptiveAvgPool2DGrad(Primitive):
+    """Gradients of the adaptive avg pool 2D operation."""
+
+    @prim_attr_register
+    def __init__(self):
+        """Initialize AdaptiveAvgPool2DGrad"""
+        self.init_prim_io_names(inputs=['input_grad', 'orig_input_shape'], outputs=['output_grad'])
+
+
+class AdaptiveAvgPool3DGrad(Primitive):
+    """Performs grad of AdaptiveAvgPool3D operation."""
+    @prim_attr_register
+    def __init__(self):
+        self.init_prim_io_names(inputs=['y_grad', 'orig_input_shape'], outputs=['x_grad'])
+
+
+class AvgPool3DGrad(Primitive):
+    """Gradients of the avg pool3d operation."""
+
+    @prim_attr_register
+    def __init__(self, kernel_size=1, strides=1, pads=0, ceil_mode=False,
+                 count_include_pad=True, divisor_override=0, data_format="NCDHW", pad_mode="pad"):
+        self.init_prim_io_names(inputs=['origin_input_shape', 'grads'], outputs=['output'])
+        self.kernel_size = _check_3d_int_or_tuple('kernel_size', kernel_size, self.name, allow_five=True, ret_five=True)
+        self.add_prim_attr('kernel_size', self.kernel_size)
+        self.strides = _check_3d_int_or_tuple('strides', strides, self.name, allow_five=True, ret_five=True)
+        self.add_prim_attr('strides', self.strides)
+        validator.check_value_type('pad_mode', pad_mode, [str], self.name)
+        self.pad_mode = validator.check_string(pad_mode.upper(), ['VALID', 'SAME', 'PAD'], 'pad_mode', self.name)
+        validator.check_value_type('pads', pads, (int, tuple), self.name)
+        if isinstance(pads, int):
+            pads = (pads,) * 6
+        validator.check_equal_int(len(pads), 6, 'pad size', self.name)
+        for item in pads:
+            validator.check_non_negative_int(item, 'pad item', self.name)
+        self.add_prim_attr('pad_list', pads)
+        self.ceil_mode = validator.check_value_type('ceil_mode', ceil_mode, bool, self.name)
+        self.count_include_pad = validator.check_value_type('count_include_pad', count_include_pad, bool, self.name)
+        self.divisor_override = validator.check_value_type('divisor_override', divisor_override, int, self.name)
+        self.format = validator.check_string(data_format, ['NCDHW'], 'format', self.name)
+
+
+class AdaptiveMaxPool2DGrad(Primitive):
+    """Gradients of the adaptive max pool 2D operation."""
+    @prim_attr_register
+    def __init__(self):
+        """Initialize AdaptiveMaxPool2DGrad"""
+        self.init_prim_io_names(inputs=['y_grad', 'x', 'argmax'], outputs=['x_grad'])
+
+
+class MaxPoolGrad(_PoolGrad):
+    """Performs gradients of the max pool operation."""
+
+    @prim_attr_register
+    def __init__(self, kernel_size=1, strides=1, pad_mode="VALID", data_format="NCHW"):
+        super(MaxPoolGrad, self).__init__(kernel_size, strides, pad_mode, data_format)
+
+    def infer_shape(self, x1_shape, x2_shape, grad_shape):
+        return x1_shape
+
+    def infer_dtype(self, x1_dtype, x2_dtype, grad_dtype):
+        return x1_dtype
+
+
+class MaxPoolGradV1(Primitive):
+    """Performs gradients of the MaxPoolV1 operation."""
+
+    @prim_attr_register
+    def __init__(self, kernel_size=1, strides=1, pad_mode="VALID", data_format="NCHW"):
+        self.init_prim_io_names(
+            inputs=['x_origin', 'out_origin', 'grad'], outputs=['output'])
+
+        validator.check_value_type('kernel_size', kernel_size, [int, tuple], self.name)
+        validator.check_value_type('strides', strides, [int, tuple], self.name)
+        self.pad_mode = validator.check_string(
+            pad_mode.upper(), ['VALID', 'SAME'], 'pad_mode', self.name)
+        self.add_prim_attr("pad_mode", self.pad_mode)
+        self.format = validator.check_string(
+            data_format, ['NCHW', 'NHWC'], 'format', self.name)
+        self.add_prim_attr('data_format', self.format)
+
+        def _grad_check_int_or_tuple(arg_name, arg_val):
+            validator.check_value_type(
+                arg_name, arg_val, (int, tuple), self.name)
+            error_msg = ValueError(f"For '{self.name}' the '{arg_name}' should be an positive int number "
+                                   f"or a tuple of two or four positive int numbers, but got {arg_val}")
+            if isinstance(arg_val, int):
+                ret = (1, 1, arg_val, arg_val)
+            elif len(arg_val) == 2:
+                ret = (1, 1, arg_val[0], arg_val[1])
+            elif len(arg_val) == 4:
+                ret = arg_val
+            else:
+                raise error_msg
+            # whether all elements of tuple are positive integers
+            for item in ret:
+                if not isinstance(item, int) or item <= 0:
+                    raise error_msg
+            return ret
+
+        self.kernel_size = _grad_check_int_or_tuple("kernel_size", kernel_size)
+        self.strides = _grad_check_int_or_tuple("strides", strides)
+
+        kernel_size_adapted = self.kernel_size if self.format == 'NCHW' else (
+            self.kernel_size[0], self.kernel_size[2], self.kernel_size[3], self.kernel_size[1])
+        strides_adapted = self.strides if self.format == 'NCHW' else (
+            self.strides[0], self.strides[2], self.strides[3], self.strides[1])
+
+        if len(kernel_size) == 4:
+            kernel_size_adapted = kernel_size
+        if len(strides) == 4:
+            strides_adapted = strides
+
+        self.add_prim_attr("kernel_size", kernel_size_adapted)
+        self.add_prim_attr("strides", strides_adapted)
+
+
+class MaxPoolGradGrad(_PoolGrad):
+    r"""
+    Performs gradients of the MaxPoolGrad operation.
+
+    Args:
+        kernel_size (Union[int, tuple[int]]): The size of kernel used to take the maximum value,
+            is an int number that represents height and width are both kernel_size, or a tuple
+            of two int numbers that represent height and width respectively. Default: 1.
+        strides (Union[int, tuple[int]]): The distance of kernel moving, an int number that represents
+            the height and width of movement are both strides, or a tuple of two int numbers that
+            represent height and width of movement respectively. Default: 1.
+        pad_mode (str): The optional value for pad mode, is "same" or "valid", not case sensitive.
+            Default: "valid".
+
+            - same: Adopts the way of completion. The height and width of the output will be the same as
+              the input. The total number of padding will be calculated in horizontal and vertical
+              directions and evenly distributed to top and bottom, left and right if possible.
+              Otherwise, the last extra padding will be done from the bottom and the right side.
+
+            - valid: Adopts the way of discarding. The possible largest height and width of output
+              will be returned without padding. Extra pixels will be discarded.
+
+    Inputs:
+        - **origin_input** (Tensor) - Tensor with data format "NCHW".
+          For Ascend, data type must be float16. For CPU and GPU, data type support float16 and float32.
+        - **origin_output** (Tensor) - Data type same as `origin_input`.
+        - **grad** (Tensor) - Data type and shape same as `origin_input`.
+
+    Outputs:
+        Tensor, with data type same as `origin_input`. Shape same as `origin_output`.
+
+    Raises:
+        TypeError: If kernel_size is neither int nor a tuple of 2/4 int numbers.
+        TypeError: If strides is neither int nor a tuple of 2/4 int numbers.
+        TypeError: If pad_mode is not string.
+        ValueError: If pad_mode is neither "same" nor "valid"(not case sensitive).
+        TypeError: For Ascend, input data type is not float16. For CPU or GPU, input data type is neither
+        float16 nor float32.
+        ValueError: If the rank of `origin_input`, `origin_output` or `grad` is not equal to 4.
+        ValueError: If data types of all inputs are not equal.
+        ValueError: If the shapes of `origin_input` and `grad` are not equal.
+
+    Supported Platforms:
+        ``Ascend`` ``GPU`` ``CPU``
+    """
+
+    @prim_attr_register
+    def __init__(self, kernel_size=1, strides=1, pad_mode="VALID"):
+        super(MaxPoolGradGrad, self).__init__(kernel_size, strides, pad_mode)
+
+    def infer_shape(self, x1_shape, x2_shape, grad_shape):
+        return x2_shape
+
+    def infer_dtype(self, x1_dtype, x2_dtype, grad_dtype):
+        args = {'x1_dtype': x1_dtype, 'x2_dtype': x2_dtype, 'grad_dtype': grad_dtype}
+        validator.check_tensors_dtypes_same_and_valid(args, [mstype.float16], self.name)
+        return x2_dtype
+
+
+def _get_max_pool3d_grad_pads_by_pad_mode(input_shape, kernel_size, strides, pad_mode):
+    """
+    helper for get max pool3d grad pads by pad_mode
+    """
+
+    def get_pad(origin_shape, ksize, stride):
+        tail = origin_shape % stride
+        pad = (ksize - tail) if tail > 0 else (ksize - stride)
+        pad = max(pad, 0)
+        pad1 = int(pad / 2)
+        pad2 = int(pad / 2) + pad % 2
+        return pad1, pad2
+
+    _, _, d, h, w = input_shape
+    _, _, kd, kh, kw = kernel_size
+    _, _, strd, strh, strw = strides
+
+    pads = (0, 0, 0, 0, 0, 0)
+    if pad_mode == 'SAME':
+        pads_d = get_pad(d, kd, strd)
+        pads_h = get_pad(h, kh, strh)
+        pads_w = get_pad(w, kw, strw)
+        pads = pads_d + pads_h + pads_w
+    return pads
+
+
+class MaxPool3DGrad(Primitive):
+    """Gradients of the max pool3d operation."""
+
+    @prim_attr_register
+    def __init__(self, kernel_size=(1, 1, 1, 1, 1), strides=(1, 1, 1, 1, 1),
+                 pad_mode='VALID', pad_list=0, data_format="NCDHW"):
+        self.init_prim_io_names(inputs=['x_origin', 'out_origin', 'grad'], outputs=['output'])
+        validator.check_value_type('kernel_size', kernel_size, [int, tuple], self.name)
+        validator.check_value_type('strides', strides, [int, tuple], self.name)
+        validator.check_value_type('pad_mode', pad_mode, [str], self.name)
+        self.format = validator.check_string(data_format, ['NCDHW'], 'format', self.name)
+        if pad_mode.upper() == 'PAD':
+            pad_mode = 'CALCULATED'
+        self.pad_mode = validator.check_string(pad_mode.upper(), ['VALID', 'SAME', 'CALCULATED'], 'pad_mode', self.name)
+        self.kernel_size = _check_3d_int_or_tuple("kernel_size", kernel_size, self.name,
+                                                  allow_five=True, ret_five=True)
+        self.add_prim_attr("kernel_size", self.kernel_size)
+        self.strides = _check_3d_int_or_tuple("strides", strides, self.name, allow_five=True, ret_five=True)
+        self.add_prim_attr("strides", self.strides)
+        validator.check_value_type('pad_list', pad_list, (int, tuple), self.name)
+        self.pad_list = pad_list
+        if isinstance(self.pad_list, int):
+            self.pad_list = (self.pad_list,) * 6
+        if len(self.pad_list) == 3:
+            self.pad_list = (pad_list[0], pad_list[0], pad_list[1], pad_list[1], pad_list[2], pad_list[3])
+        if len(self.pad_list) != 3 and len(self.pad_list) != 6:
+            raise ValueError(f"For `maxpool3d` attr 'pad_list' must be an positive int number or a tuple of "
+                             f"three or six positive int numbers, but got `{len(self.pad_list)}` numbers.")
+        if self.pad_mode != 'CALCULATED' and self.pad_list != (0, 0, 0, 0, 0, 0):
+            raise ValueError(f"For '{self.name}', when pad_list is not 0, pad_mode must be set as 'pad'.")
+        if self.pad_mode == 'CALCULATED':
+            for item in self.pad_list:
+                validator.check_non_negative_int(item, 'pad_list item', self.name)
+        self.add_prim_attr("pad_list", self.pad_list)
+
+
+class MaxPool3DGradGrad(PrimitiveWithInfer):
+    r"""Gradients of the max pool3d grad operation.
+
+    Args:
+        kernel_size (Union[int, tuple[int]]): The size of kernel used to take the maximum value,
+            is an int number that represents depth, height and width are both kernel_size, or a tuple
+            of two int numbers that represent depth, height and width respectively. Default: 1.
+        strides (Union[int, tuple[int]]): The distance of kernel moving, an int number that represents
+            the depth, height and width of movement are both strides, or a tuple of two int numbers that
+            represent depth, height and width of movement respectively. Default: 1.
+        pad_mode (str): The optional value for pad mode, is "same" or "valid", not case sensitive.
+            Default: "valid".
+
+            - same: Adopts the way of completion. The depth, height and width of the output will be the
+              same as the input. The total number of padding will be calculated in depth, horizontal and
+              vertical directions and evenly distributed to front and back, top and bottom, left and
+              right if possible. Otherwise, the last extra padding will be done from the back, the bottom
+              and the right side.
+
+            - valid: Adopts the way of discarding. The possible largest height and width of output
+              will be returned without padding. Extra pixels will be discarded.
+
+    Inputs:
+        - **origin_input** (Tensor) - Tensor with data format "NCDHW".
+          For Ascend, data type must be float16. For CPU and GPU, data type support float16 and float32.
+        - **origin_output** (Tensor) - Data type same as `origin_input`.
+        - **grad** (Tensor) - Data type and shape same as `origin_input`.
+
+    Outputs:
+        Tensor, with data type same as `origin_input`. Shape same as `origin_output`.
+
+    Raises:
+        TypeError: If kernel_size is neither int nor a tuple of 3/5 int numbers.
+        TypeError: If strides is neither int nor a tuple of 3/5 int numbers.
+        TypeError: If pad_mode is not string.
+        ValueError: If pad_mode is neither "same" nor "valid"(not case sensitive).
+        TypeError: For Ascend, input data type is not float16. For CPU or GPU, input data type is neither
+        float16 nor float32.
