mindspore 2.0.0rc1__cp38-cp38-manylinux1_x86_64.whl → 2.2.0__cp38-cp38-manylinux1_x86_64.whl

mindspore/nn/layer/normalization.py

@@ -62,7 +62,8 @@ class _BatchNorm(Cell):
                  moving_mean_init='zeros',
                  moving_var_init='ones',
                  use_batch_statistics=None,
-                 data_format='NCHW'):
+                 data_format='NCHW',
+                 dtype=mstype.float32):
         """Initialize _BatchNorm."""
         super(_BatchNorm, self).__init__()
         validator.check_value_type('num_features', num_features, [int], self.cls_name)
@@ -87,13 +88,13 @@ class _BatchNorm(Cell):
         self.moving_mean_init = moving_mean_init
         self.moving_var_init = moving_var_init
         self.moving_mean = Parameter(initializer(
-            moving_mean_init, num_features), name="mean", requires_grad=False)
+            moving_mean_init, num_features, dtype=dtype), name="mean", requires_grad=False)
         self.moving_variance = Parameter(initializer(
-            moving_var_init, num_features), name="variance", requires_grad=False)
+            moving_var_init, num_features, dtype=dtype), name="variance", requires_grad=False)
         self.gamma = Parameter(initializer(
-            gamma_init, num_features), name="gamma", requires_grad=affine)
+            gamma_init, num_features, dtype=dtype), name="gamma", requires_grad=affine)
         self.beta = Parameter(initializer(
-            beta_init, num_features), name="beta", requires_grad=affine)
+            beta_init, num_features, dtype=dtype), name="beta", requires_grad=affine)
 
         self.parallel_mode = context.get_auto_parallel_context("parallel_mode")
 
@@ -188,28 +189,39 @@ class BatchNorm1d(_BatchNorm):
 
     Args:
         num_features (int): number of features or channels `C` of the input `x` .
-        eps (float): :math:`\epsilon` added to the denominator for numerical stability. Default: 1e-5.
+        eps (float): :math:`\epsilon` added to the denominator for numerical stability. Default: ``1e-5`` .
         momentum (float): A floating hyperparameter of the momentum for the
-            running_mean and running_var computation. Default: 0.9.
-        affine (bool): A bool value. When set to True, :math:`\gamma` and :math:`\beta` can be learned. Default: True.
+            running_mean and running_var computation. Default: ``0.9`` .
+        affine (bool): A bool value. When set to ``True`` , :math:`\gamma` and :math:`\beta` can be learned.
+            Default: ``True`` .
         gamma_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the :math:`\gamma` weight.
-            The values of str refer to the function `initializer` including 'zeros', 'ones', etc. Default: 'ones'.
+            The values of str refer to the function `mindspore.common.initializer
+            <https://www.mindspore.cn/docs/en/r2.2/api_python/mindspore.common.initializer.html>`_
+            including ``'zeros'`` , ``'ones'`` , etc. Default: ``'ones'`` .
         beta_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the :math:`\beta` weight.
-            The values of str refer to the function `initializer` including 'zeros', 'ones', etc. Default: 'zeros'.
+            The values of str refer to the function `mindspore.common.initializer
+            <https://www.mindspore.cn/docs/en/r2.2/api_python/mindspore.common.initializer.html>`_
+            including ``'zeros'`` , ``'ones'``, etc. Default: ``'zeros'`` .
         moving_mean_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the moving mean.
-            The values of str refer to the function `initializer` including 'zeros', 'ones', etc. Default: 'zeros'.
+            The values of str refer to the function `mindspore.common.initializer
+            <https://www.mindspore.cn/docs/en/r2.2/api_python/mindspore.common.initializer.html>`_
+            including ``'zeros'`` , ``'ones'`` , etc. Default: ``'zeros'`` .
         moving_var_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the moving variance.
-            The values of str refer to the function `initializer` including 'zeros', 'ones', etc. Default: 'ones'.
-        use_batch_statistics (bool): If true, use the mean value and variance value of current batch data. If false,
-            use the mean value and variance value of specified value. If None, the training process will use the mean
-            and variance of current batch data and track the running mean and variance, the evaluation process will use
-            the running mean and variance. Default: None.
-        data_format (str): The optional value for data format, is 'NHWC' or 'NCHW'.
-            Default: 'NCHW'.
+            The values of str refer to the function `mindspore.common.initializer
+            <https://www.mindspore.cn/docs/en/r2.2/api_python/mindspore.common.initializer.html>`_
+            including ``'zeros'`` , ``'ones'`` , etc. Default: ``'ones'`` .
+        use_batch_statistics (bool): If ``true`` , use the mean value and variance value of current batch data. If
+            ``false`` , use the mean value and variance value of specified value. If ``None`` , the training process
+            will use the mean and variance of current batch data and track the running mean and variance, the
+            evaluation process will use the running mean and variance. Default: ``None`` .
+        data_format (str): The optional value for data format, is ``'NHWC'`` or ``'NCHW'`` .
+            Default: ``'NCHW'`` .
+        dtype (:class:`mindspore.dtype`): Dtype of Parameters. Default: ``mstype.float32`` .
 
     Inputs:
         - **x** (Tensor) - Tensor of shape :math:`(N, C)` or :math:`(N, C, L)` ,
           where `N` is the batch size, `C` is the number of features or channels, and `L` is the sequence length.
+          Supported types: float16, float32.
 
