combust/mleap

MLeap is a machine learning deployment framework that serializes, stores, and executes ML models outside of their original training environments. It provides a unified bundle format to persist and load models from diverse ecosystems, including Spark MLlib, Scikit-Learn, TensorFlow, and XGBoost. The system enables high-performance inference via a core executor, supporting remote serving through gRPC, HTTP, and Spring Boot integrations. By standardizing model representation, it allows seamless model portability and execution across different runtime environments.

The project established a comprehensive serialization framework for MLeap, introducing support for JSON, Protobuf, and Avro formats while removing legacy bundle mechanisms. This period also significantly expanded parity testing and model support for Spark MLlib transformers, ensuring functional consistency with the Spark ecosystem.

48 changes

This period focused on expanding the MLeap ecosystem with a comprehensive Python SDK, enabling serialization and deserialization for Scikit-Learn, PySpark, and TensorFlow models. The work also introduced new Spark ML transformers, improved tensor serialization, and added XGBoost support, significantly broadening the framework's interoperability and deployment capabilities.

31 changes

This period focused on expanding MLeap's serving capabilities with the introduction of the MLeap Executor, gRPC, and Spring Boot integrations for remote model inference. It also added support for storing and retrieving model bundles via HDFS and S3, alongside extensive test coverage for new and existing features.

27 changes

combust/mleap was measured the same way every project in this corpus was: the same rubric, at a pinned commit, with the result published in full. Point the surveyor at a repository you know and see whether you agree with it.