diff --git a/develop/doc/_sources/howto/usage/cluster/cluster_train_en.md.txt b/develop/doc/_sources/howto/usage/cluster/cluster_train_en.md.txt
index 915405ca5b446981515e301ca4b7ee065a82a9ff..28cd1fa7903e559e33a7fc2f00172fdfbe2fdc97 100644
--- a/develop/doc/_sources/howto/usage/cluster/cluster_train_en.md.txt
+++ b/develop/doc/_sources/howto/usage/cluster/cluster_train_en.md.txt
@@ -52,7 +52,7 @@ Parameter Description
 
 - port: **required, default 7164**, port which parameter server will listen on. If ports_num greater than 1, parameter server will listen on multiple ports for more network throughput.
 - ports_num: **required, default 1**, total number of ports will listen on.
-- ports_num_for_sparse: **required, default 1**, number of ports which serves sparse parameter update.
+- ports_num_for_sparse: **required, default 0**, number of ports which serves sparse parameter update.
 - num_gradient_servers: **required, default 1**, total number of gradient servers.
 
 ### Starting trainer
@@ -98,7 +98,7 @@ Parameter Description
 - trainer_count: **required, default 1**, total count of trainers in the training job.
 - port: **required, default 7164**, port to connect to parameter server.
 - ports_num: **required, default 1**, number of ports for communication.
-- ports_num_for_sparse: **required, default 1**, number of ports for sparse type caculation.
+- ports_num_for_sparse: **required, default 0**, number of ports for sparse type caculation.
 - num_gradient_servers: **required, default 1**, total number of gradient server.
 - trainer_id: **required, default 0**, ID for every trainer, start from 0.
 - pservers: **required, default 127.0.0.1**, list of IPs of parameter servers, separated by ",".
diff --git a/develop/doc/howto/usage/cluster/cluster_train_en.html b/develop/doc/howto/usage/cluster/cluster_train_en.html
index e65bce1e95c7775ed197bc0d37c4e619abd768b7..6a9e68b486a7ad155f1dbaf01b0bab09f4a83af6 100644
--- a/develop/doc/howto/usage/cluster/cluster_train_en.html
+++ b/develop/doc/howto/usage/cluster/cluster_train_en.html
@@ -262,7 +262,7 @@ PaddlePaddle <span class="m">0</span>.10.0rc, compiled with
 <ul class="simple">
 <li>port: <strong>required, default 7164</strong>, port which parameter server will listen on. If ports_num greater than 1, parameter server will listen on multiple ports for more network throughput.</li>
 <li>ports_num: <strong>required, default 1</strong>, total number of ports will listen on.</li>
-<li>ports_num_for_sparse: <strong>required, default 1</strong>, number of ports which serves sparse parameter update.</li>
+<li>ports_num_for_sparse: <strong>required, default 0</strong>, number of ports which serves sparse parameter update.</li>
 <li>num_gradient_servers: <strong>required, default 1</strong>, total number of gradient servers.</li>
 </ul>
 </div>
@@ -303,7 +303,7 @@ python train.py
 <li>trainer_count: <strong>required, default 1</strong>, total count of trainers in the training job.</li>
 <li>port: <strong>required, default 7164</strong>, port to connect to parameter server.</li>
 <li>ports_num: <strong>required, default 1</strong>, number of ports for communication.</li>
-<li>ports_num_for_sparse: <strong>required, default 1</strong>, number of ports for sparse type caculation.</li>
+<li>ports_num_for_sparse: <strong>required, default 0</strong>, number of ports for sparse type caculation.</li>
 <li>num_gradient_servers: <strong>required, default 1</strong>, total number of gradient server.</li>
 <li>trainer_id: <strong>required, default 0</strong>, ID for every trainer, start from 0.</li>
 <li>pservers: <strong>required, default 127.0.0.1</strong>, list of IPs of parameter servers, separated by &#8221;,&#8221;.</li>
diff --git a/develop/doc_cn/_sources/howto/usage/cluster/cluster_train_cn.md.txt b/develop/doc_cn/_sources/howto/usage/cluster/cluster_train_cn.md.txt
index 9328b4cca45f77516e45d2f6e0bd7cb6873aa95d..c2fc86687d7106aac7c74d6dd16bc229353cb7c1 100644
--- a/develop/doc_cn/_sources/howto/usage/cluster/cluster_train_cn.md.txt
+++ b/develop/doc_cn/_sources/howto/usage/cluster/cluster_train_cn.md.txt
@@ -51,7 +51,7 @@ $ stdbuf -oL /usr/bin/nohup paddle pserver --port=7164 --ports_num=1 --ports_num
 
