GPU Accelerated Data Science
RAPIDS provides unmatched speed with familiar APIs that match the most popular PyData libraries. Built on the shoulders of giants including NVIDIA CUDA and Apache Arrow, it unlocks the speed of GPUs with code you already know.Learn more on the About Section
RAPIDS allows fluid, creative interaction with data for everyone from BI users to AI researchers on the cutting edge. GPU acceleration means less time and less cost moving data and training models.Find out more from RAPIDS Use Cases
RAPIDS is Open Source and available on GitHub. Our mission is to empower and advance the open-source GPU data science data engineering ecosystem.Jump to RAPIDS on GitHub
cuDF is a near drop in replacement to pandas for most use cases and has greatly improved performance.
* Benchmark on AMD EPYC 7642 (using 1x 2.3GHz CPU core) w/ 512GB and NVIDIA A100 80GB (1x GPU) w/ pandas v1.5 and cuDF v23.02
cuML brings huge speedups to ML modeling with an API that matches scikit-learn.
* Benchmark on AMD EPYC 7642 (using 1x 2.3GHz CPU core) w/ 512GB and NVIDIA A100 80GB (1x GPU) w/ scikit-learn v1.2 and cuML v23.02
cuGraph makes migration from networkX easy, accelerates graph analytics, and allows scaling far beyond existing tools.Run this benchmark yourself
* Benchmark on AMD EPYC 7642 (using 1x 2.3GHz CPU core) w/ 512GB and NVIDIA A100 80GB (1x GPU) w/ networkX v3.0 and cuGraph v23.02
RAPIDS offers several installation methods, the quickest is shown below. For more detailed guides go to our advanced install page
A. NVIDIA Pascal™ or better GPU with a compute capability 6.0 and above
B. Ubuntu 20.04 or 22.04, CentOS 7, Rocky Linux 8, or WSL2 on Windows 11
C. Recent CUDA version and NVIDIA driver pairs. Check yours with:
1. If not installed, download and run the install script.
This will install the latest miniconda:
wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh bash Miniconda3-latest-Linux-x86_64.sh
2. Then quick install RAPIDS with:
conda create -n rapids-23.04 -c rapidsai -c conda-forge -c nvidia rapids=23.04 python=3.10 cudatoolkit=11.8
1. Install via pip channels.
Not all RAPIDS libraries are currently available:
pip install cudf-cu11 dask-cudf-cu11 --extra-index-url=https://pypi.nvidia.com pip install cuml-cu11 --extra-index-url=https://pypi.nvidia.com pip install cugraph-cu11 --extra-index-url=https://pypi.nvidia.com
Check that you have the required environment and then use the install selector
Use the Windows WSL2 installation instructions
If you want to use Docker, WSL2, have an uncommon OS, hardware configuration, environment, or
need specific versions such as
CUDA 11.5 or
Python 3.8 then you should use the
installation selector in our
Advanced Installations Page
Don't have access to a GPU system right now? Try out RAPIDS with cloud based hardware from one of these featured channels:
Jump right into a GPU enabled notebook environment with required account. Also available with Conda.
Enables notebook based environments in a free trial with required account.
Use Quick Start Instances through a limited free account.
Free short term use to try and learn with hands-on lab environment.
Microsoft Azure Cloud infrastructure and services are available with RAPIDS.
Oracle Cloud infrastructure and services are available with RAPIDS.
IBM Cloud infrastructure and services are available with RAPIDS.
After installing, the best place to start is by looking through our more detailed tutorials and guides on the User Guides Page
NVIDIA's industry leading hardware provides the platform for RAPIDS high performance. Get details on the newest GPUs, server architectures, and cloud offerings in our Ecosystem Hardware Section
Find out details on featured RAPIDS projects like cuDF, cuML, cuGraph, and more. Also learn about those using our integrated with RAPIDS in our Ecosystem Software Section
Get involved with RAPIDS projects, reach out to its developers, find maintainer and contribution guides in our Ecosystem Developers Section
RAPIDS would not be possible without the collaboration of these important open source projects. Click on a logo to learn more:
Learn more about RAPIDS' start with Apache Arrow and GoAi. Also find an overview of the capabilities of RAPIDS, as well as featured projects in our About Section
Hear about success stories, resources for integrating RAPIDs workflows in your business, and deployment strategies in our Use Cases Section
Use RAPIDS directly or through NVIDIA AI Enterprise, which provides extensive optimization, certified hardware profiles, and direct IT support. Find additional business resources, community resources, and guides for RAPIDS evangelism in our Get Involved Section
Follow the latest from the RAPIDS Twitter community with @RAPIDSai
Get the full list of developer updates and notices (RSN) that may affect your projects on the RSN Documentation Page
Find our highlighted content, including videos, talks, posts, guides and more in the Featured Content Page
First Release of the Year RAPIDS Release 23.02 is live. We are most excited to highlight several quality-of-life improvements. We have a new web presence for RAPIDS; we believe this will make learning, using, and adding to RAPIDS easier and more enjoyable than ever b...
Post by Paul Mahler
Imbalanced-learn is the most popular open source library for resampling datasets with class imbalance. With more than 3 million downloads in the past month alone, it’s a critical piece of the data analysis and machine learning ecosystem. RAPIDS cuML is a suite of fas...
Post by Nick Becker
It’s now easy to switch between CPU (NumPy / Pandas) and GPU (CuPy / cuDF) in Dask. As of Dask 2022.10.0, users can optionally select the backend engine for input IO and data creation. In the short-term, the goal of the backend-configuration system is to enable Dask ...
Post by Rick Zamora
BERTopic is a topic modeling framework that brings neural network embeddings and classical machine learning techniques together into a state-of-the-art solution. The library provides a friendly user interface for many different tasks, including guided, supervised, se...
Post by Maarten Grootendorst
GPU-accelerated machine learning is producing fascinating results across a wide range of fields on a seemingly daily basis. While neural networks are usually the focus of attention, data pre-processing and preparation is just as important and often a bottleneck in ma...
Post by Ashwin Srinath
cuSpatial 22.12 focused on progress toward four primary goals: distance queries, spatial relationship queries, linestring intersection, and low-level C++ refactoring and iteration support. cuSpatial 22.12 takes steps towards rich support of spatial relationship queri...
Post by Michael Wang