     Outputs:
         Tensor, the normalized, scaled, offset tensor, of shape :math:`(N, C)` or :math:`(N, C, L)` .
@@ -225,11 +237,10 @@ class BatchNorm1d(_BatchNorm):
 
     Examples:
         >>> import numpy as np
-        >>> import mindspore.nn as nn
-        >>> from mindspore import Tensor
-        >>> net = nn.BatchNorm1d(num_features=4)
-        >>> x = Tensor(np.array([[0.7, 0.5, 0.5, 0.6],
-        ...                      [0.5, 0.4, 0.6, 0.9]]).astype(np.float32))
+        >>> import mindspore as ms
+        >>> net = ms.nn.BatchNorm1d(num_features=4)
+        >>> x = ms.Tensor(np.array([[0.7, 0.5, 0.5, 0.6],
+        ...                         [0.5, 0.4, 0.6, 0.9]]).astype(np.float32))
         >>> output = net(x)
         >>> print(output)
         [[ 0.6999965   0.4999975  0.4999975  0.59999704 ]
@@ -274,32 +285,42 @@ class BatchNorm2d(_BatchNorm):
     Args:
         num_features (int): The number of channels of the input tensor. Expected input size is :math:`(N, C, H, W)`,
             `C` represents the number of channels.
-        eps (float): :math:`\epsilon` added to the denominator for numerical stability. Default: 1e-5.
+        eps (float): :math:`\epsilon` added to the denominator for numerical stability. Default: ``1e-5`` .
         momentum (float): A floating hyperparameter of the momentum for the
-            running_mean and running_var computation. Default: 0.9.
-        affine (bool): A bool value. When set to True, :math:`\gamma` and :math:`\beta` can be learned. Default: True.
+            running_mean and running_var computation. Default: ``0.9`` .
+        affine (bool): A bool value. When set to ``True`` , :math:`\gamma` and :math:`\beta` can be learned.
+            Default: ``True`` .
         gamma_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the :math:`\gamma` weight.
-            The values of str refer to the function `initializer` including 'zeros', 'ones', etc. Default: 'ones'.
+            The values of str refer to the function `mindspore.common.initializer
+            <https://www.mindspore.cn/docs/en/r2.2/api_python/mindspore.common.initializer.html>`_
+            including ``'zeros'`` , ``'ones'`` , etc. Default: ``'ones'`` .
         beta_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the :math:`\beta` weight.
-            The values of str refer to the function `initializer` including 'zeros', 'ones', etc. Default: 'zeros'.
+            The values of str refer to the function `mindspore.common.initializer
+            <https://www.mindspore.cn/docs/en/r2.2/api_python/mindspore.common.initializer.html>`_
+            including ``'zeros'`` , ``'ones'`` , etc. Default: ``'zeros'`` .
         moving_mean_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the moving mean.
-            The values of str refer to the function `initializer` including 'zeros', 'ones', etc. Default: 'zeros'.
+            The values of str refer to the function `mindspore.common.initializer
+            <https://www.mindspore.cn/docs/en/r2.2/api_python/mindspore.common.initializer.html>`_
+            including ``'zeros'`` , ``'ones'`` , etc. Default: ``'zeros'`` .
         moving_var_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the moving variance.
-            The values of str refer to the function `initializer` including 'zeros', 'ones', etc. Default: 'ones'.
-        use_batch_statistics (bool):
+            The values of str refer to the function `mindspore.common.initializer
+            <https://www.mindspore.cn/docs/en/r2.2/api_python/mindspore.common.initializer.html>`_
+            including ``'zeros'`` , ``'ones'`` , etc. Default: ``'ones'`` .
+        use_batch_statistics (bool): Default: ``None`` .
 
-            - If true, use the mean value and variance value of current batch data and track running mean
+            - If ``true`` , use the mean value and variance value of current batch data and track running mean
               and running variance.
-            - If false, use the mean value and variance value of specified value, and not track statistical value.
-            - If None, the use_batch_statistics is automatically set to true or false according to the training
-              and evaluation mode. During training, the parameter is set to true, and during evaluation, the
-              parameter is set to false. Default: None.
+            - If ``false`` , use the mean value and variance value of specified value, and not track statistical value.
+            - If ``None`` , the use_batch_statistics is automatically set to ``true`` or ``false`` according to the
+              training and evaluation mode. During training, the parameter is set to true, and during evaluation, the
+              parameter is set to false.
 
-        data_format (str): The optional value for data format, is 'NHWC' or 'NCHW'.
-            Default: 'NCHW'.
+        data_format (str): The optional value for data format, is ``'NHWC'`` or ``'NCHW'`` .
+            Default: ``'NCHW'`` .
+        dtype (:class:`mindspore.dtype`): Dtype of Parameters. Default: ``mstype.float32`` .
 
     Inputs:
-        - **x** (Tensor) - Tensor of shape :math:`(N, C, H, W)`.
+        - **x** (Tensor) - Tensor of shape :math:`(N, C, H, W)`. Supported types: float16, float32.
 
