You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@singa.apache.org by wa...@apache.org on 2016/03/29 06:41:20 UTC
svn commit: r1736944 - in
/incubator/singa/site/trunk/content/markdown/docs/kr: architecture.md
distributed-training.md frameworks.md overview.md programming-guide.md
quick-start.md
Author: wangwei
Date: Tue Mar 29 04:41:19 2016
New Revision: 1736944
URL: http://svn.apache.org/viewvc?rev=1736944&view=rev
Log:
update korean docs
Modified:
incubator/singa/site/trunk/content/markdown/docs/kr/architecture.md
incubator/singa/site/trunk/content/markdown/docs/kr/distributed-training.md
incubator/singa/site/trunk/content/markdown/docs/kr/frameworks.md
incubator/singa/site/trunk/content/markdown/docs/kr/overview.md
incubator/singa/site/trunk/content/markdown/docs/kr/programming-guide.md
incubator/singa/site/trunk/content/markdown/docs/kr/quick-start.md
Modified: incubator/singa/site/trunk/content/markdown/docs/kr/architecture.md
URL: http://svn.apache.org/viewvc/incubator/singa/site/trunk/content/markdown/docs/kr/architecture.md?rev=1736944&r1=1736943&r2=1736944&view=diff
==============================================================================
--- incubator/singa/site/trunk/content/markdown/docs/kr/architecture.md (original)
+++ incubator/singa/site/trunk/content/markdown/docs/kr/architecture.md Tue Mar 29 04:41:19 2016
@@ -7,42 +7,41 @@
<img src = "../../images/logical.png" style="width:550px"/>
<p> <strong> Fig.1 - ìì¤í
ìí¤í
ì² </strong> </p>
-SINGAë ë¤ìí ë¶ì° [í¸ë ì´ë íë ììí¬](frameworks.html) (ë기 ëë ë¹ë기 í¸ë ì´ë)ì ì§ìíë ì ì°í 구조를 ê°ì§ê³ ììµëë¤.
-Fig.1. ìì¤í
ì 구조를 ë³´ì¬ì¤ëë¤.
-í¹ì§ì¼ë¡ë ì¬ë¬ server 그룹과 worker 그룹ì ê°ì§ê³ ìë¤.
+SINGAë ë¤ìí ë¶ì° [í¸ë ì´ë íë ììí¬](frameworks.html) (ë기 í¹ì ë¹ë기 í¸ë ì´ë)ì ì§ìíë 구조를 ê°ì§ê³ ììµëë¤.
+í¹ì§ì¼ë¡ë ì¬ë¬ server그룹 ê³¼ worker그룹ì ê°ì§ê³ ììµëë¤.
* **Server 그룹**
-  Server 그룹ì ëª¨ë¸ ë§¤ê° ë³ìì ë³µì 본ì ê°ì§ê³ worker 그룹ì ìì²ì ë°ë¼ ë§¤ê° ë³ìì ì
ë°ì´í¸ë¥¼ ë´ë¹í©ëë¤. ì¸ì í server 그룹ë¤ì ë§¤ê° ë³ì를 ì 기ì ì¼ë¡ ë기íí©ëë¤. ì¼ë°ì ì¼ë¡ íëì server 그룹ì ì¬ë¬ serverë¡ êµ¬ì±ë ê° serverë ëª¨ë¸ ë§¤ê° ë³ìì ë¶í ë ë¶ë¶ì ë´ë¹í©ëë¤.
+ Server그룹ì ëª¨ë¸ ë§¤ê° ë³ìì ë³µì 를 ê°ì§ê³ worker그룹ì ìì²ì ë°ë¼ ë§¤ê° ë³ìì ì
ë°ì´í¸ë¥¼ ë´ë¹í©ëë¤. ì¸ì ë server그룹ë¤ì ë§¤ê° ë³ì를 ì 기ì ì¼ë¡ ë기íí©ëë¤. ì¼ë°ì ì¼ë¡ íëì server그룹ì ì¬ë¬ serverë¡ êµ¬ì±ëë©°, ê° serverë ëª¨ë¸ ë§¤ê° ë³ìì ë¶í ë ë¶ë¶ì ë´ë¹í©ëë¤.
