diff --git a/doc/fluid/api_guides/fluid.rst b/doc/fluid/api_guides/fluid.rst
new file mode 100755
index 0000000000000000000000000000000000000000..65c21a14f6475509636ee7842bf5a712f954d540
--- /dev/null
+++ b/doc/fluid/api_guides/fluid.rst
@@ -0,0 +1,172 @@
+######
+fluid
+######
+.. _cn_api_fluid_AsyncExecutor:
+
+AsyncExecutor
+>>>>>>>>>>>>>>>
+
+.. py:class:: paddle.fluid.AsyncExecutor(place=None)
+
+Python中的异步执行器。AsyncExecutor利用多核处理器和数据排队的强大功能，使数据读取和融合解耦，每个线程并行运行。
+
+AsyncExecutor不是在python端读取数据，而是接受一个训练文件列表，该列表将在c++中检索，然后训练输入将被读取、解析并在c++代码中提供给训练网络。
+
+AsyncExecutor正在积极开发，API可能在不久的将来会发生变化。
+
+参数：
+	- **place** (fluid.CPUPlace|None) - 指示 executor 将在哪个设备上运行。目前仅支持CPU
+
+**代码示例：**
+
+.. code-block:: python
+
+    data_feed = fluid.DataFeedDesc('data.proto')
+    startup_program = fluid.default_startup_program()
+    main_program = fluid.default_main_program()
+    filelist = ["train_data/part-%d" % i for i in range(100)]
+    thread_num = len(filelist) / 4
+    place = fluid.CPUPlace()
+    async_executor = fluid.AsyncExecutor(place)
+    async_executor.run_startup_program(startup_program)
+    epoch = 10
+    for i in range(epoch):
+        async_executor.run(main_program,
+                           data_feed,
+                           filelist,
+                           thread_num,
+                           [acc],
+                           debug=False)
+
+.. note::
+
+	对于并行gpu调试复杂网络，您可以在executor上测试。他们有完全相同的参数，并可以得到相同的结果。
+
+	目前仅支持CPU
+
+.. _cn_api_fluid_DataFeedDesc:
+
+DataFeedDesc
+>>>>>>>>>>>>>>
+
+.. py:class:: paddle.fluid.DataFeedDesc(proto_file)
+
+数据描述符，描述输入训练数据格式。
+
+这个类目前只用于AsyncExecutor(有关类AsyncExecutor的简要介绍，请参阅注释)
+
+DataFeedDesc应由来自磁盘的有效protobuf消息初始化:
+
+.. code-block:: python
+
+	data_feed = fluid.DataFeedDesc('data.proto')
+
+可以参考 :code:`paddle/fluid/framework/data_feed.proto` 查看我们如何定义message
+
+一段典型的message可能是这样的：
+
+.. code-block:: text
+
+    name: "MultiSlotDataFeed"
+    batch_size: 2
+    multi_slot_desc {
+        slots {
+            name: "words"
+            type: "uint64"
+            is_dense: false
+            is_used: true
+        }
+        slots {
+            name: "label"
+            type: "uint64"
+            is_dense: false
+            is_used: true
+        }
+    }
+
+但是，用户通常不应该关心消息格式;相反，我们鼓励他们在将原始日志文件转换为AsyncExecutor可以接受的训练文件的过程中，使用 :code:`Data Generator` 生成有效数据描述。
+
+DataFeedDesc也可以在运行时更改。一旦你熟悉了每个字段的含义，您可以修改它以更好地满足您的需要。例如:
+
+.. code-block:: text
+
+    data_feed.set_batch_size(128)
+    data_feed.set_dense_slots('wd')  # The slot named 'wd' will be dense
+    data_feed.set_use_slots('wd')    # The slot named 'wd' will be used
+    
+    #Finally, the content can be dumped out for debugging purpose:
+    
+    print(data_feed.desc())
+
+参数：
+	- **proto_file** (string) - 包含数据feed中描述的磁盘文件
+
+
+.. py:method:: set_batch_size(self, batch_size)
+
+设置batch size，训练期间有效
+
+
+参数：
