This module contains the core API of the Deep Java Library (DJL) project. It includes the following packages:
- engine - Contains classes to load a deep learning engine
- inference - Contains classes to implement inference tasks
- metric - Contains classes to collect metrics information
- modality - Contains utility classes for each of the predefined modalities
- ndarray - Contains classes and interfaces to define an n-dimensional array
- nn - Contains classes to define neural network operations
- training - Contains classes to implement training tasks
- translate - Contains classes and interfaces to translate between java objects and NDArrays
The latest javadocs can be found on here.
You can also build the latest javadocs locally using the following command:
# for Linux/macOS:
./gradlew javadoc
# for Windows:
..\gradlew javadoc
The javadocs output is built in the build/doc/javadoc folder.
You can pull the DJL API from the central Maven repository by including the following dependency in your pom.xml
file:
<dependency>
<groupId>ai.djl</groupId>
<artifactId>api</artifactId>
<version>0.29.0</version>
</dependency>
For testing the current nightly build, use the following:
<dependency>
<groupId>ai.djl</groupId>
<artifactId>api</artifactId>
<version>0.30.0-SNAPSHOT</version>
</dependency>
Note that the nightly build is under active development and might contain a number of bugs and instabilities.