     Outputs:
         Tensor, the normalized, scaled, offset tensor, of shape :math:`(N, C, H, W)`.
@@ -316,10 +337,9 @@ class BatchNorm2d(_BatchNorm):
 
     Examples:
         >>> import numpy as np
-        >>> import mindspore.nn as nn
-        >>> from mindspore import Tensor
-        >>> net = nn.BatchNorm2d(num_features=3)
-        >>> x = Tensor(np.ones([1, 3, 2, 2]).astype(np.float32))
+        >>> import mindspore as ms
+        >>> net = ms.nn.BatchNorm2d(num_features=3)
+        >>> x = ms.Tensor(np.ones([1, 3, 2, 2]).astype(np.float32))
         >>> output = net(x)
         >>> print(output)
         [[[[ 0.999995 0.999995 ]
@@ -355,25 +375,35 @@ class BatchNorm3d(Cell):
 
     Args:
         num_features (int): `C` from an expected input of size :math:`(N, C, D, H, W)` .
-        eps (float): A value added to the denominator for numerical stability. Default: 1e-5.
+        eps (float): A value added to the denominator for numerical stability. Default: ``1e-5`` .
         momentum (float): A floating hyperparameter of the momentum for the
-            running_mean and running_var computation. Default: 0.9.
-        affine (bool): A bool value. When set to True, gamma and beta can be learned. Default: True.
+            running_mean and running_var computation. Default: ``0.9`` .
+        affine (bool): A bool value. When set to ``True`` , gamma and beta can be learned. Default: ``True`` .
         gamma_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the gamma weight.
-            The values of str refer to the function `initializer` including 'zeros', 'ones', etc. Default: 'ones'.
+            The values of str refer to the function `mindspore.common.initializer
+            <https://www.mindspore.cn/docs/en/r2.2/api_python/mindspore.common.initializer.html>`_
+            including ``'zeros'`` , ``'ones'`` , etc. Default: ``'ones'`` .
         beta_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the beta weight.
-            The values of str refer to the function `initializer` including 'zeros', 'ones', etc. Default: 'zeros'.
+            The values of str refer to the function `mindspore.common.initializer
+            <https://www.mindspore.cn/docs/en/r2.2/api_python/mindspore.common.initializer.html>`_
+            including ``'zeros'`` , ``'ones'`` , etc. Default: ``'zeros'`` .
         moving_mean_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the moving mean.
-            The values of str refer to the function `initializer` including 'zeros', 'ones', etc. Default: 'zeros'.
+            The values of str refer to the function `mindspore.common.initializer
+            <https://www.mindspore.cn/docs/en/r2.2/api_python/mindspore.common.initializer.html>`_
+            including ``'zeros'`` , ``'ones'`` , etc. Default: ``'zeros'`` .
         moving_var_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the moving variance.
-            The values of str refer to the function `initializer` including 'zeros', 'ones', etc. Default: 'ones'.
-        use_batch_statistics (bool): If true, use the mean value and variance value of current batch data. If false,
-            use the mean value and variance value of specified value. If None, the training process will use the mean
-            and variance of current batch data and track the running mean and variance, the evaluation process will use
-            the running mean and variance. Default: None.
+            The values of str refer to the function `mindspore.common.initializer
+            <https://www.mindspore.cn/docs/en/r2.2/api_python/mindspore.common.initializer.html>`_
+            including ``'zeros'`` , ``'ones'`` , etc. Default: ``'ones'`` .
+        use_batch_statistics (bool): If true, use the mean value and variance value of current batch data. If
+            ``false``, use the mean value and variance value of specified value. If ``None`` , the training process
+            will use the mean and variance of current batch data and track the running mean and variance, the
+            evaluation process will use the running mean and variance. Default: ``None`` .
+        dtype (:class:`mindspore.dtype`): Dtype of Parameters. Default: ``mstype.float32`` .
 
     Inputs:
         - **x** (Tensor) - Tensor of shape :math:`(N, C_{in}, D_{in}, H_{in}, W_{in})`.
+          Supported types: float16, float32.
 
     Outputs:
         Tensor, the normalized, scaled, offset tensor, of shape :math:`(N, C_{out}, D_{out},H_{out}, W_{out})`.
@@ -389,10 +419,9 @@ class BatchNorm3d(Cell):
 
     Examples:
         >>> import numpy as np
-        >>> import mindspore.nn as nn
-        >>> from mindspore import Tensor
-        >>> net = nn.BatchNorm3d(num_features=3)
-        >>> x = Tensor(np.ones([16, 3, 10, 32, 32]).astype(np.float32))
+        >>> import mindspore as ms
+        >>> net = ms.nn.BatchNorm3d(num_features=3)
+        >>> x = ms.Tensor(np.ones([16, 3, 10, 32, 32]).astype(np.float32))
         >>> output = net(x)
         >>> print(output.shape)
         (16, 3, 10, 32, 32)
@@ -407,7 +436,8 @@ class BatchNorm3d(Cell):
                  beta_init='zeros',
                  moving_mean_init='zeros',
                  moving_var_init='ones',
-                 use_batch_statistics=None):
+                 use_batch_statistics=None,
+                 dtype=mstype.float32):
         """Initialize BatchNorm3d."""
         super(BatchNorm3d, self).__init__()
         self.bn2d = BatchNorm2d(num_features=num_features,
@@ -419,7 +449,8 @@ class BatchNorm3d(Cell):
                                 moving_mean_init=moving_mean_init,
                                 moving_var_init=moving_var_init,
                                 use_batch_statistics=use_batch_statistics,
-                                data_format="NCHW")
+                                data_format="NCHW",
+                                dtype=dtype)
         self.shape = P.Shape()
         self.reshape = P.Reshape()
 
@@ -464,34 +495,36 @@ class SyncBatchNorm(_BatchNorm):
 
     Note:
         Currently, SyncBatchNorm only supports 2D and 4D inputs.
+        :math:`\gamma` and :math:`\beta` are trainable scale and shift.
 
