Practical overview
When it fits
Use MLeap when a supported Spark or scikit-learn pipeline needs to be serialized and executed in a JVM environment separately from the framework used to train it.
Who it is for
- Data scientists exporting pipelines trained with Spark or scikit-learn.
- Engineers integrating exported machine-learning pipelines into JVM applications or services.
Documented uses
- Export a fitted Spark pipeline into a portable bundle.
- Run an exported pipeline with MLeap Runtime without requiring Spark, scikit-learn, NumPy or pandas in the runtime environment.
- Use MLeap's Spark, PySpark or scikit-learn support in a pipeline workflow.
Requirements & limits
The project is a public repository licensed under Apache-2.0. Its documented setup uses SBT or Maven dependencies for JVM projects, or a PyPI package for PySpark integration.
- The documented runtime target is the JVM.
- MLeap supports JSON and Protobuf serialization formats.
- The README publishes tested combinations of MLeap, Spark, Scala, Java, Python, XGBoost and TensorFlow versions; other combinations should be checked against project documentation.
- Spark integration is added separately through the mleap-spark dependency.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Add MLeap Runtime to a Scala project
Set up the MLeap runtime dependency in an SBT or Maven project.
Before you start
- A project using SBT or Maven.
- Choose a MLeap version and Scala artifact that match the project; the repository lists tested compatibility configurations.
Steps
For SBT, add
libraryDependencies += "ml.combust.mleap" %% "mleap-runtime" % "0.25.2".For Maven, add the dependency
ml.combust.mleap:mleap-runtime_2.13:0.25.2.Consult the compatibility matrix and getting-started documentation when selecting versions or Scala artifacts.
Expected outcome
The project is configured to include MLeap Runtime, which provides facilities to execute and serialize ML pipelines.
Guide limitations
- MLeap Runtime does not include facilities for training ML pipelines.
- The dependency examples use version 0.25.2 and Scala 2.13; align versions with the project’s chosen compatibility configuration.
Add MLeap Spark integration to a Scala project
Set up the MLeap Spark library dependency for serializing Spark-trained pipelines to MLeap Bundles.
Before you start
- A Scala project using SBT or Maven.
- A Spark and Scala version compatible with the selected MLeap release.
Steps
For SBT, add
libraryDependencies += "ml.combust.mleap" %% "mleap-spark" % "0.25.2".For Maven, add the dependency
ml.combust.mleap:mleap-spark_2.13:0.25.2.Check the repository’s dependency compatibility matrix and Spark getting-started documentation for version selection.
Expected outcome
The project is configured with MLeap Spark integration for serializing Spark-trained ML pipelines to MLeap Bundles.
Guide limitations
- The dependency setup alone does not train or export a pipeline.
- The examples use version 0.25.2 and Scala 2.13; compatibility depends on the selected Spark, Scala, and MLeap versions.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 13 October 2026
- Original website title
- combust/mleap · Official homepage title · retained in its source language