libraft-cu12
RAFT: Reusable Algorithms Functions and other Tools (C++)
Decision gist · record as of 2026-08-14
Yes, if you have a compatible NVIDIA GPU and are building GPU-accelerated machine learning or data mining applications. The library is actively maintained, permissively licensed under Apache-2.0, and provides well-tested primitives that reduce development time. Install only if you need low-level GPU acceleration for algorithms—it is not intended for direct data science experimentation. Requires Python 3.11+, CUDA Toolkit, and medium install friction due to GPU runtime dependencies.AI-flagged interpretation of the facts on this page — verify before relying
Before you install
- Requires CUDA Toolkit, compatible NVIDIA GPU, and Python 3.11 or later.
- Medium install friction due to CUDA toolkit and NVIDIA runtime dependencies (librmm-cu12, nvidia-nccl-cu12, nvidia-nvjitlink-cu12).
- Actively maintained with recent releases; requires Python 3.11 or later.
License · maintenance · safety
Apache-2.0 (permissive) — Apache-2.0 permissive license allows commercial and private use with minimal restrictions, making it suitable for most production environments.
last release 2026-08-06 (8 days) · last repo commit 2026-08-14 · 1,037 stars
0 known vulnerabilities (OSV.dev, 2026-08-14) · 366,894 downloads/mo, #7,201 on PyPI
Alternatives
Verify before relying
pip install libraft-cu12
from libraft-cu12 import device_resources
handle = device_resources()
# Use handle with RAFT primitives for GPU-accelerated computations- Whether the C++ header-only library and optional shared library are both included in the cu12 wheel.
- Specific GPU architectures and CUDA compute capabilities supported by this release.
- Complete list of algorithms and primitives exposed in the Python API for version 26.8.0.
- Interoperability details with external libraries beyond what the description excerpt covers.
What it is and what it does
libraft-cu12 is a CUDA-accelerated C++ library providing low-level, reusable primitives for machine learning and data mining. It includes accelerated functions for dense and sparse linear algebra, matrix operations, solvers, statistical computations, and GPU infrastructure utilities. The library is designed as a building block for developers creating high-performance GPU applications, not for direct use by data scientists—it centralizes core computations to reduce maintenance burden and enable future optimizations across dependent projects.
The package offers both C++ APIs (header-only with optional shared library for faster compilation) and Python bindings through pylibraft, which wraps host-accessible runtime APIs. Python users can pass objects supporting the __cuda_array_interface__ directly to RAFT functions. The library depends on cuda-toolkit, librmm-cu12, nvidia-nccl-cu12, nvidia-nvjitlink-cu12, and rapids-logger, making it suitable for environments with compatible GPUs.
Use it for
- Accelerate pairwise distance computations between large GPU-resident arrays for nearest-neighbor search.
- Build distributed multi-GPU machine learning workflows using raft-dask infrastructure.
- Implement custom GPU-accelerated algorithms by reusing RAFT's linear algebra and solver primitives.
- Optimize sparse and dense matrix operations on GPU without writing custom CUDA kernels.
- Integrate GPU-accelerated statistical sampling and model evaluation into existing GPU pipelines.
Worth the install?
AI-flagged interpretation of the facts on this page. Verify before relying on it.
Yes, if you have a compatible NVIDIA GPU and are building GPU-accelerated machine learning or data mining applications.
The library is actively maintained, permissively licensed under Apache-2.0, and provides well-tested primitives that reduce development time. Install only if you need low-level GPU acceleration for algorithms—it is not intended for direct data science experimentation. Requires Python 3.11+, CUDA Toolkit, and medium install friction due to GPU runtime dependencies.
Install
libraft-cu12 on PyPI
Before you install
Medium install friction due to CUDA toolkit and NVIDIA runtime dependencies (librmm-cu12, nvidia-nccl-cu12, nvidia-nvjitlink-cu12). Actively maintained with recent releases; requires Python 3.11 or later.
Requires CUDA Toolkit, compatible NVIDIA GPU, and Python 3.11 or later.
License in practice
Apache-2.0 permissive license allows commercial and private use with minimal restrictions, making it suitable for most production environments.
Quickstart
pip install libraft-cu12
from libraft-cu12 import device_resources
handle = device_resources()
# Use handle with RAFT primitives for GPU-accelerated computationsVerify before relying
- Whether the C++ header-only library and optional shared library are both included in the cu12 wheel.
- Specific GPU architectures and CUDA compute capabilities supported by this release.
- Complete list of algorithms and primitives exposed in the Python API for version 26.8.0.
- Interoperability details with external libraries beyond what the description excerpt covers.
Package facts
| License | Apache-2.0 permissive |
| Python support | Supports the current Python release >=3.11 |
| Install friction | Medium. Platform-specific wheel |
| Runtime dependencies | 5 packagescuda-toolkitlibrmm-cu12nvidia-nccl-cu12nvidia-nvjitlink-cu12rapids-logger |
| Maintenance | Actively maintained 8 days since the last release |
| Last repo commit | |
| First released | |
| Downloads | 366,894 / month, #7,201 on PyPI 30-day window, as of 2026-08-14 |
| Known vulnerabilities | None known OSV.dev, checked 2026-08-14 |
| Classifiers | Intended Audience :: Developers |
Evidence: libraft_cu12-26.8.0-py3-none-manylinux_2_26_aarch64.manylinux_2_28_aarch64.whl; libraft_cu12-26.8.0-py3-none-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl
Tags
Let your AI agent find packages like this
Example. Real query, live index.
You found this page by searching. An agent finds it by wishing: SkillFed indexes 14,416 PyPI packages by what they can do, searchable in plain language.
wish › “cuda accelerated machine learning primitives”
- libraft-cu12libraft-cu12 provides CUDA-accelerated primitives and algorithms for…
- pylibraft-cu12pylibraft-cu12 provides Python bindings to RAFT's CUDA-accelerated…
- nvidia-cudnn-cu13Provides cuDNN runtime libraries for GPU-accelerated deep neural…
Give your agent the search over MCP, or paste the wish link into any chat.
More Scientific/Engineering packages
NumPy provides an N-dimensional array object and a comprehensive suite of mathematical, linear algebra, Fourier transform, and random number functions for scientific computing in Python.
pandas provides fast, flexible data structures (Series and DataFrame) for loading, cleaning, transforming, and analyzing labeled or relational data in Python.
scipy provides numerical algorithms for mathematics, science, and engineering—including optimization, integration, linear algebra, Fourier transforms, signal and image processing, and ODE solvers—built on numpy arrays.
scikit-learn provides a comprehensive Python library for supervised and unsupervised machine learning, including classification, regression, clustering, dimensionality reduction, and model evaluation tools built on NumPy and SciPy.
Install it if you need to train, evaluate, or deploy supervised or unsupervised learning models.
dill extends Python's pickle module to serialize and deserialize a much wider range of Python objects, including functions, lambdas, classes, and interpreter sessions, to byte streams for storage or network transmission.
Multiprocess is an enhanced fork of Python's standard multiprocessing library that uses dill for better serialization, allowing you to spawn processes with a threading-like API and share complex objects between them.
Install it if you use multiprocessing and encounter pickle serialization limits with lambdas or complex objects.
See also pylibraft-cu12 · raft-dask-cu12 · librmm-cu12 · cuml-cu12 · rmm-cu12 · libcuml-cu12 · libcuvs-cu12 · libcudf-cu12 · cudf-cu12 · pylibcudf-cu12