Developer Guide

This guide collects the information needed to contribute to, develop, test, document, and benchmark cuML. Start with Contributing, then use the implementation-specific guidance below.

• Contributing covers how to propose changes, prepare pull requests, run repository checks, and work with continuous integration.

• Python development covers style, testing, validation, memory management, and Python documentation.

  • Python estimator development describes the cuml.Base estimator contract and provides implementation patterns.

• C++ and CUDA development covers source layout, resources, testing, and Doxygen documentation.

  • Internal C++ API reference exposes primarily internal libcuml interfaces for developers. Prefer the supported Python API for applications.

• Benchmarking explains the benchmark CLI and manifests.