cuml

NVIDIA cuML: GPU-Accelerated Machine Learning

What is cuml?

NVIDIA cuML: GPU-Accelerated Machine Learning

Overview of cuml

What it is

NVIDIA cuML is an open-source CUDA-X Data Science library for GPU-accelerated machine learning. It provides two ways to run machine learning workloads on NVIDIA GPUs: the cuml Python API with GPU-native estimators and a scikit-learn-style interface, and cuml.accel which accelerates existing scikit-learn, UMAP, and HDBSCAN code without changing the Python code. On representative benchmarks, cuML can accelerate scikit-learn workflows by up to 50x, depending on the algorithm, dataset, and hardware.

Key features

  • GPU-native cuml API following the familiar scikit-learn fit-predict-transform pattern, keeping data and computation on the GPU.
  • cuml.accel module that accelerates existing scikit-learn, UMAP, and HDBSCAN code without modifying the original Python code.
  • Support for clustering, dimensionality reduction, regression, classification, preprocessing, model selection, model explanation, and nearest-neighbor workflows.
  • Distributed implementations via cuml.dask for multi-GPU and multi-node execution with Dask.
  • Model serialization with pickle or joblib (using cloudpickle) for inference, with security warnings about untrusted sources.
  • Compatible with scikit-learn version 1.6 or higher.

Use cases

  • Running machine learning workflows on NVIDIA GPUs using a familiar scikit-learn-like API.
  • Accelerating existing Python scripts or Jupyter notebooks that use scikit-learn, UMAP, or HDBSCAN without code changes.
  • Scaling machine learning algorithms across multiple GPUs or multiple nodes using Dask.
  • Clustering, classification, regression, and dimensionality reduction on large datasets that benefit from GPU acceleration.

Getting started

See the repository README for installation instructions. Use the installation selector at the cuML documentation site to generate a command for installing nightly or release cuML packages with conda, pip, or Docker.

License and pricing

License: Apache-2.0. Pricing: open source (free).

Based on the project's public README and repository metadata.

Add this badge to your README

Featured on YunoDB Badge styles & help

Maintain this project? Show that it is featured on YunoDB.

Markdown[![Featured on YunoDB](https://yunodb.com/badge/nvidiacuml.svg)](https://yunodb.com/tools/nvidiacuml?utm_source=github&utm_medium=badge)
HTML<a href="https://yunodb.com/tools/nvidiacuml?utm_source=github&utm_medium=badge"><img src="https://yunodb.com/badge/nvidiacuml.svg" alt="Featured on YunoDB" height="20"></a>