+        ValueError: If the rank of `origin_input`, `origin_output` or `grad` is not equal to 5.
+        ValueError: If data types of all inputs are not equal.
+        ValueError: If the shapes of `origin_input` and `grad` are not equal.
+
+    Supported Platforms:
+        ``Ascend`` ``GPU`` ``CPU``
+    """
+
+    @prim_attr_register
+    def __init__(self, kernel_size=(1, 1, 1, 1, 1), strides=(1, 1, 1, 1, 1), pad_mode='VALID', data_format="NCDHW"):
+        validator.check_value_type('kernel_size', kernel_size, [int, tuple], self.name)
+        validator.check_value_type('strides', strides, [int, tuple], self.name)
+        validator.check_value_type('pad_mode', pad_mode, [str], self.name)
+        self.format = validator.check_string(data_format, ['NCDHW'], 'format', self.name)
+        self.pad_mode = validator.check_string(pad_mode.upper(), ['VALID', 'SAME'], 'pad_mode', self.name)
+        self.kernel_size = _check_3d_int_or_tuple("kernel_size", kernel_size, self.name,
+                                                  allow_five=True, ret_five=True)
+        self.add_prim_attr("kernel_size", self.kernel_size)
+        self.strides = _check_3d_int_or_tuple("strides", strides, self.name, allow_five=True, ret_five=True)
+        self.add_prim_attr("strides", self.strides)
+
+    def infer_shape(self, x_shape, y_shape, grad_shape):
+        validator.check_equal_int(len(x_shape), 5, "x rank", self.name)
+        validator.check('x_shape', x_shape, 'grad_shape', grad_shape, prim_name=self.name)
+        pad_list = _get_max_pool3d_grad_pads_by_pad_mode(x_shape, self.kernel_size, self.strides, self.pad_mode)
+        for pad in pad_list:
+            validator.check_non_negative_int(pad, 'element of pad_list', self.name)
+        self.add_prim_attr("pad_list", pad_list)
+        return y_shape
+
+    def infer_dtype(self, x_dtype, y_dtype, grad_dtype):
+        args = {'x_dtype': x_dtype, 'y_dtype': y_dtype}
+        validator.check_tensors_dtypes_same_and_valid(args, [mstype.float16, mstype.float32], self.name)
+        validator.check_tensor_dtype_valid('grad_dtype', grad_dtype, [mstype.float16, mstype.float32], self.name)
+        return x_dtype
+
+
+class MaxPoolGradWithArgmax(Primitive):
+    """Computes the gradients of MaxPoolWithArgmax."""
+    @prim_attr_register
+    def __init__(self, kernel_size=1, strides=1, pad_mode="VALID", data_format="NCHW"):
+        self.init_prim_io_names(inputs=['x_origin', 'out_origin', 'grad'], outputs=['output'])
+        validator.check_value_type('kernel_size', kernel_size, [int, tuple], self.name)
+        validator.check_value_type('strides', strides, [int, tuple], self.name)
+        validator.check_value_type('pad_mode', pad_mode, [str], self.name)
+        self.pad_mode = validator.check_string(pad_mode.upper(), ['VALID', 'SAME'], 'pad_mode', self.name)
+        self.add_prim_attr("pad_mode", self.pad_mode)
+        self.format = validator.check_string(data_format, ['NCHW', 'NHWC'], 'format', self.name)
+        if context.get_context("device_target") != "GPU" and self.format == "NHWC":
+            raise ValueError("NHWC format only support in GPU target.")
+        self.is_maxpoolgradwithargmax = (self.name == "MaxPoolGradWithArgmax")
+        if not self.is_maxpoolgradwithargmax:
+            self.add_prim_attr('data_format', self.format)
+
+        def _grad_check_int_or_tuple(arg_name, arg_val):
+            validator.check_value_type(arg_name, arg_val, (int, tuple), self.name)
+            error_msg = ValueError(f"For '{self.name}' the '{arg_name}' must be an positive int number "
+                                   f"or a tuple of two or four positive int numbers, but got {arg_val}")
+            if isinstance(arg_val, int):
+                ret = (1, arg_val, arg_val, 1)
+            elif len(arg_val) == 2:
+                ret = (1, arg_val[0], arg_val[1], 1)
+            elif len(arg_val) == 4:
+                ret = arg_val
+            else:
+                raise error_msg
+            # whether all elements of tuple are positive integers
+            for item in ret:
+                if not isinstance(item, int) or item <= 0:
+                    raise error_msg
+            return ret
+
+        kernel_size = _grad_check_int_or_tuple("kernel_size", kernel_size)
+        self.kernel_size = kernel_size
+        self.add_prim_attr("kernel_size", self.kernel_size)
+
+        strides = _grad_check_int_or_tuple("strides", strides)
+        self.strides = strides
+        self.add_prim_attr("strides", self.strides)
+
+
+class MaxPoolGradWithArgmaxV2(Primitive):
+    """Gradients of the MaxPoolWithArgmaxV2 operation."""
+
+    @prim_attr_register
+    def __init__(self, kernel_size, strides=None, pads=0, dilation=(1, 1), ceil_mode=False, argmax_type=mstype.int64):
+        self.init_prim_io_names(inputs=['x', 'grad', 'argmax'], outputs=['y'])
+        self.kernel_size = _check_positive_int_or_tuple("kernel_size", kernel_size, self.name, allow_four=True,
+                                                        ret_four=True)
+        self.add_prim_attr('kernel_size', self.kernel_size)
+        if strides is None:
+            strides = kernel_size
+        self.strides = _check_positive_int_or_tuple("strides", strides, self.name, allow_four=True, ret_four=True)
+        self.add_prim_attr('strides', self.strides)
+        self.pads = _check_positive_int_or_tuple("pads", pads, self.name, allow_four=True, ret_four=True,
+                                                 strict_positive=False)
+        self.add_prim_attr('pads', self.pads)
+        validator.check_value_type('ceil_mode', ceil_mode, bool, self.name)
+        self.add_prim_attr('ceil_mode', self.ceil_mode)
+        self.dilation = _check_positive_int_or_tuple("dilation", dilation, self.name, allow_four=True, ret_four=True)
+        self.add_prim_attr('dilation', self.dilation)
+        self.add_prim_attr('argmax_type', self.argmax_type)
+
+
+class MaxPool3DGradWithArgmax(Primitive):
+    """Gradients of the maxpool3Dwithargmax operation."""
+
+    @prim_attr_register
+    def __init__(self, ksize, strides, pads, dilation=(1, 1, 1), ceil_mode=False, data_format="NCDHW"):
+        self.init_prim_io_names(inputs=['x', 'grads', 'argmax'], outputs=['y'])
+        validator.check_value_type('ceil_mode', ceil_mode, bool, self.name)
+        validator.check_value_type('data_format', data_format, str, self.name)
+        self.data_format = validator.check_string(data_format, ['NCDHW'], 'data_format', self.name)
+        self.ksize = _check_3d_int_or_tuple("ksize", ksize, self.name, ret_five=False)
+        self.add_prim_attr('ksize', self.ksize)
+        self.strides = _check_3d_int_or_tuple("strides", strides, self.name, ret_five=False)
+        self.add_prim_attr('strides', self.strides)
+        self.pads = _check_3d_int_or_tuple("pads", pads, self.name, greater_zero=False, ret_five=False)
+        self.add_prim_attr('pads', self.pads)
+        self.dilation = _check_3d_int_or_tuple("dilation", dilation, self.name, allow_five=True, ret_five=False)
+        self.add_prim_attr('dilation', self.dilation)
+
+
+class MaxPoolGradGradWithArgmax(_PoolGrad):
+    r"""
+    Computes the gradients of MaxPoolGradWithArgmax.
+
+    Args:
+        kernel_size (Union[int, tuple[int]]): The size of kernel used to take the maximum value,
+            is an int number that represents height and width are both kernel_size, or a tuple
+            of two int numbers that represent height and width respectively. Default: 1.
+        strides (Union[int, tuple[int]]): The distance of kernel moving, an int number that represents
+            the height and width of movement are both strides, or a tuple of two int numbers that
+            represent height and width of movement respectively. Default: 1.
+        pad_mode (str): The optional value for pad mode, is "same" or "valid", not case sensitive.
+            Default: "valid".
+
+            - same: Adopts the way of completion. The height and width of the output will be the same as
+              the input. The total number of padding will be calculated in horizontal and vertical
+              directions and evenly distributed to top and bottom, left and right if possible.
+              Otherwise, the last extra padding will be done from the bottom and the right side.
+
+            - valid: Adopts the way of discarding. The possible largest height and width of output
+              will be returned without padding. Extra pixels will be discarded.
+
+    Inputs:
+        - **x** (Tensor) - Tensor with data format "NCHW".
+          For Ascend, data type must be float16. For CPU and GPU, data type support float16 and float32.
+        - **grad** (Tensor) - Data type and shape same as `x`.
+        - **argmax** (Tensor) - Data type must be int32 or int64.
+
+    Outputs:
+        Tensor, with data type same as `x`. Shape same as `argmax`.
+
+    Raises:
+        TypeError: If kernel_size is neither int nor a tuple of 2/4 int numbers.
+        TypeError: If strides is neither int nor a tuple of 2/4 int numbers.
+        TypeError: If pad_mode is not string.
+        ValueError: If pad_mode is neither "same" nor "valid"(not case sensitive).
+        TypeError: For Ascend, the data types of `x` and `grad` are not float16.
+        For CPU or GPU, the data types of `x` and `grad` are neither float16 nor float32.
+        TypeError: The data type of `argmax` is neither int32 nor int64.
+        ValueError: If the rank of `x`, `grad` or `argmax` is not equal to 4.
+        ValueError: If the shapes of `x` and `grad` are not equal.
+
+    Supported Platforms:
+        ``Ascend`` ``GPU`` ``CPU``
+    """
+
+    @prim_attr_register
+    def __init__(self, kernel_size=1, strides=1, pad_mode="VALID"):
+        self.init_prim_io_names(inputs=['x', 'grad', 'argmax'], outputs=['output'])
+        super(MaxPoolGradGradWithArgmax, self).__init__(kernel_size, strides, pad_mode)
+
+    def infer_shape(self, x_shape, grad_shape, argmax_shape):
+        if not grad_shape:
+            raise TypeError("The dout of MaxPoolGradGradWithArgmax must be a Tensor.")
+        return x_shape
+
+    def infer_dtype(self, x_dtype, grad_dtype, argmax_dtype):
+        args = {'x_dtype': x_dtype, 'grad_dtype': grad_dtype}
+        validator.check_tensors_dtypes_same_and_valid(args, [mstype.float16], self.name)
+        return grad_dtype
+
+
+class MinimumGradGrad(Primitive):
+    """Grad for minimum_grad."""
+    @prim_attr_register
+    def __init__(self):
+        """Initialize MinimumGradGrad"""
+        super().__init__("MinimumGradGrad")
+        self.init_prim_io_names(inputs=['x1', 'x2', 'grad_y1', 'grad_y2'],
+                                outputs=['sopd_x1', 'sopd_x2', 'sopd_grads'])
+
+
+class L2NormalizeGrad(Primitive):
+    r"""
+    Gradients of L2 normalize.
+
+    Args:
+        axis (Union[list(int), tuple(int), int]): The begin axis for the input to apply L2 normalize. Default: 0.
+        epsilon (float): A small value added for numerical stability. Default: 1e-4.
+
+    Inputs:
+        - **input_x** (Tensor) - Must be the input `weight` of forward operator L2Normalize.
+        - **out** (Tensor) - Must be the output of forward operator L2Normalize.
+        - **dout** (Tensor) - The backprop of the next layer.
+
+    Outputs:
+        Tensor, gradients of L2Normalize `input_x`.
+    """
+
+    @prim_attr_register
+    def __init__(self, axis=0, epsilon=1e-4):
+        axis = [axis] if isinstance(axis, int) else axis
+        validator.check_value_type('axis', axis, [list, tuple], self.name)
+        validator.check_value_type('epsilon', epsilon, [int, float], self.name)
+        self.add_prim_attr('axis', axis)
+        self.init_attrs['axis'] = axis
+        if len(axis) != 1:
+            raise TypeError("The length of axis must be 1, later will support multiple axis!")
+
+
+class LSTMGradData(Primitive):
+    """Computes the data gradients of LSTM."""
+
+    @prim_attr_register
+    def __init__(self, input_size, hidden_size, num_layers, has_bias, bidirectional, dropout):
+        self.input_size = validator.check_positive_int(input_size, 'input_size', self.name)
+        self.hidden_size = validator.check_positive_int(hidden_size, 'hidden_size', self.name)
+        self.num_layers = validator.check_positive_int(num_layers, 'num_layers', self.name)
+        self.has_bias = validator.check_value_type('has_bias', has_bias, (bool,), self.name)
+        self.bidirectional = validator.check_value_type('bidirectional', bidirectional, (bool,), self.name)
+        self.dropout = validator.check_value_type("dropout", dropout, [float], self.name)
+        self.dropout = validator.check_float_range(dropout, 0, 1, validator.INC_BOTH, 'dropout', self.name)
+
+        if bidirectional:
+            self.num_directions = 2
+        else:
+            self.num_directions = 1
+
+
+class LSTMGradWeight(Primitive):
+    """Computes the weight gradients of LSTM."""
+
+    @prim_attr_register
+    def __init__(self, input_size, hidden_size, num_layers, has_bias, bidirectional, dropout):
+        self.input_size = validator.check_positive_int(input_size, 'input_size', self.name)
+        self.hidden_size = validator.check_positive_int(hidden_size, 'hidden_size', self.name)
+        self.num_layers = validator.check_positive_int(num_layers, 'num_layers', self.name)
+        self.has_bias = validator.check_value_type('has_bias', has_bias, (bool,), self.name)
+        self.bidirectional = validator.check_value_type('bidirectional', bidirectional, (bool,), self.name)
+        self.dropout = validator.check_value_type("dropout", dropout, [float], self.name)
+        self.dropout = validator.check_float_range(dropout, 0, 1, validator.INC_BOTH, 'dropout', self.name)
+
+        if bidirectional:
+            self.num_directions = 2
+        else:
+            self.num_directions = 1
+
+
+class LSTMGrad(Primitive):
+    """Computes the data and weight gradients of LSTM."""
