diff --git a/core/engine/cluster/cloud/k8s_cpu_job.sh.template b/core/engine/cluster/cloud/k8s_cpu_job.sh.template
index c5203fcad76b28b5a48de62067b46f4ed5bf1696..2889cd1d55008f22b7e9fb854019f996a4746f8c 100644
--- a/core/engine/cluster/cloud/k8s_cpu_job.sh.template
+++ b/core/engine/cluster/cloud/k8s_cpu_job.sh.template
@@ -9,7 +9,7 @@ job_name=<$ JOB_NAME $>
 group_name="<$ GROUP_NAME $>"               
 job_version="paddle-fluid-v1.7.1"
 start_cmd="<$ START_CMD $>"
-wall_time="10:00:00"
+wall_time="2000:00:00"
 
 k8s_priority=<$ K8S_PRIORITY $>
 k8s_trainers=<$ K8S_TRAINERS $>
diff --git a/core/engine/cluster/cloud/k8s_job.sh.template b/core/engine/cluster/cloud/k8s_job.sh.template
index 9886f11aebbbe547ed1fb433a35c653e2a77f6f3..8314e9efd0ec349bb00e28605386e34dfc601102 100644
--- a/core/engine/cluster/cloud/k8s_job.sh.template
+++ b/core/engine/cluster/cloud/k8s_job.sh.template
@@ -9,7 +9,7 @@ job_name=<$ JOB_NAME $>
 group_name="<$ GROUP_NAME $>"               
 job_version="paddle-fluid-v1.7.1"
 start_cmd="<$ START_CMD $>"
-wall_time="10:00:00"
+wall_time="2000:00:00"
 
 k8s_priority=<$ K8S_PRIORITY $>
 k8s_trainers=<$ K8S_TRAINERS $>
diff --git a/core/engine/cluster/cloud/mpi_job.sh.template b/core/engine/cluster/cloud/mpi_job.sh.template
index 46d68d2130d591c86f4a587000498c139c1e74aa..b3a3c20a02094cca68c96f527bf29d3150996228 100644
--- a/core/engine/cluster/cloud/mpi_job.sh.template
+++ b/core/engine/cluster/cloud/mpi_job.sh.template
@@ -9,7 +9,7 @@ job_name=<$ JOB_NAME $>
 group_name=<$ GROUP_NAME $>
 job_version="paddle-fluid-v1.7.1"
 start_cmd="<$ START_CMD $>"
-wall_time="2:00:00"
+wall_time="2000:00:00"
 
 # 你的ak/sk（可在paddlecloud web页面【个人中心】处获取）
 ak=<$ AK $>
diff --git a/doc/distributed_train.md b/doc/distributed_train.md
index 0f63678220adb81d68dc181dc4d517a40823155e..e27e1f7eb1d2519eedc670d0021e09a1b0bfd17a 100644
--- a/doc/distributed_train.md
+++ b/doc/distributed_train.md
@@ -35,9 +35,9 @@
 
 分布式运行首先需要更改`config.yaml`，主要调整以下内容：
 
-- workspace: 调整为在远程点运行时的工作目录，一般设置为`"./"`即可
-- runner_class: 从单机的"train"调整为"cluster_train"，单机训练->分布式训练（例外情况，k8s上单机单卡训练仍然为train）
-- fleet_mode: 选则参数服务器模式(ps)，抑或GPU的all-reduce模式(collective)
+- workspace: 调整为在远程节点运行时的工作目录，一般设置为`"./"`即可
+- runner_class: 从单机的"train"调整为"cluster_train"，单机训练->分布式训练（例外情况，k8s上单机单卡训练仍然为train，后续支持）
+- fleet_mode: 选择参数服务器模式(ps)，或者GPU的all-reduce模式(collective)
 - distribute_strategy: 可选项，选择分布式训练的策略，目前只在参数服务器模式下生效，可选项:`sync、asycn、half_async、geo`
 
 配置选项具体参数，可以参考[yaml配置说明](./yaml.md)
@@ -306,7 +306,7 @@ dataset:
 - name: dataloader_train 
   batch_size: 2
   type: DataLoader 
-  data_path: "{workspace}/train_data"
+  data_path: "{workspace}/afs/挂载数据文件夹的路径"
   sparse_slots: "click 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26"
   dense_slots: "dense_var:13"
 
@@ -390,7 +390,7 @@ dataset:
 - name: dataloader_train 
   batch_size: 2
   type: DataLoader 
-  data_path: "{workspace}/train_data"
+  data_path: "{workspace}/afs/挂载数据文件夹的路径"
   sparse_slots: "click 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26"
   dense_slots: "dense_var:13"