* **Worker 그룹**
-Â Â ê° worker 그룹ì íëì server 그룹과 íµì í©ëë¤. íëì worker 그룹ì ë§¤ê° ë³ìì 기ì¸ê¸° ê³ì°ì ë´ë¹í©ëë¤. ëí ë¶í ë ë°ì´í°ì ì¼ë¶ì ëí´ "ìì í"ëª¨ë¸ ë³µì 본ì í¸ë ì´ëí©ëë¤. 모ë worker 그룹ë¤ì í´ë¹ server 그룹ë¤ê³¼ ë¹ë기 ì ì¼ë¡ íµì í©ëë¤. ê·¸ë¬ë ê°ì worker 그룹ì workerë¤ì ë기íí©ëë¤.
+ ê° worker그룹ì íëì server그룹과 íµì í©ëë¤. íëì worker그룹ì ë§¤ê° ë³ìì 구배ê³ì°ì ë´ë¹í©ëë¤. ëí ë¶í ë ì¼ë¶ ë°ì´í°ë¥¼ ì¨ì "ì ì²´"모ë¸ì ë í리카를 í¸ë ì´ëí©ëë¤. 모ë worker그룹ë¤ì í´ë¹ server그룹ë¤ê³¼ ë¹ë기 íµì ì í©ëë¤. ê·¸ë¬ë ê°ì worker그룹ì workerë¤ì ë기í©ëë¤.
ëì¼ ê·¸ë£¹ ë´ìì workerë¤ì ë¶ì° í¸ë ì´ëìë ë§ì ë°©ë²ì´ ììµëë¤.
* **ëª¨ë¸ ë³ë ¬í**
- ê° worker 그룹ì ë°°ì ë 모ë ë°ì´í°ì ëí´ ë§¤ê° ë³ìì ë¶ë¶ ì§í©ì ê³ì°í©ëë¤.
+ ê° worker그룹ì ë°°ì ë 모ë ë°ì´í°ì ëí´ ë§¤ê° ë³ìì ë¶ë¶ ì§í©ì ê³ì°í©ëë¤.
* **ë°ì´í° ë³ë ¬í**
- ê° workerë ë°°ë¶ ë ë°ì´í°ì ë¶ë¶ ì§í©ì ëí´ ëª¨ë ë§¤ê° ë³ì를 ê³ì°í©ëë¤.
+ ê° workerë ë°°ë¶ ë ë°ì´í°ì ë¶ë¶ ì§í©ì ëí´ ëª¨ë ë§¤ê° ë³ì를 ê³ì°í©ëë¤.
* **íì´ë¸ë¦¬ë ë³ë ¬í**
- ìì ë°©ë²ì ì¡°í©í íì´ë¸ë¦¬ë ë³ë ¬í를 ì§ìí©ëë¤.
+ ìì ë°©ë²ì ì¡°í©í íì´ë¸ë¦¬ë ë³ë ¬í를 ì§ìí©ëë¤.
## ìí리ë©í
ì´ì
-SINGAìì servers ì workersë ë¤ë¥¸ ì¤ë ëìì ìì§ì´ë ì¤í ì ëì
ëë¤.
+SINGAìì servers ì workersë ë¤ë¥¸ ì¤ë ëìì ì¤íëë ì ëì
ëë¤.
-ì´ë¤ì [messages](communication.html)를 ì´ì©íì¬ íµì í©ëë¤.
-모ë íë¡ì¸ì¤ë ë©ì¸ ì¤ë ëì¸ stubì ì¤ííì¬ ë¡ì»¬ messages를 ìì§íê³ ê·¸ê²ì ëìíë receiverì ì ì¡í©ëë¤.
+ì´ë¤ì [messages](communication.html)를 ì´ì©íì¬ íµì ì í©ëë¤.