+
+    @prim_attr_register
+    def __init__(self, input_size, hidden_size, num_layers, has_bias, bidirectional, dropout, proj_size=0):
+        self.input_size = validator.check_positive_int(input_size, 'input_size', self.name)
+        self.hidden_size = validator.check_positive_int(hidden_size, 'hidden_size', self.name)
+        self.proj_size = validator.check_int_range(proj_size, 0, hidden_size, validator.INC_LEFT,
+                                                   'proj_size', self.name)
+        self.num_layers = validator.check_positive_int(num_layers, 'num_layers', self.name)
+        self.has_bias = validator.check_value_type('has_bias', has_bias, (bool,), self.name)
+        self.bidirectional = validator.check_value_type('bidirectional', bidirectional, (bool,), self.name)
+        self.dropout = validator.check_value_type("dropout", dropout, [float], self.name)
+        self.dropout = validator.check_float_range(dropout, 0, 1, validator.INC_BOTH, 'dropout', self.name)
+
+        if bidirectional:
+            self.num_directions = 2
+        else:
+            self.num_directions = 1
+
+
+class DynamicRNNGrad(Primitive):
+    """Computes the input gradients of DynamicRNN."""
+
+    @prim_attr_register
+    def __init__(self,
+                 cell_type='LSTM',
+                 direction='UNIDIRECTIONAL',
+                 cell_depth=1,
+                 use_peephole=False,
+                 keep_prob=1.0,
+                 cell_clip=-1.0,
+                 num_proj=0,
+                 time_major=True,
+                 forget_bias=0.0):
+        self.forget_bias = validator.check_value_type("forget_bias", forget_bias, [float], self.name)
+
+
+class GruGradData(PrimitiveWithInfer):
+    """Computes the data gradients of GRU."""
+
+    @prim_attr_register
+    def __init__(self, input_size, hidden_size, num_layers, has_bias, bidirectional, dropout):
+        self.input_size = validator.check_positive_int(input_size, 'input_size', self.name)
+        self.hidden_size = validator.check_positive_int(hidden_size, 'hidden_size', self.name)
+        self.num_layers = validator.check_positive_int(num_layers, 'num_layers', self.name)
+        self.has_bias = validator.check_value_type('has_bias', has_bias, (bool,), self.name)
+        self.bidirectional = validator.check_value_type('bidirectional', bidirectional, (bool,), self.name)
+        self.dropout = validator.check_value_type("dropout", dropout, [float], self.name)
+        self.dropout = validator.check_float_range(dropout, 0, 1, validator.INC_BOTH, 'dropout', self.name)
+
+        if bidirectional:
+            self.num_directions = 2
+        else:
+            self.num_directions = 1
+
+    def infer_shape(self, y_shape, dy_shape, dhy_shape, w_shape,
+                    hx_shape, reserve_shape, state_shape):
+        # dhy and dcy should be same shape
+        validator.check_equal_int(len(dhy_shape), 3, "h_shape", self.name)
+
+        validator.check_int(dhy_shape[0], self.num_layers * self.num_directions, validator.EQ, "h_shape[0]", self.name)
+        validator.check_equal_int(dhy_shape[2], self.hidden_size, "h_shape[2]", self.name)
+
+        validator.check_equal_int(len(dy_shape), 3, "dy_shape", self.name)
+        validator.check_equal_int(dy_shape[1], dhy_shape[1], "dy[1]", self.name)
+        validator.check_int(dy_shape[2], self.hidden_size * self.num_directions, validator.EQ, "dy[2]", self.name)
+
+        dx_shape = (y_shape[0], y_shape[1], self.input_size)
+        dhx_shape = dhy_shape
+
+        return (dx_shape, dhx_shape)
+
+    def infer_dtype(self, y_dtype, dy_dtype, dhy_dtype, w_dtype,
+                    hx_dtype, reserve_dtype, state_dtype):
+        args = {"dy": dy_dtype, "dhy": dhy_dtype}
+        validator.check_tensors_dtypes_same_and_valid(args, (mstype.float32, mstype.float16), self.name)
+        return (dy_dtype, dy_dtype)
+
+
+class GruGradWeight(PrimitiveWithInfer):
+    """Computes the weight gradients of GRU."""
+
+    @prim_attr_register
+    def __init__(self, input_size, hidden_size, num_layers, has_bias, bidirectional, dropout):
+        self.input_size = validator.check_positive_int(input_size, 'input_size', self.name)
+        self.hidden_size = validator.check_positive_int(hidden_size, 'hidden_size', self.name)
+        self.num_layers = validator.check_positive_int(num_layers, 'num_layers', self.name)
+        self.has_bias = validator.check_value_type('has_bias', has_bias, (bool,), self.name)
+        self.bidirectional = validator.check_value_type('bidirectional', bidirectional, (bool,), self.name)
+        self.dropout = validator.check_value_type("dropout", dropout, [float], self.name)
+        self.dropout = validator.check_float_range(dropout, 0, 1, validator.INC_BOTH, 'dropout', self.name)
+
+        if bidirectional:
+            self.num_directions = 2
+        else:
+            self.num_directions = 1
+
+    def infer_shape(self, x_shape, hx_shape, y_shape, reserve_shape, state_shape):
+        weight_size = 0
+        gate_size = 3 * self.hidden_size
+        for layer in range(self.num_layers):
+            for _ in range(self.num_directions):
+                input_layer_size = self.input_size if layer == 0 else self.hidden_size * self.num_directions
+                weight_size += gate_size * input_layer_size
+                weight_size += gate_size * self.hidden_size
+                if self.has_bias:
+                    weight_size += 2 * gate_size
+
+        return (weight_size, 1, 1)
+
+    def infer_dtype(self, x_dtype, hx_dtype, y_dtype, reserve_dtype, state_dtype):
+        return hx_dtype
+
+
+class GRUV2Grad(Primitive):
+    """Computes the grad gradients of GRU."""
+
+    @prim_attr_register
+    def __init__(self, input_size, hidden_size, num_layers, has_bias, bidirectional, dropout):
+        self.input_size = validator.check_positive_int(input_size, 'input_size', self.name)
+        self.hidden_size = validator.check_positive_int(hidden_size, 'hidden_size', self.name)
+        self.num_layers = validator.check_positive_int(num_layers, 'num_layers', self.name)
+        self.has_bias = validator.check_value_type('has_bias', has_bias, (bool,), self.name)
+        self.bidirectional = validator.check_value_type('bidirectional', bidirectional, (bool,), self.name)
+        self.dropout = validator.check_value_type("dropout", dropout, [float], self.name)
+        self.dropout = validator.check_float_range(dropout, 0, 1, validator.INC_BOTH, 'dropout', self.name)
+
+        if bidirectional:
+            self.num_directions = 2
+        else:
+            self.num_directions = 1
+
+
+class DynamicGRUV2Grad(Primitive):
+    r"""
+    Computes the input gradients of DynamicGRUV2.
+
+    Args:
+        direction (str): A string identifying the direction in the op. Default: 'UNIDIRECTIONAL'.
+            Only 'UNIDIRECTIONAL' is currently supported.
+        cell_depth (int): An integer identifying the cell depth in the op. Default: 1.
+        keep_prob (float): A float identifying the keep prob in the op. Default: 1.0.
+        cell_clip (float): A float identifying the cell clip in the op. Default: -1.0.
+        num_proj (int): An integer identifying the num proj in the op. Default: 0.
+        time_major (bool): A bool identifying the time major in the op. Default: ``True``.
+        gate_order (str): An string identifying the gate order in weight and bias. Default: 'rzh.
+            'zrh' is another option.
+        reset_after (bool): An bool identifying whether to apply reset gate after matrix multiplication.
+            Default: ``True``.
+
+    Inputs:
+        - **x** (Tensor) - Current words. Tensor of shape :math:`(num\_step, batch\_size, input\_size)`.
+          The data type must be float16 or float32.
+        - **weight_input** (Tensor) - Weight. Tensor of shape :math:`(input\_size, 3 x hidden\_size)`.
+          The data type must be float16 or float32.
+        - **weight_hidden** (Tensor) - Bias. Tensor of shape :math:`(hidden\_size, 3 x hidden\_size)`.
+          The data type must be float16 or float32.
+        - **y** (Tensor) - A Tensor of shape :math:
+          if num_proj > 0 `(num_step, batch_size, min(hidden_size, num_proj)`,
+          if num_proj == 0 `(num_step, batch_size, hidden_size)`.
+          The data type must be float16 or float32.
+        - **init_h** (Tensor) - Hidden state of initial time.
+          Tensor of shape :math:`(batch\_size, hidden\_size)`.
+          The data type must be float16 or float32.
+        - **h** (Tensor) - A Tensor of shape :math:`(num\_step, batch\_size, hidden\_size)`.
+          The data type must be float16 or float32.
+        - **dy** (Tensor) - Gradient of `y`, has the same shape and data type as `y`.
+        - **dh** (Tensor) - Gradient of `h`, has the same shape and data type as `init_h`.
+        - **update** (Tensor) - A Tensor of shape :math:`(num\_step, batch\_size, hidden\_size)`.
+          The data type must be float16 or float32.
+        - **reset** (Tensor) - A Tensor of shape :math:`(num\_step, batch\_size, hidden\_size)`.
+          The data type must be float16 or float32.
+        - **new** (Tensor) - A Tensor of shape :math:`(num\_step, batch\_size, hidden\_size)`.
+          The data type must be float16 or float32.
+        - **hidden_new** (Tensor) - A Tensor of shape :math:`(num\_step, batch\_size, hidden\_size)`.
+          The data type must be float16 or float32.
+        - **seq_length** (Tensor) - The length of each batch. Tensor of shape :math:`(batch\_size)`.
+          Only `None` is currently supported.
+        - **mask** (Tensor) - A 4-D Tensor. The data type must be float16 or float32.
+
+    Outputs:
+        - **dw_input** (Tensor) - A Tensor has the same shape as `weight_input`.
+          Has the same type with input `x`.
+        - **dw_hidden** (Tensor) - A Tensor has the same shape as `weight_hidden`.
+          Has the same type with input `x`.
+        - **db_input** (Tensor) - A Tensor of shape :math:`(3 x hidden\_size)`.
+          Has the same type with input `init\_h`.
+        - **db_hidden** (Tensor) - A Tensor of shape :math:`(3 x hidden\_size)`.
+          Has the same type with input `init\_h`.
+        - **dx** (Tensor) - A Tensor of shape :math:`(num\_step, batch\_size, hidden\_size)`.
+          Has the same type with input `x`.
+        - **dh_prev** (Tensor) - A Tensor of shape :math:`(batch\_size, hidden\_size)`.
+          Has the same type with input `init\_h`.
+    """
+
+    @prim_attr_register
+    def __init__(self,
+                 direction='UNIDIRECTIONAL',
+                 cell_depth=1,
+                 keep_prob=1.0,
+                 cell_clip=-1.0,
+                 num_proj=0,
+                 time_major=True,
+                 gate_order="rzh",
+                 reset_after=True):
+        self.cell_depth = validator.check_value_type("cell_depth", cell_depth, [int], self.name)
+        self.keep_prob = validator.check_value_type("keep_prob", keep_prob, [float], self.name)
+        self.cell_clip = validator.check_value_type("cell_clip", cell_clip, [float], self.name)
+        self.num_proj = validator.check_non_negative_int(num_proj, "num_proj", self.name)
+        self.time_major = validator.check_value_type("time_major", time_major, [bool], self.name)
+        self.direction = validator.check_string(direction, ['UNIDIRECTIONAL'], "direction", self.name)
+        self.gate_order = validator.check_string(gate_order, ['zrh', 'rzh'], "gate_order", self.name)
+        self.reset_after = validator.check_value_type("reset_after", reset_after, [bool], self.name)
+        self.init_prim_io_names(inputs=[
+            "x", "weight_input", "weight_hidden", "y", "init_h", "h", "dy",
+            "dh", "update", "reset", "new", "hidden_new", "seq_length", "mask"
+        ],
+                                outputs=[
+                                    "dw_input", "dw_hidden", "db_input",
+                                    "db_hidden", "dx", "dh_prev"
+                                ])
+
+
+class RandomGammaGrad(Primitive):
+    r"""
+    Computes the derivative of a random sample of Gamma with respect to alpha.:
+
+    Inputs:
+        - **alpha** (Tensor) - α is the shape parameter of RandomGamma distribution.
+        It must be greater than 0. Must be one of the following types: float32, float64.
+        - **sample** (Tensor) - The sample of random gamma tensor. Must be one of the
+        following types: float32, float64.
+
+    Outputs:
+        The dtype is the same type as alpha.
+        The output shape is derived from the input through broadcasting.
+
+    Raises:
+        TypeError: If data type of `alpha` and `sample` is not float32 or float64.
+        TypeError: If data type of `alpha` and `sample` is not same.
+        ValueError: If the shape last dim of `sample` and `alpha` is not equal.
+
+    Supported Platforms:
+        ``GPU``
+
+    Examples:
+        >>> alpha = Tensor(np.array([1., 0.6, 3., 26.]), mstype.float32)
+        >>> sample = Tensor(np.array([6., 7, 11., 0.5]), mstype.float32)
+        >>> randomgammagrad = ops.RandomGammaGrad()
+        >>> output = randomgammagrad(alpha, sample)
+        >>> print(output)
+        [2.5142431 3.4334087 1.8847835 0.07780622]
+    """
+
+    @prim_attr_register
+    def __init__(self):
+        """Initialize RandomGammaGrad"""
+        self.init_prim_io_names(inputs=['alpha', 'sample'], outputs=['output'])
+        self.add_prim_attr("side_effect_hidden", True)
+
+
+class ROIAlignGrad(Primitive):
+    """
+    ROIAlignGrad operator.
+
+    Args:
+       pooled_height (int): The output feature height.
+       pooled_width (int): The output feature width.
+       spatial_scale (float): The feature stride.
+       sample_num (int): Number of sampling points. Default: 2.