 - port：**必选，默认7164**，pserver监听的起始端口，根据ports_num决定总端口个数，从起始端口监听多个端口用于通信
 - ports_num：**必选，默认1**，监听的端口个数
-- ports_num_for_sparse：**必选，默认1**，用于稀疏类型参数通信的端口个数
+- ports_num_for_sparse：**必选，默认0**，用于稀疏类型参数通信的端口个数
 - num_gradient_servers：**必选，默认1**，当前训练任务pserver总数
 
 ### 启动计算节点
@@ -95,7 +95,7 @@ paddle.init(
 - trainer_count：**必选，默认1**，当前训练任务trainer总个数
 - port：**必选，默认7164**，连接到pserver的端口
 - ports_num：**必选，默认1**，连接到pserver的端口个数
-- ports_num_for_sparse：**必选，默认1**，和pserver之间用于稀疏类型参数通信的端口个数
+- ports_num_for_sparse：**必选，默认0**，和pserver之间用于稀疏类型参数通信的端口个数
 - num_gradient_servers：**必选，默认1**，当前训练任务pserver总数
 - trainer_id：**必选，默认0**，每个trainer的唯一ID，从0开始的整数
 - pservers：**必选，默认127.0.0.1**，当前训练任务启动的pserver的IP列表，多个IP使用“,”隔开
diff --git a/develop/doc_cn/howto/usage/cluster/cluster_train_cn.html b/develop/doc_cn/howto/usage/cluster/cluster_train_cn.html
index 36e61c712ba628f2e445fc25943f77fcd2d79e2f..bb31a7c310788760a12a42e6b35f9948722c5e2a 100644
--- a/develop/doc_cn/howto/usage/cluster/cluster_train_cn.html
+++ b/develop/doc_cn/howto/usage/cluster/cluster_train_cn.html
@@ -275,7 +275,7 @@ PaddlePaddle <span class="m">0</span>.10.0, compiled with
 <ul class="simple">
 <li>port：<strong>必选，默认7164</strong>，pserver监听的起始端口，根据ports_num决定总端口个数，从起始端口监听多个端口用于通信</li>
 <li>ports_num：<strong>必选，默认1</strong>，监听的端口个数</li>
-<li>ports_num_for_sparse：<strong>必选，默认1</strong>，用于稀疏类型参数通信的端口个数</li>
+<li>ports_num_for_sparse：<strong>必选，默认0</strong>，用于稀疏类型参数通信的端口个数</li>
 <li>num_gradient_servers：<strong>必选，默认1</strong>，当前训练任务pserver总数</li>
 </ul>
 </div>
@@ -315,7 +315,7 @@ PaddlePaddle <span class="m">0</span>.10.0, compiled with
 <li>trainer_count：<strong>必选，默认1</strong>，当前训练任务trainer总个数</li>
 <li>port：<strong>必选，默认7164</strong>，连接到pserver的端口</li>
 <li>ports_num：<strong>必选，默认1</strong>，连接到pserver的端口个数</li>
-<li>ports_num_for_sparse：<strong>必选，默认1</strong>，和pserver之间用于稀疏类型参数通信的端口个数</li>
+<li>ports_num_for_sparse：<strong>必选，默认0</strong>，和pserver之间用于稀疏类型参数通信的端口个数</li>
 <li>num_gradient_servers：<strong>必选，默认1</strong>，当前训练任务pserver总数</li>
 <li>trainer_id：<strong>必选，默认0</strong>，每个trainer的唯一ID，从0开始的整数</li>
 <li>pservers：<strong>必选，默认127.0.0.1</strong>，当前训练任务启动的pserver的IP列表，多个IP使用“,”隔开</li>