+모ë íë¡ì¸ì¤ë ë©ì¸ ì¤ë ëì¸ stubì ì¤ííì¬ ë¡ì»¬ messages를 ìì§íê³ ëìëë receiverì ì ì¡í©ëë¤.
-ê° server 그룹과 worker 그룹ì "ì ì²´"ëª¨ë¸ ë íë¦¬ì¹´ì¸ *ParamShard* 를 ê°ì§ëë¤.
-ë§ì½ workers ì servers ê° ëì¼í íë¡ì¸ì¤ìì ì¤íëë¤ë©´,
+ê° server그룹과 worker그룹ì "ì ì²´"ëª¨ë¸ ë íë¦¬ì¹´ì¸ *ParamShard* 를 ê°ì§ëë¤.
+ë§ì½ workers ì servers ê° ëì¼í íë¡ì¸ì¤ìì ì¤í ëë¤ë©´,
ê·¸ *ParamShard* (íí°ì
)ì ë©ëª¨ë¦¬ ê³µê°ì ê³µì íëë¡ ì¤ì ë©ëë¤.
ì´ ê²½ì° ë¤ë¥¸ ì¤í ì ë ì¬ì´ë¥¼ ì¤ê°ë messagesë íµì ë¹ì©ì ì¤ì´ê¸° ìí´ ë°ì´í°ì í¬ì¸í° ë§ í¬í¨í©ëë¤.
-íë¡ì¸ì¤ ê° íµì ì ê²½ì°ìë ë¬ë¦¬ messsagesë íë¼ë¯¸í° ê°ì í¬í¨í©ëë¤.
+íë¡ì¸ì¤ ê° íµì ì ê²½ì°ìë ë¬ë¦¬ messagesë íë¼ë¯¸í° ê°ì í¬í¨í©ëë¤.
Modified: incubator/singa/site/trunk/content/markdown/docs/kr/distributed-training.md
URL: http://svn.apache.org/viewvc/incubator/singa/site/trunk/content/markdown/docs/kr/distributed-training.md?rev=1736944&r1=1736943&r2=1736944&view=diff
==============================================================================
--- incubator/singa/site/trunk/content/markdown/docs/kr/distributed-training.md (original)
+++ incubator/singa/site/trunk/content/markdown/docs/kr/distributed-training.md Tue Mar 29 04:41:19 2016
@@ -14,7 +14,7 @@ SINGAë ëê·ëª¨ ë°ì�
ëª¨ë¸ í¸ë ì´ëì ë³ë ¬íí기 ìí´ ë¤ìí ë³ë ¬ë°©ì (ë°ì´í° ë³ë ¬, ëª¨ë¸ ë³ë ¬, íì´ë¸ë¦¬ë ë³ë ¬ ë±)ì ì§ìí©ëë¤.
-* íì´ë¸ë¦¬ë ë³ë ¬í (hybrid.html)
+* [íì´ë¸ë¦¬ë ë³ë ¬í](hybrid.html)
íì¬ SINGAë Mesosê³¼ íµí©ëì´ ì기 ë문ì ë¶ì° í¸ë ì´ëì Mesos íë ììí¬ë¡ ì¤íí ì ììµëë¤.
Mesos í´ë¬ì¤í°ë SINGA 컨í
ì´ëìì ì¤ì í ì ììµëë¤.
Modified: incubator/singa/site/trunk/content/markdown/docs/kr/frameworks.md
URL: http://svn.apache.org/viewvc/incubator/singa/site/trunk/content/markdown/docs/kr/frameworks.md?rev=1736944&r1=1736943&r2=1736944&view=diff
==============================================================================
--- incubator/singa/site/trunk/content/markdown/docs/kr/frameworks.md (original)
+++ incubator/singa/site/trunk/content/markdown/docs/kr/frameworks.md Tue Mar 29 04:41:19 2016
@@ -1,13 +1,13 @@
-# Distributed Training Framework
+# ë¶ì° í¸ë ì´ë
---
-## Cluster Topology Configuration
+## Cluster Topology ì¤ì
-Here we describe how to configure SINGA's cluster topology to support
-different distributed training frameworks.