+    """
+
+    @prim_attr_register
+    def __init__(self, pooled_height, pooled_width, spatial_scale, sample_num=2):
+        """Initialize ROIAlignGrad"""
+        self.init_prim_io_names(inputs=["dy", "rois", "xdiff_shape"], outputs=["dx"])
+        validator.check_value_type("pooled_height", pooled_height, [int], self.name)
+        validator.check_value_type("pooled_width", pooled_width, [int], self.name)
+        validator.check_value_type("spatial_scale", spatial_scale, [float], self.name)
+        validator.check_value_type("sample_num", sample_num, [int], self.name)
+        self.pooled_height = pooled_height
+        self.pooled_width = pooled_width
+        self.spatial_scale = spatial_scale
+        self.sample_num = sample_num
+
+
+class PsROIPoolingGrad(PrimitiveWithInfer):
+    """
+    PsROIPoolingGrad operator.
+    """
+
+    @prim_attr_register
+    def __init__(self, batch_size, channels, height, width, num_rois,
+                 pooled_height, pooled_width, spatial_scale, out_dim):
+        """Initialize PsROIPoolingGrad"""
+        validator.check_value_type("batch_size", batch_size, [int], self.name)
+        validator.check_value_type("channels", channels, [int], self.name)
+        validator.check_value_type("height", height, [int], self.name)
+        validator.check_value_type("width", width, [int], self.name)
+        validator.check_value_type("num_rois", num_rois, [int], self.name)
+        validator.check_value_type("pooled_height", pooled_height, [int], self.name)
+        validator.check_value_type("pooled_width", pooled_width, [int], self.name)
+        validator.check_value_type("spatial_scale", spatial_scale, [float], self.name)
+        validator.check_value_type("out_dim", out_dim, [int], self.name)
+        self.batch_size = batch_size
+        self.channels = channels
+        self.height = height
+        self.width = width
+        self.num_rois = num_rois
+        self.pooled_height = pooled_height
+        self.pooled_width = pooled_width
+        self.spatial_scale = spatial_scale
+        self.out_dim = out_dim
+
+    def infer_shape(self, ydiff_shape, rois_shape, mapping_channel_shape):
+        return [self.batch_size, self.channels, self.height, self.width]
+
+    def infer_dtype(self, ydiff_type, rois_type, mapping_channel_type):
+        return ydiff_type
+
+
+class _ActivationGrad(PrimitiveWithInfer):
+    """_ActivationGrad base class."""
+
+    @prim_attr_register
+    def __init__(self):
+        self.init_prim_io_names(inputs=['y_grad', 'x'], outputs=['output'])
+
+    def infer_shape(self, y_grad_shape, x_shape):
+        return x_shape
+
+    def infer_dtype(self, y_grad_dtype, x_dtype):
+        valid_dtypes = (mstype.float16, mstype.float32)
+        validator.check_tensor_dtype_valid("y_grad", y_grad_dtype, valid_dtypes, self.name)
+        validator.check_tensor_dtype_valid("x", x_dtype, valid_dtypes, self.name)
+        return x_dtype
+
+
+class SigmoidCrossEntropyWithLogitsGrad(Primitive):
+    """Computes the gradients of `SigmoidCrossEntropyWithLogits`."""
+
+    @prim_attr_register
+    def __init__(self):
+        """Initialize SigmoidCrossEntropyWithLogitsGrad"""
+        self.init_prim_io_names(inputs=['x', 'y', 'dout'], outputs=['x_grad'])
+
+
+class SliceGrad(PrimitiveWithInfer):
+    """Reverse of slice."""
+
+    @prim_attr_register
+    def __init__(self):
+        """Initialize SliceGrad"""
+        self.init_prim_io_names(inputs=['dy', 'x', 'begin', 'size'], outputs=['dx'])
+
+    def __infer__(self, dy, x, begin, size):
+        dy_shape, x_shape, size_value, begin_v = dy['shape'], x['shape'], size['value'], begin['value']
+        dy_shape_len = len(dy_shape)
+        if size_value is not None and not is_shape_unknown(x_shape) and not is_shape_unknown(dy_shape):
+            size_value = list(size_value)
+            for i in range(dy_shape_len):
+                if size_value[i] == -1:
+                    size_value[i] = x_shape[i] - begin_v[i]
+                validator.check(f'dy_shape[{i}]', dy_shape[i], f'x_shape[{i}]', x_shape[i], validator.LE, self.name)
+                validator.check(f'dy_shape[{i}]', dy_shape[i], f'size_shape[{i}]',
+                                size_value[i], validator.EQ, self.name)
+
+        return {'shape': x_shape,
+                'dtype': x['dtype'],
+                'value': None}
+
+
+class SmoothL1LossGrad(Primitive):
+    """Computes gradient for prediction on SmoothL1Loss."""
+
+    @prim_attr_register
+    def __init__(self, beta=1.0, reduction='none'):
+        self.add_prim_attr('sigma', self.beta)
+        self.reduction = validator.check_string(
+            reduction, ['none', 'sum', 'mean'], 'reduction', self.name)
+
+
+class SoftMarginLossGrad(Primitive):
+    """Computes gradient for prediction on SoftMarginLoss."""
+
+    @prim_attr_register
+    def __init__(self, reduction="mean"):
+        self.init_prim_io_names(inputs=['predict', 'label', "dout"], outputs=['gradient'])
+        self.reduction = validator.check_string(reduction, ['none', 'sum', 'mean'], 'reduction', self.name)
+
+
+class StridedSliceGrad(Primitive):
+    """
+    Performs grad of StridedSlice operation.
+
+    Args:
+        begin_mask (int): Start indexing the slice. Default: 0.
+        end_mask (int): End indexing the slice. Default: 0.
+        ellipsis_mask (int): An int32 mask. Default: 0.
+        new_axis_mask (int): An int32 mask. Default: 0.
+        shrink_axis_mask (int): An int32 mask. Default: 0.
+
+    Returns:
+        Tensor, has the same shape of input.
+    """
+
+    @prim_attr_register
+    def __init__(self,
+                 begin_mask=0,
+                 end_mask=0,
+                 ellipsis_mask=0,
+                 new_axis_mask=0,
+                 shrink_axis_mask=0):
+        """Initialize StridedSliceGrad"""
+        validator.check_value_type('begin_mask', begin_mask, [int], self.name)
+        validator.check_value_type('end_mask', end_mask, [int], self.name)
+        validator.check_value_type('ellipsis_mask', ellipsis_mask, [int], self.name)
+        validator.check_value_type('new_axis_mask', new_axis_mask, [int], self.name)
+        validator.check_value_type('shrink_axis_mask', shrink_axis_mask, [int], self.name)
+        self.init_prim_io_names(inputs=['dy', 'shapex', 'begin', 'end', 'strides'], outputs=['output'])
+
+
+class SoftplusGrad(Primitive):
+    """Computes gradient for the Softplus activation."""
+
+    @prim_attr_register
+    def __init__(self):
+        self.init_prim_io_names(inputs=['gradients', 'features'], outputs=['backprops'])
+
+
+class TanhGrad(Primitive):
+    """Computes gradient of hyperbolic tangent of input element-wise."""
+
+    @prim_attr_register
+    def __init__(self):
+        """Initialize TanhGrad"""
+        self.init_prim_io_names(inputs=['y', 'dy'], outputs=['z'])
+
+
+class MirrorPadGrad(Primitive):
+    """Gradients of MirrorPad operation."""
+
+    @prim_attr_register
+    def __init__(self, mode="REFLECT"):
+        """Initialize MirrorPad"""
+        self.init_prim_io_names(inputs=['dy', 'paddings'], outputs=['output'])
+        validator.check_string(mode, ['REFLECT', 'SYMMETRIC'], 'mode', self.name)
+        self.mode = mode
+
+
+class PadV3Grad(Primitive):
+    """Gradients of PadV3 operation."""
+
+    @prim_attr_register
+    def __init__(self, mode='reflect', paddings_contiguous=True):
+        """Initialize Padv3Grad"""
+        self.add_prim_attr("cust_aicpu", self.name)
+        self.init_prim_io_names(inputs=['x', 'paddings'], outputs=['y'])
+        validator.check_string(mode, ['reflect', 'edge', 'circular'], 'mode', self.name)
+        validator.check_bool(paddings_contiguous, "paddings_contiguous", self.name)
+        self.mode = mode
+        self.paddings_contiguous = paddings_contiguous
+
+
+class EmbeddingLookupCommGrad(PrimitiveWithInfer):
+    """
+    Performs the gradient for the communication part of EmbeddingLookup operator.
+
+    This works ONLY when 'reduce_scatter_flag' is True in 'EmbeddingLookup'. Roughly speaking,
+    this primitive is implemented by StridedSlice --> _HostAllGather --> Concat. This primitive runs on host.
+    """
+
+    @prim_attr_register
+    def __init__(self):
+        self.init_prim_io_names(inputs=['dy', 'split_num'], outputs=['output'])
+        self.set_device('CPU')
+        self.tuple_setitem = Primitive('tuple_setitem')
+
+    def __infer__(self, dy, split_num):
+        """
+        This primitive is implemented by three steps:
+            1) Splits the 'dy' along dimension 0 into 'split_num' parts.
+            2) For each part, perform _HostAllGather((0, 1, 2, 3, 4, 5, 6, 7)) on the host.
+            3) After _HostAllGather, there are still 'split_num' parts in each process. Then, perform Concat on them
+              along dimension 0.
+
+        The output shape of this primitive: shape(output)[0] == shape(dy)[0] * 8
+        """
+        dy_shape = tuple(dy['shape'])
+        split_num_value = split_num['value']
+        validator.check_value_type("split_num_value", split_num_value, [int], self.name)
+        dy_shape_all = self.tuple_setitem(dy_shape, 0, dy_shape[0] * 8)
+        return {'shape': dy_shape_all,
+                'dtype': dy['dtype'],
+                'value': None}
+
+
+class RefToEmbed(Primitive):
+    r"""
+    Make a key from Ref.
+
+    The Key is a symbolic_key, is a embedding on Parameter, which is used as a key of the variable in env_type,
+    and get items by operation `EnvironGet` with the symbolic_key instance. The `Parameter` is a ref.
+
+    Inputs:
+        - **input** (Ref) - Target ref, ref is short for reference. The value of a Parameter is a ref.
+
+    Outputs:
+        symbolic_key, made from the Ref.
+
+    Examples:
+        >>> class Net(nn.Cell):
+        >>>     def __init__(self):
+        >>>         super(Net, self).__init__()
+        >>>         self.weight = mindspore.Parameter(1.0, name='weight')
+        >>>
+        >>>     def construct(self):
+        >>>         key = RefToEmbed()(self.weight)
+        >>>         return key, self.weight
+    """
+    __mindspore_signature__ = (
+        sig.make_sig('variable', sig.sig_rw.RW_REF),
+    )
+
+    @prim_attr_register
+    def __init__(self):
+        pass
+
+
+class BasicLSTMCellCStateGrad(PrimitiveWithInfer):
+    """Computes the state gradients of BasicLSTMCell."""
+
+    @prim_attr_register
+    def __init__(self, forget_bias, activation):
+        self.forget_bias = validator.check_value_type("forget_bias", forget_bias, [float], self.name)
+        self.activation = validator.check_string(activation, ['tanh'], "activation", self.name)
+
+    def infer_shape(self, c_shape, dht_shape, dct_shape, it_shape, jt_shape, ft_shape, ot_shape, tanhct_shape):
+        # dhy and dcy should be same shape
+        validator.check_equal_int(len(c_shape), 2, "c rank", self.name)
+        validator.check("dht rank", len(dht_shape), "c rank", len(c_shape), validator.EQ, self.name)
+        validator.check("dct rank", len(dct_shape), "c rank", len(c_shape), validator.EQ, self.name)
+        validator.check("it rank", len(it_shape), "c rank", len(c_shape), validator.EQ, self.name)
+        validator.check("jt rank", len(jt_shape), "c rank", len(c_shape), validator.EQ, self.name)
+        validator.check("ft rank", len(ft_shape), "c rank", len(c_shape), validator.EQ, self.name)
+        validator.check("ot rank", len(ot_shape), "c rank", len(c_shape), validator.EQ, self.name)
+        validator.check("tanhct rank", len(tanhct_shape), "c rank", len(c_shape), validator.EQ, self.name)
+        validator.check("dht shape", dht_shape, "c shape", c_shape, validator.EQ, self.name)
+        validator.check("dct shape", dct_shape, "c shape", c_shape, validator.EQ, self.name)
+        validator.check("it shape", it_shape, "c shape", c_shape, validator.EQ, self.name)
+        validator.check("jt shape", jt_shape, "c shape", c_shape, validator.EQ, self.name)
+        validator.check("ft shape", ft_shape, "c shape", c_shape, validator.EQ, self.name)
+        validator.check("ot shape", ot_shape, "c shape", c_shape, validator.EQ, self.name)
+        validator.check("tanhct shape", tanhct_shape, "c shape", c_shape, validator.EQ, self.name)
+
+        dgate_shape = (c_shape[0], 4 * c_shape[1])
+        dct_1_shape = c_shape
+
+        return (dgate_shape, dct_1_shape)
+
+    def infer_dtype(self, c_dtype, dht_dtype, dct_dtype, it_dtype, jt_dtype, ft_dtype, ot_dtype, tanhct_dtype):
+        validator.check_subclass("c", c_dtype, [mstype.tensor_type], self.name)
+        validator.check_subclass("dht", dht_dtype, [mstype.tensor_type], self.name)
+        validator.check_subclass("dct", dct_dtype, [mstype.tensor_type], self.name)
+        validator.check_subclass("it", it_dtype, [mstype.tensor_type], self.name)
+        validator.check_subclass("jt", jt_dtype, [mstype.tensor_type], self.name)
+        validator.check_subclass("ft", ft_dtype, [mstype.tensor_type], self.name)
+        validator.check_subclass("ot", ot_dtype, [mstype.tensor_type], self.name)
+        validator.check_subclass("tanhct", tanhct_dtype, [mstype.tensor_type], self.name)
+        validator.check_type_name("c", c_dtype, [mstype.float16, mstype.float32], self.name)
+        validator.check_type_name("dht", dht_dtype, [mstype.float16, mstype.float32], self.name)
+        validator.check_type_name("dct", dct_dtype, [mstype.float16, mstype.float32], self.name)
+        validator.check_type_name("it", it_dtype, [mstype.float16, mstype.float32], self.name)
+        validator.check_type_name("jt", jt_dtype, [mstype.float16, mstype.float32], self.name)
+        validator.check_type_name("ft", ft_dtype, [mstype.float16, mstype.float32], self.name)
+        validator.check_type_name("ot", ot_dtype, [mstype.float16, mstype.float32], self.name)
+        validator.check_type_name("tanhct", tanhct_dtype, [mstype.float16, mstype.float32], self.name)
+        return (c_dtype, c_dtype)
+
+
+class BasicLSTMCellWeightGrad(PrimitiveWithInfer):
+    """Computes the weight gradients of BasicLSTM."""
