Databricks Runtime 10.4 LTS for Machine Learning

Databricks Runtime 10.4 LTS for Machine Learning provides a ready-to-go environment for machine learning and data science based on Databricks Runtime 10.4 LTS. Databricks Runtime ML contains many popular machine learning libraries, including TensorFlow, PyTorch, and XGBoost. Databricks Runtime ML includes AutoML, a tool to automatically train machine learning pipelines. Databricks Runtime ML also supports distributed deep learning training using Horovod.

For more information, including instructions for creating a Databricks Runtime ML cluster, see AI and machine learning on Databricks.

Databricks Runtime 10.4 LTS ML is built on top of Databricks Runtime 10.4 LTS. For information on what's new in Databricks Runtime 10.4 LTS, including Apache Spark MLlib and SparkR, see the Databricks Runtime 10.4 LTS release notes.

The following enhancements have been made to AutoML.

Starting with Databricks Runtime 10.4 LTS ML, AutoML is generally available.

You can now specify how null values are imputed. By default, AutoML selects an imputation method based on the column type and content. See Impute missing values for details.

For classification and regression problems, you can now use the UI in addition to the API to specify columns that AutoML should ignore during its calculations. See Column selection.

AutoML now supports numerical array types.

You can now specify a location in the workspace where AutoML should save generated notebooks and experiments. Use the experiment_dir parameter. See AutoML Python API reference.

The following enhancements have been made to Databricks Feature Store.

• You can now register an existing Delta table as a feature table.

The system environment in Databricks Runtime 10.4 LTS ML differs from Databricks Runtime 10.4 LTS as follows:

• DBUtils: Databricks Runtime ML does not include Library utility (dbutils.library) (legacy). Use %pip commands instead. See Notebook-scoped Python libraries.
• For GPU clusters, Databricks Runtime ML includes the following NVIDIA GPU libraries:
  • CUDA 11.0
  • cuDNN 8.0.5.39
  • NCCL 2.10.3
  • TensorRT 7.2.2

The following sections list the libraries included in Databricks Runtime 10.4 LTS ML that differ from those included in Databricks Runtime 10.4 LTS.

• Top-tier libraries
• Python libraries
• R libraries
• Java and Scala libraries (Scala 2.12 cluster)

Databricks Runtime 10.4 LTS ML includes the following top-tier libraries:

• GraphFrames
• Horovod and HorovodRunner
• MLflow
• PyTorch
• spark-tensorflow-connector
• TensorFlow
• TensorBoard

Databricks Runtime 10.4 LTS ML uses Virtualenv for Python package management and includes many popular ML packages.

In addition to the packages specified in the in the following sections, Databricks Runtime 10.4 LTS ML also includes the following packages:

• hyperopt 0.2.7.db1
• sparkdl 2.2.0-db5
• feature_store 0.3.8
• automl 1.7.2

To reproduce the Databricks Runtime ML Python environment in your local Python virtual environment, download the requirements-10.4.txt file and run pip install -r requirements-10.4.txt. This command installs all of the open source libraries that Databricks Runtime ML uses, but does not install Azure Databricks developed libraries, such as databricks-automl, databricks-feature-store, or the Databricks fork of hyperopt.

The R libraries are identical to the R Libraries in Databricks Runtime 10.4 LTS.

In addition to Java and Scala libraries in Databricks Runtime 10.4 LTS, Databricks Runtime 10.4 LTS ML contains the following JARs: