下载:v26.10.02v26.10.02 ↓26.10.022026-10-20
+1
日增长
9,802
stars
1,162
forks
148
watchers
17,032
prs
8,186
issues
76
releases

README

NVIDIA cuDF: A GPU-accelerated DataFrame library for tabular data processing

NVIDIA cuDF (pronounced "KOO-dee-eff") is an Apache 2.0 licensed, GPU-accelerated DataFrame library
for tabular data processing. The cuDF library is one part of the NVIDIA
CUDA-X suite of GPU
Accelerated libraries.

 About

cuDF is composed of multiple libraries including:

 libcudf: A CUDA C++ library with Apache Arrow compliant
data structures and fundamental algorithms for tabular data.
 pylibcudf: A Python library providing Cython bindings for libcudf.
 cudf: A Python library providing
    - A DataFrame library mirroring the pandas API
    - A zero-code change accelerator, cudf.pandas, for existing pandas code.
 cudf-polars: A Python library providing a GPU engine for Polars
 dask-cudf: A Python library providing a GPU backend for Dask DataFrames

Examples of projects that use cuDF:

 NVIDIA cuDF plugin for Apache Spark: A GPU accelerator plugin for Apache Spark
 cuML (NVIDIA): cuDF DataFrame inputs and outputs for GPU machine-learning workflows.
 cuGraph (NVIDIA): cuDF edge-list inputs for GPU graph construction and analytics.
 NeMo Curator (NVIDIA): cuDF-backed GPU processing for text curation and deduplication.
 Velox-cuDF: An experimental Velox extension that executes supported query operators on GPUs using libcudf.
 Polars: A cuDF-backed GPU engine for supported Lazy API queries.
 Sirius: A GPU-native SQL engine providing extensions for libraries like DuckDB
 Starburst Torin: GPU execution for Starburst Enterprise SQL using cuDF.
 Presto GPU: Experimental GPU-accelerated Presto query processing through Velox and cuDF, with public evaluation notebooks.
 TornadoVM: A GPU programming framework for Java that JIT-compiles Java to CUDA and runs libcudf operations as tasks alongside those kernels; used as a GPU engine for Apache Flink
 HoloViz: Datashader, HoloViews and hvPlot provide cuDF-backed aggregation, data exploration and plotting interfaces.
 PyGraphistry: cuDF-backed graph queries and exploration through GFQL.
 Ray Data: Experimental cuDF batch support for GPU transforms in distributed data pipelines.
 XGBoost: Native cuDF DataFrame and Series inputs for supported gradient-boosting workflows.

This list is not exhaustive. If your project uses cuDF, you're welcome to submit a pull request with a project link, a short description of the integration, and public documentation or an example.

 Installation

 System Requirements

Operating System, GPU driver, and supported CUDA version information can be found at the system requirements

 pip

A stable release of each cudf library is available on PyPI. You will need to match the major version number of your installed CUDA version with a -cu suffix when installing from PyPI.

A development version of each library is available as a nightly release by including the -i https://pypi.anaconda.org/rapidsai-wheels-nightly/simple/ index.

bash
 CUDA 13
pip install libcudf-cu13
pip install pylibcudf-cu13
pip install cudf-cu

Top issues

Discussions

Latest release

v26.10.02 v26.10.02 2026-10-20
What's Changed
 🛠️ Improvements
 Require only major.minor rapidsmpf version in libcudf_streaming (24719) by @josephine-wolf-oberholtzer in https://github.com/NVIDIA/cudf/pull/24721


Full Changelog: https://github.com/NVIDIA/cudf/compare/v26.10.01...v26.10.02

Snapshot: 2026-11-16T09:23:14Z · Data license Apache-2.0 · JSON