+
+    @prim_attr_register
+    def __init__(self):
+        pass
+
+    def infer_shape(self, x_shape, h_shape, dgate_shape):
+        validator.check_equal_int(len(x_shape), 2, "x rank", self.name)
+        validator.check("h rank", len(h_shape), " x rank", len(x_shape), validator.EQ, self.name)
+        validator.check("dgate rank", len(dgate_shape), "x rank", len(x_shape), validator.EQ, self.name)
+        validator.check("h_shape[0]", h_shape[0], "x_shape[0]", x_shape[0], validator.EQ, self.name)
+        validator.check("dgate_shape[0]", dgate_shape[0], "h_shape[0]", h_shape[0], validator.EQ, self.name)
+        validator.check("dgate_shape[1]", dgate_shape[1], "4*h_shape[1]", 4 * h_shape[1], validator.EQ, self.name)
+        input_size = x_shape[1]
+        hidden_size = h_shape[1]
+        dw_shape = (input_size + hidden_size, 4 * hidden_size)
+        db_shape = (4 * hidden_size,)
+        return (dw_shape, db_shape)
+
+    def infer_dtype(self, x_dtype, h_dtype, dgate_dtype):
+        validator.check_subclass("x", x_dtype, mstype.tensor_type, self.name)
+        validator.check_subclass("h", h_dtype, mstype.tensor_type, self.name)
+        validator.check_subclass("dgate", dgate_dtype, mstype.tensor_type, self.name)
+        validator.check_type_name("x", x_dtype, [mstype.float16, mstype.float32], self.name)
+        validator.check_type_name("h", h_dtype, [mstype.float16, mstype.float32], self.name)
+        validator.check_type_name("dgate", dgate_dtype, [mstype.float16, mstype.float32], self.name)
+        return (x_dtype, x_dtype)
+
+
+class BasicLSTMCellInputGrad(PrimitiveWithInfer):
+    """Computes the input gradients of BasicLSTM."""
+
+    @prim_attr_register
+    def __init__(self, keep_prob):
+        self.keep_prob = validator.check_value_type("keep_prob", keep_prob, [float], self.name)
+        self.keep_prob = validator.check_float_range(keep_prob, 0.0, 1.0, validator.INC_BOTH, "keep_prob", self.name)
+
+    def infer_shape(self, dgate_shape, w_shape):
+        validator.check_equal_int(len(dgate_shape), 2, "dgate rank", self.name)
+        validator.check_equal_int(len(w_shape), 2, "w rank", self.name)
+        validator.check("dgate_shape[1]", dgate_shape[1], "w_shape[1]", w_shape[1], validator.EQ, self.name)
+        batch_size = dgate_shape[0]
+        hidden_size = dgate_shape[1] // 4
+        input_size = w_shape[0] - hidden_size
+        dxt_shape = (batch_size, input_size)
+        dht_shape = (batch_size, hidden_size)
+        return (dxt_shape, dht_shape)
+
+    def infer_dtype(self, dgate_dtype, w_dtype):
+        validator.check_subclass("dgate", dgate_dtype, mstype.tensor_type, self.name)
+        validator.check_subclass("w", w_dtype, mstype.tensor_type, self.name)
+        validator.check_type_name("dgate", dgate_dtype, [mstype.float16, mstype.float32], self.name)
+        validator.check_type_name("w", w_dtype, [mstype.float16, mstype.float32], self.name)
+        return (dgate_dtype, dgate_dtype)
+
+
+class InvGrad(Primitive):
+    """Computes gradients for inv operation."""
+
+    @prim_attr_register
+    def __init__(self):
+        self.init_prim_io_names(inputs=['x', 'grad'], outputs=['y'])
+
+
+class LRNGrad(Primitive):
+    """Computes gradients for LRN operation."""
+
+    @prim_attr_register
+    def __init__(self, depth_radius=5, bias=1.0, alpha=1.0, beta=0.5):
+        self.init_prim_io_names(inputs=['grads', 'x', 'y'], outputs=['z'])
+        validator.check_value_type("depth_radius", depth_radius, [int], self.name)
+        validator.check_value_type("bias", bias, [float], self.name)
+        validator.check_value_type("alpha", alpha, [float], self.name)
+        validator.check_value_type("beta", beta, [float], self.name)
+
+
+class MvlgammaGrad(Primitive):
+    r"""
+    Computes gradients for Mvlgamma.
+
+    The following tex shows the mathematical calculation process of Mvlgamma:
+
+    .. math::
+
+        \log (\Gamma_{p}(a))=C+\sum_{i=1}^{p} \log (\Gamma(a-\frac{i-1}{2}))
+
+    where :math:`C = \log(\pi) \times \frac{p(p-1)}{4}` and :math:`\Gamma(\cdot)` is the Gamma function.
+
+    Args:
+        p(int): The number of dimensions. And the value of `p` must be greater than or equal to 1.
+
+    Inputs:
+        - **y_grad** (Tensor) - The input gradient.
+        - **x** (Tensor) - The input of Mvlgamma with data type of float32 or float64.
+
+    Outputs:
+        Tensor, has the same shape and type as `x`.
+
+    Raises:
+        TypeError: If dtype of `y_grad or `x` is neither float32 nor float64.
+        TypeError: If `p` is not an int.
+        ValueError: If p is not greater than or equal to 1.
+        ValueError: If all elements of `x` are not greater than (p-1)/2.
+
+    Supported Platforms:
+        ``Ascend`` ``CPU``
+    """
+
+    @prim_attr_register
+    def __init__(self, p):
+        self.init_prim_io_names(inputs=['y_grad', 'x'], outputs=['x_grad'])
+        self.p = validator.check_value_type('p', p, [int], self.name)
+
+
+class CdistGrad(Primitive):
+    """Computes gradient for Cdist."""
+
+    @prim_attr_register
+    def __init__(self, p=2.0):
+        validator.check_value_type("p", p, [float], self.name)
+        self.init_prim_io_names(inputs=['grad', 'input_x', 'input_y', 'cdist'], outputs=['output'])
+
+
+class PdistGrad(Primitive):
+    """Computes gradient for Pdist operation.
+
+    Args:
+        p (float): the p value for the Pdist formulation. Default: 2.0.
+
+    Inputs:
+        - **y_grad** (Tensor) - The gradients of loss to output of Pdist function.
+        - **x** (Tensor) - Input tensor of shape :math:`(N, M)`.
+        Must be the input `x` of the forward operator Pdist.
+        - **y** (Tensor) - Input tensor of shape :math:`(N*(N-1)/2)`.
+        Must be the output `y` of the forward operator Pdist.
+
+    Outputs:
+        Tensor, with the same shape and dtype as `x`.
+
+    Raises:
+        TypeError: If one of `y_grad`, `x` and `y` is not a Tensor.
+        TypeError: If dtype of `y_grad`, `x` and `y` are not all float16, float32 or float64.
+        TypeError: If `p` is not a float.
+        ValueError: If `p` is a negative float.
+        ValueError: If shape of `y_grad` is not same as `y`.
+        ValueError: If dimension of `x` is not 2.
+
+    Supported Platforms:
+        ``Ascend`` ``GPU`` ``CPU``
+    """
+
+    @prim_attr_register
+    def __init__(self, p=2.0):
+        validator.check_value_type("p", p, [float], self.name)
+        if p < 0:
+            raise ValueError('Pdist p must be a non-negative value, but got `{p}`.')
+        self.init_prim_io_names(inputs=['y_grad', 'x', 'y'], outputs=['x_grad'])
+
+
+class MultilabelMarginLossGrad(Primitive):
+    """
+    Compute the gradients of MultilabelMarginLoss operation.
+
+    Args:
+        reduction (str, optional): Apply specific reduction method to the output: ``'none'`` , ``'mean'`` ,
+            ``'sum'`` . Default: ``'mean'`` .
+
+            - ``'none'``: no reduction will be applied.
+            - ``'mean'``: compute and return the mean of elements in the output.
+            - ``'sum'``: the output elements will be summed.
+
+    Inputs:
+        - **y_grad** (Tensor) - The gradients of loss to output of MultilabelMarginLoss function, with
+          the same shape and data type as forward output `y`.
+        - **x** (Tensor) - Predict data. Tensor of shape :math:`(C)` or :math:`(N, C)`, where :math:`N`
+          is the batch size and :math:`C` is the number of classes. Data type must be float16 or float32.
+        - **target** (Tensor) - Ground truth data, with the same shape as `x`, data type must be int32 and
+          label targets padded by -1.
+        - **is_target** (Tensor) - Forward output tensor for backward input, with the same shape and
+          data type as `target`.
+
+    Outputs:
+        The shape of output :math:`(C)` or :math:`(N, C)`, with the same shape and data type as `x`.
+
+    Raises:
+        TypeError: If `x` or `target` or `y_grad` is not a Tensor.
+        TypeError: If dtype of `x` is neither float16 nor float32.
+        TypeError: If dtype of `target` is not int32.
+        TypeError: If dtype of `y_grad` is not the same as `x`.
+        ValueError: If length of shape of `x` is neither 1 nor 2.
+        ValueError: If shape of `x` is not the same as `target`.
+        ValueError: If `reduction` is not one of ``'none'``, ``'mean'``, ``'sum'``.
+        ValueError: If shape of `y_grad` is not the same as forward output `y`.
+
+    Supported Platforms:
+        ``Ascend``
+    """
+
+    @prim_attr_register
+    def __init__(self, reduction="mean"):
+        """Initialize MultilabelMarginLossGrad"""
+        self.reduction = validator.check_string(reduction, ['none', 'sum', 'mean'], 'reduction', self.name)
+        self.init_prim_io_names(inputs=['y_grad', 'x', 'target', 'is_target'], outputs=['x_grad'])
+
+
+class Dilation2DBackpropInput(Primitive):
+    """
+    Computes the gradient of morphological 2-D dilation with respect to the input.
+
+    .. warning::
+        This operator is an experimental operator, which has some accuracy problems for some inputs.
+
+    Args:
+        stride (Union[int, tuple[int]]): The distance of filter moving, an int number that represents
+            the height and width of movement are both strides, a tuple of two int numbers that
+            represent height and width of movement respectively, or a tuple of four int numbers which
+            should be :math:`(1, 1, H_{stride}, W_{stride})`.
+        dilation (Union[int, tuple[int]]): The input stride for atrous morphological dilation.The data
+            type is int or a tuple of 2 or 4 integers. Its value must be greater or equal to 1 and bounded
+            by the height and width of the input `x`.
+        pad_mode (str): Specifies padding mode. The optional values are "same", "valid".
+            Default: "same". Both upper and lower case are supported.
+        data_format (str): The format of input and output data. Only NCHW format is supported at present.
+            Default:'NCHW'
+
+    Inputs:
+        - **x** (Tensor) - Input data. A four dimension tensor with float16 or float32 data type. The shape must be
+          :math:`(N, C_{in}, H_{in}, W_{in})`.
+        - **filter** (Tensor) - A three dimension tensor with the same type as input. The shape must be
+          :math:`(C_{in}, H_{filter}, W_{filter})`.
+        - **out_backprop** (Tensor) - The gradients with respect to the output of the convolution.
+          A four dimension tensor with float16 or float32 data type. The shape must be
+          :math:`(N, C_{in}, H_{out}, W_{out})`.
+
+    outputs:
+        Tensor, the gradients with respect to the input of convolution. It has the same shape and type as the input `x`.
+
+    Raises:
+        TypeError: If type of `x` or `filter` is not the tpye in [uint8, uint16, uint32, uint64, int8, int16,
+                                  int32, int64, float16, float32, float64].
+        TypeError: If type of `out_backprop` is not the tpye in [uint8, uint16, uint32, uint64, int8, int16,
+                                  int32, int64, float16, float32, float64].
+        TypeError: If `stride` or `dilation` is not an int number or a tuple of two or four int numbers.
+        ValueError: If the length of `stride` or `dilation` is neither two nor four when they are tuples.
+        ValueError: If `stride` or `dilation` is not (1, 1, height, width) when it is a tuple of four int numbers.
+        ValueError: If `stride` is not in the range of [1, 255].
+        ValueError: If `dilation` is less than 1.
+        ValueError: If `pad_mode` is not a str of 'same', 'valid', 'SAME' or 'VALID'.
+        ValueError: If `data_format` is not the str of 'NCHW'.
+
+    Supported Platforms:
+        ``Ascend`` ``GPU`` ``CPU``
+
+    Examples:
+        (pad_mode="SAME", data_format="NCHW")
+        >>> out_backprop = Tensor(np.ones([1, 3, 4, 4]), mstype.float32)
+        >>> filter = Tensor(np.ones([3 , 2 , 2]), mstype.float32)
+        >>> x = Tensor(np.ones([1, 3, 4, 4]), mstype.float32)
+        >>> dilation_backprop_input = G.Dilation2DBackpropInput(stride=1, dilation=1)
+        >>> output = dilation_backprop_input(x, filter, out_backprop)
+        >>> print(output)
+        [[[[1. 1. 1. 1.]
+           [1. 1. 1. 1.]
+           [1. 1. 1. 1.]
+           [1. 1. 1. 1.]]
+          [[1. 1. 1. 1.]
+           [1. 1. 1. 1.]
+           [1. 1. 1. 1.]
+           [1. 1. 1. 1.]]
+          [[1. 1. 1. 1.]
+           [1. 1. 1. 1.]
+           [1. 1. 1. 1.]
+           [1. 1. 1. 1.]]]]