+	- batch_size：batch size
+
+**代码示例：**
+
+.. code-block:: python
+	
+	data_feed = fluid.DataFeedDesc('data.proto')
+	data_feed.set_batch_size(128)
+
+.. py:method:: set_dense_slots(self, dense_slots_name)
+
+指定slot经过设置后将变成密集的slot，仅在训练期间有效。
+
+密集slot的特征将被输入一个Tensor，而稀疏slot的特征将被输入一个lodTensor
+
+
+参数：
+	- **dense_slots_name** : slot名称的列表，这些slot将被设置为密集的
+
+**代码示例：**
+
+.. code-block:: python
+	
+	data_feed = fluid.DataFeedDesc('data.proto')
+	data_feed.set_dense_slots(['words'])
+
+.. note:: 
+
+	默认情况下，所有slot都是稀疏的
+
+.. py:method:: set_use_slots(self, use_slots_name)
+
+
+设置一个特定的slot是否用于训练。一个数据集包含了很多特征，通过这个函数可以选择哪些特征将用于指定的模型。
+
+参数：
+	- **use_slots_name** :将在训练中使用的slot名列表
+
+**代码示例：**
+
+.. code-block:: python
+
+	data_feed = fluid.DataFeedDesc('data.proto')
+	data_feed.set_use_slots(['words'])
+
+.. note::
+	
+	默认值不用于所有slot
+
+
+.. py:method:: desc(self)
+
+返回此DataFeedDesc的protobuf信息
+
+返回：一个message字符串
+
+**代码示例：**
+
+.. code-block:: python
+
+	data_feed = fluid.DataFeedDesc('data.proto')
+	print(data_feed.desc())
\ No newline at end of file
diff --git a/doc/fluid/api_guides/hy-3.rst b/doc/fluid/api_guides/hy-3.rst
new file mode 100755
index 0000000000000000000000000000000000000000..e08cc3d384acaaae8b5581079c4431634eafecbc
--- /dev/null
+++ b/doc/fluid/api_guides/hy-3.rst
@@ -0,0 +1,57 @@
+
+.. _cn_api_fluid_layers_linear_chain_crf:
+
+linear_chain_crf
+>>>>>>>>>>>>>>>>>>
+
+.. py:class:: paddle.fluid.layers.linear_chain_crf(input, label, param_attr=None)
+
+线性链条件随机场（Linear Chain CRF）
+
+条件随机场定义间接概率图，节点代表随机变量，边代表两个变量之间的依赖。CRF学习条件概率 :math:`P\left ( Y|X \right )` ， :math:`X = \left ( x_{1},x_{2},...,x_{n} \right )` 是结构性输入，:math:`Y = \left ( y_{1},y_{2},...,y_{n} \right )` 为输入标签。
+
+线性链条件随机场（Linear Chain CRF)是特殊的条件随机场（CRF），有利于序列标注任务。序列标注任务不为输入设定许多条件依赖。唯一的限制是输入和输出必须是现行序列。因此类似CRF的图是一个简单的链或者线，也就是线性链随机场（linear chain CRF）。
+
+该操作符实现了线性链条件随机场（linear chain CRF）的前后向算法。详情请参照http://www.cs.columbia.edu/~mcollins/fb.pdf和http://cseweb.ucsd.edu/~elkan/250Bwinter2012/loglinearCRFs.pdf。
+
+1.这里x代表Emission
+
+2.Transition的第一维度值，代表起始权重，这里用a表示
+
+3.Transition的下一维值，代表末尾权重，这里用b表示
+
+4.Transition剩下的值，代表转移权重，这里用w表示
+
+5.Label用s表示
+
+长度为L的序列s的概率定义如下：
+
+.. math::
+
+    P(s) = (1/Z)exp(a_{s_{1}}+b_{s_{L}}+sum_{L}^{l=1}x_{s+{l}}+sum_{L}^{l=2}w_{s_{l-1},s_{l}})
+
+其中Z是正则化值，所有可能序列的P(s)之和为1，x是线性链条件随机场（linear chain CRF）的发射（emission）特征权重。
+
+线性链条件随机场最终输出mini-batch每个训练样本的条件概率的对数
+
+注意：
+    1.条件随机场（CRF）的特征函数由发射特征(emission feature）和转移特征（transition feature）组成。发射特征（emission feature）权重在调用函数前计算，而不在函数里计算。
+
+    2.由于该函数对所有可能序列的进行全局正则化，发射特征（emission feature）权重应是未缩放的。因此如果该函数带有发射特征（emission feature），并且发射特征是任意非线性激活函数的输出，则请勿调用该函数。
+
+    3.Emission的第二维度必须和标记数字（tag number）相同
+
+参数：
+    - **input** (Variable，LoDTensor，默认float类型LoDTensor) - 一个二维LoDTensor，shape为[N*D]，N是mini-batch的大小，D是总标记数。线性链条件随机场的未缩放发射权重矩阵
+    - **input** (Tensor，默认float类型LoDTensor) - 一个二维张量，shape为[(D+2)*D]。linear_chain_crf操作符的可学习参数。更多详情见operator注释