-The cluster topology is configured in the `cluster` field in `JobProto`.
-The `cluster` is of type `ClusterProto`:
+SINGA ìì ë¤ìí ë¶ì° í¸ë ì´ë íë ììí¬ë¥¼ ì¤ííë ë°©ë²ì ì¤ëª
í©ëë¤.
+
+cluster topology ë `JobProto` ìì `cluster` field 를 ì¤ì í´ì¤ëë¤.
+`cluster` ë `ClusterProto` íì
ì
ëë¤. ì를 ë¤ì´
message ClusterProto {
optional int32 nworker_groups = 1;
@@ -24,32 +24,28 @@ The `cluster` is of type `ClusterProto`:
}
-The mostly used fields are as follows:
+ì주 ì¬ì©ëë field ë ë¤ìê³¼ ê°ìµëë¤:
+
+* `nworkers_per_group` and `nworkers_per_procs`:
+ decide the partitioning of worker side ParamShard.
+
+* `nservers_per_group` and `nservers_per_procs`:
+ decide the partitioning of server side ParamShard.
+
+* `server_worker_separate`:
+ separate servers and workers in different processes.
+
+## ë¤ìí í¸ë ì´ë íë ììí¬
- * `nworkers_per_group` and `nworkers_per_procs`:
- decide the partitioning of worker side ParamShard.
- * `nservers_per_group` and `nservers_per_procs`:
- decide the partitioning of server side ParamShard.
- * `server_worker_separate`:
- separate servers and workers in different processes.
-
-## Different Training Frameworks
-
-In SINGA, worker groups run asynchronously and
-workers within one group run synchronously.
-Users can leverage this general design to run
-both **synchronous** and **asynchronous** training frameworks.
-Here we illustrate how to configure
-popular distributed training frameworks in SINGA.
+SINGA ìì worker groups ë¤ì ë¹ë기ì ì¼ë¡, group ììì workers ë¤ì ë기ì ì¼ë¡ ì¤íë©ëë¤. ì ì ë ì´ ì¼ë°ëì ì¸ì ì´ì©í´ì **synchronous** ì **asynchronous** í¸ë ì´ë íë ììí¬ë¥¼ ì¤í í ì ììµëë¤. ë리 ìë ¤ì§ ë¶ì° í¸ë ì´ëì ì´ë»ê² ì¤ì íê³ ì¤ííëì§ ì¤ëª
íê² ìµëë¤.
<img src="../../images/frameworks.png" style="width: 800px"/>
-<p><strong> Fig.1 - Training frameworks in SINGA</strong></p>
+<p><strong> Fig.1 - ë¤ìí í¸ë ì´ë íë ììí¬</strong></p>
###Sandblaster
-This is a **synchronous** framework used by Google Brain.
-Fig.2(a) shows the Sandblaster framework implemented in SINGA.
-Its configuration is as follows:
+Google Brain ìì ì°ì´ë **synchronous** íë ììí¬.
+Fig.2(a) ë SINGAìì Sandblaster íë ììí¬ë¥¼ ì¤íí기 ìí cluster ì ì¤ì ìì
ëë¤.
cluster {
nworker_groups: 1
@@ -59,16 +55,14 @@ Its configuration is as follows:
server_worker_separate: true
}
-A single server group is launched to handle all requests from workers.
-A worker computes on its partition of the model,
-and only communicates with servers handling related parameters.
+ê° server group ë 모ë workers ì requests 를 ì²ë¦¬í©ëë¤.
+ê° worker ë ë´ë´ë¤í¸ 모ë¸ì í ë¶ë¶ì ë´ë¹íì¬ ê³ì°ì íê³ , 모ë servers ì íµì ì íì¬ ê´ë ¨ parameters ê°ì ì»ìµëë¤.
###AllReduce
-This is a **synchronous** framework used by Baidu's DeepImage.
