Skip to content

Commit

Permalink
[DOCS] 1.1.8 to 1.0.0
Browse files Browse the repository at this point in the history
  • Loading branch information
andyyehoo committed Jun 14, 2017
1 parent f8eb682 commit 6766410
Show file tree
Hide file tree
Showing 4 changed files with 24 additions and 24 deletions.
2 changes: 1 addition & 1 deletion docs/deploy/local_run.md
Original file line number Diff line number Diff line change
Expand Up @@ -6,7 +6,7 @@

* Hadoop >= 2.2.0
* Java 1.8版本
* Angel发布包 angel-1.1.8-bin.zip
* Angel发布包 angel-1.0.0-bin.zip

配置好HADOOP_HOME和JAVA_HOME环境变量,解压Angel发布包,就可以以LOCAL模式运行Angel任务了。

Expand Down
34 changes: 17 additions & 17 deletions docs/deploy/run_on_yarn.md
Original file line number Diff line number Diff line change
Expand Up @@ -4,17 +4,17 @@

由于业界很多公司的大数据平台,都是基于Yarn搭建,所以Angel目前的分布式运行是基于Yarn,方便用户复用现网环境,而无需任何修改。

1. **运行环境准备**
1. **运行环境准备**

Angel的分布式Yarn运行模式需要的环境,其实也非常简单:

1. 一个可以正常运行Hadoop集群,包括Yarn和HDFS
* Hadoop >= 2.2.0

2. 一个用于提交Angel任务的客户端Gateway
* Java >= 1.8
* 可以提交MR作业
* Angel发布包:angel-1.1.8-bin.zip
* Angel发布包:angel-1.0.0-bin.zip


### 2. **Angel任务运行示例**
Expand All @@ -35,7 +35,7 @@
hadoop fs -put data/exampledata/LRLocalExampleData/a9a.train hdfs://my-nn:54310/test/lr_data
```
2. **提交任务**

* 在发布包的bin目录下有Angel的提交脚本angel-submit,使用它将任务提交到Hadoop集群

```
Expand All @@ -49,18 +49,18 @@
--ml.feature.num 1024 \
--angel.job.name LR_test
```

**参数含义如下**


| 名称 | 含义 |
| --- | --- |
| action.type | 计算类型,目前支持"train"和"predict"两种,分别表示模型训练和预测 |
| angel.app.submit.class | 算法运行类,每个算法都对应一个运行类|
| angel.train.data.path | 训练数据路径 |
| angel.log.path | 算法指标日志输出路径 |
| angel.save.model.path | 模型保存路径 |
| ml.data.type | 训练数据格式,默认支持两种格式libsvm和dummy |
| action.type | 计算类型,目前支持"train"和"predict"两种,分别表示模型训练和预测 |
| angel.app.submit.class | 算法运行类,每个算法都对应一个运行类|
| angel.train.data.path | 训练数据路径 |
| angel.log.path | 算法指标日志输出路径 |
| angel.save.model.path | 模型保存路径 |
| ml.data.type | 训练数据格式,默认支持两种格式libsvm和dummy |
| ml.feature.num | 模型维度 |
| angel.job.name | 任务名|

Expand All @@ -74,15 +74,15 @@


任务提交之后,会在控制台打印出任务运行信息,如URL和迭代进度等,如下图所示:

![][1]

打开URL信息就可以看到Angel任务每一个组件的详细运行信息和算法相关日志:

![][2]

目前的监控页面有点简陋,后续会进一步优化,围绕PS的本质,提高美观程度和用户可用度


[1]: ../img/angel_client_log.png
[2]: ../img/lr_worker_log.png
[1]: ../img/angel_client_log.png
[2]: ../img/lr_worker_log.png
8 changes: 4 additions & 4 deletions docs/deploy/source_compile.md
Original file line number Diff line number Diff line change
Expand Up @@ -12,12 +12,12 @@
```git clone https://github.com/tencent/angel```

3. **编译**

进入源码根目录,执行命令:

```mvn clean package -Dmaven.test.skip=true```
编译完成后,在源码根目录`dist/target`目录下会生成一个发布包:`angel-1.1.8-bin.zip`

编译完成后,在源码根目录`dist/target`目录下会生成一个发布包:`angel-1.0.0-bin.zip`

4. **发布包**

Expand Down
4 changes: 2 additions & 2 deletions docs/tutorials/spark_on_angel_quick_start.md
Original file line number Diff line number Diff line change
Expand Up @@ -36,7 +36,7 @@ $SPARK_HOME/bin/spark-submit \
--executor-cores 2 \
--executor-memory 4g \
--class com.tencent.angel.spark.examples.ml.BreezeSGD \
spark-on-angel-examples-1.1.8.jar
spark-on-angel-examples-1.0.0.jar
```
- YARN将会出现两个Application,一个是Spark Application, 一个是Angel-PS Application。

Expand Down Expand Up @@ -72,4 +72,4 @@ println("feature sum:" + w.mkRemote.pull())

gradient.delete()
w.delete()
```
```

0 comments on commit 6766410

Please sign in to comment.