     Args:
         num_features (int): `C` from an expected input of size :math:`(N, C, H, W)`.
-        eps (float): :math:`\epsilon`, a value added to the denominator for numerical stability. Default: 1e-5.
+        eps (float): :math:`\epsilon`, a value added to the denominator for numerical stability. Default: ``1e-5`` .
         momentum (float): A floating hyperparameter of the momentum for the
-            running_mean and running_var computation. Default: 0.9.
-        affine (bool): A bool value. When set to True, :math:`\gamma` and :math:`\beta` can be learned.
-            Default: True.
+            running_mean and running_var computation. Default: ``0.9`` .
+        affine (bool): A bool value. When set to ``True`` , :math:`\gamma` and :math:`\beta` can be learned.
+            Default: ``True`` .
         gamma_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the :math:`\gamma` weight.
-            The values of str refer to the function `initializer` including 'zeros', 'ones', 'xavier_uniform',
-            'he_uniform', etc. Default: 'ones'.
+            The values of str refer to the function `initializer` including ``'zeros'`` , ``'ones'`` ,
+            ``'xavier_uniform'`` , ``'he_uniform'`` , etc. Default: ``'ones'`` .
         beta_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the :math:`\beta` weight.
-            The values of str refer to the function `initializer` including 'zeros', 'ones', 'xavier_uniform',
-            'he_uniform', etc. Default: 'zeros'.
+            The values of str refer to the function `initializer` including ``'zeros'`` , ``'ones'`` ,
+            ``'xavier_uniform'`` , ``'he_uniform'`` , etc. Default: ``'zeros'`` .
         moving_mean_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the moving mean.
-            The values of str refer to the function `initializer` including 'zeros', 'ones', 'xavier_uniform',
-            'he_uniform', etc. Default: 'zeros'.
+            The values of str refer to the function `initializer` including ``'zeros'`` , ``'ones'`` ,
+            ``'xavier_uniform'`` , ``'he_uniform'`` , etc. Default: ``'zeros'`` .
         moving_var_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the moving variance.
-            The values of str refer to the function `initializer` including 'zeros', 'ones', 'xavier_uniform',
-            'he_uniform', etc. Default: 'ones'.
-        use_batch_statistics (bool): If true, use the mean value and variance value of current batch data. If false,
-            use the mean value and variance value of specified value. If None, training process will use the mean and
-            variance of current batch data and track the running mean and variance, eval process will use the running
-            mean and variance. Default: None.
+            The values of str refer to the function `initializer` including ``'zeros'`` , ``'ones'`` ,
+            ``'xavier_uniform'`` , ``'he_uniform'`` , etc. Default: ``'ones'`` .
+        use_batch_statistics (bool): If ``true`` , use the mean value and variance value of current batch data. If
+            ``false`` , use the mean value and variance value of specified value. If ``None`` , training process will
+            use the mean and variance of current batch data and track the running mean and variance, eval process will
+            use the running mean and variance. Default: ``None`` .
         process_groups (list): A list to divide devices into different sync groups, containing N subtraction lists.
             Each subtraction list contains int numbers identifying rank ids which need to be synchronized in the same
-            group. All int values must be in [0, rank_size) and different from each other. Default: None, indicating
-            synchronization across all devices.
+            group. All int values must be in [0, rank_size) and different from each other. Default: ``None`` ,
+            indicating synchronization across all devices.
+        dtype (:class:`mindspore.dtype`): Dtype of Parameters. Default: ``mstype.float32`` .
 
     Inputs:
         - **x** (Tensor) - Tensor of shape :math:`(N, C_{in}, H_{in}, W_{in})`.
@@ -516,27 +549,27 @@ class SyncBatchNorm(_BatchNorm):
 
             For the Ascend devices, users need to prepare the rank table, set rank_id and device_id.
             Please see the `Ascend tutorial
-            <https://www.mindspore.cn/tutorials/experts/en/r2.0/parallel/train_ascend.html#preparations>`_
+            <https://www.mindspore.cn/tutorials/experts/en/r2.2/parallel/rank_table.html>`_
             for more details.
 
-            For the GPU devices, users need to prepare the host file and mpi, please see the `GPU tutorial
-            <https://www.mindspore.cn/tutorials/experts/en/r2.0/parallel/train_gpu.html#preparation>`_ .
+            For the GPU devices, users need to prepare the host file and mpi, please see the `mpirun Startup
+            <https://www.mindspore.cn/tutorials/experts/en/r2.2/parallel/mpirun.html>`_ .
+
+            For the CPU device, users need to write a dynamic cluster startup script, please see the `Dynamic Cluster
+            Startup <https://www.mindspore.cn/tutorials/experts/en/r2.2/parallel/dynamic_cluster.html>`_ .
 
             This example should be run with multiple devices.
 
