PSA: TFLite Artifacts Moved To MavenCentral

[Xunkai Zhang]

Hi everyone,

As JCenter & Bintray are no longer the recommended services for hosting binaries for opensource projects (Bintray sunset announcement), we have moved all TensorFlow Lite prebuilt artifacts to the Maven Central Repository, and will use Maven Central exclusively for future releases (including the 2.5.0 release). We will not push new artifacts to Bintray/JCenter in the future.

JFrog has stated that artifacts already hosted on JCenter will be accessible until 2022. However, we strongly encourage all clients to migrate to using the MavenCentral artifacts promptly, particularly if you reference nightly builds that are no longer actively updated.

For Gradle users, please follow the instructions below to use packages on Maven Central.

• To use versioned packages on Maven Central, you will need to change the jcenter() in your root Gradle build file to mavenCentral().

• To use nightly built snapshots on Maven Central, you will need an extra configuration explicitly with the Url to Sonatype snapshot repo.

 

// root build.gradle

 

allprojects {

  repositories {

    google()

    mavenCentral ()  // For versioned releases

    maven {          // For snapshot releases (optional)

      name 'ossrh-snapshot'

      url 'http://oss.sonatype.org/content/repositories/snapshots'

    }

  }

}

 

You don’t need to change other parts of the Gradle configuration.

For users who build TFLite-dependent projects with other tools, please follow the respective tool instructions for setting up remote repositories.

The change applies to all relevant TensorFlow Lite artifacts, including:

• tensorflow-lite

• tensorflow-lite-gpu

• tensorflow-lite-select-tf-ops

• tensorflow-lite-hexagon

• tensorflow-lite-metadata

• tensorflow-lite-support

• tensorflow-lite-task-base

• tensorflow-lite-task-vision

• tensorflow-lite-task-text

If you have any trouble using or referencing the packages, please let us know @ tfl...@tensorflow.org. 

Thanks!

Xunkai, on behalf of the TensorFlow Lite team