cuDF - GPU DataFrames

📢 cuDF can now be used as a no-code-change accelerator for pandas! To learn more, see here!

cuDF (pronounced "KOO-dee-eff") is a GPU DataFrame library for loading, joining, aggregating, filtering, and otherwise manipulating data. cuDF leverages libcudf, a blazing-fast C++/CUDA dataframe library and the Apache Arrow columnar format to provide a GPU-accelerated pandas API.

You can import cudf directly and use it like pandas:

import cudf
import requests
from io import StringIO

url = "https://github.com/plotly/datasets/raw/master/tips.csv"
content = requests.get(url).content.decode("utf-8")

tips_df = cudf.read_csv(StringIO(content))
tips_df["tip_percentage"] = tips_df["tip"] / tips_df["total_bill"] * 100

# display average tip by dining party size
print(tips_df.groupby("size").tip_percentage.mean())

Or, you can use cuDF as a no-code-change accelerator for pandas, using cudf.pandas. cudf.pandas supports 100% of the pandas API, utilizing cuDF for supported operations and falling back to pandas when needed:

%load_ext cudf.pandas  # pandas operations now use the GPU!

import pandas as pd
import requests
from io import StringIO

url = "https://github.com/plotly/datasets/raw/master/tips.csv"
content = requests.get(url).content.decode("utf-8")

tips_df = pd.read_csv(StringIO(content))
tips_df["tip_percentage"] = tips_df["tip"] / tips_df["total_bill"] * 100

# display average tip by dining party size
print(tips_df.groupby("size").tip_percentage.mean())

Resources

• Try cudf.pandas now: Explore cudf.pandas on a free GPU enabled instance on Google Colab!
• Install: Instructions for installing cuDF and other RAPIDS libraries.
• cudf (Python) documentation
• libcudf (C++/CUDA) documentation
• RAPIDS Community: Get help, contribute, and collaborate.

See the RAPIDS install page for the most up-to-date information and commands for installing cuDF and other RAPIDS packages.

Installation

CUDA/GPU requirements

• CUDA 11.2+
• NVIDIA driver 450.80.02+
• Volta architecture or better (Compute Capability >=7.0)

Pip

cuDF can be installed via pip from the NVIDIA Python Package Index. Be sure to select the appropriate cuDF package depending on the major version of CUDA available in your environment:

For CUDA 11.x:

pip install --extra-index-url=https://pypi.nvidia.com cudf-cu11

For CUDA 12.x:

pip install --extra-index-url=https://pypi.nvidia.com cudf-cu12

Conda

cuDF can be installed with conda (via miniconda or the full Anaconda distribution from the rapidsai channel:

conda install -c rapidsai -c conda-forge -c nvidia \
    cudf=24.06 python=3.11 cuda-version=12.2

We also provide nightly Conda packages built from the HEAD of our latest development branch.

Note: cuDF is supported only on Linux, and with Python versions 3.9 and later.

See the RAPIDS installation guide for more OS and version info.

Build/Install from Source

See build instructions.

Contributing

Please see our guide for contributing to cuDF.