         >>> import numpy as np
         >>> import mindspore as ms
         >>> from mindspore.communication import init
-        >>> from mindspore import Tensor
-        >>> from mindspore import nn
-        >>> from mindspore import dtype as mstype
         >>>
         >>> ms.set_context(mode=ms.GRAPH_MODE)
         >>> init()
         >>> ms.reset_auto_parallel_context()
         >>> ms.set_auto_parallel_context(parallel_mode=ms.ParallelMode.DATA_PARALLEL)
-        >>> sync_bn_op = nn.SyncBatchNorm(num_features=3, process_groups=[[0, 1], [2, 3]])
-        >>> x = Tensor(np.ones([1, 3, 2, 2]), mstype.float32)
+        >>> sync_bn_op = ms.nn.SyncBatchNorm(num_features=3, process_groups=[[0, 1], [2, 3]])
+        >>> x = ms.Tensor(np.ones([1, 3, 2, 2]), ms.float32)
         >>> output = sync_bn_op(x)
         >>> print(output)
         [[[[ 0.999995 0.999995 ]
@@ -557,7 +590,8 @@ class SyncBatchNorm(_BatchNorm):
                  moving_mean_init='zeros',
                  moving_var_init='ones',
                  use_batch_statistics=None,
-                 process_groups=None):
+                 process_groups=None,
+                 dtype=mstype.float32):
         """Initialize SyncBatchNorm."""
         super(SyncBatchNorm, self).__init__(num_features,
                                             eps,
@@ -567,7 +601,8 @@ class SyncBatchNorm(_BatchNorm):
                                             beta_init,
                                             moving_mean_init,
                                             moving_var_init,
-                                            use_batch_statistics)
+                                            use_batch_statistics,
+                                            dtype=dtype)
         self.is_global = False
         self.group_name = None
         self.process_groups = process_groups
@@ -642,27 +677,28 @@ class LayerNorm(Cell):
     normalization on a mini-batch of inputs for each single training case as described
     in the paper `Layer Normalization <https://arxiv.org/pdf/1607.06450.pdf>`_. Unlike Batch
     Normalization, Layer Normalization performs exactly the same computation at training and
-    testing time. It is applied across all channels
-    and pixel but only one batch size. It can be described using the following formula:
+    testing time. It is applied across all channels and pixel but only one batch size.
+    :math:`\gamma` and :math:`\beta` are trainable scale and shift.
+    It can be described using the following formula:
 
     .. math::
         y = \frac{x - \mathrm{E}[x]}{\sqrt{\mathrm{Var}[x] + \epsilon}} * \gamma + \beta
 
     Args:
         normalized_shape (Union(tuple[int], list[int])): The normalization is performed over axis
-            `begin_norm_axis ... R - 1`.
+            `begin_norm_axis ... R - 1`. R is the dimension size of input `x`.
         begin_norm_axis (int): The first normalization dimension: normalization will be performed along dimensions
-            `begin_norm_axis: rank(inputs)`, the value should be in [-1, rank(input)). Default: -1.
-        begin_params_axis (int): The first parameter(beta, gamma)dimension: scale and centering parameters
-            will have dimensions `begin_params_axis: rank(inputs)` and will be broadcast with
-            the normalized inputs accordingly, the value should be in [-1, rank(input)). Default: -1.
+            `begin_norm_axis: R`, the value should be in [-1, R). Default: ``-1`` .
+        begin_params_axis (int): The begin axis of the parameter input :math:`(\gamma, \beta)` to
+            apply LayerNorm, the value should be in [-1, R). Default: ``-1`` .
         gamma_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the :math:`\gamma` weight.
-            The values of str refer to the function `initializer` including 'zeros', 'ones', 'xavier_uniform',
-            'he_uniform', etc. Default: 'ones'.
+            The values of str refer to the function `initializer` including ``'zeros'`` , ``'ones'`` ,
+            ``'xavier_uniform'`` , ``'he_uniform'`` , etc. Default: ``'ones'`` .
         beta_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the :math:`\beta` weight.
-            The values of str refer to the function `initializer` including 'zeros', 'ones', 'xavier_uniform',
-            'he_uniform', etc. Default: 'zeros'.
-        epsilon (float): :math:`\epsilon` added to the denominator for numerical stability. Default: 1e-7.
+            The values of str refer to the function `initializer` including ``'zeros'`` , ``'ones'`` ,
+            ``'xavier_uniform'`` , ``'he_uniform'`` , etc. Default: ``'zeros'`` .
+        epsilon (float): A value added to the denominator for numerical stability(:math:`\epsilon`). Default: ``1e-7`` .
+        dtype (:class:`mindspore.dtype`): Dtype of Parameters. Default: ``mstype.float32`` .
 
     Inputs:
         - **x** (Tensor) - The shape of `x` is :math:`(x_1, x_2, ..., x_R)`,
@@ -680,9 +716,11 @@ class LayerNorm(Cell):
         ``Ascend`` ``GPU`` ``CPU``
 