+    - **label** (Variable，LoDTensor，默认int64类型LoDTensor） - shape为[N*10的LoDTensor，N是mini-batch的总元素数
+    - **param_attr** (ParamAttr) - 可学习参数的属性
+
+返回：
+    output(Variable，Tensor，默认float类型Tensor)：shape为[N*D]的二维张量。Emission的指数。这是前向计算中的中间计算结果，在后向计算中还会复用
+
+    output(Variable，Tensor，默认float类型Tensor)：shape为[(D+2)*D]的二维张量。Transition的指数。这是前向计算中的中间计算结果，在后向计算中还会复用 
+
+    output(Variable,Tensor，默认float类型Tensor)：mini-batch每个训练样本的条件概率的对数。这是一个shape为[S*1]的二维张量，S是mini-batch的序列数。注：S等同于mini-batch的序列数。输出不再是LoDTensor
+
+返回类型：output（Variable）
diff --git a/doc/fluid/api_guides/hy-second.rst b/doc/fluid/api_guides/hy-second.rst
new file mode 100755
index 0000000000000000000000000000000000000000..f34447cfe5e78fb3eb29f8929607e207c46003e1
--- /dev/null
+++ b/doc/fluid/api_guides/hy-second.rst
@@ -0,0 +1,152 @@
+################
+hy_initializer
+################
+
+.. _cn_api_fluid_initializer_XavierInitializer:
+
+XavierInitializer
+>>>>>>>>>>>>>>>>>>>
+.. py:class:: paddle.fluid.initializer.XavierInitializer(uniform=True, fan_in=None, fan_out=None, seed=0)
+
+该类实现Xavier权重初始化方法（ Xavier weight initializer），Xavier权重初始化方法出自Xavier Glorot和Yoshua Bengio的论文 `Understanding the difficulty of training deep feedforward neural networks <http://proceedings.mlr.press/v9/glorot10a/glorot10a.pdf>`_
+
+该初始化函数用于保持所有层的梯度尺度几乎一致。
+
+在均匀分布的情况下，取值范围为[-x,x]，其中：
+
+.. math::
+
+    x = \sqrt{\frac{6.0}{fan\_in+fan\_out}}
+
+正态分布的情况下，均值为0，标准差为：
+
+.. math::
+    
+    x = \sqrt{\frac{2.0}{fan\_in+fan\_out}}
+
+参数：
+    - **uniform** (bool) - 是否用均匀分布或者正态分布
+    - **fan_in** (float) - 用于Xavier初始化的fan_in。如果为None，fan_in沿伸自变量
+    - **fan_out** (float) - 用于Xavier初始化的fan_out。如果为None，fan_out沿伸自变量
+    - **seed** (int) - 随机种子
+
+.. note::
+
+    在大多数情况下推荐将fan_in和fan_out设置为None
+
+**代码示例**：
+
+.. code-block:: python
+
+    fc = fluid.layers.fc(
+        input=queries, size=10,
+        param_attr=fluid.initializer.Xavier(uniform=False))
+
+
+.. _cn_api_fluid_initializer_MSRAInitializer:
+
+MSRAInitializer
+>>>>>>>>>>>>>>>>>
+
+.. py:class:: paddle.fluid.initializer.MSRAInitializer(uniform=True, fan_in=None, seed=0)
+
+实现MSRA初始化（a.k.a. Kaiming初始化）
+
+该类实现权重初始化方法，方法来自Kaiming He，Xiangyu Zhang，Shaoqing Ren 和 Jian Sun所写的论文: `Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification <https://arxiv.org/abs/1502.01852>`_ 。这是一个鲁棒性特别强的初始化方法，并且适应了非线性激活函数（rectifier nonlinearities）。
+
+ 在均匀分布中，范围为[-x,x]，其中：
+
+.. math::
+
+    x = \sqrt{\frac{6.0}{fan\_in}}
+
+在正态分布中，均值为0，标准差为：
+
+.. math::
+
+    \sqrt{\frac{2.0}{fan\_in}}
+
+参数：
+    - **uniform** (bool) - 是否用均匀分布或正态分布
+    - **fan_in** (float) - MSRAInitializer的fan_in。如果为None，fan_in沿伸自变量
+    - **seed** (int) - 随机种子
+
+.. note:: 
+
+    在大多数情况下推荐设置fan_in为None
+
+**代码示例**：
+
+.. code-block:: python
+
+    fc = fluid.layers.fc(
+        input=queries, size=10,
+        param_attr=fluid.initializer.MSRA(uniform=False))