-Fig.2(b) shows the AllReduce framework implemented in SINGA.
-Its configuration is as follows:
+Baidu's DeepImage ìì ì°ì´ë **synchronous** íë ììí¬.
+Fig.2(b) ë SINGAìì AllReduce íë ììí¬ë¥¼ ì¤íí기 ìí cluster ì ì¤ì ìì
ëë¤.
cluster {
nworker_groups: 1
@@ -78,15 +72,12 @@ Its configuration is as follows:
server_worker_separate: false
}
-We bind each worker with a server on the same node, so that each
-node is responsible for maintaining a partition of parameters and
-collecting updates from all other nodes.
+ê° node ìì 1 worker ì 1 server 를 ì¤ííì¬, ê° node ê° parameters ì í ë¶ë¶ì ë´ë¹íê³ ê³ì°ì íëë¡ ì¤ì í©ëë¤. ë¤ë¥¸ nodes ì ì
ë°ì´í¸ ë ì 보를 êµíí©ëë¤.
###Downpour
-This is a **asynchronous** framework used by Google Brain.
-Fig.2(c) shows the Downpour framework implemented in SINGA.
-Its configuration is as follows:
+Google Brain ìì ì°ì´ë **asynchronous** íë ììí¬.
+Fig.2(c) ë SINGAìì Downpour íë ììí¬ë¥¼ ì¤íí기 ìí cluster ì ì¤ì ìì
ëë¤.
cluster {
nworker_groups: 2
@@ -96,16 +87,12 @@ Its configuration is as follows:
server_worker_separate: true
}
-Similar to the synchronous Sandblaster, all workers send
-requests to a global server group. We divide workers into several
-worker groups, each running independently and working on parameters
-from the last *update* response.
+synchronous Sandblaster ì ë¹ì·íê², 모ë workers ë 1 server group ì requests 를 ë³´ë
ëë¤. ì¬ê¸°ìë workers ë¤ì ì¬ë¬ worker groups ì¼ë¡ ëëì´ì, ê° worker ê° *update* reply ìì ë°ì ìµì parameters 를 ì¨ì ê³ì° íëë¡ ì¤ì íììµëë¤.
###Distributed Hogwild
-This is a **asynchronous** framework used by Caffe.
-Fig.2(d) shows the Distributed Hogwild framework implemented in SINGA.
-Its configuration is as follows:
+Caffe ìì ì°ì´ë **asynchronous** íë ììí¬.
+Fig.2(d) ë SINGAìì Hogwild íë ììí¬ë¥¼ ì¤íí기 ìí cluster ì ì¤ì ìì
ëë¤.
cluster {
nworker_groups: 3
@@ -115,8 +102,5 @@ Its configuration is as follows:
server_worker_separate: false
}
-Each node contains a complete server group and a complete worker group.
-Parameter updates are done locally, so that communication cost
-during each training step is minimized.
-However, the server group must periodically synchronize with
-neighboring groups to improve the training convergence.
+ê° node ë 1 server group ì 1 worker group 를 ì¤íí©ëë¤.
+Parameter updates 를 node ìì ê°ê° ì¤íìí´ì¼ë¡ì¨ íµì ì½ì¤í¸ì í¸ë ì´ë ì¤í
ì ìµìí í©ëë¤. ê·¸ë¬ë server groups ë¤ì ì 기ì ì¼ë¡ ë¤ì´ë²ë§ groups ë¤ê³¼ ë기 ìì¼ì¼ ë©ëë¤.
Modified: incubator/singa/site/trunk/content/markdown/docs/kr/overview.md
URL: http://svn.apache.org/viewvc/incubator/singa/site/trunk/content/markdown/docs/kr/overview.md?rev=1736944&r1=1736943&r2=1736944&view=diff
==============================================================================
--- incubator/singa/site/trunk/content/markdown/docs/kr/overview.md (original)
+++ incubator/singa/site/trunk/content/markdown/docs/kr/overview.md Tue Mar 29 04:41:19 2016
@@ -24,8 +24,11 @@ SINGAë ëê·ëª¨ ë°ì�
íì¥ì±ì ë¶ì° ë¥ë¬ëìì ì¤ìí ì°êµ¬ ê³¼ì ì
ëë¤.