+    """
+
+    @prim_attr_register
+    def __init__(self, stride, dilation, pad_mode="SAME", data_format="NCHW"):
+        """Initialize Dilation2DBackpropInput"""
+
+        def _check_format_stride_or_dilation(arg_name, arg_value, prim_name, data_format):
+            validator.check_value_type(arg_name, arg_value, (int, tuple), prim_name)
+            if isinstance(arg_value, int):
+                ret_value = (1, arg_value, arg_value, 1) if data_format == "NHWC" else (1, 1, arg_value, arg_value)
+            elif len(arg_value) == 2:
+                ret_value = (1, arg_value[0], arg_value[1], 1) if data_format == "NHWC" else \
+                    (1, 1, arg_value[0], arg_value[1])
+            elif len(arg_value) == 4:
+                if data_format == "NHWC" and (arg_value[0] != 1 or arg_value[3] != 1):
+                    raise ValueError(f"For '{prim_name}' attr '{arg_name}' should be "
+                                     f"[1, {arg_name}_height, {arg_name}_weigth, 1] when data_format is 'NHWC', "
+                                     f"but got {arg_value}")
+                if data_format == "NCHW" and (arg_value[0] != 1 or arg_value[1] != 1):
+                    raise ValueError(
+                        f"For '{prim_name}' attr '{arg_name}' should be [1, 1, {arg_name}_height, {arg_name}_weigth]"
+                        f"when data_format is 'NCHW', but got {arg_value}")
+                ret_value = arg_value
+            else:
+                raise ValueError(
+                    f"For '{prim_name}' attr '{arg_name}' should be an positive int number or a tuple of two "
+                    f"or four positive int numbers, but got {arg_value}")
+            for item in ret_value:
+                if isinstance(item, int) and not isinstance(item, bool) and item > 0:
+                    continue
+                raise ValueError(
+                    f"For '{prim_name}' attr '{arg_name}' should be an positive int number or a tuple of two "
+                    f"or four positive int numbers, but got {arg_value}")
+            return ret_value
+
+        if data_format == 'NHWC':
+            raise ValueError(f"For '{self.name}', NHWC format is not supported at present.")
+        self.data_format = validator.check_string(self.data_format, ['NCHW', 'NHWC'], 'data_format', self.name)
+        self.add_prim_attr("data_format", self.data_format)
+        self.pad_mode = validator.check_string(self.pad_mode, ["SAME", "VALID", 'same', "valid"], "pad_mode", self.name)
+        self.add_prim_attr("pad_mode", self.pad_mode.upper())
+        self.stride = _check_format_stride_or_dilation("stride", stride, self.name, self.data_format)
+        self.add_prim_attr("stride", self.stride)
+        self.dilation = _check_format_stride_or_dilation("dilation", dilation, self.name, self.data_format)
+        self.add_prim_attr("dilation", self.dilation)
+
+
+class Dilation2DBackpropFilter(Primitive):
+    """
+    Computes the gradient of morphological 2-D dilation with respect to the filter.
+
+    .. warning::
+        This operator is an experimental operator, which has some accuracy problems for some inputs.
+
+    Args:
+        stride (Union[int, tuple[int]]): The distance of kernel moving, an int number that represents
+            the height and width of movement are both strides, a tuple of two int numbers that
+            represent height and width of movement respectively, or a tuple of four int numbers which
+            should be :math:`(1, 1, H_{stride}, W_{stride})`.
+        dilation (Union(int, tuple[int])): The data type is int or a tuple of 2 integers or a tuple of 4 integers.
+            Specifies the dilation rate to use for dilated convolution.
+            If set to be :math:`k > 1`, there will be :math:`k - 1` pixels skipped for each sampling location.
+            Its value must be greater or equal to 1 and bounded by the height and width of the input `x`.
+        pad_mode (str): Specifies padding mode. The optional values are "same", "valid".
+            Default: "same". Both upper and lower case are supported.
+        data_format (str): The format of input and output data. Only NCHW format is supported at present.
+            Default:'NCHW'
+
+    Inputs:
+        - **x** (Tensor) - Input data. A four dimension tensor with float16 or float32 data type. The shape must be
+          :math:`(N, C_{in}, H_{in}, W_{in})`.
+        - **filter** (Tensor) - A three dimension tensor with the same type as input. The shape must be
+          :math:`(C_{in}, H_{filter}, W_{filter})`.
+        - **out_backprop** (Tensor) - The gradients with respect to the output of the convolution.
+          A four dimension tensor with float16 or float32 data type. The shape must be
+          :math:`(N, C_{in}, H_{out}, W_{out})`.
+
+    outputs:
+        Tensor, the gradients with respect to the input of convolution. It has the same shape and type as the input `x`.
+
+    Raises:
+        TypeError: If type of `x` or `filter` is not the tpye in [uint8, uint16, uint32, uint64, int8, int16,
+                                  int32, int64, float16, float32, float64].
+        TypeError: If type of `out_backprop` is not the tpye in [uint8, uint16, uint32, uint64, int8, int16,
+                                  int32, int64, float16, float32, float64].
+        TypeError: If `stride` or `dilation` is not an int number or a tuple of two or four int numbers.
+        ValueError: If the length of `stride` or `dilation` is neither two nor four when they are tuples.
+        ValueError: If `stride` or `dilation` is not (1, 1, height, width) when it is a tuple of four int numbers.
+        ValueError: If `stride` is not in the range of [1, 255].
+        ValueError: If `dilation` is less than 1.
+        ValueError: If `pad_mode` is not a str of 'same', 'valid', 'SAME' or 'VALID'.
+        ValueError: If `data_format` is not the str of 'NCHW'.
+
+
+    Supported Platforms:
+        ``Ascend`` ``GPU`` ``CPU``
+
+    Examples:
+        (pad_mode="SAME", data_format="NCHW")
+        >>> x = Tensor(np.ones([2, 3, 4, 4]), mstype.float32)
+        >>> filter = Tensor(np.ones([3,2,2]), mstype.float32)
+        >>> out_backprop = Tensor(np.ones([2,3,2,2]), mstype.float32)
+        >>> dilation_backprop_filter = G.Dilation2DBackpropFilter(stride=2, dilation=1)
+        >>> output = dilation_backprop_filter(x, filter, out_backprop)
+        >>> print(output)
+        [[[8. 8. 8.]
+          [0. 0. 0.]]
+         [[0. 0. 0.]
+          [0. 0. 0.]]]
+    """
+
+    @prim_attr_register
+    def __init__(self, stride, dilation, pad_mode="SAME", data_format="NCHW"):
+        """Initialize Dilation2DBackpropFilter"""
+
+        def _check_format_stride_or_dilation(arg_name, arg_value, prim_name, data_format):
+            validator.check_value_type(arg_name, arg_value, (int, tuple), prim_name)
+            if isinstance(arg_value, int):
+                ret_value = (1, arg_value, arg_value, 1) if data_format == "NHWC" else (1, 1, arg_value, arg_value)
+            elif len(arg_value) == 2:
+                ret_value = (1, arg_value[0], arg_value[1], 1) if data_format == "NHWC" else \
+                    (1, 1, arg_value[0], arg_value[1])
+            elif len(arg_value) == 4:
+                if data_format == "NHWC" and (arg_value[0] != 1 or arg_value[3] != 1):
+                    raise ValueError(
+                        f"For '{prim_name}' attr '{arg_name}' should be [1, {arg_name}_height, {arg_name}_weigth, 1]"
+                        f"when data_format is 'NHWC', but got {arg_value}")
+                if data_format == "NCHW" and (arg_value[0] != 1 or arg_value[1] != 1):
+                    raise ValueError(
+                        f"For '{prim_name}' attr '{arg_name}' should be [1, 1, {arg_name}_height, {arg_name}_weigth]"
+                        f"when data_format is 'NCHW', but got {arg_value}")
+                ret_value = arg_value
+            else:
+                raise ValueError(
+                    f"For '{prim_name}' attr '{arg_name}' should be an positive int number or a tuple of two "
+                    f"or four positive int numbers, but got {arg_value}")
+            for item in ret_value:
+                if isinstance(item, int) and not isinstance(item, bool) and item > 0:
+                    continue
+                raise ValueError(
+                    f"For '{prim_name}' attr '{arg_name}' should be an positive int number or a tuple of two "
+                    f"or four positive int numbers, but got {arg_value}")
+            return ret_value
+
+        if data_format == 'NHWC':
+            raise ValueError(f"For '{self.name}', NHWC format is not supported at present.")
+        self.data_format = validator.check_string(self.data_format, ['NCHW', 'NHWC'], 'data_format', self.name)
+        self.add_prim_attr("data_format", self.data_format)
+        self.pad_mode = validator.check_string(self.pad_mode, ["SAME", "VALID", 'same', "valid"], "pad_mode", self.name)
+        self.add_prim_attr("pad_mode", self.pad_mode.upper())
+        self.stride = _check_format_stride_or_dilation("stride", stride, self.name, self.data_format)
+        def is_in_range(x):
+            return 1 <= x <= 255
+        if not is_in_range(self.stride[2]) or not is_in_range(self.stride[3]):
+            raise ValueError(f"For '{self.name}', size of stride is not supported, "
+                             f'stride should be in the range of [1, 255], '
+                             f'but got stride_h: `{self.stride[2]}`, stride_w: `{self.stride[3]}`.')
+        self.add_prim_attr("stride", self.stride)
+        self.dilation = _check_format_stride_or_dilation("dilation", dilation, self.name, self.data_format)
+        self.add_prim_attr("dilation", self.dilation)
+
+
+class ParallelResizeBilinearGrad(PrimitiveWithInfer):
+    """ParallelResizeBilinearGrad ops"""
+
+    @prim_attr_register
+    def __init__(self, ori_image_size, src_start_w, dst_start_w, align_corners):
+        """Initialize ParallelResizeBilinearGrad."""
+        self.init_prim_io_names(inputs=["grad", "x", "size"], outputs=['y'])
+        validator.check_value_type("ori_image_size", ori_image_size, [tuple, list], self.name)
+        validator.check_value_type("src_start_w", src_start_w, [int], self.name)
+        validator.check_value_type("dst_start_w", dst_start_w, [int], self.name)
+        validator.check_value_type("align_corners", align_corners, [bool], self.name)
+        self.ori_image_size = list(ori_image_size)
+        self.src_start_w = src_start_w
+        self.dst_start_w = dst_start_w
+        self.align_corners = align_corners
+        self.half_pixel_centers = False
+        self.add_prim_attr('ori_image_size', self.ori_image_size)
+        self.add_prim_attr('src_start_w', self.src_start_w)
+        self.add_prim_attr('dst_start_w', self.dst_start_w)
+        self.add_prim_attr('align_corners', self.align_corners)
+        self.add_prim_attr('half_pixel_centers', self.half_pixel_centers)
+
+    def __infer__(self, grad, x, size):
+        size_val = size['value']
+        grad_shape = grad['shape']
+        grad_dtype = grad['dtype']
+        x_shape = x['shape']
+        x_dtype = x['dtype']
+        validator.check_tensor_dtype_valid("grad_dtype", grad_dtype, [mstype.float16, mstype.float32], self.name)
+        validator.check_tensor_dtype_valid("x_dtype", x_dtype, [mstype.float16, mstype.float32], self.name)
+        if size_val is None:
+            raise ValueError("size must be const input")
+        output_shape = [grad_shape[0], grad_shape[1], x_shape[2], x_shape[3]]
+
+        return {'shape': output_shape,
+                'dtype': x_dtype,
+                'value': None}
+
+
+class MultiMarginLossGrad(Primitive):
+    """
+    Compute the gradients of MultiMarginLoss operation
+
+    Args:
+        p (int): Optional. The norm degree for pairwise distance.Should be 1 or 2. Default: 1.
+        margin (float): Optional. A parameter to change pairwise distance. Default: 1.0.
+        reduction (str, optional): Apply specific reduction method to the output: ``'none'`` , ``'mean'`` ,
+            ``'sum'`` . Default: ``'mean'`` .
+
+            - ``'none'``: no reduction will be applied.
+            - ``'mean'``: compute and return the weighted mean of elements in the output.
+            - ``'sum'``: the output elements will be summed.
+
+    Inputs:
+        - **y_grad** (Tensor) - If it's not a scalar, the shape of 'y_grad' :math:`(N, C)`.
+          Data type only support float32 or float16,float64.
+        - **x** (Tensor) - Input x, with shape :math:`(N, C)`. Data type only support float32, float16 or float64.
+        - **target** (Tensor) - Ground truth labels, with shape :math:`(N,)`. Data type only support int64. The
+          value of target should be non-negative, less than C.
+        - **weight** (Tensor, optional) - The rescaling weight to each class with shape :math:`(C,)`. Data type only
+          support float32, float16 or float64. Default: ``None``.
+
+    Outputs:
+        The shape of output :math:`(N, C)`. Data type only support float32 or float16, float64.
+        Has the same data type with 'x'.
+
+    Raises:
+        TypeError: If dtype of `p` and `target` is not int.
+        TypeError: If dtype of `margin` is not float.
+        TypeError: If dtype of `reduction` is not str.
+        TypeError: If dtype of `x` is not float16, float or float64.
+        TypeError: If dtype of `weight` and `x` is not the same.
+        ValueError: If 'p' is not 1 or 2.
+        ValueError: If 'reduction' is not one of {'none','sum','mean'}.
+        ValueError: If shape[0] of `x` is not equal to shape[0] of `target`.
+        ValueError: If shape[1] of `x` is not equal to shape[0] of `weight`.
+        ValueError: IF rank of `weight` is not 1.
+        ValueError: If rank of `x` is not 2 or rank of 'target' is not 1.
