Implementations of the Feedzai OpenML API to allow support for machine learning models in Java.
All models may be used in nodes with AMD64 and ARM64 cpu architectures except for H2O-xgboost that can be used only with AMD64 cpu architectures.
This is a Maven project which you can build using the following command:
mvn clean installIf the build fails compiling for ARM64 you may need to run:
docker run --rm --privileged multiarch/qemu-user-static --reset -p yesTo use the models on Operating Systems with musl (supported only for AMD64 architectures), the ENV variable FDZ_OPENML_JAVA_LIBC
must be set to musl. This variable can take the values glibc and musl. H2O-xgboost can't be used with musl.
For all releases, as the hotfix branch is ready all that's needed to actually release is to create an annotated tag pointing to the hotfix branch head.
The project is maintained in two parallel release lineages, each tracked by its own branch, so that clients on different Java baselines can keep receiving features:
| Lineage | Branch | Java baseline | Purpose |
|---|---|---|---|
| 3.X | master |
Java 25 | Default lineage. Carries the Java 25 changes. |
| 2.X | hf-2.X |
pre-Java 25 (Java 8+) | Hotfix lineage for clients that cannot yet run Java 25. |
Both lineages ship the same providers (for example, the native XGBoost provider); the only difference is the
Java baseline each one targets. When preparing a release, make the tag on the branch that matches the lineage:
master for 3.x.y tags and hf-2.X for 2.x.y tags.
For a public release you need to create and push an annotated tag. See the example below for releasing version 1.2.29:
# Ensure the tag is made on the updated branch
git fetch -a
git checkout origin/hf-1.2.X
git tag -a 1.2.29
# Your EDITOR will open. Write a good message and save as it is used on Github as a release message
git push origin 1.2.29Alternatively, create a new release directly in GitHub, choosing
"Create new tag on publish": the tag it creates is lightweight, so the deploy workflow annotates
plain x.y.z tags before building (on the runner only; the tag in the repository is not changed).
In both cases create (or edit) the release for this tag on GitHub to publish the release notes.
There is a "Generate release notes" button available to create the release notes, just organize them according to the previous ones.
For experimental releases, the process is the same as above but with a lightweight tag instead of annotated. The tag name must have a
suffix (for example 1.5.0-xp): the deploy workflow annotates plain x.y.z tags, which makes them public releases.
#the xp suffix is needed: a plain x.y.z tag is turned into a public release by the deploy workflow. The tag must not be annotated
git tag 1.5.0-xp
git push origin 1.5.0-xpThe openml-h2o module contains a provider that allows to load and train models with H2O.
Pull the provider from Maven Central:
<dependency>
<groupId>com.feedzai</groupId>
<artifactId>openml-h2o</artifactId>
<!-- See project tags for latest version -->
<version>1.2.0</version>
</dependency>The openml-datarobot module contains a provider that allows to load models trained with DataRobot.
Pull this module from Maven Central:
<dependency>
<groupId>com.feedzai</groupId>
<artifactId>openml-datarobot</artifactId>
<!-- See project tags for latest version -->
<version>1.2.0</version>
</dependency>The openml-lightgbm module contains a provider that allows to load models trained with Microsoft LightGBM.
Pull this module from Maven Central:
<dependency>
<groupId>com.feedzai</groupId>
<artifactId>openml-lightgbm</artifactId>
<!-- See project tags for latest version -->
<version>1.2.0</version>
</dependency>