+
+##################
+nets
+##################
+
+.. _cn_api_fluid_nets_glu:
+glu
+>>>>
+.. py:class:: paddle.fluid.nets.glu(input, dim=-1)
+The Gated Linear Units(GLU)由切分（split），sigmoid激活函数和按元素相乘组成。沿着给定维将input拆分成两个大小相同的部分，a和b，计算如下：
+
+.. math::
+
+    GLU(a,b) = a\bigotimes \sigma (b)
+
+参考论文: `Language Modeling with Gated Convolutional Networks <https://arxiv.org/pdf/1612.08083.pdf>`_
+
+参数：
+    - **input** (Variable) - 输入变量，张量或者LoDTensor
+    - **dim** (int) - 拆分的维度。如果 :math: `dim<0`，拆分的维为 :math: `rank(input)+dim`。默认为-1
+
+返回：变量 —— 变量的大小为输入的一半
+
+返回类型：变量（Variable）
+
+**代码示例：**
+
+.. code-block:: python
+
+    data = fluid.layers.data(name="words", shape=[3, 6, 9], dtype="float32")
+    output = fluid.nets.glu(input=data, dim=1)  # shape of output: [3, 3, 9]
+
+
+.. _cn_api_fluid_nets_sequence_conv_pool:
+sequence_conv_pool
+>>>>>>>>>>>>>>>>>>>
+
+.. py:class:: paddle.fluid.nets.sequence_conv_pool(input, num_filters, filter_size, param_attr=None, act='sigmoid', pool_type='max')
+
+sequence_conv_pool由序列卷积和池化组成
+
+参数：
+    - **input** (Variable) - sequence_conv的输入，支持变量时间长度输入序列。当前输入为shape为（T，N）的矩阵，T是mini-batch中的总时间步数，N是input_hidden_size
+    - **num_filters** （int）- 滤波器数
+    - **filter_size** （int）- 滤波器大小
+    - **param_attr** （ParamAttr) - Sequence_conv层的参数。默认：None
+    - **act** （str） - Sequence_conv层的激活函数类型。默认：sigmoid
+    - **pool_type** （str）- 池化类型。可以是max-pooling的max，average-pooling的average，sum-pooling的sum，sqrt-pooling的sqrt。默认max
+
+返回：序列卷积（Sequence Convolution）和池化（Pooling）的结果
+
+
+返回类型：变量（Variable）
+
+**代码示例**：
+
+.. code-block:: python
+
+    input_dim = len(word_dict)
+    emb_dim = 128
+    hid_dim = 512
+    data = fluid.layers.data( ame="words", shape=[1], dtype="int64", lod_level=1)
+    emb = fluid.layers.embedding(input=data, size=[input_dim, emb_dim], is_sparse=True)
+    seq_conv = fluid.nets.sequence_conv_pool(input=emb,
+                                         num_filters=hid_dim,
+                                         filter_size=3,
+                                         act="tanh",
+                                         pool_type="sqrt")
\ No newline at end of file
diff --git a/doc/fluid/book/fit_a_line/README.md b/doc/fluid/book/fit_a_line/README.md
new file mode 120000
index 0000000000000000000000000000000000000000..bd2a37a3c4f2897d1f62ec205c5e0f2f51000b11
--- /dev/null
+++ b/doc/fluid/book/fit_a_line/README.md
@@ -0,0 +1 @@
+../../../../external/book/01.fit_a_line/README.md
\ No newline at end of file
diff --git a/doc/fluid/book/fit_a_line/image b/doc/fluid/book/fit_a_line/image
new file mode 120000
index 0000000000000000000000000000000000000000..d05fb9cdba7453d935f59f0eda3519234ede4459
--- /dev/null
+++ b/doc/fluid/book/fit_a_line/image
@@ -0,0 +1 @@
+../../../../external/book/01.fit_a_line/image/
\ No newline at end of file
