MLeap is a serialization format and execution engine for machine learning pipelines. You can find the package at the MLeap GitHub repo. It is an open source package and employs the Apache License 2.0.
- Only Databricks Runtime 6.4 Extended Support and Databricks Runtime 5.5 LTS support the open source version of MLeap.
- A custom version of MLeap is included in all versions of Databricks Runtime for Machine Learning, a Databricks Runtime runtime that provides a ready-to-go environment for machine learning and data science. To use MLeap with Databricks Runtime 7.3 LTS and above, you must create a cluster running Databricks Runtime ML.
- MLeap does not support XGBoost4J-Spark models in Databricks Runtime 10.0 ML.
The following notebook provides guidance on installing MLeap on Databricks and give an example of a model export workflow.