SINGAë ë¤ìí í¸ë ì´ë íë ììí¬ì íì¥ì±ì ì ì§í ì ìëë¡ ì¤ê³ëì´ ììµëë¤.
+
* Synchronous (ë기) : í¸ë ì´ëì 1ë¨ê³ìì ì»ì ììë í¨ê³¼ë¥¼ ëì
ëë¤.
+
* Asynchronous (ë¹ë기) : í¸ë ì´ëì ìë ´ ìë를 í¥ììíµëë¤.
+
* Hybrid (íì´ë¸ë¦¬ë) : ì½ì¤í¸ ë° ë¦¬ìì¤ (í´ë¬ì¤í° í¬ê¸° ë±)ì ë§ë í¨ê³¼ì ìë ´ ìëì ê· íì ì¡ê³ íì¥ì±ì í¥ììíµëë¤.
SINGAë ë¥ë¬ë 모ë¸ì ë¤í¸ìí¬ "ë ì´ì´" ê°ë
ì ë°ë¼ ì§ê´ì ì¼ë¡ íë¡ê·¸ëë°ì í ì ìëë¡ ëìì¸ëì´ ììµëë¤. ë¤ìí 모ë¸ì ì½ê² 구ì¶íê³ í¸ë ì´ë í ì ììµëë¤.
@@ -33,16 +36,16 @@ SINGAë ë¥ë¬ë 모ë�
## ìì¤í
ê°ì
<img src = "../../images/sgd.png" align="center" width="400px"/>
-<span> <strong> Figure 1 - SGD Flow </strong> </span>
+<span><strong> Figure 1 - SGD Flow </strong></span>
-"ë¥ë¬ë 모ë¸ì íìµíë¤"ë ê²ì í¹ì ìì
(ë¶ë¥, ì측 ë±)ì ë¬ì±í기 ìíì¬ ì¬ì©ëë í¹ì§ë(feature)ì ìì±íë ë³í í¨ìì ìµì íë¼ë¯¸í°ë¥¼ ì°¾ë ê²ì
ëë¤.
+"ë¥ë¬ë 모ë¸ì íìµíë¤"ë ê²ì í¹ì ìì
(ë¶ë¥, ì측 ë±)ì ë¬ì±í기 ìíì¬ ì¬ì©ëë í¹ì§ë(feature)ì ìì±íë ë³í í¨ìì ìµì íë¼ë¯¸í°ë¥¼ ì°¾ëë¤ë ê²ì
ëë¤.
ê·¸ ë³ìì ì¢ê³ ëì¨ì, Cross-Entropy Loss (https://en.wikipedia.org/wiki/Cross_entropy) ë±ì loss function (ìì¤ í¨ì)ì¼ë¡ íì¸í©ëë¤. ì´ í¨ìë ì¼ë°ì ì¼ë¡ ë¹ì í ëë ë¹ ë³¼ë¡ í¨ìì´ë¯ë¡ é解ì ì°¾ê¸°ê° íëëë¤.
ê·¸ëì Stochastic Gradient Descent (íë¥ ì 구배ê°íë²)ì ì´ì©í©ëë¤.
-Figure 1ê³¼ ê°ì´ 무ììë¡ ì´ê¸°í ë íë¼ë¯¸í° ê°ì ìì¤ í¨ìê° ìì ì§ëë¡ ë°ë³µ ì
ë°ì´í¸íê³ ììµëë¤.
+Figure 1ê³¼ ê°ì´ ëë¤ì¼ë¡ ì´ê¸°í ë íë¼ë¯¸í° ê°ì, ìì¤ í¨ìê° ìì ì§ëë¡ ë°ë³µ ì
ë°ì´í¸íê³ ììµëë¤.
