This repository contains the wrapper for both the stand-alone and servlet-compatible DL4j UiServers.
To build it from source, the following commands should be run:
git clone https://github.com/deeplearning4j/UiServer.git
cd UiServer
mvn clean install -DskipTests=true
As result, you'll see a few files in your target folder, including:
deeplearning4j-ui-dropwizard.jar
deeplearning4j-ui-servlet.war
The first file is a stand-alone executable jar, with an embedded http server powered by DropWizard.
The other is a standard WAR package, which can be deployed to any servlet server compatible with a Java EE environment.
To use a remote UiServer with your DL4j tasks, here's some example code:
...
MultiLayerNetwork model = new MultiLayerNetwork(conf);
model.init();
Note here that the port is 8250 please override this in the dropwizard.yml under src/main/resources
to 0 if you want dynamic binding. We do this so we can make assumptions about how to run ports.
UiConnectionInfo connectionInfo = new UiConnectionInfo.Builder()
.setAddress("8.8.8.8")
.setPort(8080)
.setPath("deeplearning4j-ui-servlet")
.build();
model.setListeners(Collections.singletonList((IterationListener) new HistogramIterationListener(connectionInfo,listenerFreq)));
In this example, we assume that the external web service is used located at 8.8.8.8:8080, and a servlet is deployed at deeplearning4j-ui-servlet path.
If you're going to use a dropwizard-based executable, then the setPath() call should be omitted, since everything will be deployed within the root folder.
Obviously, the network address passed into UiConnectionInfo should be accessible from the host where DL4j is running.