     Examples:
-        >>> x = Tensor(np.ones([20, 5, 10, 10]), mindspore.float32)
+        >>> import mindspore as ms
+        >>> import numpy as np
+        >>> x = ms.Tensor(np.ones([20, 5, 10, 10]), ms.float32)
         >>> shape1 = x.shape[1:]
-        >>> m = nn.LayerNorm(shape1,  begin_norm_axis=1, begin_params_axis=1)
+        >>> m = ms.nn.LayerNorm(shape1,  begin_norm_axis=1, begin_params_axis=1)
         >>> output = m(x).shape
         >>> print(output)
         (20, 5, 10, 10)
@@ -694,21 +732,27 @@ class LayerNorm(Cell):
                  begin_params_axis=-1,
                  gamma_init='ones',
                  beta_init='zeros',
-                 epsilon=1e-7
+                 epsilon=1e-7,
+                 dtype=mstype.float32
                  ):
         """Initialize LayerNorm."""
         super(LayerNorm, self).__init__()
         if not isinstance(normalized_shape, (tuple, list)):
             raise TypeError(f"For '{self.cls_name}', the type of 'normalized_shape' must be tuple[int] or list[int], "
                             f"but got {normalized_shape} and the type is {type(normalized_shape)}.")
+        if not normalized_shape:
+            raise ValueError(
+                f"Expected normalized_shape to be at least 1-dimensional, i.e., containing at "
+                f"least one element, but got normalized_shape = {normalized_shape}"
+            )
         self.normalized_shape = normalized_shape
         self.begin_norm_axis = begin_norm_axis
         self.begin_params_axis = begin_params_axis
         self.epsilon = epsilon
         self.gamma = Parameter(initializer(
-            gamma_init, normalized_shape), name="gamma")
+            gamma_init, normalized_shape, dtype=dtype), name="gamma")
         self.beta = Parameter(initializer(
-            beta_init, normalized_shape), name="beta")
+            beta_init, normalized_shape, dtype=dtype), name="beta")
         self.layer_norm = P.LayerNorm(begin_norm_axis=self.begin_norm_axis,
                                       begin_params_axis=self.begin_params_axis,
                                       epsilon=self.epsilon)
@@ -731,7 +775,8 @@ class _InstanceNorm(Cell):
                  momentum=0.1,
                  affine=True,
                  gamma_init='ones',
-                 beta_init='zeros'):
+                 beta_init='zeros',
+                 dtype=mstype.float32):
         """Initialize Normalization base class."""
         super(_InstanceNorm, self).__init__()
         validator.check_value_type('num_features', num_features, [int], self.cls_name)
@@ -748,12 +793,13 @@ class _InstanceNorm(Cell):
                              f"but got {momentum}.")
         self.num_features = num_features
         self.eps = eps
-        self.moving_mean = Parameter(initializer('zeros', num_features), name="mean", requires_grad=False)
-        self.moving_variance = Parameter(initializer('ones', num_features), name="variance", requires_grad=False)
+        self.moving_mean = Parameter(initializer('zeros', num_features, dtype=dtype), name="mean", requires_grad=False)
+        self.moving_variance = Parameter(initializer('ones', num_features, dtype=dtype), name="variance",
+                                         requires_grad=False)
         self.gamma = Parameter(initializer(
-            gamma_init, num_features), name="gamma", requires_grad=affine)
+            gamma_init, num_features, dtype=dtype), name="gamma", requires_grad=affine)
         self.beta = Parameter(initializer(
-            beta_init, num_features), name="beta", requires_grad=affine)
+            beta_init, num_features, dtype=dtype), name="beta", requires_grad=affine)
 
         self.shape = P.Shape()
         self.momentum = momentum
@@ -807,16 +853,17 @@ class InstanceNorm1d(_InstanceNorm):
 
     Args:
         num_features (int): `C` from an expected input of size :math:`(N, C, L)`.
-        eps (float): A value added to the denominator for numerical stability. Default: 1e-5.
+        eps (float): A value added to the denominator for numerical stability. Default: ``1e-5`` .
         momentum (float): A floating hyperparameter of the momentum for the
-            running_mean and running_var computation. Default: 0.1.
-        affine (bool): A bool value. When set to True, gamma and beta can be learned. Default: True.
+            running_mean and running_var computation. Default: ``0.1`` .
+        affine (bool): A bool value. When set to True, gamma and beta can be learned. Default: ``True`` .
         gamma_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the gamma weight.
-            The values of str refer to the function `initializer` including 'zeros', 'ones', etc.
-            When initialized with Tensor, the shape should be :math:`(C)`. Default: 'zeros'.
+            The values of str refer to the function `initializer` including ``'zeros'`` , ``'ones'`` , etc.
+            When initialized with Tensor, the shape should be :math:`(C)`. Default: ``'ones'`` .
         beta_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the beta weight.
-            The values of str refer to the function `initializer` including 'zeros', 'ones', etc.
-            When initialized with Tensor, the shape should be :math:`(C)`. Default: 'zeros'.
+            The values of str refer to the function `initializer` including ``'zeros'`` , ``'ones'`` , etc.
+            When initialized with Tensor, the shape should be :math:`(C)`. Default: ``'zeros'`` .
+        dtype (:class:`mindspore.dtype`): Dtype of Parameters. Default: ``mstype.float32`` .
 
     Inputs:
         - **x** (Tensor) - Tensor of shape :math:`(N, C, L)`. Data type: float16 or float32.
@@ -842,12 +889,10 @@ class InstanceNorm1d(_InstanceNorm):
         ``GPU``
 
     Examples:
-        >>> import mindspore
+        >>> import mindspore as ms
         >>> import numpy as np
-        >>> import mindspore.nn as nn
-        >>> from mindspore import Tensor
-        >>> net = nn.InstanceNorm1d(3)
-        >>> x = Tensor(np.ones([2, 3, 5]), mindspore.float32)
+        >>> net = ms.nn.InstanceNorm1d(3)
+        >>> x = ms.Tensor(np.ones([2, 3, 5]), ms.float32)
         >>> output = net(x)
         >>> print(output.shape)
         (2, 3, 5)
@@ -886,16 +931,17 @@ class InstanceNorm2d(_InstanceNorm):
 