<img src = "../../images/overview.png" align="center" width="400px"/>
-<span> <strong> Figure 2 - SINGA ê°ì </strong> </span>
+<span> <strong> Figure 2 - SINGA Overview </strong> </span>
í¸ë ì´ëì íìí ìí¬ë¡ëë workers ì servers ì ë¶ì°ë©ëë¤.
Figure 2ì ê°ì´ 루í(iteration)ë§ë¤ workers ë *TrainOneBatch* í¨ì를 ë¶ë¬ì íë¼ë¯¸í° 구배를 ê³ì°í©ëë¤.
Modified: incubator/singa/site/trunk/content/markdown/docs/kr/programming-guide.md
URL: http://svn.apache.org/viewvc/incubator/singa/site/trunk/content/markdown/docs/kr/programming-guide.md?rev=1736944&r1=1736943&r2=1736944&view=diff
==============================================================================
--- incubator/singa/site/trunk/content/markdown/docs/kr/programming-guide.md (original)
+++ incubator/singa/site/trunk/content/markdown/docs/kr/programming-guide.md Tue Mar 29 04:41:19 2016
@@ -9,31 +9,23 @@ Figure 1ì ê·¸ë ¤ì§ ë¤
* [Updater](updater.html) : serverìì ë§¤ê° ë³ì를 ì
ë°ì´í¸íë ë°©ë²ì 기ì í©ëë¤.
* [Cluster Topology](distributed-training.html) : workersì servers ë¶ì° í í´ë¡ì§ë¥¼ 기ì í©ëë¤.
-The *Basic user guide* section describes how to submit a training job using
-built-in components; while the *Advanced user guide* section presents details
-on writing user's own main function to register components implemented by
-themselves. In addition, the training data must be prepared, which has the same
-[process](data.html) for both advanced users and basic users.
+*Basic ì ì ê°ì´ë* ìì built-in components 를 ì¨ì í¸ë ì´ëì ììíë ë°©ë²ì ì¤ëª
í©ëë¤. *Advanced ì ì ê°ì´ë* ììë ì ì ê° ìí리ë©í¸í 모ë¸, í¨ì, ìê³ ë¦¬ë¬ì ì¨ì í¸ë ì´ëì ììíë ë°©ë²ì ì¤ë³í©ëë¤. í¸ë ì´ë ë°ì´íë [process](data.html) 를 ì°¸ê³ ë¡ ì¤ë¹ë¥¼ í´ì£¼ì¸ì.
<img src="../../images/overview.png" align="center" width="400px"/>
-<span><strong>Figure 1 - SINGA ê°ì </strong></span>
+<span><strong>Figure 1 - SINGA Overview </strong></span>
## Basic ì ì ê°ì´ë
-Users can use the default main function provided SINGA to submit the training
-job. For this case, a job configuration file written as a google protocol
-buffer message for the [JobProto](../api/classsinga_1_1JobProto.html) must be provided in the command line,
+SINGA ìì ì¤ë¹ë main í¨ì를 ì¨ì ì½ê² í¸ë ì´ëì ììí ì ììµëë¤.
+ì´ ê²½ì° [JobProto](../api/classsinga_1_1JobProto.html) 를 ìíì¬ google protocol buffer message ë¡ ìì¬ì§ job configuration íì¼ì ì¤ë¹í©ëë¤. ê·¸ë¦¬ê³ ìëì 커맨ëë¼ì¸ì ì¤íí©ëë¤.
./bin/singa-run.sh -conf <path to job conf> [-resume]
-`-resume` is for continuing the training from last
-[checkpoint](checkpoint.html).
-The [MLP](mlp.html) and [CNN](cnn.html)
-examples use built-in components. Please read the corresponding pages for their
-job configuration files. The subsequent pages will illustrate the details on
-each component of the configuration.
+`-resume` ë í¸ë ì´ëì ì ë² [checkpoint](checkpoint.html) ë¶í° ë¤ì ê³ìí ë ì°ë ì¸ì ì
ëë¤.