+
+    Supported Platforms:
+        ``Ascend``  ``CPU``
+    """
+    __mindspore_signature__ = (
+        sig.make_sig('y_grad'),
+        sig.make_sig('x'),
+        sig.make_sig('target'),
+        sig.make_sig('weight', default=None)
+    )
+
+    @prim_attr_register
+    def __init__(self, p=1, margin=1.0, reduction="mean"):
+        """Initialize MultiMarginLossGrad"""
+        self.p = validator.check_value_type('p', p, [int], self.name)
+        validator.check_int(p, {1, 2}, validator.IN, 'p', self.name)
+        self.margin = validator.check_value_type('margin', margin, [float], self.name)
+        self.reduction = validator.check_string(reduction, ['none', 'sum', 'mean'], 'reduction', self.name)
+        self.init_prim_io_names(inputs=['y_grad', 'x', 'target', 'weight'], outputs=['x_grad'])
+
+    def __call__(self, y_grad, x, target, weight=None):
+        return super().__call__(y_grad, x, target, weight)
+
+
+class SparseSegmentMeanGrad(Primitive):
+    """
+    Compute gradients for SparseSegmentMeanGrad operation.
+
+    Inputs:
+        - **x** (Tensor) - A Tensor of the first input of SparseSegmentMeanGrad.
+        - **indices** (Tensor) - Indices is a 1-D tensor with indices into `x`. Must be one of the following
+          types: int32, int64. Has same rank as `segment_ids`. The shape should be :math:`(N,)`.
+        - **segment_ids** (Tensor) - Segment_ids is a 1-D tensor with indices into the output `y`. Must be one of the
+          following types: int32, int64. Values should be sorted and can be repeated. The shape should be :math:`(N,)`.
+        - **output_dim0** (Tensor) - Output_dim0 is a 0-D tensor. Dimension 0 of `x` passed to SparseSegmentMean op.
+
+    Outputs:
+        A Tensor. Has the same type as `x` .
+        Has same shape as `x`, except for dimension 0 which is the value of `output_dim0`.
+
+    Raises:
+        TypeError: If `x` or `indices` or `segment_ids` is not a tensor.
+        TypeError: If the dtype of `x` is not any of the following data types: {float32, float64}.
+        TypeError: If the dtype of `indices` is not int32.
+        TypeError: If the dtype of `segment_ids` is not int32.
+        TypeError: If the dtype of `output_dim0` is not int32.
+        ValueError: If dimension size of `x` is less than 1.
+        ValueError: If rank of `indices` or `segment_ids` is not 1.
+        ValueError: If dimension size of `output_dim0` is not 0.
+        ValueError: If the first dimension of `indices` is not equal to the first dimension of `segment_ids`.
+        ValueError: If `segment_ids` is not sorted.
+        ValueError: If `indices` is out of range of `output_dim0`.
+
+    Supported Platforms:
+        ``Ascend`` ``GPU`` ``CPU``
+    """
+
+    @prim_attr_register
+    def __init__(self):
+        """Initialize SparseSegmentMeanGrad"""
+        self.init_prim_io_names(inputs=['x', 'indices', 'segment_ids', 'output_dim0'], outputs=['y'])
+
+
+class FractionalMaxPoolGrad(Primitive):
+    """Computes gradients for FractionalMaxPool operation."""
+
+    @prim_attr_register
+    def __init__(self, overlapping=False):
+        self.init_prim_io_names(inputs=["orig_input", "orig_output", "out_backprop",
+                                        "row_pooling_sequence", "col_pooling_sequence"],
+                                outputs=["y"])
+        validator.check_value_type("overlapping", overlapping, [bool], self.name)
+
+
+class FractionalMaxPool3DGradWithFixedKsize(Primitive):
+    """Computes gradients for FractionalMaxPool3DWithFixedKsize operation."""
+
+    @prim_attr_register
+    def __init__(self, data_format="NCDHW"):
+        self.init_prim_io_names(inputs=["origin_input", "out_backprop", "argmax"], outputs=["y"])
+        self.data_format = validator.check_string(data_format, ['NCDHW', "NDHWC"], 'data_format', self.name)
+
+
+class MaxUnpool2DGrad(Primitive):
+    r"""
+    Gradients for MaxUnpool2D operation.
+    """
+
+    @prim_attr_register
+    def __init__(self, ksize, strides=0, pads=0, output_shape=(), data_format="NCHW"):
+        """Initialize MaxUnpool2DGrad."""
+        self.init_prim_io_names(inputs=['x', 'grads', 'argmax'], outputs=['y'])
+        validator.check_value_type("ksize", ksize, [int, tuple], self.name)
+        validator.check_value_type("strides", strides, [int, tuple], self.name)
+        validator.check_value_type("pads", pads, [int, tuple], self.name)
+        validator.check_value_type("output_shape", output_shape, [tuple], self.name)
+        validator.check_string(data_format, ['NCHW', 'NHWC'], 'data_format', self.name)
+        validator.check_int(len(ksize), 4, validator.EQ, "ksize rank", self.name)
+        validator.check_int(len(strides), 4, validator.EQ, "strides rank", self.name)
+        validator.check_int(len(pads), 4, validator.EQ, "pads rank", self.name)
+
+
+class MaxUnpool3DGrad(Primitive):
+    r"""
+    Gradients for MaxUnpool3D operation.
+    """
+
+    @prim_attr_register
+    def __init__(self, ksize, strides=0, pads=0, output_shape=(), data_format="NCDHW"):
+        """Initialize MaxUnpool3DGrad."""
+        self.init_prim_io_names(inputs=['x', 'grads', 'argmax'], outputs=['y'])
+        validator.check_value_type("ksize", ksize, [int, tuple], self.name)
+        validator.check_value_type("strides", strides, [int, tuple], self.name)
+        validator.check_value_type("pads", pads, [int, tuple], self.name)
+        validator.check_value_type("output_shape", output_shape, [tuple], self.name)
+        validator.check_string(data_format, ['NCDHW', 'NDHWC'], 'data_format', self.name)
+        validator.check_int(len(ksize), 5, validator.EQ, "ksize rank", self.name)
+        validator.check_int(len(strides), 5, validator.EQ, "strides rank", self.name)
+        validator.check_int(len(pads), 5, validator.EQ, "pads rank", self.name)
+
+
+class FractionalAvgPoolGrad(Primitive):
+    """Computes gradients for FractionalAvgPool operation."""
+
+    @prim_attr_register
+    def __init__(self, overlapping=False):
+        self.add_prim_attr("max_length", 1000000)
+        self.init_prim_io_names(inputs=["orig_input_tensor_shape", "out_backprop", "row_pooling_sequence",
+                                        "col_pooling_sequence"],
+                                outputs=["y"])
+        validator.check_value_type("overlapping", overlapping, [bool], self.name)
+
+
+class PSROIPoolingGrad(Primitive):
+    """Computes gradients for PSROIPooling operation."""
+
+    @prim_attr_register
+    def __init__(self, input_size, spatial_scale, group_size, output_dim):
+        """Initialize PSROIPoolingGrad."""
+        self.init_prim_io_names(inputs=["x", "rois"], outputs=['y'])
+        validator.check_value_type("input_size", input_size, [int, tuple], self.name)
+        validator.check_positive_float(spatial_scale, "spatial_scale", self.name)
+        validator.check_positive_int(group_size, "group_size", self.name)
+        validator.check_positive_int(output_dim, "output_dim", self.name)
+
+        if isinstance(input_size, int):
+            self.input_size = [input_size, input_size]
+        else:
+            self.input_size = list(input_size)
+
+        validator.check_positive_int_sequence(self.input_size, "input_size", self.name)
+        self.spatial_scale = spatial_scale
+        self.group_size = group_size
+        self.output_dim = output_dim
+
+        self.add_prim_attr('input_size', self.input_size)
+        self.add_prim_attr('spatial_scale', self.spatial_scale)
+        self.add_prim_attr('group_size', self.group_size)
+        self.add_prim_attr('output_dim', self.output_dim)
+
+
+class AdaptiveMaxPool3DGrad(Primitive):
+    """Computes gradients for AdaptiveMaxPool3D operation."""
+
+    @prim_attr_register
+    def __init__(self):
+        """Initialize AdaptiveMaxPool3DGrad"""
+        self.init_prim_io_names(inputs=['input_grad', 'x', 'argmax'], outputs=['output_grad'])
+
+
+class TraceGrad(Primitive):
+    """
+    Computes grad for Trace operation.
+
+    Inputs:
+        - **y_grad** (Tensor) - the grad of trace to output of Trace function.
+          Currently grad data type support float16, float32, int8, int16, int32, int64,
+          uint8, uint16, uint32, uint64, float64.
+        - **x_shape** (Tensor) - the shape of trace to output of Trace function.
+          Currently shape data type support int32, int64.
+
+    Outputs:
+        x_grad - Tensor, with the same data type as 'y_grad' and shape is x_shape.
+
+    Raises:
+        TypeError: If `x_shape` is not a Tensor.
+        TypeError: If the dtype of `x_shape` is neither int32 nor int64.
+        ValueError: If `x_shape` is not a 1D Tensor.
+        ValueError: If length of shape of `x_shape` is not equal to 2.
+
+    Support Platforms:
+        ``Ascend`` ``GPU`` ``CPU``
+    """
+
+    @prim_attr_register
+    def __init__(self):
+        self.init_prim_io_names(inputs=['y_grad', 'x_shape'], outputs=['x_grad'])
+
+
+class IgammaGradA(Primitive):
+    r"""
+    Computes the gradient of igamma(a, x) wrt a.
+
+    Inputs:
+        - **a** (Tensor) - The input tensor. With float32 or float64 data type.
+        - **x** (Tensor) - The input tensor. With float32 data or float64 type. `x` should have
+          the same dtype with `a`.
+
+    Outputs:
+        Tensor, has the same dtype as `a` and `x`.
+
+    Raises:
+        TypeError: If a or grad is not a Tensor.
+        TypeError: If dtype of input x and a is not float32 nor float64.
+        TypeError: If x has different dtype with a.
+        ValueError: If `a` could not be broadcast to a tensor with shape of `x`.
+
+    Supported Platforms:
+        ``Ascend`` ``GPU`` ``CPU``
+
+    Examples:
+        >>> a = Tensor(np.array([2.0, 4.0, 6.0, 8.0]).astype(np.float32))
+        >>> x = Tensor(np.array([2.0, 3.0, 4.0, 5.0]).astype(np.float32))
+        >>> igammagrada = G.IgammaGradA()
+        >>> output = igammagrada(a, x)
+        >>> print (output)
+        [-0.2940046  -0.20153049 -0.13028376 -0.08352186]
+    """
+
+    @prim_attr_register
+    def __init__(self):
+        """Initialize IgammaGradA"""
+        self.init_prim_io_names(inputs=['a', 'x'], outputs=['z'])
+
+
+class DeformableOffsetsGrad(Primitive):
+    r"""
+    Computes gradients of DeformableOffsets operation.
+    Args:
+        strides (tuple[int, int ,int ,int]): A tuple of 4 integers. The stride of sliding windows for height
+            and width for H/W dimension.
+        pads (tuple[int, int ,int ,int]): A tuple of 4 integers.Padding added to H/W dimension of the input.The number
+            of pixels to add to each (top, bottom, left,right) side of the input
+        kernel_size (tuple[int, int]): Kernel size, a tuple of 2 integers.
+        dilations (tuple[int, int, int, int]): A tuple of 4 integers. The dilation factor for each dimension of
+            input. Default:(1, 1, 1, 1)
+        data_format (str): An optional string from:"NCHW", "NHWC".Specify the data format of the input x. Default:
+            "NCHW".
+        deformable_groups (int): Specify the C-axis grouping number of input x. Default: 1.
+        modulated (bool): Specify version of DeformableOffsetsGrad, true means v2, false means v1. Default: ``True``.
+
+    Inputs:
+        - **grad** (Tensor) - The input grad tensor. With float16 or float32 data type.
+        - **x** (Tensor) - The input `x` of DeformableOffsets with data type of float16 or float32.
+        - **offsets** (Tensor) - The input 'offsets' of DeformableOffsets with data type of float16 or float32.
+
+    Outputs:
+        - **grad_x** (Tensor) - The output grad of input `x`. With same dtype and shape of input `x`.
+        - ""grad_offsets** (Tensor) - The output grad of input `offsets`. With same dtype and shape of input `offsets`.
+
+    Supported Platforms:
+        ``Ascend````GPU````CPU``
+    """
+
+    @prim_attr_register
+    def __init__(self,
+                 strides,
+                 pads,
+                 kernel_size,
+                 dilations=(1, 1, 1, 1),
+                 data_format="NCHW",
+                 deformable_groups=1,
+                 modulated=True):
+        """Initialize DeformableOffsetsGrad"""
+        self.init_prim_io_names(inputs=['out_backprop', 'input', 'offsets'], outputs=['out_grad'])
+
+        self.strides = _check_positive_int_or_tuple('strides', strides, self.name, allow_four=True, ret_four=True)
+        self.add_prim_attr('strides', self.strides)
+
+        self.pads = pads
+        self.add_prim_attr('pads', self.pads)
+
+        self.kernel_size = _check_positive_int_or_tuple('kernel_size', kernel_size, self.name, allow_four=True,
+                                                        ret_four=False)
+        self.add_prim_attr('ksize', self.kernel_size)
+
+        self.dilations = _check_positive_int_or_tuple('dilations', dilations, self.name, allow_four=True, ret_four=True)
+        self.add_prim_attr('dilations', dilations)
+
+        self.data_format = validator.check_string(data_format, ['NCHW', 'NHWC'], 'format', self.name)
+        self.add_prim_attr('data_format', self.data_format)
+
+        self.deformable_groups = validator.check_positive_int(deformable_groups, 'deformable_groups', self.name)
+        self.add_prim_attr('deformable_groups', self.deformable_groups)
+
+        self.modulated = validator.check_bool(modulated, 'modulated', self.name)
+        self.add_prim_attr('modulated', self.modulated)
+
+
+class MedianGrad(Primitive):
+    """
+    Computes gradient for Median operation.
+
+    .. warning::
+        When attr `global_median` is True, the value of Median's second output Tensor `indices` value is meaningless.
+
+    Args:
+        global_median (bool): Whether the output tensor is the global median of all input tensor elements
+            or not in Median operation.
+        axis (int): The dimension need to reduce in Median operation.
+        keep_dims (bool): Whether the output tensor need to retain `axis` dimension or not in Median operation.
+
+    Inputs:
+        - **y_grad** (Tensor) - The gradients of loss to output of Median function.
+        - **x** (Tensor) - The first input is a tensor whose data type is number.
+          The dtype is one of the following: int16, int32, int64, float32, double.