diff --git a/doc/fluid/book/image_classification/README.md b/doc/fluid/book/image_classification/README.md
new file mode 120000
index 0000000000000000000000000000000000000000..52413a256afad4df447e622495b62ddb7a5eef3f
--- /dev/null
+++ b/doc/fluid/book/image_classification/README.md
@@ -0,0 +1 @@
+../../../../external/book/03.image_classification/README.md
\ No newline at end of file
diff --git a/doc/fluid/book/image_classification/image b/doc/fluid/book/image_classification/image
new file mode 120000
index 0000000000000000000000000000000000000000..afd28996d9fca9c55d5ea34043ac24fcc86c213c
--- /dev/null
+++ b/doc/fluid/book/image_classification/image
@@ -0,0 +1 @@
+../../../../external/book/03.image_classification/image/
\ No newline at end of file
diff --git a/doc/fluid/book/index_en.rst b/doc/fluid/book/index_en.rst
index d7710a791f77b06b0dcffb5103b6eaaa9e1d5b68..2242e72520a098c75a33ed2ee53df30f698451f5 100644
--- a/doc/fluid/book/index_en.rst
+++ b/doc/fluid/book/index_en.rst
@@ -2,7 +2,13 @@ Book
 ======
 
 ..  toctree::
-  :maxdepth: 1
+    :maxdepth: 1
 
-  ../beginners_guide/basics/index.rst
-  ../beginners_guide/quick_start/index.rst
+    fit_a_line/README.md
+    recognize_digits/README.md
+    image_classification/README.md
+    word2vec/README.md
+    recommender_system/README.md
+    understand_sentiment/README.md
+    label_semantic_roles/README.md
+    machine_translation/README.md
diff --git a/doc/fluid/book/label_semantic_roles/README.md b/doc/fluid/book/label_semantic_roles/README.md
new file mode 120000
index 0000000000000000000000000000000000000000..0b3329adad969993a9ebf770b1fc34363bf900b2
--- /dev/null
+++ b/doc/fluid/book/label_semantic_roles/README.md
@@ -0,0 +1 @@
+../../../../external/book/07.label_semantic_roles/README.md
\ No newline at end of file
diff --git a/doc/fluid/book/label_semantic_roles/image b/doc/fluid/book/label_semantic_roles/image
new file mode 120000
index 0000000000000000000000000000000000000000..0eb76d55feab0c443e7b1fab3da870b193f166b2
--- /dev/null
+++ b/doc/fluid/book/label_semantic_roles/image
@@ -0,0 +1 @@
+../../../../external/book/07.label_semantic_roles/image/
\ No newline at end of file
diff --git a/doc/fluid/book/machine_translation/README.md b/doc/fluid/book/machine_translation/README.md
new file mode 120000
index 0000000000000000000000000000000000000000..a3507c5a5f4181e2279c1e256920b7e6c93b6c05
--- /dev/null
+++ b/doc/fluid/book/machine_translation/README.md
@@ -0,0 +1 @@
+../../../../external/book/08.machine_translation/README.md
\ No newline at end of file
diff --git a/doc/fluid/book/machine_translation/image b/doc/fluid/book/machine_translation/image
new file mode 120000
index 0000000000000000000000000000000000000000..ed4d2366ec7d3fb3d03e0e4643cf9025bed4f698
--- /dev/null
+++ b/doc/fluid/book/machine_translation/image
@@ -0,0 +1 @@
+../../../../external/book/08.machine_translation/image/
\ No newline at end of file