     Args:
         num_features (int): `C` from an expected input of size :math:`(N, C, H, W)`.
-        eps (float): A value added to the denominator for numerical stability. Default: 1e-5.
+        eps (float): A value added to the denominator for numerical stability. Default: ``1e-5`` .
         momentum (float): A floating hyperparameter of the momentum for the
-            running_mean and running_var computation. Default: 0.1.
-        affine (bool): A bool value. When set to True, gamma and beta can be learned. Default: True.
+            running_mean and running_var computation. Default: ``0.1`` .
+        affine (bool): A bool value. When set to ``True`` , gamma and beta can be learned. Default: ``True`` .
         gamma_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the gamma weight.
-            The values of str refer to the function `initializer` including 'zeros', 'ones', etc.
-            When initialized with Tensor, the shape should be :math:`(C)`. Default: 'zeros'.
+            The values of str refer to the function `initializer` including ``'zeros'`` , ``'ones'`` , etc.
+            When initialized with Tensor, the shape should be :math:`(C)`. Default: ``'ones'`` .
         beta_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the beta weight.
-            The values of str refer to the function `initializer` including 'zeros', 'ones', etc.
-            When initialized with Tensor, the shape should be :math:`(C)`. Default: 'zeros'.
+            The values of str refer to the function `initializer` including ``'zeros'`` , ``'ones'`` , etc.
+            When initialized with Tensor, the shape should be :math:`(C)`. Default: ``'zeros'`` .
+        dtype (:class:`mindspore.dtype`): Dtype of Parameters. Default: ``mstype.float32`` .
 
     Inputs:
         - **x** (Tensor) - Tensor of shape :math:`(N, C, H, W)`. Data type: float16 or float32.
@@ -921,12 +967,10 @@ class InstanceNorm2d(_InstanceNorm):
         ``GPU``
 
     Examples:
-        >>> import mindspore
+        >>> import mindspore as ms
         >>> import numpy as np
-        >>> import mindspore.nn as nn
-        >>> from mindspore import Tensor
-        >>> net = nn.InstanceNorm2d(3)
-        >>> x = Tensor(np.ones([2, 3, 2, 2]), mindspore.float32)
+        >>> net = ms.nn.InstanceNorm2d(3)
+        >>> x = ms.Tensor(np.ones([2, 3, 2, 2]), ms.float32)
         >>> output = net(x)
         >>> print(output.shape)
         (2, 3, 2, 2)
@@ -964,16 +1008,17 @@ class InstanceNorm3d(_InstanceNorm):
 
     Args:
         num_features (int): `C` from an expected input of size :math:`(N, C, D, H, W)`.
-        eps (float): A value added to the denominator for numerical stability. Default: 1e-5.
+        eps (float): A value added to the denominator for numerical stability. Default: ``1e-5`` .
         momentum (float): A floating hyperparameter of the momentum for the
-            running_mean and running_var computation. Default: 0.1.
-        affine (bool): A bool value. When set to True, gamma and beta can be learned. Default: True.
+            running_mean and running_var computation. Default: ``0.1`` .
+        affine (bool): A bool value. When set to ``True`` , gamma and beta can be learned. Default: ``True`` .
         gamma_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the gamma weight.
-            The values of str refer to the function `initializer` including 'zeros', 'ones', etc.
-            When initialized with Tensor, the shape should be :math:`(C)`. Default: 'zeros'.
+            The values of str refer to the function `initializer` including ``'zeros'`` , ``'ones'`` , etc.
+            When initialized with Tensor, the shape should be :math:`(C)`. Default: ``'ones'`` .
         beta_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the beta weight.
-            The values of str refer to the function `initializer` including 'zeros', 'ones', etc.
-            When initialized with Tensor, the shape should be :math:`(C)`. Default: 'zeros'.
+            The values of str refer to the function `initializer` including ``'zeros'`` , ``'ones'`` , etc.
+            When initialized with Tensor, the shape should be :math:`(C)`. Default: ``'zeros'`` .
+        dtype (:class:`mindspore.dtype`): Dtype of Parameters. Default: ``mstype.float32`` .
 
     Inputs:
         - **x** (Tensor) - Tensor of shape :math:`(N, C, D, H, W)`. Data type: float16 or float32.
@@ -999,12 +1044,10 @@ class InstanceNorm3d(_InstanceNorm):
         ``GPU``
 
     Examples:
-        >>> import mindspore
+        >>> import mindspore as ms
         >>> import numpy as np
-        >>> import mindspore.nn as nn
-        >>> from mindspore import Tensor
-        >>> net = nn.InstanceNorm3d(3)
-        >>> x = Tensor(np.ones([2, 3, 5, 2, 2]), mindspore.float32)
+        >>> net = ms.nn.InstanceNorm3d(3)
+        >>> x = ms.Tensor(np.ones([2, 3, 5, 2, 2]), ms.float32)
         >>> output = net(x)
         >>> print(output.shape)
         (2, 3, 5, 2, 2)
@@ -1025,7 +1068,9 @@ class GroupNorm(Cell):
     normalization on a mini-batch of inputs for each single training case as described
     in the paper `Group Normalization <https://arxiv.org/pdf/1803.08494.pdf>`_. Group Normalization
     divides the channels into groups and computes within each group the mean and variance for normalization,
-    and it performs very stable over a wide range of batch size. It can be described using the following formula:
+    and it performs very stable over a wide range of batch size. :math:`\gamma` and :math:`\beta` are trainable scale
+    and shift.
+    It can be described using the following formula:
 