+        - **y** (Tensor) - The first output of Median function, which datatype is same as `x`.
+        - **indices** (Tensor) - The second output of Median function, which datatype is int64.
+
+    Outputs:
+        x_grad - Tensor, has the same shape as the `x`, dtype is double only when dtype of `x` is double.
+        Otherwise, dtype of `x_grad` is float32.
+
+    Raises:
+        TypeError: If dtype of `y_grad` is not the same as `x`.
+        ValueError: If shape of `y_grad` is not the same as `y`.
+
+    Supported Platforms:
+        ``Ascend`` ``CPU``
+    """
+
+    @prim_attr_register
+    def __init__(self, global_median=False, axis=0, keep_dims=False):
+        validator.check_value_type("global_median", global_median, [bool], self.name)
+        self.global_median = global_median
+        if global_median is False:
+            validator.check_value_type("axis", axis, [int], self.name)
+            validator.check_value_type("keep_dims", keep_dims, [bool], self.name)
+        self.init_prim_io_names(inputs=['y_grad', 'x', 'y', 'indices'], outputs=['x_grad'])
+
+
+class SparseSegmentSumGrad(Primitive):
+    """
+    Computes gradients for SparseSegmentSumGrad operation.
+
+    Inputs:
+        - **grad** (Tensor) - A tensor.
+        - **indices** (Tensor) - Indices is a 1-D tensor. Must be one of the following types: int32, int64.
+          Has same rank as segment_ids. The shape should be :math:`(N,)`.
+        - **segment_ids** (Tensor) - Segment_ids is a 1-D tensor. Must be one of the following types: int32, int64.
+          Values should be sorted and can be repeated. The shape should be :math:`(N,)`.
+        - **output_dim0** (Tensor) - Output_dim0 is a 0-D tensor. Dimension 0 of `x` passed to SparseSegmentSum op.
+
+    Outputs:
+        A Tensor. Has the same type as `grad` .
+        Has same shape as `grad`, except for dimension 0 which is the value of `output_dim0`.
+
+    Raises:
+        TypeError: If `grad` or `indices` or `segment_ids` or `output_dim0` is not a tensor.
+        TypeError: If the dtype of `grad` is not any of the following data types: {float16, float32, float64}.
+        TypeError: If the dtype of `indices` and `segment_ids` and `output_dim0` is not int32 or int64.
+        ValueError: If dimension size of `grad` less than 1.
+        ValueError: If rank of `indices` or `segment_ids` is not 1.
+        ValueError: If dimension size of `output_dim0` is not 0.
+        ValueError: If shape[0] of `indices` is not corresponding to shape[0] of `segment_ids`.
+        ValueError: If `segment_ids` is not sorted.
+        ValueError: If the last number of `segment_ids` is out of range of grad's first shape.
+        ValueError: If `indices` is bigger than or equal to `output_dim0`.
+
+    Supported Platforms:
+        ``GPU``
+    """
+    __mindspore_signature__ = (
+        sig.make_sig('grad', dtype=sig.sig_dtype.T1),
+        sig.make_sig('indices', dtype=sig.sig_dtype.T),
+        sig.make_sig('segment_ids', dtype=sig.sig_dtype.T),
+        sig.make_sig('output_dim0', dtype=sig.sig_dtype.T)
+    )
+
+    @prim_attr_register
+    def __init__(self):
+        """Initialize SparseSegmentSumGrad"""
+        self.init_prim_io_names(inputs=['grad', 'indices', 'segment_ids', 'output_dim0'], outputs=['y'])
+
+
+class SparseSegmentSqrtNGrad(Primitive):
+    """
+    Computes gradients for SparseSegmentSqrtNGrad operation.
+
+    Inputs:
+        - **x** (Tensor) - A tensor. It's rank must be more than or equal to one.
+        - **indices** (Tensor) - Indices is a 1-D tensor with indices into `x`. Must be one of the following
+          types: int32, int64. Has same rank as segment_ids. The shape should be :math:`(N,)`.
+        - **segment_ids** (Tensor) - Segment_ids is a 1-D tensor with indices into the output `y`. Must be one
+          of the following types: int32, int64. Values should be sorted and can be repeated. The shape should
+          be :math:`(N,)`.
+        - **output_dim0** (Tensor) - Output_dim0 is a 0-D tensor. Dimension 0 of `x` passed to SparseSegmentSqrtN op.
+
+    Outputs:
+        A Tensor. Has the same type as `x` .
+        Has same shape as `x`, except for dimension 0 which is the value of `output_dim0`.
+
+    Raises:
+        TypeError: If `x` or `indices` or `segment_ids` or `output_dim0` is not a tensor.
+        TypeError: If the dtype of `x` is not any of the following data types: {float16, float32, float64}.
+        TypeError: If the dtype of `indices` is not int32.
+        TypeError: If the dtype of `segment_ids` is not int32.
+        TypeError: If the dtype of `output_dim0` is not int32.
+        ValueError: If dimension size of `x` is less than 1.
+        ValueError: If rank of `indices` or `segment_ids` is not 1.
+        ValueError: If dimension size of `output_dim0` is not 0.
+        ValueError: If shape[0] of `indices` is not corresponding to shape[0] of `segment_ids`.
+        ValueError: If `segment_ids` is not sorted.
+        ValueError: If the last number of `segment_ids` is out of range of x's first shape.
+        ValueError: If `indices` is bigger than or equal to `output_dim0`.
+
+    Supported Platforms:
+        ``Ascend`` ``GPU`` ``CPU``
+    """
+
+    @prim_attr_register
+    def __init__(self):
+        """Initialize SparseSegmentSqrtNGrad"""
+        self.init_prim_io_names(inputs=['x', 'indices', 'segment_ids', 'output_dim0'], outputs=['y'])
+
+
+class SparseSliceGrad(Primitive):
+    r"""
+    Computes gradients for SparseSlice operation.
+
+    Inputs:
+        - **backprop_val_grad** (Tensor) - A 1D Tensor.
+          The shape should be :math:`(N,)`.
+        - **indices** (Tensor) - A 2D Tensor (N x R matrix) of type int64. The indices of the SparseTensor.
+          Support int64, each element value should be a non-negative int number. This tensor should be sorted.
+          The shape is :math:`(N, R)`.
+        - **start** (Tensor) - A 1D Tensor of type int64, represents the start of the indices.
+          The shape should be :math:`(R,)`.
+        - **new_indices** (Tensor) - A 2D Tensor (N x C matrix) of type int64. The indices of the SparseTensor.
+          Support int64, each element value should be a non-negative int number. This tensor should be sorted.
+          The shape is :math:`(N, C)`.
+
+    Outputs:
+        - *y_grad_val: A Tensor. Has the same type as `backprop_val_grad`.
+          Has the same number as `indices`.
+
+    Raises:
+        TypeError: If the dtype of `indices`, `start`, `new_indices` are not int64.
+        ValueError: If `indices`, `new_indices` are not 2-D tensor.
+        ValueError: If `backprop_val_grad`, `start` is not a 1-D tensor.
+        ValueError: If the number of `backprop_val_grad` is not corresponding to the number of `new_indices`.
+        ValueError: If the shape of `indices[1]` is not corresponding to `start[1]`.
+        ValueError: If the shape of `indices[1]` is not corresponding to `new_indices[1]`.
+        RuntimeError: If the `backprop_val_grad` is not all backpropagated, because `indices` or `new_indices`
+        is not sorted.
+
+    Supported Platforms:
+        ``Ascend`` ``GPU`` ``CPU``
+    Examples:
+        >>> backprop_val_grad = Tensor(np.array([1, 2, 3, 4]).astype(np.int64))
+        >>> indices = Tensor(np.array([[0, 0], [0, 2], [1, 2], [1, 3], [2, 3], [2, 4]]).astype(np.int64))
+        >>> start = Tensor(np.array([0, 0]).astype(np.int64))
+        >>> new_indices = Tensor(np.array([[0, 2], [1, 2], [1, 3], [2, 4]]).astype(np.int64))
+        >>> grad = SparseSliceGrad()
+        >>> output = grad(backprop_val_grad, indices, start, new_indices)
+        >>> print(output)
+        [0 1 2 3 0 4]
+    """
+
+    @prim_attr_register
+    def __init__(self):
+        """Initialize SparseSliceGrad."""
+        self.init_prim_io_names(inputs=['backprop_val_grad', 'indices', 'start', 'new_indices'], outputs=['y_grad'])
+
+
+class FractionalMaxPoolGradWithFixedKsize(Primitive):
+    """
+    Computes the gradients of FractionalMaxPoolWithFixedKsize.
+
+    Args:
+        data_format (str): The optional value for data format, is 'NCHW'. Default: "NCHW".
+
+    Inputs:
+        - **origin_input** (Tensor) - Tensor with data format "NCHW", data type must be int32 or int64.
+        - **out_backprop** (Tensor) - The gradients with respect to the output of FractionalMaxPoolWithFixedKsize
+        function. Tensor with data format "NCHW", whose data type is float16, float32, float64, int32 or int64.
+        - **argmax** (Tensor) - The second output of FractionalMaxPoolWithFixedKsize function, whose data
+        type is int64.
+
+    Outputs:
+        - **y** (Tensor) - Tensor, with the same shape as `origin_input`, and the same data type as
+        the input `out_backprop`.
+
+    Raises:
+        TypeError: If data type of `out_backprop` is not one of the following: float16, float32, float64, int32, int64.
+        TypeError: If data type of `argmax` is not int64.
+        ValueError: If the shape of `out_backprop` and `argmax` is not equal.
+        ValueError: If the first dimension size of `origin_input` and `out_backprop` is not equal.
+        ValueError: If the second dimension size of `origin_input` and `out_backprop` is not equal.
+
+    Supported Platforms:
+        ``Ascend`` ``GPU`` ``CPU``
+    """
+
+    @prim_attr_register
+    def __init__(self, data_format="NCHW"):
+        self.data_format = validator.check_string(data_format, ['NCHW'], 'data_format', self.name)
+        self.add_prim_attr("data_format", self.data_format)
+        self.init_prim_io_names(inputs=['origin_input', 'out_backprop', 'argmax'], outputs=['y'])
+
+
+class AffineGridGrad(Primitive):
+    r"""
+    Computes gradients for AffineGrid operation.
+
+    Args:
+        align_corners (bool): if True, consider -1 and 1 to refer to the centers
+            of the corner pixels rather than the image corners. Default: ``False``.
+
+    Inputs:
+        - **y_grad** (Tensor) - Data type must be float16 or float32.
+        - **x_size** (tuple) - Data type must be int32 or int64.
+
+    Outputs:
+        Tensor, with data type same as `y_grad`.
+
+    Supported Platforms:
+        ``CPU``
+
+    Examples:
+        >>> import mindspore.ops.operations._grad_ops as _grad_ops
+        >>> affinegridgrad = _grad_ops.AffineGridGrad()
+        >>> y_grad = Tensor(np.ones([1, 2, 2, 2]), mindspore.float32)
+        >>> x_size = (1, 2, 2, 2)
+        >>> x_grad = affinegridgrad(y_grad, x_size)
+        >>> print(x_grad)
+        [[[0. 0. 4.]
+          [0. 0. 4.]]]
+    """
+
+    @prim_attr_register
+    def __init__(self, align_corners=False):
+        """Initialize AffineGridGrad."""
+        validator.check_value_type("align_corners", align_corners, [bool], self.name)
+        self.init_prim_io_names(inputs=['y_grad', 'x_size'], outputs=['x_grad'])
+
+
+
+class GluGrad(Primitive):
+    """
+    Computes grad for Glu operation.
+    """
+
+    @prim_attr_register
+    def __init__(self, axis):
+        self.add_prim_attr("cust_aicpu", self.name)
+        self.init_prim_io_names(inputs=["grads", "x"], outputs=["y"])
+        validator.check_value_type("axis", axis, [int], self.name)
+
+
+class MapTensorGetGrad(Primitive):
+    """
+    Computes gradients for MapTensorGet operation.
+
+    Inputs:
+        - **map_tensor** (MapTensor) - The input `map_tensor` of the forward operator MapTensorGet.
+        - **key_tensor** (Tensor) - The input `key_tensor` of the forward operator MapTensorGet.
+        - **default_value** (Scalar) - The input `default_value` of the forward operator MapTensorGet.
+        - **grad** (Tensor) - The grad value according the forward operator MapTensorGet.
+
+    Outputs:
+        - **output** (MapTensor) -  MapTensor with grad values.
+    """
+    @prim_attr_register
+    def __init__(self):
+        """Initialize MapTensorGetGrad"""
+        self.init_prim_io_names(inputs=['map_tensor', 'key_tensor', 'default_value', 'grad'], outputs=['output'])
+        self.add_prim_attr('side_effect_mem', True)
+
+
+class ResizeV2Grad(Primitive):
+    r"""
+    Calculates the gradient of ResizeV2 operation.
+
+    Supported Platforms:
+        ``CPU``
+    """
+
+    @prim_attr_register
+    def __init__(self, coordinate_transformation_mode="half_pixel", mode="nearest"):
+        """Initialize ResizeV2Grad."""
+        self.init_prim_io_names(inputs=["grads", "roi", "scales", "original_size"], outputs=["y"])
+        self.add_prim_attr("nearest_mode", "floor")
+        self.add_prim_attr("cubic_coeff_a", -0.75)
+        validator.check_value_type(
+            "coordinate_transformation_mode", coordinate_transformation_mode, [str], self.name)
+        validator.check_string(coordinate_transformation_mode,
+                               ["align_corners", "half_pixel"], "coordinate_transformation_mode", self.name)
+        validator.check_value_type("mode", mode, [str], self.name)
+        validator.check_string(mode, ["nearest", "linear", "cubic"], "mode", self.name)
+
+
+class WKVGrad(Primitive):
+    r"""
+    Calculates the gradient of WKV operation.
+
+    Supported Platforms:
+        ``Ascend``
+    """
+
+    @prim_attr_register
+    def __init__(self):
+        """Initialize WKVGrad."""
+        self.init_prim_io_names(inputs=["time_first", "time_decay", "key", "value", "gy"],
+                                outputs=["gw", "gu", "gk", "gv"])