MLeap Bundle Serialization

Serialization and deserialization is fully functional for all Spark and MLeap transformers. These is partial support for Scikit-learn serialization and Tensorflow serialization/deserialization is still in planning. Once you have exported an MLeap Bundle, as long as all of the operations in it are supported by your target platform, you can load it into that platform for use. In its current state, this means that you can serialize from Spark, PySpark, Scikit-learn or MLeap and deserialize the same pipeline back into Spark, PySpark or MLeap. It's also worth noting that Tensorflow graphs are supported by MLeap, so you can include them in your ML pipelines.

Serialization Formats

Format Description
JSON Serialize all attributes and models as JSON
Protobuf Serialize all attributes and models as protobuf objects
Mixed Serialize small attributes and models as JSON, all large attributes and models will be serialized as Protobuf.

results matching ""

    No results matching ""