CuPy – A NumPy-compatible array library accelerated by CUDA

• Overview
• Installation
  • Requirements
  • Installing CuPy
  • Installing CuPy from Conda-Forge
  • Installing CuPy from Source
  • Uninstalling CuPy
  • Upgrading CuPy
  • Reinstalling CuPy
  • Using CuPy inside Docker
  • FAQ
• Using CuPy on AMD GPU (experimental)
  • Requirements
  • Environment Variables
  • Docker
  • Installing Binary Packages
  • Building CuPy for ROCm From Source
• User Guide
  • Basics of CuPy
  • User-Defined Kernels
  • Accessing CUDA Functionalities
  • Memory Management
  • Performance Best Practices
  • Interoperability
  • Difference between CuPy and NumPy
  • API Compatibility Policy
• API Reference
  • Multi-Dimensional Array (ndarray)
  • Universal Functions (ufunc)
  • Routines
  • SciPy-compatible Routines
  • CuPy-specific Functions
  • Low-Level CUDA Support
  • Custom kernels
  • Profiling
  • Environment variables
  • Comparison Table

Development

• Contribution Guide
  • Classification of Contributions
  • Development Cycle
  • Issues and Pull Requests
  • Coding Guidelines
  • Unit Testing
  • Documentation
  • Tips for Developers

Misc Notes

• Upgrade Guide
  • CuPy v9
  • CuPy v8
  • CuPy v7
  • CuPy v6
  • CuPy v5
  • CuPy v4
  • CuPy v2
• License
  • NumPy
  • SciPy

Overview