Third-Party Machine Learning Integrations

This section provides instructions and examples of how to install, configure, and run some of the most popular third-party ML tools in Databricks. Databricks provides these examples on a best-effort basis. Because they are external libraries, they may change in ways that are not easy to predict. If you need additional support for third-party tools, consult the documentation, mailing lists, forums, or other support options provided by the library vendor or maintainer.

H2O Sparkling Water

H2O is an open source project for distributed machine learning. This section describes how to integrate H2O using the Sparkling Water module.

H2O Flow

The H2O Flow UI provides user-friendly clickable interface to machine learning and also provides some useful visualizations to view your jobs. To enable H2O Flow on a Databricks cluster, you must set up an ssh tunnel to the Apache Spark driver. After H2O starts, run the following on your Spark driver:

ssh ubuntu@<hostname> -p 2200 -i <private-key> -L 54321:localhost:54321

You should be able to access H2O Flow on localhost:54321.

Python notebook

Note

Databricks Runtime for Machine Learning installs XGBoost, which conflicts with the XGBoost packaged in PySparkling. To use PySparkling on Databricks Runtime ML, you must remove XGBoost using an init script:

#!/bin/bash

rm /databricks/jars/spark--maven-trees--ml--xgboost*

H2O Sparking Water Python notebook

scikit-learn

scikit-learn, a well-known Python machine learning library, is included in Databricks Runtime. See Databricks Runtime Release Notes for the scikit-learn library version included with your cluster’s runtime.

DataRobot

The DataRobot modeling engine is a commercial product that supports massively parallel modeling applications, building and optimizing models of many different types, and evaluating and ranking their relative performance. This modeling engine exists in a variety of implementations, some cloud-based, accessed via the Internet, and others residing in customer-specific on-premises computing environments. Read more at DataRobot.

XGBoost

XGBoost is a popular machine learning library designed specifically for training decision trees and random forests. You can train XGBoost models on individual machines or in a distributed fashion. Read more in the XGBoost documentation.

Warning

If you use XGBoost 0.90 for training and the training job fails, the shared Spark context will be killed and the only way to recover is to restart the cluster. This is a bug in XGBoost.

Install XGBoost

Install XGBoost on Databricks Runtime ML

XGBoost is included in Databricks Runtime ML. The Python version is included in Databricks Runtime 5.4 ML and above. You can use these libraries in Databricks Runtime ML without installing any packages. See Databricks Runtime for Machine Learning. The XGBoost versions included are:

Databricks Runtime ML Version Python Version Scala/Java Version
5.5 LTS 0.90 0.90
5.4 0.80 0.81
5.1 - 5.3 N/A 0.81
5.0 N/A 0.81

To install other Python versions in Databricks Runtime ML, install XGBoost as a Databricks PyPI library. Specify it as the following and replace <xgboost version> with the desired version.

xgboost==<xgboost version>

Install XGBoost on Databricks Runtime

  • Python package: Use Databricks Library Utilities by executing the command below in a notebook cell and replace <xgboost version> with the desired version:

    dbutils.library.installPyPI("xgboost", version="<xgboost version>" )
    
  • Scala/Java packages: Install as a Databricks library with the Spark Package name xgboost-linux64.

Single node training in Python

The Python package allows you to train only single node workloads.

Distributed training in Scala

To perform distributed training, you must use XGBoost’s Scala/Java packages.

Integrate XGBoost with ML pipelines

XGBoost classification notebook

Integrate XGBoost with cross validation

XGBoost regression notebook