Suite of libraries for data analytics and machine learning on GPUs.


5.5.2022 Due to Mahti's update to Red Hat Enterprise Linux 8 (RHEL8), older versions of RAPIDS are no longer fully supported. Please contact our servicedesk if you really need access to older versions.

4.2.2022 All old RAPIDS versions which were based on direct Conda installations have been deprecated, and we encourage users to move to newer versions. Read more on our separate Conda deprecation page.


Rapids is available on Puhti and Mahti. Some older versions are available on Puhti only. Currently supported RAPIDS versions:

  • 22.04 using Singularity, based on RAPIDS official Docker images: 22.04
  • 21.12 using Singularity, based on RAPIDS official Docker images: 21.12 (Puhti only)
  • 0.16 using Singularity: 0.16-sng (Puhti only)
  • 0.15 using Singularity: 0.15-sng (Puhti only)
  • 0.14 using Singularity: 0.14-sng (Puhti only)
  • (deprecated) 0.11 (Puhti only)

Contains the RAPIDS suite (including cuDF, cuML, cuGraph, and XGBoost) for Python with GPU support via CUDA.

Modules ending with -sng are based on NVIDIA's optimized container images from NGC.

If you find that some package is missing, you can often install it yourself with pip install --user. See our Python documentation for more information on how to install packages yourself. If you think that some important RAPIDS-related package should be included in the module provided by CSC, please contact our servicedesk.

Some modules are Singularity-based. Wrapper scripts have been provided so that common commands such as python, python3, pip and pip3 should work as normal. For more information, see CSC's general instructions on how to run Singularity containers.


RAPIDS is licensed under Apache License 2.0


To use this software, initialize it with:

module load rapids

to access the default version.

This will show all available versions:

module avail rapids

To check the exact packages and versions included in the loaded module you can run:



Note that login nodes are not intended for heavy computing, please use slurm batch jobs instead. See our instructions on how to use the batch job system.

Local storage

The GPU nodes have fast local storage which is useful for IO-intensive applications. See our general instructions on how to take the fast local storage into use.

More information

Last edited Mon Apr 25 2022