+[MLP](mlp.html) ì [CNN](cnn.html) ìíë¤ì built-in ì»´í¬ëí¸ë¥¼ ì°ê³ ììµëë¤.
+Please read the corresponding pages for their job configuration files. The subsequent pages will illustrate the details on each component of the configuration.
## Advanced ì ì ê°ì´ë
@@ -42,11 +34,11 @@ If a user's model contains some user-def
register these components. It is similar to Hadoop's main function. Generally,
the main function should
- * initialize SINGA, e.g., setup logging.
+* SINGA ì´ê¸°í, e.g., setup logging.
- * register user-defined components.
+* ì ì ì»´í¬ëí¸ì ë±ë¡
- * create and pass the job configuration to SINGA driver
+* job configuration ì ìì±íê³ SINGA driver ìì ì¤ì
main í¨ìì ìíì
ëë¤.
@@ -71,13 +63,9 @@ main í¨ìì ìíì
return 0;
}
-The Driver class' `Init` method will load a job configuration file provided by
-users as a command line argument (`-conf <job conf>`). It contains at least the
-cluster topology and returns the `jobConf` for users to update or fill in
-configurations of neural net, updater, etc. If users define subclasses of
-Layer, Updater, Worker and Param, they should register them through the driver.
-Finally, the job configuration is submitted to the driver which starts the
-training.
+Driver class' `Init` method ë 커맨ëë¼ì¸ ì¸ì `-conf <job conf>` ìì 주ì´ì§ job configuration íì¼ì ì½ìµëë¤. ê·¸ íì¼ìë cluster topology ì ë³´ê° ê¸°ì ëì´ìê³ , ì ì ê° neural net, updater ë±ì ì
ë°ì´í¸ í¹ì ì¤ì í기ìí `jobConf`를 리í´í©ëë¤.
+ì ì ê° Layer, Updater, Worker, Param ë±ì subclass를 ì ìíë©´, driver ì ë±ë¡ì í´ì¼í©ëë¤.
+í¸ë ì´ëì ììí기 ìíì¬ job configuration ì¦ `jobConf`를 driver.Train ì ë겨ì¤ëë¤.
We will provide helper functions to make the configuration easier in the
future, like [keras](https://github.com/fchollet/keras).
@@ -89,5 +77,4 @@ path of the *mysinga* and base job confi
./bin/singa-run.sh -conf <path to job conf> -exec <path to mysinga> [other arguments]
-The [RNN application](rnn.html) provides a full example of
-implementing the main function for training a specific RNN model.
+[RNN application](rnn.html) ìì RNN 모ë¸ì í¸ë ì´ëì ìí í¨ìì íë¡ê·¸ë¨ ì를 ì¤ëª
í©ëë¤.
Modified: incubator/singa/site/trunk/content/markdown/docs/kr/quick-start.md
URL: http://svn.apache.org/viewvc/incubator/singa/site/trunk/content/markdown/docs/kr/quick-start.md?rev=1736944&r1=1736943&r2=1736944&view=diff
==============================================================================
--- incubator/singa/site/trunk/content/markdown/docs/kr/quick-start.md (original)
+++ incubator/singa/site/trunk/content/markdown/docs/kr/quick-start.md Tue Mar 29 04:41:19 2016
@@ -115,7 +115,7 @@ Cluster Topologyì 기본ê°�
ì¤í¬ë¦½í¸ ì¤í :
-Â Â Â Â ./bin/singa-run.sh -conf examples/cifar10/job.conf
+ ./bin/singa-run.sh -conf examples/cifar10/job.conf
#### ë기í ë³ë ¬ í¸ë ì´ë
@@ -165,7 +165,7 @@ e.g.,
ì¤í¬ë¦½í¸ì ì¤íì "Single ë
¸ë í¸ë ì´ë"ê³¼ ëì¼í©ëë¤.
-Â Â Â Â ./bin/singa-run.sh -conf examples/cifar10/job.conf
+ ./bin/singa-run.sh -conf examples/cifar10/job.conf
## Mesosìì ì¤í