1 # TensorFlow for Java 2 3 > *WARNING*: The TensorFlow Java API is not currently covered by the TensorFlow 4 > [API stability guarantees](https://www.tensorflow.org/programmers_guide/version_semantics). 5 > 6 > For using TensorFlow on Android refer instead to 7 > [contrib/android](https://www.tensorflow.org/code/tensorflow/contrib/android), 8 > [makefile](https://www.tensorflow.org/code/tensorflow/contrib/makefile#android) 9 > and/or the [Android demo](https://www.tensorflow.org/code/tensorflow/examples/android). 10 11 ## Quickstart 12 13 - Refer to [Installing TensorFlow for Java](https://www.tensorflow.org/install/install_java) 14 - [Javadoc](https://www.tensorflow.org/api_docs/java/reference/org/tensorflow/package-summary) 15 - [](https://maven-badges.herokuapp.com/maven-central/org.tensorflow/tensorflow) 16 17 ## Building from source 18 19 If the quickstart instructions above do not work out, the TensorFlow Java and 20 native libraries will need to be built from source. 21 22 1. Install [bazel](https://www.bazel.build/versions/master/docs/install.html) 23 24 2. Setup the environment to build TensorFlow from source code 25 ([Linux](https://www.tensorflow.org/install/install_sources#PrepareLinux) 26 or [Mac OS 27 X](https://www.tensorflow.org/install/install_sources#PrepareMac)). 28 If you'd like to skip reading those details and do not care about GPU 29 support, try the following: 30 31 ```sh 32 # On Linux 33 sudo apt-get install python swig python-numpy 34 35 # On Mac OS X with homebrew 36 brew install swig 37 ``` 38 39 3. [Configure](https://www.tensorflow.org/install/install_sources#configure_the_installation) 40 (e.g., enable GPU support) and build: 41 42 ```sh 43 ./configure 44 bazel build --config opt \ 45 //tensorflow/java:tensorflow \ 46 //tensorflow/java:libtensorflow_jni 47 ``` 48 49 The command above will produce two files in the `bazel-bin/tensorflow/java` 50 directory: 51 52 * An archive of Java classes: `libtensorflow.jar` 53 * A native library: `libtensorflow_jni.so` on Linux, `libtensorflow_jni.dylib` 54 on OS X, or `tensorflow_jni.dll` on Windows. 55 56 To compile Java code that uses the TensorFlow Java API, include 57 `libtensorflow.jar` in the classpath. For example: 58 59 ```sh 60 javac -cp bazel-bin/tensorflow/java/libtensorflow.jar ... 61 ``` 62 63 To execute the compiled program, include `libtensorflow.jar` in the classpath 64 and the native library in the library path. For example: 65 66 ```sh 67 java -cp bazel-bin/tensorflow/java/libtensorflow.jar \ 68 -Djava.library.path=bazel-bin/tensorflow/java \ 69 ... 70 ``` 71 72 Installation on Windows requires the more experimental [bazel on 73 Windows](https://bazel.build/versions/master/docs/windows.html). Details are 74 omitted here, but find inspiration in the script used for building the release 75 archive: 76 [`tensorflow/tools/ci_build/windows/libtensorflow_cpu.sh`](https://www.tensorflow.org/code/tensorflow/tools/ci_build/windows/libtensorflow_cpu.sh). 77 78 ### Maven 79 80 Details of the release process for Maven Central are in 81 [`maven/README.md`](https://www.tensorflow.org/code/tensorflow/java/maven/README.md). 82 However, for development, you can push the library built from source to a local 83 Maven repository with: 84 85 ```sh 86 bazel build -c opt //tensorflow/java:pom 87 mvn install:install-file \ 88 -Dfile=../../bazel-bin/tensorflow/java/libtensorflow.jar \ 89 -DpomFile=../../bazel-bin/tensorflow/java/pom.xml 90 ``` 91 92 And then refer to this library in a project's `pom.xml` with: (replacing 93 VERSION with the appropriate version of TensorFlow): 94 95 ```xml 96 <dependency> 97 <groupId>org.tensorflow</groupId> 98 <artifactId>libtensorflow</artifactId> 99 <version>VERSION</version> 100 </dependency> 101 ``` 102 103 ### Bazel 104 105 If your project uses bazel for builds, add a dependency on 106 `//tensorflow/java:tensorflow` to the `java_binary` or `java_library` rule. For 107 example: 108 109 ```sh 110 bazel run -c opt //tensorflow/java/src/main/java/org/tensorflow/examples:label_image 111 ``` 112
