NOTE: For the latest stable README.md ensure you are on the main branch.
The RAPIDS Accelerator for Apache Spark provides a set of plugins for Apache Spark that leverage GPUs to accelerate processing via the RAPIDS libraries.
Documentation on the current release can be found here.
To get started and try the plugin out use the getting started guide.
The SQL plugin tries to produce results that are bit for bit identical with Apache Spark. Operator compatibility is documented here
To get started tuning your job and get the most performance out of it please start with the tuning guide.
The plugin has a set of Spark configs that control its behavior and are documented here.
We use github to track bugs, feature requests, and answer questions. File an issue for a bug or feature request. Ask or answer a question on the discussion board.
The jar files for the most recent release can be retrieved from the download page.
See the build instructions in the contributing guide.
Tests are described here.
The RAPIDS Accelerator For Apache Spark does provide some APIs for doing zero copy data transfer into other GPU enabled applications. It is described here.
Currently, we are working with XGBoost to try to provide this integration out of the box.
You may need to disable RMM caching when exporting data to an ML library as that library will likely want to use all of the GPU's memory and if it is not aware of RMM it will not have access to any of the memory that RMM is holding.
The Qualification and Profiling tools have been moved to nvidia/spark-rapids-tools repo.
Please refer to Qualification tool documentation and Profiling tool documentation for more details on how to use the tools.
If you need to develop some functionality on top of RAPIDS Accelerator For Apache Spark (we currently
limit support to GPU-accelerated UDFs) we recommend you declare our distribution artifact
as a provided dependency.
<dependency>
<groupId>com.nvidia</groupId>
<artifactId>rapids-4-spark_2.12</artifactId>
<version>24.10.1</version>
<scope>provided</scope>
</dependency>