nvmath-python: Unleashing the Full Capabilities of NVIDIA Math Libraries within Python

Welcome to the nvmath-python documentation!

nvmath-python is a Python library to enable cutting edge performance, productivity, and interoperability within the Python computational ecosystem through NVIDIA’s high-performance math libraries.

To quickly get started with nvmath-python, take a look at our Getting Started manual. Refer to our Installation Guide for detailed instructions on the various installation choices available.

Contents

User Guide

• Getting Started
  • Installation
  • Examples
• Overview
  • Architecture
  • Host APIs
  • Host APIs with Callbacks
  • Device APIs
  • Compatibility Policy
• Installation
  • Install nvmath-python
  • Run nvmath-python
  • Troubleshooting

Examples and tutorials

• Linear Algebra Host APIs Tutorial
  • Introduction to GEMM with nvmath-python
  • Fused Epilogs
  • Implementing a simple neural network
  • Narrow-precision operations
• Examples on GitHub

API Reference

• Host APIs
  • Contents
    • Linear Algebra
    • Fast Fourier Transform
    • Host API Utilities
• Device APIs
  • Device API utilities
    • Overview
    • API Reference
  • cuBLASDx
    • Overview
    • API Reference
  • cuFFTDx
    • Overview
    • API Reference
  • cuRAND Device APIs
    • Overview
    • API Reference
• Bindings
  • Overview
  • Naming & Calling Convention
  • Memory management
    • Pointer and data lifetime
  • API Reference
    • cuBLAS (nvmath.bindings.cublas)
    • cuBLASLt (nvmath.bindings.cublaslt)
    • cuFFT (nvmath.bindings.cufft)
    • cuSOLVER (nvmath.bindings.cusolver)
    • cuSOLVERDn (nvmath.bindings.cusolverDn)
    • cuSPARSE (nvmath.bindings.cusparse)
    • cuRAND (nvmath.bindings.curand)

Other Resources

• Release Notes
• Code of Conduct
• Contributing
• License
• GitHub Repository