From 4fa06dd9a29eb561a0f2e0b108a6858ae01f2b74 Mon Sep 17 00:00:00 2001
From: wanyiming <lovelychurch@163.com>
Date: Thu, 24 Feb 2022 11:11:55 +0800
Subject: [PATCH] code_docs_init

---
 .../mindspore.common.initializer.rst          | 80 +++++++++++++++++++
 .../python/mindspore/common/initializer.py    | 22 ++---
 2 files changed, 93 insertions(+), 9 deletions(-)

diff --git a/docs/api/api_python/mindspore.common.initializer.rst b/docs/api/api_python/mindspore.common.initializer.rst
index 9777d4b7c09..59ef53c8f03 100644
--- a/docs/api/api_python/mindspore.common.initializer.rst
+++ b/docs/api/api_python/mindspore.common.initializer.rst
@@ -132,3 +132,83 @@ mindspore.common.initializer
 
 .. automodule:: mindspore.common.initializer
     :members:
+
+
+.. py:class:: mindspore.common.initializer.Identity(**kwargs)
+
+    生成一个2维的单位阵用于初始化Tensor。
+
+    **异常：**
+
+    - **ValueError** - 被初始化的Tensor的维度不等于2。
+
+
+.. py:class:: mindspore.common.initializer.Sparse(sparsity, sigma=0.01)
+
+    生成一个2维的稀疏矩阵用于初始化Tensor。矩阵非0的位置的值服从正态分布N(0, 0.01)。
+
+    **参数：**
+
+    **sparsity** (float) - 矩阵每列中元素被置0的比例。
+    **sigma** (float) - 正态分布的标准差，默认值为0.01。
+
+    **异常：**
+
+    - **ValueError** - 被初始化的Tensor的维度不等于2。
+
+
+
+.. py:class:: mindspore.common.initializer.Dirac(group=1)
+
+    利用Dirac delta函数生成一个array用于初始化Tensor。这种初始化方式将会保留卷积层的输入。对于group
+    卷积，通道的每个分组会被分别保留。
+
+    **参数：**
+
+    **group** (int) - 卷积层中的分组，默认值为1。
+
+    **异常：**
+
+    - **ValueError** - group不在[3, 4, 5]的范围内。
+    - **ValueError** - 初始化的Tensor的第一个维度不能被group整除。
+
+
+
+.. py:class:: mindspore.common.initializer.Orthogonal(gain=1.)
+
+    生成一个(半)正交矩阵用于初始化Tensor。被初始化的Tensor的维度至少为2。
+    如果维度大于2，多余的维度将会被展平。
+
+    **参数：**
+
+    **gain** (float) - 可选的比例因子，默认值为1。
+
+    **异常：**
+
+    - **ValueError** - 被初始化的Tensor的维度小于2。
+
+
+.. py:class:: mindspore.common.initializer.VarianceScaling(scale=1.0, mode="fan_in", distribution="truncated_normal")
+
+    生成一个随机的array用于初始化Tensor。
+    当distribution是"truncated_normal"或者"untruncated_normal"时，array中的值将服从均值为0，标准差
+    为 :math:`stddev = sqrt(scale/n)` 的截断或者非截断正太分布。如果mode是"fan_in"， :math:`n` 是输入单元的数量；
+    如果mode是"fan_out"， :math:`n` 是输出单元的数量；如果mode是"fan_avg"， :math:`n` 是输入输出单元数量的均值。
+    当distribution是"uniform"时，array中的值将服从均匀分布[`-sqrt(3*scale/n)`, `sqrt(3*scale/n)`]。
+
+    **参数：**
+
+    **scale** (float) - 比例因子，默认值为1.0。
+    **mode** (str) - 其值应为"fan_in"，"fan_out"或者"fan_avg"，默认值为"fan_in"。
+    **distribution** (str) - 用于采样的分布类型。它可以是"uniform"，"truncated_normal"或"untruncated_normal"，
+    默认值为"truncated_normal"。
+
+
+    **异常：**
+
+    - **ValueError** - scale小于等于0。
+    - **ValueError** - mode不是"fan_in"，"fan_out"或者"fan_avg"。
+    - **ValueError** - distribution不是"truncated_normal"，"untruncated_normal"或者"uniform"。
+
+
+    
diff --git a/mindspore/python/mindspore/common/initializer.py b/mindspore/python/mindspore/common/initializer.py
index 1c7c1b875b6..5a0f585b3bb 100644
--- a/mindspore/python/mindspore/common/initializer.py
+++ b/mindspore/python/mindspore/common/initializer.py
@@ -379,7 +379,7 @@ class Constant(Initializer):
 @_register()
 class Identity(Initializer):
     """
-    Initialize a 2 dimension identity matrix to fill the input tensor.
+    Generates a 2 dimension identity matrix array in order to initialize a tensor.
 
