h2o-pysparkling-3.1

What it is and what it does

PySparkling is a Python library that bridges H2O's machine learning engine with Apache Spark, enabling distributed model training and scoring across Spark clusters. It provides APIs to work with H2O-3 models and MOJO (Model Object, Optimized) artifacts, which are H2O's portable, language-agnostic model format. The package also supports scoring Driverless AI MOJO models.

The library is designed for data scientists and engineers working in big-data environments who need to integrate H2O's statistical and machine learning capabilities into Spark-based workflows. It has no runtime dependencies listed in the fact sheet, meaning it relies entirely on an external Spark installation and H2O runtime to function. The package is classified as Production/Stable and supports Python 3.6 through 3.10, though the last release was in November 2024.

Use it for:

• Train H2O machine learning models on large distributed datasets using Spark's cluster computing.
• Score pre-trained MOJO models in batch or streaming Spark jobs for inference at scale.
• Integrate Driverless AI MOJO models into Spark-based data pipelines for automated predictions.
• Build end-to-end machine learning workflows combining Spark data preparation with H2O model training.
• Deploy H2O models to production Spark clusters for real-time or batch scoring on big data.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

PySparkling provides Python bindings to train and score H2O-3 and Sparkling Water machine learning models on Apache Spark, with support for MOJO model formats.

Yes, with conditions. PySparkling is worth installing if you are already committed to an Apache Spark infrastructure and need H2O's machine learning capabilities in that environment. The permissive Apache license and Production/Stable status support production use. However, high install friction, aging maintenance (633 days since last release), and the requirement for a pre-configured Spark cluster mean this is not a lightweight addition—evaluate whether H2O's specific algorithms justify the operational overhead in your stack.

Install

h2o-pysparkling-3-1 on PyPI

pip install h2o-pysparkling-3-1

uv add h2o-pysparkling-3-1

poetry add h2o-pysparkling-3-1

Installing h2o-pysparkling-3.1

Before you install

Installation friction is high due to the package's large compiled artifact (h2o_pysparkling_3.1-3.46.0.6.post1.tar.gz). The package is in aging maintenance status with the last release 633 days old, though the repository remains active with recent commits as of 2025-11-05.

License in practice

Licensed under Apache v2 (permissive), which allows commercial use, modification, and distribution with minimal restrictions—suitable for most production environments.

Quickstart

pip install h2o-pysparkling-3.1==3.46.0.6.post1

from pysparkling.ml import H2OMOJOModel
model = H2OMOJOModel.load('path/to/model.mojo')

Requires Apache Spark to be installed and configured; PySparkling is a Spark integration layer and cannot function standalone.

Verify before relying

• Exact Python version compatibility beyond the classifiers (3.6–3.10 listed); whether newer Python versions are supported.
• Whether the 633-day gap since last release indicates active maintenance or dormancy; commit date alone does not clarify release cadence.
• Performance characteristics and scalability limits for large MOJO model scoring workloads.
• Whether Driverless AI MOJO scoring support is feature-complete or has known limitations.

Package facts

Evidence: h2o_pysparkling_3.1-3.46.0.6.post1.tar.gz

Keywords: machine learning, data mining, statistical analysis, modeling, big data, distributed, parallel

Tags

More Build Tools packages