diff --git a/doc/fluid/book/recognize_digits/README.md b/doc/fluid/book/recognize_digits/README.md
new file mode 120000
index 0000000000000000000000000000000000000000..7106e07a423d3a8184b62da2401eb7bb3f307031
--- /dev/null
+++ b/doc/fluid/book/recognize_digits/README.md
@@ -0,0 +1 @@
+../../../../external/book/02.recognize_digits/README.md
\ No newline at end of file
diff --git a/doc/fluid/book/recognize_digits/image b/doc/fluid/book/recognize_digits/image
new file mode 120000
index 0000000000000000000000000000000000000000..925f1617634b19104d7795071cbaae7b55bd84df
--- /dev/null
+++ b/doc/fluid/book/recognize_digits/image
@@ -0,0 +1 @@
+../../../../external/book/02.recognize_digits/image/
\ No newline at end of file
diff --git a/doc/fluid/book/recommender_system/README.md b/doc/fluid/book/recommender_system/README.md
new file mode 120000
index 0000000000000000000000000000000000000000..8a1e846978fcb3d96e2d226e1baea1797cd1761c
--- /dev/null
+++ b/doc/fluid/book/recommender_system/README.md
@@ -0,0 +1 @@
+../../../../external/book/05.recommender_system/README.md
\ No newline at end of file
diff --git a/doc/fluid/book/recommender_system/image b/doc/fluid/book/recommender_system/image
new file mode 120000
index 0000000000000000000000000000000000000000..a4e97ee1ae98c330733670be080c86d5aa24fc8d
--- /dev/null
+++ b/doc/fluid/book/recommender_system/image
@@ -0,0 +1 @@
+../../../../external/book/05.recommender_system/image/
\ No newline at end of file
diff --git a/doc/fluid/book/understand_sentiment/README.md b/doc/fluid/book/understand_sentiment/README.md
new file mode 120000
index 0000000000000000000000000000000000000000..1133fe46ddd0d6c12f16dc3adc2ee7372cfb1f95
--- /dev/null
+++ b/doc/fluid/book/understand_sentiment/README.md
@@ -0,0 +1 @@
+../../../../external/book/06.understand_sentiment/README.md
\ No newline at end of file
diff --git a/doc/fluid/book/understand_sentiment/image b/doc/fluid/book/understand_sentiment/image
new file mode 120000
index 0000000000000000000000000000000000000000..c519bc94e3b50cc5bb31e7f601f6c274400345a2
--- /dev/null
+++ b/doc/fluid/book/understand_sentiment/image
@@ -0,0 +1 @@
+../../../../external/book/06.understand_sentiment/image/
\ No newline at end of file
diff --git a/doc/fluid/book/word2vec/README.md b/doc/fluid/book/word2vec/README.md
new file mode 120000
index 0000000000000000000000000000000000000000..7dfb712d27ab15cb33b2557b1b886d7d369d44f0
--- /dev/null
+++ b/doc/fluid/book/word2vec/README.md
@@ -0,0 +1 @@
+../../../../external/book/04.word2vec/README.md
\ No newline at end of file
diff --git a/doc/fluid/book/word2vec/images b/doc/fluid/book/word2vec/images
new file mode 120000
index 0000000000000000000000000000000000000000..dd59e54f0c4cf9058d375650889b52a8996d1418
--- /dev/null
+++ b/doc/fluid/book/word2vec/images
@@ -0,0 +1 @@
+../../../../external/book/04.word2vec/image/
\ No newline at end of file