     .. math::
         y = \frac{x - \mathrm{E}[x]}{\sqrt{\mathrm{Var}[x] + \epsilon}} * \gamma + \beta
@@ -1033,14 +1078,18 @@ class GroupNorm(Cell):
     Args:
         num_groups (int): The number of groups to be divided along the channel dimension.
         num_channels (int): The number of input channels.
-        eps (float): A value added to the denominator for numerical stability. Default: 1e-5.
-        affine (bool): A bool value, this layer will have learnable affine parameters when set to true. Default: True.
+        eps (float): A value added to the denominator for numerical stability. Default: ``1e-05`` .
+        affine (bool): A bool value, this layer will have learnable affine parameters when set to ``true`` .
+            Default: ``True`` .
         gamma_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the gamma weight.
-            The values of str refer to the function `initializer` including 'zeros', 'ones', 'xavier_uniform',
-            'he_uniform', etc. Default: 'ones'. If gamma_init is a Tensor, the shape must be :math:`(num\_channels)`.
+            The values of str refer to the function `initializer` including ``'zeros'`` , ``'ones'`` ,
+            ``'xavier_uniform'`` , ``'he_uniform'`` , etc. Default: ``'ones'`` . If gamma_init is a Tensor, the shape
+            must be :math:`(num\_channels)`.
         beta_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the beta weight.
-            The values of str refer to the function `initializer` including 'zeros', 'ones', 'xavier_uniform',
-            'he_uniform', etc. Default: 'zeros'. If beta_init is a Tensor, the shape must be :math:`(num\_channels)`.
+            The values of str refer to the function `initializer` including ``'zeros'`` , ``'ones'`` ,
+            ``'xavier_uniform'`` , ``'he_uniform'`` , etc. Default: ``'zeros'`` . If beta_init is a Tensor, the shape
+            must be :math:`(num\_channels)`.
+        dtype (:class:`mindspore.dtype`): Dtype of Parameters. Default: ``mstype.float32`` .
 
     Inputs:
         - **x** (Tensor) - The input feature with shape :math:`(N, C, H, W)` .
@@ -1059,8 +1108,10 @@ class GroupNorm(Cell):
         ``Ascend`` ``GPU`` ``CPU``
 
     Examples:
-        >>> group_norm_op = nn.GroupNorm(2, 2)
-        >>> x = Tensor(np.ones([1, 2, 4, 4], np.float32))
+        >>> import mindspore as ms
+        >>> import numpy as np
+        >>> group_norm_op = ms.nn.GroupNorm(2, 2)
+        >>> x = ms.Tensor(np.ones([1, 2, 4, 4], np.float32))
         >>> output = group_norm_op(x)
         >>> print(output)
         [[[[0. 0. 0. 0.]
@@ -1073,7 +1124,8 @@ class GroupNorm(Cell):
            [0. 0. 0. 0.]]]]
     """
 
-    def __init__(self, num_groups, num_channels, eps=1e-05, affine=True, gamma_init='ones', beta_init='zeros'):
+    def __init__(self, num_groups, num_channels, eps=1e-05, affine=True, gamma_init='ones', beta_init='zeros',
+                 dtype=mstype.float32):
         """Initialize GroupNorm."""
         super(GroupNorm, self).__init__()
         self.num_groups = validator.check_positive_int(num_groups, "num_groups", self.cls_name)
@@ -1085,27 +1137,27 @@ class GroupNorm(Cell):
         self.affine = validator.check_bool(affine, arg_name="affine", prim_name=self.cls_name)
 
         self.gamma = Parameter(initializer(
-            gamma_init, num_channels), name="gamma", requires_grad=affine)
+            gamma_init, num_channels, dtype=dtype), name="gamma", requires_grad=affine)
         self.beta = Parameter(initializer(
-            beta_init, num_channels), name="beta", requires_grad=affine)
+            beta_init, num_channels, dtype=dtype), name="beta", requires_grad=affine)
+        self.reduce_mean = P.ReduceMean(keep_dims=True)
+        self.reduce_sum = P.ReduceSum(keep_dims=True)
         self.shape = F.shape
         self.reshape = F.reshape
-        self.reduce_mean = P.ReduceMean(keep_dims=True)
         self.square = F.square
-        self.reduce_sum = P.ReduceSum(keep_dims=True)
         self.sqrt = P.Sqrt()
 
     def _cal_output(self, x):
         """calculate groupnorm output"""
-        batch, channel, height, width = self.shape(x)
+        batch, channel, height, width = F.shape(x)
         self._channel_check(channel, self.num_channels, self.cls_name)
-        x = self.reshape(x, (batch, self.num_groups, -1))
+        x = F.reshape(x, (batch, self.num_groups, -1))
         mean = self.reduce_mean(x, 2)
-        var = self.reduce_sum(self.square(x - mean), 2) / (channel * height * width / self.num_groups)
+        var = F.div(self.reduce_sum(F.square(F.sub(x, mean)), 2), (channel * height * width / self.num_groups))
         std = self.sqrt(var + self.eps)
-        x = (x - mean) / std
-        x = self.reshape(x, (batch, channel, height, width))
-        output = x * self.reshape(self.gamma, (-1, 1, 1)) + self.reshape(self.beta, (-1, 1, 1))
+        x = F.div(F.sub(x, mean), std)
+        x = F.reshape(x, (batch, channel, height, width))
+        output = F.add(x * F.reshape(self.gamma, (-1, 1, 1)), F.reshape(self.beta, (-1, 1, 1)))
         return output
 
     @staticmethod
@@ -1133,7 +1185,7 @@ class GroupNorm(Cell):
         return 'num_groups={}, num_channels={}'.format(self.num_groups, self.num_channels)
 
     def construct(self, x):
-        self._check_input_dim(self.shape(x), self.cls_name)
+        self._check_input_dim(F.shape(x), self.cls_name)
         self._check_dtype(x.dtype, [mstype.float16, mstype.float32], self.cls_name)
         output = self._cal_output(x)
         return output