     Raises:
         ValueError: If the dimension of input tensor is not equal to 2.
@@ -401,8 +401,8 @@ class Identity(Initializer):
 @_register()
 class Sparse(Initializer):
     """
-    Initialize a 2 dimension sparse matrix to fill the input tensor. The non-zero positions will be filled with
-    the value sampled from the normal distribution :math:`{N}(0, 0.01)`
+    Generates a 2 dimension sparse matrix array in order to initialize a tensor. The non-zero positions
+    will be filled with the value sampled from the normal distribution :math:`{N}(0, 0.01)`
 
     Args:
          sparsity (float): The fraction of elements being set to zero in each column.
@@ -436,8 +436,10 @@ class Sparse(Initializer):
 
 @_register()
 class Dirac(Initializer):
-    """Initialize input tensor with the Dirac delta function. It tries to preserves the identity of
-    input for convolution layers. For group convolution, each group of channels will be preserved respectively.
+    """
+    Generates an array with the Dirac delta function in order to initialize a tensor.
+    It tries to preserves the identity of input for convolution layers.
+    For group convolution, each group of channels will be preserved respectively.
 
     Args:
         groups (int): The number of group in convolution layer. Default: 1.
@@ -487,8 +489,9 @@ class Dirac(Initializer):
 @_register()
 class Orthogonal(Initializer):
     r"""
-    Initialize a (semi) orthogonal matrix to fill the input tensor. The dimension of input tensor must have at least 2
-    dimensions. If the dimension is greater than 2, the trailing dimensions will be flattened.
+    Generates a (semi) orthogonal matrix array in order to initialize a tensor.
+    The dimension of input tensor must have at least 2 dimensions.
+    If the dimension is greater than 2, the trailing dimensions will be flattened.
 
     Args:
          gain (float): An optional scaling factor. Default: 1.
@@ -532,7 +535,7 @@ class Orthogonal(Initializer):
 @_register()
 class VarianceScaling(Initializer):
     r"""
-    Randomly initialize an array with scaling to fill the input tensor.
+    Generates an random array with scaling in order to initialize a tensor.
     When distribution is truncated_normal or untruncated_normal, the value will be sampled from truncated or
     untruncated normal distribution with a mean of 0 and a scaled standard deviation :math:`stddev = sqrt(scale/n)`.
     :math:`n` will be the number of input units if mode is fan_in, the number of output units if mode is fan_out,
@@ -543,7 +546,8 @@ class VarianceScaling(Initializer):
     Args:
         scale (float): The scaling factor. Default: 1.0.
         mode (str): Should be 'fan_in', 'fan_out' or 'fan_avg'. Default: 'fan_in'.
-        distribution(str): The type of distribution chose to sample values. Default: 'truncated_normal'.
+        distribution(str): The type of distribution chose to sample values. It should be
+        'uniform', 'truncated_normal' or 'untruncated_normal'. Default: 'truncated_normal'.
 
     Raises:
         ValueError: If scale is not greater than 0.